[f2fs-dev] [PATCH v6 0/2] f2fs: Check write pointers of zoned block devices

[f2fs-dev] [PATCH v6 0/2] f2fs: Check write pointers of zoned block devices

[Shin'ichiro Kawasaki]

On sudden f2fs shutdown, zoned block device status and f2fs meta data can be
inconsistent. When f2fs shutdown happens during write operations, write pointers
on the device go forward but the f2fs meta data does not reflect the write
pointer progress. This inconsistency will eventually cause "Unaligned write
command" error when restarting write operation after the next mount.

This error is observed with xfstests test case generic/388, which enforces
sudden shutdown during write operation and checks the file system recovery.

This patch series adds a feature to f2fs to check and fix the write pointer
consistency when zoned block devices get mounted. Per discussion on the list,
implement two checks: a check for open zones and a check for non-open zones.
The first patch add the check for open zones that current segments point to.
Check write pointer consistency with current segment positions recorded in CP,
and if they are inconsistent, assign a new zone to the current segment. When
fsync data exists, check and fix the consistency after fsync data recovery.

The second patch adds the check for non-open zones that current segments do not
point to. Compare write pointers with valid block maps in SIT, and if they are
inconsistent, report the failure or fix by resetting the write pointer.

Thank goes to Chao Yu for the detailed discussion on the list.

Changes from v5:
* Removed unused variable cs_block that YueHaibing pointed out

Changes from v4:
* Replaced blkdev_zone_mgmt() function calls with __f2fs_issue_discard_zone()

Changes from v3:
* Rebased to linux master tip and fixed conflicts for blkdev_report_zones()
* Discard zones with write pointers not at zone start but without valid blocks,
  resetting the write pointers instead of fsck run by users
* Improved readability of open-zone's check condition and indent with tab

Changes from v2:
* 1st patch: Fix write pointer in SBI_POR_DOING guard
             Removed __set_inuse() and f2fs_stop_checkpoint()
* 2nd patch: Reversed bitmap search and removed __set_inuse()
             Changed condition for disable_roll_forward/norecovery mount options

Changes from v1:
* 2nd patch: Added check to ensure zones are in main segments

Shin'ichiro Kawasaki (2):
  f2fs: Check write pointer consistency of open zones
  f2fs: Check write pointer consistency of non-open zones

 fs/f2fs/f2fs.h     |   2 +
 fs/f2fs/recovery.c |  20 +++-
 fs/f2fs/segment.c  | 257 +++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/super.c    |  11 ++
 4 files changed, 288 insertions(+), 2 deletions(-)

-- 
2.24.1



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

[f2fs-dev] [PATCH v6 1/2] f2fs: Check write pointer consistency of open zones

[Shin'ichiro Kawasaki]

On sudden f2fs shutdown, write pointers of zoned block devices can go
further but f2fs meta data keeps current segments at positions before the
write operations. After remounting the f2fs, this inconsistency causes
write operations not at write pointers and "Unaligned write command"
error is reported.

To avoid the error, compare current segments with write pointers of open
zones the current segments point to, during mount operation. If the write
pointer position is not aligned with the current segment position, assign
a new zone to the current segment. Also check the newly assigned zone has
write pointer at zone start. If not, reset write pointer of the zone.

Perform the consistency check during fsync recovery. Not to lose the
fsync data, do the check after fsync data gets restored and before
checkpoint commit which flushes data at current segment positions. Not to
cause conflict with kworker's dirfy data/node flush, do the fix within
SBI_POR_DOING protection.

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Cc: YueHaibing <yuehaibing@huawei.com>
---
 fs/f2fs/f2fs.h     |   1 +
 fs/f2fs/recovery.c |  20 ++++++-
 fs/f2fs/segment.c  | 131 +++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 150 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 5a888a063c7f..002c417b0a53 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -3155,6 +3155,7 @@ void f2fs_write_node_summaries(struct f2fs_sb_info *sbi, block_t start_blk);
 int f2fs_lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
 			unsigned int val, int alloc);
 void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc);
+int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi);
 int f2fs_build_segment_manager(struct f2fs_sb_info *sbi);
 void f2fs_destroy_segment_manager(struct f2fs_sb_info *sbi);
 int __init f2fs_create_segment_manager_caches(void);
diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
index 76477f71d4ee..763d5c0951d1 100644
--- a/fs/f2fs/recovery.c
+++ b/fs/f2fs/recovery.c
@@ -723,6 +723,7 @@ int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only)
 	int ret = 0;
 	unsigned long s_flags = sbi->sb->s_flags;
 	bool need_writecp = false;
+	bool fix_curseg_write_pointer = false;
 #ifdef CONFIG_QUOTA
 	int quota_enabled;
 #endif
@@ -774,6 +775,8 @@ int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only)
 		sbi->sb->s_flags = s_flags;
 	}
 skip:
+	fix_curseg_write_pointer = !check_only || list_empty(&inode_list);
+
 	destroy_fsync_dnodes(&inode_list, err);
 	destroy_fsync_dnodes(&tmp_inode_list, err);
 
@@ -784,9 +787,22 @@ int f2fs_recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only)
 	if (err) {
 		truncate_inode_pages_final(NODE_MAPPING(sbi));
 		truncate_inode_pages_final(META_MAPPING(sbi));
-	} else {
-		clear_sbi_flag(sbi, SBI_POR_DOING);
 	}
+
+	/*
+	 * If fsync data succeeds or there is no fsync data to recover,
+	 * and the f2fs is not read only, check and fix zoned block devices'
+	 * write pointer consistency.
+	 */
+	if (!err && fix_curseg_write_pointer && !f2fs_readonly(sbi->sb) &&
+			f2fs_sb_has_blkzoned(sbi)) {
+		err = f2fs_fix_curseg_write_pointer(sbi);
+		ret = err;
+	}
+
+	if (!err)
+		clear_sbi_flag(sbi, SBI_POR_DOING);
+
 	mutex_unlock(&sbi->cp_mutex);
 
 	/* let's drop all the directory inodes for clean checkpoint */
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 56e81447e2f3..5c3018743a9f 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -4368,6 +4368,137 @@ static int sanity_check_curseg(struct f2fs_sb_info *sbi)
 	return 0;
 }
 
+#ifdef CONFIG_BLK_DEV_ZONED
+
+static struct f2fs_dev_info *get_target_zoned_dev(struct f2fs_sb_info *sbi,
+						  block_t zone_blkaddr)
+{
+	int i;
+
+	for (i = 0; i < sbi->s_ndevs; i++) {
+		if (!bdev_is_zoned(FDEV(i).bdev))
+			continue;
+		if (sbi->s_ndevs == 1 || (FDEV(i).start_blk <= zone_blkaddr &&
+				zone_blkaddr <= FDEV(i).end_blk))
+			return &FDEV(i);
+	}
+
+	return NULL;
+}
+
+static int report_one_zone_cb(struct blk_zone *zone, unsigned int idx,
+			      void *data) {
+	memcpy(data, zone, sizeof(struct blk_zone));
+	return 0;
+}
+
+static int fix_curseg_write_pointer(struct f2fs_sb_info *sbi, int type)
+{
+	struct curseg_info *cs = CURSEG_I(sbi, type);
+	struct f2fs_dev_info *zbd;
+	struct blk_zone zone;
+	unsigned int cs_section, wp_segno, wp_blkoff, wp_sector_off;
+	block_t cs_zone_block, wp_block;
+	unsigned int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
+	sector_t zone_sector;
+	int err;
+
+	cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
+	cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
+
+	zbd = get_target_zoned_dev(sbi, cs_zone_block);
+	if (!zbd)
+		return 0;
+
+	/* report zone for the sector the curseg points to */
+	zone_sector = (sector_t)(cs_zone_block - zbd->start_blk)
+		<< log_sectors_per_block;
+	err = blkdev_report_zones(zbd->bdev, zone_sector, 1,
+				  report_one_zone_cb, &zone);
+	if (err != 1) {
+		f2fs_err(sbi, "Report zone failed: %s errno=(%d)",
+			 zbd->path, err);
+		return err;
+	}
+
+	if (zone.type != BLK_ZONE_TYPE_SEQWRITE_REQ)
+		return 0;
+
+	wp_block = zbd->start_blk + (zone.wp >> log_sectors_per_block);
+	wp_segno = GET_SEGNO(sbi, wp_block);
+	wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
+	wp_sector_off = zone.wp & GENMASK(log_sectors_per_block - 1, 0);
+
+	if (cs->segno == wp_segno && cs->next_blkoff == wp_blkoff &&
+		wp_sector_off == 0)
+		return 0;
+
+	f2fs_notice(sbi, "Unaligned curseg[%d] with write pointer: "
+		    "curseg[0x%x,0x%x] wp[0x%x,0x%x]",
+		    type, cs->segno, cs->next_blkoff, wp_segno, wp_blkoff);
+
+	f2fs_notice(sbi, "Assign new section to curseg[%d]: "
+		    "curseg[0x%x,0x%x]", type, cs->segno, cs->next_blkoff);
+	allocate_segment_by_default(sbi, type, true);
+
+	/* check newly assigned zone */
+	cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
+	cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
+
+	zbd = get_target_zoned_dev(sbi, cs_zone_block);
+	if (!zbd)
+		return 0;
+
+	zone_sector = (sector_t)(cs_zone_block - zbd->start_blk)
+		<< log_sectors_per_block;
+	err = blkdev_report_zones(zbd->bdev, zone_sector, 1,
+				  report_one_zone_cb, &zone);
+	if (err != 1) {
+		f2fs_err(sbi, "Report zone failed: %s errno=(%d)",
+			 zbd->path, err);
+		return err;
+	}
+
+	if (zone.type != BLK_ZONE_TYPE_SEQWRITE_REQ)
+		return 0;
+
+	if (zone.wp != zone.start) {
+		f2fs_notice(sbi,
+			    "New zone for curseg[%d] is not yet discarded. "
+			    "Reset the zone: curseg[0x%x,0x%x]",
+			    type, cs->segno, cs->next_blkoff);
+		err = __f2fs_issue_discard_zone(sbi, zbd->bdev,
+				zone_sector >> log_sectors_per_block,
+				zone.len >> log_sectors_per_block);
+		if (err) {
+			f2fs_err(sbi, "Discard zone failed: %s (errno=%d)",
+				 zbd->path, err);
+			return err;
+		}
+	}
+
+	return 0;
+}
+
+int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
+{
+	int i, ret;
+
+	for (i = 0; i < NO_CHECK_TYPE; i++) {
+		ret = fix_curseg_write_pointer(sbi, i);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+#else
+int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
+{
+	return 0;
+}
+#endif
+
 /*
  * Update min, max modified time for cost-benefit GC algorithm
  */
-- 
2.24.1



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

[f2fs-dev] [PATCH v6 2/2] f2fs: Check write pointer consistency of non-open zones

[Shin'ichiro Kawasaki]

To catch f2fs bugs in write pointer handling code for zoned block
devices, check write pointers of non-open zones that current segments do
not point to. Do this check at mount time, after the fsync data recovery
and current segments' write pointer consistency fix. Or when fsync data
recovery is disabled by mount option, do the check when there is no fsync
data.

Check two items comparing write pointers with valid block maps in SIT.
The first item is check for zones with no valid blocks. When there is no
valid blocks in a zone, the write pointer should be at the start of the
zone. If not, next write operation to the zone will cause unaligned write
error. If write pointer is not at the zone start, reset the write pointer
to place at the zone start.

The second item is check between the write pointer position and the last
valid block in the zone. It is unexpected that the last valid block
position is beyond the write pointer. In such a case, report as a bug.
Fix is not required for such zone, because the zone is not selected for
next write operation until the zone get discarded.

Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
---
 fs/f2fs/f2fs.h    |   1 +
 fs/f2fs/segment.c | 126 ++++++++++++++++++++++++++++++++++++++++++++++
 fs/f2fs/super.c   |  11 ++++
 3 files changed, 138 insertions(+)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 002c417b0a53..23a84d7f17b8 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -3156,6 +3156,7 @@ int f2fs_lookup_journal_in_cursum(struct f2fs_journal *journal, int type,
 			unsigned int val, int alloc);
 void f2fs_flush_sit_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc);
 int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi);
+int f2fs_check_write_pointer(struct f2fs_sb_info *sbi);
 int f2fs_build_segment_manager(struct f2fs_sb_info *sbi);
 void f2fs_destroy_segment_manager(struct f2fs_sb_info *sbi);
 int __init f2fs_create_segment_manager_caches(void);
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 5c3018743a9f..3ec912bd5504 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -4370,6 +4370,90 @@ static int sanity_check_curseg(struct f2fs_sb_info *sbi)
 
 #ifdef CONFIG_BLK_DEV_ZONED
 
+static int check_zone_write_pointer(struct f2fs_sb_info *sbi,
+				    struct f2fs_dev_info *fdev,
+				    struct blk_zone *zone)
+{
+	unsigned int wp_segno, wp_blkoff, zone_secno, zone_segno, segno;
+	block_t zone_block, wp_block, last_valid_block;
+	unsigned int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
+	int i, s, b, ret;
+	struct seg_entry *se;
+
+	if (zone->type != BLK_ZONE_TYPE_SEQWRITE_REQ)
+		return 0;
+
+	wp_block = fdev->start_blk + (zone->wp >> log_sectors_per_block);
+	wp_segno = GET_SEGNO(sbi, wp_block);
+	wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
+	zone_block = fdev->start_blk + (zone->start >> log_sectors_per_block);
+	zone_segno = GET_SEGNO(sbi, zone_block);
+	zone_secno = GET_SEC_FROM_SEG(sbi, zone_segno);
+
+	if (zone_segno >= MAIN_SEGS(sbi))
+		return 0;
+
+	/*
+	 * Skip check of zones cursegs point to, since
+	 * fix_curseg_write_pointer() checks them.
+	 */
+	for (i = 0; i < NO_CHECK_TYPE; i++)
+		if (zone_secno == GET_SEC_FROM_SEG(sbi,
+						   CURSEG_I(sbi, i)->segno))
+			return 0;
+
+	/*
+	 * Get last valid block of the zone.
+	 */
+	last_valid_block = zone_block - 1;
+	for (s = sbi->segs_per_sec - 1; s >= 0; s--) {
+		segno = zone_segno + s;
+		se = get_seg_entry(sbi, segno);
+		for (b = sbi->blocks_per_seg - 1; b >= 0; b--)
+			if (f2fs_test_bit(b, se->cur_valid_map)) {
+				last_valid_block = START_BLOCK(sbi, segno) + b;
+				break;
+			}
+		if (last_valid_block >= zone_block)
+			break;
+	}
+
+	/*
+	 * If last valid block is beyond the write pointer, report the
+	 * inconsistency. This inconsistency does not cause write error
+	 * because the zone will not be selected for write operation until
+	 * it get discarded. Just report it.
+	 */
+	if (last_valid_block >= wp_block) {
+		f2fs_notice(sbi, "Valid block beyond write pointer: "
+			    "valid block[0x%x,0x%x] wp[0x%x,0x%x]",
+			    GET_SEGNO(sbi, last_valid_block),
+			    GET_BLKOFF_FROM_SEG0(sbi, last_valid_block),
+			    wp_segno, wp_blkoff);
+		return 0;
+	}
+
+	/*
+	 * If there is no valid block in the zone and if write pointer is
+	 * not at zone start, reset the write pointer.
+	 */
+	if (last_valid_block + 1 == zone_block && zone->wp != zone->start) {
+		f2fs_notice(sbi,
+			    "Zone without valid block has non-zero write "
+			    "pointer. Reset the write pointer: wp[0x%x,0x%x]",
+			    wp_segno, wp_blkoff);
+		ret = __f2fs_issue_discard_zone(sbi, fdev->bdev, zone_block,
+					zone->len >> log_sectors_per_block);
+		if (ret) {
+			f2fs_err(sbi, "Discard zone failed: %s (errno=%d)",
+				 fdev->path, ret);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
 static struct f2fs_dev_info *get_target_zoned_dev(struct f2fs_sb_info *sbi,
 						  block_t zone_blkaddr)
 {
@@ -4441,6 +4525,10 @@ static int fix_curseg_write_pointer(struct f2fs_sb_info *sbi, int type)
 		    "curseg[0x%x,0x%x]", type, cs->segno, cs->next_blkoff);
 	allocate_segment_by_default(sbi, type, true);
 
+	/* check consistency of the zone curseg pointed to */
+	if (check_zone_write_pointer(sbi, zbd, &zone))
+		return -EIO;
+
 	/* check newly assigned zone */
 	cs_section = GET_SEC_FROM_SEG(sbi, cs->segno);
 	cs_zone_block = START_BLOCK(sbi, GET_SEG_FROM_SEC(sbi, cs_section));
@@ -4492,11 +4580,49 @@ int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
 
 	return 0;
 }
+
+struct check_zone_write_pointer_args {
+	struct f2fs_sb_info *sbi;
+	struct f2fs_dev_info *fdev;
+};
+
+static int check_zone_write_pointer_cb(struct blk_zone *zone, unsigned int idx,
+				      void *data) {
+	struct check_zone_write_pointer_args *args;
+	args = (struct check_zone_write_pointer_args *)data;
+
+	return check_zone_write_pointer(args->sbi, args->fdev, zone);
+}
+
+int f2fs_check_write_pointer(struct f2fs_sb_info *sbi)
+{
+	int i, ret;
+	struct check_zone_write_pointer_args args;
+
+	for (i = 0; i < sbi->s_ndevs; i++) {
+		if (!bdev_is_zoned(FDEV(i).bdev))
+			continue;
+
+		args.sbi = sbi;
+		args.fdev = &FDEV(i);
+		ret = blkdev_report_zones(FDEV(i).bdev, 0, BLK_ALL_ZONES,
+					  check_zone_write_pointer_cb, &args);
+		if (ret < 0)
+			return ret;
+	}
+
+	return 0;
+}
 #else
 int f2fs_fix_curseg_write_pointer(struct f2fs_sb_info *sbi)
 {
 	return 0;
 }
+
+int f2fs_check_write_pointer(struct f2fs_sb_info *sbi)
+{
+	return 0;
+}
 #endif
 
 /*
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 5111e1ffe58a..755ad57c795b 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -3544,6 +3544,17 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 			goto free_meta;
 		}
 	}
+
+	/*
+	 * If the f2fs is not readonly and fsync data recovery succeeds,
+	 * check zoned block devices' write pointer consistency.
+	 */
+	if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
+		err = f2fs_check_write_pointer(sbi);
+		if (err)
+			goto free_meta;
+	}
+
 reset_checkpoint:
 	/* f2fs_recover_fsync_data() cleared this already */
 	clear_sbi_flag(sbi, SBI_POR_DOING);
-- 
2.24.1



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel