[PATCH] Race conditions in QUEUE handling on SMP builds

[Rob Thomas]

I don't mean to sound a bit pedantic here, but..

On Fri, 21 Jul 2000, Alexander Demenshin wrote:

> 	It fixes all problems in QUEUE handling reported by me
> 	(and don't introduce new, I hope - at least I didn't found
> 	found more problems).

This is the one-line patch jamesm sent to you, and the list? I don't
understand the need for the skb get and free either.  James, Rusty?
(BTW, last time I looked, james's patch had been put into CVS, so if you
had been watching the CVS tree, you would have saved a lot of time on your
behalf)

> @@ -521,6 +536,7 @@
>  	case NF_QUEUE:
>  		nf_queue(skb, elem, info->pf, info->hook, 
>  			 info->indev, info->outdev, info->okfn);
> +		break;
>  
>  	case NF_DROP:
>  		kfree_skb(skb);
> 

I woule never have spotted that myself, so I'm guessing that james put a
-lot- of time and effort into searching that one down.

I'm a bit grumpy because you immediately responded to james's patch with a
'No, that won't work', and then suddenly you arrive with this magic 'Ooh,
look, this fixes it!'

I think I'll shut up now, it's friday night. Why am I working?

--Rob