Re: [PATCH] PR15789/gettimeofday: do not bail out if cpufreq_register_notifier fails

[David Smith <dsmith at redhat dot com>]

On 07/30/2013 03:10 AM, Timo Juhani Lindfors wrote:
> Linux commit d5aaffa9dd531c978c6f3fea06a2972653bd7fc8 (cpufreq: handle
> cpufreq being disabled for all exported function.) causes
> cpufreq_register_notifier to fail if cpufreq gets disabled at
> runtime. This patch simply ignores the failure of
> cpufreq_register_notifier so that gettimeofday can continue to work.
> Once cpufreq is disabled it can not be enabled at runtime so this
> should be safe.
> ---
>  runtime/time.c |    6 ++----
>  1 files changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/runtime/time.c b/runtime/time.c
> index 34df72a..a01d922 100644
> --- a/runtime/time.c
> +++ b/runtime/time.c
> @@ -253,10 +253,8 @@ _stp_init_time(void)
>  
>  #ifdef CONFIG_CPU_FREQ
>      if (!ret && !__stp_constant_freq()) {
> -        ret = cpufreq_register_notifier(&__stp_time_notifier,
> -                CPUFREQ_TRANSITION_NOTIFIER);
> -
> -        if (!ret) {
> +        if (!cpufreq_register_notifier(&__stp_time_notifier,
> +                                       CPUFREQ_TRANSITION_NOTIFIER)) {
>              for_each_online_cpu(cpu) {
>                  unsigned long flags;
>                  int freq_khz = cpufreq_get(cpu);
> 

I've looked at this patch, and while the patch looks simple enough, I
can't really test it (since I don't run xen anywhere). It seemed to work
on my fedora kernel (3.11.0-0.rc3.git1.1.fc20.x86_64).

I went ahead and committed it (with a minor tweak, to only unregister
the notifier it the registration worked).

If the patch doesn't fix your issue, let us know.

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)