This patch fixes the error path of function read_rindex_entry
so that it correctly gives up its glock reference in cases where
there is a race to re-read the rindex after gfs2_grow.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch adds a call to gfs2_rindex_update from function gfs2_blk2rgrpd
and removes calls to it that are made redundant by it. The problem is
that a gfs2_grow can add rgrps to the rindex, then put those rgrps into
use, thus rendering the rindex we read in at mount time incomplete.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Over time, we've slowly eliminated the use of sd_rindex_mutex.
Up to this point, it was only used in two places: function
gfs2_ri_total (which totals the file system size by reading
and parsing the rindex file) and function gfs2_rindex_update
which updates the rgrps in memory. Both of these functions have
the rindex glock to protect them, so the rindex is unnecessary.
Since gfs2_grow writes to the rindex via the meta_fs, the mutex
is in the wrong order according to the normal rules. This patch
eliminates the mutex entirely to avoid the problem.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch fixes an error path in function gfs2_rindex_update
that leaves the rindex mutex held.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

The FITRIM ioctl provides an alternative way to send discard requests to
the underlying device. Using the discard mount option results in every
freed block generating a discard request to the block device. This can
be slow, since many block devices can only process discard requests of
larger sizes, and also such operations can be time consuming.

Rather than using the discard mount option, FITRIM allows a sweep of the
filesystem on an occasional basis, and also to optionally avoid sending
down discard requests for smaller regions.

In GFS2 FITRIM will work at resource group granularity. There is a flag
for each resource group which keeps track of which resource groups have
been trimmed. This flag is reset whenever a deallocation occurs in the
resource group, and set whenever a successful FITRIM of that resource
group has taken place. This helps to reduce repeated discard requests
for the same block ranges, again improving performance.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This makes mount take slightly longer, but at the same time, the first
write to the filesystem will be faster too. It also means that if there
is a problem in the resource index, then we can refuse to mount rather
than having to try and report that when the first write occurs.

In addition, to avoid recursive locking, we hvae to take account of
instances when the rindex glock may already be held when we are
trying to update the rbtree of resource groups.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

In function gfs2_inplace_release it was trying to unlock a gfs2_holder
structure associated with a reservation, after said reservation was
freed. The problem is that the statements have the wrong order.
This patch corrects the order so that the reservation is freed after
the gfs2_holder is unlocked.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Clean up gfs2_alloc_blocks so that it takes the full extent length
rather than just the number of non-inode blocks as an argument. That
will only make a difference in the inode allocation case for now.

Also, this fixes the extent length handling around gfs2_alloc_extent() so
that multi block allocations will work again.

The rd_last_alloc block is set to the final block in the allocated
extent (as per the update to i_goal, but referenced to a different
start point).

This also removes the dinode argument to rgblk_search() which is no
longer used.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch separates the code pertaining to allocations into two
parts: quota-related information and block reservations.
This patch also moves all the block reservation structure allocations to
function gfs2_inplace_reserve to simplify the code, and moves
the frees to function gfs2_inplace_release.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch splits function rgblk_search into a function that finds
blocks to allocate (rgblk_search) and a function that assigns those
blocks (gfs2_alloc_extent).
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@rehat.com>

The trace point should take extlen and not *ndata as the
extent length.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch is a revision of the one I previously posted.
I tried to integrate all the suggestions Steve gave.
The purpose of the patch is to change function gfs2_alloc_block
(allocate either a dinode block or an extent of data blocks)
to a more generic gfs2_alloc_blocks function that can
allocate both a dinode _and_ an extent of data blocks in the
same call. This will ultimately help us create a multi-block
reservation scheme to reduce file fragmentation.

This patch moves more toward a generic multi-block allocator that
takes a pointer to the number of data blocks to allocate, plus whether
or not to allocate a dinode. In theory, it could be called to allocate
(1) a single dinode block, (2) a group of one or more data blocks, or
(3) a dinode plus several data blocks.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch removes the vestigial variable al_alloced from
the gfs2_alloc structure. This is another baby step toward
multi-block reservations.

My next planned step is to decouple the quota variables
from the gfs2_alloc structure so we can use a different
method for allocations.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

GFS2 functions gfs2_alloc_block and gfs2_alloc_di do basically
the same things, with a few exceptions. This patch combines
the two functions into a slightly more generic gfs2_alloc_block.
Having one centralized block allocation function will reduce
code redundancy and make it easier to implement multi-block
reservations to reduce file fragmentation in the future.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This upstream patch had what I believe is an unintended consequence:

http://git.kernel.org/?p=linux/kernel/git/steve/gfs2-3.0-nmw.git;a=commitdiff;h=beca42486749c1538a5ed58fe9dcc9f26d428c93

The patch changed function get_local_rgrp such that it ONLY
used TRY locks for RGRP searches. Prior to that patch, the code
used TRY locks during the first loop, and if that was unsuccessful,
it used normal blocking locks on subsequent searches. This patch
changes it back to the old way.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

The two variables being initialised in gfs2_inplace_reserve
to track the file & line number of the caller are never
used, so we might as well remove them.

If something does go wrong, then a stack trace is probably
more useful anyway.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Bob reported:

I found an off-by-one problem with how I coded this section:
It should be:

+ else if (blk >= cur->rd_data0 + cur->rd_data)

In fact, cur->rd_data0 + cur->rd_data is the start of the next
rgrp (the next ri_addr), so without the "=" check it can land on
the wrong rgrp.

In all normal cases, this won't be a problem: you're searching
for a block _within_ the rgrp, which will pass the test properly.
Where it gets into trouble is if you search the rgrps for the
block exactly equal to ri_addr.  I don't think anything in the
kernel does this, but I found a place in gfs2-utils gfs2_edit
where it does.  So I definitely need to fix it in libgfs2.  I'd
like to suggest we fix it in the kernel as well for the sake of
keeping the functions similar.

So this patch fixes the above mentioned off by one error as well
as removing the unused parent pointer.
Reported-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

The new goal block should be set to the end of the newly
allocated extent, not the start of it.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Each block which is deallocated, requires a call to gfs2_rlist_add()
and each of those calls was calling gfs2_blk2rgrpd() in order to
figure out which rgrp the block belonged in. This can be speeded up
by making use of the rgrp cached in the inode. We also reset this
cached rgrp in case the block has changed rgrp. This should provide
a big reduction in gfs2_blk2rgrpd() calls during deallocation.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Given that a resource group has been locked, there is no reason why
we should not be able to allocate as many blocks as are free. The
al_requested parameter should really be considered as a minimum
number of blocks to be available. Should this limit be overshot,
there are other mechanisms which will prevent over allocation.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This means that after the initial allocation for any inode, the
last used resource group is cached in the inode for future use.
This drastically reduces the number of lookups of resource
groups in the common case, and this the contention on that
data structure.

The allocation algorithm is the same as previously, except that we
always check to see if the goal block is within the cached rgrp
first before going to the rbtree to look one up.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Since we have ruled out supporting online filesystem shrink,
it is possible to make the resource group list append only
during the life of a super block. This gives several benefits:

Firstly, we only need to read new rindex elements as they are added
rather than needing to reread the whole rindex file each time one
element is added.

Secondly, the rindex glock can be held for much shorter periods of
time, and is completely removed from the fast path for allocations.
The lock is taken in shared mode only when updating the resource
groups when the first allocation occurs, and after a grow has
taken place.

Thirdly, this results in a reduction in code size, and everything
gets a lot simpler to understand in this area.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Here is an update of Bob's original rbtree patch which, in addition, also
resolves the rather strange ref counting that was being done relating to
the bitmap blocks.

Originally we had a dual system for journaling resource groups. The metadata
blocks were journaled and also the rgrp itself was added to a list. The reason
for adding the rgrp to the list in the journal was so that the "repolish
clones" code could be run to update the free space, and potentially send any
discard requests when the log was flushed. This was done by comparing the
"cloned" bitmap with what had been written back on disk during the transaction
commit.

Due to this, there was a requirement to hang on to the rgrps' bitmap buffers
until the journal had been flushed. For that reason, there was a rather
complicated set up in the ->go_lock ->go_unlock functions for rgrps involving
both a mutex and a spinlock (the ->sd_rindex_spin) to maintain a reference
count on the buffers.

However, the journal maintains a reference count on the buffers anyway, since
they are being journaled as metadata buffers. So by moving the code which deals
with the post-journal accounting for bitmap blocks to the metadata journaling
code, we can entirely dispense with the rather strange buffer ref counting
scheme and also the requirement to journal the rgrps.

The net result of all this is that the ->sd_rindex_spin is left to do exactly
one job, and that is to look after the rbtree or rgrps.

This patch is designed to be a stepping stone towards using RCU for the rbtree
of resource groups, however the reduction in the number of uses of the
->sd_rindex_spin is likely to have benefits for multi-threaded workloads,
anyway.

The patch retains ->go_lock and ->go_unlock for rgrps, however these maybe also
be removed in future in favour of calling the functions directly where required
in the code. That will allow locking of resource groups without needing to
actually read them in - something that could be useful in speeding up statfs.

In the mean time though it is valid to dereference ->bi_bh only when the rgrp
is locked. This is basically the same rule as before, modulo the references not
being valid until the following journal flush.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Cc: Benjamin Marzinski <bmarzins@redhat.com>

__gfs2_free_data and __gfs2_free_meta are almost identical, and
can be trivially combined.

[This is as per Eric's original patch minus gfs2_free_data() which had
 no callers left and plus the conversion of the bmap.c calls to these
 functions. All in all, a nice clean up]
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

The deallocation code for directories in GFS2 is largely divided into
two parts. The first part deallocates any directory leaf blocks and
marks the directory as being a regular file when that is complete. The
second stage was identical to deallocating regular files.

Regular files have their data blocks in a different
address space to directories, and thus what would have been normal data
blocks in a regular file (the hash table in a GFS2 directory) were
deallocated correctly. However, a reference to these blocks was left in the
journal (assuming of course that some previous activity had resulted in
those blocks being in the journal or ail list).

This patch uses the i_depth as a test of whether the inode is an
exhash directory (we cannot test the inode type as that has already
been changed to a regular file at this stage in deallocation)

The original issue was reported by Chris Hertel as an issue he encountered
running bonnie++
Reported-by: NChristopher R. Hertel <crh@samba.org>
Cc: Abhijith Das <adas@redhat.com>
Signed-off-by: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

On rare occasions we encounter gfs2 problems where an
invalid bitmap state transition is attempted.  For example,
trying to "unlink" a free block.  In these cases, there
is really no useful information logged to debug the problem.
This patch adds more debug details that should allow us to
more closely examine the problem and possibly solve it.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch fixes a deadlock in GFS2 where two processes are trying
to reclaim an unlinked dinode:
One holds the inode glock and calls gfs2_lookup_by_inum trying to look
up the inode, which it can't, due to I_FREEING.  The other has set
I_FREEING from vfs and is at the beginning of gfs2_delete_inode
waiting for the glock, which is held by the first.  The solution is to
add a new non_block parameter to the gfs2_iget function that causes it
to return -ENOENT if the inode is being freed.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch is a performance improvement to GFS2's dealloc code.
Rather than update the quota file and statfs file for every
single block that's stripped off in unlink function do_strip,
this patch keeps track and updates them once for every layer
that's stripped.  This is done entirely inside the existing
transaction, so there should be no risk of corruption.
The other functions that deallocate blocks will be unaffected
because they are using wrapper functions that do the same
thing that they do today.

I tested this code on my roth cluster by creating 200
files in a directory, each of which is 100MB, then on
four nodes, I simultaneously deleted the files, thus competing
for GFS2 resources (but different files).  The commands
I used were:

[root@roth-01]# time for i in `seq 1 4 200` ; do rm /mnt/gfs2/bigdir/gfs2.$i; done
[root@roth-02]# time for i in `seq 2 4 200` ; do rm /mnt/gfs2/bigdir/gfs2.$i; done
[root@roth-03]# time for i in `seq 3 4 200` ; do rm /mnt/gfs2/bigdir/gfs2.$i; done
[root@roth-05]# time for i in `seq 4 4 200` ; do rm /mnt/gfs2/bigdir/gfs2.$i; done

The performance increase was significant:

             roth-01     roth-02     roth-03     roth-05
             ---------   ---------   ---------   ---------
old: real    0m34.027    0m25.021s   0m23.906s   0m35.646s
new: real    0m22.379s   0m24.362s   0m24.133s   0m18.562s

Total time spent deleting:
old: 118.6s
new:  89.4

For this particular case, this showed a 25% performance increase for
GFS2 unlinks.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

When you do gfs2_grow it failed to take the very last
rgrp into account when adding up the new free space due
to an off-by-one error.  It was not reading the last
rgrp from the rindex because of a check for "<=" that
should have been "<".  Therefore, fsck.gfs2 was finding
(and fixing) an error with the system statfs file.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>

When you truncate the rindex file, you need to avoid calling gfs2_rindex_hold,
since you already hold it.  However, if you haven't already read in the
resource groups, you need to do that.
Signed-off-by: NBenjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

When GFS2 grew the filesystem, it was never rereading the rindex file during
the grow. This is necessary for large grows when the filesystem is almost full,
and GFS2 needs to use some of the space allocated earlier in the grow to
complete it.  Now, if GFS2 fails to reserve the necessary space and the rindex
file is not uptodate, it rereads it.  Also, the only difference between
gfs2_ri_update() and gfs2_ri_update_special() was that gfs2_ri_update_special()
didn't clear out the existing resource groups, since you knew that it was only
called when there were no resource groups.  Attempting to clear out the
resource groups when there are none takes almost no time, and rarely happens,
so I simply removed gfs2_ri_update_special().
Signed-off-by: NBenjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

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: NSteven Whitehouse <swhiteho@redhat.com>

This patch fixes a GFS2 problem whereby the first rename after a
mount can result in a file system consistency error being flagged
improperly and cause the file system to withdraw.  The problem is
that the rename code tries to run the rgrp list with function
gfs2_blk2rgrpd before the rgrp list is guaranteed to be read in
from disk.  The patch makes the rename function hold the rindex
glock (as the gfs2_unlink code does today) which reads in the rgrp
list if need be.  There were a total of three places in the rename
code that improperly referenced the rgrp list without the rindex
glock and this patch fixes all three.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch adds support for fallocate to gfs2.  Since the gfs2 does not support
uninitialized data blocks, it must write out zeros to all the blocks.  However,
since it does not need to lock any pages to read from, gfs2 can write out the
zero blocks much more efficiently.  On a moderately full filesystem, fallocate
works around 5 times faster on average.  The fallocate call also allows gfs2 to
add blocks to the file without changing the filesize, which will make it
possible for gfs2 to preallocate space for the rindex file, so that gfs2 can
grow a completely full filesystem.
Signed-off-by: NBenjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This adds a check to ensure that if we reach the block allocator
that we don't try and proceed if there is no alloc structure
hanging off the inode. This should only happen if there is a bug
in GFS2. The error return code is distinctive in order that it
will be easily spotted.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

With the update of the truncate code, ip->i_disksize and
inode->i_size are merely copies of each other. This means
we can remove ip->i_disksize and use inode->i_size exclusively
reducing the size of a GFS2 inode by 8 bytes.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

All the blkdev_issue_* helpers can only sanely be used for synchronous
caller.  To issue cache flushes or barriers asynchronously the caller needs
to set up a bio by itself with a completion callback to move the asynchronous
state machine ahead.  So drop the BLKDEV_IFL_WAIT flag that is always
specified when calling blkdev_issue_* and also remove the now unused flags
argument to blkdev_issue_flush and blkdev_issue_zeroout.  For
blkdev_issue_discard we need to keep it for the secure discard flag, which
gains a more descriptive name and loses the bitops vs flag confusion.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJens Axboe <jaxboe@fusionio.com>

Switch to the WRITE_FLUSH_FUA flag for log writes, remove the EOPNOTSUPP
detection for barriers and stop setting the barrier flag for discards.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NSteven Whitehouse <swhiteho@redhat.com>
Acked-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <jaxboe@fusionio.com>

The previous patch I wrote for reclaiming unlinked dinodes
had some shortcomings and did not prevent all hangs.
This version is much cleaner and more logical, and has
passed very difficult testing.  Sorry for the churn.
Signed-off-by: NBob Peterson <rpeterso@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>