On 2019/11/28 12:07, Shinichiro Kawasaki wrote:
> On Nov 25, 2019 / 14:59, Chao Yu wrote:
>> On 2019/11/14 16:19, Shin'ichiro Kawasaki wrote:
>>> 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, make mount fail and ask users to run
>>> fsck.
>>>
>>> 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 <[email protected]>
>>> ---
>>> fs/f2fs/f2fs.h | 1 +
>>> fs/f2fs/recovery.c | 17 ++++++-
>>> fs/f2fs/segment.c | 120 +++++++++++++++++++++++++++++++++++++++++++++
>>> 3 files changed, 136 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
>>> index 4024790028aa..a2e24718c13b 100644
>>> --- a/fs/f2fs/f2fs.h
>>> +++ b/fs/f2fs/f2fs.h
>>> @@ -3136,6 +3136,7 @@ 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 783773e4560d..712054ed8d64 100644
>>> --- a/fs/f2fs/recovery.c
>>> +++ b/fs/f2fs/recovery.c
>>> @@ -784,9 +784,22 @@ 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 (!ret && !err && !f2fs_readonly(sbi->sb)
>>
>> Using !check_only will be more readable?
>>
>> if (!err && !check_only && !f2fs_readonly(sbi->sb)
>
> When check_only is on and there is no fsync data, I think we should fix the
> write pointer inconsistency. With the condition you suggested, this case can
> not be covered.
Alright.
>
> Having said that, my expression with !ret is not good from readability point
> of view. How about this?
>
>
> bool fix_curseg_write_pointer;
> fix_curseg_write_pointer = !check_only || (check_only &&
> list_empty(&inode_list));
fix_curseg_write_pointer = !check_only || list_empty(&inode_list); is enough.
>
> ...
>
> if (!err && fix_curseg_write_pointer && !f2fs_readonly(sbi->sb)
> && f2fs_sb_has_blkzoned(sbi)) {
> err = f2fs_fix_curseg_write_pointer(sbi);
> ret = err;
> }
It's okay to me.
Thanks,
>
>
>>> + && 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 808709581481..6ece146dab34 100644
>>> --- a/fs/f2fs/segment.c
>>> +++ b/fs/f2fs/segment.c
>>> @@ -4331,6 +4331,126 @@ 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 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, nr_zones, wp_sector_off;
>>> + block_t cs_zone_block, wp_block, cs_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));
>>> + cs_block = START_BLOCK(sbi, cs->segno) + cs->next_blkoff;
>>> +
>>> + 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;
>>> + nr_zones = 1;
>>> + err = blkdev_report_zones(zbd->bdev, zone_sector, &zone, &nr_zones);
>>> + if (err) {
>>> + 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)
>>
>> We uses indent instead of space in f2fs coding style, please keep line
>> with it.
>
> Noted this f2fs conding style. Will replace the spaces with tab indent.
> Thanks!
>
> --
> Best Regards,
> Shin'ichiro Kawasaki.
>
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
