| Age | Commit message (Collapse) | Author |
|---|
|
continuously-running tasks
We typically update a task_group's shares within the dequeue/enqueue
path. However, continuously running tasks sharing a CPU are not
subject to these updates as they are only put/picked. Unfortunately,
when we reverted f269ae046 (in 17bc14b7), we lost the augmenting
periodic update that was supposed to account for this; resulting in a
potential loss of fairness.
To fix this, re-introduce the explicit update in
update_cfs_rq_blocked_load() [called via entity_tick()].
Reported-by: Max Hailperin <max@gustavus.edu>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: Paul Turner <pjt@google.com>
Link: http://lkml.kernel.org/n/tip-9545m3apw5d93ubyrotrj31y@git.kernel.org
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
The ARM architected system counter has at least 56 usable bits.
Add support for counters with more than 32 bits to the generic
sched_clock implementation so we can increase the time between
wakeups due to dealing with wrap-around on these devices while
benefiting from the irqtime accounting and suspend/resume
handling that the generic sched_clock code already has. On my
system using 56 bits over 32 bits changes the wraparound time
from a few minutes to an hour. For faster running counters (GHz
range) this is even more important because we may not be able to
execute the timer in time to deal with the wraparound if only 32
bits are used.
We choose a maxsec value of 3600 seconds because we assume no
system will go idle for more than an hour. In the future we may
need to increase this value.
Note: All users should switch over to the 64-bit read function so
we can remove setup_sched_clock() in favor of sched_clock_register().
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
In the next patch we're going to increase the number of bits that
the generic sched_clock can handle to be greater than 32. With
more than 32 bits the wraparound time can be larger than what can
fit into the units that msecs_to_jiffies takes (unsigned int).
Luckily, the wraparound is initially calculated in nanoseconds
which we can easily use with hrtimers, so switch to using an
hrtimer.
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
[jstultz: Fixup hrtimer intitialization order issue]
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
We're going to increase the cyc value to 64 bits in the near
future. Doing that is going to break the custom seqcount
implementation in the sched_clock code because 64 bit numbers
aren't guaranteed to be atomic. Replace the cyc_copy with a
seqcount to avoid this problem.
Cc: Russell King <linux@arm.linux.org.uk>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
We need to calculate the same number in the clocksource code and
the sched_clock code, so extract this code into its own function.
We also drop the min_t and just use min() because the two types
are the same.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
On Wed, Jun 12, 2013 at 11:14:40AM -0700, Kent Overstreet wrote:
> On Mon, Apr 15, 2013 at 02:40:55PM +0300, Octavian Purdila wrote:
> > When using a large number of threads performing AIO operations the
> > IOCTX list may get a significant number of entries which will cause
> > significant overhead. For example, when running this fio script:
> >
> > rw=randrw; size=256k ;directory=/mnt/fio; ioengine=libaio; iodepth=1
> > blocksize=1024; numjobs=512; thread; loops=100
> >
> > on an EXT2 filesystem mounted on top of a ramdisk we can observe up to
> > 30% CPU time spent by lookup_ioctx:
> >
> > 32.51% [guest.kernel] [g] lookup_ioctx
> > 9.19% [guest.kernel] [g] __lock_acquire.isra.28
> > 4.40% [guest.kernel] [g] lock_release
> > 4.19% [guest.kernel] [g] sched_clock_local
> > 3.86% [guest.kernel] [g] local_clock
> > 3.68% [guest.kernel] [g] native_sched_clock
> > 3.08% [guest.kernel] [g] sched_clock_cpu
> > 2.64% [guest.kernel] [g] lock_release_holdtime.part.11
> > 2.60% [guest.kernel] [g] memcpy
> > 2.33% [guest.kernel] [g] lock_acquired
> > 2.25% [guest.kernel] [g] lock_acquire
> > 1.84% [guest.kernel] [g] do_io_submit
> >
> > This patchs converts the ioctx list to a radix tree. For a performance
> > comparison the above FIO script was run on a 2 sockets 8 core
> > machine. This are the results (average and %rsd of 10 runs) for the
> > original list based implementation and for the radix tree based
> > implementation:
> >
> > cores 1 2 4 8 16 32
> > list 109376 ms 69119 ms 35682 ms 22671 ms 19724 ms 16408 ms
> > %rsd 0.69% 1.15% 1.17% 1.21% 1.71% 1.43%
> > radix 73651 ms 41748 ms 23028 ms 16766 ms 15232 ms 13787 ms
> > %rsd 1.19% 0.98% 0.69% 1.13% 0.72% 0.75%
> > % of radix
> > relative 66.12% 65.59% 66.63% 72.31% 77.26% 83.66%
> > to list
> >
> > To consider the impact of the patch on the typical case of having
> > only one ctx per process the following FIO script was run:
> >
> > rw=randrw; size=100m ;directory=/mnt/fio; ioengine=libaio; iodepth=1
> > blocksize=1024; numjobs=1; thread; loops=100
> >
> > on the same system and the results are the following:
> >
> > list 58892 ms
> > %rsd 0.91%
> > radix 59404 ms
> > %rsd 0.81%
> > % of radix
> > relative 100.87%
> > to list
>
> So, I was just doing some benchmarking/profiling to get ready to send
> out the aio patches I've got for 3.11 - and it looks like your patch is
> causing a ~1.5% throughput regression in my testing :/
... <snip>
I've got an alternate approach for fixing this wart in lookup_ioctx()...
Instead of using an rbtree, just use the reserved id in the ring buffer
header to index an array pointing the ioctx. It's not finished yet, and
it needs to be tidied up, but is most of the way there.
-ben
--
"Thought is the essence of where you are now."
--
kmo> And, a rework of Ben's code, but this was entirely his idea
kmo> -Kent
bcrl> And fix the code to use the right mm_struct in kill_ioctx(), actually
free memory.
Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
|
|
Calling freeze_processes sets a global flag that will cause any
process that calls try_to_freeze to enter the refrigerator. It
skips sending a signal to the current task, but if the current
task ever hits try_to_freeze, all threads will be frozen and the
system will deadlock.
Set a new flag, PF_SUSPEND_TASK, on the task that calls
freeze_processes. The flag notifies the freezer that the thread
is involved in suspend and should not be frozen. Also add a
WARN_ON in thaw_processes if the caller does not have the
PF_SUSPEND_TASK flag set to catch if a different task calls
thaw_processes than the one that called freeze_processes, leaving
a task with PF_SUSPEND_TASK permanently set on it.
Threads that spawn off a task with PF_SUSPEND_TASK set (which
swsusp does) will also have PF_SUSPEND_TASK set, preventing them
from freezing while they are helping with suspend, but they need
to be dead by the time suspend is triggered, otherwise they may
run when userspace is expected to be frozen. Add a WARN_ON in
thaw_processes if more than one thread has the PF_SUSPEND_TASK
flag set.
Reported-and-tested-by: Michael Leun <lkml20130126@newton.leun.net>
Signed-off-by: Colin Cross <ccross@android.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
When ftrace ops modifies the functions that it will trace, the update
to the function mcount callers may need to be modified. Consolidate
the two places that do the checks to see if an update is required
with a wrapper function for those checks.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Change remove_event_file_dir() to clear ->i_private for every
file we are going to remove.
We need to check file->dir != NULL because event_create_dir()
can fail. debugfs_remove_recursive(NULL) is fine but the patch
moves it under the same check anyway for readability.
spin_lock(d_lock) and "d_inode != NULL" check are not needed
afaics, but I do not understand this code enough.
tracing_open_generic_file() and tracing_release_generic_file()
can go away, ftrace_enable_fops and ftrace_event_filter_fops()
use tracing_open_generic() but only to check tracing_disabled.
This fixes all races with event_remove() or instance_delete().
f_op->read/write/whatever can never use the freed file/call,
all event/* files were changed to check and use ->i_private
under event_mutex.
Note: this doesn't not fix other problems, event_remove() can
destroy the active ftrace_event_call, we need more changes but
those changes are completely orthogonal.
Link: http://lkml.kernel.org/r/20130728183527.GB16723@redhat.com
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Preparation for the next patch. Extract the common code from
remove_event_from_tracers() and __trace_remove_event_dirs()
into the new helper, remove_event_file_dir().
The patch looks more complicated than it actually is, it also
moves remove_subsystem() up to avoid the forward declaration.
Link: http://lkml.kernel.org/r/20130726172547.GA3629@redhat.com
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
trace_format_open() and trace_format_seq_ops are racy, nothing
protects ftrace_event_call from trace_remove_event_call().
Change f_start() to take event_mutex and verify i_private != NULL,
change f_stop() to drop this lock.
This fixes nothing, but now we can change debugfs_remove("format")
callers to nullify ->i_private and fix the the problem.
Note: the usage of event_mutex is sub-optimal but simple, we can
change this later.
Link: http://lkml.kernel.org/r/20130726172543.GA3622@redhat.com
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
event_filter_read/write() are racy, ftrace_event_call can be already
freed by trace_remove_event_call() callers.
1. Shift mutex_lock(event_mutex) from print/apply_event_filter to
the callers.
2. Change the callers, event_filter_read() and event_filter_write()
to read i_private under this mutex and abort if it is NULL.
This fixes nothing, but now we can change debugfs_remove("filter")
callers to nullify ->i_private and fix the the problem.
Link: http://lkml.kernel.org/r/20130726172540.GA3619@redhat.com
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
tracing_open_generic_file() is racy, ftrace_event_file can be
already freed by rmdir or trace_remove_event_call().
Change event_enable_read() and event_disable_read() to read and
verify "file = i_private" under event_mutex.
This fixes nothing, but now we can change debugfs_remove("enable")
callers to nullify ->i_private and fix the the problem.
Link: http://lkml.kernel.org/r/20130726172536.GA3612@redhat.com
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
event_id_read() is racy, ftrace_event_call can be already freed
by trace_remove_event_call() callers.
Change event_create_dir() to pass "data = call->event.type", this
is all event_id_read() needs. ftrace_event_id_fops no longer needs
tracing_open_generic().
We add the new helper, event_file_data(), to read ->i_private, it
will have more users.
Note: currently ACCESS_ONCE() and "id != 0" check are not needed,
but we are going to change event_remove/rmdir to clear ->i_private.
Link: http://lkml.kernel.org/r/20130726172532.GA3605@redhat.com
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Currently, RCU tracepoints save only a pointer to strings in the
ring buffer. When displayed via the /sys/kernel/debug/tracing/trace file
they are referenced like the printf "%s" that looks at the address
in the ring buffer and prints out the string it points too. This requires
that the strings are constant and persistent in the kernel.
The problem with this is for tools like trace-cmd and perf that read the
binary data from the buffers but have no access to the kernel memory to
find out what string is represented by the address in the buffer.
By using the tracepoint_string infrastructure, the RCU tracepoint strings
can be exported such that userspace tools can map the addresses to
the strings.
# cat /sys/kernel/debug/tracing/printk_formats
0xffffffff81a4a0e8 : "rcu_preempt"
0xffffffff81a4a0f4 : "rcu_bh"
0xffffffff81a4a100 : "rcu_sched"
0xffffffff818437a0 : "cpuqs"
0xffffffff818437a6 : "rcu_sched"
0xffffffff818437a0 : "cpuqs"
0xffffffff818437b0 : "rcu_bh"
0xffffffff818437b7 : "Start context switch"
0xffffffff818437cc : "End context switch"
0xffffffff818437a0 : "cpuqs"
[...]
Now userspaces tools can display:
rcu_utilization: Start context switch
rcu_dyntick: Start 1 0
rcu_utilization: End context switch
rcu_batch_start: rcu_preempt CBs=0/5 bl=10
rcu_dyntick: End 0 140000000000000
rcu_invoke_callback: rcu_preempt rhp=0xffff880071c0d600 func=proc_i_callback
rcu_invoke_callback: rcu_preempt rhp=0xffff880077b5b230 func=__d_free
rcu_dyntick: Start 140000000000000 0
rcu_invoke_callback: rcu_preempt rhp=0xffff880077563980 func=file_free_rcu
rcu_batch_end: rcu_preempt CBs-invoked=3 idle=>c<>c<>c<>c<
rcu_utilization: End RCU core
rcu_grace_period: rcu_preempt 9741 start
rcu_dyntick: Start 1 0
rcu_dyntick: End 0 140000000000000
rcu_dyntick: Start 140000000000000 0
Instead of:
rcu_utilization: ffffffff81843110
rcu_future_grace_period: ffffffff81842f1d 9939 9939 9940 0 0 3 ffffffff81842f32
rcu_batch_start: ffffffff81842f1d CBs=0/4 bl=10
rcu_future_grace_period: ffffffff81842f1d 9939 9939 9940 0 0 3 ffffffff81842f3c
rcu_grace_period: ffffffff81842f1d 9939 ffffffff81842f80
rcu_invoke_callback: ffffffff81842f1d rhp=0xffff88007888aac0 func=file_free_rcu
rcu_grace_period: ffffffff81842f1d 9939 ffffffff81842f95
rcu_invoke_callback: ffffffff81842f1d rhp=0xffff88006aeb4600 func=proc_i_callback
rcu_future_grace_period: ffffffff81842f1d 9939 9939 9940 0 0 3 ffffffff81842f32
rcu_future_grace_period: ffffffff81842f1d 9939 9939 9940 0 0 3 ffffffff81842f3c
rcu_invoke_callback: ffffffff81842f1d rhp=0xffff880071cb9fc0 func=__d_free
rcu_grace_period: ffffffff81842f1d 9939 ffffffff81842f80
rcu_invoke_callback: ffffffff81842f1d rhp=0xffff88007888ae80 func=file_free_rcu
rcu_batch_end: ffffffff81842f1d CBs-invoked=4 idle=>c<>c<>c<>c<
rcu_utilization: ffffffff8184311f
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
The RCU_STATE_INITIALIZER() macro is used only in the rcutree.c file
as well as the rcutree_plugin.h file. It is passed as a rvalue to
a variable of a similar name. A per_cpu variable is also created
with a similar name as well.
The uses of RCU_STATE_INITIALIZER() can be simplified to remove some
of the duplicate code that is done. Currently the three users of this
macro has this format:
struct rcu_state rcu_sched_state =
RCU_STATE_INITIALIZER(rcu_sched, call_rcu_sched);
DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
Notice that "rcu_sched" is called three times. This is the same with
the other two users. This can be condensed to just:
RCU_STATE_INITIALIZER(rcu_sched, call_rcu_sched);
by moving the rest into the macro itself.
This also opens the door to allow the RCU tracepoint strings and
their addresses to be exported so that userspace tracing tools can
translate the contents of the pointers of the RCU tracepoints.
The change will allow for helper code to be placed in the
RCU_STATE_INITIALIZER() macro to export the name that is used.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
All the RCU tracepoints and functions that reference char pointers do
so with just 'char *' even though they do not modify the contents of
the string itself. This will cause warnings if a const char * is used
in one of these functions.
The RCU tracepoints store the pointer to the string to refer back to them
when the trace output is displayed. As this can be minutes, hours or
even days later, those strings had better be constant.
This change also opens the door to allow the RCU tracepoint strings and
their addresses to be exported so that userspace tracing tools can
translate the contents of the pointers of the RCU tracepoints.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Comment for cpuset_css_offline() was on top of cpuset_css_free().
Move it.
Signed-off-by: Zhao Hongjiang <zhaohongjiang@huawei.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
get rid of the useless forward declaration of the struct cpuset cause the
below define it.
Signed-off-by: Zhao Hongjiang <zhaohongjiang@huawei.com>
Acked-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Revert commit 69a37bea (cpuidle: Quickly notice prediction failure for
repeat mode), because it has been identified as the source of a
significant performance regression in v3.8 and later as explained by
Jeremy Eder:
We believe we've identified a particular commit to the cpuidle code
that seems to be impacting performance of variety of workloads.
The simplest way to reproduce is using netperf TCP_RR test, so
we're using that, on a pair of Sandy Bridge based servers. We also
have data from a large database setup where performance is also
measurably/positively impacted, though that test data isn't easily
share-able.
Included below are test results from 3 test kernels:
kernel reverts
-----------------------------------------------------------
1) vanilla upstream (no reverts)
2) perfteam2 reverts e11538d1f03914eb92af5a1a378375c05ae8520c
3) test reverts 69a37beabf1f0a6705c08e879bdd5d82ff6486c4
e11538d1f03914eb92af5a1a378375c05ae8520c
In summary, netperf TCP_RR numbers improve by approximately 4%
after reverting 69a37beabf1f0a6705c08e879bdd5d82ff6486c4. When
69a37beabf1f0a6705c08e879bdd5d82ff6486c4 is included, C0 residency
never seems to get above 40%. Taking that patch out gets C0 near
100% quite often, and performance increases.
The below data are histograms representing the %c0 residency @
1-second sample rates (using turbostat), while under netperf test.
- If you look at the first 4 histograms, you can see %c0 residency
almost entirely in the 30,40% bin.
- The last pair, which reverts 69a37beabf1f0a6705c08e879bdd5d82ff6486c4,
shows %c0 in the 80,90,100% bins.
Below each kernel name are netperf TCP_RR trans/s numbers for the
particular kernel that can be disclosed publicly, comparing the 3
test kernels. We ran a 4th test with the vanilla kernel where
we've also set /dev/cpu_dma_latency=0 to show overall impact
boosting single-threaded TCP_RR performance over 11% above
baseline.
3.10-rc2 vanilla RX + c0 lock (/dev/cpu_dma_latency=0):
TCP_RR trans/s 54323.78
-----------------------------------------------------------
3.10-rc2 vanilla RX (no reverts)
TCP_RR trans/s 48192.47
Receiver %c0
0.0000 - 10.0000 [ 1]: *
10.0000 - 20.0000 [ 0]:
20.0000 - 30.0000 [ 0]:
30.0000 - 40.0000 [ 59]:
***********************************************************
40.0000 - 50.0000 [ 1]: *
50.0000 - 60.0000 [ 0]:
60.0000 - 70.0000 [ 0]:
70.0000 - 80.0000 [ 0]:
80.0000 - 90.0000 [ 0]:
90.0000 - 100.0000 [ 0]:
Sender %c0
0.0000 - 10.0000 [ 1]: *
10.0000 - 20.0000 [ 0]:
20.0000 - 30.0000 [ 0]:
30.0000 - 40.0000 [ 11]: ***********
40.0000 - 50.0000 [ 49]:
*************************************************
50.0000 - 60.0000 [ 0]:
60.0000 - 70.0000 [ 0]:
70.0000 - 80.0000 [ 0]:
80.0000 - 90.0000 [ 0]:
90.0000 - 100.0000 [ 0]:
-----------------------------------------------------------
3.10-rc2 perfteam2 RX (reverts commit
e11538d1f03914eb92af5a1a378375c05ae8520c)
TCP_RR trans/s 49698.69
Receiver %c0
0.0000 - 10.0000 [ 1]: *
10.0000 - 20.0000 [ 1]: *
20.0000 - 30.0000 [ 0]:
30.0000 - 40.0000 [ 59]:
***********************************************************
40.0000 - 50.0000 [ 0]:
50.0000 - 60.0000 [ 0]:
60.0000 - 70.0000 [ 0]:
70.0000 - 80.0000 [ 0]:
80.0000 - 90.0000 [ 0]:
90.0000 - 100.0000 [ 0]:
Sender %c0
0.0000 - 10.0000 [ 1]: *
10.0000 - 20.0000 [ 0]:
20.0000 - 30.0000 [ 0]:
30.0000 - 40.0000 [ 2]: **
40.0000 - 50.0000 [ 58]:
**********************************************************
50.0000 - 60.0000 [ 0]:
60.0000 - 70.0000 [ 0]:
70.0000 - 80.0000 [ 0]:
80.0000 - 90.0000 [ 0]:
90.0000 - 100.0000 [ 0]:
-----------------------------------------------------------
3.10-rc2 test RX (reverts 69a37beabf1f0a6705c08e879bdd5d82ff6486c4
and e11538d1f03914eb92af5a1a378375c05ae8520c)
TCP_RR trans/s 47766.95
Receiver %c0
0.0000 - 10.0000 [ 1]: *
10.0000 - 20.0000 [ 1]: *
20.0000 - 30.0000 [ 0]:
30.0000 - 40.0000 [ 27]: ***************************
40.0000 - 50.0000 [ 2]: **
50.0000 - 60.0000 [ 0]:
60.0000 - 70.0000 [ 2]: **
70.0000 - 80.0000 [ 0]:
80.0000 - 90.0000 [ 0]:
90.0000 - 100.0000 [ 28]: ****************************
Sender:
0.0000 - 10.0000 [ 1]: *
10.0000 - 20.0000 [ 0]:
20.0000 - 30.0000 [ 0]:
30.0000 - 40.0000 [ 11]: ***********
40.0000 - 50.0000 [ 0]:
50.0000 - 60.0000 [ 1]: *
60.0000 - 70.0000 [ 0]:
70.0000 - 80.0000 [ 3]: ***
80.0000 - 90.0000 [ 7]: *******
90.0000 - 100.0000 [ 38]: **************************************
These results demonstrate gaining back the tendency of the CPU to
stay in more responsive, performant C-states (and thus yield
measurably better performance), by reverting commit
69a37beabf1f0a6705c08e879bdd5d82ff6486c4.
Requested-by: Jeremy Eder <jeder@redhat.com>
Tested-by: Len Brown <len.brown@intel.com>
Cc: 3.8+ <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing fixes from Steven Rostedt:
"Oleg is working on fixing a very tight race between opening a event
file and deleting that event at the same time (both must be done as
root).
I also found a bug while testing Oleg's patches which has to do with a
race with kprobes using the function tracer.
There's also a deadlock fix that was introduced with the previous
fixes"
* tag 'trace-fixes-3.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Remove locking trace_types_lock from tracing_reset_all_online_cpus()
ftrace: Add check for NULL regs if ops has SAVE_REGS set
tracing: Kill trace_cpu struct/members
tracing: Change tracing_fops/snapshot_fops to rely on tracing_get_cpu()
tracing: Change tracing_entries_fops to rely on tracing_get_cpu()
tracing: Change tracing_stats_fops to rely on tracing_get_cpu()
tracing: Change tracing_buffers_fops to rely on tracing_get_cpu()
tracing: Change tracing_pipe_fops() to rely on tracing_get_cpu()
tracing: Introduce trace_create_cpu_file() and tracing_get_cpu()
|
|
When (integer) sysctl values are expressed in ms and have to be
represented internally as jiffies. The msecs_to_jiffies function
returns an unsigned long, which gets assigned to the integer.
This patch prevents the value to be assigned if bigger than
INT_MAX, done in a similar way as in cba9f3 ("Range checking in
do_proc_dointvec_(userhz_)jiffies_conv").
Signed-off-by: Francesco Fusco <ffusco@redhat.com>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: linux-kernel@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There are several tracepoints (mostly in RCU), that reference a string
pointer and uses the print format of "%s" to display the string that
exists in the kernel, instead of copying the actual string to the
ring buffer (saves time and ring buffer space).
But this has an issue with userspace tools that read the binary buffers
that has the address of the string but has no access to what the string
itself is. The end result is just output that looks like:
rcu_dyntick: ffffffff818adeaa 1 0
rcu_dyntick: ffffffff818adeb5 0 140000000000000
rcu_dyntick: ffffffff818adeb5 0 140000000000000
rcu_utilization: ffffffff8184333b
rcu_utilization: ffffffff8184333b
The above is pretty useless when read by the userspace tools. Ideally
we would want something that looks like this:
rcu_dyntick: Start 1 0
rcu_dyntick: End 0 140000000000000
rcu_dyntick: Start 140000000000000 0
rcu_callback: rcu_preempt rhp=0xffff880037aff710 func=put_cred_rcu 0/4
rcu_callback: rcu_preempt rhp=0xffff880078961980 func=file_free_rcu 0/5
rcu_dyntick: End 0 1
The trace_printk() which also only stores the address of the string
format instead of recording the string into the buffer itself, exports
the mapping of kernel addresses to format strings via the printk_format
file in the debugfs tracing directory.
The tracepoint strings can use this same method and output the format
to the same file and the userspace tools will be able to decipher
the address without any modification.
The tracepoint strings need its own section to save the strings because
the trace_printk section will cause the trace_printk() buffers to be
allocated if anything exists within the section. trace_printk() is only
used for debugging and should never exist in the kernel, we can not use
the trace_printk sections.
Add a new tracepoint_str section that will also be examined by the output
of the printk_format file.
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Commit a82274151af "tracing: Protect ftrace_trace_arrays list in trace_events.c"
added taking the trace_types_lock mutex in trace_events.c as there were
several locations that needed it for protection. Unfortunately, it also
encapsulated a call to tracing_reset_all_online_cpus() which also takes
the trace_types_lock, causing a deadlock.
This happens when a module has tracepoints and has been traced. When the
module is removed, the trace events module notifier will grab the
trace_types_lock, do a bunch of clean ups, and also clears the buffer
by calling tracing_reset_all_online_cpus. This doesn't happen often
which explains why it wasn't caught right away.
Commit a82274151af was marked for stable, which means this must be
sent to stable too.
Link: http://lkml.kernel.org/r/51EEC646.7070306@broadcom.com
Reported-by: Arend van Spril <arend@broadcom.com>
Tested-by: Arend van Spriel <arend@broadcom.com>
Cc: Alexander Z Lam <azl@google.com>
Cc: Vaibhav Nagarnaik <vnagarnaik@google.com>
Cc: David Sharp <dhsharp@google.com>
Cc: stable@vger.kernel.org # 3.10
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Change where ftrace is disabled and re-enabled during system
suspend/resume to allow tracing of device driver pm callbacks.
Ftrace will now be turned off when suspend reaches
disable_nonboot_cpus() instead of at the very beginning of system
suspend.
Ftrace was disabled during suspend/resume back in 2008 by
Steven Rostedt as he discovered there was a conflict in the
enable_nonboot_cpus() call (see commit f42ac38 "ftrace: disable
tracing for suspend to ram"). This change preserves his fix by
disabling ftrace, but only at the function where it is known
to cause problems.
The new change allows tracing of the device level code for better
debug.
[rjw: Changelog]
Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
|
|
Fengguang reported the following warning when optimistic
spinning is disabled (ie: make allnoconfig):
kernel/mutex.c:599:1: warning: label 'done' defined but not used
Remove the 'done' label altogether.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks into timers/urgent
Pull nohz fixes from Frederic Weisbecker.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
cpu is not used after commit 5b8621a68fdcd2baf1d3b413726f913a5254d46a
Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Li Zhong <zhong@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Kevin Hilman <khilman@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
|
|
If the user enables CONFIG_NO_HZ_FULL and runs the kernel on a machine
with an unstable TSC, it will produce a WARN_ON dump as well as taint
the kernel. This is a bit extreme for a kernel that just enables a
feature but doesn't use it.
The warning should only happen if the user tries to use the feature by
either adding nohz_full to the kernel command line, or by enabling
CONFIG_NO_HZ_FULL_ALL that makes nohz used on all CPUs at boot up. Note,
this second feature should not (yet) be used by distros or anyone that
doesn't care if NO_HZ is used or not.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Li Zhong <zhong@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Kevin Hilman <khilman@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
|
|
If the @fn call work_on_cpu() again, the lockdep will complain:
> [ INFO: possible recursive locking detected ]
> 3.11.0-rc1-lockdep-fix-a #6 Not tainted
> ---------------------------------------------
> kworker/0:1/142 is trying to acquire lock:
> ((&wfc.work)){+.+.+.}, at: [<ffffffff81077100>] flush_work+0x0/0xb0
>
> but task is already holding lock:
> ((&wfc.work)){+.+.+.}, at: [<ffffffff81075dd9>] process_one_work+0x169/0x610
>
> other info that might help us debug this:
> Possible unsafe locking scenario:
>
> CPU0
> ----
> lock((&wfc.work));
> lock((&wfc.work));
>
> *** DEADLOCK ***
It is false-positive lockdep report. In this sutiation,
the two "wfc"s of the two work_on_cpu() are different,
they are both on stack. flush_work() can't be deadlock.
To fix this, we need to avoid the lockdep checking in this case,
thus we instroduce a internal __flush_work() which skip the lockdep.
tj: Minor comment adjustment.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Reported-by: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com>
Reported-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
If a ftrace ops is registered with the SAVE_REGS flag set, and there's
already a ops registered to one of its functions but without the
SAVE_REGS flag, there's a small race window where the SAVE_REGS ops gets
added to the list of callbacks to call for that function before the
callback trampoline gets set to save the regs.
The problem is, the function is not currently saving regs, which opens
a small race window where the ops that is expecting regs to be passed
to it, wont. This can cause a crash if the callback were to reference
the regs, as the SAVE_REGS guarantees that regs will be set.
To fix this, we add a check in the loop case where it checks if the ops
has the SAVE_REGS flag set, and if so, it will ignore it if regs is
not set.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
After the previous changes trace_array_cpu->trace_cpu and
trace_array->trace_cpu becomes write-only. Remove these members
and kill "struct trace_cpu" as well.
As a side effect this also removes memset(per_cpu_memory, 0).
It was not needed, alloc_percpu() returns zero-filled memory.
Link: http://lkml.kernel.org/r/20130723152613.GA23741@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
tracing_open() and tracing_snapshot_open() are racy, the memory
inode->i_private points to can be already freed.
Convert these last users of "inode->i_private == trace_cpu" to
use "i_private = trace_array" and rely on tracing_get_cpu().
v2: incorporate the fix from Steven, tracing_release() must not
blindly dereference file->private_data unless we know that
the file was opened for reading.
Link: http://lkml.kernel.org/r/20130723152610.GA23737@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
tracing_open_generic_tc() is racy, the memory inode->i_private
points to can be already freed.
1. Change its last user, tracing_entries_fops, to use
tracing_*_generic_tr() instead.
2. Change debugfs_create_file("buffer_size_kb", data) callers
to pass "data = tr".
3. Change tracing_entries_read() and tracing_entries_write() to
use tracing_get_cpu().
4. Kill the no longer used tracing_open_generic_tc() and
tracing_release_generic_tc().
Link: http://lkml.kernel.org/r/20130723152606.GA23730@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
tracing_open_generic_tc() is racy, the memory inode->i_private
points to can be already freed.
1. Change one of its users, tracing_stats_fops, to use
tracing_*_generic_tr() instead.
2. Change trace_create_cpu_file("stats", data) to pass "data = tr".
3. Change tracing_stats_read() to use tracing_get_cpu().
Link: http://lkml.kernel.org/r/20130723152603.GA23727@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
tracing_buffers_open() is racy, the memory inode->i_private points
to can be already freed.
Change debugfs_create_file("trace_pipe_raw", data) caller to pass
"data = tr", tracing_buffers_open() can use tracing_get_cpu().
Change debugfs_create_file("snapshot_raw_fops", data) caller too,
this file uses tracing_buffers_open/release.
Link: http://lkml.kernel.org/r/20130723152600.GA23720@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
tracing_open_pipe() is racy, the memory inode->i_private points to
can be already freed.
Change debugfs_create_file("trace_pipe", data) callers to to pass
"data = tr", tracing_open_pipe() can use tracing_get_cpu().
Link: http://lkml.kernel.org/r/20130723152557.GA23717@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
Every "file_operations" used by tracing_init_debugfs_percpu is buggy.
f_op->open/etc does:
1. struct trace_cpu *tc = inode->i_private;
struct trace_array *tr = tc->tr;
2. trace_array_get(tr) or fail;
3. do_something(tc);
But tc (and tr) can be already freed before trace_array_get() is called.
And it doesn't matter whether this file is per-cpu or it was created by
init_tracer_debugfs(), free_percpu() or kfree() are equally bad.
Note that even 1. is not safe, the freed memory can be unmapped. But even
if it was safe trace_array_get() can wrongly succeed if we also race with
the next new_instance_create() which can re-allocate the same tr, or tc
was overwritten and ->tr points to the valid tr. In this case 3. uses the
freed/reused memory.
Add the new trivial helper, trace_create_cpu_file() which simply calls
trace_create_file() and encodes "cpu" in "struct inode". Another helper,
tracing_get_cpu() will be used to read cpu_nr-or-RING_BUFFER_ALL_CPUS.
The patch abuses ->i_cdev to encode the number, it is never used unless
the file is S_ISCHR(). But we could use something else, say, i_bytes or
even ->d_fsdata. In any case this hack is hidden inside these 2 helpers,
it would be trivial to change them if needed.
This patch only changes tracing_init_debugfs_percpu() to use the new
trace_create_cpu_file(), the next patches will change file_operations.
Note: tracing_get_cpu(inode) is always safe but you can't trust the
result unless trace_array_get() was called, without trace_types_lock
which acts as a barrier it can wrongly return RING_BUFFER_ALL_CPUS.
Link: http://lkml.kernel.org/r/20130723152554.GA23710@redhat.com
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup changes from Tejun Heo:
"This contains two patches, both of which aren't fixes per-se but I
think it'd be better to fast-track them.
One removes bcache_subsys_id which was added without proper review
through the block tree. Fortunately, bcache cgroup code is
unconditionally disabled, so this was never exposed to userland. The
cgroup subsys_id is removed. Kent will remove the affected (disabled)
code through bcache branch.
The other simplifies task_group_path_from_hierarchy(). The function
doesn't currently have in-kernel users but there are external code and
development going on dependent on the function and making the function
available for 3.11 would make things go smoother"
* 'for-3.11-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup: replace task_cgroup_path_from_hierarchy() with task_cgroup_path()
cgroup: remove bcache_subsys_id which got added stealthily
|
|
Fix __wait_on_atomic_t() so that it calls the action func if the counter != 0
rather than if the counter is 0 so as to be analogous to __wait_on_bit().
Thanks to Yacine who found this by visual inspection.
This will affect FS-Cache in that it will could fail to sleep correctly when
trying to clean up after a netfs cookie is withdrawn.
Reported-by: Yacine Belkadi <yacine.belkadi.1@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
cc: Milosz Tanski <milosz@adfin.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Smart wake-affine is using node-size as the factor currently, but the overhead
of the mask operation is high.
Thus, this patch introduce the 'sd_llc_size' percpu variable, which will record
the highest cache-share domain size, and make it to be the new factor, in order
to reduce the overhead and make it more reasonable.
Tested-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Link: http://lkml.kernel.org/r/51D5008E.6030102@linux.vnet.ibm.com
[ Tidied up the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
The wake-affine scheduler feature is currently always trying to pull
the wakee close to the waker. In theory this should be beneficial if
the waker's CPU caches hot data for the wakee, and it's also beneficial
in the extreme ping-pong high context switch rate case.
Testing shows it can benefit hackbench up to 15%.
However, the feature is somewhat blind, from which some workloads
such as pgbench suffer. It's also time-consuming algorithmically.
Testing shows it can damage pgbench up to 50% - far more than the
benefit it brings in the best case.
So wake-affine should be smarter and it should realize when to
stop its thankless effort at trying to find a suitable CPU to wake on.
This patch introduces 'wakee_flips', which will be increased each
time the task flips (switches) its wakee target.
So a high 'wakee_flips' value means the task has more than one
wakee, and the bigger the number, the higher the wakeup frequency.
Now when making the decision on whether to pull or not, pay attention to
the wakee with a high 'wakee_flips', pulling such a task may benefit
the wakee. Also imply that the waker will face cruel competition later,
it could be very cruel or very fast depends on the story behind
'wakee_flips', waker therefore suffers.
Furthermore, if waker also has a high 'wakee_flips', that implies that
multiple tasks rely on it, then waker's higher latency will damage all
of them, so pulling wakee seems to be a bad deal.
Thus, when 'waker->wakee_flips / wakee->wakee_flips' becomes
higher and higher, the cost of pulling seems to be worse and worse.
The patch therefore helps the wake-affine feature to stop its pulling
work when:
wakee->wakee_flips > factor &&
waker->wakee_flips > (factor * wakee->wakee_flips)
The 'factor' here is the number of CPUs in the current CPU's NUMA node,
so a bigger node will lead to more pulling since the trial becomes more
severe.
After applying the patch, pgbench shows up to 40% improvements and no regressions.
Tested with 12 cpu x86 server and tip 3.10.0-rc7.
The percentages in the final column highlight the areas with the biggest wins,
all other areas improved as well:
pgbench base smart
| db_size | clients | tps | | tps |
+---------+---------+-------+ +-------+
| 22 MB | 1 | 10598 | | 10796 |
| 22 MB | 2 | 21257 | | 21336 |
| 22 MB | 4 | 41386 | | 41622 |
| 22 MB | 8 | 51253 | | 57932 |
| 22 MB | 12 | 48570 | | 54000 |
| 22 MB | 16 | 46748 | | 55982 | +19.75%
| 22 MB | 24 | 44346 | | 55847 | +25.93%
| 22 MB | 32 | 43460 | | 54614 | +25.66%
| 7484 MB | 1 | 8951 | | 9193 |
| 7484 MB | 2 | 19233 | | 19240 |
| 7484 MB | 4 | 37239 | | 37302 |
| 7484 MB | 8 | 46087 | | 50018 |
| 7484 MB | 12 | 42054 | | 48763 |
| 7484 MB | 16 | 40765 | | 51633 | +26.66%
| 7484 MB | 24 | 37651 | | 52377 | +39.11%
| 7484 MB | 32 | 37056 | | 51108 | +37.92%
| 15 GB | 1 | 8845 | | 9104 |
| 15 GB | 2 | 19094 | | 19162 |
| 15 GB | 4 | 36979 | | 36983 |
| 15 GB | 8 | 46087 | | 49977 |
| 15 GB | 12 | 41901 | | 48591 |
| 15 GB | 16 | 40147 | | 50651 | +26.16%
| 15 GB | 24 | 37250 | | 52365 | +40.58%
| 15 GB | 32 | 36470 | | 50015 | +37.14%
Signed-off-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/51D50057.9000809@linux.vnet.ibm.com
[ Improved the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
The bad thing about update_h_load(), which computes hierarchical load
factor for task groups, is that it is called for each task group in the
system before every load balancer run, and since rebalance can be
triggered very often, this function can eat really a lot of cpu time if
there are many cpu cgroups in the system.
Although the situation was improved significantly by commit a35b646
('sched, cgroup: Reduce rq->lock hold times for large cgroup
hierarchies'), the problem still can arise under some kinds of loads,
e.g. when cpus are switching from idle to busy and back very frequently.
For instance, when I start 1000 of processes that wake up every
millisecond on my 8 cpus host, 'top' and 'perf top' show:
Cpu(s): 17.8%us, 24.3%sy, 0.0%ni, 57.9%id, 0.0%wa, 0.0%hi, 0.0%si
Events: 243K cycles
7.57% [kernel] [k] __schedule
7.08% [kernel] [k] timerqueue_add
6.13% libc-2.12.so [.] usleep
Then if I create 10000 *idle* cpu cgroups (no processes in them), cpu
usage increases significantly although the 'wakers' are still executing
in the root cpu cgroup:
Cpu(s): 19.1%us, 48.7%sy, 0.0%ni, 31.6%id, 0.0%wa, 0.0%hi, 0.7%si
Events: 230K cycles
24.56% [kernel] [k] tg_load_down
5.76% [kernel] [k] __schedule
This happens because this particular kind of load triggers 'new idle'
rebalance very frequently, which requires calling update_h_load(),
which, in turn, calls tg_load_down() for every *idle* cpu cgroup even
though it is absolutely useless, because idle cpu cgroups have no tasks
to pull.
This patch tries to improve the situation by making h_load calculation
proceed only when h_load is really necessary. To achieve this, it
substitutes update_h_load() with update_cfs_rq_h_load(), which computes
h_load only for a given cfs_rq and all its ascendants, and makes the
load balancer call this function whenever it considers if a task should
be pulled, i.e. it moves h_load calculations directly to task_h_load().
For h_load of the same cfs_rq not to be updated multiple times (in case
several tasks in the same cgroup are considered during the same balance
run), the patch keeps the time of the last h_load update for each cfs_rq
and breaks calculation when it finds h_load to be uptodate.
The benefit of it is that h_load is computed only for those cfs_rq's,
which really need it, in particular all idle task groups are skipped.
Although this, in fact, moves h_load calculation under rq lock, it
should not affect latency much, because the amount of work done under rq
lock while trying to pull tasks is limited by sched_nr_migrate.
After the patch applied with the setup described above (1000 wakers in
the root cgroup and 10000 idle cgroups), I get:
Cpu(s): 16.9%us, 24.8%sy, 0.0%ni, 58.4%id, 0.0%wa, 0.0%hi, 0.0%si
Events: 242K cycles
7.57% [kernel] [k] __schedule
6.70% [kernel] [k] timerqueue_add
5.93% libc-2.12.so [.] usleep
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1373896159-1278-1-git-send-email-vdavydov@parallels.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
Due to a discussion with Adrian I had a good look at the perf_event_type record
layout and found the documentation to be somewhat unclear.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20130716150907.GL23818@dyad.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
Upon entering the slowpath, we immediately attempt to acquire
the lock by checking if it is already unlocked. If we are lucky
enough that this is the case, then we don't need to deal with
any waiter related logic.
Furthermore any checks for an empty wait_list are unnecessary as
we already know that count is non-negative and hence no one is
waiting for the lock.
Move the count check and xchg calls to be done before any
waiters are setup - including waiter debugging. Upon failure to
acquire the lock, the xchg sets the counter to 0, instead of -1
as it was originally. This can be done here since we set it back
to -1 right at the beginning of the loop so other waiters are
woken up when the lock is released.
When tested on a 8-socket (80 core) system against a vanilla
3.10-rc1 kernel, this patch provides some small performance
benefits (+2-6%). While it could be considered in the noise
level, the average percentages were stable across multiple runs
and no performance regressions were seen. Two big winners, for
small amounts of users (10-100), were the short and compute
workloads had a +19.36% and +%15.76% in jobs per minute.
Also change some break statements to 'goto slowpath', which IMO
makes a little more intuitive to read.
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1372450398.2106.1.camel@buesod1.americas.hpqcorp.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
Merge in Linux 3.11-rc2 before moving on with new work.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
In fd4363fff3d96 ("x86: Introduce int3 (breakpoint)-based
instruction patching"), the mechanism that was introduced for
notifying alternatives code from int3 exception handler that and
exception occured was die_notifier.
This is however problematic, as early code might be using jump
labels even before the notifier registration has been performed,
which will then lead to an oops due to unhandled exception. One
of such occurences has been encountered by Fengguang:
int3: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
Modules linked in:
CPU: 1 PID: 0 Comm: swapper/1 Not tainted 3.11.0-rc1-01429-g04bf576 #8
task: ffff88000da1b040 ti: ffff88000da1c000 task.ti: ffff88000da1c000
RIP: 0010:[<ffffffff811098cc>] [<ffffffff811098cc>] ttwu_do_wakeup+0x28/0x225
RSP: 0000:ffff88000dd03f10 EFLAGS: 00000006
RAX: 0000000000000000 RBX: ffff88000dd12940 RCX: ffffffff81769c40
RDX: 0000000000000002 RSI: 0000000000000000 RDI: 0000000000000001
RBP: ffff88000dd03f28 R08: ffffffff8176a8c0 R09: 0000000000000002
R10: ffffffff810ff484 R11: ffff88000dd129e8 R12: ffff88000dbc90c0
R13: ffff88000dbc90c0 R14: ffff88000da1dfd8 R15: ffff88000da1dfd8
FS: 0000000000000000(0000) GS:ffff88000dd00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00000000ffffffff CR3: 0000000001c88000 CR4: 00000000000006e0
Stack:
ffff88000dd12940 ffff88000dbc90c0 ffff88000da1dfd8 ffff88000dd03f48
ffffffff81109e2b ffff88000dd12940 0000000000000000 ffff88000dd03f68
ffffffff81109e9e 0000000000000000 0000000000012940 ffff88000dd03f98
Call Trace:
<IRQ>
[<ffffffff81109e2b>] ttwu_do_activate.constprop.56+0x6d/0x79
[<ffffffff81109e9e>] sched_ttwu_pending+0x67/0x84
[<ffffffff8110c845>] scheduler_ipi+0x15a/0x2b0
[<ffffffff8104dfb4>] smp_reschedule_interrupt+0x38/0x41
[<ffffffff8173bf5d>] reschedule_interrupt+0x6d/0x80
<EOI>
[<ffffffff810ff484>] ? __atomic_notifier_call_chain+0x5/0xc1
[<ffffffff8105cc30>] ? native_safe_halt+0xd/0x16
[<ffffffff81015f10>] default_idle+0x147/0x282
[<ffffffff81017026>] arch_cpu_idle+0x3d/0x5d
[<ffffffff81127d6a>] cpu_idle_loop+0x46d/0x5db
[<ffffffff81127f5c>] cpu_startup_entry+0x84/0x84
[<ffffffff8104f4f8>] start_secondary+0x3c8/0x3d5
[...]
Fix this by directly calling poke_int3_handler() from the int3
exception handler (analogically to what ftrace has been doing
already), instead of relying on notifier, registration of which
might not have yet been finalized by the time of the first trap.
Reported-and-tested-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/alpine.LNX.2.00.1307231007490.14024@pobox.suse.cz
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing fixes and cleanups from Steven Rostedt:
"This contains fixes, optimizations and some clean ups
Some of the fixes need to go back to 3.10. They are minor, and deal
mostly with incorrect ref counting in accessing event files.
There was a couple of optimizations that should have perf perform a
bit better when accessing trace events.
And some various clean ups. Some of the clean ups are necessary to
help in a fix to a theoretical race between opening a event file and
deleting that event"
* tag 'trace-3.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Kill the unbalanced tr->ref++ in tracing_buffers_open()
tracing: Kill trace_array->waiter
tracing: Do not (ab)use trace_seq in event_id_read()
tracing: Simplify the iteration logic in f_start/f_next
tracing: Add ref_data to function and fgraph tracer structs
tracing: Miscellaneous fixes for trace_array ref counting
tracing: Fix error handling to ensure instances can always be removed
tracing/kprobe: Wait for disabling all running kprobe handlers
tracing/perf: Move the PERF_MAX_TRACE_SIZE check into perf_trace_buf_prepare()
tracing/syscall: Avoid perf_trace_buf_*() if sys_data->perf_events is empty
tracing/function: Avoid perf_trace_buf_*() if event_function.perf_events is empty
tracing: Typo fix on ring buffer comments
tracing: Use trace_seq_puts()/trace_seq_putc() where possible
tracing: Use correct config guard CONFIG_STACK_TRACER
|
|
The expression '(1 << 32)' happens to evaluate as 0 on ARM, but
it evaluates as 1 on xtensa and x86_64. This zeros sched_clock_mask,
and breaks sched_clock().
Set the type of 1 to 'unsigned long long' to get the value we need.
Reported-by: Max Filippov <jcmvbkbc@gmail.com>
Tested-by: Max Filippov <jcmvbkbc@gmail.com>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
|
If I explicitly disable the clocksource watchdog in the x86 Kconfig,
the x86 kernel will not compile unless this is properly defined.
Cc: John Stultz <john.stultz@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: x86@kernel.org
Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
|
