| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <20210329134013.9516-1-boon.leong.ong@intel.com>
Date: Mon, 29 Mar 2021 21:40:07 +0800
From: Ong Boon Leong <boon.leong.ong@...el.com>
To: Giuseppe Cavallaro <peppe.cavallaro@...com>,
Alexandre Torgue <alexandre.torgue@...com>,
Jose Abreu <joabreu@...opsys.com>,
"David S . Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>,
Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
Jesper Dangaard Brouer <hawk@...nel.org>,
John Fastabend <john.fastabend@...il.com>
Cc: Maxime Coquelin <mcoquelin.stm32@...il.com>,
Andrii Nakryiko <andrii@...nel.org>,
Martin KaFai Lau <kafai@...com>,
Song Liu <songliubraving@...com>, Yonghong Song <yhs@...com>,
KP Singh <kpsingh@...nel.org>, netdev@...r.kernel.org,
linux-stm32@...md-mailman.stormreply.com,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
bpf@...r.kernel.org, Ong Boon Leong <boon.leong.ong@...el.com>
Subject: [PATCH net-next 0/6] stmmac: Add XDP support
Hi,
This is the patch series to adds XDP support to stmmac driver.
Summary of the changes as follow:-
1/6: Add IRQ affinity hint to make RXQ and TXQ to be by default serviced
by the same CPU. This is required for Intel mGbE controller which
has independent per DMA channel RX & and TX IRQ resources.
2/6: To add capability to enable/disable Split Header (SPH) dynamically.
Disable SPH for XDP and revert back to the SPH capability of the IP
when XDP exits.
3/6: To rearrange TX tail pointer update into common function for reuse
across driver.
4/6: Add basic XDP framework into stmmac driver. This patch only support
XDP_DROP only.
5/6: Add support for XDP_TX.
6/6: Add support for XDP_REDIRECT and ndo_xdp_xmit() implementation.
To send burst traffics to DUT, we use samples/pktgen script called
pktgen_sample03_burst_single_flow.sh. Following shows the test steps
and results that are obtained using Intel mGbE controller under
preempt-rt environment. In such environment, I also have sent a patch
that fixes an issue found in xdp_return_frame() earlier:
https://patchwork.kernel.org/project/netdevbpf/list/?series=457139
########################################################################
root@...el-corei7-64:~# ./xdp1 eth0
libbpf: elf: skipping unrecognized data section(16) .eh_frame
libbpf: elf: skipping relo section(17) .rel.eh_frame for section(16) .eh_frame
proto 17: 421844 pkt/s
proto 17: 684396 pkt/s
proto 17: 684332 pkt/s
proto 17: 684074 pkt/s
proto 17: 683837 pkt/s
proto 17: 684357 pkt/s
proto 17: 684371 pkt/s
proto 17: 684419 pkt/s
proto 17: 683996 pkt/s
root@...el-corei7-64:~# ./xdp1 -S eth0
libbpf: elf: skipping unrecognized data section(16) .eh_frame
libbpf: elf: skipping relo section(17) .rel.eh_frame for section(16) .eh_frame
proto 0: 1 pkt/s
proto 17: 1 pkt/s
proto 17: 1 pkt/s
proto 17: 65564 pkt/s
proto 17: 117450 pkt/s
proto 17: 117600 pkt/s
proto 17: 118108 pkt/s
proto 17: 118032 pkt/s
proto 17: 118092 pkt/s
root@...el-corei7-64:~# ./xdp2 eth0
libbpf: elf: skipping unrecognized data section(16) .eh_frame
libbpf: elf: skipping relo section(17) .rel.eh_frame for section(16) .eh_frame
proto 17: 0 pkt/s
proto 17: 49338 pkt/s
proto 17: 714255 pkt/s
proto 17: 714495 pkt/s
proto 17: 710120 pkt/s
proto 17: 712892 pkt/s
proto 17: 712585 pkt/s
proto 17: 713883 pkt/s
root@...el-corei7-64:~# ./xdp2 -S eth0
libbpf: elf: skipping unrecognized data section(16) .eh_frame
libbpf: elf: skipping relo section(17) .rel.eh_frame for section(16) .eh_frame
proto 0: 0 pkt/s
proto 17: 89483 pkt/s
proto 17: 89540 pkt/s
proto 17: 89358 pkt/s
proto 17: 89797 pkt/s
proto 0: 1 pkt/s
proto 17: 89672 pkt/s
root@...el-corei7-64:~# ./xdp_redirect eth0 eth1
input: 7 output: 8
libbpf: elf: skipping unrecognized data section(20) .eh_frame
libbpf: elf: skipping relo section(21) .rel.eh_frame for section(20) .eh_frame
ifindex 8: 143805 pkt/s
ifindex 8: 676900 pkt/s
ifindex 8: 676801 pkt/s
ifindex 8: 677657 pkt/s
ifindex 8: 677916 pkt/s
root@...el-corei7-64:~# ./xdp_redirect -S eth0 eth1
input: 7 output: 8
libbpf: elf: skipping unrecognized data section(20) .eh_frame
libbpf: elf: skipping relo section(21) .rel.eh_frame for section(20) .eh_frame
ifindex 8: 2 pkt/s
ifindex 8: 0 pkt/s
ifindex 8: 33505 pkt/s
ifindex 8: 73536 pkt/s
ifindex 8: 52512 pkt/s
ifindex 8: 97600 pkt/s
ifindex 8: 96928 pkt/s
ifindex 8: 96480 pkt/s
ifindex 8: 96760 pkt/s
ifindex 8: 96949 pkt/s
ifindex 8: 96591 pkt/s
root@...el-corei7-64:~# ./xdp_redirect_cpu --dev eth0 --cpu 0
libbpf: elf: skipping unrecognized data section(39) .eh_frame
libbpf: elf: skipping relo section(40) .rel.eh_frame for section(39) .eh_frame
libbpf: elf: skipping unrecognized data section(20) .eh_frame
libbpf: elf: skipping relo section(21) .rel.eh_frame for section(20) .eh_frame
Add-new CPU:0 as idx:0 qsize:192 prog_fd: 33 (cpus_count:1)
Running XDP/eBPF prog_name:xdp_cpu_map5_lb_hash_ip_pairs
XDP-cpumap CPU:to pps drop-pps extra-info
XDP-RX 0 667,952 0 0
XDP-RX total 667,952 0
cpumap-enqueue 0:0 667,952 550,150 7.82 bulk-average
cpumap-enqueue sum:0 667,952 550,150 7.82 bulk-average
cpumap_kthread 0 117,821 0 0
cpumap_kthread total 117,821 0 0
redirect_err total 0 0
xdp_exception total 0 0
2nd remote XDP/eBPF prog_name: xdp_redirect_dummy
XDP-cpumap CPU:to xdp-pass xdp-drop xdp-redir
xdp-in-kthread 0 117,822 0 0
xdp-in-kthread total 117,822 0 0
Running XDP/eBPF prog_name:xdp_cpu_map5_lb_hash_ip_pairs
XDP-cpumap CPU:to pps drop-pps extra-info
XDP-RX 0 667,871 0 0
XDP-RX total 667,871 0
cpumap-enqueue 0:0 667,877 550,820 7.82 bulk-average
cpumap-enqueue sum:0 667,877 550,820 7.82 bulk-average
cpumap_kthread 0 117,041 0 0
cpumap_kthread total 117,041 0 0
redirect_err total 0 0
xdp_exception total 0 0
2nd remote XDP/eBPF prog_name: xdp_redirect_dummy
XDP-cpumap CPU:to xdp-pass xdp-drop xdp-redir
xdp-in-kthread 0 117,041 0 0
xdp-in-kthread total 117,041 0 0
Running XDP/eBPF prog_name:xdp_cpu_map5_lb_hash_ip_pairs
XDP-cpumap CPU:to pps drop-pps extra-info
XDP-RX 0 667,856 0 0
XDP-RX total 667,856 0
cpumap-enqueue 0:0 667,849 549,672 7.82 bulk-average
cpumap-enqueue sum:0 667,849 549,672 7.82 bulk-average
cpumap_kthread 0 118,177 0 0
cpumap_kthread total 118,177 0 0
redirect_err total 0 0
xdp_exception total 0 0
2nd remote XDP/eBPF prog_name: xdp_redirect_dummy
XDP-cpumap CPU:to xdp-pass xdp-drop xdp-redir
xdp-in-kthread 0 118,177 0 0
xdp-in-kthread total 118,177 0 0
########################################################################
As stmmac driver is using page_pool mechanism, I waited > 1min after
each of above apps is terminated to make sure there is no stalled
pool warning prints on the terminal, and I found none which looks good
on Intel mGbE platform.
It will be great if community help tes out these patch series on your
platform and provide me feedback.
Thank you very much,
Boon Leong
Ong Boon Leong (6):
stmmac: intel: set IRQ affinity hint for multi MSI vectors
net: stmmac: make SPH enable/disable to be configurable
net: stmmac: arrange Tx tail pointer update to
stmmac_flush_tx_descriptors
net: stmmac: Add initial XDP support
net: stmmac: Add support for XDP_TX action
net: stmmac: Add support for XDP_REDIRECT action
drivers/net/ethernet/stmicro/stmmac/Makefile | 1 +
.../net/ethernet/stmicro/stmmac/dwmac-intel.c | 7 +
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 35 +-
.../net/ethernet/stmicro/stmmac/stmmac_main.c | 502 +++++++++++++++---
.../net/ethernet/stmicro/stmmac/stmmac_xdp.c | 40 ++
.../net/ethernet/stmicro/stmmac/stmmac_xdp.h | 12 +
6 files changed, 523 insertions(+), 74 deletions(-)
create mode 100644 drivers/net/ethernet/stmicro/stmmac/stmmac_xdp.c
create mode 100644 drivers/net/ethernet/stmicro/stmmac/stmmac_xdp.h
--
2.25.1
Powered by blists - more mailing lists