Re: [dpdk-stable] [dpdk-dev] [PATCH] net: fix unneeded replacement of 0 by ffff for TCP checksum

["Morten Brørup" <mb@smartsharesystems.com>]

> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Olivier Matz
> Sent: Friday, July 10, 2020 3:41 PM
> 
> On Fri, Jul 10, 2020 at 03:29:36PM +0200, Morten Brørup wrote:
> > > From: Olivier Matz [mailto:olivier.matz@6wind.com]
> > > Sent: Friday, July 10, 2020 3:16 PM
> > >
> > > On Fri, Jul 10, 2020 at 03:10:34PM +0200, Morten Brørup wrote:
> > > > > From: Olivier Matz [mailto:olivier.matz@6wind.com]
> > > > > Sent: Friday, July 10, 2020 2:41 PM
> > > > >
> > > > > On Fri, Jul 10, 2020 at 02:55:51PM +0800, Hongzhi Guo wrote:
> > > > > > Per RFC768:
> > > > > > If the computed checksum is zero, it is transmitted as all
> ones.
> > > > > > An all zero transmitted checksum value means that the
> transmitter
> > > > > > generated no checksum.
> > > > > >
> > > > > > RFC793 for TCP has no such special treatment for the checksum
> of
> > > > > zero.
> > > > > >
> > > > > > Fixes: 6006818cfb26 ("net: new checksum functions")
> > > > > > Cc: stable@dpdk.org
> > > > > >
> > > > > > Signed-off-by: Hongzhi Guo <guohongzhi1@huawei.com>
> > > > > > ---
> > > > > > v2:
> > > > > > * Fixed commit tile
> > > > > > * Fixed the API comment
> > > > > > ---
> > > > > > ---
> > > > > >  lib/librte_net/rte_ip.h | 18 +++++++++++++++---
> > > > > >  1 file changed, 15 insertions(+), 3 deletions(-)
> > > > > >
> > > > > > diff --git a/lib/librte_net/rte_ip.h
> b/lib/librte_net/rte_ip.h
> > > > > > index 292f63fd7..d03c77120 100644
> > > > > > --- a/lib/librte_net/rte_ip.h
> > > > > > +++ b/lib/librte_net/rte_ip.h
> > > > > > @@ -325,7 +325,7 @@ rte_ipv4_phdr_cksum(const struct
> rte_ipv4_hdr
> > > > > *ipv4_hdr, uint64_t ol_flags)
> > > > > >   *   The pointer to the beginning of the L4 header.
> > > > > >   * @return
> > > > > >   *   The complemented checksum to set in the IP packet
> > > > > > - *   or 0 on error
> > > > > > + *   or 0 if the IP length is invalid in the header.
> > > > > >   */
> > > > > >  static inline uint16_t
> > > > > >  rte_ipv4_udptcp_cksum(const struct rte_ipv4_hdr *ipv4_hdr,
> const
> > > > > void *l4_hdr)
> > > >
> > > > 0 is a valid return value, so I suggest omitting it from the
> return
> > > value description:
> > > >
> > > >   * @return
> > > > - *   The complemented checksum to set in the IP packet
> > > > - *   or 0 on error
> > > > + *   The complemented checksum to set in the IP packet.
> > > >
> > > > The comparison "if (l3_len < sizeof(struct rte_ipv4_hdr))" is
> only
> > > there to protect against invalid input; it prevents l4_len from
> > > becoming negative.
> > >
> > > I don't get why "0 if the IP length is invalid in the header"
> should
> > > be removed from the comment: 0 is both a valid return value and
> > > the value returned on invalid packet.
> >
> > To avoid confusion. We do not want people to add error handling for a
> return value of 0.
> >
> > 0 is not a special value or an error, so it does not deserve explicit
> mentioning.
> >
> > If we want to mention the return value for garbage input, we should
> not use the wording "or 0", because this suggests that 0 is not a
> normal return value.
> 
> Ok, got it.
> 
> So maybe this?
> 
>  The complemented checksum to set in the IP packet. If
>  the IP length is invalid in the header, it returns 0.
> 
It still mentions 0 as a special value, increasing the risk of the defensive user adding "error handling" for a return value of 0.

How about this?

 The complemented checksum to set in the IP packet. If
 the IP length is invalid in the header, the return value
 is undefined.

> 
> >
> > >
> > > > For the same reason, unlikely() should be added to this
> comparison.
> > >
> > > Maybe yes, but that's another story I think.
> >
> > Agree. I was just mentioning it so it can be done when modifying the
> function anyway.
> >
> > >
> > > > Otherwise,
> > > >
> > > > Acked-by: Morten Brørup <mb@smartsharesystems.com>
> > > >
> >