Re: Regression in 2.6.27-rc1 for set_cpus_allowed_ptr

["Dmitry Adamushko" <dmitry.adamushko@gmail.com>]

2008/8/8 Rafael J. Wysocki <rjw@sisk.pl>:
> [Adding CCs]
>
> On Friday, 8 of August 2008, Langsdorf, Mark wrote:
>> One of my co-workers noticed that the powernow-k8
>> driver no longer restarts when a CPU core is
>> hot-disabled and then hot-enabled on AMD quad-core
>> systems.
>>
>> The following comands work fine on 2.6.26 and fail
>> on 2.6.27-rc1:
>>
>> echo 0 > /sys/devices/system/cpu/cpu3/online
>> echo 1 > /sys/devices/system/cpu/cpu3/online
>> find /sys -name cpufreq
>>
>> For 2.6.26, the find will return a cpufreq
>> directory for each processor.  In 2.6.27-rc1,
>> the cpu3 directory is missing.
>>
>> After digging through the code, the following
>> logic is failing when the core is hot-enabled
>> at runtime.  The code works during the boot
>> sequence.
>>
>>       cpumask_t = current->cpus_allowed;
>>       set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
>>       if (smp_processor_id() != cpu)
>>               return -ENODEV;
>>


if it gets called from any of the cpu-hotplug handlers, it won't work
now (x86-microcode is another victim).

Please give a try to the following patch: http://lkml.org/lkml/2008/7/30/171

does it help?

(the explanation is also available in this thread).

well, provided we may guarantee that load-balancing has been
initialized (it's ok in our case) by the moment CPU_ONLINE gets
called, this approach is not that bad perhaps... (and it looks like
there is plenty of code that relies on set_cpus_allowed_ptr() being
workable in cpu-hotplug-handlers).

Although, I personally don't like that much this particular use-case
of set_cpus_allowed_ptr() (I posted patches for x86-microcode). btw.,
last time I briefly looked at various places, there seemed to be a few
where

old_mask = p->cpus_allowed;
set_cpus_allowed_ptr(p, target_cpu);
// do something
set_cpus_allowed_ptr(p, old_mask);

is used just wrongly. e.g. it may race with sched_setaffinity() and
negate its effect.


>> -Mark Langdsorf
>> Operating System Research Center
>> AMD

-- 
Best regards,
Dmitry Adamushko