On 02/12, guoweichao wrote: > Hi Jaegeuk, > > On 2018/2/12 7:32, Jaegeuk Kim wrote: > > On 02/06, Weichao Guo wrote: > >> There is a potential inconsistent metadata case due to a cp block > >> crc invalid in the latest checkpoint caused by hardware issues: > >> 1) write nodes into segment x; > >> 2) write checkpoint A; > >> 3) remove nodes in segment x; > >> 4) write checkpoint B; > >> 5) issue discard or write datas into segment x; > >> 6) sudden power-cut; > >> 7) use checkpoint A after reboot as checkpoint B is invalid > >> > >> This inconsistency may be found after several reboots long time later > >> and the kernel log about cp block crc invalid has disappeared. This > >> makes the root cause of the inconsistency is hard to locate. Let us > >> separate such other part issues from f2fs logical bugs in debug version. > >> > >> Signed-off-by: Weichao Guo <guoweic...@huawei.com> > >> --- > >> fs/f2fs/checkpoint.c | 8 ++++++-- > >> 1 file changed, 6 insertions(+), 2 deletions(-) > >> > >> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > >> index 8b0945b..16ba96a 100644 > >> --- a/fs/f2fs/checkpoint.c > >> +++ b/fs/f2fs/checkpoint.c > >> @@ -737,13 +737,17 @@ static int get_checkpoint_version(struct > >> f2fs_sb_info *sbi, block_t cp_addr, > >> crc_offset = le32_to_cpu((*cp_block)->checksum_offset); > >> if (crc_offset > (blk_size - sizeof(__le32))) { > >> f2fs_msg(sbi->sb, KERN_WARNING, > >> - "invalid crc_offset: %zu", crc_offset); > >> + "invalid crc_offset: %zu at blk_addr: 0x%x", > >> + crc_offset, cp_addr); > >> + f2fs_bug_on(sbi, 1); > > > > I don't think we can use bug_on here, since we're easily getting this when > > power-cut happened in the middle of checkpoint pack writes, which is an > > expected > > behavior. Hmm, we need to consider another way to detect that. > We only check CP block crc here. The two CP blocks may have different CP > versions when > power-cut happened, but their crc value should be valid. IMO, this patch will > trigger a > bug_on only when some external issues cause CP block crc invalid as one 4K > page is > persisted atomically.
Huh? This checks crc_offset, not crc? Unfortunately, my simple fault injection test gave this bug_on within a day. The below bug_on seems what you're saying about tho. > > Thanks, > > > > Thanks, > > > >> return -EINVAL; > >> } > >> > >> crc = cur_cp_crc(*cp_block); > >> if (!f2fs_crc_valid(sbi, crc, *cp_block, crc_offset)) { > >> - f2fs_msg(sbi->sb, KERN_WARNING, "invalid crc value"); > >> + f2fs_msg(sbi->sb, KERN_WARNING, > >> + "invalid crc value at blk_addr: 0x%x", cp_addr); > >> + f2fs_bug_on(sbi, 1); > >> return -EINVAL; > >> } > >> > >> -- > >> 2.10.1 > > > > . > > ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel