Skip to content

tick: Optimize broadcast handling and prevent expiry ping pong

Thomas Gleixner linux-arm-kernel lkml 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest7 patches 4 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/7tick: Call tick_init lateinit/main.c+1 -10 replies
2/7tick: Convert broadcast cpu bitmaps to cpumask_var_tkernel/time/tick-broadcast.c kernel/time/tick-common.c kernel/time/tick-internal.h+46 -441 reply
3/7tick: Avoid programming the local cpu timer if broadcast pendingkernel/time/tick-broadcast.c+29 -40 replies
4/7tick: Handle broadcast wakeup of multiple cpuskernel/time/tick-broadcast.c+58 -12 replies
5/7tick: Provide a check for a forced broadcast pendinginclude/linux/clockchips.h kernel/time/tick-broadcast.c+18 -00 replies
6/7arm: Use tick broadcast expired checkarch/arm/kernel/process.c+10 -10 replies
7/7x86: Use tick broadcast expired checkarch/x86/kernel/process.c+12 -00 replies

Reviews trailers per patch

None of the 7 patches has a Reviewed-by or Acked-by.

PatchSantosh ShilimkarLorenzo Pieralisi
SeriesT-b
1/7
2/7
3/7
4/7T-b
5/7
6/7
7/7

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

Read the v1 thread

Cover letter [patch 0/7] tick: Optimize broadcast handling and prevent expiry ping pong

Jason decoded a problem related to the broadcast timer mode. The reprogramming of the cpu local timer causes a huge number of retries. Also there is a situation where the CPU which does not handle the broadcast timer interrupt exits and reenters broadcast mode before the broadcast interrupt got handled by another CPU. This can lead to an interesting ping pong of the broadcast and the cpu local timer code.

This series addresses these problems. The first two patches convert the broadcast code to proper cpumask_var_t instead of adding more bitmaps later.

The rest of the series is adopted from the quick patches which I posted earlier while discussing the issue with Jason et. al.

Please give it a proper testing on your affected hardware.

Thanks,

tglx