linux.git/kernel/sched/core.c, branch v4.18-rc2

sched/core / kcov: avoid kcov_area during task switch

2018-06-14T22:55:24+00:00

During a context switch, we first switch_mm() to the next task's mm,
then switch_to() that new task.  This means that vmalloc'd regions which
had previously been faulted in can transiently disappear in the context
of the prev task.

Functions instrumented by KCOV may try to access a vmalloc'd kcov_area
during this window, and as the fault handling code is instrumented, this
results in a recursive fault.

We must avoid accessing any kcov_area during this window.  We can do so
with a new flag in kcov_mode, set prior to switching the mm, and cleared
once the new task is live.  Since task_struct::kcov_mode isn't always a
specific enum kcov_mode value, this is made an unsigned int.

The manipulation is hidden behind kcov_{prepare,finish}_switch() helpers,
which are empty for !CONFIG_KCOV kernels.

The code uses macros because I can't use static inline functions without a
circular include dependency between  and ,
since the definition of task_struct uses things defined in 

Link: http://lkml.kernel.org/r/20180504135535.53744-4-mark.rutland@arm.com
Signed-off-by: Mark Rutland 
Acked-by: Andrey Ryabinin 
Cc: Dmitry Vyukov 
Cc: Ingo Molnar 
Cc: Peter Zijlstra 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

rseq: Introduce restartable sequences system call

2018-06-06T09:58:31+00:00

Expose a new system call allowing each thread to register one userspace
memory area to be used as an ABI between kernel and user-space for two
purposes: user-space restartable sequences and quick access to read the
current CPU number value from user-space.

* Restartable sequences (per-cpu atomics)

Restartables sequences allow user-space to perform update operations on
per-cpu data without requiring heavy-weight atomic operations.

The restartable critical sections (percpu atomics) work has been started
by Paul Turner and Andrew Hunter. It lets the kernel handle restart of
critical sections. [1] [2] The re-implementation proposed here brings a
few simplifications to the ABI which facilitates porting to other
architectures and speeds up the user-space fast path.

Here are benchmarks of various rseq use-cases.

Test hardware:

arm32: ARMv7 Processor rev 4 (v7l) "Cubietruck", 2-core
x86-64: Intel E5-2630 v3@2.40GHz, 16-core, hyperthreading

The following benchmarks were all performed on a single thread.

* Per-CPU statistic counter increment

                getcpu+atomic (ns/op)    rseq (ns/op)    speedup
arm32:                344.0                 31.4          11.0
x86-64:                15.3                  2.0           7.7

* LTTng-UST: write event 32-bit header, 32-bit payload into tracer
             per-cpu buffer

                getcpu+atomic (ns/op)    rseq (ns/op)    speedup
arm32:               2502.0                 2250.0         1.1
x86-64:               117.4                   98.0         1.2

* liburcu percpu: lock-unlock pair, dereference, read/compare word

                getcpu+atomic (ns/op)    rseq (ns/op)    speedup
arm32:                751.0                 128.5          5.8
x86-64:                53.4                  28.6          1.9

* jemalloc memory allocator adapted to use rseq

Using rseq with per-cpu memory pools in jemalloc at Facebook (based on
rseq 2016 implementation):

The production workload response-time has 1-2% gain avg. latency, and
the P99 overall latency drops by 2-3%.

* Reading the current CPU number

Speeding up reading the current CPU number on which the caller thread is
running is done by keeping the current CPU number up do date within the
cpu_id field of the memory area registered by the thread. This is done
by making scheduler preemption set the TIF_NOTIFY_RESUME flag on the
current thread. Upon return to user-space, a notify-resume handler
updates the current CPU value within the registered user-space memory
area. User-space can then read the current CPU number directly from
memory.

Keeping the current cpu id in a memory area shared between kernel and
user-space is an improvement over current mechanisms available to read
the current CPU number, which has the following benefits over
alternative approaches:

- 35x speedup on ARM vs system call through glibc
- 20x speedup on x86 compared to calling glibc, which calls vdso
  executing a "lsl" instruction,
- 14x speedup on x86 compared to inlined "lsl" instruction,
- Unlike vdso approaches, this cpu_id value can be read from an inline
  assembly, which makes it a useful building block for restartable
  sequences.
- The approach of reading the cpu id through memory mapping shared
  between kernel and user-space is portable (e.g. ARM), which is not the
  case for the lsl-based x86 vdso.

On x86, yet another possible approach would be to use the gs segment
selector to point to user-space per-cpu data. This approach performs
similarly to the cpu id cache, but it has two disadvantages: it is
not portable, and it is incompatible with existing applications already
using the gs segment selector for other purposes.

Benchmarking various approaches for reading the current CPU number:

ARMv7 Processor rev 4 (v7l)
Machine model: Cubietruck
- Baseline (empty loop):                                    8.4 ns
- Read CPU from rseq cpu_id:                               16.7 ns
- Read CPU from rseq cpu_id (lazy register):               19.8 ns
- glibc 2.19-0ubuntu6.6 getcpu:                           301.8 ns
- getcpu system call:                                     234.9 ns

x86-64 Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz:
- Baseline (empty loop):                                    0.8 ns
- Read CPU from rseq cpu_id:                                0.8 ns
- Read CPU from rseq cpu_id (lazy register):                0.8 ns
- Read using gs segment selector:                           0.8 ns
- "lsl" inline assembly:                                   13.0 ns
- glibc 2.19-0ubuntu6 getcpu:                              16.6 ns
- getcpu system call:                                      53.9 ns

- Speed (benchmark taken on v8 of patchset)

Running 10 runs of hackbench -l 100000 seems to indicate, contrary to
expectations, that enabling CONFIG_RSEQ slightly accelerates the
scheduler:

Configuration: 2 sockets * 8-core Intel(R) Xeon(R) CPU E5-2630 v3 @
2.40GHz (directly on hardware, hyperthreading disabled in BIOS, energy
saving disabled in BIOS, turboboost disabled in BIOS, cpuidle.off=1
kernel parameter), with a Linux v4.6 defconfig+localyesconfig,
restartable sequences series applied.

* CONFIG_RSEQ=n

avg.:      41.37 s
std.dev.:   0.36 s

* CONFIG_RSEQ=y

avg.:      40.46 s
std.dev.:   0.33 s

- Size

On x86-64, between CONFIG_RSEQ=n/y, the text size increase of vmlinux is
567 bytes, and the data size increase of vmlinux is 5696 bytes.

[1] https://lwn.net/Articles/650333/
[2] http://www.linuxplumbersconf.org/2013/ocw/system/presentations/1695/original/LPC%20-%20PerCpu%20Atomics.pdf

Signed-off-by: Mathieu Desnoyers 
Signed-off-by: Thomas Gleixner 
Acked-by: Peter Zijlstra (Intel) 
Cc: Joel Fernandes 
Cc: Catalin Marinas 
Cc: Dave Watson 
Cc: Will Deacon 
Cc: Andi Kleen 
Cc: "H . Peter Anvin" 
Cc: Chris Lameter 
Cc: Russell King 
Cc: Andrew Hunter 
Cc: Michael Kerrisk 
Cc: "Paul E . McKenney" 
Cc: Paul Turner 
Cc: Boqun Feng 
Cc: Josh Triplett 
Cc: Steven Rostedt 
Cc: Ben Maurer 
Cc: Alexander Viro 
Cc: linux-api@vger.kernel.org
Cc: Andy Lutomirski 
Cc: Andrew Morton 
Cc: Linus Torvalds 
Link: http://lkml.kernel.org/r/20151027235635.16059.11630.stgit@pjt-glaptop.roam.corp.google.com
Link: http://lkml.kernel.org/r/20150624222609.6116.86035.stgit@kitami.mtv.corp.google.com
Link: https://lkml.kernel.org/r/20180602124408.8430-3-mathieu.desnoyers@efficios.com

Merge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2018-06-05T00:45:38+00:00

Pull scheduler updates from Ingo Molnar:

 - power-aware scheduling improvements (Patrick Bellasi)

 - NUMA balancing improvements (Mel Gorman)

 - vCPU scheduling fixes (Rohit Jain)

* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/fair: Update util_est before updating schedutil
  sched/cpufreq: Modify aggregate utilization to always include blocked FAIR utilization
  sched/deadline/Documentation: Add overrun signal and GRUB-PA documentation
  sched/core: Distinguish between idle_cpu() calls based on desired effect, introduce available_idle_cpu()
  sched/wait: Include  in 
  sched/numa: Stagger NUMA balancing scan periods for new threads
  sched/core: Don't schedule threads on pre-empted vCPUs
  sched/fair: Avoid calling sync_entity_load_avg() unnecessarily
  sched/fair: Rearrange select_task_rq_fair() to optimize it

Merge branch 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2018-06-04T22:54:04+00:00

Pull RCU updates from Ingo Molnar:

 - updates to the handling of expedited grace periods

 - updates to reduce lock contention in the rcu_node combining tree

   [ These are in preparation for the consolidation of RCU-bh,
     RCU-preempt, and RCU-sched into a single flavor, which was
     requested by Linus in response to a security flaw whose root cause
     included confusion between the multiple flavors of RCU ]

 - torture-test updates that save their users some time and effort

 - miscellaneous fixes

* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (44 commits)
  rcu/x86: Provide early rcu_cpu_starting() callback
  torture: Make kvm-find-errors.sh find build warnings
  rcutorture: Abbreviate kvm.sh summary lines
  rcutorture: Print end-of-test state in kvm.sh summary
  rcutorture: Print end-of-test state
  torture: Fold parse-torture.sh into parse-console.sh
  torture: Add a script to edit output from failed runs
  rcu: Update list of rcu_future_grace_period() trace events
  rcu: Drop early GP request check from rcu_gp_kthread()
  rcu: Simplify and inline cpu_needs_another_gp()
  rcu: The rcu_gp_cleanup() function does not need cpu_needs_another_gp()
  rcu: Make rcu_start_this_gp() check for out-of-range requests
  rcu: Add funnel locking to rcu_start_this_gp()
  rcu: Make rcu_start_future_gp() caller select grace period
  rcu: Inline rcu_start_gp_advanced() into rcu_start_future_gp()
  rcu: Clear request other than RCU_GP_FLAG_INIT at GP end
  rcu: Cleanup, don't put ->completed into an int
  rcu: Switch __rcu_process_callbacks() to rcu_accelerate_cbs()
  rcu: Avoid __call_rcu_core() root rcu_node ->lock acquisition
  rcu: Make rcu_migrate_callbacks wake GP kthread when needed
  ...

sched/core: Require cpu_active() in select_task_rq(), for user tasks

2018-05-31T10:24:25+00:00

select_task_rq() is used in a few paths to select the CPU upon which a
thread should be run - for example it is used by try_to_wake_up() & by
fork or exec balancing. As-is it allows use of any online CPU that is
present in the task's cpus_allowed mask.

This presents a problem because there is a period whilst CPUs are
brought online where a CPU is marked online, but is not yet fully
initialized - ie. the period where CPUHP_AP_ONLINE_IDLE <= state <
CPUHP_ONLINE. Usually we don't run any user tasks during this window,
but there are corner cases where this can happen. An example observed
is:

  - Some user task A, running on CPU X, forks to create task B.

  - sched_fork() calls __set_task_cpu() with cpu=X, setting task B's
    task_struct::cpu field to X.

  - CPU X is offlined.

  - Task A, currently somewhere between the __set_task_cpu() in
    copy_process() and the call to wake_up_new_task(), is migrated to
    CPU Y by migrate_tasks() when CPU X is offlined.

  - CPU X is onlined, but still in the CPUHP_AP_ONLINE_IDLE state. The
    scheduler is now active on CPU X, but there are no user tasks on
    the runqueue.

  - Task A runs on CPU Y & reaches wake_up_new_task(). This calls
    select_task_rq() with cpu=X, taken from task B's task_struct,
    and select_task_rq() allows CPU X to be returned.

  - Task A enqueues task B on CPU X's runqueue, via activate_task() &
    enqueue_task().

  - CPU X now has a user task on its runqueue before it has reached the
    CPUHP_ONLINE state.

In most cases, the user tasks that schedule on the newly onlined CPU
have no idea that anything went wrong, but one case observed to be
problematic is if the task goes on to invoke the sched_setaffinity
syscall. The newly onlined CPU reaches the CPUHP_AP_ONLINE_IDLE state
before the CPU that brought it online calls stop_machine_unpark(). This
means that for a portion of the window of time between
CPUHP_AP_ONLINE_IDLE & CPUHP_ONLINE the newly onlined CPU's struct
cpu_stopper has its enabled field set to false. If a user thread is
executed on the CPU during this window and it invokes sched_setaffinity
with a CPU mask that does not include the CPU it's running on, then when
__set_cpus_allowed_ptr() calls stop_one_cpu() intending to invoke
migration_cpu_stop() and perform the actual migration away from the CPU
it will simply return -ENOENT rather than calling migration_cpu_stop().
We then return from the sched_setaffinity syscall back to the user task
that is now running on a CPU which it just asked not to run on, and
which is not present in its cpus_allowed mask.

This patch resolves the problem by having select_task_rq() enforce that
user tasks run on CPUs that are active - the same requirement that
select_fallback_rq() already enforces. This should ensure that newly
onlined CPUs reach the CPUHP_AP_ACTIVE state before being able to
schedule user tasks, and also implies that bringup_wait_for_ap() will
have called stop_machine_unpark() which resolves the sched_setaffinity
issue above.

I haven't yet investigated them, but it may be of interest to review
whether any of the actions performed by hotplug states between
CPUHP_AP_ONLINE_IDLE & CPUHP_AP_ACTIVE could have similar unintended
effects on user tasks that might schedule before they are reached, which
might widen the scope of the problem from just affecting the behaviour
of sched_setaffinity.

Signed-off-by: Paul Burton 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Linus Torvalds 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/20180526154648.11635-2-paul.burton@mips.com
Signed-off-by: Ingo Molnar

sched/core: Fix rules for running on online && !active CPUs

2018-05-31T10:24:24+00:00

As already enforced by the WARN() in __set_cpus_allowed_ptr(), the rules
for running on an online && !active CPU are stricter than just being a
kthread, you need to be a per-cpu kthread.

If you're not strictly per-CPU, you have better CPUs to run on and
don't need the partially booted one to get your work done.

The exception is to allow smpboot threads to bootstrap the CPU itself
and get kernel 'services' initialized before we allow userspace on it.

Signed-off-by: Peter Zijlstra (Intel) 
Cc: Linus Torvalds 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Steven Rostedt 
Cc: Tejun Heo 
Cc: Thomas Gleixner 
Fixes: 955dbdf4ce87 ("sched: Allow migrating kthreads into online but inactive CPUs")
Link: http://lkml.kernel.org/r/20170725165821.cejhb7v2s3kecems@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar

Merge branch 'for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu into core/rcu

2018-05-16T07:34:51+00:00

 - Updates to the handling of expedited grace periods, perhaps most
   notably parallelizing their initialization.  Other changes
   include fixes from Boqun Feng.

 - Miscellaneous fixes.  These include an nvme fix from Nitzan Carmi
   that I am carrying because it depends on a new SRCU function
   cleanup_srcu_struct_quiesced().  This branch also includes fixes
   from Byungchul Park and Yury Norov.

 - Updates to reduce lock contention in the rcu_node combining tree.
   These are in preparation for the consolidation of RCU-bh,
   RCU-preempt, and RCU-sched into a single flavor, which was
   requested by Linus Torvalds in response to a security flaw
   whose root cause included confusion between the multiple flavors
   of RCU.

 - Torture-test updates that save their users some time and effort.

Conflicts:
	drivers/nvme/host/core.c

Signed-off-by: Ingo Molnar

softirq: Eliminate unused cond_resched_softirq() macro

2018-05-15T17:27:35+00:00

The cond_resched_softirq() macro is not used anywhere in mainline, so
this commit simplifies the kernel by eliminating it.

Suggested-by: Eric Dumazet 
Signed-off-by: Paul E. McKenney 
Cc: Ingo Molnar 
Acked-by: Peter Zijlstra (Intel) 
Reviewed-by: Eric Dumazet 
Tested-by: Nicholas Piggin

sched/core: Distinguish between idle_cpu() calls based on desired effect, introduce available_idle_cpu()

2018-05-14T07:12:26+00:00

In the following commit:

  247f2f6f3c70 ("sched/core: Don't schedule threads on pre-empted vCPUs")

... we distinguish between idle_cpu() when the vCPU is not running for
scheduling threads.

However, the idle_cpu() function is used in other places for
actually checking whether the state of the CPU is idle or not.

Hence split the use of that function based on the desired return value,
by introducing the available_idle_cpu() function.

This fixes a (slight) regression in that initial vCPU commit, because
some code paths (like the load-balancer) don't care and shouldn't care
if the vCPU is preempted or not, they just want to know if there's any
tasks on the CPU.

Signed-off-by: Rohit Jain 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Linus Torvalds 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: dhaval.giani@oracle.com
Cc: linux-kernel@vger.kernel.org
Cc: matt@codeblueprint.co.uk
Cc: steven.sistare@oracle.com
Cc: subhra.mazumdar@oracle.com
Link: http://lkml.kernel.org/r/1525883988-10356-1-git-send-email-rohit.k.jain@oracle.com
Signed-off-by: Ingo Molnar

sched/numa: Stagger NUMA balancing scan periods for new threads

2018-05-14T07:12:24+00:00

Threads share an address space and each can change the protections of the
same address space to trap NUMA faults. This is redundant and potentially
counter-productive as any thread doing the update will suffice. Potentially
only one thread is required but that thread may be idle or it may not have
any locality concerns and pick an unsuitable scan rate.

This patch uses independent scan period but they are staggered based on
the number of address space users when the thread is created.  The intent
is that threads will avoid scanning at the same time and have a chance
to adapt their scan rate later if necessary. This reduces the total scan
activity early in the lifetime of the threads.

The different in headline performance across a range of machines and
workloads is marginal but the system CPU usage is reduced as well as overall
scan activity.  The following is the time reported by NAS Parallel Benchmark
using unbound openmp threads and a D size class:

			      4.17.0-rc1             4.17.0-rc1
				 vanilla           stagger-v1r1
	Time bt.D      442.77 (   0.00%)      419.70 (   5.21%)
	Time cg.D      171.90 (   0.00%)      180.85 (  -5.21%)
	Time ep.D       33.10 (   0.00%)       32.90 (   0.60%)
	Time is.D        9.59 (   0.00%)        9.42 (   1.77%)
	Time lu.D      306.75 (   0.00%)      304.65 (   0.68%)
	Time mg.D       54.56 (   0.00%)       52.38 (   4.00%)
	Time sp.D     1020.03 (   0.00%)      903.77 (  11.40%)
	Time ua.D      400.58 (   0.00%)      386.49 (   3.52%)

Note it's not a universal win but we have no prior knowledge of which
thread matters but the number of threads created often exceeds the size
of the node when the threads are not bound. However, there is a reducation
of overall system CPU usage:

				    4.17.0-rc1             4.17.0-rc1
				       vanilla           stagger-v1r1
	sys-time-bt.D         48.78 (   0.00%)       48.22 (   1.15%)
	sys-time-cg.D         25.31 (   0.00%)       26.63 (  -5.22%)
	sys-time-ep.D          1.65 (   0.00%)        0.62 (  62.42%)
	sys-time-is.D         40.05 (   0.00%)       24.45 (  38.95%)
	sys-time-lu.D         37.55 (   0.00%)       29.02 (  22.72%)
	sys-time-mg.D         47.52 (   0.00%)       34.92 (  26.52%)
	sys-time-sp.D        119.01 (   0.00%)      109.05 (   8.37%)
	sys-time-ua.D         51.52 (   0.00%)       45.13 (  12.40%)

NUMA scan activity is also reduced:

	NUMA alloc local               1042828     1342670
	NUMA base PTE updates        140481138    93577468
	NUMA huge PMD updates           272171      180766
	NUMA page range updates      279832690   186129660
	NUMA hint faults               1395972     1193897
	NUMA hint local faults          877925      855053
	NUMA hint local percent             62          71
	NUMA pages migrated           12057909     9158023

Similar observations are made for other thread-intensive workloads. System
CPU usage is lower even though the headline gains in performance tend to be
small. For example, specjbb 2005 shows almost no difference in performance
but scan activity is reduced by a third on a 4-socket box. I didn't find
a workload (thread intensive or otherwise) that suffered badly.

Signed-off-by: Mel Gorman 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Linus Torvalds 
Cc: Matt Fleming 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/20180504154109.mvrha2qo5wdl65vr@techsingularity.net
Signed-off-by: Ingo Molnar