At some stage, a mutex was added to gfs2_glock_put() without
checking all its call sites. Two of them were called from
under a spinlock causing random delays at various points and
crashes.
Signed-off-by: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

A typo in the directory code was causing postmark to fail
somewhere in the allocation code, since it was unable to
find newly allocated directory leaf blocks under certain
circumstances.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

A small update to the journaling code to change the way that
the "extra" blocks are accounted for in the journal. These are
used at a rate of one per 503 metadata blocks or one per 251
journaled data blocks (or just one if the total number of journaled
blocks in the transaction is smaller). Since we are using them at
two different rates the old method of accounting for them no longer
works and we count them up as required.

Since the "per transaction" accounting can't handle this (there is no
fixed number of header blocks per transaction) we have to account for
it in the general journal code. We now require that each transaction
reserves more blocks than it actually needs to take account of the
possible extra blocks.

Also a final fix to dir.c to ensure that all ref counts are handled
correctly.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

The last patch missed some other instances of incorrect ref counting,
this fixes all of those too.
Signed-off-by: NSteven 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: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

Some interfaces have changed. In particular one of the posix
locking functions has changed prototype, along with the
address space operation invalidatepage and the block getting
callback to the direct IO function.

In addition add the splice file operations. These will need to
be updated to support AOP_TRUNCATED_PAGE before they will be
of much use to us.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This is designed as a fs independent way to set flags on a
particular inode. The values of the ioctl() and flags are
designed to be identical to the ext2/3 values. Assuming that
this plan is acceptable to people in general, the plan is to
then move other fs across to using the same set of #defines,
etc.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

In some cases 16 bit functions were being used rather than 32 bit
functions.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

A conditional had missing {} around the two following
statements. Now added.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

We didn't properly check that leaf splitting was allowed. We do
now.
Signed-off-by: NSteven 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: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

Look for "nodir" in the hostdata mount option which is used to
set the NODIR flag in dlm_new_lockspace().
Signed-off-by: NDavid Teigland <teigland@redhat.com>
Signed-off-by: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

The various flags on inodes will in future be set and queried
via the extended attributes interface, so this interface is no
longer required.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Due to a typo, the dir leaf split operation was (for the first
split in a directory) writing the new hash vaules at the
wrong offset. This is now fixed.

Also some other tidy ups are included:

 - We use GFS2's hash function for dentries (see ops_dentry.c) so that
   we don't have to keep recalculating the hash values.
 - A lot of common code is eliminated between the various directory
   lookup routines.
 - Better error checking on directory lookup (previously different
   routines checked for different errors)
 - The leaf split operation has a couple of redundant operations
   removed from it, so it should be faster.

There is still further scope for further clean ups in the directory
code, and readdir in particular could do with slimming down a bit.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

In order to separate out the filesystem's metadata from "normal"
files and directories, a new filesystem type has been created.
It is called gfs2meta and mounting it gives access to the files
that were previously under .gfs2_admin (well still are until mkfs
is altered, which is next on the adgenda).

Its not currently possible to mount both gfs2 and gfs2meta on the
same block device at the same time. A future patch will allow that
to happen.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Fix a bug I introduced earlier with a kfree() and usage of
a structure in the wrong order. Also try and get the counts
of the journaled data buffers "more correct". Still some work
to do in this area though.
Signed-off-by: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

For every filesystem operation where we need a transaction, we
now make one less memory allocation.
Signed-off-by: NSteven 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: NSteven Whitehouse <swhiteho@redhat.com>

As requested by:
Pavel Machek <pavel@suse.cz>

Pavel's other comments will be dealt with in later patches.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Requested by:
Prarit Bhargava <prarit@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Requested by:
Prarit Bhargava <prarit@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This issue was raised by:
Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

All printk calls now have KERN_ set where required and a couple of
kmalloc(), memset(.., 0, ...) calls changed to kzalloc().

This is in response to comments from:
Pekka Enberg <penberg@cs.helsinki.fi> and
Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Signed-off-by: NDavid Teigland <teigland@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Signed-off-by: NDavid Teigland <teigland@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This is the second of two patches removing support for range
locks from the DLM
Signed-off-by: NDavid Teigland <teigland@redhat.com>
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

One line which should have been deleted in the last patch
was missed.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Heinz had spotted that I'd forgotten to test in databuf_lo_add()
that the data buffer in question hadn't already been added to
the list. This was causing an infinite loop later on in the
"before commit" routine.

This means that GFS2 is now ready to be tested by everybody.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

As well as a number of minor bug fixes, this patch changes GFS
to use mutices rather than semaphores. This results in better
information in case there are any locking problems.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

There was a bug in the unstuffing logic which caused a crash
under certain circumstances. This is now fixed.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Two internal files which are read through the gfs2_internal_read()
routine were already locked when the routine was called and this
do not need locking at the redapages level.

This patch introduces a struct file which is used as a sentinal
so that readpage will only perform locking in the case that the
struct file passed to it is _not_ equal to this sentinal.

Since the comments in the generic kernel code indicate that the
struct file will never be used for anything other than passing
straight through to readpage(), this should be ok.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Various endianess changes required in the directory code.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

This patch adds back O_DIRECT support with various caveats
attached:

 1. Journaled data can be read via O_DIRECT since its now the
    same on disk format as normal data files.
 2. Journaled data writes with O_DIRECT will be failed sliently
    back to normal writes (should we really do this I wonder or
    should we return an error instead?)
 3. Stuffed files will be failed back to normal buffered I/O
 4. All the usual corner cases (write beyond current end of file,
    write to an unallocated block) will also revert to normal buffered I/O.

The I/O path is slightly odd as reads arrive at the page cache layer
with the lock for the file already held, but writes arrive unlocked.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

There were one or two fields in structures which didn't get changed
last time back to their gfs1 sizes and alignments. One or two constants
have also changed back to their original values which were missed the
first time.

Its possible that indirect pointer blocks might need to change. If
they don't we'll have to rewrite them all on upgrade due to a change
in the amount of padding that they use.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>

Umount is now working correctly again. The bug was due to
not getting an extra ref count when mounting the fs. We
should have bumped it by two (once for the internal pointer
to the root inode from the super block and once for the
inode hanging off the dcache entry for root).

Also this patch tidys up the code dealing with looking up
and creating inodes. We now pass Linux inodes (with gfs2_inodes
attached) rather than the other way around and this reduces code
duplication in various places.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>