mirror of
https://github.com/torvalds/linux.git
synced 2024-12-05 02:23:16 +00:00
71634d7f92
Adds support for fq's Earliest Departure Time to HBM (Host Bandwidth Manager). Includes a new BPF program supporting EDT, and also updates corresponding programs. It will drop packets with an EDT of more than 500us in the future unless the packet belongs to a flow with less than 2 packets in flight. This is done so each flow has at least 2 packets in flight, so they will not starve, and also to help prevent delayed ACK timeouts. It will also work with ECN enabled traffic, where the packets will be CE marked if their EDT is more than 50us in the future. The table below shows some performance numbers. The flows are back to back RPCS. One server sending to another, either 2 or 4 flows. One flow is a 10KB RPC, the rest are 1MB RPCs. When there are more than one flow of a given RPC size, the numbers represent averages. The rate limit applies to all flows (they are in the same cgroup). Tests ending with "-edt" ran with the new BPF program supporting EDT. Tests ending with "-hbt" ran on top HBT qdisc with the specified rate (i.e. no HBM). The other tests ran with the HBM BPF program included in the HBM patch-set. EDT has limited value when using DCTCP, but it helps in many cases when using Cubic. It usually achieves larger link utilization and lower 99% latencies for the 1MB RPCs. HBM ends up queueing a lot of packets with its default parameter values, reducing the goodput of the 10KB RPCs and increasing their latency. Also, the RTTs seen by the flows are quite large. Aggr 10K 10K 10K 1MB 1MB 1MB Limit rate drops RTT rate P90 P99 rate P90 P99 Test rate Flows Mbps % us Mbps us us Mbps ms ms -------- ---- ----- ---- ----- --- ---- ---- ---- ---- ---- ---- cubic 1G 2 904 0.02 108 257 511 539 647 13.4 24.5 cubic-edt 1G 2 982 0.01 156 239 656 967 743 14.0 17.2 dctcp 1G 2 977 0.00 105 324 408 744 653 14.5 15.9 dctcp-edt 1G 2 981 0.01 142 321 417 811 660 15.7 17.0 cubic-htb 1G 2 919 0.00 1825 40 2822 4140 879 9.7 9.9 cubic 200M 2 155 0.30 220 81 532 655 74 283 450 cubic-edt 200M 2 188 0.02 222 87 1035 1095 101 84 85 dctcp 200M 2 188 0.03 111 77 912 939 111 76 325 dctcp-edt 200M 2 188 0.03 217 74 1416 1738 114 76 79 cubic-htb 200M 2 188 0.00 5015 8 14ms 15ms 180 48 50 cubic 1G 4 952 0.03 110 165 516 546 262 38 154 cubic-edt 1G 4 973 0.01 190 111 1034 1314 287 65 79 dctcp 1G 4 951 0.00 103 180 617 905 257 37 38 dctcp-edt 1G 4 967 0.00 163 151 732 1126 272 43 55 cubic-htb 1G 4 914 0.00 3249 13 7ms 8ms 300 29 34 cubic 5G 4 4236 0.00 134 305 490 624 1310 10 17 cubic-edt 5G 4 4865 0.00 156 306 425 759 1520 10 16 dctcp 5G 4 4936 0.00 128 485 221 409 1484 7 9 dctcp-edt 5G 4 4924 0.00 148 390 392 623 1508 11 26 v1 -> v2: Incorporated Andrii's suggestions v2 -> v3: Incorporated Yonghong's suggestions v3 -> v4: Removed credit update that is not needed Signed-off-by: Lawrence Brakmo <brakmo@fb.com> Acked-by: Yonghong Song <yhs@fb.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
169 lines
5.3 KiB
C
169 lines
5.3 KiB
C
// SPDX-License-Identifier: GPL-2.0
|
|
/* Copyright (c) 2019 Facebook
|
|
*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of version 2 of the GNU General Public
|
|
* License as published by the Free Software Foundation.
|
|
*
|
|
* Sample Host Bandwidth Manager (HBM) BPF program.
|
|
*
|
|
* A cgroup skb BPF egress program to limit cgroup output bandwidth.
|
|
* It uses a modified virtual token bucket queue to limit average
|
|
* egress bandwidth. The implementation uses credits instead of tokens.
|
|
* Negative credits imply that queueing would have happened (this is
|
|
* a virtual queue, so no queueing is done by it. However, queueing may
|
|
* occur at the actual qdisc (which is not used for rate limiting).
|
|
*
|
|
* This implementation uses 3 thresholds, one to start marking packets and
|
|
* the other two to drop packets:
|
|
* CREDIT
|
|
* - <--------------------------|------------------------> +
|
|
* | | | 0
|
|
* | Large pkt |
|
|
* | drop thresh |
|
|
* Small pkt drop Mark threshold
|
|
* thresh
|
|
*
|
|
* The effect of marking depends on the type of packet:
|
|
* a) If the packet is ECN enabled and it is a TCP packet, then the packet
|
|
* is ECN marked.
|
|
* b) If the packet is a TCP packet, then we probabilistically call tcp_cwr
|
|
* to reduce the congestion window. The current implementation uses a linear
|
|
* distribution (0% probability at marking threshold, 100% probability
|
|
* at drop threshold).
|
|
* c) If the packet is not a TCP packet, then it is dropped.
|
|
*
|
|
* If the credit is below the drop threshold, the packet is dropped. If it
|
|
* is a TCP packet, then it also calls tcp_cwr since packets dropped by
|
|
* by a cgroup skb BPF program do not automatically trigger a call to
|
|
* tcp_cwr in the current kernel code.
|
|
*
|
|
* This BPF program actually uses 2 drop thresholds, one threshold
|
|
* for larger packets (>= 120 bytes) and another for smaller packets. This
|
|
* protects smaller packets such as SYNs, ACKs, etc.
|
|
*
|
|
* The default bandwidth limit is set at 1Gbps but this can be changed by
|
|
* a user program through a shared BPF map. In addition, by default this BPF
|
|
* program does not limit connections using loopback. This behavior can be
|
|
* overwritten by the user program. There is also an option to calculate
|
|
* some statistics, such as percent of packets marked or dropped, which
|
|
* a user program, such as hbm, can access.
|
|
*/
|
|
|
|
#include "hbm_kern.h"
|
|
|
|
SEC("cgroup_skb/egress")
|
|
int _hbm_out_cg(struct __sk_buff *skb)
|
|
{
|
|
long long delta = 0, delta_send;
|
|
unsigned long long curtime, sendtime;
|
|
struct hbm_queue_stats *qsp = NULL;
|
|
unsigned int queue_index = 0;
|
|
bool congestion_flag = false;
|
|
bool ecn_ce_flag = false;
|
|
struct hbm_pkt_info pkti = {};
|
|
struct hbm_vqueue *qdp;
|
|
bool drop_flag = false;
|
|
bool cwr_flag = false;
|
|
int len = skb->len;
|
|
int rv = ALLOW_PKT;
|
|
|
|
qsp = bpf_map_lookup_elem(&queue_stats, &queue_index);
|
|
|
|
// Check if we should ignore loopback traffic
|
|
if (qsp != NULL && !qsp->loopback && (skb->ifindex == 1))
|
|
return ALLOW_PKT;
|
|
|
|
hbm_get_pkt_info(skb, &pkti);
|
|
|
|
// We may want to account for the length of headers in len
|
|
// calculation, like ETH header + overhead, specially if it
|
|
// is a gso packet. But I am not doing it right now.
|
|
|
|
qdp = bpf_get_local_storage(&queue_state, 0);
|
|
if (!qdp)
|
|
return ALLOW_PKT;
|
|
if (qdp->lasttime == 0)
|
|
hbm_init_edt_vqueue(qdp, 1024);
|
|
|
|
curtime = bpf_ktime_get_ns();
|
|
|
|
// Begin critical section
|
|
bpf_spin_lock(&qdp->lock);
|
|
delta = qdp->lasttime - curtime;
|
|
// bound bursts to 100us
|
|
if (delta < -BURST_SIZE_NS) {
|
|
// negative delta is a credit that allows bursts
|
|
qdp->lasttime = curtime - BURST_SIZE_NS;
|
|
delta = -BURST_SIZE_NS;
|
|
}
|
|
sendtime = qdp->lasttime;
|
|
delta_send = BYTES_TO_NS(len, qdp->rate);
|
|
__sync_add_and_fetch(&(qdp->lasttime), delta_send);
|
|
bpf_spin_unlock(&qdp->lock);
|
|
// End critical section
|
|
|
|
// Set EDT of packet
|
|
skb->tstamp = sendtime;
|
|
|
|
// Check if we should update rate
|
|
if (qsp != NULL && (qsp->rate * 128) != qdp->rate)
|
|
qdp->rate = qsp->rate * 128;
|
|
|
|
// Set flags (drop, congestion, cwr)
|
|
// last packet will be sent in the future, bound latency
|
|
if (delta > DROP_THRESH_NS || (delta > LARGE_PKT_DROP_THRESH_NS &&
|
|
len > LARGE_PKT_THRESH)) {
|
|
drop_flag = true;
|
|
if (pkti.is_tcp && pkti.ecn == 0)
|
|
cwr_flag = true;
|
|
} else if (delta > MARK_THRESH_NS) {
|
|
if (pkti.is_tcp)
|
|
congestion_flag = true;
|
|
else
|
|
drop_flag = true;
|
|
}
|
|
|
|
if (congestion_flag) {
|
|
if (bpf_skb_ecn_set_ce(skb)) {
|
|
ecn_ce_flag = true;
|
|
} else {
|
|
if (pkti.is_tcp) {
|
|
unsigned int rand = bpf_get_prandom_u32();
|
|
|
|
if (delta >= MARK_THRESH_NS +
|
|
(rand % MARK_REGION_SIZE_NS)) {
|
|
// Do congestion control
|
|
cwr_flag = true;
|
|
}
|
|
} else if (len > LARGE_PKT_THRESH) {
|
|
// Problem if too many small packets?
|
|
drop_flag = true;
|
|
congestion_flag = false;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (pkti.is_tcp && drop_flag && pkti.packets_out <= 1) {
|
|
drop_flag = false;
|
|
cwr_flag = true;
|
|
congestion_flag = false;
|
|
}
|
|
|
|
if (qsp != NULL && qsp->no_cn)
|
|
cwr_flag = false;
|
|
|
|
hbm_update_stats(qsp, len, curtime, congestion_flag, drop_flag,
|
|
cwr_flag, ecn_ce_flag, &pkti, (int) delta);
|
|
|
|
if (drop_flag) {
|
|
__sync_add_and_fetch(&(qdp->lasttime), -delta_send);
|
|
rv = DROP_PKT;
|
|
}
|
|
|
|
if (cwr_flag)
|
|
rv |= CWR;
|
|
return rv;
|
|
}
|
|
char _license[] SEC("license") = "GPL";
|