xen-netfront sets partial checksum at wrong offset

xen-netfront sets partial checksum at wrong offset

[Venkat Venkatsubra]

Hello All,

When handle_incoming_queue (xen-netfront.c) calls checksum_setup()
the transport header didn't seem to be initialized yet.
skb_checksum_setup_ipv4 ends up initializing the checksum at incorrect offset. 
        if (recalculate)
                *csum = ~csum_tcpudp_magic(ip_hdr(skb)->saddr,
                                           ip_hdr(skb)->daddr,
                                           skb->len - off,
                                           ip_hdr(skb)->protocol, 0);

where csum is &tcp_hdr(skb)->check.

This problem got exposed when LRO was ON on the host.
It hits this case in checksum_setup().
        /*
         * A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
         * peers can fail to set NETRXF_csum_blank when sending a GSO
         * frame. In this case force the SKB to CHECKSUM_PARTIAL and
         * recalculate the partial checksum.
         */

skb->ip_summed was CHECKSUM_UNNECESSARY. 
The guest was at 3.8.13 kernel.

I can send a patch if you think the problem exists in current releases too.

Thanks.

Venkat

Re: [Xen-devel] xen-netfront sets partial checksum at wrong offset

[Konrad Rzeszutek Wilk]

On Thu, May 07, 2015 at 12:44:24PM -0700, Venkat Venkatsubra wrote:
> Hello All,
> 
> When handle_incoming_queue (xen-netfront.c) calls checksum_setup()
> the transport header didn't seem to be initialized yet.
> skb_checksum_setup_ipv4 ends up initializing the checksum at incorrect offset. 
>         if (recalculate)
>                 *csum = ~csum_tcpudp_magic(ip_hdr(skb)->saddr,
>                                            ip_hdr(skb)->daddr,
>                                            skb->len - off,
>                                            ip_hdr(skb)->protocol, 0);
> 
> where csum is &tcp_hdr(skb)->check.
> 
> This problem got exposed when LRO was ON on the host.
> It hits this case in checksum_setup().
>         /*
>          * A GSO SKB must be CHECKSUM_PARTIAL. However some buggy
>          * peers can fail to set NETRXF_csum_blank when sending a GSO
>          * frame. In this case force the SKB to CHECKSUM_PARTIAL and
>          * recalculate the partial checksum.
>          */
> 
> skb->ip_summed was CHECKSUM_UNNECESSARY. 
> The guest was at 3.8.13 kernel.
> 
> I can send a patch if you think the problem exists in current releases too.

Please CC the maintainers of the driver. You can get that from 'scripts/get_maintainer.pl'

I've done that for you.
> 
> Thanks.
> 
> Venkat
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

RE: [Xen-devel] xen-netfront sets partial checksum at wrong offset

[Venkat Venkatsubra]

> 
> Please CC the maintainers of the driver. You can get that from
> 'scripts/get_maintainer.pl'
> 
> I've done that for you.

Thanks, Konrad.

I am copying Wei too who had fixed the below problem earlier.
It fixed the incorrect ip_hdr(). tcp_hdr() still needs to fixed.

commit d554f73df6bc35ac8f6a65e5560bf1d31dfebed9
Author: Wei Liu <wei.liu2@citrix.com>
Date:   Wed Feb 19 18:48:34 2014 +0000

    xen-netfront: reset skb network header before checksum

    In ed1f50c3a ("net: add skb_checksum_setup") we introduced some checksum
    functions in core driver. Subsequent change b5cf66cd1 ("xen-netfront:
    use new skb_checksum_setup function") made use of those functions to
    replace its own implementation.

    However with that change netfront is broken. It sees a lot of checksum
    error. That's because its own implementation of checksum function was a
    bit hacky (dereferencing skb->data directly) while the new function was
    implemented using ip_hdr(). The network header is not reset before skb
    is passed to the new function. When the new function tries to do its
    job, it's confused and reports error.

    The fix is simple, we need to reset network header before passing skb to
    checksum function. Netback is not affected as it already does the right
    thing.

    Reported-by: Sander Eikelenboom <linux@eikelenboom.it>
    Signed-off-by: Wei Liu <wei.liu2@citrix.com>
    Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
    Cc: Paul Durrant <paul.durrant@citrix.com>
    Tested-By: Sander Eikelenboom <linux@eikelenboom.it>
    Signed-off-by: David S. Miller <davem@davemloft.net>

Thanks.

Venkat

Re: [Xen-devel] xen-netfront sets partial checksum at wrong offset

[Jan Beulich]

>>> On 11.05.15 at 19:25, <venkat.x.venkatsubra@oracle.com> wrote:
>>  
>> Please CC the maintainers of the driver. You can get that from
>> 'scripts/get_maintainer.pl'
>> 
>> I've done that for you.
> 
> Thanks, Konrad.
> 
> I am copying Wei too who had fixed the below problem earlier.
> It fixed the incorrect ip_hdr(). tcp_hdr() still needs to fixed.
> 
> commit d554f73df6bc35ac8f6a65e5560bf1d31dfebed9
> Author: Wei Liu <wei.liu2@citrix.com>
> Date:   Wed Feb 19 18:48:34 2014 +0000
> 
>     xen-netfront: reset skb network header before checksum

So no response at all so far from the maintainers made me look into
this: I first thought what we need would be calls to
skb_probe_transport_header() in skb_checksum_setup_ip() after
each of the skb_maybe_pull_tail() functions. But
skb_partial_csum_set() already calls skb_set_transport_header(),
so I now think things ought to be fine without any change. Can
you clarify what you think is missing? Or is this an issue in just the
old (3.8.x) kernel you're using?

(In either case netback's xenvif_tx_submit() calling
skb_probe_transport_header() would seem pointless, as
skb_checksum_setup() - with or without a fix - ought to be taking
care of this anyway.)

Jan

Re: [Xen-devel] xen-netfront sets partial checksum at wrong offset

[Wei Liu]

On Fri, May 29, 2015 at 11:34:07AM +0100, Jan Beulich wrote:
> >>> On 11.05.15 at 19:25, <venkat.x.venkatsubra@oracle.com> wrote:
> >>  
> >> Please CC the maintainers of the driver. You can get that from
> >> 'scripts/get_maintainer.pl'
> >> 
> >> I've done that for you.
> > 
> > Thanks, Konrad.
> > 
> > I am copying Wei too who had fixed the below problem earlier.
> > It fixed the incorrect ip_hdr(). tcp_hdr() still needs to fixed.
> > 
> > commit d554f73df6bc35ac8f6a65e5560bf1d31dfebed9
> > Author: Wei Liu <wei.liu2@citrix.com>
> > Date:   Wed Feb 19 18:48:34 2014 +0000
> > 
> >     xen-netfront: reset skb network header before checksum
> 
> So no response at all so far from the maintainers made me look into
> this: I first thought what we need would be calls to
> skb_probe_transport_header() in skb_checksum_setup_ip() after
> each of the skb_maybe_pull_tail() functions. But
> skb_partial_csum_set() already calls skb_set_transport_header(),
> so I now think things ought to be fine without any change. Can
> you clarify what you think is missing? Or is this an issue in just the
> old (3.8.x) kernel you're using?
> 

I think this is the follow-up thread <20150512013424.GA7960@oracle.com>

And the conclusion is 3.8 is too old so the fix is not there.

> (In either case netback's xenvif_tx_submit() calling
> skb_probe_transport_header() would seem pointless, as
> skb_checksum_setup() - with or without a fix - ought to be taking
> care of this anyway.)
> 

Patches welcome. :-)

Wei.

> Jan

Re: [Xen-devel] xen-netfront sets partial checksum at wrong offset

[Jan Beulich]

>>> On 29.05.15 at 12:39, <wei.liu2@citrix.com> wrote:
> On Fri, May 29, 2015 at 11:34:07AM +0100, Jan Beulich wrote:
>> >>> On 11.05.15 at 19:25, <venkat.x.venkatsubra@oracle.com> wrote:
>> >>  
>> >> Please CC the maintainers of the driver. You can get that from
>> >> 'scripts/get_maintainer.pl'
>> >> 
>> >> I've done that for you.
>> > 
>> > Thanks, Konrad.
>> > 
>> > I am copying Wei too who had fixed the below problem earlier.
>> > It fixed the incorrect ip_hdr(). tcp_hdr() still needs to fixed.
>> > 
>> > commit d554f73df6bc35ac8f6a65e5560bf1d31dfebed9
>> > Author: Wei Liu <wei.liu2@citrix.com>
>> > Date:   Wed Feb 19 18:48:34 2014 +0000
>> > 
>> >     xen-netfront: reset skb network header before checksum
>> 
>> So no response at all so far from the maintainers made me look into
>> this: I first thought what we need would be calls to
>> skb_probe_transport_header() in skb_checksum_setup_ip() after
>> each of the skb_maybe_pull_tail() functions. But
>> skb_partial_csum_set() already calls skb_set_transport_header(),
>> so I now think things ought to be fine without any change. Can
>> you clarify what you think is missing? Or is this an issue in just the
>> old (3.8.x) kernel you're using?
>> 
> 
> I think this is the follow-up thread <20150512013424.GA7960@oracle.com>
> 
> And the conclusion is 3.8 is too old so the fix is not there.

Ah, right. Seems like I failed to remove the earlier mail from my list of
things to look at when this came through. Sorry for the noise then.

Jan