Skip to content

hrtimer: Reprogramming and clock_was_set() overhaul

Thomas Gleixner lkml 2 versions – Applied

Status from this message: The following commit has been merged into the timers/core branch of tip:

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v2 Latest10 patches 9 repliesApplied
  2. v11 patch 6 repliesSuperseded

Changes in v2 from the cover letter

- Pick up Peter's consolidation patch
- Polish a few things

Patches in v2 Applied

#Subject and filesSizeRepliesApplied
1/10hrtimer: Avoid double reprogramming in __hrtimer_start_range_ns()kernel/time/hrtimer.c+53 -70 replies
2/10hrtimer: Consolidate reprogramming codekernel/time/hrtimer.c+29 -4313 replies
3/10hrtimer: Ensure timerfd notification for HIGHRES=ninclude/linux/hrtimer.h kernel/time/hrtimer.c kernel/time/tick-internal.h+19 -210 replies
4/10hrtimer: Force clock_was_set() handling for the HIGHRES=n, NOHZ=y casekernel/time/hrtimer.c+59 -280 replies
5/10timerfd: Provide timerfd_resume()fs/timerfd.c include/linux/hrtimer.h+18 -00 replies
6/10timekeeping: Distangle resume and clock-was-set eventsinclude/linux/hrtimer.h kernel/time/hrtimer.c kernel/time/tick-common.c +2 more+18 -110 replies
7/10time/timekeeping: Avoid invoking clock_was_set() twicekernel/time/timekeeping.c+10 -80 replies
8/10hrtimer: Add bases argument to clock_was_set()kernel/time/hrtimer.c kernel/time/tick-internal.h kernel/time/timekeeping.c+17 -100 replies
9/10hrtimer: Avoid unnecessary SMP function calls in clock_was_set()kernel/time/hrtimer.c+33 -20 replies
10/10hrtimer: Avoid more SMP function calls in clock_was_set()kernel/time/hrtimer.c+65 -90 replies

Reviews trailers per patch

None of the 10 patches has a Reviewed-by or Acked-by.

PatchMike GalbraithMarek Szyprowski
1/10
2/10T-b T-b
3/10
4/10
5/10
6/10
7/10
8/10
9/10
10/10

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v2 thread

Cover letter [patch V2 00/10] hrtimer: Reprogramming and clock_was_set() overhaul

Folks!

This is the second version of the clock_was_set() overhaul together with the avoidance of double reprogramming when rearming an enqueued timer.

Version 1 of those patches can be found here:

https://lore.kernel.org/r/[email protected]
and
https://lore.kernel.org/r/[email protected]

Both the reprogramming logic and the clock_was_set() logic are not really optimized. The series addresses this.

- Avoid double reprogramming when an enqueued timer is rearmed

- Consolidate the reprogramming variants (Peter Zijlstra)

- Fix the cases where the clock_was_set() handling is incorrect depending on configuration or runtime conditions.

- Distangle the resume notification and the clock-was-set mechanism to prepare for IPI avoidance

- Adopt Marcelo's patch to the modified code and add some more smarts on top.

The series is also available from git:

git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git hrtimer

Changes vs V1:

- Pick up Peter's consolidation patch
- Polish a few things

Thanks,

tglx

---
 fs/timerfd.c                |   16 ++
 include/linux/hrtimer.h     |    8 -
 kernel/time/hrtimer.c       |  337 +++++++++++++++++++++++++++++++-------------
 kernel/time/tick-common.c   |    7 
 kernel/time/tick-internal.h |   12 +
 kernel/time/timekeeping.c   |   36 ++--
 6 files changed, 299 insertions(+), 117 deletions(-)