https://bugzilla.novell.com/show_bug.cgi?id=668194
https://bugzilla.novell.com/show_bug.cgi?id=668194#c11
Olaf Kirch changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEEDINFO |NEW
CC| |okir@novell.com
InfoProvider|okir@novell.com |
--- Comment #11 from Olaf Kirch 2011-02-17 09:34:50 UTC ---
I think turning off checksums is not the right approach. Offloading
checksum generation to hardware is an important performance improvement,
even more so on a virtualization host, where CPU cycles are often the
most contested resource.
The bug here is that the packet with the partial checksum is delivered to
the application. The most obvious fix would be to have the AF_PACKET recvmsg
function in the kernel could fill in the checksum if it sees it's missing.
It seems this approach was considered and rejected by Herbert Xu and Dave
Miller,
and instead they chose the alternative approach in
http://git.kernel.org/linus/8dc4194
I don't remember off the top of my head whether dhcpd uses libpcap to
capture packets; if that is the case, we should enhance the linux pcap
routines to receive the PACKET_AUXDATA information, and check whether the
TP_STATUS_CSUMNOTREADY flags is set. If that's the case, fake the correct
checksum.
If dhcpd uses its own packet capture functions, we should patch these.
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.