Re: [PATCH] remove rwsem lock from CPUFREQ_GOV_STOP call (second call site)

[Dave Young]

On Mon, Jun 8, 2009 at 11:23 PM, Mathieu
Desnoyers<mathieu.desnoyers@xxxxxxxxxx> wrote:
> * Dave Jones (davej@xxxxxxxxxx) wrote:
>> On Mon, Jun 08, 2009 at 08:48:45AM -0400, Mathieu Desnoyers wrote:
>>
>> Â> > > >> Bug-Entry    : http://bugzilla.kernel.org/show_bug.cgi?id=13475
>> Â> > > >> Subject     : suspend/hibernate lockdep warning
>> Â> > > >> References   Â: http://marc.info/?l=linux-kernel&m=124393723321241&w=4
>> Â> > >
>> Â> > > I suspect the following commit, after revert this patch I test 5 times
>> Â> > > without lockdep warnings.
>> Â> > >
>> Â> > > commit b14893a62c73af0eca414cfed505b8c09efc613c
>> Â> > > Author: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxx>
>> Â> > > Date: Â Sun May 17 10:30:45 2009 -0400
>> Â> > >
>> Â> > > Â Â Â Â[CPUFREQ] fix timer teardown in ondemand governor
>> Â> >
>> Â> > The patch is probably not at fault here. I suspect it's some latent bug
>> Â> > that simply got exposed by the change to cancel_delayed_work_sync(). In
>> Â> > any case, Mathieu, can you take a look at this please?
>> Â>
>> Â> Yes, it's been looked at and discussed on the cpufreq ML. The short
>> Â> answer is that they plan to re-engineer cpufreq and remove the policy
>> Â> rwlock taken around almost every operations at the cpufreq level.
>> Â>
>> Â> The short-term solution, which is recognised as ugly, would be do to the
>> Â> following before doing the cancel_delayed_work_sync() :
>> Â>
>> Â> unlock policy rwlock write lock
>> Â>
>> Â> lock policy rwlock write lock
>> Â>
>> Â> It basically works because this rwlock is unneeded for teardown, hence
>> Â> the future re-work planned.
>> Â>
>> Â> I'm sorry I cannot prepare a patch current... I've got quite a few pages
>> Â> of Ph.D. thesis due for the beginning of July.
>>
>> I'm kinda scared to touch this code at all for .30 due to the number of
>> unexpected gotchas we seem to run into every time we touch something
>> locking related. ÂSo I'm inclined to just live with the lockdep warning
>> for .30, and see how the real fixes look for .31, and push them back
>> as -stable updates if they work out.
>>
>>
>> Venki, what are your thoughts?
>>
>
> Hi Dave,
>
> I've looked through the cpufreq code, and the following patch should
> address the call site I've missed in commit
> 42a06f2166f2f6f7bf04f32b4e823eacdceafdc9. I've followed all
> __cpufreq_set_policy call sites within cpufreq.c to make sure they all
> hold the rwsem write lock. An extra round of review would be good
> though.
>
> Can someone try the following patch and see if it fixes the regression ?

Bad news, I have tried the patch and It does not fix the regression.

> My test machine is currently busy running long formal verifications, and
> therefore unavailable for kernel patch testing. It compiles fine on a
> 2.6.30-rc5 kernel with my (now mainlined) cpufreq patches applied.
>
> Mathieu
>
>
> remove rwsem lock from CPUFREQ_GOV_STOP call (second call site)
>
> commit Â42a06f2166f2f6f7bf04f32b4e823eacdceafdc9
>
> Missed a call site for CPUFREQ_GOV_STOP to remove the rwlock taken around the
> teardown. To make a long story short, the rwlock write-lock causes a circular
> dependency with cancel_delayed_work_sync(), because the timer handler takes the
> read lock.
>
> Note that all callers to __cpufreq_set_policy are taking the rwsem. All sysfs
> callers (writers) hold the write rwsem at the earliest sysfs calling stage.
>
> However, the rwlock write-lock is not needed upon governor stop.
>
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxx>
> CC: rjw@xxxxxxx
> CC: mingo@xxxxxxx
> CC: Shaohua Li <shaohua.li@xxxxxxxxx>
> CC: Pekka Enberg <penberg@xxxxxxxxxxxxxx>
> CC: Dave Young <hidave.darkstar@xxxxxxxxx>
> CC: "Rafael J. Wysocki" <rjw@xxxxxxx>
> CC: Rusty Russell <rusty@xxxxxxxxxxxxxxx>
> CC: trenn@xxxxxxx
> CC: sven.wegener@xxxxxxxxxxx
> CC: Venkatesh Pallipadi <venkatesh.pallipadi@xxxxxxxxx>
> CC: cpufreq@xxxxxxxxxxxxxxx
> ---
> Âdrivers/cpufreq/cpufreq.c | Â 11 ++++++++++-
> Â1 file changed, 10 insertions(+), 1 deletion(-)
>
> Index: linux-2.6-lttng/drivers/cpufreq/cpufreq.c
> ===================================================================
> --- linux-2.6-lttng.orig/drivers/cpufreq/cpufreq.c   Â2009-06-08 10:20:48.000000000 -0400
> +++ linux-2.6-lttng/drivers/cpufreq/cpufreq.c  2009-06-08 10:48:52.000000000 -0400
> @@ -1697,8 +1697,17 @@ static int __cpufreq_set_policy(struct c
> Â Â Â Â Â Â Â Â Â Â Â Âdprintk("governor switch\n");
>
> Â Â Â Â Â Â Â Â Â Â Â Â/* end old governor */
> - Â Â Â Â Â Â Â Â Â Â Â if (data->governor)
> + Â Â Â Â Â Â Â Â Â Â Â if (data->governor) {
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â /*
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â* Need to release the rwsem around governor
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â* stop due to lock dependency between
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â* cancel_delayed_work_sync and the read lock
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â* taken in the delayed work handler.
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â*/
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â unlock_policy_rwsem_write(data->cpu);
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â__cpufreq_governor(data, CPUFREQ_GOV_STOP);
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â lock_policy_rwsem_write(data->cpu);
> + Â Â Â Â Â Â Â Â Â Â Â }
>
> Â Â Â Â Â Â Â Â Â Â Â Â/* start new governor */
> Â Â Â Â Â Â Â Â Â Â Â Âdata->governor = policy->governor;
>
>
> --
> Mathieu Desnoyers
> OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F ÂBA06 3F25 A8FE 3BAE 9A68
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at Âhttp://vger.kernel.org/majordomo-info.html
> Please read the FAQ at Âhttp://www.tux.org/lkml/
>



-- 
Regards
dave
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/