8 years agoBtrfs: clean up search_extent_mapping()
Li Zefan [Thu, 14 Jul 2011 03:18:03 +0000]
Btrfs: clean up search_extent_mapping()

rb_node returned by __tree_search() can be a valid pointer or NULL,
but won't be some errno.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: remove redundant code for dir item lookup
Li Zefan [Thu, 14 Jul 2011 03:17:52 +0000]
Btrfs: remove redundant code for dir item lookup

When we search a dir item with a specific hash code, we can
just return NULL without further checking if btrfs_search_slot()
returns 1.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: make acl functions really no-op if acl is not enabled
Li Zefan [Thu, 14 Jul 2011 03:17:39 +0000]
Btrfs: make acl functions really no-op if acl is not enabled

So there's no overhead for something we don't use.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: remove remaining ref-cache code
Li Zefan [Thu, 14 Jul 2011 03:17:27 +0000]
Btrfs: remove remaining ref-cache code

Since commit f2a97a9dbd86eb1ef956bdf20e05c507b32beb96
("btrfs: remove all unused functions"), there's no extern functions
at all in ref-cache.c, so just remove the remaining dead code.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: remove a BUG_ON() in btrfs_commit_transaction()
Li Zefan [Thu, 14 Jul 2011 03:17:14 +0000]
Btrfs: remove a BUG_ON() in btrfs_commit_transaction()

wait_for_commit() always returns 0.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: use wait_event()
Li Zefan [Thu, 14 Jul 2011 03:17:00 +0000]
Btrfs: use wait_event()

Use wait_event() when possible to avoid code duplication.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: check the nodatasum flag when writing compressed files
Li Zefan [Thu, 14 Jul 2011 03:16:47 +0000]
Btrfs: check the nodatasum flag when writing compressed files

If mounting with nodatasum option, we won't csum file data for
general write or direct-io write, and this rule should also be
applied when writing compressed files.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: copy string correctly in INO_LOOKUP ioctl
Li Zefan [Thu, 14 Jul 2011 03:16:00 +0000]
Btrfs: copy string correctly in INO_LOOKUP ioctl

Memory areas [ptr, ptr+total_len] and [name, name+total_len]
may overlap, so it's wrong to use memcpy().

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: don't print the leaf if we had an error
Josef Bacik [Wed, 13 Jul 2011 15:03:50 +0000]
Btrfs: don't print the leaf if we had an error

In __btrfs_free_extent we will print the leaf if we fail to find the extent we
wanted, but the problem is if we get an error we won't have a leaf so often this
leads to a NULL pointer dereference and we lose the error that actually
occurred.  So only print the leaf if ret > 0, which means we didn't find the
item we were looking for but we didn't error either.  This way the error is
preserved.

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agobtrfs: make btrfs_set_root_node void
Mark Fasheh [Thu, 14 Jul 2011 21:23:06 +0000]
btrfs: make btrfs_set_root_node void

This is fairly trivial - btrfs_set_root_node() - always returns zero so we
can just make it void.  All callers ignore the return code now anyway.  I
also made sure to check that none of the functions that
btrfs_set_root_node() calls returns an error that we might have needed to
catch and pass back.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: fix oops while writing data to SSD partitions
liubo [Wed, 27 Jul 2011 09:49:18 +0000]
Btrfs: fix oops while writing data to SSD partitions

Here I have a two SSD-partitions btrfs, and they are defaultly set to
"data=raid0, metadata=raid1", then I try to fill my btrfs partition
till "No space left on device", via "dd if=/dev/zero of=/mnt/btrfs/tmp".

I get an oops panic from kernel BUG at fs/btrfs/extent-tree.c:5199!, which
refers to find_free_extent's
BUG_ON(index != get_block_group_index(block_group));

In SSD mode, in order to find enough space to alloc, we may check the
block_group cache which has been checked sometime before, but the index is not
updated, where it hits the BUG_ON.

Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Acked-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: Protect the readonly flag of block group
WuBo [Tue, 26 Jul 2011 03:30:11 +0000]
Btrfs: Protect the readonly flag of block group

The access for ro in btrfs_block_group_cache should be protected
because of the racy lock in relocation.

Signed-off-by: Wu Bo <wu.bo@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agobtrfs: Make extent-io callbacks that never fail return void
Jeff Mahoney [Thu, 21 Jul 2011 16:56:09 +0000]
btrfs: Make extent-io callbacks that never fail return void

The set/clear bit and the extent split/merge hooks only ever return 0.

 Changing them to return void simplifies the error handling cases later.

 This patch changes the hook prototypes, the single implementation of each,
 and the functions that call them to return void instead.

 Since all four of these hooks execute under a spinlock, they're necessarily
 simple.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: fix readahead in file defrag
Li Zefan [Wed, 20 Jul 2011 03:46:35 +0000]
Btrfs: fix readahead in file defrag

We passed the wrong value to btrfs_force_ra(). Fix this by changing
the argument of btrfs_force_ra() from last_index to nr_page.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: return error to caller when btrfs_unlink() failes
Tsutomu Itoh [Tue, 19 Jul 2011 07:27:20 +0000]
Btrfs: return error to caller when btrfs_unlink() failes

When btrfs_unlink_inode() and btrfs_orphan_add() in btrfs_unlink()
are error, the error code is returned to the caller instead of
BUG_ON().

Signed-off-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs:don't check the return value of __btrfs_add_inode_defrag
Wanlong Gao [Mon, 18 Jul 2011 12:19:35 +0000]
Btrfs:don't check the return value of __btrfs_add_inode_defrag

Don't need to check the return value of __btrfs_add_inode_defrag(),
since it will always return 0.

Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoMerge branch 'alloc_path' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh...
Chris Mason [Mon, 1 Aug 2011 18:27:34 +0000]
Merge branch 'alloc_path' of git://git./linux/kernel/git/mfasheh/btrfs-error-handling into for-linus

8 years agoMerge branch 'integration' into for-linus
Chris Mason [Wed, 27 Jul 2011 20:13:10 +0000]
Merge branch 'integration' into for-linus

8 years agoBtrfs: make sure reserve_metadata_bytes doesn't leak out strange errors
Chris Mason [Wed, 27 Jul 2011 19:57:44 +0000]
Btrfs: make sure reserve_metadata_bytes doesn't leak out strange errors

The btrfs transaction code will return any errors that come from
reserve_metadata_bytes.  We need to make sure we don't return funny
things like 1 or EAGAIN.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: use the commit_root for reading free_space_inode crcs
Chris Mason [Tue, 26 Jul 2011 19:35:09 +0000]
Btrfs: use the commit_root for reading free_space_inode crcs

Now that we are using regular file crcs for the free space cache,
we can deadlock if we try to read the free_space_inode while we are
updating the crc tree.

This commit fixes things by using the commit_root to read the crcs.  This is
safe because we the free space cache file would already be loaded if
that block group had been changed in the current transaction.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: reduce extent_state lock contention for metadata
Chris Mason [Mon, 25 Jul 2011 10:50:50 +0000]
Btrfs: reduce extent_state lock contention for metadata

For metadata buffers that don't straddle pages (all of them), btrfs
can safely use the page uptodate bits and extent_buffer uptodate bit
instead of needing to use the extent_state tree.

This greatly reduces contention on the state tree lock.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: remove lockdep magic from btrfs_next_leaf
Chris Mason [Tue, 26 Jul 2011 20:01:59 +0000]
Btrfs: remove lockdep magic from btrfs_next_leaf

Before the reader/writer locks, btrfs_next_leaf needed to keep
the path blocking to avoid making lockdep upset.

Now that btrfs_next_leaf only takes read locks, this isn't required.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: make a lockdep class for each root
Chris Mason [Tue, 26 Jul 2011 20:11:19 +0000]
Btrfs: make a lockdep class for each root

This patch was originally from Tejun Heo.  lockdep complains about the btrfs
locking because we sometimes take btree locks from two different trees at the
same time.  The current classes are based only on level in the btree, which
isn't enough information for lockdep to figure out if the lock is safe.

This patch makes a class for each type of tree, and lumps all the FS trees that
actually have files and directories into the same class.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: switch the btrfs tree locks to reader/writer
Chris Mason [Sat, 16 Jul 2011 19:23:14 +0000]
Btrfs: switch the btrfs tree locks to reader/writer

The btrfs metadata btree is the source of significant
lock contention, especially in the root node.   This
commit changes our locking to use a reader/writer
lock.

The lock is built on top of rw spinlocks, and it
extends the lock tracking to remember if we have a
read lock or a write lock when we go to blocking.  Atomics
count the number of blocking readers or writers at any
given time.

It removes all of the adaptive spinning from the old code
and uses only the spinning/blocking hints inside of btrfs
to decide when it should continue spinning.

In read heavy workloads this is dramatically faster.  In write
heavy workloads we're still faster because of less contention
on the root node lock.

We suffer slightly in dbench because we schedule more often
during write locks, but all other benchmarks so far are improved.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: fix deadlock when throttling transactions
Josef Bacik [Sun, 24 Jul 2011 19:45:34 +0000]
Btrfs: fix deadlock when throttling transactions

Hit this nice little deadlock.  What happens is this

__btrfs_end_transaction with throttle set, --use_count so it equals 0
  btrfs_commit_transaction
    <somebody else actually manages to start the commit>
    btrfs_end_transaction --use_count so now its -1 <== BAD
      we just return and wait on the transaction

This is bad because we just return after our use_count is -1 and don't let go
of our num_writer count on the transaction, so the guy committing the
transaction just sits there forever.  Fix this by inc'ing our use_count if we're
going to call commit_transaction so that if we call btrfs_end_transaction it's
valid.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: stop using highmem for extent_buffers
Chris Mason [Tue, 19 Jul 2011 16:04:14 +0000]
Btrfs: stop using highmem for extent_buffers

The extent_buffers have a very complex interface where
we use HIGHMEM for metadata and try to cache a kmap mapping
to access the memory.

The next commit adds reader/writer locks, and concurrent use
of this kmap cache would make it even more complex.

This commit drops the ability to use HIGHMEM with extent buffers,
and rips out all of the related code.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: fix BUG_ON() caused by ENOSPC when relocating space
Miao Xie [Fri, 15 Jul 2011 10:34:36 +0000]
Btrfs: fix BUG_ON() caused by ENOSPC when relocating space

When we balanced the chunks across the devices, BUG_ON() in
__finish_chunk_alloc() was triggered.

------------[ cut here ]------------
kernel BUG at fs/btrfs/volumes.c:2568!
[SNIP]
Call Trace:
 [<ffffffffa049525e>] btrfs_alloc_chunk+0x8e/0xa0 [btrfs]
 [<ffffffffa04546b0>] do_chunk_alloc+0x330/0x3a0 [btrfs]
 [<ffffffffa045c654>] btrfs_reserve_extent+0xb4/0x1f0 [btrfs]
 [<ffffffffa045c86b>] btrfs_alloc_free_block+0xdb/0x350 [btrfs]
 [<ffffffffa048a8d8>] ? read_extent_buffer+0xd8/0x1d0 [btrfs]
 [<ffffffffa04476fd>] __btrfs_cow_block+0x14d/0x5e0 [btrfs]
 [<ffffffffa044660d>] ? read_block_for_search+0x14d/0x4d0 [btrfs]
 [<ffffffffa0447c9b>] btrfs_cow_block+0x10b/0x240 [btrfs]
 [<ffffffffa044dd5e>] btrfs_search_slot+0x49e/0x7a0 [btrfs]
 [<ffffffffa044f07d>] btrfs_insert_empty_items+0x8d/0xf0 [btrfs]
 [<ffffffffa045e973>] insert_with_overflow+0x43/0x110 [btrfs]
 [<ffffffffa045eb0d>] btrfs_insert_dir_item+0xcd/0x1f0 [btrfs]
 [<ffffffffa0489bd0>] ? map_extent_buffer+0xb0/0xc0 [btrfs]
 [<ffffffff812276ad>] ? rb_insert_color+0x9d/0x160
 [<ffffffffa046cc40>] ? inode_tree_add+0xf0/0x150 [btrfs]
 [<ffffffffa0474801>] btrfs_add_link+0xc1/0x1c0 [btrfs]
 [<ffffffff811dacac>] ? security_inode_init_security+0x1c/0x30
 [<ffffffffa04a28aa>] ? btrfs_init_acl+0x4a/0x180 [btrfs]
 [<ffffffffa047492f>] btrfs_add_nondir+0x2f/0x70 [btrfs]
 [<ffffffffa046af16>] ? btrfs_init_inode_security+0x46/0x60 [btrfs]
 [<ffffffffa0474ac0>] btrfs_create+0x150/0x1d0 [btrfs]
 [<ffffffff81159c63>] ? generic_permission+0x23/0xb0
 [<ffffffff8115b415>] vfs_create+0xa5/0xc0
 [<ffffffff8115ce6e>] do_last+0x5fe/0x880
 [<ffffffff8115dc0d>] path_openat+0xcd/0x3d0
 [<ffffffff8115e029>] do_filp_open+0x49/0xa0
 [<ffffffff8116a965>] ? alloc_fd+0x95/0x160
 [<ffffffff8114f0c7>] do_sys_open+0x107/0x1e0
 [<ffffffff810bcc3f>] ? audit_syscall_entry+0x1bf/0x1f0
 [<ffffffff8114f1e0>] sys_open+0x20/0x30
 [<ffffffff81484ec2>] system_call_fastpath+0x16/0x1b
[SNIP]
RIP  [<ffffffffa049444a>] __finish_chunk_alloc+0x20a/0x220 [btrfs]

The reason is:
Task1 Space balance task
do_chunk_alloc()
  __finish_chunk_alloc()
    update device info
    in the chunk tree
      alloc system metadata block
relocate system metadata block group
  set system metadata block group
  readonly, This block group is the
  only one that can allocate space. So
  there is no free space that can be
  allocated now.
        find no space and don't try
        to alloc new chunk, and then
        return ENOSPC
  BUG_ON() in __finish_chunk_alloc()
  was triggered.

Fix this bug by allocating a new system metadata chunk before relocating the
old one if we find there is no free space which can be allocated after setting
the old block group to be read-only.

Reported-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: tag pages for writeback in sync
Josef Bacik [Fri, 15 Jul 2011 21:26:38 +0000]
Btrfs: tag pages for writeback in sync

Everybody else does this, we need to do it too.  If we're syncing, we need to
tag the pages we're going to write for writeback so we don't end up writing the
same stuff over and over again if somebody is constantly redirtying our file.
This will keep us from having latencies with heavy sync workloads.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: fix enospc problems with delalloc
Josef Bacik [Fri, 15 Jul 2011 15:16:44 +0000]
Btrfs: fix enospc problems with delalloc

So I had this brilliant idea to use atomic counters for outstanding and reserved
extents, but this turned out to be a bad idea.  Consider this where we have 1
outstanding extent and 1 reserved extent

Reserver Releaser
atomic_dec(outstanding) now 0
atomic_read(outstanding)+1 get 1
atomic_read(reserved) get 1
don't actually reserve anything because
they are the same
atomic_cmpxchg(reserved, 1, 0)
atomic_inc(outstanding)
atomic_add(0, reserved)
free reserved space for 1 extent

Then the reserver now has no actual space reserved for it, and when it goes to
finish the ordered IO it won't have enough space to do it's allocation and you
get those lovely warnings.

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: don't flush delalloc arbitrarily
Josef Bacik [Fri, 15 Jul 2011 16:01:03 +0000]
Btrfs: don't flush delalloc arbitrarily

Kill the check to see if we have 512mb of reserved space in delalloc and
shrink_delalloc if we do.  This causes unexpected latencies and we have other
logic to see if we need to throttle.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

8 years agoBtrfs: use find_or_create_page instead of grab_cache_page
Josef Bacik [Mon, 11 Jul 2011 14:47:06 +0000]
Btrfs: use find_or_create_page instead of grab_cache_page

grab_cache_page will use mapping_gfp_mask(), which for all inodes is set to
GFP_HIGHUSER_MOVABLE.  So instead use find_or_create_page in all cases where we
need GFP_NOFS so we don't deadlock.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

8 years agoBtrfs: use a worker thread to do caching
Josef Bacik [Thu, 30 Jun 2011 18:42:28 +0000]
Btrfs: use a worker thread to do caching

A user reported a deadlock when copying a bunch of files.  This is because they
were low on memory and kthreadd got hung up trying to migrate pages for an
allocation when starting the caching kthread.  The page was locked by the person
starting the caching kthread.  To fix this we just need to use the async thread
stuff so that the threads are already created and we don't have to worry about
deadlocks.  Thanks,

Reported-by: Roman Mamedov <rm@romanrm.ru>
Signed-off-by: Josef Bacik <josef@redhat.com>

8 years agobtrfs: don't BUG_ON allocation errors in btrfs_drop_snapshot
Mark Fasheh [Wed, 13 Jul 2011 17:59:59 +0000]
btrfs: don't BUG_ON allocation errors in btrfs_drop_snapshot

In addition to properly handling allocation failure from btrfs_alloc_path, I
also fixed up the kzalloc error handling code immediately below it.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>

8 years agobtrfs: Don't BUG_ON alloc_path errors in find_next_chunk
Mark Fasheh [Tue, 12 Jul 2011 17:57:59 +0000]
btrfs: Don't BUG_ON alloc_path errors in find_next_chunk

I also removed the BUG_ON from error return of find_next_chunk in
init_first_rw_device(). It turns out that the only caller of
init_first_rw_device() also BUGS on any nonzero return so no actual behavior
change has occurred here.

do_chunk_alloc() also needed an update since it calls btrfs_alloc_chunk()
which can now return -ENOMEM. Instead of setting space_info->full on any
error from btrfs_alloc_chunk() I catch and return every error value _except_
-ENOSPC. Thanks goes to Tsutomu Itoh for pointing that issue out.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>

8 years agoLinux 3.0
Linus Torvalds [Fri, 22 Jul 2011 02:17:23 +0000]
Linux 3.0

8 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jwessel...
Linus Torvalds [Fri, 22 Jul 2011 00:20:57 +0000]
Merge branch 'for_linus' of git://git./linux/kernel/git/jwessel/linux-2.6-kgdb

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jwessel/linux-2.6-kgdb:
  sparc,kgdbts: fix compile regression with kgdb test suite

8 years agosparc,kgdbts: fix compile regression with kgdb test suite
Jason Wessel [Tue, 19 Jul 2011 20:43:16 +0000]
sparc,kgdbts: fix compile regression with kgdb test suite

Commit 63ab25ebbc (kgdbts: unify/generalize gdb breakpoint adjustment)
introduced a compile regression on sparc.

kgdbts.c: In function 'check_and_rewind_pc':
kgdbts.c:307: error: implicit declaration of function 'instruction_pointer_set'

Simply add the correct macro definition for instruction pointer on the
Sparc architecture.

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: David S. Miller <davem@davemloft.net>

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Thu, 21 Jul 2011 21:28:01 +0000]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  CIFS: Fix wrong length in cifs_iovec_read

8 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 21 Jul 2011 19:25:39 +0000]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Make Dell Latitude E6420 use reboot=pci
  x86: Make Dell Latitude E5420 use reboot=pci

8 years agox86: Make Dell Latitude E6420 use reboot=pci
H. Peter Anvin [Thu, 21 Jul 2011 18:22:21 +0000]
x86: Make Dell Latitude E6420 use reboot=pci

Yet another variant of the Dell Latitude series which requires
reboot=pci.

From the E5420 bug report by Daniel J Blueman:

> The E6420 is affected also (same platform, different casing and
> features), which provides an external confirmation of the issue; I can
> submit a patch for that later or include it if you prefer:
> http://linux.koolsolutions.com/2009/08/04/howto-fix-linux-hangfreeze-during-reboots-and-restarts/

Reported-by: Daniel J Blueman <daniel.blueman@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@kernel.org>

8 years agox86: Make Dell Latitude E5420 use reboot=pci
Daniel J Blueman [Fri, 13 May 2011 01:04:59 +0000]
x86: Make Dell Latitude E5420 use reboot=pci

Rebooting on the Dell E5420 often hangs with the keyboard or ACPI
methods, but is reliable via the PCI method.

[ hpa: this was deferred because we believed for a long time that the
  recent reshuffling of the boot priorities in commit
  660e34cebf0a11d54f2d5dd8838607452355f321 fixed this platform.
  Unfortunately that turned out to be incorrect. ]

Signed-off-by: Daniel J Blueman <daniel.blueman@gmail.com>
Link: http://lkml.kernel.org/r/1305248699-2347-1-git-send-email-daniel.blueman@gmail.com
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@kernel.org>

8 years agoMerge branch 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keith...
Linus Torvalds [Thu, 21 Jul 2011 18:07:18 +0000]
Merge branch 'drm-intel-fixes' of git://git./linux/kernel/git/keithp/linux-2.6

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keithp/linux-2.6:
  drm/i915: Fix unfenced alignment on pre-G33 hardware
  drm/i915: Add quirk to disable SSC on Lenovo U160 LVDS

8 years agovfs: drop conditional inode prefetch in __do_lookup_rcu
Linus Torvalds [Thu, 21 Jul 2011 18:01:42 +0000]
vfs: drop conditional inode prefetch in __do_lookup_rcu

It seems to hurt performance in real life.  Yes, the inode will be used
later, but the conditional doesn't seem to predict all that well
(negative dentries are not uncommon) and it looks like the cost of
prefetching is simply higher than depending on the cache doing the right
thing.

As usual.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoFS-Cache: Fix __fscache_uncache_all_inode_pages()'s outer loop
Jan Beulich [Thu, 21 Jul 2011 14:02:43 +0000]
FS-Cache: Fix __fscache_uncache_all_inode_pages()'s outer loop

The compiler, at least for ix86 and m68k, validly warns that the
comparison:

next <= (loff_t)-1

is always true (and it's always true also for x86-64 and probably all
other arches - as long as pgoff_t isn't wider than loff_t).  The
intention appears to be to avoid wrapping of "next", so rather than
eliminating the pointless comparison, fix the loop to indeed get exited
when "next" would otherwise wrap.

On m68k the following warning is observed:

  fs/fscache/page.c: In function '__fscache_uncache_all_inode_pages':
  fs/fscache/page.c:979: warning: comparison is always false due to limited range of data type

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reported-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: Suresh Jayaraman <sjayaraman@suse.de>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoCIFS: Fix wrong length in cifs_iovec_read
Pavel Shilovsky [Wed, 20 Jul 2011 14:24:09 +0000]
CIFS: Fix wrong length in cifs_iovec_read

Signed-off-by: Pavel Shilovsky <piastryyy@gmail.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>

8 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 20 Jul 2011 22:56:25 +0000]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  signal: align __lock_task_sighand() irq disabling and RCU
  softirq,rcu: Inform RCU of irq_exit() activity
  sched: Add irq_{enter,exit}() to scheduler_ipi()
  rcu: protect __rcu_read_unlock() against scheduler-using irq handlers
  rcu: Streamline code produced by __rcu_read_unlock()
  rcu: Fix RCU_BOOST race handling current->rcu_read_unlock_special
  rcu: decrease rcu_report_exp_rnp coupling with scheduler

8 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 20 Jul 2011 22:55:48 +0000]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: Avoid creating superfluous NUMA domains on non-NUMA systems
  sched: Allow for overlapping sched_domain spans
  sched: Break out cpu_power from the sched_group structure

8 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 20 Jul 2011 22:33:59 +0000]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86. reboot: Make Dell Latitude E6320 use reboot=pci
  x86, doc only: Correct real-mode kernel header offset for init_size
  x86: Disable AMD_NUMA for 32bit for now

8 years agoMerge branch 'rcu/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck...
Ingo Molnar [Wed, 20 Jul 2011 18:59:26 +0000]
Merge branch 'rcu/urgent' of git://git./linux/kernel/git/paulmck/linux-2.6-rcu into core/urgent

8 years agosignal: align __lock_task_sighand() irq disabling and RCU
Paul E. McKenney [Tue, 19 Jul 2011 10:25:36 +0000]
signal: align __lock_task_sighand() irq disabling and RCU

The __lock_task_sighand() function calls rcu_read_lock() with interrupts
and preemption enabled, but later calls rcu_read_unlock() with interrupts
disabled.  It is therefore possible that this RCU read-side critical
section will be preempted and later RCU priority boosted, which means that
rcu_read_unlock() will call rt_mutex_unlock() in order to deboost itself, but
with interrupts disabled. This results in lockdep splats, so this commit
nests the RCU read-side critical section within the interrupt-disabled
region of code.  This prevents the RCU read-side critical section from
being preempted, and thus prevents the attempt to deboost with interrupts
disabled.

It is quite possible that a better long-term fix is to make rt_mutex_unlock()
disable irqs when acquiring the rt_mutex structure's ->wait_lock.

Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

8 years agosoftirq,rcu: Inform RCU of irq_exit() activity
Peter Zijlstra [Tue, 19 Jul 2011 22:32:00 +0000]
softirq,rcu: Inform RCU of irq_exit() activity

The rcu_read_unlock_special() function relies on in_irq() to exclude
scheduler activity from interrupt level.  This fails because exit_irq()
can invoke the scheduler after clearing the preempt_count() bits that
in_irq() uses to determine that it is at interrupt level.  This situation
can result in failures as follows:

 $task IRQ SoftIRQ

 rcu_read_lock()

 /* do stuff */

 <preempt> |= UNLOCK_BLOCKED

 rcu_read_unlock()
   --t->rcu_read_lock_nesting

irq_enter();
/* do stuff, don't use RCU */
irq_exit();
  sub_preempt_count(IRQ_EXIT_OFFSET);
  invoke_softirq()

ttwu();
  spin_lock_irq(&pi->lock)
  rcu_read_lock();
  /* do stuff */
  rcu_read_unlock();
    rcu_read_unlock_special()
      rcu_report_exp_rnp()
        ttwu()
          spin_lock_irq(&pi->lock) /* deadlock */

   rcu_read_unlock_special(t);

Ed can simply trigger this 'easy' because invoke_softirq() immediately
does a ttwu() of ksoftirqd/# instead of doing the in-place softirq stuff
first, but even without that the above happens.

Cure this by also excluding softirqs from the
rcu_read_unlock_special() handler and ensuring the force_irqthreads
ksoftirqd/# wakeup is done from full softirq context.

[ Alternatively, delaying the ->rcu_read_lock_nesting decrement
  until after the special handling would make the thing more robust
  in the face of interrupts as well.  And there is a separate patch
  for that. ]

Cc: Thomas Gleixner <tglx@linutronix.de>
Reported-and-tested-by: Ed Tomlinson <edt@aei.ca>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

8 years agosched: Add irq_{enter,exit}() to scheduler_ipi()
Peter Zijlstra [Tue, 19 Jul 2011 22:07:25 +0000]
sched: Add irq_{enter,exit}() to scheduler_ipi()

Ensure scheduler_ipi() calls irq_{enter,exit} when it does some actual
work. Traditionally we never did any actual work from the resched IPI
and all magic happened in the return from interrupt path.

Now that we do do some work, we need to ensure irq_{enter,exit} are
called so that we don't confuse things.

This affects things like timekeeping, NO_HZ and RCU, basically
everything with a hook in irq_enter/exit.

Explicit examples of things going wrong are:

  sched_clock_cpu() -- has a callback when leaving NO_HZ state to take
                    a new reading from GTOD and TSC. Without this
                    callback, time is stuck in the past.

  RCU -- needs in_irq() to work in order to avoid some nasty deadlocks

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

8 years agorcu: protect __rcu_read_unlock() against scheduler-using irq handlers
Paul E. McKenney [Mon, 18 Jul 2011 04:14:35 +0000]
rcu: protect __rcu_read_unlock() against scheduler-using irq handlers

The addition of RCU read-side critical sections within runqueue and
priority-inheritance lock critical sections introduced some deadlock
cycles, for example, involving interrupts from __rcu_read_unlock()
where the interrupt handlers call wake_up().  This situation can cause
the instance of __rcu_read_unlock() invoked from interrupt to do some
of the processing that would otherwise have been carried out by the
task-level instance of __rcu_read_unlock().  When the interrupt-level
instance of __rcu_read_unlock() is called with a scheduler lock held
from interrupt-entry/exit situations where in_irq() returns false,
deadlock can result.

This commit resolves these deadlocks by using negative values of
the per-task ->rcu_read_lock_nesting counter to indicate that an
instance of __rcu_read_unlock() is in flight, which in turn prevents
instances from interrupt handlers from doing any special processing.
This patch is inspired by Steven Rostedt's earlier patch that similarly
made __rcu_read_unlock() guard against interrupt-mediated recursion
(see https://lkml.org/lkml/2011/7/15/326), but this commit refines
Steven's approach to avoid the need for preemption disabling on the
__rcu_read_unlock() fastpath and to also avoid the need for manipulating
a separate per-CPU variable.

This patch avoids need for preempt_disable() by instead using negative
values of the per-task ->rcu_read_lock_nesting counter.  Note that nested
rcu_read_lock()/rcu_read_unlock() pairs are still permitted, but they will
never see ->rcu_read_lock_nesting go to zero, and will therefore never
invoke rcu_read_unlock_special(), thus preventing them from seeing the
RCU_READ_UNLOCK_BLOCKED bit should it be set in ->rcu_read_unlock_special.
This patch also adds a check for ->rcu_read_unlock_special being negative
in rcu_check_callbacks(), thus preventing the RCU_READ_UNLOCK_NEED_QS
bit from being set should a scheduling-clock interrupt occur while
__rcu_read_unlock() is exiting from an outermost RCU read-side critical
section.

Of course, __rcu_read_unlock() can be preempted during the time that
->rcu_read_lock_nesting is negative.  This could result in the setting
of the RCU_READ_UNLOCK_BLOCKED bit after __rcu_read_unlock() checks it,
and would also result it this task being queued on the corresponding
rcu_node structure's blkd_tasks list.  Therefore, some later RCU read-side
critical section would enter rcu_read_unlock_special() to clean up --
which could result in deadlock if that critical section happened to be in
the scheduler where the runqueue or priority-inheritance locks were held.

This situation is dealt with by making rcu_preempt_note_context_switch()
check for negative ->rcu_read_lock_nesting, thus refraining from
queuing the task (and from setting RCU_READ_UNLOCK_BLOCKED) if we are
already exiting from the outermost RCU read-side critical section (in
other words, we really are no longer actually in that RCU read-side
critical section).  In addition, rcu_preempt_note_context_switch()
invokes rcu_read_unlock_special() to carry out the cleanup in this case,
which clears out the ->rcu_read_unlock_special bits and dequeues the task
(if necessary), in turn avoiding needless delay of the current RCU grace
period and needless RCU priority boosting.

It is still illegal to call rcu_read_unlock() while holding a scheduler
lock if the prior RCU read-side critical section has ever had either
preemption or irqs enabled.  However, the common use case is legal,
namely where then entire RCU read-side critical section executes with
irqs disabled, for example, when the scheduler lock is held across the
entire lifetime of the RCU read-side critical section.

Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

8 years agosched: Avoid creating superfluous NUMA domains on non-NUMA systems
Peter Zijlstra [Wed, 20 Jul 2011 16:42:57 +0000]
sched: Avoid creating superfluous NUMA domains on non-NUMA systems

When creating sched_domains, stop when we've covered the entire
target span instead of continuing to create domains, only to
later find they're redundant and throw them away again.

This avoids single node systems from touching funny NUMA
sched_domain creation code and reduces the risks of the new
SD_OVERLAP code.

Requested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Anton Blanchard <anton@samba.org>
Cc: mahesh@linux.vnet.ibm.com
Cc: benh@kernel.crashing.org
Cc: linuxppc-dev@lists.ozlabs.org
Link: http://lkml.kernel.org/r/1311180177.29152.57.camel@twins
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agosched: Allow for overlapping sched_domain spans
Peter Zijlstra [Fri, 15 Jul 2011 08:35:52 +0000]
sched: Allow for overlapping sched_domain spans

Allow for sched_domain spans that overlap by giving such domains their
own sched_group list instead of sharing the sched_groups amongst
each-other.

This is needed for machines with more than 16 nodes, because
sched_domain_node_span() will generate a node mask from the
16 nearest nodes without regard if these masks have any overlap.

Currently sched_domains have a sched_group that maps to their child
sched_domain span, and since there is no overlap we share the
sched_group between the sched_domains of the various CPUs. If however
there is overlap, we would need to link the sched_group list in
different ways for each cpu, and hence sharing isn't possible.

In order to solve this, allocate private sched_groups for each CPU's
sched_domain but have the sched_groups share a sched_group_power
structure such that we can uniquely track the power.

Reported-and-tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-08bxqw9wis3qti9u5inifh3y@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agosched: Break out cpu_power from the sched_group structure
Peter Zijlstra [Thu, 14 Jul 2011 11:00:06 +0000]
sched: Break out cpu_power from the sched_group structure

In order to prepare for non-unique sched_groups per domain, we need to
carry the cpu_power elsewhere, so put a level of indirection in.

Reported-and-tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-qkho2byuhe4482fuknss40ad@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Wed, 20 Jul 2011 05:10:28 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  ceph: fix file mode calculation

8 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/linux-arm-soc
Linus Torvalds [Wed, 20 Jul 2011 05:10:05 +0000]
Merge branch 'fixes' of git://git./linux/kernel/git/arm/linux-arm-soc

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/linux-arm-soc:
  davinci: DM365 EVM: fix video input mux bits
  ARM: davinci: Check for NULL return from irq_alloc_generic_chip
  arm: davinci: Fix low level gpio irq handlers' argument

8 years agovmscan: fix a livelock in kswapd
Shaohua Li [Tue, 19 Jul 2011 15:49:26 +0000]
vmscan: fix a livelock in kswapd

I'm running a workload which triggers a lot of swap in a machine with 4
nodes.  After I kill the workload, I found a kswapd livelock.  Sometimes
kswapd3 or kswapd2 are keeping running and I can't access filesystem,
but most memory is free.

This looks like a regression since commit 08951e545918c159 ("mm: vmscan:
correct check for kswapd sleeping in sleeping_prematurely").

Node 2 and 3 have only ZONE_NORMAL, but balance_pgdat() will return 0
for classzone_idx.  The reason is end_zone in balance_pgdat() is 0 by
default, if all zones have watermark ok, end_zone will keep 0.

Later sleeping_prematurely() always returns true.  Because this is an
order 3 wakeup, and if classzone_idx is 0, both balanced_pages and
present_pages in pgdat_balanced() are 0.  We add a special case here.
If a zone has no page, we think it's balanced.  This fixes the livelock.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agofs/libfs.c: fix simple_attr_write() on 32bit machines
Akinobu Mita [Tue, 19 Jul 2011 15:49:25 +0000]
fs/libfs.c: fix simple_attr_write() on 32bit machines

Assume that /sys/kernel/debug/dummy64 is debugfs file created by
debugfs_create_x64().

# cd /sys/kernel/debug
# echo 0x1234567812345678 > dummy64
# cat dummy64
0x0000000012345678

# echo 0x80000000 > dummy64
# cat dummy64
0xffffffff80000000

A value larger than INT_MAX cannot be written to the debugfs file created
by debugfs_create_u64 or debugfs_create_x64 on 32bit machine.  Because
simple_attr_write() uses simple_strtol() for the conversion.

To fix this, use simple_strtoll() instead.

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Wed, 20 Jul 2011 04:50:21 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  vfs: fix race in rcu lookup of pruned dentry
  Fix cifs_get_root()

[ Edited the last commit to get rid of a 'unused variable "seq"'
  warning due to Al editing the patch.  - Linus ]

8 years agovfs: fix race in rcu lookup of pruned dentry
Linus Torvalds [Mon, 18 Jul 2011 22:43:29 +0000]
vfs: fix race in rcu lookup of pruned dentry

Don't update *inode in __follow_mount_rcu() until we'd verified that
there is mountpoint there.  Kudos to Hugh Dickins for catching that
one in the first place and eventually figuring out the solution (and
catching a braino in the earlier version of patch).

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

8 years agorcu: Streamline code produced by __rcu_read_unlock()
Paul E. McKenney [Sat, 21 May 2011 12:57:18 +0000]
rcu: Streamline code produced by __rcu_read_unlock()

Given some common flag combinations, particularly -Os, gcc will inline
rcu_read_unlock_special() despite its being in an unlikely() clause.
Use noinline to prohibit this misoptimization.

In addition, move the second barrier() in __rcu_read_unlock() so that
it is not on the common-case code path.  This will allow the compiler to
generate better code for the common-case path through __rcu_read_unlock().

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

8 years agorcu: Fix RCU_BOOST race handling current->rcu_read_unlock_special
Paul E. McKenney [Thu, 14 Jul 2011 19:24:11 +0000]
rcu: Fix RCU_BOOST race handling current->rcu_read_unlock_special

The RCU_BOOST commits for TREE_PREEMPT_RCU introduced an other-task
write to a new RCU_READ_UNLOCK_BOOSTED bit in the task_struct structure's
->rcu_read_unlock_special field, but, as noted by Steven Rostedt, without
correctly synchronizing all accesses to ->rcu_read_unlock_special.
This could result in bits in ->rcu_read_unlock_special being spuriously
set and cleared due to conflicting accesses, which in turn could result
in deadlocks between the rcu_node structure's ->lock and the scheduler's
rq and pi locks.  These deadlocks would result from RCU incorrectly
believing that the just-ended RCU read-side critical section had been
preempted and/or boosted.  If that RCU read-side critical section was
executed with either rq or pi locks held, RCU's ensuing (incorrect)
calls to the scheduler would cause the scheduler to attempt to once
again acquire the rq and pi locks, resulting in deadlock.  More complex
deadlock cycles are also possible, involving multiple rq and pi locks
as well as locks from multiple rcu_node structures.

This commit fixes synchronization by creating ->rcu_boosted field in
task_struct that is accessed and modified only when holding the ->lock
in the rcu_node structure on which the task is queued (on that rcu_node
structure's ->blkd_tasks list).  This results in tasks accessing only
their own current->rcu_read_unlock_special fields, making unsynchronized
access once again legal, and keeping the rcu_read_unlock() fastpath free
of atomic instructions and memory barriers.

The reason that the rcu_read_unlock() fastpath does not need to access
the new current->rcu_boosted field is that this new field cannot
be non-zero unless the RCU_READ_UNLOCK_BLOCKED bit is set in the
current->rcu_read_unlock_special field.  Therefore, rcu_read_unlock()
need only test current->rcu_read_unlock_special: if that is zero, then
current->rcu_boosted must also be zero.

This bug does not affect TINY_PREEMPT_RCU because this implementation
of RCU accesses current->rcu_read_unlock_special with irqs disabled,
thus preventing races on the !SMP systems that TINY_PREEMPT_RCU runs on.

Maybe-reported-by: Dave Jones <davej@redhat.com>
Maybe-reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Steven Rostedt <rostedt@goodmis.org>

8 years agorcu: decrease rcu_report_exp_rnp coupling with scheduler
Paul E. McKenney [Sun, 17 Jul 2011 09:05:49 +0000]
rcu: decrease rcu_report_exp_rnp coupling with scheduler

PREEMPT_RCU read-side critical sections blocking an expedited grace
period invoke rcu_report_exp_rnp().  When the last such critical section
has completed, rcu_report_exp_rnp() invokes the scheduler to wake up the
task that invoked synchronize_rcu_expedited() -- needlessly holding the
root rcu_node structure's lock while doing so, thus needlessly providing
a way for RCU and the scheduler to deadlock.

This commit therefore releases the root rcu_node structure's lock before
calling wake_up().

Reported-by: Ed Tomlinson <edt@aei.ca>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

8 years agoceph: fix file mode calculation
Sage Weil [Tue, 19 Jul 2011 18:25:04 +0000]
ceph: fix file mode calculation

open(2) must always include one of O_RDONLY, O_WRONLY, or O_RDWR.  No need
for any O_APPEND special case.

Passing O_WRONLY|O_RDWR is undefined according to the man page, but the
Linux VFS interprets this as O_RDWR, so we'll do the same.

This fixes open(2) with flags O_RDWR|O_APPEND, which was incorrectly being
translated to readonly.

Reported-by: Fyodor Ustinov <ufm@ufm.su>
Signed-off-by: Sage Weil <sage@newdream.net>

8 years agodavinci: DM365 EVM: fix video input mux bits
Jon Povey [Tue, 19 Jul 2011 03:30:11 +0000]
davinci: DM365 EVM: fix video input mux bits

Video input mux settings for tvp7002 and imager inputs were swapped.
Comment was correct.

Tested on EVM with tvp7002 input.

Signed-off-by: Jon Povey <jon.povey@racelogic.co.uk>
Acked-by: Manjunath Hadli <manjunath.hadli@ti.com>
Cc: stable@kernel.org
Signed-off-by: Sekhar Nori <nsekhar@ti.com>

8 years agoARM: davinci: Check for NULL return from irq_alloc_generic_chip
Todd Poynor [Sun, 17 Jul 2011 05:39:35 +0000]
ARM: davinci: Check for NULL return from irq_alloc_generic_chip

Avoid NULL dereference of irq_alloc_generic_chip return in low
memory conditions.

Signed-off-by: Todd Poynor <toddpoynor@google.com>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>

8 years agodrm/i915: Fix unfenced alignment on pre-G33 hardware
Chris Wilson [Mon, 18 Jul 2011 20:11:49 +0000]
drm/i915: Fix unfenced alignment on pre-G33 hardware

Align unfenced buffers on older hardware to the power-of-two object
size.  The docs suggest that it should be possible to align only to a
power-of-two tile height, but using the already computed fence size is
easier and always correct. We also have to make sure that we unbind
misaligned buffers upon tiling changes.

In order to prevent a repetition of this bug, we change the interface
to the alignment computation routines to force the caller to provide
the requested alignment and size of the GTT binding rather than assume
the current values on the object.

Reported-and-tested-by: Sitosfe Wheeler <sitsofe@yahoo.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=36326
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Keith Packard <keithp@keithp.com>

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Mon, 18 Jul 2011 20:29:26 +0000]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  pppoe: Must flush connections when MAC address changes too.
  include/linux/sdla.h: remove the prototype of sdla()
  tulip: dmfe: Remove old log spamming pr_debugs

8 years agopppoe: Must flush connections when MAC address changes too.
David S. Miller [Mon, 18 Jul 2011 18:48:28 +0000]
pppoe: Must flush connections when MAC address changes too.

Kernel bugzilla: 39252

Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoinclude/linux/sdla.h: remove the prototype of sdla()
WANG Cong [Sat, 16 Jul 2011 22:22:20 +0000]
include/linux/sdla.h: remove the prototype of sdla()

`make headers_check` complains that

linux-2.6/usr/include/linux/sdla.h:116: userspace cannot reference
function or variable defined in the kernel

this is due to that there is no such a kernel function,

void sdla(void *cfg_info, char *dev, struct frad_conf *conf, int quiet);

I don't know why we have it in a kernel header, so remove it.

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoFix cifs_get_root()
Al Viro [Mon, 18 Jul 2011 17:50:40 +0000]
Fix cifs_get_root()

Add missing ->i_mutex, convert to lookup_one_len() instead of
(broken) open-coded analog, cope with getting something like
a//b as relative pathname.  Simplify the hell out of it, while
we are there...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Reviewed-by: Jeff Layton <jlayton@redhat.com>

8 years agotulip: dmfe: Remove old log spamming pr_debugs
Joe Perches [Mon, 18 Jul 2011 17:44:44 +0000]
tulip: dmfe: Remove old log spamming pr_debugs

Commit 726b65ad444d ("tulip: Convert uses of KERN_DEBUG") enabled
some old previously inactive uses of pr_debug converted by
commit dde7c8ef1679 ("tulip/dmfe.c: Use dev_<level> and pr_<level>").

Remove these pr_debugs.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agosi4713-i2c: avoid potential buffer overflow on si4713
Mauro Carvalho Chehab [Sun, 17 Jul 2011 03:24:37 +0000]
si4713-i2c: avoid potential buffer overflow on si4713

While compiling it with Fedora 15, I noticed this issue:

  inlined from ‘si4713_write_econtrol_string’ at drivers/media/radio/si4713-i2c.c:1065:24:
  arch/x86/include/asm/uaccess_32.h:211:26: error: call to ‘copy_from_user_overflow’ declared with attribute error: copy_from_user() buffer size is not provably correct

Cc: stable@kernel.org
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Acked-by: Sakari Ailus <sakari.ailus@maxwell.research.nokia.com>
Acked-by: Eduardo Valentin <edubezval@gmail.com>
Reviewed-by: Eugene Teo <eugeneteo@kernel.sg>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoMerge branch 'fix/asoc' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Mon, 18 Jul 2011 16:05:59 +0000]
Merge branch 'fix/asoc' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'fix/asoc' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ASoC: Correct WM8994 MICBIAS supply widget hookup
  ASoC: Fix shift in WM8958 accessory detection default implementation
  ASoC: sh: fsi-hdmi: fixup snd_soc_card name
  ASoC: sh: fsi-da7210: fixup snd_soc_card name
  ASoC: sh: fsi-ak4642: fixup snd_soc_card name

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Mon, 18 Jul 2011 16:03:15 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  hppfs_lookup(): don't open-code lookup_one_len()
  hppfs: fix dentry leak
  cramfs: get_cramfs_inode() returns ERR_PTR() on failure
  ufs should use d_splice_alias()
  fix exofs ->get_parent()
  ceph analog of cifs build_path_from_dentry() race fix
  cifs: build_path_from_dentry() race fix

8 years agoMerge branch 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelv...
Linus Torvalds [Mon, 18 Jul 2011 16:02:58 +0000]
Merge branch 'hwmon-for-linus' of git://git./linux/kernel/git/jdelvare/staging

* 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging:
  hwmon: (max1111) Fix race condition causing NULL pointer exception
  hwmon: (it87) Fix label group removal
  hwmon: (asus_atk0110) Fix memory leak

8 years agohppfs_lookup(): don't open-code lookup_one_len()
Al Viro [Mon, 18 Jul 2011 02:27:22 +0000]
hppfs_lookup(): don't open-code lookup_one_len()

... and it's getting it wrong, too - missing ->d_revalidate() calls when
it's dealing with filesystem (procfs) that has non-trivial ->d_revalidate()...

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

8 years agohppfs: fix dentry leak
Al Viro [Mon, 18 Jul 2011 02:24:15 +0000]
hppfs: fix dentry leak

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

8 years agocramfs: get_cramfs_inode() returns ERR_PTR() on failure
Al Viro [Sun, 17 Jul 2011 23:04:14 +0000]
cramfs: get_cramfs_inode() returns ERR_PTR() on failure

... and we want to report these failures in ->lookup() anyway.

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

8 years agoufs should use d_splice_alias()
Al Viro [Sun, 17 Jul 2011 14:07:34 +0000]
ufs should use d_splice_alias()

it's NFS-exportable, so...

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

8 years agofix exofs ->get_parent()
Al Viro [Sat, 9 Jul 2011 00:56:55 +0000]
fix exofs ->get_parent()

NULL is not a possible return value for that method, TYVM...

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

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Sun, 17 Jul 2011 19:49:55 +0000]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  [CIFS] update cifs to version 1.74
  [CIFS] update limit for snprintf in cifs_construct_tcon
  cifs: Fix signing failure when server mandates signing for NTLMSSP

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Sun, 17 Jul 2011 19:49:28 +0000]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  Bluetooth: Fix crash with incoming L2CAP connections
  Bluetooth: Fix regression in L2CAP connection procedure
  gianfar: rx parser
  r6040: only disable RX interrupt if napi_schedule_prep is successful
  net: remove NETIF_F_ALL_TX_OFFLOADS
  net: sctp: fix checksum marking for outgoing packets

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog
Linus Torvalds [Sun, 17 Jul 2011 19:48:52 +0000]
Merge git://git./linux/kernel/git/wim/linux-2.6-watchdog

* git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  watchdog: hpwdt depends on PCI
  watchdog: fix hpwdt Kconfig regression in 3.0-rc

8 years agoMerge branch 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Sun, 17 Jul 2011 19:48:18 +0000]
Merge branch 'v4l_for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6:
  [media] tuner-core: fix a 2.6.39 regression with mt20xx
  [media] dvb_frontend: fix race condition in stopping/starting frontend
  [media] media: fix radio-sf16fmr2 build when SND is not enabled
  [media] MEDIA: Fix non-ISA_DMA_API link failure of sound code
  [media] nuvoton-cir: make idle timeout more sane
  [media] mceusb: increase default timeout to 100ms
  [media] mceusb: Timeout unit corrections
  [media] Revert "V4L/DVB: cx23885: Enable Message Signaled Interrupts(MSI)"

8 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
Linus Torvalds [Sun, 17 Jul 2011 19:47:47 +0000]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux-acpi-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6:
  ACPI: Fixes device power states array overflow
  ACPI, APEI, HEST, Detect duplicated hardware error source ID
  ACPI: Fix lockdep false positives in acpi_power_off()

8 years agoMerge branch 'pm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/suspe...
Linus Torvalds [Sun, 17 Jul 2011 19:47:27 +0000]
Merge branch 'pm-fixes' of git://git./linux/kernel/git/rafael/suspend-2.6

* 'pm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/suspend-2.6:
  PM / MIPS: Convert i8259.c to using syscore_ops

8 years agoMerge branch 's5p-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 17 Jul 2011 19:47:11 +0000]
Merge branch 's5p-fixes-for-linus' of git://git./linux/kernel/git/kgene/linux-samsung

* 's5p-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
  ARM: SAMSUNG: DMA Cleanup as per sparse
  ARM: SAMSUNG: Check NULL return from irq_alloc_generic_chip

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Sun, 17 Jul 2011 19:43:58 +0000]
Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sparc: sun4m SMP: fix wrong shift instruction in IPI handler
  sparc32,leon: Added __init declaration to leon_flush_needed()
  sparc/irqs: Do not trace arch_local_{*,irq_*} functions

8 years agohwmon: (max1111) Fix race condition causing NULL pointer exception
Pavel Herrmann [Sun, 17 Jul 2011 16:39:19 +0000]
hwmon: (max1111) Fix race condition causing NULL pointer exception

spi_sync call uses its spi_message parameter to keep completion information,
using a drvdata structure is not thread-safe. Use a mutex to prevent
multiple access to shared driver data.

Signed-off-by: Pavel Herrmann <morpheus.ibis@gmail.com>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Marek Vasut <marek.vasut@gmail.com>
Acked-by: Cyril Hrubis <metan@ucw.cz>
Tested-by: Stanislav Brabec <utx@penguin.cz>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: stable@kernel.org

8 years agohwmon: (it87) Fix label group removal
Jean Delvare [Sun, 17 Jul 2011 16:39:19 +0000]
hwmon: (it87) Fix label group removal

A copy-and-paste error caused it87_attributes_vid to be referenced
where it87_attributes_label should be. Thankfully the group is only
used for attribute removal, not attribute creation, so the effects of
this bug are limited, but let's fix it still.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: stable@kernel.org
Acked-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (asus_atk0110) Fix memory leak
Luca Tettamanti [Sun, 17 Jul 2011 16:39:18 +0000]
hwmon: (asus_atk0110) Fix memory leak

The object returned by atk_gitm is dynamically allocated and must be
freed.

Signed-off-by: Luca Tettamanti <kronos.it@gmail.com>
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Cc: stable@kernel.org

8 years agowatchdog: hpwdt depends on PCI
Randy Dunlap [Sat, 16 Jul 2011 19:25:49 +0000]
watchdog: hpwdt depends on PCI

hpwdt is a PCI driver so it should depend on PCI.
Fixes these build errors:

drivers/watchdog/hpwdt.c:762: error: implicit declaration of function 'pci_iomap'
drivers/watchdog/hpwdt.c:762: warning: assignment makes pointer from integer without a cast
drivers/watchdog/hpwdt.c:797: error: implicit declaration of function 'pci_iounmap'

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Cc: Thomas Mingarelli <thomas.mingarelli@hp.com>

8 years agoASoC: Correct WM8994 MICBIAS supply widget hookup
Mark Brown [Thu, 14 Jul 2011 09:21:37 +0000]
ASoC: Correct WM8994 MICBIAS supply widget hookup

The WM8994 and WM8958 series of devices have two MICBIAS supplies rather
than one, the current widget actually manages the microphone detection
control register bit (which is managed separately by the relevant API).

Fix this, hooking the relevant supplies up to the MICBIAS1 and MICBIAS2
widgets.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Cc: stable@kernel.org

8 years agoceph analog of cifs build_path_from_dentry() race fix
Al Viro [Sun, 17 Jul 2011 03:43:58 +0000]
ceph analog of cifs build_path_from_dentry() race fix

... unfortunately, cifs bug got copied.  Fix is essentially the same.

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

8 years agocifs: build_path_from_dentry() race fix
Al Viro [Sun, 17 Jul 2011 03:37:20 +0000]
cifs: build_path_from_dentry() race fix

deal with d_move() races properly; rename_lock read-retry loop,
rcu_read_lock() held while walking to root, d_lock held over
subtraction from namelen and copying the component to stabilize
->d_name.

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

8 years agosparc: sun4m SMP: fix wrong shift instruction in IPI handler
Will Simoneau [Sat, 16 Jul 2011 17:45:12 +0000]
sparc: sun4m SMP: fix wrong shift instruction in IPI handler

This shift instruction appears to be shifting in the wrong direction.
Without this change, my SparcStation-20MP hangs just after bringing up
the second CPU:

Entering SMP Mode...
Starting CPU 2 at f02b4e90
Brought up 2 CPUs
Total of 2 processors activated (99.52 BogoMIPS).
   *** stuck ***

Signed-off-by: Will Simoneau <simoneau@ele.uri.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoBluetooth: Fix crash with incoming L2CAP connections
Ilia Kolomisnky [Fri, 15 Jul 2011 18:30:21 +0000]
Bluetooth: Fix crash with incoming L2CAP connections

Another regression fix considering incomming l2cap connections with
defer_setup enabled. In situations when incomming connection is
extracted with l2cap_sock_accept, it's bt_sock info will have
'parent' member zerroed, but 'parent' may be used unconditionally
in l2cap_conn_start() and l2cap_security_cfm() when defer_setup
is enabled.

Backtrace:
[<bf02d5ac>] (l2cap_security_cfm+0x0/0x2ac [bluetooth]) from [<bf01f01c>] (hci_event_pac
ket+0xc2c/0x4aa4 [bluetooth])
[<bf01e3f0>] (hci_event_packet+0x0/0x4aa4 [bluetooth]) from [<bf01a844>] (hci_rx_task+0x
cc/0x27c [bluetooth])
[<bf01a778>] (hci_rx_task+0x0/0x27c [bluetooth]) from [<c008eee4>] (tasklet_action+0xa0/
0x15c)
[<c008ee44>] (tasklet_action+0x0/0x15c) from [<c008f38c>] (__do_softirq+0x98/0x130)
 r7:00000101 r6:00000018 r5:00000001 r4:efc46000
[<c008f2f4>] (__do_softirq+0x0/0x130) from [<c008f524>] (do_softirq+0x4c/0x58)
[<c008f4d8>] (do_softirq+0x0/0x58) from [<c008f5e0>] (run_ksoftirqd+0xb0/0x1b4)
 r4:efc46000 r3:00000001
[<c008f530>] (run_ksoftirqd+0x0/0x1b4) from [<c009f2a8>] (kthread+0x84/0x8c)
 r7:00000000 r6:c008f530 r5:efc47fc4 r4:efc41f08
[<c009f224>] (kthread+0x0/0x8c) from [<c008cc84>] (do_exit+0x0/0x5f0)

Signed-off-by: Ilia Kolomisnky <iliak@ti.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: David S. Miller <davem@davemloft.net>