netdev - Re: [PATCH] xen/netfront: improve truesize tracking

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 18 Dec 2012 08:35:48 -0800
From:	Eric Dumazet <erdnetdev@...il.com>
To:	Ian Campbell <Ian.Campbell@...rix.com>
Cc:	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	Sander Eikelenboom <linux@...elenboom.it>,
	Konrad Rzeszutek Wilk <konrad.wilk@...cle.com>,
	annie li <annie.li@...cle.com>,
	"xen-devel@...ts.xensource.com" <xen-devel@...ts.xensource.com>
Subject: Re: [PATCH] xen/netfront: improve truesize tracking

On Tue, 2012-12-18 at 16:22 +0000, Ian Campbell wrote:
> On Tue, 2012-12-18 at 16:13 +0000, Eric Dumazet wrote:
> > On Tue, 2012-12-18 at 15:26 +0000, Ian Campbell wrote:
> > 
> > > So actually we want += PAGE_SIZE * skb_shinfo(skb)->nr_frags ?
> > > 
> > 
> > I dont know what are the real frag sizes in your case.
> 
> I think it's a page, see xennet_alloc_rx_buffers and the alloc_page
> therein.
> 

If they are order-0 pages, then PAGE_SIZE * nr_frags  is OK.


> > Some drivers allocate a full page for an ethernet frame, others use half
> > of a page, it really depends.
> > 

> > As the frag ABI doesnt contain real size, its ok in this case to account
> > the actual frag size.
> > 
> > (skb->data_len in your driver)
> 
> I guess I'm a bit confused by what truesize means again then ;-),
> because in that case the original patch is correct although it would
> have been less confusing to do:
> 	skb->truesize += skb->data_len; 
> in xennet_poll() and then do the subtraction of
> NETFRONT_SKB_CB(skb)->pull_to in handle_incoming_queue() where we
> actually do the pull up.
> 
> Unless __pskb_pull_tail does that adjustment for us, but if it does I
> can't see where.

Thats because skb frags only contain :

a page pointer.
An offset
A size. (Exact number or used bytes in this frag)

And not the 'originally allocated size. It could be 256, 768, 2048,
4096, 65536 bytes, nobody but the driver really knows.

So when we pull X bytes from a fragment to skb->head, there is no way to
remember what was the original size of the fragment.

Only the driver allocating the frag knows its truesize.

Once skb is given to the stack, we lose this information, and rely on
skb->truesize being an accurate estimation.



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