1#!/bin/bash
2# SPDX-License-Identifier: GPL-2.0
3#
4# This test injects a 10-MB burst of traffic with VLAN tag and 802.1p priority
5# of 1. This stream is consistently prioritized as priority 1, is put to PG
6# buffer 1, and scheduled at TC 1.
7#
8# - the stream first ingresses through $swp1, where it is forwarded to $swp3
9#
10# - then it ingresses through $swp4. Here it is put to a lossless buffer and put
11#   to a small pool ("PFC pool"). The traffic is forwarded to $swp2, which is
12#   shaped, and thus the PFC pool eventually fills, therefore the headroom
13#   fills, and $swp3 is paused.
14#
15# - since $swp3 now can't send traffic, the traffic ingressing $swp1 is kept at
16#   a pool ("overflow pool"). The overflow pool needs to be large enough to
17#   contain the whole burst.
18#
19# - eventually the PFC pool gets some traffic out, headroom therefore gets some
20#   traffic to the pool, and $swp3 is unpaused again. This way the traffic is
21#   gradually forwarded from the overflow pool, through the PFC pool, out of
22#   $swp2, and eventually to $h2.
23#
24# - if PFC works, all lossless flow packets that ingress through $swp1 should
25#   also be seen ingressing $h2. If it doesn't, there will be drops due to
26#   discrepancy between the speeds of $swp1 and $h2.
27#
28# - it should all play out relatively quickly, so that SLL and HLL will not
29#   cause drops.
30#
31# +-----------------------+
32# | H1                    |
33# |   + $h1.111           |
34# |   | 192.0.2.33/28     |
35# |   |                   |
36# |   + $h1               |
37# +---|-------------------+  +--------------------+
38#     |                      |                    |
39# +---|----------------------|--------------------|---------------------------+
40# |   + $swp1          $swp3 +                    + $swp4                     |
41# |   | iPOOL1        iPOOL0 |                    | iPOOL2                    |
42# |   | ePOOL4        ePOOL5 |                    | ePOOL4                    |
43# |   |        PFC:enabled=1 |                    | PFC:enabled=1             |
44# | +-|----------------------|-+                +-|------------------------+  |
45# | | + $swp1.111  $swp3.111 + |                | + $swp4.111              |  |
46# | |                          |                |                          |  |
47# | | BR1                      |                | BR2                      |  |
48# | |                          |                |                          |  |
49# | |                          |                |         + $swp2.111      |  |
50# | +--------------------------+                +---------|----------------+  |
51# |                                                       |                   |
52# | iPOOL0: 500KB dynamic                                 |                   |
53# | iPOOL1: 10MB static                                   |                   |
54# | iPOOL2: 1MB static                                    + $swp2             |
55# | ePOOL4: 500KB dynamic                                 | iPOOL0            |
56# | ePOOL5: 10MB static                                   | ePOOL6            |
57# | ePOOL6: "infinite" static                             | 200Mbps shaper    |
58# +-------------------------------------------------------|-------------------+
59#                                                         |
60#                                                     +---|-------------------+
61#                                                     |   + $h2            H2 |
62#                                                     |   |                   |
63#                                                     |   + $h2.111           |
64#                                                     |     192.0.2.34/28     |
65#                                                     +-----------------------+
66#
67# iPOOL0+ePOOL4 is a helper pool for control traffic etc.
68# iPOOL1+ePOOL5 are overflow pools.
69# iPOOL2+ePOOL6 are PFC pools.
70
71ALL_TESTS="
72	ping_ipv4
73	test_qos_pfc
74"
75
76lib_dir=$(dirname $0)/../../../net/forwarding
77
78NUM_NETIFS=6
79source $lib_dir/lib.sh
80source $lib_dir/devlink_lib.sh
81
82_1KB=1000
83_100KB=$((100 * _1KB))
84_500KB=$((500 * _1KB))
85_1MB=$((1000 * _1KB))
86_10MB=$((10 * _1MB))
87
88h1_create()
89{
90	simple_if_init $h1
91	mtu_set $h1 10000
92
93	vlan_create $h1 111 v$h1 192.0.2.33/28
94}
95
96h1_destroy()
97{
98	vlan_destroy $h1 111
99
100	mtu_restore $h1
101	simple_if_fini $h1
102}
103
104h2_create()
105{
106	simple_if_init $h2
107	mtu_set $h2 10000
108
109	vlan_create $h2 111 v$h2 192.0.2.34/28
110}
111
112h2_destroy()
113{
114	vlan_destroy $h2 111
115
116	mtu_restore $h2
117	simple_if_fini $h2
118}
119
120switch_create()
121{
122	local lanes_swp4
123	local pg1_size
124
125	# pools
126	# -----
127
128	devlink_pool_size_thtype_save 0
129	devlink_pool_size_thtype_save 4
130	devlink_pool_size_thtype_save 1
131	devlink_pool_size_thtype_save 5
132	devlink_pool_size_thtype_save 2
133	devlink_pool_size_thtype_save 6
134
135	devlink_port_pool_th_save $swp1 1
136	devlink_port_pool_th_save $swp2 6
137	devlink_port_pool_th_save $swp3 5
138	devlink_port_pool_th_save $swp4 2
139
140	devlink_tc_bind_pool_th_save $swp1 1 ingress
141	devlink_tc_bind_pool_th_save $swp2 1 egress
142	devlink_tc_bind_pool_th_save $swp3 1 egress
143	devlink_tc_bind_pool_th_save $swp4 1 ingress
144
145	# Control traffic pools. Just reduce the size. Keep them dynamic so that
146	# we don't need to change all the uninteresting quotas.
147	devlink_pool_size_thtype_set 0 dynamic $_500KB
148	devlink_pool_size_thtype_set 4 dynamic $_500KB
149
150	# Overflow pools.
151	devlink_pool_size_thtype_set 1 static $_10MB
152	devlink_pool_size_thtype_set 5 static $_10MB
153
154	# PFC pools. As per the writ, the size of egress PFC pool should be
155	# infinice, but actually it just needs to be large enough to not matter
156	# in practice, so reuse the 10MB limit.
157	devlink_pool_size_thtype_set 2 static $_1MB
158	devlink_pool_size_thtype_set 6 static $_10MB
159
160	# $swp1
161	# -----
162
163	ip link set dev $swp1 up
164	mtu_set $swp1 10000
165	vlan_create $swp1 111
166	ip link set dev $swp1.111 type vlan ingress-qos-map 0:0 1:1
167
168	devlink_port_pool_th_set $swp1 1 $_10MB
169	devlink_tc_bind_pool_th_set $swp1 1 ingress 1 $_10MB
170
171	# Configure qdisc so that we can configure PG and therefore pool
172	# assignment.
173	tc qdisc replace dev $swp1 root handle 1: \
174	   ets bands 8 strict 8 priomap 7 6
175	dcb buffer set dev $swp1 prio-buffer all:0 1:1
176
177	# $swp2
178	# -----
179
180	ip link set dev $swp2 up
181	mtu_set $swp2 10000
182	vlan_create $swp2 111
183	ip link set dev $swp2.111 type vlan egress-qos-map 0:0 1:1
184
185	devlink_port_pool_th_set $swp2 6 $_10MB
186	devlink_tc_bind_pool_th_set $swp2 1 egress 6 $_10MB
187
188	# prio 0->TC0 (band 7), 1->TC1 (band 6). TC1 is shaped.
189	tc qdisc replace dev $swp2 root handle 1: \
190	   ets bands 8 strict 8 priomap 7 6
191	tc qdisc replace dev $swp2 parent 1:7 handle 17: \
192	   tbf rate 200Mbit burst 131072 limit 1M
193
194	# $swp3
195	# -----
196
197	ip link set dev $swp3 up
198	mtu_set $swp3 10000
199	vlan_create $swp3 111
200	ip link set dev $swp3.111 type vlan egress-qos-map 0:0 1:1
201
202	devlink_port_pool_th_set $swp3 5 $_10MB
203	devlink_tc_bind_pool_th_set $swp3 1 egress 5 $_10MB
204
205	# prio 0->TC0 (band 7), 1->TC1 (band 6)
206	tc qdisc replace dev $swp3 root handle 1: \
207	   ets bands 8 strict 8 priomap 7 6
208
209	# Need to enable PFC so that PAUSE takes effect. Therefore need to put
210	# the lossless prio into a buffer of its own. Don't bother with buffer
211	# sizes though, there is not going to be any pressure in the "backward"
212	# direction.
213	dcb buffer set dev $swp3 prio-buffer all:0 1:1
214	dcb pfc set dev $swp3 prio-pfc all:off 1:on
215
216	# $swp4
217	# -----
218
219	ip link set dev $swp4 up
220	mtu_set $swp4 10000
221	vlan_create $swp4 111
222	ip link set dev $swp4.111 type vlan ingress-qos-map 0:0 1:1
223
224	devlink_port_pool_th_set $swp4 2 $_1MB
225	devlink_tc_bind_pool_th_set $swp4 1 ingress 2 $_1MB
226
227	# Configure qdisc so that we can hand-tune headroom.
228	tc qdisc replace dev $swp4 root handle 1: \
229	   ets bands 8 strict 8 priomap 7 6
230	dcb buffer set dev $swp4 prio-buffer all:0 1:1
231	dcb pfc set dev $swp4 prio-pfc all:off 1:on
232	# PG0 will get autoconfigured to Xoff, give PG1 arbitrarily 100K, which
233	# is (-2*MTU) about 80K of delay provision.
234	pg1_size=$_100KB
235
236	setup_wait_dev_with_timeout $swp4
237
238	lanes_swp4=$(ethtool $swp4 | grep 'Lanes:')
239	lanes_swp4=${lanes_swp4#*"Lanes: "}
240
241	# 8-lane ports use two buffers among which the configured buffer
242	# is split, so double the size to get twice (20K + 80K).
243	if [[ $lanes_swp4 -eq 8 ]]; then
244		pg1_size=$((pg1_size * 2))
245	fi
246
247	dcb buffer set dev $swp4 buffer-size all:0 1:$pg1_size
248
249	# bridges
250	# -------
251
252	ip link add name br1 type bridge vlan_filtering 0
253	ip link set dev $swp1.111 master br1
254	ip link set dev $swp3.111 master br1
255	ip link set dev br1 up
256
257	ip link add name br2 type bridge vlan_filtering 0
258	ip link set dev $swp2.111 master br2
259	ip link set dev $swp4.111 master br2
260	ip link set dev br2 up
261}
262
263switch_destroy()
264{
265	# Do this first so that we can reset the limits to values that are only
266	# valid for the original static / dynamic setting.
267	devlink_pool_size_thtype_restore 6
268	devlink_pool_size_thtype_restore 5
269	devlink_pool_size_thtype_restore 4
270	devlink_pool_size_thtype_restore 2
271	devlink_pool_size_thtype_restore 1
272	devlink_pool_size_thtype_restore 0
273
274	# bridges
275	# -------
276
277	ip link set dev br2 down
278	ip link set dev $swp4.111 nomaster
279	ip link set dev $swp2.111 nomaster
280	ip link del dev br2
281
282	ip link set dev br1 down
283	ip link set dev $swp3.111 nomaster
284	ip link set dev $swp1.111 nomaster
285	ip link del dev br1
286
287	# $swp4
288	# -----
289
290	dcb buffer set dev $swp4 buffer-size all:0
291	dcb pfc set dev $swp4 prio-pfc all:off
292	dcb buffer set dev $swp4 prio-buffer all:0
293	tc qdisc del dev $swp4 root
294
295	devlink_tc_bind_pool_th_restore $swp4 1 ingress
296	devlink_port_pool_th_restore $swp4 2
297
298	vlan_destroy $swp4 111
299	mtu_restore $swp4
300	ip link set dev $swp4 down
301
302	# $swp3
303	# -----
304
305	dcb pfc set dev $swp3 prio-pfc all:off
306	dcb buffer set dev $swp3 prio-buffer all:0
307	tc qdisc del dev $swp3 root
308
309	devlink_tc_bind_pool_th_restore $swp3 1 egress
310	devlink_port_pool_th_restore $swp3 5
311
312	vlan_destroy $swp3 111
313	mtu_restore $swp3
314	ip link set dev $swp3 down
315
316	# $swp2
317	# -----
318
319	tc qdisc del dev $swp2 parent 1:7
320	tc qdisc del dev $swp2 root
321
322	devlink_tc_bind_pool_th_restore $swp2 1 egress
323	devlink_port_pool_th_restore $swp2 6
324
325	vlan_destroy $swp2 111
326	mtu_restore $swp2
327	ip link set dev $swp2 down
328
329	# $swp1
330	# -----
331
332	dcb buffer set dev $swp1 prio-buffer all:0
333	tc qdisc del dev $swp1 root
334
335	devlink_tc_bind_pool_th_restore $swp1 1 ingress
336	devlink_port_pool_th_restore $swp1 1
337
338	vlan_destroy $swp1 111
339	mtu_restore $swp1
340	ip link set dev $swp1 down
341}
342
343setup_prepare()
344{
345	h1=${NETIFS[p1]}
346	swp1=${NETIFS[p2]}
347
348	swp2=${NETIFS[p3]}
349	h2=${NETIFS[p4]}
350
351	swp3=${NETIFS[p5]}
352	swp4=${NETIFS[p6]}
353
354	h2mac=$(mac_get $h2)
355
356	vrf_prepare
357
358	h1_create
359	h2_create
360	switch_create
361}
362
363cleanup()
364{
365	pre_cleanup
366
367	switch_destroy
368	h2_destroy
369	h1_destroy
370
371	vrf_cleanup
372}
373
374ping_ipv4()
375{
376	ping_test $h1 192.0.2.34
377}
378
379test_qos_pfc()
380{
381	RET=0
382
383	# 10M pool, each packet is 8K of payload + headers
384	local pkts=$((_10MB / 8050))
385	local size=$((pkts * 8050))
386	local in0=$(ethtool_stats_get $swp1 rx_octets_prio_1)
387	local out0=$(ethtool_stats_get $swp2 tx_octets_prio_1)
388
389	$MZ $h1 -p 8000 -Q 1:111 -A 192.0.2.33 -B 192.0.2.34 \
390		-a own -b $h2mac -c $pkts -t udp -q
391	sleep 2
392
393	local in1=$(ethtool_stats_get $swp1 rx_octets_prio_1)
394	local out1=$(ethtool_stats_get $swp2 tx_octets_prio_1)
395
396	local din=$((in1 - in0))
397	local dout=$((out1 - out0))
398
399	local pct_in=$((din * 100 / size))
400
401	((pct_in > 95 && pct_in < 105))
402	check_err $? "Relative ingress out of expected bounds, $pct_in% should be 100%"
403
404	((dout == din))
405	check_err $? "$((din - dout)) bytes out of $din ingressed got lost"
406
407	log_test "PFC"
408}
409
410bail_on_lldpad "configure DCB" "configure Qdiscs"
411
412trap cleanup EXIT
413setup_prepare
414setup_wait
415tests_run
416
417exit $EXIT_STATUS
418