2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* An async IO implementation for Linux
|
|
|
|
* Written by Benjamin LaHaise <bcrl@kvack.org>
|
|
|
|
*
|
|
|
|
* Implements an efficient asynchronous io interface.
|
|
|
|
*
|
|
|
|
* Copyright 2000, 2001, 2002 Red Hat, Inc. All Rights Reserved.
|
|
|
|
*
|
|
|
|
* See ../COPYING for licensing terms.
|
|
|
|
*/
|
2013-05-07 23:18:35 +00:00
|
|
|
#define pr_fmt(fmt) "%s: " fmt, __func__
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/time.h>
|
|
|
|
#include <linux/aio_abi.h>
|
2011-11-17 04:57:37 +00:00
|
|
|
#include <linux/export.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/syscalls.h>
|
2009-10-29 12:59:26 +00:00
|
|
|
#include <linux/backing-dev.h>
|
2006-10-01 06:28:46 +00:00
|
|
|
#include <linux/uio.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/file.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/mman.h>
|
2009-09-22 00:03:51 +00:00
|
|
|
#include <linux/mmu_context.h>
|
2013-04-26 00:58:39 +00:00
|
|
|
#include <linux/percpu.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/timer.h>
|
|
|
|
#include <linux/aio.h>
|
|
|
|
#include <linux/highmem.h>
|
|
|
|
#include <linux/workqueue.h>
|
|
|
|
#include <linux/security.h>
|
signal/timer/event: KAIO eventfd support example
This is an example about how to add eventfd support to the current KAIO code,
in order to enable KAIO to post readiness events to a pollable fd (hence
compatible with POSIX select/poll). The KAIO code simply signals the eventfd
fd when events are ready, and this triggers a POLLIN in the fd. This patch
uses a reserved for future use member of the struct iocb to pass an eventfd
file descriptor, that KAIO will use to post events every time a request
completes. At that point, an aio_getevents() will return the completed result
to a struct io_event. I made a quick test program to verify the patch, and it
runs fine here:
http://www.xmailserver.org/eventfd-aio-test.c
The test program uses poll(2), but it'd, of course, work with select and epoll
too.
This can allow to schedule both block I/O and other poll-able devices
requests, and wait for results using select/poll/epoll. In a typical
scenario, an application would submit KAIO request using aio_submit(), and
will also use epoll_ctl() on the whole other class of devices (that with the
addition of signals, timers and user events, now it's pretty much complete),
and then would:
epoll_wait(...);
for_each_event {
if (curr_event_is_kaiofd) {
aio_getevents();
dispatch_aio_events();
} else {
dispatch_epoll_event();
}
}
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-11 05:23:21 +00:00
|
|
|
#include <linux/eventfd.h>
|
2009-10-02 22:57:36 +00:00
|
|
|
#include <linux/blkdev.h>
|
2010-05-26 21:44:26 +00:00
|
|
|
#include <linux/compat.h>
|
2013-07-16 09:56:16 +00:00
|
|
|
#include <linux/anon_inodes.h>
|
|
|
|
#include <linux/migrate.h>
|
|
|
|
#include <linux/ramfs.h>
|
2013-05-28 22:14:48 +00:00
|
|
|
#include <linux/percpu-refcount.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
#include <asm/kmap_types.h>
|
|
|
|
#include <asm/uaccess.h>
|
|
|
|
|
2013-06-19 11:26:04 +00:00
|
|
|
#include "internal.h"
|
|
|
|
|
2013-05-07 23:18:33 +00:00
|
|
|
#define AIO_RING_MAGIC 0xa10a10a1
|
|
|
|
#define AIO_RING_COMPAT_FEATURES 1
|
|
|
|
#define AIO_RING_INCOMPAT_FEATURES 0
|
|
|
|
struct aio_ring {
|
|
|
|
unsigned id; /* kernel internal index number */
|
|
|
|
unsigned nr; /* number of io_events */
|
|
|
|
unsigned head;
|
|
|
|
unsigned tail;
|
|
|
|
|
|
|
|
unsigned magic;
|
|
|
|
unsigned compat_features;
|
|
|
|
unsigned incompat_features;
|
|
|
|
unsigned header_length; /* size of aio_ring */
|
|
|
|
|
|
|
|
|
|
|
|
struct io_event io_events[0];
|
|
|
|
}; /* 128 bytes + ring size */
|
|
|
|
|
|
|
|
#define AIO_RING_PAGES 8
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
struct kioctx_cpu {
|
|
|
|
unsigned reqs_available;
|
|
|
|
};
|
|
|
|
|
2013-05-07 23:18:33 +00:00
|
|
|
struct kioctx {
|
2013-05-28 22:14:48 +00:00
|
|
|
struct percpu_ref users;
|
2013-05-07 23:18:41 +00:00
|
|
|
atomic_t dead;
|
2013-05-07 23:18:33 +00:00
|
|
|
|
|
|
|
/* This needs improving */
|
|
|
|
unsigned long user_id;
|
|
|
|
struct hlist_node list;
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
struct __percpu kioctx_cpu *cpu;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For percpu reqs_available, number of slots we move to/from global
|
|
|
|
* counter at a time:
|
|
|
|
*/
|
|
|
|
unsigned req_batch;
|
2013-05-07 23:18:51 +00:00
|
|
|
/*
|
|
|
|
* This is what userspace passed to io_setup(), it's not used for
|
|
|
|
* anything but counting against the global max_reqs quota.
|
|
|
|
*
|
2013-05-07 23:18:55 +00:00
|
|
|
* The real limit is nr_events - 1, which will be larger (see
|
2013-05-07 23:18:51 +00:00
|
|
|
* aio_setup_ring())
|
|
|
|
*/
|
2013-05-07 23:18:33 +00:00
|
|
|
unsigned max_reqs;
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
/* Size of ringbuffer, in units of struct io_event */
|
|
|
|
unsigned nr_events;
|
2013-05-07 23:18:33 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
unsigned long mmap_base;
|
|
|
|
unsigned long mmap_size;
|
|
|
|
|
|
|
|
struct page **ring_pages;
|
|
|
|
long nr_pages;
|
|
|
|
|
2013-05-07 23:18:56 +00:00
|
|
|
struct rcu_head rcu_head;
|
2013-05-28 22:14:48 +00:00
|
|
|
struct work_struct free_work;
|
2013-05-07 23:18:56 +00:00
|
|
|
|
|
|
|
struct {
|
2013-04-26 00:58:39 +00:00
|
|
|
/*
|
|
|
|
* This counts the number of available slots in the ringbuffer,
|
|
|
|
* so we avoid overflowing it: it's decremented (if positive)
|
|
|
|
* when allocating a kiocb and incremented when the resulting
|
|
|
|
* io_event is pulled off the ringbuffer.
|
2013-04-26 00:58:39 +00:00
|
|
|
*
|
|
|
|
* We batch accesses to it with a percpu version.
|
2013-04-26 00:58:39 +00:00
|
|
|
*/
|
|
|
|
atomic_t reqs_available;
|
2013-05-07 23:18:56 +00:00
|
|
|
} ____cacheline_aligned_in_smp;
|
|
|
|
|
|
|
|
struct {
|
|
|
|
spinlock_t ctx_lock;
|
|
|
|
struct list_head active_reqs; /* used for cancellation */
|
|
|
|
} ____cacheline_aligned_in_smp;
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
struct {
|
|
|
|
struct mutex ring_lock;
|
2013-05-07 23:18:56 +00:00
|
|
|
wait_queue_head_t wait;
|
|
|
|
} ____cacheline_aligned_in_smp;
|
2013-05-07 23:18:55 +00:00
|
|
|
|
|
|
|
struct {
|
|
|
|
unsigned tail;
|
|
|
|
spinlock_t completion_lock;
|
2013-05-07 23:18:56 +00:00
|
|
|
} ____cacheline_aligned_in_smp;
|
2013-05-07 23:18:55 +00:00
|
|
|
|
|
|
|
struct page *internal_pages[AIO_RING_PAGES];
|
2013-07-16 09:56:16 +00:00
|
|
|
struct file *aio_ring_file;
|
2013-05-07 23:18:33 +00:00
|
|
|
};
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*------ sysctl variables----*/
|
2005-11-07 08:59:31 +00:00
|
|
|
static DEFINE_SPINLOCK(aio_nr_lock);
|
|
|
|
unsigned long aio_nr; /* current system wide number of aio requests */
|
|
|
|
unsigned long aio_max_nr = 0x10000; /* system wide maximum number of aio requests */
|
2005-04-16 22:20:36 +00:00
|
|
|
/*----end sysctl variables---*/
|
|
|
|
|
2006-12-07 04:33:20 +00:00
|
|
|
static struct kmem_cache *kiocb_cachep;
|
|
|
|
static struct kmem_cache *kioctx_cachep;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* aio_setup
|
|
|
|
* Creates the slab caches used by the aio routines, panic on
|
|
|
|
* failure as this is done early during the boot sequence.
|
|
|
|
*/
|
|
|
|
static int __init aio_setup(void)
|
|
|
|
{
|
2007-05-06 21:49:57 +00:00
|
|
|
kiocb_cachep = KMEM_CACHE(kiocb, SLAB_HWCACHE_ALIGN|SLAB_PANIC);
|
|
|
|
kioctx_cachep = KMEM_CACHE(kioctx,SLAB_HWCACHE_ALIGN|SLAB_PANIC);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("sizeof(struct page) = %zu\n", sizeof(struct page));
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2009-09-22 23:43:53 +00:00
|
|
|
__initcall(aio_setup);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
static void aio_free_ring(struct kioctx *ctx)
|
|
|
|
{
|
2013-07-16 09:56:16 +00:00
|
|
|
int i;
|
|
|
|
struct file *aio_ring_file = ctx->aio_ring_file;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-07-16 09:56:16 +00:00
|
|
|
for (i = 0; i < ctx->nr_pages; i++) {
|
|
|
|
pr_debug("pid(%d) [%d] page->count=%d\n", current->pid, i,
|
|
|
|
page_count(ctx->ring_pages[i]));
|
2013-05-07 23:18:55 +00:00
|
|
|
put_page(ctx->ring_pages[i]);
|
2013-07-16 09:56:16 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
if (ctx->ring_pages && ctx->ring_pages != ctx->internal_pages)
|
|
|
|
kfree(ctx->ring_pages);
|
2013-07-16 09:56:16 +00:00
|
|
|
|
|
|
|
if (aio_ring_file) {
|
|
|
|
truncate_setsize(aio_ring_file->f_inode, 0);
|
|
|
|
pr_debug("pid(%d) i_nlink=%u d_count=%d d_unhashed=%d i_count=%d\n",
|
|
|
|
current->pid, aio_ring_file->f_inode->i_nlink,
|
|
|
|
aio_ring_file->f_path.dentry->d_count,
|
|
|
|
d_unhashed(aio_ring_file->f_path.dentry),
|
|
|
|
atomic_read(&aio_ring_file->f_inode->i_count));
|
|
|
|
fput(aio_ring_file);
|
|
|
|
ctx->aio_ring_file = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int aio_ring_mmap(struct file *file, struct vm_area_struct *vma)
|
|
|
|
{
|
|
|
|
vma->vm_ops = &generic_file_vm_ops;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct file_operations aio_ring_fops = {
|
|
|
|
.mmap = aio_ring_mmap,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int aio_set_page_dirty(struct page *page)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-07-17 13:34:24 +00:00
|
|
|
#if IS_ENABLED(CONFIG_MIGRATION)
|
2013-07-16 09:56:16 +00:00
|
|
|
static int aio_migratepage(struct address_space *mapping, struct page *new,
|
|
|
|
struct page *old, enum migrate_mode mode)
|
|
|
|
{
|
|
|
|
struct kioctx *ctx = mapping->private_data;
|
|
|
|
unsigned long flags;
|
|
|
|
unsigned idx = old->index;
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
/* Writeback must be complete */
|
|
|
|
BUG_ON(PageWriteback(old));
|
|
|
|
put_page(old);
|
|
|
|
|
|
|
|
rc = migrate_page_move_mapping(mapping, new, old, NULL, mode);
|
|
|
|
if (rc != MIGRATEPAGE_SUCCESS) {
|
|
|
|
get_page(old);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
get_page(new);
|
|
|
|
|
|
|
|
spin_lock_irqsave(&ctx->completion_lock, flags);
|
|
|
|
migrate_page_copy(new, old);
|
|
|
|
ctx->ring_pages[idx] = new;
|
|
|
|
spin_unlock_irqrestore(&ctx->completion_lock, flags);
|
|
|
|
|
|
|
|
return rc;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2013-07-17 13:34:24 +00:00
|
|
|
#endif
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-07-16 09:56:16 +00:00
|
|
|
static const struct address_space_operations aio_ctx_aops = {
|
|
|
|
.set_page_dirty = aio_set_page_dirty,
|
2013-07-17 13:34:24 +00:00
|
|
|
#if IS_ENABLED(CONFIG_MIGRATION)
|
2013-07-16 09:56:16 +00:00
|
|
|
.migratepage = aio_migratepage,
|
2013-07-17 13:34:24 +00:00
|
|
|
#endif
|
2013-07-16 09:56:16 +00:00
|
|
|
};
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
static int aio_setup_ring(struct kioctx *ctx)
|
|
|
|
{
|
|
|
|
struct aio_ring *ring;
|
|
|
|
unsigned nr_events = ctx->max_reqs;
|
2013-05-07 23:18:25 +00:00
|
|
|
struct mm_struct *mm = current->mm;
|
2013-02-23 00:32:47 +00:00
|
|
|
unsigned long size, populate;
|
2005-04-16 22:20:36 +00:00
|
|
|
int nr_pages;
|
2013-07-16 09:56:16 +00:00
|
|
|
int i;
|
|
|
|
struct file *file;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* Compensate for the ring buffer's head/tail overlap entry */
|
|
|
|
nr_events += 2; /* 1 is required, 2 for good luck */
|
|
|
|
|
|
|
|
size = sizeof(struct aio_ring);
|
|
|
|
size += sizeof(struct io_event) * nr_events;
|
|
|
|
|
2013-07-16 09:56:16 +00:00
|
|
|
nr_pages = PFN_UP(size);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (nr_pages < 0)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2013-07-16 09:56:16 +00:00
|
|
|
file = anon_inode_getfile_private("[aio]", &aio_ring_fops, ctx, O_RDWR);
|
|
|
|
if (IS_ERR(file)) {
|
|
|
|
ctx->aio_ring_file = NULL;
|
|
|
|
return -EAGAIN;
|
|
|
|
}
|
|
|
|
|
|
|
|
file->f_inode->i_mapping->a_ops = &aio_ctx_aops;
|
|
|
|
file->f_inode->i_mapping->private_data = ctx;
|
|
|
|
file->f_inode->i_size = PAGE_SIZE * (loff_t)nr_pages;
|
|
|
|
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page;
|
|
|
|
page = find_or_create_page(file->f_inode->i_mapping,
|
|
|
|
i, GFP_HIGHUSER | __GFP_ZERO);
|
|
|
|
if (!page)
|
|
|
|
break;
|
|
|
|
pr_debug("pid(%d) page[%d]->count=%d\n",
|
|
|
|
current->pid, i, page_count(page));
|
|
|
|
SetPageUptodate(page);
|
|
|
|
SetPageDirty(page);
|
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
ctx->aio_ring_file = file;
|
|
|
|
nr_events = (PAGE_SIZE * nr_pages - sizeof(struct aio_ring))
|
|
|
|
/ sizeof(struct io_event);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->ring_pages = ctx->internal_pages;
|
2005-04-16 22:20:36 +00:00
|
|
|
if (nr_pages > AIO_RING_PAGES) {
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->ring_pages = kcalloc(nr_pages, sizeof(struct page *),
|
|
|
|
GFP_KERNEL);
|
|
|
|
if (!ctx->ring_pages)
|
2005-04-16 22:20:36 +00:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->mmap_size = nr_pages * PAGE_SIZE;
|
|
|
|
pr_debug("attempting mmap of %lu bytes\n", ctx->mmap_size);
|
2013-07-16 09:56:16 +00:00
|
|
|
|
2013-05-07 23:18:25 +00:00
|
|
|
down_write(&mm->mmap_sem);
|
2013-07-16 09:56:16 +00:00
|
|
|
ctx->mmap_base = do_mmap_pgoff(ctx->aio_ring_file, 0, ctx->mmap_size,
|
|
|
|
PROT_READ | PROT_WRITE,
|
|
|
|
MAP_SHARED | MAP_POPULATE, 0, &populate);
|
2013-05-07 23:18:55 +00:00
|
|
|
if (IS_ERR((void *)ctx->mmap_base)) {
|
2013-05-07 23:18:25 +00:00
|
|
|
up_write(&mm->mmap_sem);
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->mmap_size = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
aio_free_ring(ctx);
|
|
|
|
return -EAGAIN;
|
|
|
|
}
|
2013-07-16 09:56:16 +00:00
|
|
|
up_write(&mm->mmap_sem);
|
|
|
|
|
|
|
|
mm_populate(ctx->mmap_base, populate);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
pr_debug("mmap address: 0x%08lx\n", ctx->mmap_base);
|
|
|
|
ctx->nr_pages = get_user_pages(current, mm, ctx->mmap_base, nr_pages,
|
|
|
|
1, 0, ctx->ring_pages, NULL);
|
2013-07-16 09:56:16 +00:00
|
|
|
for (i = 0; i < ctx->nr_pages; i++)
|
|
|
|
put_page(ctx->ring_pages[i]);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
if (unlikely(ctx->nr_pages != nr_pages)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
aio_free_ring(ctx);
|
|
|
|
return -EAGAIN;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->user_id = ctx->mmap_base;
|
|
|
|
ctx->nr_events = nr_events; /* trusted copy */
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ring = kmap_atomic(ctx->ring_pages[0]);
|
2005-04-16 22:20:36 +00:00
|
|
|
ring->nr = nr_events; /* user copy */
|
|
|
|
ring->id = ctx->user_id;
|
|
|
|
ring->head = ring->tail = 0;
|
|
|
|
ring->magic = AIO_RING_MAGIC;
|
|
|
|
ring->compat_features = AIO_RING_COMPAT_FEATURES;
|
|
|
|
ring->incompat_features = AIO_RING_INCOMPAT_FEATURES;
|
|
|
|
ring->header_length = sizeof(struct aio_ring);
|
2011-11-25 15:14:27 +00:00
|
|
|
kunmap_atomic(ring);
|
2013-05-07 23:18:55 +00:00
|
|
|
flush_dcache_page(ctx->ring_pages[0]);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
#define AIO_EVENTS_PER_PAGE (PAGE_SIZE / sizeof(struct io_event))
|
|
|
|
#define AIO_EVENTS_FIRST_PAGE ((PAGE_SIZE - sizeof(struct aio_ring)) / sizeof(struct io_event))
|
|
|
|
#define AIO_EVENTS_OFFSET (AIO_EVENTS_PER_PAGE - AIO_EVENTS_FIRST_PAGE)
|
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
void kiocb_set_cancel_fn(struct kiocb *req, kiocb_cancel_fn *cancel)
|
|
|
|
{
|
|
|
|
struct kioctx *ctx = req->ki_ctx;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&ctx->ctx_lock, flags);
|
|
|
|
|
|
|
|
if (!req->ki_list.next)
|
|
|
|
list_add(&req->ki_list, &ctx->active_reqs);
|
|
|
|
|
|
|
|
req->ki_cancel = cancel;
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&ctx->ctx_lock, flags);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(kiocb_set_cancel_fn);
|
|
|
|
|
2013-05-07 23:18:31 +00:00
|
|
|
static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb,
|
|
|
|
struct io_event *res)
|
|
|
|
{
|
2013-05-07 23:18:49 +00:00
|
|
|
kiocb_cancel_fn *old, *cancel;
|
2013-05-07 23:18:31 +00:00
|
|
|
int ret = -EINVAL;
|
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
/*
|
|
|
|
* Don't want to set kiocb->ki_cancel = KIOCB_CANCELLED unless it
|
|
|
|
* actually has a cancel function, hence the cmpxchg()
|
|
|
|
*/
|
|
|
|
|
|
|
|
cancel = ACCESS_ONCE(kiocb->ki_cancel);
|
|
|
|
do {
|
|
|
|
if (!cancel || cancel == KIOCB_CANCELLED)
|
|
|
|
return ret;
|
2013-05-07 23:18:31 +00:00
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
old = cancel;
|
|
|
|
cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED);
|
|
|
|
} while (cancel != old);
|
2013-05-07 23:18:31 +00:00
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
atomic_inc(&kiocb->ki_users);
|
|
|
|
spin_unlock_irq(&ctx->ctx_lock);
|
|
|
|
|
|
|
|
memset(res, 0, sizeof(*res));
|
|
|
|
res->obj = (u64)(unsigned long)kiocb->ki_obj.user;
|
|
|
|
res->data = kiocb->ki_user_data;
|
|
|
|
ret = cancel(kiocb, res);
|
|
|
|
|
|
|
|
spin_lock_irq(&ctx->ctx_lock);
|
2013-05-07 23:18:31 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
static void free_ioctx_rcu(struct rcu_head *head)
|
|
|
|
{
|
|
|
|
struct kioctx *ctx = container_of(head, struct kioctx, rcu_head);
|
2013-04-26 00:58:39 +00:00
|
|
|
|
|
|
|
free_percpu(ctx->cpu);
|
2013-05-07 23:18:41 +00:00
|
|
|
kmem_cache_free(kioctx_cachep, ctx);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* When this function runs, the kioctx has been removed from the "hash table"
|
|
|
|
* and ctx->users has dropped to 0, so we know no more kiocbs can be submitted -
|
|
|
|
* now it's safe to cancel any that need to be.
|
|
|
|
*/
|
2013-05-28 22:14:48 +00:00
|
|
|
static void free_ioctx(struct work_struct *work)
|
2013-05-07 23:18:41 +00:00
|
|
|
{
|
2013-05-28 22:14:48 +00:00
|
|
|
struct kioctx *ctx = container_of(work, struct kioctx, free_work);
|
2013-05-07 23:18:51 +00:00
|
|
|
struct aio_ring *ring;
|
2013-05-07 23:18:41 +00:00
|
|
|
struct io_event res;
|
|
|
|
struct kiocb *req;
|
2013-04-26 00:58:39 +00:00
|
|
|
unsigned cpu, head, avail;
|
2013-05-07 23:18:41 +00:00
|
|
|
|
|
|
|
spin_lock_irq(&ctx->ctx_lock);
|
|
|
|
|
|
|
|
while (!list_empty(&ctx->active_reqs)) {
|
|
|
|
req = list_first_entry(&ctx->active_reqs,
|
|
|
|
struct kiocb, ki_list);
|
|
|
|
|
|
|
|
list_del_init(&req->ki_list);
|
|
|
|
kiocb_cancel(ctx, req, &res);
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(&ctx->ctx_lock);
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
struct kioctx_cpu *kcpu = per_cpu_ptr(ctx->cpu, cpu);
|
|
|
|
|
|
|
|
atomic_add(kcpu->reqs_available, &ctx->reqs_available);
|
|
|
|
kcpu->reqs_available = 0;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ring = kmap_atomic(ctx->ring_pages[0]);
|
2013-05-07 23:18:51 +00:00
|
|
|
head = ring->head;
|
|
|
|
kunmap_atomic(ring);
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
while (atomic_read(&ctx->reqs_available) < ctx->nr_events - 1) {
|
2013-05-24 22:55:38 +00:00
|
|
|
wait_event(ctx->wait,
|
2013-04-26 00:58:39 +00:00
|
|
|
(head != ctx->tail) ||
|
|
|
|
(atomic_read(&ctx->reqs_available) >=
|
|
|
|
ctx->nr_events - 1));
|
2013-05-07 23:18:51 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
|
2013-05-07 23:18:51 +00:00
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
atomic_add(avail, &ctx->reqs_available);
|
2013-05-07 23:18:51 +00:00
|
|
|
head += avail;
|
2013-05-07 23:18:55 +00:00
|
|
|
head %= ctx->nr_events;
|
2013-05-07 23:18:51 +00:00
|
|
|
}
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
WARN_ON(atomic_read(&ctx->reqs_available) > ctx->nr_events - 1);
|
2013-05-07 23:18:41 +00:00
|
|
|
|
|
|
|
aio_free_ring(ctx);
|
|
|
|
|
|
|
|
pr_debug("freeing %p\n", ctx);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Here the call_rcu() is between the wait_event() for reqs_active to
|
|
|
|
* hit 0, and freeing the ioctx.
|
|
|
|
*
|
|
|
|
* aio_complete() decrements reqs_active, but it has to touch the ioctx
|
|
|
|
* after to issue a wakeup so we use rcu.
|
|
|
|
*/
|
|
|
|
call_rcu(&ctx->rcu_head, free_ioctx_rcu);
|
|
|
|
}
|
|
|
|
|
2013-05-28 22:14:48 +00:00
|
|
|
static void free_ioctx_ref(struct percpu_ref *ref)
|
2013-05-07 23:18:41 +00:00
|
|
|
{
|
2013-05-28 22:14:48 +00:00
|
|
|
struct kioctx *ctx = container_of(ref, struct kioctx, users);
|
|
|
|
|
|
|
|
INIT_WORK(&ctx->free_work, free_ioctx);
|
|
|
|
schedule_work(&ctx->free_work);
|
2013-05-07 23:18:41 +00:00
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* ioctx_alloc
|
|
|
|
* Allocates and initializes an ioctx. Returns an ERR_PTR if it failed.
|
|
|
|
*/
|
|
|
|
static struct kioctx *ioctx_alloc(unsigned nr_events)
|
|
|
|
{
|
2013-05-07 23:18:25 +00:00
|
|
|
struct mm_struct *mm = current->mm;
|
2005-04-16 22:20:36 +00:00
|
|
|
struct kioctx *ctx;
|
2012-03-06 19:33:22 +00:00
|
|
|
int err = -ENOMEM;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
/*
|
|
|
|
* We keep track of the number of available ringbuffer slots, to prevent
|
|
|
|
* overflow (reqs_available), and we also use percpu counters for this.
|
|
|
|
*
|
|
|
|
* So since up to half the slots might be on other cpu's percpu counters
|
|
|
|
* and unavailable, double nr_events so userspace sees what they
|
|
|
|
* expected: additionally, we move req_batch slots to/from percpu
|
|
|
|
* counters at a time, so make sure that isn't 0:
|
|
|
|
*/
|
|
|
|
nr_events = max(nr_events, num_possible_cpus() * 4);
|
|
|
|
nr_events *= 2;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* Prevent overflows */
|
|
|
|
if ((nr_events > (0x10000000U / sizeof(struct io_event))) ||
|
|
|
|
(nr_events > (0x10000000U / sizeof(struct kiocb)))) {
|
|
|
|
pr_debug("ENOMEM: nr_events too high\n");
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
}
|
|
|
|
|
2012-03-11 04:10:35 +00:00
|
|
|
if (!nr_events || (unsigned long)nr_events > aio_max_nr)
|
2005-04-16 22:20:36 +00:00
|
|
|
return ERR_PTR(-EAGAIN);
|
|
|
|
|
2007-02-10 09:45:03 +00:00
|
|
|
ctx = kmem_cache_zalloc(kioctx_cachep, GFP_KERNEL);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!ctx)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
|
|
|
ctx->max_reqs = nr_events;
|
|
|
|
|
2013-05-28 22:14:48 +00:00
|
|
|
if (percpu_ref_init(&ctx->users, free_ioctx_ref))
|
|
|
|
goto out_freectx;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
spin_lock_init(&ctx->ctx_lock);
|
2013-05-07 23:18:49 +00:00
|
|
|
spin_lock_init(&ctx->completion_lock);
|
2013-05-07 23:18:55 +00:00
|
|
|
mutex_init(&ctx->ring_lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
init_waitqueue_head(&ctx->wait);
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&ctx->active_reqs);
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
ctx->cpu = alloc_percpu(struct kioctx_cpu);
|
|
|
|
if (!ctx->cpu)
|
2013-05-28 22:14:48 +00:00
|
|
|
goto out_freeref;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
if (aio_setup_ring(ctx) < 0)
|
|
|
|
goto out_freepcpu;
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
atomic_set(&ctx->reqs_available, ctx->nr_events - 1);
|
2013-04-26 00:58:39 +00:00
|
|
|
ctx->req_batch = (ctx->nr_events - 1) / (num_possible_cpus() * 4);
|
|
|
|
BUG_ON(!ctx->req_batch);
|
2013-04-26 00:58:39 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* limit the number of system wide aios */
|
2012-03-11 04:14:05 +00:00
|
|
|
spin_lock(&aio_nr_lock);
|
2012-03-11 04:10:35 +00:00
|
|
|
if (aio_nr + nr_events > aio_max_nr ||
|
|
|
|
aio_nr + nr_events < aio_nr) {
|
2012-03-11 04:14:05 +00:00
|
|
|
spin_unlock(&aio_nr_lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
goto out_cleanup;
|
2012-03-11 04:10:35 +00:00
|
|
|
}
|
|
|
|
aio_nr += ctx->max_reqs;
|
2012-03-11 04:14:05 +00:00
|
|
|
spin_unlock(&aio_nr_lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_get(&ctx->users); /* io_setup() will drop this ref */
|
|
|
|
|
2008-04-29 08:03:48 +00:00
|
|
|
/* now link into global list. */
|
2008-12-09 07:11:22 +00:00
|
|
|
spin_lock(&mm->ioctx_lock);
|
|
|
|
hlist_add_head_rcu(&ctx->list, &mm->ioctx_list);
|
|
|
|
spin_unlock(&mm->ioctx_lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("allocated ioctx %p[%ld]: mm=%p mask=0x%x\n",
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx, ctx->user_id, mm, ctx->nr_events);
|
2005-04-16 22:20:36 +00:00
|
|
|
return ctx;
|
|
|
|
|
|
|
|
out_cleanup:
|
2012-03-06 19:33:22 +00:00
|
|
|
err = -EAGAIN;
|
|
|
|
aio_free_ring(ctx);
|
2013-04-26 00:58:39 +00:00
|
|
|
out_freepcpu:
|
|
|
|
free_percpu(ctx->cpu);
|
2013-05-28 22:14:48 +00:00
|
|
|
out_freeref:
|
|
|
|
free_percpu(ctx->users.pcpu_count);
|
2005-04-16 22:20:36 +00:00
|
|
|
out_freectx:
|
2013-07-16 09:56:16 +00:00
|
|
|
if (ctx->aio_ring_file)
|
|
|
|
fput(ctx->aio_ring_file);
|
2005-04-16 22:20:36 +00:00
|
|
|
kmem_cache_free(kioctx_cachep, ctx);
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("error allocating ioctx %d\n", err);
|
2012-03-06 19:33:22 +00:00
|
|
|
return ERR_PTR(err);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
/* kill_ioctx
|
|
|
|
* Cancels all outstanding aio requests on an aio context. Used
|
|
|
|
* when the processes owning a context have all exited to encourage
|
|
|
|
* the rapid destruction of the kioctx.
|
|
|
|
*/
|
|
|
|
static void kill_ioctx(struct kioctx *ctx)
|
|
|
|
{
|
|
|
|
if (!atomic_xchg(&ctx->dead, 1)) {
|
|
|
|
hlist_del_rcu(&ctx->list);
|
2013-05-28 22:14:48 +00:00
|
|
|
/* percpu_ref_kill() will do the necessary call_rcu() */
|
|
|
|
wake_up_all(&ctx->wait);
|
2007-02-03 09:13:45 +00:00
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
/*
|
2013-06-12 21:04:59 +00:00
|
|
|
* It'd be more correct to do this in free_ioctx(), after all
|
|
|
|
* the outstanding kiocbs have finished - but by then io_destroy
|
|
|
|
* has already returned, so io_setup() could potentially return
|
|
|
|
* -EAGAIN with no ioctxs actually in use (as far as userspace
|
|
|
|
* could tell).
|
2013-05-07 23:18:41 +00:00
|
|
|
*/
|
2013-06-12 21:04:59 +00:00
|
|
|
spin_lock(&aio_nr_lock);
|
|
|
|
BUG_ON(aio_nr - ctx->max_reqs > aio_nr);
|
|
|
|
aio_nr -= ctx->max_reqs;
|
|
|
|
spin_unlock(&aio_nr_lock);
|
|
|
|
|
|
|
|
if (ctx->mmap_size)
|
|
|
|
vm_munmap(ctx->mmap_base, ctx->mmap_size);
|
|
|
|
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_kill(&ctx->users);
|
2013-05-07 23:18:41 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* wait_on_sync_kiocb:
|
|
|
|
* Waits on the given sync kiocb to complete.
|
|
|
|
*/
|
2008-02-08 12:19:52 +00:00
|
|
|
ssize_t wait_on_sync_kiocb(struct kiocb *iocb)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:18:39 +00:00
|
|
|
while (atomic_read(&iocb->ki_users)) {
|
2005-04-16 22:20:36 +00:00
|
|
|
set_current_state(TASK_UNINTERRUPTIBLE);
|
2013-05-07 23:18:39 +00:00
|
|
|
if (!atomic_read(&iocb->ki_users))
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
2007-10-17 06:27:20 +00:00
|
|
|
io_schedule();
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
__set_current_state(TASK_RUNNING);
|
|
|
|
return iocb->ki_user_data;
|
|
|
|
}
|
2009-09-22 23:43:53 +00:00
|
|
|
EXPORT_SYMBOL(wait_on_sync_kiocb);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
/*
|
|
|
|
* exit_aio: called when the last user of mm goes away. At this point, there is
|
|
|
|
* no way for any new requests to be submited or any of the io_* syscalls to be
|
|
|
|
* called on the context.
|
|
|
|
*
|
|
|
|
* There may be outstanding kiocbs, but free_ioctx() will explicitly wait on
|
|
|
|
* them.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-02-08 12:19:52 +00:00
|
|
|
void exit_aio(struct mm_struct *mm)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-12-09 07:11:22 +00:00
|
|
|
struct kioctx *ctx;
|
2013-05-07 23:18:41 +00:00
|
|
|
struct hlist_node *n;
|
2008-12-09 07:11:22 +00:00
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
hlist_for_each_entry_safe(ctx, n, &mm->ioctx_list, list) {
|
2012-04-21 01:49:41 +00:00
|
|
|
/*
|
|
|
|
* We don't need to bother with munmap() here -
|
|
|
|
* exit_mmap(mm) is coming and it'll unmap everything.
|
|
|
|
* Since aio_free_ring() uses non-zero ->mmap_size
|
|
|
|
* as indicator that it needs to unmap the area,
|
|
|
|
* just set it to 0; aio_free_ring() is the only
|
|
|
|
* place that uses ->mmap_size, so it's safe.
|
|
|
|
*/
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->mmap_size = 0;
|
2013-05-07 23:18:41 +00:00
|
|
|
|
2013-06-12 21:04:59 +00:00
|
|
|
kill_ioctx(ctx);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
static void put_reqs_available(struct kioctx *ctx, unsigned nr)
|
|
|
|
{
|
|
|
|
struct kioctx_cpu *kcpu;
|
|
|
|
|
|
|
|
preempt_disable();
|
|
|
|
kcpu = this_cpu_ptr(ctx->cpu);
|
|
|
|
|
|
|
|
kcpu->reqs_available += nr;
|
|
|
|
while (kcpu->reqs_available >= ctx->req_batch * 2) {
|
|
|
|
kcpu->reqs_available -= ctx->req_batch;
|
|
|
|
atomic_add(ctx->req_batch, &ctx->reqs_available);
|
|
|
|
}
|
|
|
|
|
|
|
|
preempt_enable();
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool get_reqs_available(struct kioctx *ctx)
|
|
|
|
{
|
|
|
|
struct kioctx_cpu *kcpu;
|
|
|
|
bool ret = false;
|
|
|
|
|
|
|
|
preempt_disable();
|
|
|
|
kcpu = this_cpu_ptr(ctx->cpu);
|
|
|
|
|
|
|
|
if (!kcpu->reqs_available) {
|
|
|
|
int old, avail = atomic_read(&ctx->reqs_available);
|
|
|
|
|
|
|
|
do {
|
|
|
|
if (avail < ctx->req_batch)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
old = avail;
|
|
|
|
avail = atomic_cmpxchg(&ctx->reqs_available,
|
|
|
|
avail, avail - ctx->req_batch);
|
|
|
|
} while (avail != old);
|
|
|
|
|
|
|
|
kcpu->reqs_available += ctx->req_batch;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = true;
|
|
|
|
kcpu->reqs_available--;
|
|
|
|
out:
|
|
|
|
preempt_enable();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* aio_get_req
|
2013-05-07 23:18:39 +00:00
|
|
|
* Allocate a slot for an aio request. Increments the ki_users count
|
2005-04-16 22:20:36 +00:00
|
|
|
* of the kioctx so that the kioctx stays around until all requests are
|
|
|
|
* complete. Returns NULL if no requests are free.
|
|
|
|
*
|
2013-05-07 23:18:39 +00:00
|
|
|
* Returns with kiocb->ki_users set to 2. The io submit code path holds
|
2005-04-16 22:20:36 +00:00
|
|
|
* an extra reference while submitting the i/o.
|
|
|
|
* This prevents races between the aio code path referencing the
|
|
|
|
* req (after submitting it) and aio_complete() freeing the req.
|
|
|
|
*/
|
2013-05-07 23:18:53 +00:00
|
|
|
static inline struct kiocb *aio_get_req(struct kioctx *ctx)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:18:53 +00:00
|
|
|
struct kiocb *req;
|
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
if (!get_reqs_available(ctx))
|
2013-05-07 23:18:53 +00:00
|
|
|
return NULL;
|
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(!req))
|
2013-05-07 23:18:53 +00:00
|
|
|
goto out_put;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:39 +00:00
|
|
|
atomic_set(&req->ki_users, 2);
|
2005-04-16 22:20:36 +00:00
|
|
|
req->ki_ctx = ctx;
|
2011-11-02 20:40:10 +00:00
|
|
|
return req;
|
2013-05-07 23:18:53 +00:00
|
|
|
out_put:
|
2013-04-26 00:58:39 +00:00
|
|
|
put_reqs_available(ctx, 1);
|
2013-05-07 23:18:53 +00:00
|
|
|
return NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:39 +00:00
|
|
|
static void kiocb_free(struct kiocb *req)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:18:37 +00:00
|
|
|
if (req->ki_filp)
|
|
|
|
fput(req->ki_filp);
|
2009-06-30 18:41:11 +00:00
|
|
|
if (req->ki_eventfd != NULL)
|
|
|
|
eventfd_ctx_put(req->ki_eventfd);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (req->ki_dtor)
|
|
|
|
req->ki_dtor(req);
|
2006-10-01 06:28:49 +00:00
|
|
|
if (req->ki_iovec != &req->ki_inline_vec)
|
|
|
|
kfree(req->ki_iovec);
|
2005-04-16 22:20:36 +00:00
|
|
|
kmem_cache_free(kiocb_cachep, req);
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:29 +00:00
|
|
|
void aio_put_req(struct kiocb *req)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:18:39 +00:00
|
|
|
if (atomic_dec_and_test(&req->ki_users))
|
|
|
|
kiocb_free(req);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2009-09-22 23:43:53 +00:00
|
|
|
EXPORT_SYMBOL(aio_put_req);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-04-29 07:58:57 +00:00
|
|
|
static struct kioctx *lookup_ioctx(unsigned long ctx_id)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-12-09 07:11:22 +00:00
|
|
|
struct mm_struct *mm = current->mm;
|
2009-03-19 00:04:21 +00:00
|
|
|
struct kioctx *ctx, *ret = NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-12-09 07:11:22 +00:00
|
|
|
rcu_read_lock();
|
|
|
|
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-28 01:06:00 +00:00
|
|
|
hlist_for_each_entry_rcu(ctx, &mm->ioctx_list, list) {
|
2013-05-07 23:18:41 +00:00
|
|
|
if (ctx->user_id == ctx_id) {
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_get(&ctx->users);
|
2009-03-19 00:04:21 +00:00
|
|
|
ret = ctx;
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
|
|
|
}
|
2008-12-09 07:11:22 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-12-09 07:11:22 +00:00
|
|
|
rcu_read_unlock();
|
2009-03-19 00:04:21 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* aio_complete
|
|
|
|
* Called when the io request on the given iocb is complete.
|
|
|
|
*/
|
2013-05-07 23:18:29 +00:00
|
|
|
void aio_complete(struct kiocb *iocb, long res, long res2)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct kioctx *ctx = iocb->ki_ctx;
|
|
|
|
struct aio_ring *ring;
|
2013-05-07 23:18:47 +00:00
|
|
|
struct io_event *ev_page, *event;
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long flags;
|
2013-05-07 23:18:47 +00:00
|
|
|
unsigned tail, pos;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2005-11-14 00:07:33 +00:00
|
|
|
/*
|
|
|
|
* Special case handling for sync iocbs:
|
|
|
|
* - events go directly into the iocb for fast handling
|
|
|
|
* - the sync task with the iocb in its stack holds the single iocb
|
|
|
|
* ref, no other paths have a way to get another ref
|
|
|
|
* - the sync task helpfully left a reference to itself in the iocb
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
|
|
|
if (is_sync_kiocb(iocb)) {
|
2013-05-07 23:18:39 +00:00
|
|
|
BUG_ON(atomic_read(&iocb->ki_users) != 1);
|
2005-04-16 22:20:36 +00:00
|
|
|
iocb->ki_user_data = res;
|
2013-05-07 23:18:39 +00:00
|
|
|
atomic_set(&iocb->ki_users, 0);
|
2005-04-16 22:20:36 +00:00
|
|
|
wake_up_process(iocb->ki_obj.tsk);
|
2013-05-07 23:18:29 +00:00
|
|
|
return;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
/*
|
|
|
|
* Take rcu_read_lock() in case the kioctx is being destroyed, as we
|
2013-04-26 00:58:39 +00:00
|
|
|
* need to issue a wakeup after incrementing reqs_available.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2013-05-07 23:18:41 +00:00
|
|
|
rcu_read_lock();
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
if (iocb->ki_list.next) {
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&ctx->ctx_lock, flags);
|
|
|
|
list_del(&iocb->ki_list);
|
|
|
|
spin_unlock_irqrestore(&ctx->ctx_lock, flags);
|
|
|
|
}
|
2013-05-07 23:18:39 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* cancelled requests don't get events, userland was given one
|
|
|
|
* when the event got cancelled.
|
|
|
|
*/
|
2013-05-07 23:18:49 +00:00
|
|
|
if (unlikely(xchg(&iocb->ki_cancel,
|
2013-05-07 23:18:51 +00:00
|
|
|
KIOCB_CANCELLED) == KIOCB_CANCELLED)) {
|
2013-04-26 00:58:39 +00:00
|
|
|
/*
|
|
|
|
* Can't use the percpu reqs_available here - could race with
|
|
|
|
* free_ioctx()
|
|
|
|
*/
|
2013-04-26 00:58:39 +00:00
|
|
|
atomic_inc(&ctx->reqs_available);
|
2013-05-07 23:18:51 +00:00
|
|
|
/* Still need the wake_up in case free_ioctx is waiting */
|
2005-04-16 22:20:36 +00:00
|
|
|
goto put_rq;
|
2013-05-07 23:18:51 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
/*
|
|
|
|
* Add a completion event to the ring buffer. Must be done holding
|
2013-07-03 22:09:16 +00:00
|
|
|
* ctx->completion_lock to prevent other code from messing with the tail
|
2013-05-07 23:18:49 +00:00
|
|
|
* pointer since we might be called from irq context.
|
|
|
|
*/
|
|
|
|
spin_lock_irqsave(&ctx->completion_lock, flags);
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
tail = ctx->tail;
|
2013-05-07 23:18:47 +00:00
|
|
|
pos = tail + AIO_EVENTS_OFFSET;
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
if (++tail >= ctx->nr_events)
|
2005-05-01 15:59:15 +00:00
|
|
|
tail = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
|
2013-05-07 23:18:47 +00:00
|
|
|
event = ev_page + pos % AIO_EVENTS_PER_PAGE;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
event->obj = (u64)(unsigned long)iocb->ki_obj.user;
|
|
|
|
event->data = iocb->ki_user_data;
|
|
|
|
event->res = res;
|
|
|
|
event->res2 = res2;
|
|
|
|
|
2013-05-07 23:18:47 +00:00
|
|
|
kunmap_atomic(ev_page);
|
2013-05-07 23:18:55 +00:00
|
|
|
flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
|
2013-05-07 23:18:47 +00:00
|
|
|
|
|
|
|
pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
|
2013-05-07 23:18:35 +00:00
|
|
|
ctx, tail, iocb, iocb->ki_obj.user, iocb->ki_user_data,
|
|
|
|
res, res2);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/* after flagging the request as done, we
|
|
|
|
* must never even look at it again
|
|
|
|
*/
|
|
|
|
smp_wmb(); /* make event visible before updating tail */
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ctx->tail = tail;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ring = kmap_atomic(ctx->ring_pages[0]);
|
2013-05-07 23:18:47 +00:00
|
|
|
ring->tail = tail;
|
2011-11-25 15:14:27 +00:00
|
|
|
kunmap_atomic(ring);
|
2013-05-07 23:18:55 +00:00
|
|
|
flush_dcache_page(ctx->ring_pages[0]);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:49 +00:00
|
|
|
spin_unlock_irqrestore(&ctx->completion_lock, flags);
|
|
|
|
|
2013-05-07 23:18:47 +00:00
|
|
|
pr_debug("added to ring %p at [%u]\n", iocb, tail);
|
2008-04-11 04:29:19 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the user asked us to deliver the result through an
|
|
|
|
* eventfd. The eventfd_signal() function is safe to be called
|
|
|
|
* from IRQ context.
|
|
|
|
*/
|
2009-03-19 00:04:19 +00:00
|
|
|
if (iocb->ki_eventfd != NULL)
|
2008-04-11 04:29:19 +00:00
|
|
|
eventfd_signal(iocb->ki_eventfd, 1);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
put_rq:
|
|
|
|
/* everything turned out well, dispose of the aiocb. */
|
2013-05-07 23:18:39 +00:00
|
|
|
aio_put_req(iocb);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
aio: bad AIO race in aio_complete() leads to process hang
My group ran into a AIO process hang on a 2.6.24 kernel with the process
sleeping indefinitely in io_getevents(2) waiting for the last wakeup to come
and it never would.
We ran the tests on x86_64 SMP. The hang only occurred on a Xeon box
("Clovertown") but not a Core2Duo ("Conroe"). On the Xeon, the L2 cache isn't
shared between all eight processors, but is L2 is shared between between all
two processors on the Core2Duo we use.
My analysis of the hang is if you go down to the second while-loop
in read_events(), what happens on processor #1:
1) add_wait_queue_exclusive() adds thread to ctx->wait
2) aio_read_evt() to check tail
3) if aio_read_evt() returned 0, call [io_]schedule() and sleep
In aio_complete() with processor #2:
A) info->tail = tail;
B) waitqueue_active(&ctx->wait)
C) if waitqueue_active() returned non-0, call wake_up()
The way the code is written, step 1 must be seen by all other processors
before processor 1 checks for pending events in step 2 (that were recorded by
step A) and step A by processor 2 must be seen by all other processors
(checked in step 2) before step B is done.
The race I believed I was seeing is that steps 1 and 2 were
effectively swapped due to the __list_add() being delayed by the L2
cache not shared by some of the other processors. Imagine:
proc 2: just before step A
proc 1, step 1: adds to ctx->wait, but is not visible by other processors yet
proc 1, step 2: checks tail and sees no pending events
proc 2, step A: updates tail
proc 1, step 3: calls [io_]schedule() and sleeps
proc 2, step B: checks ctx->wait, but sees no one waiting, skips wakeup
so proc 1 sleeps indefinitely
My patch adds a memory barrier between steps A and B. It ensures that the
update in step 1 gets seen on processor 2 before continuing. If processor 1
was just before step 1, the memory barrier makes sure that step A (update
tail) gets seen by the time processor 1 makes it to step 2 (check tail).
Before the patch our AIO process would hang virtually 100% of the time. After
the patch, we have yet to see the process ever hang.
Signed-off-by: Quentin Barnes <qbarnes+linux@yahoo-inc.com>
Reviewed-by: Zach Brown <zach.brown@oracle.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: <stable@kernel.org>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ We should probably disallow that "if (waitqueue_active()) wake_up()"
coding pattern, because it's so often buggy wrt memory ordering ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-20 00:00:39 +00:00
|
|
|
/*
|
|
|
|
* We have to order our ring_info tail store above and test
|
|
|
|
* of the wait list below outside the wait lock. This is
|
|
|
|
* like in wake_up_bit() where clearing a bit has to be
|
|
|
|
* ordered with the unlocked test.
|
|
|
|
*/
|
|
|
|
smp_mb();
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (waitqueue_active(&ctx->wait))
|
|
|
|
wake_up(&ctx->wait);
|
|
|
|
|
2013-05-07 23:18:41 +00:00
|
|
|
rcu_read_unlock();
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2009-09-22 23:43:53 +00:00
|
|
|
EXPORT_SYMBOL(aio_complete);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
/* aio_read_events
|
|
|
|
* Pull an event off of the ioctx's event ring. Returns the number of
|
|
|
|
* events fetched
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2013-05-07 23:18:45 +00:00
|
|
|
static long aio_read_events_ring(struct kioctx *ctx,
|
|
|
|
struct io_event __user *event, long nr)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct aio_ring *ring;
|
2013-05-07 23:18:45 +00:00
|
|
|
unsigned head, pos;
|
|
|
|
long ret = 0;
|
|
|
|
int copy_ret;
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
mutex_lock(&ctx->ring_lock);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ring = kmap_atomic(ctx->ring_pages[0]);
|
2013-05-07 23:18:45 +00:00
|
|
|
head = ring->head;
|
|
|
|
kunmap_atomic(ring);
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
pr_debug("h%u t%u m%u\n", head, ctx->tail, ctx->nr_events);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
if (head == ctx->tail)
|
2005-04-16 22:20:36 +00:00
|
|
|
goto out;
|
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
while (ret < nr) {
|
|
|
|
long avail;
|
|
|
|
struct io_event *ev;
|
|
|
|
struct page *page;
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
|
|
|
|
if (head == ctx->tail)
|
2013-05-07 23:18:45 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
avail = min(avail, nr - ret);
|
|
|
|
avail = min_t(long, avail, AIO_EVENTS_PER_PAGE -
|
|
|
|
((head + AIO_EVENTS_OFFSET) % AIO_EVENTS_PER_PAGE));
|
|
|
|
|
|
|
|
pos = head + AIO_EVENTS_OFFSET;
|
2013-05-07 23:18:55 +00:00
|
|
|
page = ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE];
|
2013-05-07 23:18:45 +00:00
|
|
|
pos %= AIO_EVENTS_PER_PAGE;
|
|
|
|
|
|
|
|
ev = kmap(page);
|
|
|
|
copy_ret = copy_to_user(event + ret, ev + pos,
|
|
|
|
sizeof(*ev) * avail);
|
|
|
|
kunmap(page);
|
|
|
|
|
|
|
|
if (unlikely(copy_ret)) {
|
|
|
|
ret = -EFAULT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret += avail;
|
|
|
|
head += avail;
|
2013-05-07 23:18:55 +00:00
|
|
|
head %= ctx->nr_events;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
ring = kmap_atomic(ctx->ring_pages[0]);
|
2013-05-07 23:18:45 +00:00
|
|
|
ring->head = head;
|
2013-04-26 03:03:53 +00:00
|
|
|
kunmap_atomic(ring);
|
2013-05-07 23:18:55 +00:00
|
|
|
flush_dcache_page(ctx->ring_pages[0]);
|
2013-05-07 23:18:45 +00:00
|
|
|
|
2013-05-07 23:18:55 +00:00
|
|
|
pr_debug("%li h%u t%u\n", ret, head, ctx->tail);
|
2013-05-07 23:18:51 +00:00
|
|
|
|
2013-04-26 00:58:39 +00:00
|
|
|
put_reqs_available(ctx, ret);
|
2013-05-07 23:18:45 +00:00
|
|
|
out:
|
2013-05-07 23:18:55 +00:00
|
|
|
mutex_unlock(&ctx->ring_lock);
|
2013-05-07 23:18:45 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
static bool aio_read_events(struct kioctx *ctx, long min_nr, long nr,
|
|
|
|
struct io_event __user *event, long *i)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:18:45 +00:00
|
|
|
long ret = aio_read_events_ring(ctx, event + *i, nr - *i);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
if (ret > 0)
|
|
|
|
*i += ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
if (unlikely(atomic_read(&ctx->dead)))
|
|
|
|
ret = -EINVAL;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
if (!*i)
|
|
|
|
*i = ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
return ret < 0 || *i >= min_nr;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
static long read_events(struct kioctx *ctx, long min_nr, long nr,
|
2005-04-16 22:20:36 +00:00
|
|
|
struct io_event __user *event,
|
|
|
|
struct timespec __user *timeout)
|
|
|
|
{
|
2013-05-07 23:18:45 +00:00
|
|
|
ktime_t until = { .tv64 = KTIME_MAX };
|
|
|
|
long ret = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
if (timeout) {
|
|
|
|
struct timespec ts;
|
2013-05-07 23:18:45 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(copy_from_user(&ts, timeout, sizeof(ts))))
|
2013-05-07 23:18:45 +00:00
|
|
|
return -EFAULT;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
until = timespec_to_ktime(ts);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
/*
|
|
|
|
* Note that aio_read_events() is being called as the conditional - i.e.
|
|
|
|
* we're calling it after prepare_to_wait() has set task state to
|
|
|
|
* TASK_INTERRUPTIBLE.
|
|
|
|
*
|
|
|
|
* But aio_read_events() can block, and if it blocks it's going to flip
|
|
|
|
* the task state back to TASK_RUNNING.
|
|
|
|
*
|
|
|
|
* This should be ok, provided it doesn't flip the state back to
|
|
|
|
* TASK_RUNNING and return 0 too much - that causes us to spin. That
|
|
|
|
* will only happen if the mutex_lock() call blocks, and we then find
|
|
|
|
* the ringbuffer empty. So in practice we should be ok, but it's
|
|
|
|
* something to be aware of when touching this code.
|
|
|
|
*/
|
|
|
|
wait_event_interruptible_hrtimeout(ctx->wait,
|
|
|
|
aio_read_events(ctx, min_nr, nr, event, &ret), until);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
if (!ret && signal_pending(current))
|
|
|
|
ret = -EINTR;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:18:45 +00:00
|
|
|
return ret;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* sys_io_setup:
|
|
|
|
* Create an aio_context capable of receiving at least nr_events.
|
|
|
|
* ctxp must not point to an aio_context that already exists, and
|
|
|
|
* must be initialized to 0 prior to the call. On successful
|
|
|
|
* creation of the aio_context, *ctxp is filled in with the resulting
|
|
|
|
* handle. May fail with -EINVAL if *ctxp is not initialized,
|
|
|
|
* if the specified nr_events exceeds internal limits. May fail
|
|
|
|
* with -EAGAIN if the specified nr_events exceeds the user's limit
|
|
|
|
* of available events. May fail with -ENOMEM if insufficient kernel
|
|
|
|
* resources are available. May fail with -EFAULT if an invalid
|
|
|
|
* pointer is passed for ctxp. Will fail with -ENOSYS if not
|
|
|
|
* implemented.
|
|
|
|
*/
|
2009-01-14 13:14:18 +00:00
|
|
|
SYSCALL_DEFINE2(io_setup, unsigned, nr_events, aio_context_t __user *, ctxp)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct kioctx *ioctx = NULL;
|
|
|
|
unsigned long ctx;
|
|
|
|
long ret;
|
|
|
|
|
|
|
|
ret = get_user(ctx, ctxp);
|
|
|
|
if (unlikely(ret))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
ret = -EINVAL;
|
2005-11-07 08:59:31 +00:00
|
|
|
if (unlikely(ctx || nr_events == 0)) {
|
|
|
|
pr_debug("EINVAL: io_setup: ctx %lu nr_events %u\n",
|
|
|
|
ctx, nr_events);
|
2005-04-16 22:20:36 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ioctx = ioctx_alloc(nr_events);
|
|
|
|
ret = PTR_ERR(ioctx);
|
|
|
|
if (!IS_ERR(ioctx)) {
|
|
|
|
ret = put_user(ioctx->user_id, ctxp);
|
2012-03-20 20:27:57 +00:00
|
|
|
if (ret)
|
2013-05-07 23:18:41 +00:00
|
|
|
kill_ioctx(ioctx);
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_put(&ioctx->users);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* sys_io_destroy:
|
|
|
|
* Destroy the aio_context specified. May cancel any outstanding
|
|
|
|
* AIOs and block on completion. Will fail with -ENOSYS if not
|
2010-08-05 18:23:11 +00:00
|
|
|
* implemented. May fail with -EINVAL if the context pointed to
|
2005-04-16 22:20:36 +00:00
|
|
|
* is invalid.
|
|
|
|
*/
|
2009-01-14 13:14:18 +00:00
|
|
|
SYSCALL_DEFINE1(io_destroy, aio_context_t, ctx)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct kioctx *ioctx = lookup_ioctx(ctx);
|
|
|
|
if (likely(NULL != ioctx)) {
|
2013-05-07 23:18:41 +00:00
|
|
|
kill_ioctx(ioctx);
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_put(&ioctx->users);
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
pr_debug("EINVAL: io_destroy: invalid context id\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2006-10-01 06:28:49 +00:00
|
|
|
static void aio_advance_iovec(struct kiocb *iocb, ssize_t ret)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-10-01 06:28:49 +00:00
|
|
|
struct iovec *iov = &iocb->ki_iovec[iocb->ki_cur_seg];
|
|
|
|
|
|
|
|
BUG_ON(ret <= 0);
|
|
|
|
|
|
|
|
while (iocb->ki_cur_seg < iocb->ki_nr_segs && ret > 0) {
|
|
|
|
ssize_t this = min((ssize_t)iov->iov_len, ret);
|
|
|
|
iov->iov_base += this;
|
|
|
|
iov->iov_len -= this;
|
|
|
|
iocb->ki_left -= this;
|
|
|
|
ret -= this;
|
|
|
|
if (iov->iov_len == 0) {
|
|
|
|
iocb->ki_cur_seg++;
|
|
|
|
iov++;
|
2005-09-30 18:58:55 +00:00
|
|
|
}
|
2006-10-01 06:28:49 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-10-01 06:28:49 +00:00
|
|
|
/* the caller should not have done more io than what fit in
|
|
|
|
* the remaining iovecs */
|
|
|
|
BUG_ON(ret > 0 && iocb->ki_left == 0);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
typedef ssize_t (aio_rw_op)(struct kiocb *, const struct iovec *,
|
|
|
|
unsigned long, loff_t);
|
|
|
|
|
|
|
|
static ssize_t aio_rw_vect_retry(struct kiocb *iocb, int rw, aio_rw_op *rw_op)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct file *file = iocb->ki_filp;
|
2006-10-01 06:28:49 +00:00
|
|
|
struct address_space *mapping = file->f_mapping;
|
|
|
|
struct inode *inode = mapping->host;
|
2005-04-16 22:20:36 +00:00
|
|
|
ssize_t ret = 0;
|
|
|
|
|
2008-02-08 12:20:15 +00:00
|
|
|
/* This matches the pread()/pwrite() logic */
|
|
|
|
if (iocb->ki_pos < 0)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
if (rw == WRITE)
|
2013-03-20 01:01:03 +00:00
|
|
|
file_start_write(file);
|
2005-09-30 18:58:55 +00:00
|
|
|
do {
|
2006-10-01 06:28:49 +00:00
|
|
|
ret = rw_op(iocb, &iocb->ki_iovec[iocb->ki_cur_seg],
|
|
|
|
iocb->ki_nr_segs - iocb->ki_cur_seg,
|
|
|
|
iocb->ki_pos);
|
|
|
|
if (ret > 0)
|
|
|
|
aio_advance_iovec(iocb, ret);
|
|
|
|
|
|
|
|
/* retry all partial writes. retry partial reads as long as its a
|
|
|
|
* regular file. */
|
|
|
|
} while (ret > 0 && iocb->ki_left > 0 &&
|
2013-05-07 23:19:11 +00:00
|
|
|
(rw == WRITE ||
|
2006-10-01 06:28:49 +00:00
|
|
|
(!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))));
|
2013-05-07 23:19:11 +00:00
|
|
|
if (rw == WRITE)
|
2013-03-20 01:01:03 +00:00
|
|
|
file_end_write(file);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-10-01 06:28:49 +00:00
|
|
|
/* This means we must have transferred all that we could */
|
|
|
|
/* No need to retry anymore */
|
2005-04-16 22:20:36 +00:00
|
|
|
if ((ret == 0) || (iocb->ki_left == 0))
|
|
|
|
ret = iocb->ki_nbytes - iocb->ki_left;
|
|
|
|
|
2008-02-08 12:20:14 +00:00
|
|
|
/* If we managed to write some out we return that, rather than
|
|
|
|
* the eventual error. */
|
2013-05-07 23:19:11 +00:00
|
|
|
if (rw == WRITE
|
2013-05-07 23:18:25 +00:00
|
|
|
&& ret < 0 && ret != -EIOCBQUEUED
|
2008-02-08 12:20:14 +00:00
|
|
|
&& iocb->ki_nbytes - iocb->ki_left)
|
|
|
|
ret = iocb->ki_nbytes - iocb->ki_left;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
static ssize_t aio_setup_vectored_rw(int rw, struct kiocb *kiocb, bool compat)
|
2006-10-01 06:28:49 +00:00
|
|
|
{
|
|
|
|
ssize_t ret;
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
kiocb->ki_nr_segs = kiocb->ki_nbytes;
|
|
|
|
|
2010-05-26 21:44:26 +00:00
|
|
|
#ifdef CONFIG_COMPAT
|
|
|
|
if (compat)
|
2013-05-07 23:19:11 +00:00
|
|
|
ret = compat_rw_copy_check_uvector(rw,
|
2010-05-26 21:44:26 +00:00
|
|
|
(struct compat_iovec __user *)kiocb->ki_buf,
|
2013-05-07 23:19:11 +00:00
|
|
|
kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
|
2012-05-31 23:26:42 +00:00
|
|
|
&kiocb->ki_iovec);
|
2010-05-26 21:44:26 +00:00
|
|
|
else
|
|
|
|
#endif
|
2013-05-07 23:19:11 +00:00
|
|
|
ret = rw_copy_check_uvector(rw,
|
2010-05-26 21:44:26 +00:00
|
|
|
(struct iovec __user *)kiocb->ki_buf,
|
2013-05-07 23:19:11 +00:00
|
|
|
kiocb->ki_nr_segs, 1, &kiocb->ki_inline_vec,
|
2012-05-31 23:26:42 +00:00
|
|
|
&kiocb->ki_iovec);
|
2006-10-01 06:28:49 +00:00
|
|
|
if (ret < 0)
|
2013-05-07 23:19:11 +00:00
|
|
|
return ret;
|
2012-05-21 23:06:20 +00:00
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
/* ki_nbytes now reflect bytes instead of segs */
|
2006-10-01 06:28:49 +00:00
|
|
|
kiocb->ki_nbytes = ret;
|
2013-05-07 23:19:11 +00:00
|
|
|
return 0;
|
2006-10-01 06:28:49 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
static ssize_t aio_setup_single_vector(int rw, struct kiocb *kiocb)
|
2006-10-01 06:28:49 +00:00
|
|
|
{
|
2013-05-07 23:19:11 +00:00
|
|
|
if (unlikely(!access_ok(!rw, kiocb->ki_buf, kiocb->ki_nbytes)))
|
|
|
|
return -EFAULT;
|
2012-05-21 23:06:20 +00:00
|
|
|
|
2006-10-01 06:28:49 +00:00
|
|
|
kiocb->ki_iovec = &kiocb->ki_inline_vec;
|
|
|
|
kiocb->ki_iovec->iov_base = kiocb->ki_buf;
|
2013-05-07 23:19:11 +00:00
|
|
|
kiocb->ki_iovec->iov_len = kiocb->ki_nbytes;
|
2006-10-01 06:28:49 +00:00
|
|
|
kiocb->ki_nr_segs = 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* aio_setup_iocb:
|
|
|
|
* Performs the initial checks and aio retry method
|
|
|
|
* setup for the kiocb at the time of io submission.
|
|
|
|
*/
|
2013-05-07 23:19:11 +00:00
|
|
|
static ssize_t aio_run_iocb(struct kiocb *req, bool compat)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:19:11 +00:00
|
|
|
struct file *file = req->ki_filp;
|
|
|
|
ssize_t ret;
|
|
|
|
int rw;
|
|
|
|
fmode_t mode;
|
|
|
|
aio_rw_op *rw_op;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
switch (req->ki_opcode) {
|
2005-04-16 22:20:36 +00:00
|
|
|
case IOCB_CMD_PREAD:
|
2006-10-01 06:28:49 +00:00
|
|
|
case IOCB_CMD_PREADV:
|
2013-05-07 23:19:11 +00:00
|
|
|
mode = FMODE_READ;
|
|
|
|
rw = READ;
|
|
|
|
rw_op = file->f_op->aio_read;
|
|
|
|
goto rw_common;
|
|
|
|
|
|
|
|
case IOCB_CMD_PWRITE:
|
2006-10-01 06:28:49 +00:00
|
|
|
case IOCB_CMD_PWRITEV:
|
2013-05-07 23:19:11 +00:00
|
|
|
mode = FMODE_WRITE;
|
|
|
|
rw = WRITE;
|
|
|
|
rw_op = file->f_op->aio_write;
|
|
|
|
goto rw_common;
|
|
|
|
rw_common:
|
|
|
|
if (unlikely(!(file->f_mode & mode)))
|
|
|
|
return -EBADF;
|
|
|
|
|
|
|
|
if (!rw_op)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
ret = (req->ki_opcode == IOCB_CMD_PREADV ||
|
|
|
|
req->ki_opcode == IOCB_CMD_PWRITEV)
|
|
|
|
? aio_setup_vectored_rw(rw, req, compat)
|
|
|
|
: aio_setup_single_vector(rw, req);
|
2006-10-01 06:28:49 +00:00
|
|
|
if (ret)
|
2013-05-07 23:19:11 +00:00
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = rw_verify_area(rw, file, &req->ki_pos, req->ki_nbytes);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
req->ki_nbytes = ret;
|
|
|
|
req->ki_left = ret;
|
|
|
|
|
|
|
|
ret = aio_rw_vect_retry(req, rw, rw_op);
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
2013-05-07 23:19:11 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
case IOCB_CMD_FDSYNC:
|
2013-05-07 23:19:11 +00:00
|
|
|
if (!file->f_op->aio_fsync)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
ret = file->f_op->aio_fsync(req, 1);
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
2013-05-07 23:19:11 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
case IOCB_CMD_FSYNC:
|
2013-05-07 23:19:11 +00:00
|
|
|
if (!file->f_op->aio_fsync)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
ret = file->f_op->aio_fsync(req, 0);
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
2013-05-07 23:19:11 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
default:
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("EINVAL: no operation provided\n");
|
2013-05-07 23:19:11 +00:00
|
|
|
return -EINVAL;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
if (ret != -EIOCBQUEUED) {
|
|
|
|
/*
|
|
|
|
* There's no easy way to restart the syscall since other AIO's
|
|
|
|
* may be already running. Just fail this IO with EINTR.
|
|
|
|
*/
|
|
|
|
if (unlikely(ret == -ERESTARTSYS || ret == -ERESTARTNOINTR ||
|
|
|
|
ret == -ERESTARTNOHAND ||
|
|
|
|
ret == -ERESTART_RESTARTBLOCK))
|
|
|
|
ret = -EINTR;
|
|
|
|
aio_complete(req, ret, 0);
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-04-29 07:58:57 +00:00
|
|
|
static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
|
2013-05-07 23:18:53 +00:00
|
|
|
struct iocb *iocb, bool compat)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct kiocb *req;
|
|
|
|
ssize_t ret;
|
|
|
|
|
|
|
|
/* enforce forwards compatibility on users */
|
signal/timer/event: KAIO eventfd support example
This is an example about how to add eventfd support to the current KAIO code,
in order to enable KAIO to post readiness events to a pollable fd (hence
compatible with POSIX select/poll). The KAIO code simply signals the eventfd
fd when events are ready, and this triggers a POLLIN in the fd. This patch
uses a reserved for future use member of the struct iocb to pass an eventfd
file descriptor, that KAIO will use to post events every time a request
completes. At that point, an aio_getevents() will return the completed result
to a struct io_event. I made a quick test program to verify the patch, and it
runs fine here:
http://www.xmailserver.org/eventfd-aio-test.c
The test program uses poll(2), but it'd, of course, work with select and epoll
too.
This can allow to schedule both block I/O and other poll-able devices
requests, and wait for results using select/poll/epoll. In a typical
scenario, an application would submit KAIO request using aio_submit(), and
will also use epoll_ctl() on the whole other class of devices (that with the
addition of signals, timers and user events, now it's pretty much complete),
and then would:
epoll_wait(...);
for_each_event {
if (curr_event_is_kaiofd) {
aio_getevents();
dispatch_aio_events();
} else {
dispatch_epoll_event();
}
}
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-11 05:23:21 +00:00
|
|
|
if (unlikely(iocb->aio_reserved1 || iocb->aio_reserved2)) {
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("EINVAL: reserve field set\n");
|
2005-04-16 22:20:36 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* prevent overflows */
|
|
|
|
if (unlikely(
|
|
|
|
(iocb->aio_buf != (unsigned long)iocb->aio_buf) ||
|
|
|
|
(iocb->aio_nbytes != (size_t)iocb->aio_nbytes) ||
|
|
|
|
((ssize_t)iocb->aio_nbytes < 0)
|
|
|
|
)) {
|
|
|
|
pr_debug("EINVAL: io_submit: overflow check\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
req = aio_get_req(ctx);
|
2013-05-07 23:18:37 +00:00
|
|
|
if (unlikely(!req))
|
2005-04-16 22:20:36 +00:00
|
|
|
return -EAGAIN;
|
2013-05-07 23:18:37 +00:00
|
|
|
|
|
|
|
req->ki_filp = fget(iocb->aio_fildes);
|
|
|
|
if (unlikely(!req->ki_filp)) {
|
|
|
|
ret = -EBADF;
|
|
|
|
goto out_put_req;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2013-05-07 23:18:37 +00:00
|
|
|
|
signal/timer/event: KAIO eventfd support example
This is an example about how to add eventfd support to the current KAIO code,
in order to enable KAIO to post readiness events to a pollable fd (hence
compatible with POSIX select/poll). The KAIO code simply signals the eventfd
fd when events are ready, and this triggers a POLLIN in the fd. This patch
uses a reserved for future use member of the struct iocb to pass an eventfd
file descriptor, that KAIO will use to post events every time a request
completes. At that point, an aio_getevents() will return the completed result
to a struct io_event. I made a quick test program to verify the patch, and it
runs fine here:
http://www.xmailserver.org/eventfd-aio-test.c
The test program uses poll(2), but it'd, of course, work with select and epoll
too.
This can allow to schedule both block I/O and other poll-able devices
requests, and wait for results using select/poll/epoll. In a typical
scenario, an application would submit KAIO request using aio_submit(), and
will also use epoll_ctl() on the whole other class of devices (that with the
addition of signals, timers and user events, now it's pretty much complete),
and then would:
epoll_wait(...);
for_each_event {
if (curr_event_is_kaiofd) {
aio_getevents();
dispatch_aio_events();
} else {
dispatch_epoll_event();
}
}
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-11 05:23:21 +00:00
|
|
|
if (iocb->aio_flags & IOCB_FLAG_RESFD) {
|
|
|
|
/*
|
|
|
|
* If the IOCB_FLAG_RESFD flag of aio_flags is set, get an
|
|
|
|
* instance of the file* now. The file descriptor must be
|
|
|
|
* an eventfd() fd, and will be signaled for each completed
|
|
|
|
* event using the eventfd_signal() function.
|
|
|
|
*/
|
2009-06-30 18:41:11 +00:00
|
|
|
req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd);
|
2008-04-29 08:03:09 +00:00
|
|
|
if (IS_ERR(req->ki_eventfd)) {
|
signal/timer/event: KAIO eventfd support example
This is an example about how to add eventfd support to the current KAIO code,
in order to enable KAIO to post readiness events to a pollable fd (hence
compatible with POSIX select/poll). The KAIO code simply signals the eventfd
fd when events are ready, and this triggers a POLLIN in the fd. This patch
uses a reserved for future use member of the struct iocb to pass an eventfd
file descriptor, that KAIO will use to post events every time a request
completes. At that point, an aio_getevents() will return the completed result
to a struct io_event. I made a quick test program to verify the patch, and it
runs fine here:
http://www.xmailserver.org/eventfd-aio-test.c
The test program uses poll(2), but it'd, of course, work with select and epoll
too.
This can allow to schedule both block I/O and other poll-able devices
requests, and wait for results using select/poll/epoll. In a typical
scenario, an application would submit KAIO request using aio_submit(), and
will also use epoll_ctl() on the whole other class of devices (that with the
addition of signals, timers and user events, now it's pretty much complete),
and then would:
epoll_wait(...);
for_each_event {
if (curr_event_is_kaiofd) {
aio_getevents();
dispatch_aio_events();
} else {
dispatch_epoll_event();
}
}
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-11 05:23:21 +00:00
|
|
|
ret = PTR_ERR(req->ki_eventfd);
|
2009-03-19 00:04:19 +00:00
|
|
|
req->ki_eventfd = NULL;
|
signal/timer/event: KAIO eventfd support example
This is an example about how to add eventfd support to the current KAIO code,
in order to enable KAIO to post readiness events to a pollable fd (hence
compatible with POSIX select/poll). The KAIO code simply signals the eventfd
fd when events are ready, and this triggers a POLLIN in the fd. This patch
uses a reserved for future use member of the struct iocb to pass an eventfd
file descriptor, that KAIO will use to post events every time a request
completes. At that point, an aio_getevents() will return the completed result
to a struct io_event. I made a quick test program to verify the patch, and it
runs fine here:
http://www.xmailserver.org/eventfd-aio-test.c
The test program uses poll(2), but it'd, of course, work with select and epoll
too.
This can allow to schedule both block I/O and other poll-able devices
requests, and wait for results using select/poll/epoll. In a typical
scenario, an application would submit KAIO request using aio_submit(), and
will also use epoll_ctl() on the whole other class of devices (that with the
addition of signals, timers and user events, now it's pretty much complete),
and then would:
epoll_wait(...);
for_each_event {
if (curr_event_is_kaiofd) {
aio_getevents();
dispatch_aio_events();
} else {
dispatch_epoll_event();
}
}
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-11 05:23:21 +00:00
|
|
|
goto out_put_req;
|
|
|
|
}
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-07 23:19:10 +00:00
|
|
|
ret = put_user(KIOCB_KEY, &user_iocb->aio_key);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(ret)) {
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("EFAULT: aio_key\n");
|
2005-04-16 22:20:36 +00:00
|
|
|
goto out_put_req;
|
|
|
|
}
|
|
|
|
|
|
|
|
req->ki_obj.user = user_iocb;
|
|
|
|
req->ki_user_data = iocb->aio_data;
|
|
|
|
req->ki_pos = iocb->aio_offset;
|
|
|
|
|
|
|
|
req->ki_buf = (char __user *)(unsigned long)iocb->aio_buf;
|
|
|
|
req->ki_left = req->ki_nbytes = iocb->aio_nbytes;
|
|
|
|
req->ki_opcode = iocb->aio_lio_opcode;
|
|
|
|
|
2013-05-07 23:19:11 +00:00
|
|
|
ret = aio_run_iocb(req, compat);
|
2013-05-07 23:18:25 +00:00
|
|
|
if (ret)
|
2011-02-25 22:44:27 +00:00
|
|
|
goto out_put_req;
|
2013-05-07 23:18:25 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
aio_put_req(req); /* drop extra ref to req */
|
|
|
|
return 0;
|
|
|
|
out_put_req:
|
2013-04-26 00:58:39 +00:00
|
|
|
put_reqs_available(ctx, 1);
|
2005-04-16 22:20:36 +00:00
|
|
|
aio_put_req(req); /* drop extra ref to req */
|
|
|
|
aio_put_req(req); /* drop i/o ref to req */
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-05-26 21:44:26 +00:00
|
|
|
long do_io_submit(aio_context_t ctx_id, long nr,
|
|
|
|
struct iocb __user *__user *iocbpp, bool compat)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct kioctx *ctx;
|
|
|
|
long ret = 0;
|
2011-11-02 20:40:10 +00:00
|
|
|
int i = 0;
|
2010-07-01 05:55:01 +00:00
|
|
|
struct blk_plug plug;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
if (unlikely(nr < 0))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2010-09-10 21:16:00 +00:00
|
|
|
if (unlikely(nr > LONG_MAX/sizeof(*iocbpp)))
|
|
|
|
nr = LONG_MAX/sizeof(*iocbpp);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(!access_ok(VERIFY_READ, iocbpp, (nr*sizeof(*iocbpp)))))
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
ctx = lookup_ioctx(ctx_id);
|
|
|
|
if (unlikely(!ctx)) {
|
2013-05-07 23:18:35 +00:00
|
|
|
pr_debug("EINVAL: invalid context id\n");
|
2005-04-16 22:20:36 +00:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2010-07-01 05:55:01 +00:00
|
|
|
blk_start_plug(&plug);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* AKPM: should this return a partial result if some of the IOs were
|
|
|
|
* successfully submitted?
|
|
|
|
*/
|
|
|
|
for (i=0; i<nr; i++) {
|
|
|
|
struct iocb __user *user_iocb;
|
|
|
|
struct iocb tmp;
|
|
|
|
|
|
|
|
if (unlikely(__get_user(user_iocb, iocbpp + i))) {
|
|
|
|
ret = -EFAULT;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(copy_from_user(&tmp, user_iocb, sizeof(tmp)))) {
|
|
|
|
ret = -EFAULT;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2013-05-07 23:18:53 +00:00
|
|
|
ret = io_submit_one(ctx, user_iocb, &tmp, compat);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
}
|
2010-07-01 05:55:01 +00:00
|
|
|
blk_finish_plug(&plug);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_put(&ctx->users);
|
2005-04-16 22:20:36 +00:00
|
|
|
return i ? i : ret;
|
|
|
|
}
|
|
|
|
|
2010-05-26 21:44:26 +00:00
|
|
|
/* sys_io_submit:
|
|
|
|
* Queue the nr iocbs pointed to by iocbpp for processing. Returns
|
|
|
|
* the number of iocbs queued. May return -EINVAL if the aio_context
|
|
|
|
* specified by ctx_id is invalid, if nr is < 0, if the iocb at
|
|
|
|
* *iocbpp[0] is not properly initialized, if the operation specified
|
|
|
|
* is invalid for the file descriptor in the iocb. May fail with
|
|
|
|
* -EFAULT if any of the data structures point to invalid data. May
|
|
|
|
* fail with -EBADF if the file descriptor specified in the first
|
|
|
|
* iocb is invalid. May fail with -EAGAIN if insufficient resources
|
|
|
|
* are available to queue any iocbs. Will return 0 if nr is 0. Will
|
|
|
|
* fail with -ENOSYS if not implemented.
|
|
|
|
*/
|
|
|
|
SYSCALL_DEFINE3(io_submit, aio_context_t, ctx_id, long, nr,
|
|
|
|
struct iocb __user * __user *, iocbpp)
|
|
|
|
{
|
|
|
|
return do_io_submit(ctx_id, nr, iocbpp, 0);
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* lookup_kiocb
|
|
|
|
* Finds a given iocb for cancellation.
|
|
|
|
*/
|
2005-04-25 15:18:14 +00:00
|
|
|
static struct kiocb *lookup_kiocb(struct kioctx *ctx, struct iocb __user *iocb,
|
|
|
|
u32 key)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct list_head *pos;
|
2005-11-14 00:07:34 +00:00
|
|
|
|
|
|
|
assert_spin_locked(&ctx->ctx_lock);
|
|
|
|
|
2013-05-07 23:19:10 +00:00
|
|
|
if (key != KIOCB_KEY)
|
|
|
|
return NULL;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* TODO: use a hash or array, this sucks. */
|
|
|
|
list_for_each(pos, &ctx->active_reqs) {
|
|
|
|
struct kiocb *kiocb = list_kiocb(pos);
|
2013-05-07 23:19:10 +00:00
|
|
|
if (kiocb->ki_obj.user == iocb)
|
2005-04-16 22:20:36 +00:00
|
|
|
return kiocb;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* sys_io_cancel:
|
|
|
|
* Attempts to cancel an iocb previously passed to io_submit. If
|
|
|
|
* the operation is successfully cancelled, the resulting event is
|
|
|
|
* copied into the memory pointed to by result without being placed
|
|
|
|
* into the completion queue and 0 is returned. May fail with
|
|
|
|
* -EFAULT if any of the data structures pointed to are invalid.
|
|
|
|
* May fail with -EINVAL if aio_context specified by ctx_id is
|
|
|
|
* invalid. May fail with -EAGAIN if the iocb specified was not
|
|
|
|
* cancelled. Will fail with -ENOSYS if not implemented.
|
|
|
|
*/
|
2009-01-14 13:14:18 +00:00
|
|
|
SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb,
|
|
|
|
struct io_event __user *, result)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2013-05-07 23:18:31 +00:00
|
|
|
struct io_event res;
|
2005-04-16 22:20:36 +00:00
|
|
|
struct kioctx *ctx;
|
|
|
|
struct kiocb *kiocb;
|
|
|
|
u32 key;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = get_user(key, &iocb->aio_key);
|
|
|
|
if (unlikely(ret))
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
ctx = lookup_ioctx(ctx_id);
|
|
|
|
if (unlikely(!ctx))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
spin_lock_irq(&ctx->ctx_lock);
|
2013-05-07 23:18:31 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
kiocb = lookup_kiocb(ctx, iocb, key);
|
2013-05-07 23:18:31 +00:00
|
|
|
if (kiocb)
|
|
|
|
ret = kiocb_cancel(ctx, kiocb, &res);
|
|
|
|
else
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
spin_unlock_irq(&ctx->ctx_lock);
|
|
|
|
|
2013-05-07 23:18:31 +00:00
|
|
|
if (!ret) {
|
|
|
|
/* Cancellation succeeded -- copy the result
|
|
|
|
* into the user's buffer.
|
|
|
|
*/
|
|
|
|
if (copy_to_user(result, &res, sizeof(res)))
|
|
|
|
ret = -EFAULT;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_put(&ctx->users);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* io_getevents:
|
|
|
|
* Attempts to read at least min_nr events and up to nr events from
|
2010-08-05 18:23:11 +00:00
|
|
|
* the completion queue for the aio_context specified by ctx_id. If
|
|
|
|
* it succeeds, the number of read events is returned. May fail with
|
|
|
|
* -EINVAL if ctx_id is invalid, if min_nr is out of range, if nr is
|
|
|
|
* out of range, if timeout is out of range. May fail with -EFAULT
|
|
|
|
* if any of the memory specified is invalid. May return 0 or
|
|
|
|
* < min_nr if the timeout specified by timeout has elapsed
|
|
|
|
* before sufficient events are available, where timeout == NULL
|
|
|
|
* specifies an infinite timeout. Note that the timeout pointed to by
|
2013-05-24 22:55:24 +00:00
|
|
|
* timeout is relative. Will fail with -ENOSYS if not implemented.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2009-01-14 13:14:18 +00:00
|
|
|
SYSCALL_DEFINE5(io_getevents, aio_context_t, ctx_id,
|
|
|
|
long, min_nr,
|
|
|
|
long, nr,
|
|
|
|
struct io_event __user *, events,
|
|
|
|
struct timespec __user *, timeout)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
struct kioctx *ioctx = lookup_ioctx(ctx_id);
|
|
|
|
long ret = -EINVAL;
|
|
|
|
|
|
|
|
if (likely(ioctx)) {
|
2011-01-13 01:01:08 +00:00
|
|
|
if (likely(min_nr <= nr && min_nr >= 0))
|
2005-04-16 22:20:36 +00:00
|
|
|
ret = read_events(ioctx, min_nr, nr, events, timeout);
|
2013-05-28 22:14:48 +00:00
|
|
|
percpu_ref_put(&ioctx->users);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|