Re: [PATCH] netpoll: use non-BH variant of RCU

[David Miller <davem@davemloft.net>]

From: "John W. Linville" <linville@tuxdriver.com>
Date: Wed, 11 Aug 2010 11:37:36 -0400

> On Wed, Aug 11, 2010 at 07:27:06AM -0400, Herbert Xu wrote:
>> On Wed, Aug 11, 2010 at 07:03:30AM -0400, Herbert Xu wrote:
>> >
>> > To use rcu_read_lock safely we'd also need to add do synchronize_rcu
>> > in addition of synchronize_rcu_bh, right Paul?
>> > 
>> > Of course as we were doing this unsafely prior to my patch anyway
>> > I'm also fine with just reverting it.
>> 
>> Actually, we could just disable IRQs for stable.  How about this
>> patch (untested)?
>> 
>> netpoll: Disable IRQ around RCU dereference in netpoll_rx
>> 
>> We cannot use rcu_dereference_bh safely in netpoll_rx as we may
>> be called with IRQs disabled.  We could however simply disable
>> IRQs as that too causes BH to be disabled and is safe in either
>> case.
>> 
>> Thanks to John Linville for discovering this bug and providing
>> a patch.
>> 
>> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
> 
> This also seems to work for me.

I'll queue up this variant for -stable.

And we'll go with making use of Paul's new RCU irqsoff interface
approach for Linus's tree.

Thanks.