A customer reported the following softlockup:
[899688.160002] NMI watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [test.sh:16464]
[899688.160002] CPU: 0 PID: 16464 Comm: test.sh Not tainted 4.12.14-6.23-azure #1 SLE12-SP4
[899688.160002] RIP: 0010:up_write+0x1a/0x30
[899688.160002] Kernel panic - not syncing: softlockup: hung tasks
[899688.160002] RIP: 0010:up_write+0x1a/0x30
[899688.160002] RSP: 0018:ffffa86784d4fde8 EFLAGS: 00000257 ORIG_RAX: ffffffffffffff12
[899688.160002] RAX: ffffffff970fea00 RBX: 0000000000000001 RCX: 0000000000000000
[899688.160002] RDX: ffffffff00000001 RSI: 0000000000000080 RDI: ffffffff970fea00
[899688.160002] RBP: ffffffffffffffff R08: ffffffffffffffff R09: 0000000000000000
[899688.160002] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8b59014720d8
[899688.160002] R13: ffff8b59014720c0 R14: ffff8b5901471090 R15: ffff8b5901470000
[899688.160002] tracing_read_pipe+0x336/0x3c0
[899688.160002] __vfs_read+0x26/0x140
[899688.160002] vfs_read+0x87/0x130
[899688.160002] SyS_read+0x42/0x90
[899688.160002] do_syscall_64+0x74/0x160
It caught the process in the middle of trace_access_unlock(). There is
no loop. So, it must be looping in the caller tracing_read_pipe()
via the "waitagain" label.
Crashdump analyze uncovered that iter->seq was completely zeroed
at this point, including iter->seq.seq.size. It means that
print_trace_line() was never able to print anything and
there was no forward progress.
The culprit seems to be in the code:
/* reset all but tr, trace, and overruns */
memset(&iter->seq, 0,
sizeof(struct trace_iterator) -
offsetof(struct trace_iterator, seq));
It was added by the commit 53d0aa7730 ("ftrace:
add logic to record overruns"). It was v2.6.27-rc1.
It was the time when iter->seq looked like:
struct trace_seq {
unsigned char buffer[PAGE_SIZE];
unsigned int len;
};
There was no "size" variable and zeroing was perfectly fine.
The solution is to reinitialize the structure after or without
zeroing.
Link: http://lkml.kernel.org/r/20191011142134.11997-1-pmladek@suse.com
Signed-off-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
nmi_total_ts is supposed to record the total time spent in *all* NMIs
that occur on the given CPU during the (active portion of the)
sampling window. However, the code seems to be overwriting this
variable for each NMI, thereby only recording the time spent in the
most recent NMI. Fix it by accumulating the duration instead.
Link: http://lkml.kernel.org/r/157073343544.17189.13911783866738671133.stgit@srivatsa-ubuntu
Fixes: 7b2c862501 ("tracing: Add NMI tracing in hwlat detector")
Cc: stable@vger.kernel.org
Signed-off-by: Srivatsa S. Bhat (VMware) <srivatsa@csail.mit.edu>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Added various checks on open tracefs calls to see if tracefs is in lockdown
mode, and if so, to return -EPERM.
Note, the event format files (which are basically standard on all machines)
as well as the enabled_functions file (which shows what is currently being
traced) are not lockde down. Perhaps they should be, but it seems counter
intuitive to lockdown information to help you know if the system has been
modified.
Link: http://lkml.kernel.org/r/CAHk-=wj7fGPKUspr579Cii-w_y60PtRaiDgKuxVtBAMK0VNNkA@mail.gmail.com
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Currently, most files in the tracefs directory test if tracing_disabled is
set. If so, it should return -ENODEV. The tracing_disabled is called when
tracing is found to be broken. Originally it was done in case the ring
buffer was found to be corrupted, and we wanted to prevent reading it from
crashing the kernel. But it's also called if a tracing selftest fails on
boot. It's a one way switch. That is, once it is triggered, tracing is
disabled until reboot.
As most tracefs files can also be used by instances in the tracefs
directory, they need to be carefully done. Each instance has a trace_array
associated to it, and when the instance is removed, the trace_array is
freed. But if an instance is opened with a reference to the trace_array,
then it requires looking up the trace_array to get its ref counter (as there
could be a race with it being deleted and the open itself). Once it is
found, a reference is added to prevent the instance from being removed (and
the trace_array associated with it freed).
Combine the two checks (tracing_disabled and trace_array_get()) into a
single helper function. This will also make it easier to add lockdown to
tracefs later.
Link: http://lkml.kernel.org/r/20191011135458.7399da44@gandalf.local.home
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Instead of having the trace events system open call open code the taking of
the trace_array descriptor (with trace_array_get()) and then calling
trace_open_generic(), have it use the tracing_open_generic_tr() that does
the combination of the two. This requires making tracing_open_generic_tr()
global.
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
As instances may have different tracers available, we need to look at the
trace_array descriptor that shows the list of the available tracers for the
instance. But there's a race between opening the file and an admin
deleting the instance. The trace_array_get() needs to be called before
accessing the trace_array.
Cc: stable@vger.kernel.org
Fixes: 607e2ea167 ("tracing: Set up infrastructure to allow tracers for instances")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
The ftrace set_ftrace_filter and set_ftrace_notrace files are specific for
an instance now. They need to take a reference to the instance otherwise
there could be a race between accessing the files and deleting the instance.
It wasn't until the :mod: caching where these file operations started
referencing the trace_array directly.
Cc: stable@vger.kernel.org
Fixes: 673feb9d76 ("ftrace: Add :mod: caching infrastructure to trace_array")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Pull scheduler fixes from Ingo Molnar:
"Two fixes: a guest-cputime accounting fix, and a cgroup bandwidth
quota precision fix"
* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/vtime: Fix guest/system mis-accounting on task switch
sched/fair: Scale bandwidth quota and period without losing quota/period ratio precision
Pull perf fixes from Ingo Molnar:
"Mostly tooling fixes, but also a couple of updates for new Intel
models (which are technically hw-enablement, but to users it's a fix
to perf behavior on those new CPUs - hope this is fine), an AUX
inheritance fix, event time-sharing fix, and a fix for lost non-perf
NMI events on AMD systems"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (36 commits)
perf/x86/cstate: Add Tiger Lake CPU support
perf/x86/msr: Add Tiger Lake CPU support
perf/x86/intel: Add Tiger Lake CPU support
perf/x86/cstate: Update C-state counters for Ice Lake
perf/x86/msr: Add new CPU model numbers for Ice Lake
perf/x86/cstate: Add Comet Lake CPU support
perf/x86/msr: Add Comet Lake CPU support
perf/x86/intel: Add Comet Lake CPU support
perf/x86/amd: Change/fix NMI latency mitigation to use a timestamp
perf/core: Fix corner case in perf_rotate_context()
perf/core: Rework memory accounting in perf_mmap()
perf/core: Fix inheritance of aux_output groups
perf annotate: Don't return -1 for error when doing BPF disassembly
perf annotate: Return appropriate error code for allocation failures
perf annotate: Fix arch specific ->init() failure errors
perf annotate: Propagate the symbol__annotate() error return
perf annotate: Fix the signedness of failure returns
perf annotate: Propagate perf_env__arch() error
perf evsel: Fall back to global 'perf_env' in perf_evsel__env()
perf tools: Propagate get_cpuid() error
...
Pull block fixes from Jens Axboe:
- Fix wbt performance regression introduced with the blk-rq-qos
refactoring (Harshad)
- Fix io_uring fileset removal inadvertently killing the workqueue (me)
- Fix io_uring typo in linked command nonblock submission (Pavel)
- Remove spurious io_uring wakeups on request free (Pavel)
- Fix null_blk zoned command error return (Keith)
- Don't use freezable workqueues for backing_dev, also means we can
revert a previous libata hack (Mika)
- Fix nbd sysfs mutex dropped too soon at removal time (Xiubo)
* tag 'for-linus-20191010' of git://git.kernel.dk/linux-block:
nbd: fix possible sysfs duplicate warning
null_blk: Fix zoned command return code
io_uring: only flush workqueues on fileset removal
io_uring: remove wait loop spurious wakeups
blk-wbt: fix performance regression in wbt scale_up/scale_down
Revert "libata, freezer: avoid block device removal while system is frozen"
bdi: Do not use freezable workqueue
io_uring: fix reversed nonblock flag for link submission
ptrace_stop() does preempt_enable_no_resched() to avoid the preemption,
but after that cgroup_enter_frozen() does spin_lock/unlock and this adds
another preemption point.
Reported-and-tested-by: Bruce Ashfield <bruce.ashfield@gmail.com>
Fixes: 76f969e894 ("cgroup: cgroup v2 freezer")
Cc: stable@vger.kernel.org # v5.2+
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roman Gushchin <guro@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Maps that are read-only both from BPF program side and user space side
have their contents constant, so verifier can track referenced values
precisely and use that knowledge for dead code elimination, branch
pruning, etc. This patch teaches BPF verifier how to do this.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20191009201458.2679171-2-andriin@fb.com
This allows the seccomp notifier to continue a syscall. A positive
discussion about this feature was triggered by a post to the
ksummit-discuss mailing list (cf. [3]) and took place during KSummit
(cf. [1]) and again at the containers/checkpoint-restore
micro-conference at Linux Plumbers.
Recently we landed seccomp support for SECCOMP_RET_USER_NOTIF (cf. [4])
which enables a process (watchee) to retrieve an fd for its seccomp
filter. This fd can then be handed to another (usually more privileged)
process (watcher). The watcher will then be able to receive seccomp
messages about the syscalls having been performed by the watchee.
This feature is heavily used in some userspace workloads. For example,
it is currently used to intercept mknod() syscalls in user namespaces
aka in containers.
The mknod() syscall can be easily filtered based on dev_t. This allows
us to only intercept a very specific subset of mknod() syscalls.
Furthermore, mknod() is not possible in user namespaces toto coelo and
so intercepting and denying syscalls that are not in the whitelist on
accident is not a big deal. The watchee won't notice a difference.
In contrast to mknod(), a lot of other syscall we intercept (e.g.
setxattr()) cannot be easily filtered like mknod() because they have
pointer arguments. Additionally, some of them might actually succeed in
user namespaces (e.g. setxattr() for all "user.*" xattrs). Since we
currently cannot tell seccomp to continue from a user notifier we are
stuck with performing all of the syscalls in lieu of the container. This
is a huge security liability since it is extremely difficult to
correctly assume all of the necessary privileges of the calling task
such that the syscall can be successfully emulated without escaping
other additional security restrictions (think missing CAP_MKNOD for
mknod(), or MS_NODEV on a filesystem etc.). This can be solved by
telling seccomp to resume the syscall.
One thing that came up in the discussion was the problem that another
thread could change the memory after userspace has decided to let the
syscall continue which is a well known TOCTOU with seccomp which is
present in other ways already.
The discussion showed that this feature is already very useful for any
syscall without pointer arguments. For any accidentally intercepted
non-pointer syscall it is safe to continue.
For syscalls with pointer arguments there is a race but for any cautious
userspace and the main usec cases the race doesn't matter. The notifier
is intended to be used in a scenario where a more privileged watcher
supervises the syscalls of lesser privileged watchee to allow it to get
around kernel-enforced limitations by performing the syscall for it
whenever deemed save by the watcher. Hence, if a user tricks the watcher
into allowing a syscall they will either get a deny based on
kernel-enforced restrictions later or they will have changed the
arguments in such a way that they manage to perform a syscall with
arguments that they would've been allowed to do anyway.
In general, it is good to point out again, that the notifier fd was not
intended to allow userspace to implement a security policy but rather to
work around kernel security mechanisms in cases where the watcher knows
that a given action is safe to perform.
/* References */
[1]: https://linuxplumbersconf.org/event/4/contributions/560
[2]: https://linuxplumbersconf.org/event/4/contributions/477
[3]: https://lore.kernel.org/r/20190719093538.dhyopljyr5ns33qx@brauner.io
[4]: commit 6a21cc50f0 ("seccomp: add a return code to trap to userspace")
Co-developed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Reviewed-by: Tycho Andersen <tycho@tycho.ws>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Will Drewry <wad@chromium.org>
CC: Tyler Hicks <tyhicks@canonical.com>
Link: https://lore.kernel.org/r/20190920083007.11475-2-christian.brauner@ubuntu.com
Signed-off-by: Kees Cook <keescook@chromium.org>
In kdb when you do 'btc' (back trace on CPU) it doesn't necessarily
give you the right info. Specifically on many architectures
(including arm64, where I tested) you can't dump the stack of a
"running" process that isn't the process running on the current CPU.
This can be seen by this:
echo SOFTLOCKUP > /sys/kernel/debug/provoke-crash/DIRECT
# wait 2 seconds
<sysrq>g
Here's what I see now on rk3399-gru-kevin. I see the stack crawl for
the CPU that handled the sysrq but everything else just shows me stuck
in __switch_to() which is bogus:
======
[0]kdb> btc
btc: cpu status: Currently on cpu 0
Available cpus: 0, 1-3(I), 4, 5(I)
Stack traceback for pid 0
0xffffff801101a9c0 0 0 1 0 R 0xffffff801101b3b0 *swapper/0
Call trace:
dump_backtrace+0x0/0x138
...
kgdb_compiled_brk_fn+0x34/0x44
...
sysrq_handle_dbg+0x34/0x5c
Stack traceback for pid 0
0xffffffc0f175a040 0 0 1 1 I 0xffffffc0f175aa30 swapper/1
Call trace:
__switch_to+0x1e4/0x240
0xffffffc0f65616c0
Stack traceback for pid 0
0xffffffc0f175d040 0 0 1 2 I 0xffffffc0f175da30 swapper/2
Call trace:
__switch_to+0x1e4/0x240
0xffffffc0f65806c0
Stack traceback for pid 0
0xffffffc0f175b040 0 0 1 3 I 0xffffffc0f175ba30 swapper/3
Call trace:
__switch_to+0x1e4/0x240
0xffffffc0f659f6c0
Stack traceback for pid 1474
0xffffffc0dde8b040 1474 727 1 4 R 0xffffffc0dde8ba30 bash
Call trace:
__switch_to+0x1e4/0x240
__schedule+0x464/0x618
0xffffffc0dde8b040
Stack traceback for pid 0
0xffffffc0f17b0040 0 0 1 5 I 0xffffffc0f17b0a30 swapper/5
Call trace:
__switch_to+0x1e4/0x240
0xffffffc0f65dd6c0
===
The problem is that 'btc' eventually boils down to
show_stack(task_struct, NULL);
...and show_stack() doesn't work for "running" CPUs because their
registers haven't been stashed.
On x86 things might work better (I haven't tested) because kdb has a
special case for x86 in kdb_show_stack() where it passes the stack
pointer to show_stack(). This wouldn't work on arm64 where the stack
crawling function seems needs the "fp" and "pc", not the "sp" which is
presumably why arm64's show_stack() function totally ignores the "sp"
parameter.
NOTE: we _can_ get a good stack dump for all the cpus if we manually
switch each one to the kdb master and do a back trace. AKA:
cpu 4
bt
...will give the expected trace. That's because now arm64's
dump_backtrace will now see that "tsk == current" and go through a
different path.
In this patch I fix the problems by catching a request to stack crawl
a task that's running on a CPU and then I ask that CPU to do the stack
crawl.
NOTE: this will (presumably) change what stack crawls are printed for
x86 machines. Now kdb functions will show up in the stack crawl.
Presumably this is OK but if it's not we can go back and add a special
case for x86 again.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
I noticed that when I did "btc <cpu>" and the CPU I passed in hadn't
rounded up that I'd crash. I was going to copy the same fix from
commit 162bc7f5af ("kdb: Don't back trace on a cpu that didn't round
up") into the "not all the CPUs" case, but decided it'd be better to
clean things up a little bit.
This consolidates the two code paths. It is _slightly_ wasteful in in
that the checks for "cpu" being too small or being offline isn't
really needed when we're iterating over all online CPUs, but that
really shouldn't hurt. Better to have the same code path.
While at it, eliminate at least one slightly ugly (and totally
needless) recursive use of kdb_parse().
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
The kdb_bt1() had a mysterious "argcount" parameter passed in (always
the number 5, by the way) and never used. Presumably this is just old
cruft. Remove it. While at it, upgrade the btaprompt parameter to a
full fledged bool instead of an int.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
From doing a 'git log --patch kernel/debug', it looks as if DCPU_SSTEP
has never been used. Presumably it used to be used back when kgdb was
out of tree and nobody thought to delete the definition when the usage
went away. Delete.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
Include the <linux/runtime_pm.h> for the definition of
pm_wq to avoid the following warning:
kernel/power/main.c:890:25: warning: symbol 'pm_wq' was not declared. Should it be static?
Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
In perf_rotate_context(), when the first cpu flexible event fail to
schedule, cpu_rotate is 1, while cpu_event is NULL. Since cpu_event is
NULL, perf_rotate_context will _NOT_ call cpu_ctx_sched_out(), thus
cpuctx->ctx.is_active will have EVENT_FLEXIBLE set. Then, the next
perf_event_sched_in() will skip all cpu flexible events because of the
EVENT_FLEXIBLE bit.
In the next call of perf_rotate_context(), cpu_rotate stays 1, and
cpu_event stays NULL, so this process repeats. The end result is, flexible
events on this cpu will not be scheduled (until another event being added
to the cpuctx).
Here is an easy repro of this issue. On Intel CPUs, where ref-cycles
could only use one counter, run one pinned event for ref-cycles, one
flexible event for ref-cycles, and one flexible event for cycles. The
flexible ref-cycles is never scheduled, which is expected. However,
because of this issue, the cycles event is never scheduled either.
$ perf stat -e ref-cycles:D,ref-cycles,cycles -C 5 -I 1000
time counts unit events
1.000152973 15,412,480 ref-cycles:D
1.000152973 <not counted> ref-cycles (0.00%)
1.000152973 <not counted> cycles (0.00%)
2.000486957 18,263,120 ref-cycles:D
2.000486957 <not counted> ref-cycles (0.00%)
2.000486957 <not counted> cycles (0.00%)
To fix this, when the flexible_active list is empty, try rotate the
first event in the flexible_groups. Also, rename ctx_first_active() to
ctx_event_to_rotate(), which is more accurate.
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <kernel-team@fb.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sasha Levin <sashal@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 8d5bce0c37 ("perf/core: Optimize perf_rotate_context() event scheduling")
Link: https://lkml.kernel.org/r/20191008165949.920548-1-songliubraving@fb.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
On context switch we are locking the vtime seqcount of the scheduling-out
task twice:
* On vtime_task_switch_common(), when we flush the pending vtime through
vtime_account_system()
* On arch_vtime_task_switch() to reset the vtime state.
This is pointless as these actions can be performed without the need
to unlock/lock in the middle. The reason these steps are separated is to
consolidate a very small amount of common code between
CONFIG_VIRT_CPU_ACCOUNTING_GEN and CONFIG_VIRT_CPU_ACCOUNTING_NATIVE.
Performance in this fast path is definitely a priority over artificial
code factorization so split the task switch code between GEN and
NATIVE and mutualize the parts than can run under a single seqcount
locked block.
As a side effect, vtime_account_idle() becomes included in the seqcount
protection. This happens to be a welcome preparation in order to
properly support kcpustat under vtime in the future and fetch
CPUTIME_IDLE without race.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpengli@tencent.com>
Cc: Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
Link: https://lkml.kernel.org/r/20191003161745.28464-3-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
vtime_account_system() assumes that the target task to account cputime
to is always the current task. This is most often true indeed except on
task switch where we call:
vtime_common_task_switch(prev)
vtime_account_system(prev)
Here prev is the scheduling-out task where we account the cputime to. It
doesn't match current that is already the scheduling-in task at this
stage of the context switch.
So we end up checking the wrong task flags to determine if we are
accounting guest or system time to the previous task.
As a result the wrong task is used to check if the target is running in
guest mode. We may then spuriously account or leak either system or
guest time on task switch.
Fix this assumption and also turn vtime_guest_enter/exit() to use the
task passed in parameter as well to avoid future similar issues.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wanpeng Li <wanpengli@tencent.com>
Fixes: 2a42eb9594 ("sched/cputime: Accumulate vtime on top of nsec clocksource")
Link: https://lkml.kernel.org/r/20190925214242.21873-1-frederic@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The quota/period ratio is used to ensure a child task group won't get
more bandwidth than the parent task group, and is calculated as:
normalized_cfs_quota() = [(quota_us << 20) / period_us]
If the quota/period ratio was changed during this scaling due to
precision loss, it will cause inconsistency between parent and child
task groups.
See below example:
A userspace container manager (kubelet) does three operations:
1) Create a parent cgroup, set quota to 1,000us and period to 10,000us.
2) Create a few children cgroups.
3) Set quota to 1,000us and period to 10,000us on a child cgroup.
These operations are expected to succeed. However, if the scaling of
147/128 happens before step 3, quota and period of the parent cgroup
will be changed:
new_quota: 1148437ns, 1148us
new_period: 11484375ns, 11484us
And when step 3 comes in, the ratio of the child cgroup will be
104857, which will be larger than the parent cgroup ratio (104821),
and will fail.
Scaling them by a factor of 2 will fix the problem.
Tested-by: Phil Auld <pauld@redhat.com>
Signed-off-by: Xuewei Zhang <xueweiz@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Phil Auld <pauld@redhat.com>
Cc: Anton Blanchard <anton@ozlabs.org>
Cc: Ben Segall <bsegall@google.com>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Fixes: 2e8e192263 ("sched/fair: Limit sched_cfs_period_timer() loop to avoid hard lockup")
Link: https://lkml.kernel.org/r/20191004001243.140897-1-xueweiz@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Merge misc fixes from Andrew Morton:
"The usual shower of hotfixes.
Chris's memcg patches aren't actually fixes - they're mature but a few
niggling review issues were late to arrive.
The ocfs2 fixes are quite old - those took some time to get reviewer
attention.
Subsystems affected by this patch series: ocfs2, hotfixes, mm/memcg,
mm/slab-generic"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
mm, sl[aou]b: guarantee natural alignment for kmalloc(power-of-two)
mm, sl[ou]b: improve memory accounting
mm, memcg: make scan aggression always exclude protection
mm, memcg: make memory.emin the baseline for utilisation determination
mm, memcg: proportional memory.{low,min} reclaim
mm/vmpressure.c: fix a signedness bug in vmpressure_register_event()
mm/page_alloc.c: fix a crash in free_pages_prepare()
mm/z3fold.c: claim page in the beginning of free
kernel/sysctl.c: do not override max_threads provided by userspace
memcg: only record foreign writebacks with dirty pages when memcg is not disabled
mm: fix -Wmissing-prototypes warnings
writeback: fix use-after-free in finish_writeback_work()
mm/memremap: drop unused SECTION_SIZE and SECTION_MASK
panic: ensure preemption is disabled during panic()
fs: ocfs2: fix a possible null-pointer dereference in ocfs2_info_scan_inode_alloc()
fs: ocfs2: fix a possible null-pointer dereference in ocfs2_write_end_nolock()
fs: ocfs2: fix possible null-pointer dereferences in ocfs2_xa_prepare_entry()
ocfs2: clear zero in unaligned direct IO
Partially revert 16db3d3f11 ("kernel/sysctl.c: threads-max observe
limits") because the patch is causing a regression to any workload which
needs to override the auto-tuning of the limit provided by kernel.
set_max_threads is implementing a boot time guesstimate to provide a
sensible limit of the concurrently running threads so that runaways will
not deplete all the memory. This is a good thing in general but there
are workloads which might need to increase this limit for an application
to run (reportedly WebSpher MQ is affected) and that is simply not
possible after the mentioned change. It is also very dubious to
override an admin decision by an estimation that doesn't have any direct
relation to correctness of the kernel operation.
Fix this by dropping set_max_threads from sysctl_max_threads so any
value is accepted as long as it fits into MAX_THREADS which is important
to check because allowing more threads could break internal robust futex
restriction. While at it, do not use MIN_THREADS as the lower boundary
because it is also only a heuristic for automatic estimation and admin
might have a good reason to stop new threads to be created even when
below this limit.
This became more severe when we switched x86 from 4k to 8k kernel
stacks. Starting since 6538b8ea88 ("x86_64: expand kernel stack to
16K") (3.16) we use THREAD_SIZE_ORDER = 2 and that halved the auto-tuned
value.
In the particular case
3.12
kernel.threads-max = 515561
4.4
kernel.threads-max = 200000
Neither of the two values is really insane on 32GB machine.
I am not sure we want/need to tune the max_thread value further. If
anything the tuning should be removed altogether if proven not useful in
general. But we definitely need a way to override this auto-tuning.
Link: http://lkml.kernel.org/r/20190922065801.GB18814@dhcp22.suse.cz
Fixes: 16db3d3f11 ("kernel/sysctl.c: threads-max observe limits")
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Heinrich Schuchardt <xypron.glpk@gmx.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Calling 'panic()' on a kernel with CONFIG_PREEMPT=y can leave the
calling CPU in an infinite loop, but with interrupts and preemption
enabled. From this state, userspace can continue to be scheduled,
despite the system being "dead" as far as the kernel is concerned.
This is easily reproducible on arm64 when booting with "nosmp" on the
command line; a couple of shell scripts print out a periodic "Ping"
message whilst another triggers a crash by writing to
/proc/sysrq-trigger:
| sysrq: Trigger a crash
| Kernel panic - not syncing: sysrq triggered crash
| CPU: 0 PID: 1 Comm: init Not tainted 5.2.15 #1
| Hardware name: linux,dummy-virt (DT)
| Call trace:
| dump_backtrace+0x0/0x148
| show_stack+0x14/0x20
| dump_stack+0xa0/0xc4
| panic+0x140/0x32c
| sysrq_handle_reboot+0x0/0x20
| __handle_sysrq+0x124/0x190
| write_sysrq_trigger+0x64/0x88
| proc_reg_write+0x60/0xa8
| __vfs_write+0x18/0x40
| vfs_write+0xa4/0x1b8
| ksys_write+0x64/0xf0
| __arm64_sys_write+0x14/0x20
| el0_svc_common.constprop.0+0xb0/0x168
| el0_svc_handler+0x28/0x78
| el0_svc+0x8/0xc
| Kernel Offset: disabled
| CPU features: 0x0002,24002004
| Memory Limit: none
| ---[ end Kernel panic - not syncing: sysrq triggered crash ]---
| Ping 2!
| Ping 1!
| Ping 1!
| Ping 2!
The issue can also be triggered on x86 kernels if CONFIG_SMP=n,
otherwise local interrupts are disabled in 'smp_send_stop()'.
Disable preemption in 'panic()' before re-enabling interrupts.
Link: http://lkml.kernel.org/r/20191002123538.22609-1-will@kernel.org
Link: https://lore.kernel.org/r/BX1W47JXPMR8.58IYW53H6M5N@dragonstone
Signed-off-by: Will Deacon <will@kernel.org>
Reported-by: Xogium <contact@xogium.me>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Feng Tang <feng.tang@intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are reports of users who use thread migrations between cgroups and
they report performance drop after d59cfc09c3 ("sched, cgroup: replace
signal_struct->group_rwsem with a global percpu_rwsem"). The effect is
pronounced on machines with more CPUs.
The migration is affected by forking noise happening in the background,
after the mentioned commit a migrating thread must wait for all
(forking) processes on the system, not only of its threadgroup.
There are several places that need to synchronize with migration:
a) do_exit,
b) de_thread,
c) copy_process,
d) cgroup_update_dfl_csses,
e) parallel migration (cgroup_{proc,thread}s_write).
In the case of self-migrating thread, we relax the synchronization on
cgroup_threadgroup_rwsem to avoid the cost of waiting. d) and e) are
excluded with cgroup_mutex, c) does not matter in case of single thread
migration and the executing thread cannot exec(2) or exit(2) while it is
writing into cgroup.threads. In case of do_exit because of signal
delivery, we either exit before the migration or finish the migration
(of not yet PF_EXITING thread) and die afterwards.
This patch handles only the case of self-migration by writing "0" into
cgroup.threads. For simplicity, we always take cgroup_threadgroup_rwsem
with numeric PIDs.
This change improves migration dependent workload performance similar
to per-signal_struct state.
Signed-off-by: Michal Koutný <mkoutny@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
We no longer take cgroup_mutex in cgroup_exit and the exiting tasks are
not moved to init_css_set, reflect that in several comments to prevent
confusion.
Signed-off-by: Michal Koutný <mkoutny@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Like commit 13d82fb77a ("cgroup: short-circuit cset_cgroup_from_root() on
the default hierarchy"), short-circuit current_cgns_cgroup_from_root() on
the default hierarchy.
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Pull dma-mapping regression fix from Christoph Hellwig:
"Revert an incorret hunk from a patch that caused problems on various
arm boards (Andrey Smirnov)"
* tag 'dma-mapping-5.4-1' of git://git.infradead.org/users/hch/dma-mapping:
dma-mapping: fix false positive warnings in dma_common_free_remap()
This reverts commit 85fbd722ad.
The commit was added as a quick band-aid for a hang that happened when a
block device was removed during system suspend. Now that bdi_wq is not
freezable anymore the hang should not be possible and we can get rid of
this hack by reverting it.
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull Kbuild fixes from Masahiro Yamada:
- remove unneeded ar-option and KBUILD_ARFLAGS
- remove long-deprecated SUBDIRS
- fix modpost to suppress false-positive warnings for UML builds
- fix namespace.pl to handle relative paths to ${objtree}, ${srctree}
- make setlocalversion work for /bin/sh
- make header archive reproducible
- fix some Makefiles and documents
* tag 'kbuild-fixes-v5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
kheaders: make headers archive reproducible
kbuild: update compile-test header list for v5.4-rc2
kbuild: two minor updates for Documentation/kbuild/modules.rst
scripts/setlocalversion: clear local variable to make it work for sh
namespace: fix namespace.pl script to support relative paths
video/logo: do not generate unneeded logo C files
video/logo: remove unneeded *.o pattern from clean-files
integrity: remove pointless subdir-$(CONFIG_...)
integrity: remove unneeded, broken attempt to add -fshort-wchar
modpost: fix static EXPORT_SYMBOL warnings for UML build
kbuild: correct formatting of header in kbuild module docs
kbuild: remove SUBDIRS support
kbuild: remove ar-option and KBUILD_ARFLAGS
Including rwlock.h directly will cause kernel builds to fail
if CONFIG_PREEMPT_RT is defined. The correct header file
(rwlock_rt.h OR rwlock.h) will be included by spinlock.h which
is included by locktorture.c anyway.
Remove the include of linux/rwlock.h.
Signed-off-by: Wolfgang M. Reimer <linuxball@gmail.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The rcu_torture_fwd_prog_nr() tests the ability of RCU to tolerate
in-kernel busy loops. It invokes rcu_torture_fwd_prog_cond_resched()
within its delay loop, which, in PREEMPT && NO_HZ_FULL kernels results
in the occasional direct call to schedule(). Now, this direct call to
schedule() is appropriate for call_rcu() flood testing, in which either
the kernel should restrain itself or userspace transitions will supply
the needed restraint. But in pure in-kernel loops, the occasional
cond_resched() should do the job.
This commit therefore makes rcu_torture_fwd_prog_nr() use cond_resched()
instead of rcu_torture_fwd_prog_cond_resched() in order to increase the
brutality of this aspect of rcutorture testing.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Currently, each of six different types of failure triggers a
single WARN_ON_ONCE(), and it is then necessary to stare at the
rcu_torture_stats(), Reader Pipe, and Reader Batch lines looking for
inappropriately non-zero values. This can be annoying and error-prone,
so this commit provides a separate WARN_ON_ONCE() for each of the
six error conditions and adds short comments to each to ease error
identification.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The variable rcu_perf_writer_state is declared and initialized,
but is never actually referenced. Remove it to clean code.
Signed-off-by: Ethan Hansen <1ethanhansen@gmail.com>
[ paulmck: Also removed unused macros assigned to that variable. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The strncmp() function is error-prone because it is easy to get the
length wrong, especially if the string is subject to change, especially
given the need to account for the terminating nul byte. This commit
therefore substitutes the newly introduced str_has_prefix(), which
does not require a separately specified length.
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
The function rcutorture_record_progress() is declared in rcu.h, but is
never used. This commit therefore removes rcutorture_record_progress()
to clean code.
Signed-off-by: Ethan Hansen <1ethanhansen@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
During an actual call_rcu() flood, there would be frequent trips to
userspace (in-kernel call_rcu() floods must be otherwise housebroken).
Userspace execution on nohz_full CPUs implies an RCU dyntick idle/not-idle
transition pair, so this commit adds emulation of that pair.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
CPU-hotplug removal operations run the multi_cpu_stop() function, which
relies on the scheduler to gain control from whatever is running on the
various online CPUs, including any nohz_full CPUs running long loops in
kernel-mode code. Lack of the scheduler-clock interrupt on such CPUs
can delay multi_cpu_stop() for several minutes and can also result in
RCU CPU stall warnings. This commit therefore causes CPU-hotplug removal
operations to enable the scheduler-clock interrupt on all online CPUs.
[ paulmck: Apply Joel Fernandes TICK_DEP_MASK_RCU->TICK_DEP_BIT_RCU fix. ]
[ paulmck: Apply simplifications suggested by Frederic Weisbecker. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
When multi_cpu_stop() loops waiting for other tasks, it can trigger an RCU
CPU stall warning. This can be misleading because what is instead needed
is information on whatever task is blocking multi_cpu_stop(). This commit
therefore inserts an RCU quiescent state into the multi_cpu_stop()
function's waitloop.
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Readers and callback flooders in the rcutorture stress-test suite run for
extended time periods by design. They do take pains to relinquish the
CPU from time to time, but in some cases this relies on the scheduler
being active, which in turn relies on the scheduler-clock interrupt
firing from time to time.
This commit therefore forces scheduling-clock interrupts within
these loops. While in the area, this commit also prevents
rcu_torture_reader()'s occasional timed sleeps from delaying shutdown.
[ paulmck: Apply Joel Fernandes TICK_DEP_MASK_RCU->TICK_DEP_BIT_RCU fix. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Callback invocation can run for a significant time period, and within
CONFIG_NO_HZ_FULL=y kernels, this period will be devoid of scheduler-clock
interrupts. In-kernel execution without such interrupts can cause all
manner of malfunction, with RCU CPU stall warnings being but one result.
This commit therefore forces scheduling-clock interrupts on whenever more
than a few RCU callbacks are invoked. Because offloaded callback invocation
can be preempted, this forcing is withdrawn on each context switch. This
in turn requires that the loop invoking RCU callbacks reiterate the forcing
periodically.
[ paulmck: Apply Joel Fernandes TICK_DEP_MASK_RCU->TICK_DEP_BIT_RCU fix. ]
[ paulmck: Remove NO_HZ_FULL check per Frederic Weisbecker feedback. ]
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
It turns out that rcutorture needs to ensure that the scheduling-clock
interrupt is enabled in CONFIG_NO_HZ_FULL kernels before starting on
CPU-bound in-kernel processing. This commit therefore exports
tick_nohz_dep_set_task(), tick_nohz_dep_clear_task(), and
tick_nohz_full_setup() to GPL kernel modules.
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
If a nohz_full CPU is looping in the kernel, the scheduling-clock tick
might nevertheless remain disabled. In !PREEMPT kernels, this can
prevent RCU's attempts to enlist the aid of that CPU's executions of
cond_resched(), which can in turn result in an arbitrarily delayed grace
period and thus an OOM. RCU therefore needs a way to enable a holdout
nohz_full CPU's scheduler-clock interrupt.
This commit therefore provides a new TICK_DEP_BIT_RCU value which RCU can
pass to tick_dep_set_cpu() and friends to force on the scheduler-clock
interrupt for a specified CPU or task. In some cases, rcutorture needs
to turn on the scheduler-clock tick, so this commit also exports the
relevant symbols to GPL-licensed modules.
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
