Re: [PATCH 0/7] cputime: Full dynticks task/cputime accounting v5

From: Paul E. McKenney
Date: Fri Jan 25 2013 - 12:42:40 EST


On Thu, Jan 17, 2013 at 02:24:28PM +0100, Frederic Weisbecker wrote:
> Hi,
>
> This patchset was formerly called "Generic virtual based cputime accounting"
> but I changed the name to better reflect what it's all about.
>
> This is the part of the full dynticks patchset that handles the task and CPU
> time accounting without the tick.
>
> It's available at:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks.git
> vtime/generic-v5
>
> There have been a lot of changes since latest full dynticks patchset:
>
> - Use sched_clock() instead of jiffies for better precision. And because
> jiffies granularity was requiring too much dirty hacks to make it working.
> (Thanks Steven for the suggestion).
>
> - To fully support sched_clock(), use a per nanosec granularity in cputime_t
>
> - Fix a deadlock with tsacct
>
> - Better description of the Kconfig
>
> - Don't scale generic vtime accounting against CFS precise task time accounting.
> Now that we use sched_clock() we should have precise enough statistics.
>
> - Consolidate some code in the reader side
>
> - Make the reader irqsafe (fix potential deadlock against writers)
>
>
> Most of the work is done. There are just a few small things I need to care about
> to complete the full cputime support:
>
> * Also handle dynticks guest time accounting (p->gtime and p->utime). This
> require to bring a new accessor: task_gtime() and to add guest_enter() and
> guest_exit() on the context tracking subsystem. Then make the same treatment
> than with utime and stime. No big deal.
>
> * New tasks start their cputime accounting from schedule() on sched-in time
> (see arch_vtime_task_switch()). But I fear that idle tasks don't fall into this
> category. That's a small detail but I hope I can find some generic way to handle it.
> The boot init idle task is however handled.
>
> * Per nsec granularity requires that cputime_t directly maps to cputime64_t. This
> is fine on 64 bits archs and mostly fine in 32 bits archs as well because readers
> use seqlock to synchronize against writers. But reading struct kernel_cputstat
> is not protected. Not sure yet how to solve that.
>
> So I'll likely focus on these items on the next version. Let's hope we can make
> it ready for the next merge window.

Queued for 3.9, thank you!

Thanx, Paul

> Thanks.
>
> ---
> Frederic Weisbecker (7):
> context_tracking: Export context state for generic vtime
> cputime: Librarize per nsecs resolution cputime definitions
> cputime: Move default nsecs_to_cputime() to jiffies based cputime file
> cputime: Generic on-demand virtual cputime accounting
> cputime: Allow dynamic switch between tick/virtual based cputime accounting
> cputime: Use accessors to read task cputime stats
> cputime: Safely read cputime of full dynticks CPUs
>
>
> arch/alpha/kernel/osf_sys.c | 6 +-
> arch/ia64/include/asm/cputime.h | 92 +----------
> arch/ia64/include/asm/thread_info.h | 4 +-
> arch/ia64/include/asm/xen/minstate.h | 2 +-
> arch/ia64/kernel/asm-offsets.c | 2 +-
> arch/ia64/kernel/entry.S | 16 +-
> arch/ia64/kernel/fsys.S | 4 +-
> arch/ia64/kernel/head.S | 4 +-
> arch/ia64/kernel/ivt.S | 8 +-
> arch/ia64/kernel/minstate.h | 2 +-
> arch/ia64/kernel/time.c | 4 +-
> arch/powerpc/configs/chroma_defconfig | 2 +-
> arch/powerpc/configs/corenet64_smp_defconfig | 2 +-
> arch/powerpc/configs/pasemi_defconfig | 2 +-
> arch/powerpc/include/asm/cputime.h | 6 +-
> arch/powerpc/include/asm/lppaca.h | 2 +-
> arch/powerpc/include/asm/ppc_asm.h | 4 +-
> arch/powerpc/kernel/entry_64.S | 4 +-
> arch/powerpc/kernel/time.c | 4 +-
> arch/powerpc/platforms/pseries/dtl.c | 6 +-
> arch/powerpc/platforms/pseries/setup.c | 6 +-
> arch/s390/kernel/vtime.c | 6 +-
> arch/x86/kernel/apm_32.c | 11 +-
> drivers/isdn/mISDN/stack.c | 7 +-
> fs/binfmt_elf.c | 8 +-
> fs/binfmt_elf_fdpic.c | 7 +-
> include/asm-generic/cputime.h | 66 +-------
> include/asm-generic/cputime_jiffies.h | 72 +++++++++
> include/asm-generic/cputime_nsecs.h | 104 ++++++++++++
> include/linux/context_tracking.h | 28 ++++
> include/linux/hardirq.h | 4 +-
> include/linux/init_task.h | 11 ++
> include/linux/kernel_stat.h | 2 +-
> include/linux/sched.h | 34 ++++
> include/linux/tsacct_kern.h | 3 +
> include/linux/vtime.h | 51 ++++--
> init/Kconfig | 23 +++-
> kernel/acct.c | 6 +-
> kernel/context_tracking.c | 22 +--
> kernel/cpu.c | 4 +-
> kernel/delayacct.c | 7 +-
> kernel/exit.c | 6 +-
> kernel/fork.c | 14 ++
> kernel/posix-cpu-timers.c | 28 +++-
> kernel/sched/cputime.c | 219 +++++++++++++++++++++++---
> kernel/signal.c | 12 +-
> kernel/softirq.c | 6 +-
> kernel/time/tick-sched.c | 5 +-
> kernel/tsacct.c | 44 ++++--
> 49 files changed, 696 insertions(+), 296 deletions(-)
> create mode 100644 include/asm-generic/cputime_jiffies.h
> create mode 100644 include/asm-generic/cputime_nsecs.h
>
> --
> 1.7.5.4
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/