Re: [dpdk-dev] [PATCH] lib/librte_net: fix bug for ipv4 checksumcalculating

[Olivier Matz <olivier.matz@6wind.com>]

Hi,

On Thu, May 14, 2020 at 02:56:41PM +0200, Morten Brørup wrote:
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of guohongzhi
> > Sent: Thursday, May 14, 2020 3:27 AM
> > 
> > The function of rte_ipv4_cksum for calculating the
> > checksum of IPv4 header is incorrect.
> > This function will return checksum value like 0xffff.
> > This value, however, is considered an illegal checksum on some
> > switches(like Trident3).
> > 
> > RFC 1624 specifies the IPv4 checksum as follows:
> > https://tools.ietf.org/rfc/rfc1624
> > Since there is guaranteed to be at least one
> >    non-zero field in the IP header, and the checksum field in the
> >    protocol header is the complement of the sum, the checksum field can
> >    never contain ~(+0), which is -0 (0xFFFF).  It can, however, contain
> >    ~(-0), which is +0 (0x0000).
> > 
> > ---
> >  lib/librte_net/rte_ip.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/lib/librte_net/rte_ip.h b/lib/librte_net/rte_ip.h
> > index 1ceb7b7..ece2e43 100644
> > --- a/lib/librte_net/rte_ip.h
> > +++ b/lib/librte_net/rte_ip.h
> > @@ -267,7 +267,7 @@ rte_ipv4_cksum(const struct rte_ipv4_hdr *ipv4_hdr)
> >  {
> >  	uint16_t cksum;
> >  	cksum = rte_raw_cksum(ipv4_hdr, sizeof(struct rte_ipv4_hdr));
> > -	return (cksum == 0xffff) ? cksum : (uint16_t)~cksum;
> > +	return (uint16_t)~cksum;
> >  }
> > 
> >  /**
> > --
> > 2.21.0.windows.1
> > 
> > 
> 
> Well spotted!

Indeed.

> Reviewed-By: Morten Brørup <mb@smartsharesystems.com>

Fixes: 6006818cfb26 ("net: new checksum functions")
Cc: stable@dpdk.org

Acked-by: Olivier Matz <olivier.matz@6wind.com>

> Would you consider writing another patch splitting
> rte_ipv4_udptcp_cksum() up into rte_ipv4_udp_cksum() and
> rte_ipv4_tcp_cksum(), so the TCP checksum will be calculated
> correctly?
> 
> RFC 768 for UDP specifies:
> 
> If the computed checksum is zero, it is transmitted as all ones (the
> equivalent in one's complement arithmetic).  An all zero transmitted
> checksum value means that the transmitter generated no checksum (for
> debugging or for higher level protocols that don't care).
>
> RFC 793 for TCP has no such special treatment for the checksum of
> zero, but rte_ipv4_udptcp_cksum() implements the UDP special treatment
> anyway.

I agree the following test is useless in case of TCPv4 and TCPv6:

	if (cksum == 0)
                cksum =	0xffff;

For UDPv4, it is needed because 0 means "no checksum".
For UDPv6, it is needed because 0 is forbidden.

So yes, I think we could have specific csum functions for tcp and udp
checksum as Morten suggests (as soon as we keep the backward compat).