openwrt/staging/blogic.git
11 years agosched/numa: Remove check that skips small VMAs
Mel Gorman [Mon, 7 Oct 2013 10:29:06 +0000 (11:29 +0100)]
sched/numa: Remove check that skips small VMAs

task_numa_work skips small VMAs. At the time the logic was to reduce the
scanning overhead which was considerable. It is a dubious hack at best.
It would make much more sense to cache where faults have been observed
and only rescan those regions during subsequent PTE scans. Remove this
hack as motivation to do it properly in the future.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-29-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: numa: Scan pages with elevated page_mapcount
Mel Gorman [Mon, 7 Oct 2013 10:29:05 +0000 (11:29 +0100)]
mm: numa: Scan pages with elevated page_mapcount

Currently automatic NUMA balancing is unable to distinguish between false
shared versus private pages except by ignoring pages with an elevated
page_mapcount entirely. This avoids shared pages bouncing between the
nodes whose task is using them but that is ignored quite a lot of data.

This patch kicks away the training wheels in preparation for adding support
for identifying shared/private pages is now in place. The ordering is so
that the impact of the shared/private detection can be easily measured. Note
that the patch does not migrate shared, file-backed within vmas marked
VM_EXEC as these are generally shared library pages. Migrating such pages
is not beneficial as there is an expectation they are read-shared between
caches and iTLB and iCache pressure is generally low.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-28-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Check current->mm before allocating NUMA faults
Mel Gorman [Mon, 7 Oct 2013 10:29:04 +0000 (11:29 +0100)]
sched/numa: Check current->mm before allocating NUMA faults

task_numa_placement checks current->mm but after buffers for faults
have already been uselessly allocated. Move the check earlier.

[peterz@infradead.org: Identified the problem]

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-27-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Add infrastructure for split shared/private accounting of NUMA hinting...
Mel Gorman [Mon, 7 Oct 2013 10:29:03 +0000 (11:29 +0100)]
sched/numa: Add infrastructure for split shared/private accounting of NUMA hinting faults

Ideally it would be possible to distinguish between NUMA hinting faults
that are private to a task and those that are shared.  This patch prepares
infrastructure for separately accounting shared and private faults by
allocating the necessary buffers and passing in relevant information. For
now, all faults are treated as private and detection will be introduced
later.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-26-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Reschedule task on preferred NUMA node once selected
Mel Gorman [Mon, 7 Oct 2013 10:29:02 +0000 (11:29 +0100)]
sched/numa: Reschedule task on preferred NUMA node once selected

A preferred node is selected based on the node the most NUMA hinting
faults was incurred on. There is no guarantee that the task is running
on that node at the time so this patch rescheules the task to run on
the most idle CPU of the selected node when selected. This avoids
waiting for the balancer to make a decision.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-25-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Resist moving tasks towards nodes with fewer hinting faults
Mel Gorman [Mon, 7 Oct 2013 10:29:01 +0000 (11:29 +0100)]
sched/numa: Resist moving tasks towards nodes with fewer hinting faults

Just as "sched: Favour moving tasks towards the preferred node" favours
moving tasks towards nodes with a higher number of recorded NUMA hinting
faults, this patch resists moving tasks towards nodes with lower faults.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-24-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Favour moving tasks towards the preferred node
Mel Gorman [Mon, 7 Oct 2013 10:29:00 +0000 (11:29 +0100)]
sched/numa: Favour moving tasks towards the preferred node

This patch favours moving tasks towards NUMA node that recorded a higher
number of NUMA faults during active load balancing.  Ideally this is
self-reinforcing as the longer the task runs on that node, the more faults
it should incur causing task_numa_placement to keep the task running on that
node. In reality a big weakness is that the nodes CPUs can be overloaded
and it would be more efficient to queue tasks on an idle node and migrate
to the new node. This would require additional smarts in the balancer so
for now the balancer will simply prefer to place the task on the preferred
node for a PTE scans which is controlled by the numa_balancing_settle_count
sysctl. Once the settle_count number of scans has complete the schedule
is free to place the task on an alternative node if the load is imbalanced.

[srikar@linux.vnet.ibm.com: Fixed statistics]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
[ Tunable and use higher faults instead of preferred. ]
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-23-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Update NUMA hinting faults once per scan
Mel Gorman [Mon, 7 Oct 2013 10:28:59 +0000 (11:28 +0100)]
sched/numa: Update NUMA hinting faults once per scan

NUMA hinting fault counts and placement decisions are both recorded in the
same array which distorts the samples in an unpredictable fashion. The values
linearly accumulate during the scan and then decay creating a sawtooth-like
pattern in the per-node counts. It also means that placement decisions are
time sensitive. At best it means that it is very difficult to state that
the buffer holds a decaying average of past faulting behaviour. At worst,
it can confuse the load balancer if it sees one node with an artifically high
count due to very recent faulting activity and may create a bouncing effect.

This patch adds a second array. numa_faults stores the historical data
which is used for placement decisions. numa_faults_buffer holds the
fault activity during the current scan window. When the scan completes,
numa_faults decays and the values from numa_faults_buffer are copied
across.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-22-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Select a preferred node with the most numa hinting faults
Mel Gorman [Mon, 7 Oct 2013 10:28:58 +0000 (11:28 +0100)]
sched/numa: Select a preferred node with the most numa hinting faults

This patch selects a preferred node for a task to run on based on the
NUMA hinting faults. This information is later used to migrate tasks
towards the node during balancing.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-21-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Track NUMA hinting faults on per-node basis
Mel Gorman [Mon, 7 Oct 2013 10:28:57 +0000 (11:28 +0100)]
sched/numa: Track NUMA hinting faults on per-node basis

This patch tracks what nodes numa hinting faults were incurred on.
This information is later used to schedule a task on the node storing
the pages most frequently faulted by the task.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-20-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Slow scan rate if no NUMA hinting faults are being recorded
Mel Gorman [Mon, 7 Oct 2013 10:28:56 +0000 (11:28 +0100)]
sched/numa: Slow scan rate if no NUMA hinting faults are being recorded

NUMA PTE scanning slows if a NUMA hinting fault was trapped and no page
was migrated. For long-lived but idle processes there may be no faults
but the scan rate will be high and just waste CPU. This patch will slow
the scan rate for processes that are not trapping faults.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-19-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Set the scan rate proportional to the memory usage of the task being...
Mel Gorman [Mon, 7 Oct 2013 10:28:55 +0000 (11:28 +0100)]
sched/numa: Set the scan rate proportional to the memory usage of the task being scanned

The NUMA PTE scan rate is controlled with a combination of the
numa_balancing_scan_period_min, numa_balancing_scan_period_max and
numa_balancing_scan_size. This scan rate is independent of the size
of the task and as an aside it is further complicated by the fact that
numa_balancing_scan_size controls how many pages are marked pte_numa and
not how much virtual memory is scanned.

In combination, it is almost impossible to meaningfully tune the min and
max scan periods and reasoning about performance is complex when the time
to complete a full scan is is partially a function of the tasks memory
size. This patch alters the semantic of the min and max tunables to be
about tuning the length time it takes to complete a scan of a tasks occupied
virtual address space. Conceptually this is a lot easier to understand. There
is a "sanity" check to ensure the scan rate is never extremely fast based on
the amount of virtual memory that should be scanned in a second. The default
of 2.5G seems arbitrary but it is to have the maximum scan rate after the
patch roughly match the maximum scan rate before the patch was applied.

On a similar note, numa_scan_period is in milliseconds and not
jiffies. Properly placed pages slow the scanning rate but adding 10 jiffies
to numa_scan_period means that the rate scanning slows depends on HZ which
is confusing. Get rid of the jiffies_to_msec conversion and treat it as ms.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-18-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Initialise numa_next_scan properly
Mel Gorman [Mon, 7 Oct 2013 10:28:54 +0000 (11:28 +0100)]
sched/numa: Initialise numa_next_scan properly

Scan delay logic and resets are currently initialised to start scanning
immediately instead of delaying properly. Initialise them properly at
fork time and catch when a new mm has been allocated.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-17-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agoRevert "mm: sched: numa: Delay PTE scanning until a task is scheduled on a new node"
Mel Gorman [Mon, 7 Oct 2013 10:28:53 +0000 (11:28 +0100)]
Revert "mm: sched: numa: Delay PTE scanning until a task is scheduled on a new node"

PTE scanning and NUMA hinting fault handling is expensive so commit
5bca2303 ("mm: sched: numa: Delay PTE scanning until a task is scheduled
on a new node") deferred the PTE scan until a task had been scheduled on
another node. The problem is that in the purely shared memory case that
this may never happen and no NUMA hinting fault information will be
captured. We are not ruling out the possibility that something better
can be done here but for now, this patch needs to be reverted and depend
entirely on the scan_delay to avoid punishing short-lived processes.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-16-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Continue PTE scanning even if migrate rate limited
Peter Zijlstra [Mon, 7 Oct 2013 10:28:52 +0000 (11:28 +0100)]
sched/numa: Continue PTE scanning even if migrate rate limited

Avoiding marking PTEs pte_numa because a particular NUMA node is migrate rate
limited sees like a bad idea. Even if this node can't migrate anymore other
nodes might and we want up-to-date information to do balance decisions.
We already rate limit the actual migrations, this should leave enough
bandwidth to allow the non-migrating scanning. I think its important we
keep up-to-date information if we're going to do placement based on it.

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1381141781-10992-15-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Mitigate chance that same task always updates PTEs
Peter Zijlstra [Mon, 7 Oct 2013 10:28:51 +0000 (11:28 +0100)]
sched/numa: Mitigate chance that same task always updates PTEs

With a trace_printk("working\n"); right after the cmpxchg in
task_numa_work() we can see that of a 4 thread process, its always the
same task winning the race and doing the protection change.

This is a problem since the task doing the protection change has a
penalty for taking faults -- it is busy when marking the PTEs. If its
always the same task the ->numa_faults[] get severely skewed.

Avoid this by delaying the task doing the protection change such that
it is unlikely to win the privilege again.

Before:

root@interlagos:~# grep "thread 0/.*working" /debug/tracing/trace | tail -15
      thread 0/0-3232  [022] ....   212.787402: task_numa_work: working
      thread 0/0-3232  [022] ....   212.888473: task_numa_work: working
      thread 0/0-3232  [022] ....   212.989538: task_numa_work: working
      thread 0/0-3232  [022] ....   213.090602: task_numa_work: working
      thread 0/0-3232  [022] ....   213.191667: task_numa_work: working
      thread 0/0-3232  [022] ....   213.292734: task_numa_work: working
      thread 0/0-3232  [022] ....   213.393804: task_numa_work: working
      thread 0/0-3232  [022] ....   213.494869: task_numa_work: working
      thread 0/0-3232  [022] ....   213.596937: task_numa_work: working
      thread 0/0-3232  [022] ....   213.699000: task_numa_work: working
      thread 0/0-3232  [022] ....   213.801067: task_numa_work: working
      thread 0/0-3232  [022] ....   213.903155: task_numa_work: working
      thread 0/0-3232  [022] ....   214.005201: task_numa_work: working
      thread 0/0-3232  [022] ....   214.107266: task_numa_work: working
      thread 0/0-3232  [022] ....   214.209342: task_numa_work: working

After:

root@interlagos:~# grep "thread 0/.*working" /debug/tracing/trace | tail -15
      thread 0/0-3253  [005] ....   136.865051: task_numa_work: working
      thread 0/2-3255  [026] ....   136.965134: task_numa_work: working
      thread 0/3-3256  [024] ....   137.065217: task_numa_work: working
      thread 0/3-3256  [024] ....   137.165302: task_numa_work: working
      thread 0/3-3256  [024] ....   137.265382: task_numa_work: working
      thread 0/0-3253  [004] ....   137.366465: task_numa_work: working
      thread 0/2-3255  [026] ....   137.466549: task_numa_work: working
      thread 0/0-3253  [004] ....   137.566629: task_numa_work: working
      thread 0/0-3253  [004] ....   137.666711: task_numa_work: working
      thread 0/1-3254  [028] ....   137.766799: task_numa_work: working
      thread 0/0-3253  [004] ....   137.866876: task_numa_work: working
      thread 0/2-3255  [026] ....   137.966960: task_numa_work: working
      thread 0/1-3254  [028] ....   138.067041: task_numa_work: working
      thread 0/2-3255  [026] ....   138.167123: task_numa_work: working
      thread 0/3-3256  [024] ....   138.267207: task_numa_work: working

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1381141781-10992-14-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: numa: Do not migrate or account for hinting faults on the zero page
Mel Gorman [Mon, 7 Oct 2013 10:28:50 +0000 (11:28 +0100)]
mm: numa: Do not migrate or account for hinting faults on the zero page

The zero page is not replicated between nodes and is often shared between
processes. The data is read-only and likely to be cached in local CPUs
if heavily accessed meaning that the remote memory access cost is less
of a concern. This patch prevents trapping faults on the zero pages. For
tasks using the zero page this will reduce the number of PTE updates,
TLB flushes and hinting faults.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
[ Correct use of is_huge_zero_page]
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-13-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: Only flush TLBs if a transhuge PMD is modified for NUMA pte scanning
Mel Gorman [Mon, 7 Oct 2013 10:28:49 +0000 (11:28 +0100)]
mm: Only flush TLBs if a transhuge PMD is modified for NUMA pte scanning

NUMA PTE scanning is expensive both in terms of the scanning itself and
the TLB flush if there are any updates. The TLB flush is avoided if no
PTEs are updated but there is a bug where transhuge PMDs are considered
to be updated even if they were already pmd_numa. This patch addresses
the problem and TLB flushes should be reduced.

Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-12-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: Do not flush TLB during protection change if !pte_present && !migration_entry
Mel Gorman [Mon, 7 Oct 2013 10:28:48 +0000 (11:28 +0100)]
mm: Do not flush TLB during protection change if !pte_present && !migration_entry

NUMA PTE scanning is expensive both in terms of the scanning itself and
the TLB flush if there are any updates. Currently non-present PTEs are
accounted for as an update and incurring a TLB flush where it is only
necessary for anonymous migration entries. This patch addresses the
problem and should reduce TLB flushes.

Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-11-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: Account for a THP NUMA hinting update as one PTE update
Mel Gorman [Mon, 7 Oct 2013 10:28:47 +0000 (11:28 +0100)]
mm: Account for a THP NUMA hinting update as one PTE update

A THP PMD update is accounted for as 512 pages updated in vmstat.  This is
large difference when estimating the cost of automatic NUMA balancing and
can be misleading when comparing results that had collapsed versus split
THP. This patch addresses the accounting issue.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-10-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: Close races between THP migration and PMD numa clearing
Mel Gorman [Mon, 7 Oct 2013 10:28:46 +0000 (11:28 +0100)]
mm: Close races between THP migration and PMD numa clearing

THP migration uses the page lock to guard against parallel allocations
but there are cases like this still open

  Task A Task B
  --------------------- ---------------------
  do_huge_pmd_numa_page do_huge_pmd_numa_page
  lock_page
  mpol_misplaced == -1
  unlock_page
  goto clear_pmdnuma
lock_page
mpol_misplaced == 2
migrate_misplaced_transhuge
  pmd = pmd_mknonnuma
  set_pmd_at

During hours of testing, one crashed with weird errors and while I have
no direct evidence, I suspect something like the race above happened.
This patch extends the page lock to being held until the pmd_numa is
cleared to prevent migration starting in parallel while the pmd_numa is
being cleared. It also flushes the old pmd entry and orders pagetable
insertion before rmap insertion.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-9-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: numa: Sanitize task_numa_fault() callsites
Mel Gorman [Mon, 7 Oct 2013 10:28:45 +0000 (11:28 +0100)]
mm: numa: Sanitize task_numa_fault() callsites

There are three callers of task_numa_fault():

 - do_huge_pmd_numa_page():
     Accounts against the current node, not the node where the
     page resides, unless we migrated, in which case it accounts
     against the node we migrated to.

 - do_numa_page():
     Accounts against the current node, not the node where the
     page resides, unless we migrated, in which case it accounts
     against the node we migrated to.

 - do_pmd_numa_page():
     Accounts not at all when the page isn't migrated, otherwise
     accounts against the node we migrated towards.

This seems wrong to me; all three sites should have the same
sementaics, furthermore we should accounts against where the page
really is, we already know where the task is.

So modify all three sites to always account; we did after all receive
the fault; and always account to where the page is after migration,
regardless of success.

They all still differ on when they clear the PTE/PMD; ideally that
would get sorted too.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-8-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: Prevent parallel splits during THP migration
Mel Gorman [Mon, 7 Oct 2013 10:28:44 +0000 (11:28 +0100)]
mm: Prevent parallel splits during THP migration

THP migrations are serialised by the page lock but on its own that does
not prevent THP splits. If the page is split during THP migration then
the pmd_same checks will prevent page table corruption but the unlock page
and other fix-ups potentially will cause corruption. This patch takes the
anon_vma lock to prevent parallel splits during migration.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-7-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: Wait for THP migrations to complete during NUMA hinting faults
Mel Gorman [Mon, 7 Oct 2013 10:28:43 +0000 (11:28 +0100)]
mm: Wait for THP migrations to complete during NUMA hinting faults

The locking for migrating THP is unusual. While normal page migration
prevents parallel accesses using a migration PTE, THP migration relies on
a combination of the page_table_lock, the page lock and the existance of
the NUMA hinting PTE to guarantee safety but there is a bug in the scheme.

If a THP page is currently being migrated and another thread traps a
fault on the same page it checks if the page is misplaced. If it is not,
then pmd_numa is cleared. The problem is that it checks if the page is
misplaced without holding the page lock meaning that the racing thread
can be migrating the THP when the second thread clears the NUMA bit
and faults a stale page.

This patch checks if the page is potentially being migrated and stalls
using the lock_page if it is potentially being migrated before checking
if the page is misplaced or not.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-6-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: numa: Do not account for a hinting fault if we raced
Mel Gorman [Mon, 7 Oct 2013 10:28:42 +0000 (11:28 +0100)]
mm: numa: Do not account for a hinting fault if we raced

If another task handled a hinting fault in parallel then do not double
account for it.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-5-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/numa: Fix comments
Peter Zijlstra [Mon, 7 Oct 2013 10:28:41 +0000 (11:28 +0100)]
sched/numa: Fix comments

Fix a 80 column violation and a PTE vs PMD reference.

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1381141781-10992-4-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agomm: numa: Document automatic NUMA balancing sysctls
Mel Gorman [Mon, 7 Oct 2013 10:28:40 +0000 (11:28 +0100)]
mm: numa: Document automatic NUMA balancing sysctls

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1381141781-10992-3-git-send-email-mgorman@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agoMerge tag 'v3.12-rc4' into sched/core
Ingo Molnar [Wed, 9 Oct 2013 10:36:13 +0000 (12:36 +0200)]
Merge tag 'v3.12-rc4' into sched/core

Merge Linux v3.12-rc4 to fix a conflict and also to refresh the tree
before applying more scheduler patches.

Conflicts:
arch/avr32/include/asm/Kbuild

Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agoLinux 3.12-rc4
Linus Torvalds [Sun, 6 Oct 2013 21:00:20 +0000 (14:00 -0700)]
Linux 3.12-rc4

11 years agonet: Update the sysctl permissions handler to test effective uid/gid
Eric W. Biederman [Sat, 5 Oct 2013 20:15:30 +0000 (13:15 -0700)]
net: Update the sysctl permissions handler to test effective uid/gid

Modify the code to use current_euid(), and in_egroup_p, as in done
in fs/proc/proc_sysctl.c:test_perm()

Cc: stable@vger.kernel.org
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reported-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending
Linus Torvalds [Sun, 6 Oct 2013 20:38:31 +0000 (13:38 -0700)]
Merge git://git./linux/kernel/git/nab/target-pending

Pull SCSI target fixes from Nicholas Bellinger:
 "Here are the outstanding target fixes queued up for v3.12-rc4 code.

  The highlights include:

   - Make vhost/scsi tag percpu_ida_alloc() use GFP_ATOMIC
   - Allow sess_cmd_map allocation failure fallback to use vzalloc
   - Fix COMPARE_AND_WRITE se_cmd->data_length bug with FILEIO backends
   - Fixes for COMPARE_AND_WRITE callback recursive failure OOPs + non
     zero scsi_status bug
   - Make iscsi-target do acknowledgement tag release from RX context
   - Setup iscsi-target with extra (cmdsn_depth / 2) percpu_ida tags

  Also included is a iscsi-target patch CC'ed for v3.10+ that avoids
  legacy wait_for_task=true release during fast-past StatSN
  acknowledgement, and two other SRP target related patches that address
  long-standing issues that are CC'ed for v3.3+.

  Extra thanks to Thomas Glanzmann for his testing feedback with
  COMPARE_AND_WRITE + EXTENDED_COPY VAAI logic"

* git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending:
  iscsi-target; Allow an extra tag_num / 2 number of percpu_ida tags
  iscsi-target: Perform release of acknowledged tags from RX context
  iscsi-target: Only perform wait_for_tasks when performing shutdown
  target: Fail on non zero scsi_status in compare_and_write_callback
  target: Fix recursive COMPARE_AND_WRITE callback failure
  target: Reset data_length for COMPARE_AND_WRITE to NoLB * block_size
  ib_srpt: always set response for task management
  target: Fall back to vzalloc upon ->sess_cmd_map kzalloc failure
  vhost/scsi: Use GFP_ATOMIC with percpu_ida_alloc for obtaining tag
  ib_srpt: Destroy cm_id before destroying QP.
  target: Fix xop->dbl assignment in target_xcopy_parse_segdesc_02

11 years agoMerge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 6 Oct 2013 20:35:15 +0000 (13:35 -0700)]
Merge branch 'fixes' of git://git.infradead.org/users/vkoul/slave-dma

Pull slave-dmaengine fixes from Vinod Koul:
 "Here is the slave dmanegine fixes.  We have the fix for deadlock issue
  on imx-dma by Michael and Josh's edma config fix along with author
  change"

* 'fixes' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: imx-dma: fix callback path in tasklet
  dmaengine: imx-dma: fix lockdep issue between irqhandler and tasklet
  dmaengine: imx-dma: fix slow path issue in prep_dma_cyclic
  dma/Kconfig: Make TI_EDMA select TI_PRIV_EDMA
  edma: Update author email address

11 years agosched/rt: Remove redundant nr_cpus_allowed test
Shawn Bohrer [Fri, 4 Oct 2013 19:24:53 +0000 (14:24 -0500)]
sched/rt: Remove redundant nr_cpus_allowed test

In 76854c7e8f3f4172fef091e78d88b3b751463ac6 ("sched: Use
rt.nr_cpus_allowed to recover select_task_rq() cycles") an
optimization was added to select_task_rq_rt() that immediately
returns when p->nr_cpus_allowed == 1 at the beginning of the
function.

This makes the latter p->nr_cpus_allowed > 1 check redundant,
which can now be removed.

Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Reviewed-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Mike Galbraith <mgalbraith@suse.de>
Cc: tomk@rgmadvisors.com
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1380914693-24634-1-git-send-email-shawn.bohrer@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sat, 5 Oct 2013 19:17:24 +0000 (12:17 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "This is a small collection of fixes, including a regression fix from
  Liu Bo that solves rare crashes with compression on.

  I've merged my for-linus up to 3.12-rc3 because the top commit is only
  meant for 3.12.  The rest of the fixes are also available in my master
  branch on top of my last 3.11 based pull"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: Fix crash due to not allocating integrity data for a bioset
  Btrfs: fix a use-after-free bug in btrfs_dev_replace_finishing
  Btrfs: eliminate races in worker stopping code
  Btrfs: fix crash of compressed writes
  Btrfs: fix transid verify errors when recovering log tree

11 years agoMerge tag 'gpio-v3.12-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Sat, 5 Oct 2013 19:11:40 +0000 (12:11 -0700)]
Merge tag 'gpio-v3.12-2' of git://git./linux/kernel/git/linusw/linux-gpio

Pull GPIO fixes from Linus Walleij:
 "Two patches for the OMAP driver, dealing with setting up IRQs properly
  on the device tree boot path"

* tag 'gpio-v3.12-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
  gpio/omap: auto-setup a GPIO when used as an IRQ
  gpio/omap: maintain GPIO and IRQ usage separately

11 years agoMerge tag 'usb-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 5 Oct 2013 18:54:10 +0000 (11:54 -0700)]
Merge tag 'usb-3.12-rc4' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are none fixes for various USB driver problems.  The majority are
  gadget/musb fixes, but there are some new device ids in here as well"

* tag 'usb-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: chipidea: add Intel Clovertrail pci id
  usb: gadget: s3c-hsotg: fix can_write limit for non-periodic endpoints
  usb: gadget: f_fs: fix error handling
  usb: musb: dsps: do not bind to "musb-hdrc"
  USB: serial: option: Ignore card reader interface on Huawei E1750
  usb: musb: gadget: fix otg active status flag
  usb: phy: gpio-vbus: fix deferred probe from __init
  usb: gadget: pxa25x_udc: fix deferred probe from __init
  usb: musb: fix otg default state

11 years agoMerge tag 'tty-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sat, 5 Oct 2013 18:26:19 +0000 (11:26 -0700)]
Merge tag 'tty-3.12-rc4' of git://git./linux/kernel/git/gregkh/tty

Pull tty fixes from Greg KH:
 "Here are two tty driver fixes for 3.12-rc4.

  One fixes the reported regression in the n_tty code that a number of
  people found recently, and the other one fixes an issue with xen
  consoles that broke in 3.10"

* tag 'tty-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  xen/hvc: allow xenboot console to be used again
  tty: Fix pty master read() after slave closes

11 years agoMerge tag 'staging-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 5 Oct 2013 18:25:38 +0000 (11:25 -0700)]
Merge tag 'staging-3.12-rc4' of git://git./linux/kernel/git/gregkh/staging

Pull staging fixes from Greg KH:
 "Here are 4 tiny staging and iio driver fixes for 3.12-rc4.  Nothing
  major, just some small fixes for reported issues"

* tag 'staging-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: comedi: ni_65xx: (bug fix) confine insn_bits to one subdevice
  iio:magnetometer: Bugfix magnetometer default output registers
  iio: Remove debugfs entries in iio_device_unregister()
  iio: amplifiers: ad8366: Remove regulator_put

11 years agobtrfs: Fix crash due to not allocating integrity data for a bioset
Darrick J. Wong [Fri, 20 Sep 2013 03:37:07 +0000 (20:37 -0700)]
btrfs: Fix crash due to not allocating integrity data for a bioset

When btrfs creates a bioset, we must also allocate the integrity data pool.
Otherwise btrfs will crash when it tries to submit a bio to a checksumming
disk:

 BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
 IP: [<ffffffff8111e28a>] mempool_alloc+0x4a/0x150
 PGD 2305e4067 PUD 23063d067 PMD 0
 Oops: 0000 [#1] PREEMPT SMP
 Modules linked in: btrfs scsi_debug xfs ext4 jbd2 ext3 jbd mbcache
sch_fq_codel eeprom lpc_ich mfd_core nfsd exportfs auth_rpcgss af_packet
raid6_pq xor zlib_deflate libcrc32c [last unloaded: scsi_debug]
 CPU: 1 PID: 4486 Comm: mount Not tainted 3.12.0-rc1-mcsum #2
 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
 task: ffff8802451c9720 ti: ffff880230698000 task.ti: ffff880230698000
 RIP: 0010:[<ffffffff8111e28a>]  [<ffffffff8111e28a>] mempool_alloc+0x4a/0x150
 RSP: 0018:ffff880230699688  EFLAGS: 00010286
 RAX: 0000000000000001 RBX: 0000000000000000 RCX: 00000000005f8445
 RDX: 0000000000000001 RSI: 0000000000000010 RDI: 0000000000000000
 RBP: ffff8802306996f8 R08: 0000000000011200 R09: 0000000000000008
 R10: 0000000000000020 R11: ffff88009d6e8000 R12: 0000000000011210
 R13: 0000000000000030 R14: ffff8802306996b8 R15: ffff8802451c9720
 FS:  00007f25b8a16800(0000) GS:ffff88024fc80000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
 CR2: 0000000000000018 CR3: 0000000230576000 CR4: 00000000000007e0
 Stack:
  ffff8802451c9720 0000000000000002 ffffffff81a97100 0000000000281250
  ffffffff81a96480 ffff88024fc99150 ffff880228d18200 0000000000000000
  0000000000000000 0000000000000040 ffff880230e8c2e8 ffff8802459dc900
 Call Trace:
  [<ffffffff811b2208>] bio_integrity_alloc+0x48/0x1b0
  [<ffffffff811b26fc>] bio_integrity_prep+0xac/0x360
  [<ffffffff8111e298>] ? mempool_alloc+0x58/0x150
  [<ffffffffa03e8041>] ? alloc_extent_state+0x31/0x110 [btrfs]
  [<ffffffff81241579>] blk_queue_bio+0x1c9/0x460
  [<ffffffff8123e58a>] generic_make_request+0xca/0x100
  [<ffffffff8123e639>] submit_bio+0x79/0x160
  [<ffffffffa03f865e>] btrfs_map_bio+0x48e/0x5b0 [btrfs]
  [<ffffffffa03c821a>] btree_submit_bio_hook+0xda/0x110 [btrfs]
  [<ffffffffa03e7eba>] submit_one_bio+0x6a/0xa0 [btrfs]
  [<ffffffffa03ef450>] read_extent_buffer_pages+0x250/0x310 [btrfs]
  [<ffffffff8125eef6>] ? __radix_tree_preload+0x66/0xf0
  [<ffffffff8125f1c5>] ? radix_tree_insert+0x95/0x260
  [<ffffffffa03c66f6>] btree_read_extent_buffer_pages.constprop.128+0xb6/0x120
[btrfs]
  [<ffffffffa03c8c1a>] read_tree_block+0x3a/0x60 [btrfs]
  [<ffffffffa03caefd>] open_ctree+0x139d/0x2030 [btrfs]
  [<ffffffffa03a282a>] btrfs_mount+0x53a/0x7d0 [btrfs]
  [<ffffffff8113ab0b>] ? pcpu_alloc+0x8eb/0x9f0
  [<ffffffff81167305>] ? __kmalloc_track_caller+0x35/0x1e0
  [<ffffffff81176ba0>] mount_fs+0x20/0xd0
  [<ffffffff81191096>] vfs_kern_mount+0x76/0x120
  [<ffffffff81193320>] do_mount+0x200/0xa40
  [<ffffffff81135cdb>] ? strndup_user+0x5b/0x80
  [<ffffffff81193bf0>] SyS_mount+0x90/0xe0
  [<ffffffff8156d31d>] system_call_fastpath+0x1a/0x1f
 Code: 4c 8d 75 a8 4c 89 6d e8 45 89 e0 4c 8d 6f 30 48 89 5d d8 41 83 e0 af 48
89 fb 49 83 c6 18 4c 89 7d f8 65 4c 8b 3c 25 c0 b8 00 00 <48> 8b 73 18 44 89 c7
44 89 45 98 ff 53 20 48 85 c0 48 89 c2 74
 RIP  [<ffffffff8111e28a>] mempool_alloc+0x4a/0x150
  RSP <ffff880230699688>
 CR2: 0000000000000018
 ---[ end trace 7a96042017ed21e2 ]---

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
11 years agoMerge branch 'for-linus' into for-linus-3.12
Chris Mason [Sat, 5 Oct 2013 14:51:32 +0000 (10:51 -0400)]
Merge branch 'for-linus' into for-linus-3.12

11 years agoMerge branch 'for-linus' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 5 Oct 2013 03:50:16 +0000 (20:50 -0700)]
Merge branch 'for-linus' of git://git.samba.org/sfrench/cifs-2.6

Pull CIFS fixes from Steve French:
 "Small set of cifs fixes.  Most important is Jeff's fix that works
  around disconnection problems which can be caused by simultaneous use
  of user space tools (starting a long running smbclient backup then
  doing a cifs kernel mount) or multiple cifs mounts through a NAT, and
  Jim's fix to deal with reexport of cifs share.

  I expect to send two more cifs fixes next week (being tested now) -
  fixes to address an SMB2 unmount hang when server dies and a fix for
  cifs symlink handling of Windows "NFS" symlinks"

* 'for-linus' of git://git.samba.org/sfrench/cifs-2.6:
  [CIFS] update cifs.ko version
  [CIFS] Remove ext2 flags that have been moved to fs.h
  [CIFS] Provide sane values for nlink
  cifs: stop trying to use virtual circuits
  CIFS: FS-Cache: Uncache unread pages in cifs_readpages() before freeing them

11 years agoMerge tag 'pci-v3.12-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Sat, 5 Oct 2013 03:48:20 +0000 (20:48 -0700)]
Merge tag 'pci-v3.12-fixes-1' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fix from Bjorn Helgaas:
 "We merged what was intended to be an MMCONFIG cleanup, but in fact,
  for systems without _CBA (which is almost everything), it broke
  extended config space for domain 0 and it broke all config space for
  other domains.

  This reverts the change"

* tag 'pci-v3.12-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  Revert "x86/PCI: MMCONFIG: Check earlier for MMCONFIG region at address zero"

11 years agoRevert "x86/PCI: MMCONFIG: Check earlier for MMCONFIG region at address zero"
Bjorn Helgaas [Fri, 4 Oct 2013 22:14:30 +0000 (16:14 -0600)]
Revert "x86/PCI: MMCONFIG: Check earlier for MMCONFIG region at address zero"

This reverts commit 07f9b61c3915e8eb156cb4461b3946736356ad02.

07f9b61c was intended to be a cleanup that didn't change anything, but in
fact, for systems without _CBA (which is almost everything), it broke
extended config space for domain 0 and all config space for other domains.

Reference: http://lkml.kernel.org/r/20131004011806.GE20450@dangermouse.emea.sgi.com
Reported-by: Hedi Berriche <hedi@sgi.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
11 years agoMerge tag 'pm+acpi-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 4 Oct 2013 22:03:42 +0000 (15:03 -0700)]
Merge tag 'pm+acpi-3.12-rc4' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI and power management fixes from Rafael Wysocki:

 - The resume part of user space driven hibernation (s2disk) is now
   broken after the change that moved the creation of memory bitmaps to
   after the freezing of tasks, because I forgot that the resume utility
   loaded the image before freezing tasks and needed the bitmaps for
   that.  The fix adds special handling for that case.

 - One of recent commits changed the export of acpi_bus_get_device() to
   EXPORT_SYMBOL_GPL(), which was technically correct but broke existing
   binary modules using that function including one in particularly
   widespread use.  Change it back to EXPORT_SYMBOL().

 - The intel_pstate driver sometimes fails to disable turbo if its
   no_turbo sysfs attribute is set.  Fix from Srinivas Pandruvada.

 - One of recent cpufreq fixes forgot to update a check in cpufreq-cpu0
   which still (incorrectly) treats non-NULL as non-error.  Fix from
   Philipp Zabel.

 - The SPEAr cpufreq driver uses a wrong variable type in one place
   preventing it from catching errors returned by one of the functions
   called by it.  Fix from Sachin Kamat.

* tag 'pm+acpi-3.12-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI: Use EXPORT_SYMBOL() for acpi_bus_get_device()
  intel_pstate: fix no_turbo
  cpufreq: cpufreq-cpu0: NULL is a valid regulator, part 2
  cpufreq: SPEAr: Fix incorrect variable type
  PM / hibernate: Fix user space driven resume regression

11 years agoMerge tag 'xfs-for-linus-v3.12-rc4' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Fri, 4 Oct 2013 21:47:22 +0000 (14:47 -0700)]
Merge tag 'xfs-for-linus-v3.12-rc4' of git://oss.sgi.com/xfs/xfs

Pull xfs bugfixes from Ben Myers:
 "There are lockdep annotations for project quotas, a fix for dirent
  dtype support on v4 filesystems, a fix for a memory leak in recovery,
  and a fix for the build error that resulted from it.  D'oh"

* tag 'xfs-for-linus-v3.12-rc4' of git://oss.sgi.com/xfs/xfs:
  xfs: Use kmem_free() instead of free()
  xfs: fix memory leak in xlog_recover_add_to_trans
  xfs: dirent dtype presence is dependent on directory magic numbers
  xfs: lockdep needs to know about 3 dquot-deep nesting

11 years agoselinux: remove 'flags' parameter from avc_audit()
Linus Torvalds [Fri, 4 Oct 2013 21:05:38 +0000 (14:05 -0700)]
selinux: remove 'flags' parameter from avc_audit()

Now avc_audit() has no more users with that parameter. Remove it.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoselinux: avc_has_perm_flags has no more users
Linus Torvalds [Fri, 4 Oct 2013 19:57:22 +0000 (12:57 -0700)]
selinux: avc_has_perm_flags has no more users

.. so get rid of it.  The only indirect users were all the
avc_has_perm() callers which just expanded to have a zero flags
argument.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoBtrfs: fix a use-after-free bug in btrfs_dev_replace_finishing
Ilya Dryomov [Wed, 2 Oct 2013 17:41:01 +0000 (20:41 +0300)]
Btrfs: fix a use-after-free bug in btrfs_dev_replace_finishing

free_device rcu callback, scheduled from btrfs_rm_dev_replace_srcdev,
can be processed before btrfs_scratch_superblock is called, which would
result in a use-after-free on btrfs_device contents.  Fix this by
zeroing the superblock before the rcu callback is registered.

Cc: Stefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
11 years agoBtrfs: eliminate races in worker stopping code
Ilya Dryomov [Wed, 2 Oct 2013 16:39:50 +0000 (19:39 +0300)]
Btrfs: eliminate races in worker stopping code

The current implementation of worker threads in Btrfs has races in
worker stopping code, which cause all kinds of panics and lockups when
running btrfs/011 xfstest in a loop.  The problem is that
btrfs_stop_workers is unsynchronized with respect to check_idle_worker,
check_busy_worker and __btrfs_start_workers.

E.g., check_idle_worker race flow:

       btrfs_stop_workers():            check_idle_worker(aworker):
- grabs the lock
- splices the idle list into the
  working list
- removes the first worker from the
  working list
- releases the lock to wait for
  its kthread's completion
                                  - grabs the lock
                                  - if aworker is on the working list,
                                    moves aworker from the working list
                                    to the idle list
                                  - releases the lock
- grabs the lock
- puts the worker
- removes the second worker from the
  working list
                              ......
        btrfs_stop_workers returns, aworker is on the idle list
                 FS is umounted, memory is freed
                              ......
              aworker is waken up, fireworks ensue

With this applied, I wasn't able to trigger the problem in 48 hours,
whereas previously I could reliably reproduce at least one of these
races within an hour.

Reported-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
11 years agoBtrfs: fix crash of compressed writes
Liu Bo [Tue, 1 Oct 2013 15:49:49 +0000 (23:49 +0800)]
Btrfs: fix crash of compressed writes

The crash[1] is found by xfstests/generic/208 with "-o compress",
it's not reproduced everytime, but it does panic.

The bug is quite interesting, it's actually introduced by a recent commit
(573aecafca1cf7a974231b759197a1aebcf39c2a,
Btrfs: actually limit the size of delalloc range).

Btrfs implements delay allocation, so during writeback, we
(1) get a page A and lock it
(2) search the state tree for delalloc bytes and lock all pages within the range
(3) process the delalloc range, including find disk space and create
    ordered extent and so on.
(4) submit the page A.

It runs well in normal cases, but if we're in a racy case, eg.
buffered compressed writes and aio-dio writes,
sometimes we may fail to lock all pages in the 'delalloc' range,
in which case, we need to fall back to search the state tree again with
a smaller range limit(max_bytes = PAGE_CACHE_SIZE - offset).

The mentioned commit has a side effect, that is, in the fallback case,
we can find delalloc bytes before the index of the page we already have locked,
so we're in the case of (delalloc_end <= *start) and return with (found > 0).

This ends with not locking delalloc pages but making ->writepage still
process them, and the crash happens.

This fixes it by just thinking that we find nothing and returning to caller
as the caller knows how to deal with it properly.

[1]:
------------[ cut here ]------------
kernel BUG at mm/page-writeback.c:2170!
[...]
CPU: 2 PID: 11755 Comm: btrfs-delalloc- Tainted: G           O 3.11.0+ #8
[...]
RIP: 0010:[<ffffffff810f5093>]  [<ffffffff810f5093>] clear_page_dirty_for_io+0x1e/0x83
[...]
[ 4934.248731] Stack:
[ 4934.248731]  ffff8801477e5dc8 ffffea00049b9f00 ffff8801869f9ce8 ffffffffa02b841a
[ 4934.248731]  0000000000000000 0000000000000000 0000000000000fff 0000000000000620
[ 4934.248731]  ffff88018db59c78 ffffea0005da8d40 ffffffffa02ff860 00000001810016c0
[ 4934.248731] Call Trace:
[ 4934.248731]  [<ffffffffa02b841a>] extent_range_clear_dirty_for_io+0xcf/0xf5 [btrfs]
[ 4934.248731]  [<ffffffffa02a8889>] compress_file_range+0x1dc/0x4cb [btrfs]
[ 4934.248731]  [<ffffffff8104f7af>] ? detach_if_pending+0x22/0x4b
[ 4934.248731]  [<ffffffffa02a8bad>] async_cow_start+0x35/0x53 [btrfs]
[ 4934.248731]  [<ffffffffa02c694b>] worker_loop+0x14b/0x48c [btrfs]
[ 4934.248731]  [<ffffffffa02c6800>] ? btrfs_queue_worker+0x25c/0x25c [btrfs]
[ 4934.248731]  [<ffffffff810608f5>] kthread+0x8d/0x95
[ 4934.248731]  [<ffffffff81060868>] ? kthread_freezable_should_stop+0x43/0x43
[ 4934.248731]  [<ffffffff814fe09c>] ret_from_fork+0x7c/0xb0
[ 4934.248731]  [<ffffffff81060868>] ? kthread_freezable_should_stop+0x43/0x43
[ 4934.248731] Code: ff 85 c0 0f 94 c0 0f b6 c0 59 5b 5d c3 0f 1f 44 00 00 55 48 89 e5 41 54 53 48 89 fb e8 2c de 00 00 49 89 c4 48 8b 03 a8 01 75 02 <0f> 0b 4d 85 e4 74 52 49 8b 84 24 80 00 00 00 f6 40 20 01 75 44
[ 4934.248731] RIP  [<ffffffff810f5093>] clear_page_dirty_for_io+0x1e/0x83
[ 4934.248731]  RSP <ffff8801869f9c48>
[ 4934.280307] ---[ end trace 36f06d3f8750236a ]---

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
11 years agoBtrfs: fix transid verify errors when recovering log tree
Josef Bacik [Mon, 30 Sep 2013 18:10:43 +0000 (14:10 -0400)]
Btrfs: fix transid verify errors when recovering log tree

If we crash with a log, remount and recover that log, and then crash before we
can commit another transaction we will get transid verify errors on the next
mount.  This is because we were not zero'ing out the log when we committed the
transaction after recovery.  This is ok as long as we commit another transaction
at some point in the future, but if you abort or something else goes wrong you
can end up in this weird state because the recovery stuff says that the tree log
should have a generation+1 of the super generation, which won't be the case of
the transaction that was started for recovery.  Fix this by removing the check
and _always_ zero out the log portion of the super when we commit a transaction.
This fixes the transid verify issues I was seeing with my force errors tests.
Thanks,

Signed-off-by: Josef Bacik <jbacik@fusionio.com>
11 years agoselinux: remove 'flags' parameter from inode_has_perm
Linus Torvalds [Fri, 4 Oct 2013 19:54:11 +0000 (12:54 -0700)]
selinux: remove 'flags' parameter from inode_has_perm

Every single user passes in '0'.  I think we had non-zero users back in
some stone age when selinux_inode_permission() was implemented in terms
of inode_has_perm(), but that complicated case got split up into a
totally separate code-path so that we could optimize the much simpler
special cases.

See commit 2e33405785d3 ("SELinux: delay initialization of audit data in
selinux_inode_permission") for example.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoxfs: Use kmem_free() instead of free()
Thierry Reding [Tue, 1 Oct 2013 14:47:53 +0000 (16:47 +0200)]
xfs: Use kmem_free() instead of free()

This fixes a build failure caused by calling the free() function which
does not exist in the Linux kernel.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit aaaae98022efa4f3c31042f1fdf9e7a0c5f04663)

11 years agoxfs: fix memory leak in xlog_recover_add_to_trans
tinguely@sgi.com [Fri, 27 Sep 2013 14:00:55 +0000 (09:00 -0500)]
xfs: fix memory leak in xlog_recover_add_to_trans

Free the memory in error path of xlog_recover_add_to_trans().
Normally this memory is freed in recovery pass2, but is leaked
in the error path.

Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 519ccb81ac1c8e3e4eed294acf93be00b43dcad6)

11 years agoxfs: dirent dtype presence is dependent on directory magic numbers
Dave Chinner [Sun, 29 Sep 2013 23:37:04 +0000 (09:37 +1000)]
xfs: dirent dtype presence is dependent on directory magic numbers

The determination of whether a directory entry contains a dtype
field originally was dependent on the filesystem having CRCs
enabled. This meant that the format for dtype beign enabled could be
determined by checking the directory block magic number rather than
doing a feature bit check. This was useful in that it meant that we
didn't need to pass a struct xfs_mount around to functions that
were already supplied with a directory block header.

Unfortunately, the introduction of dtype fields into the v4
structure via a feature bit meant this "use the directory block
magic number" method of discriminating the dirent entry sizes is
broken. Hence we need to convert the places that use magic number
checks to use feature bit checks so that they work correctly and not
by chance.

The current code works on v4 filesystems only because the dirent
size roundup covers the extra byte needed by the dtype field in the
places where this problem occurs.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 367993e7c6428cb7617ab7653d61dca54e2fdede)

11 years agoxfs: lockdep needs to know about 3 dquot-deep nesting
Dave Chinner [Sun, 29 Sep 2013 23:37:03 +0000 (09:37 +1000)]
xfs: lockdep needs to know about 3 dquot-deep nesting

Michael Semon reported that xfs/299 generated this lockdep warning:

=============================================
[ INFO: possible recursive locking detected ]
3.12.0-rc2+ #2 Not tainted
---------------------------------------------
touch/21072 is trying to acquire lock:
 (&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64

but task is already holding lock:
 (&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&xfs_dquot_other_class);
  lock(&xfs_dquot_other_class);

 *** DEADLOCK ***

 May be due to missing lock nesting notation

7 locks held by touch/21072:
 #0:  (sb_writers#10){++++.+}, at: [<c11185b6>] mnt_want_write+0x1e/0x3e
 #1:  (&type->i_mutex_dir_key#4){+.+.+.}, at: [<c11078ee>] do_last+0x245/0xe40
 #2:  (sb_internal#2){++++.+}, at: [<c122c9e0>] xfs_trans_alloc+0x1f/0x35
 #3:  (&(&ip->i_lock)->mr_lock/1){+.+...}, at: [<c126cd1b>] xfs_ilock+0x100/0x1f1
 #4:  (&(&ip->i_lock)->mr_lock){++++-.}, at: [<c126cf52>] xfs_ilock_nowait+0x105/0x22f
 #5:  (&dqp->q_qlock){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
 #6:  (&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64

The lockdep annotation for dquot lock nesting only understands
locking for user and "other" dquots, not user, group and quota
dquots. Fix the annotations to match the locking heirarchy we now
have.

Reported-by: Michael L. Semon <mlsemon35@gmail.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit f112a049712a5c07de25d511c3c6587a2b1a015e)

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi...
Linus Torvalds [Fri, 4 Oct 2013 16:06:13 +0000 (09:06 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mszeredi/fuse

Pull fuse bugfixes from Miklos Szeredi:
 "This contains two more fixes by Maxim for writeback/truncate races and
  fixes for RCU walk in fuse_dentry_revalidate()"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
  fuse: no RCU mode in fuse_access()
  fuse: readdirplus: fix RCU walk
  fuse: don't check_submounts_and_drop() in RCU walk
  fuse: fix fallocate vs. ftruncate race
  fuse: wait for writeback in fuse_file_fallocate()

11 years agoMerge tag 'iommu-fixes-v3.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 4 Oct 2013 16:05:12 +0000 (09:05 -0700)]
Merge tag 'iommu-fixes-v3.12-rc3' of git://git./linux/kernel/git/joro/iommu

Pull iommu fixes from Joerg Roedel:
 "A couple of fixes from the IOMMU side:

   - some small fixes for the new ARM-SMMU driver
   - a register offset correction for VT-d
   - add MAINTAINERS entry for drivers/iommu

  Overall no really big or intrusive changes"

* tag 'iommu-fixes-v3.12-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  x86/iommu: correct ICS register offset
  MAINTAINERS: add overall IOMMU section
  iommu/arm-smmu: don't enable SMMU device until probing has completed
  iommu/arm-smmu: fix iommu_present() test in init
  iommu/arm-smmu: fix a signedness bug

11 years agoMerge tag 'arm64-stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmarinas...
Linus Torvalds [Fri, 4 Oct 2013 16:04:26 +0000 (09:04 -0700)]
Merge tag 'arm64-stable' of git://git./linux/kernel/git/cmarinas/linux-aarch64

Pull ARM64 fixes/updates from Catalin Marinas:
 - Bug-fixes (get_user/put_user, incorrect register width for ASID,
   FPSIMD initialisation)
 - Kconfig clean-up
 - defconfig update

* tag 'arm64-stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmarinas/linux-aarch64:
  arm64: Remove duplicate DEBUG_STACK_USAGE config
  arm64: include VIRTIO_{MMIO,BLK} in defconfig
  arm64: include EXT4 in defconfig
  arm64: fix possible invalid FPSIMD initialization state
  arm64: use correct register width when retrieving ASID
  arm64: avoid multiple evaluation of ptr in get_user/put_user()

11 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Fri, 4 Oct 2013 16:03:51 +0000 (09:03 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:
 "Two small fixes for 3.12 only this week.  I have a few more fixes
  pending but those are conceptually more complex so will have to wait
  for a bit longer"

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: Fix forgotten preempt_enable() when CPU has inclusive pcaches
  MIPS: Alchemy: MTX-1: fix incorrect placement of __initdata tag

11 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 4 Oct 2013 16:03:07 +0000 (09:03 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "Two simplefb fixes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/simplefb: Mark framebuffer mem-resources as IORESOURCE_BUSY to avoid bootup warning
  x86/simplefb: Fix overflow causing bogus fall-back

11 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 4 Oct 2013 16:02:35 +0000 (09:02 -0700)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Ingo Molnar:
 "Frederic's minimal fix for hardirq/softirq nesting crashes"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irq: Force hardirq exit's softirq processing on its own stack

11 years agosched/wait: Clean up wait.h details a bit
Ingo Molnar [Fri, 4 Oct 2013 08:24:49 +0000 (10:24 +0200)]
sched/wait: Clean up wait.h details a bit

Since we are changing wait.h profoundly, use the opportunity to:

 - add a sentence to explain what this file is about
 - remove whitespace noise
 - prettify weird looking line break fixup attempts
 - standardize type definition and initialization sequences
 - use consistent style details

No code is changed.

Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-O8dIie5swnctqpupakatvqyq@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agodmaengine: imx-dma: fix callback path in tasklet
Michael Grzeschik [Tue, 17 Sep 2013 13:56:08 +0000 (15:56 +0200)]
dmaengine: imx-dma: fix callback path in tasklet

We need to free the ld_active list head before jumping into the callback
routine. Otherwise the callback could run into issue_pending and change
our ld_active list head we just going to free. This will run the channel
list into an currupted and undefined state.

Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodmaengine: imx-dma: fix lockdep issue between irqhandler and tasklet
Michael Grzeschik [Tue, 17 Sep 2013 13:56:07 +0000 (15:56 +0200)]
dmaengine: imx-dma: fix lockdep issue between irqhandler and tasklet

The tasklet and irqhandler are using spin_lock while other routines are
using spin_lock_irqsave/restore. This leads to lockdep issues as
described bellow. This patch is changing the code to use
spinlock_irq_save/restore in both code pathes.

As imxdma_xfer_desc always gets called with spin_lock_irqsave lock held,
this patch also removes the spare call inside the routine to avoid
double locking.

[  403.358162] =================================
[  403.362549] [ INFO: inconsistent lock state ]
[  403.366945] 3.10.0-20130823+ #904 Not tainted
[  403.371331] ---------------------------------
[  403.375721] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
[  403.381769] swapper/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
[  403.386762]  (&(&imxdma->lock)->rlock){?.-...}, at: [<c019d77c>] imxdma_tasklet+0x20/0x134
[  403.395201] {IN-HARDIRQ-W} state was registered at:
[  403.400108]   [<c004b264>] mark_lock+0x2a0/0x6b4
[  403.404798]   [<c004d7c8>] __lock_acquire+0x650/0x1a64
[  403.410004]   [<c004f15c>] lock_acquire+0x94/0xa8
[  403.414773]   [<c02f74e4>] _raw_spin_lock+0x54/0x8c
[  403.419720]   [<c019d094>] dma_irq_handler+0x78/0x254
[  403.424845]   [<c0061124>] handle_irq_event_percpu+0x38/0x1b4
[  403.430670]   [<c00612e4>] handle_irq_event+0x44/0x64
[  403.435789]   [<c0063a70>] handle_level_irq+0xd8/0xf0
[  403.440903]   [<c0060a20>] generic_handle_irq+0x28/0x38
[  403.446194]   [<c0009cc4>] handle_IRQ+0x68/0x8c
[  403.450789]   [<c0008714>] avic_handle_irq+0x3c/0x48
[  403.455811]   [<c0008f84>] __irq_svc+0x44/0x74
[  403.460314]   [<c0040b04>] cpu_startup_entry+0x88/0xf4
[  403.465525]   [<c02f00d0>] rest_init+0xb8/0xe0
[  403.470045]   [<c03e07dc>] start_kernel+0x28c/0x2d4
[  403.474986]   [<a0008040>] 0xa0008040
[  403.478709] irq event stamp: 50854
[  403.482140] hardirqs last  enabled at (50854): [<c001c6b8>] tasklet_action+0x38/0xdc
[  403.489954] hardirqs last disabled at (50853): [<c001c6a0>] tasklet_action+0x20/0xdc
[  403.497761] softirqs last  enabled at (50850): [<c001bc64>] _local_bh_enable+0x14/0x18
[  403.505741] softirqs last disabled at (50851): [<c001c268>] irq_exit+0x88/0xdc
[  403.513026]
[  403.513026] other info that might help us debug this:
[  403.519593]  Possible unsafe locking scenario:
[  403.519593]
[  403.525548]        CPU0
[  403.528020]        ----
[  403.530491]   lock(&(&imxdma->lock)->rlock);
[  403.534828]   <Interrupt>
[  403.537474]     lock(&(&imxdma->lock)->rlock);
[  403.541983]
[  403.541983]  *** DEADLOCK ***
[  403.541983]
[  403.547951] no locks held by swapper/0.
[  403.551813]
[  403.551813] stack backtrace:
[  403.556222] CPU: 0 PID: 0 Comm: swapper Not tainted 3.10.0-20130823+ #904
[  403.563039] Backtrace:
[  403.565581] [<c000b98c>] (dump_backtrace+0x0/0x10c) from [<c000bb28>] (show_stack+0x18/0x1c)
[  403.574054]  r6:00000000 r5:c05c51d8 r4:c040bd58 r3:00200000
[  403.579872] [<c000bb10>] (show_stack+0x0/0x1c) from [<c02f398c>] (dump_stack+0x20/0x28)
[  403.587955] [<c02f396c>] (dump_stack+0x0/0x28) from [<c02f29c8>] (print_usage_bug.part.28+0x224/0x28c)
[  403.597340] [<c02f27a4>] (print_usage_bug.part.28+0x0/0x28c) from [<c004b404>] (mark_lock+0x440/0x6b4)
[  403.606682]  r8:c004a41c r7:00000000 r6:c040bd58 r5:c040c040 r4:00000002
[  403.613566] [<c004afc4>] (mark_lock+0x0/0x6b4) from [<c004d844>] (__lock_acquire+0x6cc/0x1a64)
[  403.622244] [<c004d178>] (__lock_acquire+0x0/0x1a64) from [<c004f15c>] (lock_acquire+0x94/0xa8)
[  403.631010] [<c004f0c8>] (lock_acquire+0x0/0xa8) from [<c02f74e4>] (_raw_spin_lock+0x54/0x8c)
[  403.639614] [<c02f7490>] (_raw_spin_lock+0x0/0x8c) from [<c019d77c>] (imxdma_tasklet+0x20/0x134)
[  403.648434]  r6:c3847010 r5:c040e890 r4:c38470d4
[  403.653194] [<c019d75c>] (imxdma_tasklet+0x0/0x134) from [<c001c70c>] (tasklet_action+0x8c/0xdc)
[  403.662013]  r8:c0599160 r7:00000000 r6:00000000 r5:c040e890 r4:c3847114 r3:c019d75c
[  403.670042] [<c001c680>] (tasklet_action+0x0/0xdc) from [<c001bd4c>] (__do_softirq+0xe4/0x1f0)
[  403.678687]  r7:00000101 r6:c0402000 r5:c059919c r4:00000001
[  403.684498] [<c001bc68>] (__do_softirq+0x0/0x1f0) from [<c001c268>] (irq_exit+0x88/0xdc)
[  403.692652] [<c001c1e0>] (irq_exit+0x0/0xdc) from [<c0009cc8>] (handle_IRQ+0x6c/0x8c)
[  403.700514]  r4:00000030 r3:00000110
[  403.704192] [<c0009c5c>] (handle_IRQ+0x0/0x8c) from [<c0008714>] (avic_handle_irq+0x3c/0x48)
[  403.712664]  r5:c0403f28 r4:c0593ebc
[  403.716343] [<c00086d8>] (avic_handle_irq+0x0/0x48) from [<c0008f84>] (__irq_svc+0x44/0x74)
[  403.724733] Exception stack(0xc0403f28 to 0xc0403f70)
[  403.729841] 3f20:                   00000001 00000004 00000000 20000013 c0402000 c04104a8
[  403.738078] 3f40: 00000002 c0b69620 a0004000 41069264 a03fb5f4 c0403f7c c0403f40 c0403f70
[  403.746301] 3f60: c004b92c c0009e74 20000013 ffffffff
[  403.751383]  r6:ffffffff r5:20000013 r4:c0009e74 r3:c004b92c
[  403.757210] [<c0009e30>] (arch_cpu_idle+0x0/0x4c) from [<c0040b04>] (cpu_startup_entry+0x88/0xf4)
[  403.766161] [<c0040a7c>] (cpu_startup_entry+0x0/0xf4) from [<c02f00d0>] (rest_init+0xb8/0xe0)
[  403.774753] [<c02f0018>] (rest_init+0x0/0xe0) from [<c03e07dc>] (start_kernel+0x28c/0x2d4)
[  403.783051]  r6:c03fc484 r5:ffffffff r4:c040a0e0
[  403.787797] [<c03e0550>] (start_kernel+0x0/0x2d4) from [<a0008040>] (0xa0008040)

Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodmaengine: imx-dma: fix slow path issue in prep_dma_cyclic
Michael Grzeschik [Tue, 17 Sep 2013 13:56:06 +0000 (15:56 +0200)]
dmaengine: imx-dma: fix slow path issue in prep_dma_cyclic

When perparing cyclic_dma buffers by the sound layer, it will dump the
following lockdep trace. The leading snd_pcm_action_single get called
with read_lock_irq called. To fix this, we change the kcalloc call from
GFP_KERNEL to GFP_ATOMIC.

WARNING: at kernel/lockdep.c:2740 lockdep_trace_alloc+0xcc/0x114()
DEBUG_LOCKS_WARN_ON(irqs_disabled_flags(flags))
Modules linked in:
CPU: 0 PID: 832 Comm: aplay Not tainted 3.11.0-20130823+ #903
Backtrace:
[<c000b98c>] (dump_backtrace+0x0/0x10c) from [<c000bb28>] (show_stack+0x18/0x1c)
 r6:c004c090 r5:00000009 r4:c2e0bd18 r3:00404000
[<c000bb10>] (show_stack+0x0/0x1c) from [<c02f397c>] (dump_stack+0x20/0x28)
[<c02f395c>] (dump_stack+0x0/0x28) from [<c001531c>] (warn_slowpath_common+0x54/0x70)
[<c00152c8>] (warn_slowpath_common+0x0/0x70) from [<c00153dc>] (warn_slowpath_fmt+0x38/0x40)
 r8:00004000 r7:a3b90000 r6:000080d0 r5:60000093 r4:c2e0a000 r3:00000009
[<c00153a4>] (warn_slowpath_fmt+0x0/0x40) from [<c004c090>] (lockdep_trace_alloc+0xcc/0x114)
 r3:c03955d8 r2:c03907db
[<c004bfc4>] (lockdep_trace_alloc+0x0/0x114) from [<c008f16c>] (__kmalloc+0x34/0x118)
 r6:000080d0 r5:c3800120 r4:000080d0 r3:c040a0f8
[<c008f138>] (__kmalloc+0x0/0x118) from [<c019c95c>] (imxdma_prep_dma_cyclic+0x64/0x168)
 r7:a3b90000 r6:00000004 r5:c39d8420 r4:c3847150
[<c019c8f8>] (imxdma_prep_dma_cyclic+0x0/0x168) from [<c024618c>] (snd_dmaengine_pcm_trigger+0xa8/0x160)
[<c02460e4>] (snd_dmaengine_pcm_trigger+0x0/0x160) from [<c0241fa8>] (soc_pcm_trigger+0x90/0xb4)
 r8:c058c7b0 r7:c3b8140c r6:c39da560 r5:00000001 r4:c3b81000
[<c0241f18>] (soc_pcm_trigger+0x0/0xb4) from [<c022ece4>] (snd_pcm_do_start+0x2c/0x38)
 r7:00000000 r6:00000003 r5:c058c7b0 r4:c3b81000
[<c022ecb8>] (snd_pcm_do_start+0x0/0x38) from [<c022e958>] (snd_pcm_action_single+0x40/0x6c)
[<c022e918>] (snd_pcm_action_single+0x0/0x6c) from [<c022ea64>] (snd_pcm_action_lock_irq+0x7c/0x9c)
 r7:00000003 r6:c3b810f0 r5:c3b810f0 r4:c3b81000
[<c022e9e8>] (snd_pcm_action_lock_irq+0x0/0x9c) from [<c023009c>] (snd_pcm_common_ioctl1+0x7f8/0xfd0)
 r8:c3b7f888 r7:005407b8 r6:c2c991c0 r5:c3b81000 r4:c3b81000 r3:00004142
[<c022f8a4>] (snd_pcm_common_ioctl1+0x0/0xfd0) from [<c023117c>] (snd_pcm_playback_ioctl1+0x464/0x488)
[<c0230d18>] (snd_pcm_playback_ioctl1+0x0/0x488) from [<c02311d4>] (snd_pcm_playback_ioctl+0x34/0x40)
 r8:c3b7f888 r7:00004142 r6:00000004 r5:c2c991c0 r4:005407b8
[<c02311a0>] (snd_pcm_playback_ioctl+0x0/0x40) from [<c00a14a4>] (vfs_ioctl+0x30/0x44)
[<c00a1474>] (vfs_ioctl+0x0/0x44) from [<c00a1fe8>] (do_vfs_ioctl+0x55c/0x5c0)
[<c00a1a8c>] (do_vfs_ioctl+0x0/0x5c0) from [<c00a208c>] (SyS_ioctl+0x40/0x68)
[<c00a204c>] (SyS_ioctl+0x0/0x68) from [<c0009380>] (ret_fast_syscall+0x0/0x44)
 r8:c0009544 r7:00000036 r6:bedeaa58 r5:00000000 r4:000000c0

Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoMAINTAINERS, sched: Update file pattern
Peter Zijlstra [Wed, 2 Oct 2013 09:49:47 +0000 (11:49 +0200)]
MAINTAINERS, sched: Update file pattern

Took a while to sort out these bits and we'd like to be Cc:-ed on
future modifications to the waitqueue APIs and all that.

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-0ix315c7qcz88slmnrpshvmf@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Make the __wait_event*() interface more friendly
Peter Zijlstra [Wed, 2 Oct 2013 09:22:33 +0000 (11:22 +0200)]
sched/wait: Make the __wait_event*() interface more friendly

Change all __wait_event*() implementations to match the corresponding
wait_event*() signature for convenience.

In particular this does away with the weird 'ret' logic. Since there
are __wait_event*() users this requires we update them too.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092529.042563462@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_hrtimeout()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:32 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_hrtimeout()

While not a whole-sale replacement like the others we can still reduce
the size of __wait_event_hrtimeout() considerably by noting that the
actual core of __wait_event_hrtimeout() is identical to what
___wait_event() generates.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.972793648@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_killable()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:31 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_killable()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.898691966@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_interruptible_tty()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:30 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_interruptible_tty()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.831085521@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_interruptible_lock_irq_timeout()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:29 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_interruptible_lock_irq_timeout()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.759956109@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_interruptible_lock_irq()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:28 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_interruptible_lock_irq()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.686006009@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_lock_irq()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:27 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_lock_irq()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.612813379@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_interruptible_exclusive()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:26 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_interruptible_exclusive()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.541716442@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_interruptible_timeout()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:25 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_interruptible_timeout()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.469616907@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_interruptible()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:24 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_interruptible()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.396949919@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event_timeout()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:23 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event_timeout()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.325264677@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Collapse __wait_event()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:22 +0000 (11:22 +0200)]
sched/wait: Collapse __wait_event()

Reduce macro complexity by using the new ___wait_event() helper.
No change in behaviour, identical generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.254863348@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Introduce ___wait_event()
Peter Zijlstra [Wed, 2 Oct 2013 09:22:21 +0000 (11:22 +0200)]
sched/wait: Introduce ___wait_event()

There's far too much duplication in the __wait_event macros; in order
to fix this introduce ___wait_event() a macro with the capability to
replace most other macros.

With the previous patches changing the various __wait_event*()
implementations to be more uniform; we can now collapse the lot
without also changing generated code.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.181897111@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Change the wait_exclusive control flow
Peter Zijlstra [Wed, 2 Oct 2013 09:22:20 +0000 (11:22 +0200)]
sched/wait: Change the wait_exclusive control flow

Purely a preparatory patch; it changes the control flow to match what
will soon be generated by generic code so that that patch can be a
unity transform.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.107994763@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Change timeout logic
Peter Zijlstra [Wed, 2 Oct 2013 09:22:19 +0000 (11:22 +0200)]
sched/wait: Change timeout logic

Commit 4c663cf ("wait: fix false timeouts when using
wait_event_timeout()") introduced an additional condition check after
a timeout but there's a few issues;

 - it forgot one site
 - it put the check after the main loop; not at the actual timeout
   check.

Cure both; by wrapping the condition (as suggested by Oleg), this
avoids double evaluation of 'condition' which could be quite big.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092528.028892896@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agosched/wait: Make the signal_pending() checks consistent
Peter Zijlstra [Wed, 2 Oct 2013 09:22:18 +0000 (11:22 +0200)]
sched/wait: Make the signal_pending() checks consistent

There's two patterns to check signals in the __wait_event*() macros:

  if (!signal_pending(current)) {
schedule();
continue;
  }
  ret = -ERESTARTSYS;
  break;

And the more natural:

  if (signal_pending(current)) {
ret = -ERESTARTSYS;
break;
  }
  schedule();

Change them all into the latter form.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131002092527.956416254@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
11 years agoxen/hvc: allow xenboot console to be used again
David Vrabel [Tue, 1 Oct 2013 18:00:49 +0000 (19:00 +0100)]
xen/hvc: allow xenboot console to be used again

Commit d0380e6c3c0f6edb986d8798a23acfaf33d5df23 (early_printk:
consolidate random copies of identical code) added in 3.10 introduced
a check for con->index == -1 in early_console_register().

Initialize index to -1 for the xenboot console so earlyprintk=xen
works again.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
11 years agousb: chipidea: add Intel Clovertrail pci id
David Cohen [Tue, 1 Oct 2013 21:32:58 +0000 (14:32 -0700)]
usb: chipidea: add Intel Clovertrail pci id

Also clean up the last item of the pci id list to be "cleaner".

Signed-off-by: David Cohen <david.a.cohen@linux.intel.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
11 years agostaging: comedi: ni_65xx: (bug fix) confine insn_bits to one subdevice
Ian Abbott [Wed, 2 Oct 2013 13:57:51 +0000 (14:57 +0100)]
staging: comedi: ni_65xx: (bug fix) confine insn_bits to one subdevice

The `insn_bits` handler `ni_65xx_dio_insn_bits()` has a `for` loop that
currently writes (optionally) and reads back up to 5 "ports" consisting
of 8 channels each.  It reads up to 32 1-bit channels but can only read
and write a whole port at once - it needs to handle up to 5 ports as the
first channel it reads might not be aligned on a port boundary.  It
breaks out of the loop early if the next port it handles is beyond the
final port on the card.  It also breaks out early on the 5th port in the
loop if the first channel was aligned.  Unfortunately, it doesn't check
that the current port it is dealing with belongs to the comedi subdevice
the `insn_bits` handler is acting on.  That's a bug.

Redo the `for` loop to terminate after the final port belonging to the
subdevice, changing the loop variable in the process to simplify things
a bit.  The `for` loop could now try and handle more than 5 ports if the
subdevice has more than 40 channels, but the test `if (bitshift >= 32)`
ensures it will break out early after 4 or 5 ports (depending on whether
the first channel is aligned on a port boundary).  (`bitshift` will be
between -7 and 7 inclusive on the first iteration, increasing by 8 for
each subsequent operation.)

Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Cc: <stable@vger.kernel.org> # 3.10.y 3.11.y 3.12.y
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
11 years agoiscsi-target; Allow an extra tag_num / 2 number of percpu_ida tags
Nicholas Bellinger [Thu, 3 Oct 2013 21:03:59 +0000 (14:03 -0700)]
iscsi-target; Allow an extra tag_num / 2 number of percpu_ida tags

This patch bumps the default number of tags allocated per session by
iscsi-target via transport_alloc_session_tags() -> percpu_ida_init()
by another (tag_num / 2).

This is done to take into account the tags waiting to be acknowledged
and released in iscsit_ack_from_expstatsn(), but who's number are not
directly limited by the CmdSN Window queue_depth being enforced by
the target.

Using a larger value here is also useful to prevent percpu_ida_alloc()
from having to steal tags from other CPUs when no tags are available
on the local CPU, while waiting for unacknowledged tags to be released.

Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agoiscsi-target: Perform release of acknowledged tags from RX context
Nicholas Bellinger [Thu, 3 Oct 2013 20:56:14 +0000 (13:56 -0700)]
iscsi-target: Perform release of acknowledged tags from RX context

This patch converts iscsit_ack_from_expstatsn() to populate a local
ack_list of commands, and call iscsit_free_cmd() directly from RX
thread context, instead of using iscsit_add_cmd_to_immediate_queue()
to queue the acknowledged commands to be released from TX thread
context.

It is helpful to release the acknowledge commands as quickly as
possible, along with the associated percpu_ida tags, in order to
prevent percpu_ida_alloc() from having to steal tags from other
CPUs while waiting for iscsit_free_cmd() to happen from TX thread
context.

Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agoiscsi-target: Only perform wait_for_tasks when performing shutdown
Nicholas Bellinger [Thu, 3 Oct 2013 20:37:21 +0000 (13:37 -0700)]
iscsi-target: Only perform wait_for_tasks when performing shutdown

This patch changes transport_generic_free_cmd() to only wait_for_tasks
when shutdown=true is passed to iscsit_free_cmd().

With the advent of >= v3.10 iscsi-target code using se_cmd->cmd_kref,
the extra wait_for_tasks with shutdown=false is unnecessary, and may
end up causing an extra context switch when releasing WRITEs.

Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agoMerge git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Thu, 3 Oct 2013 15:56:41 +0000 (08:56 -0700)]
Merge git://git.kernel.org/pub/scm/virt/kvm/kvm

Pull kvm NULL deref fix from Gleb Natapov.

* git://git.kernel.org/pub/scm/virt/kvm/kvm:
  Fix NULL dereference in gfn_to_hva_prot()

11 years agoMerge branch 'for-curr' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc
Linus Torvalds [Thu, 3 Oct 2013 15:55:50 +0000 (08:55 -0700)]
Merge branch 'for-curr' of git://git./linux/kernel/git/vgupta/arc

Pull ARC fix from Vineet Gupta:
 "Chrisitian found/fixed issue with SA_SIGINFO based signal handler
  corrupting the user space registers post after signal handling"

* 'for-curr' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: Fix signal frame management for SA_SIGINFO

11 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Thu, 3 Oct 2013 15:54:39 +0000 (08:54 -0700)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

Pull powerpc fixes from Ben Herrenschmidt:
 "Here are a few powerpc fixes, all aimed at -stable, found in part
  thanks to the ramping up of a major distro testing and in part thanks
  to the LE guys hitting all sort interesting corner cases.

  The most scary are probably the register clobber issues in
  csum_partial_copy_generic(), especially since Anton even had a test
  case for that thing, which didn't manage to hit the bugs :-)

  Another highlight is that memory hotplug should work again with these
  fixes.

  Oh and the vio modalias one is worse than the cset implies as it
  upsets distro installers, so I've been told at least, which is why I'm
  shooting it to stable"

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc/tm: Switch out userspace PPR and DSCR sooner
  powerpc/tm: Turn interrupts hard off in tm_reclaim()
  powerpc/perf: Fix handling of FAB events
  powerpc/vio: Fix modalias_show return values
  powerpc/iommu: Use GFP_KERNEL instead of GFP_ATOMIC in iommu_init_table()
  powerpc/sysfs: Disable writing to PURR in guest mode
  powerpc: Restore registers on error exit from csum_partial_copy_generic()
  powerpc: Fix parameter clobber in csum_partial_copy_generic()
  powerpc: Fix memory hotplug with sparse vmemmap

11 years agotarget: Fail on non zero scsi_status in compare_and_write_callback
Nicholas Bellinger [Wed, 2 Oct 2013 00:04:40 +0000 (17:04 -0700)]
target: Fail on non zero scsi_status in compare_and_write_callback

This patch addresses a bug for backends such as IBLOCK that perform
asynchronous completion via transport_complete_cmd(), that will call
target_complete_failure_work() -> transport_generic_request_failure(),
upon exception status and invoke cmd->transport_complete_callback()
-> compare_and_write_callback() incorrectly during the failure case.

It adds a check for a non zero se_cmd->scsi_status within the first
invocation of compare_and_write_callback(), and will jump to out plus
up se_device->caw_sem before exiting the callback.

Reported-by: Thomas Glanzmann <thomas@glanzmann.de>
Tested-by: Thomas Glanzmann <thomas@glanzmann.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agotarget: Fix recursive COMPARE_AND_WRITE callback failure
Nicholas Bellinger [Tue, 1 Oct 2013 23:53:10 +0000 (16:53 -0700)]
target: Fix recursive COMPARE_AND_WRITE callback failure

This patch addresses a bug when compare_and_write_callback() invoked from
target_complete_ok_work() hits an failure from __target_execute_cmd() ->
cmd->execute_cmd(), that ends up calling transport_generic_request_failure()
-> compare_and_write_post(), thus causing SCF_COMPARE_AND_WRITE_POST to
incorrectly be set.

The result of this bug is that target_complete_ok_work() no longer hits
the if (!rc && !(cmd->se_cmd_flags & SCF_COMPARE_AND_WRITE_POST) check
that forces an immediate return, and instead double completes the se_cmd
in question, triggering an OOPs in the process.

This patch changes compare_and_write_post() to only set this bit when a
failure has not already occured to ensure the immediate return from within
target_complete_ok_work(), and thus allow transport_generic_request_failure()
to handle the sending of the CHECK_CONDITION exception status.

Reported-by: Thomas Glanzmann <thomas@glanzmann.de>
Tested-by: Thomas Glanzmann <thomas@glanzmann.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agotarget: Reset data_length for COMPARE_AND_WRITE to NoLB * block_size
Nicholas Bellinger [Tue, 1 Oct 2013 23:46:37 +0000 (16:46 -0700)]
target: Reset data_length for COMPARE_AND_WRITE to NoLB * block_size

This patch resets se_cmd->data_length for COMPARE_AND_WRITE emulation
within sbc_compare_and_write() to NoLB * block_size in order to address
a bug with FILEIO backends where a I/O failure will occur when data_length
does not match the I/O size being actually dispatched for the individual
per block READs + WRITEs.

This is done late enough in sbc_compare_and_write() after the memory
allocations have occured in transport_generic_new_cmd() to not cause
any unwanted side-effects.

Reported-by: Thomas Glanzmann <thomas@glanzmann.de>
Tested-by: Thomas Glanzmann <thomas@glanzmann.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agoib_srpt: always set response for task management
Jack Wang [Mon, 30 Sep 2013 08:09:05 +0000 (10:09 +0200)]
ib_srpt: always set response for task management

The SRP specification requires:

  "Response data shall be provided in any SRP_RSP response that is sent in
   response to an SRP_TSK_MGMT request (see 6.7). The information in the
   RSP_CODE field (see table 24) shall indicate the completion status of
   the task management function."

So fix this to avoid the SRP initiator interprets task management functions
that succeeded as failed.

Signed-off-by: Jack Wang <jinpu.wang@profitbricks.com>
Cc: stable@vger.kernel.org # 3.3+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
11 years agopowerpc/tm: Switch out userspace PPR and DSCR sooner
Michael Neuling [Thu, 26 Sep 2013 03:29:09 +0000 (13:29 +1000)]
powerpc/tm: Switch out userspace PPR and DSCR sooner

When we do a treclaim or trecheckpoint we end up running with userspace
PPR and DSCR values.  Currently we don't do anything special to avoid
running with user values which could cause a severe performance
degradation.

This patch moves the PPR and DSCR save and restore around treclaim and
trecheckpoint so that we run with user values for a much shorter period.
More care is taken with the PPR as it's impact is greater than the DSCR.

This is similar to user exceptions, where we run HTM_MEDIUM early to
ensure that we don't run with a userspace PPR values in the kernel.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Cc: <stable@vger.kernel.org> # 3.9+
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agopowerpc/tm: Turn interrupts hard off in tm_reclaim()
Michael Neuling [Wed, 2 Oct 2013 07:15:15 +0000 (17:15 +1000)]
powerpc/tm: Turn interrupts hard off in tm_reclaim()

We can't take IRQs in tm_reclaim as we might have a bogus r13 and r1.

This turns IRQs hard off in this function.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Cc: <stable@vger.kernel.org> # 3.9+
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agopowerpc/perf: Fix handling of FAB events
Michael Ellerman [Wed, 2 Oct 2013 08:04:06 +0000 (18:04 +1000)]
powerpc/perf: Fix handling of FAB events

Commit 4df4899 "Add power8 EBB support" included a bug in the handling
of the FAB_CRESP_MATCH and FAB_TYPE_MATCH fields.

These values are pulled out of the event code using EVENT_THR_CTL_SHIFT,
however we were then or'ing that value directly into MMCR1.

This meant we were failing to set the FAB fields correctly, and also
potentially corrupting the value for PMC4SEL. Leading to no counts for
the FAB events and incorrect counts for PMC4.

The fix is simply to shift left the FAB value correctly before or'ing it
with MMCR1.

Reported-by: Sooraj Ravindran Nair <soonair3@in.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Cc: <stable@vger.kernel.org> # 3.10+
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
11 years agopowerpc/vio: Fix modalias_show return values
Prarit Bhargava [Mon, 23 Sep 2013 13:33:36 +0000 (09:33 -0400)]
powerpc/vio: Fix modalias_show return values

modalias_show() should return an empty string on error, not -ENODEV.

This causes the following false and annoying error:

> find /sys/devices -name modalias -print0 | xargs -0 cat >/dev/null
cat: /sys/devices/vio/4000/modalias: No such device
cat: /sys/devices/vio/4001/modalias: No such device
cat: /sys/devices/vio/4002/modalias: No such device
cat: /sys/devices/vio/4004/modalias: No such device
cat: /sys/devices/vio/modalias: No such device

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: <stable@vger.kernel.org>