This patch is a performance improvement for GFS2 in a clustered
environment. It makes the glock hold time self-adjusting.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Change each shrinker's API by consolidating the existing parameters into
shrink_control struct. This will simplify any further features added w/o
touching each file of shrinker.
[akpm@linux-foundation.org: fix build]
[akpm@linux-foundation.org: fix warning]
[kosaki.motohiro@jp.fujitsu.com: fix up new shrinker API]
[akpm@linux-foundation.org: fix xfs warning]
[akpm@linux-foundation.org: update gfs2]
Signed-off-by: Ying Han <yinghan@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Pavel Emelyanov <xemul@openvz.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch fixes a race in the GFS2 glock state machine that may
result in lockups. The symptom is that all nodes but one will
hang, waiting for a particular glock. All the holder records
will have the "W" (Waiting) bit set. The other node will
typically have the glock stuck in Exclusive mode (EX) with no
holder records, but the dinode will be cached. In other words,
an entry with "I:" will appear in the glock dump for that glock,
but nothing else.
The race has to do with the glock "Pending Demote" bit, which
can be set, then immediately reset, thus losing the fact that
another node needs the glock. The sequence of events is:
1. Something schedules the glock workqueue (e.g. glock request from fs)
2. The glock workqueue gets to the point between the test of the reply pending
bit and the spin lock:
if (test_and_clear_bit(GLF_REPLY_PENDING, &gl->gl_flags)) {
finish_xmote(gl, gl->gl_reply);
drop_ref = 1;
}
down_read(&gfs2_umount_flush_sem); <---- i.e. here
spin_lock(&gl->gl_spin);
3. In comes (a) the reply to our EX lock request setting GLF_REPLY_PENDING and
(b) the demote request which sets GLF_PENDING_DEMOTE
4. The following test is executed:
if (test_and_clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) &&
gl->gl_state != LM_ST_UNLOCKED &&
gl->gl_demote_state != LM_ST_EXCLUSIVE) {
This resets the pending demote flag, and gl->gl_demote_state is not equal to
exclusive, however because the reply from the dlm arrived after we checked for
the GLF_REPLY_PENDING flag, gl->gl_state is still equal to unlocked, so
although we reset the GLF_PENDING_DEMOTE flag, we didn't then set the
GLF_DEMOTE flag or reinstate the GLF_PENDING_DEMOTE_FLAG.
The patch closes the timing window by only transitioning the
"Pending demote" bit to the "demote" flag once we know the
other conditions (not unlocked and not exclusive) are met.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw: (32 commits)
GFS2: Move all locking inside the inode creation function
GFS2: Clean up symlink creation
GFS2: Clean up mkdir
GFS2: Use UUID field in generic superblock
GFS2: Rename ops_inode.c to inode.c
GFS2: Inode.c is empty now, remove it
GFS2: Move final part of inode.c into super.c
GFS2: Move most of the remaining inode.c into ops_inode.c
GFS2: Move gfs2_refresh_inode() and friends into glops.c
GFS2: Remove gfs2_dinode_print() function
GFS2: When adding a new dir entry, inc link count if it is a subdir
GFS2: Make gfs2_dir_del update link count when required
GFS2: Don't use gfs2_change_nlink in link syscall
GFS2: Don't use a try lock when promoting to a higher mode
GFS2: Double check link count under glock
GFS2: Improve bug trap code in ->releasepage()
GFS2: Fix ail list traversal
GFS2: make sure fallocate bytes is a multiple of blksize
GFS2: Add an AIL writeback tracepoint
GFS2: Make writeback more responsive to system conditions
...
Previously we marked all locks being promoted to a higher mode
with the try flag to avoid any potential deadlocks issues. The
DLM is able to detect these and report them in way that GFS2 can
deal with them correctly. So we can just request the required mode
and wait for a response without needing to perform this check.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Now that the whole dcache_hash_bucket crap is gone, go all the way and
also remove the weird locking layering violations for locking the hash
buckets. Add hlist_bl_lock/unlock helpers to move the locking into the
list abstraction instead of requiring each caller to open code it.
After all allowing for the bit locks is the whole point of these helpers
over the plain hlist variant.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch adds writeback_control to writing back the AIL
list. This means that we can then take advantage of the
information we get in ->write_inode() in order to set off
some pre-emptive writeback.
In addition, the AIL code is cleaned up a bit to make it
a bit simpler to understand.
There is still more which can usefully be done in this area,
but this is a good start at least.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The GLF_LRU flag introduced in the previous patch can be
used to check if a glock is on the lru list when a new
holder is queued and if so remove it, without having first
to get the lru_lock.
The main purpose of this patch however is to optimise the
glocks left over when an inode at end of life is being
evicted. Previously such glocks were left with the GLF_LFLUSH
flag set, so that when reclaimed, each one required a log flush.
This patch resets the GLF_LFLUSH flag when there is nothing
left to flush thus preventing later log flushes as glocks are
reused or demoted.
In order to do this, we need to keep track of the number of
revokes which are outstanding, and also to clear the GLF_LFLUSH
bit after a log commit when only revokes have been processed.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This adds support for two new flags. One keeps track of whether
the glock is on the LRU list or not. The other isn't really a
flag as such, but an indication of whether the glock has an
attached object or not. This indication is reported without
any locking, which is ok since we do not dereference the object
pointer but merely report whether it is NULL or not.
Also, this fixes one place where a tracepoint was missing, which
was at the point we remove deallocated blocks from the journal.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Rather than allowing the glocks to be scheduled for possible
reclaim as soon as they have exited the journal, this patch
delays their entry to the list until the glocks in question
are no longer in use.
This means that we will rely on the vm for writeback of all
dirty data and metadata from now on. When glocks are added
to the lru list they should be freeable much faster since all
the I/O required to free them should have already been completed.
This should lead to much better I/O patterns under low memory
conditions.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
As per RCU glock patch review comments, don't use the _raw
version of this function here.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This is a small patch that optimizes multiple glock dequeue
operations. It changes the unlock order to be more efficient
and makes it easier for lock debugging tools to unravel. It
also eliminates the need for the temp variable x, although
that would likely be optimized out.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch fixes a race in deallocating glocks which was introduced
in the RCU glock patch. We need to ensure that the glock count is
kept correct even in the case that there is a race to add a new
glock into the hash table. Also, to avoid having to wait for an
RCU grace period, the glock counter can be decremented before
call_rcu() is called.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
There are two spellings in use for 'freeze' + 'able' - 'freezable' and
'freezeable'. The former is the more prominent one. The latter is
mostly used by workqueue and in a few other odd places. Unify the
spelling to 'freezable'.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Steven Whitehouse <swhiteho@redhat.com>
This has a number of advantages:
- Reduces contention on the hash table lock
- Makes the code smaller and simpler
- Should speed up glock dumps when under load
- Removes ref count changing in examine_bucket
- No longer need hash chain lock in glock_put() in common case
There are some further changes which this enables and which
we may do in the future. One is to look at using SLAB_RCU,
and another is to look at using a per-cpu counter for the
per-sb glock counter, since that is touched twice in the
lifetime of each glock (but only used at umount time).
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
We can only merge the fields into a bitfield if the locking
rules for them are the same. In this case gl_spin covers all
of the fields (write side) but a couple of them are used
with GLF_LOCK as the read side lock, which should be ok
since we know that the field in question won't be changing
at the time.
The gl_req setting has to be done earlier (in glock.c) in order
to place it under gl_spin. The gl_reply setting also has to be
brought under gl_spin in order to comply with the new rules.
This saves 4*sizeof(unsigned int) per glock.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Bob Peterson <rpeterso@redhat.com>
The DLM never returns -EAGAIN in response to dlm_lock(), and even
if it did, the test in gdlm_lock() was wrong anyway. Once that
test is removed, it is possible to greatly simplify this code
by simply using a "normal" error return code (0 for success).
We then no longer need the LM_OUT_ASYNC return code which can
be removed.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Using %pV reduces the number of printk calls and
eliminates any possible message interleaving from
other printk calls.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The WQ_RESCUER flag should only be used internally to the
workqueue implementation.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
This area of the code has always been a bit delicate due to the
subtleties of lock ordering. The problem is that for "normal"
alloc/dealloc, we always grab the inode locks first and the rgrp lock
later.
In order to ensure no races in looking up the unlinked, but still
allocated inodes, we need to hold the rgrp lock when we do the lookup,
which means that we can't take the inode glock.
The solution is to borrow the technique already used by NFS to solve
what is essentially the same problem (given an inode number, look up
the inode carefully, checking that it really is in the expected
state).
We cannot do that directly from the allocation code (lock ordering
again) so we give the job to the pre-existing delete workqueue and
carry on with the allocation as normal.
If we find there is no space, we do a journal flush (required anyway
if space from a deallocation is to be released) which should block
against the pending deallocations, so we should always get the space
back.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The tests further down the recovery function relating to
unlocking the journal need to be updated to match the
intial test. Also, a test in the umount code which was
surplus to requirements has been removed. Umounting
spectator mounts now works correctly, as expected.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The recovery workqueue can be freezable since
we want it to finish what it is doing if the system is to
be frozen (although why you'd want to freeze a cluster node
is beyond me since it will result in it being ejected from
the cluster). It does still make sense for single node
GFS2 filesystems though.
The glock workqueue will benefit from being able to run more
work items concurrently. A test running postmark shows
improved performance and multi-threaded workloads are likely
to benefit even more. It needs to be high priority because
the latency directly affects the latency of filesystem glock
operations.
The delete workqueue is similar to the recovery workqueue in
that it must not get blocked by memory allocations, and may
run for a long time.
Potentially other GFS2 threads might also be converted to
workqueues, but I'll leave that for a later patch.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Due to the design of the VFS, it is quite usual for operations on GFS2
to consist of a lookup (requiring a shared lock) followed by an
operation requiring an exclusive lock. If a remote node has cached an
exclusive lock, then it will receive two demote events in rapid succession
firstly for a shared lock and then to unlocked. The existing min hold time
code was triggering in this case, even if the node was otherwise idle
since the state change time was being updated by the initial demote.
This patch introduces logic to skip the min hold timer in the case that
a "double demote" of this kind has occurred. The min hold timer will
still be used in all other cases.
A new glock flag is introduced which is used to keep track of whether
there have been any newly queued holders since the last glock state
change. The min hold time is only applied if the flag is set.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Tested-by: Abhijith Das <adas@redhat.com>
This is a clean up of the code which deals with LM_FLAG_NOEXP
which aims to remove any possible race conditions by using
gl_spin to cover the gap between testing for the LM_FLAG_NOEXP
and the GL_FROZEN flag.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This reverts commit b7dc2df572.
The initial patch didn't quite work since it doesn't cover all
the possible routes by which the GLF_FROZEN flag might be set.
A revised fix is coming up in the next patch.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This looks like a big change, but in reality its only a single line of actual
code change, the rest is just moving a function to before its new caller.
The "try" flag for glocks is a rather subtle and delicate setting since it
requires that the state machine tries just hard enough to ensure that it has
a good chance of getting the requested lock, but no so hard that the
request can land up blocked behind another.
The patch adds in an additional check which will fail any queued try
locks if there is another request blocking the try lock request which
is not granted and compatible, nor in progress already. The check is made
only after all pending locks which may be granted have been granted.
I've checked this with the reproducer for the reported flock bug which
this is intended to fix, and it now passes.
Signed-off-by: Steven Whitehouse <swhiteho@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>
This patch fixes bugzilla bug #590878: GFS2: recovery stuck on
transaction lock. We set the frozen flag on the glock when we receive
a completion that cannot be delivered due to blocked locks. At that
point we check to see whether the first waiting holder has the noexp
flag set. If the noexp lock is queued later, then we need to unfreeze
the glock at that point in time, namely, in the glock work function.
This patch was originally written by Steve Whitehouse, but since
he's on holiday, I'm submitting it. It's been well tested with a
complex recovery test called revolver.
Signed-off-by: Steve Whitehouse <swhiteho@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch fixes a couple gfs2 problems with the reclaiming of
unlinked dinodes. First, there were a couple of livelocks where
everything would come to a halt waiting for a glock that was
seemingly held by a process that no longer existed. In fact, the
process did exist, it just had the wrong pid number in the holder
information. Second, there was a lock ordering problem between
inode locking and glock locking. Third, glock/inode contention
could sometimes cause inodes to be improperly marked invalid by
iget_failed.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
This patch changes glock numbers from printing in decimal to hex.
Since DLM prints corresponding resource IDs in hex, it makes debugging
easier.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
As a consequence of the previous patch, we can now remove the
loop which used to be required due to the circular dependency
between the inodes and glocks. Instead we can just invalidate
the inodes, and then clear up any glocks which are left.
Also we no longer need the rwsem since there is no longer any
danger of the inode invalidation calling back into the glock
code (and from there back into the inode code).
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Since the start of GFS2, an "extra" inode has been used to store
the metadata belonging to each inode. The only reason for using
this inode was to have an extra address space, the other fields
were unused. This means that the memory usage was rather inefficient.
The reason for keeping each inode's metadata in a separate address
space is that when glocks are requested on remote nodes, we need to
be able to efficiently locate the data and metadata which relating
to that glock (inode) in order to sync or sync and invalidate it
(depending on the remotely requested lock mode).
This patch adds a new type of glock, which has in addition to
its normal fields, has an address space. This applies to all
inode and rgrp glocks (but to no other glock types which remain
as before). As a result, we no longer need to have the second
inode.
This results in three major improvements:
1. A saving of approx 25% of memory used in caching inodes
2. A removal of the circular dependency between inodes and glocks
3. No confusion between "normal" and "metadata" inodes in super.c
Although the first of these is the more immediately apparent, the
second is just as important as it now enables a number of clean
ups at umount time. Those will be the subject of future patches.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Although all glocks are, by the time of the umount glock wait,
scheduled for demotion, some of them haven't made it far
enough through the process for the original set of waiting
code to wait for them.
This extends the ref count to the whole glock lifetime in order
to ensure that the waiting does catch all glocks. It does make
it a bit more invasive, but it seems the only sensible solution
at the moment.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch fixes some ref counting issues. Firstly by moving
the point at which we drop the ref count after a dlm lock
operation has completed we ensure that we never call
gfs2_glock_hold() on a lock with a zero ref count.
Secondly, by using atomic_dec_and_lock() in gfs2_glock_put()
we ensure that at no time will a glock with zero ref count
appear on the lru_list. That means that we can remove the
check for this in our shrinker (which was racy).
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
We need to be careful of the ordering between clearing the
GLF_LOCK bit and scheduling the workqueue.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
When a file is deleted from a gfs2 filesystem on one node, a dcache
entry for it may still exist on other nodes in the cluster. If this
happens, gfs2 will be unable to free this file on disk. Because of this,
it's possible to have a gfs2 filesystem with no files on it and no free
space. With this patch, when a node receives a callback notifying it
that the file is being deleted on another node, it schedules a new
workqueue thread to remove the file's dcache entry.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
GFS2 was placing far too many glocks on the reclaim list that were not good
candidates for freeing up from cache. These locks would sit there and
repeatedly get scanned to see if they could be reclaimed, wasting a lot
of time when there was memory pressure. This fix does more checks on the
locks to see if they are actually likely to be removable from cache.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
It is possible for gfs2_shrink_glock_memory() to check a glock for
demotion
that's in the process of being freed by gfs2_glock_put(). In this case,
gfs2_shrink_glock_memory() will acquire a new reference to this glock,
and
then try to free the glock itself when it drops the refernce. To solve
this, gfs2_shrink_glock_memory() just needs to check if the glock is in
the process of being freed, and if so skip it without ever unlocking the
lru_lock.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Acked-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch removes some of the special cases that the shrinker
was trying to deal with. As a result we leave fewer items on
the list and none at all which cannot be demoted. This makes
the list scanning more efficient and solves some issues seen
with large numbers of inodes.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch adds the ability to trace various aspects of the GFS2
filesystem. The trace points are divided into three groups,
glocks, logging and bmap. These points have been chosen because
they allow inspection of the major internal functions of GFS2
and they are also generic enough that they are unlikely to need
any major changes as the filesystem evolves.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch fixes a race condition where we can receive recovery
requests part way through processing a umount. This was causing
problems since the recovery thread had already gone away.
Looking in more detail at the recovery code, it was really trying
to implement a slight variation on a work queue, and that happens to
align nicely with the recently introduced slow-work subsystem. As a
result I've updated the code to use slow-work, rather than its own home
grown variety of work queue.
When using the wait_on_bit() function, I noticed that the wait function
that was supplied as an argument was appearing in the WCHAN field, so
I've updated the function names in order to produce more meaningful
output.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Depending on the ordering of events as we go around the
glock shrinker loop, it is possible to drop the ref count
of a glock incorrectly. It doesn't happen very often. This
patch corrects the got_ref variable, fixing the problem.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The rwsem, used only on umount, is in the wrong place in glock.c.
This patch moves it up a bit so that it does not get called under
a spinlock.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This adds a sysfs file called demote_rq to GFS2's
per filesystem directory. Its possible to use this
file to demote arbitrary glocks in exactly the same
way as if a request had come in from a remote node.
This is intended for testing issues relating to caching
of data under glocks. Despite that, the interface is
generic enough to send requests to any type of glock,
but be careful as its not always safe to send an
arbitrary message to an arbitrary glock. For that reason
and to prevent DoS, this interface is restricted to root
only.
The messages look like this:
<type>:<glocknumber> <mode>
Example:
echo -n "2:13324 EX" >/sys/fs/gfs2/unity:myfs/demote_rq
Which means "please demote inode glock (type 2) number 13324 so that
I can get an EX (exclusive) lock". The lock modes are those which
would normally be sent by a remote node in its callback so if you
want to unlock a glock, you use EX, to demote to shared, use SH or PR
(depending on whether you like GFS2 or DLM lock modes better!).
If the glock doesn't exist, you'll get -ENOENT returned. If the
arguments don't make sense, you'll get -EINVAL returned.
The plan is that this interface will be used in combination with
the blktrace patch which I recently posted for comments although
it is, of course, still useful in its own right.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch fixes a deadlock when the journal is flushed and there
are dirty inodes other than the one which caused the journal flush.
Originally the journal flushing code was trying to obtain the
transaction glock while running the flush code for an inode glock.
We no longer require the transaction glock at this point in time
since we know that any attempt to get the transaction glock from
another node will result in a journal flush. So if we are flushing
the journal, we can be sure that the transaction lock is still
cached from when the transaction was started.
By inlining a version of gfs2_trans_begin() (minus the bit which
gets the transaction glock) we can avoid the deadlock problems
caused if there is a demote request queued up on the transaction
glock.
In addition I've also moved the umount rwsem so that it covers
the glock workqueue, since it all demotions are done by this
workqueue now. That fixes a bug on umount which I came across
while fixing the original problem.
Reported-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Bob Peterson