On 07/14/2017 02:58 PM, Bill O'Donnell wrote: > On Fri, Jul 14, 2017 at 02:11:01PM -0500, Bill O'Donnell wrote: >> On Fri, Jul 14, 2017 at 02:01:31PM -0500, Eric Sandeen wrote: >>> >>> >>> On 07/14/2017 11:47 AM, Bill O'Donnell wrote: >>>> On Fri, Jul 14, 2017 at 10:42:57AM -0500, Eric Sandeen wrote: >>> >>> >>>>>> +# for different sector sizes, ensure no CRC errors are falsely reported >>>>>> + >>>>>> +# Supported types include: agf, agfl, agi, attr3, bmapbta, >>>>>> +# bmapbtd, bnobt, cntbt, data, dir3, dqblk, inobt, inodata, >>>>>> +# inode, log, rtbitmap, rtsummary, sb, symlink, text, finobt. >>>>> >>>>> This leaves me wondering why we don't test most of the above ;) >>>>> >>>>>> +# For various sector sizes, test some types that involve type size. >>>>>> +for SECTOR_SIZE in 512 1024 2048 4096; do >>>>>> + $MKFS_XFS_PROG -f -s size=$SECTOR_SIZE $SCRATCH_DEV > /dev/null >>>>>> + for TYPE in agf agi agfl sb; do >>>>>> + DADDR=`$XFS_DB_PROG -c "$TYPE" -c "daddr" $SCRATCH_DEV | _filter_dbval` >>>>>> + $XFS_DB_PROG -c "$TYPE" -c "daddr 42" -c "daddr $DADDR" -c "type $TYPE" $SCRATCH_DEV >>>>> >>>>> Hm is there a reason for the first "-c $TYPE" or even the "daddr 42" in the line just above? >>>> It's drawn from the reproducer test case. >>>> The first "-c $TYPE" sets the object, "daddr 42" is an arbitrary address set. Without >>>> the daddr change and change-back, the test will pass even without the recent xfs_db >>>> xfsprogs change. >>> >>> Really? Not here, xfsprogs-4.5.0 : >>> >>> # xfs_db -c "daddr 2" -c "type agi" fsfile >>> Metadata CRC error detected at xfs_agi block 0x2/0x200 >> >> Ahh, ok, you're correct - I'll modify the test accordingly. > > Wait. I still think the object has to get set with "-c $TYPE". If not > the crc error occurs even with the new xfs_db code: > # xfs_db -V > xfs_db version 4.12.0-rc2 > # xfs_db /dev/sda6 > xfs_db> daddr > current daddr is 0 > xfs_db> daddr 2 > xfs_db> type agi > Metadata CRC error detected at xfs_agi block 0x2/0x1000 > Metadata CRC error detected at xfs_agi block 0x2/0x1000 You've set daddr to 2... presumably on a 4k sector fs ... > > And error is gone when the object is set and the daddr gets > set to the correct value. > [root@localhost xfsprogs-dev3]# xfs_db /dev/sda6 > xfs_db> agi > xfs_db> daddr > current daddr is 16 ... but the agi is at daddr 16, not 2. IOWS, you've told it the wrong type for daddr 2 (which is halfway through the first superblock; daddr 2 is not the beginning of any on-disk structure) so if you ask it to tell you about that block as an AGI, it'll tell you that it's corrupted, because it's /not/ an agi. > xfs_db> daddr 2 > xfs_db> daddr 16 If you correctly tell it to show you block 16 as an AGI... > xfs_db> type agi ... it works. -Eric -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
