lkml.org 
[lkml]   [2006]   [Oct]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [rfc] [patch] mm: Slab - Eliminate lock_cpu_hotplug from slab
On Fri, Oct 27, 2006 at 06:19:19PM -0700, Ravikiran G Thirumalai wrote:
> This patch also takes the cache_chain_sem at kmem_cache_shrink to
> protect sanity of cpu_online_map at __cache_shrink, as viewed by slab.
> (kmem_cache_shrink->__cache_shrink->drain_cpu_caches). But, really,

drain_cpu_caches uses on_each_cpu() ..which does a preempt_disable()
before using the cpu_online_map. That should be a safe enough access to the
bitmap?

> kmem_cache_shrink is used at just one place in the acpi subsystem!
> Do we really need to keep kmem_cache_shrink at all?
>
> Another note. Looks like a cpu hotplug event can send CPU_UP_CANCELED to
> a registered subsystem even if the subsystem did not receive CPU_UP_PREPARE.
> This could be due to a subsystem registered for notification earlier than
> the current subsystem crapping out with NOTIFY_BAD. Badness can occur with
> in the CPU_UP_CANCELED code path at slab if this happens (The same would
> apply for workqueue.c as well). To overcome this, we might have to use either
> a) a per subsystem flag and avoid handling of CPU_UP_CANCELED, or
> b) Use a special notifier events like LOCK_ACQUIRE/RELEASE as Gautham was
> using in his experiments, or
> c) Do not send CPU_UP_CANCELED to a subsystem which did not receive
> CPU_UP_PREPARE.
>
> I would prefer c).

I think we need both b) and c).

Let me explain.

The need for c) is straightforward.

The need for b) comes from the fact that _cpu_down messes with the
tsk->cpus_allowed mask (to possibly jump off the dying CPU). This would cause
sched_getaffinity() to potentially return a false value back to the user and
hence it was modified to take lock_cpu_hotplug() before reading
tsk->cpus_allowed.

If we are discarding this whole lock_cpu_hotplug(), then IMO, we should
use LOCK_ACQUIRE/RELEASE, where ACQUIRE notification is sent *before*
messing with tsk->cpus_allowed and RELEASE notification sent *after*
restoring tsk->cpus_allowed (something like below):

@@ -186,13 +186,14 @@ int cpu_down(unsigned int cpu)
{
int err = 0;

- mutex_lock(&cpu_add_remove_lock);
+ blocking_notifier_call_chain(&cpu_chain, CPU_LOCK_ACQUIRE,
+ (void *)(long)cpu);
if (cpu_hotplug_disabled)
err = -EBUSY;
else
err = _cpu_down(cpu);
-
- mutex_unlock(&cpu_add_remove_lock);
+ blocking_notifier_call_chain(&cpu_chain, CPU_LOCK_RELEASE,
+ (void *)(long)cpu);
return err;
}


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

\
 
 \ /
  Last update: 2006-10-30 11:19    [W:0.056 / U:3.344 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site