Under heavy metadata load we are seeing log hangs. The AIL has items in it
ready to be pushed, and they are within the push target window. However,
we are not pushing them when the last pushed LSN is less than the LSN of
the first log item on the AIL. This is a regression introduced by the AIL
push cursor modifications.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32409a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NTim Shimmin <tes@sgi.com>

To make sure we free the security data inodes need to be freed using the
proper VFS helper (which we also need to export for this). We mark these
inodes bad so we can skip the flush path for them.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32398a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>

To make sure we free the security data inodes need to be freed using
the proper VFS helper (which we also need to export for this). We mark
these inodes bad so we can skip the flush path for them.
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>

xfs_bulkstat only wants the dinode, offset and buffer from a given inode
number. Instead of using xfs_itobp on a fake inode which is complicated
and currently leads to leaks of the security data just use xfs_inotobp
which is designed to do exactly the kind of lookup xfs_bulkstat wants. The
only thing that's missing in xfs_inotobp is a flags paramter that let's us
pass down XFS_IMAP_BULKSTAT, but that can easily added.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32397a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>

If we are syncing data in xfs_sync_inodes_ag(), the VFS inode must still
be referencable as the dirty data state is carried on the VFS inode. hence
if we can't get a reference via igrab(), the inode must be in reclaim
which implies that it has no dirty data attached.

Leave such inodes to the reclaim code to flush the dirty inode state to
disk and so avoid attempting to access the VFS inode when it may not exist
in xfs_sync_inodes_ag().

Version 4:
o don't reference linux inode until after igrab() succeeds

Version 3:
o converted unlock/rele to an xfs_iput() call.

Version 2:
o change igrab logic to be more linear
o remove initial reclaimable inode check now that we are using
  igrab() failure to find reclaimable inodes
o assert that igrab failure occurs only on reclaimable inodes
o clean up inode locking - only grab the iolock if we are doing
  a SYNC_DELWRI call and we have a dirty inode.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32391a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NPeter Leckie <pleckie@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

When we are inside a radix tree preload region, we cannot sleep. Recently
we moved the inode locking inside the preload region for the inode radix
tree. Fix that, and fix a missed unlock in another error path in the same
code at the same time.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32385a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

The dp to ip comment should be for the unconditional xfs_droplink call,
and the "." link obviously only exists for directories, so it should be in
the is_dir conditional.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32374a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDonald Douwsma <donaldd@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

iosizelog shouldn't be the same as iosize but the logarithm of it. Then
again the current biosize option doesn't make much sense to me as it
doesn't set the preferred I/O size as mentioned in the comment next to it
but rather the allocation size and thus is identical to the allocsize
option (except for the missing logarithm). It's also not documented in
Documentation/filesystems/xfs.txt or the mount manpage.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32373a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDonald Douwsma <donaldd@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Noquota should clear all mount options, and not just user and group quota.
Probably doesn't matter very much in real life.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32372a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDonald Douwsma <donaldd@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

No need to parse the mount option into a structure before applying it to
struct xfs_mount.

The content of xfs_start_flags gets merged into xfs_parseargs. Calls
inbetween don't care and can use mount members instead of the args struct.

This patch uncovered that the mount option for shared filesystems wasn't
ever exposed on Linux. The code to handle it is #if 0'ed in this patch
pending a decision on this feature. I'll send a writeup about it to the
list soon.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32371a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDonald Douwsma <donaldd@sgi.com>
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

When we are about to add a new item to a transaction in recovery, we need
to check that it is valid first. Currently we just assert that header
magic number matches, but in production systems that is not present and we
add a corrupted transaction to the list to be processed. This results in a
kernel oops later when processing the corrupted transaction.

Instead, if we detect a corrupted transaction, abort recovery and leave
the user to clean up the mess that has occurred.

SGI-PV: 988145

SGI-Modid: xfs-linux-melb:xfs-kern:32356a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Change all the remaining AIL API functions that are passed struct
xfs_mount pointers to pass pointers directly to the struct xfs_ail being
used. With this conversion, all external access to the AIL is via the
struct xfs_ail. Hence the operation and referencing of the AIL is almost
entirely independent of the xfs_mount that is using it - it is now much
more tightly tied to the log and the items it is tracking in the log than
it is tied to the xfs_mount.

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32353a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Add an xfs_ail pointer to log items so that the log items can reference
the AIL directly during callbacks without needed a struct xfs_mount.

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32352a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

When we need to go from the log to the AIL, we have to go via the
xfs_mount. Add a xfs_ail pointer to the log so we can go directly to the
AIL associated with the log.

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32351a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Bring the ail lock inside the struct xfs_ail. This means the AIL can be
entirely manipulated via the struct xfs_ail rather than needing both the
struct xfs_mount and the struct xfs_ail.

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32350a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

When copying lsn's from the log item to the inode or dquot flush lsn, we
currently grab the AIL lock. We do this because the LSN is a 64 bit
quantity and it needs to be read atomically. The lock is used to guarantee
atomicity for 32 bit platforms.

Make the LSN copying a small function, and make the function used
conditional on BITS_PER_LONG so that 64 bit machines don't need to take
the AIL lock in these places.

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32349a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

With the new cursor interface, it makes sense to make all the traversing
code use the cursor interface and make the old one go away. This means
more of the AIL interfacing is done by passing struct xfs_ail pointers
around the place instead of struct xfs_mount pointers.

We can replace the use of xfs_trans_first_ail() in xfs_log_need_covered()
as it is only checking if the AIL is empty. We can do that with a call to
xfs_trans_ail_tail() instead, where a zero LSN returned indicates and
empty AIL...

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32348a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

To replace the current generation number ensuring sanity of the AIL
traversal, replace it with an external cursor that is linked to the AIL.

Basically, we store the next item in the cursor whenever we want to drop
the AIL lock to do something to the current item. When we regain the lock.
the current item may already be free, so we can't reference it, but the
next item in the traversal is already held in the cursor.

When we move or delete an object, we search all the active cursors and if
there is an item match we clear the cursor(s) that point to the object.
This forces the traversal to restart transparently.

We don't invalidate the cursor on insert because the cursor still points
to a valid item. If the intem is inserted between the current item and the
cursor it does not matter; the traversal is considered to be past the
insertion point so it will be picked up in the next traversal.

Hence traversal restarts pretty much disappear altogether with this method
of traversal, which should substantially reduce the overhead of pushing on
a busy AIL.

Version 2 o add restart logic o comment cursor interface o minor cleanups

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32347a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Rather than embedding the struct xfs_ail in the struct xfs_mount, allocate
it during AIL initialisation. Add a back pointer to the struct xfs_ail so
that we can pass around the xfs_ail and still be able to access the
xfs_mount if need be. This is th first step involved in isolating the AIL
implementation from the surrounding filesystem code.

SGI-PV: 988143

SGI-Modid: xfs-linux-melb:xfs-kern:32346a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

When we create a directory, we reserve a number of blocks for the maximum
possible expansion of of the directory due to various btree splits,
freespace allocation, etc. Unfortunately, each allocation is not reflected
in the total number of blocks still available to the transaction, so the
maximal reservation is used over and over again.

This leads to problems where an allocation group has only enough blocks
for *some* of the allocations required for the directory modification.
After the first N allocations, the remaining blocks in the allocation
group drops below the total reservation, and subsequent allocations fail
because the allocator will not allow the allocation to proceed if the AG
does not have the enough blocks available for the entire allocation total.

This results in an ENOSPC occurring after an allocation has already
occurred. This results in aborting the directory operation (leaving the
directory in an inconsistent state) and cancelling a dirty transaction,
which results in a filesystem shutdown.

Avoid the problem by reflecting the number of blocks allocated in any
directory expansion in the total number of blocks available to the
modification in progress. This prevents a directory modification from
being aborted part way through with an ENOSPC.

SGI-PV: 988144

SGI-Modid: xfs-linux-melb:xfs-kern:32340a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

If the last block of the AG has inodes in it and the AG is an exactly
power-of-2 size then the last inode in the AG points to the last block in
the AG. If we try to find the next inode in the AG by adding one to the
inode number, we increment the inode number past the size of the AG. The
result is that the macro XFS_INO_TO_AGINO() will strip the AG portion of
the inode number and return an inode number of zero.

That is, instead of terminating the lookup loop because we hit the inode
number went outside the valid range for the AG, the search index returns
to zero and we start traversing the radix tree from the start again. This
results in an endless loop in xfs_sync_inodes_ag().

Fix it be detecting if the new search index decreases as a result of
incrementing the current inode number. That indicate an overflow and hence
that we have finished processing the AG so we can terminate the loop.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32335a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Now that the deleted inodes list is unused, kill it. This also removes the
i_reclaim list head from the xfs_inode, shrinking it by two pointers.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32334a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Use the reclaim tag to walk the radix tree and find the inodes under
reclaim. This was the only user of the deleted inode list.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32333a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Prepare for removing the deleted inode list by marking inodes for reclaim
in the inode radix trees so that we can use the radix trees to find
reclaimable inodes.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32331a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

The function names xfs_finish_reclaim and xfs_finish_reclaim_all are not
very descriptive of what they are reclaiming. Rename to
xfs_reclaim_inode[s] to match the xfs_sync_inodes() function.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32330a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Background inode reclaim is run by the xfssyncd. Move the reclaim worker
functions to be close to the sync code as the are very similar in
structure and are both run from the same background thread.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32329a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32325a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

With the combined linux and XFS inode, we need to ensure that the combined
structure is not freed before the generic code is finished with the inode.
As it turns out, there is a case where the XFS inode is freed before the
linux inode - when xfs_reclaim() is called from ->clear_inode() on a clean
inode, the xfs inode is freed during that call. The generic code
references the inode after the ->clear_inode() call, so this is a use
after free situation.

Fix the problem by moving the xfs_reclaim() call to ->destroy_inode()
instead of in ->clear_inode(). This ensures the combined inode structure
is not freed until after the generic code has finished with it.

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32324a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

To avoid issues with different lifecycles of XFS and Linux inodes, embedd
the linux inode inside the XFS inode. This means that the linux inode has
the same lifecycle as the XFS inode, even when it has been released by the
OS. XFS inodes don't live much longer than this (a short stint in reclaim
at most), so there isn't significant memory usage penalties here.

Version 3 o kill xfs_icount()

Version 2 o remove unused commented out code from xfs_iget(). o kill
useless cast in VFS_I()

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32323a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

To allow XFS to combine the XFS and linux inodes into a single
structure, we need to drive inode lookup from the XFS inode cache,
not the generic inode cache. This means that we need initialise a
struct inode from a context outside alloc_inode() as it is no longer
used by XFS.

After inode allocation and initialisation, we need to add the inode
to the superblock list, the in-use list, hash it and do some
accounting. This all needs to be done with the inode_lock held and
there are already several places in fs/inode.c that do this list
manipulation.  Factor out the common code, add a locking wrapper and
export the function so ti can be called from XFS.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

To allow XFS to combine the XFS and linux inodes into a single
structure, we need to drive inode lookup from the XFS inode cache,
not the generic inode cache. This means that we need initialise a
struct inode from a context outside alloc_inode() as it is no longer
used by XFS.

Factor and export the struct inode initialisation code from
alloc_inode() to inode_init_always() as a counterpart to
inode_init_once().  i.e. we have to call this init function for each
inode instantiation (always), as opposed inode_init_once() which is
only called on slab object instantiation (once).
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Once the Linux inode and the XFS inode are combined, we cannot rely on
just check if the linux inode exists as a method of determining if it is
valid or not. Hence we should always call xfs_mark_inode_dirty_sync()
instead as it does the correct checks to determine if the liinux inode is
in a valid state or not.

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32318a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

There are really two cases in xfs_iget_core(). The first is the cache hit
case, the second is the miss case. They share very little code, and hence
can easily be factored out into separate functions. This makes the code
much easier to understand and subsequently modify.

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32317a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

We can only read inode->i_count if the inode is actually there and not a
NULL pointer. This was introduced in one of the recent sync patches.

SGI-PV: 988255

SGI-Modid: xfs-linux-melb:xfs-kern:32315a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

With all the other filesystem sync code it in xfs_sync.c including the
data quiesce code, it makes sense to move the remaining quiesce code to
the same place.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32312a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

There are no more callers to xfs_sync() now, so remove the function
altogther.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32311a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

SYNC_CLOSE is only ever used and checked in conjunction with SYNC_WAIT,
and this only done in one spot. The only thing this does is make
XFS_bflush() calls to the data buftargs.

This will happen very shortly afterwards the xfs_sync() call anyway in the
unmount path via the xfs_close_devices(), so this code is redundant and
can be removed. That only user of SYNC_CLOSE is now gone, so kill the flag
completely.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32310a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Continue to de-multiplex xfs_sync be replacing all SYNC_DELWRI callers
with direct calls functions that do the work. Isolate the data quiesce
case to a function in xfs_sync.c. Isolate the FSDATA case with explicit
calls to xfs_sync_fsdata().

Version 2: o Push delwri related log forces into xfs_sync_inodes().

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32309a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Continue to de-multiplex xfs_sync be replacing all SYNC_ATTR callers with
direct calls xfs_sync_inodes(). Add an assert into xfs_sync() to ensure we
caught all the SYNC_ATTR callers.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32308a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Start de-multiplexing xfs_sync() by making xfs_sync_worker() call the
specific sync functions it needs. This is only a small, unique subset of
the entire xfs_sync() code so is easier to follow.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32307a
Signed-off-by: NDavid Chinner <david@fromorbit.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>