This patch doesn't make any changes to the ordering of the various
operations related to glocking, but it does tidy up the calls to the
glops.c functions to make the structure more obvious.
The two functions: gfs2_glock_xmote_th() and gfs2_glock_drop_th() can be
made static within glock.c since they are called by every set of glock
operations. The xmote_th and drop_th glock operations are then made
conditional upon those two routines existing and called from the
previously mentioned functions in glock.c respectively.
Also it can be seen that the go_sync operation isn't needed since it can
easily be replaced by calls to xmote_bh and drop_bh respectively. This
results in no longer (confusingly) calling back into routines in glock.c
from glops.c and also reducing the glock operations by one member.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The "greedy" code was an attempt to retain glocks for a minimum length
of time when they relate to mmap()ed files. The current implementation
of this feature is not, however, ideal in that it required allocating
memory in order to do this and its overly complicated.
It also misses the mark by ignoring the other I/O operations which are
just as likely to suffer from the same problem. So the plan is to remove
this now and then add the functionality back as part of the glock state
machine at a later date (and thus take into account all the possible
users of this feature)
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Here is something I spotted (while looking for something entirely
different) the other day.
Rather than using a completion in each and every struct gfs2_holder,
this removes it in favour of hashed wait queues, thus saving a
considerable amount of memory both on the stack (where a number of
gfs2_holder structures are allocated) and in particular in the
gfs2_inode which has 8 gfs2_holder structures embedded within it.
As a result on x86_64 the gfs2_inode shrinks from 2488 bytes to
1912 bytes, a saving of 576 bytes per inode (no thats not a typo!).
In actual practice we get a much better result than that since
now that a gfs2_inode is under the 2048 byte barrier, we get two
per 4k slab page effectively halving the amount of memory required
to store gfs2_inodes.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This removes the extra filldir callback which gfs2 was using to
enclose an attempt at readahead for inodes during readdir. The
code was too complicated and also hurts performance badly in the
case that the getdents64/readdir call isn't being followed by
stat() and it wasn't even getting it right all the time when it
was.
As a result, on my test box an "ls" of a directory containing 250000
files fell from about 7mins (freshly mounted, so nothing cached) to
between about 15 to 25 seconds. When the directory content was cached,
the time taken fell from about 3mins to about 4 or 5 seconds.
Interestingly in the cached case, running "ls -l" once reduced the time
taken for subsequent runs of "ls" to about 6 secs even without this
patch. Now it turns out that there was a special case of glocks being
used for prefetching the metadata, but because of the timeouts for these
locks (set to 10 secs) the metadata was being timed out before it was
being used and this the prefetch code was constantly trying to prefetch
the same data over and over.
Calling "ls -l" meant that the inodes were brought into memory and once
the inodes are cached, the glocks are not disposed of until the inodes
are pushed out of the cache, thus extending the lifetime of the glocks,
and thus bringing down the time for subsequent runs of "ls"
considerably.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The go_sync callback took two flags, but one of them was set on every
call, so this patch removes once of the flags and makes the previously
conditional operations (on this flag), unconditional.
The go_inval callback took three flags, each of which was set on every
call to it. This patch removes the flags and makes the operations
unconditional, which makes the logic rather more obvious.
Two now unused flags are also removed from incore.h.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This shrinks the size of the gfs2_inode by 8 bytes by
replacing the version counter with a one bit valid/invalid
flag.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Annotated scalar fields, dropped unused ones. Note that
it's not at all obvious that we want to convert all of them
to host-endian...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Fix a bug in the directory reading code, where we might have dereferenced
a NULL pointer in case of OOM. Updated the directory code to use the new
& improved version of gfs2_meta_ra() which now returns the first block
that was being read. Previously it was releasing it requiring following
code to grab the block again at each point it was called.
Also turned off readahead on directory lookups since we are reading a
hash table, and therefore reading the entries in order is very
unlikely. Readahead is still used for all other calls to the
directory reading function (e.g. when growing the hash table).
Removed the DIO_START constant. Everywhere this was used, it was
used to unconditionally start i/o aside from a couple of places, so
I've removed it and made the couple of exceptions to this rule into
separate functions.
Also hunted through the other DIO flags and removed them as arguments
from functions which were always called with the same combination of
arguments.
Updated gfs2_meta_indirect_buffer to be a bit more efficient and
hopefully also be a bit easier to read.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Use atomic_t as the ref count in glocks rather than a kref.
This is another step towards using RCU for the glock hash.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This results in smaller list heads, so that we can have more chains
in the same amount of memory (twice as many). I've multiplied the
size of the table by four though - this is because we are saving
memory by not having one lock per chain any more. So we land up
using about the same amount of memory for the hash table as we
did before I started these changes, the difference being that we
now have four times as many hash chains.
The reason that I say "about the same amount of memory" is that the
actual amount now depends upon the NR_CPUS and some of the config
variables, so that its not exact and in some cases we do use more
memory. Eventually we might want to scale the hash table size
according to the size of physical ram as measured on module load.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
The existing implementation of this function in glock.c was not
very efficient as it relied upon keeping a cursor element upon the
hash chain in question and moving it along. This new version improves
upon this by using the current element as a cursor. This is possible
since we only look at the "next" element in the list after we've
taken the read_lock() subsequent to calling the examiner function.
Obviously we have to eventually drop the ref count that we are then
left with and we cannot do that while holding the read_lock, so we
do that next time we drop the lock. That means either just before
we examine another glock, or when the loop has terminated.
The new implementation has several advantages: it uses only a
read_lock() rather than a write_lock(), so it can run simnultaneously
with other code, it doesn't need a "plug" element, so that it removes
a test not only from this list iterator, but from all the other glock
list iterators too. So it makes things faster and smaller.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Make the number of locks used for hash chains in glock.c
proportional to NR_CPUS. Also move constants for the number
of hash chains into glock.c from incore.h since they are
not used outside of glock.c.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This splits the rwlocks guarding the hash chains of the glock hash
table into their own array. This will reduce memory usage in some
cases due to better alignment, although the real reason for doing it
is to allow the two tables to be different sizes in future (i.e.
the locks will be sized proportionally with the max number of CPUs
and the hash chains sized proportinally with the size of physical memory)
In order to allow this, the gl_bucket member of struct gfs2_glock has
now become gl_hash, so we record the hash rather than a pointer to the
bucket itself.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
As requested by Jan Engelhardt, this removes the typedefs in the
locking module interface and replaces them with void *. Also
since we are changing the interface, I've added a few consts
as well.
Cc: Jan Engelhardt <jengelh@linux01.gwdg.de>
Cc: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
There are several reasons why we want to do this:
- Firstly its large and thus we'll scale better with multiple
GFS2 fs mounted at the same time
- Secondly its easier to scale its size as required (thats a plan
for later patches)
- Thirdly, we can use kzalloc rather than vmalloc when allocating
the superblock (its now only 4888 bytes)
- Fourth its all part of my plan to eventually be able to use RCU
with the glock hash.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
As per Jan Engelhardt's comments, this should make all the headers
compile on their own by including and/or declaring structures
early.
Cc: Jan Engelhardt <jengelh@linux01.gwdg.de>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This makes all fixed size types have consistent names.
Cc: Jan Engelhardt <jengelh@linux01.gwdg.de>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
As per comments from Jan Engelhardt <jengelh@linux01.gwdg.de> this
updates the copyright message to say "version" in full rather than
"v.2". Also incore.h has been updated to remove forward structure
declarations which are not required.
The gfs2_quota_lvb structure has now had endianess annotations added
to it. Also quota.c has been updated so that we now store the
lvb data locally in endian independant format to avoid needing
a structure in host endianess too. As a result the endianess
conversions are done as required at various points and thus the
conversion routines in lvb.[ch] are no longer required. I've
moved the one remaining constant in lvb.h thats used into lm.h
and removed the unused lvb.[ch].
I have not changed the HIF_ constants. That is left to a later patch
which I hope will unify the gh_flags and gh_iflags fields of the
struct gfs2_holder.
Cc: Jan Engelhardt <jengelh@linux01.gwdg.de>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Remove the unused sync feature from glocks. This is currently done by
calling the required functions to sync pages/blocks directly so this
code isn't needed.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
For all the usual reasons of enforcing correctness and potentially
reducing code size, this patch makes the glock operations const.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch allows the simultaneous mounting of gfs2meta and gfs2
filesystems. A restriction however is that a gfs2meta fs may only be
mounted if its corresponding gfs2 filesystem is also mounted. Also, a
gfs2 filesystem cannot be unmounted before its gfs2meta filesystem.
Signed-off-by: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
I noticed the gfs2_scand seemed to be taking a lot of CPU,
so in order to cut that down a bit, here is a patch. Firstly
the type of a glock is a constant during its lifetime, so that
its possible to check this without needing locking. I've moved
the (common) case of testing for an inode glock outside of
the glmutex lock.
Also there was a mutex left over from when the glock cache was
master of the inode cache. That isn't required any more so I've
removed that too.
There is probably scope for further speed ups in the future
in this area.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Mmapped files were able to trigger a lock ordering bug. Private
maps do not need to take the glock so early on. Shared maps do
unfortunately, however we can get around that by adding a flag
into the flags for the struct gfs2_file. This only works because
we are taking an exclusive lock at this point, so we know that
nobody else can be racing with us.
Fixes Red Hat bugzilla: #201196
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
fs/gfs2/glock.c: In function ‘gfs2_holder_get’:
fs/gfs2/glock.c:439: warning: passing argument 2 of ‘set_bit’ from incompatible pointer type
fs/gfs2/glock.c: In function ‘rq_promote’:
fs/gfs2/glock.c:512: warning: passing argument 2 of ‘set_bit’ from incompatible pointer type
fs/gfs2/glock.c:526: warning: passing argument 2 of ‘set_bit’ from incompatible pointer type
...
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
This patch fixes the way we have been dealing with unlinked,
but still open files. It removes all limits (other than memory
for inodes, as per every other filesystem) on numbers of these
which we can support on GFS2. It also means that (like other
fs) its the responsibility of the last process to close the file
to deallocate the storage, rather than the person who did the
unlinking. Note that with GFS2, those two events might take place
on different nodes.
Also there are a number of other changes:
o We use the Linux inode subsystem as it was intended to be
used, wrt allocating GFS2 inodes
o The Linux inode cache is now the point which we use for
local enforcement of only holding one copy of the inode in
core at once (previous to this we used the glock layer).
o We no longer use the unlinked "special" file. We just ignore it
completely. This makes unlinking more efficient.
o We now use the 4th block allocation state. The previously unused
state is used to track unlinked but still open inodes.
o gfs2_inoded is no longer needed
o Several fields are now no longer needed (and removed) from the in
core struct gfs2_inode
o Several fields are no longer needed (and removed) from the in core
superblock
There are a number of future possible optimisations and clean ups
which have been made possible by this patch.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
We can reclaim some space by moving fields in some structures
in order to allow them to pack better on 64 bit architectures.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This adds some extra debugging to glock.c and changes
inode.c's deallocation code to call the debugging code
at a suitable moment. I'm chasing down a particular bug
to do with deallocation at the moment and the code can
go again once the bug is fixed.
Also this includes the first part of some changes to unify
the Linux struct inode and GFS2's struct gfs2_inode. This
transformation will happen in small parts over the next short
period.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This patch changes the last user of recursive locking so that
it no longer needs this feature and removes it from the glock
layer. This makes the glock code a lot simpler and easier to
understand. Its also a prerequsite to adding support for the
AOP_TRUNCATED_PAGE return code (or at least it is if you don't
want your brain to melt in the process)
I've left in a couple of checks just in case there is some place
else in the code which is still using this feature that I didn't
spot yet, but they can probably be removed long term.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
When allocating memory to sort directory entries, use vmalloc()
rather than kmalloc() since for larger directories, the required
size can easily be graeter than the 128k maximum of kmalloc().
Also adding the first steps towards getting the AOP_TRUNCATED_PAGE
return code get in the glock code by flagging all places where we
request a glock and we are holding a page lock.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This fixes a ref count bug that sometimes showed up a umount time
(causing it to hang) but it otherwise mostly harmless. At the same
time there are some clean ups including making the log operations
structures const, moving a memory allocation so that its not done
in the fast path of checking to see if there is an outstanding
transaction related to a particular glock.
Removes the sd_log_wrap varaible which was updated, but never actually
used anywhere. Updates the gfs2 ioctl() to run without the kernel lock
(which it never needed anyway). Removes the "invalidate inodes" loop
from GFS2's put_super routine. This is done in kill super anyway so
we don't need to do it here. The loop was also bogus in that if there
are any inodes "stuck" at this point its a bug and we need to know
about it rather than hide it by hanging forever.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This puts the finishing touches to the ioctl support and also
removes a couple of unused fields from GFS2's private per file
structure.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Update the debugging code in trans.c and at the same time improve
the debugging code for gfs2_holders. The new code should be pretty
fast during the normal case and provide just as much information
in case of errors (or more).
One small function from glock.c has moved to glock.h as a static inline so
that its return address won't get in the way of the debugging.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Replace the lock_for_trans()/lock_for_flush() functions with an rwsem.
In fact the sd_log_flush_lock becomes an rwsem (the write part of it)
and is extended slightly to cover everything that the lock_for_flush()
used to cover. The read part of the lock is instead of lock_for_trans().
This corrects the races in the original code and reduces the code size.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
This reduces the size of the directory code by about 3k and gets
readdir() to use the functions which were introduced in the previous
directory code update.
Two memory allocations are merged into one. Eliminates zeroing of some
buffers which were never used before they were initialised by
other data.
There is still scope for further improvement in the directory code.
On the logging side, a hand created mutex has been replaced by a
standard Linux mutex in the log allocation code.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
We no longer lookup ".gfs2_admin" in the root directory in order to
find it, but instead use the inode number given in the superblock.
Both the root directory and the admin directory are now looked up using
the same routine, so the redundant code is removed.
Also, there is no longer a reference to the root inode in the
GFS2 super block. When required this can be retreived via
sb->s_root->d_inode instead.
Assuming that we introduce a metadata filesystem type for GFS, then
this is a first step towards that goal.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
For every filesystem operation where we need a transaction, we
now make one less memory allocation.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
As suggested by Pekka Enberg <penberg@cs.helsinki.fi>.
The DIV_RU macro is renamed DIV_ROUND_UP and and moved to kernel.h
The other macros are gone from gfs2.h as (although not requested
by Pekka Enberg) are a number of included header file which are now
included individually. The inode number comparison function is
now an inline function.
The DT2IF and IF2DT may be addressed in a future patch.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
