This is based on commit d1f5273e
ext4: return 32/64-bit dir name hash according to usage type
by Fan Yong <yong.fan@whamcloud.com>

Traditionally ext2/3/4 has returned a 32-bit hash value from llseek()
to appease NFSv2, which can only handle a 32-bit cookie for seekdir()
and telldir().  However, this causes problems if there are 32-bit hash
collisions, since the NFSv2 server can get stuck resending the same
entries from the directory repeatedly.

Allow ext3 to return a full 64-bit hash (both major and minor) for
telldir to decrease the chance of hash collisions.

This patch does implement a new ext3_dir_llseek op, because with 64-bit
hashes, nfs will attempt to seek to a hash "offset" which is much
larger than ext3's s_maxbytes.  So for dx dirs, we call
generic_file_llseek_size() with the appropriate max hash value as the
maximum seekable size.  Otherwise we just pass through to
generic_file_llseek().
Patch-updated-by: NBernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Patch-updated-by: NEric Sandeen <sandeen@redhat.com>
(blame us if something is not correct)
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NJan Kara <jack@suse.cz>

So far i_mutex was ranking above dqonoff_mutex and i_mutex on quota files
was special and ranking below dqonoff_mutex (and several other locks).
However there's no real need for i_mutex on quota files to be special.
IO on quota files is serialized by dqio_mutex anyway so we don't need to
take i_mutex when writing to quota files. Other places where we take i_mutex
on quota file can accomodate standard i_mutex lock ranking, we only need
to change the lock ranking to be dqonoff_mutex > i_mutex which is a matter
of changing documentation because there's no place which would enforce
ordering in the other direction.
Signed-off-by: NJan Kara <jack@suse.cz>

Signed-off-by: NJan Kara <jack@suse.cz>

We don't need i_mutex in ext2_quota_write() because writes to quota file
are serialized by dqio_mutex anyway. Changes to quota files outside of quota
code are forbidded and enforced by NOATIME and IMMUTABLE bits.
Signed-off-by: NJan Kara <jack@suse.cz>

We don't need i_mutex in reiserfs_quota_write() because writes to quota file
are serialized by dqio_mutex anyway. Changes to quota files outside of quota
code are forbidded and enforced by NOATIME and IMMUTABLE bits.
Signed-off-by: NJan Kara <jack@suse.cz>

We don't need i_mutex in ext4_quota_write() because writes to quota file
are serialized by dqio_mutex anyway. Changes to quota files outside of quota
code are forbidded and enforced by NOATIME and IMMUTABLE bits.
Signed-off-by: NJan Kara <jack@suse.cz>

We don't need i_mutex in ext3_quota_write() because writes to quota file
are serialized by dqio_mutex anyway. Changes to quota files outside of quota
code are forbidded and enforced by NOATIME and IMMUTABLE bits.
Signed-off-by: NJan Kara <jack@suse.cz>

When CONFIG_QUOTA_DEBUG is enabled we call inode_get_rsv_space() from
add_dquot_ref() while holding i_lock. But inode_get_rsv_space() is trying
to get i_lock as well resulting in double lock.

Fix the problem by moving inode_get_rsv_space() call out of i_lock.
Reported-and-analyzed-by: NJie Liu <jeff.liu@oracle.com>
Signed-off-by: NJan Kara <jack@suse.cz>

If journal superblock is written only in disk's caches and other transaction
starts reusing space of the transaction cleaned from the log, it can happen
blocks of a new transaction reach the disk before journal superblock. When
power failure happens in such case, subsequent journal replay would still try
to replay the old transaction but some of it's blocks may be already
overwritten by the new transaction. For this reason we must use WRITE_FUA when
updating log tail and we must first write new log tail to disk and update
in-memory information only after that.
Signed-off-by: NJan Kara <jack@suse.cz>

There are some log tail updates that are not protected by j_checkpoint_mutex.
Some of these are harmless because they happen during startup or shutdown but
updates in journal_commit_transaction() and journal_flush() can really race
with other log tail updates (e.g. someone doing journal_flush() with someone
running cleanup_journal_tail()). So protect all log tail updates with
j_checkpoint_mutex.
Signed-off-by: NJan Kara <jack@suse.cz>

There are three case of updating journal superblock. In the first case, we want
to mark journal as empty (setting s_sequence to 0), in the second case we want
to update log tail, in the third case we want to update s_errno. Split these
cases into separate functions. It makes the code slightly more straightforward
and later patches will make the distinction even more important.
Signed-off-by: NJan Kara <jack@suse.cz>

Jan Kara removed 'sb->s_dirt' VFS flag references, so we do not need to
register the ext2 'ext2_write_super()' method in the VFS superblock operations,
because 'sb->s_dirt' won't be ever set to 1 and VFS won't ever call
'->write_super()' anyway. Thus, remove the method.

Tested using xfstests.
Signed-off-by: NArtem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: NJan Kara <jack@suse.cz>

Places which modify superblock feature / state fields mark the superblock
buffer dirty so it is written out by flusher thread. Thus there's no need to
set s_dirt there.

The only other fields changing in the superblock are the numbers of free
blocks, free inodes and s_wtime. There's no real need to write (or even
compute) these periodically. Free blocks / inodes counters are recomputed on
every mount from group counters anyway and value of s_wtime is only
informational and imprecise anyway. So it should be enough to write these
opportunistically on mount, remount, umount, and sync_fs times.
Signed-off-by: NJan Kara <jack@suse.cz>

Currently on unmount if we are mounted R/W, we first write the superblock to
the media if it is dirty, and then write it again, which is not optimal. This
patch makes ext2 write the superblock on unmount less times.
Signed-off-by: NArtem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: NJan Kara <jack@suse.cz>

max_debt, involved variables and calculations
are no longer needed, clean them up.
Signed-off-by: NAkira Fujita <a-fujita@rs.jp.nec.com>
Signed-off-by: NJan Kara <jack@suse.cz>

Currently we write out all journal buffers in WRITE_SYNC mode. This improves
performance for fsync heavy workloads but hinders performance when writes
are mostly asynchronous, most noticably it slows down readers and users
complain about slow desktop response etc.

So submit writes as asynchronous in the normal case and only submit writes as
WRITE_SYNC if we detect someone is waiting for current transaction commit.

I've gathered some numbers to back this change. The first is the read latency
test. It measures time to read 1 MB after several seconds of sleeping in
presence of streaming writes.

Top 10 times (out of 90) in us:
Before		After
2131586		697473
1709932		557487
1564598		535642
1480462		347573
1478579		323153
1408496		222181
1388960		181273
1329565		181070
1252486		172832
1223265		172278

Average:
619377		82180

So the improvement in both maximum and average latency is massive.

I've measured fsync throughput by:
fs_mark -n 100 -t 1 -s 16384 -d /mnt/fsync/ -S 1 -L 4

in presence of streaming reader. The numbers (fsyncs/s) are:
Before		After
9.9		6.3
6.8		6.0
6.3		6.2
5.8		6.1

So fsync performance seems unharmed by this change.
Signed-off-by: NJan Kara <jack@suse.cz>

I have a new optimized x86 "strncpy_from_user()" that will use these
same helper functions for all the same reasons the name lookup code uses
them.  This is preparation for that.

This moves them into an architecture-specific header file.  It's
architecture-specific for two reasons:

 - some of the functions are likely to want architecture-specific
   implementations.  Even if the current code happens to be "generic" in
   the sense that it should work on any little-endian machine, it's
   likely that the "multiply by a big constant and shift" implementation
   is less than optimal for an architecture that has a guaranteed fast
   bit count instruction, for example.

 - I expect that if architectures like sparc want to start playing
   around with this, we'll need to abstract out a few more details (in
   particular the actual unaligned accesses).  So we're likely to have
   more architecture-specific stuff if non-x86 architectures start using
   this.

   (and if it turns out that non-x86 architectures don't start using
   this, then having it in an architecture-specific header is still the
   right thing to do, of course)
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

commit 2f533844 (tcp: allow splice() to build full TSO packets) added
a regression for splice() calls using SPLICE_F_MORE.

We need to call tcp_flush() at the end of the last page processed in
tcp_sendpages(), or else transmits can be deferred and future sends
stall.

Add a new internal flag, MSG_SENDPAGE_NOTLAST, acting like MSG_MORE, but
with different semantic.

For all sendpage() providers, its a transparent change. Only
sock_sendpage() and tcp_sendpages() can differentiate the two different
flags provided by pipe_to_sendpage()
Reported-by: NTom Herbert <therbert@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: H.K. Jerry Chu <hkchu@google.com>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail&gt;com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Many users of debugfs copy the implementation of default_open() when
they want to support a custom read/write function op.  This leads to a
proliferation of the default_open() implementation across the entire
tree.

Now that the common implementation has been consolidated into libfs we
can replace all the users of this function with simple_open().

This replacement was done with the following semantic patch:

<smpl>
@ open @
identifier open_f != simple_open;
identifier i, f;
@@
-int open_f(struct inode *i, struct file *f)
-{
(
-if (i->i_private)
-f->private_data = i->i_private;
|
-f->private_data = i->i_private;
)
-return 0;
-}

@ has_open depends on open @
identifier fops;
identifier open.open_f;
@@
struct file_operations fops = {
...
-.open = open_f,
+.open = simple_open,
...
};
</smpl>

[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: NStephen Boyd <sboyd@codeaurora.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Acked-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

debugfs and a few other drivers use an open-coded version of
simple_open() to pass a pointer from the file to the read/write file
ops.  Add support for this simple case to libfs so that we can remove
the many duplicate copies of this simple function.
Signed-off-by: NStephen Boyd <sboyd@codeaurora.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

It was introduced by d1d5e05f ("hugetlbfs: return error code when
initializing module") but as Al pointed out, is a bad idea.

Quoted comments from Al:
 "Note that unregister_filesystem() in module init is *always* wrong;
  it's not an issue here (it's done too early to care about and
  realistically the box is not going anywhere - it'll panic when attempt
  to exec /sbin/init fails, if not earlier), but it's a damn bad
  example.

  Consider a normal fs module.  Somebody loads it and in parallel with
  that we get a mount attempt on that fs type.  It comes between
  register and failure exits that causes unregister; at that point we
  are screwed since grabbing a reference to module as done by mount is
  enough to prevent exit, but not to prevent the failure of init.  As
  the result, module will get freed when init fails, mounted fs of that
  type be damned."

So remove it.
Signed-off-by: NHillf Danton <dhillf@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This allocation can be as large as 64k.

 - Add __GFP_NOWARN so the a falied kmalloc() is silent

 - Fall back to vmalloc() if the kmalloc() failed

Cc: Dave Chinner <david@fromorbit.com>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: David Rientjes <rientjes@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This allocation can be as large as 64k.  As David points out, "falling
back to vmalloc here is much better solution than failing to retreive
the attribute - it will work no matter how fragmented memory gets.  That
means we don't get incomplete backups occurring after days or months of
uptime and successful backups".

Cc: Dave Chinner <david@fromorbit.com>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: David Rientjes <rientjes@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This size is user controllable, up to a maximum of XATTR_LIST_MAX (64k).
So it's trivial for someone to trigger a stream of order:4 page
allocation errors.
Signed-off-by: NDave Jones <davej@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Chinner <david@fromorbit.com>
Acked-by: NDavid Rientjes <rientjes@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The proc_parse_options() call from proc_mount() runs only once at boot
time.  So on any later mount attempt, any mount options are ignored
because ->s_root is already initialized.

As a consequence, "mount -o <options>" will ignore the options.  The
only way to change mount options is "mount -o remount,<options>".

To fix this, parse the mount options unconditionally.
Signed-off-by: NVasiliy Kulikov <segoon@openwall.com>
Reported-by: NArkadiusz Miskiewicz <a.miskiewicz@gmail.com>
Tested-by: NArkadiusz Miskiewicz <a.miskiewicz@gmail.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The code cleanup of cifs_parse_mount_options resulted in a new bug being
introduced in the parsing of the UNC. This results in vol->UNC being
modified before vol->UNC was allocated.
Reported-by: NSteve French <smfrench@gmail.com>
Signed-off-by: NSachin Prabhu <sprabhu@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

The password parser has an unnecessary check for a NULL value which
triggers warnings in source checking tools. The code contains artifacts
from the old parsing code which are no longer required.
Signed-off-by: NSachin Prabhu <sprabhu@redhat.com>
Reviewed-by: NJeff Layton <jlayton@redhat.com>
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

We can deadlock if we have a write oplock and two processes
use the same file handle. In this case the first process can't
unlock its lock if the second process blocked on the lock in the
same time.

Fix it by using posix_lock_file rather than posix_lock_file_wait
under cinode->lock_mutex. If we request a blocking lock and
posix_lock_file indicates that there is another lock that prevents
us, wait untill that lock is released and restart our call.

Cc: stable@kernel.org
Acked-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NPavel Shilovsky <piastry@etersoft.ru>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

Revert previous version of patch to incorporate feedback
so that we can merge version 3 of the patch instead.w

This reverts commit b5efb978.

"s6->sin6_scope_id" is an int bits but strict_strtoul() writes a long
so this can corrupt memory on 64 bit systems.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

gcc-4.7.0 has started throwing these warnings when building cifs.ko.

  CC [M]  fs/cifs/cifssmb.o
fs/cifs/cifssmb.c: In function ‘CIFSSMBSetCIFSACL’:
fs/cifs/cifssmb.c:3905:9: warning: array subscript is above array bounds [-Warray-bounds]
fs/cifs/cifssmb.c: In function ‘CIFSSMBSetFileInfo’:
fs/cifs/cifssmb.c:5711:8: warning: array subscript is above array bounds [-Warray-bounds]
fs/cifs/cifssmb.c: In function ‘CIFSSMBUnixSetFileInfo’:
fs/cifs/cifssmb.c:6001:25: warning: array subscript is above array bounds [-Warray-bounds]

This patch cleans up the code a bit by using the offsetof macro instead
of the funky "&pSMB->hdr.Protocol" construct.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

We can deadlock if we have a write oplock and two processes
use the same file handle. In this case the first process can't
unlock its lock if another process blocked on the lock in the
same time.

Fix this by removing lock_mutex protection from waiting on a
blocked lock and protect only posix_lock_file call.

Cc: stable@kernel.org
Signed-off-by: NPavel Shilovsky <piastry@etersoft.ru>
Acked-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NSteve French <sfrench@us.ibm.com>

64252c75 "vfs: remove dget() from
dentry_unhash()" changed the implementation but not the comment.

Cc: Sage Weil <sage@newdream.net>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Split __lookup_hash into two component functions:

 lookup_dcache - tries cached lookup, returns whether real lookup is needed
 lookup_real - calls i_op->lookup

This eliminates code duplication between d_alloc_and_lookup() and
d_inode_lookup().
Signed-off-by: NMiklos Szeredi <mszeredi@suse.cz>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

now we have __lookup_hash() open-coded if !dentry case;
just call the damn thing instead...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Reorder if-else cases for starters...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Everything arriving into if (!dentry) will have need_reval = 1.
Indeed, the only way to get there with need_reval reset to 0 would
be via
	if (unlikely(d_need_lookup(dentry)))
		goto unlazy;
	if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE)) {
		status = d_revalidate(dentry, nd);
	if (unlikely(status <= 0)) {
		if (status != -ECHILD)
			need_reval = 0;
		goto unlazy;
...
unlazy:
	/* no assignments to dentry */
	if (dentry && unlikely(d_need_lookup(dentry))) {
		dput(dentry);
		dentry = NULL;
	}
and if d_need_lookup() had already been false the first time around, it
will remain false on the second call as well.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>