LinuxLists.cc - [ppc] Disparity between sys_clock

2008-01-27 14:02:53

Subject: [ppc] Disparity between sys_clock_getres and vdso implementation

Hi Paul,

On PPC, I see a disparity between clock_getres implementations in the
vdso and syscall. I am using a IBM Openpower hardware and 2.6.24 kernel
with CONFIG_HIGH_RES_TIMERS=y.

clock_getres call for CLOCK_REALTIME returns 1 millisecond. However,
when I edit arch/powerpc/kernel/vdso*/gettimeofday.S to force it to use
sys_clock_getres, I get 1 nanosecond resolution. The code in vdso seems
to be returning some pre-defined (incorrect) variables.

Could you please let me know the reason for this? Is it something that
should be fixed in vdso?

Thanks,
Sripathi.

2008-02-04 06:48:20

by Tony Breeds

[permalink] [raw]

Subject: Re: [ppc] Disparity between sys_clock_getres and vdso implementation

2008-02-05 05:18:56

by Tony Breeds

[permalink] [raw]

Subject: [PATCH] [POWERPC] Use a sensible default for clock_getres() in the vdso.

On Sun, Jan 27, 2008 at 07:32:59PM +0530, Sripathi Kodi wrote:
> Hi Paul,
>
> On PPC, I see a disparity between clock_getres implementations in the
> vdso and syscall. I am using a IBM Openpower hardware and 2.6.24 kernel
> with CONFIG_HIGH_RES_TIMERS=y.
>
> clock_getres call for CLOCK_REALTIME returns 1 millisecond. However,
> when I edit arch/powerpc/kernel/vdso*/gettimeofday.S to force it to use
> sys_clock_getres, I get 1 nanosecond resolution. The code in vdso seems
> to be returning some pre-defined (incorrect) variables.
>
> Could you please let me know the reason for this? Is it something that
> should be fixed in vdso?

Can you try the attached patch and see it if works for you?

From: Tony Breeds <[email protected]>
Subject: [PATCH] [POWERPC] Use a sensible default for clock_getres() in the vdso.

This ensures that the syscall and the (fast) vdso versions of clock_getres()
will return the same resolution.

Signed-off-by: Tony Breeds <[email protected]>
---
arch/powerpc/kernel/asm-offsets.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index ed083fe..e6e4928 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -22,6 +22,7 @@
#include <linux/mman.h>
#include <linux/mm.h>
#include <linux/suspend.h>
+#include <linux/hrtimer.h>
#ifdef CONFIG_PPC64
#include <linux/time.h>
#include <linux/hardirq.h>
@@ -312,7 +313,7 @@ int main(void)
DEFINE(CLOCK_REALTIME, CLOCK_REALTIME);
DEFINE(CLOCK_MONOTONIC, CLOCK_MONOTONIC);
DEFINE(NSEC_PER_SEC, NSEC_PER_SEC);
- DEFINE(CLOCK_REALTIME_RES, TICK_NSEC);
+ DEFINE(CLOCK_REALTIME_RES, (KTIME_MONOTONIC_RES).tv64);

#ifdef CONFIG_BUG
DEFINE(BUG_ENTRY_SIZE, sizeof(struct bug_entry));
--
1.5.4

Yours Tony

linux.conf.au http://linux.conf.au/ || http://lca2008.linux.org.au/
Jan 28 - Feb 02 2008 The Australian Linux Technical Conference!

2008-02-05 09:08:04

by Chirag Jog

[permalink] [raw]

Subject: Re: [PATCH] [POWERPC] Use a sensible default for clock_getres() in the vdso.

* Tony Breeds <[email protected]> [2008-02-05 16:16:48]:

> On Sun, Jan 27, 2008 at 07:32:59PM +0530, Sripathi Kodi wrote:
> > Hi Paul,
> >
> > On PPC, I see a disparity between clock_getres implementations in the
> > vdso and syscall. I am using a IBM Openpower hardware and 2.6.24 kernel
> > with CONFIG_HIGH_RES_TIMERS=y.
> >
> > clock_getres call for CLOCK_REALTIME returns 1 millisecond. However,
> > when I edit arch/powerpc/kernel/vdso*/gettimeofday.S to force it to use
> > sys_clock_getres, I get 1 nanosecond resolution. The code in vdso seems
> > to be returning some pre-defined (incorrect) variables.
> >
> > Could you please let me know the reason for this? Is it something that
> > should be fixed in vdso?
>
> Can you try the attached patch and see it if works for you?
I tried this patch.
It seems to solve the reported problem and give a 1ns resolution.
Thanks.

--
Cheers,
Chirag Jog