[Intel-wired-lan] [RFC PATCH bpf-next v2 0/9] Introduce preferred busy-polling
Björn Töpel
bjorn.topel at gmail.com
Thu Nov 5 10:28:03 UTC 2020
This series introduces three new features:
1. A new "heavy traffic" busy-polling variant that works in concert
with the existing napi_defer_hard_irqs and gro_flush_timeout knobs.
2. A new socket option that let a user change the busy-polling NAPI
budget.
3. Allow busy-polling to be performed on XDP sockets.
The existing busy-polling mode, enabled by the SO_BUSY_POLL socket
option or system-wide using the /proc/sys/net/core/busy_read knob, is
an opportunistic. That means that if the NAPI context is not
scheduled, it will poll it. If, after busy-polling, the budget is
exceeded the busy-polling logic will schedule the NAPI onto the
regular softirq handling.
One implication of the behavior above is that a busy/heavy loaded NAPI
context will never enter/allow for busy-polling. Some applications
prefer that most NAPI processing would be done by busy-polling.
This series adds a new socket option, SO_PREFER_BUSY_POLL, that works
in concert with the napi_defer_hard_irqs and gro_flush_timeout
knobs. The napi_defer_hard_irqs and gro_flush_timeout knobs were
introduced in commit 6f8b12d661d0 ("net: napi: add hard irqs deferral
feature"), and allows for a user to defer interrupts to be enabled and
instead schedule the NAPI context from a watchdog timer. When a user
enables the SO_PREFER_BUSY_POLL, again with the other knobs enabled,
and the NAPI context is being processed by a softirq, the softirq NAPI
processing will exit early to allow the busy-polling to be performed.
If the application stops performing busy-polling via a system call,
the watchdog timer defined by gro_flush_timeout will timeout, and
regular softirq handling will resume.
In summary; Heavy traffic applications that prefer busy-polling over
softirq processing should use this option.
Example usage:
$ echo 2 | sudo tee /sys/class/net/ens785f1/napi_defer_hard_irqs
$ echo 200000 | sudo tee /sys/class/net/ens785f1/gro_flush_timeout
Note that the timeout should be larger than the userspace processing
window, otherwise the watchdog will timeout and fall back to regular
softirq processing.
Enable the SO_BUSY_POLL/SO_PREFER_BUSY_POLL options on your socket.
Performance netperf UDP_RR:
Note that netperf UDP_RR is not a heavy traffic tests, and preferred
busy-polling is not typically something we want to use here.
$ echo 20 | sudo tee /proc/sys/net/core/busy_read
$ netperf -H 192.168.1.1 -l 30 -t UDP_RR -v 2 -- \
-o min_latency,mean_latency,max_latency,stddev_latency,transaction_rate
busy-polling blocking sockets: 12,13.33,224,0.63,74731.177
I hacked netperf to use non-blocking sockets and re-ran:
busy-polling non-blocking sockets: 12,13.46,218,0.72,73991.172
prefer busy-polling non-blocking sockets: 12,13.62,221,0.59,73138.448
Using the preferred busy-polling mode does not impact performance.
Performance XDP sockets:
Today, running XDP sockets sample on the same core as the softirq
handling, performance tanks mainly because we do not yield to
user-space when the XDP socket Rx queue is full.
# taskset -c 5 ./xdpsock -i ens785f1 -q 5 -n 1 -r
Rx: 64Kpps
# # biased busy-polling, budget 8
# taskset -c 5 ./xdpsock -i ens785f1 -q 5 -n 1 -r -B -b 8
Rx 9.9Mpps
# # biased busy-polling, budget 64
# taskset -c 5 ./xdpsock -i ens785f1 -q 5 -n 1 -r -B -b 64
Rx: 19.3Mpps
# # biased busy-polling, budget 256
# taskset -c 5 ./xdpsock -i ens785f1 -q 5 -n 1 -r -B -b 256
Rx: 21.4Mpps
# # biased busy-polling, budget 512
# taskset -c 5 ./xdpsock -i ens785f1 -q 5 -n 1 -r -B -b 512
Rx: 21.7Mpps
Compared to the two-core case:
# taskset -c 4 ./xdpsock -i ens785f1 -q 20 -n 1 -r
Rx: 20.7Mpps
We're getting better single-core performance than two, for this naïve
drop scenario.
The above tests was done for the 'ice' driver.
Thanks to Jakub for suggesting this busy-polling addition [1], and
Eric for the input on the v1 RFC.
Some outstanding questions:
* Currently busy-polling for UDP/TCP is only wired up in the recvmsg()
path. Does it make sense to extend that to sendmsg() as well?
* Extending xdp_rxq_info_reg() with napi_id touches a lot of drivers,
and I've only verified the Intel ones. Some drivers initialize NAPI
(generating the napi_id) after the xdp_rxq_info_reg() call, which
maybe would open up for another API. I did not send this RFC to all
the driver authors. I'll do that for a patch proper series.
* Today, enabling busy-polling require CAP_NET_ADMIN. For a NAPI
context that services multiple socket, this makes sense because one
socket can affect performance of other sockets. Now, for a
*dedicated* queue for say XDP socket, would it be OK to drop
CAP_NET_ADMIN, because it cannot affect other sockets/users.
Changes:
rfc-v1 [2] -> rfc-v2:
* Changed name from bias to prefer.
* Base the work on Eric's/Luigi's defer irq/gro timeout work.
* Proper GRO flushing.
* Build issues for some XDP drivers.
[1] https://lore.kernel.org/netdev/20200925120652.10b8d7c5@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com/
[2] https://lore.kernel.org/bpf/20201028133437.212503-1-bjorn.topel@gmail.com/
Björn Töpel (9):
net: introduce preferred busy-polling
net: add SO_BUSY_POLL_BUDGET socket option
xsk: add support for recvmsg()
xsk: check need wakeup flag in sendmsg()
xsk: add busy-poll support for {recv,send}msg()
xsk: propagate napi_id to XDP socket Rx path
samples/bpf: use recvfrom() in xdpsock
samples/bpf: add busy-poll support to xdpsock
samples/bpf: add option to set the busy-poll budget
arch/alpha/include/uapi/asm/socket.h | 3 +
arch/mips/include/uapi/asm/socket.h | 3 +
arch/parisc/include/uapi/asm/socket.h | 3 +
arch/sparc/include/uapi/asm/socket.h | 3 +
drivers/net/ethernet/amazon/ena/ena_netdev.c | 2 +-
drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 +-
.../ethernet/cavium/thunder/nicvf_queues.c | 2 +-
.../net/ethernet/freescale/dpaa2/dpaa2-eth.c | 2 +-
drivers/net/ethernet/intel/i40e/i40e_txrx.c | 2 +-
drivers/net/ethernet/intel/ice/ice_base.c | 4 +-
drivers/net/ethernet/intel/ice/ice_txrx.c | 2 +-
drivers/net/ethernet/intel/igb/igb_main.c | 2 +-
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 2 +-
.../net/ethernet/intel/ixgbevf/ixgbevf_main.c | 2 +-
drivers/net/ethernet/marvell/mvneta.c | 2 +-
.../net/ethernet/marvell/mvpp2/mvpp2_main.c | 4 +-
drivers/net/ethernet/mellanox/mlx4/en_rx.c | 2 +-
.../net/ethernet/mellanox/mlx5/core/en_main.c | 2 +-
.../ethernet/netronome/nfp/nfp_net_common.c | 2 +-
drivers/net/ethernet/qlogic/qede/qede_main.c | 2 +-
drivers/net/ethernet/sfc/rx_common.c | 2 +-
drivers/net/ethernet/socionext/netsec.c | 2 +-
drivers/net/ethernet/ti/cpsw_priv.c | 2 +-
drivers/net/hyperv/netvsc.c | 2 +-
drivers/net/tun.c | 2 +-
drivers/net/veth.c | 2 +-
drivers/net/virtio_net.c | 2 +-
drivers/net/xen-netfront.c | 2 +-
fs/eventpoll.c | 3 +-
include/linux/netdevice.h | 35 +++++---
include/net/busy_poll.h | 27 ++++--
include/net/sock.h | 4 +
include/net/xdp.h | 3 +-
include/uapi/asm-generic/socket.h | 3 +
net/core/dev.c | 89 ++++++++++++++-----
net/core/sock.c | 19 ++++
net/core/xdp.c | 3 +-
net/xdp/xsk.c | 36 +++++++-
net/xdp/xsk_buff_pool.c | 13 ++-
samples/bpf/xdpsock_user.c | 53 ++++++++---
40 files changed, 262 insertions(+), 90 deletions(-)
base-commit: d0b3d2d7e50de5ce121f77a16df4c17e91b09421
--
2.27.0
More information about the Intel-wired-lan
mailing list