Re: [Xen-devel] [0/5] [NET]: Add TSO support - Herbert Xu - com.xensource.lists.xen-devel

53 messages in com.xensource.lists.xen-develRe: [Xen-devel] [0/5] [NET]: Add TSO ...

From	Sent On	Attachments
Herbert Xu	27 Jun 2006 05:02
Herbert Xu	27 Jun 2006 05:04
Herbert Xu	27 Jun 2006 05:07
Herbert Xu	27 Jun 2006 05:07
Herbert Xu	27 Jun 2006 05:08
Herbert Xu	27 Jun 2006 05:09
Keir Fraser	27 Jun 2006 06:26
Herbert Xu	27 Jun 2006 07:02
Herbert Xu	27 Jun 2006 20:57
Herbert Xu	27 Jun 2006 20:58
Herbert Xu	27 Jun 2006 20:59
Herbert Xu	27 Jun 2006 20:59
Herbert Xu	27 Jun 2006 21:00
Herbert Xu	27 Jun 2006 21:00
Keir Fraser	28 Jun 2006 05:29
Herbert Xu	28 Jun 2006 07:02
Keir Fraser	28 Jun 2006 07:24
Herbert Xu	28 Jun 2006 17:19
Herbert Xu	28 Jun 2006 17:21
Herbert Xu	28 Jun 2006 17:21
Herbert Xu	28 Jun 2006 21:44
Keir Fraser	29 Jun 2006 02:01
Herbert Xu	29 Jun 2006 02:40
Keir Fraser	29 Jun 2006 03:32
Herbert Xu	29 Jun 2006 05:45
Keir Fraser	29 Jun 2006 06:19
Herbert Xu	29 Jun 2006 06:25
Keir Fraser	29 Jun 2006 06:41
Herbert Xu	29 Jun 2006 16:08
Herbert Xu	29 Jun 2006 19:40
Herbert Xu	29 Jun 2006 19:41
Herbert Xu	29 Jun 2006 19:42
Herbert Xu	29 Jun 2006 19:42
Herbert Xu	30 Jun 2006 05:46
Herbert Xu	30 Jun 2006 05:47
Herbert Xu	30 Jun 2006 05:47
Herbert Xu	30 Jun 2006 05:48
Keir Fraser	30 Jun 2006 06:06
Herbert Xu	30 Jun 2006 06:21
Keir Fraser	30 Jun 2006 07:36
Herbert Xu	30 Jun 2006 20:25
Herbert Xu	30 Jun 2006 20:32
Keir Fraser	01 Jul 2006 01:16
Keir Fraser	01 Jul 2006 01:23
Herbert Xu	01 Jul 2006 02:59
Keir Fraser	01 Jul 2006 05:17
Herbert Xu	01 Jul 2006 05:38
Herbert Xu	02 Jul 2006 21:44
Herbert Xu	02 Jul 2006 21:45
Herbert Xu	02 Jul 2006 21:45
Herbert Xu	02 Jul 2006 21:46
Keir Fraser	03 Jul 2006 01:12
Herbert Xu	03 Jul 2006 01:13

Subject:	Re: [Xen-devel] [0/5] [NET]: Add TSO support
From:	Herbert Xu (herb...@gondor.apana.org.au)
Date:	06/29/2006 05:45:51 AM
List:	com.xensource.lists.xen-devel

On Thu, Jun 29, 2006 at 11:32:43AM +0100, Keir Fraser wrote:

What if gso_segs, or any other gso parameter (e.g., type), is set incorrectly? The backend cannot trust frontend clients, so I'm rather worried that we could make the backend network stack crash!

gso_type is a key so the OS (Linux) must deal with all possible values. Any packet with a protocol or feature bit that it does not recognise will be dropped.

However, you have a very good point regarding gso_segs. I'll change it so that it is always recalculated for SKB_GSO_DODGY packets.

Also we should probably only define a XEN_GSO_TCPV4 flag for now, since we support only plain TSO right now. No skbuffs should get passed to netfront that have other GSO flags set, right? Seince we only advertise NETIF_F_TSO.

While it shouldn't have any adverse effects for the reason above, I will add a mask so that only bits in the mask are allowed through to gso_type.

Cheers,