openwrt/staging/blogic.git
13 years agomd/raid1: Mark device want_replacement when we see a write error.
NeilBrown [Thu, 22 Dec 2011 23:17:57 +0000 (10:17 +1100)]
md/raid1: Mark device want_replacement when we see a write error.

Now that WantReplacement drives are replaced cleanly, mark a drive
as want_replacement when we see a write error.  It might get failed soon so
the WantReplacement flag is irrelevant, but if the write error is recorded
in the bad block log, we still want to activate any spare that might
be available.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid1: If there is a spare and a want_replacement device, start replacement.
NeilBrown [Thu, 22 Dec 2011 23:17:57 +0000 (10:17 +1100)]
md/raid1: If there is a spare and a want_replacement device, start replacement.

When attempting to add a spare to a RAID1 array, also consider
adding it as a replacement for a want_replacement device.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid1: recognise replacements when assembling arrays.
NeilBrown [Thu, 22 Dec 2011 23:17:57 +0000 (10:17 +1100)]
md/raid1: recognise replacements when assembling arrays.

If a Replacement is seen, file it as such.

If we see two replacements (or two normal devices) for the one slot,
abort.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid1: handle activation of replacement device when recovery completes.
NeilBrown [Thu, 22 Dec 2011 23:17:57 +0000 (10:17 +1100)]
md/raid1: handle activation of replacement device when recovery completes.

When recovery completes ->spare_active is called.
This checks if the replacement is ready and if so it fails
the original.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid1: Allow a failed replacement device to be removed.
NeilBrown [Thu, 22 Dec 2011 23:17:56 +0000 (10:17 +1100)]
md/raid1: Allow a failed replacement device to be removed.

Replacement devices are stored at a different offset, so look
there too.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid1: Allocate spare to store replacement devices and their bios.
NeilBrown [Thu, 22 Dec 2011 23:17:56 +0000 (10:17 +1100)]
md/raid1: Allocate spare to store replacement devices and their bios.

In RAID1, a replacement is much like a normal device, so we just
double the size of the relevant arrays and look at all possible
devices for reads and writes.

This means that the array looks like it is now double the size in some
way - we need to be careful about that.
In particular, we checking if the array is still degraded while
creating a recovery request we need to only consider the first 'half'
- i.e. the real (non-replacement) devices.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid1: Replace use of mddev->raid_disks with conf->raid_disks.
NeilBrown [Thu, 22 Dec 2011 23:17:56 +0000 (10:17 +1100)]
md/raid1:  Replace use of mddev->raid_disks with conf->raid_disks.

In general mddev->raid_disks can change unexpectedly while
conf->raid_disks will only change in a very controlled way.  So change
some uses of one to the other.

The use of mddev->raid_disks will not cause actually problems but
this way is more consistent and safer in the long term.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: If there is a spare and a want_replacement device, start replacement.
NeilBrown [Thu, 22 Dec 2011 23:17:56 +0000 (10:17 +1100)]
md/raid10: If there is a spare and a want_replacement device, start replacement.

When attempting to add a spare to a RAID10 array, also consider
adding it as a replacement for a want_replacement device.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: recognise replacements when assembling array.
NeilBrown [Thu, 22 Dec 2011 23:17:55 +0000 (10:17 +1100)]
md/raid10: recognise replacements when assembling array.

If a Replacement is seen, file it as such.

If we see two replacements (or two normal devices) for the one slot,
abort.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: Allow replacement device to be replace old drive.
NeilBrown [Thu, 22 Dec 2011 23:17:55 +0000 (10:17 +1100)]
md/raid10: Allow replacement device to be replace old drive.

When recovery finish and spare_active is called, check for a
replace that might have just become fully synced and mark it
as such, marking the original as failed.

Then when the original is removed, move the replacement into
its position.

This means that 'replacement' and spontaneously become NULL in some
situations.  Make sure we check for those.
It also means that 'rdev' and 'replacement' could appear to be
identical - check for that too.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: handle recovery of replacement devices.
NeilBrown [Thu, 22 Dec 2011 23:17:55 +0000 (10:17 +1100)]
md/raid10: handle recovery of replacement devices.

If there is a replacement device, then recover to it,
reading from any drives - maybe the one being replaced, maybe not.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: Handle replacement devices during resync.
NeilBrown [Thu, 22 Dec 2011 23:17:55 +0000 (10:17 +1100)]
md/raid10:  Handle replacement devices during resync.

If we need to resync an array which has replacement devices,
we always write any block checked to every replacement.

If the resync was bitmap-based resync we will then complete the
replacement normally.
If it was a full resync, we mark the replacements as fully recovered
when the resync finishes so no further recovery is needed.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: writes should get directed to replacement as well as original.
NeilBrown [Thu, 22 Dec 2011 23:17:55 +0000 (10:17 +1100)]
md/raid10: writes should get directed to replacement as well as original.

When writing, we need to submit two writes, one to the original,
and one to the replacements - if there is a replacement.

If the write to the replacement results in a write error we just
fail the device.  We only try to record write errors to the
original.

This only handles writing new data.  Writing for resync/recovery
will come later.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: allow removal of failed replacement devices.
NeilBrown [Thu, 22 Dec 2011 23:17:54 +0000 (10:17 +1100)]
md/raid10: allow removal of failed replacement devices.

Enhance raid10_remove_disk to be able to remove ->replacement
as well as ->rdev

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: preferentially read from replacement device if possible.
NeilBrown [Thu, 22 Dec 2011 23:17:54 +0000 (10:17 +1100)]
md/raid10: preferentially read from replacement device if possible.

When reading (for array reads, not for recovery etc) we read from the
replacement device if it has recovered far enough.
This requires storing the chosen rdev in the 'r10_bio' so we can make
sure to drop the ref on the right device when the read finishes.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: change read_balance to return an rdev
NeilBrown [Thu, 22 Dec 2011 23:17:54 +0000 (10:17 +1100)]
md/raid10:  change read_balance to return an rdev

It makes more sense to return an rdev than just an index as
read_balance() gets a reference to the rdev and so returning
the pointer make this more idiomatic.

This will be needed in a future patch when we might return
a 'replacement' rdev instead of the main rdev.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid10: prepare data structures for handling replacement.
NeilBrown [Thu, 22 Dec 2011 23:17:54 +0000 (10:17 +1100)]
md/raid10: prepare data structures for handling replacement.

Allow each slot in the RAID10 to have 2 devices, the want_replacement
and the replacement.

Also an r10bio to have 2 bios, and for resync/recovery allocate the
second bio if there are any replacement devices.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: Mark device want_replacement when we see a write error.
NeilBrown [Thu, 22 Dec 2011 23:17:54 +0000 (10:17 +1100)]
md/raid5: Mark device want_replacement when we see a write error.

Now that WantReplacement drives are replaced cleanly, mark a drive
as WantReplacement when we see a write error.  It might get failed soon so
the WantReplacement flag is irrelevant, but if the write error is recorded
in the bad block log, we still want to activate any spare that might
be available.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: If there is a spare and a want_replacement device, start replacement.
NeilBrown [Thu, 22 Dec 2011 23:17:53 +0000 (10:17 +1100)]
md/raid5: If there is a spare and a want_replacement device, start replacement.

When attempting to add a spare to a RAID[456] array, also consider
adding it as a replacement for a want_replacement device.

This requires that common md code attempt hot_add even when the array
is not formally degraded.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: recognise replacements when assembling array.
NeilBrown [Thu, 22 Dec 2011 23:17:53 +0000 (10:17 +1100)]
md/raid5: recognise replacements when assembling array.

If a Replacement is seen, file it as such.

If we see two replacements (or two normal devices) for the one slot,
abort.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: handle activation of replacement device when recovery completes.
NeilBrown [Thu, 22 Dec 2011 23:17:53 +0000 (10:17 +1100)]
md/raid5: handle activation of replacement device when recovery completes.

When recovery completes - as reported by a call to ->spare_active,
we clear In_sync on the original and set it on the replacement.

Then when the original gets removed we move the replacement from
'replacement' to 'rdev'.

This could race with other code that is looking at these pointers,
so we use memory barriers and careful ordering to ensure that
a reader might see one device twice, but never no devices.
Then the readers guard against using both devices, which could
only happen when writing.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: detect and handle replacements during recovery.
NeilBrown [Thu, 22 Dec 2011 23:17:53 +0000 (10:17 +1100)]
md/raid5:  detect and handle replacements during recovery.

During recovery we want to write to the replacement but not
the original.  So we have two new flags
 - R5_NeedReplace if this stripe has a replacement that needs to
   be written at some stage
 - R5_WantReplace if NeedReplace, and the data is available, and
   a 'sync' has been requested on this stripe.

We also distinguish between 'sync and replace' which need to read
all other devices, and 'replace' which only needs to read the
devices being replaced.

Note that during resync we always write to any replacement device.
It might not need to be written to, but as we don't read to compare,
we have to write to be sure.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: writes should get directed to replacement as well as original.
NeilBrown [Thu, 22 Dec 2011 23:17:53 +0000 (10:17 +1100)]
md/raid5: writes should get directed to replacement as well as original.

When writing, we need to submit two writes, one to the original, and
one to the replacement - if there is a replacement.

If the write to the replacement results in a write error, we just fail
the device.  We only try to record write errors to the original.

When writing for recovery, we shouldn't write to the original.  This
will be addressed in a subsequent patch that generally addresses
recovery.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: allow removal for failed replacement devices.
NeilBrown [Thu, 22 Dec 2011 23:17:52 +0000 (10:17 +1100)]
md/raid5: allow removal for failed replacement devices.

Enhance raid5_remove_disk to be able to remove ->replacement
as well as ->rdev.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: preferentially read from replacement device if possible.
NeilBrown [Thu, 22 Dec 2011 23:17:52 +0000 (10:17 +1100)]
md/raid5: preferentially read from replacement device if possible.

If a replacement device is present and has been recovered far enough,
then use it for reading into the stripe cache.

If we get an error we don't try to repair it, we just fail the device.
A replacement device that gives errors does not sound sensible.

This requires removing the setting of R5_ReadError when we get
a read error during a read that bypasses the cache.  It was probably
a bad idea anyway as we don't know that every block in the read
caused an error, and it could cause ReadError to be set for the
replacement device, which is bad.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: remove redundant bio initialisations.
NeilBrown [Thu, 22 Dec 2011 23:17:52 +0000 (10:17 +1100)]
md/raid5: remove redundant bio initialisations.

We current initialise some fields of a bio when preparing a
stripe_head, and again just before submitting the request.

Remove the duplication by only setting the fields that lower level
devices don't touch in raid5_build_block, and only set the changeable
fields in ops_run_io.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: raid5.h cleanup
NeilBrown [Thu, 22 Dec 2011 23:17:52 +0000 (10:17 +1100)]
md/raid5: raid5.h cleanup

Remove some #defines that are no longer used, and replace some
others with an enum.
And remove an unused field.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: allow each slot to have an extra replacement device
NeilBrown [Thu, 22 Dec 2011 23:17:52 +0000 (10:17 +1100)]
md/raid5: allow each slot to have an extra replacement device

Just enhance data structures to record a second device per slot to be
used as a 'replacement' device, replacing the original.
We also have a second bio in each slot in each stripe_head.  This will
only be used when writing to the array - we need to write to both the
original and the replacement at the same time, so will need two bios.

For now, only try using the replacement drive for aligned-reads.
In this case, we prefer the replacement if it has been recovered far
enough, otherwise use the original.

This includes a small enhancement.  Previously we would only do
aligned reads if the target device was fully recovered.  Now we also
do them if it has recovered far enough.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: create externally visible flags for supporting hot-replace.
NeilBrown [Thu, 22 Dec 2011 23:17:51 +0000 (10:17 +1100)]
md: create externally visible flags for supporting hot-replace.

hot-replace is a feature being added to md which will allow a
device to be replaced without removing it from the array first.

With hot-replace a spare can be activated and recovery can start while
the original device is still in place, thus allowing a transition from
an unreliable device to a reliable device without leaving the array
degraded during the transition.  It can also be use when the original
device is still reliable but it not wanted for some reason.

This will eventually be supported in RAID4/5/6 and RAID10.

This patch adds a super-block flag to distinguish the replacement
device.  If an old kernel sees this flag it will reject the device.

It also adds two per-device flags which are viewable and settable via
sysfs.
   "want_replacement" can be set to request that a device be replaced.
   "replacement" is set to show that this device is replacing another
   device.

The "rd%d" links in /sys/block/mdXx/md only apply to the original
device, not the replacement.  We currently don't make links for the
replacement - there doesn't seem to be a need.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: change hot_remove_disk to take an rdev rather than a number.
NeilBrown [Thu, 22 Dec 2011 23:17:51 +0000 (10:17 +1100)]
md: change hot_remove_disk to take an rdev rather than a number.

Soon an array will be able to have multiple devices with the
same raid_disk number (an original and a replacement).  So removing
a device based on the number won't work.  So pass the actual device
handle instead.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: remove test for duplicate device when setting slot number.
NeilBrown [Thu, 22 Dec 2011 23:17:51 +0000 (10:17 +1100)]
md: remove test for duplicate device when setting slot number.

When setting the slot number on a device in an active array we
currently check that the number is not already in use.
We then call into the personality's hot_add_disk function
which performs the same test and returns the same error.

Thus the common test is not needed.

As we will shortly be changing some personalities to allow duplicates
in some cases (to support hot-replace), the common test will become
inconvenient.

So remove the common test.

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/bitmap: be more consistent when setting new bits in memory bitmap.
NeilBrown [Thu, 22 Dec 2011 23:17:51 +0000 (10:17 +1100)]
md/bitmap: be more consistent when setting new bits in memory bitmap.

For each active region corresponding to a bit in the bitmap with have
a 14bit counter (and some flags).
This counts
   number of active writes + bit in the on-disk bitmap + delay-needed.

The "delay-needed" is because we always want a delay before clearing a
bit.  So the number here is normally number of active writes plus 2.
If there have been no writes for a while, we drop to 1.
If still no writes we clear the bit and drop to 0.

So for consistency, when setting bit from the on-disk bitmap or by
request from user-space it is best to set the counter to '2' to start
with.

In particular we might also set the NEEDED_MASK flag at this time, and
in all other cases NEEDED_MASK is only set when the counter is 2 or
more.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: Fix userspace free_pages() macro
Steven Rostedt [Thu, 22 Dec 2011 23:17:51 +0000 (10:17 +1100)]
md: Fix userspace free_pages() macro

While using etags to find free_pages(), I stumbled across this debug
definition of free_pages() that is to be used while debugging some raid
code in userspace. The __get_free_pages() allocates the correct size,
but the free_pages() does not match. free_pages(), like
__get_free_pages(), takes an order and not a size.

Acked-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: be more thorough in calculating 'degraded' value.
NeilBrown [Thu, 22 Dec 2011 23:17:50 +0000 (10:17 +1100)]
md/raid5: be more thorough in calculating 'degraded' value.

When an array is being reshaped to change the number of devices,
the two halves can be differently degraded.  e.g. one could be
missing a device and the other not.

So we need to be more careful about calculating the 'degraded'
attribute.

Instead of just inc/dec at appropriate times, perform a full
re-calculation examining both possible cases.  This doesn't happen
often so it not a big cost, and we already have most of the code to
do it.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/bitmap: daemon_work cleanup.
NeilBrown [Thu, 22 Dec 2011 23:17:50 +0000 (10:17 +1100)]
md/bitmap: daemon_work cleanup.

We have a variable 'mddev' in this function, but repeatedly get the
same value by dereferencing bitmap->mddev.
There is room for simplification here...

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: allow non-privileged uses to GET_*_INFO about raid arrays.
NeilBrown [Thu, 22 Dec 2011 23:17:26 +0000 (10:17 +1100)]
md: allow non-privileged uses to GET_*_INFO about raid arrays.

The info is already available in /proc/mdstat and /sys/block in
an accessible form so there is no point in putting a road-block in
the ioctl for information gathering.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/bitmap: It is OK to clear bits during recovery.
NeilBrown [Thu, 22 Dec 2011 22:57:48 +0000 (09:57 +1100)]
md/bitmap: It is OK to clear bits during recovery.

commit d0a4bb492772ce5c4bdfba3744a99ed6f6fb238f introduced a
regression which is annoying but fairly harmless.

When writing to an array that is undergoing recovery (a spare
in being integrated into the array), writing to the array will
set bits in the bitmap, but they will not be cleared when the
write completes.

For bits covering areas that have not been recovered yet this is not a
problem as the recovery will clear the bits.  However bits set in
already-recovered region will stay set and never be cleared.
This doesn't risk data integrity.  The only negatives are:
 - next time there is a crash, more resyncing than necessary will
   be done.
 - the bitmap doesn't look clean, which is confusing.

While an array is recovering we don't want to update the
'events_cleared' setting in the bitmap but we do still want to clear
bits that have very recently been set - providing they were written to
the recovering device.

So split those two needs - which previously both depended on 'success'
and always clear the bit of the write went to all devices.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: don't give up looking for spares on first failure-to-add
NeilBrown [Thu, 22 Dec 2011 22:57:19 +0000 (09:57 +1100)]
md: don't give up looking for spares on first failure-to-add

Before performing a recovery we try to remove any spares that
might not be working, then add any that might have become relevant.

Currently we abort on the first spare that cannot be added.
This is a false optimisation.
It is conceivable that - depending on rules in the personality - a
subsequent spare might be accepted.
Also the loop does other things like count the available spares and
reset the 'recovery_offset' value.

If we abort early these might not happen properly.

So remove the early abort.

In particular if you have an array what is undergoing recovery and
which has extra spares, then the recovery may not restart after as
reboot as the could of 'spares' might end up as zero.

Reported-by: Anssi Hannula <anssi.hannula@iki.fi>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: ensure correct assessment of drives during degraded reshape.
NeilBrown [Thu, 22 Dec 2011 22:57:00 +0000 (09:57 +1100)]
md/raid5: ensure correct assessment of drives during degraded reshape.

While reshaping a degraded array (as when reshaping a RAID0 by first
converting it to a degraded RAID4) we currently get confused about
which devices are in_sync.  In most cases we get it right, but in the
region that is being reshaped we need to treat non-failed devices as
in-sync when we have the data but haven't actually written it out yet.

Reported-by: Adam Kwolek <adam.kwolek@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/linear: fix hot-add of devices to linear arrays.
NeilBrown [Thu, 22 Dec 2011 22:56:55 +0000 (09:56 +1100)]
md/linear: fix hot-add of devices to linear arrays.

commit d70ed2e4fafdbef0800e73942482bb075c21578b
broke hot-add to a linear array.
After that commit, metadata if not written to devices until they
have been fully integrated into the array as determined by
saved_raid_disk.  That patch arranged to clear that field after
a recovery completed.

However for linear arrays, there is no recovery - the integration is
instantaneous.  So we need to explicitly clear the saved_raid_disk
field.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: raid5 crash during degradation
Adam Kwolek [Fri, 9 Dec 2011 03:26:11 +0000 (14:26 +1100)]
md: raid5 crash during degradation

NULL pointer access causes crash in raid5 module.

Signed-off-by: Adam Kwolek <adam.kwolek@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: never wait for bad-block acks on failed device.
NeilBrown [Thu, 8 Dec 2011 05:27:57 +0000 (16:27 +1100)]
md/raid5: never wait for bad-block acks on failed device.

Once a device is failed we really want to completely ignore it.
It should go away soon anyway.

In particular the presence of bad blocks on it should not cause us to
block as we won't be trying to write there anyway.

So as soon as we can check if a device is Faulty, do so and pretend
that it is already gone if it is Faulty.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: ensure new badblocks are handled promptly.
NeilBrown [Thu, 8 Dec 2011 05:26:08 +0000 (16:26 +1100)]
md: ensure new badblocks are handled promptly.

When we mark blocks as bad we need them to be acknowledged by the
metadata handler promptly.

For an in-kernel metadata handler that was already being done.  But
for an external metadata handler we need to alert it of the change by
sending a notification through the sysfs file.  This adds that
notification.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: bad blocks shouldn't cause a Blocked status on a Faulty device.
NeilBrown [Thu, 8 Dec 2011 05:22:48 +0000 (16:22 +1100)]
md: bad blocks shouldn't cause a Blocked status on a Faulty device.

Once a device is marked Faulty the badblocks - whether acknowledged or
not - become irrelevant.  So they shouldn't cause the device to be
marked as Blocked.

Without this patch, a process might write "-blocked" to clear the
Blocked status, but while that will correctly fail the device, it
won't remove the apparent 'blocked' status.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: take a reference to mddev during sysfs access.
NeilBrown [Thu, 8 Dec 2011 04:49:46 +0000 (15:49 +1100)]
md: take a reference to mddev during sysfs access.

When we are accessing an mddev via sysfs we know that the
mddev cannot disappear because it has an embedded kobj which
is refcounted by sysfs.
And we also take the mddev_lock.
However this is not enough.

The final mddev_put could have been called and the
mddev_delayed_delete is waiting for sysfs to let go so it can destroy
the kobj and mddev.
In this state there are a lot of changes that should not be attempted.

To to guard against this we:
 - initialise mddev->all_mddevs in on last put so the state can be
   easily detected.
 - in md_attr_show and md_attr_store, check ->all_mddevs under
   all_mddevs_lock and mddev_get the mddev if it still appears to
   be active.

This means that if we get to sysfs as the mddev is being deleted we
will get -EBUSY.

rdev_attr_store and rdev_attr_show are similar but already have
sufficient protection.  They check that rdev->mddev still points to
mddev after taking mddev_lock.  As this is cleared  before delayed
removal which can only be requested under the mddev_lock, this
ensure the rdev and mddev are still alive.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd: refine interpretation of "hold_active == UNTIL_IOCTL".
NeilBrown [Thu, 8 Dec 2011 04:49:12 +0000 (15:49 +1100)]
md: refine interpretation of "hold_active == UNTIL_IOCTL".

We like md devices to disappear when they really are not needed.
However it is not possible to tell from the current state whether it
is needed or not.  We can only tell from recent history of changes.

In particular immediately after we create an md device it looks very
similar to immediately after we have finished with it.

So we always preserve a newly created md device until something
significant happens.  This state is stored in 'hold_active'.

The normal case is to keep it until an ioctl happens, as that will
normally either activate it, or explicitly de-activate it.  If it
doesn't then it was probably created by mistake and it is now time to
get rid of it.

We can also modify an array via sysfs (instead of via ioctl) and we
currently treat any change via sysfs like an ioctl as a sign that if
it now isn't more active, it should be destroyed.
However this is not appropriate as changes made via sysfs are more
gradual so we should look for a more definitive change.

So this patch only clears 'hold_active' from UNTIL_IOCTL to clear when
the array_state is changed via sysfs.  Other changes via sysfs
are ignored.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/lock: ensure updates to page_attrs are properly locked.
NeilBrown [Tue, 22 Nov 2011 23:18:52 +0000 (10:18 +1100)]
md/lock: ensure updates to page_attrs are properly locked.

Page attributes are set using __set_bit rather than set_bit as
it normally called under a spinlock so the extra atomicity is not
needed.

However there are two places where we might set or clear page
attributes without holding the spinlock.
So add the spinlock in those cases.

This might be the cause of occasional reports that bits a aren't
getting clear properly - theory is that BITMAP_PAGE_PENDING gets lost
when BITMAP_PAGE_NEEDWRITE is set or cleared.  This is an
inconvenience, not a threat to data safety.

Signed-off-by: NeilBrown <neilb@suse.de>
13 years agomd/raid5: STRIPE_ACTIVE has lock semantics, add barriers
Dan Williams [Tue, 8 Nov 2011 05:22:06 +0000 (16:22 +1100)]
md/raid5: STRIPE_ACTIVE has lock semantics, add barriers

All updates that occur under STRIPE_ACTIVE should be globally visible
when STRIPE_ACTIVE clears.  test_and_set_bit() implies a barrier, but
clear_bit() does not.

This is suitable for 3.1-stable.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Cc: stable@kernel.org
13 years agomd/raid5: abort any pending parity operations when array fails.
NeilBrown [Tue, 8 Nov 2011 05:22:01 +0000 (16:22 +1100)]
md/raid5: abort any pending parity operations when array fails.

When the number of failed devices exceeds the allowed number
we must abort any active parity operations (checks or updates) as they
are no longer meaningful, and can lead to a BUG_ON in
handle_parity_checks6.

This bug was introduce by commit 6c0069c0ae9659e3a91b68eaed06a5c6c37f45c8
in 2.6.29.

Reported-by: Manish Katiyar <mkatiyar@gmail.com>
Tested-by: Manish Katiyar <mkatiyar@gmail.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Cc: stable@kernel.org
13 years agoLinux 3.2-rc1
Linus Torvalds [Tue, 8 Nov 2011 00:16:02 +0000 (16:16 -0800)]
Linux 3.2-rc1

.. with new name.  Because nothing says "really solid kernel release"
like naming it after an extinct animal that just happened to be in the
news lately.

13 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux...
Linus Torvalds [Tue, 8 Nov 2011 00:14:26 +0000 (16:14 -0800)]
Merge branch 'fixes' of git://git./linux/kernel/git/tmlind/linux-omap

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap: (31 commits)
  ARM: OMAP: Fix export.h or module.h includes
  ARM: OMAP: omap_device: Include linux/export.h
  ARM: OMAP2: Fix H4 matrix keyboard warning
  ARM: OMAP1: Remove unused omap-alsa.h
  ARM: OMAP1: Fix warnings about enabling 32 KiHz timer
  ARM: OMAP2+: timer: Remove omap_device_pm_latency
  ARM: OMAP2+: clock data: Remove redundant timer clkdev
  ARM: OMAP: Devkit8000: Remove double omap_mux_init_gpio
  ARM: OMAP: usb: musb: OMAP: Delete unused function
  MAINTAINERS: Update linux-omap git repository
  ARM: OMAP: change get_context_loss_count ret value to int
  ARM: OMAP4: hsmmc: configure SDMMC1_DR0 properly
  ARM: OMAP4: hsmmc: Fix Pbias configuration on regulator OFF
  ARM: OMAP3: hwmod: fix variant registration and remove SmartReflex from common list
  ARM: OMAP: I2C: Fix omap_register_i2c_bus() return value on success
  ARM: OMAP: dmtimer: Include linux/module.h
  ARM: OMAP2+: l3-noc: Include linux/module.h
  ARM: OMAP2+: devices: Fixes for McPDM
  ARM: OMAP: Fix errors and warnings when building for one board
  ARM: OMAP3: PM: restrict erratum i443 handling to OMAP3430 only
  ...

13 years agoVFS: we need to set LOOKUP_JUMPED on mountpoint crossing
Al Viro [Mon, 7 Nov 2011 21:21:26 +0000 (21:21 +0000)]
VFS: we need to set LOOKUP_JUMPED on mountpoint crossing

Mountpoint crossing is similar to following procfs symlinks - we do
not get ->d_revalidate() called for dentry we have arrived at, with
unpleasant consequences for NFS4.

Simple way to reproduce the problem in mainline:

    cat >/tmp/a.c <<'EOF'
    #include <unistd.h>
    #include <fcntl.h>
    #include <stdio.h>
    main()
    {
            struct flock fl = {.l_type = F_RDLCK, .l_whence = SEEK_SET, .l_len = 1};
            if (fcntl(0, F_SETLK, &fl))
                    perror("setlk");
    }
    EOF
    cc /tmp/a.c -o /tmp/test

then on nfs4:

    mount --bind file1 file2
    /tmp/test < file1 # ok
    /tmp/test < file2 # spews "setlk: No locks available"...

What happens is the missing call of ->d_revalidate() after mountpoint
crossing and that's where NFS4 would issue OPEN request to server.

The fix is simple - treat mountpoint crossing the same way we deal with
following procfs-style symlinks.  I.e.  set LOOKUP_JUMPED...

Cc: stable@kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 7 Nov 2011 20:38:11 +0000 (12:38 -0800)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf top: Fix live annotation in the --stdio interface
  perf top tui: Don't recalc column widths considering just the first page
  perf report: Add progress bar when processing time ordered events
  perf hists browser: Warn about lost events
  perf tools: Fix a typo of command name as trace-cmd
  perf hists: Fix recalculation of total_period when sorting entries
  perf header: Fix build on old systems
  perf ui browser: Handle K_RESIZE in dialog windows
  perf ui browser: No need to switch char sets that often
  perf hists browser: Use K_TIMER
  perf ui: Rename ui__warning_paranoid to ui__error_paranoid
  perf ui: Reimplement the popup windows using libslang
  perf ui: Reimplement ui__popup_menu using ui__browser
  perf ui: Reimplement ui_helpline using libslang
  perf ui: Improve handling sigwinch a bit
  perf ui progress: Reimplement using slang
  perf evlist: Fix grouping of multiple events

13 years agoMerge branch 'fixes-modulesplit' into fixes
Tony Lindgren [Mon, 7 Nov 2011 20:27:23 +0000 (12:27 -0800)]
Merge branch 'fixes-modulesplit' into fixes

13 years agoARM: OMAP: Fix export.h or module.h includes
Tony Lindgren [Mon, 7 Nov 2011 20:27:10 +0000 (12:27 -0800)]
ARM: OMAP: Fix export.h or module.h includes

Commit 32aaeffbd4a7457bf2f7448b33b5946ff2a960eb (Merge branch
'modsplit-Oct31_2011'...) caused some build errors. Fix these
and make sure we always have export.h or module.h included
for MODULE_ and EXPORT_SYMBOL users:

$ grep -rl ^MODULE_ arch/arm/*omap*/*.c | xargs \
  grep -L linux/module.h
  arch/arm/mach-omap2/dsp.c
  arch/arm/mach-omap2/mailbox.c
  arch/arm/mach-omap2/omap-iommu.c
  arch/arm/mach-omap2/smartreflex.c

Also check we either have export.h or module.h included
for the files exporting symbols:

$ grep -rl EXPORT_SYMBOL arch/arm/*omap*/*.c | xargs \
  grep -L linux/export.h | xargs grep -L linux/module.h

Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Tony Lindgren <tony@atomide.com>
13 years agoARM: OMAP: omap_device: Include linux/export.h
Axel Lin [Mon, 7 Nov 2011 20:27:10 +0000 (12:27 -0800)]
ARM: OMAP: omap_device: Include linux/export.h

Include linux/export.h to fix below build warning:

  CC      arch/arm/plat-omap/omap_device.o
arch/arm/plat-omap/omap_device.c:1055: warning: data definition has no type or storage class
arch/arm/plat-omap/omap_device.c:1055: warning: type defaults to 'int' in declaration of 'EXPORT_SYMBOL'
arch/arm/plat-omap/omap_device.c:1055: warning: parameter names (without types) in function declaration

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Mon, 7 Nov 2011 18:55:33 +0000 (10:55 -0800)]
Merge git://git./linux/kernel/git/davem/net

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (47 commits)
  forcedeth: fix a few sparse warnings (variable shadowing)
  forcedeth: Improve stats counters
  forcedeth: remove unneeded stats updates
  forcedeth: Acknowledge only interrupts that are being processed
  forcedeth: fix race when unloading module
  MAINTAINERS/rds: update maintainer
  wanrouter: Remove kernel_lock annotations
  usbnet: fix oops in usbnet_start_xmit
  ixgbe: Fix compile for kernel without CONFIG_PCI_IOV defined
  etherh: Add MAINTAINERS entry for etherh
  bonding: comparing a u8 with -1 is always false
  sky2: fix regression on Yukon Optima
  netlink: clarify attribute length check documentation
  netlink: validate NLA_MSECS length
  i825xx:xscale:8390:freescale: Fix Kconfig dependancies
  macvlan: receive multicast with local address
  tg3: Update version to 3.121
  tg3: Eliminate timer race with reset_task
  tg3: Schedule at most one tg3_reset_task run
  tg3: Obtain PCI function number from device
  ...

13 years agovfs: d_invalidate() should leave mountpoints alone
Al Viro [Mon, 7 Nov 2011 16:39:57 +0000 (16:39 +0000)]
vfs: d_invalidate() should leave mountpoints alone

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoforcedeth: fix a few sparse warnings (variable shadowing)
david decotigny [Sat, 5 Nov 2011 14:38:24 +0000 (14:38 +0000)]
forcedeth: fix a few sparse warnings (variable shadowing)

This fixes the following sparse warnings:
drivers/net/ethernet/nvidia/forcedeth.c:2113:7: warning: symbol 'size' shadows an earlier one
drivers/net/ethernet/nvidia/forcedeth.c:2102:6: originally declared here
drivers/net/ethernet/nvidia/forcedeth.c:2155:7: warning: symbol 'size' shadows an earlier one
drivers/net/ethernet/nvidia/forcedeth.c:2102:6: originally declared here
drivers/net/ethernet/nvidia/forcedeth.c:2227:7: warning: symbol 'size' shadows an earlier one
drivers/net/ethernet/nvidia/forcedeth.c:2215:6: originally declared here
drivers/net/ethernet/nvidia/forcedeth.c:2271:7: warning: symbol 'size' shadows an earlier one
drivers/net/ethernet/nvidia/forcedeth.c:2215:6: originally declared here
drivers/net/ethernet/nvidia/forcedeth.c:2986:20: warning: symbol 'addr' shadows an earlier one
drivers/net/ethernet/nvidia/forcedeth.c:2963:6: originally declared here

Signed-off-by: David Decotigny <david.decotigny@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoforcedeth: Improve stats counters
Mandeep Baines [Sat, 5 Nov 2011 14:38:23 +0000 (14:38 +0000)]
forcedeth: Improve stats counters

Rx byte count was off; instead use the hardware's count.  Tx packet
count was counting pre-TSO packets; instead count on-the-wire packets.
Report hardware dropped frame count as rx_fifo_errors.

- The count of transmitted packets reported by the forcedeth driver
  reports pre-TSO (TCP Segmentation Offload) packet counts and not the
  count of the number of packets sent on the wire. This change fixes
  the forcedeth driver to report the correct count. Fixed the code by
  copying the count stored in the NIC H/W to the value reported by the
  driver.

- Count rx_drop_frame errors as rx_fifo_errors:
  We see a lot of rx_drop_frame errors if we disable the rx bottom-halves
  for too long.  Normally, rx_fifo_errors would be counted in this case.
  The rx_drop_frame error count is private to forcedeth and is not
  reported by ifconfig or sysfs.  The rx_fifo_errors count is currently
  unused in the forcedeth driver.  It is reported by ifconfig as overruns.
  This change reports rx_drop_frame errors as rx_fifo_errors.

Signed-off-by: David Decotigny <david.decotigny@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoforcedeth: remove unneeded stats updates
david decotigny [Sat, 5 Nov 2011 14:38:22 +0000 (14:38 +0000)]
forcedeth: remove unneeded stats updates

Function ndo_get_stats() updates most of the stats from hardware
registers, making the manual updates un-needed. This change removes
these manual updates. Main exception is rx_missed_errors which needs
manual update.

Another exception is rx_packets, still updated manually in this commit
to make sure this patch doesn't change behavior of driver. This will
be addressed by a future patch.

Signed-off-by: David Decotigny <david.decotigny@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoforcedeth: Acknowledge only interrupts that are being processed
Mike Ditto [Sat, 5 Nov 2011 14:38:21 +0000 (14:38 +0000)]
forcedeth: Acknowledge only interrupts that are being processed

This is to avoid a race, accidentally acknowledging an interrupt that
we didn't notice and won't immediately process.  This is based solely
on code inspection; it is not known if there was an actual bug here.

Signed-off-by: David Decotigny <david.decotigny@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoforcedeth: fix race when unloading module
david decotigny [Sat, 5 Nov 2011 14:38:20 +0000 (14:38 +0000)]
forcedeth: fix race when unloading module

When forcedeth module is unloaded, there exists a path that can lead
to mod_timer() after del_timer_sync(), causing an oops. This patch
short-circuits this unneeded path, which originates in
nv_get_ethtool_stats().

Tested:
  x86_64 16-way + 3 ethtool -S infinite loops + 100Mbps incoming traffic
  + rmmod/modprobe/ifconfig in a loop

Initial-Author: Salman Qazi <sqazi@google.com>
Discussion: http://patchwork.ozlabs.org/patch/123548/

Signed-off-by: David Decotigny <david.decotigny@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agodevice-mapper: using EXPORT_SYBOL in dm-space-map-checker.c needs export.h
Stephen Rothwell [Tue, 1 Nov 2011 09:27:43 +0000 (20:27 +1100)]
device-mapper: using EXPORT_SYBOL in dm-space-map-checker.c needs export.h

Reported-by: Witold Baryluk <baryluk@smp.if.uj.edu.pl>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agodevice-mapper: dm-bufio.c needs to include module.h
Stephen Rothwell [Tue, 1 Nov 2011 07:30:49 +0000 (18:30 +1100)]
device-mapper: dm-bufio.c needs to include module.h

since it uses the module facilities.

Reported-by: Witold Baryluk <baryluk@smp.if.uj.edu.pl>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agodrivers/md: change module.h -> export.h in persistent-data/dm-*
Paul Gortmaker [Wed, 28 Sep 2011 22:29:32 +0000 (18:29 -0400)]
drivers/md: change module.h -> export.h in persistent-data/dm-*

For the files which are not themselves modular, we can change
them to include only the smaller export.h since all they are
doing is looking for EXPORT_SYMBOL.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoarm: Add export.h to recently added files for EXPORT_SYMBOL
Paul Gortmaker [Sun, 9 Oct 2011 03:24:48 +0000 (23:24 -0400)]
arm: Add export.h to recently added files for EXPORT_SYMBOL

These files didn't exist at the time of the module.h split, and
so were not fixed by the commits on that baseline.  Since they use
the EXPORT_SYMBOL and/or THIS_MODULE macros, they will need the
new export.h file included that provides them.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMAINTAINERS/rds: update maintainer
Or Gerlitz [Mon, 7 Nov 2011 18:28:20 +0000 (13:28 -0500)]
MAINTAINERS/rds: update maintainer

update for the actual maintainer

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agowanrouter: Remove kernel_lock annotations
Richard Weinberger [Mon, 7 Nov 2011 18:27:30 +0000 (13:27 -0500)]
wanrouter: Remove kernel_lock annotations

The BKL is gone, these annotations are useless.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agousbnet: fix oops in usbnet_start_xmit
Konstantin Khlebnikov [Mon, 7 Nov 2011 05:54:58 +0000 (05:54 +0000)]
usbnet: fix oops in usbnet_start_xmit

This patch fixes the bug added in commit v3.1-rc7-1055-gf9b491e
SKB can be NULL at this point, at least for cdc-ncm.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoixgbe: Fix compile for kernel without CONFIG_PCI_IOV defined
Rose, Gregory V [Mon, 7 Nov 2011 07:44:17 +0000 (07:44 +0000)]
ixgbe: Fix compile for kernel without CONFIG_PCI_IOV defined

Fix compiler errors and warnings with CONFIG_PCI_IOV defined and not
defined.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux
Linus Torvalds [Mon, 7 Nov 2011 18:13:52 +0000 (10:13 -0800)]
Merge branch 'release' of git://git./linux/kernel/git/lenb/linux

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux:
  cpuidle: Single/Global registration of idle states
  cpuidle: Split cpuidle_state structure and move per-cpu statistics fields
  cpuidle: Remove CPUIDLE_FLAG_IGNORE and dev->prepare()
  cpuidle: Move dev->last_residency update to driver enter routine; remove dev->last_state
  ACPI: Fix CONFIG_ACPI_DOCK=n compiler warning
  ACPI: Export FADT pm_profile integer value to userspace
  thermal: Prevent polling from happening during system suspend
  ACPI: Drop ACPI_NO_HARDWARE_INIT
  ACPI atomicio: Convert width in bits to bytes in __acpi_ioremap_fast()
  PNPACPI: Simplify disabled resource registration
  ACPI: Fix possible recursive locking in hwregs.c
  ACPI: use kstrdup()
  mrst pmu: update comment
  tools/power turbostat: less verbose debugging

13 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Mon, 7 Nov 2011 18:01:56 +0000 (10:01 -0800)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (40 commits)
  vmwgfx: Snoop DMA transfers with non-covering sizes
  vmwgfx: Move the prefered mode first in the list
  vmwgfx: Unreference surface on cursor error path
  vmwgfx: Free prefered mode on error path
  vmwgfx: Use pointer return error codes
  vmwgfx: Fix hw cursor position
  vmwgfx: Infrastructure for explicit placement
  vmwgfx: Make the preferred autofit mode have a 60Hz vrefresh
  vmwgfx: Remove screen object active list
  vmwgfx: Screen object cleanups
  drm/radeon/kms: consolidate GART code, fix segfault after GPU lockup V2
  drm/radeon/kms: don't poll forever if MC GDDR link training fails
  drm/radeon/kms: fix DP setup on TRAVIS bridges
  drm/radeon/kms: set HPD polarity in hpd_init()
  drm/radeon/kms: add MSI module parameter
  drm/radeon/kms: Add MSI quirk for Dell RS690
  drm/radeon/kms: Add MSI quirk for HP RS690
  drm/radeon/kms: split MSI check into a separate function
  vmwgfx: Reinstate the update_layout ioctl
  drm/radeon/kms: always do extended edid probe
  ...

13 years agoMerge branch 'urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 7 Nov 2011 17:59:02 +0000 (09:59 -0800)]
Merge branch 'urgent-for-linus' of git://git./linux/kernel/git/wfg/linux

* 'urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux:
  writeback: fix uninitialized task_ratelimit

13 years agoMerge git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Mon, 7 Nov 2011 17:56:22 +0000 (09:56 -0800)]
Merge git://git.samba.org/sfrench/cifs-2.6

* git://git.samba.org/sfrench/cifs-2.6:
  CIFS: Cleanup byte-range locking code style
  CIFS: Simplify setlk error handling for mandatory locking

13 years agoMerge git://git.infradead.org/mtd-2.6
Linus Torvalds [Mon, 7 Nov 2011 17:11:16 +0000 (09:11 -0800)]
Merge git://git.infradead.org/mtd-2.6

* git://git.infradead.org/mtd-2.6: (226 commits)
  mtd: tests: annotate as DANGEROUS in Kconfig
  mtd: tests: don't use mtd0 as a default
  mtd: clean up usage of MTD_DOCPROBE_ADDRESS
  jffs2: add compr=lzo and compr=zlib options
  jffs2: implement mount option parsing and compression overriding
  mtd: nand: initialize ops.mode
  mtd: provide an alias for the redboot module name
  mtd: m25p80: don't probe device which has status of 'disabled'
  mtd: nand_h1900 never worked
  mtd: Add DiskOnChip G3 support
  mtd: m25p80: add EON flash EN25Q32B into spi flash id table
  mtd: mark block device queue as non-rotational
  mtd: r852: make r852_pm_ops static
  mtd: m25p80: add support for at25df321a spi data flash
  mtd: mxc_nand: preset_v1_v2: unlock all NAND flash blocks
  mtd: nand: switch `check_pattern()' to standard `memcmp()'
  mtd: nand: invalidate cache on unaligned reads
  mtd: nand: do not scan bad blocks with NAND_BBT_NO_OOB set
  mtd: nand: wait to set BBT version
  mtd: nand: scrub BBT on ECC errors
  ...

Fix up trivial conflicts:
 - arch/arm/mach-at91/board-usb-a9260.c
Merged into board-usb-a926x.c
 - drivers/mtd/maps/lantiq-flash.c
add_mtd_partitions -> mtd_device_register vs changed to use
mtd_device_parse_register.

13 years agoMerge branch 'linux-next' of git://git.infradead.org/ubifs-2.6
Linus Torvalds [Mon, 7 Nov 2011 16:52:19 +0000 (08:52 -0800)]
Merge branch 'linux-next' of git://git.infradead.org/ubifs-2.6

* 'linux-next' of git://git.infradead.org/ubifs-2.6:
  UBIFS: fix the dark space calculation
  UBIFS: introduce a helper to dump scanning info

13 years agovmwgfx: Snoop DMA transfers with non-covering sizes
Jakob Bornecrantz [Thu, 3 Nov 2011 20:03:08 +0000 (21:03 +0100)]
vmwgfx: Snoop DMA transfers with non-covering sizes

Enough to get cursors working under Wayland.

Signed-off-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Move the prefered mode first in the list
Jakob Bornecrantz [Thu, 3 Nov 2011 20:03:07 +0000 (21:03 +0100)]
vmwgfx: Move the prefered mode first in the list

Signed-off-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Unreference surface on cursor error path
Jakob Bornecrantz [Thu, 3 Nov 2011 20:03:06 +0000 (21:03 +0100)]
vmwgfx: Unreference surface on cursor error path

Signed-off-by: Jakob Bornecrantz <jakob@vmware.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Free prefered mode on error path
Jakob Bornecrantz [Thu, 3 Nov 2011 20:03:05 +0000 (21:03 +0100)]
vmwgfx: Free prefered mode on error path

Signed-off-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Use pointer return error codes
Jakob Bornecrantz [Thu, 3 Nov 2011 20:03:04 +0000 (21:03 +0100)]
vmwgfx: Use pointer return error codes

Signed-off-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Fix hw cursor position
Thomas Hellstrom [Wed, 2 Nov 2011 08:43:12 +0000 (09:43 +0100)]
vmwgfx: Fix hw cursor position

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Infrastructure for explicit placement
Thomas Hellstrom [Wed, 2 Nov 2011 08:43:11 +0000 (09:43 +0100)]
vmwgfx: Infrastructure for explicit placement

Make it possible to use explicit placement
(although not hooked up with a user-space interface yet)
and relax the single framebuffer limit to only apply to implicit placement.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Make the preferred autofit mode have a 60Hz vrefresh
Thomas Hellstrom [Wed, 2 Nov 2011 08:43:10 +0000 (09:43 +0100)]
vmwgfx: Make the preferred autofit mode have a 60Hz vrefresh

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Remove screen object active list
Thomas Hellstrom [Wed, 2 Nov 2011 08:43:09 +0000 (09:43 +0100)]
vmwgfx: Remove screen object active list

It isn't used for anything. Replace with an active bool.

Also make a couple of functions return void instead of int
since their return value wasn't checked anyway.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakbo Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agovmwgfx: Screen object cleanups
Thomas Hellstrom [Wed, 2 Nov 2011 08:43:08 +0000 (09:43 +0100)]
vmwgfx: Screen object cleanups

Remove unused member.
No need to pin / unpin fb.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agowriteback: fix uninitialized task_ratelimit
Wu Fengguang [Mon, 7 Nov 2011 11:19:28 +0000 (19:19 +0800)]
writeback: fix uninitialized task_ratelimit

In balance_dirty_pages() task_ratelimit may be not initialized
(initialization skiped by goto pause), and then used when calling
tracing hook.

Fix it by moving the task_ratelimit assignment before goto pause.

Reported-by: Witold Baryluk <baryluk@smp.if.uj.edu.pl>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
13 years agoRevert "hvc_console: display printk messages on console."
Linus Torvalds [Mon, 7 Nov 2011 06:22:16 +0000 (22:22 -0800)]
Revert "hvc_console: display printk messages on console."

This reverts commit 361162459f62dc0826b82c9690a741a940f457f0.

It causes an infinite loop when booting Linux under Xen, as so:

  [    2.382984] console [hvc0] enabled
  [    2.382984] console [hvc0] enabled
  [    2.382984] console [hvc0] enabled
  ...

as reported by Konrad Rzeszutek Wilk.  And Rusty reports the same for
lguest.  He goes on to say:

   "This is not a concurrency problem: the issue seems to be that
    calling register_console() twice on the same struct console is a bad
    idea."

and Greg says he'll fix it up properly at some point later. Revert for now.

Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Reported-by: Rusty Russell <rusty@ozlabs.org>
Requested-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Miche Baker-Harvey <miche@google.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agopowerpc: fix building hvc_opal.c
Michael Neuling [Mon, 7 Nov 2011 06:12:28 +0000 (17:12 +1100)]
powerpc: fix building hvc_opal.c

Fix building following build error:

  drivers/tty/hvc/hvc_opal.c:244:12: error: 'THIS_MODULE' undeclared here (not in a function)

Signed-off-by: Michael Neuling <mikey@neuling.org>
[ New file from powerpc tree not following the new rules from the
  module.h split, both of which were merged today.  - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'upstream/jump-label-noearly' of git://git.kernel.org/pub/scm/linux...
Linus Torvalds [Mon, 7 Nov 2011 04:20:46 +0000 (20:20 -0800)]
Merge branch 'upstream/jump-label-noearly' of git://git./linux/kernel/git/jeremy/xen

* 'upstream/jump-label-noearly' of git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen:
  jump-label: initialize jump-label subsystem much earlier
  x86/jump_label: add arch_jump_label_transform_static()
  s390/jump-label: add arch_jump_label_transform_static()
  jump_label: add arch_jump_label_transform_static() to optimise non-live code updates
  sparc/jump_label: drop arch_jump_label_text_poke_early()
  x86/jump_label: drop arch_jump_label_text_poke_early()
  jump_label: if a key has already been initialized, don't nop it out
  stop_machine: make stop_machine safe and efficient to call early
  jump_label: use proper atomic_t initializer

Conflicts:
 - arch/x86/kernel/jump_label.c
Added __init_or_module to arch_jump_label_text_poke_early vs
removal of that function entirely
 - kernel/stop_machine.c
same patch ("stop_machine: make stop_machine safe and efficient
to call early") merged twice, with whitespace fix in one version

13 years agoMerge branch 'upstream/xen-settime' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Mon, 7 Nov 2011 04:15:05 +0000 (20:15 -0800)]
Merge branch 'upstream/xen-settime' of git://git./linux/kernel/git/jeremy/xen

* 'upstream/xen-settime' of git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen:
  xen/dom0: set wallclock time in Xen
  xen: add dom0_op hypercall
  xen/acpi: Domain0 acpi parser related platform hypercall

13 years agoMerge branch 'stable/cleanups-3.2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 7 Nov 2011 04:13:34 +0000 (20:13 -0800)]
Merge branch 'stable/cleanups-3.2' of git://git./linux/kernel/git/konrad/xen

* 'stable/cleanups-3.2' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen: use static initializers in xen-balloon.c
  Xen: fix braces and tabs coding style issue in xenbus_probe.c
  Xen: fix braces coding style issue in xenbus_probe.h
  Xen: fix whitespaces,tabs coding style issue in drivers/xen/pci.c
  Xen: fix braces coding style issue in gntdev.c and grant-table.c
  Xen: fix whitespaces,tabs coding style issue in drivers/xen/events.c
  Xen: fix whitespaces,tabs coding style issue in drivers/xen/balloon.c

Fix up trivial whitespace-conflicts in
 drivers/xen/{balloon.c,pci.c,xenbus/xenbus_probe.c}

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Mon, 7 Nov 2011 04:03:41 +0000 (20:03 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (114 commits)
  Btrfs: check for a null fs root when writing to the backup root log
  Btrfs: fix race during transaction joins
  Btrfs: fix a potential btrfs_bio leak on scrub fixups
  Btrfs: rename btrfs_bio multi -> bbio for consistency
  Btrfs: stop leaking btrfs_bios on readahead
  Btrfs: stop the readahead threads on failed mount
  Btrfs: fix extent_buffer leak in the metadata IO error handling
  Btrfs: fix the new inspection ioctls for 32 bit compat
  Btrfs: fix delayed insertion reservation
  Btrfs: ClearPageError during writepage and clean_tree_block
  Btrfs: be smarter about committing the transaction in reserve_metadata_bytes
  Btrfs: make a delayed_block_rsv for the delayed item insertion
  Btrfs: add a log of past tree roots
  btrfs: separate superblock items out of fs_info
  Btrfs: use the global reserve when truncating the free space cache inode
  Btrfs: release metadata from global reserve if we have to fallback for unlink
  Btrfs: make sure to flush queued bios if write_cache_pages waits
  Btrfs: fix extent pinning bugs in the tree log
  Btrfs: make sure btrfs_remove_free_space doesn't leak EAGAIN
  Btrfs: don't wait as long for more batches during SSD log commit
  ...

13 years agoMerge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 7 Nov 2011 03:44:47 +0000 (19:44 -0800)]
Merge branch 'modsplit-Oct31_2011' of git://git./linux/kernel/git/paulg/linux

* 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux: (230 commits)
  Revert "tracing: Include module.h in define_trace.h"
  irq: don't put module.h into irq.h for tracking irqgen modules.
  bluetooth: macroize two small inlines to avoid module.h
  ip_vs.h: fix implicit use of module_get/module_put from module.h
  nf_conntrack.h: fix up fallout from implicit moduleparam.h presence
  include: replace linux/module.h with "struct module" wherever possible
  include: convert various register fcns to macros to avoid include chaining
  crypto.h: remove unused crypto_tfm_alg_modname() inline
  uwb.h: fix implicit use of asm/page.h for PAGE_SIZE
  pm_runtime.h: explicitly requires notifier.h
  linux/dmaengine.h: fix implicit use of bitmap.h and asm/page.h
  miscdevice.h: fix up implicit use of lists and types
  stop_machine.h: fix implicit use of smp.h for smp_processor_id
  of: fix implicit use of errno.h in include/linux/of.h
  of_platform.h: delete needless include <linux/module.h>
  acpi: remove module.h include from platform/aclinux.h
  miscdevice.h: delete unnecessary inclusion of module.h
  device_cgroup.h: delete needless include <linux/module.h>
  net: sch_generic remove redundant use of <linux/module.h>
  net: inet_timewait_sock doesnt need <linux/module.h>
  ...

Fix up trivial conflicts (other header files, and  removal of the ab3550 mfd driver) in
 - drivers/media/dvb/frontends/dibx000_common.c
 - drivers/media/video/{mt9m111.c,ov6650.c}
 - drivers/mfd/ab3550-core.c
 - include/linux/dmaengine.h

13 years agoMerge branches 'acpi', 'idle', 'mrst-pmu' and 'pm-tools' into next
Len Brown [Mon, 7 Nov 2011 03:14:50 +0000 (22:14 -0500)]
Merge branches 'acpi', 'idle', 'mrst-pmu' and 'pm-tools' into next

13 years agoMerge branches 'atomic-io', 'lockdep', 'misc', 'pm-profile' and 'pnp-cleanup' into...
Len Brown [Mon, 7 Nov 2011 03:10:14 +0000 (22:10 -0500)]
Merge branches 'atomic-io', 'lockdep', 'misc', 'pm-profile' and 'pnp-cleanup' into acpi

13 years agoMerge branch 'writeback-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 7 Nov 2011 03:02:23 +0000 (19:02 -0800)]
Merge branch 'writeback-for-linus' of git://git./linux/kernel/git/wfg/linux

* 'writeback-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux:
  writeback: Add a 'reason' to wb_writeback_work
  writeback: send work item to queue_io, move_expired_inodes
  writeback: trace event balance_dirty_pages
  writeback: trace event bdi_dirty_ratelimit
  writeback: fix ppc compile warnings on do_div(long long, unsigned long)
  writeback: per-bdi background threshold
  writeback: dirty position control - bdi reserve area
  writeback: control dirty pause time
  writeback: limit max dirty pause time
  writeback: IO-less balance_dirty_pages()
  writeback: per task dirty rate limit
  writeback: stabilize bdi->dirty_ratelimit
  writeback: dirty rate control
  writeback: add bg_threshold parameter to __bdi_update_bandwidth()
  writeback: dirty position control
  writeback: account per-bdi accumulated dirtied pages

13 years agoMerge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target...
Linus Torvalds [Mon, 7 Nov 2011 03:00:42 +0000 (19:00 -0800)]
Merge branch 'for-next' of git://git./linux/kernel/git/nab/target-pending

* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  target: use ->exectute_task for all CDB emulation
  target: remove SCF_EMULATE_CDB_ASYNC
  target: refactor transport_emulate_control_cdb
  target: pass the se_task to the CDB emulation callback
  target: split core_scsi3_emulate_pr
  target: split core_scsi2_emulate_crh
  target: Add generic active I/O shutdown logic
  target: add back error handling in transport_complete_task
  target/pscsi: blk_make_request() returns an ERR_PTR()
  target: Remove core TRANSPORT_FREE_CMD_INTR usage
  target: Make TFO->check_stop_free return free status
  iscsi-target: Fix non-immediate TMR handling
  iscsi-target: Add missing CMDSN_LOWER_THAN_EXP check in iscsit_handle_scsi_cmd
  target: Avoid double list_del for aborted se_tmr_req
  target: Minor cleanups to core_tmr_drain_tmr_list
  target: Fix wrong se_tmr being added to drain_tmr_list
  target: Fix incorrect se_cmd assignment in core_tmr_drain_tmr_list
  target: Check -ENOMEM to signal QUEUE_FULL from fabric callbacks
  tcm_loop: Add explict read buffer memset for SCF_SCSI_CONTROL_SG_IO_CDB
  target: Fix compile warning w/ missing module.h include

13 years agoMerge branch 'trivial' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild
Linus Torvalds [Mon, 7 Nov 2011 02:54:53 +0000 (18:54 -0800)]
Merge branch 'trivial' of git://git./linux/kernel/git/mmarek/kbuild

* 'trivial' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  scsi: drop unused Kconfig symbol
  pci: drop unused Kconfig symbol
  stmmac: drop unused Kconfig symbol
  x86: drop unused Kconfig symbol
  powerpc: drop unused Kconfig symbols
  powerpc: 40x: drop unused Kconfig symbol
  mips: drop unused Kconfig symbols
  openrisc: drop unused Kconfig symbols
  arm: at91: drop unused Kconfig symbol
  samples: drop unused Kconfig symbol
  m32r: drop unused Kconfig symbol
  score: drop unused Kconfig symbols
  sh: drop unused Kconfig symbol
  um: drop unused Kconfig symbol
  sparc: drop unused Kconfig symbol
  alpha: drop unused Kconfig symbol

Fix up trivial conflict in drivers/net/ethernet/stmicro/stmmac/Kconfig
as per Michal: the STMMAC_DUAL_MAC config variable is still unused and
should be deleted.