handle_lockup() currently calls scx_verror() but ignores its return value,
always returning true when the scheduler is enabled. Make it capture and return
the result from scx_verror(). This prepares for hardlockup handling.
Cc: Dan Schatzberg <schatzberg.dan@gmail.com>
Cc: Emil Tsalapatis <etsal@meta.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
scx_rcu_cpu_stall() and scx_softlockup() share the same pattern: check if the
scheduler is enabled under RCU read lock and trigger an error if so. Extract
the common pattern into handle_lockup() helper. Add scx_verror() macro and use
guard(rcu)().
This simplifies both handlers, reduces code duplication, and prepares for
hardlockup handling.
Reviewed-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Cc: Emil Tsalapatis <etsal@meta.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Make scx_exit() and scx_vexit() return bool indicating whether the calling
thread successfully claimed the exit. This will be used by the abort mechanism
added in a later patch.
Reviewed-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Cc: Emil Tsalapatis <etsal@meta.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
62dcbab8b0 ("sched_ext: Avoid live-locking bypass mode switching") introduced
the breather mechanism to inject delays during bypass mode switching. It
maintains operation semantics unchanged while reducing lock contention to avoid
live-locks on large NUMA systems.
However, the breather only activates when exiting the scheduler, so there's no
need to maintain operation semantics. Simplify by exiting dispatch and move
operations immediately when scx_aborting is set. In consume_dispatch_q(), break
out of the task iteration loop. In scx_dsq_move(), return early before
acquiring locks.
This also fixes cases the breather mechanism cannot handle. When a large system
has many runnable threads affinitized to different CPU subsets and the BPF
scheduler places them all into a single DSQ, many CPUs can scan the DSQ
concurrently for tasks they can run. This can cause DSQ and RQ locks to be held
for extended periods, leading to various failure modes. The breather cannot
solve this because once in the consume loop, there's no exit. The new mechanism
fixes this by exiting the loop immediately.
The bypass DSQ is exempted to ensure the bypass mechanism itself can make
progress.
v2: Use READ_ONCE() when reading scx_aborting (Andrea Righi).
Reported-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Reviewed-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Cc: Andrea Righi <arighi@nvidia.com>
Cc: Emil Tsalapatis <etsal@meta.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The breather mechanism was introduced in 62dcbab8b0 ("sched_ext: Avoid
live-locking bypass mode switching") and e32c260195 ("sched_ext: Enable the
ops breather and eject BPF scheduler on softlockup") to prevent live-locks by
injecting delays when CPUs are trapped in dispatch paths.
Currently, it uses scx_breather_depth (atomic_t) and scx_in_softlockup
(unsigned long) with separate increment/decrement and cleanup operations. The
breather is only activated when aborting, so tie it directly to the exit
mechanism. Replace both variables with scx_aborting flag set when exit is
claimed and cleared after bypass is enabled. Introduce scx_claim_exit() to
consolidate exit_kind claiming and breather enablement. This eliminates
scx_clear_softlockup() and simplifies scx_softlockup() and scx_bypass().
The breather mechanism will be replaced by a different abort mechanism in a
future patch. This simplification prepares for that change.
Reviewed-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Acked-by: Andrea Righi <arighi@nvidia.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Bypass mode routes tasks through fallback dispatch queues. Originally a single
global DSQ, b7b3b2dbae ("sched_ext: Split the global DSQ per NUMA node")
changed this to per-node DSQs to resolve NUMA-related livelocks.
Dan Schatzberg found per-node DSQs can still livelock when many threads are
pinned to different small CPU subsets: each CPU must scan many incompatible
tasks to find runnable ones, causing severe contention with high CPU counts.
Switch to per-CPU bypass DSQs. Each task queues on its current CPU. Default
idle CPU selection and direct dispatch handle most cases well.
This introduces a failure mode when tasks concentrate on one CPU in
over-saturated systems. If the BPF scheduler severely skews placement before
triggering bypass, that CPU's queue may be too long to drain, causing RCU
stalls. A load balancer in a future patch will address this. The bypass DSQ is
separate from local DSQ to enable load balancing: local DSQs use rq locks,
preventing efficient scanning and transfer across CPUs, especially problematic
when systems are already contended.
v2: Clarified why bypass DSQ is separate from local DSQ (Andrea Righi).
Reported-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Reviewed-by: Dan Schatzberg <schatzberg.dan@gmail.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The local and global DSQ enqueue paths in do_enqueue_task() share the same
slice refill logic. Factor out the common code into a shared enqueue label.
This makes adding new enqueue cases easier. No functional changes.
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
There have been reported cases of bypass mode not making forward progress fast
enough. The 20ms default slice is unnecessarily long for bypass mode where the
primary goal is ensuring all tasks can make forward progress.
Introduce SCX_SLICE_BYPASS set to 5ms and make the scheduler automatically
switch to it when entering bypass mode. Also make the bypass slice value
tunable through the slice_bypass_us module parameter (adjustable between 100us
and 100ms) to make it easier to test whether slice durations are a factor in
problem cases.
v3: Use READ_ONCE/WRITE_ONCE for scx_slice_dfl access (Dan).
v2: Removed slice_dfl_us module parameter. Fixed typos (Andrea).
Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Reviewed-by: Andrea Righi <arighi@nvidia.com>
Cc: Dan Schatzberg <schatzberg.dan@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Currently the function tracer's options are saved via a global mask when
it should be per instance. Use the new infrastructure to define a
"default_flags" field in the tracer structure that is used for the top
level instance as well as new ones.
Currently the global mask causes confusion:
# cd /sys/kernel/tracing
# mkdir instances/foo
# echo function > instances/foo/current_tracer
# echo 1 > options/func-args
# echo function > current_tracer
# cat trace
[..]
<idle>-0 [005] d..3. 1050.656187: rcu_needs_cpu() <-tick_nohz_next_event
<idle>-0 [005] d..3. 1050.656188: get_next_timer_interrupt(basej=0x10002dbad, basem=0xf45fd7d300) <-tick_nohz_next_event
<idle>-0 [005] d..3. 1050.656189: _raw_spin_lock(lock=0xffff8944bdf5de80) <-__get_next_timer_interrupt
<idle>-0 [005] d..4. 1050.656190: do_raw_spin_lock(lock=0xffff8944bdf5de80) <-__get_next_timer_interrupt
<idle>-0 [005] d..4. 1050.656191: _raw_spin_lock_nested(lock=0xffff8944bdf5f140, subclass=1) <-__get_next_timer_interrupt
# cat instances/foo/options/func-args
1
# cat instances/foo/trace
[..]
kworker/4:1-88 [004] ...1. 298.127735: next_zone <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127736: first_online_pgdat <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127738: next_online_pgdat <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127739: fold_diff <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127741: round_jiffies_relative <-vmstat_update
[..]
The above shows that updating the "func-args" option at the top level
instance also updates the "func-args" option in the instance but because
the update is only done by the instance that gets changed (as it should),
it's confusing to see that the option is already set in the other instance.
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: https://patch.msgid.link/20251111232429.470883736@kernel.org
Fixes: f20a580627 ("ftrace: Allow instances to use function tracing")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Tracers can add specify options to modify them. This logic was added
before instances were created and the tracer flags were global variables.
After instances were created where a tracer may exist in more than one
instance, the flags were not updated from being global into instance
specific. This causes confusion with these options. For example, the
function tracer has an option to enable function arguments:
# cd /sys/kernel/tracing
# mkdir instances/foo
# echo function > instances/foo/current_tracer
# echo 1 > options/func-args
# echo function > current_tracer
# cat trace
[..]
<idle>-0 [005] d..3. 1050.656187: rcu_needs_cpu() <-tick_nohz_next_event
<idle>-0 [005] d..3. 1050.656188: get_next_timer_interrupt(basej=0x10002dbad, basem=0xf45fd7d300) <-tick_nohz_next_event
<idle>-0 [005] d..3. 1050.656189: _raw_spin_lock(lock=0xffff8944bdf5de80) <-__get_next_timer_interrupt
<idle>-0 [005] d..4. 1050.656190: do_raw_spin_lock(lock=0xffff8944bdf5de80) <-__get_next_timer_interrupt
<idle>-0 [005] d..4. 1050.656191: _raw_spin_lock_nested(lock=0xffff8944bdf5f140, subclass=1) <-__get_next_timer_interrupt
# cat instances/foo/options/func-args
1
# cat instances/foo/trace
[..]
kworker/4:1-88 [004] ...1. 298.127735: next_zone <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127736: first_online_pgdat <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127738: next_online_pgdat <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127739: fold_diff <-refresh_cpu_vm_stats
kworker/4:1-88 [004] ...1. 298.127741: round_jiffies_relative <-vmstat_update
[..]
The above shows that setting "func-args" in the top level instance also
set it in the instance "foo", but since the interface of the trace flags
are per instance, the update didn't take affect in the "foo" instance.
Update the infrastructure to allow tracers to add a "default_flags" field
in the tracer structure that can be set instead of "flags" which will make
the flags per instance. If a tracer needs to keep the flags global (like
blktrace), keeping the "flags" field set will keep the old behavior.
This does not update function or the function graph tracers. That will be
handled later.
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: https://patch.msgid.link/20251111232429.305317942@kernel.org
Fixes: f20a580627 ("ftrace: Allow instances to use function tracing")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Previously, update_cpumasks_hier() used need_rebuild_sched_domains to
decide whether to invoke rebuild_sched_domains_locked(). Now that
rebuild_sched_domains_locked() only sets force_rebuild, the flag is
redundant. Hence, remove it.
Signed-off-by: Chen Ridong <chenridong@huawei.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The remote_children list is used to track all remote partitions attached
to a cpuset. However, it serves no other purpose. Using a boolean flag to
indicate whether a cpuset is a remote partition is a more direct approach,
making remote_children unnecessary.
This patch replaces the list with a remote_partition flag in the cpuset
structure and removes remote_children entirely.
Signed-off-by: Chen Ridong <chenridong@huawei.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
There is no need to jump to the 'done' label upon failure, as no cleanup
is required. Return the error code directly instead.
Signed-off-by: Chen Ridong <chenridong@huawei.com>
Reviewed-by: Waiman Long <longman@redhat.com>
Reviewed-by: Michal Koutný <mkoutny@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
To compile cgroup.c with PREEMPT_RT=y include header which declares
struct irq_work.
Fixes: 9311e6c29b ("cgroup: Fix sleeping from invalid context warning on PREEMPT_RT")
Signed-off-by: Bert Karwatzki <spasswolf@web.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
cpumask_subset(a,b) -> cpumask_weight(a) should be same as cpumask_weight_and(a,b)
for_each_cpu_and(a,b) to count cpus could be replaced by cpumask_weight_and(a,b)
No Functional Change. It could save a few cycles since cpumask_weight_and
would be more efficient. Plus one less stack variable.
Signed-off-by: Shrikanth Hegde <sshegde@linux.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://patch.msgid.link/20251014100342.978936-3-sshegde@linux.ibm.com
For now, we will use ftrace for the fprobe if fp->exit_handler not exists
and CONFIG_DYNAMIC_FTRACE_WITH_REGS is enabled.
However, CONFIG_DYNAMIC_FTRACE_WITH_REGS is not supported by some arch,
such as arm. What we need in the fprobe is the function arguments, so we
can use ftrace for fprobe if CONFIG_DYNAMIC_FTRACE_WITH_ARGS is enabled.
Therefore, use ftrace if CONFIG_DYNAMIC_FTRACE_WITH_REGS or
CONFIG_DYNAMIC_FTRACE_WITH_ARGS enabled.
Link: https://lore.kernel.org/all/20251103063434.47388-1-dongml2@chinatelecom.cn/
Signed-off-by: Menglong Dong <dongml2@chinatelecom.cn>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
For now, fgraph is used for the fprobe, even if we need trace the entry
only. However, the performance of ftrace is better than fgraph, and we
can use ftrace_ops for this case.
Then performance of kprobe-multi increases from 54M to 69M. Before this
commit:
$ ./benchs/run_bench_trigger.sh kprobe-multi
kprobe-multi : 54.663 ± 0.493M/s
After this commit:
$ ./benchs/run_bench_trigger.sh kprobe-multi
kprobe-multi : 69.447 ± 0.143M/s
Mitigation is disable during the bench testing above.
Link: https://lore.kernel.org/all/20251015083238.2374294-2-dongml2@chinatelecom.cn/
Signed-off-by: Menglong Dong <dongml2@chinatelecom.cn>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Reactors can be called from any context through tracepoints.
When developing reactors care needs to be taken to only call APIs which
are safe. As the tracepoints used during testing may not actually be
called from restrictive contexts lockdep may not be helpful.
Add explicit overrides to help lockdep find invalid code patterns.
The usage of LD_WAIT_FREE will trigger lockdep warnings in the panic
reactor. These are indeed valid warnings but they are out of scope for
RV and will instead be fixed by the printk subsystem.
Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Acked-by: Nam Cao <namcao@linutronix.de>
Link: https://lore.kernel.org/r/20251014-rv-lockdep-v1-3-0b9e51919ea8@linutronix.de
Signed-off-by: Gabriele Monaco <gmonaco@redhat.com>
Gabriel reported that the dl_server doesn't stop as expected.
The problem was found to be the fact that idle time and fair runtime are
treated equally. Both will count towards dl_server runtime and push the
activation forwards when it is in the zero-laxity wait state.
Notably:
dl_server_update_idle()
update_curr_dl_se()
if (dl_defer && dl_throttled && dl_runtime_exceeded())
hrtimer_try_to_cancel(); // stop timer
replenish_dl_new_period()
deadline = now + dl_deadline; // fwd period
runtime = dl_runtime;
start_dl_timer(); // restart timer
And while we do want idle time accounted towards the *current* activation of
the dl_server -- after all, a fair task could've ran if we had any -- we don't
necessarily want idle time to cause or push forward an activation.
Introduce dl_defer_idle to make this distinction. It will be set once idle time
pushed the activation forward, once set idle time will only be allowed to
consume any runtime but not push the activation. This will then cause
dl_server_timer() to fire, which will stop the dl_server.
Any non-idle time accounting during this phase will clear dl_defer_idle, so
only a full period of idle will cause the dl_server to stop.
Reported-by: Gabriele Monaco <gmonaco@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20251101000057.GA2184199@noisy.programming.kicks-ass.net
The dl_server time accounting code is a little odd. The normal scheduler
pattern is to update curr before doing something, such that the old state is
fully accounted before changing state.
Notably, the dl_server_timer() needs to propagate the current time accounting
since the current task could be ran by dl_server and thus this can affect
dl_se->runtime. Similarly for dl_server_start().
And since the (deferred) dl_server wants idle time accounted, rework
sched_idle_class time accounting to be more like all the others.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20251020141130.GJ3245006@noisy.programming.kicks-ass.net
Since commit d4c64207b8 ("sched: Cleanup the sched_change NOCLOCK usage"),
update_rq_clock() is called in do_set_cpus_allowed() -> sched_change_begin()
to update the rq clock. This results in a duplicate call update_rq_clock()
in __set_cpus_allowed_ptr_locked().
While holding the rq lock and before calling do_set_cpus_allowed(),
there is nothing that depends on an updated rq_clock.
Therefore, remove the redundant update_rq_clock() in
__set_cpus_allowed_ptr_locked() to avoid the warning about double
rq clock updates.
Fixes: d4c64207b8 ("sched: Cleanup the sched_change NOCLOCK usage")
Signed-off-by: Hao Jia <jiahao1@lixiang.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: K Prateek Nayak <kprateek.nayak@amd.com>
Link: https://patch.msgid.link/20251029093655.31252-1-jiahao.kernel@gmail.com
Basically, from the constraint that the sum of lag is zero, you can
infer that the 0-lag point is the weighted average of the individual
vruntime, which is what we're trying to compute:
\Sum w_i * v_i
avg = --------------
\Sum w_i
Now, since vruntime takes the whole u64 (worse, it wraps), this
multiplication term in the numerator is not something we can compute;
instead we do the min_vruntime (v0 henceforth) thing like:
v_i = (v_i - v0) + v0
This does two things:
- it keeps the key: (v_i - v0) 'small';
- it creates a relative 0-point in the modular space.
If you do that subtitution and work it all out, you end up with:
\Sum w_i * (v_i - v0)
avg = --------------------- + v0
\Sum w_i
Since you cannot very well track a ratio like that (and not suffer
terrible numerical problems) we simpy track the numerator and
denominator individually and only perform the division when strictly
needed.
Notably, the numerator lives in cfs_rq->avg_vruntime and the denominator
lives in cfs_rq->avg_load.
The one extra 'funny' is that these numbers track the entities in the
tree, and current is typically outside of the tree, so avg_vruntime()
adds current when needed before doing the division.
(vruntime_eligible() elides the division by cross-wise multiplication)
Anyway, as mentioned above, we currently use the CFS era min_vruntime
for this purpose. However, this thing can only move forward, while the
above avg can in fact move backward (when a non-eligible task leaves,
the average becomes smaller), this can cause trouble when through
happenstance (or construction) these values drift far enough apart to
wreck the game.
Replace cfs_rq::min_vruntime with cfs_rq::zero_vruntime which is kept
near/at avg_vruntime, following its motion.
The down-side is that this requires computing the avg more often.
Fixes: 147f3efaa2 ("sched/fair: Implement an EEVDF-like scheduling policy")
Reported-by: Zicheng Qu <quzicheng@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://patch.msgid.link/20251106111741.GC4068168@noisy.programming.kicks-ass.net
Cc: stable@vger.kernel.org
Bring in the shared branch with the kbuild tree to enable
'-fms-extensions' for 6.19. Further namespace cleanup work
requires this extension.
Signed-off-by: Christian Brauner <brauner@kernel.org>
Exports three necessary symbols for implementing struct_ops with
tristate subsystem.
To hold or release refcnt of struct_ops refcnt by inline funcs
bpf_try_module_get and bpf_module_put which use bpf_struct_ops_get(put)
conditionally.
And to copy obj name from one to the other with effective checks by
bpf_obj_name_cpy.
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://patch.msgid.link/20251107035632.115950-2-alibuda@linux.alibaba.com
Currently bpf_dynptr_from_skb_meta() marks the dynptr as read-only when
the skb is cloned, preventing writes to metadata.
Remove this restriction and unclone the skb head on bpf_dynptr_write() to
metadata, now that the metadata is preserved during uncloning. This makes
metadata dynptr consistent with skb dynptr, allowing writes regardless of
whether the skb is cloned.
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://patch.msgid.link/20251105-skb-meta-rx-path-v4-3-5ceb08a9b37b@cloudflare.com
The decision whether some more space is needed is tricky in the printk
ring buffer code:
1. The given lpos values might overflow. A subtraction must be used
instead of a simple "lower than" check.
2. Another CPU might reuse the space in the mean time. It can be
detected when the subtraction is bigger than DATA_SIZE(data_ring).
3. There is exactly enough space when the result of the subtraction
is zero. But more space is needed when the result is exactly
DATA_SIZE(data_ring).
Add a helper function to make sure that the check is done correctly
in all situations. Also it helps to make the code consistent and
better documented.
Suggested-by: John Ogness <john.ogness@linutronix.de>
Link: https://lore.kernel.org/r/87tsz7iea2.fsf@jogness.linutronix.de
Reviewed-by: John Ogness <john.ogness@linutronix.de>
Link: https://patch.msgid.link/20251107194720.1231457-3-pmladek@suse.com
[pmladek@suse.com: Updated wording as suggested by John]
Signed-off-by: Petr Mladek <pmladek@suse.com>
