third_party_f2fs-tools/fsck/quotaio_v2.c
Jaegeuk Kim 83c64019e8 fsck.f2fs: don't allocate new blocks on unclean shutdown
We have to keep data for roll-forward recovery. Without this patch, we're
able to lose there-in data by quota overwrites.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-11-21 11:38:23 -08:00

285 lines
7.6 KiB
C

/*
* Implementation of new quotafile format
*
* Jan Kara <jack@suse.cz> - sponsored by SuSE CR
* Hyojun Kim <hyojun@google.com> - Ported to f2fs-tools
*/
#include "config.h"
#include <sys/types.h>
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "common.h"
#include "quotaio_v2.h"
#include "dqblk_v2.h"
#include "quotaio_tree.h"
static int v2_check_file(struct quota_handle *h, int type);
static int v2_init_io(struct quota_handle *h);
static int v2_new_io(struct quota_handle *h);
static int v2_write_info(struct quota_handle *h);
static struct dquot *v2_read_dquot(struct quota_handle *h, qid_t id);
static int v2_commit_dquot(struct dquot *dquot);
static int v2_scan_dquots(struct quota_handle *h,
int (*process_dquot) (struct dquot *dquot,
void *data),
void *data);
static int v2_report(struct quota_handle *h, int verbose);
struct quotafile_ops quotafile_ops_2 = {
.check_file = v2_check_file,
.init_io = v2_init_io,
.new_io = v2_new_io,
.write_info = v2_write_info,
.read_dquot = v2_read_dquot,
.commit_dquot = v2_commit_dquot,
.scan_dquots = v2_scan_dquots,
.report = v2_report,
};
/*
* Copy dquot from disk to memory
*/
static void v2r1_disk2memdqblk(struct dquot *dquot, void *dp)
{
struct util_dqblk *m = &dquot->dq_dqb;
struct v2r1_disk_dqblk *d = dp, empty;
dquot->dq_id = le32_to_cpu(d->dqb_id);
m->dqb_ihardlimit = le64_to_cpu(d->dqb_ihardlimit);
m->dqb_isoftlimit = le64_to_cpu(d->dqb_isoftlimit);
m->dqb_bhardlimit = le64_to_cpu(d->dqb_bhardlimit);
m->dqb_bsoftlimit = le64_to_cpu(d->dqb_bsoftlimit);
m->dqb_curinodes = le64_to_cpu(d->dqb_curinodes);
m->dqb_curspace = le64_to_cpu(d->dqb_curspace);
m->dqb_itime = le64_to_cpu(d->dqb_itime);
m->dqb_btime = le64_to_cpu(d->dqb_btime);
memset(&empty, 0, sizeof(struct v2r1_disk_dqblk));
empty.dqb_itime = cpu_to_le64(1);
if (!memcmp(&empty, dp, sizeof(struct v2r1_disk_dqblk)))
m->dqb_itime = 0;
}
/*
* Copy dquot from memory to disk
*/
static void v2r1_mem2diskdqblk(void *dp, struct dquot *dquot)
{
struct util_dqblk *m = &dquot->dq_dqb;
struct v2r1_disk_dqblk *d = dp;
d->dqb_ihardlimit = cpu_to_le64(m->dqb_ihardlimit);
d->dqb_isoftlimit = cpu_to_le64(m->dqb_isoftlimit);
d->dqb_bhardlimit = cpu_to_le64(m->dqb_bhardlimit);
d->dqb_bsoftlimit = cpu_to_le64(m->dqb_bsoftlimit);
d->dqb_curinodes = cpu_to_le64(m->dqb_curinodes);
d->dqb_curspace = cpu_to_le64(m->dqb_curspace);
d->dqb_itime = cpu_to_le64(m->dqb_itime);
d->dqb_btime = cpu_to_le64(m->dqb_btime);
d->dqb_id = cpu_to_le32(dquot->dq_id);
if (qtree_entry_unused(&dquot->dq_h->qh_info.u.v2_mdqi.dqi_qtree, dp))
d->dqb_itime = cpu_to_le64(1);
}
static int v2r1_is_id(void *dp, struct dquot *dquot)
{
struct v2r1_disk_dqblk *d = dp;
struct qtree_mem_dqinfo *info =
&dquot->dq_h->qh_info.u.v2_mdqi.dqi_qtree;
if (qtree_entry_unused(info, dp))
return 0;
return le32_to_cpu(d->dqb_id) == dquot->dq_id;
}
static struct qtree_fmt_operations v2r1_fmt_ops = {
.mem2disk_dqblk = v2r1_mem2diskdqblk,
.disk2mem_dqblk = v2r1_disk2memdqblk,
.is_id = v2r1_is_id,
};
/*
* Copy dqinfo from disk to memory
*/
static inline void v2_disk2memdqinfo(struct util_dqinfo *m,
struct v2_disk_dqinfo *d)
{
m->dqi_bgrace = le32_to_cpu(d->dqi_bgrace);
m->dqi_igrace = le32_to_cpu(d->dqi_igrace);
m->u.v2_mdqi.dqi_flags = le32_to_cpu(d->dqi_flags) & V2_DQF_MASK;
m->u.v2_mdqi.dqi_qtree.dqi_blocks = le32_to_cpu(d->dqi_blocks);
m->u.v2_mdqi.dqi_qtree.dqi_free_blk =
le32_to_cpu(d->dqi_free_blk);
m->u.v2_mdqi.dqi_qtree.dqi_free_entry =
le32_to_cpu(d->dqi_free_entry);
}
/*
* Copy dqinfo from memory to disk
*/
static inline void v2_mem2diskdqinfo(struct v2_disk_dqinfo *d,
struct util_dqinfo *m)
{
d->dqi_bgrace = cpu_to_le32(m->dqi_bgrace);
d->dqi_igrace = cpu_to_le32(m->dqi_igrace);
d->dqi_flags = cpu_to_le32(m->u.v2_mdqi.dqi_flags & V2_DQF_MASK);
d->dqi_blocks = cpu_to_le32(m->u.v2_mdqi.dqi_qtree.dqi_blocks);
d->dqi_free_blk =
cpu_to_le32(m->u.v2_mdqi.dqi_qtree.dqi_free_blk);
d->dqi_free_entry =
cpu_to_le32(m->u.v2_mdqi.dqi_qtree.dqi_free_entry);
}
static int v2_read_header(struct quota_handle *h, struct v2_disk_dqheader *dqh)
{
if (h->read(&h->qh_qf, 0, dqh, sizeof(struct v2_disk_dqheader)) !=
sizeof(struct v2_disk_dqheader))
return 0;
return 1;
}
/*
* Check whether given quota file is in our format
*/
static int v2_check_file(struct quota_handle *h, int type)
{
struct v2_disk_dqheader dqh;
int file_magics[] = INITQMAGICS;
int be_magic;
if (!v2_read_header(h, &dqh))
return 0;
be_magic = be32_to_cpu((__force __be32)dqh.dqh_magic);
if (be_magic == file_magics[type]) {
log_err("Your quota file is stored in wrong endianity");
return 0;
}
if (V2_VERSION != le32_to_cpu(dqh.dqh_version))
return 0;
return 1;
}
/*
* Open quotafile
*/
static int v2_init_io(struct quota_handle *h)
{
struct v2_disk_dqinfo ddqinfo;
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_entry_size =
sizeof(struct v2r1_disk_dqblk);
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_ops = &v2r1_fmt_ops;
/* Read information about quotafile */
if (h->read(&h->qh_qf, V2_DQINFOOFF, &ddqinfo,
sizeof(ddqinfo)) != sizeof(ddqinfo))
return -1;
v2_disk2memdqinfo(&h->qh_info, &ddqinfo);
return 0;
}
/*
* Initialize new quotafile
*/
static int v2_new_io(struct quota_handle *h)
{
int file_magics[] = INITQMAGICS;
struct v2_disk_dqheader ddqheader;
struct v2_disk_dqinfo ddqinfo;
if (h->qh_fmt != QFMT_VFS_V1)
return -1;
/* Write basic quota header */
ddqheader.dqh_magic = cpu_to_le32(file_magics[h->qh_type]);
ddqheader.dqh_version = cpu_to_le32(V2_VERSION);
if (h->write(&h->qh_qf, 0, &ddqheader, sizeof(ddqheader)) !=
sizeof(ddqheader))
return -1;
/* Write information about quotafile */
h->qh_info.dqi_bgrace = MAX_DQ_TIME;
h->qh_info.dqi_igrace = MAX_IQ_TIME;
h->qh_info.u.v2_mdqi.dqi_flags = 0;
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_blocks = QT_TREEOFF + 1;
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_free_blk = 0;
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_free_entry = 0;
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_entry_size =
sizeof(struct v2r1_disk_dqblk);
h->qh_info.u.v2_mdqi.dqi_qtree.dqi_ops = &v2r1_fmt_ops;
v2_mem2diskdqinfo(&ddqinfo, &h->qh_info);
if (h->write(&h->qh_qf, V2_DQINFOOFF, &ddqinfo,
sizeof(ddqinfo)) !=
sizeof(ddqinfo))
return -1;
return 0;
}
/*
* Write information (grace times to file)
*/
static int v2_write_info(struct quota_handle *h)
{
struct v2_disk_dqinfo ddqinfo;
v2_mem2diskdqinfo(&ddqinfo, &h->qh_info);
if (h->write(&h->qh_qf, V2_DQINFOOFF, &ddqinfo, sizeof(ddqinfo)) !=
sizeof(ddqinfo))
return -1;
return 0;
}
/*
* Read dquot from disk
*/
static struct dquot *v2_read_dquot(struct quota_handle *h, qid_t id)
{
return qtree_read_dquot(h, id);
}
/*
* Commit changes of dquot to disk - it might also mean deleting it when quota
* became fake one and user has no blocks.
* User can process use 'errno' to detect errstr.
*/
static int v2_commit_dquot(struct dquot *dquot)
{
struct util_dqblk *b = &dquot->dq_dqb;
if (!b->dqb_curspace && !b->dqb_curinodes && !b->dqb_bsoftlimit &&
!b->dqb_isoftlimit && !b->dqb_bhardlimit && !b->dqb_ihardlimit)
{
qtree_delete_dquot(dquot);
} else {
return qtree_write_dquot(dquot);
}
return 0;
}
static int v2_scan_dquots(struct quota_handle *h,
int (*process_dquot) (struct dquot *, void *),
void *data)
{
return qtree_scan_dquots(h, process_dquot, data);
}
/* Report information about quotafile.
* TODO: Not used right now, but we should be able to use this when we add
* support to debugfs to read quota files.
*/
static int v2_report(struct quota_handle *UNUSED(h), int UNUSED(verbose))
{
log_err("Not Implemented.");
return -1;
}