openharmony/third_party_littlefs
1
0
Fork 0
mirror of https://gitee.com/openharmony/third_party_littlefs synced 2024-11-23 14:59:50 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
0982020fb3
third_party_littlefs/lfs.c

2157 lines
56 KiB
C
Raw Normal View History

Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
/*
* The little filesystem
*
* Copyright (c) 2017 Christopher Haster
Adopted the Apache 2.0 license
2017-07-08 16:49:40 +00:00
* Distributed under the Apache 2.0 license
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
*/
#include "lfs.h"
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
#include "lfs_util.h"
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
#include <string.h>
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
#include <stdlib.h>
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
#include <assert.h>
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
/// Caching block device operations ///
static int lfs_cache_read(lfs_t *lfs, lfs_cache_t *rcache,
const lfs_cache_t *pcache, lfs_block_t block,
Adopted more conventional buffer parameter ordering Adopted buffer followed by size. The other order was original chosen due to some other functions with a more complicated parameter list. This convention is important, as the bd api is one of the main apis facing porting efforts.
2017-04-24 04:49:21 +00:00
lfs_off_t off, void *buffer, lfs_size_t size) {
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
uint8_t *data = buffer;
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
assert(block < lfs->cfg->block_count);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
while (size > 0) {
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (pcache && block == pcache->block && off >= pcache->off &&
off < pcache->off + lfs->cfg->prog_size) {
// is already in pcache?
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
lfs_size_t diff = lfs_min(size,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs->cfg->prog_size - (off-pcache->off));
memcpy(data, &pcache->buffer[off-pcache->off], diff);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
data += diff;
off += diff;
size -= diff;
continue;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
}
if (block == rcache->block && off >= rcache->off &&
off < rcache->off + lfs->cfg->read_size) {
// is already in rcache?
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
lfs_size_t diff = lfs_min(size,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs->cfg->read_size - (off-rcache->off));
memcpy(data, &rcache->buffer[off-rcache->off], diff);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
data += diff;
off += diff;
size -= diff;
continue;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (off % lfs->cfg->read_size == 0 && size >= lfs->cfg->read_size) {
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
// bypass cache?
lfs_size_t diff = size - (size % lfs->cfg->read_size);
Adopted more conventional buffer parameter ordering Adopted buffer followed by size. The other order was original chosen due to some other functions with a more complicated parameter list. This convention is important, as the bd api is one of the main apis facing porting efforts.
2017-04-24 04:49:21 +00:00
int err = lfs->cfg->read(lfs->cfg, block, off, data, diff);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (err) {
return err;
}
data += diff;
off += diff;
size -= diff;
continue;
}
// load to cache, first condition can no longer fail
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
rcache->block = block;
rcache->off = off - (off % lfs->cfg->read_size);
int err = lfs->cfg->read(lfs->cfg, rcache->block,
rcache->off, rcache->buffer, lfs->cfg->read_size);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (err) {
return err;
}
}
return 0;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
static int lfs_cache_cmp(lfs_t *lfs, lfs_cache_t *rcache,
const lfs_cache_t *pcache, lfs_block_t block,
lfs_off_t off, const void *buffer, lfs_size_t size) {
const uint8_t *data = buffer;
for (lfs_off_t i = 0; i < size; i++) {
uint8_t c;
int err = lfs_cache_read(lfs, rcache, pcache,
block, off+i, &c, 1);
if (err) {
return err;
}
if (c != data[i]) {
return false;
}
}
return true;
}
static int lfs_cache_crc(lfs_t *lfs, lfs_cache_t *rcache,
const lfs_cache_t *pcache, lfs_block_t block,
lfs_off_t off, lfs_size_t size, uint32_t *crc) {
for (lfs_off_t i = 0; i < size; i++) {
uint8_t c;
int err = lfs_cache_read(lfs, rcache, pcache,
block, off+i, &c, 1);
if (err) {
return err;
}
lfs_crc(crc, &c, 1);
}
return 0;
}
static int lfs_cache_flush(lfs_t *lfs,
lfs_cache_t *pcache, lfs_cache_t *rcache) {
if (pcache->block != 0xffffffff) {
int err = lfs->cfg->prog(lfs->cfg, pcache->block,
pcache->off, pcache->buffer, lfs->cfg->prog_size);
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (err) {
return err;
}
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if (rcache) {
int res = lfs_cache_cmp(lfs, rcache, NULL, pcache->block,
pcache->off, pcache->buffer, lfs->cfg->prog_size);
if (res < 0) {
return res;
}
if (!res) {
return LFS_ERR_CORRUPT;
}
}
pcache->block = 0xffffffff;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
return 0;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
static int lfs_cache_prog(lfs_t *lfs, lfs_cache_t *pcache,
lfs_cache_t *rcache, lfs_block_t block,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_off_t off, const void *buffer, lfs_size_t size) {
const uint8_t *data = buffer;
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
assert(block < lfs->cfg->block_count);
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
while (size > 0) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if (block == pcache->block && off >= pcache->off &&
off < pcache->off + lfs->cfg->prog_size) {
// is already in pcache?
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
lfs_size_t diff = lfs_min(size,
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
lfs->cfg->prog_size - (off-pcache->off));
memcpy(&pcache->buffer[off-pcache->off], data, diff);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
data += diff;
off += diff;
size -= diff;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (off % lfs->cfg->prog_size == 0) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
// eagerly flush out pcache if we fill up
int err = lfs_cache_flush(lfs, pcache, rcache);
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (err) {
return err;
}
}
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
continue;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
// pcache must have been flushed, either by programming and
// entire block or manually flushing the pcache
assert(pcache->block == 0xffffffff);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (off % lfs->cfg->prog_size == 0 &&
size >= lfs->cfg->prog_size) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
// bypass pcache?
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
lfs_size_t diff = size - (size % lfs->cfg->prog_size);
Adopted more conventional buffer parameter ordering Adopted buffer followed by size. The other order was original chosen due to some other functions with a more complicated parameter list. This convention is important, as the bd api is one of the main apis facing porting efforts.
2017-04-24 04:49:21 +00:00
int err = lfs->cfg->prog(lfs->cfg, block, off, data, diff);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (err) {
return err;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if (rcache) {
int res = lfs_cache_cmp(lfs, rcache, NULL,
block, off, data, diff);
if (res < 0) {
return res;
}
if (!res) {
return LFS_ERR_CORRUPT;
}
}
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
data += diff;
off += diff;
size -= diff;
continue;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
// prepare pcache, first condition can no longer fail
pcache->block = block;
pcache->off = off - (off % lfs->cfg->prog_size);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
}
return 0;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
/// General lfs block device operations ///
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
static int lfs_bd_read(lfs_t *lfs, lfs_block_t block,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_off_t off, void *buffer, lfs_size_t size) {
// if we ever do more than writes to alternating pairs,
// this may need to consider pcache
return lfs_cache_read(lfs, &lfs->rcache, NULL,
block, off, buffer, size);
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
static int lfs_bd_prog(lfs_t *lfs, lfs_block_t block,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_off_t off, const void *buffer, lfs_size_t size) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
return lfs_cache_prog(lfs, &lfs->pcache, NULL,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
block, off, buffer, size);
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
static int lfs_bd_cmp(lfs_t *lfs, lfs_block_t block,
lfs_off_t off, const void *buffer, lfs_size_t size) {
return lfs_cache_cmp(lfs, &lfs->rcache, NULL, block, off, buffer, size);
}
static int lfs_bd_crc(lfs_t *lfs, lfs_block_t block,
lfs_off_t off, lfs_size_t size, uint32_t *crc) {
return lfs_cache_crc(lfs, &lfs->rcache, NULL, block, off, size, crc);
}
static int lfs_bd_erase(lfs_t *lfs, lfs_block_t block) {
Simplified config Before, the lfs had multiple paths to determine config options: - lfs_config struct passed during initialization - lfs_bd_info struct passed during block device initialization - compile time options This allowed different developers to provide their own needs to the filesystem, such as the block device capabilities and the higher level user's own tweaks. However, this comes with additional complexity and action required when the configurations are incompatible. For now, this has been reduced to all information (including block device function pointers) being passed through the lfs_config struct. We just defer more complicated handling of configuration options to the top level user. This simplifies configuration handling and gives the top level user the responsibility to handle configuration, which they probably would have wanted to do anyways.
2017-04-22 16:42:05 +00:00
return lfs->cfg->erase(lfs->cfg, block);
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
static int lfs_bd_sync(lfs_t *lfs) {
Fixed dirty rcache during directory commit An interesting side-effect of adding internal checks to the littlefs for block errors, is that the littlefs starts to cover up its own flaws. Probably out of embarrassment. In this case, the relocation logic for directories left the littlefs rcache dirty with invalid data. The littlefs detected the error, treated it as a corrupted write, and just moved the "corrupted" block to a new block, which as a side-effect flushes the rcache. Since committing a dir will end up flushing the rcache to check for errors anyways, we can just drop the rcache in lfs_bd_sync.
2017-06-25 21:21:14 +00:00
lfs->rcache.block = 0xffffffff;
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_cache_flush(lfs, &lfs->pcache, NULL);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (err) {
return err;
}
Simplified config Before, the lfs had multiple paths to determine config options: - lfs_config struct passed during initialization - lfs_bd_info struct passed during block device initialization - compile time options This allowed different developers to provide their own needs to the filesystem, such as the block device capabilities and the higher level user's own tweaks. However, this comes with additional complexity and action required when the configurations are incompatible. For now, this has been reduced to all information (including block device function pointers) being passed through the lfs_config struct. We just defer more complicated handling of configuration options to the top level user. This simplifies configuration handling and gives the top level user the responsibility to handle configuration, which they probably would have wanted to do anyways.
2017-04-22 16:42:05 +00:00
return lfs->cfg->sync(lfs->cfg);
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
/// Internal operations predeclared here ///
int lfs_traverse(lfs_t *lfs, int (*cb)(void*, lfs_block_t), void *data);
static int lfs_pred(lfs_t *lfs, const lfs_block_t dir[2], lfs_dir_t *pdir);
static int lfs_parent(lfs_t *lfs, const lfs_block_t dir[2],
lfs_dir_t *parent, lfs_entry_t *entry);
static int lfs_relocate(lfs_t *lfs,
const lfs_block_t oldpair[2], const lfs_block_t newpair[2]);
int lfs_deorphan(lfs_t *lfs);
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
/// Block allocator ///
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
static int lfs_alloc_lookahead(void *p, lfs_block_t block) {
lfs_t *lfs = p;
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
lfs_block_t off = (block - lfs->free.start) % lfs->cfg->block_count;
if (off < lfs->cfg->lookahead) {
lfs->free.lookahead[off / 32] |= 1U << (off % 32);
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
return 0;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
static int lfs_alloc(lfs_t *lfs, lfs_block_t *block) {
// deorphan if we haven't yet, only needed once after poweron
if (!lfs->deorphaned) {
int err = lfs_deorphan(lfs);
if (err) {
return err;
}
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
while (true) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
while (true) {
// check if we have looked at all blocks since last ack
if (lfs->free.start + lfs->free.off == lfs->free.end) {
LFS_WARN("No more free space %d", lfs->free.end);
return LFS_ERR_NOSPC;
}
if (lfs->free.off >= lfs->cfg->lookahead) {
break;
}
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
lfs_block_t off = lfs->free.off;
lfs->free.off += 1;
if (!(lfs->free.lookahead[off / 32] & (1U << (off % 32)))) {
// found a free block
*block = (lfs->free.start + off) % lfs->cfg->block_count;
return 0;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
}
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
lfs->free.start += lfs->cfg->lookahead;
lfs->free.off = 0;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
// find mask of free blocks from tree
memset(lfs->free.lookahead, 0, lfs->cfg->lookahead/8);
int err = lfs_traverse(lfs, lfs_alloc_lookahead, lfs);
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
if (err) {
return err;
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
static void lfs_alloc_ack(lfs_t *lfs) {
lfs->free.end = lfs->free.start + lfs->free.off + lfs->cfg->block_count;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
/// Metadata pair and directory operations ///
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
static inline void lfs_pairswap(lfs_block_t pair[2]) {
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
lfs_block_t t = pair[0];
pair[0] = pair[1];
pair[1] = t;
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
static inline bool lfs_pairisnull(const lfs_block_t pair[2]) {
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
return pair[0] == 0xffffffff || pair[1] == 0xffffffff;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
static inline int lfs_paircmp(
const lfs_block_t paira[2],
const lfs_block_t pairb[2]) {
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
return !(paira[0] == pairb[0] || paira[1] == pairb[1] ||
paira[0] == pairb[1] || paira[1] == pairb[0]);
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
static inline bool lfs_pairsync(
const lfs_block_t paira[2],
const lfs_block_t pairb[2]) {
return (paira[0] == pairb[0] && paira[1] == pairb[1]) ||
(paira[0] == pairb[1] && paira[1] == pairb[0]);
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
static int lfs_dir_alloc(lfs_t *lfs, lfs_dir_t *dir) {
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// allocate pair of dir blocks
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
for (int i = 0; i < 2; i++) {
int err = lfs_alloc(lfs, &dir->pair[i]);
if (err) {
return err;
}
}
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// rather than clobbering one of the blocks we just pretend
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
// the revision may be valid
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_bd_read(lfs, dir->pair[0], 0, &dir->d.rev, 4);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (err) {
return err;
}
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// set defaults
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
dir->d.rev += 1;
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
dir->d.size = sizeof(dir->d)+4;
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
dir->d.tail[0] = -1;
dir->d.tail[1] = -1;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
dir->off = sizeof(dir->d);
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// don't write out yet, let caller take care of that
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
return 0;
}
static int lfs_dir_fetch(lfs_t *lfs,
lfs_dir_t *dir, const lfs_block_t pair[2]) {
// copy out pair, otherwise may be aliasing dir
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
const lfs_block_t tpair[2] = {pair[0], pair[1]};
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
bool valid = false;
// check both blocks for the most recent revision
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
for (int i = 0; i < 2; i++) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
struct lfs_disk_dir test;
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_bd_read(lfs, tpair[i], 0, &test, sizeof(test));
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
if (err) {
return err;
}
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (valid && lfs_scmp(test.rev, dir->d.rev) < 0) {
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
continue;
}
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if ((0x7fffffff & test.size) < sizeof(test)+4 ||
(0x7fffffff & test.size) > lfs->cfg->block_size) {
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
continue;
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
uint32_t crc = 0xffffffff;
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
lfs_crc(&crc, &test, sizeof(test));
err = lfs_bd_crc(lfs, tpair[i], sizeof(test),
(0x7fffffff & test.size) - sizeof(test), &crc);
if (err) {
return err;
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
if (crc != 0) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
continue;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
valid = true;
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
// setup dir in case it's valid
dir->pair[0] = tpair[(i+0) % 2];
dir->pair[1] = tpair[(i+1) % 2];
dir->off = sizeof(dir->d);
dir->d = test;
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (!valid) {
LFS_ERROR("Corrupted dir pair at %d %d", tpair[0], tpair[1]);
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_CORRUPT;
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
}
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
return 0;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
struct lfs_region {
lfs_off_t oldoff;
lfs_size_t oldlen;
const void *newdata;
lfs_size_t newlen;
};
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
static int lfs_dir_commit(lfs_t *lfs, lfs_dir_t *dir,
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
const struct lfs_region *regions, int count) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
dir->d.rev += 1;
lfs_pairswap(dir->pair);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
for (int i = 0; i < count; i++) {
dir->d.size += regions[i].newlen - regions[i].oldlen;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
const lfs_block_t oldpair[2] = {dir->pair[0], dir->pair[1]};
bool relocated = false;
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
while (true) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_bd_erase(lfs, dir->pair[0]);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return err;
}
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
uint32_t crc = 0xffffffff;
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
lfs_crc(&crc, &dir->d, sizeof(dir->d));
err = lfs_bd_prog(lfs, dir->pair[0], 0, &dir->d, sizeof(dir->d));
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
int i = 0;
lfs_off_t oldoff = sizeof(dir->d);
lfs_off_t newoff = sizeof(dir->d);
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
while (newoff < (0x7fffffff & dir->d.size)-4) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (i < count && regions[i].oldoff == oldoff) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
lfs_crc(&crc, regions[i].newdata, regions[i].newlen);
int err = lfs_bd_prog(lfs, dir->pair[0],
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
newoff, regions[i].newdata, regions[i].newlen);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (err) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
oldoff += regions[i].oldlen;
newoff += regions[i].newlen;
i += 1;
} else {
uint8_t data;
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_bd_read(lfs, oldpair[1], oldoff, &data, 1);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
return err;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
lfs_crc(&crc, &data, 1);
err = lfs_bd_prog(lfs, dir->pair[0], newoff, &data, 1);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
oldoff += 1;
newoff += 1;
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
err = lfs_bd_prog(lfs, dir->pair[0], newoff, &crc, 4);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (err) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
return err;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
err = lfs_bd_sync(lfs);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
// successful commit, check checksum to make sure
crc = 0xffffffff;
err = lfs_bd_crc(lfs, dir->pair[0], 0, 0x7fffffff & dir->d.size, &crc);
if (err) {
return err;
}
if (crc == 0) {
break;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
relocate:
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
//commit was corrupted
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
LFS_DEBUG("Bad block at %d", dir->pair[0]);
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// drop caches and prepare to relocate block
relocated = true;
lfs->pcache.block = 0xffffffff;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// can't relocate superblock, filesystem is now frozen
if (lfs_paircmp(oldpair, (const lfs_block_t[2]){0, 1}) == 0) {
LFS_WARN("Superblock %d has become unwritable", oldpair[0]);
return LFS_ERR_CORRUPT;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// relocate half of pair
err = lfs_alloc(lfs, &dir->pair[0]);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (err) {
return err;
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if (relocated) {
// update references if we relocated
LFS_DEBUG("Relocating %d %d to %d %d",
oldpair[0], oldpair[1], dir->pair[0], dir->pair[1]);
return lfs_relocate(lfs, oldpair, dir->pair);
}
return 0;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
static int lfs_dir_update(lfs_t *lfs, lfs_dir_t *dir,
const lfs_entry_t *entry, const void *data) {
return lfs_dir_commit(lfs, dir, (struct lfs_region[]){
{entry->off, sizeof(entry->d), &entry->d, sizeof(entry->d)},
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
{entry->off+sizeof(entry->d), entry->d.nlen, data, entry->d.nlen}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}, data ? 2 : 1);
Adopted ctz skip-list structure earlier than expected The primary data structure backing the little fs was planned to be a little ctz based skip-list for O(logn) lookup and O(1) append. Was initially planning to start with a simple linked list of index blocks, but was having trouble implementing the free-list on top of the structure. Went ahead and adopted the skip-list structure since it may have actually been easier.
2017-02-27 03:00:39 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
static int lfs_dir_append(lfs_t *lfs, lfs_dir_t *dir,
lfs_entry_t *entry, const void *data) {
// check if we fit, if top bit is set we do not and move on
while (true) {
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
if (dir->d.size + 4+entry->d.elen+entry->d.alen+entry->d.nlen
<= lfs->cfg->block_size) {
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
entry->off = dir->d.size - 4;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return lfs_dir_commit(lfs, dir, (struct lfs_region[]){
{entry->off, 0, &entry->d, sizeof(entry->d)},
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
{entry->off, 0, data, entry->d.nlen}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}, 2);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// we need to allocate a new dir block
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (!(0x80000000 & dir->d.size)) {
lfs_dir_t newdir;
int err = lfs_dir_alloc(lfs, &newdir);
if (err) {
return err;
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
newdir.d.tail[0] = dir->d.tail[0];
newdir.d.tail[1] = dir->d.tail[1];
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
entry->off = newdir.d.size - 4;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
err = lfs_dir_commit(lfs, &newdir, (struct lfs_region[]){
{entry->off, 0, &entry->d, sizeof(entry->d)},
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
{entry->off, 0, data, entry->d.nlen}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}, 2);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (err) {
return err;
}
dir->d.size |= 0x80000000;
dir->d.tail[0] = newdir.pair[0];
dir->d.tail[1] = newdir.pair[1];
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return lfs_dir_commit(lfs, dir, NULL, 0);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
int err = lfs_dir_fetch(lfs, dir, dir->d.tail);
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
if (err) {
return err;
}
}
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
static int lfs_dir_remove(lfs_t *lfs, lfs_dir_t *dir, lfs_entry_t *entry) {
// either shift out the one entry or remove the whole dir block
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
if (dir->d.size == sizeof(dir->d)+4) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
lfs_dir_t pdir;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
int res = lfs_pred(lfs, dir->pair, &pdir);
if (res < 0) {
return res;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
if (!(pdir.d.size & 0x80000000)) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return lfs_dir_commit(lfs, dir, (struct lfs_region[]){
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
{entry->off, 4+entry->d.elen+entry->d.alen+entry->d.nlen,
NULL, 0},
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}, 1);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
} else {
pdir.d.tail[0] = dir->d.tail[0];
pdir.d.tail[1] = dir->d.tail[1];
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return lfs_dir_commit(lfs, dir, NULL, 0);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
} else {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return lfs_dir_commit(lfs, dir, (struct lfs_region[]){
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
{entry->off, 4+entry->d.elen+entry->d.alen+entry->d.nlen,
NULL, 0},
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}, 1);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
Added better handling for metadata pairs The core algorithim that backs this filesystem's goal of fault tolerance is the alternating of "metadata pairs". Backed by a simple core function for reading and writing, makes heavy use of c99 designated initializers for passing info about multiple chunks in an erase block.
2017-03-12 20:11:52 +00:00
}
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
static int lfs_dir_next(lfs_t *lfs, lfs_dir_t *dir, lfs_entry_t *entry) {
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
while (dir->off + sizeof(entry->d) > (0x7fffffff & dir->d.size)-4) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (!(0x80000000 & dir->d.size)) {
entry->off = dir->off;
return LFS_ERR_NOENT;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
int err = lfs_dir_fetch(lfs, dir, dir->d.tail);
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
if (err) {
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
dir->off = sizeof(dir->d);
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
dir->pos += sizeof(dir->d) + 4;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_bd_read(lfs, dir->pair[0], dir->off,
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
&entry->d, sizeof(entry->d));
if (err) {
return err;
}
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
entry->off = dir->off;
dir->off += 4+entry->d.elen+entry->d.alen+entry->d.nlen;
dir->pos += 4+entry->d.elen+entry->d.alen+entry->d.nlen;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return 0;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
static int lfs_dir_find(lfs_t *lfs, lfs_dir_t *dir,
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
lfs_entry_t *entry, const char **path) {
Added dir navigation without needing parent entries This should be the last step to removing the need for parent entries. Parent entries cause all sort of problems with atomic directory updates, especially related to moving/deleting directories. I couldn't figure out a parser for '..' entries without, O(n^2) runtime, a stack, or modifying the path itself. Since the goal is constant memory consumption, I went with the O(n^2) runtime solution, but this may need to be optimized later.
2017-04-14 23:27:06 +00:00
const char *pathname = *path;
size_t pathlen;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
while (true) {
Added dir navigation without needing parent entries This should be the last step to removing the need for parent entries. Parent entries cause all sort of problems with atomic directory updates, especially related to moving/deleting directories. I couldn't figure out a parser for '..' entries without, O(n^2) runtime, a stack, or modifying the path itself. Since the goal is constant memory consumption, I went with the O(n^2) runtime solution, but this may need to be optimized later.
2017-04-14 23:27:06 +00:00
nextname:
// skip slashes
pathname += strspn(pathname, "/");
pathlen = strcspn(pathname, "/");
// skip '.' and root '..'
if ((pathlen == 1 && memcmp(pathname, ".", 1) == 0) ||
(pathlen == 2 && memcmp(pathname, "..", 2) == 0)) {
pathname += pathlen;
goto nextname;
}
// skip if matched by '..' in name
const char *suffix = pathname + pathlen;
size_t sufflen;
int depth = 1;
while (true) {
suffix += strspn(suffix, "/");
sufflen = strcspn(suffix, "/");
if (sufflen == 0) {
break;
}
if (sufflen == 2 && memcmp(suffix, "..", 2) == 0) {
depth -= 1;
if (depth == 0) {
pathname = suffix + sufflen;
goto nextname;
}
} else {
depth += 1;
}
suffix += sufflen;
}
// find path
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
while (true) {
int err = lfs_dir_next(lfs, dir, entry);
if (err) {
return err;
}
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
if ((entry->d.type != LFS_TYPE_REG &&
entry->d.type != LFS_TYPE_DIR) ||
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
entry->d.nlen != pathlen) {
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
continue;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int res = lfs_bd_cmp(lfs, dir->pair[0],
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
entry->off + 4+entry->d.elen+entry->d.alen,
pathname, pathlen);
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if (res < 0) {
return res;
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// found match
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
if (res) {
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
break;
}
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
pathname += pathlen;
pathname += strspn(pathname, "/");
if (pathname[0] == '\0') {
return 0;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added dir navigation without needing parent entries This should be the last step to removing the need for parent entries. Parent entries cause all sort of problems with atomic directory updates, especially related to moving/deleting directories. I couldn't figure out a parser for '..' entries without, O(n^2) runtime, a stack, or modifying the path itself. Since the goal is constant memory consumption, I went with the O(n^2) runtime solution, but this may need to be optimized later.
2017-04-14 23:27:06 +00:00
// continue on if we hit a directory
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
if (entry->d.type != LFS_TYPE_DIR) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_NOTDIR;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
int err = lfs_dir_fetch(lfs, dir, entry->d.u.dir);
if (err) {
return err;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
*path = pathname;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added path iteration and chained directories All path iteration all goes through the lfs_dir_find function, which manages the syntax of paths and updates the path pointer to just the name stored in the dir entry. Also added directory chaining, which allows more than one block per directory. This is a simple linked list.
2017-04-01 15:09:17 +00:00
return 0;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
/// Top level directory operations ///
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
int lfs_mkdir(lfs_t *lfs, const char *path) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
// fetch parent directory
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
lfs_dir_t cwd;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
int err = lfs_dir_fetch(lfs, &cwd, lfs->root);
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
if (err) {
return err;
}
lfs_entry_t entry;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
err = lfs_dir_find(lfs, &cwd, &entry, &path);
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err != LFS_ERR_NOENT) {
return err ? err : LFS_ERR_EXISTS;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// build up new directory
lfs_alloc_ack(lfs);
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
lfs_dir_t dir;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
err = lfs_dir_alloc(lfs, &dir);
if (err) {
return err;
}
dir.d.tail[0] = cwd.d.tail[0];
dir.d.tail[1] = cwd.d.tail[1];
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
err = lfs_dir_commit(lfs, &dir, NULL, 0);
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
if (err) {
return err;
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
entry.d.type = LFS_TYPE_DIR;
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
entry.d.elen = sizeof(entry.d) - 4;
entry.d.alen = 0;
entry.d.nlen = strlen(path);
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
entry.d.u.dir[0] = dir.pair[0];
entry.d.u.dir[1] = dir.pair[1];
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
cwd.d.tail[0] = dir.pair[0];
cwd.d.tail[1] = dir.pair[1];
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
err = lfs_dir_append(lfs, &cwd, &entry, path);
if (err) {
return err;
}
lfs_alloc_ack(lfs);
return 0;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
int lfs_dir_open(lfs_t *lfs, lfs_dir_t *dir, const char *path) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
dir->pair[0] = lfs->root[0];
dir->pair[1] = lfs->root[1];
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
int err = lfs_dir_fetch(lfs, dir, dir->pair);
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
if (err) {
return err;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
if (strspn(path, "/.") == strlen(path)) {
// can only be something like '/././../.'
dir->head[0] = dir->pair[0];
dir->head[1] = dir->pair[1];
dir->pos = sizeof(dir->d) - 2;
dir->off = sizeof(dir->d);
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
return 0;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
lfs_entry_t entry;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
err = lfs_dir_find(lfs, dir, &entry, &path);
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
if (err) {
return err;
} else if (entry.d.type != LFS_TYPE_DIR) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_NOTDIR;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
err = lfs_dir_fetch(lfs, dir, entry.d.u.dir);
if (err) {
return err;
}
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
// setup head dir
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
// special offset for '.' and '..'
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
dir->head[0] = dir->pair[0];
dir->head[1] = dir->pair[1];
dir->pos = sizeof(dir->d) - 2;
dir->off = sizeof(dir->d);
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
return 0;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
}
int lfs_dir_close(lfs_t *lfs, lfs_dir_t *dir) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// do nothing, dir is always synchronized
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
return 0;
}
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
int lfs_dir_read(lfs_t *lfs, lfs_dir_t *dir, struct lfs_info *info) {
memset(info, 0, sizeof(*info));
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
// special offset for '.' and '..'
if (dir->pos == sizeof(dir->d) - 2) {
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
info->type = LFS_TYPE_DIR;
strcpy(info->name, ".");
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
dir->pos += 1;
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
return 1;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
} else if (dir->pos == sizeof(dir->d) - 1) {
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
info->type = LFS_TYPE_DIR;
strcpy(info->name, "..");
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
dir->pos += 1;
Removed .. and . entries No longer need to be stored on disk, can be simulated on the chip side. As mentioned in other commits, the parent entries had dozens of problems with atomic updates, as well as making everything just a bit more complex than is needed.
2017-04-15 16:26:37 +00:00
return 1;
}
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
lfs_entry_t entry;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
while (true) {
int err = lfs_dir_next(lfs, dir, &entry);
if (err) {
return (err == LFS_ERR_NOENT) ? 0 : err;
}
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
if (entry.d.type == LFS_TYPE_REG ||
entry.d.type == LFS_TYPE_DIR) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
break;
}
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
}
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
info->type = entry.d.type;
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
if (info->type == LFS_TYPE_REG) {
info->size = entry.d.u.file.size;
}
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
int err = lfs_bd_read(lfs, dir->pair[0],
entry.off + 4+entry.d.elen+entry.d.alen,
info->name, entry.d.nlen);
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
if (err) {
return err;
}
return 1;
}
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
int lfs_dir_seek(lfs_t *lfs, lfs_dir_t *dir, lfs_off_t off) {
// simply walk from head dir
int err = lfs_dir_rewind(lfs, dir);
if (err) {
return err;
}
dir->pos = off;
while (off > (0x7fffffff & dir->d.size)) {
off -= 0x7fffffff & dir->d.size;
if (!(0x80000000 & dir->d.size)) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
int err = lfs_dir_fetch(lfs, dir, dir->d.tail);
if (err) {
return err;
}
}
dir->off = off;
return 0;
}
lfs_soff_t lfs_dir_tell(lfs_t *lfs, lfs_dir_t *dir) {
return dir->pos;
}
int lfs_dir_rewind(lfs_t *lfs, lfs_dir_t *dir) {
// reload the head dir
int err = lfs_dir_fetch(lfs, dir, dir->head);
if (err) {
return err;
}
dir->pair[0] = dir->head[0];
dir->pair[1] = dir->head[1];
dir->pos = sizeof(dir->d) - 2;
dir->off = sizeof(dir->d);
return 0;
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
/// File index list operations ///
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
static int lfs_index(lfs_t *lfs, lfs_off_t *off) {
lfs_off_t i = 0;
Removed words variable from lfs struct
2017-04-30 18:12:44 +00:00
lfs_size_t words = lfs->cfg->block_size / 4;
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
while (*off >= lfs->cfg->block_size) {
i += 1;
*off -= lfs->cfg->block_size;
Removed words variable from lfs struct
2017-04-30 18:12:44 +00:00
*off += 4*lfs_min(lfs_ctz(i)+1, words-1);
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
}
return i;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
static int lfs_index_find(lfs_t *lfs,
lfs_cache_t *rcache, const lfs_cache_t *pcache,
lfs_block_t head, lfs_size_t size,
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
lfs_size_t pos, lfs_block_t *block, lfs_off_t *off) {
if (size == 0) {
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
*block = -1;
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
*off = 0;
return 0;
}
lfs_off_t current = lfs_index(lfs, &(lfs_off_t){size-1});
lfs_off_t target = lfs_index(lfs, &pos);
Removed words variable from lfs struct
2017-04-30 18:12:44 +00:00
lfs_size_t words = lfs->cfg->block_size / 4;
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
while (current > target) {
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
lfs_size_t skip = lfs_min(
lfs_npw2(current-target+1) - 1,
Removed words variable from lfs struct
2017-04-30 18:12:44 +00:00
lfs_min(lfs_ctz(current)+1, words-1) - 1);
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
int err = lfs_cache_read(lfs, rcache, pcache, head, 4*skip, &head, 4);
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
if (err) {
return err;
}
Fixed issue with cold-write after seek to block boundary This off-by-one error was caused by a slight difference between the pos argument to lfs_index_find and lfs_index_extend. When pos is on a block boundary, lfs_index_extend expects block to point before pos, as it would when writing a file linearly. But when seeking to that pos, the lfs_index_find to warm things up just supplies the block it expects pos to be in. Fixed the off-by-one error and added a test case for several of these cold seek+writes.
2017-09-17 17:53:18 +00:00
assert(head >= 2 && head <= lfs->cfg->block_count);
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
current -= 1 << skip;
}
*block = head;
*off = pos;
return 0;
}
static int lfs_index_extend(lfs_t *lfs,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_cache_t *rcache, lfs_cache_t *pcache,
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
lfs_block_t head, lfs_size_t size,
lfs_off_t *block, lfs_block_t *off) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
while (true) {
// go ahead and grab a block
int err = lfs_alloc(lfs, block);
if (err) {
return err;
}
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
err = lfs_bd_erase(lfs, *block);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (size == 0) {
*off = 0;
return 0;
}
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
size -= 1;
lfs_off_t index = lfs_index(lfs, &size);
size += 1;
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// just copy out the last block if it is incomplete
if (size != lfs->cfg->block_size) {
for (lfs_off_t i = 0; i < size; i++) {
uint8_t data;
int err = lfs_cache_read(lfs, rcache, NULL, head, i, &data, 1);
if (err) {
return err;
}
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
err = lfs_cache_prog(lfs, pcache, rcache, *block, i, &data, 1);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
*off = size;
return 0;
}
// append block
index += 1;
lfs_size_t words = lfs->cfg->block_size / 4;
lfs_size_t skips = lfs_min(lfs_ctz(index)+1, words-1);
for (lfs_off_t i = 0; i < skips; i++) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_cache_prog(lfs, pcache, rcache,
*block, 4*i, &head, 4);
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
if (err) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (i != skips-1) {
err = lfs_cache_read(lfs, rcache, NULL, head, 4*i, &head, 4);
if (err) {
return err;
}
}
Fixed issue with cold-write after seek to block boundary This off-by-one error was caused by a slight difference between the pos argument to lfs_index_find and lfs_index_extend. When pos is on a block boundary, lfs_index_extend expects block to point before pos, as it would when writing a file linearly. But when seeking to that pos, the lfs_index_find to warm things up just supplies the block it expects pos to be in. Fixed the off-by-one error and added a test case for several of these cold seek+writes.
2017-09-17 17:53:18 +00:00
assert(head >= 2 && head <= lfs->cfg->block_count);
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
*off = 4*skips;
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
return 0;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
relocate:
LFS_DEBUG("Bad block at %d", *block);
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// just clear cache and try a new block
pcache->block = 0xffffffff;
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
}
}
static int lfs_index_traverse(lfs_t *lfs,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_cache_t *rcache, const lfs_cache_t *pcache,
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
lfs_block_t head, lfs_size_t size,
int (*cb)(void*, lfs_block_t), void *data) {
if (size == 0) {
return 0;
}
lfs_off_t index = lfs_index(lfs, &(lfs_off_t){size-1});
while (true) {
int err = cb(data, head);
if (err) {
return err;
}
if (index == 0) {
return 0;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
err = lfs_cache_read(lfs, rcache, pcache, head, 0, &head, 4);
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
if (err) {
return err;
}
index -= 1;
}
return 0;
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
/// Top level file operations ///
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
int lfs_file_open(lfs_t *lfs, lfs_file_t *file,
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
const char *path, int flags) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// allocate entry for file if it doesn't exist
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
lfs_dir_t cwd;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
int err = lfs_dir_fetch(lfs, &cwd, lfs->root);
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
if (err) {
return err;
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
lfs_entry_t entry;
err = lfs_dir_find(lfs, &cwd, &entry, &path);
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err && err != LFS_ERR_NOENT) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
return err;
}
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err == LFS_ERR_NOENT) {
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
if (!(flags & LFS_O_CREAT)) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_NOENT;
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
// create entry to remember name
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
entry.d.type = LFS_TYPE_REG;
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
entry.d.elen = sizeof(entry.d) - 4;
entry.d.alen = 0;
entry.d.nlen = strlen(path);
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
entry.d.u.file.head = -1;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
entry.d.u.file.size = 0;
err = lfs_dir_append(lfs, &cwd, &entry, path);
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
if (err) {
return err;
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
} else if (entry.d.type == LFS_TYPE_DIR) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_ISDIR;
Removed a layer of indirection for index-list lookup Files are now stored directly in the index-list, instead of being referenced by pointers that used to live there. This somewhat reduces the complexity around handling files, while still keeping the O(logn) lookup cost.
2017-04-23 00:48:31 +00:00
} else if (flags & LFS_O_EXCL) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_EXISTS;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// setup file struct
file->pair[0] = cwd.pair[0];
file->pair[1] = cwd.pair[1];
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->poff = entry.off;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
file->head = entry.d.u.file.head;
file->size = entry.d.u.file.size;
file->flags = flags;
file->pos = 0;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
if (flags & LFS_O_TRUNC) {
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
file->head = -1;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
file->size = 0;
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
// allocate buffer if needed
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
file->cache.block = 0xffffffff;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (lfs->cfg->file_buffer) {
file->cache.buffer = lfs->cfg->file_buffer;
} else if ((file->flags & 3) == LFS_O_RDONLY) {
file->cache.buffer = malloc(lfs->cfg->read_size);
if (!file->cache.buffer) {
return LFS_ERR_NOMEM;
}
} else {
file->cache.buffer = malloc(lfs->cfg->prog_size);
if (!file->cache.buffer) {
return LFS_ERR_NOMEM;
}
}
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
// add to list of files
file->next = lfs->files;
lfs->files = file;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
return 0;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
}
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
int lfs_file_close(lfs_t *lfs, lfs_file_t *file) {
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
int err = lfs_file_sync(lfs, file);
// remove from list of files
for (lfs_file_t **p = &lfs->files; *p; p = &(*p)->next) {
if (*p == file) {
*p = file->next;
break;
}
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
// clean up memory
if (!lfs->cfg->file_buffer) {
free(file->cache.buffer);
}
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
return err;
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Fixed relocation bug when a file is closed with lingering caches This bug required larger cache sizes to notice, since block errors usually get detected in the early stages of writing to files. Since the fix here requires both lfs_file_write and lfs_file_flush relocate file blocks, the relocation logic was moved out into a seperate function.
2017-06-25 19:01:33 +00:00
static int lfs_file_relocate(lfs_t *lfs, lfs_file_t *file) {
relocate:
LFS_DEBUG("Bad block at %d", file->block);
// just relocate what exists into new block
lfs_block_t nblock;
int err = lfs_alloc(lfs, &nblock);
if (err) {
return err;
}
Fixed missing erase during file relocation This was an easy fix, but highlighted the fact that the current testing framework doesn't detect when a block is written to without an associated erase. Added a quick solution that creates an empty file during an erase.
2017-06-25 21:56:12 +00:00
err = lfs_bd_erase(lfs, nblock);
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
Fixed relocation bug when a file is closed with lingering caches This bug required larger cache sizes to notice, since block errors usually get detected in the early stages of writing to files. Since the fix here requires both lfs_file_write and lfs_file_flush relocate file blocks, the relocation logic was moved out into a seperate function.
2017-06-25 19:01:33 +00:00
// either read from dirty cache or disk
for (lfs_off_t i = 0; i < file->off; i++) {
uint8_t data;
Adopted redundant cache read in lfs_file_relocate Previously had some custom logic that could be reduced
2017-06-25 22:23:40 +00:00
err = lfs_cache_read(lfs, &lfs->rcache, &file->cache,
file->block, i, &data, 1);
if (err) {
return err;
Fixed relocation bug when a file is closed with lingering caches This bug required larger cache sizes to notice, since block errors usually get detected in the early stages of writing to files. Since the fix here requires both lfs_file_write and lfs_file_flush relocate file blocks, the relocation logic was moved out into a seperate function.
2017-06-25 19:01:33 +00:00
}
err = lfs_cache_prog(lfs, &lfs->pcache, &lfs->rcache,
nblock, i, &data, 1);
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
}
// copy over new state of file
memcpy(file->cache.buffer, lfs->pcache.buffer, lfs->cfg->prog_size);
file->cache.block = lfs->pcache.block;
file->cache.off = lfs->pcache.off;
lfs->pcache.block = 0xffffffff;
file->block = nblock;
return 0;
}
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
static int lfs_file_flush(lfs_t *lfs, lfs_file_t *file) {
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (file->flags & LFS_F_READING) {
// just drop read cache
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
file->cache.block = 0xffffffff;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->flags &= ~LFS_F_READING;
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (file->flags & LFS_F_WRITING) {
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
lfs_off_t pos = file->pos;
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// copy over anything after current branch
lfs_file_t orig = {
.head = file->head,
.size = file->size,
.flags = LFS_O_RDONLY,
.pos = file->pos,
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
.cache = lfs->rcache,
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
};
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
lfs->rcache.block = 0xffffffff;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
while (file->pos < file->size) {
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
// copy over a byte at a time, leave it up to caching
// to make this efficient
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
uint8_t data;
lfs_ssize_t res = lfs_file_read(lfs, &orig, &data, 1);
if (res < 0) {
return res;
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
res = lfs_file_write(lfs, file, &data, 1);
if (res < 0) {
return res;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
// keep our reference to the rcache in sync
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
if (lfs->rcache.block != 0xffffffff) {
orig.cache.block = 0xffffffff;
lfs->rcache.block = 0xffffffff;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
}
}
// write out what we have
Fixed relocation bug when a file is closed with lingering caches This bug required larger cache sizes to notice, since block errors usually get detected in the early stages of writing to files. Since the fix here requires both lfs_file_write and lfs_file_flush relocate file blocks, the relocation logic was moved out into a seperate function.
2017-06-25 19:01:33 +00:00
while (true) {
int err = lfs_cache_flush(lfs, &file->cache, &lfs->rcache);
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
break;
relocate:
err = lfs_file_relocate(lfs, file);
if (err) {
return err;
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// actual file updates
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->head = file->block;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
file->size = file->pos;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->flags &= ~LFS_F_WRITING;
file->flags |= LFS_F_DIRTY;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
file->pos = pos;
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
return 0;
}
int lfs_file_sync(lfs_t *lfs, lfs_file_t *file) {
int err = lfs_file_flush(lfs, file);
if (err) {
return err;
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
Added proper handling for removing open files Conveniently we previously added a linked-list of files for things like this. This should handle most of the corner cases where files are open during strange operations. This also brings up the point that we aren't doing anything similar for directories and don't even have a dir linked-list. After thinking about it for a while, I've decided to leave out this handling for dirs. It will likely be very complicated, with little gains as directories are less used in embedded systems. Additionally, dirs are only open for reading, and corruption will probably just cause the dir iteration to terminate. If needed, correct handling of open directories can be added later.
2017-04-30 17:35:50 +00:00
if ((file->flags & LFS_F_DIRTY) && !lfs_pairisnull(file->pair)) {
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// update dir entry
lfs_dir_t cwd;
int err = lfs_dir_fetch(lfs, &cwd, file->pair);
if (err) {
return err;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_entry_t entry = {.off = file->poff};
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
err = lfs_bd_read(lfs, cwd.pair[0], entry.off,
Adopted more conventional buffer parameter ordering Adopted buffer followed by size. The other order was original chosen due to some other functions with a more complicated parameter list. This convention is important, as the bd api is one of the main apis facing porting efforts.
2017-04-24 04:49:21 +00:00
&entry.d, sizeof(entry.d));
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
if (err) {
return err;
}
if (entry.d.type != LFS_TYPE_REG) {
// sanity check valid entry
return LFS_ERR_INVAL;
}
entry.d.u.file.head = file->head;
entry.d.u.file.size = file->size;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
err = lfs_dir_update(lfs, &cwd, &entry, NULL);
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
if (err) {
return err;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->flags &= ~LFS_F_DIRTY;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
return 0;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
}
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
lfs_ssize_t lfs_file_read(lfs_t *lfs, lfs_file_t *file,
void *buffer, lfs_size_t size) {
uint8_t *data = buffer;
lfs_size_t nsize = size;
if ((file->flags & 3) == LFS_O_WRONLY) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (file->flags & LFS_F_WRITING) {
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// flush out any writes
int err = lfs_file_flush(lfs, file);
if (err) {
return err;
}
}
size = lfs_min(size, file->size - file->pos);
nsize = size;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
while (nsize > 0) {
// check if we need a new block
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (!(file->flags & LFS_F_READING) ||
file->off == lfs->cfg->block_size) {
int err = lfs_index_find(lfs, &file->cache, NULL,
file->head, file->size,
file->pos, &file->block, &file->off);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
if (err) {
return err;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->flags |= LFS_F_READING;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
}
// read as much as we can in current block
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_size_t diff = lfs_min(nsize, lfs->cfg->block_size - file->off);
int err = lfs_cache_read(lfs, &file->cache, NULL,
file->block, file->off, data, diff);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
if (err) {
return err;
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
file->pos += diff;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->off += diff;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
data += diff;
nsize -= diff;
}
return size;
}
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
lfs_ssize_t lfs_file_write(lfs_t *lfs, lfs_file_t *file,
const void *buffer, lfs_size_t size) {
const uint8_t *data = buffer;
lfs_size_t nsize = size;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
if ((file->flags & 3) == LFS_O_RDONLY) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (file->flags & LFS_F_READING) {
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// drop any reads
int err = lfs_file_flush(lfs, file);
if (err) {
return err;
}
}
if ((file->flags & LFS_O_APPEND) && file->pos < file->size) {
file->pos = file->size;
}
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
while (nsize > 0) {
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
// check if we need a new block
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (!(file->flags & LFS_F_WRITING) ||
file->off == lfs->cfg->block_size) {
Fixed issue with cold-write after seek to block boundary This off-by-one error was caused by a slight difference between the pos argument to lfs_index_find and lfs_index_extend. When pos is on a block boundary, lfs_index_extend expects block to point before pos, as it would when writing a file linearly. But when seeking to that pos, the lfs_index_find to warm things up just supplies the block it expects pos to be in. Fixed the off-by-one error and added a test case for several of these cold seek+writes.
2017-09-17 17:53:18 +00:00
if (!(file->flags & LFS_F_WRITING) && file->pos > 0) {
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
// find out which block we're extending from
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
int err = lfs_index_find(lfs, &file->cache, NULL,
file->head, file->size,
Fixed issue with cold-write after seek to block boundary This off-by-one error was caused by a slight difference between the pos argument to lfs_index_find and lfs_index_extend. When pos is on a block boundary, lfs_index_extend expects block to point before pos, as it would when writing a file linearly. But when seeking to that pos, the lfs_index_find to warm things up just supplies the block it expects pos to be in. Fixed the off-by-one error and added a test case for several of these cold seek+writes.
2017-09-17 17:53:18 +00:00
file->pos-1, &file->block, &file->off);
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
if (err) {
return err;
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
// mark cache as dirty since we may have read data into it
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
file->cache.block = 0xffffffff;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
// extend file with new blocks
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs_alloc_ack(lfs);
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
int err = lfs_index_extend(lfs, &lfs->rcache, &file->cache,
file->block, file->pos,
&file->block, &file->off);
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
if (err) {
return err;
}
Fixed issue with cold-write after seek to block boundary This off-by-one error was caused by a slight difference between the pos argument to lfs_index_find and lfs_index_extend. When pos is on a block boundary, lfs_index_extend expects block to point before pos, as it would when writing a file linearly. But when seeking to that pos, the lfs_index_find to warm things up just supplies the block it expects pos to be in. Fixed the off-by-one error and added a test case for several of these cold seek+writes.
2017-09-17 17:53:18 +00:00
file->flags |= LFS_F_WRITING;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
}
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
// program as much as we can in current block
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
lfs_size_t diff = lfs_min(nsize, lfs->cfg->block_size - file->off);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
while (true) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_cache_prog(lfs, &file->cache, &lfs->rcache,
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
file->block, file->off, data, diff);
if (err) {
if (err == LFS_ERR_CORRUPT) {
goto relocate;
}
return err;
}
break;
relocate:
Fixed relocation bug when a file is closed with lingering caches This bug required larger cache sizes to notice, since block errors usually get detected in the early stages of writing to files. Since the fix here requires both lfs_file_write and lfs_file_flush relocate file blocks, the relocation logic was moved out into a seperate function.
2017-06-25 19:01:33 +00:00
err = lfs_file_relocate(lfs, file);
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (err) {
return err;
}
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
file->pos += diff;
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
file->off += diff;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
data += diff;
nsize -= diff;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs_alloc_ack(lfs);
Added correct handling of file syncing around overwrites Now all of the open flags are correctly handled Even annoying cases where we can't trust the blocks that are already on file, such as appending existing files and writing to the middle of files.
2017-04-23 02:42:22 +00:00
}
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
return size;
}
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
lfs_soff_t lfs_file_seek(lfs_t *lfs, lfs_file_t *file,
lfs_soff_t off, int whence) {
// write out everything beforehand, may be noop if rdonly
int err = lfs_file_flush(lfs, file);
if (err) {
return err;
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
// update pos
lfs_off_t pos = file->pos;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
if (whence == LFS_SEEK_SET) {
file->pos = off;
} else if (whence == LFS_SEEK_CUR) {
file->pos = file->pos + off;
} else if (whence == LFS_SEEK_END) {
file->pos = file->size + off;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
return pos;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
lfs_soff_t lfs_file_tell(lfs_t *lfs, lfs_file_t *file) {
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
return file->pos;
Added support for full seek operations A rather involved upgrade for both files and directories, seek and related functions are now completely supported: - lfs_file_seek - lfs_file_tell - lfs_file_rewind - lfs_file_size - lfs_dir_seek - lfs_dir_tell - lfs_dir_rewind This change also highlighted the concern that lfs_off_t is unsigned, whereas off_t is traditionally signed. Unfortunately, lfs_off_t is already used intensively through the codebase, so in focusing on moving forward and avoiding getting bogged down by details, I'm going to keep it as is and use the signed type lfs_soff_t where necessary.
2017-04-23 04:11:13 +00:00
}
int lfs_file_rewind(lfs_t *lfs, lfs_file_t *file) {
lfs_soff_t res = lfs_file_seek(lfs, file, 0, LFS_SEEK_SET);
if (res < 0) {
return res;
}
return 0;
}
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
lfs_soff_t lfs_file_size(lfs_t *lfs, lfs_file_t *file) {
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
return lfs_max(file->pos, file->size);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
}
/// General fs oprations ///
int lfs_stat(lfs_t *lfs, const char *path, struct lfs_info *info) {
lfs_dir_t cwd;
int err = lfs_dir_fetch(lfs, &cwd, lfs->root);
if (err) {
return err;
}
lfs_entry_t entry;
err = lfs_dir_find(lfs, &cwd, &entry, &path);
if (err) {
return err;
}
memset(info, 0, sizeof(*info));
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
info->type = entry.d.type;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
if (info->type == LFS_TYPE_REG) {
info->size = entry.d.u.file.size;
}
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
err = lfs_bd_read(lfs, cwd.pair[0],
entry.off + 4+entry.d.elen+entry.d.alen,
info->name, entry.d.nlen);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
if (err) {
return err;
}
return 0;
}
int lfs_remove(lfs_t *lfs, const char *path) {
lfs_dir_t cwd;
int err = lfs_dir_fetch(lfs, &cwd, lfs->root);
if (err) {
return err;
}
lfs_entry_t entry;
err = lfs_dir_find(lfs, &cwd, &entry, &path);
if (err) {
return err;
}
lfs_dir_t dir;
if (entry.d.type == LFS_TYPE_DIR) {
// must be empty before removal, checking size
// without masking top bit checks for any case where
// dir is not empty
int err = lfs_dir_fetch(lfs, &dir, entry.d.u.dir);
if (err) {
return err;
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
} else if (dir.d.size != sizeof(dir.d)+4) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
}
}
// remove the entry
err = lfs_dir_remove(lfs, &cwd, &entry);
if (err) {
return err;
}
Added proper handling for removing open files Conveniently we previously added a linked-list of files for things like this. This should handle most of the corner cases where files are open during strange operations. This also brings up the point that we aren't doing anything similar for directories and don't even have a dir linked-list. After thinking about it for a while, I've decided to leave out this handling for dirs. It will likely be very complicated, with little gains as directories are less used in embedded systems. Additionally, dirs are only open for reading, and corruption will probably just cause the dir iteration to terminate. If needed, correct handling of open directories can be added later.
2017-04-30 17:35:50 +00:00
// shift over any files that are affected
for (lfs_file_t *f = lfs->files; f; f = f->next) {
if (lfs_paircmp(f->pair, cwd.pair) == 0) {
if (f->poff == entry.off) {
f->pair[0] = 0xffffffff;
f->pair[1] = 0xffffffff;
} else if (f->poff > entry.off) {
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
f->poff -= 4 + entry.d.elen + entry.d.alen + entry.d.nlen;
Added proper handling for removing open files Conveniently we previously added a linked-list of files for things like this. This should handle most of the corner cases where files are open during strange operations. This also brings up the point that we aren't doing anything similar for directories and don't even have a dir linked-list. After thinking about it for a while, I've decided to leave out this handling for dirs. It will likely be very complicated, with little gains as directories are less used in embedded systems. Additionally, dirs are only open for reading, and corruption will probably just cause the dir iteration to terminate. If needed, correct handling of open directories can be added later.
2017-04-30 17:35:50 +00:00
}
}
}
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
// if we were a directory, just run a deorphan step, this should
// collect us, although is expensive
if (entry.d.type == LFS_TYPE_DIR) {
int err = lfs_deorphan(lfs);
if (err) {
return err;
}
}
return 0;
}
int lfs_rename(lfs_t *lfs, const char *oldpath, const char *newpath) {
// find old entry
lfs_dir_t oldcwd;
int err = lfs_dir_fetch(lfs, &oldcwd, lfs->root);
if (err) {
return err;
}
lfs_entry_t oldentry;
err = lfs_dir_find(lfs, &oldcwd, &oldentry, &oldpath);
if (err) {
return err;
}
// allocate new entry
lfs_dir_t newcwd;
err = lfs_dir_fetch(lfs, &newcwd, lfs->root);
if (err) {
return err;
}
lfs_entry_t preventry;
err = lfs_dir_find(lfs, &newcwd, &preventry, &newpath);
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err && err != LFS_ERR_NOENT) {
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
return err;
}
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
bool prevexists = (err != LFS_ERR_NOENT);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
// must have same type
if (prevexists && preventry.d.type != oldentry.d.type) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
}
lfs_dir_t dir;
if (prevexists && preventry.d.type == LFS_TYPE_DIR) {
// must be empty before removal, checking size
// without masking top bit checks for any case where
// dir is not empty
int err = lfs_dir_fetch(lfs, &dir, preventry.d.u.dir);
if (err) {
return err;
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
} else if (dir.d.size != sizeof(dir.d)+4) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
}
}
// move to new location
lfs_entry_t newentry = preventry;
newentry.d = oldentry.d;
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
newentry.d.nlen = strlen(newpath);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
if (prevexists) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
int err = lfs_dir_update(lfs, &newcwd, &newentry, newpath);
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
if (err) {
return err;
}
} else {
int err = lfs_dir_append(lfs, &newcwd, &newentry, newpath);
if (err) {
return err;
}
}
// fetch again in case newcwd == oldcwd
err = lfs_dir_fetch(lfs, &oldcwd, oldcwd.pair);
if (err) {
return err;
}
err = lfs_dir_find(lfs, &oldcwd, &oldentry, &oldpath);
if (err) {
return err;
}
// remove from old location
err = lfs_dir_remove(lfs, &oldcwd, &oldentry);
if (err) {
return err;
}
Added proper handling for removing open files Conveniently we previously added a linked-list of files for things like this. This should handle most of the corner cases where files are open during strange operations. This also brings up the point that we aren't doing anything similar for directories and don't even have a dir linked-list. After thinking about it for a while, I've decided to leave out this handling for dirs. It will likely be very complicated, with little gains as directories are less used in embedded systems. Additionally, dirs are only open for reading, and corruption will probably just cause the dir iteration to terminate. If needed, correct handling of open directories can be added later.
2017-04-30 17:35:50 +00:00
// shift over any files that are affected
for (lfs_file_t *f = lfs->files; f; f = f->next) {
if (lfs_paircmp(f->pair, oldcwd.pair) == 0) {
if (f->poff == oldentry.off) {
f->pair[0] = 0xffffffff;
f->pair[1] = 0xffffffff;
} else if (f->poff > oldentry.off) {
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
f->poff -= 4+oldentry.d.elen+oldentry.d.alen+oldentry.d.nlen;
Added proper handling for removing open files Conveniently we previously added a linked-list of files for things like this. This should handle most of the corner cases where files are open during strange operations. This also brings up the point that we aren't doing anything similar for directories and don't even have a dir linked-list. After thinking about it for a while, I've decided to leave out this handling for dirs. It will likely be very complicated, with little gains as directories are less used in embedded systems. Additionally, dirs are only open for reading, and corruption will probably just cause the dir iteration to terminate. If needed, correct handling of open directories can be added later.
2017-04-30 17:35:50 +00:00
}
}
}
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
// if we were a directory, just run a deorphan step, this should
// collect us, although is expensive
if (prevexists && preventry.d.type == LFS_TYPE_DIR) {
int err = lfs_deorphan(lfs);
if (err) {
return err;
}
}
return 0;
}
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
/// Filesystem operations ///
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
static int lfs_init(lfs_t *lfs, const struct lfs_config *cfg) {
lfs->cfg = cfg;
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
// setup read cache
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
lfs->rcache.block = 0xffffffff;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (lfs->cfg->read_buffer) {
lfs->rcache.buffer = lfs->cfg->read_buffer;
} else {
lfs->rcache.buffer = malloc(lfs->cfg->read_size);
if (!lfs->rcache.buffer) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_NOMEM;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
}
}
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
// setup program cache
Adopted 0xffffffff as null pointer - Default value of most flash-based storage - Avoids 0 == superblock/dir issue - Usually causes assertions in bd driver layer - Easier to notice in hex dumps
2017-04-30 16:54:27 +00:00
lfs->pcache.block = 0xffffffff;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (lfs->cfg->prog_buffer) {
lfs->pcache.buffer = lfs->cfg->prog_buffer;
} else {
lfs->pcache.buffer = malloc(lfs->cfg->prog_size);
if (!lfs->pcache.buffer) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_NOMEM;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
}
}
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
// setup lookahead
if (lfs->cfg->lookahead_buffer) {
lfs->free.lookahead = lfs->cfg->lookahead_buffer;
} else {
lfs->free.lookahead = malloc(lfs->cfg->lookahead/8);
if (!lfs->free.lookahead) {
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_NOMEM;
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
}
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// setup default state
lfs->root[0] = 0xffffffff;
lfs->root[1] = 0xffffffff;
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
lfs->files = NULL;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs->deorphaned = false;
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
return 0;
}
static int lfs_deinit(lfs_t *lfs) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// free allocated memory
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
if (!lfs->cfg->read_buffer) {
free(lfs->rcache.buffer);
}
if (!lfs->cfg->prog_buffer) {
free(lfs->pcache.buffer);
}
Fixed memory leak for lookahead buffer
2017-04-29 17:50:23 +00:00
if (!lfs->cfg->lookahead_buffer) {
free(lfs->free.lookahead);
}
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
return 0;
}
int lfs_format(lfs_t *lfs, const struct lfs_config *cfg) {
int err = lfs_init(lfs, cfg);
if (err) {
return err;
}
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// create free lookahead
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
memset(lfs->free.lookahead, 0, lfs->cfg->lookahead/8);
lfs->free.start = 0;
lfs->free.off = 0;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs->free.end = lfs->free.start + lfs->cfg->block_count;
Added support for the basic file operation Missing seek, but these are the core filesystem operations provided by this filesystem: - Read a file - Append to a file Additional work is needed around freeing the previous file, so right now it's limited to appending to existing files, a real append only filesystem. Unfortunately the overhead of the free list with multiple open files is becoming tricky.
2017-03-20 03:00:56 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// create superblock dir
lfs_alloc_ack(lfs);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
lfs_dir_t superdir;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
err = lfs_dir_alloc(lfs, &superdir);
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (err) {
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// write root directory
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
lfs_dir_t root;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
err = lfs_dir_alloc(lfs, &root);
if (err) {
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
err = lfs_dir_commit(lfs, &root, NULL, 0);
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
if (err) {
return err;
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
lfs->root[0] = root.pair[0];
lfs->root[1] = root.pair[1];
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// write superblocks
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
lfs_superblock_t superblock = {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
.off = sizeof(superdir.d),
.d.type = LFS_TYPE_SUPERBLOCK,
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
.d.elen = sizeof(superblock.d) - sizeof(superblock.d.magic) - 4,
.d.nlen = sizeof(superblock.d.magic),
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
.d.version = 0x00010001,
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
.d.magic = {"littlefs"},
Simplified config Before, the lfs had multiple paths to determine config options: - lfs_config struct passed during initialization - lfs_bd_info struct passed during block device initialization - compile time options This allowed different developers to provide their own needs to the filesystem, such as the block device capabilities and the higher level user's own tweaks. However, this comes with additional complexity and action required when the configurations are incompatible. For now, this has been reduced to all information (including block device function pointers) being passed through the lfs_config struct. We just defer more complicated handling of configuration options to the top level user. This simplifies configuration handling and gives the top level user the responsibility to handle configuration, which they probably would have wanted to do anyways.
2017-04-22 16:42:05 +00:00
.d.block_size = lfs->cfg->block_size,
.d.block_count = lfs->cfg->block_count,
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
.d.root = {lfs->root[0], lfs->root[1]},
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
};
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
superdir.d.tail[0] = root.pair[0];
superdir.d.tail[1] = root.pair[1];
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
superdir.d.size = sizeof(superdir.d) + sizeof(superblock.d) + 4;
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// write both pairs to be safe
bool valid = false;
Restructured the major interfaces of the filesystem
2017-03-25 21:20:31 +00:00
for (int i = 0; i < 2; i++) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
int err = lfs_dir_commit(lfs, &superdir, (struct lfs_region[]){
{sizeof(superdir.d), sizeof(superblock.d),
&superblock.d, sizeof(superblock.d)}
}, 1);
if (err && err != LFS_ERR_CORRUPT) {
Added initial superblock definition Really started working out how the internal structure of the driver will be organized. There are a few hazy lines between the intended data structures with the goal of code reuse, so the function boundaries may end up a bit weird.
2017-03-05 20:11:52 +00:00
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
valid = valid || !err;
}
if (!valid) {
return LFS_ERR_CORRUPT;
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
}
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
// sanity check that fetch works
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
err = lfs_dir_fetch(lfs, &superdir, (const lfs_block_t[2]){0, 1});
if (err) {
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs_alloc_ack(lfs);
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
return lfs_deinit(lfs);
Initial commit of progress, minimal formatting niave free list
2017-02-27 00:05:27 +00:00
}
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
int lfs_mount(lfs_t *lfs, const struct lfs_config *cfg) {
int err = lfs_init(lfs, cfg);
if (err) {
return err;
}
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
// setup free lookahead
lfs->free.start = -lfs->cfg->lookahead;
lfs->free.off = lfs->cfg->lookahead;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs->free.end = lfs->free.start + lfs->cfg->block_count;
Cleaned up block allocator Removed scanning for stride - Adds complexity with questionable benefit - Can be added as an optimization later Fixed handling around device boundaries and where lookahead may not be a factor of the device size (consider small devices with only a few blocks) Added support for configuration with optional dynamic memory as found in the caching configuration
2017-04-22 19:56:12 +00:00
// load superblock
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
lfs_dir_t dir;
lfs_superblock_t superblock;
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
err = lfs_dir_fetch(lfs, &dir, (const lfs_block_t[2]){0, 1});
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
if (!err) {
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
err = lfs_bd_read(lfs, dir.pair[0], sizeof(dir.d),
&superblock.d, sizeof(superblock.d));
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
lfs->root[0] = superblock.d.root[0];
lfs->root[1] = superblock.d.root[1];
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err == LFS_ERR_CORRUPT ||
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
memcmp(superblock.d.magic, "littlefs", 8) != 0) {
LFS_ERROR("Invalid superblock at %d %d", dir.pair[0], dir.pair[1]);
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_CORRUPT;
Added limited support for directories This comes with a lot of scafolding put into place around the core of the filesystem. Added operations: - append an entry to a directory - find an entry in a directory - iterate over entries in a directory Some to do: - Chaining multiple directory blocks - Recursion on directory operations
2017-03-13 00:41:08 +00:00
}
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
if (superblock.d.version > (0x00010001 | 0x0000ffff)) {
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
LFS_ERROR("Invalid version %d.%d\n",
0xffff & (superblock.d.version >> 16),
0xffff & (superblock.d.version >> 0));
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
return LFS_ERR_INVAL;
Restructured directory code After quite a bit of prototyping, settled on the following functions: - lfs_dir_alloc - create a new dir - lfs_dir_fetch - load and check a dir pair from disk - lfs_dir_commit - save a dir pair to disk - lfs_dir_shift - shrink a dir pair to disk - lfs_dir_append - add a dir entry, creating dirs if needed - lfs_dir_remove - remove a dir entry, dropping dirs if needed Additionally, followed through with a few other tweaks
2017-04-18 03:27:06 +00:00
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
return err;
}
int lfs_unmount(lfs_t *lfs) {
Added optional block-level caching This adds caching of the most recent read/program blocks, allowing support of devices that don't have byte-level read+writes, along with reduced device access on devices that do support byte-level read+writes. Note: The current implementation is a bit eager to drop caches where it simplifies the cache layer. This layer is already complex enough. Note: It may be worthwhile to add a compile switch for caching to reduce code size, note sure. Note: This does add a dependency on malloc, which could have a porting layer, but I'm just using the functions from stdlib for now. These can be overwritten with noops if the user controls the system, and keeps things simple for now.
2017-04-22 18:30:40 +00:00
return lfs_deinit(lfs);
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
Structured some of the bulk of the codebase - Removed lfs_config.h, distributed between lfs.h and lfs_util.h - Moved some functions that felt out of place
2017-04-24 02:40:03 +00:00
/// Littlefs specific operations ///
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
int lfs_traverse(lfs_t *lfs, int (*cb)(void*, lfs_block_t), void *data) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (lfs_pairisnull(lfs->root)) {
return 0;
}
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
// iterate over metadata pairs
Added dir tests, test fixes, config
2017-03-25 23:11:45 +00:00
lfs_dir_t dir;
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
lfs_entry_t entry;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
lfs_block_t cwd[2] = {0, 1};
while (true) {
for (int i = 0; i < 2; i++) {
int err = cb(data, cwd[i]);
if (err) {
return err;
}
}
int err = lfs_dir_fetch(lfs, &dir, cwd);
if (err) {
return err;
}
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
// iterate over contents
Shrinked on-disk directory program size Directories still consume two full erase blocks, but now only program the exact on-disk region to store the directory contents. This results in a decent improvement in the amount of data written and read to the device when doing directory operations. Calculating the checksum of dynamically sized data is surprisingly tricky, since the size of the data could also contain errors. For the littlefs, we can assume the data size must fit in an erase block. If the data size is invalid, we can just treat the block as corrupted.
2017-06-24 01:03:44 +00:00
while (dir.off + sizeof(entry.d) <= (0x7fffffff & dir.d.size)-4) {
Added internal check of data written to disk Before, the littlefs relied on the underlying block device to report corruption that occurs when writing data to disk. This requirement is easy to miss or implement incorrectly, since the error detection is only required when a block becomes corrupted, which is very unlikely to happen until late in the block device's lifetime. The littlefs can detect corruption itself by reading back written data. This requires a bit of care to reuse the available buffers, and may rely on checksums to avoid additional RAM requirements. This does have a runtime penalty with the extra read operations, but should make the littlefs much more robust to different implementations.
2017-06-24 05:43:05 +00:00
int err = lfs_bd_read(lfs, dir.pair[0], dir.off,
Adopted more conventional buffer parameter ordering Adopted buffer followed by size. The other order was original chosen due to some other functions with a more complicated parameter list. This convention is important, as the bd api is one of the main apis facing porting efforts.
2017-04-24 04:49:21 +00:00
&entry.d, sizeof(entry.d));
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
if (err) {
return err;
}
Extended entry tag to support attributes Zero attributes are actually supported at the moment, but this change will allow entry attribute to be added in a backwards compatible manner. Each dir entry is now prefixed with a 32 bit tag: 4b - entry type 4b - data structure 8b - entry len 8b - attribute len 8b - name len A full entry on disk looks a bit like this: [- 8 -|- 8 -|- 8 -|- 8 -|-- elen --|-- alen --|-- nlen --] [ type | elen | alen | nlen | entry | attrs | name ] The actually contents of the attributes section is a bit handwavey until the first attributes are implemented, but to put plans in place: Each attribute will be prefixed with only a byte that indicates the type of attribute. Attributes should be sorted based on portability, since unknown attributes will force attribute parsing to stop.
2017-07-18 07:09:35 +00:00
dir.off += 4+entry.d.elen+entry.d.alen+entry.d.nlen;
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
if ((0xf & entry.d.type) == (0xf & LFS_TYPE_REG)) {
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
int err = lfs_index_traverse(lfs, &lfs->rcache, NULL,
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
entry.d.u.file.head, entry.d.u.file.size, cb, data);
Fixed non-standard behaviour of rdwr streams Originally had two seperate positions for reading/writing, but this is inconsistent with the the posix standard, which has a single position for reading and writing. Also added proper handling of when the file is dirty, just added an internal flag for this state. Also moved the entry out of the file struct, and rearranged some members to clean things up.
2017-04-24 04:39:50 +00:00
if (err) {
return err;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
}
}
cwd[0] = dir.d.tail[0];
cwd[1] = dir.d.tail[1];
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
if (lfs_pairisnull(cwd)) {
break;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
}
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
// iterate over any open files
for (lfs_file_t *f = lfs->files; f; f = f->next) {
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (f->flags & LFS_F_DIRTY) {
int err = lfs_index_traverse(lfs, &lfs->rcache, &f->cache,
f->head, f->size, cb, data);
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
if (err) {
return err;
}
}
Added caching with managed caches at the file level This adds a fully independent layer between the rest of the filesystem and the block device. This requires some additionally logic around cache invalidation and flushing, but removes the need for any higher layer to consider read/write sizes less than what is supported by the hardware. Additionally, these caches can be used for possible speed improvements. This is left up to the user to optimize for their use cases. For very limited embedded systems with byte-level read/writes, the caches could be omitted completely, or they could even be the size of a full block for minimizing storage access. (A full block may not be the best for speed, consider if only a small portion of the read block is used, but I'll leave that evaluation as an exercise for any consumers of this library)
2017-04-30 16:19:37 +00:00
if (f->flags & LFS_F_WRITING) {
int err = lfs_index_traverse(lfs, &lfs->rcache, &f->cache,
f->block, f->pos, cb, data);
Added file list for tracking in flight allocations Needed primarily for tracking block allocations, unfortunately this prevents the freedom for the user to bitwise copy files.
2017-04-29 15:22:01 +00:00
if (err) {
return err;
}
}
}
return 0;
Added full dir list and rudimentary block allocator In writing the initial allocator, I ran into the rather difficult problem of trying to iterate through the entire filesystem cheaply and with only constant memory consumption (which prohibits recursive functions). The solution was to simply thread all directory blocks onto a massive linked-list that spans the entire filesystem. With the linked-list it was easy to create a traverse function for all blocks in use on the filesystem (which has potential for other utility), and add the rudimentary block allocator using a bit-vector. While the linked-list may add complexity (especially where needing to maintain atomic operations), the linked-list helps simplify what is currently the most expensive operation in the filesystem, with no cost to space (the linked-list can reuse the pointers used for chained directory blocks).
2017-04-01 15:44:17 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
static int lfs_pred(lfs_t *lfs, const lfs_block_t dir[2], lfs_dir_t *pdir) {
if (lfs_pairisnull(lfs->root)) {
return 0;
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
// iterate over all directory directory entries
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
int err = lfs_dir_fetch(lfs, pdir, (const lfs_block_t[2]){0, 1});
if (err) {
return err;
}
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
while (!lfs_pairisnull(pdir->d.tail)) {
if (lfs_paircmp(pdir->d.tail, dir) == 0) {
return true;
}
int err = lfs_dir_fetch(lfs, pdir, pdir->d.tail);
if (err) {
return err;
}
}
return false;
}
static int lfs_parent(lfs_t *lfs, const lfs_block_t dir[2],
lfs_dir_t *parent, lfs_entry_t *entry) {
if (lfs_pairisnull(lfs->root)) {
return 0;
}
parent->d.tail[0] = 0;
parent->d.tail[1] = 1;
// iterate over all directory directory entries
while (!lfs_pairisnull(parent->d.tail)) {
int err = lfs_dir_fetch(lfs, parent, parent->d.tail);
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
if (err) {
return err;
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
while (true) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
int err = lfs_dir_next(lfs, parent, entry);
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err && err != LFS_ERR_NOENT) {
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
return err;
}
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
Standardized error values Now matches the commonly used errno codes in name with the value encoded as the negative errno code
2017-04-24 03:10:16 +00:00
if (err == LFS_ERR_NOENT) {
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
break;
}
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
Modified entry head to include name length This provides a path for adding inlined files in the future, which requires multiple lengths to distinguish between the file data and name. As an extra bonus, the directory can now be iterated over even if the types are unknown, since the name's representation is consistent on all entry types. This does come at the cost of reducing types from 16-bits to 8-bits, but I doubt this will become a problem.
2017-06-27 04:17:14 +00:00
if (((0xf & entry->d.type) == (0xf & LFS_TYPE_DIR)) &&
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs_paircmp(entry->d.u.dir, dir) == 0) {
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
return true;
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
}
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
}
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
return false;
}
static int lfs_relocate(lfs_t *lfs,
const lfs_block_t oldpair[2], const lfs_block_t newpair[2]) {
// find parent
lfs_dir_t parent;
lfs_entry_t entry;
int res = lfs_parent(lfs, oldpair, &parent, &entry);
if (res < 0) {
return res;
}
if (res) {
// update disk, this creates a desync
entry.d.u.dir[0] = newpair[0];
entry.d.u.dir[1] = newpair[1];
int err = lfs_dir_update(lfs, &parent, &entry, NULL);
if (err) {
return err;
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// update internal root
if (lfs_paircmp(oldpair, lfs->root) == 0) {
LFS_DEBUG("Relocating root %d %d", newpair[0], newpair[1]);
lfs->root[0] = newpair[0];
lfs->root[1] = newpair[1];
}
// clean up bad block, which should now be a desync
return lfs_deorphan(lfs);
}
// find pred
res = lfs_pred(lfs, oldpair, &parent);
if (res < 0) {
return res;
}
if (res) {
// just replace bad pair, no desync can occur
parent.d.tail[0] = newpair[0];
parent.d.tail[0] = newpair[0];
return lfs_dir_commit(lfs, &parent, NULL, 0);
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// couldn't find dir, must be new
return 0;
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
}
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
int lfs_deorphan(lfs_t *lfs) {
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs->deorphaned = true;
if (lfs_pairisnull(lfs->root)) {
return 0;
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
lfs_dir_t pdir;
lfs_dir_t cdir;
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// skip superblock
int err = lfs_dir_fetch(lfs, &pdir, (const lfs_block_t[2]){0, 1});
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
if (err) {
return err;
}
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
// iterate over all directories
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
while (!lfs_pairisnull(pdir.d.tail)) {
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
int err = lfs_dir_fetch(lfs, &cdir, pdir.d.tail);
if (err) {
return err;
}
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// only check head blocks
if (!(0x80000000 & pdir.d.size)) {
// check if we have a parent
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
lfs_dir_t parent;
lfs_entry_t entry;
int res = lfs_parent(lfs, pdir.d.tail, &parent, &entry);
if (res < 0) {
return res;
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
if (!res) {
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
// we are an orphan
LFS_DEBUG("Orphan %d %d", pdir.d.tail[0], pdir.d.tail[1]);
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
pdir.d.tail[0] = cdir.d.tail[0];
pdir.d.tail[1] = cdir.d.tail[1];
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
Added support for handling corrupted blocks This provides a limited form of wear leveling. While wear is not actually balanced across blocks, the filesystem can recover from corrupted blocks and extend the lifetime of a device nearly as much as dynamic wear leveling. For use-cases where wear is important, it would be better to use a full form of dynamic wear-leveling at the block level. (or consider a logging filesystem). Corrupted block handling was simply added on top of the existing logic in place for the filesystem, so it's a bit more noodly than it may have to be, but it gets the work done.
2017-05-14 17:01:45 +00:00
err = lfs_dir_commit(lfs, &pdir, NULL, 0);
if (err) {
return err;
}
break;
}
if (!lfs_pairsync(entry.d.u.dir, pdir.d.tail)) {
// we have desynced
LFS_DEBUG("Desync %d %d", entry.d.u.dir[0], entry.d.u.dir[1]);
pdir.d.tail[0] = entry.d.u.dir[0];
pdir.d.tail[1] = entry.d.u.dir[1];
err = lfs_dir_commit(lfs, &pdir, NULL, 0);
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
if (err) {
return err;
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
Fixed allocation bugs near the end of storage Also added better testing specifically for these corner cases around the end of storage
2017-04-29 17:41:53 +00:00
break;
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
}
memcpy(&pdir, &cdir, sizeof(pdir));
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
}
Moved to brute-force deorphan without parent pointers Removing the dependency to the parent pointer solves many issues with non-atomic updates of children's parent pointers with respect to any move operations. However, this comes with an embarrassingly terrible runtime as the only other option is to exhaustively check every dir entry to find a child's parent. Fortunately, deorphaning should be a relatively rare operation.
2017-04-14 22:33:36 +00:00
return 0;
Added proper handling of orphans Unfortunately, threading all dir blocks in a linked-list did not come without problems. While it's possible to atomically add a dir to the linked list (by adding the new dir into the linked-list position immediately after it's parent, requiring only one atomic update to the parent block), it is not easy to make sure the linked-list is in a state that always allows atomic removal of dirs. The simple solution is to allow this non-atomic removal, with an additional step to remove any orphans that could have been created by a power-loss. This deorphan step is only run if the normal allocator has failed.
2017-04-01 17:23:15 +00:00
}
Reference in New Issue Copy Permalink