Re: 2.6.34-rc1: rcu lockdep bug?

["Américo Wang" <xiyou.wangcong@gmail.com>]

(Cc'ing netdev)

On Fri, Mar 12, 2010 at 12:17 AM, Américo Wang <xiyou.wangcong@gmail.com> wrote:
> On Thu, Mar 11, 2010 at 05:45:56AM -0800, Paul E. McKenney wrote:
>>On Thu, Mar 11, 2010 at 06:05:38PM +0800, Américo Wang wrote:
>>> Hello, Paul and Peter,
>>>
>>> Attached is the lockdep warning that I triggered today.
>>>
>>> I am not sure if this is a bug of rcu lockdep, because I am
>>> testing my patch when this occurred. However, in the backtrace,
>>> there is none of the functions that I touched, weird.
>>>
>>> So, please help to check if this is a bug of rcu lockdep.
>>
>>This sort of thing is caused by acquiring the same lock with softirq
>>(AKA BH) blocked and not, which can result in self-deadlock.
>>
>>There was such a bug in the RCU lockdep stuff in -tip, but it has long
>>since been fixed.  If you were seeing that bug, rcu_do_batch() would
>>be on the stack, which it does not appear to be.
>>
>>So does your patch involve the usbfs_mutex?  Or attempt to manipulate
>>vfs/fs state from withing networking softirq/BH context?
>>
>
> Nope, it is a patch for netpoll, nothing related with usb, nor vfs.
>

Ok, after decoding the lockdep output, it looks like that
netif_receive_skb() should call rcu_read_lock_bh() instead of rcu_read_lock()?
But I don't know if all callers of netif_receive_skb() are in softirq context.

Paul, what do you think?

Thank you.