- 03 Sep, 2019 4 commits
-
-
Patrick Bellasi authored
The cgroup CPU bandwidth controller allows to assign a specified (maximum) bandwidth to the tasks of a group. However this bandwidth is defined and enforced only on a temporal base, without considering the actual frequency a CPU is running on. Thus, the amount of computation completed by a task within an allocated bandwidth can be very different depending on the actual frequency the CPU is running that task. The amount of computation can be affected also by the specific CPU a task is running on, especially when running on asymmetric capacity systems like Arm's big.LITTLE. With the availability of schedutil, the scheduler is now able to drive frequency selections based on actual task utilization. Moreover, the utilization clamping support provides a mechanism to bias the frequency selection operated by schedutil depending on constraints assigned to the tasks currently RUNNABLE on a CPU. Giving the mechanisms described above, it is now possible to extend the cpu controller to specify the minimum (or maximum) utilization which should be considered for tasks RUNNABLE on a cpu. This makes it possible to better defined the actual computational power assigned to task groups, thus improving the cgroup CPU bandwidth controller which is currently based just on time constraints. Extend the CPU controller with a couple of new attributes uclamp.{min,max} which allow to enforce utilization boosting and capping for all the tasks in a group. Specifically: - uclamp.min: defines the minimum utilization which should be considered i.e. the RUNNABLE tasks of this group will run at least at a minimum frequency which corresponds to the uclamp.min utilization - uclamp.max: defines the maximum utilization which should be considered i.e. the RUNNABLE tasks of this group will run up to a maximum frequency which corresponds to the uclamp.max utilization These attributes: a) are available only for non-root nodes, both on default and legacy hierarchies, while system wide clamps are defined by a generic interface which does not depends on cgroups. This system wide interface enforces constraints on tasks in the root node. b) enforce effective constraints at each level of the hierarchy which are a restriction of the group requests considering its parent's effective constraints. Root group effective constraints are defined by the system wide interface. This mechanism allows each (non-root) level of the hierarchy to: - request whatever clamp values it would like to get - effectively get only up to the maximum amount allowed by its parent c) have higher priority than task-specific clamps, defined via sched_setattr(), thus allowing to control and restrict task requests. Add two new attributes to the cpu controller to collect "requested" clamp values. Allow that at each non-root level of the hierarchy. Keep it simple by not caring now about "effective" values computation and propagation along the hierarchy. Update sysctl_sched_uclamp_handler() to use the newly introduced uclamp_mutex so that we serialize system default updates with cgroup relate updates. Signed-off-by:Patrick Bellasi <patrick.bellasi@arm.com> Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by:
Michal Koutny <mkoutny@suse.com> Acked-by:
Tejun Heo <tj@kernel.org> Cc: Alessio Balsini <balsini@android.com> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com> Cc: Joel Fernandes <joelaf@google.com> Cc: Juri Lelli <juri.lelli@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Morten Rasmussen <morten.rasmussen@arm.com> Cc: Paul Turner <pjt@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Quentin Perret <quentin.perret@arm.com> Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com> Cc: Steve Muckle <smuckle@google.com> Cc: Suren Baghdasaryan <surenb@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Todd Kjos <tkjos@google.com> Cc: Vincent Guittot <vincent.guittot@linaro.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Link: https://lkml.kernel.org/r/20190822132811.31294-2-patrick.bellasi@arm.comSigned-off-by:
Ingo Molnar <mingo@kernel.org>
-
Matt Fleming authored
SD_BALANCE_{FORK,EXEC} and SD_WAKE_AFFINE are stripped in sd_init() for any sched domains with a NUMA distance greater than 2 hops (RECLAIM_DISTANCE). The idea being that it's expensive to balance across domains that far apart. However, as is rather unfortunately explained in: commit 32e45ff4 ("mm: increase RECLAIM_DISTANCE to 30") the value for RECLAIM_DISTANCE is based on node distance tables from 2011-era hardware. Current AMD EPYC machines have the following NUMA node distances: node distances: node 0 1 2 3 4 5 6 7 0: 10 16 16 16 32 32 32 32 1: 16 10 16 16 32 32 32 32 2: 16 16 10 16 32 32 32 32 3: 16 16 16 10 32 32 32 32 4: 32 32 32 32 10 16 16 16 5: 32 32 32 32 16 10 16 16 6: 32 32 32 32 16 16 10 16 7: 32 32 32 32 16 16 16 10 where 2 hops is 32. The result is that the scheduler fails to load balance properly across NUMA nodes on different sockets -- 2 hops apart. For example, pinning 16 busy threads to NUMA nodes 0 (CPUs 0-7) and 4 (CPUs 32-39) like so, $ numactl -C 0-7,32-39 ./spinner 16 causes all threads to fork and remain on node 0 until the active balancer kicks in after a few seconds and forcibly moves some threads to node 4. Override node_reclaim_distance for AMD Zen. Signed-off-by:Matt Fleming <matt@codeblueprint.co.uk> Signed-off-by:
Mel Gorman <mgorman@techsingularity.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@surriel.com> Cc: Suravee.Suthikulpanit@amd.com Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Thomas.Lendacky@amd.com Cc: Tony Luck <tony.luck@intel.com> Link: https://lkml.kernel.org/r/20190808195301.13222-3-matt@codeblueprint.co.ukSigned-off-by:
-
Matt Fleming authored
While it does make sense to allow CONFIG_NUMA and !CONFIG_SMP in theory, it doesn't make much sense in practice. Follow other architectures and make CONFIG_NUMA select CONFIG_SMP. The motivation for this patch is to allow a new NUMA variable to be initialised in kernel/sched/topology.c. Signed-off-by:
-
Peter Zijlstra authored
The below entries are a little unorthodox; I've not found other entries in MAINTAINER that subdivide responsibilities like this, and certainly the lovely get_maintainers.pl script will not get it, but I'm thinking to a human it should be plenty clear and we're all very good at ignoring email anyway. Signed-off-by:
Juri Lelli <juri.lelli@redhat.com> Acked-by:
Mark Rutland <mark.rutland@arm.com> Acked-by:
Steven Rostedt (VMware) <rostedt@goodmis.org> Acked-by:
Vincent Guittot <vincent.guittot@linaro.org> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Ben Segall <bsegall@google.com> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
-
- 12 Aug, 2019 1 commit
-
-
Phil Auld authored
Enabling WARN_DOUBLE_CLOCK in /sys/kernel/debug/sched_features causes warning to fire in update_rq_clock. This seems to be caused by onlining a new fair sched group not using the rq lock wrappers. [] rq->clock_update_flags & RQCF_UPDATED [] WARNING: CPU: 5 PID: 54385 at kernel/sched/core.c:210 update_rq_clock+0xec/0x150 [] Call Trace: [] online_fair_sched_group+0x53/0x100 [] cpu_cgroup_css_online+0x16/0x20 [] online_css+0x1c/0x60 [] cgroup_apply_control_enable+0x231/0x3b0 [] cgroup_mkdir+0x41b/0x530 [] kernfs_iop_mkdir+0x61/0xa0 [] vfs_mkdir+0x108/0x1a0 [] do_mkdirat+0x77/0xe0 [] do_syscall_64+0x55/0x1d0 [] entry_SYSCALL_64_after_hwframe+0x44/0xa9 Using the wrappers in online_fair_sched_group instead of the raw locking removes this warning. [ tglx: Use rq_*lock_irq() ] Signed-off-by:
Phil Auld <pauld@redhat.com> Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Vincent Guittot <vincent.guittot@linaro.org> Cc: Ingo Molnar <mingo@kernel.org> Link: https://lkml.kernel.org/r/20190801133749.11033-1-pauld@redhat.com
-
- 08 Aug, 2019 12 commits
-
-
Qais Yousef authored
scale_irq_capacity() call in schedutil_cpu_util() does util *= (max - irq) util /= max But the comment says util *= (1 - irq) util /= max Fix the comment to match what the scaling function does. Signed-off-by:
Qais Yousef <qais.yousef@arm.com> Signed-off-by:
Viresh Kumar <viresh.kumar@linaro.org> Acked-by:
-
Peter Zijlstra authored
Avoid the RETRY_TASK case in the pick_next_task() slow path. By doing the put_prev_task() early, we get the rt/deadline pull done, and by testing rq->nr_running we know if we need newidle_balance(). This then gives a stable state to pick a task from. Since the fast-path is fair only; it means the other classes will always have pick_next_task(.prev=NULL, .rf=NULL) and we can simplify. Signed-off-by:
-
Peter Zijlstra authored
Currently the pick_next_task() loop is convoluted and ugly because of how it can drop the rq->lock and needs to restart the picking. For the RT/Deadline classes, it is put_prev_task() where we do balancing, and we could do this before the picking loop. Make this possible. Signed-off-by:
-
Peter Zijlstra authored
For pick_next_task_fair() it is the newidle balance that requires dropping the rq->lock; provided we do put_prev_task() early, we can also detect the condition for doing newidle early. Signed-off-by:
-
Peter Zijlstra authored
In preparation of further separating pick_next_task() and set_curr_task() we have to pass the actual task into it, while there, rename the thing to better pair with put_prev_task(). Signed-off-by:
-
Peter Zijlstra authored
The CPU hotplug task selection is the only place where we used put_prev_task() on a task that is not current. While looking at that, it occured to me that we can simplify all that by by using a custom pick loop. Since we don't need to put current, we can do away with the fake task too. Signed-off-by: -
Peter Zijlstra authored
Because pick_next_task() implies set_curr_task() and some of the details haven't mattered too much, some of what _should_ be in set_curr_task() ended up in pick_next_task, correct this. This prepares the way for a pick_next_task() variant that does not affect the current state; allowing remote picking. Signed-off-by:
-
Peter Zijlstra authored
Signed-off-by:
-
Peter Zijlstra authored
Make sure the entire for loop has stop_cpus_in_progress set. Signed-off-by:
-
Dave Chiluk authored
It has been observed, that highly-threaded, non-cpu-bound applications running under cpu.cfs_quota_us constraints can hit a high percentage of periods throttled while simultaneously not consuming the allocated amount of quota. This use case is typical of user-interactive non-cpu bound applications, such as those running in kubernetes or mesos when run on multiple cpu cores. This has been root caused to cpu-local run queue being allocated per cpu bandwidth slices, and then not fully using that slice within the period. At which point the slice and quota expires. This expiration of unused slice results in applications not being able to utilize the quota for which they are allocated. The non-expiration of per-cpu slices was recently fixed by 'commit 512ac999 ("sched/fair: Fix bandwidth timer clock drift condition")'. Prior to that it appears that this had been broken since at least 'commit 51f2176d ("sched/fair: Fix unlocked reads of some cfs_b->quota/period")' which was introduced in v3.16-rc1 in 2014. That added the following conditional which resulted in slices never being expired. if (cfs_rq->runtime_expires != cfs_b->runtime_expires) { /* extend local deadline, drift is bounded above by 2 ticks */ cfs_rq->runtime_expires += TICK_NSEC; Because this was broken for nearly 5 years, and has recently been fixed and is now being noticed by many users running kubernetes (https://github.com/kubernetes/kubernetes/issues/67577) it is my opinion that the mechanisms around expiring runtime should be removed altogether. This allows quota already allocated to per-cpu run-queues to live longer than the period boundary. This allows threads on runqueues that do not use much CPU to continue to use their remaining slice over a longer period of time than cpu.cfs_period_us. However, this helps prevent the above condition of hitting throttling while also not fully utilizing your cpu quota. This theoretically allows a machine to use slightly more than its allotted quota in some periods. This overflow would be bounded by the remaining quota left on each per-cpu runqueueu. This is typically no more than min_cfs_rq_runtime=1ms per cpu. For CPU bound tasks this will change nothing, as they should theoretically fully utilize all of their quota in each period. For user-interactive tasks as described above this provides a much better user/application experience as their cpu utilization will more closely match the amount they requested when they hit throttling. This means that cpu limits no longer strictly apply per period for non-cpu bound applications, but that they are still accurate over longer timeframes. This greatly improves performance of high-thread-count, non-cpu bound applications with low cfs_quota_us allocation on high-core-count machines. In the case of an artificial testcase (10ms/100ms of quota on 80 CPU machine), this commit resulted in almost 30x performance improvement, while still maintaining correct cpu quota restrictions. That testcase is available at https://github.com/indeedeng/fibtest. Fixes: 512ac999 ("sched/fair: Fix bandwidth timer clock drift condition") Signed-off-by:
Dave Chiluk <chiluk+linux@indeed.com> Signed-off-by:
Ben Segall <bsegall@google.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: John Hammond <jhammond@indeed.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Kyle Anderson <kwa@yelp.com> Cc: Gabriel Munos <gmunoz@netflix.com> Cc: Peter Oskolkov <posk@posk.io> Cc: Cong Wang <xiyou.wangcong@gmail.com> Cc: Brendan Gregg <bgregg@netflix.com> Link: https://lkml.kernel.org/r/1563900266-19734-2-git-send-email-chiluk+linux@indeed.com
-
Peter Zijlstra authored
The current active_mm reference counting is confusing and sub-optimal. Rewrite the code to explicitly consider the 4 separate cases: user -> user When switching between two user tasks, all we need to consider is switch_mm(). user -> kernel When switching from a user task to a kernel task (which doesn't have an associated mm) we retain the last mm in our active_mm. Increment a reference count on active_mm. kernel -> kernel When switching between kernel threads, all we need to do is pass along the active_mm reference. kernel -> user When switching between a kernel and user task, we must switch from the last active_mm to the next mm, hoping of course that these are the same. Decrement a reference on the active_mm. The code keeps a different order, because as you'll note, both 'to user' cases require switch_mm(). And where the old code would increment/decrement for the 'kernel -> kernel' case, the new code observes this is a neutral operation and avoids touching the reference count. Signed-off-by:Rik van Riel <riel@surriel.com> Reviewed-by:
Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: luto@kernel.org
-
Peter Zijlstra authored
A rather embarrasing mistake had us call sched_setscheduler() before initializing the parameters passed to it. Fixes: 1a763fd7 ("rcu/tree: Call setschedule() gp ktread to SCHED_FIFO outside of atomic region") Signed-off-by:
Paul E. McKenney <paulmck@linux.ibm.com> Cc: Juri Lelli <juri.lelli@redhat.com>
-
- 25 Jul, 2019 23 commits
-
-
Qian Cai authored
Compiling a kernel with both FAIR_GROUP_SCHED=n and RT_GROUP_SCHED=n will generate a compiler warning: kernel/sched/core.c: In function 'sched_init': kernel/sched/core.c:5906:32: warning: variable 'ptr' set but not used It is unnecessary to have both "alloc_size" and "ptr", so just combine them. Signed-off-by:
Qian Cai <cai@lca.pw> Signed-off-by:
-
Juri Lelli authored
On !CONFIG_RT_GROUP_SCHED configurations it is currently not possible to move RT tasks between cgroups to which CPU controller has been attached; but it is oddly possible to first move tasks around and then make them RT (setschedule to FIFO/RR). E.g.: # mkdir /sys/fs/cgroup/cpu,cpuacct/group1 # chrt -fp 10 $$ # echo $$ > /sys/fs/cgroup/cpu,cpuacct/group1/tasks bash: echo: write error: Invalid argument # chrt -op 0 $$ # echo $$ > /sys/fs/cgroup/cpu,cpuacct/group1/tasks # chrt -fp 10 $$ # cat /sys/fs/cgroup/cpu,cpuacct/group1/tasks 2345 2598 # chrt -p 2345 pid 2345's current scheduling policy: SCHED_FIFO pid 2345's current scheduling priority: 10 Also, as Michal noted, it is currently not possible to enable CPU controller on unified hierarchy with !CONFIG_RT_GROUP_SCHED (if there are any kernel RT threads in root cgroup, they can't be migrated to the newly created CPU controller's root in cgroup_update_dfl_csses()). Existing code comes with a comment saying the "we don't support RT-tasks being in separate groups". Such comment is however stale and belongs to pre-RT_GROUP_SCHED times. Also, it doesn't make much sense for !RT_GROUP_ SCHED configurations, since checks related to RT bandwidth are not performed at all in these cases. Make moving RT tasks between CPU controller groups viable by removing special case check for RT (and DEADLINE) tasks. Signed-off-by:
Daniel Bristot de Oliveira <bristot@redhat.com> Acked-by:
-
Juri Lelli authored
No synchronisation mechanism exists between the cpuset subsystem and calls to function __sched_setscheduler(). As such, it is possible that new root domains are created on the cpuset side while a deadline acceptance test is carried out in __sched_setscheduler(), leading to a potential oversell of CPU bandwidth. Grab cpuset_rwsem read lock from core scheduler, so to prevent situations such as the one described above from happening. The only exception is normalize_rt_tasks() which needs to work under tasklist_lock and can't therefore grab cpuset_rwsem. We are fine with this, as this function is only called by sysrq and, if that gets triggered, DEADLINE guarantees are already gone out of the window anyway. Tested-by:
Dietmar Eggemann <dietmar.eggemann@arm.com> Signed-off-by:
-
Juri Lelli authored
sched_setscheduler() needs to acquire cpuset_rwsem, but it is currently called from an invalid (atomic) context by rcu_spawn_gp_kthread(). Fix that by simply moving sched_setscheduler_nocheck() call outside of the atomic region, as it doesn't actually require to be guarded by rcu_node lock. Suggested-by:
-
Juri Lelli authored
cpuset_rwsem is going to be acquired from sched_setscheduler() with a following patch. There are however paths (e.g., spawn_ksoftirqd) in which sched_scheduler() is eventually called while holding hotplug lock; this creates a dependecy between hotplug lock (to be always acquired first) and cpuset_rwsem (to be always acquired after hotplug lock). Fix paths which currently take the two locks in the wrong order (after a following patch is applied). Tested-by:
-
Juri Lelli authored
Holding cpuset_mutex means that cpusets are stable (only the holder can make changes) and this is required for fixing a synchronization issue between cpusets and scheduler core. However, grabbing cpuset_mutex from setscheduler() hotpath (as implemented in a later patch) is a no-go, as it would create a bottleneck for tasks concurrently calling setscheduler(). Convert cpuset_mutex to be a percpu_rwsem (cpuset_rwsem), so that setscheduler() will then be able to read lock it and avoid concurrency issues. Tested-by:
-
Juri Lelli authored
If a task happens to be throttled while the CPU it was running on gets hotplugged off, the bandwidth associated with the task is not correctly migrated with it when the replenishment timer fires (offline_migration). Fix things up, for this_bw, running_bw and total_bw, when replenishment timer fires and task is migrated (dl_task_offline_migration()). Tested-by:
-
Mathieu Poirier authored
When the topology of root domains is modified by CPUset or CPUhotplug operations information about the current deadline bandwidth held in the root domain is lost. This patch addresses the issue by recalculating the lost deadline bandwidth information by circling through the deadline tasks held in CPUsets and adding their current load to the root domain they are associated with. Tested-by:
Mathieu Poirier <mathieu.poirier@linaro.org> Signed-off-by:
-
Mathieu Poirier authored
Calls to task_rq_unlock() are done several times in the __sched_setscheduler() function. This is fine when only the rq lock needs to be handled but not so much when other locks come into play. This patch streamlines the release of the rq lock so that only one location need to be modified when dealing with more than one lock. No change of functionality is introduced by this patch. Tested-by:
-
Mathieu Poirier authored
Introduce the partition_sched_domains_locked() function by taking the mutex locking code out of the original function. That way the work done by partition_sched_domains_locked() can be reused without dropping the mutex lock. No change of functionality is introduced by this patch. Tested-by:
-
Viresh Kumar authored
The same formula to check utilization against capacity (after considering capacity_margin) is already used at 5 different locations. This patch creates a new macro, fits_capacity(), which can be used from all these locations without exposing the details of it and hence simplify code. All the 5 code locations are updated as well to use it.. Signed-off-by:
-
Wanpeng Li authored
In real product setup, there will be houseeking CPUs in each nodes, it is prefer to do housekeeping from local node, fallback to global online cpumask if failed to find houseeking CPU from local node. Signed-off-by:
Wanpeng Li <wanpengli@tencent.com> Signed-off-by:
Frederic Weisbecker <frederic@kernel.org> Reviewed-by:
Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: https://lkml.kernel.org/r/1561711901-4755-2-git-send-email-wanpengli@tencent.comSigned-off-by:
-
Yi Wang authored
sched_info_on() is called with unlikely hint, however, the test is to be a constant(1) on which compiler will do nothing when make defconfig, so remove the hint. Also, fix a lack of {}. Signed-off-by:Yi Wang <wang.yi59@zte.com.cn> Signed-off-by:
-
Matthew Wilcox (Oracle) authored
Returning the pointer that was passed in allows us to write slightly more idiomatic code. Convert a few users. Signed-off-by:
Matthew Wilcox (Oracle) <willy@infradead.org> Signed-off-by:
-
Viresh Kumar authored
We try to find an idle CPU to run the next task, but in case we don't find an idle CPU it is better to pick a CPU which will run the task the soonest, for performance reason. A CPU which isn't idle but has only SCHED_IDLE activity queued on it should be a good target based on this criteria as any normal fair task will most likely preempt the currently running SCHED_IDLE task immediately. In fact, choosing a SCHED_IDLE CPU over a fully idle one shall give better results as it should be able to run the task sooner than an idle CPU (which requires to be woken up from an idle state). This patch updates both fast and slow paths with this optimization. Signed-off-by:
-
Viresh Kumar authored
Track how many tasks are present with SCHED_IDLE policy in each cfs_rq. This will be used by later commits. Signed-off-by:
-
Paul E. McKenney authored
time/tick-broadcast: Fix tick_broadcast_offline() lockdep complaint The TASKS03 and TREE04 rcutorture scenarios produce the following lockdep complaint: WARNING: inconsistent lock state 5.2.0-rc1+ #513 Not tainted -------------------------------- inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage. migration/1/14 [HC0[0]:SC0[0]:HE1:SE1] takes: (____ptrval____) (tick_broadcast_lock){?...}, at: tick_broadcast_offline+0xf/0x70 {IN-HARDIRQ-W} state was registered at: lock_acquire+0xb0/0x1c0 _raw_spin_lock_irqsave+0x3c/0x50 tick_broadcast_switch_to_oneshot+0xd/0x40 tick_switch_to_oneshot+0x4f/0xd0 hrtimer_run_queues+0xf3/0x130 run_local_timers+0x1c/0x50 update_process_times+0x1c/0x50 tick_periodic+0x26/0xc0 tick_handle_periodic+0x1a/0x60 smp_apic_timer_interrupt+0x80/0x2a0 apic_timer_interrupt+0xf/0x20 _raw_spin_unlock_irqrestore+0x4e/0x60 rcu_nocb_gp_kthread+0x15d/0x590 kthread+0xf3/0x130 ret_from_fork+0x3a/0x50 irq event stamp: 171 hardirqs last enabled at (171): [<ffffffff8a201a37>] trace_hardirqs_on_thunk+0x1a/0x1c hardirqs last disabled at (170): [<ffffffff8a201a53>] trace_hardirqs_off_thunk+0x1a/0x1c softirqs last enabled at (0): [<ffffffff8a264ee0>] copy_process.part.56+0x650/0x1cb0 softirqs last disabled at (0): [<0000000000000000>] 0x0 [...] To reproduce, run the following rcutorture test: $ tools/testing/selftests/rcutorture/bin/kvm.sh --duration 5 --kconfig "CONFIG_DEBUG_LOCK_ALLOC=y CONFIG_PROVE_LOCKING=y" --configs "TASKS03 TREE04" It turns out that tick_broadcast_offline() was an innocent bystander. After all, interrupts are supposed to be disabled throughout take_cpu_down(), and therefore should have been disabled upon entry to tick_offline_cpu() and thus to tick_broadcast_offline(). This suggests that one of the CPU-hotplug notifiers was incorrectly enabling interrupts, and leaving them enabled on return. Some debugging code showed that the culprit was sched_cpu_dying(). It had irqs enabled after return from sched_tick_stop(). Which in turn had irqs enabled after return from cancel_delayed_work_sync(). Which is a wrapper around __cancel_work_timer(). Which can sleep in the case where something else is concurrently trying to cancel the same delayed work, and as Thomas Gleixner pointed out on IRC, sleeping is a decidedly bad idea when you are invoked from take_cpu_down(), regardless of the state you leave interrupts in upon return. Code inspection located no reason why the delayed work absolutely needed to be canceled from sched_tick_stop(): The work is not bound to the outgoing CPU by design, given that the whole point is to collect statistics without disturbing the outgoing CPU. This commit therefore simply drops the cancel_delayed_work_sync() from sched_tick_stop(). Instead, a new ->state field is added to the tick_work structure so that the delayed-work handler function sched_tick_remote() can avoid reposting itself. A cpu_is_offline() check is also added to sched_tick_remote() to avoid mucking with the state of an offlined CPU (though it does appear safe to do so). The sched_tick_start() and sched_tick_stop() functions also update ->state, and sched_tick_start() also schedules the delayed work if ->state indicates that it is not already in flight. Signed-off-by: -
Vincent Guittot authored
The load_balance() has a dedicated mecanism to detect when an imbalance is due to CPU affinity and must be handled at parent level. In this case, the imbalance field of the parent's sched_group is set. The description of sg_imbalanced() gives a typical example of two groups of 4 CPUs each and 4 tasks each with a cpumask covering 1 CPU of the first group and 3 CPUs of the second group. Something like: { 0 1 2 3 } { 4 5 6 7 } * * * * But the load_balance fails to fix this UC on my octo cores system made of 2 clusters of quad cores. Whereas the load_balance is able to detect that the imbalanced is due to CPU affinity, it fails to fix it because the imbalance field is cleared before letting parent level a chance to run. In fact, when the imbalance is detected, the load_balance reruns without the CPU with pinned tasks. But there is no other running tasks in the situation described above and everything looks balanced this time so the imbalance field is immediately cleared. The imbalance field should not be cleared if there is no other task to move when the imbalance is detected. Signed-off-by: -
Valentin Schneider authored
There are no callers outside of fair.c. Signed-off-by:
Valentin Schneider <valentin.schneider@arm.com> Signed-off-by:
-
Valentin Schneider authored
We only need to set the callback_head worker function once, do it during sched_fork(). While at it, move the comment regarding double task_work addition to init_numa_balancing(), since the double add sentinel is first set there. Suggested-by:
-
Valentin Schneider authored
To reference task_numa_work() from within init_numa_balancing(), we need the former to be declared before the latter. Do just that. This is a pure code movement. Signed-off-by:
-
Jann Horn authored
The old code used RCU annotations and accessors inconsistently for ->numa_group, which can lead to use-after-frees and NULL dereferences. Let all accesses to ->numa_group use proper RCU helpers to prevent such issues. Signed-off-by:
Jann Horn <jannh@google.com> Signed-off-by:
-
Jann Horn authored
When going through execve(), zero out the NUMA fault statistics instead of freeing them. During execve, the task is reachable through procfs and the scheduler. A concurrent /proc/*/sched reader can read data from a freed ->numa_faults allocation (confirmed by KASAN) and write it back to userspace. I believe that it would also be possible for a use-after-free read to occur through a race between a NUMA fault and execve(): task_numa_fault() can lead to task_numa_compare(), which invokes task_weight() on the currently running task of a different CPU. Another way to fix this would be to make ->numa_faults RCU-managed or add extra locking, but it seems easier to wipe the NUMA fault statistics on execve. Signed-off-by:
-
