Re: [RFC net-next v4 8/8] selftests/bpf: Add TX side to xdp_hw_metadata

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 07/25, Willem de Bruijn wrote:
> Stanislav Fomichev wrote:
> > When we get packets on port 9091, we swap src/dst and send it out.
> > At this point, we also request the timestamp and plumb it back
> > to the userspace. The userspace simply prints the timestamp.
> > 
> > Also print current UDP checksum, rewrite it with the pseudo-header
> > checksum and offload TX checksum calculation to devtx. Upon
> > completion, report TX checksum back (mlx5 doesn't put it back, so
> > I've used tcpdump to confirm that the checksum is correct).
> > 
> > Some other related changes:
> > - switched to zerocopy mode by default; new flag can be used to force
> >   old behavior
> > - request fixed TX_METADATA_LEN headroom
> > - some other small fixes (umem size, fill idx+i, etc)
> > 
> > mvbz3:~# ./xdp_hw_metadata eth3 -c mlx5e_devtx_complete_xdp -s mlx5e_devtx_submit_xd
> > attach rx bpf program...
> > ...
> > 0x206d298: rx_desc[0]->addr=80100 addr=80100 comp_addr=80100
> > rx_hash: 0x2BFB7FEC with RSS type:0x2A
> > rx_timestamp:  1690238278345877848 (sec:1690238278.3459)
> > XDP RX-time:   1690238278538397674 (sec:1690238278.5384) delta sec:0.1925 (192519.826 usec)
> > AF_XDP time:   1690238278538515250 (sec:1690238278.5385) delta sec:0.0001 (117.576 usec)
> > 0x206d298: ping-pong with csum=8e3b (want 57c9) csum_start=54 csum_offset=6
> > 0x206d298: complete tx idx=0 addr=10
> > 0x206d298: tx_timestamp:  1690238278577008140 (sec:1690238278.5770)
> > 0x206d298: complete rx idx=128 addr=80100
> > 
> > mvbz4:~# nc  -Nu -q1 ${MVBZ3_LINK_LOCAL_IP}%eth3 9091
> > 
> > mvbz4:~# tcpdump -vvx -i eth3 udp
> > tcpdump: listening on eth3, link-type EN10MB (Ethernet), snapshot length 262144 bytes
> > 10:12:43.901436 IP6 (flowlabel 0x7a5d2, hlim 127, next-header UDP (17) payload length: 11) fe80::1270:fdff:fe48:1087.44339 > fe80::1270:fdff:fe48:1077.9091: [bad udp cksum 0x3b8e -> 0x0b4b!] UDP, length 3
> >         0x0000:  6007 a5d2 000b 117f fe80 0000 0000 0000
> >         0x0010:  1270 fdff fe48 1087 fe80 0000 0000 0000
> >         0x0020:  1270 fdff fe48 1077 ad33 2383 000b 3b8e
> >         0x0030:  7864 70
> > 10:12:43.902125 IP6 (flowlabel 0x7a5d2, hlim 127, next-header UDP (17) payload length: 11) fe80::1270:fdff:fe48:1077.9091 > fe80::1270:fdff:fe48:1087.44339: [udp sum ok] UDP, length 3
> >         0x0000:  6007 a5d2 000b 117f fe80 0000 0000 0000
> >         0x0010:  1270 fdff fe48 1077 fe80 0000 0000 0000
> >         0x0020:  1270 fdff fe48 1087 2383 ad33 000b 0b4b
> >         0x0030:  7864 70
> > 
> > Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
> > ---
> >  tools/testing/selftests/bpf/xdp_hw_metadata.c | 201 +++++++++++++++++-
> >  1 file changed, 191 insertions(+), 10 deletions(-)
> > 
> 
> > +static void usage(const char *prog)
> > +{
> > +	fprintf(stderr,
> > +		"usage: %s [OPTS] <ifname>\n"
> > +		"OPTS:\n"
> > +		"    -T    don't generate AF_XDP reply (rx metadata only)\n"
> > +		"    -Z    run in copy mode\n",
> 
> nit: makes more sense to call copy mode 'C', rather than 'Z'

Initially I had -c and -s for completion/submission bpf hooks. Now that
these are gone, can actually use -c. Capital letter here actually means
'not'. Z - not zerocopy. T - no tx.

I'll rename to:
-r - rx only
-c - copy mode

LMK if it doesn't make sense..




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux