On 03/23/2018 07:15 AM, Jesper Dangaard Brouer wrote:
> On Fri, 23 Mar 2018 06:29:55 -0700
> Eric Dumazet <eric.duma...@gmail.com> wrote:
>
>> On 03/23/2018 05:18 AM, Jesper Dangaard Brouer wrote:
>>
>>> +
>>> +void page_pool_destroy_rcu(struct page_pool *pool)
>>> +{
>>> + call_rcu(&pool->rcu, __page_pool_destroy_rcu);
>>> +}
>>> +EXPORT_SYMBOL(page_pool_destroy_rcu);
>>>
>>
>>
>> Why do we need to respect one rcu grace period before destroying a page pool
>> ?
>
> Due to previous allocator ID patch, which can have a pointer reference
> to a page_pool, and the allocator ID lookup uses RCU.
>
I am not convinced. How comes a patch that is _before_ this one can have any
impact ?
Normally, we put first infrastructure, then something using it.
rcu grace period before freeing huge quantitites of pages is problematic and
could
be used by syzbot to OOM the host.
