Re: 3.7.3+: Bad paging request in ip_rcv_finish while running NFS traffic.

[Ben Greear <greearb@candelatech.com>]

On 01/22/2013 10:11 PM, Eric Dumazet wrote:
> On Tue, 2013-01-22 at 18:32 -0800, Ben Greear wrote:
>
>> diff --git a/net/core/dst.c b/net/core/dst.c
>> index ee6153e..234b168 100644
>> --- a/net/core/dst.c
>> +++ b/net/core/dst.c
>> @@ -245,6 +245,7 @@ again:
>>                   dst->ops->destroy(dst);
>>           if (dst->dev)
>>                   dev_put(dst->dev);
>> +       dst->input = dst->output = 0xdeadbeef;
>>           kmem_cache_free(dst->ops->kmem_cachep, dst);
>
> Great !
>
> You could comment the kmem_cache_free() as well to get better chances to
> hit the bug, and maybe start a bisection to find the faulty commit ?

I suspect the bug goes back at least as far as 3.3.  And since
I need the NFS patches for this test case, bisecting will be pure hell.

I'll work on some more code instrumentation tomorrow.

One thing that came to mind while I was looking at the code today:

How are the non-ref-counted dst objects used safely?  Any chance
that tearing down the IP protocol on a device (or deleting a device)
could delete a dst that is referenced by an skb (and thus crashes as
I see)?

Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com