Jan Beulich
2013-Apr-17 08:08 UTC
Attempting to checksum a non-TCP/UDP packet, dropping a protocol 1 packet
Pasi, in http://readlist.com/lists/lists.xensource.com/xen-users/10/50495.html you mention a netback fix without identifying which one. In going over the changes as well as looking at the code, I can''t spot anything related, or see how non-TCP, non-UDP packets could pass checksum_setup(), and hence I don''t really understand how this problem can be considered fixed (for, in the case here, ICMP, the header of which does have a checksum field). Jan
Pasi Kärkkäinen
2013-Apr-17 16:26 UTC
Re: Attempting to checksum a non-TCP/UDP packet, dropping a protocol 1 packet
On Wed, Apr 17, 2013 at 09:08:24AM +0100, Jan Beulich wrote:> Pasi, >Hi,> in http://readlist.com/lists/lists.xensource.com/xen-users/10/50495.html > you mention a netback fix without identifying which one. In going > over the changes as well as looking at the code, I can''t spot > anything related, or see how non-TCP, non-UDP packets could > pass checksum_setup(), and hence I don''t really understand how > this problem can be considered fixed (for, in the case here, ICMP, > the header of which does have a checksum field). >That''s almost 3 years ago.. I wonder if it was this one: "xen/netback: correctly setup skb->ip_summed on receive": http://git.kernel.org/cgit/linux/kernel/git/jeremy/xen.git/commit/?h=xen/next-2.6.32&id=7ec9d1e805338fd18bbce73e8c72e4d09da8f2d5 That''s at least in 2.6.32.13 in Jeremy''s xen.git. -- Pasi
Jan Beulich
2013-Apr-18 08:03 UTC
Re: Attempting to checksum a non-TCP/UDP packet, dropping a protocol 1 packet
>>> On 17.04.13 at 18:26, Pasi Kärkkäinen<pasik@iki.fi> wrote: > On Wed, Apr 17, 2013 at 09:08:24AM +0100, Jan Beulich wrote: >> in http://readlist.com/lists/lists.xensource.com/xen-users/10/50495.html >> you mention a netback fix without identifying which one. In going >> over the changes as well as looking at the code, I can't spot >> anything related, or see how non-TCP, non-UDP packets could >> pass checksum_setup(), and hence I don't really understand how >> this problem can be considered fixed (for, in the case here, ICMP, >> the header of which does have a checksum field). >> > > That's almost 3 years ago.. I wonder if it was this one: > > "xen/netback: correctly setup skb->ip_summed on receive": > http://git.kernel.org/cgit/linux/kernel/git/jeremy/xen.git/commit/?h=xen/nex > t-2.6.32&id=7ec9d1e805338fd18bbce73e8c72e4d09da8f2d5 > > That's at least in 2.6.32.13 in Jeremy's xen.git.Yeah, that seemed the most likely one, but still not really related, since there's nothing here affecting non-TCP, non-UDP packets. I.e. it could only have an effect for such packets if there was a false attempt to checksum them when no checksumming was needed. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
Pasi Kärkkäinen
2013-Apr-18 16:19 UTC
Re: Attempting to checksum a non-TCP/UDP packet, dropping a protocol 1 packet
On Thu, Apr 18, 2013 at 09:03:58AM +0100, Jan Beulich wrote:> >>> On 17.04.13 at 18:26, Pasi Kärkkäinen<pasik@iki.fi> wrote: > > On Wed, Apr 17, 2013 at 09:08:24AM +0100, Jan Beulich wrote: > >> in http://readlist.com/lists/lists.xensource.com/xen-users/10/50495.html > >> you mention a netback fix without identifying which one. In going > >> over the changes as well as looking at the code, I can''t spot > >> anything related, or see how non-TCP, non-UDP packets could > >> pass checksum_setup(), and hence I don''t really understand how > >> this problem can be considered fixed (for, in the case here, ICMP, > >> the header of which does have a checksum field). > >> > > > > That''s almost 3 years ago.. I wonder if it was this one: > > > > "xen/netback: correctly setup skb->ip_summed on receive": > > http://git.kernel.org/cgit/linux/kernel/git/jeremy/xen.git/commit/?h=xen/nex > > t-2.6.32&id=7ec9d1e805338fd18bbce73e8c72e4d09da8f2d5 > > > > That''s at least in 2.6.32.13 in Jeremy''s xen.git. > > Yeah, that seemed the most likely one, but still not really related, > since there''s nothing here affecting non-TCP, non-UDP packets. > I.e. it could only have an effect for such packets if there was a > false attempt to checksum them when no checksumming was > needed. >Ok. It''s also possible it was already fixed in an earlier version, I have some vague memory of that.. so 2.6.32.12 or 2.6.32.11 in Jeremy''s tree. -- Pasi
Seemingly Similar Threads
- [PATCH net] xen-netback: fix fragment detection in checksum setup
- [PATCH] Fix checksum errors when firewalling in domU
- [PATCH next] xen: Use more current logging styles
- [PATCH next] xen: Use more current logging styles
- [PATCH next] xen: Use more current logging styles