The following commit:
c89970202a ("cputime: remove cputime_to_nsecs fallback")
Removed an <asm/cputime.h> inclusion from <linux/sched/cputime.h>, but this
broke the IA64 build:
arch/ia64/kernel/time.c:110:6: warning: no previous prototype for 'arch_vtime_task_switch' [-Wmissing-prototypes]
Add in the missing <asm/cputime.h> header to fix it.
Fixes: c89970202a ("cputime: remove cputime_to_nsecs fallback")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The max CPU capacity is the same for all CPUs sharing frequency domain.
There is a way to avoid heavy operations in a loop for each CPU by
leveraging this knowledge. Thus, simplify the looping code in the
sugov_next_freq_shared() and drop heavy multiplications. Instead, use
simple max() to get the highest utilization from these CPUs.
This is useful for platforms with many (4 or 6) little CPUs. We avoid
heavy 2*PD_CPU_NUM multiplications in that loop, which is called billions
of times, since it's not limited by the schedutil time delta filter in
sugov_should_update_freq(). When there was no need to change frequency
the code bailed out, not updating the sg_policy::last_freq_update_time.
Then every visit after delta_ns time longer than the
sg_policy::freq_update_delay_ns goes through and triggers the next
frequency calculation code. Although, if the next frequency, as outcome
of that, would be the same as current frequency, we won't update the
sg_policy::last_freq_update_time and the story will be repeated (in
a very short period, sometimes a few microseconds).
The max CPU capacity must be fetched every time we are called, due to
difficulties during the policy setup, where we are not able to get the
normalized CPU capacity at the right time.
The fetched CPU capacity value is than used in sugov_iowait_apply() to
calculate the right boost. This required a few changes in the local
functions and arguments. The capacity value should hopefully be fetched
once when needed and then passed over CPU registers to those functions.
Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20221208160256.859-2-lukasz.luba@arm.com
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
ttwu_do_activate() is used for a complete wakeup, in which we will
activate_task() and use ttwu_do_wakeup() to mark the task runnable
and perform wakeup-preemption, also call class->task_woken() callback
and update the rq->idle_stamp.
Since ttwu_runnable() is not a complete wakeup, don't need all those
done in ttwu_do_wakeup(), so we can move those to ttwu_do_activate()
to simplify ttwu_do_wakeup(), making it only mark the task runnable
to be reused in ttwu_runnable() and try_to_wake_up().
This patch should not have any functional changes.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20221223103257.4962-2-zhouchengming@bytedance.com
ttwu_runnable() is used as a fast wakeup path when the wakee task
is running on CPU or runnable on RQ, in both cases we can just
set its state to TASK_RUNNING to prevent a sleep.
If the wakee task is on_cpu running, we don't need to update_rq_clock()
or check_preempt_curr().
But if the wakee task is on_rq && !on_cpu (e.g. an IRQ hit before
the task got to schedule() and the task been preempted), we should
check_preempt_curr() to see if it can preempt the current running.
This also removes the class->task_woken() callback from ttwu_runnable(),
which wasn't required per the RT/DL implementations: any required push
operation would have been queued during class->set_next_task() when p
got preempted.
ttwu_runnable() also loses the update to rq->idle_stamp, as by definition
the rq cannot be idle in this scenario.
Suggested-by: Valentin Schneider <vschneid@redhat.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20221223103257.4962-1-zhouchengming@bytedance.com
When select_idle_capacity() starts scanning for an idle CPU, it starts
with target CPU that has already been checked in select_idle_sibling().
So we start checking from the next CPU and try the target CPU at the end.
Similarly for task_numa_assign(), we have just checked numa_migrate_on
of dst_cpu, so start from the next CPU. This also works for
steal_cookie_task(), the first scan must fail and start directly
from the next one.
Signed-off-by: Hao Jia <jiahao.os@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lore.kernel.org/r/20221216062406.7812-3-jiahao.os@bytedance.com
In update_numa_stats() we try to find an idle cpu on the NUMA node,
preferably an idle core. we can stop looking for the next idle core
or idle cpu after finding an idle core. But we can't stop the
whole loop of scanning the CPU, because we need to calculate
approximate NUMA stats at a point in time. For example,
the src and dst nr_running is needed by task_numa_find_cpu().
Signed-off-by: Hao Jia <jiahao.os@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lore.kernel.org/r/20221216062406.7812-2-jiahao.os@bytedance.com
With a modified container_of() that preserves constness, the compiler
finds some pointers which should have been marked as const. task_of()
also needs to become const-preserving for the !FAIR_GROUP_SCHED case so
that cfs_rq_of() can take a const argument. No change to generated code.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20221212144946.2657785-1-willy@infradead.org
Introduce a rseq-x86-bits.h template header which is internally included
to generate the static inline functions covering:
- relaxed and release memory ordering,
- per-cpu-id and per-mm-cid per-cpu data access.
This introduces changes to the rseq.h selftests API which require to
update the rseq selftest programs. Similar API/templating changes need
to be done for other architectures.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221122203932.231377-12-mathieu.desnoyers@efficios.com
This feature allows the scheduler to expose a per-memory map concurrency
ID to user-space. This concurrency ID is within the possible cpus range,
and is temporarily (and uniquely) assigned while threads are actively
running within a memory map. If a memory map has fewer threads than
cores, or is limited to run on few cores concurrently through sched
affinity or cgroup cpusets, the concurrency IDs will be values close
to 0, thus allowing efficient use of user-space memory for per-cpu
data structures.
This feature is meant to be exposed by a new rseq thread area field.
The primary purpose of this feature is to do the heavy-lifting needed
by memory allocators to allow them to use per-cpu data structures
efficiently in the following situations:
- Single-threaded applications,
- Multi-threaded applications on large systems (many cores) with limited
cpu affinity mask,
- Multi-threaded applications on large systems (many cores) with
restricted cgroup cpuset per container.
One of the key concern from scheduler maintainers is the overhead
associated with additional spin locks or atomic operations in the
scheduler fast-path. This is why the following optimization is
implemented.
On context switch between threads belonging to the same memory map,
transfer the mm_cid from prev to next without any atomic ops. This
takes care of use-cases involving frequent context switch between
threads belonging to the same memory map.
Additional optimizations can be done if the spin locks added when
context switching between threads belonging to different memory maps end
up being a performance bottleneck. Those are left out of this patch
though. A performance impact would have to be clearly demonstrated to
justify the added complexity.
The credit goes to Paul Turner (Google) for the original virtual cpu id
idea. This feature is implemented based on the discussions with Paul
Turner and Peter Oskolkov (Google), but I took the liberty to implement
scheduler fast-path optimizations and my own NUMA-awareness scheme. The
rumor has it that Google have been running a rseq vcpu_id extension
internally in production for a year. The tcmalloc source code indeed has
comments hinting at a vcpu_id prototype extension to the rseq system
call [1].
The following benchmarks do not show any significant overhead added to
the scheduler context switch by this feature:
* perf bench sched messaging (process)
Baseline: 86.5±0.3 ms
With mm_cid: 86.7±2.6 ms
* perf bench sched messaging (threaded)
Baseline: 84.3±3.0 ms
With mm_cid: 84.7±2.6 ms
* hackbench (process)
Baseline: 82.9±2.7 ms
With mm_cid: 82.9±2.9 ms
* hackbench (threaded)
Baseline: 85.2±2.6 ms
With mm_cid: 84.4±2.9 ms
[1] https://github.com/google/tcmalloc/blob/master/tcmalloc/internal/linux_syscall_support.h#L26
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221122203932.231377-8-mathieu.desnoyers@efficios.com
Adding the NUMA node id to struct rseq is a straightforward thing to do,
and a good way to figure out if anything in the user-space ecosystem
prevents extending struct rseq.
This NUMA node id field allows memory allocators such as tcmalloc to
take advantage of fast access to the current NUMA node id to perform
NUMA-aware memory allocation.
It can also be useful for implementing fast-paths for NUMA-aware
user-space mutexes.
It also allows implementing getcpu(2) purely in user-space.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221122203932.231377-5-mathieu.desnoyers@efficios.com
Introduce the extensible rseq ABI, where the feature size supported by
the kernel and the required alignment are communicated to user-space
through ELF auxiliary vectors.
This allows user-space to call rseq registration with a rseq_len of
either 32 bytes for the original struct rseq size (which includes
padding), or larger.
If rseq_len is larger than 32 bytes, then it must be large enough to
contain the feature size communicated to user-space through ELF
auxiliary vectors.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221122203932.231377-4-mathieu.desnoyers@efficios.com
When linking the selftests against a libc which does not handle rseq
registration (before 2.35), rseq thread registration silently succeed
even with CONFIG_RSEQ=n because it erroneously thinks that libc is
handling rseq registration.
This is caused by setting the rseq ownership flag only after the
rseq_available() check. It should rather be set before the
rseq_available() check.
Set the rseq_size to 0 (error value) immediately after the
rseq_available() check fails rather than in the thread registration
functions.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221122203932.231377-2-mathieu.desnoyers@efficios.com
CFS bandwidth currently distributes new runtime and unthrottles cfs_rq's
inline in an hrtimer callback. Runtime distribution is a per-cpu
operation, and unthrottling is a per-cgroup operation, since a tg walk
is required. On machines with a large number of cpus and large cgroup
hierarchies, this cpus*cgroups work can be too much to do in a single
hrtimer callback: since IRQ are disabled, hard lockups may easily occur.
Specifically, we've found this scalability issue on configurations with
256 cpus, O(1000) cgroups in the hierarchy being throttled, and high
memory bandwidth usage.
To fix this, we can instead unthrottle cfs_rq's asynchronously via a
CSD. Each cpu is responsible for unthrottling itself, thus sharding the
total work more fairly across the system, and avoiding hard lockups.
Signed-off-by: Josh Don <joshdon@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20221117005418.3499691-1-joshdon@google.com
Due to several bugs caused by timers being re-armed after they are
shutdown and just before they are freed, a new state of timers was added
called "shutdown". After a timer is set to this state, then it can no
longer be re-armed.
The following script was run to find all the trivial locations where
del_timer() or del_timer_sync() is called in the same function that the
object holding the timer is freed. It also ignores any locations where
the timer->function is modified between the del_timer*() and the free(),
as that is not considered a "trivial" case.
This was created by using a coccinelle script and the following
commands:
$ cat timer.cocci
@@
expression ptr, slab;
identifier timer, rfield;
@@
(
- del_timer(&ptr->timer);
+ timer_shutdown(&ptr->timer);
|
- del_timer_sync(&ptr->timer);
+ timer_shutdown_sync(&ptr->timer);
)
... when strict
when != ptr->timer
(
kfree_rcu(ptr, rfield);
|
kmem_cache_free(slab, ptr);
|
kfree(ptr);
)
$ spatch timer.cocci . > /tmp/t.patch
$ patch -p1 < /tmp/t.patch
Link: https://lore.kernel.org/lkml/20221123201306.823305113@linutronix.de/
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Acked-by: Pavel Machek <pavel@ucw.cz> [ LED ]
Acked-by: Kalle Valo <kvalo@kernel.org> [ wireless ]
Acked-by: Paolo Abeni <pabeni@redhat.com> [ networking ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull spi fix from Mark Brown:
"One driver specific change here which handles the case where a SPI
device for some reason tries to change the bus speed during a message
on fsl_spi hardware, this should be very unusual"
* tag 'spi-fix-v6.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
spi: fsl_spi: Don't change speed while chipselect is active
Pull regulator fixes from Mark Brown:
"Two core fixes here, one for a long standing race which some Qualcomm
systems have started triggering with their UFS driver and another
fixing a problem with supply lookup introduced by the fixes for devm
related use after free issues that were introduced in this merge
window"
* tag 'regulator-fix-v6.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
regulator: core: fix deadlock on regulator enable
regulator: core: Fix resolve supply lookup issue
Pull coccicheck update from Julia Lawall:
"Modernize use of grep in coccicheck:
Use 'grep -E' instead of 'egrep'"
* tag 'coccinelle-6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/jlawall/linux:
scripts: coccicheck: use "grep -E" instead of "egrep"
Pull kernel hardening fixes from Kees Cook:
- Fix CFI failure with KASAN (Sami Tolvanen)
- Fix LKDTM + CFI under GCC 7 and 8 (Kristina Martsenko)
- Limit CONFIG_ZERO_CALL_USED_REGS to Clang > 15.0.6 (Nathan
Chancellor)
- Ignore "contents" argument in LoadPin's LSM hook handling
- Fix paste-o in /sys/kernel/warn_count API docs
- Use READ_ONCE() consistently for oops/warn limit reading
* tag 'hardening-v6.2-rc1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
cfi: Fix CFI failure with KASAN
exit: Use READ_ONCE() for all oops/warn limit reads
security: Restrict CONFIG_ZERO_CALL_USED_REGS to gcc or clang > 15.0.6
lkdtm: cfi: Make PAC test work with GCC 7 and 8
docs: Fix path paste-o for /sys/kernel/warn_count
LoadPin: Ignore the "contents" argument of the LSM hooks
