| Age | Commit message (Collapse) | Author |
|---|
|
proc_cpuset_show() has a spurious -EINVAL assignment which does
nothing. Remove it.
This patch doesn't make any functional difference.
tj: Rewrote patch description.
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
5d21cc2db040d01f8c19b8602f6987813e1176b4 ("cpuset: replace
cgroup_mutex locking with cpuset internal locking") incorrectly
converted proc_cpuset_show() from cgroup_lock() to cpuset_mutex.
proc_cpuset_show() is accessing cgroup hierarchy proper to determine
cgroup path which can't be protected by cpuset_mutex. This triggered
the following RCU warning.
===============================
[ INFO: suspicious RCU usage. ]
3.8.0-rc3-next-20130114-sasha-00016-ga107525-dirty #262 Tainted: G W
-------------------------------
include/linux/cgroup.h:534 suspicious rcu_dereference_check() usage!
other info that might help us debug this:
rcu_scheduler_active = 1, debug_locks = 1
2 locks held by trinity/7514:
#0: (&p->lock){+.+.+.}, at: [<ffffffff812b06aa>] seq_read+0x3a/0x3e0
#1: (cpuset_mutex){+.+...}, at: [<ffffffff811abae4>] proc_cpuset_show+0x84/0x190
stack backtrace:
Pid: 7514, comm: trinity Tainted: G W
+3.8.0-rc3-next-20130114-sasha-00016-ga107525-dirty #262
Call Trace:
[<ffffffff81182cab>] lockdep_rcu_suspicious+0x10b/0x120
[<ffffffff811abb71>] proc_cpuset_show+0x111/0x190
[<ffffffff812b0827>] seq_read+0x1b7/0x3e0
[<ffffffff812b0670>] ? seq_lseek+0x110/0x110
[<ffffffff8128b4fb>] do_loop_readv_writev+0x4b/0x90
[<ffffffff8128b776>] do_readv_writev+0xf6/0x1d0
[<ffffffff8128b8ee>] vfs_readv+0x3e/0x60
[<ffffffff8128b960>] sys_readv+0x50/0xd0
[<ffffffff83d33d18>] tracesys+0xe1/0xe6
The operation can be performed under RCU read lock. Replace
cpuset_mutex locking with RCU read locking.
tj: Rewrote patch description.
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing regression fixes from Steven Rostedt:
"The clean up patch commit 0fb9656d957d "tracing: Make tracing_enabled
be equal to tracing_on" caused two regressions.
1) The irqs off latency tracer no longer starts if tracing_on is off
when the tracer is set, and then tracing_on is enabled. The
tracing_on file needs the hook that tracing_enabled had to enable
tracers if they request it (call the tracer's start() method).
2) That commit had a separate change that really should have been a
separate patch, but it must have been added accidently with the -a
option of git commit. But as the change is still related to the
commit it wasn't noticed in review. That change, changed the way
blocking is done by the trace_pipe file with respect to the
tracing_on settings. I've been told that this change breaks
current userspace, and this specific change is being reverted."
* tag 'trace-3.8-rc3-regression-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Fix regression of trace_pipe
tracing: Fix regression with irqsoff tracer and tracing_on file
|
|
Nothing's protected by RCU in rebind_subsystems(), and I can't think
of a reason why it is needed.
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
These 2 syncronize_rcu()s make attaching a task to a cgroup
quite slow, and it can't be ignored in some situations.
A real case from Colin Cross: Android uses cgroups heavily to
manage thread priorities, putting threads in a background group
with reduced cpu.shares when they are not visible to the user,
and in a foreground group when they are. Some RPCs from foreground
threads to background threads will temporarily move the background
thread into the foreground group for the duration of the RPC.
This results in many calls to cgroup_attach_task.
In cgroup_attach_task() it's task->cgroups that is protected by RCU,
and put_css_set() calls kfree_rcu() to free it.
If we remove this synchronize_rcu(), there can be threads in RCU-read
sections accessing their old cgroup via current->cgroups with
concurrent rmdir operation, but this is safe.
# time for ((i=0; i<50; i++)) { echo $$ > /mnt/sub/tasks; echo $$ > /mnt/tasks; }
real 0m2.524s
user 0m0.008s
sys 0m0.004s
With this patch:
real 0m0.004s
user 0m0.004s
sys 0m0.000s
tj: These synchronize_rcu()s are utterly confused. synchornize_rcu()
necessarily has to come between two operations to guarantee that
the changes made by the former operation are visible to all rcu
readers before proceeding to the latter operation. Here,
synchornize_rcu() are at the end of attach operations with nothing
beyond it. Its only effect would be delaying completion of
write(2) to sysfs tasks/procs files until all rcu readers see the
change, which doesn't mean anything.
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Colin Cross <ccross@google.com>
|
|
Clockevent cleanup series from Shawn Guo.
Resolved move/change conflict in mach-pxa/time.c due to the sys_timer
cleanup.
* clocksource/cleanup:
clocksource: use clockevents_config_and_register() where possible
ARM: use clockevents_config_and_register() where possible
clockevents: export clockevents_config_and_register for module use
+ sync to Linux 3.8-rc3
Signed-off-by: Olof Johansson <olof@lixom.net>
Conflicts:
arch/arm/mach-pxa/time.c
|
|
Commit 0fb9656d "tracing: Make tracing_enabled be equal to tracing_on"
changes the behaviour of trace_pipe, ie. it makes trace_pipe return if
we've read something and tracing is enabled, and this means that we have
to 'cat trace_pipe' again and again while running tests.
IMO the right way is if tracing is enabled, we always block and wait for
ring buffer, or we may lose what we want since ring buffer's size is limited.
Link: http://lkml.kernel.org/r/1358132051-5410-1-git-send-email-bo.li.liu@oracle.com
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
clockevents_config_and_register is a handy helper for clockevent
drivers, some of which might support module build, so export the symbol.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Olof Johansson <olof@lixom.net>
|
|
bio_{front|back}_merge tracepoints report a bio merging into an
existing request but didn't specify which request the bio is being
merged into. Add @req to it. This makes it impossible to share the
event template with block_bio_queue - split it out.
@req isn't used or exported to userland at this point and there is no
userland visible behavior change. Later changes will make use of the
extra parameter.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
bio completion didn't kick block_bio_complete TP. Only dm was
explicitly triggering the TP on IO completion. This makes
block_bio_complete TP useless for tracers which want to know about
bios, and all other bio based drivers skip generating blktrace
completion events.
This patch makes all bio completions via bio_endio() generate
block_bio_complete TP.
* Explicit trace_block_bio_complete() invocation removed from dm and
the trace point is unexported.
* @rq dropped from trace_block_bio_complete(). bios may fly around
w/o queue associated. Verifying and accessing the assocaited queue
belongs to TP probes.
* blktrace now gets both request and bio completions. Make it ignore
bio completions if request completion path is happening.
This makes all bio based drivers generate blktrace completion events
properly and makes the block_bio_complete TP actually useful.
v2: With this change, block_bio_complete TP could be invoked on sg
commands which have bio's with %NULL bi_bdev. Update TP
assignment code to check whether bio->bi_bdev is %NULL before
dereferencing.
Signed-off-by: Tejun Heo <tj@kernel.org>
Original-patch-by: Namhyung Kim <namhyung@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Alasdair Kergon <agk@redhat.com>
Cc: dm-devel@redhat.com
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Prarit's excellent bug report:
> In recent Fedora releases (F17 & F18) some users have reported seeing
> messages similar to
>
> [ 15.478160] kvm: Could not allocate 304 bytes percpu data
> [ 15.478174] PERCPU: allocation failed, size=304 align=32, alloc from
> reserved chunk failed
>
> during system boot. In some cases, users have also reported seeing this
> message along with a failed load of other modules.
>
> What is happening is systemd is loading an instance of the kvm module for
> each cpu found (see commit e9bda3b). When the module load occurs the kernel
> currently allocates the modules percpu data area prior to checking to see
> if the module is already loaded or is in the process of being loaded. If
> the module is already loaded, or finishes load, the module loading code
> releases the current instance's module's percpu data.
Now we have a new state MODULE_STATE_UNFORMED, we can insert the
module into the list (and thus guarantee its uniqueness) before we
allocate the per-cpu region.
Reported-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Tested-by: Prarit Bhargava <prarit@redhat.com>
|
|
You should never look at such a module, so it's excised from all paths
which traverse the modules list.
We add the state at the end, to avoid gratuitous ABI break (ksplice).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
|
|
audit_log_start() performs the same jiffies comparison in two places.
If sufficient time has elapsed between the two comparisons, the second
one produces a negative sleep duration:
schedule_timeout: wrong timeout value fffffffffffffff0
Pid: 6606, comm: trinity-child1 Not tainted 3.8.0-rc1+ #43
Call Trace:
schedule_timeout+0x305/0x340
audit_log_start+0x311/0x470
audit_log_exit+0x4b/0xfb0
__audit_syscall_exit+0x25f/0x2c0
sysret_audit+0x17/0x21
Fix it by performing the comparison a single time.
Reported-by: Dave Jones <davej@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
It's possible for audit_log_start() to return NULL. Handle it in the
various callers.
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Julien Tinnes <jln@google.com>
Cc: Will Drewry <wad@google.com>
Cc: Steve Grubb <sgrubb@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
The seccomp path was using AUDIT_ANOM_ABEND from when seccomp mode 1
could only kill a process. While we still want to make sure an audit
record is forced on a kill, this should use a separate record type since
seccomp mode 2 introduces other behaviors.
In the case of "handled" behaviors (process wasn't killed), only emit a
record if the process is under inspection. This change also fixes
userspace examination of seccomp audit events, since it was considered
malformed due to missing fields of the AUDIT_ANOM_ABEND event type.
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Julien Tinnes <jln@google.com>
Acked-by: Will Drewry <wad@chromium.org>
Acked-by: Steve Grubb <sgrubb@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
down_write_nest_lock() provides a means to annotate locking scenario
where an outer lock is guaranteed to serialize the order nested locks
are being acquired.
This is analogoue to already existing mutex_lock_nest_lock() and
spin_lock_nest_lock().
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: Rik van Riel <riel@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mel Gorman <mel@csn.ul.ie>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Commit 02404baf1b47 "tracing: Remove deprecated tracing_enabled file"
removed the tracing_enabled file as it never worked properly and
the tracing_on file should be used instead. But the tracing_on file
didn't call into the tracers start/stop routines like the
tracing_enabled file did. This caused trace-cmd to break when it
enabled the irqsoff tracer.
If you just did "echo irqsoff > current_tracer" then it would work
properly. But the tool trace-cmd disables tracing first by writing
"0" into the tracing_on file. Then it writes "irqsoff" into
current_tracer and then writes "1" into tracing_on. Unfortunately,
the above commit changed the irqsoff tracer to check the tracing_on
status instead of the tracing_enabled status. If it's disabled then
it does not start the tracer internals.
The problem is that writing "1" into tracing_on does not call the
tracers "start" routine like writing "1" into tracing_enabled did.
This makes the irqsoff tracer not start when using the trace-cmd
tool, and is a regression for userspace.
Simple fix is to have the tracing_on file call the tracers start()
method when being enabled (and the stop() method when disabled).
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
The CONFIG_EXPERIMENTAL config item has not carried much meaning for a
while now and is almost always enabled by default. As agreed during the
Linux kernel summit, remove it from any "depends on" lines in Kconfigs.
Cc: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Acked-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
|
|
In various network workloads, __do_softirq() latencies can be up
to 20 ms if HZ=1000, and 200 ms if HZ=100.
This is because we iterate 10 times in the softirq dispatcher,
and some actions can consume a lot of cycles.
This patch changes the fallback to ksoftirqd condition to :
- A time limit of 2 ms.
- need_resched() being set on current task
When one of this condition is met, we wakeup ksoftirqd for further
softirq processing if we still have pending softirqs.
Using need_resched() as the only condition can trigger RCU stalls,
as we can keep BH disabled for too long.
I ran several benchmarks and got no significant difference in
throughput, but a very significant reduction of latencies (one order
of magnitude) :
In following bench, 200 antagonist "netperf -t TCP_RR" are started in
background, using all available cpus.
Then we start one "netperf -t TCP_RR", bound to the cpu handling the NIC
IRQ (hard+soft)
Before patch :
# netperf -H 7.7.7.84 -t TCP_RR -T2,2 -- -k
RT_LATENCY,MIN_LATENCY,MAX_LATENCY,P50_LATENCY,P90_LATENCY,P99_LATENCY,MEAN_LATENCY,STDDEV_LATENCY
MIGRATED TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET
to 7.7.7.84 () port 0 AF_INET : first burst 0 : cpu bind
RT_LATENCY=550110.424
MIN_LATENCY=146858
MAX_LATENCY=997109
P50_LATENCY=305000
P90_LATENCY=550000
P99_LATENCY=710000
MEAN_LATENCY=376989.12
STDDEV_LATENCY=184046.92
After patch :
# netperf -H 7.7.7.84 -t TCP_RR -T2,2 -- -k
RT_LATENCY,MIN_LATENCY,MAX_LATENCY,P50_LATENCY,P90_LATENCY,P99_LATENCY,MEAN_LATENCY,STDDEV_LATENCY
MIGRATED TCP REQUEST/RESPONSE TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET
to 7.7.7.84 () port 0 AF_INET : first burst 0 : cpu bind
RT_LATENCY=40545.492
MIN_LATENCY=9834
MAX_LATENCY=78366
P50_LATENCY=33583
P90_LATENCY=59000
P99_LATENCY=69000
MEAN_LATENCY=38364.67
STDDEV_LATENCY=12865.26
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: David Miller <davem@davemloft.net>
Cc: Tom Herbert <therbert@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fix new kernel-doc warning in auditfilter.c:
Warning(kernel/auditfilter.c:1157): Excess function parameter 'uid' description in 'audit_receive_filter'
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Cc: linux-audit@redhat.com (subscribers-only)
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Switch cgroup to use the new hashtable implementation. No functional changes.
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing regression fix from Steven Rostedt:
"A change that came in this merge window broke the writing to the
trace_options file. It causes garbage to be read during the compare
of option names, and breaks setting options via the trace_options
file, although options can still be set via the options/<option>
files."
* tag 'trace-3.8-rc2-regression-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Fix regression of trace_options file setting
|
|
The latest change to allow trace options to be set on the command
line also broke the trace_options file.
The zeroing of the last byte of the option name that is echoed into
the trace_option file was removed with the consolidation of some
of the code. The compare between the option and what was written to
the trace_options file fails because the string holding the data
written doesn't terminate with a null character.
A zero needs to be added to the end of the string copied from
user space.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
IA64 defines /proc/sys/kernel/ignore-unaligned-usertrap to control
verbose warnings on unaligned access emulation.
Although the exact mechanics of what to do with sysctl (ignore/shout)
are arch specific, this change enables the sysctl to be usable cross-arch.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
|
|
This commit adds event tracing for callback acceleration to allow better
tracking of callbacks through the system.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
|
|
Currently, callbacks are advanced each time the corresponding CPU
notices a change in its leaf rcu_node structure's ->completed value
(this value counts grace-period completions). This approach has worked
quite well, but with the advent of RCU_FAST_NO_HZ, we cannot count on
a given CPU seeing all the grace-period completions. When a CPU misses
a grace-period completion that occurs while it is in dyntick-idle mode,
this will delay invocation of its callbacks.
In addition, acceleration of callbacks (when RCU realizes that a given
callback need only wait until the end of the next grace period, rather
than having to wait for a partial grace period followed by a full
grace period) must be carried out extremely carefully. Insufficient
acceleration will result in unnecessarily long grace-period latencies,
while excessive acceleration will result in premature callback invocation.
Changes that involve this tradeoff are therefore among the most
nerve-wracking changes to RCU.
This commit therefore explicitly tags groups of callbacks with the
number of the grace period that they are waiting for. This means that
callback-advancement and callback-acceleration functions are idempotent,
so that excessive acceleration will merely waste a few CPU cycles. This
also allows a CPU to take full advantage of any grace periods that have
elapsed while it has been in dyntick-idle mode. It should also enable
simulataneous simplifications to and optimizations of RCU_FAST_NO_HZ.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
|
|
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
|
|
It turns out that gcc 4.8 warns on array indexes being out of bounds
unless it can prove otherwise. It gives this warning on some RCU
initialization code. Because this is far from any fastpath, add
an explicit check for array bounds and panic if so. This gives the
compiler enough information to figure out that the array index is never
out of bounds.
However, if a similar false positive occurs on a fastpath, it will
probably be necessary to tell the compiler to keep its array-index
anxieties to itself. ;-)
Markus Trippelsdorf <markus@trippelsdorf.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
|
|
This patch uses the real new value of dynticks_nesting instead of 0 in
rcu_eqs_enter_common().
Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
|
|
Both rcutiny and rcutree define a helper function named
rcu_is_cpu_rrupt_from_idle(), each used exactly once, later in the
same file. This commit therefore declares these helper functions static.
Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
|
|
Currently, rcutorture traces every read-side access. This can be
problematic because even a two-minute rcutorture run on a two-CPU system
can generate 28,853,363 reads. Normally, only a failing read is of
interest, so this commit traces adjusts rcutorture's tracing to only
trace failing reads. The resulting event tracing records the time
and the ->completed value captured at the beginning of the RCU read-side
critical section, allowing correlation with other event-tracing messages.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
[ paulmck: Add fix to build problem located by Randy Dunlap based on
diagnosis by Steven Rostedt. ]
|
|
The rcutorture tests need to be able to trace the time of the
beginning of an RCU read-side critical section, and thus need access
to trace_clock_local(). This commit therefore adds a the needed
EXPORT_SYMBOL_GPL().
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
|
|
The as-documented rcu_nocb_poll will fail to enable this feature
for two reasons. (1) there is an extra "s" in the documented
name which is not in the code, and (2) since it uses module_param,
it really is expecting a prefix, akin to "rcutree.fanout_leaf"
and the prefix isn't documented.
However, there are several reasons why we might not want to
simply fix the typo and add the prefix:
1) we'd end up with rcutree.rcu_nocb_poll, and rather probably make
a change to rcutree.nocb_poll
2) if we did #1, then the prefix wouldn't be consistent with the
rcu_nocbs=<cpumap> parameter (i.e. one with, one without prefix)
3) the use of module_param in a header file is less than desired,
since it isn't immediately obvious that it will get processed
via rcutree.c and get the prefix from that (although use of
module_param_named() could clarify that.)
4) the implied export of /sys/module/rcutree/parameters/rcu_nocb_poll
data to userspace via module_param() doesn't really buy us anything,
as it is read-only and we can tell if it is enabled already without
it, since there is a printk at early boot telling us so.
In light of all that, just change it from a module_param() to an
early_setup() call, and worry about adding it to /sys later on if
we decide to allow a dynamic setting of it.
Also change the variable to be tagged as read_mostly, since it
will only ever be fiddled with at most, once at boot.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
|
|
The wait_event() at the head of the rcu_nocb_kthread() can result in
soft-lockup complaints if the CPU in question does not register RCU
callbacks for an extended period. This commit therefore changes
the wait_event() to a wait_event_interruptible().
Reported-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/swarren/linux-tegra into next/cleanup
From Stephen Warren:
ARM/...: timer and clock events cleanup, and remove struct sys_timer
This branch contains a number of cleanups and unifications to various
timer- clock-events- and ARM timer code. The main points are:
1) Convert arch_gettimeoffset to a pointer, so that architectures with
multiple timer implementations can simply set this standard pointer
rather than maintaining their own arch-specific pointers for the
same purpose. Various architectures are converted to using this new
feature.
2) Conversion of ARM timer implementations to use clock_event_devices's
suspend/resume operations, rather than the ARM-specific sys_timer
versions. Thus, the ARM code begins to use more common infra-structure
rather than arch-specific code.
3) Removal of ARM's struct sys_timer completely, now that everything uses
common code.
4) Introduction of drivers/clocksource/clksrc-of.c, which allows ARM clock
source implementations to be moved into drivers/clocksource, with the
need to add SoC-specific header files for each timer initialization
function; instead, all enabled implementations are registered into a
table which a single core function iterates over, and calls the
relevant initialization functions based on device tree. At least the
Tegra and BCM2835 clocksource implementations will use this feature in
the 3.9 kernel cycle.
* tag 'swarren-for-3.9-arm-timer-rework' of git://git.kernel.org/pub/scm/linux/kernel/git/swarren/linux-tegra:
clocksource: add common of_clksrc_init() function
ARM: delete struct sys_timer
ARM: remove struct sys_timer suspend and resume fields
ARM: samsung: register syscore_ops for timer resume directly
ARM: ux500: convert timer suspend/resume to clock_event_device
ARM: sa1100: convert timer suspend/resume to clock_event_device
ARM: pxa: convert timer suspend/resume to clock_event_device
ARM: at91: convert timer suspend/resume to clock_event_device
ARM: set arch_gettimeoffset directly
m68k: set arch_gettimeoffset directly
time: convert arch_gettimeoffset to a pointer
cris: move usec/nsec conversion to do_slow_gettimeoffset
Signed-off-by: Olof Johansson <olof@lixom.net>
|
|
cgroup already tracks the hierarchy. Follow cgroup->parent to find
the parent and drop cpuset->parent.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
Implement cpuset_for_each_descendant_pre() and replace the
cpuset-specific tree walking using cpuset->stack_list with it.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
Supposedly for historical reasons, cpuset depends on cgroup core for
locking. It depends on cgroup_mutex in cgroup callbacks and grabs
cgroup_mutex from other places where it wants to be synchronized.
This is majorly messy and highly prone to introducing circular locking
dependency especially because cgroup_mutex is supposed to be one of
the outermost locks.
As previous patches already plugged possible races which may happen by
decoupling from cgroup_mutex, replacing cgroup_mutex with cpuset
specific cpuset_mutex is mostly straight-forward. Introduce
cpuset_mutex, replace all occurrences of cgroup_mutex with it, and add
cpuset_mutex locking to places which inherited cgroup_mutex from
cgroup core.
The only complication is from cpuset wanting to initiate task
migration when a cpuset loses all cpus or memory nodes. Task
migration may go through full cgroup and all subsystem locking and
should be initiated without holding any cpuset specific lock; however,
a previous patch already made hotplug handled asynchronously and
moving the task migration part outside other locks is easy.
cpuset_propagate_hotplug_workfn() now invokes
remove_tasks_in_empty_cpuset() without holding any lock.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
cpuset is scheduled to be decoupled from cgroup_lock which will make
hotplug handling race with task migration. cpus or mems will be
allowed to go offline between ->can_attach() and ->attach(). If
hotplug takes down all cpus or mems of a cpuset while attach is in
progress, ->attach() may end up putting tasks into an empty cpuset.
This patchset makes ->attach() schedule hotplug propagation if the
cpuset is empty after attaching is complete. This will move the tasks
to the nearest ancestor which can execute and the end result would be
as if hotplug handling happened after the tasks finished attaching.
cpuset_write_resmask() now also flushes cpuset_propagate_hotplug_wq to
wait for propagations scheduled directly by cpuset_attach().
This currently doesn't make any functional difference as everything is
protected by cgroup_mutex but enables decoupling the locking.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
cpuset is scheduled to be decoupled from cgroup_lock which will make
configuration updates race with task migration. Any config update
will be allowed to happen between ->can_attach() and ->attach(). If
such config update removes either all cpus or mems, by the time
->attach() is called, the condition verified by ->can_attach(), that
the cpuset is capable of hosting the tasks, is no longer true.
This patch adds cpuset->attach_in_progress which is incremented from
->can_attach() and decremented when the attach operation finishes
either successfully or not. validate_change() treats cpusets w/
non-zero ->attach_in_progress like cpusets w/ tasks and refuses to
remove all cpus or mems from it.
This currently doesn't make any functional difference as everything is
protected by cgroup_mutex but enables decoupling the locking.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
cpuset_hotplug_workfn() has been invoking cpuset_propagate_hotplug()
directly to propagate hotplug updates to !root cpusets; however, this
has the following problems.
* cpuset locking is scheduled to be decoupled from cgroup_mutex,
cgroup_mutex will be unexported, and cgroup_attach_task() will do
cgroup locking internally, so propagation can't synchronously move
tasks to a parent cgroup while walking the hierarchy.
* We can't use cgroup generic tree iterator because propagation to
each cpuset may sleep. With propagation done asynchronously, we can
lose the rather ugly cpuset specific iteration.
Convert cpuset_propagate_hotplug() to
cpuset_propagate_hotplug_workfn() and execute it from newly added
cpuset->hotplug_work. The work items are run on an ordered workqueue,
so the propagation order is preserved. cpuset_hotplug_workfn()
schedules all propagations while holding cgroup_mutex and waits for
completion without cgroup_mutex. Each in-flight propagation holds a
reference to the cpuset->css.
This patch doesn't cause any functional difference.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
In general, we want to make cgroup_mutex one of the outermost locks
and be able to use get_online_cpus() and friends from cgroup methods.
With cpuset hotplug made async, get_online_cpus() can now be nested
inside cgroup_mutex.
Currently, cpuset avoids nesting get_online_cpus() inside cgroup_mutex
by bouncing sched_domain rebuilding to a work item. As such nesting
is allowed now, remove the workqueue bouncing code and always rebuild
sched_domains synchronously. This also nests sched_domains_mutex
inside cgroup_mutex, which is intended and should be okay.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
CPU / memory hotplug path currently grabs cgroup_mutex from hotplug
event notifications. We want to separate cpuset locking from cgroup
core and make cgroup_mutex outer to hotplug synchronization so that,
among other things, mechanisms which depend on get_online_cpus() can
be used from cgroup callbacks. In general, we want to keep
cgroup_mutex the outermost lock to minimize locking interactions among
different controllers.
Convert cpuset_handle_hotplug() to cpuset_hotplug_workfn() and
schedule it from the hotplug notifications. As the function can
already handle multiple mixed events without any input, converting it
to a work function is mostly trivial; however, one complication is
that cpuset_update_active_cpus() needs to update sched domains
synchronously to reflect an offlined cpu to avoid confusing the
scheduler. This is worked around by falling back to the the default
single sched domain synchronously before scheduling the actual hotplug
work. This makes sched domain rebuilt twice per CPU hotplug event but
the operation isn't that heavy and a lot of the second operation would
be noop for systems w/ single sched domain, which is the common case.
This decouples cpuset hotplug handling from the notification callbacks
and there can be an arbitrary delay between the actual event and
updates to cpusets. Scheduler and mm can handle it fine but moving
tasks out of an empty cpuset may race against writes to the cpuset
restoring execution resources which can lead to confusing behavior.
Flush hotplug work item from cpuset_write_resmask() to avoid such
confusions.
v2: Synchronous sched domain rebuilding using the fallback sched
domain added. This fixes various issues caused by confused
scheduler putting tasks on a dead CPU, including the one reported
by Li Zefan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
Reorganize hotplug path to prepare for async hotplug handling.
* Both CPU and memory hotplug handlings are collected into a single
function - cpuset_handle_hotplug(). It doesn't take any argument
but compares the current setttings of top_cpuset against what's
actually available to determine what happened. This function
directly updates top_cpuset. If there are CPUs or memory nodes
which are taken down, cpuset_propagate_hotplug() in invoked on all
!root cpusets.
* cpuset_propagate_hotplug() is responsible for updating the specified
cpuset so that it doesn't include any resource which isn't available
to top_cpuset. If no CPU or memory is left after update, all tasks
are moved to the nearest ancestor with both resources.
* update_tasks_cpumask() and update_tasks_nodemask() are now always
called after cpus or mems masks are updated even if the cpuset
doesn't have any task. This is for brevity and not expected to have
any measureable effect.
* cpu_active_mask and N_HIGH_MEMORY are read exactly once per
cpuset_handle_hotplug() invocation, all cpusets share the same view
of what resources are available, and cpuset_handle_hotplug() can
handle multiple resources going up and down. These properties will
allow async operation.
The reorganization, while drastic, is equivalent and shouldn't cause
any behavior difference. This will enable making hotplug handling
async and remove get_online_cpus() -> cgroup_mutex nesting.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
cpuset_can_attach() prepare global variables cpus_attach and
cpuset_attach_nodemask_{to|from} which are used by cpuset_attach().
There is no reason to prepare in cpuset_can_attach(). The same
information can be accessed from cpuset_attach().
Move the prepartion logic from cpuset_can_attach() to cpuset_attach()
and make the global variables static ones inside cpuset_attach().
With this change, there's no reason to keep
cpuset_attach_nodemask_{from|to} global. Move them inside
cpuset_attach(). Unfortunately, we need to keep cpus_attach global as
it can't be allocated from cpuset_attach().
v2: cpus_attach not converted to cpumask_t as per Li Zefan and Rusty
Russell.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
|
|
Instead of iterating cgroup->children directly, introduce and use
cpuset_for_each_child() which wraps cgroup_for_each_child() and
performs online check. As it uses the generic iterator, it requires
RCU read locking too.
As cpuset is currently protected by cgroup_mutex, non-online cpusets
aren't visible to all the iterations and this patch currently doesn't
make any functional difference. This will be used to de-couple cpuset
locking from cgroup core.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
Add CS_ONLINE which is set from css_online() and cleared from
css_offline(). This will enable using generic cgroup iterator while
allowing decoupling cpuset from cgroup internal locking.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
Add cpuset_css_on/offline() and rearrange css init/exit such that,
* Allocation and clearing to the default values happen in css_alloc().
Allocation now uses kzalloc().
* Config inheritance and registration happen in css_online().
* css_offline() undoes what css_online() did.
* css_free() frees.
This doesn't introduce any visible behavior changes. This will help
cleaning up locking.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
The function isn't that hot, the overhead of missing the fast exit is
low, the test itself depends heavily on cgroup internals, and it's
gonna be a hindrance when trying to decouple cpuset locking from
cgroup core. Remove the fast exit path.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
|
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
|
