xfs_iomap passes a xfs_bmbt_irec pointer to xfs_iomap_write_direct and
xfs_iomap_write_allocate to give them the results of our read-only
xfs_bmapi query. Instead of allocating a new xfs_bmbt_irec on stack
for the next call to xfs_bmapi re use the one we got passed as it's not
used after this point.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
We already rely on the fact that the sync code will cause a synchronous
log force later on (currently via xfs_fs_sync_fs -> xfs_quiesce_data ->
xfs_sync_data), so no need to do this here. This allows us to avoid
a lot of synchronous log forces during sync, which pays of especially
with delayed logging enabled. Some compilebench numbers that show
this:
xfs (delayed logging, 256k logbufs)
===================================
intial create 25.94 MB/s 25.75 MB/s 25.64 MB/s
create 8.54 MB/s 9.12 MB/s 9.15 MB/s
patch 2.47 MB/s 2.47 MB/s 3.17 MB/s
compile 29.65 MB/s 30.51 MB/s 27.33 MB/s
clean 90.92 MB/s 98.83 MB/s 128.87 MB/s
read tree 11.90 MB/s 11.84 MB/s 8.56 MB/s
read compiled 28.75 MB/s 29.96 MB/s 24.25 MB/s
delete tree 8.39 seconds 8.12 seconds 8.46 seconds
delete compiled 8.35 seconds 8.44 seconds 5.11 seconds
stat tree 6.03 seconds 5.59 seconds 5.19 seconds
stat compiled tree 9.00 seconds 9.52 seconds 8.49 seconds
xfs + write_inode log_force removal
===================================
intial create 25.87 MB/s 25.76 MB/s 25.87 MB/s
create 15.18 MB/s 14.80 MB/s 14.94 MB/s
patch 3.13 MB/s 3.14 MB/s 3.11 MB/s
compile 36.74 MB/s 37.17 MB/s 36.84 MB/s
clean 226.02 MB/s 222.58 MB/s 217.94 MB/s
read tree 15.14 MB/s 15.02 MB/s 15.14 MB/s
read compiled tree 29.30 MB/s 29.31 MB/s 29.32 MB/s
delete tree 6.22 seconds 6.14 seconds 6.15 seconds
delete compiled tree 5.75 seconds 5.92 seconds 5.81 seconds
stat tree 4.60 seconds 4.51 seconds 4.56 seconds
stat compiled tree 4.07 seconds 3.87 seconds 3.96 seconds
In addition to that also remove the delwri inode flush that is unessecary
now that bulkstat is always coherent.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
The writepage implementation in XFS still tries to deal with dirty but
unmapped buffers which used to caused by writes through shared mmaps. Since
the introduction of ->page_mkwrite these can't happen anymore, so remove the
code dealing with them.
Note that the all_bh variable which causes us to start I/O on all buffers on
the pages was controlled by the count of unmapped buffers, which also
included those not actually dirty. It's now unconditionally initialized to
0 but set to 1 for the case of small file size extensions. It probably can
be removed entirely, but that's left for another patch.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Currently the xfs releasepage implementation has code to deal with converting
delayed allocated and unwritten space. But we never get called for those as
we always convert delayed and unwritten space when cleaning a page, or drop
the state from the buffers in block_invalidatepage. We still keep a WARN_ON
on those cases for now, but remove all the case dealing with it, which allows
to fold xfs_page_state_convert into xfs_vm_writepage and remove the !startio
case from the whole writeback path.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
xfstests 194 first truncats a file back and then extends it again by
truncating it to a larger size. This causes discard_buffer to drop
the mapped, but not the uptodate bit and thus creates something that
xfs_page_state_convert takes for unmapped space created by mmap because
it doesn't check for the dirty bit, which also gets cleared by
discard_buffer and checked by other ->writepage implementations like
block_write_full_page. Handle this kind of buffers early, and unlike
Eric's first version of the patch simply ASSERT that the buffers is
dirty, given that the mmap write case can't happen anymore since the
introduction of ->page_mkwrite. The now dead code dealing with that
will be deleted in a follow on patch.
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
This code was introduced four years ago in commit
3e57ecf640 without any review and has
been unused since. Remove it just as the rest of the code introduced
in that commit to reduce that stack usage and complexity in this central
piece of code.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
By making this member a void pointer we can get rid of a lot of pointless
casts.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Currently we need to either call IHOLD or xfs_trans_ihold on an inode when
joining it to a transaction via xfs_trans_ijoin.
This patches instead makes xfs_trans_ijoin usable on it's own by doing
an implicity xfs_trans_ihold, which also allows us to drop the third
argument. For the case where we want to hold a reference on the inode
a xfs_trans_ijoin_ref wrapper is added which does the IHOLD and marks
the inode for needing an xfs_iput. In addition to the cleaner interface
to the caller this also simplifies the implementation.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.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>
Stop the function pointer casting madness and give all the li_cb instances
correct prototype.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Stop the function pointer casting madness and give all the xfs_item_ops the
correct prototypes.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
The unpin_remove item operation instances always share most of the
implementation with the respective unpin implementation. So instead
of keeping two different entry points add a remove flag to the unpin
operation and share the code more easily.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Currently we track log item descriptor belonging to a transaction using a
complex opencoded chunk allocator. This code has been there since day one
and seems to work around the lack of an efficient slab allocator.
This patch replaces it with dynamically allocated log item descriptors
from a dedicated slab pool, linked to the transaction by a linked list.
This allows to greatly simplify the log item descriptor tracking to the
point where it's just a couple hundred lines in xfs_trans.c instead of
a separate file. The external API has also been simplified while we're
at it - the xfs_trans_add_item and xfs_trans_del_item functions to add/
delete items from a transaction have been simplified to the bare minium,
and the xfs_trans_find_item function is replaced with a direct dereference
of the li_desc field. All debug code walking the list of log items in
a transaction is down to a simple list_for_each_entry.
Note that we could easily use a singly linked list here instead of the
double linked list from list.h as the fastpath only does deletion from
sequential traversal. But given that we don't have one available as
a library function yet I use the list.h functions for simplicity.
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>
https://bugzilla.kernel.org/show_bug.cgi?id=16348
When the filesystem grows to a large number of allocation groups,
the summing of recalimable inodes gets expensive. In many cases,
most AGs won't have any reclaimable inodes and so we are wasting CPU
time aggregating over these AGs. This is particularly important for
the inode shrinker that gets called frequently under memory
pressure.
To avoid the overhead, track AGs with reclaimable inodes in the
per-ag radix tree so that we can find all the AGs with reclaimable
inodes via a simple gang tag lookup. This involves setting the tag
when the first reclaimable inode is tracked in the AG, and removing
the tag when the last reclaimable inode is removed from the tree.
Then the summation process becomes a loop walking the radix tree
summing AGs with the reclaim tag set.
This significantly reduces the overhead of scanning - a 6400 AG
filesystea now only uses about 25% of a cpu in kswapd while slab
reclaim progresses instead of being permanently stuck at 100% CPU
and making little progress. Clean filesystems filesystems will see
no overhead and the overhead only increases linearly with the number
of dirty AGs.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Now the shrinker passes us a context, wire up a shrinker context per
filesystem. This allows us to remove the global mount list and the
locking problems that introduced. It also means that a shrinker call
does not need to traverse clean filesystems before finding a
filesystem with reclaimable inodes. This significantly reduces
scanning overhead when lots of filesystems are present.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
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>
The block number comes from bulkstat based inode lookups to shortcut
the mapping calculations. We ar enot able to trust anything from
bulkstat, so drop the block number as well so that the correct
lookups and mappings are always done.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Inode numbers may come from somewhere external to the filesystem
(e.g. file handles, bulkstat information) and so are inherently
untrusted. Rename the flag we use for these lookups to make it
obvious we are doing a lookup of an untrusted inode number and need
to verify it completely before trying to read it from disk.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
When we decode a handle or do a bulkstat lookup, we are using an
inode number we cannot trust to be valid. If we are deleting inode
chunks from disk (default noikeep mode), then we cannot trust the on
disk inode buffer for any given inode number to correctly reflect
whether the inode has been unlinked as the di_mode nor the
generation number may have been updated on disk.
This is due to the fact that when we delete an inode chunk, we do
not write the clusters back to disk when they are removed - instead
we mark them stale to avoid them being written back potentially over
the top of something that has been subsequently allocated at that
location. The result is that we can have locations of disk that look
like they contain valid inodes but in reality do not. Hence we
cannot simply convert the inode number to a block number and read
the location from disk to determine if the inode is valid or not.
As a result, and XFS_IGET_BULKSTAT lookup needs to actually look the
inode up in the inode allocation btree to determine if the inode
number is valid or not.
It should be noted even on ikeep filesystems, there is the
possibility that blocks on disk may look like valid inode clusters.
e.g. if there are filesystem images hosted on the filesystem. Hence
even for ikeep filesystems we really need to validate that the inode
number is valid before issuing the inode buffer read.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
The non-coherent bulkstat versionsthat look directly at the inode
buffers causes various problems with performance optimizations that
make increased use of just logging inodes. This patch makes bulkstat
always use iget, which should be fast enough for normal use with the
radix-tree based inode cache introduced a while ago.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
This patch prevents user "foo" from using the SWAPEXT ioctl to swap
a write-only file owned by user "bar" into a file owned by "foo" and
subsequently reading it. It does so by checking that the file
descriptors passed to the ioctl are also opened for reading.
Signed-off-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Now that the background flush code has been fixed, we shouldn't need to
silently multiply the wbc->nr_to_write to get good writeback. Remove
that code.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use rwsem_is_locked to make the assertations for shared locks work.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Allowing writeback from reclaim context causes massive problems with stack
overflows as we can call into the writeback code which tends to be a heavy
stack user both in the generic code and XFS from random contexts that
perform memory allocations.
Follow the example of btrfs (and in slightly different form ext4) and refuse
to write out data from reclaim context. This issue should really be handled
by the VM so that we can tune better for this case, but until we get it
sorted out there we have to hack around this in each filesystem with a
complex writeback path.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
When an inode cluster is freed, it needs to mark all inodes in memory as
XFS_ISTALE before marking the buffer as stale. This is eeded because the inodes
have a different life cycle to the buffer, and once the buffer is torn down
during transaction completion, we must ensure none of the inodes get written
back (which is what XFS_ISTALE does).
Unfortunately, xfs_ifree_cluster() has some bugs that lead to inodes not being
marked with XFS_ISTALE. This shows up when xfs_iflush() is called on these
inodes either during inode reclaim or tail pushing on the AIL. The buffer is
read back, but no longer contains inodes and so triggers assert failures and
shutdowns. This was reproducable with at run.dbench10 invocation from xfstests.
There are two main causes of xfs_ifree_cluster() failing. The first is simple -
it checks in-memory inodes it finds in the per-ag icache to see if they are
clean without holding the flush lock. if they are clean it skips them
completely. However, If an inode is flushed delwri, it will
appear clean, but is not guaranteed to be written back until the flush lock has
been dropped. Hence we may have raced on the clean check and the inode may
actually be dirty. Hence always mark inodes found in memory stale before we
check properly if they are clean.
The second is more complex, and makes the first problem easier to hit.
Basically the in-memory inode scan is done with full knowledge it can be racing
with inode flushing and AIl tail pushing, which means that inodes that it can't
get the flush lock on might not be attached to the buffer after then in-memory
inode scan due to IO completion occurring. This is actually documented in the
code as "needs better interlocking". i.e. this is a zero-day bug.
Effectively, the in-memory scan must be done while the inode buffer is locked
and Io cannot be issued on it while we do the in-memory inode scan. This
ensures that inodes we couldn't get the flush lock on are guaranteed to be
attached to the cluster buffer, so we can then catch all in-memory inodes and
mark them stale.
Now that the inode cluster buffer is locked before the in-memory scan is done,
there is no need for the two-phase update of the in-memory inodes, so simplify
the code into two loops and remove the allocation of the temporary buffer used
to hold locked inodes across the phases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
If a filesystem is mounted without the inode64 mount option we
should still be able to access inodes not fitting into 32 bits, just
not created new ones. For this to work we need to make sure the
inode cache radix tree is initialized for all allocation groups, not
just those we plan to allocate inodes from. This patch makes sure
we initialize the inode cache radix tree for all allocation groups,
and also cleans xfs_initialize_perag up a bit to separate the
inode32 logical from the general perag structure setup.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
The use of radix_tree_preload() only works if the radix tree was
initialised without the __GFP_WAIT flag. The per-ag tree uses
GFP_NOFS, so does not trigger allocation of new tree nodes from the
preloaded array. Hence it enters the allocator with a spinlock held
and triggers the might_sleep() warnings.
Reported-by; Chris Mason <chris.mason@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Add a mutex_unlock missing on the error path. The use of this lock
is balanced elsewhere in the file.
The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
expression E1;
@@
* mutex_lock(E1,...);
<+... when != E1
if (...) {
... when != E1
* return ...;
}
...+>
* mutex_unlock(E1,...);
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Alex Elder <aelder@sgi.com>
Use DECLARE_EVENT_CLASS, and save ~15K:
text data bss dec hex filename
171949 43028 48 215025 347f1 fs/xfs/linux-2.6/xfs_trace.o.orig
156521 43028 36 199585 30ba1 fs/xfs/linux-2.6/xfs_trace.o
No change in functionality.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Alex Elder <aelder@sgi.com>
The new xfsqa test 228 tries to preallocate more space than the
filesystem contains. it should fail, but instead triggers an assert
about lock flags. The failure is due to the size extension failing
in vmtruncate() due to rlimit being set. Check this before we start
the preallocation to avoid allocating space that will never be used.
Also the path through xfs_vn_allocate already holds the IO lock, so
it should not be present in the lock flags when the setattr fails.
Hence the assert needs to take this into account. This will prevent
other such callers from hitting this incorrect ASSERT.
(Fixed a reference to "newsize" to read "new_size". -Alex)
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Add suggested cleanups to commit 29db3370a1369541d58d692fbfb168b8a0bd7f41
from review that didn't end up being commited.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Instead of having small helper functions calling big macros do the
calculations for the log reservations directly in the functions.
These are mostly 1:1 from the macros execept that the macros kept
the quota calculations in their callers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Recent testers were slightly confused that a realtime mount failed
due to missing CONFIG_XFS_RT; we can make that a little more
obvious.
V2: drop the else as suggested by Christoph
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Many places in the xfs code return E2BIG when they really mean
EFBIG; trying to grow past 16T on a 32 bit machine, for example,
says "Argument list too long" rather than "File too large" which is
not particularly helpful.
Some of these don't make perfect sense as EFBIG either, but still
better than E2BIG IMHO.
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
With delayed logging, we can get inode allocation buffers in the
same transaction inode unlink buffers. We don't currently mark inode
allocation buffers in the log, so inode unlink buffers take
precedence over allocation buffers.
The result is that when they are combined into the same checkpoint,
only the unlinked inode chain fields are replayed, resulting in
uninitialised inode buffers being detected when the next inode
modification is replayed.
To fix this, we need to ensure that we do not set the inode buffer
flag in the buffer log item format flags if the inode allocation has
not already hit the log. To avoid requiring a change to log
recovery, we really need to make this a modification that relies
only on in-memory sate.
We can do this by checking during buffer log formatting (while the
CIL cannot be flushed) if we are still in the same sequence when we
commit the unlink transaction as the inode allocation transaction.
If we are, then we do not add the inode buffer flag to the buffer
log format item flags. This means the entire buffer will be
replayed, not just the unlinked fields. We do this while
CIL flusheѕ are locked out to ensure that we don't race with the
sequence numbers changing and hence fail to put the inode buffer
flag in the buffer format flags when we really need to.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
If we let the CIL grow without bound, it will grow large enough to violate
recovery constraints (must be at least one complete transaction in the log at
all times) or take forever to write out through the log buffers. Hence we need
a check during asynchronous transactions as to whether the CIL needs to be
pushed.
We track the amount of log space the CIL consumes, so it is relatively simple
to limit it on a pure size basis. Make the limit the minimum of just under half
the log size (recovery constraint) or 8MB of log space (which is an awful lot
of metadata).
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
If the filesystem is being shut down and the there is no log error,
the current code forces out the current log buffers. This code now needs
to push the CIL before it forces out the log buffers to acheive the same
result.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
The delayed logging code only changes in-memory structures and as
such can be enabled and disabled with a mount option. Add the mount
option and emit a warning that this is an experimental feature that
should not be used in production yet.
We also need infrastructure to track committed items that have not
yet been written to the log. This is what the Committed Item List
(CIL) is for.
The log item also needs to be extended to track the current log
vector, the associated memory buffer and it's location in the Commit
Item List. Extend the log item and log vector structures to enable
this tracking.
To maintain the current log format for transactions with delayed
logging, we need to introduce a checkpoint transaction and a context
for tracking each checkpoint from initiation to transaction
completion. This includes adding a log ticket for tracking space
log required/used by the context checkpoint.
To track all the changes we need an io vector array per log item,
rather than a single array for the entire transaction. Using the new
log vector structure for this requires two passes - the first to
allocate the log vector structures and chain them together, and the
second to fill them out. This log vector chain can then be passed
to the CIL for formatting, pinning and insertion into the CIL.
Formatting of the log vector chain is relatively simple - it's just
a loop over the iovecs on each log vector, but it is made slightly
more complex because we re-write the iovec after the copy to point
back at the memory buffer we just copied into.
This code also needs to pin log items. If the log item is not
already tracked in this checkpoint context, then it needs to be
pinned. Otherwise it is already pinned and we don't need to pin it
again.
The only other complexity is calculating the amount of new log space
the formatting has consumed. This needs to be accounted to the
transaction in progress, and the accounting is made more complex
becase we need also to steal space from it for log metadata in the
checkpoint transaction. Calculate all this at insert time and update
all the tickets, counters, etc correctly.
Once we've formatted all the log items in the transaction, attach
the busy extents to the checkpoint context so the busy extents live
until checkpoint completion and can be processed at that point in
time. Transactions can then be freed at this point in time.
Now we need to issue checkpoints - we are tracking the amount of log space
used by the items in the CIL, so we can trigger background checkpoints when the
space usage gets to a certain threshold. Otherwise, checkpoints need ot be
triggered when a log synchronisation point is reached - a log force event.
Because the log write code already handles chained log vectors, writing the
transaction is trivial, too. Construct a transaction header, add it
to the head of the chain and write it into the log, then issue a
commit record write. Then we can release the checkpoint log ticket
and attach the context to the log buffer so it can be called during
Io completion to complete the checkpoint.
We also need to allow for synchronising multiple in-flight
checkpoints. This is needed for two things - the first is to ensure
that checkpoint commit records appear in the log in the correct
sequence order (so they are replayed in the correct order). The
second is so that xfs_log_force_lsn() operates correctly and only
flushes and/or waits for the specific sequence it was provided with.
To do this we need a wait variable and a list tracking the
checkpoint commits in progress. We can walk this list and wait for
the checkpoints to change state or complete easily, an this provides
the necessary synchronisation for correct operation in both cases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
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>
