[XFS] Long btree pointers are still 64 bit on disk
On 32 bit machines with CONFIG_LBD=n, XFS reduces the
in memory size of xfs_fsblock_t to 32 bits so that it
will fit within 32 bit addressing. However, the disk format
for long btree pointers are still 64 bits in size.
The recent btree rewrite failed to take this into account
when initialising new btree blocks, setting sibling pointers
to NULL and checking if they are NULL. Hence checking whether
a 64 bit NULL was the same as a 32 bit NULL was failingi
resulting in NULL sibling pointers failing to be detected
correctly. This showed up as WANT_CORRUPTED_GOTO shutdowns
in xfs_btree_delrec.
Fix this by making all the comparisons and setting of long
pointer btree NULL blocks to the disk format, not the
in memory format. i.e. use NULLDFSBNO.
Reported-by: Alexander Beregalov <a.beregalov@gmail.com>
Reported-by: Jacek Luczak <difrost.kernel@gmail.com>
Reported-by: Danny ter Haar <dth@dth.net>
Tested-by: Jacek Luczak <difrost.kernel@gmail.com>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Felix Blyakher <felixb@sgi.com>
Remove the last of the macros-defined-to-static-functions.
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Recently we have quite a few kerneloops reports about dereferencing a NULL
if_data in the attribute fork. From looking over the code this can only
happen if we pass a 0 size argument to xfs_iformat_local. This implies some
sort of corruption and in fact the only mailinglist report about this from
earlier this year was after a powerfail presumably on a system with write
cache and without barriers.
Add a quick sanity check for the attr fork size in xfs_iformat to catch
these early and without an oops.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Currently the bad_features2 fixup and the alignment updates in the superblock
are skipped if we mount a filesystem read-only. But for the root filesystem
the typical case is to mount read-only first and only later remount writeable
so we'll never perform this update at all. It's not a big problem but means
the logs of people needing the fixup get spammed at every boot because they
never happen on disk.
Reported-by: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
We can have both a user and a group/project dquot locked at the same time,
as long as the user dquot is locked first. Tell lockdep about that fact
by making the group/project dquots a different lock class.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
xfs_dqlock2 locks two xfs_dquots, which is fine as it always locks the
dquot with the lower id first. Use mutex_lock_nested to tell lockdep
about this fact.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
We can have both a a quota hash chain and the per-mount list locked at
the same time. But given that both use the same struct dqhash as list
head we have to tell lockdep that they are different lock classes.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
The compat version of the attrmulti ioctl needs to ask for and then
later release write access to the mount just like the native version,
otherwise we could potentially write to read-only mounts.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Open by handle just grabs an inode by handle and then creates itself
a dentry for it. While this works for regular files it is horribly
broken for directories, where the VFS locking relies on the fact that
there is only just one single dentry for a given inode, and that
these are always connected to the root of the filesystem so that
it's locking algorithms work (see Documentations/filesystems/Locking)
Remove all the existing open by handle code and replace it with a small
wrapper around the exportfs code which deals with all these issues.
At the same time we also make the checks for a valid handle strict
enough to reject all not perfectly well formed handles - given that
we never hand out others that's okay and simplifies the code.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
It removes XFS specific ioctl interfaces and request codes
for freeze feature.
This patch has been supplied by David Chinner.
Signed-off-by: Dave Chinner <dgc@sgi.com>
Signed-off-by: Takashi Sato <t-sato@yk.jp.nec.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: <xfs-masters@oss.sgi.com>
Cc: <linux-ext4@vger.kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Kleikamp <shaggy@austin.ibm.com>
Cc: Alasdair G Kergon <agk@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently, ext3 in mainline Linux doesn't have the freeze feature which
suspends write requests. So, we cannot take a backup which keeps the
filesystem's consistency with the storage device's features (snapshot and
replication) while it is mounted.
In many case, a commercial filesystem (e.g. VxFS) has the freeze feature
and it would be used to get the consistent backup.
If Linux's standard filesystem ext3 has the freeze feature, we can do it
without a commercial filesystem.
So I have implemented the ioctls of the freeze feature.
I think we can take the consistent backup with the following steps.
1. Freeze the filesystem with the freeze ioctl.
2. Separate the replication volume or create the snapshot
with the storage device's feature.
3. Unfreeze the filesystem with the unfreeze ioctl.
4. Take the backup from the separated replication volume
or the snapshot.
This patch:
VFS:
Changed the type of write_super_lockfs and unlockfs from "void"
to "int" so that they can return an error.
Rename write_super_lockfs and unlockfs of the super block operation
freeze_fs and unfreeze_fs to avoid a confusion.
ext3, ext4, xfs, gfs2, jfs:
Changed the type of write_super_lockfs and unlockfs from "void"
to "int" so that write_super_lockfs returns an error if needed,
and unlockfs always returns 0.
reiserfs:
Changed the type of write_super_lockfs and unlockfs from "void"
to "int" so that they always return 0 (success) to keep a current behavior.
Signed-off-by: Takashi Sato <t-sato@yk.jp.nec.com>
Signed-off-by: Masayuki Hamaguchi <m-hamaguchi@ys.jp.nec.com>
Cc: <xfs-masters@oss.sgi.com>
Cc: <linux-ext4@vger.kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Kleikamp <shaggy@austin.ibm.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Alasdair G Kergon <agk@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Implement XFS's large buffer support with the new vmap APIs. See the vmap
rewrite (db64fe02) for some numbers. The biggest improvement that comes from
using the new APIs is avoiding the global KVA allocation lock on every call.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
XFS's vmap batching simply defers a number (up to 64) of vunmaps, and keeps
track of them in a list. To purge the batch, it just goes through the list and
calls vunamp on each one. This is pretty poor: a global TLB flush is generally
still performed on each vunmap, with the most expensive parts of the operation
being the broadcast IPIs and locking involved in the SMP callouts, and the
locking involved in the vmap management -- none of these are avoided by just
batching up the calls. I'm actually surprised it ever made much difference.
(Now that the lazy vmap allocator is upstream, this description is not quite
right, but the vunmap batching still doesn't seem to do much)
Rip all this logic out of XFS completely. I will improve vmap performance
and scalability directly in subsequent patch.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Currently xfs_ino_t is defined as a u64 which can either be an unsigned
long long or on some 64 bit platforms and unsigned long. Just making
it and unsigned long long mean's it's still always 64 bits wide, but we
don't need to resort to cases to print it.
Fixes a warning regression on 64 bit powerpc in current git.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
John Stanley reported EOVERFLOW errors in readdir from his self-build
glibc. I traced this down to glibc enabling d_off overflow checks
in one of the about five million different getdents implementations.
In 2.6.28 Dave Woodhouse moved our readdir double buffering required
for NFS4 readdirplus into nfsd and at that point we lost the capping
of the directory offsets to 32 bit signed values. Johns glibc used
getdents64 to even implement readdir for normal 32 bit offset dirents,
and failed with EOVERFLOW only if this happens on the first dirent in
a getdents call. I managed to come up with a testcase that uses
raw getdents and does the EOVERFLOW check manually. We always hit
it with our last entry due to the special end of directory marker.
The patch below is a dumb version of just putting back the masking,
to make sure we have the same behavior as in 2.6.27 and earlier.
I will work on a better and cleaner fix for 2.6.30.
Reported-by: John Stanley <jpsinthemix@verizon.net>
Tested-by: John Stanley <jpsinthemix@verizon.net>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Change the left/right variables to the proper always 64bit xfs_dfsbo_t
type because otherwise compilation fails for Geert on m68k without
CONFIG_LBD:
| fs/xfs/xfs_btree.c: In function 'xfs_btree_readahead_lblock':
| fs/xfs/xfs_btree.c:736: warning: comparison is always true due to limited range of data type
| fs/xfs/xfs_btree.c:741: warning: comparison is always true due to limited range of data type
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
NFS clients or users of the handle ioctls can pass us arbitrary inode
numbers through the exportfs interface. Make sure we use the
XFS_IGET_BULKSTAT so that these don't cause shutdowns due to the corruption
checks. Also translate the EINVAL we get back for invalid inode clusters
into an ESTALE which is more appropinquate, and remove the useless check
for a NULL inode on a successfull xfs_iget return.
I have a testcase to reproduce this using the handle interface which
I will submit to xfsqa.
Reported-by: Mario Becroft <mb@gem.win.co.nz>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
The iolock is dropped and re-acquired around the call to XFS_SEND_NAMESP().
While the iolock is released the file can become cached. We then
'goto retry' and - if we are doing direct I/O - mapping->nrpages may now be
non zero but need_i_mutex will be zero and we will hit the WARN_ON().
Since we have dropped the I/O lock then the file size may have also changed
so what we need to do here is 'goto start' like we do for the XFS_SEND_DATA()
DMAPI event.
We also need to update the filesize before releasing the iolock so that
needs to be done before the XFS_SEND_NAMESP event. If we drop the iolock
before setting the filesize we could race with a truncate.
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
In libxfs xfs_bmbt_disk_get_all needs to handle unaligned data and thus
has been updated to use get_unaligned_be64. In kernelspace we don't strictly
need it as the routine is only used for tracing and xfsidbg, but let's keep
the two implementations in sync.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
xfs_fs_vcmn_err can be called under a spinlock, but does a sleeping memory
allocation to create buffer for it's internal sprintf. Fortunately it's
the only caller of icmn_err, so we can merge the two and have one single
static buffer and spinlock protecting it. While we're at it make sure
we proper __attribute__ format annotations so that the compiler can detect
mismatched format strings.
Reported-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Speculative allocation beyond eof doesn't work properly. It was
broken some time ago after a code cleanup that moved what is now
xfs_iomap_eof_align_last_fsb() and xfs_iomap_eof_want_preallocate()
out of xfs_iomap_write_delay() into separate functions. The code
used to use the current file size in various checks but got changed
to be max(file_size, i_new_size). Since i_new_size is the result
of 'offset + count' then in xfs_iomap_eof_want_preallocate() the
check for '(offset + count) <= isize' will always be true.
ie if 'offset + count' is > ip->i_size then isize will be i_new_size
and equal to 'offset + count'.
This change fixes all the places that used to use the current file
size.
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
We should be using the incore inode size here not the linux inode
size. The incore inode size is always up to date for directories
whereas the linux inode size is not updated for directories.
We've hit assertions in xfs_bmap() and traced it back to the linux
inode size being zero but the incore size being correct.
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Preserve any error returned by the bio layer.
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Tim Shimmin <tes@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Instead of implementing our own checks use inode_change_ok to check for
necessary permission in setattr. There is a slight change in behaviour
as inode_change_ok doesn't allow i_mode updates to add the suid or sgid
without superuser privilegues while the old XFS code just stripped away
those bits from the file mode.
(First sent on Semptember 29th)
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
XFS has a mode called invisble I/O that doesn't update any of the
timestamps. It's used for HSM-style applications and exposed through
the nasty open by handle ioctl.
Instead of doing directly assignment of file operations that set an
internal flag for it add a new FMODE_NOCMTIME flag that we can check
in the normal file operations.
(addition of the generic VFS flag has been ACKed by Al as an interims
solution)
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
- xfs_sb.h add the XFS_SB_VERSION2_PARENTBIT features2 that has been
around in userspace for some time
- xfs_inode.h: move a few things out of __KERNEL__ that are needed by
userspace
- xfs_mount.h: only include xfs_sync.h under __KERNEL__
- xfs_inode.c: minor whitespace fixup. I accidentaly changes this when
importing this file for use by userspace.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Check for the project ID after attaching all inodes to the transaction.
That way the unlock in the error case is done by the transaction subsystem,
which guaratees that is uses the right flags (which was wrong from day one
of this check), and avoids having special code unlocking an array of inodes
with potential duplicates. Attaching the inode first is the method used
by xfs_rename and the other namespace methods all other error that require
multiple locked inodes.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Replace the b_fspriv pointer and it's ugly accessors with a properly types
xfs_mount pointer. Also switch log reocvery over to it instead of using
b_fspriv for the mount pointer.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
None of this code appears to be used anywhere so remove it.
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
When project quota is active and is being used for directory tree
quota control, we disallow rename outside the current directory
tree. This requires a check to be made after all the inodes
involved in the rename are locked. We fail to unlock the inodes
correctly if we disallow the rename when the target is outside the
current directory tree. This results in a hang on the next access
to the inodes involved in failed rename.
Reported-by: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Tested-by: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Hit this assert because an inode was tagged with XFS_ICI_RECLAIM_TAG but
not XFS_IRECLAIMABLE|XFS_IRECLAIM. This is because xfs_iget_cache_hit()
first clears XFS_IRECLAIMABLE and then calls __xfs_inode_clear_reclaim_tag()
while only holding the pag_ici_lock in read mode so we can race with
xfs_reclaim_inodes_ag(). Looks like xfs_reclaim_inodes_ag() will do the
right thing anyway so just remove the assert.
Thanks to Christoph for pointing out where the problem was.
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
entries_size is probably left over from when we used to pass the
size to kmem_free().
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
We check the return value of all other calls to xfs_buf_get_noaddr().
Make sense to do it here too.
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
When project quota is active and is being used for directory tree
quota control, we disallow rename outside the current directory
tree. This requires a check to be made after all the inodes
involved in the rename are locked. We fail to unlock the inodes
correctly if we disallow the rename when the target is outside the
current directory tree. This results in a hang on the next access
to the inodes involved in failed rename.
Reported-by: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Tested-by: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Move the inode tracing into xfs_iget.c / xfs_inode.h and kill xfs_vnode.c
now that it's empty.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
The whole machinery to wait on I/O completion is related to the I/O path
and should be there instead of in xfs_vnode.c. Also give the functions
more descriptive names.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
There's just one caller of this helper, and it's much cleaner to just merge
the xfs_do_force_shutdown call into it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
There's almost nothing left in this function, instead remove the IRELE
on the real times inodes and the call to XFS_QM_UNMOUNT into xfs_unmountfs.
For the regular unmount case that means it now also happenes after dmapi
notification, but otherwise there is no difference in behaviour.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Currently we explicitly call xfs_iflush on the quota, real-time and root
inodes from xfs_unmount_flush. But we just called xfs_sync_inodes with
SYNC_ATTR and do an XFS_bflush aka xfs_flush_buftarg to make sure all inodes
are on disk already, so there is no need for these special cases.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
Use xfs_trans_ijoin in xfs_trans_iget in case we need to join an inode into
a transaction instead of opencoding it. Based on a discussion with and an
incomplete patch from Niv Sardi.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Niv Sardi <xaiki@sgi.com>
