firefly-linux-kernel-4.4.55.git
12 years agodrbd: Consider the disk-timeout also for meta-data IO operations
Philipp Reisner [Wed, 14 Dec 2011 17:01:21 +0000 (18:01 +0100)]
drbd: Consider the disk-timeout also for meta-data IO operations

If the backing device is already frozen during attach, we failed
to recognize that. The current disk-timeout code works on top
of the drbd_request objects. During attach we do not allow IO
and therefore never generate a drbd_request object but block
before that in drbd_make_request().

This patch adds the timeout to all drbd_md_sync_page_io().

Before this patch we used to go from D_ATTACHING directly
to D_DISKLESS if IO failed during attach. We can no longer
do this since we have to stay in D_FAILED until all IO
ops issued to the backing device returned.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Do not send state packets while lower than C_CONNECTED cstate
Philipp Reisner [Tue, 13 Dec 2011 09:31:32 +0000 (10:31 +0100)]
drbd: Do not send state packets while lower than C_CONNECTED cstate

I.e. in C_WF_REPORT_PARAMS or in C_WF_CONNECTION.
Sending may already work in these cstates, but the peer still expects
the HandShake / ConnectionFeatures packet.

Actually triggered by the Testuite on kugel.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: fix race between disconnect and receive_state
Lars Ellenberg [Mon, 5 Dec 2011 13:39:25 +0000 (14:39 +0100)]
drbd: fix race between disconnect and receive_state

If the asender thread, or request_timer_fn(), or some other part of
the code, decided to drop the connection (because of timeout or other),
but the receiver just now was processing a P_STATE packet, there was a
chance that receive_state() would do a hard state change
"re-establishing" an already failed connection without additional handshake.

Log excerpt:
  Remote failed to finish a request within ko-count * timeout
  peer( Secondary -> Unknown ) conn( Connected -> Timeout ) pdsk( UpToDate -> DUnknown )
  asender terminated
  ...
  peer( Unknown -> Secondary ) conn( Timeout -> Connected ) pdsk( DUnknown -> UpToDate ) peer_isp( 0 -> 1 )
  ...
  Connection closed
  peer( Secondary -> Unknown ) conn( Connected -> Unconnected ) pdsk( UpToDate -> DUnknown ) peer_isp( 1 -> 0 )
  receiver terminated

Impact:
while the connection state is erroneously "Connected",
requests may be queued and even sent,
which would never be acknowledged,
and may have been missed by the cleanup.
These requests would never be completed.

The next drbd_suspend_io() will then lock up,
waiting forever for these requests to complete.

Fixed in several code paths:
  Make sure the connection state is NetworkFailure or worse
  before starting the cleanup in drbd_disconnect().
  This should make sure the cleanup won't miss any requests.

  Disallow receive_state() to "upgrade" the connection state
  from an error state. This will make sure the "illegal" state
  transition won't happen.

  For all connection failure states,
  relax the safe-guard in sanitize_state() again
  to silently mask out those state changes
  (e.g. Timeout -> Connected becomes Timeout -> Timeout).

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: fix potential spinlock deadlock
Lars Ellenberg [Wed, 2 Nov 2011 15:29:45 +0000 (16:29 +0100)]
drbd: fix potential spinlock deadlock

drbd_try_clear_on_disk_bm() has a sanity check for the number of blocks
left to be resynced (rs_left) in the current resync extent.
If it detects a mismatch, it complains, and forces a disconnect using
drbd_force_state(mdev, NS(conn, C_DISCONNECTING));

Unfortunately, this may be called while holding the req_lock,
and drbd_force_state() want's to aquire that lock itself. Deadlock.

Don't force a disconnect, but fix up rs_left by recounting and
reassigning the number of dirty blocks in that extent.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Fixed an obvious copy-n-paste mistake
Philipp Reisner [Wed, 9 Nov 2011 20:04:03 +0000 (21:04 +0100)]
drbd: Fixed an obvious copy-n-paste mistake

This bug might have caused troubles if disk-barriers and the ahead-behind
more are enabled at the same time.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: send intermediate state change results to the peer
Lars Ellenberg [Thu, 27 Oct 2011 14:52:30 +0000 (16:52 +0200)]
drbd: send intermediate state change results to the peer

DRBD state changes schedule after_state_ch() actions to a worker thread,
which decides on the old and new states of that change, whether to send
an informational state update packet (P_STATE) to the peer.
If it decides to drbd_send_state(), it would however always send the
_curent_ state, which, if a second state change happens before the
after_state_ch() of the first ran, may "fast-forward" the peer's view
about this node.  In most cases that is harmless, but sometimes this can
confuse DRBD, for example into not actually starting a necessary resync
if you do a very tight detach/attach loop on a Connected Secondary.

Fix this by always sending the "new" state of the respective state
transition which scheduled this after_state_ch() work.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: fix spurious meta data IO "error"
Lars Ellenberg [Thu, 6 Oct 2011 15:30:26 +0000 (17:30 +0200)]
drbd: fix spurious meta data IO "error"

When detaching, even cleanly detaching due to administrator request,
we always go through D_FAILED before we become D_DISKLESS.

Don't let that state change race with an in-flight meta data IO,
or that one might think it actually experienced an IO error.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Fixed a race condition between detach and start of resync
Philipp Reisner [Thu, 6 Oct 2011 16:29:14 +0000 (18:29 +0200)]
drbd: Fixed a race condition between detach and start of resync

drbd_state_lock() is only there to serialize cluster wide state
changes. Testing the local disk state needs to happen while
holding the global_state_lock.

Otherwise you might see something like this (Oct 6 on kugel)
14:20:24 drbd0: conn( WFSyncUUID -> Connected ) disk( Inconsistent -> Failed )
14:20:24 drbd0: helper command: /sbin/drbdadm before-resync-target minor-0 exit code 0 (0x0)
14:20:24 drbd0: conn( Connected -> SyncTarget ) disk( Failed -> Inconsistent )

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: fix harmless race to not trigger an ASSERT
Lars Ellenberg [Thu, 6 Oct 2011 15:10:34 +0000 (17:10 +0200)]
drbd: fix harmless race to not trigger an ASSERT

We have one pre-allocated page to do certain synchronous meta data IO with,
using it is serialized like so:
drbd_md_get_buffer();
drbd_md_sync_page_io();
drbd_md_sync_page_io();
...
drbd_md_put_buffer();

In drbd_md_sync_page_io() there is an
ASSERT(atomic_read(&mdev->md_io_in_use) == 1);

We want to be able to timeout on unresponsive lower level devices, so we
can "detach" in that case. Inside drbd_md_sync_page_io() we grab an extra
reference, to not have a dangling pointer in case a delayed IO eventually
does still complete, even after we "detached" already.

We need to put the extra reference before we signal completion from the
completion handler, or the second drbd_md_sync_page_io() above may
trigger the assert (reference count still 2).

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Derive sync-UUIDs only from the bitmap-uuid if it is non-zero
Philipp Reisner [Wed, 5 Oct 2011 13:54:18 +0000 (15:54 +0200)]
drbd: Derive sync-UUIDs only from the bitmap-uuid if it is non-zero

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: drbd_nl_resize(): Fix missing put_ldev() on error path
Andreas Gruenbacher [Wed, 28 Sep 2011 20:15:04 +0000 (22:15 +0200)]
drbd: drbd_nl_resize(): Fix missing put_ldev() on error path

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: fix "stalled" empty resync
Lars Ellenberg [Mon, 26 Sep 2011 13:24:56 +0000 (15:24 +0200)]
drbd: fix "stalled" empty resync

With sync-after dependencies, given "lucky" timing of pause/unpause
events, and the end of an empty (0 bits set) resync was sometimes not
detected on the SyncTarget, leading to a "stalled" SyncSource state.

Fixed this by expecting not only "Inconsistent -> UpToDate" but also
"Consistent -> UpToDate" transitions for the peer disk state
to end a resync.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Bugfix for the connection behavior
Philipp Reisner [Thu, 4 Aug 2011 08:33:08 +0000 (10:33 +0200)]
drbd: Bugfix for the connection behavior

If we get into the C_BROKEN_PIPE cstate once, the state engine set the
thi->t_state of the receiver thread to restarting.  But with the while loop
in drbdd_init() a new connection gets established. After the call into
drbdd() returns immediately since the thi->t_state is not RUNNING.  The
restart of drbd_init() then resets thi->t_state to RUNNING.

I.e. after entering C_BROKEN_PIPE once, the next successful established
connection gets wasted.

The two parts of the fix:
  * Do not cause the thread to restart if we detect the issue
    with the sockets while we are in C_WF_CONNECTION.

  * Make sure that all actions that would have set us to C_BROKEN_PIPE
    happen before the state change to C_WF_REPORT_PARAMS.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Cleanup all epoch objects upon connection loss
Philipp Reisner [Mon, 18 Jul 2011 13:45:15 +0000 (15:45 +0200)]
drbd: Cleanup all epoch objects upon connection loss

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: detach must not try to abort non-local requests from drbd-8.4
Philipp Reisner [Mon, 18 Jul 2011 14:25:15 +0000 (16:25 +0200)]
drbd: detach must not try to abort non-local requests from drbd-8.4

Cherry picked form 8.4

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Consider that the no-data-condition could be in connected state
Philipp Reisner [Fri, 15 Jul 2011 16:44:26 +0000 (18:44 +0200)]
drbd: Consider that the no-data-condition could be in connected state

...when the peer has inconsistent data. In that case we failed to
clear the susp_nod flag. When the local disk was attached again

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Fixed current UUID generation
Philipp Reisner [Fri, 15 Jul 2011 10:14:27 +0000 (12:14 +0200)]
drbd: Fixed current UUID generation

Now, the new edition of the clause only fires if a diskless
peer gets promoted.

This is a fixup for "drbd: Delayed creation of current-UUID".

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: change some GFP_KERNEL to GFP_NOIO
Lars Ellenberg [Mon, 11 Jul 2011 15:32:26 +0000 (17:32 +0200)]
drbd: change some GFP_KERNEL to GFP_NOIO

Bitmap IO may happend in the context of an application write,
in the generic block IO path.  We need to use GFP_NOIO.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Implemented the disk-timeout option
Philipp Reisner [Wed, 29 Jun 2011 12:06:08 +0000 (14:06 +0200)]
drbd: Implemented the disk-timeout option

When the disk-timeout is active, and it expires for a single request,
we consider the local disk as D_FAILED. Note: With this change,
I made both timeout based state transitions HARD state transitions.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Force flag for the detach operation
Philipp Reisner [Mon, 14 Mar 2011 10:54:47 +0000 (11:54 +0100)]
drbd: Force flag for the detach operation

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Allow new IOs while the local disk in in FAILED state
Philipp Reisner [Tue, 28 Jun 2011 15:01:19 +0000 (17:01 +0200)]
drbd: Allow new IOs while the local disk in in FAILED state

The last bunch of commits prepared the 'detach from tar pit' feature.
With that we can be for long time in disk state FAILED. We need
to accept new IO requests during that time.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Bitmap IO functions can now return prematurely if the disk breaks
Philipp Reisner [Mon, 27 Jun 2011 13:29:16 +0000 (15:29 +0200)]
drbd: Bitmap IO functions can now return prematurely if the disk breaks

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Added a kref to bm_aio_ctx
Philipp Reisner [Mon, 27 Jun 2011 12:56:52 +0000 (14:56 +0200)]
drbd: Added a kref to bm_aio_ctx

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Hold a reference to ldev while doing meta-data IO
Philipp Reisner [Mon, 27 Jun 2011 10:23:33 +0000 (12:23 +0200)]
drbd: Hold a reference to ldev while doing meta-data IO

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Keep a reference to the bio until the completion handler finished
Philipp Reisner [Mon, 4 Jul 2011 09:14:31 +0000 (11:14 +0200)]
drbd: Keep a reference to the bio until the completion handler finished

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Implemented wait_until_done_or_disk_failure()
Philipp Reisner [Sun, 26 Jun 2011 20:26:31 +0000 (22:26 +0200)]
drbd: Implemented wait_until_done_or_disk_failure()

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Replaced md_io_mutex by an atomic: md_io_in_use
Philipp Reisner [Mon, 27 Jun 2011 09:51:46 +0000 (11:51 +0200)]
drbd: Replaced md_io_mutex by an atomic: md_io_in_use

The new function drbd_md_get_buffer() aborts waiting for the buffer
in case the disk failes in the meantime.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: moved md_io into mdev
Philipp Reisner [Sun, 26 Jun 2011 09:20:27 +0000 (11:20 +0200)]
drbd: moved md_io into mdev

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Immediately allow completion of IOs, that wait for IO completions on a failed...
Philipp Reisner [Mon, 14 Mar 2011 12:01:50 +0000 (13:01 +0100)]
drbd: Immediately allow completion of IOs, that wait for IO completions on a failed disk

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Keep a reference to barrier acked requests
Philipp Reisner [Tue, 15 Mar 2011 09:25:18 +0000 (10:25 +0100)]
drbd: Keep a reference to barrier acked requests

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Improve compatibility with drbd's older than 8.3.7
Philipp Reisner [Thu, 30 Jun 2011 13:43:06 +0000 (15:43 +0200)]
drbd: Improve compatibility with drbd's older than 8.3.7

Regression introduced with 8.3.11 commit:
drbd: Take a more conservative approach when deciding max_bio_size

Never ever tell an older drbd, that we support more than 32KiB
in a single data request (packet).
Never believe an older drbd, that is supports more than 32KiB
in a single data request (packet)

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Only print sanitize state's warnings, if the state change happens
Philipp Reisner [Wed, 29 Jun 2011 08:49:13 +0000 (10:49 +0200)]
drbd: Only print sanitize state's warnings, if the state change happens

The reason for this change is that, with when doing
'drbdadm invalidate' on a disconnected resource caused
an "implicitly set pdsk from UpToDate to DUnknown" message,
which was missleading.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: downgraded error printk to info
Lars Ellenberg [Mon, 20 Jun 2011 23:13:37 +0000 (01:13 +0200)]
drbd: downgraded error printk to info

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: allow ping-timeout of up to 30 seconds
Lars Ellenberg [Mon, 20 Jun 2011 20:21:19 +0000 (22:21 +0200)]
drbd: allow ping-timeout of up to 30 seconds

Allow up to 300 centi-seconds to be configured for the "ping timeout".
There may be setups where heavy congestion, huge buffers, and asymmetric
bandwidth limitations may need a "huge" ping-timeout as work-around
for "spurious connection loss" problems.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agoDRBD: Fix comparison always false warning due to long/long long compare
David Howells [Tue, 14 Jun 2011 23:59:04 +0000 (00:59 +0100)]
DRBD: Fix comparison always false warning due to long/long long compare

Fix warnings of the following nature in the drbd header:

In file included from drivers/block/drbd/drbd_bitmap.c:32:
drivers/block/drbd/drbd_int.h: In function 'drbd_get_syncer_progress':
drivers/block/drbd/drbd_int.h:2234: warning: comparison is always false due to limited range of data

where mdev->rs_total (an unsigned long) is being compared to 1ULL << 32, which
is always false on a 32-bit machine.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
12 years agodrbd: spelling fix: too small
Lars Ellenberg [Mon, 6 Jun 2011 13:36:04 +0000 (15:36 +0200)]
drbd: spelling fix: too small

It is not "to small", but "too small".

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: cosmetic: fix accidental division instead of modulo when pretty printing
Lars Ellenberg [Fri, 3 Jun 2011 19:13:17 +0000 (21:13 +0200)]
drbd: cosmetic: fix accidental division instead of modulo when pretty printing

For large resync rates, seq_printf_with_thousands_grouping()
accidentally only produced Y,000,00Y, instead of the real numbers.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agodrbd: Lower log priority for an event that is definitely not an error
Philipp Reisner [Wed, 25 May 2011 09:03:04 +0000 (11:03 +0200)]
drbd: Lower log priority for an event that is definitely not an error

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
12 years agoMerge tag 'v3.4-rc5' into for-3.5/core
Jens Axboe [Tue, 1 May 2012 12:29:55 +0000 (14:29 +0200)]
Merge tag 'v3.4-rc5' into for-3.5/core

The core branch is behind driver commits that we want to build
on for 3.5, hence I'm pulling in a later -rc.

Linux 3.4-rc5

Conflicts:
Documentation/feature-removal-schedule.txt

Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agoLinux 3.4-rc5
Linus Torvalds [Sun, 29 Apr 2012 22:19:10 +0000 (15:19 -0700)]
Linux 3.4-rc5

12 years agoMerge tag 'pm-for-3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Sun, 29 Apr 2012 22:00:44 +0000 (15:00 -0700)]
Merge tag 'pm-for-3.4-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael J. Wysocki:
 "Fix for an issue causing hibernation to hang on systems with highmem
  (that practically means i386) due to broken memory management (bug
  introduced in 3.2, so -stable material) and PM documentation update
  making the freezer documentation follow the code again after some
  recent updates."

* tag 'pm-for-3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM / Freezer / Docs: Update documentation about freezing of tasks
  PM / Hibernate: fix the number of pages used for hibernate/thaw buffering

12 years agoautofs: make the autofsv5 packet file descriptor use a packetized pipe
Linus Torvalds [Sun, 29 Apr 2012 20:30:08 +0000 (13:30 -0700)]
autofs: make the autofsv5 packet file descriptor use a packetized pipe

The autofs packet size has had a very unfortunate size problem on x86:
because the alignment of 'u64' differs in 32-bit and 64-bit modes, and
because the packet data was not 8-byte aligned, the size of the autofsv5
packet structure differed between 32-bit and 64-bit modes despite
looking otherwise identical (300 vs 304 bytes respectively).

We first fixed that up by making the 64-bit compat mode know about this
problem in commit a32744d4abae ("autofs: work around unhappy compat
problem on x86-64"), and that made a 32-bit 'systemd' work happily on a
64-bit kernel because everything then worked the same way as on a 32-bit
kernel.

But it turned out that 'automount' had actually known and worked around
this problem in user space, so fixing the kernel to do the proper 32-bit
compatibility handling actually *broke* 32-bit automount on a 64-bit
kernel, because it knew that the packet sizes were wrong and expected
those incorrect sizes.

As a result, we ended up reverting that compatibility mode fix, and
thus breaking systemd again, in commit fcbf94b9dedd.

With both automount and systemd doing a single read() system call, and
verifying that they get *exactly* the size they expect but using
different sizes, it seemed that fixing one of them inevitably seemed to
break the other.  At one point, a patch I seriously considered applying
from Michael Tokarev did a "strcmp()" to see if it was automount that
was doing the operation.  Ugly, ugly.

However, a prettier solution exists now thanks to the packetized pipe
mode.  By marking the communication pipe as being packetized (by simply
setting the O_DIRECT flag), we can always just write the bigger packet
size, and if user-space does a smaller read, it will just get that
partial end result and the extra alignment padding will simply be thrown
away.

This makes both automount and systemd happy, since they now get the size
they asked for, and the kernel side of autofs simply no longer needs to
care - it could pad out the packet arbitrarily.

Of course, if there is some *other* user of autofs (please, please,
please tell me it ain't so - and we haven't heard of any) that tries to
read the packets with multiple writes, that other user will now be
broken - the whole point of the packetized mode is that one system call
gets exactly one packet, and you cannot read a packet in pieces.

Tested-by: Michael Tokarev <mjt@tls.msk.ru>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: David Miller <davem@davemloft.net>
Cc: Ian Kent <raven@themaw.net>
Cc: Thomas Meyer <thomas@m3y3r.de>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoPM / Freezer / Docs: Update documentation about freezing of tasks
Marcos Paulo de Souza [Sun, 29 Apr 2012 20:29:30 +0000 (22:29 +0200)]
PM / Freezer / Docs: Update documentation about freezing of tasks

The file Documentation/power/freezing-of-tasks.txt was still referencing
the TIF_FREEZE flag, that was removed by the commit
d88e4cb67197d007fb778d62fe17360e970d5bfa(freezer: remove now unused
TIF_FREEZE).

This patch removes all the references of TIF_FREEZE that were left
behind.

Signed-off-by: Marcos Paulo de Souza <marcos.souza.org@gmail.com>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
12 years agopipes: add a "packetized pipe" mode for writing
Linus Torvalds [Sun, 29 Apr 2012 20:12:42 +0000 (13:12 -0700)]
pipes: add a "packetized pipe" mode for writing

The actual internal pipe implementation is already really about
individual packets (called "pipe buffers"), and this simply exposes that
as a special packetized mode.

When we are in the packetized mode (marked by O_DIRECT as suggested by
Alan Cox), a write() on a pipe will not merge the new data with previous
writes, so each write will get a pipe buffer of its own.  The pipe
buffer is then marked with the PIPE_BUF_FLAG_PACKET flag, which in turn
will tell the reader side to break the read at that boundary (and throw
away any partial packet contents that do not fit in the read buffer).

End result: as long as you do writes less than PIPE_BUF in size (so that
the pipe doesn't have to split them up), you can now treat the pipe as a
packet interface, where each read() system call will read one packet at
a time.  You can just use a sufficiently big read buffer (PIPE_BUF is
sufficient, since bigger than that doesn't guarantee atomicity anyway),
and the return value of the read() will naturally give you the size of
the packet.

NOTE! We do not support zero-sized packets, and zero-sized reads and
writes to a pipe continue to be no-ops.  Also note that big packets will
currently be split at write time, but that the size at which that
happens is not really specified (except that it's bigger than PIPE_BUF).
Currently that limit is the system page size, but we might want to
explicitly support bigger packets some day.

The main user for this is going to be the autofs packet interface,
allowing us to stop having to care so deeply about exact packet sizes
(which have had bugs with 32/64-bit compatibility modes).  But user
space can create packetized pipes with "pipe2(fd, O_DIRECT)", which will
fail with an EINVAL on kernels that do not support this interface.

Tested-by: Michael Tokarev <mjt@tls.msk.ru>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: David Miller <davem@davemloft.net>
Cc: Ian Kent <raven@themaw.net>
Cc: Thomas Meyer <thomas@m3y3r.de>
Cc: stable@kernel.org # needed for systemd/autofs interaction fix
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge tag 'staging-3.4-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sun, 29 Apr 2012 19:19:13 +0000 (12:19 -0700)]
Merge tag 'staging-3.4-rc4' of git://git./linux/kernel/git/gregkh/staging

Pull staging tree fixes from Greg Kroah-Hartman:
 "Here are some tiny drivers/staging/ bugfixes.  Some build fixes that
  were recently reported, as well as one kfree bug that is hitting a
  number of users."

* tag 'staging-3.4-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: ozwpan: Fix bug where kfree is called twice.
  staging: octeon-ethernet: fix build errors by including interrupt.h
  staging: zcache: fix Kconfig crypto dependency
  staging: tidspbridge: remove usage of OMAP2_L4_IO_ADDRESS

12 years agoMerge tag 'usb-3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 29 Apr 2012 19:17:54 +0000 (12:17 -0700)]
Merge tag 'usb-3.4-rc5' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg Kroah-Hartman:
 "Here are a number of small USB fixes for 3.4-rc5.

  Nothing major, as before, some USB gadget fixes.  There's a crash fix
  for a number of ASUS laptops on resume that had been reported by a
  number of different people.  We think the fix might also pertain to
  other machines, as this was a BIOS bug, and they seem to travel to
  different models and manufacturers quite easily.  Other than that,
  some other reported problems fixed as well."

* tag 'usb-3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: gadget: udc-core: fix incompatibility with dummy-hcd
  usb: gadget: udc-core: fix wrong call order
  USB: cdc-wdm: fix race leading leading to memory corruption
  USB: EHCI: fix crash during suspend on ASUS computers
  usb gadget: uvc: uvc_request_data::length field must be signed
  usb: gadget: dummy: do not call pullup() on udc_stop()
  usb: musb: davinci.c: add missing unregister
  usb: musb: drop __deprecated flag
  USB: gadget: storage gadgets send wrong error code for unknown commands
  usb: otg: gpio_vbus: Add otg transceiver events and notifiers

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sat, 28 Apr 2012 16:30:07 +0000 (09:30 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "This has our collection of bug fixes.  I missed the last rc because I
  thought our patches were making NFS crash during my xfs test runs.
  Turns out it was an NFS client bug fixed by someone else while I tried
  to bisect it.

  All of these fixes are small, but some are fairly high impact.  The
  biggest are fixes for our mount -o remount handling, a deadlock due to
  GFP_KERNEL allocations in readdir, and a RAID10 error handling bug.

  This was tested against both 3.3 and Linus' master as of this morning."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (26 commits)
  Btrfs: reduce lock contention during extent insertion
  Btrfs: avoid deadlocks from GFP_KERNEL allocations during btrfs_real_readdir
  Btrfs: Fix space checking during fs resize
  Btrfs: fix block_rsv and space_info lock ordering
  Btrfs: Prevent root_list corruption
  Btrfs: fix repair code for RAID10
  Btrfs: do not start delalloc inodes during sync
  Btrfs: fix that check_int_data mount option was ignored
  Btrfs: don't count CRC or header errors twice while scrubbing
  Btrfs: fix btrfs_ioctl_dev_info() crash on missing device
  btrfs: don't return EINTR
  Btrfs: double unlock bug in error handling
  Btrfs: always store the mirror we read the eb from
  fs/btrfs/volumes.c: add missing free_fs_devices
  btrfs: fix early abort in 'remount'
  Btrfs: fix max chunk size check in chunk allocator
  Btrfs: add missing read locks in backref.c
  Btrfs: don't call free_extent_buffer twice in iterate_irefs
  Btrfs: Make free_ipath() deal gracefully with NULL pointers
  Btrfs: avoid possible use-after-free in clear_extent_bit()
  ...

12 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sat, 28 Apr 2012 16:28:43 +0000 (09:28 -0700)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "Nothing controversial, just another batch of fixes:

   - Samsung/exynos fixes for more merge window fallout: build errors
     and warnings mostly, but also some clock/device setup issues on
     exynos4/5
   - PXA bug and warning fixes related to gpio and pinmux
   - IRQ domain conversion bugfixes for U300 and MSM
   - A regulator setup fix for U300"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: PXA2xx: MFP: fix potential direction bug
  ARM: PXA2xx: MFP: fix bug with MFP_LPM_KEEP_OUTPUT
  arm/sa1100: fix sa1100-rtc memory resource
  ARM: pxa: fix gpio wakeup setting
  ARM: SAMSUNG: add missing MMC_CAP2_BROKEN_VOLTAGE capability
  ARM: EXYNOS: Fix compilation error when CONFIG_OF is not defined
  ARM: EXYNOS: Fix resource on dev-dwmci.c
  ARM: S3C24XX: Fix build warning for S3C2410_PM
  ARM: mini2440_defconfig: Fix build error
  ARM: msm: Fix gic irqdomain support
  ARM: EXYNOS: Fix incorrect initialization of GIC
  ARM: EXYNOS: use 'exynos4-sdhci' as device name for sdhci controllers
  ARM: u300: bump all IRQ numbers by one
  ARM: ux300: Fix unimplementable regulation constraints

12 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 28 Apr 2012 16:27:07 +0000 (09:27 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "As soon as I sent the non-urgent stack, two important fixes come in:

   - i915: fixes SNB GPU hangs in a number of 3D apps

   - radeon: initial fix for VGA on LLano system, 3 or 4 of us have
     spent time debugging this, and Jerome finally figured out the magic
     bit the BIOS/fglrx set that we didn't.  This at least should get
     things working, there may be future reliability fixes."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/i915: Set the Stencil Cache eviction policy to non-LRA mode.
  drm/radeon/kms: need to set up ss on DP bridges as well

12 years agoRevert "autofs: work around unhappy compat problem on x86-64"
Linus Torvalds [Sat, 28 Apr 2012 15:29:56 +0000 (08:29 -0700)]
Revert "autofs: work around unhappy compat problem on x86-64"

This reverts commit a32744d4abae24572eff7269bc17895c41bd0085.

While that commit was technically the right thing to do, and made the
x86-64 compat mode work identically to native 32-bit mode (and thus
fixing the problem with a 32-bit systemd install on a 64-bit kernel), it
turns out that the automount binaries had workarounds for this compat
problem.

Now, the workarounds are disgusting: doing an "uname()" to find out the
architecture of the kernel, and then comparing it for the 64-bit cases
and fixing up the size of the read() in automount for those.  And they
were confused: it's not actually a generic 64-bit issue at all, it's
very much tied to just x86-64, which has different alignment for an
'u64' in 64-bit mode than in 32-bit mode.

But the end result is that fixing the compat layer actually breaks the
case of a 32-bit automount on a x86-64 kernel.

There are various approaches to fix this (including just doing a
"strcmp()" on current->comm and comparing it to "automount"), but I
think that I will do the one that teaches pipes about a special "packet
mode", which will allow user space to not have to care too deeply about
the padding at the end of the autofs packet.

That change will make the compat workaround unnecessary, so let's revert
it first, and get automount working again in compat mode.  The
packetized pipes will then fix autofs for systemd.

Reported-and-requested-by: Michael Tokarev <mjt@tls.msk.ru>
Cc: Ian Kent <raven@themaw.net>
Cc: stable@kernel.org # for 3.3
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrm/i915: Set the Stencil Cache eviction policy to non-LRA mode.
Kenneth Graunke [Fri, 27 Apr 2012 19:44:41 +0000 (12:44 -0700)]
drm/i915: Set the Stencil Cache eviction policy to non-LRA mode.

Clearing bit 5 of CACHE_MODE_0 is necessary to prevent GPU hangs in
OpenGL programs such as Google MapsGL, Google Earth, and gzdoom when
using separate stencil buffers.  Without it, the GPU tries to use the
LRA eviction policy, which isn't supported.  This was supposed to be off
by default, but seems to be on for many machines.

This cannot be done in gen6_init_clock_gating with most of the other
workaround bits; the render ring needs to exist.  Otherwise, the
register write gets dropped on the floor (one printk will show it
changed, but a second printk immediately following shows the value
reverts to the old one).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=47535
Cc: stable@vger.kernel.org
Cc: Rob Castle <futuredub@gmail.com>
Cc: Eric Appleman <erappleman@gmail.com>
Cc: aaron667@gmx.net
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm/radeon/kms: need to set up ss on DP bridges as well
Alex Deucher [Fri, 27 Apr 2012 21:18:59 +0000 (17:18 -0400)]
drm/radeon/kms: need to set up ss on DP bridges as well

Makes Nutmeg DP to VGA bridges work for me.

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=42490

Noticed by Jerome Glisse (after weeks of debugging).

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agoMerge git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 28 Apr 2012 03:56:54 +0000 (20:56 -0700)]
Merge git://git.samba.org/sfrench/cifs-2.6

Pull CIFS fixes from Steve French.

* git://git.samba.org/sfrench/cifs-2.6:
  Use correct conversion specifiers in cifs_show_options
  CIFS: Show backupuid/gid in /proc/mounts
  cifs: fix offset handling in cifs_iovec_write

12 years agoMerge tag 'stable/for-linus-3.4-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 28 Apr 2012 02:56:22 +0000 (19:56 -0700)]
Merge tag 'stable/for-linus-3.4-rc4-tag' of git://git./linux/kernel/git/konrad/xen

Pull Xen fixes from Konrad Rzeszutek Wilk:
 "Some of these had been in existence since the 2.6.27 days, some since
  3.0 - and some due to new features added in v3.4.

  The one that is most interesting is David's one - in the low-level
  assembler code we had be checking events needlessly.  With his patch
  now we do it when the appropriate flag is set - with the added benefit
  that we can process events faster.  Stefano's is fixing a mistake
  where the Linux IRQ numbers were ACK-ed instead of the Xen IRQ,
  resulting in missing interrupts.  The other ones are bootup related
  that can show up on various hardware."

 - In the low-level assembler code we would jump to check events even if
   none were present.  This incorrect behavior had been there since
   2.6.27 days!
 - When using the fast-path for ACK-ing interrupts we were using the
   Linux IRQ numbers instead of the Xen ones (and they can differ) and
   missing interrupts in process.
 - Fix bootup crashes when ACPI hotplug CPUs were present and they would
   expand past the set number of CPUs we were allocated.
 - Deal with broken BIOSes when uploading C-states to the hypervisor.
 - Disable the cpuid check for MWAIT_LEAF if the ACPI PAD driver is
   loaded.  If the ACPI PAD driver is used it will crash, so lets not
   export the functionality so the ACPI PAD driver won't load.

* tag 'stable/for-linus-3.4-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen: correctly check for pending events when restoring irq flags
  xen/acpi: Workaround broken BIOSes exporting non-existing C-states.
  xen/smp: Fix crash when booting with ACPI hotplug CPUs.
  xen: use the pirq number to check the pirq_eoi_map
  xen/enlighten: Disable MWAIT_LEAF so that acpi-pad won't be loaded.

12 years agoMerge tag 'spi-for-linus' of git://git.secretlab.ca/git/linux-2.6
Linus Torvalds [Sat, 28 Apr 2012 02:52:30 +0000 (19:52 -0700)]
Merge tag 'spi-for-linus' of git://git.secretlab.ca/git/linux-2.6

Pull misc SPI device driver bug fixes from Grant Likely.

* tag 'spi-for-linus' of git://git.secretlab.ca/git/linux-2.6:
  spi/spi-bfin5xx: Fix flush of last bit after each spi transfer
  spi/spi-bfin5xx: fix reversed if condition in interrupt mode
  spi/spi_bfin_sport: drop bits_per_word from client data
  spi/bfin_spi: drop bits_per_word from client data
  spi/spi-bfin-sport: move word length setup to transfer handler
  spi/bfin5xx: rename config macro name for bfin5xx spi controller driver
  spi/pl022: Allow request for higher frequency than maximum possible
  spi/bcm63xx: set master driver mode_bits.
  spi/bcm63xx: don't use the stopping state
  spi/bcm63xx: convert to the pump message infrastructure
  spi/spi-ep93xx.c: use dma_transfer_direction instead of dma_data_direction
  spi: fix spi.h kernel-doc warning
  spi/pl022: Fix calculate_effective_freq()
  spi/pl022: Fix range checking for bits per word

12 years agoMerge tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck...
Linus Torvalds [Sat, 28 Apr 2012 02:50:56 +0000 (19:50 -0700)]
Merge tag 'hwmon-for-linus' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon patches from Guenter Roeck:
 - Fix build warning in ad7314 driver
 - Fix pci_device_id array access in fam15h_power driver, introduced by
   commit 00250ec90963 ("hwmon: fam15h_power: fix bogus values with
   current BIOSes")

* tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (fam15h_power) Fix pci_device_id array
  hwmon: (ad7314) Fix build warning

12 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 28 Apr 2012 02:46:31 +0000 (19:46 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "For your Friday pull request stack, nothing astounding or shattering
  this week some exynos, some intel, some radeon fixes.  One intel fix
  for a regression somwehere back in 2.6.35 land."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/radeon/kms: use frac fb div on APUs
  drm/radeon: add a missing entry to encoder_names
  drm/i915: handle input/output sdvo timings separately in mode_set
  drm/i915: fix integer overflow in i915_gem_do_execbuffer()
  drm/i915: fix integer overflow in i915_gem_execbuffer2()
  drm/exynos: added missed vm area region mapping type.
  drm/exynos: fixed exynos_drm_gem_map_pages bug.
  drm/exynos: fixed duplicatd memory allocation bug.
  drm/i915: fixup load-detect on enabled, but not active pipe

12 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 28 Apr 2012 02:40:56 +0000 (19:40 -0700)]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull RCU fix from Ingo Molnar.

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  rcu: Permit call_rcu() from CPU_DYING notifiers

12 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 28 Apr 2012 02:40:17 +0000 (19:40 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar.

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/apic: Use x2apic physical mode based on FADT setting
  x86/mrst: Quiet sparse noise about plain integer as NULL pointer
  x86, intel_cacheinfo: Fix error return code in amd_set_l3_disable_slot()

12 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 28 Apr 2012 02:37:00 +0000 (19:37 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar.

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched: Fix OOPS when build_sched_domains() percpu allocation fails
  sched: Fix more load-balancing fallout

12 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 28 Apr 2012 02:35:50 +0000 (19:35 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar.

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf: Fix perf_event_for_each() to use sibling
  perf symbols: Read plt symbols from proper symtab_type binary
  tracing: Fix stacktrace of latency tracers (irqsoff and friends)
  perf tools: Add 'G' and 'H' modifiers to event parsing
  tracing: Fix regression with tracing_on
  perf tools: Drop CROSS_COMPILE from flex and bison calls
  perf report: Fix crash showing warning related to kernel maps
  tracing: Fix build breakage without CONFIG_PERF_EVENTS (again)

12 years agoMerge branch 'for-v3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg...
Linus Torvalds [Sat, 28 Apr 2012 02:32:37 +0000 (19:32 -0700)]
Merge branch 'for-v3.4-rc5' of git://git./linux/kernel/git/paulg/linux

Pull build fixes for less mainstream architectures from Paul Gortmaker:
 "These are fixes for frv(1), blackfin(2), powerpc(1) and xtensa(4).

  Fortunately the touches are nearly all specific to files just used by
  the arch in question.  The two touches to shared/common files
  [kernel/irq/debug.h and drivers/pci/Makefile] are trivial to assess as
  no risk to anyone.

  Half of them relate to xtensa directly.  It was only when I fixed the
  last xtensa issue that I realized that the arch has been broken for a
  significant time, and isn't a specific v3.4 regression.  So if you
  wanted, we could leave xtensa lying bleeding in the street for a
  couple more weeks and queue those for 3.5.  But given they are no risk
  to anyone outside of xtensa, I figured to just leave them in.

  If you are OK with taking the xtensa fixes, then please pull to get:

   - one last implicit include uncovered by system.h that is in a file
     specific to just one powerpc defconfig.  (I'd sync'd with BenH).

   - fix an oversight in the PCI makefile where shared code wasn't being
     compiled for ARCH=frv

   - fix a missing include for GPIO in blackfin framebuffer.

   - audit and tag endif in blackfin ezkit board file, in order to find
     and fix the misplaced endif masking a block of code.

   - fix irq/debug.h choice of temporary macro names to be more internal
     so they don't conflict with names used by xtensa.

   - fix a reference to an undeclared local var in xtensa's signal.c

   - fix an implicit bug.h usage in xtensa's asm/io.h uncovered by my
     removing bug.h from kernel.h

   - fix xtensa to properly indicate it is using asm-generic/hardirq.h
     in order to resolve the link error - undefined ack_bad_irq

  The xtensa still fails final link as my latest binutils does something
  evil when ld forward-relocates unlikely() blocks, but in theory people
  who have older/valid toolchains could now use the thing."

* 'for-v3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
  xtensa: fix build fail on undefined ack_bad_irq
  blackfin: fix ifdef fustercluck in mach-bf538/boards/ezkit.c
  blackfin: fix compile error in bfin-lq035q1-fb.c
  pci: frv architecture needs generic setup-bus infrastructure
  irq: hide debug macros so they don't collide with others.
  xtensa: fix build error in xtensa/include/asm/io.h
  xtensa: fix build failure in xtensa/kernel/signal.c
  powerpc: fix system.h fallout in sysdev/scom.c [chroma_defconfig]

12 years agoMerge tag 'sh-for-linus' of git://github.com/pmundt/linux-sh
Linus Torvalds [Sat, 28 Apr 2012 02:31:10 +0000 (19:31 -0700)]
Merge tag 'sh-for-linus' of git://github.com/pmundt/linux-sh

Pull SuperH fixes from Paul Mundt.

* tag 'sh-for-linus' of git://github.com/pmundt/linux-sh:
  sh: Fix up tracepoint build fallout from static key introduction.
  sh: Fix error synchronising kernel page tables

12 years agoMerge branch 'docs-3.4' of git://git.samba.org/jlayton/linux
Linus Torvalds [Sat, 28 Apr 2012 02:27:26 +0000 (19:27 -0700)]
Merge branch 'docs-3.4' of git://git.samba.org/jlayton/linux

Pull security key doc update from Jeff Layton:
 "Ordinarily, I send my patches through others' trees, but David
  suggested I just send this one to you directly since it's just a
  Documentation/ update"

* 'docs-3.4' of git://git.samba.org/jlayton/linux:
  keys: update the documentation with info about "logon" keys

12 years agoxen: correctly check for pending events when restoring irq flags
David Vrabel [Thu, 26 Apr 2012 18:44:06 +0000 (19:44 +0100)]
xen: correctly check for pending events when restoring irq flags

In xen_restore_fl_direct(), xen_force_evtchn_callback() was being
called even if no events were pending.  This resulted in (depending on
workload) about a 100 times as many xen_version hypercalls as
necessary.

Fix this by correcting the sense of the conditional jump.

This seems to give a significant performance benefit for some
workloads.

There is some subtle tricksy "..since the check here is trying to
check both pending and masked in a single cmpw, but I think this is
correct. It will call check_events now only when the combined
mask+pending word is 0x0001 (aka unmasked, pending)." (Ian)

CC: stable@kernel.org
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoBtrfs: reduce lock contention during extent insertion
Chris Mason [Fri, 27 Apr 2012 18:31:29 +0000 (14:31 -0400)]
Btrfs: reduce lock contention during extent insertion

We're spending huge amounts of time on lock contention during
end_io processing because we unconditionally assume we are overwriting
an existing extent in the file for each IO.

This checks to see if we are outside i_size, and if so, it uses a
less expensive readonly search of the btree to look for existing
extents.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: avoid deadlocks from GFP_KERNEL allocations during btrfs_real_readdir
Chris Mason [Fri, 27 Apr 2012 18:23:22 +0000 (14:23 -0400)]
Btrfs: avoid deadlocks from GFP_KERNEL allocations during btrfs_real_readdir

Btrfs has an optimization where it will preallocate dentries during
readdir to fill in enough information to open the inode without an extra
lookup.

But, we're calling d_alloc, which is doing GFP_KERNEL allocations, and
that leads to deadlocks because our readdir code has tree locks held.

For now, disable this optimization.  We'll fix the gfp mask in the next
merge window.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agospi/spi-bfin5xx: Fix flush of last bit after each spi transfer
Scott Jiang [Mon, 23 Apr 2012 22:18:13 +0000 (18:18 -0400)]
spi/spi-bfin5xx: Fix flush of last bit after each spi transfer

This patch ensures that the last bit of a transfer gets correctly
flushed out of the register.

Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/spi-bfin5xx: fix reversed if condition in interrupt mode
Scott Jiang [Mon, 23 Apr 2012 22:18:12 +0000 (18:18 -0400)]
spi/spi-bfin5xx: fix reversed if condition in interrupt mode

This condition is used to determine 8 bits or 16 and 32 bits transfer.
Obviously it is reversed.

Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/spi_bfin_sport: drop bits_per_word from client data
Scott Jiang [Mon, 23 Apr 2012 22:18:11 +0000 (18:18 -0400)]
spi/spi_bfin_sport: drop bits_per_word from client data

Since the member was dropped from the common Blackfin header, we need
to stop using it in the SPORT driver too.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/bfin_spi: drop bits_per_word from client data
Scott Jiang [Mon, 23 Apr 2012 22:18:10 +0000 (18:18 -0400)]
spi/bfin_spi: drop bits_per_word from client data

No other SPI controller has this field, and SPI clients should be setting
this up in their own drivers.  So drop it from the Blackfin controller to
keep people from using it.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/spi-bfin-sport: move word length setup to transfer handler
Scott Jiang [Mon, 23 Apr 2012 22:18:09 +0000 (18:18 -0400)]
spi/spi-bfin-sport: move word length setup to transfer handler

Each transfer may have its own bits per word.

Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/bfin5xx: rename config macro name for bfin5xx spi controller driver
Scott Jiang [Mon, 23 Apr 2012 22:18:08 +0000 (18:18 -0400)]
spi/bfin5xx: rename config macro name for bfin5xx spi controller driver

This controller is only for blackfin 5xx soc, so rename it to BFIN5XX

Signed-off-by: Scott Jiang <scott.jiang.linux@gmail.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/pl022: Allow request for higher frequency than maximum possible
Viresh Kumar [Thu, 19 Apr 2012 06:18:15 +0000 (11:48 +0530)]
spi/pl022: Allow request for higher frequency than maximum possible

Currently, if we request for frequency greater than maximum possible, spi driver
returns error.

For example, if the spi block src frequency is 333/4 MHz, i.e. 83.33.. MHz,
maximum frequency programmable would be src/2. Which would come around 41.6...

It is difficult to pass frequency in these figures. We normally try to program
in round figures, like 42 MHz and it should get programmed to <=
requested_frequency, i.e. 41.6...

For this to happen, we must not return error even if requested freq is higher
than max possible. But should program it to max possible.

Reported-by: Vinit Kamalaksha Shenoy <vinit.shenoy@st.com>
Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agoBtrfs: Fix space checking during fs resize
Daniel J Blueman [Fri, 27 Apr 2012 16:41:46 +0000 (12:41 -0400)]
Btrfs: Fix space checking during fs resize

Fix out-of-space checking, addressing a warning and potential resource
leak when resizing the filesystem down while allocating blocks.

Signed-off-by: Daniel J Blueman <daniel@quora.org>
Reviewed-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: fix block_rsv and space_info lock ordering
Stefan Behrens [Fri, 27 Apr 2012 16:41:46 +0000 (12:41 -0400)]
Btrfs: fix block_rsv and space_info lock ordering

may_commit_transaction() calls
        spin_lock(&space_info->lock);
        spin_lock(&delayed_rsv->lock);
and update_global_block_rsv() calls
        spin_lock(&block_rsv->lock);
        spin_lock(&sinfo->lock);

Lockdep complains about this at run time.
Everywhere except in update_global_block_rsv(), the space_info lock is
the outer lock, therefore the locking order in update_global_block_rsv()
is changed.

Signed-off-by: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: Prevent root_list corruption
Daniel J Blueman [Fri, 27 Apr 2012 16:41:46 +0000 (12:41 -0400)]
Btrfs: Prevent root_list corruption

I was seeing root_list corruption on unmount during fs resize in 3.4-rc4; add
correct locking to address this.

Signed-off-by: Daniel J Blueman <daniel@quora.org>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: fix repair code for RAID10
Jan Schmidt [Fri, 27 Apr 2012 16:41:45 +0000 (12:41 -0400)]
Btrfs: fix repair code for RAID10

btrfs_map_block sets mirror_num, so that the repair code knows eventually
which device gave us the read error. For RAID10, mirror_num must be 1 or 2.
Before this fix mirror_num was incorrectly related to our stripe index.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agoBtrfs: do not start delalloc inodes during sync
Josef Bacik [Tue, 24 Apr 2012 00:35:03 +0000 (20:35 -0400)]
Btrfs: do not start delalloc inodes during sync

btrfs_start_delalloc_inodes will just walk the list of delalloc inodes and
start writing them out, but it doesn't splice the list or anything so as
long as somebody is doing work on the box you could end up in this section
_forever_.  So just remove it, it's not needed anyway since sync will start
writeback on all inodes anyway, all we need to do is wait for ordered
extents and then we can commit the transaction.  In my horrible torture test
sync goes from taking 4 minutes to about 1.5 minutes.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
12 years agospi/bcm63xx: set master driver mode_bits.
Florian Fainelli [Fri, 20 Apr 2012 13:37:35 +0000 (15:37 +0200)]
spi/bcm63xx: set master driver mode_bits.

We were not properly advertising the MODE bits supported by this driver, fix
that.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/bcm63xx: don't use the stopping state
Florian Fainelli [Fri, 20 Apr 2012 13:37:34 +0000 (15:37 +0200)]
spi/bcm63xx: don't use the stopping state

We do not need to use a flag to indicate if the master driver is stopping
it is sufficient to perform spi master unregistering in the platform
driver's remove function.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/bcm63xx: convert to the pump message infrastructure
Florian Fainelli [Fri, 20 Apr 2012 13:37:33 +0000 (15:37 +0200)]
spi/bcm63xx: convert to the pump message infrastructure

This patch converts the bcm63xx SPI driver to use the SPI infrastructure
pump message queue. Since we were previously sleeping in the SPI
driver's transfer() function (which is not allowed) this is now fixed as well.

To complete that conversion a certain number of changes have been made:
- the transfer len is split into multiple hardware transfers in case its
  size is bigger than the hardware FIFO size
- the FIFO refill is no longer done in the interrupt context, which was a
  bad idea leading to quick interrupt handler re-entrancy

Tested-by: Tanguy Bouzeloc <tanguy.bouzeloc@efixo.com>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/spi-ep93xx.c: use dma_transfer_direction instead of dma_data_direction
H Hartley Sweeten [Wed, 18 Apr 2012 01:46:36 +0000 (18:46 -0700)]
spi/spi-ep93xx.c: use dma_transfer_direction instead of dma_data_direction

A new enum indicating the dma channel direction was introduced by:

commit 49920bc66984a512f4bcc7735a61642cd0e4d6f2
    dmaengine: add new enum dma_transfer_direction

The following commit changed spi-ep93xx to use the new enum:

commit a485df4b4404379786c4bdd258bc528b2617449d
    spi, serial: move to dma_transfer_direction

In doing so a sparse warning was introduced:

warning: mixing different enum types
   int enum dma_data_direction  versus
   int enum dma_transfer_direction

This is produced because the 'dir' passed in ep93xx_spi_dma_prepare
is an enum dma_data_direction and is being used to set the
dma_slave_config 'direction' which is now an enum dma_transfer_direction.

Fix this by converting spi-ep93xx to use the new enum type in all
places.

Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Acked-by: Mika Westerberg <mika.westerberg@iki.fi>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi: fix spi.h kernel-doc warning
Randy Dunlap [Wed, 18 Apr 2012 00:03:50 +0000 (17:03 -0700)]
spi: fix spi.h kernel-doc warning

Fix kernel-doc warning in spi.h (copy/paste):

Warning(include/linux/spi/spi.h:365): No description found for parameter 'unprepare_transfer_hardware'

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agospi/pl022: Fix calculate_effective_freq()
Viresh Kumar [Thu, 19 Apr 2012 09:14:21 +0000 (14:44 +0530)]
spi/pl022: Fix calculate_effective_freq()

calculate_effective_freq() was still not optimized and there were cases when it
returned without error and with values of cpsr and scr as zero.

Also, the variable named found is not used well.

This patch targets to optimize and correct this routine. Tested for SPEAr.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Tested-by: Vinit Kamalaksha Shenoy <vinit.shenoy@st.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
12 years agoMerge branch 'v3.4-samsung-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel...
Olof Johansson [Fri, 27 Apr 2012 16:40:06 +0000 (09:40 -0700)]
Merge branch 'v3.4-samsung-fixes-3' of git://git./linux/kernel/git/kgene/linux-samsung into fixes

* 'v3.4-samsung-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
  ARM: SAMSUNG: add missing MMC_CAP2_BROKEN_VOLTAGE capability
  ARM: EXYNOS: Fix compilation error when CONFIG_OF is not defined
  ARM: EXYNOS: Fix resource on dev-dwmci.c
  ARM: S3C24XX: Fix build warning for S3C2410_PM
  ARM: mini2440_defconfig: Fix build error
  ARM: EXYNOS: Fix incorrect initialization of GIC
  ARM: EXYNOS: use 'exynos4-sdhci' as device name for sdhci controllers

12 years agoMerge branch 'fixes' of git://github.com/hzhuang1/linux into fixes
Olof Johansson [Fri, 27 Apr 2012 16:31:02 +0000 (09:31 -0700)]
Merge branch 'fixes' of git://github.com/hzhuang1/linux into fixes

* 'fixes' of git://github.com/hzhuang1/linux:
  ARM: PXA2xx: MFP: fix potential direction bug
  ARM: PXA2xx: MFP: fix bug with MFP_LPM_KEEP_OUTPUT
  arm/sa1100: fix sa1100-rtc memory resource
  ARM: pxa: fix gpio wakeup setting

12 years agoMerge tag 'fixes-for-v3.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/balbi...
Greg Kroah-Hartman [Fri, 27 Apr 2012 14:50:34 +0000 (07:50 -0700)]
Merge tag 'fixes-for-v3.4-rc5' of git://git./linux/kernel/git/balbi/usb into usb-linus

usb: fixes for v3.4-rc cycle

A few more fixes for v3.4-rc cycle.

It includes a couple of fixes to the ordering of the methods in udc-core.c.
Without these two patches, we will have issues when either unregistering a
gadget driver (triggered with dummy_hcd only) or issuing a device-initiated
disconnect through sysfs.

There's also a fix on dummy_hcd to not call ->pullup() from udc_stop() because
udc-core.c already handles that.

A fix to MUSB as promised, to kill the compile warnings regarding deprecated
interfaces. We are essentially dropping the __deprecated flag because it
doesn't look like we will ever be able to live without it when we consider the
amount of silicon issues we find on different MUSB instantiations.

A couple of other fixes are also available, one adding the missing transceiver
events to gpio_vbus and another adding a missing unregister call to MUSB's
davinci glue layer.

12 years agousb: gadget: udc-core: fix incompatibility with dummy-hcd
Alan Stern [Thu, 26 Apr 2012 15:31:57 +0000 (11:31 -0400)]
usb: gadget: udc-core: fix incompatibility with dummy-hcd

This patch (as1548) fixes a recently-introduced incompatibility
between the UDC core and the dummy-hcd driver.  Commit
8ae8090c82eb407267001f75b3d256b3bd4ae691 (usb: gadget: udc-core: fix
asymmetric calls in remove_driver) moved the usb_gadget_udc_stop()
call in usb_gadget_remove_driver() below the usb_gadget_disconnect()
call.

As a result, usb_gadget_disconnect() gets called at a time when the
gadget driver believes it has been unbound but dummy-hcd believes
it has not.  A nasty error ensues when dummy-hcd calls the gadget
driver's disconnect method a second time.

To fix the problem, this patch moves the gadget driver's unbind
notification after the usb_gadget_disconnect() call.  Now nothing
happens between the two unbind notifications, so nothing goes wrong.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
12 years agousb: gadget: udc-core: fix wrong call order
Felipe Balbi [Fri, 27 Apr 2012 08:02:15 +0000 (11:02 +0300)]
usb: gadget: udc-core: fix wrong call order

commit 6d258a4 (usb: gadget: udc-core: stop UDC on device-initiated
disconnect) introduced another case of asymmetric calls when issuing
a device-initiated disconnect. Fix it.

Reported-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Felipe Balbi <balbi@ti.com>
12 years agodrm/radeon/kms: use frac fb div on APUs
Alex Deucher [Thu, 19 Apr 2012 14:48:38 +0000 (10:48 -0400)]
drm/radeon/kms: use frac fb div on APUs

Seems to be more stable on certain monitors.

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=48880

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm/radeon: add a missing entry to encoder_names
Ilija Hadzic [Thu, 19 Apr 2012 16:22:20 +0000 (12:22 -0400)]
drm/radeon: add a missing entry to encoder_names

An entry for INTERNAL_VCE encoder was missing. Add it.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agoMerge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel...
Dave Airlie [Fri, 27 Apr 2012 07:21:28 +0000 (08:21 +0100)]
Merge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel into drm-fixes

From Daniel Vetter

- VGA load-detect fix. This bug seems to be as old as the load-detect code
  (2.6.30), but needs stupid userspace (upowerd trying to detect
  connectors on dpms-off outputs) to actually kill the machine. And
  obviously a machine without VGA-hotplug, otherwise we don't do load
  detect.
- 2 interger overflow fixes for unpriviledged ioctls from Xi Wang.
- Fix SDVO regression for low-res (pixelclock < 100MHz) digital outputs,
 introduce in 2.6.36.

* 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel:
  drm/i915: handle input/output sdvo timings separately in mode_set
  drm/i915: fix integer overflow in i915_gem_do_execbuffer()
  drm/i915: fix integer overflow in i915_gem_execbuffer2()
  drm/i915: fixup load-detect on enabled, but not active pipe

12 years agoMerge branch 'exynos-drm-fixes' of git://git.infradead.org/users/kmpark/linux-2.6...
Dave Airlie [Fri, 27 Apr 2012 07:20:27 +0000 (08:20 +0100)]
Merge branch 'exynos-drm-fixes' of git://git.infradead.org/users/kmpark/linux-2.6-samsung into drm-fixes

From Inki Dae:

this patch set fixes gem allocation and mapping issue between user space and
physical memory region.

* 'exynos-drm-fixes' of git://git.infradead.org/users/kmpark/linux-2.6-samsung:
  drm/exynos: added missed vm area region mapping type.
  drm/exynos: fixed exynos_drm_gem_map_pages bug.
  drm/exynos: fixed duplicatd memory allocation bug.

12 years agoARM: PXA2xx: MFP: fix potential direction bug
Igor Grinberg [Thu, 12 Apr 2012 12:43:29 +0000 (15:43 +0300)]
ARM: PXA2xx: MFP: fix potential direction bug

Pins configured as input and have MFP_LPM_DRIVE_* flag set, can have a
wrong output value for some period of time (spike) during the suspend
sequence.
This can happen because the direction of the pins (GPDR) is set by
software and the output level is set by hardware (PGSR) at a later
stage.

Fix the above potential bug by setting the output levels first.
Also save the actual levels of the pins before the suspend and restore
them after the resume, but before the direction settings take place, so
the same bug as described above will not happen in the resume sequence.

Reported-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Igor Grinberg <grinberg@compulab.co.il>
Tested-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@gmail.com>
12 years agoARM: PXA2xx: MFP: fix bug with MFP_LPM_KEEP_OUTPUT
Igor Grinberg [Thu, 12 Apr 2012 12:43:28 +0000 (15:43 +0300)]
ARM: PXA2xx: MFP: fix bug with MFP_LPM_KEEP_OUTPUT

Pins that have MFP_LPM_KEEP_OUTPUT set and are configured for output
must retain the output state in low power mode.
Currently, the pin direction configuration is overrided with values
in gpdr_lpm[] array and do not obey the MFP_LPM_KEEP_OUTPUT setting.

Fix the above bug and add some documentation to clarify the
MFP_LPM_KEEP_OUTPUT setting purpose.

Reported-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Igor Grinberg <grinberg@compulab.co.il>
Tested-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@gmail.com>
12 years agoarm/sa1100: fix sa1100-rtc memory resource
Dmitry Artamonow [Sat, 14 Apr 2012 06:26:19 +0000 (10:26 +0400)]
arm/sa1100: fix sa1100-rtc memory resource

DEFINE_RES_MEM() takes the size of resource as a second argument,
not the end address. Passing end address leads to following error
in runtime during device registration:
sa1100-rtc: failed to claim resource 0

Fix it.

Signed-off-by: Dmitry Artamonow <mad_soft@inbox.ru>
Signed-off-by: Haojian Zhuang <haojian.zhuang@gmail.com>
12 years agoARM: pxa: fix gpio wakeup setting
Robert Jarzmik [Sun, 22 Apr 2012 11:37:24 +0000 (13:37 +0200)]
ARM: pxa: fix gpio wakeup setting

In 3.3, gpio wakeup setting was broken. The call
enable_irq_wake() didn't set up the PXA gpio registers
(PWER, ...) anymore.

Fix it at least for pxa27x. The driver doesn't seem to be
used in pxa25x (weird ...), and the fix doesn't extend to
pxa3xx and pxa95x (which don't have a gpio_set_wake()
available).

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Haojian Zhuang <haojian.zhuang@gmail.com>
12 years agoxen/acpi: Workaround broken BIOSes exporting non-existing C-states.
Konrad Rzeszutek Wilk [Thu, 26 Apr 2012 18:22:33 +0000 (14:22 -0400)]
xen/acpi: Workaround broken BIOSes exporting non-existing C-states.

We did a similar check for the P-states but did not do it for
the C-states. What we want to do is ignore cases where the DSDT
has definition for sixteen CPUs, but the machine only has eight
CPUs and we get:
xen-acpi-processor: (CX): Hypervisor error (-22) for ACPI CPU14

Reported-by: Tobias Geiger <tobias.geiger@vido.info>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen/smp: Fix crash when booting with ACPI hotplug CPUs.
Konrad Rzeszutek Wilk [Thu, 26 Apr 2012 17:50:03 +0000 (13:50 -0400)]
xen/smp: Fix crash when booting with ACPI hotplug CPUs.

When we boot on a machine that can hotplug CPUs and we
are using 'dom0_max_vcpus=X' on the Xen hypervisor line
to clip the amount of CPUs available to the initial domain,
we get this:

(XEN) Command line: com1=115200,8n1 dom0_mem=8G noreboot dom0_max_vcpus=8 sync_console mce_verbosity=verbose console=com1,vga loglvl=all guest_loglvl=all
.. snip..
DMI: Intel Corporation S2600CP/S2600CP, BIOS SE5C600.86B.99.99.x032.072520111118 07/25/2011
.. snip.
SMP: Allowing 64 CPUs, 32 hotplug CPUs
installing Xen timer for CPU 7
cpu 7 spinlock event irq 361
NMI watchdog: disabled (cpu7): hardware events not enabled
Brought up 8 CPUs
.. snip..
[acpi processor finds the CPUs are not initialized and starts calling
arch_register_cpu, which creates /sys/devices/system/cpu/cpu8/online]
CPU 8 got hotplugged
CPU 9 got hotplugged
CPU 10 got hotplugged
.. snip..
initcall 1_acpi_battery_init_async+0x0/0x1b returned 0 after 406 usecs
calling  erst_init+0x0/0x2bb @ 1

[and the scheduler sticks newly started tasks on the new CPUs, but
said CPUs cannot be initialized b/c the hypervisor has limited the
amount of vCPUS to 8 - as per the dom0_max_vcpus=8 flag.
The spinlock tries to kick the other CPU, but the structure for that
is not initialized and we crash.]
BUG: unable to handle kernel paging request at fffffffffffffed8
IP: [<ffffffff81035289>] xen_spin_lock+0x29/0x60
PGD 180d067 PUD 180e067 PMD 0
Oops: 0002 [#1] SMP
CPU 7
Modules linked in:

Pid: 1, comm: swapper/0 Not tainted 3.4.0-rc2upstream-00001-gf5154e8 #1 Intel Corporation S2600CP/S2600CP
RIP: e030:[<ffffffff81035289>]  [<ffffffff81035289>] xen_spin_lock+0x29/0x60
RSP: e02b:ffff8801fb9b3a70  EFLAGS: 00010282

With this patch, we cap the amount of vCPUS that the initial domain
can run, to exactly what dom0_max_vcpus=X has specified.

In the future, if there is a hypercall that will allow a running
domain to expand past its initial set of vCPUS, this patch should
be re-evaluated.

CC: stable@kernel.org
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>