Re: [patch 00/15] clocksource / timekeeping rework V4 (resend V3 + bug fix)

[john stultz <johnstul@us.ibm.com>]

On Fri, 2009-08-14 at 15:47 +0200, Martin Schwidefsky wrote:
> Greetings,
> version 4 of the clocksource / timekeeping cleanup patches. If you
> wonder why version 3 did not show up on the mailing list: I had a
> hiccup with quilt which did strange things to the message ids and
> the post to lkml got rejected. Sorry for the spamming of the people
> on Cc.
> 
> John pointed out a subtle bug in v3 of the patch set with the reset of
> the cycle_last value in timekeeping_resume. Fixed in v4.
> 
> The latest additions are:
> 1) Change read_persistent_clock to return a struct timespec instead of
>    an unsigned long with the number of seconds.
> 2) Introduce read_boot_clock to initialize wall_to_monotonic.
> 3) Move the reset of cycle_last to zero from the timekeeping code to
>    the TSC code.
> 
> The patch set is based on todays upstream tree plus the patches from
> the tip tree, if anyone wants to try them you need to pull from the
> master branch of
>     git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip
> 
> Keeping the fingers crossed that this is now really done.
> John already gave his thumbs up for the patch series, so the next
> logical question is who is going to sit on the patches until the merge
> window is opened? I can certainly create a git branch on git390 for
> them, but the patches depend on some patches in the x86-tip tree.
> Thomas, any thoughts?

Just as a heads up to Thomas and Martin, I finally got a box with a few
clocksources (TSC, HPET, ACPI PM) and tested switching between them with
this patch set.

Unfortunately when I try to switch away form the TSC, the write to the
sysfs entry seems to hang. The dmesg shows a switch message to hpet, but
then immdiately follows with ah switch message to tsc. The box still
runs, but the shells I ran the echo command from are hung.

I'm bisecting down the issue and will get you some more info as soon as
I can.

thanks
-john