On Wed, 2013-12-18 at 11:48 +0000, Morten Østergaard wrote:
> On Wed, 2013-12-18 at 14:14 +0400, Eugene Shatokhin wrote:
> > Could you also add a call to dump_stack() in your patch so as to get the 
> > call trace in the system log when that strange event happens, not only 
> > the warning?
> I added dump_stack():
> ------------------------------------------------------------
> diff -Naur linux-3.10.10.orig/net/core/skbuff.c
> linux-3.10.10/net/core/skbuff.c
> --- linux-3.10.10.orig/net/core/skbuff.c        2013-08-29
> 18:47:51.000000000 +0200
> +++ linux-3.10.10/net/core/skbuff.c     2013-12-18 12:16:37.946854138
> +0100
> @@ -1282,8 +1282,11 @@
>         SKB_LINEAR_ASSERT(skb);
>         skb->tail += len;
>         skb->len  += len;
> -       if (unlikely(skb->tail > skb->end))
> -               skb_over_panic(skb, len, __builtin_return_address(0));
> +       if (unlikely(skb->tail > skb->end)) {
> +         net_warn_ratelimited("Warning: skb_put l:%i t:%#lx e:%#lx",
> +          len, (unsigned long)skb->tail,(unsigned long)skb->end);
> +          dump_stack();
> +        }
>         return tmp;
>  }
>  EXPORT_SYMBOL(skb_put);
> ------------------------------------------------------------
> 
> And here is the dump in my log: (sorry for line wrapping)
> ...

> skbuff: Warning: skb_put l:9250 t:0xdd425c62 e:0xdd423f40


could you be receiving jumbo frames on this network? that 9250 byte
packet looks suspiciously like a reasonable MTU size for a possible
large (jumbo) packet.

As for the statements about this being a race, does it always happen
*during* the up?

If you have a reproducible case, doing something like instrumenting with
logging using printk at the beginning and end of e1000e_open, and at the
beginning/end of e1000e_close would give us a bit more info.

The other thing to add might be a printk of the packet data (skb->data)
using print_hex_data function when the issue occurs (inside your "if"
statement)


------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business. Most IT 
organizations don't have a clear picture of how application performance 
affects their revenue. With AppDynamics, you get 100% visibility into your 
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit 
http://communities.intel.com/community/wired

Reply via email to