| #
bd6c11bc |
| 29-Aug-2023 |
Linus Torvalds <torvalds@linux-foundation.org> |
Merge tag 'net-next-6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next
Pull networking updates from Paolo Abeni:
"Core:
- Increase size limits for to-be-sent skb frag allocat
Merge tag 'net-next-6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next
Pull networking updates from Paolo Abeni:
"Core:
- Increase size limits for to-be-sent skb frag allocations. This
allows tun, tap devices and packet sockets to better cope with
large writes operations
- Store netdevs in an xarray, to simplify iterating over netdevs
- Refactor nexthop selection for multipath routes
- Improve sched class lifetime handling
- Add backup nexthop ID support for bridge
- Implement drop reasons support in openvswitch
- Several data races annotations and fixes
- Constify the sk parameter of routing functions
- Prepend kernel version to netconsole message
Protocols:
- Implement support for TCP probing the peer being under memory
pressure
- Remove hard coded limitation on IPv6 specific info placement inside
the socket struct
- Get rid of sysctl_tcp_adv_win_scale and use an auto-estimated per
socket scaling factor
- Scaling-up the IPv6 expired route GC via a separated list of
expiring routes
- In-kernel support for the TLS alert protocol
- Better support for UDP reuseport with connected sockets
- Add NEXT-C-SID support for SRv6 End.X behavior, reducing the SR
header size
- Get rid of additional ancillary per MPTCP connection struct socket
- Implement support for BPF-based MPTCP packet schedulers
- Format MPTCP subtests selftests results in TAP
- Several new SMC 2.1 features including unique experimental options,
max connections per lgr negotiation, max links per lgr negotiation
BPF:
- Multi-buffer support in AF_XDP
- Add multi uprobe BPF links for attaching multiple uprobes and usdt
probes, which is significantly faster and saves extra fds
- Implement an fd-based tc BPF attach API (TCX) and BPF link support
on top of it
- Add SO_REUSEPORT support for TC bpf_sk_assign
- Support new instructions from cpu v4 to simplify the generated code
and feature completeness, for x86, arm64, riscv64
- Support defragmenting IPv(4|6) packets in BPF
- Teach verifier actual bounds of bpf_get_smp_processor_id() and fix
perf+libbpf issue related to custom section handling
- Introduce bpf map element count and enable it for all program types
- Add a BPF hook in sys_socket() to change the protocol ID from
IPPROTO_TCP to IPPROTO_MPTCP to cover migration for legacy
- Introduce bpf_me_mcache_free_rcu() and fix OOM under stress
- Add uprobe support for the bpf_get_func_ip helper
- Check skb ownership against full socket
- Support for up to 12 arguments in BPF trampoline
- Extend link_info for kprobe_multi and perf_event links
Netfilter:
- Speed-up process exit by aborting ruleset validation if a fatal
signal is pending
- Allow NLA_POLICY_MASK to be used with BE16/BE32 types
Driver API:
- Page pool optimizations, to improve data locality and cache usage
- Introduce ndo_hwtstamp_get() and ndo_hwtstamp_set() to avoid the
need for raw ioctl() handling in drivers
- Simplify genetlink dump operations (doit/dumpit) providing them the
common information already populated in struct genl_info
- Extend and use the yaml devlink specs to [re]generate the split ops
- Introduce devlink selective dumps, to allow SF filtering SF based
on handle and other attributes
- Add yaml netlink spec for netlink-raw families, allow route, link
and address related queries via the ynl tool
- Remove phylink legacy mode support
- Support offload LED blinking to phy
- Add devlink port function attributes for IPsec
New hardware / drivers:
- Ethernet:
- Broadcom ASP 2.0 (72165) ethernet controller
- MediaTek MT7988 SoC
- Texas Instruments AM654 SoC
- Texas Instruments IEP driver
- Atheros qca8081 phy
- Marvell 88Q2110 phy
- NXP TJA1120 phy
- WiFi:
- MediaTek mt7981 support
- Can:
- Kvaser SmartFusion2 PCI Express devices
- Allwinner T113 controllers
- Texas Instruments tcan4552/4553 chips
- Bluetooth:
- Intel Gale Peak
- Qualcomm WCN3988 and WCN7850
- NXP AW693 and IW624
- Mediatek MT2925
Drivers:
- Ethernet NICs:
- nVidia/Mellanox:
- mlx5:
- support UDP encapsulation in packet offload mode
- IPsec packet offload support in eswitch mode
- improve aRFS observability by adding new set of counters
- extends MACsec offload support to cover RoCE traffic
- dynamic completion EQs
- mlx4:
- convert to use auxiliary bus instead of custom interface
logic
- Intel
- ice:
- implement switchdev bridge offload, even for LAG
interfaces
- implement SRIOV support for LAG interfaces
- igc:
- add support for multiple in-flight TX timestamps
- Broadcom:
- bnxt:
- use the unified RX page pool buffers for XDP and non-XDP
- use the NAPI skb allocation cache
- OcteonTX2:
- support Round Robin scheduling HTB offload
- TC flower offload support for SPI field
- Freescale:
- add XDP_TX feature support
- AMD:
- ionic: add support for PCI FLR event
- sfc:
- basic conntrack offload
- introduce eth, ipv4 and ipv6 pedit offloads
- ST Microelectronics:
- stmmac: maximze PTP timestamping resolution
- Virtual NICs:
- Microsoft vNIC:
- batch ringing RX queue doorbell on receiving packets
- add page pool for RX buffers
- Virtio vNIC:
- add per queue interrupt coalescing support
- Google vNIC:
- add queue-page-list mode support
- Ethernet high-speed switches:
- nVidia/Mellanox (mlxsw):
- add port range matching tc-flower offload
- permit enslavement to netdevices with uppers
- Ethernet embedded switches:
- Marvell (mv88e6xxx):
- convert to phylink_pcs
- Renesas:
- r8A779fx: add speed change support
- rzn1: enables vlan support
- Ethernet PHYs:
- convert mv88e6xxx to phylink_pcs
- WiFi:
- Qualcomm Wi-Fi 7 (ath12k):
- extremely High Throughput (EHT) PHY support
- RealTek (rtl8xxxu):
- enable AP mode for: RTL8192FU, RTL8710BU (RTL8188GU),
RTL8192EU and RTL8723BU
- RealTek (rtw89):
- Introduce Time Averaged SAR (TAS) support
- Connector:
- support for event filtering"
* tag 'net-next-6.6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1806 commits)
net: ethernet: mtk_wed: minor change in wed_{tx,rx}info_show
net: ethernet: mtk_wed: add some more info in wed_txinfo_show handler
net: stmmac: clarify difference between "interface" and "phy_interface"
r8152: add vendor/device ID pair for D-Link DUB-E250
devlink: move devlink_notify_register/unregister() to dev.c
devlink: move small_ops definition into netlink.c
devlink: move tracepoint definitions into core.c
devlink: push linecard related code into separate file
devlink: push rate related code into separate file
devlink: push trap related code into separate file
devlink: use tracepoint_enabled() helper
devlink: push region related code into separate file
devlink: push param related code into separate file
devlink: push resource related code into separate file
devlink: push dpipe related code into separate file
devlink: move and rename devlink_dpipe_send_and_alloc_skb() helper
devlink: push shared buffer related code into separate file
devlink: push port related code into separate file
devlink: push object register/unregister notifications into separate helpers
inet: fix IP_TRANSPARENT error handling
...
show more ...
|
| #
54f00cce |
| 10-Aug-2023 |
William Tu <u9012063@gmail.com> |
vmxnet3: Add XDP support.
The patch adds native-mode XDP support: XDP DROP, PASS, TX, and REDIRECT.
Background:
The vmxnet3 rx consists of three rings: ring0, ring1, and dataring.
For r0 and r1, bu
vmxnet3: Add XDP support.
The patch adds native-mode XDP support: XDP DROP, PASS, TX, and REDIRECT.
Background:
The vmxnet3 rx consists of three rings: ring0, ring1, and dataring.
For r0 and r1, buffers at r0 are allocated using alloc_skb APIs and dma
mapped to the ring's descriptor. If LRO is enabled and packet size larger
than 3K, VMXNET3_MAX_SKB_BUF_SIZE, then r1 is used to mapped the rest of
the buffer larger than VMXNET3_MAX_SKB_BUF_SIZE. Each buffer in r1 is
allocated using alloc_page. So for LRO packets, the payload will be in one
buffer from r0 and multiple from r1, for non-LRO packets, only one
descriptor in r0 is used for packet size less than 3k.
When receiving a packet, the first descriptor will have the sop (start of
packet) bit set, and the last descriptor will have the eop (end of packet)
bit set. Non-LRO packets will have only one descriptor with both sop and
eop set.
Other than r0 and r1, vmxnet3 dataring is specifically designed for
handling packets with small size, usually 128 bytes, defined in
VMXNET3_DEF_RXDATA_DESC_SIZE, by simply copying the packet from the backend
driver in ESXi to the ring's memory region at front-end vmxnet3 driver, in
order to avoid memory mapping/unmapping overhead. In summary, packet size:
A. < 128B: use dataring
B. 128B - 3K: use ring0 (VMXNET3_RX_BUF_SKB)
C. > 3K: use ring0 and ring1 (VMXNET3_RX_BUF_SKB + VMXNET3_RX_BUF_PAGE)
As a result, the patch adds XDP support for packets using dataring
and r0 (case A and B), not the large packet size when LRO is enabled.
XDP Implementation:
When user loads and XDP prog, vmxnet3 driver checks configurations, such
as mtu, lro, and re-allocate the rx buffer size for reserving the extra
headroom, XDP_PACKET_HEADROOM, for XDP frame. The XDP prog will then be
associated with every rx queue of the device. Note that when using dataring
for small packet size, vmxnet3 (front-end driver) doesn't control the
buffer allocation, as a result we allocate a new page and copy packet
from the dataring to XDP frame.
The receive side of XDP is implemented for case A and B, by invoking the
bpf program at vmxnet3_rq_rx_complete and handle its returned action.
The vmxnet3_process_xdp(), vmxnet3_process_xdp_small() function handles
the ring0 and dataring case separately, and decides the next journey of
the packet afterward.
For TX, vmxnet3 has split header design. Outgoing packets are parsed
first and protocol headers (L2/L3/L4) are copied to the backend. The
rest of the payload are dma mapped. Since XDP_TX does not parse the
packet protocol, the entire XDP frame is dma mapped for transmission
and transmitted in a batch. Later on, the frame is freed and recycled
back to the memory pool.
Performance:
Tested using two VMs inside one ESXi vSphere 7.0 machine, using single
core on each vmxnet3 device, sender using DPDK testpmd tx-mode attached
to vmxnet3 device, sending 64B or 512B UDP packet.
VM1 txgen:
$ dpdk-testpmd -l 0-3 -n 1 -- -i --nb-cores=3 \
--forward-mode=txonly --eth-peer=0,<mac addr of vm2>
option: add "--txonly-multi-flow"
option: use --txpkts=512 or 64 byte
VM2 running XDP:
$ ./samples/bpf/xdp_rxq_info -d ens160 -a <options> --skb-mode
$ ./samples/bpf/xdp_rxq_info -d ens160 -a <options>
options: XDP_DROP, XDP_PASS, XDP_TX
To test REDIRECT to cpu 0, use
$ ./samples/bpf/xdp_redirect_cpu -d ens160 -c 0 -e drop
Single core performance comparison with skb-mode.
64B: skb-mode -> native-mode
XDP_DROP: 1.6Mpps -> 2.4Mpps
XDP_PASS: 338Kpps -> 367Kpps
XDP_TX: 1.1Mpps -> 2.3Mpps
REDIRECT-drop: 1.3Mpps -> 2.3Mpps
512B: skb-mode -> native-mode
XDP_DROP: 863Kpps -> 1.3Mpps
XDP_PASS: 275Kpps -> 376Kpps
XDP_TX: 554Kpps -> 1.2Mpps
REDIRECT-drop: 659Kpps -> 1.2Mpps
Demo: https://youtu.be/4lm1CSCi78Q
Future work:
- XDP frag support
- use napi_consume_skb() instead of dev_kfree_skb_any at unmap
- stats using u64_stats_t
- using bitfield macro BIT()
- optimization for DMA synchronization using actual frame length,
instead of always max_len
Signed-off-by: William Tu <u9012063@gmail.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
show more ...
|