Pull scheduler updates from Ingo Molnar:
"The main scheduler changes in this cycle were:
- various sched/deadline fixes and enhancements
- rescheduling latency fixes/cleanups
- rework the rq->clock code to be more consistent and more robust.
- minor micro-optimizations
- ->avg.decay_count fixes
- add a stack overflow check to might_sleep()
- idle-poll handler fix, possibly resulting in power savings
- misc smaller updates and fixes"
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/Documentation: Remove unneeded word
sched/wait: Introduce wait_on_bit_timeout()
sched: Pull resched loop to __schedule() callers
sched/deadline: Remove cpu_active_mask from cpudl_find()
sched: Fix hrtick_start() on UP
sched/deadline: Avoid pointless __setscheduler()
sched/deadline: Fix stale yield state
sched/deadline: Fix hrtick for a non-leftmost task
sched/deadline: Modify cpudl::free_cpus to reflect rd->online
sched/idle: Add missing checks to the exit condition of cpu_idle_poll()
sched: Fix missing preemption opportunity
sched/rt: Reduce rq lock contention by eliminating locking of non-feasible target
sched/debug: Print rq->clock_task
sched/core: Rework rq->clock update skips
sched/core: Validate rq_clock*() serialization
sched/core: Remove check of p->sched_class
sched/fair: Fix sched_entity::avg::decay_count initialization
sched/debug: Fix potential call to __ffs(0) in sched_show_task()
sched/debug: Check for stack overflow in ___might_sleep()
sched/fair: Fix the dealing with decay_count in __synchronize_entity_decay()
We explicitly mark the task running after returning from
a __rt_mutex_slowlock() call, which does the actual sleeping
via wait-wake-trylocking. As such, this patch does two things:
(1) refactors the code so that setting current to TASK_RUNNING
is done by __rt_mutex_slowlock(), and not by the callers. The
downside to this is that it becomes a bit unclear when at what
point we block. As such I've added a comment that the task
blocks when calling __rt_mutex_slowlock() so readers can figure
out when it is running again.
(2) relaxes setting current's state through __set_current_state(),
instead of it's more expensive barrier alternative. There was no
need for the implied barrier as we're obviously not planning on
blocking.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1422857784.18096.1.camel@stgolabs.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Both mutexes and rwsems took a performance hit when we switched
over from the original mcs code to the cancelable variant (osq).
The reason being the use of smp_load_acquire() when polling for
node->locked. This is not needed as reordering is not an issue,
as such, relax the barrier semantics. Paul describes the scenario
nicely: https://lkml.org/lkml/2013/11/19/405
- If we start polling before the insertion is complete, all that
happens is that the first few polls have no chance of seeing a lock
grant.
- Ordering the polling against the initialization -- the above
xchg() is already doing that for us.
The smp_load_acquire() when unqueuing make sense. In addition,
we don't need to worry about leaking the critical region as
osq is only used internally.
This impacts both regular and large levels of concurrency,
ie on a 40 core system with a disk intensive workload:
disk-1 804.83 ( 0.00%) 828.16 ( 2.90%)
disk-61 8063.45 ( 0.00%) 18181.82 (125.48%)
disk-121 7187.41 ( 0.00%) 20119.17 (179.92%)
disk-181 6933.32 ( 0.00%) 20509.91 (195.82%)
disk-241 6850.81 ( 0.00%) 20397.80 (197.74%)
disk-301 6815.22 ( 0.00%) 20287.58 (197.68%)
disk-361 7080.40 ( 0.00%) 20205.22 (185.37%)
disk-421 7076.13 ( 0.00%) 19957.33 (182.04%)
disk-481 7083.25 ( 0.00%) 19784.06 (179.31%)
disk-541 7038.39 ( 0.00%) 19610.92 (178.63%)
disk-601 7072.04 ( 0.00%) 19464.53 (175.23%)
disk-661 7010.97 ( 0.00%) 19348.23 (175.97%)
disk-721 7069.44 ( 0.00%) 19255.33 (172.37%)
disk-781 7007.58 ( 0.00%) 19103.14 (172.61%)
disk-841 6981.18 ( 0.00%) 18964.22 (171.65%)
disk-901 6968.47 ( 0.00%) 18826.72 (170.17%)
disk-961 6964.61 ( 0.00%) 18708.02 (168.62%)
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1420573509-24774-7-git-send-email-dave@stgolabs.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We have two flavors of the MCS spinlock: standard and cancelable (OSQ).
While each one is independent of the other, we currently mix and match
them. This patch:
- Moves the OSQ code out of mcs_spinlock.h (which only deals with the traditional
version) into include/linux/osq_lock.h. No unnecessary code is added to the
more global header file, anything locks that make use of OSQ must include
it anyway.
- Renames mcs_spinlock.c to osq_lock.c. This file only contains osq code.
- Introduces a CONFIG_LOCK_SPIN_ON_OWNER in order to only build osq_lock
if there is support for it.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Waiman Long <Waiman.Long@hp.com>
Link: http://lkml.kernel.org/r/1420573509-24774-5-git-send-email-dave@stgolabs.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
We're going to make might_sleep() test for TASK_RUNNING, because
blocking without TASK_RUNNING will destroy the task state by setting
it to TASK_RUNNING.
There are a few occasions where its 'valid' to call blocking
primitives (and mutex_lock in particular) and not have TASK_RUNNING,
typically such cases are right before we set TASK_RUNNING anyhow.
Robustify the code by not assuming this; this has the beneficial side
effect of allowing optional code emission for fixing the above
might_sleep() false positives.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: tglx@linutronix.de
Cc: ilya.dryomov@inktank.com
Cc: umgwanakikbuti@gmail.com
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140924082241.988560063@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull core locking updates from Ingo Molnar:
"The main updates in this cycle were:
- mutex MCS refactoring finishing touches: improve comments, refactor
and clean up code, reduce debug data structure footprint, etc.
- qrwlock finishing touches: remove old code, self-test updates.
- small rwsem optimization
- various smaller fixes/cleanups"
* 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
locking/lockdep: Revert qrwlock recusive stuff
locking/rwsem: Avoid double checking before try acquiring write lock
locking/rwsem: Move EXPORT_SYMBOL() lines to follow function definition
locking/rwlock, x86: Delete unused asm/rwlock.h and rwlock.S
locking/rwlock, x86: Clean up asm/spinlock*.h to remove old rwlock code
locking/semaphore: Resolve some shadow warnings
locking/selftest: Support queued rwlock
locking/lockdep: Restrict the use of recursive read_lock() with qrwlock
locking/spinlocks: Always evaluate the second argument of spin_lock_nested()
locking/Documentation: Update locking/mutex-design.txt disadvantages
locking/Documentation: Move locking related docs into Documentation/locking/
locking/mutexes: Use MUTEX_SPIN_ON_OWNER when appropriate
locking/mutexes: Refactor optimistic spinning code
locking/mcs: Remove obsolete comment
locking/mutexes: Document quick lock release when unlocking
locking/mutexes: Standardize arguments in lock/unlock slowpaths
locking: Remove deprecated smp_mb__() barriers
Commit f0bab73cb5 ("locking/lockdep: Restrict the use of recursive
read_lock() with qrwlock") changed lockdep to try and conform to the
qrwlock semantics which differ from the traditional rwlock semantics.
In particular qrwlock is fair outside of interrupt context, but in
interrupt context readers will ignore all fairness.
The problem modeling this is that read and write side have different
lock state (interrupts) semantics but we only have a single
representation of these. Therefore lockdep will get confused, thinking
the lock can cause interrupt lock inversions.
So revert it for now; the old rwlock semantics were already imperfectly
modeled and the qrwlock extra won't fit either.
If we want to properly fix this, I think we need to resurrect the work
by Gautham did a few years ago that split the read and write state of
locks:
http://lwn.net/Articles/332801/
FWIW the locking selftest that would've failed (and was reported by
Borislav earlier) is something like:
RL(X1); /* IRQ-ON */
LOCK(A);
UNLOCK(A);
RU(X1);
IRQ_ENTER();
RL(X1); /* IN-IRQ */
RU(X1);
IRQ_EXIT();
At which point it would report that because A is an IRQ-unsafe lock we
can suffer the following inversion:
CPU0 CPU1
lock(A)
lock(X1)
lock(A)
<IRQ>
lock(X1)
And this is 'wrong' because X1 can recurse (assuming the above lock are
in fact read-lock) but lockdep doesn't know about this.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Waiman Long <Waiman.Long@hp.com>
Cc: ego@linux.vnet.ibm.com
Cc: bp@alien8.de
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20140930132600.GA7444@worktop.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Commit 9b0fc9c09f ("rwsem: skip initial trylock in rwsem_down_write_failed")
checks for if there are known active lockers in order to avoid write trylocking
using expensive cmpxchg() when it likely wouldn't get the lock.
However, a subsequent patch was added such that we directly
check for sem->count == RWSEM_WAITING_BIAS right before trying
that cmpxchg().
Thus, commit 9b0fc9c09f now just adds overhead.
This patch modifies it so that we only do a check for if
count == RWSEM_WAITING_BIAS.
Also, add a comment on why we do an "extra check" of count
before the cmpxchg().
Signed-off-by: Jason Low <jason.low2@hp.com>
Acked-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Aswin Chandramouleeswaran <aswin@hp.com>
Cc: Chegu Vinod <chegu_vinod@hp.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1410913017.2447.22.camel@j-VirtualBox
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Remove some unnecessary ones and explicitly include rwsem.h
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Its quite easy to get mixed up with the names -- 'torture_spinlock_irq'
is not actually a valid spinlock name.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Add a "rw_lock" torture test to stress kernel rwlocks and their irq
variant. Reader critical regions are 5x longer than writers. As such
a similar ratio of lock acquisitions is seen in the statistics. In the
case of massive contention, both hold the lock for 1/10 of a second.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The amount of global variables is getting pretty ugly. Group variables
related to the execution (ie: not parameters) in a new context structure.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
We can easily do so with our new reader lock support. Just an arbitrary
design default: readers have higher (5x) critical region latencies than
writers: 50 ms and 10 ms, respectively.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Most of it is based on what we already have for writers. This allows
readers to be very independent (and thus configurable), enabling
future module parameters to control things such as rw distribution.
Furthermore, readers have their own delaying function, allowing us
to test different rw critical region latencies, and stress locking
internals. Similarly, statistics, for now will only serve for the
number of lock acquisitions -- as opposed to writers, readers have
no failure detection.
In addition, introduce a new nreaders_stress module parameter. The
default number of readers will be the same number of writers threads.
Writer threads are interleaved with readers. Documentation is updated,
respectively.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
When performing module cleanups by calling torture_cleanup() the
'torture_type' string in nullified However, callers are not necessarily
done, and might still need to reference the variable. This impacts
both rcutorture and locktorture, causing printing things like:
[ 94.226618] (null)-torture: Stopping lock_torture_writer task
[ 94.226624] (null)-torture: Stopping lock_torture_stats task
Thus delay this operation until the very end of the cleanup process.
The consequence (which shouldn't matter for this kid of program) is,
of course, that we delay the window between rmmod and modprobing,
for instance in module_torture_begin().
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The statistics structure can serve well for both reader and writer
locks, thus simply rename some fields that mention 'write' and leave
the declaration of lwsa.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Regular locks are very different than locks with debugging. For instance
for mutexes, debugging forces to only take the slowpaths. As such, the
locktorture module should take this into account when printing related
information -- specifically when printing user passed parameters, it seems
the right place for such info.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
