On 8/2/17 12:29 PM, Darrick J. Wong wrote: > On Tue, Aug 01, 2017 at 10:45:06PM -0500, Eric Sandeen wrote: >> xfs_metadump by default sanitizes the image so that all strings >> longer than 5 chars are obfusccated, and all stale data in metadata >> blocks (i.e. unused/unwritten data) is zeroed out. We didn't have >> a test for this, though, so this does it. >> >> It patterns 256M of the scratch device, then uses djwong's >> populate infrastructure to write all types of metadata, >> metadumps & mdrestores it, then looks for either the leaked >> pre-pattern or any leaked strings or filenames. >> >> The strings we look for are, unfortunately, a bit ad-hoc based on >> what is currently used in the populate routines. >> >> Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx> >> --- >> >> V2: include .out file, minor whitespace & typo edits >> V3: remove populate debug stuff that snuck in >> +echo "Silence is golden" >> + >> +# Pattern the scratch disk, mkfs, and restore. >> +$XFS_IO_PROG -d -c "pwrite 0 256M" $SCRATCH_DEV > $seqres.full 2>&1 > > pwrite -b 1m to speed this up a bit... ok >> +_scratch_mkfs_sized $((256 * 1024 * 1024)) >> $seqres.full 2>&1 >> +_scratch_populate nofill >> $seqres.full 2>&1 >> + >> +# populate unmounts the fs for us >> + >> +_scratch_metadump $METADUMP_FILE >> +xfs_mdrestore $METADUMP_FILE $MDRESTORE_FILE >> + >> +# Grep for stale data (leaked cd cd pattern) or strings >> +# from populate routine >> + >> +hexdump -C $MDRESTORE_FILE | grep \ > > _require_command /usr/bin/hexdump "hexdump" ? sure? other tests don't bother, but ok. > >> +"cd cd cd cd\| \ >> +41 41 41 41\|\ >> +42 42 42 42\|\ >> +43 43 43 43\|\ >> +44 44 44 44\|\ >> +61 61 61 61\|\ >> +62 62 62 62\|\ >> +63 63 63 63\|\ >> +64 64 64 64\|\ >> +dummy\|\ >> +S_IF\|\ >> +FMT_\|\ >> +INLINE\|\ >> +BLOCK\|\ >> +LEAF\|\ >> +NODE\|\ >> +BTREE\|\ >> +LOCAL\|\ >> +EXTENTS\|\ >> +REMOTE\|\ >> +ATTR\|\ >> +SYSTEM\|\ >> +TRUSTED\|\ >> +SECURITY\|\ >> +attrvalfile\|\ >> +unused\|\ >> +BNOBT\|\ >> +RMAPBT\|\ >> +RTRMAPBT\|\ >> +REFCOUNTBT" && echo "Leaked data found; see comments in test to debug" > > /me wonders if this grep pattern ought to be some kind of _filter > function in common/populate? That way if the xfs populate command adds > more weird strings, the search pattern is right there in the rc file, > not buried in tests/xfs/ somewhere. > > OTOH a good counterargument is that this is the only test that cares > about that, so why make it common code? <sigh> um, I totally agree with both your arguments... :/ i'll put it in populate I guess. > > --D -- 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
