Re: [PATCH 3/3] slub: release a lock if freeing object with a lock isfailed in __slab_free()

[JoonSoo Kim]

2012/7/6 Christoph Lameter <cl@xxxxxxxxx>:
> On Fri, 6 Jul 2012, JoonSoo Kim wrote:
>
>> For example,
>> When we try to free object A at cpu 1, another process try to free
>> object B at cpu 2 at the same time.
>> object A, B is in same slab, and this slab is in full list.
>>
>> CPU 1                           CPU 2
>> prior = page->freelist;    prior = page->freelist
>> ....                                  ...
>> new.inuse--;                   new.inuse--;
>> taking lock                      try to take the lock, but failed, so
>> spinning...
>> free success                   spinning...
>> add_partial
>> release lock                    taking lock
>>                                        fail cmpxchg_double_slab
>>                                        retry
>>                                        currently, we don't need lock
>>
>> At CPU2, we don't need lock anymore, because this slab already in partial list.
>
> For that scenario we could also simply do a trylock there and redo
> the loop if we fail. But still what guarantees that another process will
> not modify the page struct between fetching the data and a successful
> trylock?


I'm not familiar with English, so take my ability to understand into
consideration.

we don't need guarantees that another process will not modify
the page struct between fetching the data and a successful trylock.

As I understand, do u ask below scenario?

CPU A               CPU B
lock
cmpxchg fail
retry
unlock
...                       modify page strcut
...
cmpxchg~~

In this case, cmpxchg will fail and just redo the loop.
If we need the lock again during redo, re-take the lock.
But I think this is not common case.
--
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/