Re: TCP Checksum offload in 2.4.19

From: Bernd Eckenfels (ecki@lina.inka.de)
Date: Fri Sep 06 2002 - 18:35:27 EST


On Fri, Sep 06, 2002 at 04:16:36PM -0700, David S. Miller wrote:
> You don't understand. Because it is too complicated to zerocopy
> arbitrary anonymous user pages, we have to copy it into a stable
> kernel buffer. Because we have to copy, we checksum at the
> same time.

well, which does not make the claim false. It would make sence to have
hardware checksum with zerocopy sendmsg. It is just not easy with the
current api to have zero copy user/kernel calls (besides the mmap stuff,
which i am not sure if it is zero copy or bothering IO, eighter).

> If you want zerocopy for user created data, create a file for
> the process, mmap() it, write your contents, then sendfile()
> the thing to the socket. End of story.

i dont like it, but on the other hand I see the need for transparent TCP
streams vanish quickly anyway. I for one do not use them anywhere. Even my
HTTP streams are mostly chunked.

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



This archive was generated by hypermail 2b29 : Sat Sep 07 2002 - 22:00:02 EST