* 'for-linus' of git://oss.sgi.com/xfs/xfs:
xfs: stop using the page cache to back the buffer cache
xfs: register the inode cache shrinker before quotachecks
xfs: xfs_trans_read_buf() should return an error on failure
xfs: introduce inode cluster buffer trylocks for xfs_iflush
vmap: flush vmap aliases when mapping fails
xfs: preallocation transactions do not need to be synchronous
Fix up trivial conflicts in fs/xfs/linux-2.6/xfs_buf.c due to plug removal.
Now that the buffer cache has it's own LRU, we do not need to use
the page cache to provide persistent caching and reclaim
infrastructure. Convert the buffer cache to use alloc_pages()
instead of the page cache. This will remove all the overhead of page
cache management from setup and teardown of the buffers, as well as
needing to mark pages accessed as we find buffers in the buffer
cache.
By avoiding the page cache, we also remove the need to keep state in
the page_private(page) field for persistant storage across buffer
free/buffer rebuild and so all that code can be removed. This also
fixes the long-standing problem of not having enough bits in the
page_private field to track all the state needed for a 512
sector/64k page setup.
It also removes the need for page locking during reads as the pages
are unique to the buffer and nobody else will be attempting to
access them.
Finally, it removes the buftarg address space lock as a point of
global contention on workloads that allocate and free buffers
quickly such as when creating or removing large numbers of inodes in
parallel. This remove the 16TB limit on filesystem size on 32 bit
machines as the page index (32 bit) is no longer used for lookups
of metadata buffers - the buffer cache is now solely indexed by disk
address which is stored in a 64 bit field in the buffer.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Alex Elder <aelder@sgi.com>
On 32 bit systems, vmalloc space is limited and XFS can chew through
it quickly as the vmalloc space is lazily freed. This can result in
failure to map buffers, even when there is apparently large amounts
of vmalloc space available. Hence, if we fail to map a buffer, purge
the aliases that have not yet been freed to hopefuly free up enough
vmalloc space to allow a retry to succeed.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
* 'for-2.6.39/core' of git://git.kernel.dk/linux-2.6-block: (65 commits)
Documentation/iostats.txt: bit-size reference etc.
cfq-iosched: removing unnecessary think time checking
cfq-iosched: Don't clear queue stats when preempt.
blk-throttle: Reset group slice when limits are changed
blk-cgroup: Only give unaccounted_time under debug
cfq-iosched: Don't set active queue in preempt
block: fix non-atomic access to genhd inflight structures
block: attempt to merge with existing requests on plug flush
block: NULL dereference on error path in __blkdev_get()
cfq-iosched: Don't update group weights when on service tree
fs: assign sb->s_bdi to default_backing_dev_info if the bdi is going away
block: Require subsystems to explicitly allocate bio_set integrity mempool
jbd2: finish conversion from WRITE_SYNC_PLUG to WRITE_SYNC and explicit plugging
jbd: finish conversion from WRITE_SYNC_PLUG to WRITE_SYNC and explicit plugging
fs: make fsync_buffers_list() plug
mm: make generic_writepages() use plugging
blk-cgroup: Add unaccounted time to timeslice_used.
block: fixup plugging stubs for !CONFIG_BLOCK
block: remove obsolete comments for blkdev_issue_zeroout.
blktrace: Use rq->cmd_flags directly in blk_add_trace_rq.
...
Fix up conflicts in fs/{aio.c,super.c}
* 'for-linus' of git://oss.sgi.com/xfs/xfs: (23 commits)
xfs: don't name variables "panic"
xfs: factor agf counter updates into a helper
xfs: clean up the xfs_alloc_compute_aligned calling convention
xfs: kill support/debug.[ch]
xfs: Convert remaining cmn_err() callers to new API
xfs: convert the quota debug prints to new API
xfs: rename xfs_cmn_err_fsblock_zero()
xfs: convert xfs_fs_cmn_err to new error logging API
xfs: kill xfs_fs_mount_cmn_err() macro
xfs: kill xfs_fs_repair_cmn_err() macro
xfs: convert xfs_cmn_err to xfs_alert_tag
xfs: Convert xlog_warn to new logging interface
xfs: Convert linux-2.6/ files to new logging interface
xfs: introduce new logging API.
xfs: zero proper structure size for geometry calls
xfs: enable delaylog by default
xfs: more sensible inode refcounting for ialloc
xfs: stop using xfs_trans_iget in the RT allocator
xfs: check if device support discard in xfs_ioc_trim()
xfs: prevent leaking uninitialized stack memory in FSGEOMETRY_V1
...
Code has been converted over to the new explicit on-stack plugging,
and delay users have been converted to use the new API for that.
So lets kill off the old plugging along with aops->sync_page().
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Convert the files in fs/xfs/linux-2.6/ to use the new xfs_<level>
logging format that replaces the old Irix inherited cmn_err()
interfaces. While there, also convert naked printk calls to use the
relevant xfs logging function to standardise output format.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Alex Elder <aelder@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Convert from create[_singlethread]_workqueue() to alloc_workqueue().
* xfsdatad_workqueue and xfsconvertd_workqueue are identity converted.
Using higher concurrency limit might be useful but given the
complexity of workqueue usage in xfs, proceeding cautiously seems
better.
* xfs_mru_reap_wq is converted to non-ordered workqueue with max
concurrency of 1 as the work items don't require any specific
ordering and already have proper synchronization. It seems it was
singlethreaded to save worker threads, which is no longer a concern.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Alex Elder <aelder@sgi.com>
Cc: xfs-masters@oss.sgi.com
Cc: Christoph Hellwig <hch@infradead.org>
If we get an IO error on a synchronous superblock write, we attach an
error release function to it so that when the last reference goes away
the release function is called and the buffer is invalidated and
unlocked. The buffer is left locked until the release function is
called so that other concurrent users of the buffer will be locked out
until the buffer error is fully processed.
Unfortunately, for the superblock buffer the filesyetm itself holds a
reference to the buffer which prevents the reference count from
dropping to zero and the release function being called. As a result,
once an IO error occurs on a sync write, the buffer will never be
unlocked and all future attempts to lock the buffer will hang.
To make matters worse, this problems is not unique to such buffers;
if there is a concurrent _xfs_buf_find() running, the lookup will grab
a reference to the buffer and then wait on the buffer lock, preventing
the reference count from ever falling to zero and hence unlocking the
buffer.
As such, the whole b_relse function implementation is broken because it
cannot rely on the buffer reference count falling to zero to unlock the
errored buffer. The synchronous write error path is the only path that
uses this callback - it is used to ensure that the synchronous waiter
gets the buffer error before the error state is cleared from the buffer
by the release function.
Given that the only sychronous buffer writes now go through xfs_bwrite
and the error path in question can only occur for a write of a dirty,
logged buffer, we can move most of the b_relse processing to happen
inline in xfs_buf_iodone_callbacks, just like a normal I/O completion.
In addition to that we make sure the error is not cleared in
xfs_buf_iodone_callbacks, so that xfs_bwrite can reliably check it.
Given that xfs_bwrite keeps the buffer locked until it has waited for
it and checked the error this allows to reliably propagate the error
to the caller, and make sure that the buffer is reliably unlocked.
Given that xfs_buf_iodone_callbacks was the only instance of the
b_relse callback we can remove it entirely.
Based on earlier patches by Dave Chinner and Ajeet Yadav.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Ajeet Yadav <ajeet.yadav.77@gmail.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
Introduce a per-buftarg LRU for memory reclaim to operate on. This
is the last piece we need to put in place so that we can fully
control the buffer lifecycle. This allows XFS to be responsibile for
maintaining the working set of buffers under memory pressure instead
of relying on the VM reclaim not to take pages we need out from
underneath us.
The implementation introduces a b_lru_ref counter into the buffer.
This is currently set to 1 whenever the buffer is referenced and so is used to
determine if the buffer should be added to the LRU or not when freed.
Effectively it allows lazy LRU initialisation of the buffer so we do not need
to touch the LRU list and locks in xfs_buf_find().
Instead, when the buffer is being released and we drop the last
reference to it, we check the b_lru_ref count and if it is none zero
we re-add the buffer reference and add the inode to the LRU. The
b_lru_ref counter is decremented by the shrinker, and whenever the
shrinker comes across a buffer with a zero b_lru_ref counter, if
released the LRU reference on the buffer. In the absence of a lookup
race, this will result in the buffer being freed.
This counting mechanism is used instead of a reference flag so that
it is simple to re-introduce buffer-type specific reclaim reference
counts to prioritise reclaim more effectively. We still have all
those hooks in the XFS code, so this will provide the infrastructure
to re-implement that functionality.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Before we introduce per-buftarg LRU lists, split the shrinker
implementation into per-buftarg shrinker callbacks. At the moment
we wake all the xfsbufds to run the delayed write queues to free
the dirty buffers and make their pages available for reclaim.
However, with an LRU, we want to be able to free clean, unused
buffers as well, so we need to separate the xfsbufd from the
shrinker callbacks.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
As reported by Nick Piggin, XFS is suffering from long pauses under
highly concurrent workloads when hosted on ramdisks. The problem is
that an inode buffer is stuck in the pinned state in memory and as a
result either the inode buffer or one of the inodes within the
buffer is stopping the tail of the log from being moved forward.
The system remains in this state until a periodic log force issued
by xfssyncd causes the buffer to be unpinned. The main problem is
that these are stale buffers, and are hence held locked until the
transaction/checkpoint that marked them state has been committed to
disk. When the filesystem gets into this state, only the xfssyncd
can cause the async transactions to be committed to disk and hence
unpin the inode buffer.
This problem was encountered when scaling the busy extent list, but
only the blocking lock interface was fixed to solve the problem.
Extend the same fix to the buffer trylock operations - if we fail to
lock a pinned, stale buffer, then force the log immediately so that
when the next attempt to lock it comes around, it will have been
unpinned.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
The delayed write buffer split trace currently issues a trace for
every buffer it scans. These buffers are not necessarily queued for
delayed write. Indeed, when buffers are pinned, there can be
thousands of traces of buffers that aren't actually queued for
delayed write and the ones that are are lost in the noise. Move the
trace point to record only buffers that are split out for IO to be
issued on.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Instead of always assigning an increasing inode number in new_inode
move the call to assign it into those callers that actually need it.
For now callers that need it is estimated conservatively, that is
the call is added to all filesystems that do not assign an i_ino
by themselves. For a few more filesystems we can avoid assigning
any inode number given that they aren't user visible, and for others
it could be done lazily when an inode number is actually needed,
but that's left for later patches.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
* 'for-linus' of git://oss.sgi.com/xfs/xfs: (36 commits)
xfs: semaphore cleanup
xfs: Extend project quotas to support 32bit project ids
xfs: remove xfs_buf wrappers
xfs: remove xfs_cred.h
xfs: remove xfs_globals.h
xfs: remove xfs_version.h
xfs: remove xfs_refcache.h
xfs: fix the xfs_trans_committed
xfs: remove unused t_callback field in struct xfs_trans
xfs: fix bogus m_maxagi check in xfs_iget
xfs: do not use xfs_mod_incore_sb_batch for per-cpu counters
xfs: do not use xfs_mod_incore_sb for per-cpu counters
xfs: remove XFS_MOUNT_NO_PERCPU_SB
xfs: pack xfs_buf structure more tightly
xfs: convert buffer cache hash to rbtree
xfs: serialise inode reclaim within an AG
xfs: batch inode reclaim lookup
xfs: implement batched inode lookups for AG walking
xfs: split out inode walk inode grabbing
xfs: split inode AG walking into separate code for reclaim
...
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: remove in_workqueue_context()
workqueue: Clarify that schedule_on_each_cpu is synchronous
memory_hotplug: drop spurious calls to flush_scheduled_work()
shpchp: update workqueue usage
pciehp: update workqueue usage
isdn/eicon: don't call flush_scheduled_work() from diva_os_remove_soft_isr()
workqueue: add and use WQ_MEM_RECLAIM flag
workqueue: fix HIGHPRI handling in keep_working()
workqueue: add queue_work and activate_work trace points
workqueue: prepare for more tracepoints
workqueue: implement flush[_delayed]_work_sync()
workqueue: factor out start_flush_work()
workqueue: cleanup flush/cancel functions
workqueue: implement alloc_ordered_workqueue()
Fix up trivial conflict in fs/gfs2/main.c as per Tejun
Get rid of init_MUTEX[_LOCKED]() and use sema_init() instead.
(Ported to current XFS code by <aelder@sgi.com>.)
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Stop having two different names for many buffer functions and use
the more descriptive xfs_buf_* names directly.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
The buffer cache hash is showing typical hash scalability problems.
In large scale testing the number of cached items growing far larger
than the hash can efficiently handle. Hence we need to move to a
self-scaling cache indexing mechanism.
I have selected rbtrees for indexing becuse they can have O(log n)
search scalability, and insert and remove cost is not excessive,
even on large trees. Hence we should be able to cache large numbers
of buffers without incurring the excessive cache miss search
penalties that the hash is imposing on us.
To ensure we still have parallel access to the cache, we need
multiple trees. Rather than hashing the buffers by disk address to
select a tree, it seems more sensible to separate trees by typical
access patterns. Most operations use buffers from within a single AG
at a time, so rather than searching lots of different lists,
separate the buffer indexes out into per-AG rbtrees. This means that
searches during metadata operation have a much higher chance of
hitting cache resident nodes, and that updates of the tree are less
likely to disturb trees being accessed on other CPUs doing
independent operations.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
For RT and external log devices, we never use hashed buffers on them
now. Remove the buftarg hash tables that are set up for them.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
Filesystem level managed buffers are buffers that have their
lifecycle controlled by the filesystem layer, not the buffer cache.
We currently cache these buffers, which makes cleanup and cache
walking somewhat troublesome. Convert the fs managed buffers to
uncached buffers obtained by via xfs_buf_get_uncached(), and remove
the XBF_FS_MANAGED special cases from the buffer cache.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
Each buffer contains both a buftarg pointer and a mount pointer. If
we add a mount pointer into the buftarg, we can avoid needing the
b_mount field in every buffer and grab it from the buftarg when
needed instead. This shrinks the xfs_buf by 8 bytes.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
To avoid the need to use cached buffers for single-shot or buffers
cached at the filesystem level, introduce a new buffer read
primitive that bypasses the cache an reads directly from disk.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
xfs_buf_get_nodaddr() is really used to allocate a buffer that is
uncached. While it is not directly assigned a disk address, the fact
that they are not cached is a more important distinction. With the
upcoming uncached buffer read primitive, we should be consistent
with this disctinction.
While there, make page allocation in xfs_buf_get_nodaddr() safe
against memory reclaim re-entrancy into the filesystem by allowing
a flags parameter to be passed.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
Add WQ_MEM_RECLAIM flag which currently maps to WQ_RESCUER, mark
WQ_RESCUER as internal and replace all external WQ_RESCUER usages to
WQ_MEM_RECLAIM.
This makes the API users express the intent of the workqueue instead
of indicating the internal mechanism used to guarantee forward
progress. This is also to make it cleaner to add more semantics to
WQ_MEM_RECLAIM. For example, if deemed necessary, memory reclaim
workqueues can be made highpri.
This patch doesn't introduce any functional change.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
The workqueue implementation in 2.6.36-rcX has changed, resulting
in the workqueues no longer having dedicated threads for work
processing. This has caused severe livelocks under heavy parallel
create workloads because the log IO completions have been getting
held up behind metadata IO completions. Hence log commits would
stall, memory allocation would stall because pages could not be
cleaned, and lock contention on the AIL during inode IO completion
processing was being seen to slow everything down even further.
By making the log Io completion workqueue a high priority workqueue,
they are queued ahead of all data/metadata IO completions and
processed before the data/metadata completions. Hence the log never
gets stalled, and operations needed to clean memory can continue as
quickly as possible. This avoids the livelock conditions and allos
the system to keep running under heavy load as per normal.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Switch to the WRITE_FLUSH_FUA flag for log writes and remove the EOPNOTSUPP
detection for barriers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
When doing large parallel file creates on a 16p machines, large amounts of
time is being spent in _xfs_buf_find(). A system wide profile with perf top
shows this:
1134740.00 19.3% _xfs_buf_find
733142.00 12.5% __ticket_spin_lock
The problem is that the hash contains 45,000 buffers, and the hash table width
is only 256 buffers. That means we've got around 200 buffers per chain, and
searching it is quite expensive. The hash table size needs to increase.
Secondly, every time we do a lookup, we promote the buffer we find to the head
of the hash chain. This is causing cachelines to be dirtied and causes
invalidation of cachelines across all CPUs that may have walked the hash chain
recently. hence every walk of the hash chain is effectively a cold cache walk.
Remove the promotion to avoid this invalidation.
The results are:
1045043.00 21.2% __ticket_spin_lock
326184.00 6.6% _xfs_buf_find
A 70% drop in the CPU usage when looking up buffers. Unfortunately that does
not result in an increase in performance underthis workload as contention on
the inode_lock soaks up most of the reduction in CPU usage.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
The b_strat callback is used by xfs_buf_iostrategy to perform additional
checks before submitting a buffer. It is used in xfs_bwrite and when
writing out delayed buffers. In xfs_bwrite it we can de-virtualize the
call easily as b_strat is set a few lines above the call to
xfs_buf_iostrategy. For the delayed buffers the rationale is a bit
more complicated:
- there are three callers of xfs_buf_delwri_queue, which places buffers
on the delwri list:
(1) xfs_bdwrite - this sets up b_strat, so it's fine
(2) xfs_buf_iorequest. None of the callers can have XBF_DELWRI set:
- xlog_bdstrat is only used for log buffers, which are never delwri
- _xfs_buf_read explicitly clears the delwri flag
- xfs_buf_iodone_work retries log buffers only
- xfsbdstrat - only used for reads, superblock writes without the
delwri flag, log I/O and file zeroing with explicitly allocated
buffers.
- xfs_buf_iostrategy - only calls xfs_buf_iorequest if b_strat is
not set
(3) xfs_buf_unlock
- only puts the buffer on the delwri list if the DELWRI flag is
already set. The DELWRI flag is only ever set in xfs_bwrite,
xfs_buf_iodone_callbacks, or xfs_trans_log_buf. For
xfs_buf_iodone_callbacks and xfs_trans_log_buf we require
an initialized buf item, which means b_strat was set to
xfs_bdstrat_cb in xfs_buf_item_init.
Conclusion: we can just get rid of the callback and replace it with
explicit calls to xfs_bdstrat_cb.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
xfs_buf_read() fails to detect dispatch errors before attempting to
wait on sychronous IO. If there was an error, it will get stuck
forever, waiting for an I/O that was never started. Make sure the
error is detected correctly.
Further, such a failure can leave locked pages in the page cache
which will cause a later operation to hang on the page. Ensure that
we correctly process pages in the buffers when we get a dispatch
error.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Get rid of the xfs_buf_pin/xfs_buf_unpin/xfs_buf_ispin helpers and opencode
them in their only callers, just like we did for the inode pinning a while
ago. Also remove duplicate trace points - the bufitem tracepoints cover
all the information that is present in a buffer tracepoint.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Dmapi support was never merged upstream, but we still have a lot of hooks
bloating XFS for it, all over the fast pathes of the filesystem.
This patch drops over 700 lines of dmapi overhead. If we'll ever get HSM
support in mainline at least the namespace events can be done much saner
in the VFS instead of the individual filesystem, so it's not like this
is much help for future work.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
The current shrinker implementation requires the registered callback
to have global state to work from. This makes it difficult to shrink
caches that are not global (e.g. per-filesystem caches). Pass the shrinker
structure to the callback so that users can embed the shrinker structure
in the context the shrinker needs to operate on and get back to it in the
callback via container_of().
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
xfs_bwrite is used with the intention of synchronously writing out
buffers, but currently it does not actually clear the async flag if
that's left from previous writes but instead implements async
behaviour if it finds it. Remove the code handling asynchronous
writes as we've got rid of those entirely outside of the log and
delwri buffers, and make sure that we clear the async and read flags
before writing the buffer.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
This allows to see in `ps` and similar tools which kthreads are
allotted to which block device/filesystem, similar to what jbd2
does. As the process name is a fixed 16-char array, no extra
space is needed in tasks.
PID TTY STAT TIME COMMAND
2 ? S 0:00 [kthreadd]
197 ? S 0:00 \_ [jbd2/sda2-8]
198 ? S 0:00 \_ [ext4-dio-unwrit]
204 ? S 0:00 \_ [flush-8:0]
2647 ? S 0:00 \_ [xfs_mru_cache]
2648 ? S 0:00 \_ [xfslogd/0]
2649 ? S 0:00 \_ [xfsdatad/0]
2650 ? S 0:00 \_ [xfsconvertd/0]
2651 ? S 0:00 \_ [xfsbufd/ram0]
2652 ? S 0:00 \_ [xfsaild/ram0]
2653 ? S 0:00 \_ [xfssyncd/ram0]
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Re-apply a commit that had been reverted due to regressions
that have since been fixed.
From 95f8e302c0 Mon Sep 17 00:00:00 2001
From: Nick Piggin <npiggin@suse.de>
Date: Tue, 6 Jan 2009 14:43:09 +1100
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>
Only modifications here were a minor reformat, plus making the patch
apply given the new use of xfs_buf_is_vmapped().
Modified-by: Alex Elder <aelder@sgi.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Re-apply a commit that had been reverted due to regressions
that have since been fixed.
Original commit: d2859751cd
Author: Nick Piggin <npiggin@suse.de>
Date: Tue, 6 Jan 2009 14:40:44 +1100
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>
The only change I made was to use the "new" xfs_buf_is_vmapped()
function in a place it had been open-coded in the original.
Modified-by: Alex Elder <aelder@sgi.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
* 'for-linus' of git://oss.sgi.com/xfs/xfs: (52 commits)
fs/xfs: Correct NULL test
xfs: optimize log flushing in xfs_fsync
xfs: only clear the suid bit once in xfs_write
xfs: kill xfs_bawrite
xfs: log changed inodes instead of writing them synchronously
xfs: remove invalid barrier optimization from xfs_fsync
xfs: kill the unused XFS_QMOPT_* flush flags V2
xfs: Use delay write promotion for dquot flushing
xfs: Sort delayed write buffers before dispatch
xfs: Don't issue buffer IO direct from AIL push V2
xfs: Use delayed write for inodes rather than async V2
xfs: Make inode reclaim states explicit
xfs: more reserved blocks fixups
xfs: turn off sign warnings
xfs: don't hold onto reserved blocks on remount,ro
xfs: quota limit statvfs available blocks
xfs: replace KM_LARGE with explicit vmalloc use
xfs: cleanup up xfs_log_force calling conventions
xfs: kill XLOG_VEC_SET_TYPE
xfs: remove duplicate buffer flags
...
xfs_buf.c includes what is essentially a hand rolled version of
blk_rq_map_kern(). In order to work properly with the vmalloc buffers
that xfs uses, this hand rolled routine must also implement the flushing
API for vmap/vmalloc areas.
[style updates from hch@lst.de]
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
There are no more users of this function left in the XFS code
now that we've switched everything to delayed write flushing.
Remove it.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Currently when the xfsbufd writes delayed write buffers, it pushes
them to disk in the order they come off the delayed write list. If
there are lots of buffers ѕpread widely over the disk, this results
in overwhelming the elevator sort queues in the block layer and we
end up losing the posibility of merging adjacent buffers to minimise
the number of IOs.
Use the new generic list_sort function to sort the delwri dispatch
queue before issue to ensure that the buffers are pushed in the most
friendly order possible to the lower layers.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
All buffers logged into the AIL are marked as delayed write.
When the AIL needs to push the buffer out, it issues an async write of the
buffer. This means that IO patterns are dependent on the order of
buffers in the AIL.
Instead of flushing the buffer, promote the buffer in the delayed
write list so that the next time the xfsbufd is run the buffer will
be flushed by the xfsbufd. Return the state to the xfsaild that the
buffer was promoted so that the xfsaild knows that it needs to cause
the xfsbufd to run to flush the buffers that were promoted.
Using the xfsbufd for issuing the IO allows us to dispatch all
buffer IO from the one queue. This means that we can make much more
enlightened decisions on what order to flush buffers to disk as
we don't have multiple places issuing IO. Optimisations to xfsbufd
will be in a future patch.
Version 2
- kill XFS_ITEM_FLUSHING as it is now unused.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
We use the KM_LARGE flag to make kmem_alloc and friends use vmalloc
if necessary. As we only need this for a few boot/mount time
allocations just switch to explicit vmalloc calls there.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Currently we define aliases for the buffer flags in various
namespaces, which only adds confusion. Remove all but the XBF_
flags to clean this up a bit.
Note that we still abuse XFS_B_ASYNC/XBF_ASYNC for some non-buffer
uses, but I'll clean that up later.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
xfs_buf_iomove() uses xfs_caddr_t as it's parameter types, but it doesn't
care about the signedness of the variables as it is just copying the
data. Change the prototype to use void * so that we don't get sign
warnings at call sites.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Move xfsbdstrat and xfs_bdstrat_cb from xfs_lrw.c and xfs_bioerror
and xfs_bioerror_relse from xfs_rw.c into xfs_buf.c. This also
means xfs_bioerror and xfs_bioerror_relse can be marked static now.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Fold XFS_bwrite into it's only caller, xfs_bwrite and move it into
xfs_buf.c instead of leaving it as a fairly large inline function.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Linus Torvalds
Dave Chinner
Jens Axboe
Tejun Heo
Christoph Hellwig
Christoph Hellwig
Thomas Gleixner
Dave Chinner
Jan Engelhardt
Alex Elder
James Bottomley