Reader optimistic spinning is helpful when the reader critical section
is short and there aren't that many readers around. It makes readers
relatively more preferred than writers. When a writer times out spinning
on a reader-owned lock and set the nospinnable bits, there are two main
reasons for that.
1) The reader critical section is long, perhaps the task sleeps after
acquiring the read lock.
2) There are just too many readers contending the lock causing it to
take a while to service all of them.
In the former case, long reader critical section will impede the progress
of writers which is usually more important for system performance.
In the later case, reader optimistic spinning tends to make the reader
groups that contain readers that acquire the lock together smaller
leading to more of them. That may hurt performance in some cases. In
other words, the setting of nonspinnable bits indicates that reader
optimistic spinning may not be helpful for those workloads that cause it.
Therefore, any writers that have observed the setting of the writer
nonspinnable bit for a given rwsem after they fail to acquire the lock
via optimistic spinning will set the reader nonspinnable bit once they
acquire the write lock. Similarly, readers that observe the setting
of reader nonspinnable bit at slowpath entry will also set the reader
nonspinnable bit when they acquire the read lock via the wakeup path.
Once the reader nonspinnable bit is on, it will only be reset when
a writer is able to acquire the rwsem in the fast path or somehow a
reader or writer in the slowpath doesn't observe the nonspinable bit.
This is to discourage reader optmistic spinning on that particular
rwsem and make writers more preferred. This adaptive disabling of reader
optimistic spinning will alleviate some of the negative side effect of
this feature.
In addition, this patch tries to make readers in the spinning queue
follow the phase-fair principle after quitting optimistic spinning
by checking if another reader has somehow acquired a read lock after
this reader enters the optimistic spinning queue. If so and the rwsem
is still reader-owned, this reader is in the right read-phase and can
attempt to acquire the lock.
On a 2-socket 40-core 80-thread Skylake system, the page_fault1 test of
the will-it-scale benchmark was run with various number of threads. The
number of operations done before reader optimistic spinning patches,
this patch and after this patch were:
Threads Before rspin Before patch After patch %change
------- ------------ ------------ ----------- -------
20 5541068 5345484 5455667 -3.5%/ +2.1%
40 10185150 7292313 9219276 -28.5%/+26.4%
60 8196733 6460517 7181209 -21.2%/+11.2%
80 9508864 6739559 8107025 -29.1%/+20.3%
This patch doesn't recover all the lost performance, but it is more
than half. Given the fact that reader optimistic spinning does benefit
some workloads, this is a good compromise.
Using the rwsem locking microbenchmark with very short critical section,
this patch doesn't have too much impact on locking performance as shown
by the locking rates (kops/s) below with equal numbers of readers and
writers before and after this patch:
# of Threads Pre-patch Post-patch
------------ --------- ----------
2 4,730 4,969
4 4,814 4,786
8 4,866 4,815
16 4,715 4,511
32 3,338 3,500
64 3,212 3,389
80 3,110 3,044
When running the locking microbenchmark with 40 dedicated reader and writer
threads, however, the reader performance is curtailed to favor the writer.
Before patch:
40 readers, Iterations Min/Mean/Max = 204,026/234,309/254,816
40 writers, Iterations Min/Mean/Max = 88,515/95,884/115,644
After patch:
40 readers, Iterations Min/Mean/Max = 33,813/35,260/36,791
40 writers, Iterations Min/Mean/Max = 95,368/96,565/97,798
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-16-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When the rwsem is owned by reader, writers stop optimistic spinning
simply because there is no easy way to figure out if all the readers
are actively running or not. However, there are scenarios where
the readers are unlikely to sleep and optimistic spinning can help
performance.
This patch provides a simple mechanism for spinning on a reader-owned
rwsem by a writer. It is a time threshold based spinning where the
allowable spinning time can vary from 10us to 25us depending on the
condition of the rwsem.
When the time threshold is exceeded, the nonspinnable bits will be set
in the owner field to indicate that no more optimistic spinning will
be allowed on this rwsem until it becomes writer owned again. Not even
readers is allowed to acquire the reader-locked rwsem by optimistic
spinning for fairness.
We also want a writer to acquire the lock after the readers hold the
lock for a relatively long time. In order to give preference to writers
under such a circumstance, the single RWSEM_NONSPINNABLE bit is now split
into two - one for reader and one for writer. When optimistic spinning
is disabled, both bits will be set. When the reader count drop down
to 0, the writer nonspinnable bit will be cleared to allow writers to
spin on the lock, but not the readers. When a writer acquires the lock,
it will write its own task structure pointer into sem->owner and clear
the reader nonspinnable bit in the process.
The time taken for each iteration of the reader-owned rwsem spinning
loop varies. Below are sample minimum elapsed times for 16 iterations
of the loop.
System Time for 16 Iterations
------ ----------------------
1-socket Skylake ~800ns
4-socket Broadwell ~300ns
2-socket ThunderX2 (arm64) ~250ns
When the lock cacheline is contended, we can see up to almost 10X
increase in elapsed time. So 25us will be at most 500, 1300 and 1600
iterations for each of the above systems.
With a locking microbenchmark running on 5.1 based kernel, the total
locking rates (in kops/s) on a 8-socket IvyBridge-EX system with
equal numbers of readers and writers before and after this patch were
as follows:
# of Threads Pre-patch Post-patch
------------ --------- ----------
2 1,759 6,684
4 1,684 6,738
8 1,074 7,222
16 900 7,163
32 458 7,316
64 208 520
128 168 425
240 143 474
This patch gives a big boost in performance for mixed reader/writer
workloads.
With 32 locking threads, the rwsem lock event data were:
rwsem_opt_fail=79850
rwsem_opt_nospin=5069
rwsem_opt_rlock=597484
rwsem_opt_wlock=957339
rwsem_sleep_reader=57782
rwsem_sleep_writer=55663
With 64 locking threads, the data looked like:
rwsem_opt_fail=346723
rwsem_opt_nospin=6293
rwsem_opt_rlock=1127119
rwsem_opt_wlock=1400628
rwsem_sleep_reader=308201
rwsem_sleep_writer=72281
So a lot more threads acquired the lock in the slowpath and more threads
went to sleep.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-15-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This patch enables readers to optimistically spin on a
rwsem when it is owned by a writer instead of going to sleep
directly. The rwsem_can_spin_on_owner() function is extracted
out of rwsem_optimistic_spin() and is called directly by
rwsem_down_read_slowpath() and rwsem_down_write_slowpath().
With a locking microbenchmark running on 5.1 based kernel, the total
locking rates (in kops/s) on a 8-socket IvyBrige-EX system with equal
numbers of readers and writers before and after the patch were as
follows:
# of Threads Pre-patch Post-patch
------------ --------- ----------
4 1,674 1,684
8 1,062 1,074
16 924 900
32 300 458
64 195 208
128 164 168
240 149 143
The performance change wasn't significant in this case, but this change
is required by a follow-on patch.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-13-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When the front of the wait queue is a reader, other readers
immediately following the first reader will also be woken up at the
same time. However, if there is a writer in between. Those readers
behind the writer will not be woken up.
Because of optimistic spinning, the lock acquisition order is not FIFO
anyway. The lock handoff mechanism will ensure that lock starvation
will not happen.
Assuming that the lock hold times of the other readers still in the
queue will be about the same as the readers that are being woken up,
there is really not much additional cost other than the additional
latency due to the wakeup of additional tasks by the waker. Therefore
all the readers up to a maximum of 256 in the queue are woken up when
the first waiter is a reader to improve reader throughput. This is
somewhat similar in concept to a phase-fair R/W lock.
With a locking microbenchmark running on 5.1 based kernel, the total
locking rates (in kops/s) on a 8-socket IvyBridge-EX system with
equal numbers of readers and writers before and after this patch were
as follows:
# of Threads Pre-Patch Post-patch
------------ --------- ----------
4 1,641 1,674
8 731 1,062
16 564 924
32 78 300
64 38 195
240 50 149
There is no performance gain at low contention level. At high contention
level, however, this patch gives a pretty decent performance boost.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-11-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
An RT task can do optimistic spinning only if the lock holder is
actually running. If the state of the lock holder isn't known, there
is a possibility that high priority of the RT task may block forward
progress of the lock holder if it happens to reside on the same CPU.
This will lead to deadlock. So we have to make sure that an RT task
will not spin on a reader-owned rwsem.
When the owner is temporarily set to NULL, there are two cases
where we may want to continue spinning:
1) The lock owner is in the process of releasing the lock, sem->owner
is cleared but the lock has not been released yet.
2) The lock was free and owner cleared, but another task just comes
in and acquire the lock before we try to get it. The new owner may
be a spinnable writer.
So an RT task is now made to retry one more time to see if it can
acquire the lock or continue spinning on the new owning writer.
When testing on a 8-socket IvyBridge-EX system, the one additional retry
seems to improve locking performance of RT write locking threads under
heavy contentions. The table below shows the locking rates (in kops/s)
with various write locking threads before and after the patch.
Locking threads Pre-patch Post-patch
--------------- --------- -----------
4 2,753 2,608
8 2,529 2,520
16 1,727 1,918
32 1,263 1,956
64 889 1,343
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-10-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Because of writer lock stealing, it is possible that a constant
stream of incoming writers will cause a waiting writer or reader to
wait indefinitely leading to lock starvation.
This patch implements a lock handoff mechanism to disable lock stealing
and force lock handoff to the first waiter or waiters (for readers)
in the queue after at least a 4ms waiting period unless it is a RT
writer task which doesn't need to wait. The waiting period is used to
avoid discouraging lock stealing too much to affect performance.
The setting and clearing of the handoff bit is serialized by the
wait_lock. So racing is not possible.
A rwsem microbenchmark was run for 5 seconds on a 2-socket 40-core
80-thread Skylake system with a v5.1 based kernel and 240 write_lock
threads with 5us sleep critical section.
Before the patch, the min/mean/max numbers of locking operations for
the locking threads were 1/7,792/173,696. After the patch, the figures
became 5,842/6,542/7,458. It can be seen that the rwsem became much
more fair, though there was a drop of about 16% in the mean locking
operations done which was a tradeoff of having better fairness.
Making the waiter set the handoff bit right after the first wakeup can
impact performance especially with a mixed reader/writer workload. With
the same microbenchmark with short critical section and equal number of
reader and writer threads (40/40), the reader/writer locking operation
counts with the current patch were:
40 readers, Iterations Min/Mean/Max = 1,793/1,794/1,796
40 writers, Iterations Min/Mean/Max = 1,793/34,956/86,081
By making waiter set handoff bit immediately after wakeup:
40 readers, Iterations Min/Mean/Max = 43/44/46
40 writers, Iterations Min/Mean/Max = 43/1,263/3,191
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-8-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
After merging all the relevant rwsem code into one single file, there
are a number of optimizations and cleanups that can be done:
1) Remove all the EXPORT_SYMBOL() calls for functions that are not
accessed elsewhere.
2) Remove all the __visible tags as none of the functions will be
called from assembly code anymore.
3) Make all the internal functions static.
4) Remove some unneeded blank lines.
5) Remove the intermediate rwsem_down_{read|write}_failed*() functions
and rename __rwsem_down_{read|write}_failed_common() to
rwsem_down_{read|write}_slowpath().
6) Remove "__" prefix of __rwsem_mark_wake().
7) Use atomic_long_try_cmpxchg_acquire() as much as possible.
8) Remove the rwsem_rtrylock and rwsem_wtrylock lock events as they
are not that useful.
That enables the compiler to do better optimization and reduce code
size. The text+data size of rwsem.o on an x86-64 machine with gcc8 was
reduced from 10237 bytes to 5030 bytes with this change.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-6-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Now we only have one implementation of rwsem. Even though we still use
xadd to handle reader locking, we use cmpxchg for writer instead. So
the filename rwsem-xadd.c is not strictly correct. Also no one outside
of the rwsem code need to know the internal implementation other than
function prototypes for two internal functions that are called directly
from percpu-rwsem.c.
So the rwsem-xadd.c and rwsem.h files are now merged into rwsem.c in
the following order:
<upper part of rwsem.h>
<rwsem-xadd.c>
<lower part of rwsem.h>
<rwsem.c>
The rwsem.h file now contains only 2 function declarations for
__up_read() and __down_read().
This is a code relocation patch with no code change at all except
making __up_read() and __down_read() non-static functions so they
can be used by percpu-rwsem.c.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-5-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The current way of using various reader, writer and waiting biases
in the rwsem code are confusing and hard to understand. I have to
reread the rwsem count guide in the rwsem-xadd.c file from time to
time to remind myself how this whole thing works. It also makes the
rwsem code harder to be optimized.
To make rwsem more sane, a new locking scheme similar to the one in
qrwlock is now being used. The atomic long count has the following
bit definitions:
Bit 0 - writer locked bit
Bit 1 - waiters present bit
Bits 2-7 - reserved for future extension
Bits 8-X - reader count (24/56 bits)
The cmpxchg instruction is now used to acquire the write lock. The read
lock is still acquired with xadd instruction, so there is no change here.
This scheme will allow up to 16M/64P active readers which should be
more than enough. We can always use some more reserved bits if necessary.
With that change, we can deterministically know if a rwsem has been
write-locked. Looking at the count alone, however, one cannot determine
for certain if a rwsem is owned by readers or not as the readers that
set the reader count bits may be in the process of backing out. So we
still need the reader-owned bit in the owner field to be sure.
With a locking microbenchmark running on 5.1 based kernel, the total
locking rates (in kops/s) of the benchmark on a 8-socket 120-core
IvyBridge-EX system before and after the patch were as follows:
Before Patch After Patch
# of Threads wlock rlock wlock rlock
------------ ----- ----- ----- -----
1 30,659 31,341 31,055 31,283
2 8,909 16,457 9,884 17,659
4 9,028 15,823 8,933 20,233
8 8,410 14,212 7,230 17,140
16 8,217 25,240 7,479 24,607
The locking rates of the benchmark on a Power8 system were as follows:
Before Patch After Patch
# of Threads wlock rlock wlock rlock
------------ ----- ----- ----- -----
1 12,963 13,647 13,275 13,601
2 7,570 11,569 7,902 10,829
4 5,232 5,516 5,466 5,435
8 5,233 3,386 5,467 3,168
The locking rates of the benchmark on a 2-socket ARM64 system were
as follows:
Before Patch After Patch
# of Threads wlock rlock wlock rlock
------------ ----- ----- ----- -----
1 21,495 21,046 21,524 21,074
2 5,293 10,502 5,333 10,504
4 5,325 11,463 5,358 11,631
8 5,391 11,712 5,470 11,680
The performance are roughly the same before and after the patch. There
are run-to-run variations in performance. Runs with higher variances
usually have higher throughput.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-4-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
After the following commit:
59aabfc7e9 ("locking/rwsem: Reduce spinlock contention in wakeup after up_read()/up_write()")
the rwsem_wake() forgoes doing a wakeup if the wait_lock cannot be directly
acquired and an optimistic spinning locker is present. This can help performance
by avoiding spinning on the wait_lock when it is contended.
With the later commit:
133e89ef5e ("locking/rwsem: Enable lockless waiter wakeup(s)")
the performance advantage of the above optimization diminishes as the average
wait_lock hold time become much shorter.
With a later patch that supports rwsem lock handoff, we can no
longer relies on the fact that the presence of an optimistic spinning
locker will ensure that the lock will be acquired by a task soon and
rwsem_wake() will be called later on to wake up waiters. This can lead
to missed wakeup and application hang.
So the original 59aabfc7e9 commit has to be reverted.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Link: https://lkml.kernel.org/r/20190520205918.22251-3-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Make pcc_cpufreq_init() return error codes when the driver cannot be
registered. Otherwise the driver can shows up loaded via lsmod even
though it failed initialization. This is confusing to the user.
Signed-off-by: David Arcari <darcari@redhat.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
When a cfs_rq sleeps and returns its quota, we delay for 5ms before
waking any throttled cfs_rqs to coalesce with other cfs_rqs going to
sleep, as this has to be done outside of the rq lock we hold.
The current code waits for 5ms without any sleeps, instead of waiting
for 5ms from the first sleep, which can delay the unthrottle more than
we want. Switch this around so that we can't push this forward forever.
This requires an extra flag rather than using hrtimer_active, since we
need to start a new timer if the current one is in the process of
finishing.
Signed-off-by: Ben Segall <bsegall@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Xunlei Pang <xlpang@linux.alibaba.com>
Acked-by: Phil Auld <pauld@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/xm26a7euy6iq.fsf_-_@bsegall-linux.svl.corp.google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Jens reported that significant performance can be had on some block
workloads by special casing local wakeups. That is, wakeups on the
current task before it schedules out.
Given something like the normal wait pattern:
for (;;) {
set_current_state(TASK_UNINTERRUPTIBLE);
if (cond)
break;
schedule();
}
__set_current_state(TASK_RUNNING);
Any wakeup (on this CPU) after set_current_state() and before
schedule() would benefit from this.
Normal wakeups take p->pi_lock, which serializes wakeups to the same
task. By eliding that we gain concurrency on:
- ttwu_stat(); we already had concurrency on rq stats, this now also
brings it to task stats. -ENOCARE
- tracepoints; it is now possible to get multiple instances of
trace_sched_waking() (and possibly trace_sched_wakeup()) for the
same task. Tracers will have to learn to cope.
Furthermore, p->pi_lock is used by set_special_state(), to order
against TASK_RUNNING stores from other CPUs. But since this is
strictly CPU local, we don't need the lock, and set_special_state()'s
disabling of IRQs is sufficient.
After the normal wakeup takes p->pi_lock it issues
smp_mb__after_spinlock(), in order to ensure the woken task must
observe prior stores before we observe the p->state. If this is CPU
local, this will be satisfied with a compiler barrier, and we rely on
try_to_wake_up() being a funcation call, which implies such.
Since, when 'p == current', 'p->on_rq' must be true, the normal wakeup
would continue into the ttwu_remote() branch, which normally is
concerned with exactly this wakeup scenario, except from a remote CPU.
IOW we're waking a task that is still running. In this case, we can
trivially avoid taking rq->lock, all that's left from this is to set
p->state.
This then yields an extremely simple and fast path for 'p == current'.
Reported-by: Jens Axboe <axboe@kernel.dk>
Tested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qian Cai <cai@lca.pw>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: akpm@linux-foundation.org
Cc: gkohli@codeaurora.org
Cc: hch@lst.de
Cc: oleg@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
runnable_avg_yN_inv[] is only used in kernel/sched/pelt.c but was
included in several other places because they need other macros all
came from kernel/sched/sched-pelt.h which was generated by
Documentation/scheduler/sched-pelt. As the result, it causes compilation
a lot of warnings,
kernel/sched/sched-pelt.h:4:18: warning: 'runnable_avg_yN_inv' defined but not used [-Wunused-const-variable=]
kernel/sched/sched-pelt.h:4:18: warning: 'runnable_avg_yN_inv' defined but not used [-Wunused-const-variable=]
kernel/sched/sched-pelt.h:4:18: warning: 'runnable_avg_yN_inv' defined but not used [-Wunused-const-variable=]
...
Silence it by appending the __maybe_unused attribute for it, so all
generated variables and macros can still be kept in the same file.
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/1559596304-31581-1-git-send-email-cai@lca.pw
Signed-off-by: Ingo Molnar <mingo@kernel.org>
cfs_rq_has_blocked() and others_have_blocked() are only used within
update_blocked_averages(). The !CONFIG_FAIR_GROUP_SCHED version of the
latter calls them within a #define CONFIG_NO_HZ_COMMON block, whereas
the CONFIG_FAIR_GROUP_SCHED one calls them unconditionnally.
As reported by Qian, the above leads to this warning in
!CONFIG_NO_HZ_COMMON configs:
kernel/sched/fair.c: In function 'update_blocked_averages':
kernel/sched/fair.c:7750:7: warning: variable 'done' set but not used [-Wunused-but-set-variable]
It wouldn't be wrong to keep cfs_rq_has_blocked() and
others_have_blocked() as they are, but since their only current use is
to figure out when we can stop calling update_blocked_averages() on
fully decayed NOHZ idle CPUs, we can give them a new definition for
!CONFIG_NO_HZ_COMMON.
Change the definition of cfs_rq_has_blocked() and
others_have_blocked() for !CONFIG_NO_HZ_COMMON so that the
NOHZ-specific blocks of update_blocked_averages() become no-ops and
the 'done' variable gets optimised out.
While at it, remove the CONFIG_NO_HZ_COMMON block from the
!CONFIG_FAIR_GROUP_SCHED definition of update_blocked_averages() by
using the newly-introduced update_blocked_load_status() helper.
No change in functionality intended.
[ Additions by Peter Zijlstra. ]
Reported-by: Qian Cai <cai@lca.pw>
Signed-off-by: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190603115424.7951-1-valentin.schneider@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Recent probing at the Linux Kernel Memory Model uncovered a
'surprise'. Strongly ordered architectures where the atomic RmW
primitive implies full memory ordering and
smp_mb__{before,after}_atomic() are a simple barrier() (such as x86)
fail for:
*x = 1;
atomic_inc(u);
smp_mb__after_atomic();
r0 = *y;
Because, while the atomic_inc() implies memory order, it
(surprisingly) does not provide a compiler barrier. This then allows
the compiler to re-order like so:
atomic_inc(u);
*x = 1;
smp_mb__after_atomic();
r0 = *y;
Which the CPU is then allowed to re-order (under TSO rules) like:
atomic_inc(u);
r0 = *y;
*x = 1;
And this very much was not intended. Therefore strengthen the atomic
RmW ops to include a compiler barrier.
NOTE: atomic_{or,and,xor} and the bitops already had the compiler
barrier.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
All callers of lockdep_assert_held_exclusive() use it to verify the
correct locking state of either a semaphore (ldisc_sem in tty,
mmap_sem for perf events, i_rwsem of inode for dax) or rwlock by
apparmor. Thus it makes sense to rename _exclusive to _write since
that's the semantics callers care. Additionally there is already
lockdep_assert_held_read(), which this new naming is more consistent with.
No functional changes.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190531100651.3969-1-nborisov@suse.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Currently, the jump label of a static key is transformed via the arch
specific function:
void arch_jump_label_transform(struct jump_entry *entry,
enum jump_label_type type)
The new approach (batch mode) uses two arch functions, the first has the
same arguments of the arch_jump_label_transform(), and is the function:
bool arch_jump_label_transform_queue(struct jump_entry *entry,
enum jump_label_type type)
Rather than transforming the code, it adds the jump_entry in a queue of
entries to be updated. This functions returns true in the case of a
successful enqueue of an entry. If it returns false, the caller must to
apply the queue and then try to queue again, for instance, because the
queue is full.
This function expects the caller to sort the entries by the address before
enqueueuing then. This is already done by the arch independent code, though.
After queuing all jump_entries, the function:
void arch_jump_label_transform_apply(void)
Applies the changes in the queue.
Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Chris von Recklinghausen <crecklin@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jason Baron <jbaron@akamai.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott Wood <swood@redhat.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/57b4caa654bad7e3b066301c9a9ae233dea065b5.1560325897.git.bristot@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
If the architecture supports the batching of jump label updates, use it!
An easy way to see the benefits of this patch is switching the
schedstats on and off. For instance:
-------------------------- %< ----------------------------
#!/bin/sh
while [ true ]; do
sysctl -w kernel.sched_schedstats=1
sleep 2
sysctl -w kernel.sched_schedstats=0
sleep 2
done
-------------------------- >% ----------------------------
while watching the IPI count:
-------------------------- %< ----------------------------
# watch -n1 "cat /proc/interrupts | grep Function"
-------------------------- >% ----------------------------
With the current mode, it is possible to see +- 168 IPIs each 2 seconds,
while with this patch the number of IPIs goes to 3 each 2 seconds.
Regarding the performance impact of this patch set, I made two measurements:
The time to update a key (the task that is causing the change)
The time to run the int3 handler (the side effect on a thread that
hits the code being changed)
The schedstats static key was chosen as the key to being switched on and off.
The reason being is that it is used in more than 56 places, in a hot path. The
change in the schedstats static key will be done with the following command:
while [ true ]; do
sysctl -w kernel.sched_schedstats=1
usleep 500000
sysctl -w kernel.sched_schedstats=0
usleep 500000
done
In this way, they key will be updated twice per second. To force the hit of the
int3 handler, the system will also run a kernel compilation with two jobs per
CPU. The test machine is a two nodes/24 CPUs box with an Intel Xeon processor
@2.27GHz.
Regarding the update part, on average, the regular kernel takes 57 ms to update
the schedstats key, while the kernel with the batch updates takes just 1.4 ms
on average. Although it seems to be too good to be true, it makes sense: the
schedstats key is used in 56 places, so it was expected that it would take
around 56 times to update the keys with the current implementation, as the
IPIs are the most expensive part of the update.
Regarding the int3 handler, the non-batch handler takes 45 ns on average, while
the batch version takes around 180 ns. At first glance, it seems to be a high
value. But it is not, considering that it is doing 56 updates, rather than one!
It is taking four times more, only. This gain is possible because the patch
uses a binary search in the vector: log2(56)=5.8. So, it was expected to have
an overhead within four times.
(voice of tv propaganda) But, that is not all! As the int3 handler keeps on for
a shorter period (because the update part is on for a shorter time), the number
of hits in the int3 handler decreased by 10%.
The question then is: Is it worth paying the price of "135 ns" more in the int3
handler?
Considering that, in this test case, we are saving the handling of 53 IPIs,
that takes more than these 135 ns, it seems to be a meager price to be paid.
Moreover, the test case was forcing the hit of the int3, in practice, it
does not take that often. While the IPI takes place on all CPUs, hitting
the int3 handler or not!
For instance, in an isolated CPU with a process running in user-space
(nohz_full use-case), the chances of hitting the int3 handler is barely zero,
while there is no way to avoid the IPIs. By bounding the IPIs, we are improving
a lot this scenario.
Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Chris von Recklinghausen <crecklin@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jason Baron <jbaron@akamai.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott Wood <swood@redhat.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/acc891dbc2dbc9fd616dd680529a2337b1d1274c.1560325897.git.bristot@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Currently, the patch of an address is done in three steps:
-- Pseudo-code #1 - Current implementation ---
1) add an int3 trap to the address that will be patched
sync cores (send IPI to all other CPUs)
2) update all but the first byte of the patched range
sync cores (send IPI to all other CPUs)
3) replace the first byte (int3) by the first byte of replacing opcode
sync cores (send IPI to all other CPUs)
-- Pseudo-code #1 ---
When a static key has more than one entry, these steps are called once for
each entry. The number of IPIs then is linear with regard to the number 'n' of
entries of a key: O(n*3), which is O(n).
This algorithm works fine for the update of a single key. But we think
it is possible to optimize the case in which a static key has more than
one entry. For instance, the sched_schedstats jump label has 56 entries
in my (updated) fedora kernel, resulting in 168 IPIs for each CPU in
which the thread that is enabling the key is _not_ running.
With this patch, rather than receiving a single patch to be processed, a vector
of patches is passed, enabling the rewrite of the pseudo-code #1 in this
way:
-- Pseudo-code #2 - This patch ---
1) for each patch in the vector:
add an int3 trap to the address that will be patched
sync cores (send IPI to all other CPUs)
2) for each patch in the vector:
update all but the first byte of the patched range
sync cores (send IPI to all other CPUs)
3) for each patch in the vector:
replace the first byte (int3) by the first byte of replacing opcode
sync cores (send IPI to all other CPUs)
-- Pseudo-code #2 - This patch ---
Doing the update in this way, the number of IPI becomes O(3) with regard
to the number of keys, which is O(1).
The batch mode is done with the function text_poke_bp_batch(), that receives
two arguments: a vector of "struct text_to_poke", and the number of entries
in the vector.
The vector must be sorted by the addr field of the text_to_poke structure,
enabling the binary search of a handler in the poke_int3_handler function
(a fast path).
Signed-off-by: Daniel Bristot de Oliveira <bristot@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Chris von Recklinghausen <crecklin@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jason Baron <jbaron@akamai.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott Wood <swood@redhat.com>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/ca506ed52584c80f64de23f6f55ca288e5d079de.1560325897.git.bristot@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Add support for 4 bit instances on TI's J721E devices. Because these
instances have no DLL, introduce a DLL_PRESENT flag and make sure DLL
related registers are only accessed when it is present. Also add a
separate set_clock callback for this compatible.
Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The 8 bit IP on the TI's J721E device departs from the AM654x IP in some
ways which require special handling. Create a driver_data structure
which holds the pltfm_data and a flags field which is used to indicate
these differences. These are the following:
1. The pins are not muxed with anything else inside the SoC and hence the
IOMUX_ENABLE field does not exist. Add a flag which is used to
indicate the presence of the field.
2. The register field used to select DLL frequency is 3 bit wide as
compared to 2 bits in AM65x. Add another flag which differentiates
between 3 bit and 2 bit fields.
3. The strobe select field is 8 bit wide as compared to 4 bits for
AM65x. Add yet another flag to indicate this difference. Strobe select
is used only for HS400 speed mode, support for which has not yet been
added in AM65x.
Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Add binding documentation for mmc host controllers present on TI's J721E
SOC. The 4 bit IP on J721E doesn't have a phy DLL so make DLL related
properties as optional for that compatible. Also add an optional
strobe-sel property used for HS400 speed mode.
Signed-off-by: Faiz Abbas <faiz_abbas@ti.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The problem is that on 64bit systems then we don't clear the higher
bits of the "pending" variable. So when we do:
ack = pending & ~BIT(STMFX_REG_IRQ_SRC_EN_GPIO);
if (ack) {
the if (ack) condition relies on uninitialized data. The fix it that
I've changed "pending" from an unsigned long to a u32. I changed "n" as
well, because that's a number in the 0-10 range and it fits easily
inside an int. We do need to add a cast to "pending" when we use it in
the for_each_set_bit() loop, but that doesn't cause a problem, it's
fine.
Fixes: 06252ade91 ("mfd: Add ST Multi-Function eXpander (STMFX) core driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Amelie Delaunay <amelie.delaunay@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
After commit b38ff4075a, the following command does not work anymore:
$ ip xfrm state add src 10.125.0.2 dst 10.125.0.1 proto esp spi 34 reqid 1 \
mode tunnel enc 'cbc(aes)' 0xb0abdba8b782ad9d364ec81e3a7d82a1 auth-trunc \
'hmac(sha1)' 0xe26609ebd00acb6a4d51fca13e49ea78a72c73e6 96 flag align4
In fact, the selector is not mandatory, allow the user to provide an empty
selector.
Fixes: b38ff4075a ("xfrm: Fix xfrm sel prefix length validation")
CC: Anirudh Gupta <anirudh.gupta@sophos.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
CONFIG_ARM_GIC_MAX_NR is enabled by default.
It is redundant in x86 and IA-64 where is
without GIC.
Signed-off-by: Jiangfeng Xiao <xiaojiangfeng@huawei.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
At Rob's request, we're starting to migrate our DT binding
documentation to json-schema YAML format. Start by converting our cpu
binding documentation. While doing so, document more properties and
nodes. This includes adding binding documentation support for the E51
and U54 CPU cores ("harts") that are present on this SoC. These cores
are described in:
https://static.dev.sifive.com/FU540-C000-v1.0.pdf
This cpus.yaml file is intended to be a starting point and to
evolve over time. It passes dt-doc-validate as of the yaml-bindings
commit 4c79d42e9216.
This patch was originally based on the ARM json-schema binding
documentation as added by commit 672951cbd1 ("dt-bindings: arm: Convert
cpu binding to json-schema").
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
Signed-off-by: Paul Walmsley <paul@pwsan.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: devicetree@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: linux-riscv@lists.infradead.org
