[ Upstream commit e81859fe64ad42dccefe134d1696e0635f78d763 ]
Apparently despite it being marked inline, the compiler
may not inline __down_write_common() which makes it difficult
to identify the cause of lock contention, as the wchan of the
blocked function will always be listed as __down_write_common().
So add __always_inline annotation to the common function (as
well as the inlined helper callers) to force it to be inlined
so a more useful blocking function will be listed (via wchan).
This mirrors commit 92cc5d00a431 ("locking/rwsem: Add
__always_inline annotation to __down_read_common() and inlined
callers") which did the same for __down_read_common.
I sort of worry that I'm playing wack-a-mole here, and talking
with compiler people, they tell me inline means nothing, which
makes me want to cry a little. So I'm wondering if we need to
replace all the inlines with __always_inline, or remove them
because either we mean something by it, or not.
Fixes: c995e638cc ("locking/rwsem: Fold __down_{read,write}*()")
Reported-by: Tim Murray <timmurray@google.com>
Signed-off-by: John Stultz <jstultz@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Waiman Long <longman@redhat.com>
Link: https://lkml.kernel.org/r/20240709060831.495366-1-jstultz@google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 92cc5d00a431e96e5a49c0b97e5ad4fa7536bd4b upstream.
Apparently despite it being marked inline, the compiler
may not inline __down_read_common() which makes it difficult
to identify the cause of lock contention, as the blocked
function in traceevents will always be listed as
__down_read_common().
So this patch adds __always_inline annotation to the common
function (as well as the inlined helper callers) to force it to
be inlined so the blocking function will be listed (via Wchan)
in traceevents.
Fixes: c995e638cc ("locking/rwsem: Fold __down_{read,write}*()")
Reported-by: Tim Murray <timmurray@google.com>
Signed-off-by: John Stultz <jstultz@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Waiman Long <longman@redhat.com>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20230503023351.2832796-1-jstultz@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b613c7f31476c44316bfac1af7cac714b7d6bef9 upstream.
A non-first waiter can potentially spin in the for loop of
rwsem_down_write_slowpath() without sleeping but fail to acquire the
lock even if the rwsem is free if the following sequence happens:
Non-first RT waiter First waiter Lock holder
------------------- ------------ -----------
Acquire wait_lock
rwsem_try_write_lock():
Set handoff bit if RT or
wait too long
Set waiter->handoff_set
Release wait_lock
Acquire wait_lock
Inherit waiter->handoff_set
Release wait_lock
Clear owner
Release lock
if (waiter.handoff_set) {
rwsem_spin_on_owner(();
if (OWNER_NULL)
goto trylock_again;
}
trylock_again:
Acquire wait_lock
rwsem_try_write_lock():
if (first->handoff_set && (waiter != first))
return false;
Release wait_lock
A non-first waiter cannot really acquire the rwsem even if it mistakenly
believes that it can spin on OWNER_NULL value. If that waiter happens
to be an RT task running on the same CPU as the first waiter, it can
block the first waiter from acquiring the rwsem leading to live lock.
Fix this problem by making sure that a non-first waiter cannot spin in
the slowpath loop without sleeping.
Fixes: d257cc8cb8 ("locking/rwsem: Make handoff bit handling more consistent")
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Mukesh Ojha <quic_mojha@quicinc.com>
Reviewed-by: Mukesh Ojha <quic_mojha@quicinc.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230126003628.365092-2-longman@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 3f5245538a1964ae186ab7e1636020a41aa63143 ]
Commit:
91d2a812df ("locking/rwsem: Make handoff writer optimistically spin on owner")
... assumes that when the owner field is changed to NULL, the lock will
become free soon. But commit:
48dfb5d256 ("locking/rwsem: Disable preemption while trying for rwsem lock")
... disabled preemption when acquiring rwsem for write.
However, preemption has not yet been disabled when acquiring a read lock
on a rwsem. So a reader can add a RWSEM_READER_BIAS to count without
setting owner to signal a reader, got preempted out by a RT task which
then spins in the writer slowpath as owner remains NULL leading to live lock.
One easy way to fix this problem is to disable preemption at all the
down_read*() and up_read() code paths as implemented in this patch.
Fixes: 91d2a812df ("locking/rwsem: Make handoff writer optimistically spin on owner")
Reported-by: Mukesh Ojha <quic_mojha@quicinc.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20230126003628.365092-3-longman@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
Make the region inside the rwsem_write_trylock non preemptible.
We observe RT task is hogging CPU when trying to acquire rwsem lock
which was acquired by a kworker task but before the rwsem owner was set.
Here is the scenario:
1. CFS task (affined to a particular CPU) takes rwsem lock.
2. CFS task gets preempted by a RT task before setting owner.
3. RT task (FIFO) is trying to acquire the lock, but spinning until
RT throttling happens for the lock as the lock was taken by CFS task.
This patch attempts to fix the above issue by disabling preemption
until owner is set for the lock. While at it also fix the issues
at the places where rwsem_{set,clear}_owner() are called.
This also adds lockdep annotation of preemption disable in
rwsem_{set,clear}_owner() on Peter Z. suggestion.
Signed-off-by: Gokul krishna Krishnakumar <quic_gokukris@quicinc.com>
Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Waiman Long <longman@redhat.com>
Link: https://lore.kernel.org/r/1662661467-24203-1-git-send-email-quic_mojha@quicinc.com
With commit d257cc8cb8 ("locking/rwsem: Make handoff bit handling more
consistent"), the writer that sets the handoff bit can be interrupted
out without clearing the bit if the wait queue isn't empty. This disables
reader and writer optimistic lock spinning and stealing.
Now if a non-first writer in the queue is somehow woken up or a new
waiter enters the slowpath, it can't acquire the lock. This is not the
case before commit d257cc8cb8 as the writer that set the handoff bit
will clear it when exiting out via the out_nolock path. This is less
efficient as the busy rwsem stays in an unlock state for a longer time.
In some cases, this new behavior may cause lockups as shown in [1] and
[2].
This patch allows a non-first writer to ignore the handoff bit if it
is not originally set or initiated by the first waiter. This patch is
shown to be effective in fixing the lockup problem reported in [1].
[1] https://lore.kernel.org/lkml/20220617134325.GC30825@techsingularity.net/
[2] https://lore.kernel.org/lkml/3f02975c-1a9d-be20-32cf-f1d8e3dfafcc@oracle.com/
Fixes: d257cc8cb8 ("locking/rwsem: Make handoff bit handling more consistent")
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: John Donnelly <john.p.donnelly@oracle.com>
Tested-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lore.kernel.org/r/20220622200419.778799-1-longman@redhat.com
For writers, the out_nolock path will always attempt to wake up waiters.
This may not be really necessary if the waiter to be removed is not the
first one.
For readers, no attempt to wake up waiter is being made. However, if
the HANDOFF bit is set and the reader to be removed is the first waiter,
the waiter behind it will inherit the HANDOFF bit and for a write lock
waiter waking it up will allow it to spin on the lock to acquire it
faster. So it can be beneficial to do a wakeup in this case.
Add a new rwsem_del_wake_waiter() helper function to do that consistently
for both reader and writer out_nolock paths.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220322152059.2182333-4-longman@redhat.com
In an analysis of a recent vmcore, a reader-owned rwsem was found with
385 readers but no writer in the wait queue. That is kind of unusual
but it may be caused by some race conditions that we have not fully
understood yet. In such a case, all the readers in the wait queue should
join the other reader-owners and acquire the read lock.
In rwsem_down_write_slowpath(), an incoming writer will try to
wake up the front readers under such circumstance. That is not
the case for rwsem_down_read_slowpath(), add a new helper function
rwsem_cond_wake_waiter() to do wakeup and use it in both reader and
writer slowpaths to have a consistent and correct behavior.
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220322152059.2182333-3-longman@redhat.com
We found that a process with 10 thousnads threads has been encountered
a regression problem from Linux-v4.14 to Linux-v5.4. It is a kind of
workload which will concurrently allocate lots of memory in different
threads sometimes. In this case, we will see the down_read_trylock()
with a high hotspot. Therefore, we suppose that rwsem has a regression
at least since Linux-v5.4. In order to easily debug this problem, we
write a simply benchmark to create the similar situation lile the
following.
```c++
#include <sys/mman.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <sched.h>
#include <cstdio>
#include <cassert>
#include <thread>
#include <vector>
#include <chrono>
volatile int mutex;
void trigger(int cpu, char* ptr, std::size_t sz)
{
cpu_set_t set;
CPU_ZERO(&set);
CPU_SET(cpu, &set);
assert(pthread_setaffinity_np(pthread_self(), sizeof(set), &set) == 0);
while (mutex);
for (std::size_t i = 0; i < sz; i += 4096) {
*ptr = '\0';
ptr += 4096;
}
}
int main(int argc, char* argv[])
{
std::size_t sz = 100;
if (argc > 1)
sz = atoi(argv[1]);
auto nproc = std::thread::hardware_concurrency();
std::vector<std::thread> thr;
sz <<= 30;
auto* ptr = mmap(nullptr, sz, PROT_READ | PROT_WRITE, MAP_ANON |
MAP_PRIVATE, -1, 0);
assert(ptr != MAP_FAILED);
char* cptr = static_cast<char*>(ptr);
auto run = sz / nproc;
run = (run >> 12) << 12;
mutex = 1;
for (auto i = 0U; i < nproc; ++i) {
thr.emplace_back(std::thread([i, cptr, run]() { trigger(i, cptr, run); }));
cptr += run;
}
rusage usage_start;
getrusage(RUSAGE_SELF, &usage_start);
auto start = std::chrono::system_clock::now();
mutex = 0;
for (auto& t : thr)
t.join();
rusage usage_end;
getrusage(RUSAGE_SELF, &usage_end);
auto end = std::chrono::system_clock::now();
timeval utime;
timeval stime;
timersub(&usage_end.ru_utime, &usage_start.ru_utime, &utime);
timersub(&usage_end.ru_stime, &usage_start.ru_stime, &stime);
printf("usr: %ld.%06ld\n", utime.tv_sec, utime.tv_usec);
printf("sys: %ld.%06ld\n", stime.tv_sec, stime.tv_usec);
printf("real: %lu\n",
std::chrono::duration_cast<std::chrono::milliseconds>(end -
start).count());
return 0;
}
```
The functionality of above program is simply which creates `nproc`
threads and each of them are trying to touch memory (trigger page
fault) on different CPU. Then we will see the similar profile by
`perf top`.
25.55% [kernel] [k] down_read_trylock
14.78% [kernel] [k] handle_mm_fault
13.45% [kernel] [k] up_read
8.61% [kernel] [k] clear_page_erms
3.89% [kernel] [k] __do_page_fault
The highest hot instruction, which accounts for about 92%, in
down_read_trylock() is cmpxchg like the following.
91.89 │ lock cmpxchg %rdx,(%rdi)
Sice the problem is found by migrating from Linux-v4.14 to Linux-v5.4,
so we easily found that the commit ddb20d1d3a ("locking/rwsem: Optimize
down_read_trylock()") caused the regression. The reason is that the
commit assumes the rwsem is not contended at all. But it is not always
true for mmap lock which could be contended with thousands threads.
So most threads almost need to run at least 2 times of "cmpxchg" to
acquire the lock. The overhead of atomic operation is higher than
non-atomic instructions, which caused the regression.
By using the above benchmark, the real executing time on a x86-64 system
before and after the patch were:
Before Patch After Patch
# of Threads real real reduced by
------------ ------ ------ ----------
1 65,373 65,206 ~0.0%
4 15,467 15,378 ~0.5%
40 6,214 5,528 ~11.0%
For the uncontended case, the new down_read_trylock() is the same as
before. For the contended cases, the new down_read_trylock() is faster
than before. The more contended, the more fast.
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Waiman Long <longman@redhat.com>
Link: https://lore.kernel.org/r/20211118094455.9068-1-songmuchun@bytedance.com
There are some inconsistency in the way that the handoff bit is being
handled in readers and writers that lead to a race condition.
Firstly, when a queue head writer set the handoff bit, it will clear
it when the writer is being killed or interrupted on its way out
without acquiring the lock. That is not the case for a queue head
reader. The handoff bit will simply be inherited by the next waiter.
Secondly, in the out_nolock path of rwsem_down_read_slowpath(), both
the waiter and handoff bits are cleared if the wait queue becomes
empty. For rwsem_down_write_slowpath(), however, the handoff bit is
not checked and cleared if the wait queue is empty. This can
potentially make the handoff bit set with empty wait queue.
Worse, the situation in rwsem_down_write_slowpath() relies on wstate,
a variable set outside of the critical section containing the ->count
manipulation, this leads to race condition where RWSEM_FLAG_HANDOFF
can be double subtracted, corrupting ->count.
To make the handoff bit handling more consistent and robust, extract
out handoff bit clearing code into the new rwsem_del_waiter() helper
function. Also, completely eradicate wstate; always evaluate
everything inside the same critical section.
The common function will only use atomic_long_andnot() to clear bits
when the wait queue is empty to avoid possible race condition. If the
first waiter with handoff bit set is killed or interrupted to exit the
slowpath without acquiring the lock, the next waiter will inherit the
handoff bit.
While at it, simplify the trylock for loop in
rwsem_down_write_slowpath() to make it easier to read.
Fixes: 4f23dbc1e6 ("locking/rwsem: Implement lock handoff to prevent lock starvation")
Reported-by: Zhenhua Ma <mazhenhua@xiaomi.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20211116012912.723980-1-longman@redhat.com
Guard the regular sleeping lock specific functionality, which is used for
rtmutex on non-RT enabled kernels and for mutex, rtmutex and semaphores on
RT enabled kernels so the code can be reused for the RT specific
implementation of spinlocks and rwlocks in a different compilation unit.
No functional change.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20210815211303.311535693@linutronix.de
The RT specific R/W semaphore implementation used to restrict the number of
readers to one, because a writer cannot block on multiple readers and
inherit its priority or budget.
The single reader restricting was painful in various ways:
- Performance bottleneck for multi-threaded applications in the page fault
path (mmap sem)
- Progress blocker for drivers which are carefully crafted to avoid the
potential reader/writer deadlock in mainline.
The analysis of the writer code paths shows that properly written RT tasks
should not take them. Syscalls like mmap(), file access which take mmap sem
write locked have unbound latencies, which are completely unrelated to mmap
sem. Other R/W sem users like graphics drivers are not suitable for RT tasks
either.
So there is little risk to hurt RT tasks when the RT rwsem implementation is
done in the following way:
- Allow concurrent readers
- Make writers block until the last reader left the critical section. This
blocking is not subject to priority/budget inheritance.
- Readers blocked on a writer inherit their priority/budget in the normal
way.
There is a drawback with this scheme: R/W semaphores become writer unfair
though the applications which have triggered writer starvation (mostly on
mmap_sem) in the past are not really the typical workloads running on a RT
system. So while it's unlikely to hit writer starvation, it's possible. If
there are unexpected workloads on RT systems triggering it, the problem
has to be revisited.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20210815211303.016885947@linutronix.de
