VFS calls '->sync_fs()' twice - first time with @wait = 0, second
time with @wait = 1. As a result, we may commit and synchronize
write-buffers twice. Avoid doing this by returning immediatelly if
@wait = 0.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

WB_SYNC_HOLD is going to be zapped so we should not use it. Use
%WB_SYNC_NONE instead. Here is what akpm said:

"I think I'll just switch that to WB_SYNC_NONE.  The `wait==0' mode is
just an advisory thing to help the fs shove lots of data into the
queues.  If some gets missed then it'll be picked up on the second
->sync_fs call, with wait==1."

Thanks to Randy Dunlap for catching this.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

- (better, more, bigger ...) then -> (...) than
Signed-off-by: NFrederik Schwarzer <schwarzerf@gmail.com>
Signed-off-by: NJiri Kosina <jkosina@suse.cz>

With the write_begin/write_end aops, page_symlink was broken because it
could no longer pass a GFP_NOFS type mask into the point where the
allocations happened.  They are done in write_begin, which would always
assume that the filesystem can be entered from reclaim.  This bug could
cause filesystem deadlocks.

The funny thing with having a gfp_t mask there is that it doesn't really
allow the caller to arbitrarily tinker with the context in which it can be
called.  It couldn't ever be GFP_ATOMIC, for example, because it needs to
take the page lock.  The only thing any callers care about is __GFP_FS
anyway, so turn that into a single flag.

Add a new flag for write_begin, AOP_FLAG_NOFS.  Filesystems can now act on
this flag in their write_begin function.  Change __grab_cache_page to
accept a nofs argument as well, to honour that flag (while we're there,
change the name to grab_cache_page_write_begin which is more instructive
and does away with random leading underscores).

This is really a more flexible way to go in the end anyway -- if a
filesystem happens to want any extra allocations aside from the pagecache
ones in ints write_begin function, it may now use GFP_KERNEL (rather than
GFP_NOFS) for common case allocations (eg.  ocfs2_alloc_write_ctxt, for a
random example).

[kosaki.motohiro@jp.fujitsu.com: fix ubifs]
[kosaki.motohiro@jp.fujitsu.com: fix fuse]
Signed-off-by: NNick Piggin <npiggin@suse.de>
Reviewed-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org>		[2.6.28.x]
Signed-off-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
[ Cleaned up the calling convention: just pass in the AOP flags
  untouched to the grab_cache_page_write_begin() function.  That
  just simplifies everybody, and may even allow future expansion of the
  logic.   - Linus ]
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Print node sizes and maximum node sizes.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

We cannot use ubifs_err() macro with DBGKEY() and DBGKEY1(),
because this is racy and holding dbg_lock is needed. Use
dbg_err() instead, which does have the lock held.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

It is fine if there is not free space - we should still allow mounting
this FS. This patch relaxes the free space requirements and adds info
dumps.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

UBIFS does not disable compression if ui->flags is non-zero, e.g.
if the file has "sync" flag. This is because of the typo which
is fixed by this patch. The patch also adds a couple of useful
debugging prints.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

These are mostly long lines and wrong indentation warning
fixes. But also there are two volatile variables and
checkpatch.pl complains about them:

WARNING: Use of volatile is usually wrong: see Documentation/volatile-considered-harmful.txt
+       volatile int gc_seq;

WARNING: Use of volatile is usually wrong: see Documentation/volatile-considered-harmful.txt
+       volatile int gced_lnum;

Well, we anyway use smp_wmb() for c->gc_seq and c->gced_lnum, so
these 'volatile' modifiers can be just dropped.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

fs/ubifs/compress.c:111:8: warning: incorrect type in argument 5 (different signedness)
fs/ubifs/compress.c:111:8:    expected unsigned int *dlen
fs/ubifs/compress.c:111:8:    got int *out_len
fs/ubifs/compress.c:175:10: warning: incorrect type in argument 5 (different signedness)
fs/ubifs/compress.c:175:10:    expected unsigned int *dlen
fs/ubifs/compress.c:175:10:    got int *out_len

Fix this by adding a cast to (unsigned int *). We guarantee that
our lengths are small and no overflow is possible.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

The 'make_free_space()' function was too complex and this patch
simplifies it. It also fixes a bug - the freespace test failed
straight away on UBI volumes with 512 bytes LEB size.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Do not force UBIFS return 0 used space when it is empty. It leads
to a situation when creating any file immediately produces tens of
used blocks, which looks very weird. It is better to be honest and
say that some blocks are used even if the FS is empty. And ext2
does the same.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

UBIFS stores uncommitted index size in c->budg_uncommitted_idx,
and this affect budgeting calculations. When mounting and
replaying, this variable is not updated, so we may end up
with "over-budgeting". This patch fixes the issue.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

UBIFS commits on unmount to make the next mount faster. Currently,
it commits only if there is more than LEB size bytes in the
journal. This is not very good, because journal size may be
large (512KiB). And there may be few deletions in the journal
which do not take much journal space, but which do introduce
a lot of TNC changes and make mount slow.

Thus, jurt remove this condition and always commit.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

UBI now has (fake for now, though) synchronization call - use
it.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Always run commit in sync_fs, because even if the journal seems
to be almost empty, there may be a deletion which removes a large
file, which affects the index greatly. And because we want
better free space predictions after 'sync_fs()', we have to
commit.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Argh. The ->sync_fs call is called _before_ all inodes are flushed.
This means we first sync write buffers and commit, then all
inodes are synced, and we end up with unflushed write buffers!

Fix this by forcing synching all indoes from 'ubifs_sync_fs()'.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

The c->min_idx_lebs constant depends on c->old_idx_sz, which
is read from the master node. This means that we have to
initialize c->min_idx_lebs only after we have read the master
node.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Do not calculate min_idx_lebs, because it is available in
c->min_idx_lebs
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

When we commit, but before we try to write anything to the flash
media, @c->min_idx_size is inaccurate, because we do not re-calculate
it after the commit. Do not forget to do this.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Instead of using do_div(), use better primitives from
linux/math64.h.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Take into account that 2 eraseblocks are never available because
they are reserved for the index. This gives more realistic count
of FS blocks.

To avoid future confusions like this, introduce a constant.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Dump available space calculated by budgeting subsystem.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

debugfs tnc dumping was broken because of an obvious typo.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

It has high bits set, not low bits set as the UBIFS code
assumed.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

This patch fixes the following section mismatch:

WARNING: fs/ubifs/ubifs.o(.init.text+0xec): Section mismatch in reference from the function init_module() to the function .exit.text:ubifs_compressors_exit()
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

It is useful to know how we got to the checking function when
hunting the bugs.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Do not forget to check whether lpt debugging is enabled before
running the check functions. This commit also makes some spelling
fixes.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

We need to have a possibility to see various UBIFS variables
and ask UBIFS to dump various information. Debugfs is what
we need.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Introduce a new data structure which contains all debugging
stuff inside. This is cleaner than having debugging stuff
directly in 'c'.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

I have a habit of compiling kernel with
EXTRA_CFLAGS="-Wextra -Wno-unused -Wno-sign-compare -Wno-missing-field-initializers"
and so fs/ubifs/key.h give lots (~10) of these every time:

CC      fs/ubifs/tnc_misc.o
In file included from fs/ubifs/ubifs.h:1725,
from fs/ubifs/tnc_misc.c:30:
fs/ubifs/key.h: In function 'key_r5_hash':
fs/ubifs/key.h:64: warning: comparison of unsigned expression >= 0 is always true
fs/ubifs/key.h: In function 'key_test_hash':
fs/ubifs/key.h:81: warning: comparison of unsigned expression >= 0 is always true

This patch fixes the warnings.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

It is very handy to be able to change default UBIFS compressor
via mount options. Introduce -o compr=<name> mount option support.
Currently only "none", "lzo" and "zlib" compressors are supported.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Save a 4 bytes of RAM per 'struct inode' by stroring inode
compression type in bit-filed, instead of using 'int'.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

If data does not compress, it is better to leave it uncompressed
because we'll read it faster then. So do not compress data if we
save less than 64 bytes.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

To avoid memory allocation failure during bulk-read, pre-allocate
a bulk-read buffer, so that if there is only one bulk-reader at
a time, it would just use the pre-allocated buffer and would not
do any memory allocation. However, if there are more than 1 bulk-
reader, then only one reader would use the pre-allocated buffer,
while the other reader would allocate the buffer for itself.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Bulk-read allocates 128KiB or more using kmalloc. The allocation
starts failing often when the memory gets fragmented. UBIFS still
works fine in this case, because it falls-back to standard
(non-optimized) read method, though. This patch teaches bulk-read
to allocate exactly the amount of memory it needs, instead of
allocating 128KiB every time.

This patch is also a preparation to the further fix where we'll
have a pre-allocated bulk-read buffer as well. For example, now
the @bu object is prepared in 'ubifs_bulk_read()', so we could
path either pre-allocated or allocated information to
'ubifs_do_bulk_read()' later. Or teaching 'ubifs_do_bulk_read()'
not to allocate 'bu->buf' if it is already there.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>

Bulk-read allocates a lot of memory with 'kmalloc()', and when it
is/gets fragmented 'kmalloc()' fails with a scarry warning. But
because bulk-read is just an optimization, UBIFS keeps working fine.
Supress the warning by passing __GFP_NOWARN option to 'kmalloc()'.

This patch also introduces a macro for the magic 128KiB constant.
This is just neater.

Note, this is not really fixes the problem we had, but just hides
the warnings. The further patches fix the problem.
Signed-off-by: NArtem Bityutskiy <Artem.Bityutskiy@nokia.com>