When many hundreds to thousands of threads all try to do simultaneous
transactions and the log is in a tail-pushing situation (i.e. full), we
can get multiple threads walking the AIL list and contending on the AIL
lock.
The AIL push is, in effect, a simple I/O dispatch algorithm complicated by
the ordering constraints placed on it by the transaction subsystem. It
really does not need multiple threads to push on it - even when only a
single CPU is pushing the AIL, it can push the I/O out far faster that
pretty much any disk subsystem can handle.
So, to avoid contention problems stemming from multiple list walkers, move
the list walk off into another thread and simply provide a "target" to
push to. When a thread requires a push, it sets the target and wakes the
push thread, then goes to sleep waiting for the required amount of space
to become available in the log.
This mechanism should also be a lot fairer under heavy load as the waiters
will queue in arrival order, rather than queuing in "who completed a push
first" order.
Also, by moving the pushing to a separate thread we can do more
effectively overload detection and prevention as we can keep context from
loop iteration to loop iteration. That is, we can push only part of the
list each loop and not have to loop back to the start of the list every
time we run. This should also help by reducing the number of items we try
to lock and/or push items that we cannot move.
Note that this patch is not intended to solve the inefficiencies in the
AIL structure and the associated issues with extremely large list
contents. That needs to be addresses separately; parallel access would
cause problems to any new structure as well, so I'm only aiming to isolate
the structure from unbounded parallelism here.
SGI-PV: 972759
SGI-Modid: xfs-linux-melb:xfs-kern:30371a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
These are mostly locking annotations, marking things static, casts where
needed and declaring stuff in header files.
SGI-PV: 971186
SGI-Modid: xfs-linux-melb:xfs-kern:30002a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Refactoring xfs_mountfs() to call sub-functions for logical chunks can
help save a bit of stack, and can make it easier to read this long
function.
The mount path is one of the longest common callchains, easily getting to
within a few bytes of the end of a 4k stack when over lvm, quotas are
enabled, and quotacheck must be done.
With this change on top of the other stack-related changes I've sent, I
can get xfs to survive a normal xfsqa run on 4k stacks over lvm.
SGI-PV: 971186
SGI-Modid: xfs-linux-melb:xfs-kern:29834a
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
remove spinlock init abstraction macro in spin.h, remove the callers, and
remove the file. Move no-op spinlock_destroy to xfs_linux.h Cleanup
spinlock locals in xfs_mount.c
SGI-PV: 970382
SGI-Modid: xfs-linux-melb:xfs-kern:29751a
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Un-obfuscate XFS_SB_LOCK, remove XFS_SB_LOCK->mutex_lock->spin_lock
macros, call spin_lock directly, remove extraneous cookie holdover from
old xfs code, and change lock type to spinlock_t.
SGI-PV: 970382
SGI-Modid: xfs-linux-melb:xfs-kern:29746a
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Currently there is an indirection called ioops in the XFS data I/O path.
Various functions are called by functions pointers, but there is no
coherence in what this is for, and of course for XFS itself it's entirely
unused. This patch removes it instead and significantly reduces source and
binary size of XFS while making maintaince easier.
SGI-PV: 970841
SGI-Modid: xfs-linux-melb:xfs-kern:29737a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
m_growlock only needs plain binary mutex semantics, so use a struct mutex
instead of a semaphore for it.
SGI-PV: 968563
SGI-Modid: xfs-linux-melb:xfs-kern:29512a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Now that struct bhv_vfs doesn't have any members left we can kill it and
go directly from the super_block to the xfs_mount everywhere.
SGI-PV: 969608
SGI-Modid: xfs-linux-melb:xfs-kern:29509a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
All flags are added to xfs_mount's m_flag instead. Note that the 32bit
inode flag was duplicated in both of them, but only cleared in the mount
when it was not nessecary due to the filesystem beeing small enough. Two
flags are still required here - one to indicate the mount option setting,
and one to indicate if it applies or not.
SGI-PV: 969608
SGI-Modid: xfs-linux-melb:xfs-kern:29507a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
vfs_altfsid was just a pointer to mp->m_fixedfsid so we can trivially
replace it with the latter. vfs_fsid also was identical to m_fixedfsid
through rather obfuscated ways so we can kill it as well and simply its
only user.
SGI-PV: 969608
SGI-Modid: xfs-linux-melb:xfs-kern:29506a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Also remove the now dead behavior code.
SGI-PV: 969608
SGI-Modid: xfs-linux-melb:xfs-kern:29505a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
In the next patch we need to look at the mount structure until just before
it's freed, so we need to be able to free it as the very last thing in
xfs_unmount.
SGI-PV: 969608
SGI-Modid: xfs-linux-melb:xfs-kern:29501a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
One of the perpetual scaling problems XFS has is indexing it's incore
inodes. We currently uses hashes and the default hash sizes chosen can
only ever be a tradeoff between memory consumption and the maximum
realistic size of the cache.
As a result, anyone who has millions of inodes cached on a filesystem
needs to tunes the size of the cache via the ihashsize mount option to
allow decent scalability with inode cache operations.
A further problem is the separate inode cluster hash, whose size is based
on the ihashsize but is smaller, and so under certain conditions (sparse
cluster cache population) this can become a limitation long before the
inode hash is causing issues.
The following patchset removes the inode hash and cluster hash and
replaces them with radix trees to avoid the scalability limitations of the
hashes. It also reduces the size of the inodes by 3 pointers....
SGI-PV: 969561
SGI-Modid: xfs-linux-melb:xfs-kern:29481a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Creates a new xfs_dsb_t that is __be annotated and keeps xfs_sb_t for the
incore one. xfs_xlatesb is renamed to xfs_sb_to_disk and only handles the
incore -> disk conversion. A new helper xfs_sb_from_disk handles the other
direction and doesn't need the slightly hacky table-driven approach
because we only ever read the full sb from disk.
The handling of shared r/o filesystems has been buggy on little endian
system and fixing this required shuffling around of some code in that
area.
SGI-PV: 968563
SGI-Modid: xfs-linux-melb:xfs-kern:29477a
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
If we fail to open the the log device buftarg, we can fall through to
error handling code that fails to check for a NULL log device buftarg
before calling xfs_free_buftarg().
This patch fixes the issue by checking mp->m_logdev_targp against NULL in
xfs_unmountfs_close() and doing the proper xfs_blkdev_put(logdev); and
xfs_blkdev_put(rtdev); on (!mp->m_rtdev_targp) in xfs_mount().
Discovered by the Coverity checker.
SGI-PV: 968563
SGI-Modid: xfs-linux-melb:xfs-kern:29328a
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Remove scaling of inode "clusters" based on machine memory; small cluster
cut-point was an unrealistic 32MB and was probably never tested.
Removes another user of xfs_physmem.
SGI-PV: 968563
SGI-Modid: xfs-linux-melb:xfs-kern:29324a
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
m_nreadaheads in the mount struct is never used; remove it and the various
macros assigned to it. Also remove a couple other unused macros in the
same areas.
Removes one user of xfs_physmem.
SGI-PV: 968563
SGI-Modid: xfs-linux-melb:xfs-kern:29322a
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
During delayed allocation extent conversion or unwritten extent
conversion, we need to reserve some blocks for transactions reservations.
We need to reserve these blocks in case a btree split occurs and we need
to allocate some blocks.
Unfortunately, we've only ever reserved the number of data blocks we are
allocating, so in both the unwritten and delalloc case we can get ENOSPC
to the transaction reservation. This is bad because in both cases we
cannot report the failure to the writing application.
The fix is two-fold:
1 - leverage the reserved block infrastructure XFS already
has to reserve a small pool of blocks by default to allow
specially marked transactions to dip into when we are at
ENOSPC.
Default setting is min(5%, 1024 blocks).
2 - convert critical transaction reservations to be allowed
to dip into this pool. Spots changed are delalloc
conversion, unwritten extent conversion and growing a
filesystem at ENOSPC.
This also allows growing the filesytsem to succeed at ENOSPC.
SGI-PV: 964468
SGI-Modid: xfs-linux-melb:xfs-kern:28865a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When we are unmounting the filesystem, we flush all the inodes to disk.
Unfortunately, if we have an inode cluster that has just been freed and
marked stale sitting in an incore log buffer (i.e. hasn't been flushed to
disk), it will be holding all the flush locks on the inodes in that
cluster.
xfs_iflush_all() which is called during unmount walks all the inodes
trying to reclaim them, and it doing so calls xfs_finish_reclaim() on each
inode. If the inode is dirty, if grabs the flush lock and flushes it.
Unfortunately, find dirty inodes that already have their flush lock held
and so we sleep.
At this point in the unmount process, we are running single-threaded.
There is nothing more that can push on the log to force the transaction
holding the inode flush locks to disk and hence we deadlock.
The fix is to issue a log force before flushing the inodes on unmount so
that all the flush locks will be released before we start flushing the
inodes.
SGI-PV: 964538
SGI-Modid: xfs-linux-melb:xfs-kern:28862a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When growing a filesystem we don't check to see if the new size overflows
the page cache index range, so we can do silly things like grow a
filesystem page 16TB on a 32bit. Check new filesystem sizes against the
limits the kernel can support.
SGI-PV: 957886
SGI-Modid: xfs-linux-melb:xfs-kern:28563a
Signed-Off-By: Nathan Scott <nscott@aconex.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Since nonboot CPUs are now disabled after tasks and devices have been
frozen and the CPU hotplug infrastructure is used for this purpose, we need
special CPU hotplug notifications that will help the CPU-hotplug-aware
subsystems distinguish normal CPU hotplug events from CPU hotplug events
related to a system-wide suspend or resume operation in progress. This
patch introduces such notifications and causes them to be used during
suspend and resume transitions. It also changes all of the
CPU-hotplug-aware subsystems to take these notifications into consideration
(for now they are handled in the same way as the corresponding "normal"
ones).
[oleg@tv-sign.ru: cleanups]
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Gautham R Shenoy <ego@in.ibm.com>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After filesystem recovery the superblock is re-read to bring in any
changes. If the per-cpu superblock counters are not re-initialized from
the superblock then the next time the per-cpu counters are disabled they
might overwrite the global counter with a bogus value.
SGI-PV: 957348
SGI-Modid: xfs-linux-melb:xfs-kern:27999a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
The block reservation mechanism has been broken since the per-cpu
superblock counters were introduced. Make the block reservation code work
with the per-cpu counters by syncing the counters, snapshotting the amount
of available space and then doing a modifcation of the counter state
according to the result. Continue in a loop until we either have no space
available or we reserve some space.
SGI-PV: 956323
SGI-Modid: xfs-linux-melb:xfs-kern:27895a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
The free block modification code has a 32bit interface, limiting the size
the filesystem can be grown even on 64 bit machines. On 32 bit machines,
there are other 32bit variables in transaction structures and interfaces
that need to be expanded to allow this to work.
SGI-PV: 959978
SGI-Modid: xfs-linux-melb:xfs-kern:27894a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
The existing per-cpu superblock counter code uses the global superblock
spin lock when we approach ENOSPC for global synchronisation. On larger
machines than this code was originally tested on this can still get
catastrophic spinlock contention due increasing rebalance frequency near
ENOSPC.
By introducing a sleeping lock that is used to serialise balances and
modifications near ENOSPC we prevent contention from needlessly from
wasting the CPU time of potentially hundreds of CPUs.
To reduce the number of balances occuring, we separate the need rebalance
case from the slow allocate case. Now, a counter running dry will trigger
a rebalance during which counters are disabled. Any thread that sees a
disabled counter enters a different path where it waits on the new mutex.
When it gets the new mutex, it checks if the counter is disabled. If the
counter is disabled, then we _know_ that we have to use the global counter
and lock and it is safe to do so immediately. Otherwise, we drop the mutex
and go back to trying the per-cpu counters which we know were re-enabled.
SGI-PV: 952227
SGI-Modid: xfs-linux-melb:xfs-kern:27612a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
gcc-4.1 and more recent aggressively inline static functions which
increases XFS stack usage by ~15% in critical paths. Prevent this from
occurring by adding noinline to the STATIC definition.
Also uninline some functions that are too large to be inlined and were
causing problems with CONFIG_FORCED_INLINING=y.
Finally, clean up all the different users of inline, __inline and
__inline__ and put them under one STATIC_INLINE macro. For debug kernels
the STATIC_INLINE macro uninlines those functions.
SGI-PV: 957159
SGI-Modid: xfs-linux-melb:xfs-kern:27585a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: David Chatterton <chatz@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
The fix for recent ENOSPC deadlocks introduced certain limitations on
allocations. The fix could cause xfssyncd to loop endlessly if we did not
leave some space free for the allocator to work correctly. Basically, we
needed to ensure that we had at least 4 blocks free for an AG free list
and a block for the inode bmap btree at all times.
However, this did not take into account the fact that each AG has a free
list that needs 4 blocks. Hence any filesystem with more than one AG could
cause oversubscription of free space and make xfssyncd spin forever trying
to allocate space needed for AG freelists that was not available in the
AG.
The following patch reserves space for the free lists in all AGs plus the
inode bmap btree which prevents oversubscription. It also prevents those
blocks from being reported as free space (as they can never be used) and
makes the SMP in-core superblock accounting code and the reserved block
ioctl respect this requirement.
SGI-PV: 955674
SGI-Modid: xfs-linux-melb:xfs-kern:26894a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: David Chatterton <chatz@sgi.com>
* git://oss.sgi.com:8090/nathans/xfs-2.6:
[XFS] Fixup whitespace damage in log_write, remove final warning.
[XFS] Rework code snippets slightly to remove remaining recent-gcc
[XFS] Fix realtime subvolume expansion, a porting bug b0rked it. Coverity
[XFS] Remove a race condition where a linked inode could BUG_ON in
[XFS] Remove redundant directory checks from inode link operation.
[XFS] Remove a couple of no-longer-used macros.
[XFS] Reduce size of xfs_trans_t structure. * remove ->t_forw, ->t_back --
[XFS] remove unused behaviour lock - shrink XFS vnode as a side effect.
[XFS] * There is trivial "inode => vnode => inode" conversion, but only
[XFS] link(2) on directory is banned in VFS.
Make use the of newly defined hotplug version of cpu_notifier functionality
wherever appropriate.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
transaction within each such operation may involve multiple locking of AGF
buffer. While the freeing extent function has sorted the extents based on
AGF number before entering into transaction, however, when the file system
space is very limited, the allocation of space would try every AGF to get
space allocated, this could potentially cause out-of-order locking, thus
deadlock could happen. This fix mitigates the scarce space for allocation
by setting aside a few blocks without reservation, and avoid deadlock by
maintaining ascending order of AGF locking.
SGI-PV: 947395
SGI-Modid: xfs-linux-melb:xfs-kern:210801a
Signed-off-by: Yingping Lu <yingping@sgi.com>
Signed-off-by: Nathan Scott <nathans@sgi.com>
a preēmpt counter overflow at 256p and above. Change the exclusion
mechanism to use atomic bit operations and busy wait loops to emulate the
spin lock exclusion mechanism but without the preempt count issues.
SGI-PV: 950027
SGI-Modid: xfs-linux-melb:xfs-kern:25338a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Nathan Scott <nathans@sgi.com>
registering a notifier callback that listens to CPU up/down events to
modify the counters appropriately.
SGI-PV: 949726
SGI-Modid: xfs-linux-melb:xfs-kern:25214a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Nathan Scott <nathans@sgi.com>
threads, the incore superblock lock becomes the limiting factor for
buffered write throughput. Make the contended fields in the incore
superblock use per-cpu counters so that there is no global lock to limit
scalability.
SGI-PV: 946630
SGI-Modid: xfs-linux-melb:xfs-kern:25106a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Nathan Scott <nathans@sgi.com>
