Nathan Scott wrote:
Hi,
Earlier changes removed most of the bogus verbosity from fsstress
runs with debug kernels, when bulkstat is asked to report on non-
inode data in its stat'ing. This resolves the one remaining case
which is the bulkstat_one vs a xfs_dilocate warning, by passing a
flag down into dilocate indicating we may be looking up garbage,
and not to spam the console in that situation.
cheers.
1. xfs_ialloc.c
+ xfs_stack_trace();
What's the stack trace for?
It is just to give more info after one of the 3 error msgs
beforehand? Happens in DEBUG. Okay.
2. xfs_inode.h:
+#define XFS_IGET_CREATE 0x1
+#define XFS_IGET_BULKSTAT 0x2
Okay.
3. xfs_itable.c
xfs_bulkstat_one_iget calls xfs_iget with XFS_IGET_BULKSTAT flag.
Okay.
4. xfs_inode.c
xfs_iread takes in extra param, imap_flags
and passes it on to call of xfs_itobp
Okay
xfs_ialloc's xfs_trans_iget()'s IGET_CREATE -> XFS_IGET_CREATE
Okay
5. xfs_iget.c
IGET_CREATE -> XFS_IGET_CREATE
ok
some cleanup
xfs_iget_core calls xfs_iread with extra imap_flags of XFS_IMAP_BULKSTAT
if it is given its flag of XFS_IGET_BULKSTAT
ok
6. xfs_dm.c
xfs_dm_bulkall_iget_one calls xfs_iget with XFS_IGET_BULKSTAT
xfs_dm_bulkattr_iget_one calls xfs_iget with XFS_IGET_BULKSTAT
ok
Seems fine.
--Tim
|