On 2017/8/16 0:42, Jaegeuk Kim wrote: > On 08/15, Chao Yu wrote: >> On 2017/8/15 11:22, Jaegeuk Kim wrote: >>> On 08/15, Chao Yu wrote: >>>> Hi Jaegeuk, >>>> >>>> On 2017/8/11 8:42, Jaegeuk Kim wrote: >>>>> If we set CP_ERROR_FLAG in roll-forward error, f2fs is no longer to proceed >>>>> any IOs due to f2fs_cp_error(). But, for example, if some stale data is involved >>>>> on roll-forward process, we're able to get -ENOENT, getting fs stuck. >>>>> If we get any error, let fill_super set SBI_NEED_FSCK and try to recover back >>>>> to stable point. >>>> >>>> Before that, we have cleaned up all node/meta page cache, so we will get back to >>>> last checkpoint status, means losing fsynced datas for ever. >>>> >>>> Would it be better to just leave message reminding user to mount with >>>> disable_roll_forward or run fsck offline. >>> >>> We can't rely on user for this, since fsck cannot recover this, resulting in >> >> If fsck has no ability to recover this, it could tag superblock in somewhere, >> then kernel could skip recovery. Comparing to fail recovery directly, it give >> user another chance to rescuer his datas. > > Huh, what do you mean? This patch let f2fs_fill_super set SBI_NEED_FSCK in > superblock and skip roll-forward in second round. Oh, just notice that additional checkpoint which sets SBI_NEED_FSCK won't destroy warn node chain, so we have chance to run fsck and recover after another mount. Sorry. Reviewed-by: Chao Yu <yuchao0@xxxxxxxxxx> > >> >> Thanks, >> >>> infinite mount failure. The only way is to disable roll-forward recovery, which >>> is same as returning error here. >>> >>> Thanks, >>> >>>> >>>> Thanks, >>>> >>>>> >>>>> Cc: <stable@xxxxxxxxxxxxxxx> >>>>> Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx> >>>>> --- >>>>> fs/f2fs/recovery.c | 2 -- >>>>> 1 file changed, 2 deletions(-) >>>>> >>>>> diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c >>>>> index a3d02613934a..f707d810c87d 100644 >>>>> --- a/fs/f2fs/recovery.c >>>>> +++ b/fs/f2fs/recovery.c >>>>> @@ -649,8 +649,6 @@ int recover_fsync_data(struct f2fs_sb_info *sbi, bool check_only) >>>>> } >>>>> >>>>> clear_sbi_flag(sbi, SBI_POR_DOING); >>>>> - if (err) >>>>> - set_ckpt_flags(sbi, CP_ERROR_FLAG); >>>>> mutex_unlock(&sbi->cp_mutex); >>>>> >>>>> /* let's drop all the directory inodes for clean checkpoint */ >>>>> >>> >>> . >>> > > . >