Re: [PATCH net v2 3/4] udp: do not transition UDP fraglist to unnecessary checksum

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Antoine Tenart wrote:
> Quoting Willem de Bruijn (2024-03-19 14:38:20)
> > Antoine Tenart wrote:
> > > udp4/6_gro_complete transition fraglist packets to CHECKSUM_UNNECESSARY
> > > and sets their checksum level based on if the packet is recognized to be
> > > a tunneled one. However there is no safe way to detect a packet is a
> > > tunneled one and in case such packet is GROed at the UDP level, setting
> > > a wrong checksum level will lead to later errors. For example if those
> > > packets are forwarded to the Tx path they could produce the following
> > > dump:
> > > 
> > >   gen01: hw csum failure
> > >   skb len=3008 headroom=160 headlen=1376 tailroom=0
> > >   mac=(106,14) net=(120,40) trans=160
> > >   shinfo(txflags=0 nr_frags=0 gso(size=0 type=0 segs=0))
> > >   csum(0xffff232e ip_summed=2 complete_sw=0 valid=0 level=0)
> > >   hash(0x77e3d716 sw=1 l4=1) proto=0x86dd pkttype=0 iif=12
> > >   ...
> > > 
> > > Fixes: 9fd1ff5d2ac7 ("udp: Support UDP fraglist GRO/GSO.")
> > > Signed-off-by: Antoine Tenart <atenart@kernel.org>
> > 
> > The original patch converted to CHECKSUM_UNNECESSARY for a reason.
> > The skb->csum of the main gso_skb is not valid?
> > 
> > Should instead only the csum_level be adjusted, to always keep
> > csum_level == 0?
> 
> The above trace is an ICMPv6 packet being tunneled and GROed at the UDP
> level, thus we have:
>   UDP(CHECKSUM_PARTIAL)/Geneve/ICMPv6(was CHECKSUM_NONE)
> csum_level would need to be 1 here; but we can't know that.

Is this a packet looped internally? Else it is not CHECKSUM_PARTIAL.
Looped packets can trivially be converted to CHECKSUM_UNNECESSARY.
It just has to be level 0 if only the outer checksum is known.

> There is another issue (no kernel trace): if a packet has partial csum
> and is being GROed that information is lost and the packet ends up with
> an invalid csum.

CHECKSUM_PARTIAL should be converted to CHECKSUM_UNNECESSARY for this
reason. CHECKSUM_PARTIAL implies the header is prepared with pseudo
header checksum. Similarly CHECKSUM_COMPLETE implies skb csum is valid.
CHECKSUM_UNNECESSARY has neither expectations.
 
> Packets with CHECKSUM_UNNECESSARY should end up with the same info. My
> impression is this checksum conversion is at best setting the same info
> and otherwise is overriding valuable csum information.
> 
> Or would packets with CSUM_NONE being GROed would benefit from the
> CHECKSUM_UNNECESSARY conversion?

Definitely. If the packet has CHECKSUM_NONE and GRO checks its
validity in software, converting it to CHECKSUM_UNNECESSARY avoids
potential additional checks at later stages in the packet path.

> 
> For reference, original commit says:
> """
> After validating the csum,  we mark ip_summed as
> CHECKSUM_UNNECESSARY for fraglist GRO packets to
> make sure that the csum is not touched.
> """
> 
> But I'm failing to see where that would happen and how the none to
> unnecessary conversion would help. WDYT?

I would appreciate the GRO and fraglist GRO authors to also review
this series and chime in.

> 
> Thanks,
> Antoine