The scx_bpf_cpuperf_set() kfunc allows a BPF program to set the relative
performance target of a specified CPU. Commit d86adb4fc0 ("sched_ext: Add
cpuperf support") defined the @cpu argument to be unsigned. Let's update it
to be signed to match the norm for the rest of ext.c and the kernel.
Note that the kfunc declaration of scx_bpf_cpuperf_set() in the
common.bpf.h header in tools/sched_ext already listed the cpu as signed, so
this also fixes the build for tools/sched_ext and the sched_ext selftests
due to kfunc declarations now being emitted in vmlinux.h based on BTF (thus
causing the compiler to error due to observing conflicting types).
Fixes: d86adb4fc0 ("sched_ext: Add cpuperf support")
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
sched_ext currently does not integrate with schedutil. When schedutil is the
governor, frequencies are left unregulated and usually get stuck close to
the highest performance level from running RT tasks.
Add CPU performance monitoring and scaling support by integrating into
schedutil. The following kfuncs are added:
- scx_bpf_cpuperf_cap(): Query the relative performance capacity of
different CPUs in the system.
- scx_bpf_cpuperf_cur(): Query the current performance level of a CPU
relative to its max performance.
- scx_bpf_cpuperf_set(): Set the current target performance level of a CPU.
This gives direct control over CPU performance setting to the BPF scheduler.
The only changes on the schedutil side are accounting for the utilization
factor from sched_ext and disabling frequency holding heuristics as it may
not apply well to sched_ext schedulers which may have a lot weaker
connection between tasks and their current / last CPU.
With cpuperf support added, there is no reason to block uclamp. Enable while
at it.
A toy implementation of cpuperf is added to scx_qmap as a demonstration of
the feature.
v2: Ignore cpu_util_cfs_boost() when scx_switched_all() in sugov_get_util()
to avoid factoring in stale util metric. (Christian)
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Cc: Christian Loehle <christian.loehle@arm.com>
sugov_cpu_is_busy() is used to avoid decreasing performance level while the
CPU is busy and called by sugov_update_single_freq() and
sugov_update_single_perf(). Both callers repeat the same pattern to first
test for uclamp and then the business. Let's refactor so that the tests
aren't repeated.
The new helper is named sugov_hold_freq() and tests both the uclamp
exception and CPU business. No functional changes. This will make adding
more exception conditions easier.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Reviewed-by: Christian Loehle <christian.loehle@arm.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
scx_next_task_picked() is used by sched_ext to notify the BPF scheduler when
a CPU is taken away by a task dispatched from a higher priority sched_class
so that the BPF scheduler can, e.g., punt the task[s] which was running or
were waiting for the CPU to other CPUs.
Replace the sched_ext specific hook scx_next_task_picked() with a new
sched_class operation switch_class().
The changes are straightforward and the code looks better afterwards.
However, when !CONFIG_SCHED_CLASS_EXT, this ends up adding an unused hook
which is unlikely to be useful to other sched_classes. For further
discussion on this subject, please refer to the following:
http://lkml.kernel.org/r/CAHk-=wjFPLqo7AXu8maAGEGnOy6reUg-F4zzFhVB0Kyu22h7pw@mail.gmail.com
Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Add Documentation/scheduler/sched-ext.rst which gives a high-level overview
and pointers to the examples.
v6: - Add paragraph explaining debug dump.
v5: - Updated to reflect /sys/kernel interface change. Kconfig options
added.
v4: - README improved, reformatted in markdown and renamed to README.md.
v3: - Added tools/sched_ext/README.
- Dropped _example prefix from scheduler names.
v2: - Apply minor edits suggested by Bagas. Caveats section dropped as all
of them are addressed.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Bagas Sanjaya <bagasdotme@gmail.com>
Currently, a dsq is always a FIFO. A task which is dispatched earlier gets
consumed or executed earlier. While this is sufficient when dsq's are used
for simple staging areas for tasks which are ready to execute, it'd make
dsq's a lot more useful if they can implement custom ordering.
This patch adds a vtime-ordered priority queue to dsq's. When the BPF
scheduler dispatches a task with the new scx_bpf_dispatch_vtime() helper, it
can specify the vtime tha the task should be inserted at and the task is
inserted into the priority queue in the dsq which is ordered according to
time_before64() comparison of the vtime values.
A DSQ can either be a FIFO or priority queue and automatically switches
between the two depending on whether scx_bpf_dispatch() or
scx_bpf_dispatch_vtime() is used. Using the wrong variant while the DSQ
already has the other type queued is not allowed and triggers an ops error.
Built-in DSQs must always be FIFOs.
This makes it very easy for the BPF schedulers to implement proper vtime
based scheduling within each dsq very easy and efficient at a negligible
cost in terms of code complexity and overhead.
scx_simple and scx_example_flatcg are updated to default to weighted
vtime scheduling (the latter within each cgroup). FIFO scheduling can be
selected with -f option.
v4: - As allowing mixing priority queue and FIFO on the same DSQ sometimes
led to unexpected starvations, DSQs now error out if both modes are
used at the same time and the built-in DSQs are no longer allowed to
be priority queues.
- Explicit type struct scx_dsq_node added to contain fields needed to be
linked on DSQs. This will be used to implement stateful iterator.
- Tasks are now always linked on dsq->list whether the DSQ is in FIFO or
PRIQ mode. This confines PRIQ related complexities to the enqueue and
dequeue paths. Other paths only need to look at dsq->list. This will
also ease implementing BPF iterator.
- Print p->scx.dsq_flags in debug dump.
v3: - SCX_TASK_DSQ_ON_PRIQ flag is moved from p->scx.flags into its own
p->scx.dsq_flags. The flag is protected with the dsq lock unlike other
flags in p->scx.flags. This led to flag corruption in some cases.
- Add comments explaining the interaction between using consumption of
p->scx.slice to determine vtime progress and yielding.
v2: - p->scx.dsq_vtime was not initialized on load or across cgroup
migrations leading to some tasks being stalled for extended period of
time depending on how saturated the machine is. Fixed.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
The core-sched support is composed of the following parts:
- task_struct->scx.core_sched_at is added. This is a timestamp which can be
used to order tasks. Depending on whether the BPF scheduler implements
custom ordering, it tracks either global FIFO ordering of all tasks or
local-DSQ ordering within the dispatched tasks on a CPU.
- prio_less() is updated to call scx_prio_less() when comparing SCX tasks.
scx_prio_less() calls ops.core_sched_before() if available or uses the
core_sched_at timestamp. For global FIFO ordering, the BPF scheduler
doesn't need to do anything. Otherwise, it should implement
ops.core_sched_before() which reflects the ordering.
- When core-sched is enabled, balance_scx() balances all SMT siblings so
that they all have tasks dispatched if necessary before pick_task_scx() is
called. pick_task_scx() picks between the current task and the first
dispatched task on the local DSQ based on availability and the
core_sched_at timestamps. Note that FIFO ordering is expected among the
already dispatched tasks whether running or on the local DSQ, so this path
always compares core_sched_at instead of calling into
ops.core_sched_before().
qmap_core_sched_before() is added to scx_qmap. It scales the
distances from the heads of the queues to compare the tasks across different
priority queues and seems to behave as expected.
v3: Fixed build error when !CONFIG_SCHED_SMT reported by Andrea Righi.
v2: Sched core added the const qualifiers to prio_less task arguments.
Explicitly drop them for ops.core_sched_before() task arguments. BPF
enforces access control through the verifier, so the qualifier isn't
actually operative and only gets in the way when interacting with
various helpers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Reviewed-by: Josh Don <joshdon@google.com>
Cc: Andrea Righi <andrea.righi@canonical.com>
PM operations freeze userspace. Some BPF schedulers have active userspace
component and may misbehave as expected across PM events. While the system
is frozen, nothing too interesting is happening in terms of scheduling and
we can get by just fine with the fallback FIFO behavior. Let's make things
easier by always bypassing the BPF scheduler while PM events are in
progress.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Add ops.cpu_online/offline() which are invoked when CPUs come online and
offline respectively. As the enqueue path already automatically bypasses
tasks to the local dsq on a deactivated CPU, BPF schedulers are guaranteed
to see tasks only on CPUs which are between online() and offline().
If the BPF scheduler doesn't implement ops.cpu_online/offline(), the
scheduler is automatically exited with SCX_ECODE_RESTART |
SCX_ECODE_RSN_HOTPLUG. Userspace can implement CPU hotpplug support
trivially by simply reinitializing and reloading the scheduler.
scx_qmap is updated to print out online CPUs on hotplug events. Other
schedulers are updated to restart based on ecode.
v3: - The previous implementation added @reason to
sched_class.rq_on/offline() to distinguish between CPU hotplug events
and topology updates. This was buggy and fragile as the methods are
skipped if the current state equals the target state. Instead, add
scx_rq_[de]activate() which are directly called from
sched_cpu_de/activate(). This also allows ops.cpu_on/offline() to
sleep which can be useful.
- ops.dispatch() could be called on a CPU that the BPF scheduler was
told to be offline. The dispatch patch is updated to bypass in such
cases.
v2: - To accommodate lock ordering change between scx_cgroup_rwsem and
cpus_read_lock(), CPU hotplug operations are put into its own SCX_OPI
block and enabled eariler during scx_ope_enable() so that
cpus_read_lock() can be dropped before acquiring scx_cgroup_rwsem.
- Auto exit with ECODE added.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Scheduler classes are strictly ordered and when a higher priority class has
tasks to run, the lower priority ones lose access to the CPU. Being able to
monitor and act on these events are necessary for use cases includling
strict core-scheduling and latency management.
This patch adds two operations ops.cpu_acquire() and .cpu_release(). The
former is invoked when a CPU becomes available to the BPF scheduler and the
opposite for the latter. This patch also implements
scx_bpf_reenqueue_local() which can be called from .cpu_release() to trigger
requeueing of all tasks in the local dsq of the CPU so that the tasks can be
reassigned to other available CPUs.
scx_pair is updated to use .cpu_acquire/release() along with
%SCX_KICK_WAIT to make the pair scheduling guarantee strict even when a CPU
is preempted by a higher priority scheduler class.
scx_qmap is updated to use .cpu_acquire/release() to empty the local
dsq of a preempted CPU. A similar approach can be adopted by BPF schedulers
that want to have a tight control over latency.
v4: Use the new SCX_KICK_IDLE to wake up a CPU after re-enqueueing.
v3: Drop the const qualifier from scx_cpu_release_args.task. BPF enforces
access control through the verifier, so the qualifier isn't actually
operative and only gets in the way when interacting with various
helpers.
v2: Add p->scx.kf_mask annotation to allow calling scx_bpf_reenqueue_local()
from ops.cpu_release() nested inside ops.init() and other sleepable
operations.
Signed-off-by: David Vernet <dvernet@meta.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
When some SCX operations are in flight, it is known that the subject task's
rq lock is held throughout which makes it safe to access certain fields of
the task - e.g. its current task_group. We want to add SCX kfunc helpers
that can make use of this guarantee - e.g. to help determining the currently
associated CPU cgroup from the task's current task_group.
As it'd be dangerous call such a helper on a task which isn't rq lock
protected, the helper should be able to verify the input task and reject
accordingly. This patch adds sched_ext_entity.kf_tasks[] that track the
tasks which are currently being operated on by a terminal SCX operation. The
new SCX_CALL_OP_[2]TASK[_RET]() can be used when invoking SCX operations
which take tasks as arguments and the scx_kf_allowed_on_arg_tasks() can be
used by kfunc helpers to verify the input task status.
Note that as sched_ext_entity.kf_tasks[] can't handle nesting, the tracking
is currently only limited to terminal SCX operations. If needed in the
future, this restriction can be removed by moving the tracking to the task
side with a couple per-task counters.
v2: Updated to reflect the addition of SCX_KF_SELECT_CPU.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Allow BPF schedulers to indicate tickless operation by setting p->scx.slice
to SCX_SLICE_INF. A CPU whose current task has infinte slice goes into
tickless operation.
scx_central is updated to use tickless operations for all tasks and
instead use a BPF timer to expire slices. This also uses the SCX_ENQ_PREEMPT
and task state tracking added by the previous patches.
Currently, there is no way to pin the timer on the central CPU, so it may
end up on one of the worker CPUs; however, outside of that, the worker CPUs
can go tickless both while running sched_ext tasks and idling.
With schbench running, scx_central shows:
root@test ~# grep ^LOC /proc/interrupts; sleep 10; grep ^LOC /proc/interrupts
LOC: 142024 656 664 449 Local timer interrupts
LOC: 161663 663 665 449 Local timer interrupts
Without it:
root@test ~ [SIGINT]# grep ^LOC /proc/interrupts; sleep 10; grep ^LOC /proc/interrupts
LOC: 188778 3142 3793 3993 Local timer interrupts
LOC: 198993 5314 6323 6438 Local timer interrupts
While scx_central itself is too barebone to be useful as a
production scheduler, a more featureful central scheduler can be built using
the same approach. Google's experience shows that such an approach can have
significant benefits for certain applications such as VM hosting.
v4: Allow operation even if BPF_F_TIMER_CPU_PIN is not available.
v3: Pin the central scheduler's timer on the central_cpu using
BPF_F_TIMER_CPU_PIN.
v2: Convert to BPF inline iterators.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Being able to track the task runnable and running state transitions are
useful for a variety of purposes including latency tracking and load factor
calculation.
Currently, BPF schedulers don't have a good way of tracking these
transitions. Becoming runnable can be determined from ops.enqueue() but
becoming quiescent can only be inferred from the lack of subsequent enqueue.
Also, as the local dsq can have multiple tasks and some events are handled
in the sched_ext core, it's difficult to determine when a given task starts
and stops executing.
This patch adds sched_ext_ops.runnable(), .running(), .stopping() and
.quiescent() operations to track the task runnable and running state
transitions. They're mostly self explanatory; however, we want to ensure
that running <-> stopping transitions are always contained within runnable
<-> quiescent transitions which is a bit different from how the scheduler
core behaves. This adds a bit of complication. See the comment in
dequeue_task_scx().
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
The dispatch path retries if the local DSQ is still empty after
ops.dispatch() either dispatched or consumed a task. This is both out of
necessity and for convenience. It has to retry because the dispatch path
might lose the tasks to dequeue while the rq lock is released while trying
to migrate tasks across CPUs, and the retry mechanism makes ops.dispatch()
implementation easier as it only needs to make some forward progress each
iteration.
However, this makes it possible for ops.dispatch() to stall CPUs by
repeatedly dispatching ineligible tasks. If all CPUs are stalled that way,
the watchdog or sysrq handler can't run and the system can't be saved. Let's
address the issue by breaking out of the dispatch loop after 32 iterations.
It is unlikely but not impossible for ops.dispatch() to legitimately go over
the iteration limit. We want to come back to the dispatch path in such cases
as not doing so risks stalling the CPU by idling with runnable tasks
pending. As the previous task is still current in balance_scx(),
resched_curr() doesn't do anything - it will just get cleared. Let's instead
use scx_kick_bpf() which will trigger reschedule after switching to the next
task which will likely be the idle task.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
It's often useful to wake up and/or trigger reschedule on other CPUs. This
patch adds scx_bpf_kick_cpu() kfunc helper that BPF scheduler can call to
kick the target CPU into the scheduling path.
As a sched_ext task relinquishes its CPU only after its slice is depleted,
this patch also adds SCX_KICK_PREEMPT and SCX_ENQ_PREEMPT which clears the
slice of the target CPU's current task to guarantee that sched_ext's
scheduling path runs on the CPU.
If SCX_KICK_IDLE is specified, the target CPU is kicked iff the CPU is idle
to guarantee that the target CPU will go through at least one full sched_ext
scheduling cycle after the kicking. This can be used to wake up idle CPUs
without incurring unnecessary overhead if it isn't currently idle.
As a demonstration of how backward compatibility can be supported using BPF
CO-RE, tools/sched_ext/include/scx/compat.bpf.h is added. It provides
__COMPAT_scx_bpf_kick_cpu_IDLE() which uses SCX_KICK_IDLE if available or
becomes a regular kicking otherwise. This allows schedulers to use the new
SCX_KICK_IDLE while maintaining support for older kernels. The plan is to
temporarily use compat helpers to ease API updates and drop them after a few
kernel releases.
v5: - SCX_KICK_IDLE added. Note that this also adds a compat mechanism for
schedulers so that they can support kernels without SCX_KICK_IDLE.
This is useful as a demonstration of how new feature flags can be
added in a backward compatible way.
- kick_cpus_irq_workfn() reimplemented so that it touches the pending
cpumasks only as necessary to reduce kicking overhead on machines with
a lot of CPUs.
- tools/sched_ext/include/scx/compat.bpf.h added.
v4: - Move example scheduler to its own patch.
v3: - Make scx_example_central switch all tasks by default.
- Convert to BPF inline iterators.
v2: - Julia Lawall reported that scx_example_central can overflow the
dispatch buffer and malfunction. As scheduling for other CPUs can't be
handled by the automatic retry mechanism, fix by implementing an
explicit overflow and retry handling.
- Updated to use generic BPF cpumask helpers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
If a BPF scheduler triggers an error, the scheduler is aborted and the
system is reverted to the built-in scheduler. In the process, a lot of
information which may be useful for figuring out what happened can be lost.
This patch adds debug dump which captures information which may be useful
for debugging including runqueue and runnable thread states at the time of
failure. The following shows a debug dump after triggering the watchdog:
root@test ~# os/work/tools/sched_ext/build/bin/scx_qmap -t 100
stats : enq=1 dsp=0 delta=1 deq=0
stats : enq=90 dsp=90 delta=0 deq=0
stats : enq=156 dsp=156 delta=0 deq=0
stats : enq=218 dsp=218 delta=0 deq=0
stats : enq=255 dsp=255 delta=0 deq=0
stats : enq=271 dsp=271 delta=0 deq=0
stats : enq=284 dsp=284 delta=0 deq=0
stats : enq=293 dsp=293 delta=0 deq=0
DEBUG DUMP
================================================================================
kworker/u32:12[320] triggered exit kind 1026:
runnable task stall (stress[1530] failed to run for 6.841s)
Backtrace:
scx_watchdog_workfn+0x136/0x1c0
process_scheduled_works+0x2b5/0x600
worker_thread+0x269/0x360
kthread+0xeb/0x110
ret_from_fork+0x36/0x40
ret_from_fork_asm+0x1a/0x30
QMAP FIFO[0]:
QMAP FIFO[1]:
QMAP FIFO[2]: 1436
QMAP FIFO[3]:
QMAP FIFO[4]:
CPU states
----------
CPU 0 : nr_run=1 ops_qseq=244
curr=swapper/0[0] class=idle_sched_class
QMAP: dsp_idx=1 dsp_cnt=0
R stress[1530] -6841ms
scx_state/flags=3/0x1 ops_state/qseq=2/20
sticky/holding_cpu=-1/-1 dsq_id=(n/a)
cpus=ff
QMAP: force_local=0
asm_sysvec_apic_timer_interrupt+0x16/0x20
CPU 2 : nr_run=2 ops_qseq=142
curr=swapper/2[0] class=idle_sched_class
QMAP: dsp_idx=1 dsp_cnt=0
R sshd[1703] -5905ms
scx_state/flags=3/0x9 ops_state/qseq=2/88
sticky/holding_cpu=-1/-1 dsq_id=(n/a)
cpus=ff
QMAP: force_local=1
__x64_sys_ppoll+0xf6/0x120
do_syscall_64+0x7b/0x150
entry_SYSCALL_64_after_hwframe+0x76/0x7e
R fish[1539] -4141ms
scx_state/flags=3/0x9 ops_state/qseq=2/124
sticky/holding_cpu=-1/-1 dsq_id=(n/a)
cpus=ff
QMAP: force_local=1
futex_wait+0x60/0xe0
do_futex+0x109/0x180
__x64_sys_futex+0x117/0x190
do_syscall_64+0x7b/0x150
entry_SYSCALL_64_after_hwframe+0x76/0x7e
CPU 3 : nr_run=2 ops_qseq=162
curr=kworker/u32:12[320] class=ext_sched_class
QMAP: dsp_idx=1 dsp_cnt=0
*R kworker/u32:12[320] +0ms
scx_state/flags=3/0xd ops_state/qseq=0/0
sticky/holding_cpu=-1/-1 dsq_id=(n/a)
cpus=ff
QMAP: force_local=0
scx_dump_state+0x613/0x6f0
scx_ops_error_irq_workfn+0x1f/0x40
irq_work_run_list+0x82/0xd0
irq_work_run+0x14/0x30
__sysvec_irq_work+0x40/0x140
sysvec_irq_work+0x60/0x70
asm_sysvec_irq_work+0x16/0x20
scx_watchdog_workfn+0x15f/0x1c0
process_scheduled_works+0x2b5/0x600
worker_thread+0x269/0x360
kthread+0xeb/0x110
ret_from_fork+0x36/0x40
ret_from_fork_asm+0x1a/0x30
R kworker/3:2[1436] +0ms
scx_state/flags=3/0x9 ops_state/qseq=2/160
sticky/holding_cpu=-1/-1 dsq_id=(n/a)
cpus=08
QMAP: force_local=0
kthread+0xeb/0x110
ret_from_fork+0x36/0x40
ret_from_fork_asm+0x1a/0x30
CPU 7 : nr_run=0 ops_qseq=76
curr=swapper/7[0] class=idle_sched_class
================================================================================
EXIT: runnable task stall (stress[1530] failed to run for 6.841s)
It shows that CPU 3 was running the watchdog when it triggered the error
condition and the scx_qmap thread has been queued on CPU 0 for over 5
seconds but failed to run. It also prints out scx_qmap specific information
- e.g. which tasks are queued on each FIFO and so on using the dump_*() ops.
This dump has proved pretty useful for developing and debugging BPF
schedulers.
Debug dump is generated automatically when the BPF scheduler exits due to an
error. The debug buffer used in such cases is determined by
sched_ext_ops.exit_dump_len and defaults to 32k. If the debug dump overruns
the available buffer, the output is truncated and marked accordingly.
Debug dump output can also be read through the sched_ext_dump tracepoint.
When read through the tracepoint, there is no length limit.
SysRq-D can be used to trigger debug dump at any time while a BPF scheduler
is loaded. This is non-destructive - the scheduler keeps running afterwards.
The output can be read through the sched_ext_dump tracepoint.
v2: - The size of exit debug dump buffer can now be customized using
sched_ext_ops.exit_dump_len.
- sched_ext_ops.dump*() added to enable dumping of BPF scheduler
specific information.
- Tracpoint output and SysRq-D triggering added.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
It would be useful to see what the sched_ext scheduler state is, and what
scheduler is running, when we're dumping a task's stack. This patch
therefore adds a new print_scx_info() function that's called in the same
context as print_worker_info() and print_stop_info(). An example dump
follows.
BUG: kernel NULL pointer dereference, address: 0000000000000999
#PF: supervisor write access in kernel mode
#PF: error_code(0x0002) - not-present page
PGD 0 P4D 0
Oops: 0002 [#1] PREEMPT SMP
CPU: 13 PID: 2047 Comm: insmod Tainted: G O 6.6.0-work-10323-gb58d4cae8e99-dirty #34
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS unknown 2/2/2022
Sched_ext: qmap (enabled+all), task: runnable_at=-17ms
RIP: 0010:init_module+0x9/0x1000 [test_module]
...
v3: - scx_ops_enable_state_str[] definition moved to an earlier patch as
it's now used by core implementation.
- Convert jiffy delta to msecs using jiffies_to_msecs() instead of
multiplying by (HZ / MSEC_PER_SEC). The conversion is implemented in
jiffies_delta_msecs().
v2: - We are now using scx_ops_enable_state_str[] outside
CONFIG_SCHED_DEBUG. Move it outside of CONFIG_SCHED_DEBUG and to the
top. This was reported by Changwoo and Andrea.
Signed-off-by: David Vernet <void@manifault.com>
Reported-by: Changwoo Min <changwoo@igalia.com>
Reported-by: Andrea Righi <andrea.righi@canonical.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
BPF schedulers might not want to schedule certain tasks - e.g. kernel
threads. This patch adds p->scx.disallow which can be set by BPF schedulers
in such cases. The field can be changed anytime and setting it in
ops.prep_enable() guarantees that the task can never be scheduled by
sched_ext.
scx_qmap is updated with the -d option to disallow a specific PID:
# echo $$
1092
# grep -E '(policy)|(ext\.enabled)' /proc/self/sched
policy : 0
ext.enabled : 0
# ./set-scx 1092
# grep -E '(policy)|(ext\.enabled)' /proc/self/sched
policy : 7
ext.enabled : 0
Run "scx_qmap -p -d 1092" in another terminal.
# cat /sys/kernel/sched_ext/nr_rejected
1
# grep -E '(policy)|(ext\.enabled)' /proc/self/sched
policy : 0
ext.enabled : 0
# ./set-scx 1092
setparam failed for 1092 (Permission denied)
- v4: Refreshed on top of tip:sched/core.
- v3: Update description to reflect /sys/kernel/sched_ext interface change.
- v2: Use atomic_long_t instead of atomic64_t for scx_kick_cpus_pnt_seqs to
accommodate 32bit archs.
Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Barret Rhoden <brho@google.com>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
The most common and critical way that a BPF scheduler can misbehave is by
failing to run runnable tasks for too long. This patch implements a
watchdog.
* All tasks record when they become runnable.
* A watchdog work periodically scans all runnable tasks. If any task has
stayed runnable for too long, the BPF scheduler is aborted.
* scheduler_tick() monitors whether the watchdog itself is stuck. If so, the
BPF scheduler is aborted.
Because the watchdog only scans the tasks which are currently runnable and
usually very infrequently, the overhead should be negligible.
scx_qmap is updated so that it can be told to stall user and/or
kernel tasks.
A detected task stall looks like the following:
sched_ext: BPF scheduler "qmap" errored, disabling
sched_ext: runnable task stall (dbus-daemon[953] failed to run for 6.478s)
scx_check_timeout_workfn+0x10e/0x1b0
process_one_work+0x287/0x560
worker_thread+0x234/0x420
kthread+0xe9/0x100
ret_from_fork+0x1f/0x30
A detected watchdog stall:
sched_ext: BPF scheduler "qmap" errored, disabling
sched_ext: runnable task stall (watchdog failed to check in for 5.001s)
scheduler_tick+0x2eb/0x340
update_process_times+0x7a/0x90
tick_sched_timer+0xd8/0x130
__hrtimer_run_queues+0x178/0x3b0
hrtimer_interrupt+0xfc/0x390
__sysvec_apic_timer_interrupt+0xb7/0x2b0
sysvec_apic_timer_interrupt+0x90/0xb0
asm_sysvec_apic_timer_interrupt+0x1b/0x20
default_idle+0x14/0x20
arch_cpu_idle+0xf/0x20
default_idle_call+0x50/0x90
do_idle+0xe8/0x240
cpu_startup_entry+0x1d/0x20
kernel_init+0x0/0x190
start_kernel+0x0/0x392
start_kernel+0x324/0x392
x86_64_start_reservations+0x2a/0x2c
x86_64_start_kernel+0x104/0x109
secondary_startup_64_no_verify+0xce/0xdb
Note that this patch exposes scx_ops_error[_type]() in kernel/sched/ext.h to
inline scx_notify_sched_tick().
v4: - While disabling, cancel_delayed_work_sync(&scx_watchdog_work) was
being called before forward progress was guaranteed and thus could
lead to system lockup. Relocated.
- While enabling, it was comparing msecs against jiffies without
conversion leading to spurious load failures on lower HZ kernels.
Fixed.
- runnable list management is now used by core bypass logic and moved to
the patch implementing sched_ext core.
v3: - bpf_scx_init_member() was incorrectly comparing ops->timeout_ms
against SCX_WATCHDOG_MAX_TIMEOUT which is in jiffies without
conversion leading to spurious load failures in lower HZ kernels.
Fixed.
v2: - Julia Lawall noticed that the watchdog code was mixing msecs and
jiffies. Fix by using jiffies for everything.
Signed-off-by: David Vernet <dvernet@meta.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Julia Lawall <julia.lawall@inria.fr>
Implement a new scheduler class sched_ext (SCX), which allows scheduling
policies to be implemented as BPF programs to achieve the following:
1. Ease of experimentation and exploration: Enabling rapid iteration of new
scheduling policies.
2. Customization: Building application-specific schedulers which implement
policies that are not applicable to general-purpose schedulers.
3. Rapid scheduler deployments: Non-disruptive swap outs of scheduling
policies in production environments.
sched_ext leverages BPF’s struct_ops feature to define a structure which
exports function callbacks and flags to BPF programs that wish to implement
scheduling policies. The struct_ops structure exported by sched_ext is
struct sched_ext_ops, and is conceptually similar to struct sched_class. The
role of sched_ext is to map the complex sched_class callbacks to the more
simple and ergonomic struct sched_ext_ops callbacks.
For more detailed discussion on the motivations and overview, please refer
to the cover letter.
Later patches will also add several example schedulers and documentation.
This patch implements the minimum core framework to enable implementation of
BPF schedulers. Subsequent patches will gradually add functionalities
including safety guarantee mechanisms, nohz and cgroup support.
include/linux/sched/ext.h defines struct sched_ext_ops. With the comment on
top, each operation should be self-explanatory. The followings are worth
noting:
- Both "sched_ext" and its shorthand "scx" are used. If the identifier
already has "sched" in it, "ext" is used; otherwise, "scx".
- In sched_ext_ops, only .name is mandatory. Every operation is optional and
if omitted a simple but functional default behavior is provided.
- A new policy constant SCHED_EXT is added and a task can select sched_ext
by invoking sched_setscheduler(2) with the new policy constant. However,
if the BPF scheduler is not loaded, SCHED_EXT is the same as SCHED_NORMAL
and the task is scheduled by CFS. When the BPF scheduler is loaded, all
tasks which have the SCHED_EXT policy are switched to sched_ext.
- To bridge the workflow imbalance between the scheduler core and
sched_ext_ops callbacks, sched_ext uses simple FIFOs called dispatch
queues (dsq's). By default, there is one global dsq (SCX_DSQ_GLOBAL), and
one local per-CPU dsq (SCX_DSQ_LOCAL). SCX_DSQ_GLOBAL is provided for
convenience and need not be used by a scheduler that doesn't require it.
SCX_DSQ_LOCAL is the per-CPU FIFO that sched_ext pulls from when putting
the next task on the CPU. The BPF scheduler can manage an arbitrary number
of dsq's using scx_bpf_create_dsq() and scx_bpf_destroy_dsq().
- sched_ext guarantees system integrity no matter what the BPF scheduler
does. To enable this, each task's ownership is tracked through
p->scx.ops_state and all tasks are put on scx_tasks list. The disable path
can always recover and revert all tasks back to CFS. See p->scx.ops_state
and scx_tasks.
- A task is not tied to its rq while enqueued. This decouples CPU selection
from queueing and allows sharing a scheduling queue across an arbitrary
subset of CPUs. This adds some complexities as a task may need to be
bounced between rq's right before it starts executing. See
dispatch_to_local_dsq() and move_task_to_local_dsq().
- One complication that arises from the above weak association between task
and rq is that synchronizing with dequeue() gets complicated as dequeue()
may happen anytime while the task is enqueued and the dispatch path might
need to release the rq lock to transfer the task. Solving this requires a
bit of complexity. See the logic around p->scx.sticky_cpu and
p->scx.ops_qseq.
- Both enable and disable paths are a bit complicated. The enable path
switches all tasks without blocking to avoid issues which can arise from
partially switched states (e.g. the switching task itself being starved).
The disable path can't trust the BPF scheduler at all, so it also has to
guarantee forward progress without blocking. See scx_ops_enable() and
scx_ops_disable_workfn().
- When sched_ext is disabled, static_branches are used to shut down the
entry points from hot paths.
v7: - scx_ops_bypass() was incorrectly and unnecessarily trying to grab
scx_ops_enable_mutex which can lead to deadlocks in the disable path.
Fixed.
- Fixed TASK_DEAD handling bug in scx_ops_enable() path which could lead
to use-after-free.
- Consolidated per-cpu variable usages and other cleanups.
v6: - SCX_NR_ONLINE_OPS replaced with SCX_OPI_*_BEGIN/END so that multiple
groups can be expressed. Later CPU hotplug operations are put into
their own group.
- SCX_OPS_DISABLING state is replaced with the new bypass mechanism
which allows temporarily putting the system into simple FIFO
scheduling mode bypassing the BPF scheduler. In addition to the shut
down path, this will also be used to isolate the BPF scheduler across
PM events. Enabling and disabling the bypass mode requires iterating
all runnable tasks. rq->scx.runnable_list addition is moved from the
later watchdog patch.
- ops.prep_enable() is replaced with ops.init_task() and
ops.enable/disable() are now called whenever the task enters and
leaves sched_ext instead of when the task becomes schedulable on
sched_ext and stops being so. A new operation - ops.exit_task() - is
called when the task stops being schedulable on sched_ext.
- scx_bpf_dispatch() can now be called from ops.select_cpu() too. This
removes the need for communicating local dispatch decision made by
ops.select_cpu() to ops.enqueue() via per-task storage.
SCX_KF_SELECT_CPU is added to support the change.
- SCX_TASK_ENQ_LOCAL which told the BPF scheudler that
scx_select_cpu_dfl() wants the task to be dispatched to the local DSQ
was removed. Instead, scx_bpf_select_cpu_dfl() now dispatches directly
if it finds a suitable idle CPU. If such behavior is not desired,
users can use scx_bpf_select_cpu_dfl() which returns the verdict in a
bool out param.
- scx_select_cpu_dfl() was mishandling WAKE_SYNC and could end up
queueing many tasks on a local DSQ which makes tasks to execute in
order while other CPUs stay idle which made some hackbench numbers
really bad. Fixed.
- The current state of sched_ext can now be monitored through files
under /sys/sched_ext instead of /sys/kernel/debug/sched/ext. This is
to enable monitoring on kernels which don't enable debugfs.
- sched_ext wasn't telling BPF that ops.dispatch()'s @prev argument may
be NULL and a BPF scheduler which derefs the pointer without checking
could crash the kernel. Tell BPF. This is currently a bit ugly. A
better way to annotate this is expected in the future.
- scx_exit_info updated to carry pointers to message buffers instead of
embedding them directly. This decouples buffer sizes from API so that
they can be changed without breaking compatibility.
- exit_code added to scx_exit_info. This is used to indicate different
exit conditions on non-error exits and will be used to handle e.g. CPU
hotplugs.
- The patch "sched_ext: Allow BPF schedulers to switch all eligible
tasks into sched_ext" is folded in and the interface is changed so
that partial switching is indicated with a new ops flag
%SCX_OPS_SWITCH_PARTIAL. This makes scx_bpf_switch_all() unnecessasry
and in turn SCX_KF_INIT. ops.init() is now called with
SCX_KF_SLEEPABLE.
- Code reorganized so that only the parts necessary to integrate with
the rest of the kernel are in the header files.
- Changes to reflect the BPF and other kernel changes including the
addition of bpf_sched_ext_ops.cfi_stubs.
v5: - To accommodate 32bit configs, p->scx.ops_state is now atomic_long_t
instead of atomic64_t and scx_dsp_buf_ent.qseq which uses
load_acquire/store_release is now unsigned long instead of u64.
- Fix the bug where bpf_scx_btf_struct_access() was allowing write
access to arbitrary fields.
- Distinguish kfuncs which can be called from any sched_ext ops and from
anywhere. e.g. scx_bpf_pick_idle_cpu() can now be called only from
sched_ext ops.
- Rename "type" to "kind" in scx_exit_info to make it easier to use on
languages in which "type" is a reserved keyword.
- Since cff9b2332a ("kernel/sched: Modify initial boot task idle
setup"), PF_IDLE is not set on idle tasks which haven't been online
yet which made scx_task_iter_next_filtered() include those idle tasks
in iterations leading to oopses. Update scx_task_iter_next_filtered()
to directly test p->sched_class against idle_sched_class instead of
using is_idle_task() which tests PF_IDLE.
- Other updates to match upstream changes such as adding const to
set_cpumask() param and renaming check_preempt_curr() to
wakeup_preempt().
v4: - SCHED_CHANGE_BLOCK replaced with the previous
sched_deq_and_put_task()/sched_enq_and_set_tsak() pair. This is
because upstream is adaopting a different generic cleanup mechanism.
Once that lands, the code will be adapted accordingly.
- task_on_scx() used to test whether a task should be switched into SCX,
which is confusing. Renamed to task_should_scx(). task_on_scx() now
tests whether a task is currently on SCX.
- scx_has_idle_cpus is barely used anymore and replaced with direct
check on the idle cpumask.
- SCX_PICK_IDLE_CORE added and scx_pick_idle_cpu() improved to prefer
fully idle cores.
- ops.enable() now sees up-to-date p->scx.weight value.
- ttwu_queue path is disabled for tasks on SCX to avoid confusing BPF
schedulers expecting ->select_cpu() call.
- Use cpu_smt_mask() instead of topology_sibling_cpumask() like the rest
of the scheduler.
v3: - ops.set_weight() added to allow BPF schedulers to track weight changes
without polling p->scx.weight.
- move_task_to_local_dsq() was losing SCX-specific enq_flags when
enqueueing the task on the target dsq because it goes through
activate_task() which loses the upper 32bit of the flags. Carry the
flags through rq->scx.extra_enq_flags.
- scx_bpf_dispatch(), scx_bpf_pick_idle_cpu(), scx_bpf_task_running()
and scx_bpf_task_cpu() now use the new KF_RCU instead of
KF_TRUSTED_ARGS to make it easier for BPF schedulers to call them.
- The kfunc helper access control mechanism implemented through
sched_ext_entity.kf_mask is improved. Now SCX_CALL_OP*() is always
used when invoking scx_ops operations.
v2: - balance_scx_on_up() is dropped. Instead, on UP, balance_scx() is
called from put_prev_taks_scx() and pick_next_task_scx() as necessary.
To determine whether balance_scx() should be called from
put_prev_task_scx(), SCX_TASK_DEQD_FOR_SLEEP flag is added. See the
comment in put_prev_task_scx() for details.
- sched_deq_and_put_task() / sched_enq_and_set_task() sequences replaced
with SCHED_CHANGE_BLOCK().
- Unused all_dsqs list removed. This was a left-over from previous
iterations.
- p->scx.kf_mask is added to track and enforce which kfunc helpers are
allowed. Also, init/exit sequences are updated to make some kfuncs
always safe to call regardless of the current BPF scheduler state.
Combined, this should make all the kfuncs safe.
- BPF now supports sleepable struct_ops operations. Hacky workaround
removed and operations and kfunc helpers are tagged appropriately.
- BPF now supports bitmask / cpumask helpers. scx_bpf_get_idle_cpumask()
and friends are added so that BPF schedulers can use the idle masks
with the generic helpers. This replaces the hacky kfunc helpers added
by a separate patch in V1.
- CONFIG_SCHED_CLASS_EXT can no longer be enabled if SCHED_CORE is
enabled. This restriction will be removed by a later patch which adds
core-sched support.
- Add MAINTAINERS entries and other misc changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Co-authored-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Cc: Andrea Righi <andrea.righi@canonical.com>
This adds dummy implementations of sched_ext interfaces which interact with
the scheduler core and hook them in the correct places. As they're all
dummies, this doesn't cause any behavior changes. This is split out to help
reviewing.
v2: balance_scx_on_up() dropped. This will be handled in sched_ext proper.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
A new BPF extensible sched_class will need to dynamically change how a task
picks its sched_class. For example, if the loaded BPF scheduler progs fail,
the tasks will be forced back on CFS even if the task's policy is set to the
new sched_class. To support such mapping, add normal_policy() which wraps
testing for %SCHED_NORMAL. This doesn't cause any behavior changes.
v2: Update the description with more details on the expected use.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
RT, DL, thermal and irq load and utilization metrics need to be decayed and
updated periodically and before consumption to keep the numbers reasonable.
This is currently done from __update_blocked_others() as a part of the fair
class load balance path. Let's factor it out to update_other_load_avgs().
Pure refactor. No functional changes.
This will be used by the new BPF extensible scheduling class to ensure that
the above metrics are properly maintained.
v2: Refreshed on top of tip:sched/core.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Factor out sched_weight_from/to_cgroup() which convert between scheduler
shares and cgroup weight. No functional change. The factored out functions
will be used by a new BPF extensible sched_class so that the weights can be
exposed to the BPF programs in a way which is consistent cgroup weights and
easier to interpret.
The weight conversions will be used regardless of cgroup usage. It's just
borrowing the cgroup weight range as it's more intuitive.
CGROUP_WEIGHT_MIN/DFL/MAX constants are moved outside CONFIG_CGROUPS so that
the conversion helpers can always be defined.
v2: The helpers are now defined regardless of COFNIG_CGROUPS.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
When a task switches to a new sched_class, the prev and new classes are
notified through ->switched_from() and ->switched_to(), respectively, after
the switching is done.
A new BPF extensible sched_class will have callbacks that allow the BPF
scheduler to keep track of relevant task states (like priority and cpumask).
Those callbacks aren't called while a task is on a different sched_class.
When a task comes back, we wanna tell the BPF progs the up-to-date state
before the task gets enqueued, so we need a hook which is called before the
switching is committed.
This patch adds ->switching_to() which is called during sched_class switch
through check_class_changing() before the task is restored. Also, this patch
exposes check_class_changing/changed() in kernel/sched/sched.h. They will be
used by the new BPF extensible sched_class to implement implicit sched_class
switching which is used e.g. when falling back to CFS when the BPF scheduler
fails or unloads.
This is a prep patch and doesn't cause any behavior changes. The new
operation and exposed functions aren't used yet.
v3: Refreshed on top of tip:sched/core.
v2: Improve patch description w/ details on planned use.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Currently, during a task weight change, sched core directly calls
reweight_task() defined in fair.c if @p is on CFS. Let's make it a proper
sched_class operation instead. CFS's reweight_task() is renamed to
reweight_task_fair() and now called through sched_class.
While it turns a direct call into an indirect one, set_load_weight() isn't
called from a hot path and this change shouldn't cause any noticeable
difference. This will be used to implement reweight_task for a new BPF
extensible sched_class so that it can keep its cached task weight
up-to-date.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
A new BPF extensible sched_class will need more control over the forking
process. It wants to be able to fail from sched_cgroup_fork() after the new
task's sched_task_group is initialized so that the loaded BPF program can
prepare the task with its cgroup association is established and reject fork
if e.g. allocation fails.
Allow sched_cgroup_fork() to fail by making it return int instead of void
and adding sched_cancel_fork() to undo sched_fork() in the error path.
sched_cgroup_fork() doesn't fail yet and this patch shouldn't cause any
behavior changes.
v2: Patch description updated to detail the expected use.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: David Vernet <dvernet@meta.com>
Acked-by: Josh Don <joshdon@google.com>
Acked-by: Hao Luo <haoluo@google.com>
Acked-by: Barret Rhoden <brho@google.com>
Currently, sched_init() checks that the sched_class'es are in the expected
order by testing each adjacency which is a bit brittle and makes it
cumbersome to add optional sched_class'es. Instead, let's verify whether
they're in the expected order using sched_class_above() which is what
matters.
Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: David Vernet <dvernet@meta.com>
When RCU-TASKS-TRACE pre-gp takes a snapshot of the current task running
on all online CPUs, no explicit ordering synchronizes properly with a
context switch. This lack of ordering can permit the new task to miss
pre-grace-period update-side accesses. The following diagram, courtesy
of Paul, shows the possible bad scenario:
CPU 0 CPU 1
----- -----
// Pre-GP update side access
WRITE_ONCE(*X, 1);
smp_mb();
r0 = rq->curr;
RCU_INIT_POINTER(rq->curr, TASK_B)
spin_unlock(rq)
rcu_read_lock_trace()
r1 = X;
/* ignore TASK_B */
Either r0==TASK_B or r1==1 is needed but neither is guaranteed.
One possible solution to solve this is to wait for an RCU grace period
at the beginning of the RCU-tasks-trace grace period before taking the
current tasks snaphot. However this would introduce large additional
latencies to RCU-tasks-trace grace periods.
Another solution is to lock the target runqueue while taking the current
task snapshot. This ensures that the update side sees the latest context
switch and subsequent context switches will see the pre-grace-period
update side accesses.
This commit therefore adds runqueue locking to cpu_curr_snapshot().
Fixes: e386b67257 ("rcu-tasks: Eliminate RCU Tasks Trace IPIs to online CPUs")
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
We observed that the overhead on trigger_load_balance(), now renamed
sched_balance_trigger(), has risen with a system's core counts.
For an OLTP workload running 6.8 kernel on a 2 socket x86 systems
having 96 cores/socket, we saw that 0.7% cpu cycles are spent in
trigger_load_balance(). On older systems with fewer cores/socket, this
function's overhead was less than 0.1%.
The cause of this overhead was that there are multiple cpus calling
kick_ilb(flags), updating the balancing work needed to a common idle
load balancer cpu. The ilb_cpu's flags field got updated unconditionally
with atomic_fetch_or(). The atomic read and writes to ilb_cpu's flags
causes much cache bouncing and cpu cycles overhead. This is seen in the
annotated profile below.
kick_ilb():
if (ilb_cpu < 0)
test %r14d,%r14d
↑ js 6c
flags = atomic_fetch_or(flags, nohz_flags(ilb_cpu));
mov $0x2d600,%rdi
movslq %r14d,%r8
mov %rdi,%rdx
add -0x7dd0c3e0(,%r8,8),%rdx
arch_atomic_read():
0.01 mov 0x64(%rdx),%esi
35.58 add $0x64,%rdx
arch_atomic_fetch_or():
static __always_inline int arch_atomic_fetch_or(int i, atomic_t *v)
{
int val = arch_atomic_read(v);
do { } while (!arch_atomic_try_cmpxchg(v, &val, val | i));
0.03 157: mov %r12d,%ecx
arch_atomic_try_cmpxchg():
return arch_try_cmpxchg(&v->counter, old, new);
0.00 mov %esi,%eax
arch_atomic_fetch_or():
do { } while (!arch_atomic_try_cmpxchg(v, &val, val | i));
or %esi,%ecx
arch_atomic_try_cmpxchg():
return arch_try_cmpxchg(&v->counter, old, new);
0.01 lock cmpxchg %ecx,(%rdx)
42.96 ↓ jne 2d2
kick_ilb():
With instrumentation, we found that 81% of the updates do not result in
any change in the ilb_cpu's flags. That is, multiple cpus are asking
the ilb_cpu to do the same things over and over again, before the ilb_cpu
has a chance to run NOHZ load balance.
Skip updates to ilb_cpu's flags if no new work needs to be done.
Such updates do not change ilb_cpu's NOHZ flags. This requires an extra
atomic read but it is less expensive than frequent unnecessary atomic
updates that generate cache bounces.
We saw that on the OLTP workload, cpu cycles from trigger_load_balance()
(or sched_balance_trigger()) got reduced from 0.7% to 0.2%.
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Chen Yu <yu.c.chen@intel.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20240531205452.65781-1-tim.c.chen@linux.intel.com
There's a random number of structure pre-declaration lines in
kernel/sched/sched.h, some of which are unnecessary duplicates.
Move them to the head & order them a bit for readability.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
core.c has become rather large, move most scheduler syscall
related functionality into a separate file, syscalls.c.
This is about ~15% of core.c's raw linecount.
Move the alloc_user_cpus_ptr(), __rt_effective_prio(),
rt_effective_prio(), uclamp_none(), uclamp_se_set()
and uclamp_bucket_id() inlines to kernel/sched/sched.h.
Internally export the __sched_setscheduler(), __sched_setaffinity(),
__setscheduler_prio(), set_load_weight(), enqueue_task(), dequeue_task(),
check_class_changed(), splice_balance_callbacks() and balance_callbacks()
methods to better facilitate this.
Move the new file's build to sched_policy.c, because it fits there
semantically, but also because it's the smallest of the 4 build units
under an allmodconfig build:
-rw-rw-r-- 1 mingo mingo 7.3M May 27 12:35 kernel/sched/core.i
-rw-rw-r-- 1 mingo mingo 6.4M May 27 12:36 kernel/sched/build_utility.i
-rw-rw-r-- 1 mingo mingo 6.3M May 27 12:36 kernel/sched/fair.i
-rw-rw-r-- 1 mingo mingo 5.8M May 27 12:36 kernel/sched/build_policy.i
This better balances build time for scheduler subsystem rebuilds.
I build-tested this new file as a standalone syscalls.o file for a bit,
to make sure all the encapsulations & abstractions are robust.
Also update/add my copyright notices to these files.
Build time measurements:
# -Before/+After:
kepler:~/tip> perf stat -e 'cycles,instructions,duration_time' --sync --repeat 5 --pre 'rm -f kernel/sched/*.o' m kernel/sched/built-in.a >/dev/null
Performance counter stats for 'm kernel/sched/built-in.a' (5 runs):
- 71,938,508,607 cycles ( +- 0.17% )
+ 71,992,916,493 cycles ( +- 0.22% )
- 106,214,780,964 instructions # 1.48 insn per cycle ( +- 0.01% )
+ 105,450,231,154 instructions # 1.46 insn per cycle ( +- 0.01% )
- 5,878,232,620 ns duration_time ( +- 0.38% )
+ 5,290,085,069 ns duration_time ( +- 0.21% )
- 5.8782 +- 0.0221 seconds time elapsed ( +- 0.38% )
+ 5.2901 +- 0.0111 seconds time elapsed ( +- 0.21% )
Build time improvement of -11.1% (duration_time) is expected: the
parallel build time of the scheduler subsystem is determined by the
largest, slowest to build object file, which is kernel/sched/core.o.
By moving ~15% of its complexity into another build unit, we reduced
build time by -11%.
Measured cycles spent on building is within its ~0.2% stddev noise envelope.
The -0.7% reduction in instructions spent on building the scheduler is
statistically reliable and somewhat surprising - I can only speculate:
maybe compilers aren't that efficient at building & optimizing 10+ KLOC files
(core.c), and it's an overall win to balance the linecount a bit.
Anyway, this might be a data point that suggests that reducing the linecount
of our largest files will improve not just code readability and maintainability,
but might also improve build times a bit.
Code generation got a bit worse, by 0.5kb text on an x86 defconfig build:
# -Before/+After:
kepler:~/tip> size vmlinux
text data bss dec hex filename
-26475475 10439178 1740804 38655457 24dd5e1 vmlinux
+26476003 10439178 1740804 38655985 24dd7f1 vmlinux
kepler:~/tip> size kernel/sched/built-in.a
text data bss dec hex filename
- 76056 30025 489 106570 1a04a kernel/sched/core.o (ex kernel/sched/built-in.a)
+ 63452 29453 489 93394 16cd2 kernel/sched/core.o (ex kernel/sched/built-in.a)
44299 2181 104 46584 b5f8 kernel/sched/fair.o (ex kernel/sched/built-in.a)
- 42764 3424 120 46308 b4e4 kernel/sched/build_policy.o (ex kernel/sched/built-in.a)
+ 55651 4044 120 59815 e9a7 kernel/sched/build_policy.o (ex kernel/sched/built-in.a)
44866 12655 2192 59713 e941 kernel/sched/build_utility.o (ex kernel/sched/built-in.a)
44866 12655 2192 59713 e941 kernel/sched/build_utility.o (ex kernel/sched/built-in.a)
This is primarily due to the extra functions exported, and the size
gets exaggerated somewhat by __pfx CFI function padding:
ffffffff810cc710 <__pfx_enqueue_task>:
ffffffff810cc710: 90 nop
ffffffff810cc711: 90 nop
ffffffff810cc712: 90 nop
ffffffff810cc713: 90 nop
ffffffff810cc714: 90 nop
ffffffff810cc715: 90 nop
ffffffff810cc716: 90 nop
ffffffff810cc717: 90 nop
ffffffff810cc718: 90 nop
ffffffff810cc719: 90 nop
ffffffff810cc71a: 90 nop
ffffffff810cc71b: 90 nop
ffffffff810cc71c: 90 nop
ffffffff810cc71d: 90 nop
ffffffff810cc71e: 90 nop
ffffffff810cc71f: 90 nop
AFAICS the cost is primarily not to core.o and fair.o though (which contain
most performance sensitive scheduler functions), only to syscalls.o
that get called with much lower frequency - so I think this is an acceptable
trade-off for better code separation.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Link: https://lore.kernel.org/r/20240407084319.1462211-2-mingo@kernel.org
Pull bitmap updates from Yury Norov:
- topology_span_sane() optimization from Kyle Meyer
- fns() rework from Kuan-Wei Chiu (used in cpumask_local_spread() and
other places)
- headers cleanup from Andy
- add a MAINTAINERS record for bitops API
* tag 'bitmap-for-6.10v2' of https://github.com/norov/linux:
usercopy: Don't use "proxy" headers
bitops: Move aligned_byte_mask() to wordpart.h
MAINTAINERS: add BITOPS API record
bitmap: relax find_nth_bit() limitation on return value
lib: make test_bitops compilable into the kernel image
bitops: Optimize fns() for improved performance
lib/test_bitops: Add benchmark test for fns()
Compiler Attributes: Add __always_used macro
sched/topology: Optimize topology_span_sane()
cpumask: Add for_each_cpu_from()
Pull sysctl updates from Joel Granados:
- Remove sentinel elements from ctl_table structs in kernel/*
Removing sentinels in ctl_table arrays reduces the build time size
and runtime memory consumed by ~64 bytes per array. Removals for
net/, io_uring/, mm/, ipc/ and security/ are set to go into mainline
through their respective subsystems making the next release the most
likely place where the final series that removes the check for
proc_name == NULL will land.
This adds to removals already in arch/, drivers/ and fs/.
- Adjust ctl_table definitions and references to allow constification
- Remove unused ctl_table function arguments
- Move non-const elements from ctl_table to ctl_table_header
- Make ctl_table pointers const in ctl_table_root structure
Making the static ctl_table structs const will increase safety by
keeping the pointers to proc_handler functions in .rodata. Though no
ctl_tables where made const in this PR, the ground work for making
that possible has started with these changes sent by Thomas
Weißschuh.
* tag 'sysctl-6.10-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/sysctl/sysctl:
sysctl: drop now unnecessary out-of-bounds check
sysctl: move sysctl type to ctl_table_header
sysctl: drop sysctl_is_perm_empty_ctl_table
sysctl: treewide: constify argument ctl_table_root::permissions(table)
sysctl: treewide: drop unused argument ctl_table_root::set_ownership(table)
bpf: Remove the now superfluous sentinel elements from ctl_table array
delayacct: Remove the now superfluous sentinel elements from ctl_table array
kprobes: Remove the now superfluous sentinel elements from ctl_table array
printk: Remove the now superfluous sentinel elements from ctl_table array
scheduler: Remove the now superfluous sentinel elements from ctl_table array
seccomp: Remove the now superfluous sentinel elements from ctl_table array
timekeeping: Remove the now superfluous sentinel elements from ctl_table array
ftrace: Remove the now superfluous sentinel elements from ctl_table array
umh: Remove the now superfluous sentinel elements from ctl_table array
kernel misc: Remove the now superfluous sentinel elements from ctl_table array
In the cgroup v2 CPU subsystem, assuming we have a
cgroup named 'test', and we set cpu.max and cpu.max.burst:
# echo 1000000 > /sys/fs/cgroup/test/cpu.max
# echo 1000000 > /sys/fs/cgroup/test/cpu.max.burst
then we check cpu.max and cpu.max.burst:
# cat /sys/fs/cgroup/test/cpu.max
1000000 100000
# cat /sys/fs/cgroup/test/cpu.max.burst
1000000
Next we set cpu.max again and check cpu.max and
cpu.max.burst:
# echo 2000000 > /sys/fs/cgroup/test/cpu.max
# cat /sys/fs/cgroup/test/cpu.max
2000000 100000
# cat /sys/fs/cgroup/test/cpu.max.burst
1000
... we find that the cpu.max.burst value changed unexpectedly.
In cpu_max_write(), the unit of the burst value returned
by tg_get_cfs_burst() is microseconds, while in cpu_max_write(),
the burst unit used for calculation should be nanoseconds,
which leads to the bug.
To fix it, get the burst value directly from tg->cfs_bandwidth.burst.
Fixes: f4183717b3 ("sched/fair: Introduce the burstable CFS controller")
Reported-by: Qixin Liao <liaoqixin@huawei.com>
Signed-off-by: Cheng Yu <serein.chengyu@huawei.com>
Signed-off-by: Zhang Qiao <zhangqiao22@huawei.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/20240424132438.514720-1-serein.chengyu@huawei.com
On 05/03/2024 15:05, Vincent Guittot wrote:
I'm fine with either and that was my first thought here, too, but it did seem like
the comment was mostly placed there to justify the 'unexpected' high utilization
when explicitly passing FREQUENCY_UTIL and the need to clamp it then.
So removing did feel slightly more natural to me anyway.
So alternatively:
From: Christian Loehle <christian.loehle@arm.com>
Date: Tue, 5 Mar 2024 09:34:41 +0000
Subject: [PATCH] sched/fair: Remove stale FREQUENCY_UTIL mention
effective_cpu_util() flags were removed, so remove mentioning of the
flag.
commit 9c0b4bb7f6 ("sched/cpufreq: Rework schedutil governor performance estimation")
reworked effective_cpu_util() removing enum cpu_util_type. Modify the
comment accordingly.
Signed-off-by: Christian Loehle <christian.loehle@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lore.kernel.org/r/0e2833ee-0939-44e0-82a2-520a585a0153@arm.com
Change se->load.weight to se_weight(se) in the calculation for the
initial util_avg to avoid unnecessarily inflating the util_avg by 1024
times.
The reason is that se->load.weight has the unit/scale as the scaled-up
load, while cfs_rg->avg.load_avg has the unit/scale as the true task
weight (as mapped directly from the task's nice/priority value). With
CONFIG_32BIT, the scaled-up load is equal to the true task weight. With
CONFIG_64BIT, the scaled-up load is 1024 times the true task weight.
Thus, the current code may inflate the util_avg by 1024 times. The
follow-up capping will not allow the util_avg value to go wild. But the
calculation should have the correct logic.
Signed-off-by: Dawei Li <daweilics@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Vishal Chourasia <vishalc@linux.ibm.com>
Link: https://lore.kernel.org/r/20240315015916.21545-1-daweilics@gmail.com
Change relax_domain_level checks so that it would be possible
to include or exclude all domains from newidle balancing.
This matches the behavior described in the documentation:
-1 no request. use system default or follow request of others.
0 no search.
1 search siblings (hyperthreads in a core).
"2" enables levels 0 and 1, level_max excludes the last (level_max)
level, and level_max+1 includes all levels.
Fixes: 1d3504fcf5 ("sched, cpuset: customize sched domains, core")
Signed-off-by: Vitalii Bursov <vitaly@bursov.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Link: https://lore.kernel.org/r/bd6de28e80073c79466ec6401cdeae78f0d4423d.1714488502.git.vitaly@bursov.com
Pull scheduler updates from Ingo Molnar:
- Add cpufreq pressure feedback for the scheduler
- Rework misfit load-balancing wrt affinity restrictions
- Clean up and simplify the code around ::overutilized and
::overload access.
- Simplify sched_balance_newidle()
- Bump SCHEDSTAT_VERSION to 16 due to a cleanup of CPU_MAX_IDLE_TYPES
handling that changed the output.
- Rework & clean up <asm/vtime.h> interactions wrt arch_vtime_task_switch()
- Reorganize, clean up and unify most of the higher level
scheduler balancing function names around the sched_balance_*()
prefix
- Simplify the balancing flag code (sched_balance_running)
- Miscellaneous cleanups & fixes
* tag 'sched-core-2024-05-13' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (50 commits)
sched/pelt: Remove shift of thermal clock
sched/cpufreq: Rename arch_update_thermal_pressure() => arch_update_hw_pressure()
thermal/cpufreq: Remove arch_update_thermal_pressure()
sched/cpufreq: Take cpufreq feedback into account
cpufreq: Add a cpufreq pressure feedback for the scheduler
sched/fair: Fix update of rd->sg_overutilized
sched/vtime: Do not include <asm/vtime.h> header
s390/irq,nmi: Include <asm/vtime.h> header directly
s390/vtime: Remove unused __ARCH_HAS_VTIME_TASK_SWITCH leftover
sched/vtime: Get rid of generic vtime_task_switch() implementation
sched/vtime: Remove confusing arch_vtime_task_switch() declaration
sched/balancing: Simplify the sg_status bitmask and use separate ->overloaded and ->overutilized flags
sched/fair: Rename set_rd_overutilized_status() to set_rd_overutilized()
sched/fair: Rename SG_OVERLOAD to SG_OVERLOADED
sched/fair: Rename {set|get}_rd_overload() to {set|get}_rd_overloaded()
sched/fair: Rename root_domain::overload to ::overloaded
sched/fair: Use helper functions to access root_domain::overload
sched/fair: Check root_domain::overload value before update
sched/fair: Combine EAS check with root_domain::overutilized access
sched/fair: Simplify the continue_balancing logic in sched_balance_newidle()
...
Optimize topology_span_sane() by removing duplicate comparisons.
Since topology_span_sane() is called inside of for_each_cpu(), each
previous CPU has already been compared against every other CPU. The
current CPU only needs to be compared against higher-numbered CPUs.
The total number of comparisons is reduced from N * (N - 1) to
N * (N - 1) / 2 on each non-NUMA scheduling domain level.
Signed-off-by: Kyle Meyer <kyle.meyer@hpe.com>
Reviewed-by: Yury Norov <yury.norov@gmail.com>
Acked-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
housekeeping_setup() checks cpumask_intersects(present, online) to ensure
that the kernel will have at least one housekeeping CPU after smp_init(),
but this doesn't work if the maxcpus= kernel parameter limits the number of
processors available after bootup.
For example, a kernel with "maxcpus=2 nohz_full=0-2" parameters crashes at
boot time on a virtual machine with 4 CPUs.
Change housekeeping_setup() to use cpumask_first_and() and check that the
returned CPU number is valid and less than setup_max_cpus.
Another corner case is "nohz_full=0" on a machine with a single CPU or with
the maxcpus=1 kernel argument. In this case non_housekeeping_mask is empty
and tick_nohz_full_setup() makes no sense. And indeed, the kernel hits the
WARN_ON(tick_nohz_full_running) in tick_sched_do_timer().
And how should the kernel interpret the "nohz_full=" parameter? It should
be silently ignored, but currently cpulist_parse() happily returns the
empty cpumask and this leads to the same problem.
Change housekeeping_setup() to check cpumask_empty(non_housekeeping_mask)
and do nothing in this case.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Phil Auld <pauld@redhat.com>
Acked-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/20240413141746.GA10008@redhat.com
Documentation/timers/no_hz.rst states that the "nohz_full=" mask must not
include the boot CPU, which is no longer true after:
08ae95f4fd ("nohz_full: Allow the boot CPU to be nohz_full").
However after:
aae17ebb53 ("workqueue: Avoid using isolated cpus' timers on queue_delayed_work")
the kernel will crash at boot time in this case; housekeeping_any_cpu()
returns an invalid CPU number until smp_init() brings the first
housekeeping CPU up.
Change housekeeping_any_cpu() to check the result of cpumask_any_and() and
return smp_processor_id() in this case.
This is just the simple and backportable workaround which fixes the
symptom, but smp_processor_id() at boot time should be safe at least for
type == HK_TYPE_TIMER, this more or less matches the tick_do_timer_boot_cpu
logic.
There is no worry about cpu_down(); tick_nohz_cpu_down() will not allow to
offline tick_do_timer_cpu (the 1st online housekeeping CPU).
Fixes: aae17ebb53 ("workqueue: Avoid using isolated cpus' timers on queue_delayed_work")
Reported-by: Chris von Recklinghausen <crecklin@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Phil Auld <pauld@redhat.com>
Acked-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/20240411143905.GA19288@redhat.com
Closes: https://lore.kernel.org/all/20240402105847.GA24832@redhat.com/
