Re: Badness at net/core/dev.c:2199

[Stephen Hemminger <shemminger@linux-foundation.org>]

On Sun, 16 Dec 2007 21:20:57 +0300
Evgeniy Polyakov <johnpol@2ka.mipt.ru> wrote:

> On Sun, Dec 16, 2007 at 07:55:55PM +0200, Meelis Roos (mroos@linux.ee) wrote:
> > Just got this trace from current 2.6.24-rc5+git running on 32-bit ppc 
> > (PReP subarch, tulip NIC's) during apt-get update (logged in via ssh so 
> > also ssh traffic):
> > 
> > ------------[ cut here ]------------
> > Badness at net/core/dev.c:2199
> 
> Please test attached patch.
> If I understood ltulip correctly, it is posible, that number of entries
> can be higher than requested budget. When work_done is equal to
> budget-1, the last skb has to be processed, after 154'th line
> work_done will become equal to budget and thus loop has to break,
> check on the same 154 line will become false, but work_done will be
> increased nevertheless, which will make work_done being equal to
> budget+1 at exit, which will fire warning you saw.
> 
> Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
> 
> diff --git a/drivers/net/tulip/interrupt.c b/drivers/net/tulip/interrupt.c
> index 3653314..9e0e97a 100644
> --- a/drivers/net/tulip/interrupt.c
> +++ b/drivers/net/tulip/interrupt.c
> @@ -151,8 +151,9 @@ int tulip_poll(struct napi_struct *napi, int budget)
>                         if (tulip_debug > 5)
>                                 printk(KERN_DEBUG "%s: In tulip_rx(), entry %d %8.8x.\n",
>                                        dev->name, entry, status);
> -		       if (work_done++ >= budget)
> +		       if (work_done >= budget)
>                                 goto not_done;
> +		       work_done++;
>  
>                         if ((status & 0x38008300) != 0x0300) {
>                                 if ((status & 0x38000300) != 0x0300) {
> 
> 

I already sendout a correct patch last week. It should pre-increment.

-- 
Stephen Hemminger <shemminger@linux-foundation.org>