[10/17] netvm: Allow skb allocation to use PFMEMALLOC reserves

On Thu, Jun 21, 2012 at 06:30:29PM +0200, Sebastian Andrzej Siewior wrote:
> > diff --git a/net/core/skbuff.c b/net/core/skbuff.c
> > index 1d6ecc8..9a58dcc 100644
> > --- a/net/core/skbuff.c
> > +++ b/net/core/skbuff.c
> > @@ -167,14 +206,19 @@ static void skb_under_panic(struct sk_buff *skb, int sz, void *here)
> >   *	%GFP_ATOMIC.
> >   */
> >  struct sk_buff *__alloc_skb(unsigned int size, gfp_t gfp_mask,
> > -			    int fclone, int node)
> > +			    int flags, int node)
> >  {
> >  	struct kmem_cache *cache;
> >  	struct skb_shared_info *shinfo;
> >  	struct sk_buff *skb;
> >  	u8 *data;
> > +	bool pfmemalloc;
> >  
> > -	cache = fclone ? skbuff_fclone_cache : skbuff_head_cache;
> > +	cache = (flags & SKB_ALLOC_FCLONE)
> > +		? skbuff_fclone_cache : skbuff_head_cache;
> > +
> > +	if (sk_memalloc_socks() && (flags & SKB_ALLOC_RX))
> > +		gfp_mask |= __GFP_MEMALLOC;
> >  
> >  	/* Get the HEAD */
> >  	skb = kmem_cache_alloc_node(cache, gfp_mask & ~__GFP_DMA, node);
> 
> This is mostly used by nic to refil their RX skb pool. You add the
> __GFP_MEMALLOC to the allocation to rise the change of a successfull refill
> for the swap case.
> A few drivers use build_skb() to create the skb. __netdev_alloc_skb()
> shouldn't be affected since the allocation happens with GFP_ATOMIC. Looking at
> TG3 it uses build_skb() and get_pages() / kmalloc(). Shouldn't this be some
> considered?
> 

While TG3 is not exactly as you describe after rebasing build_skb should
make a similar check to __alloc_skb. As it is always used for RX allocation
from the skbuff_head_cache cache the following should be suitable. Thanks.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Fri, Jun 22, 2012 at 11:54:51AM +0100, Mel Gorman wrote:
> > This is mostly used by nic to refil their RX skb pool. You add the
> > __GFP_MEMALLOC to the allocation to rise the change of a successfull refill
> > for the swap case.
> > A few drivers use build_skb() to create the skb. __netdev_alloc_skb()
> > shouldn't be affected since the allocation happens with GFP_ATOMIC. Looking at
> > TG3 it uses build_skb() and get_pages() / kmalloc(). Shouldn't this be some
> > considered?
> > 
> 
> While TG3 is not exactly as you describe after rebasing build_skb should
> make a similar check to __alloc_skb. As it is always used for RX allocation
> from the skbuff_head_cache cache the following should be suitable. Thanks.

As Eric pointed out you end up in netdev_alloc_frag() which is using
alloc_page(). This is also used by __netdev_alloc_skb().

Sebastian
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html