* [PATCH 1/3] exfat: add error check when updating dir-entries @ 2020-06-04 8:44 Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 2/3] exfat: optimize exfat_zeroed_cluster() Tetsuhiro Kohada ` (2 more replies) 0 siblings, 3 replies; 10+ messages in thread From: Tetsuhiro Kohada @ 2020-06-04 8:44 UTC (permalink / raw) To: kohada.t2 Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, Sungjong Seo, Namjae Jeon, linux-fsdevel, linux-kernel Add error check when synchronously updating dir-entries. Furthermore, add exfat_update_bhs(). It wait for write completion once instead of sector by sector. Suggested-by: Sungjong Seo <sj1557.seo@samsung.com> Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> --- fs/exfat/dir.c | 15 +++++++++------ fs/exfat/exfat_fs.h | 3 ++- fs/exfat/file.c | 5 ++++- fs/exfat/inode.c | 8 +++++--- fs/exfat/misc.c | 19 +++++++++++++++++++ 5 files changed, 39 insertions(+), 11 deletions(-) diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c index de43534aa299..3eb8386fb5f2 100644 --- a/fs/exfat/dir.c +++ b/fs/exfat/dir.c @@ -602,16 +602,19 @@ void exfat_update_dir_chksum_with_entry_set(struct exfat_entry_set_cache *es) es->modified = true; } -void exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) +int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) { - int i; + int i, err = 0; - for (i = 0; i < es->num_bh; i++) { - if (es->modified) - exfat_update_bh(es->sb, es->bh[i], sync); - brelse(es->bh[i]); + if (es->modified) { + set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(es->sb)->s_state); + err = exfat_update_bhs(es->bh, es->num_bh, sync); } + + for (i = 0; i < es->num_bh; i++) + err ? bforget(es->bh[i]):brelse(es->bh[i]); kfree(es); + return err; } static int exfat_walk_fat_chain(struct super_block *sb, diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h index 595f3117f492..f4fa0e833486 100644 --- a/fs/exfat/exfat_fs.h +++ b/fs/exfat/exfat_fs.h @@ -462,7 +462,7 @@ struct exfat_dentry *exfat_get_dentry_cached(struct exfat_entry_set_cache *es, int num); struct exfat_entry_set_cache *exfat_get_dentry_set(struct super_block *sb, struct exfat_chain *p_dir, int entry, unsigned int type); -void exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync); +int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync); int exfat_count_dir_entries(struct super_block *sb, struct exfat_chain *p_dir); /* inode.c */ @@ -515,6 +515,7 @@ void exfat_set_entry_time(struct exfat_sb_info *sbi, struct timespec64 *ts, u16 exfat_calc_chksum16(void *data, int len, u16 chksum, int type); u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type); void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int sync); +int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync); void exfat_chain_set(struct exfat_chain *ec, unsigned int dir, unsigned int size, unsigned char flags); void exfat_chain_dup(struct exfat_chain *dup, struct exfat_chain *ec); diff --git a/fs/exfat/file.c b/fs/exfat/file.c index fce03f318787..37c8f04c1f8a 100644 --- a/fs/exfat/file.c +++ b/fs/exfat/file.c @@ -153,6 +153,7 @@ int __exfat_truncate(struct inode *inode, loff_t new_size) struct timespec64 ts; struct exfat_dentry *ep, *ep2; struct exfat_entry_set_cache *es; + int err; es = exfat_get_dentry_set(sb, &(ei->dir), ei->entry, ES_ALL_ENTRIES); @@ -187,7 +188,9 @@ int __exfat_truncate(struct inode *inode, loff_t new_size) } exfat_update_dir_chksum_with_entry_set(es); - exfat_free_dentry_set(es, inode_needs_sync(inode)); + err = exfat_free_dentry_set(es, inode_needs_sync(inode)); + if (err) + return err; } /* cut off from the FAT chain */ diff --git a/fs/exfat/inode.c b/fs/exfat/inode.c index ef7cf7a6d187..c0bfd1a586aa 100644 --- a/fs/exfat/inode.c +++ b/fs/exfat/inode.c @@ -77,8 +77,7 @@ static int __exfat_write_inode(struct inode *inode, int sync) ep2->dentry.stream.size = ep2->dentry.stream.valid_size; exfat_update_dir_chksum_with_entry_set(es); - exfat_free_dentry_set(es, sync); - return 0; + return exfat_free_dentry_set(es, sync); } int exfat_write_inode(struct inode *inode, struct writeback_control *wbc) @@ -222,6 +221,7 @@ static int exfat_map_cluster(struct inode *inode, unsigned int clu_offset, if (ei->dir.dir != DIR_DELETED && modified) { struct exfat_dentry *ep; struct exfat_entry_set_cache *es; + int err; es = exfat_get_dentry_set(sb, &(ei->dir), ei->entry, ES_ALL_ENTRIES); @@ -240,7 +240,9 @@ static int exfat_map_cluster(struct inode *inode, unsigned int clu_offset, ep->dentry.stream.valid_size; exfat_update_dir_chksum_with_entry_set(es); - exfat_free_dentry_set(es, inode_needs_sync(inode)); + err = exfat_free_dentry_set(es, inode_needs_sync(inode)); + if (err) + return err; } /* end of if != DIR_DELETED */ diff --git a/fs/exfat/misc.c b/fs/exfat/misc.c index 17d41f3d3709..dc34968e99d3 100644 --- a/fs/exfat/misc.c +++ b/fs/exfat/misc.c @@ -173,6 +173,25 @@ void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int sync) sync_dirty_buffer(bh); } +int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync) +{ + int i, err = 0; + + for (i = 0; i < nr_bhs; i++) { + set_buffer_uptodate(bhs[i]); + mark_buffer_dirty(bhs[i]); + if (sync) + write_dirty_buffer(bhs[i], 0); + } + + for (i = 0; i < nr_bhs && sync; i++) { + wait_on_buffer(bhs[i]); + if (!buffer_uptodate(bhs[i])) + err = -EIO; + } + return err; +} + void exfat_chain_set(struct exfat_chain *ec, unsigned int dir, unsigned int size, unsigned char flags) { -- 2.25.1 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 2/3] exfat: optimize exfat_zeroed_cluster() 2020-06-04 8:44 [PATCH 1/3] exfat: add error check when updating dir-entries Tetsuhiro Kohada @ 2020-06-04 8:44 ` Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing Tetsuhiro Kohada 2020-06-07 0:33 ` [PATCH 1/3] exfat: add error check when updating dir-entries Namjae Jeon 2 siblings, 0 replies; 10+ messages in thread From: Tetsuhiro Kohada @ 2020-06-04 8:44 UTC (permalink / raw) To: kohada.t2 Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, Namjae Jeon, Sungjong Seo, linux-fsdevel, linux-kernel Replace part of exfat_zeroed_cluster() with exfat_update_bhs(). And remove exfat_sync_bhs(). Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> --- fs/exfat/fatent.c | 54 ++++++++++------------------------------------- 1 file changed, 11 insertions(+), 43 deletions(-) diff --git a/fs/exfat/fatent.c b/fs/exfat/fatent.c index 267e5e09eb13..5d11bc2f1b68 100644 --- a/fs/exfat/fatent.c +++ b/fs/exfat/fatent.c @@ -230,21 +230,6 @@ int exfat_find_last_cluster(struct super_block *sb, struct exfat_chain *p_chain, return 0; } -static inline int exfat_sync_bhs(struct buffer_head **bhs, int nr_bhs) -{ - int i, err = 0; - - for (i = 0; i < nr_bhs; i++) - write_dirty_buffer(bhs[i], 0); - - for (i = 0; i < nr_bhs; i++) { - wait_on_buffer(bhs[i]); - if (!err && !buffer_uptodate(bhs[i])) - err = -EIO; - } - return err; -} - int exfat_zeroed_cluster(struct inode *dir, unsigned int clu) { struct super_block *sb = dir->i_sb; @@ -266,41 +251,24 @@ int exfat_zeroed_cluster(struct inode *dir, unsigned int clu) } /* Zeroing the unused blocks on this cluster */ - n = 0; while (blknr < last_blknr) { - bhs[n] = sb_getblk(sb, blknr); - if (!bhs[n]) { - err = -ENOMEM; - goto release_bhs; - } - memset(bhs[n]->b_data, 0, sb->s_blocksize); - exfat_update_bh(sb, bhs[n], 0); - - n++; - blknr++; - - if (n == nr_bhs) { - if (IS_DIRSYNC(dir)) { - err = exfat_sync_bhs(bhs, n); - if (err) - goto release_bhs; + for (n = 0; n < nr_bhs && blknr < last_blknr; n++, blknr++) { + bhs[n] = sb_getblk(sb, blknr); + if (!bhs[n]) { + err = -ENOMEM; + goto release_bhs; } - - for (i = 0; i < n; i++) - brelse(bhs[i]); - n = 0; + memset(bhs[n]->b_data, 0, sb->s_blocksize); } - } - if (IS_DIRSYNC(dir)) { - err = exfat_sync_bhs(bhs, n); + set_bit(EXFAT_SB_DIRTY, &sbi->s_state); + err = exfat_update_bhs(bhs, n, IS_DIRSYNC(dir)); if (err) goto release_bhs; - } - - for (i = 0; i < n; i++) - brelse(bhs[i]); + for (i = 0; i < n; i++) + brelse(bhs[i]); + } return 0; release_bhs: -- 2.25.1 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing 2020-06-04 8:44 [PATCH 1/3] exfat: add error check when updating dir-entries Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 2/3] exfat: optimize exfat_zeroed_cluster() Tetsuhiro Kohada @ 2020-06-04 8:44 ` Tetsuhiro Kohada 2020-06-05 7:32 ` Namjae Jeon 2020-06-07 0:33 ` [PATCH 1/3] exfat: add error check when updating dir-entries Namjae Jeon 2 siblings, 1 reply; 10+ messages in thread From: Tetsuhiro Kohada @ 2020-06-04 8:44 UTC (permalink / raw) To: kohada.t2 Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, Namjae Jeon, Sungjong Seo, linux-fsdevel, linux-kernel Set EXFAT_SB_DIRTY flag in exfat_put_super(). In some cases, can't clear VOL_DIRTY with 'sync'. ex: VOL_DIRTY is set when rmdir starts, but when non-empty-dir is detected, return error without setting EXFAT_SB_DIRTY. If performe 'sync' in this state, VOL_DIRTY will not be cleared. Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> --- fs/exfat/balloc.c | 4 ++-- fs/exfat/dir.c | 18 ++++++++---------- fs/exfat/exfat_fs.h | 2 +- fs/exfat/fatent.c | 6 +----- fs/exfat/misc.c | 3 +-- fs/exfat/namei.c | 12 ++++++------ fs/exfat/super.c | 3 +++ 7 files changed, 22 insertions(+), 26 deletions(-) diff --git a/fs/exfat/balloc.c b/fs/exfat/balloc.c index 4055eb00ea9b..a987919686c0 100644 --- a/fs/exfat/balloc.c +++ b/fs/exfat/balloc.c @@ -158,7 +158,7 @@ int exfat_set_bitmap(struct inode *inode, unsigned int clu) b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); set_bit_le(b, sbi->vol_amap[i]->b_data); - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); return 0; } @@ -180,7 +180,7 @@ void exfat_clear_bitmap(struct inode *inode, unsigned int clu) b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); clear_bit_le(b, sbi->vol_amap[i]->b_data); - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); if (opts->discard) { int ret_discard; diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c index 3eb8386fb5f2..96c9a817d928 100644 --- a/fs/exfat/dir.c +++ b/fs/exfat/dir.c @@ -468,7 +468,7 @@ int exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, &ep->dentry.file.access_date, NULL); - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); + exfat_update_bh(bh, IS_DIRSYNC(inode)); brelse(bh); ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -478,7 +478,7 @@ int exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, exfat_init_stream_entry(ep, (type == TYPE_FILE) ? ALLOC_FAT_CHAIN : ALLOC_NO_FAT_CHAIN, start_clu, size); - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); + exfat_update_bh(bh, IS_DIRSYNC(inode)); brelse(bh); return 0; @@ -514,7 +514,7 @@ int exfat_update_dir_chksum(struct inode *inode, struct exfat_chain *p_dir, } fep->dentry.file.checksum = cpu_to_le16(chksum); - exfat_update_bh(sb, fbh, IS_DIRSYNC(inode)); + exfat_update_bh(fbh, IS_DIRSYNC(inode)); release_fbh: brelse(fbh); return ret; @@ -536,7 +536,7 @@ int exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, return -EIO; ep->dentry.file.num_ext = (unsigned char)(num_entries - 1); - exfat_update_bh(sb, bh, sync); + exfat_update_bh(bh, sync); brelse(bh); ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -545,7 +545,7 @@ int exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, ep->dentry.stream.name_len = p_uniname->name_len; ep->dentry.stream.name_hash = cpu_to_le16(p_uniname->name_hash); - exfat_update_bh(sb, bh, sync); + exfat_update_bh(bh, sync); brelse(bh); for (i = EXFAT_FIRST_CLUSTER; i < num_entries; i++) { @@ -554,7 +554,7 @@ int exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, return -EIO; exfat_init_name_entry(ep, uniname); - exfat_update_bh(sb, bh, sync); + exfat_update_bh(bh, sync); brelse(bh); uniname += EXFAT_FILE_NAME_LEN; } @@ -578,7 +578,7 @@ int exfat_remove_entries(struct inode *inode, struct exfat_chain *p_dir, return -EIO; exfat_set_entry_type(ep, TYPE_DELETED); - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); + exfat_update_bh(bh, IS_DIRSYNC(inode)); brelse(bh); } @@ -606,10 +606,8 @@ int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) { int i, err = 0; - if (es->modified) { - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(es->sb)->s_state); + if (es->modified) err = exfat_update_bhs(es->bh, es->num_bh, sync); - } for (i = 0; i < es->num_bh; i++) err ? bforget(es->bh[i]):brelse(es->bh[i]); diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h index f4fa0e833486..0e094d186612 100644 --- a/fs/exfat/exfat_fs.h +++ b/fs/exfat/exfat_fs.h @@ -514,7 +514,7 @@ void exfat_set_entry_time(struct exfat_sb_info *sbi, struct timespec64 *ts, u8 *tz, __le16 *time, __le16 *date, u8 *time_cs); u16 exfat_calc_chksum16(void *data, int len, u16 chksum, int type); u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type); -void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int sync); +void exfat_update_bh(struct buffer_head *bh, int sync); int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync); void exfat_chain_set(struct exfat_chain *ec, unsigned int dir, unsigned int size, unsigned char flags); diff --git a/fs/exfat/fatent.c b/fs/exfat/fatent.c index 5d11bc2f1b68..f8171183b4c1 100644 --- a/fs/exfat/fatent.c +++ b/fs/exfat/fatent.c @@ -75,7 +75,7 @@ int exfat_ent_set(struct super_block *sb, unsigned int loc, fat_entry = (__le32 *)&(bh->b_data[off]); *fat_entry = cpu_to_le32(content); - exfat_update_bh(sb, bh, sb->s_flags & SB_SYNCHRONOUS); + exfat_update_bh(bh, sb->s_flags & SB_SYNCHRONOUS); exfat_mirror_bh(sb, sec, bh); brelse(bh); return 0; @@ -174,7 +174,6 @@ int exfat_free_cluster(struct inode *inode, struct exfat_chain *p_chain) return -EIO; } - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); clu = p_chain->dir; if (p_chain->flags == ALLOC_NO_FAT_CHAIN) { @@ -261,7 +260,6 @@ int exfat_zeroed_cluster(struct inode *dir, unsigned int clu) memset(bhs[n]->b_data, 0, sb->s_blocksize); } - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); err = exfat_update_bhs(bhs, n, IS_DIRSYNC(dir)); if (err) goto release_bhs; @@ -326,8 +324,6 @@ int exfat_alloc_cluster(struct inode *inode, unsigned int num_alloc, } } - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); - p_chain->dir = EXFAT_EOF_CLUSTER; while ((new_clu = exfat_find_free_bitmap(sb, hint_clu)) != diff --git a/fs/exfat/misc.c b/fs/exfat/misc.c index dc34968e99d3..564718747fb2 100644 --- a/fs/exfat/misc.c +++ b/fs/exfat/misc.c @@ -163,9 +163,8 @@ u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type) return chksum; } -void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int sync) +void exfat_update_bh(struct buffer_head *bh, int sync) { - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(sb)->s_state); set_buffer_uptodate(bh); mark_buffer_dirty(bh); diff --git a/fs/exfat/namei.c b/fs/exfat/namei.c index 5b0f35329d63..e36c9fc4a5d6 100644 --- a/fs/exfat/namei.c +++ b/fs/exfat/namei.c @@ -387,7 +387,7 @@ static int exfat_find_empty_entry(struct inode *inode, ep->dentry.stream.valid_size = cpu_to_le64(size); ep->dentry.stream.size = ep->dentry.stream.valid_size; ep->dentry.stream.flags = p_dir->flags; - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); + exfat_update_bh(bh, IS_DIRSYNC(inode)); brelse(bh); if (exfat_update_dir_chksum(inode, &(ei->dir), ei->entry)) @@ -1071,7 +1071,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); ei->attr |= ATTR_ARCHIVE; } - exfat_update_bh(sb, new_bh, sync); + exfat_update_bh(new_bh, sync); brelse(old_bh); brelse(new_bh); @@ -1083,7 +1083,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, return -EIO; memcpy(epnew, epold, DENTRY_SIZE); - exfat_update_bh(sb, new_bh, sync); + exfat_update_bh(new_bh, sync); brelse(old_bh); brelse(new_bh); @@ -1100,7 +1100,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, epold->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); ei->attr |= ATTR_ARCHIVE; } - exfat_update_bh(sb, old_bh, sync); + exfat_update_bh(old_bh, sync); brelse(old_bh); ret = exfat_init_ext_entry(inode, p_dir, oldentry, num_new_entries, p_uniname); @@ -1155,7 +1155,7 @@ static int exfat_move_file(struct inode *inode, struct exfat_chain *p_olddir, epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); ei->attr |= ATTR_ARCHIVE; } - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); brelse(mov_bh); brelse(new_bh); @@ -1167,7 +1167,7 @@ static int exfat_move_file(struct inode *inode, struct exfat_chain *p_olddir, return -EIO; memcpy(epnew, epmov, DENTRY_SIZE); - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); brelse(mov_bh); brelse(new_bh); diff --git a/fs/exfat/super.c b/fs/exfat/super.c index e650e65536f8..199a1e78f9e5 100644 --- a/fs/exfat/super.c +++ b/fs/exfat/super.c @@ -104,6 +104,9 @@ int exfat_set_vol_flags(struct super_block *sb, unsigned short new_flag) struct boot_sector *p_boot = (struct boot_sector *)sbi->boot_bh->b_data; bool sync; + if (new_flag == VOL_DIRTY) + set_bit(EXFAT_SB_DIRTY, &sbi->s_state); + /* flags are not changed */ if (sbi->vol_flag == new_flag) return 0; -- 2.25.1 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* RE: [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing 2020-06-04 8:44 ` [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing Tetsuhiro Kohada @ 2020-06-05 7:32 ` Namjae Jeon 2020-06-06 9:22 ` Tetsuhiro Kohada 0 siblings, 1 reply; 10+ messages in thread From: Namjae Jeon @ 2020-06-05 7:32 UTC (permalink / raw) To: 'Tetsuhiro Kohada' Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, 'Sungjong Seo', linux-fsdevel, linux-kernel > Set EXFAT_SB_DIRTY flag in exfat_put_super(). > > In some cases, can't clear VOL_DIRTY with 'sync'. > ex: > > VOL_DIRTY is set when rmdir starts, but when non-empty-dir is detected, return error without setting > EXFAT_SB_DIRTY. > If performe 'sync' in this state, VOL_DIRTY will not be cleared. Good catch. Can you split this patch into two? (Don't set VOL_DIRTY on -ENOTEMPTY and Setting EXFAT_SB_DIRTY is merged into exfat_set_vol_flag). I need to check the second one more. Thanks! > > Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> > --- > fs/exfat/balloc.c | 4 ++-- > fs/exfat/dir.c | 18 ++++++++---------- > fs/exfat/exfat_fs.h | 2 +- > fs/exfat/fatent.c | 6 +----- > fs/exfat/misc.c | 3 +-- > fs/exfat/namei.c | 12 ++++++------ > fs/exfat/super.c | 3 +++ > 7 files changed, 22 insertions(+), 26 deletions(-) > > diff --git a/fs/exfat/balloc.c b/fs/exfat/balloc.c index 4055eb00ea9b..a987919686c0 100644 > --- a/fs/exfat/balloc.c > +++ b/fs/exfat/balloc.c > @@ -158,7 +158,7 @@ int exfat_set_bitmap(struct inode *inode, unsigned int clu) > b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); > > set_bit_le(b, sbi->vol_amap[i]->b_data); > - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); > + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); > return 0; > } > > @@ -180,7 +180,7 @@ void exfat_clear_bitmap(struct inode *inode, unsigned int clu) > b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); > > clear_bit_le(b, sbi->vol_amap[i]->b_data); > - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); > + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); > > if (opts->discard) { > int ret_discard; > diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c index 3eb8386fb5f2..96c9a817d928 100644 > --- a/fs/exfat/dir.c > +++ b/fs/exfat/dir.c > @@ -468,7 +468,7 @@ int exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, > &ep->dentry.file.access_date, > NULL); > > - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); > + exfat_update_bh(bh, IS_DIRSYNC(inode)); > brelse(bh); > > ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -478,7 +478,7 @@ int > exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, > exfat_init_stream_entry(ep, > (type == TYPE_FILE) ? ALLOC_FAT_CHAIN : ALLOC_NO_FAT_CHAIN, > start_clu, size); > - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); > + exfat_update_bh(bh, IS_DIRSYNC(inode)); > brelse(bh); > > return 0; > @@ -514,7 +514,7 @@ int exfat_update_dir_chksum(struct inode *inode, struct exfat_chain *p_dir, > } > > fep->dentry.file.checksum = cpu_to_le16(chksum); > - exfat_update_bh(sb, fbh, IS_DIRSYNC(inode)); > + exfat_update_bh(fbh, IS_DIRSYNC(inode)); > release_fbh: > brelse(fbh); > return ret; > @@ -536,7 +536,7 @@ int exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, > return -EIO; > > ep->dentry.file.num_ext = (unsigned char)(num_entries - 1); > - exfat_update_bh(sb, bh, sync); > + exfat_update_bh(bh, sync); > brelse(bh); > > ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -545,7 +545,7 @@ int > exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, > > ep->dentry.stream.name_len = p_uniname->name_len; > ep->dentry.stream.name_hash = cpu_to_le16(p_uniname->name_hash); > - exfat_update_bh(sb, bh, sync); > + exfat_update_bh(bh, sync); > brelse(bh); > > for (i = EXFAT_FIRST_CLUSTER; i < num_entries; i++) { @@ -554,7 +554,7 @@ int > exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, > return -EIO; > > exfat_init_name_entry(ep, uniname); > - exfat_update_bh(sb, bh, sync); > + exfat_update_bh(bh, sync); > brelse(bh); > uniname += EXFAT_FILE_NAME_LEN; > } > @@ -578,7 +578,7 @@ int exfat_remove_entries(struct inode *inode, struct exfat_chain *p_dir, > return -EIO; > > exfat_set_entry_type(ep, TYPE_DELETED); > - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); > + exfat_update_bh(bh, IS_DIRSYNC(inode)); > brelse(bh); > } > > @@ -606,10 +606,8 @@ int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) { > int i, err = 0; > > - if (es->modified) { > - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(es->sb)->s_state); > + if (es->modified) > err = exfat_update_bhs(es->bh, es->num_bh, sync); > - } > > for (i = 0; i < es->num_bh; i++) > err ? bforget(es->bh[i]):brelse(es->bh[i]); > diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h index f4fa0e833486..0e094d186612 100644 > --- a/fs/exfat/exfat_fs.h > +++ b/fs/exfat/exfat_fs.h > @@ -514,7 +514,7 @@ void exfat_set_entry_time(struct exfat_sb_info *sbi, struct timespec64 *ts, > u8 *tz, __le16 *time, __le16 *date, u8 *time_cs); > u16 exfat_calc_chksum16(void *data, int len, u16 chksum, int type); > u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type); -void exfat_update_bh(struct > super_block *sb, struct buffer_head *bh, int sync); > +void exfat_update_bh(struct buffer_head *bh, int sync); > int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync); void exfat_chain_set(struct > exfat_chain *ec, unsigned int dir, > unsigned int size, unsigned char flags); diff --git a/fs/exfat/fatent.c > b/fs/exfat/fatent.c index 5d11bc2f1b68..f8171183b4c1 100644 > --- a/fs/exfat/fatent.c > +++ b/fs/exfat/fatent.c > @@ -75,7 +75,7 @@ int exfat_ent_set(struct super_block *sb, unsigned int loc, > > fat_entry = (__le32 *)&(bh->b_data[off]); > *fat_entry = cpu_to_le32(content); > - exfat_update_bh(sb, bh, sb->s_flags & SB_SYNCHRONOUS); > + exfat_update_bh(bh, sb->s_flags & SB_SYNCHRONOUS); > exfat_mirror_bh(sb, sec, bh); > brelse(bh); > return 0; > @@ -174,7 +174,6 @@ int exfat_free_cluster(struct inode *inode, struct exfat_chain *p_chain) > return -EIO; > } > > - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); > clu = p_chain->dir; > > if (p_chain->flags == ALLOC_NO_FAT_CHAIN) { @@ -261,7 +260,6 @@ int exfat_zeroed_cluster(struct > inode *dir, unsigned int clu) > memset(bhs[n]->b_data, 0, sb->s_blocksize); > } > > - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); > err = exfat_update_bhs(bhs, n, IS_DIRSYNC(dir)); > if (err) > goto release_bhs; > @@ -326,8 +324,6 @@ int exfat_alloc_cluster(struct inode *inode, unsigned int num_alloc, > } > } > > - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); > - > p_chain->dir = EXFAT_EOF_CLUSTER; > > while ((new_clu = exfat_find_free_bitmap(sb, hint_clu)) != diff --git a/fs/exfat/misc.c > b/fs/exfat/misc.c index dc34968e99d3..564718747fb2 100644 > --- a/fs/exfat/misc.c > +++ b/fs/exfat/misc.c > @@ -163,9 +163,8 @@ u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type) > return chksum; > } > > -void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int sync) > +void exfat_update_bh(struct buffer_head *bh, int sync) > { > - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(sb)->s_state); > set_buffer_uptodate(bh); > mark_buffer_dirty(bh); > > diff --git a/fs/exfat/namei.c b/fs/exfat/namei.c index 5b0f35329d63..e36c9fc4a5d6 100644 > --- a/fs/exfat/namei.c > +++ b/fs/exfat/namei.c > @@ -387,7 +387,7 @@ static int exfat_find_empty_entry(struct inode *inode, > ep->dentry.stream.valid_size = cpu_to_le64(size); > ep->dentry.stream.size = ep->dentry.stream.valid_size; > ep->dentry.stream.flags = p_dir->flags; > - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); > + exfat_update_bh(bh, IS_DIRSYNC(inode)); > brelse(bh); > if (exfat_update_dir_chksum(inode, &(ei->dir), > ei->entry)) > @@ -1071,7 +1071,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, > epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); > ei->attr |= ATTR_ARCHIVE; > } > - exfat_update_bh(sb, new_bh, sync); > + exfat_update_bh(new_bh, sync); > brelse(old_bh); > brelse(new_bh); > > @@ -1083,7 +1083,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, > return -EIO; > > memcpy(epnew, epold, DENTRY_SIZE); > - exfat_update_bh(sb, new_bh, sync); > + exfat_update_bh(new_bh, sync); > brelse(old_bh); > brelse(new_bh); > > @@ -1100,7 +1100,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, > epold->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); > ei->attr |= ATTR_ARCHIVE; > } > - exfat_update_bh(sb, old_bh, sync); > + exfat_update_bh(old_bh, sync); > brelse(old_bh); > ret = exfat_init_ext_entry(inode, p_dir, oldentry, > num_new_entries, p_uniname); > @@ -1155,7 +1155,7 @@ static int exfat_move_file(struct inode *inode, struct exfat_chain *p_olddir, > epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); > ei->attr |= ATTR_ARCHIVE; > } > - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); > + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); > brelse(mov_bh); > brelse(new_bh); > > @@ -1167,7 +1167,7 @@ static int exfat_move_file(struct inode *inode, struct exfat_chain *p_olddir, > return -EIO; > > memcpy(epnew, epmov, DENTRY_SIZE); > - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); > + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); > brelse(mov_bh); > brelse(new_bh); > > diff --git a/fs/exfat/super.c b/fs/exfat/super.c index e650e65536f8..199a1e78f9e5 100644 > --- a/fs/exfat/super.c > +++ b/fs/exfat/super.c > @@ -104,6 +104,9 @@ int exfat_set_vol_flags(struct super_block *sb, unsigned short new_flag) > struct boot_sector *p_boot = (struct boot_sector *)sbi->boot_bh->b_data; > bool sync; > > + if (new_flag == VOL_DIRTY) > + set_bit(EXFAT_SB_DIRTY, &sbi->s_state); > + > /* flags are not changed */ > if (sbi->vol_flag == new_flag) > return 0; > -- > 2.25.1 ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing 2020-06-05 7:32 ` Namjae Jeon @ 2020-06-06 9:22 ` Tetsuhiro Kohada 2020-06-07 0:26 ` Namjae Jeon 0 siblings, 1 reply; 10+ messages in thread From: Tetsuhiro Kohada @ 2020-06-06 9:22 UTC (permalink / raw) To: Namjae Jeon Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, 'Sungjong Seo', linux-fsdevel, linux-kernel [-- Attachment #1: Type: text/plain, Size: 11461 bytes --] On 2020/06/05 16:32, Namjae Jeon wrote: >> Set EXFAT_SB_DIRTY flag in exfat_put_super(). >> >> In some cases, can't clear VOL_DIRTY with 'sync'. >> ex: >> >> VOL_DIRTY is set when rmdir starts, but when non-empty-dir is detected, return error without setting >> EXFAT_SB_DIRTY. >> If performe 'sync' in this state, VOL_DIRTY will not be cleared. > Good catch. > > Can you split this patch into two? (Don't set VOL_DIRTY on -ENOTEMPTY and Setting EXFAT_SB_DIRTY is > merged into exfat_set_vol_flag). I need to check the second one more. Can't do that. exfat_set_vol_flag() is called when rmdir processing begins. When Not-empty is detected, VOL_DIRTY has already been written and synced to the media. This sequence is same as other write functions. <begin writing function> set VOL_DIRTY (write to the media) preparation/state analysis update bh & set SB_DIRTY clear VOL_DIRTY (cached) <end> SB_DIRTY is set when updating bh. However, in some cases SB_DIRTY is not set. If SB_DIRTY is not set, exfat_sync_fs() does nothing. I thought there was a problem with separating VOL_DIRTY and SB_DIRTY. So I investigated the timing when these are set. Attach the caller graph. The green box is a function that sets SB_DIRTY directly. The red box is a function that calls exfat_set_vol_flag() and sets VOL_DIRTY. VOL_DIRTY is set on all paths before setting SB_DIRTY. I think VOL_DIRTY and SB_DIRTY should be set at the same time. That way, It is not necessary to set SB_DIRTY when updating bh. <begin writing function> set VOL_DIRTY (actually write on media), and set SB_DIRTY preparation/state analysis update data clear VOL_DIRTY (cached) <end> By doing this, sync is guaranteed if VOL_DIRTY is set by calling exfat_set_vol_flag. This change may still have problems, but it's little better than before, I think. BR --- Tetsuhiro Kohada <kohada.t2@gmail.com> > > Thanks! >> >> Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> >> --- >> fs/exfat/balloc.c | 4 ++-- >> fs/exfat/dir.c | 18 ++++++++---------- >> fs/exfat/exfat_fs.h | 2 +- >> fs/exfat/fatent.c | 6 +----- >> fs/exfat/misc.c | 3 +-- >> fs/exfat/namei.c | 12 ++++++------ >> fs/exfat/super.c | 3 +++ >> 7 files changed, 22 insertions(+), 26 deletions(-) >> >> diff --git a/fs/exfat/balloc.c b/fs/exfat/balloc.c index 4055eb00ea9b..a987919686c0 100644 >> --- a/fs/exfat/balloc.c >> +++ b/fs/exfat/balloc.c >> @@ -158,7 +158,7 @@ int exfat_set_bitmap(struct inode *inode, unsigned int clu) >> b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); >> >> set_bit_le(b, sbi->vol_amap[i]->b_data); >> - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); >> + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); >> return 0; >> } >> >> @@ -180,7 +180,7 @@ void exfat_clear_bitmap(struct inode *inode, unsigned int clu) >> b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); >> >> clear_bit_le(b, sbi->vol_amap[i]->b_data); >> - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); >> + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); >> >> if (opts->discard) { >> int ret_discard; >> diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c index 3eb8386fb5f2..96c9a817d928 100644 >> --- a/fs/exfat/dir.c >> +++ b/fs/exfat/dir.c >> @@ -468,7 +468,7 @@ int exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, >> &ep->dentry.file.access_date, >> NULL); >> >> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >> brelse(bh); >> >> ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -478,7 +478,7 @@ int >> exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, >> exfat_init_stream_entry(ep, >> (type == TYPE_FILE) ? ALLOC_FAT_CHAIN : ALLOC_NO_FAT_CHAIN, >> start_clu, size); >> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >> brelse(bh); >> >> return 0; >> @@ -514,7 +514,7 @@ int exfat_update_dir_chksum(struct inode *inode, struct exfat_chain *p_dir, >> } >> >> fep->dentry.file.checksum = cpu_to_le16(chksum); >> - exfat_update_bh(sb, fbh, IS_DIRSYNC(inode)); >> + exfat_update_bh(fbh, IS_DIRSYNC(inode)); >> release_fbh: >> brelse(fbh); >> return ret; >> @@ -536,7 +536,7 @@ int exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, >> return -EIO; >> >> ep->dentry.file.num_ext = (unsigned char)(num_entries - 1); >> - exfat_update_bh(sb, bh, sync); >> + exfat_update_bh(bh, sync); >> brelse(bh); >> >> ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -545,7 +545,7 @@ int >> exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, >> >> ep->dentry.stream.name_len = p_uniname->name_len; >> ep->dentry.stream.name_hash = cpu_to_le16(p_uniname->name_hash); >> - exfat_update_bh(sb, bh, sync); >> + exfat_update_bh(bh, sync); >> brelse(bh); >> >> for (i = EXFAT_FIRST_CLUSTER; i < num_entries; i++) { @@ -554,7 +554,7 @@ int >> exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, >> return -EIO; >> >> exfat_init_name_entry(ep, uniname); >> - exfat_update_bh(sb, bh, sync); >> + exfat_update_bh(bh, sync); >> brelse(bh); >> uniname += EXFAT_FILE_NAME_LEN; >> } >> @@ -578,7 +578,7 @@ int exfat_remove_entries(struct inode *inode, struct exfat_chain *p_dir, >> return -EIO; >> >> exfat_set_entry_type(ep, TYPE_DELETED); >> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >> brelse(bh); >> } >> >> @@ -606,10 +606,8 @@ int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) { >> int i, err = 0; >> >> - if (es->modified) { >> - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(es->sb)->s_state); >> + if (es->modified) >> err = exfat_update_bhs(es->bh, es->num_bh, sync); >> - } >> >> for (i = 0; i < es->num_bh; i++) >> err ? bforget(es->bh[i]):brelse(es->bh[i]); >> diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h index f4fa0e833486..0e094d186612 100644 >> --- a/fs/exfat/exfat_fs.h >> +++ b/fs/exfat/exfat_fs.h >> @@ -514,7 +514,7 @@ void exfat_set_entry_time(struct exfat_sb_info *sbi, struct timespec64 *ts, >> u8 *tz, __le16 *time, __le16 *date, u8 *time_cs); >> u16 exfat_calc_chksum16(void *data, int len, u16 chksum, int type); >> u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type); -void exfat_update_bh(struct >> super_block *sb, struct buffer_head *bh, int sync); >> +void exfat_update_bh(struct buffer_head *bh, int sync); >> int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync); void exfat_chain_set(struct >> exfat_chain *ec, unsigned int dir, >> unsigned int size, unsigned char flags); diff --git a/fs/exfat/fatent.c >> b/fs/exfat/fatent.c index 5d11bc2f1b68..f8171183b4c1 100644 >> --- a/fs/exfat/fatent.c >> +++ b/fs/exfat/fatent.c >> @@ -75,7 +75,7 @@ int exfat_ent_set(struct super_block *sb, unsigned int loc, >> >> fat_entry = (__le32 *)&(bh->b_data[off]); >> *fat_entry = cpu_to_le32(content); >> - exfat_update_bh(sb, bh, sb->s_flags & SB_SYNCHRONOUS); >> + exfat_update_bh(bh, sb->s_flags & SB_SYNCHRONOUS); >> exfat_mirror_bh(sb, sec, bh); >> brelse(bh); >> return 0; >> @@ -174,7 +174,6 @@ int exfat_free_cluster(struct inode *inode, struct exfat_chain *p_chain) >> return -EIO; >> } >> >> - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >> clu = p_chain->dir; >> >> if (p_chain->flags == ALLOC_NO_FAT_CHAIN) { @@ -261,7 +260,6 @@ int exfat_zeroed_cluster(struct >> inode *dir, unsigned int clu) >> memset(bhs[n]->b_data, 0, sb->s_blocksize); >> } >> >> - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >> err = exfat_update_bhs(bhs, n, IS_DIRSYNC(dir)); >> if (err) >> goto release_bhs; >> @@ -326,8 +324,6 @@ int exfat_alloc_cluster(struct inode *inode, unsigned int num_alloc, >> } >> } >> >> - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >> - >> p_chain->dir = EXFAT_EOF_CLUSTER; >> >> while ((new_clu = exfat_find_free_bitmap(sb, hint_clu)) != diff --git a/fs/exfat/misc.c >> b/fs/exfat/misc.c index dc34968e99d3..564718747fb2 100644 >> --- a/fs/exfat/misc.c >> +++ b/fs/exfat/misc.c >> @@ -163,9 +163,8 @@ u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type) >> return chksum; >> } >> >> -void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int sync) >> +void exfat_update_bh(struct buffer_head *bh, int sync) >> { >> - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(sb)->s_state); >> set_buffer_uptodate(bh); >> mark_buffer_dirty(bh); >> >> diff --git a/fs/exfat/namei.c b/fs/exfat/namei.c index 5b0f35329d63..e36c9fc4a5d6 100644 >> --- a/fs/exfat/namei.c >> +++ b/fs/exfat/namei.c >> @@ -387,7 +387,7 @@ static int exfat_find_empty_entry(struct inode *inode, >> ep->dentry.stream.valid_size = cpu_to_le64(size); >> ep->dentry.stream.size = ep->dentry.stream.valid_size; >> ep->dentry.stream.flags = p_dir->flags; >> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >> brelse(bh); >> if (exfat_update_dir_chksum(inode, &(ei->dir), >> ei->entry)) >> @@ -1071,7 +1071,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, >> epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); >> ei->attr |= ATTR_ARCHIVE; >> } >> - exfat_update_bh(sb, new_bh, sync); >> + exfat_update_bh(new_bh, sync); >> brelse(old_bh); >> brelse(new_bh); >> >> @@ -1083,7 +1083,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, >> return -EIO; >> >> memcpy(epnew, epold, DENTRY_SIZE); >> - exfat_update_bh(sb, new_bh, sync); >> + exfat_update_bh(new_bh, sync); >> brelse(old_bh); >> brelse(new_bh); >> >> @@ -1100,7 +1100,7 @@ static int exfat_rename_file(struct inode *inode, struct exfat_chain *p_dir, >> epold->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); >> ei->attr |= ATTR_ARCHIVE; >> } >> - exfat_update_bh(sb, old_bh, sync); >> + exfat_update_bh(old_bh, sync); >> brelse(old_bh); >> ret = exfat_init_ext_entry(inode, p_dir, oldentry, >> num_new_entries, p_uniname); >> @@ -1155,7 +1155,7 @@ static int exfat_move_file(struct inode *inode, struct exfat_chain *p_olddir, >> epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); >> ei->attr |= ATTR_ARCHIVE; >> } >> - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); >> + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); >> brelse(mov_bh); >> brelse(new_bh); >> >> @@ -1167,7 +1167,7 @@ static int exfat_move_file(struct inode *inode, struct exfat_chain *p_olddir, >> return -EIO; >> >> memcpy(epnew, epmov, DENTRY_SIZE); >> - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); >> + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); >> brelse(mov_bh); >> brelse(new_bh); >> >> diff --git a/fs/exfat/super.c b/fs/exfat/super.c index e650e65536f8..199a1e78f9e5 100644 >> --- a/fs/exfat/super.c >> +++ b/fs/exfat/super.c >> @@ -104,6 +104,9 @@ int exfat_set_vol_flags(struct super_block *sb, unsigned short new_flag) >> struct boot_sector *p_boot = (struct boot_sector *)sbi->boot_bh->b_data; >> bool sync; >> >> + if (new_flag == VOL_DIRTY) >> + set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >> + >> /* flags are not changed */ >> if (sbi->vol_flag == new_flag) >> return 0; >> -- >> 2.25.1 > > [-- Attachment #2: exfat__fs_8h_a9aac0f25e9499af92c7ac753f1bcb049_icgraph.png --] [-- Type: image/png, Size: 345222 bytes --] ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing 2020-06-06 9:22 ` Tetsuhiro Kohada @ 2020-06-07 0:26 ` Namjae Jeon 2020-06-08 5:02 ` Kohada.Tetsuhiro 0 siblings, 1 reply; 10+ messages in thread From: Namjae Jeon @ 2020-06-07 0:26 UTC (permalink / raw) To: Tetsuhiro Kohada Cc: Namjae Jeon, kohada.tetsuhiro, mori.takahiro, motai.hirotaka, Sungjong Seo, linux-fsdevel, linux-kernel 2020-06-06 18:22 GMT+09:00, Tetsuhiro Kohada <kohada.t2@gmail.com>: > On 2020/06/05 16:32, Namjae Jeon wrote: >>> Set EXFAT_SB_DIRTY flag in exfat_put_super(). >>> >>> In some cases, can't clear VOL_DIRTY with 'sync'. >>> ex: >>> >>> VOL_DIRTY is set when rmdir starts, but when non-empty-dir is detected, >>> return error without setting >>> EXFAT_SB_DIRTY. >>> If performe 'sync' in this state, VOL_DIRTY will not be cleared. >> Good catch. >> >> Can you split this patch into two? (Don't set VOL_DIRTY on -ENOTEMPTY and >> Setting EXFAT_SB_DIRTY is >> merged into exfat_set_vol_flag). I need to check the second one more. > > Can't do that. > > exfat_set_vol_flag() is called when rmdir processing begins. When Not-empty > is detected, > VOL_DIRTY has already been written and synced to the media. You can move it before calling exfat_remove_entries(). > > This sequence is same as other write functions. > <begin writing function> > set VOL_DIRTY (write to the media) > preparation/state analysis > update bh & set SB_DIRTY > clear VOL_DIRTY (cached) > <end> > > SB_DIRTY is set when updating bh. > However, in some cases SB_DIRTY is not set. > If SB_DIRTY is not set, exfat_sync_fs() does nothing. > > I thought there was a problem with separating VOL_DIRTY and SB_DIRTY. > So I investigated the timing when these are set. Attach the caller graph. > The green box is a function that sets SB_DIRTY directly. > The red box is a function that calls exfat_set_vol_flag() and sets > VOL_DIRTY. > VOL_DIRTY is set on all paths before setting SB_DIRTY. > > I think VOL_DIRTY and SB_DIRTY should be set at the same time. > That way, It is not necessary to set SB_DIRTY when updating bh. > > <begin writing function> > set VOL_DIRTY (actually write on media), and set SB_DIRTY > preparation/state analysis > update data > clear VOL_DIRTY (cached) > <end> > > By doing this, sync is guaranteed if VOL_DIRTY is set by calling > exfat_set_vol_flag. > > This change may still have problems, but it's little better than before, I > think. I need to check more if it is the best or there is more better way. Thanks! > > BR > --- > Tetsuhiro Kohada <kohada.t2@gmail.com> > > >> >> Thanks! >>> >>> Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> >>> --- >>> fs/exfat/balloc.c | 4 ++-- >>> fs/exfat/dir.c | 18 ++++++++---------- >>> fs/exfat/exfat_fs.h | 2 +- >>> fs/exfat/fatent.c | 6 +----- >>> fs/exfat/misc.c | 3 +-- >>> fs/exfat/namei.c | 12 ++++++------ >>> fs/exfat/super.c | 3 +++ >>> 7 files changed, 22 insertions(+), 26 deletions(-) >>> >>> diff --git a/fs/exfat/balloc.c b/fs/exfat/balloc.c index >>> 4055eb00ea9b..a987919686c0 100644 >>> --- a/fs/exfat/balloc.c >>> +++ b/fs/exfat/balloc.c >>> @@ -158,7 +158,7 @@ int exfat_set_bitmap(struct inode *inode, unsigned >>> int clu) >>> b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); >>> >>> set_bit_le(b, sbi->vol_amap[i]->b_data); >>> - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); >>> + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); >>> return 0; >>> } >>> >>> @@ -180,7 +180,7 @@ void exfat_clear_bitmap(struct inode *inode, unsigned >>> int clu) >>> b = BITMAP_OFFSET_BIT_IN_SECTOR(sb, ent_idx); >>> >>> clear_bit_le(b, sbi->vol_amap[i]->b_data); >>> - exfat_update_bh(sb, sbi->vol_amap[i], IS_DIRSYNC(inode)); >>> + exfat_update_bh(sbi->vol_amap[i], IS_DIRSYNC(inode)); >>> >>> if (opts->discard) { >>> int ret_discard; >>> diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c index >>> 3eb8386fb5f2..96c9a817d928 100644 >>> --- a/fs/exfat/dir.c >>> +++ b/fs/exfat/dir.c >>> @@ -468,7 +468,7 @@ int exfat_init_dir_entry(struct inode *inode, struct >>> exfat_chain *p_dir, >>> &ep->dentry.file.access_date, >>> NULL); >>> >>> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >>> brelse(bh); >>> >>> ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -478,7 >>> +478,7 @@ int >>> exfat_init_dir_entry(struct inode *inode, struct exfat_chain *p_dir, >>> exfat_init_stream_entry(ep, >>> (type == TYPE_FILE) ? ALLOC_FAT_CHAIN : ALLOC_NO_FAT_CHAIN, >>> start_clu, size); >>> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >>> brelse(bh); >>> >>> return 0; >>> @@ -514,7 +514,7 @@ int exfat_update_dir_chksum(struct inode *inode, >>> struct exfat_chain *p_dir, >>> } >>> >>> fep->dentry.file.checksum = cpu_to_le16(chksum); >>> - exfat_update_bh(sb, fbh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(fbh, IS_DIRSYNC(inode)); >>> release_fbh: >>> brelse(fbh); >>> return ret; >>> @@ -536,7 +536,7 @@ int exfat_init_ext_entry(struct inode *inode, struct >>> exfat_chain *p_dir, >>> return -EIO; >>> >>> ep->dentry.file.num_ext = (unsigned char)(num_entries - 1); >>> - exfat_update_bh(sb, bh, sync); >>> + exfat_update_bh(bh, sync); >>> brelse(bh); >>> >>> ep = exfat_get_dentry(sb, p_dir, entry + 1, &bh, §or); @@ -545,7 >>> +545,7 @@ int >>> exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, >>> >>> ep->dentry.stream.name_len = p_uniname->name_len; >>> ep->dentry.stream.name_hash = cpu_to_le16(p_uniname->name_hash); >>> - exfat_update_bh(sb, bh, sync); >>> + exfat_update_bh(bh, sync); >>> brelse(bh); >>> >>> for (i = EXFAT_FIRST_CLUSTER; i < num_entries; i++) { @@ -554,7 +554,7 >>> @@ int >>> exfat_init_ext_entry(struct inode *inode, struct exfat_chain *p_dir, >>> return -EIO; >>> >>> exfat_init_name_entry(ep, uniname); >>> - exfat_update_bh(sb, bh, sync); >>> + exfat_update_bh(bh, sync); >>> brelse(bh); >>> uniname += EXFAT_FILE_NAME_LEN; >>> } >>> @@ -578,7 +578,7 @@ int exfat_remove_entries(struct inode *inode, struct >>> exfat_chain *p_dir, >>> return -EIO; >>> >>> exfat_set_entry_type(ep, TYPE_DELETED); >>> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >>> brelse(bh); >>> } >>> >>> @@ -606,10 +606,8 @@ int exfat_free_dentry_set(struct >>> exfat_entry_set_cache *es, int sync) { >>> int i, err = 0; >>> >>> - if (es->modified) { >>> - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(es->sb)->s_state); >>> + if (es->modified) >>> err = exfat_update_bhs(es->bh, es->num_bh, sync); >>> - } >>> >>> for (i = 0; i < es->num_bh; i++) >>> err ? bforget(es->bh[i]):brelse(es->bh[i]); >>> diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h index >>> f4fa0e833486..0e094d186612 100644 >>> --- a/fs/exfat/exfat_fs.h >>> +++ b/fs/exfat/exfat_fs.h >>> @@ -514,7 +514,7 @@ void exfat_set_entry_time(struct exfat_sb_info *sbi, >>> struct timespec64 *ts, >>> u8 *tz, __le16 *time, __le16 *date, u8 *time_cs); >>> u16 exfat_calc_chksum16(void *data, int len, u16 chksum, int type); >>> u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type); >>> -void exfat_update_bh(struct >>> super_block *sb, struct buffer_head *bh, int sync); >>> +void exfat_update_bh(struct buffer_head *bh, int sync); >>> int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync); >>> void exfat_chain_set(struct >>> exfat_chain *ec, unsigned int dir, >>> unsigned int size, unsigned char flags); diff --git >>> a/fs/exfat/fatent.c >>> b/fs/exfat/fatent.c index 5d11bc2f1b68..f8171183b4c1 100644 >>> --- a/fs/exfat/fatent.c >>> +++ b/fs/exfat/fatent.c >>> @@ -75,7 +75,7 @@ int exfat_ent_set(struct super_block *sb, unsigned int >>> loc, >>> >>> fat_entry = (__le32 *)&(bh->b_data[off]); >>> *fat_entry = cpu_to_le32(content); >>> - exfat_update_bh(sb, bh, sb->s_flags & SB_SYNCHRONOUS); >>> + exfat_update_bh(bh, sb->s_flags & SB_SYNCHRONOUS); >>> exfat_mirror_bh(sb, sec, bh); >>> brelse(bh); >>> return 0; >>> @@ -174,7 +174,6 @@ int exfat_free_cluster(struct inode *inode, struct >>> exfat_chain *p_chain) >>> return -EIO; >>> } >>> >>> - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >>> clu = p_chain->dir; >>> >>> if (p_chain->flags == ALLOC_NO_FAT_CHAIN) { @@ -261,7 +260,6 @@ int >>> exfat_zeroed_cluster(struct >>> inode *dir, unsigned int clu) >>> memset(bhs[n]->b_data, 0, sb->s_blocksize); >>> } >>> >>> - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >>> err = exfat_update_bhs(bhs, n, IS_DIRSYNC(dir)); >>> if (err) >>> goto release_bhs; >>> @@ -326,8 +324,6 @@ int exfat_alloc_cluster(struct inode *inode, unsigned >>> int num_alloc, >>> } >>> } >>> >>> - set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >>> - >>> p_chain->dir = EXFAT_EOF_CLUSTER; >>> >>> while ((new_clu = exfat_find_free_bitmap(sb, hint_clu)) != diff --git >>> a/fs/exfat/misc.c >>> b/fs/exfat/misc.c index dc34968e99d3..564718747fb2 100644 >>> --- a/fs/exfat/misc.c >>> +++ b/fs/exfat/misc.c >>> @@ -163,9 +163,8 @@ u32 exfat_calc_chksum32(void *data, int len, u32 >>> chksum, int type) >>> return chksum; >>> } >>> >>> -void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int >>> sync) >>> +void exfat_update_bh(struct buffer_head *bh, int sync) >>> { >>> - set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(sb)->s_state); >>> set_buffer_uptodate(bh); >>> mark_buffer_dirty(bh); >>> >>> diff --git a/fs/exfat/namei.c b/fs/exfat/namei.c index >>> 5b0f35329d63..e36c9fc4a5d6 100644 >>> --- a/fs/exfat/namei.c >>> +++ b/fs/exfat/namei.c >>> @@ -387,7 +387,7 @@ static int exfat_find_empty_entry(struct inode >>> *inode, >>> ep->dentry.stream.valid_size = cpu_to_le64(size); >>> ep->dentry.stream.size = ep->dentry.stream.valid_size; >>> ep->dentry.stream.flags = p_dir->flags; >>> - exfat_update_bh(sb, bh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(bh, IS_DIRSYNC(inode)); >>> brelse(bh); >>> if (exfat_update_dir_chksum(inode, &(ei->dir), >>> ei->entry)) >>> @@ -1071,7 +1071,7 @@ static int exfat_rename_file(struct inode *inode, >>> struct exfat_chain *p_dir, >>> epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); >>> ei->attr |= ATTR_ARCHIVE; >>> } >>> - exfat_update_bh(sb, new_bh, sync); >>> + exfat_update_bh(new_bh, sync); >>> brelse(old_bh); >>> brelse(new_bh); >>> >>> @@ -1083,7 +1083,7 @@ static int exfat_rename_file(struct inode *inode, >>> struct exfat_chain *p_dir, >>> return -EIO; >>> >>> memcpy(epnew, epold, DENTRY_SIZE); >>> - exfat_update_bh(sb, new_bh, sync); >>> + exfat_update_bh(new_bh, sync); >>> brelse(old_bh); >>> brelse(new_bh); >>> >>> @@ -1100,7 +1100,7 @@ static int exfat_rename_file(struct inode *inode, >>> struct exfat_chain *p_dir, >>> epold->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); >>> ei->attr |= ATTR_ARCHIVE; >>> } >>> - exfat_update_bh(sb, old_bh, sync); >>> + exfat_update_bh(old_bh, sync); >>> brelse(old_bh); >>> ret = exfat_init_ext_entry(inode, p_dir, oldentry, >>> num_new_entries, p_uniname); >>> @@ -1155,7 +1155,7 @@ static int exfat_move_file(struct inode *inode, >>> struct exfat_chain *p_olddir, >>> epnew->dentry.file.attr |= cpu_to_le16(ATTR_ARCHIVE); >>> ei->attr |= ATTR_ARCHIVE; >>> } >>> - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); >>> brelse(mov_bh); >>> brelse(new_bh); >>> >>> @@ -1167,7 +1167,7 @@ static int exfat_move_file(struct inode *inode, >>> struct exfat_chain *p_olddir, >>> return -EIO; >>> >>> memcpy(epnew, epmov, DENTRY_SIZE); >>> - exfat_update_bh(sb, new_bh, IS_DIRSYNC(inode)); >>> + exfat_update_bh(new_bh, IS_DIRSYNC(inode)); >>> brelse(mov_bh); >>> brelse(new_bh); >>> >>> diff --git a/fs/exfat/super.c b/fs/exfat/super.c index >>> e650e65536f8..199a1e78f9e5 100644 >>> --- a/fs/exfat/super.c >>> +++ b/fs/exfat/super.c >>> @@ -104,6 +104,9 @@ int exfat_set_vol_flags(struct super_block *sb, >>> unsigned short new_flag) >>> struct boot_sector *p_boot = (struct boot_sector >>> *)sbi->boot_bh->b_data; >>> bool sync; >>> >>> + if (new_flag == VOL_DIRTY) >>> + set_bit(EXFAT_SB_DIRTY, &sbi->s_state); >>> + >>> /* flags are not changed */ >>> if (sbi->vol_flag == new_flag) >>> return 0; >>> -- >>> 2.25.1 >> >> > ^ permalink raw reply [flat|nested] 10+ messages in thread
* RE: [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing 2020-06-07 0:26 ` Namjae Jeon @ 2020-06-08 5:02 ` Kohada.Tetsuhiro 2020-06-09 2:02 ` Namjae Jeon 0 siblings, 1 reply; 10+ messages in thread From: Kohada.Tetsuhiro @ 2020-06-08 5:02 UTC (permalink / raw) To: 'Namjae Jeon' Cc: Namjae Jeon, Mori.Takahiro, Motai.Hirotaka, Sungjong Seo, linux-fsdevel, linux-kernel Thank you for your comment. > >> Can you split this patch into two? (Don't set VOL_DIRTY on -ENOTEMPTY and > >> Setting EXFAT_SB_DIRTY is > >> merged into exfat_set_vol_flag). I need to check the second one more. > > > > Can't do that. > > > > exfat_set_vol_flag() is called when rmdir processing begins. When Not-empty > > is detected, > > VOL_DIRTY has already been written and synced to the media. > You can move it before calling exfat_remove_entries(). Can be moved, but that doesn't solve the problem. It causes the similar problem as before. exfat_remove_entries() calls exfat_get_dentry(). If exfat_get_dentry() fails, update bh and set SB_DIRTY will not be executed. As a result, SB_DIRTY is not set and sync does not work. Similar problems occur with other writing functions. Similar problems occur when pre-write checks are added in the future. If you don't set VOL_DIRTY at the beginning, you should delay to set VOL_DIRTY until update-bh & set SB_DIRTY. This avoids unnecessary changes to VOL_DIRTY/VOL_CLEAN. I think this method is smart, but it is difficult to decide when to set VOL_CLEAN. (I tried to implement it, but gave up) > > By doing this, sync is guaranteed if VOL_DIRTY is set by calling > > exfat_set_vol_flag. > > > > This change may still have problems, but it's little better than before, I > > think. > I need to check more if it is the best or there is more better way. I think the sync-problems still exist. Let's improve little by little. :-) BR --- Kohada Tetsuhiro <Kohada.Tetsuhiro@dc.MitsubishiElectric.co.jp> ^ permalink raw reply [flat|nested] 10+ messages in thread
* RE: [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing 2020-06-08 5:02 ` Kohada.Tetsuhiro @ 2020-06-09 2:02 ` Namjae Jeon 0 siblings, 0 replies; 10+ messages in thread From: Namjae Jeon @ 2020-06-09 2:02 UTC (permalink / raw) To: Kohada.Tetsuhiro Cc: Mori.Takahiro, Motai.Hirotaka, 'Sungjong Seo', linux-fsdevel, linux-kernel, 'Namjae Jeon' > Thank you for your comment. > > > >> Can you split this patch into two? (Don't set VOL_DIRTY on > > >> -ENOTEMPTY and Setting EXFAT_SB_DIRTY is merged into > > >> exfat_set_vol_flag). I need to check the second one more. > > > > > > Can't do that. > > > > > > exfat_set_vol_flag() is called when rmdir processing begins. When > > > Not-empty is detected, VOL_DIRTY has already been written and synced > > > to the media. > > You can move it before calling exfat_remove_entries(). > > Can be moved, but that doesn't solve the problem. > It causes the similar problem as before. > > exfat_remove_entries() calls exfat_get_dentry(). > If exfat_get_dentry() fails, update bh and set SB_DIRTY will not be executed. > As a result, SB_DIRTY is not set and sync does not work. > Similar problems occur with other writing functions. > Similar problems occur when pre-write checks are added in the future. > > If you don't set VOL_DIRTY at the beginning, you should delay to set VOL_DIRTY until update-bh & set > SB_DIRTY. > This avoids unnecessary changes to VOL_DIRTY/VOL_CLEAN. Right, That's what I am going to point out. > I think this method is smart, but it is difficult to decide when to set VOL_CLEAN. > (I tried to implement it, but gave up) Okay, I'm a little busy now, but I'll give you feedback soon. Thanks for your work! > > > > By doing this, sync is guaranteed if VOL_DIRTY is set by calling > > > exfat_set_vol_flag. > > > > > > This change may still have problems, but it's little better than > > > before, I think. > > I need to check more if it is the best or there is more better way. > > I think the sync-problems still exist. > Let's improve little by little. :-) > > BR > --- > Kohada Tetsuhiro <Kohada.Tetsuhiro@dc.MitsubishiElectric.co.jp> ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH 1/3] exfat: add error check when updating dir-entries 2020-06-04 8:44 [PATCH 1/3] exfat: add error check when updating dir-entries Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 2/3] exfat: optimize exfat_zeroed_cluster() Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing Tetsuhiro Kohada @ 2020-06-07 0:33 ` Namjae Jeon 2020-06-09 8:10 ` Tetsuhiro Kohada 2 siblings, 1 reply; 10+ messages in thread From: Namjae Jeon @ 2020-06-07 0:33 UTC (permalink / raw) To: Tetsuhiro Kohada Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, Sungjong Seo, Namjae Jeon, linux-fsdevel, linux-kernel 2020-06-04 17:44 GMT+09:00, Tetsuhiro Kohada <kohada.t2@gmail.com>: Hi Tetsuhiro, > Add error check when synchronously updating dir-entries. > Furthermore, add exfat_update_bhs(). It wait for write completion once > instead of sector by sector. This patch can be split into two also ? > > Suggested-by: Sungjong Seo <sj1557.seo@samsung.com> > Signed-off-by: Tetsuhiro Kohada <kohada.t2@gmail.com> > --- > fs/exfat/dir.c | 15 +++++++++------ > fs/exfat/exfat_fs.h | 3 ++- > fs/exfat/file.c | 5 ++++- > fs/exfat/inode.c | 8 +++++--- > fs/exfat/misc.c | 19 +++++++++++++++++++ > 5 files changed, 39 insertions(+), 11 deletions(-) > > diff --git a/fs/exfat/dir.c b/fs/exfat/dir.c > index de43534aa299..3eb8386fb5f2 100644 > --- a/fs/exfat/dir.c > +++ b/fs/exfat/dir.c > @@ -602,16 +602,19 @@ void exfat_update_dir_chksum_with_entry_set(struct > exfat_entry_set_cache *es) > es->modified = true; > } > > -void exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) > +int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync) > { > - int i; > + int i, err = 0; > > - for (i = 0; i < es->num_bh; i++) { > - if (es->modified) > - exfat_update_bh(es->sb, es->bh[i], sync); > - brelse(es->bh[i]); > + if (es->modified) { > + set_bit(EXFAT_SB_DIRTY, &EXFAT_SB(es->sb)->s_state); EXFAT_SB_DIRTY set can be merged into exfat_update_bhs() ? > + err = exfat_update_bhs(es->bh, es->num_bh, sync); > } > + > + for (i = 0; i < es->num_bh; i++) > + err ? bforget(es->bh[i]):brelse(es->bh[i]); > kfree(es); > + return err; > } > > static int exfat_walk_fat_chain(struct super_block *sb, > diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h > index 595f3117f492..f4fa0e833486 100644 > --- a/fs/exfat/exfat_fs.h > +++ b/fs/exfat/exfat_fs.h > @@ -462,7 +462,7 @@ struct exfat_dentry *exfat_get_dentry_cached(struct > exfat_entry_set_cache *es, > int num); > struct exfat_entry_set_cache *exfat_get_dentry_set(struct super_block *sb, > struct exfat_chain *p_dir, int entry, unsigned int type); > -void exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync); > +int exfat_free_dentry_set(struct exfat_entry_set_cache *es, int sync); > int exfat_count_dir_entries(struct super_block *sb, struct exfat_chain > *p_dir); > > /* inode.c */ > @@ -515,6 +515,7 @@ void exfat_set_entry_time(struct exfat_sb_info *sbi, > struct timespec64 *ts, > u16 exfat_calc_chksum16(void *data, int len, u16 chksum, int type); > u32 exfat_calc_chksum32(void *data, int len, u32 chksum, int type); > void exfat_update_bh(struct super_block *sb, struct buffer_head *bh, int > sync); > +int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync); > void exfat_chain_set(struct exfat_chain *ec, unsigned int dir, > unsigned int size, unsigned char flags); > void exfat_chain_dup(struct exfat_chain *dup, struct exfat_chain *ec); > diff --git a/fs/exfat/file.c b/fs/exfat/file.c > index fce03f318787..37c8f04c1f8a 100644 > --- a/fs/exfat/file.c > +++ b/fs/exfat/file.c > @@ -153,6 +153,7 @@ int __exfat_truncate(struct inode *inode, loff_t > new_size) > struct timespec64 ts; > struct exfat_dentry *ep, *ep2; > struct exfat_entry_set_cache *es; > + int err; > > es = exfat_get_dentry_set(sb, &(ei->dir), ei->entry, > ES_ALL_ENTRIES); > @@ -187,7 +188,9 @@ int __exfat_truncate(struct inode *inode, loff_t > new_size) > } > > exfat_update_dir_chksum_with_entry_set(es); > - exfat_free_dentry_set(es, inode_needs_sync(inode)); > + err = exfat_free_dentry_set(es, inode_needs_sync(inode)); > + if (err) > + return err; > } > > /* cut off from the FAT chain */ > diff --git a/fs/exfat/inode.c b/fs/exfat/inode.c > index ef7cf7a6d187..c0bfd1a586aa 100644 > --- a/fs/exfat/inode.c > +++ b/fs/exfat/inode.c > @@ -77,8 +77,7 @@ static int __exfat_write_inode(struct inode *inode, int > sync) > ep2->dentry.stream.size = ep2->dentry.stream.valid_size; > > exfat_update_dir_chksum_with_entry_set(es); > - exfat_free_dentry_set(es, sync); > - return 0; > + return exfat_free_dentry_set(es, sync); > } > > int exfat_write_inode(struct inode *inode, struct writeback_control *wbc) > @@ -222,6 +221,7 @@ static int exfat_map_cluster(struct inode *inode, > unsigned int clu_offset, > if (ei->dir.dir != DIR_DELETED && modified) { > struct exfat_dentry *ep; > struct exfat_entry_set_cache *es; > + int err; > > es = exfat_get_dentry_set(sb, &(ei->dir), ei->entry, > ES_ALL_ENTRIES); > @@ -240,7 +240,9 @@ static int exfat_map_cluster(struct inode *inode, > unsigned int clu_offset, > ep->dentry.stream.valid_size; > > exfat_update_dir_chksum_with_entry_set(es); > - exfat_free_dentry_set(es, inode_needs_sync(inode)); > + err = exfat_free_dentry_set(es, inode_needs_sync(inode)); > + if (err) > + return err; > > } /* end of if != DIR_DELETED */ > > diff --git a/fs/exfat/misc.c b/fs/exfat/misc.c > index 17d41f3d3709..dc34968e99d3 100644 > --- a/fs/exfat/misc.c > +++ b/fs/exfat/misc.c > @@ -173,6 +173,25 @@ void exfat_update_bh(struct super_block *sb, struct > buffer_head *bh, int sync) > sync_dirty_buffer(bh); > } > > +int exfat_update_bhs(struct buffer_head **bhs, int nr_bhs, int sync) > +{ > + int i, err = 0; > + > + for (i = 0; i < nr_bhs; i++) { > + set_buffer_uptodate(bhs[i]); > + mark_buffer_dirty(bhs[i]); > + if (sync) > + write_dirty_buffer(bhs[i], 0); > + } > + > + for (i = 0; i < nr_bhs && sync; i++) { > + wait_on_buffer(bhs[i]); > + if (!buffer_uptodate(bhs[i])) > + err = -EIO; > + } > + return err; > +} > + > void exfat_chain_set(struct exfat_chain *ec, unsigned int dir, > unsigned int size, unsigned char flags) > { > -- > 2.25.1 > > ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH 1/3] exfat: add error check when updating dir-entries 2020-06-07 0:33 ` [PATCH 1/3] exfat: add error check when updating dir-entries Namjae Jeon @ 2020-06-09 8:10 ` Tetsuhiro Kohada 0 siblings, 0 replies; 10+ messages in thread From: Tetsuhiro Kohada @ 2020-06-09 8:10 UTC (permalink / raw) To: Namjae Jeon Cc: kohada.tetsuhiro, mori.takahiro, motai.hirotaka, Sungjong Seo, Namjae Jeon, linux-fsdevel, linux-kernel >> Add error check when synchronously updating dir-entries. >> Furthermore, add exfat_update_bhs(). It wait for write completion once >> instead of sector by sector. > This patch can be split into two also ? I sent a patch split into 'write multiple sectors at once' and 'add error check when updating dir-entries'. The other two patches(2nd & 3rd) are no-changed, so have not been sent. If you need the other two patches, I will send them. In that case, please tell me how to write the subject and change-log. BR --- Tetsuhiro Kohada <kohada.t2@gmail.com> ^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2020-06-09 8:10 UTC | newest] Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2020-06-04 8:44 [PATCH 1/3] exfat: add error check when updating dir-entries Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 2/3] exfat: optimize exfat_zeroed_cluster() Tetsuhiro Kohada 2020-06-04 8:44 ` [PATCH 3/3] exfat: set EXFAT_SB_DIRTY and VOL_DIRTY at the same timing Tetsuhiro Kohada 2020-06-05 7:32 ` Namjae Jeon 2020-06-06 9:22 ` Tetsuhiro Kohada 2020-06-07 0:26 ` Namjae Jeon 2020-06-08 5:02 ` Kohada.Tetsuhiro 2020-06-09 2:02 ` Namjae Jeon 2020-06-07 0:33 ` [PATCH 1/3] exfat: add error check when updating dir-entries Namjae Jeon 2020-06-09 8:10 ` Tetsuhiro Kohada
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).