On Fri, Jan 16, 2009 at 10:19:35AM +1100, Herbert Xu wrote:

quoted text
> On Fri, Jan 16, 2009 at 12:03:31AM +0100, Willy Tarreau wrote:

> >

> > I'm leaving the patch below for comments, maybe someone will spot

> > something ? Don't we need at least one kfree() somewhere to match

> > alloc_pages() ?

>

> Indeed.

>

> > >  static inline int spd_fill_page(struct splice_pipe_desc *spd, struct page *page,

> > >  				unsigned int len, unsigned int offset,

> > > -				struct sk_buff *skb)

> > > +				struct sk_buff *skb, int linear)

> > >  {

> > >  	if (unlikely(spd->nr_pages == PIPE_BUFFERS))

> > >  		return 1;

> > >

> > > +	if (linear) {

> > > +		page = linear_to_page(page, len, offset);

> > > +		if (!page)

> > > +			return 1;

> > > +	}

> > > +

> > >  	spd->pages[spd->nr_pages] = page;

> > >  	spd->partial[spd->nr_pages].len = len;

> > >  	spd->partial[spd->nr_pages].offset = offset;

> > > -	spd->partial[spd->nr_pages].private = (unsigned long) skb_get(skb);

> > >  	spd->nr_pages++;

> > > +	get_page(page);

>

> This get_page needs to be moved into an else clause of the previous

> if block.

Good catch Herbert, it's working fine now. The performance I get

(30.6 MB/s) is between the original splice version (24.1) and the

recently optimized one (35.7). That's not that bad considering

we're copying the data.
Cheers,

Willy
--

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