Since commit:
4bcc595ccd ("printk: reinstate KERN_CONT for printing continuation lines")
printk() requires KERN_CONT to continue log messages. Lots of printk()
in lockdep.c and print_ip_sym() don't have it. As the result lockdep
reports are completely messed up.
Add missing KERN_CONT and inline print_ip_sym() where necessary.
Example of a messed up report:
0-rc5+ #41 Not tainted
-------------------------------------------------------
syz-executor0/5036 is trying to acquire lock:
(
rtnl_mutex
){+.+.+.}
, at:
[<ffffffff86b3d6ac>] rtnl_lock+0x1c/0x20
but task is already holding lock:
(
&net->packet.sklist_lock
){+.+...}
, at:
[<ffffffff873541a6>] packet_diag_dump+0x1a6/0x1920
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3
(
&net->packet.sklist_lock
+.+...}
...
Without this patch all scripts that parse kernel bug reports are broken.
Signed-off-by: Dmitry Vyukov <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
David reported a futex/rtmutex state corruption. It's caused by the
following problem:
CPU0 CPU1 CPU2
l->owner=T1
rt_mutex_lock(l)
lock(l->wait_lock)
l->owner = T1 | HAS_WAITERS;
enqueue(T2)
boost()
unlock(l->wait_lock)
schedule()
rt_mutex_lock(l)
lock(l->wait_lock)
l->owner = T1 | HAS_WAITERS;
enqueue(T3)
boost()
unlock(l->wait_lock)
schedule()
signal(->T2) signal(->T3)
lock(l->wait_lock)
dequeue(T2)
deboost()
unlock(l->wait_lock)
lock(l->wait_lock)
dequeue(T3)
===> wait list is now empty
deboost()
unlock(l->wait_lock)
lock(l->wait_lock)
fixup_rt_mutex_waiters()
if (wait_list_empty(l)) {
owner = l->owner & ~HAS_WAITERS;
l->owner = owner
==> l->owner = T1
}
lock(l->wait_lock)
rt_mutex_unlock(l) fixup_rt_mutex_waiters()
if (wait_list_empty(l)) {
owner = l->owner & ~HAS_WAITERS;
cmpxchg(l->owner, T1, NULL)
===> Success (l->owner = NULL)
l->owner = owner
==> l->owner = T1
}
That means the problem is caused by fixup_rt_mutex_waiters() which does the
RMW to clear the waiters bit unconditionally when there are no waiters in
the rtmutexes rbtree.
This can be fatal: A concurrent unlock can release the rtmutex in the
fastpath because the waiters bit is not set. If the cmpxchg() gets in the
middle of the RMW operation then the previous owner, which just unlocked
the rtmutex is set as the owner again when the write takes place after the
successfull cmpxchg().
The solution is rather trivial: verify that the owner member of the rtmutex
has the waiters bit set before clearing it. This does not require a
cmpxchg() or other atomic operations because the waiters bit can only be
set and cleared with the rtmutex wait_lock held. It's also safe against the
fast path unlock attempt. The unlock attempt via cmpxchg() will either see
the bit set and take the slowpath or see the bit cleared and release it
atomically in the fastpath.
It's remarkable that the test program provided by David triggers on ARM64
and MIPS64 really quick, but it refuses to reproduce on x86-64, while the
problem exists there as well. That refusal might explain that this got not
discovered earlier despite the bug existing from day one of the rtmutex
implementation more than 10 years ago.
Thanks to David for meticulously instrumenting the code and providing the
information which allowed to decode this subtle problem.
Reported-by: David Daney <[email protected]>
Tested-by: David Daney <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Acked-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Sebastian Siewior <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: [email protected]
Fixes: 23f78d4a03 ("[PATCH] pi-futex: rt mutex core")
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Currently the wake_q data structure is defined by the WAKE_Q() macro.
This macro, however, looks like a function doing something as "wake" is
a verb. Even checkpatch.pl was confused as it reported warnings like
WARNING: Missing a blank line after declarations
#548: FILE: kernel/futex.c:3665:
+ int ret;
+ WAKE_Q(wake_q);
This patch renames the WAKE_Q() macro to DEFINE_WAKE_Q() which clarifies
what the macro is doing and eliminates the checkpatch.pl warnings.
Signed-off-by: Waiman Long <[email protected]>
Acked-by: Davidlohr Bueso <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Resolved conflict and added missing rename. ]
Signed-off-by: Ingo Molnar <[email protected]>
Reduce the size of data structure for lockdep entries by half if
PROVE_LOCKING_SMALL if defined. This is used only for sparc.
Signed-off-by: Babu Moger <[email protected]>
Acked-by: Sam Ravnborg <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
Doesn't really matter yet, but pull the HANDOFF and trylock out from
under the wait_lock.
The intention is to add an optimistic spin loop here, which requires
we do not hold the wait_lock, so shuffle code around in preparation.
Also clarify the purpose of taking the wait_lock in the wait loop, its
tempting to want to avoid it altogether, but the cancellation cases
need to to avoid losing wakeups.
Suggested-by: Waiman Long <[email protected]>
Tested-by: Jason Low <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Implement lock handoff to avoid lock starvation.
Lock starvation is possible because mutex_lock() allows lock stealing,
where a running (or optimistic spinning) task beats the woken waiter
to the acquire.
Lock stealing is an important performance optimization because waiting
for a waiter to wake up and get runtime can take a significant time,
during which everyboy would stall on the lock.
The down-side is of course that it allows for starvation.
This patch has the waiter requesting a handoff if it fails to acquire
the lock upon waking. This re-introduces some of the wait time,
because once we do a handoff we have to wait for the waiter to wake up
again.
A future patch will add a round of optimistic spinning to attempt to
alleviate this penalty, but if that turns out to not be enough, we can
add a counter and only request handoff after multiple failed wakeups.
There are a few tricky implementation details:
- accepting a handoff must only be done in the wait-loop. Since the
handoff condition is owner == current, it can easily cause
recursive locking trouble.
- accepting the handoff must be careful to provide the ACQUIRE
semantics.
- having the HANDOFF bit set on unlock requires care, we must not
clear the owner.
- we must be careful to not leave HANDOFF set after we've acquired
the lock. The tricky scenario is setting the HANDOFF bit on an
unlocked mutex.
Tested-by: Jason Low <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Waiman Long <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
The current mutex implementation has an atomic lock word and a
non-atomic owner field.
This disparity leads to a number of issues with the current mutex code
as it means that we can have a locked mutex without an explicit owner
(because the owner field has not been set, or already cleared).
This leads to a number of weird corner cases, esp. between the
optimistic spinning and debug code. Where the optimistic spinning
code needs the owner field updated inside the lock region, the debug
code is more relaxed because the whole lock is serialized by the
wait_lock.
Also, the spinning code itself has a few corner cases where we need to
deal with a held lock without an owner field.
Furthermore, it becomes even more of a problem when trying to fix
starvation cases in the current code. We end up stacking special case
on special case.
To solve this rework the basic mutex implementation to be a single
atomic word that contains the owner and uses the low bits for extra
state.
This matches how PI futexes and rt_mutex already work. By having the
owner an integral part of the lock state a lot of the problems
dissapear and we get a better option to deal with starvation cases,
direct owner handoff.
Changing the basic mutex does however invalidate all the arch specific
mutex code; this patch leaves that unused in-place, a later patch will
remove that.
Tested-by: Jason Low <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Will Deacon <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
stop_two_cpus() and stop_cpus() use stop_cpus_lock to avoid the deadlock,
we need to ensure that the stopper functions can't be queued "backwards"
from one another. This doesn't look nice; if we use lglock then we do not
really need stopper->lock, cpu_stop_queue_work() could use lg_local_lock()
under local_irq_save().
OTOH it would be even better to avoid lglock in stop_machine.c and remove
lg_double_lock(). This patch adds "bool stop_cpus_in_progress" set/cleared
by queue_stop_cpus_work(), and changes cpu_stop_queue_two_works() to busy
wait until it is cleared.
queue_stop_cpus_work() sets stop_cpus_in_progress = T lockless, but after
it queues a work on CPU1 it must be visible to stop_two_cpus(CPU1, CPU2)
which checks it under the same lock. And since stop_two_cpus() holds the
2nd lock too, queue_stop_cpus_work() can not clear stop_cpus_in_progress
if it is also going to queue a work on CPU2, it needs to take that 2nd
lock to do this.
Signed-off-by: Oleg Nesterov <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
When wanting to wakeup readers, __rwsem_mark_wakeup() currently
iterates the wait_list twice while looking to wakeup the first N
queued reader-tasks. While this can be quite inefficient, it was
there such that a awoken reader would be first and foremost
acknowledged by the lock counter.
Keeping the same logic, we can further benefit from the use of
wake_qs and avoid entirely the first wait_list iteration that sets
the counter as wake_up_process() isn't going to occur right away,
and therefore we maintain the counter->list order of going about
things.
Other than saving cycles with O(n) "scanning", this change also
nicely cleans up a good chunk of __rwsem_mark_wakeup(); both
visually and less tedious to read.
For example, the following improvements where seen on some will
it scale microbenchmarks, on a 48-core Haswell:
v4.7 v4.7-rwsem-v1
Hmean signal1-processes-8 5792691.42 ( 0.00%) 5771971.04 ( -0.36%)
Hmean signal1-processes-12 6081199.96 ( 0.00%) 6072174.38 ( -0.15%)
Hmean signal1-processes-21 3071137.71 ( 0.00%) 3041336.72 ( -0.97%)
Hmean signal1-processes-48 3712039.98 ( 0.00%) 3708113.59 ( -0.11%)
Hmean signal1-processes-79 4464573.45 ( 0.00%) 4682798.66 ( 4.89%)
Hmean signal1-processes-110 4486842.01 ( 0.00%) 4633781.71 ( 3.27%)
Hmean signal1-processes-141 4611816.83 ( 0.00%) 4692725.38 ( 1.75%)
Hmean signal1-processes-172 4638157.05 ( 0.00%) 4714387.86 ( 1.64%)
Hmean signal1-processes-203 4465077.80 ( 0.00%) 4690348.07 ( 5.05%)
Hmean signal1-processes-224 4410433.74 ( 0.00%) 4687534.43 ( 6.28%)
Stddev signal1-processes-8 6360.47 ( 0.00%) 8455.31 ( 32.94%)
Stddev signal1-processes-12 4004.98 ( 0.00%) 9156.13 (128.62%)
Stddev signal1-processes-21 3273.14 ( 0.00%) 5016.80 ( 53.27%)
Stddev signal1-processes-48 28420.25 ( 0.00%) 26576.22 ( -6.49%)
Stddev signal1-processes-79 22038.34 ( 0.00%) 18992.70 (-13.82%)
Stddev signal1-processes-110 23226.93 ( 0.00%) 17245.79 (-25.75%)
Stddev signal1-processes-141 6358.98 ( 0.00%) 7636.14 ( 20.08%)
Stddev signal1-processes-172 9523.70 ( 0.00%) 4824.75 (-49.34%)
Stddev signal1-processes-203 13915.33 ( 0.00%) 9326.33 (-32.98%)
Stddev signal1-processes-224 15573.94 ( 0.00%) 10613.82 (-31.85%)
Other runs that saw improvements include context_switch and pipe; and
as expected, this is particularly highlighted on larger thread counts
as it becomes more expensive to walk the list twice.
No change in wakeup ordering or semantics.
Signed-off-by: Davidlohr Bueso <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Currently the percpu-rwsem switches to (global) atomic ops while a
writer is waiting; which could be quite a while and slows down
releasing the readers.
This patch cures this problem by ordering the reader-state vs
reader-count (see the comments in __percpu_down_read() and
percpu_down_write()). This changes a global atomic op into a full
memory barrier, which doesn't have the global cacheline contention.
This also enables using the percpu-rwsem with rcu_sync disabled in order
to bias the implementation differently, reducing the writer latency by
adding some cost to readers.
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Oleg Nesterov <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Paul McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
[ Fixed modular build. ]
Signed-off-by: Ingo Molnar <[email protected]>
Currently there are overlap in the pvqspinlock wait_again and
spurious_wakeup stat counters. Because of lock stealing, it is
no longer possible to accurately determine if spurious wakeup has
happened in the queue head. As they track both the queue node and
queue head status, it is also hard to tell how many of those comes
from the queue head and how many from the queue node.
This patch changes the accounting rules so that spurious wakeup is
only tracked in the queue node. The wait_again count, however, is
only tracked in the queue head when the vCPU failed to acquire the
lock after a vCPU kick. This should give a much better indication of
the wait-kick dynamics in the queue node and the queue head.
Signed-off-by: Waiman Long <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Boqun Feng <[email protected]>
Cc: Douglas Hatch <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Pan Xinhui <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Scott J Norton <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
When the lock holder vCPU is racing with the queue head:
CPU 0 (lock holder) CPU1 (queue head)
=================== =================
spin_lock(); spin_lock();
pv_kick_node(): pv_wait_head_or_lock():
if (!lp) {
lp = pv_hash(lock, pn);
xchg(&l->locked, _Q_SLOW_VAL);
}
WRITE_ONCE(pn->state, vcpu_halted);
cmpxchg(&pn->state,
vcpu_halted, vcpu_hashed);
WRITE_ONCE(l->locked, _Q_SLOW_VAL);
(void)pv_hash(lock, pn);
In this case, lock holder inserts the pv_node of queue head into the
hash table and set _Q_SLOW_VAL unnecessary. This patch avoids it by
restoring/setting vcpu_hashed state after failing adaptive locking
spinning.
Signed-off-by: Wanpeng Li <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Pan Xinhui <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Waiman Long <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Pull locking updates from Ingo Molnar:
"The locking tree was busier in this cycle than the usual pattern - a
couple of major projects happened to coincide.
The main changes are:
- implement the atomic_fetch_{add,sub,and,or,xor}() API natively
across all SMP architectures (Peter Zijlstra)
- add atomic_fetch_{inc/dec}() as well, using the generic primitives
(Davidlohr Bueso)
- optimize various aspects of rwsems (Jason Low, Davidlohr Bueso,
Waiman Long)
- optimize smp_cond_load_acquire() on arm64 and implement LSE based
atomic{,64}_fetch_{add,sub,and,andnot,or,xor}{,_relaxed,_acquire,_release}()
on arm64 (Will Deacon)
- introduce smp_acquire__after_ctrl_dep() and fix various barrier
mis-uses and bugs (Peter Zijlstra)
- after discovering ancient spin_unlock_wait() barrier bugs in its
implementation and usage, strengthen its semantics and update/fix
usage sites (Peter Zijlstra)
- optimize mutex_trylock() fastpath (Peter Zijlstra)
- ... misc fixes and cleanups"
* 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (67 commits)
locking/atomic: Introduce inc/dec variants for the atomic_fetch_$op() API
locking/barriers, arch/arm64: Implement LDXR+WFE based smp_cond_load_acquire()
locking/static_keys: Fix non static symbol Sparse warning
locking/qspinlock: Use __this_cpu_dec() instead of full-blown this_cpu_dec()
locking/atomic, arch/tile: Fix tilepro build
locking/atomic, arch/m68k: Remove comment
locking/atomic, arch/arc: Fix build
locking/Documentation: Clarify limited control-dependency scope
locking/atomic, arch/rwsem: Employ atomic_long_fetch_add()
locking/atomic, arch/qrwlock: Employ atomic_fetch_add_acquire()
locking/atomic, arch/mips: Convert to _relaxed atomics
locking/atomic, arch/alpha: Convert to _relaxed atomics
locking/atomic: Remove the deprecated atomic_{set,clear}_mask() functions
locking/atomic: Remove linux/atomic.h:atomic_fetch_or()
locking/atomic: Implement atomic{,64,_long}_fetch_{add,sub,and,andnot,or,xor}{,_relaxed,_acquire,_release}()
locking/atomic: Fix atomic64_relaxed() bits
locking/atomic, arch/xtensa: Implement atomic_fetch_{add,sub,and,or,xor}()
locking/atomic, arch/x86: Implement atomic{,64}_fetch_{add,sub,and,or,xor}()
locking/atomic, arch/tile: Implement atomic{,64}_fetch_{add,sub,and,or,xor}()
locking/atomic, arch/sparc: Implement atomic{,64}_fetch_{add,sub,and,or,xor}()
...
queued_spin_lock_slowpath() should not worry about another
queued_spin_lock_slowpath() running in interrupt context and
changing node->count by accident, because node->count keeps
the same value every time we enter/leave queued_spin_lock_slowpath().
On some architectures this_cpu_dec() will save/restore irq flags,
which has high overhead. Use the much cheaper __this_cpu_dec() instead.
Signed-off-by: Pan Xinhui <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
[ Rewrote changelog. ]
Signed-off-by: Ingo Molnar <[email protected]>
None of the code actually wants a thread_info, it all wants a
task_struct, and it's just converting back and forth between the two
("ti->task" to get the task_struct from the thread_info, and
"task_thread_info(task)" to go the other way).
No semantic change.
Acked-by: Peter Zijlstra <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This new form allows using hardware assisted waiting.
Some hardware (ARM64 and x86) allow monitoring an address for changes,
so by providing a pointer we can use this to replace the cpu_relax()
with hardware optimized methods in the future.
Requested-by: Will Deacon <[email protected]>
Suggested-by: Linus Torvalds <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Without using WRITE_ONCE(), the compiler can potentially break a
write into multiple smaller ones (store tearing). So a read from the
same data by another task concurrently may return a partial result.
This can result in a kernel crash if the data is a memory address
that is being dereferenced.
This patch changes all write to rwsem->owner to use WRITE_ONCE()
to make sure that store tearing will not happen. READ_ONCE() may
not be needed for rwsem->owner as long as the value is only used for
comparison and not dereferencing.
Signed-off-by: Waiman Long <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Dave Chinner <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Douglas Hatch <[email protected]>
Cc: Jason Low <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Hurley <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Scott J Norton <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Currently, it is not possible to determine for sure if a reader
owns a rwsem by looking at the content of the rwsem data structure.
This patch adds a new state RWSEM_READER_OWNED to the owner field
to indicate that readers currently own the lock. This enables us to
address the following 2 issues in the rwsem optimistic spinning code:
1) rwsem_can_spin_on_owner() will disallow optimistic spinning if
the owner field is NULL which can mean either the readers own
the lock or the owning writer hasn't set the owner field yet.
In the latter case, we miss the chance to do optimistic spinning.
2) While a writer is waiting in the OSQ and a reader takes the lock,
the writer will continue to spin when out of the OSQ in the main
rwsem_optimistic_spin() loop as the owner field is NULL wasting
CPU cycles if some of readers are sleeping.
Adding the new state will allow optimistic spinning to go forward as
long as the owner field is not RWSEM_READER_OWNED and the owner is
running, if set, but stop immediately when that state has been reached.
On a 4-socket Haswell machine running on a 4.6-rc1 based kernel, the
fio test with multithreaded randrw and randwrite tests on the same
file on a XFS partition on top of a NVDIMM were run, the aggregated
bandwidths before and after the patch were as follows:
Test BW before patch BW after patch % change
---- --------------- -------------- --------
randrw 988 MB/s 1192 MB/s +21%
randwrite 1513 MB/s 1623 MB/s +7.3%
The perf profile of the rwsem_down_write_failed() function in randrw
before and after the patch were:
19.95% 5.88% fio [kernel.vmlinux] [k] rwsem_down_write_failed
14.20% 1.52% fio [kernel.vmlinux] [k] rwsem_down_write_failed
The actual CPU cycles spend in rwsem_down_write_failed() dropped from
5.88% to 1.52% after the patch.
The xfstests was also run and no regression was observed.
Signed-off-by: Waiman Long <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Jason Low <[email protected]>
Acked-by: Davidlohr Bueso <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Dave Chinner <[email protected]>
Cc: Douglas Hatch <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Hurley <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Scott J Norton <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
While going over the code I noticed that xchg_tail() is a RELEASE but
had no obvious pairing commented.
It pairs with a somewhat unique address dependency through
decode_tail().
So the store-release of xchg_tail() is paired by the address
dependency of the load of xchg_tail followed by the dereference from
the pointer computed from that load.
The @old -> @prev transformation itself is pure, and therefore does
not depend on external state, so that is immaterial wrt. ordering.
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Boqun Feng <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Pan Xinhui <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Waiman Long <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
While this prior commit:
54cf809b95 ("locking,qspinlock: Fix spin_is_locked() and spin_unlock_wait()")
... fixes spin_is_locked() and spin_unlock_wait() for the usage
in ipc/sem and netfilter, it does not in fact work right for the
usage in task_work and futex.
So while the 2 locks crossed problem:
spin_lock(A) spin_lock(B)
if (!spin_is_locked(B)) spin_unlock_wait(A)
foo() foo();
... works with the smp_mb() injected by both spin_is_locked() and
spin_unlock_wait(), this is not sufficient for:
flag = 1;
smp_mb(); spin_lock()
spin_unlock_wait() if (!flag)
// add to lockless list
// iterate lockless list
... because in this scenario, the store from spin_lock() can be delayed
past the load of flag, uncrossing the variables and loosing the
guarantee.
This patch reworks spin_is_locked() and spin_unlock_wait() to work in
both cases by exploiting the observation that while the lock byte
store can be delayed, the contender must have registered itself
visibly in other state contained in the word.
It also allows for architectures to override both functions, as PPC
and ARM64 have an additional issue for which we currently have no
generic solution.
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Boqun Feng <[email protected]>
Cc: Davidlohr Bueso <[email protected]>
Cc: Giovanni Gherdovich <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Pan Xinhui <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Waiman Long <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: [email protected] # v4.2 and later
Fixes: 54cf809b95 ("locking,qspinlock: Fix spin_is_locked() and spin_unlock_wait()")
Signed-off-by: Ingo Molnar <[email protected]>
Ingo Molnar