summaryrefslogtreecommitdiffstats
path: root/test/test_qos.py
blob: 02941a163b4050d1eb30831adff8f81840ca014d (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
#!/usr/bin/env python3

import unittest

from framework import VppTestCase, VppTestRunner
from vpp_sub_interface import VppDot1QSubint
from vpp_ip import DpoProto
from vpp_ip_route import VppIpRoute, VppRoutePath, VppMplsRoute, \
    VppMplsLabel, VppMplsTable, FibPathProto

import scapy.compat
from scapy.packet import Raw
from scapy.layers.l2 import Ether, Dot1Q
from scapy.layers.inet import IP, UDP
from scapy.layers.inet6 import IPv6
from scapy.contrib.mpls import MPLS
from vpp_papi import VppEnum
from vpp_qos import VppQosRecord, VppQosEgressMap, VppQosMark, VppQosStore

NUM_PKTS = 67


class TestQOS(VppTestCase):
    """ QOS Test Case """

    # Note: Since the enums aren't created dynamically until after
    #       the papi client attaches to VPP, we put it in a property to
    #       ensure it is the value at runtime, not at module load time.
    @property
    def QOS_SOURCE(self):
        return VppEnum.vl_api_qos_source_t

    @classmethod
    def setUpClass(cls):
        super(TestQOS, cls).setUpClass()

    @classmethod
    def tearDownClass(cls):
        super(TestQOS, cls).tearDownClass()

    def setUp(self):
        super(TestQOS, self).setUp()

        self.create_pg_interfaces(range(5))

        tbl = VppMplsTable(self, 0)
        tbl.add_vpp_config()

        for i in self.pg_interfaces:
            i.admin_up()
            i.config_ip4()
            i.resolve_arp()
            i.config_ip6()
            i.resolve_ndp()
            i.enable_mpls()

    def tearDown(self):
        for i in self.pg_interfaces:
            i.unconfig_ip4()
            i.unconfig_ip6()
            i.disable_mpls()

        super(TestQOS, self).tearDown()

    def test_qos_ip(self):
        """ QoS Mark/Record/Store IP """

        #
        # for table 1 map the n=0xff possible values of input QoS mark,
        # n to 1-n
        #
        output = [scapy.compat.chb(0)] * 256
        for i in range(0, 255):
            output[i] = scapy.compat.chb(255 - i)
        os = b''.join(output)
        rows = [{'outputs': os},
                {'outputs': os},
                {'outputs': os},
                {'outputs': os}]

        qem1 = VppQosEgressMap(self, 1, rows).add_vpp_config()

        #
        # For table 2 (and up) use the value n for everything
        #
        output = [scapy.compat.chb(2)] * 256
        os = b''.join(output)
        rows = [{'outputs': os},
                {'outputs': os},
                {'outputs': os},
                {'outputs': os}]

        qem2 = VppQosEgressMap(self, 2, rows).add_vpp_config()

        output = [scapy.compat.chb(3)] * 256
        os = b''.join(output)
        rows = [{'outputs': os},
                {'outputs': os},
                {'outputs': os},
                {'outputs': os}]

        qem3 = VppQosEgressMap(self, 3, rows).add_vpp_config()

        output = [scapy.compat.chb(4)] * 256
        os = b''.join(output)
        rows = [{'outputs': os},
                {'outputs': os},
                {'outputs': os},
                {'outputs': os}]

        qem4 = VppQosEgressMap(self, 4, rows).add_vpp_config()
        qem5 = VppQosEgressMap(self, 5, rows).add_vpp_config()
        qem6 = VppQosEgressMap(self, 6, rows).add_vpp_config()
        qem7 = VppQosEgressMap(self, 7, rows).add_vpp_config()

        self.assertTrue(qem7.query_vpp_config())
        self.logger.info(self.vapi.cli("sh qos eg map"))

        #
        # Bind interface pgN to table n
        #
        qm1 = VppQosMark(self, self.pg1, qem1,
                         self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()
        qm2 = VppQosMark(self, self.pg2, qem2,
                         self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()
        qm3 = VppQosMark(self, self.pg3, qem3,
                         self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()
        qm4 = VppQosMark(self, self.pg4, qem4,
                         self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()
        self.assertTrue(qm3.query_vpp_config())

        self.logger.info(self.vapi.cli("sh qos mark"))

        #
        # packets ingress on Pg0
        #
        p_v4 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                IP(src=self.pg0.remote_ip4, dst=self.pg1.remote_ip4, tos=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))
        p_v6 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                IPv6(src=self.pg0.remote_ip6, dst=self.pg1.remote_ip6,
                     tc=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        #
        # Since we have not yet enabled the recording of the input QoS
        # from the input iP header, the egress packet's ToS will be unchanged
        #
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 1)
        rx = self.send_and_expect(self.pg0, p_v6 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IPv6].tc, 1)

        #
        # Enable QoS recording on IP input for pg0
        #
        qr1 = VppQosRecord(self, self.pg0,
                           self.QOS_SOURCE.QOS_API_SOURCE_IP)
        qr1.add_vpp_config()
        self.logger.info(self.vapi.cli("sh qos record"))

        #
        # send the same packets, this time expect the input TOS of 1
        # to be mapped to pg1's egress value of 254
        #
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 254)
        rx = self.send_and_expect(self.pg0, p_v6 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IPv6].tc, 254)

        #
        # different input ToS to test the mapping
        #
        p_v4[IP].tos = 127
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 128)
        p_v6[IPv6].tc = 127
        rx = self.send_and_expect(self.pg0, p_v6 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IPv6].tc, 128)

        p_v4[IP].tos = 254
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 1)
        p_v6[IPv6].tc = 254
        rx = self.send_and_expect(self.pg0, p_v6 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IPv6].tc, 1)

        #
        # send packets out the other interfaces to test the maps are
        # correctly applied
        #
        p_v4[IP].dst = self.pg2.remote_ip4
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg2)
        for p in rx:
            self.assertEqual(p[IP].tos, 2)

        p_v4[IP].dst = self.pg3.remote_ip4
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg3)
        for p in rx:
            self.assertEqual(p[IP].tos, 3)

        p_v6[IPv6].dst = self.pg3.remote_ip6
        rx = self.send_and_expect(self.pg0, p_v6 * NUM_PKTS, self.pg3)
        for p in rx:
            self.assertEqual(p[IPv6].tc, 3)

        #
        # remove the map on pg2 and pg3, now expect an unchanged IP tos
        #
        qm2.remove_vpp_config()
        qm3.remove_vpp_config()
        self.logger.info(self.vapi.cli("sh qos mark"))

        self.assertFalse(qm3.query_vpp_config())
        self.logger.info(self.vapi.cli("sh int feat pg2"))

        p_v4[IP].dst = self.pg2.remote_ip4
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg2)
        for p in rx:
            self.assertEqual(p[IP].tos, 254)

        p_v4[IP].dst = self.pg3.remote_ip4
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg3)
        for p in rx:
            self.assertEqual(p[IP].tos, 254)

        #
        # still mapping out of pg1
        #
        p_v4[IP].dst = self.pg1.remote_ip4
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 1)

        #
        # disable the input recording on pg0
        #
        self.assertTrue(qr1.query_vpp_config())
        qr1.remove_vpp_config()

        #
        # back to an unchanged TOS value
        #
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 254)

        #
        # enable QoS stroe instead of record
        #
        qst1 = VppQosStore(self, self.pg0,
                           self.QOS_SOURCE.QOS_API_SOURCE_IP,
                           5).add_vpp_config()
        self.logger.info(self.vapi.cli("sh qos store"))

        p_v4[IP].dst = self.pg1.remote_ip4
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 250)

        #
        # disable the input storing on pg0
        #
        self.assertTrue(qst1.query_vpp_config())
        qst1.remove_vpp_config()

        #
        # back to an unchanged TOS value
        #
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 254)

        #
        # disable the egress map on pg1 and pg4
        #
        qm1.remove_vpp_config()
        qm4.remove_vpp_config()

        #
        # unchanged Tos on pg1
        #
        rx = self.send_and_expect(self.pg0, p_v4 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[IP].tos, 254)

    def test_qos_mpls(self):
        """ QoS Mark/Record MPLS """

        #
        # 255 QoS for all input values
        #
        from_ext = 7
        from_ip = 6
        from_mpls = 5
        from_vlan = 4
        output = [scapy.compat.chb(from_ext)] * 256
        os1 = b''.join(output)
        output = [scapy.compat.chb(from_vlan)] * 256
        os2 = b''.join(output)
        output = [scapy.compat.chb(from_mpls)] * 256
        os3 = b''.join(output)
        output = [scapy.compat.chb(from_ip)] * 256
        os4 = b''.join(output)
        rows = [{'outputs': os1},
                {'outputs': os2},
                {'outputs': os3},
                {'outputs': os4}]

        qem1 = VppQosEgressMap(self, 1, rows).add_vpp_config()

        #
        # a route with 1 MPLS label
        #
        route_10_0_0_1 = VppIpRoute(self, "10.0.0.1", 32,
                                    [VppRoutePath(self.pg1.remote_ip4,
                                                  self.pg1.sw_if_index,
                                                  labels=[32])])
        route_10_0_0_1.add_vpp_config()

        #
        # a route with 3 MPLS labels
        #
        route_10_0_0_3 = VppIpRoute(self, "10.0.0.3", 32,
                                    [VppRoutePath(self.pg1.remote_ip4,
                                                  self.pg1.sw_if_index,
                                                  labels=[63, 33, 34])])
        route_10_0_0_3.add_vpp_config()

        #
        # enable IP QoS recording on the input Pg0 and MPLS egress marking
        # on Pg1
        #
        qr1 = VppQosRecord(self, self.pg0,
                           self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()
        qm1 = VppQosMark(self, self.pg1, qem1,
                         self.QOS_SOURCE.QOS_API_SOURCE_MPLS).add_vpp_config()

        #
        # packet that will get one label added and 3 labels added resp.
        #
        p_1 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
               IP(src=self.pg0.remote_ip4, dst="10.0.0.1", tos=1) /
               UDP(sport=1234, dport=1234) /
               Raw(scapy.compat.chb(100) * NUM_PKTS))
        p_3 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
               IP(src=self.pg0.remote_ip4, dst="10.0.0.3", tos=1) /
               UDP(sport=1234, dport=1234) /
               Raw(scapy.compat.chb(100) * NUM_PKTS))

        rx = self.send_and_expect(self.pg0, p_1 * NUM_PKTS, self.pg1)

        #
        # only 3 bits of ToS value in MPLS make sure tos is correct
        # and the label and EOS bit have not been corrupted
        #
        for p in rx:
            self.assertEqual(p[MPLS].cos, from_ip)
            self.assertEqual(p[MPLS].label, 32)
            self.assertEqual(p[MPLS].s, 1)
        rx = self.send_and_expect(self.pg0, p_3 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[MPLS].cos, from_ip)
            self.assertEqual(p[MPLS].label, 63)
            self.assertEqual(p[MPLS].s, 0)
            h = p[MPLS].payload
            self.assertEqual(h[MPLS].cos, from_ip)
            self.assertEqual(h[MPLS].label, 33)
            self.assertEqual(h[MPLS].s, 0)
            h = h[MPLS].payload
            self.assertEqual(h[MPLS].cos, from_ip)
            self.assertEqual(h[MPLS].label, 34)
            self.assertEqual(h[MPLS].s, 1)

        #
        # enable MPLS QoS recording on the input Pg0 and IP egress marking
        # on Pg1
        #
        qr2 = VppQosRecord(
            self, self.pg0,
            self.QOS_SOURCE.QOS_API_SOURCE_MPLS).add_vpp_config()
        qm2 = VppQosMark(
            self, self.pg1, qem1,
            self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()

        #
        # MPLS x-connect - COS according to pg1 map
        #
        route_32_eos = VppMplsRoute(self, 32, 1,
                                    [VppRoutePath(self.pg1.remote_ip4,
                                                  self.pg1.sw_if_index,
                                                  labels=[VppMplsLabel(33)])])
        route_32_eos.add_vpp_config()

        p_m1 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                MPLS(label=32, cos=3, ttl=2) /
                IP(src=self.pg0.remote_ip4, dst="10.0.0.1", tos=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        rx = self.send_and_expect(self.pg0, p_m1 * NUM_PKTS, self.pg1)
        for p in rx:
            self.assertEqual(p[MPLS].cos, from_mpls)
            self.assertEqual(p[MPLS].label, 33)
            self.assertEqual(p[MPLS].s, 1)

        #
        # MPLS deag - COS is copied from MPLS to IP
        #
        route_33_eos = VppMplsRoute(self, 33, 1,
                                    [VppRoutePath("0.0.0.0",
                                                  0xffffffff,
                                                  nh_table_id=0)])
        route_33_eos.add_vpp_config()

        route_10_0_0_4 = VppIpRoute(self, "10.0.0.4", 32,
                                    [VppRoutePath(self.pg1.remote_ip4,
                                                  self.pg1.sw_if_index)])
        route_10_0_0_4.add_vpp_config()

        p_m2 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                MPLS(label=33, ttl=2, cos=3) /
                IP(src=self.pg0.remote_ip4, dst="10.0.0.4", tos=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        rx = self.send_and_expect(self.pg0, p_m2 * NUM_PKTS, self.pg1)

        for p in rx:
            self.assertEqual(p[IP].tos, from_mpls)

    def test_qos_vlan(self):
        """QoS mark/record VLAN """

        #
        # QoS for all input values
        #
        output = [scapy.compat.chb(0)] * 256
        for i in range(0, 255):
            output[i] = scapy.compat.chb(255 - i)
        os = b''.join(output)
        rows = [{'outputs': os},
                {'outputs': os},
                {'outputs': os},
                {'outputs': os}]

        qem1 = VppQosEgressMap(self, 1, rows).add_vpp_config()

        sub_if = VppDot1QSubint(self, self.pg0, 11)

        sub_if.admin_up()
        sub_if.config_ip4()
        sub_if.resolve_arp()
        sub_if.config_ip6()
        sub_if.resolve_ndp()

        #
        # enable VLAN QoS recording/marking on the input Pg0 subinterface and
        #
        qr_v = VppQosRecord(
            self, sub_if,
            self.QOS_SOURCE.QOS_API_SOURCE_VLAN).add_vpp_config()
        qm_v = VppQosMark(
            self, sub_if, qem1,
            self.QOS_SOURCE.QOS_API_SOURCE_VLAN).add_vpp_config()

        #
        # IP marking/recording on pg1
        #
        qr_ip = VppQosRecord(
            self, self.pg1,
            self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()
        qm_ip = VppQosMark(
            self, self.pg1, qem1,
            self.QOS_SOURCE.QOS_API_SOURCE_IP).add_vpp_config()

        #
        # a routes to/from sub-interface
        #
        route_10_0_0_1 = VppIpRoute(self, "10.0.0.1", 32,
                                    [VppRoutePath(sub_if.remote_ip4,
                                                  sub_if.sw_if_index)])
        route_10_0_0_1.add_vpp_config()
        route_10_0_0_2 = VppIpRoute(self, "10.0.0.2", 32,
                                    [VppRoutePath(self.pg1.remote_ip4,
                                                  self.pg1.sw_if_index)])
        route_10_0_0_2.add_vpp_config()
        route_2001_1 = VppIpRoute(self, "2001::1", 128,
                                  [VppRoutePath(sub_if.remote_ip6,
                                                sub_if.sw_if_index)])
        route_2001_1.add_vpp_config()
        route_2001_2 = VppIpRoute(self, "2001::2", 128,
                                  [VppRoutePath(self.pg1.remote_ip6,
                                                self.pg1.sw_if_index)])
        route_2001_2.add_vpp_config()

        p_v1 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                Dot1Q(vlan=11, prio=1) /
                IP(src="1.1.1.1", dst="10.0.0.2", tos=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        p_v2 = (Ether(src=self.pg1.remote_mac, dst=self.pg1.local_mac) /
                IP(src="1.1.1.1", dst="10.0.0.1", tos=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        p_v3 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                Dot1Q(vlan=11, prio=1, id=1) /
                IP(src="1.1.1.1", dst="10.0.0.2", tos=2) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        rx = self.send_and_expect(self.pg1, p_v2 * NUM_PKTS, self.pg0)

        for p in rx:
            self.assertEqual(p[Dot1Q].prio, 7)
            self.assertEqual(p[Dot1Q].id, 0)

        rx = self.send_and_expect(self.pg0, p_v3 * NUM_PKTS, self.pg1)

        for p in rx:
            self.assertEqual(p[IP].tos, 252)

        rx = self.send_and_expect(self.pg0, p_v1 * NUM_PKTS, self.pg1)

        for p in rx:
            self.assertEqual(p[IP].tos, 253)

        p_v1 = (Ether(src=self.pg0.remote_mac, dst=self.pg0.local_mac) /
                Dot1Q(vlan=11, prio=2) /
                IPv6(src="2001::1", dst="2001::2", tc=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        p_v2 = (Ether(src=self.pg1.remote_mac, dst=self.pg1.local_mac) /
                IPv6(src="3001::1", dst="2001::1", tc=1) /
                UDP(sport=1234, dport=1234) /
                Raw(scapy.compat.chb(100) * NUM_PKTS))

        rx = self.send_and_expect(self.pg1, p_v2 * NUM_PKTS, self.pg0)

        for p in rx:
            self.assertEqual(p[Dot1Q].prio, 7)
            self.assertEqual(p[Dot1Q].id, 0)

        rx = self.send_and_expect(self.pg0, p_v1 * NUM_PKTS, self.pg1)

        for p in rx:
            self.assertEqual(p[IPv6].tc, 251)

        #
        # cleanup
        #
        sub_if.unconfig_ip4()
        sub_if.unconfig_ip6()


if __name__ == '__main__':
    unittest.main(testRunner=VppTestRunner)
int i; vlib_set_queue_signal_callback (vm, memclnt_queue_callback); if ((rv = memory_api_init (am->region_name)) < 0) { clib_warning ("memory_api_init returned %d, wait for godot...", rv); vlib_process_suspend (vm, 1e70); } shm = am->shmem_hdr; ASSERT (shm); q = shm->vl_input_queue; ASSERT (q); e = vlib_call_init_exit_functions (vm, vm->api_init_function_registrations, 1 /* call_once */ ); if (e) clib_error_report (e); sleep_time = 20.0; dead_client_scan_time = vlib_time_now (vm) + 20.0; /* * Send plugin message range messages for each plugin we loaded */ for (i = 0; i < vec_len (am->msg_ranges); i++) { vl_api_msg_range_t *rp = am->msg_ranges + i; send_one_plugin_msg_ids_msg (rp->name, rp->first_msg_id, rp->last_msg_id); } /* $$$ pay attention to frame size, control CPU usage */ while (1) { uword event_type __attribute__ ((unused)); i8 *headp; int need_broadcast; /* * There's a reason for checking the queue before * sleeping. If the vlib application crashes, it's entirely * possible for a client to enqueue a connect request * during the process restart interval. * * Unless some force of physics causes the new incarnation * of the application to process the request, the client will * sit and wait for Godot... */ vector_rate = vlib_last_vector_length_per_node (vm); start_time = vlib_time_now (vm); while (1) { pthread_mutex_lock (&q->mutex); if (q->cursize == 0) { vm->api_queue_nonempty = 0; pthread_mutex_unlock (&q->mutex); if (TRACE_VLIB_MEMORY_QUEUE) { /* *INDENT-OFF* */ ELOG_TYPE_DECLARE (e) = { .format = "q-underflow: len %d", .format_args = "i4", }; /* *INDENT-ON* */ struct { u32 len; } *ed; ed = ELOG_DATA (&vm->elog_main, e); ed->len = 0; } sleep_time = 20.0; break; } headp = (i8 *) (q->data + sizeof (uword) * q->head); clib_memcpy (&mp, headp, sizeof (uword)); q->head++; need_broadcast = (q->cursize == q->maxsize / 2); q->cursize--; if (PREDICT_FALSE (q->head == q->maxsize)) q->head = 0; pthread_mutex_unlock (&q->mutex); if (need_broadcast) (void) pthread_cond_broadcast (&q->condvar); vl_msg_api_handler_with_vm_node (am, (void *) mp, vm, node); /* Allow no more than 10us without a pause */ if (vlib_time_now (vm) > start_time + 10e-6) { int index = SLEEP_400_US; if (vector_rate > 40.0) sleep_time = 400e-6; else if (vector_rate > 20.0) { index = SLEEP_200_US; sleep_time = 200e-6; } else if (vector_rate >= 1.0) { index = SLEEP_100_US; sleep_time = 100e-6; } else { index = SLEEP_10_US; sleep_time = 10e-6; } vector_rate_histogram[index] += 1; break; } } event_type = vlib_process_wait_for_event_or_clock (vm, sleep_time); vm->queue_signal_pending = 0; vlib_process_get_events (vm, 0 /* event_data */ ); if (vlib_time_now (vm) > dead_client_scan_time) { vl_api_registration_t **regpp; vl_api_registration_t *regp; unix_shared_memory_queue_t *q; static u32 *dead_indices; static u32 *confused_indices; vec_reset_length (dead_indices); vec_reset_length (confused_indices); /* *INDENT-OFF* */ pool_foreach (regpp, am->vl_clients, ({ regp = *regpp; if (regp) { q = regp->vl_input_queue; if (kill (q->consumer_pid, 0) < 0) { vec_add1(dead_indices, regpp - am->vl_clients); } } else { clib_warning ("NULL client registration index %d", regpp - am->vl_clients); vec_add1 (confused_indices, regpp - am->vl_clients); } })); /* *INDENT-ON* */ /* This should "never happen," but if it does, fix it... */ if (PREDICT_FALSE (vec_len (confused_indices) > 0)) { int i; for (i = 0; i < vec_len (confused_indices); i++) { pool_put_index (am->vl_clients, confused_indices[i]); } } if (PREDICT_FALSE (vec_len (dead_indices) > 0)) { int i; svm_region_t *svm; void *oldheap; /* Allow the application to clean up its registrations */ for (i = 0; i < vec_len (dead_indices); i++) { regpp = pool_elt_at_index (am->vl_clients, dead_indices[i]); if (regpp) { u32 handle; handle = vl_msg_api_handle_from_index_and_epoch (dead_indices[i], shm->application_restarts); (void) call_reaper_functions (handle); } } svm = am->vlib_rp; pthread_mutex_lock (&svm->mutex); oldheap = svm_push_data_heap (svm); for (i = 0; i < vec_len (dead_indices); i++) { regpp = pool_elt_at_index (am->vl_clients, dead_indices[i]); if (regpp) { /* Poison the old registration */ memset (*regpp, 0xF3, sizeof (**regpp)); clib_mem_free (*regpp); /* no dangling references, please */ *regpp = 0; } else { svm_pop_heap (oldheap); clib_warning ("Duplicate free, client index %d", regpp - am->vl_clients); oldheap = svm_push_data_heap (svm); } } svm_client_scan_this_region_nolock (am->vlib_rp); pthread_mutex_unlock (&svm->mutex); svm_pop_heap (oldheap); for (i = 0; i < vec_len (dead_indices); i++) pool_put_index (am->vl_clients, dead_indices[i]); } dead_client_scan_time = vlib_time_now (vm) + 20.0; } if (TRACE_VLIB_MEMORY_QUEUE) { /* *INDENT-OFF* */ ELOG_TYPE_DECLARE (e) = { .format = "q-awake: len %d", .format_args = "i4", }; /* *INDENT-ON* */ struct { u32 len; } *ed; ed = ELOG_DATA (&vm->elog_main, e); ed->len = q->cursize; } } return 0; } static clib_error_t * vl_api_show_histogram_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { u64 total_counts = 0; int i; for (i = 0; i < SLEEP_N_BUCKETS; i++) { total_counts += vector_rate_histogram[i]; } if (total_counts == 0) { vlib_cli_output (vm, "No control-plane activity."); return 0; } #define _(n) \ do { \ f64 percent; \ percent = ((f64) vector_rate_histogram[SLEEP_##n##_US]) \ / (f64) total_counts; \ percent *= 100.0; \ vlib_cli_output (vm, "Sleep %3d us: %llu, %.2f%%",n, \ vector_rate_histogram[SLEEP_##n##_US], \ percent); \ } while (0); foreach_histogram_bucket; #undef _ return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_histogram_command, static) = { .path = "show api histogram", .short_help = "show api histogram", .function = vl_api_show_histogram_command, }; /* *INDENT-ON* */ static clib_error_t * vl_api_clear_histogram_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { int i; for (i = 0; i < SLEEP_N_BUCKETS; i++) vector_rate_histogram[i] = 0; return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_clear_api_histogram_command, static) = { .path = "clear api histogram", .short_help = "clear api histogram", .function = vl_api_clear_histogram_command, }; /* *INDENT-ON* */ /* *INDENT-OFF* */ VLIB_REGISTER_NODE (memclnt_node,static) = { .function = memclnt_process, .type = VLIB_NODE_TYPE_PROCESS, .name = "api-rx-from-ring", .state = VLIB_NODE_STATE_DISABLED, }; /* *INDENT-ON* */ static void memclnt_queue_callback (vlib_main_t * vm) { static volatile int *cursizep; if (PREDICT_FALSE (cursizep == 0)) { api_main_t *am = &api_main; vl_shmem_hdr_t *shmem_hdr = am->shmem_hdr; unix_shared_memory_queue_t *q; if (shmem_hdr == 0) return; q = shmem_hdr->vl_input_queue; if (q == 0) return; cursizep = &q->cursize; } if (*cursizep >= 1) { vm->queue_signal_pending = 1; vm->api_queue_nonempty = 1; vlib_process_signal_event (vm, memclnt_node.index, /* event_type */ 0, /* event_data */ 0); } } void vl_enable_disable_memory_api (vlib_main_t * vm, int enable) { vlib_node_set_state (vm, memclnt_node.index, (enable ? VLIB_NODE_STATE_POLLING : VLIB_NODE_STATE_DISABLED)); } static uword api_rx_from_node (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame) { uword n_packets = frame->n_vectors; uword n_left_from; u32 *from; static u8 *long_msg; vec_validate (long_msg, 4095); n_left_from = frame->n_vectors; from = vlib_frame_args (frame); while (n_left_from > 0) { u32 bi0; vlib_buffer_t *b0; void *msg; uword msg_len; bi0 = from[0]; b0 = vlib_get_buffer (vm, bi0); from += 1; n_left_from -= 1; msg = b0->data + b0->current_data; msg_len = b0->current_length; if (b0->flags & VLIB_BUFFER_NEXT_PRESENT) { ASSERT (long_msg != 0); _vec_len (long_msg) = 0; vec_add (long_msg, msg, msg_len); while (b0->flags & VLIB_BUFFER_NEXT_PRESENT) { b0 = vlib_get_buffer (vm, b0->next_buffer); msg = b0->data + b0->current_data; msg_len = b0->current_length; vec_add (long_msg, msg, msg_len); } msg = long_msg; } vl_msg_api_handler_no_trace_no_free (msg); } /* Free what we've been given. */ vlib_buffer_free (vm, vlib_frame_args (frame), n_packets); return n_packets; } /* *INDENT-OFF* */ VLIB_REGISTER_NODE (api_rx_from_node_node,static) = { .function = api_rx_from_node, .type = VLIB_NODE_TYPE_INTERNAL, .vector_size = 4, .name = "api-rx-from-node", }; /* *INDENT-ON* */ static clib_error_t * setup_memclnt_exit (vlib_main_t * vm) { atexit (vl_unmap_shmem); return 0; } VLIB_INIT_FUNCTION (setup_memclnt_exit); static clib_error_t * vl_api_ring_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { int i; ring_alloc_t *ap; vl_shmem_hdr_t *shmem_hdr; api_main_t *am = &api_main; shmem_hdr = am->shmem_hdr; if (shmem_hdr == 0) { vlib_cli_output (vm, "Shared memory segment not initialized...\n"); return 0; } vlib_cli_output (vm, "%8s %8s %8s %8s %8s\n", "Owner", "Size", "Nitems", "Hits", "Misses"); ap = shmem_hdr->vl_rings; for (i = 0; i < vec_len (shmem_hdr->vl_rings); i++) { vlib_cli_output (vm, "%8s %8d %8d %8d %8d\n", "vlib", ap->size, ap->nitems, ap->hits, ap->misses); ap++; } ap = shmem_hdr->client_rings; for (i = 0; i < vec_len (shmem_hdr->client_rings); i++) { vlib_cli_output (vm, "%8s %8d %8d %8d %8d\n", "clnt", ap->size, ap->nitems, ap->hits, ap->misses); ap++; } vlib_cli_output (vm, "%d ring miss fallback allocations\n", am->ring_misses); vlib_cli_output (vm, "%d application restarts, %d reclaimed msgs, %d garbage collects\n", shmem_hdr->application_restarts, shmem_hdr->restart_reclaims, shmem_hdr->garbage_collects); return 0; } void dump_socket_clients (vlib_main_t * vm, api_main_t * am) __attribute__ ((weak)); void dump_socket_clients (vlib_main_t * vm, api_main_t * am) { } static clib_error_t * vl_api_client_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { vl_api_registration_t **regpp, *regp; unix_shared_memory_queue_t *q; char *health; api_main_t *am = &api_main; u32 *confused_indices = 0; if (!pool_elts (am->vl_clients)) goto socket_clients; vlib_cli_output (vm, "Shared memory clients"); vlib_cli_output (vm, "%16s %8s %14s %18s %s", "Name", "PID", "Queue Length", "Queue VA", "Health"); /* *INDENT-OFF* */ pool_foreach (regpp, am->vl_clients, ({ regp = *regpp; if (regp) { q = regp->vl_input_queue; if (kill (q->consumer_pid, 0) < 0) { health = "DEAD"; } else { health = "alive"; } vlib_cli_output (vm, "%16s %8d %14d 0x%016llx %s\n", regp->name, q->consumer_pid, q->cursize, q, health); } else { clib_warning ("NULL client registration index %d", regpp - am->vl_clients); vec_add1 (confused_indices, regpp - am->vl_clients); } })); /* *INDENT-ON* */ /* This should "never happen," but if it does, fix it... */ if (PREDICT_FALSE (vec_len (confused_indices) > 0)) { int i; for (i = 0; i < vec_len (confused_indices); i++) { pool_put_index (am->vl_clients, confused_indices[i]); } } vec_free (confused_indices); if (am->missing_clients) vlib_cli_output (vm, "%u messages with missing clients", am->missing_clients); socket_clients: dump_socket_clients (vm, am); return 0; } static clib_error_t * vl_api_status_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { api_main_t *am = &api_main; // check if rx_trace and tx_trace are not null pointers if (am->rx_trace == 0) { vlib_cli_output (vm, "RX Trace disabled\n"); } else { if (am->rx_trace->enabled == 0) vlib_cli_output (vm, "RX Trace disabled\n"); else vlib_cli_output (vm, "RX Trace enabled\n"); } if (am->tx_trace == 0) { vlib_cli_output (vm, "TX Trace disabled\n"); } else { if (am->tx_trace->enabled == 0) vlib_cli_output (vm, "TX Trace disabled\n"); else vlib_cli_output (vm, "TX Trace enabled\n"); } return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_command, static) = { .path = "show api", .short_help = "Show API information", }; /* *INDENT-ON* */ /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_ring_command, static) = { .path = "show api ring-stats", .short_help = "Message ring statistics", .function = vl_api_ring_command, }; /* *INDENT-ON* */ /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_clients_command, static) = { .path = "show api clients", .short_help = "Client information", .function = vl_api_client_command, }; /* *INDENT-ON* */ /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_status_command, static) = { .path = "show api status", .short_help = "Show API trace status", .function = vl_api_status_command, }; /* *INDENT-ON* */ static clib_error_t * vl_api_message_table_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { api_main_t *am = &api_main; int i; int verbose = 0; if (unformat (input, "verbose")) verbose = 1; if (verbose == 0) vlib_cli_output (vm, "%-4s %s", "ID", "Name"); else vlib_cli_output (vm, "%-4s %-40s %6s %7s", "ID", "Name", "Bounce", "MP-safe"); for (i = 1; i < vec_len (am->msg_names); i++) { if (verbose == 0) { vlib_cli_output (vm, "%-4d %s", i, am->msg_names[i] ? am->msg_names[i] : " [no handler]"); } else { vlib_cli_output (vm, "%-4d %-40s %6d %7d", i, am->msg_names[i] ? am->msg_names[i] : " [no handler]", am->message_bounce[i], am->is_mp_safe[i]); } } return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_message_table_command, static) = { .path = "show api message-table", .short_help = "Message Table", .function = vl_api_message_table_command, }; /* *INDENT-ON* */ static clib_error_t * vl_api_trace_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { u32 nitems = 1024; vl_api_trace_which_t which = VL_API_TRACE_RX; api_main_t *am = &api_main; while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { if (unformat (input, "rx nitems %u", &nitems) || unformat (input, "rx")) goto configure; else if (unformat (input, "tx nitems %u", &nitems) || unformat (input, "tx")) { which = VL_API_TRACE_RX; goto configure; } else if (unformat (input, "on rx")) { vl_msg_api_trace_onoff (am, VL_API_TRACE_RX, 1); } else if (unformat (input, "on tx")) { vl_msg_api_trace_onoff (am, VL_API_TRACE_TX, 1); } else if (unformat (input, "on")) { vl_msg_api_trace_onoff (am, VL_API_TRACE_RX, 1); } else if (unformat (input, "off")) { vl_msg_api_trace_onoff (am, VL_API_TRACE_RX, 0); vl_msg_api_trace_onoff (am, VL_API_TRACE_TX, 0); } else if (unformat (input, "free")) { vl_msg_api_trace_onoff (am, VL_API_TRACE_RX, 0); vl_msg_api_trace_onoff (am, VL_API_TRACE_TX, 0); vl_msg_api_trace_free (am, VL_API_TRACE_RX); vl_msg_api_trace_free (am, VL_API_TRACE_TX); } else if (unformat (input, "debug on")) { am->msg_print_flag = 1; } else if (unformat (input, "debug off")) { am->msg_print_flag = 0; } else return clib_error_return (0, "unknown input `%U'", format_unformat_error, input); } return 0; configure: if (vl_msg_api_trace_configure (am, which, nitems)) { vlib_cli_output (vm, "warning: trace configure error (%d, %d)", which, nitems); } return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (trace, static) = { .path = "set api-trace", .short_help = "API trace", .function = vl_api_trace_command, }; /* *INDENT-ON* */ clib_error_t * vlibmemory_init (vlib_main_t * vm) { api_main_t *am = &api_main; svm_map_region_args_t _a, *a = &_a; memset (a, 0, sizeof (*a)); a->root_path = am->root_path; a->name = SVM_GLOBAL_REGION_NAME; a->baseva = (am->global_baseva != 0) ? am->global_baseva : SVM_GLOBAL_REGION_BASEVA; a->size = (am->global_size != 0) ? am->global_size : SVM_GLOBAL_REGION_SIZE; a->flags = SVM_FLAGS_NODATA; a->uid = am->api_uid; a->gid = am->api_gid; a->pvt_heap_size = (am->global_pvt_heap_size != 0) ? am->global_pvt_heap_size : SVM_PVT_MHEAP_SIZE; svm_region_init_args (a); return 0; } VLIB_INIT_FUNCTION (vlibmemory_init); void vl_set_memory_region_name (const char *name) { api_main_t *am = &api_main; am->region_name = name; } static int range_compare (vl_api_msg_range_t * a0, vl_api_msg_range_t * a1) { int len0, len1, clen; len0 = vec_len (a0->name); len1 = vec_len (a1->name); clen = len0 < len1 ? len0 : len1; return (strncmp ((char *) a0->name, (char *) a1->name, clen)); } static u8 * format_api_msg_range (u8 * s, va_list * args) { vl_api_msg_range_t *rp = va_arg (*args, vl_api_msg_range_t *); if (rp == 0) s = format (s, "%-20s%9s%9s", "Name", "First-ID", "Last-ID"); else s = format (s, "%-20s%9d%9d", rp->name, rp->first_msg_id, rp->last_msg_id); return s; } static clib_error_t * vl_api_show_plugin_command (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cli_cmd) { api_main_t *am = &api_main; vl_api_msg_range_t *rp = 0; int i; if (vec_len (am->msg_ranges) == 0) { vlib_cli_output (vm, "No plugin API message ranges configured..."); return 0; } rp = vec_dup (am->msg_ranges); vec_sort_with_function (rp, range_compare); vlib_cli_output (vm, "Plugin API message ID ranges...\n"); vlib_cli_output (vm, "%U", format_api_msg_range, 0 /* header */ ); for (i = 0; i < vec_len (rp); i++) vlib_cli_output (vm, "%U", format_api_msg_range, rp + i); vec_free (rp); return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (cli_show_api_plugin_command, static) = { .path = "show api plugin", .short_help = "show api plugin", .function = vl_api_show_plugin_command, }; /* *INDENT-ON* */ static void vl_api_rpc_call_t_handler (vl_api_rpc_call_t * mp) { vl_api_rpc_call_reply_t *rmp; int (*fp) (void *); i32 rv = 0; vlib_main_t *vm = vlib_get_main (); if (mp->function == 0) { rv = -1; clib_warning ("rpc NULL function pointer"); } else { if (mp->need_barrier_sync) vlib_worker_thread_barrier_sync (vm); fp = uword_to_pointer (mp->function, int (*)(void *)); rv = fp (mp->data); if (mp->need_barrier_sync) vlib_worker_thread_barrier_release (vm); } if (mp->send_reply) { unix_shared_memory_queue_t *q = vl_api_client_index_to_input_queue (mp->client_index); if (q) { rmp = vl_msg_api_alloc_as_if_client (sizeof (*rmp)); rmp->_vl_msg_id = ntohs (VL_API_RPC_CALL_REPLY); rmp->context = mp->context; rmp->retval = rv; vl_msg_api_send_shmem (q, (u8 *) & rmp); } } if (mp->multicast) { clib_warning ("multicast not yet implemented..."); } } static void vl_api_rpc_call_reply_t_handler (vl_api_rpc_call_reply_t * mp) { clib_warning ("unimplemented"); } void vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length) { vl_api_rpc_call_t *mp; api_main_t *am = &api_main; vl_shmem_hdr_t *shmem_hdr = am->shmem_hdr; unix_shared_memory_queue_t *q; /* Main thread: call the function directly */ if (vlib_get_thread_index () == 0) { vlib_main_t *vm = vlib_get_main (); void (*call_fp) (void *); vlib_worker_thread_barrier_sync (vm); call_fp = fp; call_fp (data); vlib_worker_thread_barrier_release (vm); return; } /* Any other thread, actually do an RPC call... */ mp = vl_msg_api_alloc_as_if_client (sizeof (*mp) + data_length); memset (mp, 0, sizeof (*mp)); clib_memcpy (mp->data, data, data_length); mp->_vl_msg_id = ntohs (VL_API_RPC_CALL); mp->function = pointer_to_uword (fp); mp->need_barrier_sync = 1; /* * Use the "normal" control-plane mechanism for the main thread. * Well, almost. if the main input queue is full, we cannot * block. Otherwise, we can expect a barrier sync timeout. */ q = shmem_hdr->vl_input_queue; while (pthread_mutex_trylock (&q->mutex)) vlib_worker_thread_barrier_check (); while (PREDICT_FALSE (unix_shared_memory_queue_is_full (q))) { pthread_mutex_unlock (&q->mutex); vlib_worker_thread_barrier_check (); while (pthread_mutex_trylock (&q->mutex)) vlib_worker_thread_barrier_check (); } vl_msg_api_send_shmem_nolock (q, (u8 *) & mp); pthread_mutex_unlock (&q->mutex); } static void vl_api_trace_plugin_msg_ids_t_handler (vl_api_trace_plugin_msg_ids_t * mp) { api_main_t *am = &api_main; vl_api_msg_range_t *rp; uword *p; /* Noop (except for tracing) during normal operation */ if (am->replay_in_progress == 0) return; p = hash_get_mem (am->msg_range_by_name, mp->plugin_name); if (p == 0) { clib_warning ("WARNING: traced plugin '%s' not in current image", mp->plugin_name); return; } rp = vec_elt_at_index (am->msg_ranges, p[0]); if (rp->first_msg_id != clib_net_to_host_u16 (mp->first_msg_id)) { clib_warning ("WARNING: traced plugin '%s' first message id %d not %d", mp->plugin_name, clib_net_to_host_u16 (mp->first_msg_id), rp->first_msg_id); } if (rp->last_msg_id != clib_net_to_host_u16 (mp->last_msg_id)) { clib_warning ("WARNING: traced plugin '%s' last message id %d not %d", mp->plugin_name, clib_net_to_host_u16 (mp->last_msg_id), rp->last_msg_id); } } #define foreach_rpc_api_msg \ _(RPC_CALL,rpc_call) \ _(RPC_CALL_REPLY,rpc_call_reply) #define foreach_plugin_trace_msg \ _(TRACE_PLUGIN_MSG_IDS,trace_plugin_msg_ids) static clib_error_t * rpc_api_hookup (vlib_main_t * vm) { api_main_t *am = &api_main; #define _(N,n) \ vl_msg_api_set_handlers(VL_API_##N, #n, \ vl_api_##n##_t_handler, \ vl_noop_handler, \ vl_noop_handler, \ vl_api_##n##_t_print, \ sizeof(vl_api_##n##_t), 0 /* do not trace */); foreach_rpc_api_msg; #undef _ #define _(N,n) \ vl_msg_api_set_handlers(VL_API_##N, #n, \ vl_api_##n##_t_handler, \ vl_noop_handler, \ vl_noop_handler, \ vl_api_##n##_t_print, \ sizeof(vl_api_##n##_t), 1 /* do trace */); foreach_plugin_trace_msg; #undef _ /* No reason to halt the parade to create a trace record... */ am->is_mp_safe[VL_API_TRACE_PLUGIN_MSG_IDS] = 1; return 0; } VLIB_API_INIT_FUNCTION (rpc_api_hookup); typedef enum { DUMP, CUSTOM_DUMP, REPLAY, INITIALIZERS, } vl_api_replay_t; u8 * format_vl_msg_api_trace_status (u8 * s, va_list * args) { api_main_t *am = va_arg (*args, api_main_t *); vl_api_trace_which_t which = va_arg (*args, vl_api_trace_which_t); vl_api_trace_t *tp; char *trace_name; switch (which) { case VL_API_TRACE_TX: tp = am->tx_trace; trace_name = "TX trace"; break; case VL_API_TRACE_RX: tp = am->rx_trace; trace_name = "RX trace"; break; default: abort (); } if (tp == 0) { s = format (s, "%s: not yet configured.\n", trace_name); return s; } s = format (s, "%s: used %d of %d items, %s enabled, %s wrapped\n", trace_name, vec_len (tp->traces), tp->nitems, tp->enabled ? "is" : "is not", tp->wrapped ? "has" : "has not"); return s; } void vl_msg_api_custom_dump_configure (api_main_t * am) __attribute__ ((weak)); void vl_msg_api_custom_dump_configure (api_main_t * am) { } static void vl_msg_api_process_file (vlib_main_t * vm, u8 * filename, u32 first_index, u32 last_index, vl_api_replay_t which) { vl_api_trace_file_header_t *hp; int i, fd; struct stat statb; size_t file_size; u8 *msg; u8 endian_swap_needed = 0; api_main_t *am = &api_main; u8 *tmpbuf = 0; u32 nitems; void **saved_print_handlers = 0; fd = open ((char *) filename, O_RDONLY); if (fd < 0) { vlib_cli_output (vm, "Couldn't open %s\n", filename); return; } if (fstat (fd, &statb) < 0) { vlib_cli_output (vm, "Couldn't stat %s\n", filename); close (fd); return; } if (!(statb.st_mode & S_IFREG) || (statb.st_size < sizeof (*hp))) { vlib_cli_output (vm, "File not plausible: %s\n", filename); close (fd); return; } file_size = statb.st_size; file_size = (file_size + 4095) & ~(4096); hp = mmap (0, file_size, PROT_READ, MAP_PRIVATE, fd, 0); if (hp == (vl_api_trace_file_header_t *) MAP_FAILED) { vlib_cli_output (vm, "mmap failed: %s\n", filename); close (fd); return; } close (fd); if ((clib_arch_is_little_endian && hp->endian == VL_API_BIG_ENDIAN) || (clib_arch_is_big_endian && hp->endian == VL_API_LITTLE_ENDIAN)) endian_swap_needed = 1; if (endian_swap_needed) nitems = ntohl (hp->nitems); else nitems = hp->nitems; if (last_index == (u32) ~ 0) { last_index = nitems - 1; } if (first_index >= nitems || last_index >= nitems) { vlib_cli_output (vm, "Range (%d, %d) outside file range (0, %d)\n", first_index, last_index, nitems - 1); munmap (hp, file_size); return; } if (hp->wrapped) vlib_cli_output (vm, "Note: wrapped/incomplete trace, results may vary\n"); if (which == CUSTOM_DUMP) { saved_print_handlers = (void **) vec_dup (am->msg_print_handlers); vl_msg_api_custom_dump_configure (am); } msg = (u8 *) (hp + 1); for (i = 0; i < first_index; i++) { trace_cfg_t *cfgp; int size; u16 msg_id; size = clib_host_to_net_u32 (*(u32 *) msg); msg += sizeof (u32); if (clib_arch_is_little_endian) msg_id = ntohs (*((u16 *) msg)); else msg_id = *((u16 *) msg); cfgp = am->api_trace_cfg + msg_id; if (!cfgp) { vlib_cli_output (vm, "Ugh: msg id %d no trace config\n", msg_id); munmap (hp, file_size); return; } msg += size; } if (which == REPLAY) am->replay_in_progress = 1; for (; i <= last_index; i++) { trace_cfg_t *cfgp; u16 *msg_idp; u16 msg_id; int size; if (which == DUMP) vlib_cli_output (vm, "---------- trace %d -----------\n", i); size = clib_host_to_net_u32 (*(u32 *) msg); msg += sizeof (u32); if (clib_arch_is_little_endian) msg_id = ntohs (*((u16 *) msg)); else msg_id = *((u16 *) msg); cfgp = am->api_trace_cfg + msg_id; if (!cfgp) { vlib_cli_output (vm, "Ugh: msg id %d no trace config\n", msg_id); munmap (hp, file_size); vec_free (tmpbuf); am->replay_in_progress = 0; return; } /* Copy the buffer (from the read-only mmap'ed file) */ vec_validate (tmpbuf, size - 1 + sizeof (uword)); clib_memcpy (tmpbuf + sizeof (uword), msg, size); memset (tmpbuf, 0xf, sizeof (uword)); /* * Endian swap if needed. All msg data is supposed to be * in network byte order. All msg handlers are supposed to * know that. The generic message dumpers don't know that. * One could fix apigen, I suppose. */ if ((which == DUMP && clib_arch_is_little_endian) || endian_swap_needed) { void (*endian_fp) (void *); if (msg_id >= vec_len (am->msg_endian_handlers) || (am->msg_endian_handlers[msg_id] == 0)) { vlib_cli_output (vm, "Ugh: msg id %d no endian swap\n", msg_id); munmap (hp, file_size); vec_free (tmpbuf); am->replay_in_progress = 0; return; } endian_fp = am->msg_endian_handlers[msg_id]; (*endian_fp) (tmpbuf + sizeof (uword)); } /* msg_id always in network byte order */ if (clib_arch_is_little_endian) { msg_idp = (u16 *) (tmpbuf + sizeof (uword)); *msg_idp = msg_id; } switch (which) { case CUSTOM_DUMP: case DUMP: if (msg_id < vec_len (am->msg_print_handlers) && am->msg_print_handlers[msg_id]) { u8 *(*print_fp) (void *, void *); print_fp = (void *) am->msg_print_handlers[msg_id]; (*print_fp) (tmpbuf + sizeof (uword), vm); } else { vlib_cli_output (vm, "Skipping msg id %d: no print fcn\n", msg_id); break; } break; case INITIALIZERS: if (msg_id < vec_len (am->msg_print_handlers) && am->msg_print_handlers[msg_id]) { u8 *s; int j; u8 *(*print_fp) (void *, void *); print_fp = (void *) am->msg_print_handlers[msg_id]; vlib_cli_output (vm, "/*"); (*print_fp) (tmpbuf + sizeof (uword), vm); vlib_cli_output (vm, "*/\n"); s = format (0, "static u8 * vl_api_%s_%d[%d] = {", am->msg_names[msg_id], i, am->api_trace_cfg[msg_id].size); for (j = 0; j < am->api_trace_cfg[msg_id].size; j++) { if ((j & 7) == 0) s = format (s, "\n "); s = format (s, "0x%02x,", tmpbuf[sizeof (uword) + j]); } s = format (s, "\n};\n%c", 0); vlib_cli_output (vm, (char *) s); vec_free (s); } break; case REPLAY: if (msg_id < vec_len (am->msg_print_handlers) && am->msg_print_handlers[msg_id] && cfgp->replay_enable) { void (*handler) (void *); handler = (void *) am->msg_handlers[msg_id]; if (!am->is_mp_safe[msg_id]) vl_msg_api_barrier_sync (); (*handler) (tmpbuf + sizeof (uword)); if (!am->is_mp_safe[msg_id]) vl_msg_api_barrier_release (); } else { if (cfgp->replay_enable) vlib_cli_output (vm, "Skipping msg id %d: no handler\n", msg_id); break; } break; } _vec_len (tmpbuf) = 0; msg += size; } if (saved_print_handlers) { clib_memcpy (am->msg_print_handlers, saved_print_handlers, vec_len (am->msg_print_handlers) * sizeof (void *)); vec_free (saved_print_handlers); } munmap (hp, file_size); vec_free (tmpbuf); am->replay_in_progress = 0; } static clib_error_t * api_trace_command_fn (vlib_main_t * vm, unformat_input_t * input, vlib_cli_command_t * cmd) { u32 nitems = 256 << 10; api_main_t *am = &api_main; vl_api_trace_which_t which = VL_API_TRACE_RX; u8 *filename; u32 first = 0; u32 last = (u32) ~ 0; FILE *fp; int rv; while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { if (unformat (input, "on") || unformat (input, "enable")) { if (unformat (input, "nitems %d", &nitems)) ; vl_msg_api_trace_configure (am, which, nitems); vl_msg_api_trace_onoff (am, which, 1 /* on */ ); } else if (unformat (input, "off")) { vl_msg_api_trace_onoff (am, which, 0); } else if (unformat (input, "save %s", &filename)) { u8 *chroot_filename; if (strstr ((char *) filename, "..") || index ((char *) filename, '/')) { vlib_cli_output (vm, "illegal characters in filename '%s'", filename); return 0; } chroot_filename = format (0, "/tmp/%s%c", filename, 0); vec_free (filename); fp = fopen ((char *) chroot_filename, "w"); if (fp == NULL) { vlib_cli_output (vm, "Couldn't create %s\n", chroot_filename); return 0; } rv = vl_msg_api_trace_save (am, which, fp); fclose (fp); if (rv == -1) vlib_cli_output (vm, "API Trace data not present\n"); else if (rv == -2) vlib_cli_output (vm, "File for writing is closed\n"); else if (rv == -10) vlib_cli_output (vm, "Error while writing header to file\n"); else if (rv == -11) vlib_cli_output (vm, "Error while writing trace to file\n"); else if (rv == -12) vlib_cli_output (vm, "Error while writing end of buffer trace to file\n"); else if (rv == -13) vlib_cli_output (vm, "Error while writing start of buffer trace to file\n"); else if (rv < 0) vlib_cli_output (vm, "Unkown error while saving: %d", rv); else vlib_cli_output (vm, "API trace saved to %s\n", chroot_filename); vec_free (chroot_filename); } else if (unformat (input, "dump %s", &filename)) { vl_msg_api_process_file (vm, filename, first, last, DUMP); } else if (unformat (input, "custom-dump %s", &filename)) { vl_msg_api_process_file (vm, filename, first, last, CUSTOM_DUMP); } else if (unformat (input, "replay %s", &filename)) { vl_msg_api_process_file (vm, filename, first, last, REPLAY); } else if (unformat (input, "initializers %s", &filename)) { vl_msg_api_process_file (vm, filename, first, last, INITIALIZERS); } else if (unformat (input, "tx")) { which = VL_API_TRACE_TX; } else if (unformat (input, "first %d", &first)) { ; } else if (unformat (input, "last %d", &last)) { ; } else if (unformat (input, "status")) { vlib_cli_output (vm, "%U", format_vl_msg_api_trace_status, am, which); } else if (unformat (input, "free")) { vl_msg_api_trace_onoff (am, which, 0); vl_msg_api_trace_free (am, which); } else if (unformat (input, "post-mortem-on")) vl_msg_api_post_mortem_dump_enable_disable (1 /* enable */ ); else if (unformat (input, "post-mortem-off")) vl_msg_api_post_mortem_dump_enable_disable (0 /* enable */ ); else return clib_error_return (0, "unknown input `%U'", format_unformat_error, input); } return 0; } /* *INDENT-OFF* */ VLIB_CLI_COMMAND (api_trace_command, static) = { .path = "api trace", .short_help = "api trace [on|off][dump|save|replay <file>][status][free][post-mortem-on]", .function = api_trace_command_fn, }; /* *INDENT-ON* */ static clib_error_t * api_config_fn (vlib_main_t * vm, unformat_input_t * input) { u32 nitems = 256 << 10; vl_api_trace_which_t which = VL_API_TRACE_RX; api_main_t *am = &api_main; while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { if (unformat (input, "on") || unformat (input, "enable")) { if (unformat (input, "nitems %d", &nitems)) ; vl_msg_api_trace_configure (am, which, nitems); vl_msg_api_trace_onoff (am, which, 1 /* on */ ); vl_msg_api_post_mortem_dump_enable_disable (1 /* enable */ ); } else return clib_error_return (0, "unknown input `%U'", format_unformat_error, input); } return 0; } VLIB_CONFIG_FUNCTION (api_config_fn, "api-trace"); static clib_error_t * api_queue_config_fn (vlib_main_t * vm, unformat_input_t * input) { api_main_t *am = &api_main; u32 nitems; while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) { if (unformat (input, "length %d", &nitems) || (unformat (input, "len %d", &nitems))) { if (nitems >= 1024) am->vlib_input_queue_length = nitems; else clib_warning ("vlib input queue length %d too small, ignored", nitems); } else return clib_error_return (0, "unknown input `%U'", format_unformat_error, input); } return 0; } VLIB_CONFIG_FUNCTION (api_queue_config_fn, "api-queue"); /* * fd.io coding-style-patch-verification: ON * * Local Variables: * eval: (c-set-style "gnu") * End: */