linux-stable.git/kernel/sched/cputime.c, branch v4.7

sched/cputime: Fix steal_account_process_tick() to always return jiffies

2016-03-08T11:24:56+00:00

The callers of steal_account_process_tick() expect it to return
whether a jiffy should be considered stolen or not.

Currently the return value of steal_account_process_tick() is in
units of cputime, which vary between either jiffies or nsecs
depending on CONFIG_VIRT_CPU_ACCOUNTING_GEN.

If cputime has nsecs granularity and there is a tiny amount of
stolen time (a few nsecs, say) then we will consider the entire
tick stolen and will not account the tick on user/system/idle,
causing /proc/stats to show invalid data.

The fix is to change steal_account_process_tick() to accumulate
the stolen time and only account it once it's worth a jiffy.

(Thanks to Frederic Weisbecker for suggestions to fix a bug in my
first version of the patch.)

Signed-off-by: Chris Friesen 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Thomas Gleixner 
Cc: 
Cc: Frederic Weisbecker 
Cc: Linus Torvalds 
Cc: Peter Zijlstra 
Link: http://lkml.kernel.org/r/56DBBDB8.40305@mail.usask.ca
Signed-off-by: Ingo Molnar

sched, time: Switch VIRT_CPU_ACCOUNTING_GEN to jiffy granularity

2016-02-29T08:53:10+00:00

When profiling syscall overhead on nohz-full kernels,
after removing __acct_update_integrals() from the profile,
native_sched_clock() remains as the top CPU user. This can be
reduced by moving VIRT_CPU_ACCOUNTING_GEN to jiffy granularity.

This will reduce timing accuracy on nohz_full CPUs to jiffy
based sampling, just like on normal CPUs. It results in
totally removing native_sched_clock from the profile, and
significantly speeding up the syscall entry and exit path,
as well as irq entry and exit, and KVM guest entry & exit.

Additionally, only call the more expensive functions (and
advance the seqlock) when jiffies actually changed.

This code relies on another CPU advancing jiffies when the
system is busy. On a nohz_full system, this is done by a
housekeeping CPU.

A microbenchmark calling an invalid syscall number 10 million
times in a row speeds up an additional 30% over the numbers
with just the previous patches, for a total speedup of about
40% over 4.4 and 4.5-rc1.

Run times for the microbenchmark:

 4.4				3.8 seconds
 4.5-rc1			3.7 seconds
 4.5-rc1 + first patch		3.3 seconds
 4.5-rc1 + first 3 patches	3.1 seconds
 4.5-rc1 + all patches		2.3 seconds

A non-NOHZ_FULL cpu (not the housekeeping CPU):

 all kernels			1.86 seconds

Signed-off-by: Rik van Riel 
Signed-off-by: Peter Zijlstra (Intel) 
Reviewed-by: Thomas Gleixner 
Cc: Linus Torvalds 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: clark@redhat.com
Cc: eric.dumazet@gmail.com
Cc: fweisbec@gmail.com
Cc: luto@amacapital.net
Link: http://lkml.kernel.org/r/1455152907-18495-5-git-send-email-riel@redhat.com
Signed-off-by: Ingo Molnar

Merge tag 'for-linus-4.5-rc0-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

2016-01-12T21:05:36+00:00

Pull xen updates from David Vrabel:
 "Xen features and fixes for 4.5-rc0:

   - Stolen ticks and PV wallclock support for arm/arm64

   - Add grant copy ioctl to gntdev device"

* tag 'for-linus-4.5-rc0-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/gntdev: add ioctl for grant copy
  x86/xen: don't reset vcpu_info on a cancelled suspend
  xen/gntdev: constify mmu_notifier_ops structures
  xen/grant-table: constify gnttab_ops structure
  xen/time: use READ_ONCE
  xen/x86: convert remaining timespec to timespec64 in xen_pvclock_gtod_notify
  xen/x86: support XENPF_settime64
  xen/arm: set the system time in Xen via the XENPF_settime64 hypercall
  xen/arm: introduce xen_read_wallclock
  arm: extend pvclock_wall_clock with sec_hi
  xen: introduce XENPF_settime64
  xen/arm: introduce HYPERVISOR_platform_op on arm and arm64
  xen: rename dom0_op to platform_op
  xen/arm: account for stolen ticks
  arm64: introduce CONFIG_PARAVIRT, PARAVIRT_TIME_ACCOUNTING and pv_time_ops
  arm: introduce CONFIG_PARAVIRT, PARAVIRT_TIME_ACCOUNTING and pv_time_ops
  missing include asm/paravirt.h in cputime.c
  xen: move xen_setup_runstate_info and get_runstate_snapshot to drivers/xen/time.c

missing include asm/paravirt.h in cputime.c

2015-12-21T14:40:53+00:00

Add include asm/paravirt.h to cputime.c, as steal_account_process_tick
calls paravirt_steal_clock, which is defined in asm/paravirt.h.

The ifdef CONFIG_PARAVIRT is necessary because not all archs have an
asm/paravirt.h to include.

The reason why currently cputime.c compiles, even though include
 is missing, is that on x86 asm/paravirt.h is included
by one of the other headers included in kernel/sched/cputime.c:

On arm and arm64, where I am about to introduce asm/paravirt.h and
stolen time support, without #include  in cputime.c, I
would get an error.

Signed-off-by: Stefano Stabellini 
Acked-by: Peter Zijlstra (Intel)

sched/cputime: Convert vtime_seqlock to seqcount

2015-12-04T09:34:46+00:00

The cputime can only be updated by the current task itself, even in
vtime case. So we can safely use seqcount instead of seqlock as there
is no writer concurrency involved.

Signed-off-by: Frederic Weisbecker 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Chris Metcalf 
Cc: Christoph Lameter 
Cc: Hiroshi Shimamoto 
Cc: Linus Torvalds 
Cc: Luiz Capitulino 
Cc: Mike Galbraith 
Cc: Paul E . McKenney 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Rik van Riel 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/1447948054-28668-8-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar

sched/cputime: Introduce vtime accounting check for readers

2015-12-04T09:34:45+00:00

Readers need to know if vtime runs at all on some CPU somewhere, this
is a fast-path check to determine if we need to check further the need
to add up any tickless cputime delta.

This fast path check uses context tracking state because vtime is tied
to context tracking as of now. This check appears to be confusing though
so lets use a vtime function that deals with context tracking details
in vtime implementation instead.

Signed-off-by: Frederic Weisbecker 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Chris Metcalf 
Cc: Christoph Lameter 
Cc: Hiroshi Shimamoto 
Cc: Linus Torvalds 
Cc: Luiz Capitulino 
Cc: Mike Galbraith 
Cc: Paul E . McKenney 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Rik van Riel 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/1447948054-28668-7-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar

sched/cputime: Rename vtime_accounting_enabled() to vtime_accounting_cpu_enabled()

2015-12-04T09:34:45+00:00

vtime_accounting_enabled() checks if vtime is running on the current CPU
and is as such a misnomer. Lets rename it to a function that reflect its
locality. We are going to need the current name for a function that tells
if vtime runs at all on some CPU.

Signed-off-by: Frederic Weisbecker 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Chris Metcalf 
Cc: Christoph Lameter 
Cc: Hiroshi Shimamoto 
Cc: Linus Torvalds 
Cc: Luiz Capitulino 
Cc: Mike Galbraith 
Cc: Paul E . McKenney 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Rik van Riel 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/1447948054-28668-6-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar

sched/cputime: Correctly handle task guest time on housekeepers

2015-12-04T09:34:44+00:00

When a task runs on a housekeeper (a CPU running with the periodic tick
with neighbours running tickless), it doesn't account cputime using vtime
but relies on the tick. Such a task has its vtime_snap_whence value set
to VTIME_INACTIVE.

Readers won't handle that correctly though. As long as vtime is running
on some CPU, readers incorretly assume that vtime runs on all CPUs and
always compute the tickless cputime delta, which is only junk on
housekeepers.

So lets fix this with checking that the target runs on a vtime CPU through
the appropriate state check before computing the tickless delta.

Signed-off-by: Frederic Weisbecker 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Chris Metcalf 
Cc: Christoph Lameter 
Cc: Hiroshi Shimamoto 
Cc: Linus Torvalds 
Cc: Luiz Capitulino 
Cc: Mike Galbraith 
Cc: Paul E . McKenney 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Rik van Riel 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/1447948054-28668-5-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar

sched/cputime: Clarify vtime symbols and document them

2015-12-04T09:34:44+00:00

VTIME_SLEEPING state happens either when:

1) The task is sleeping and no tickless delta is to be added on the task
   cputime stats.
2) The CPU isn't running vtime at all, so the same properties of 1) applies.

Lets rename the vtime symbol to reflect both states.

Signed-off-by: Frederic Weisbecker 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Chris Metcalf 
Cc: Christoph Lameter 
Cc: Hiroshi Shimamoto 
Cc: Linus Torvalds 
Cc: Luiz Capitulino 
Cc: Mike Galbraith 
Cc: Paul E . McKenney 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Rik van Riel 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/1447948054-28668-4-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar

sched/cputime: Remove extra cost in task_cputime()

2015-12-04T09:34:43+00:00

There is an extra cost in task_cputime() and task_cputime_scaled() when
nohz_full is not activated. When vtime accounting is not enabled, we
don't need to get deltas of utime and stime under vtime seqlock.

This patch removes that cost with adding a shortcut route if vtime
accounting is not enabled.

Use context_tracking_is_enabled() to check if vtime is accounting on
some cpu, in which case only we need to check the tickless cputime delta.

Signed-off-by: Hiroshi Shimamoto 
Signed-off-by: Frederic Weisbecker 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Chris Metcalf 
Cc: Christoph Lameter 
Cc: Linus Torvalds 
Cc: Luiz Capitulino 
Cc: Mike Galbraith 
Cc: Paul E . McKenney 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Rik van Riel 
Cc: Thomas Gleixner 
Link: http://lkml.kernel.org/r/1447948054-28668-3-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar