forked from openvswitch/ovs
-
Notifications
You must be signed in to change notification settings - Fork 0
/
upcall_cost.py
executable file
·1787 lines (1532 loc) · 64.4 KB
/
upcall_cost.py
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
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#!/usr/bin/env python3
#
# Copyright (c) 2021 Red Hat, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at:
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# Script information:
# -------------------
# upcall_cost.py uses various user space and kernel space probes to determine
# the costs (in time) for handling the first packet in user space. It
# calculates the following costs:
#
# - Time it takes from the kernel sending the upcall till it's received by the
# ovs-vswitchd process.
# - Time it takes from ovs-vswitchd sending the execute actions command till
# the kernel receives it.
# - The total time it takes from the kernel to sent the upcall until it
# receives the packet execute command.
# - The total time of the above, minus the time it takes for the actual lookup.
#
# In addition, it will also report the number of packets batched, as OVS will
# first try to read UPCALL_MAX_BATCH(64) packets from kernel space and then
# does the flow lookups and execution. So the smaller the batch size, the more
# realistic are the cost estimates.
#
# The script does not need any options to attach to a running instance of
# ovs-vswitchd. However, it's recommended always run the script with the
# --write-events option. This way, if something does go wrong, the collected
# data is saved. Use the --help option to see all the available options.
#
# Note: In addition to the bcc tools for your specific setup, you need the
# following Python packages:
# pip install alive-progress halo psutil scapy strenum text_histogram3
#
try:
from bcc import BPF, USDT, USDTException
except ModuleNotFoundError:
print("WARNING: Can't find the BPF Compiler Collection (BCC) tools!")
print(" This is NOT problem if you analyzing previously collected"
" data.\n")
from alive_progress import alive_bar
from collections import namedtuple
from halo import Halo
from scapy.all import TCP, UDP
from scapy.layers.l2 import Ether
from strenum import StrEnum
from text_histogram3 import histogram
from time import process_time
import argparse
import ast
import psutil
import re
import struct
import subprocess
import sys
import time
#
# Global definitions
#
DP_TUNNEL_PORT = -1
#
# Actual eBPF source code
#
ebpf_source = """
#include <linux/sched.h>
#include <linux/skbuff.h>
#include <uapi/linux/bpf.h>
#define MAX_PACKET <MAX_PACKET_VAL>
#define MAX_KEY <MAX_KEY_VAL>
enum {
EVENT_RECV_UPCALL = 0,
EVENT_DP_UPCALL,
EVENT_OP_FLOW_PUT,
EVENT_OP_FLOW_EXECUTE,
EVENT_OVS_PKT_EXEC,
_EVENT_MAX_EVENT
};
#define barrier_var(var) asm volatile("" : "=r"(var) : "0"(var))
struct event_t {
u32 event;
u32 cpu;
u32 pid;
u32 upcall_type;
u64 ts;
u32 pkt_frag_size;
u32 pkt_size;
u64 key_size;
char comm[TASK_COMM_LEN];
char dpif_name[32];
char dev_name[16];
unsigned char pkt[MAX_PACKET];
unsigned char key[MAX_KEY];
};
BPF_RINGBUF_OUTPUT(events, <BUFFER_PAGE_CNT>);
BPF_TABLE("percpu_array", uint32_t, uint64_t, dropcnt, _EVENT_MAX_EVENT);
static struct event_t *init_event(u32 type)
{
struct event_t *event = events.ringbuf_reserve(sizeof(struct event_t));
if (!event) {
uint64_t *value = dropcnt.lookup(&type);
if (value)
__sync_fetch_and_add(value, 1);
return NULL;
}
event->event = type;
event->ts = bpf_ktime_get_ns();
event->cpu = bpf_get_smp_processor_id();
event->pid = bpf_get_current_pid_tgid();
bpf_get_current_comm(&event->comm, sizeof(event->comm));
return event;
}
int trace__recv_upcall(struct pt_regs *ctx) {
uint32_t upcall_type;
uint64_t addr;
uint64_t size;
bpf_usdt_readarg(2, ctx, &upcall_type);
if (upcall_type != 0)
return 0;
struct event_t *event = init_event(EVENT_RECV_UPCALL);
if (!event)
return 1;
bpf_usdt_readarg(1, ctx, &addr);
bpf_probe_read_str(&event->dpif_name, sizeof(event->dpif_name),
(void *)addr);
event->upcall_type = upcall_type;
bpf_usdt_readarg(4, ctx, &event->pkt_size);
bpf_usdt_readarg(6, ctx, &event->key_size);
if (event->pkt_size > MAX_PACKET)
size = MAX_PACKET;
else
size = event->pkt_size;
bpf_usdt_readarg(3, ctx, &addr);
bpf_probe_read(&event->pkt, size, (void *)addr);
if (event->key_size > MAX_KEY)
size = MAX_KEY;
else
size = event->key_size;
bpf_usdt_readarg(5, ctx, &addr);
bpf_probe_read(&event->key, size, (void *)addr);
events.ringbuf_submit(event, 0);
return 0;
};
int trace__op_flow_put(struct pt_regs *ctx) {
uint64_t addr;
uint64_t size;
struct event_t *event = init_event(EVENT_OP_FLOW_PUT);
if (!event) {
return 1;
}
events.ringbuf_submit(event, 0);
return 0;
};
int trace__op_flow_execute(struct pt_regs *ctx) {
uint64_t addr;
uint64_t size;
struct event_t *event = init_event(EVENT_OP_FLOW_EXECUTE);
if (!event) {
return 1;
}
bpf_usdt_readarg(4, ctx, &event->pkt_size);
if (event->pkt_size > MAX_PACKET)
size = MAX_PACKET;
else
size = event->pkt_size;
bpf_usdt_readarg(3, ctx, &addr);
bpf_probe_read(&event->pkt, size, (void *)addr);
events.ringbuf_submit(event, 0);
return 0;
};
TRACEPOINT_PROBE(openvswitch, ovs_dp_upcall) {
uint64_t size;
struct sk_buff *skb = args->skbaddr;
if (args->upcall_cmd != 1 || skb == NULL || skb->data == NULL)
return 0;
struct event_t *event = init_event(EVENT_DP_UPCALL);
if (!event) {
return 1;
}
event->upcall_type = args->upcall_cmd;
event->pkt_size = args->len;
TP_DATA_LOC_READ_CONST(&event->dpif_name, dp_name,
sizeof(event->dpif_name));
TP_DATA_LOC_READ_CONST(&event->dev_name, dev_name,
sizeof(event->dev_name));
if (skb->data_len != 0) {
event->pkt_frag_size = (skb->len - skb->data_len) & 0xfffffff;
size = event->pkt_frag_size;
} else {
event->pkt_frag_size = 0;
size = event->pkt_size;
}
/* Prevent clang from using register mirroring (or any optimization) on
* the 'size' variable. */
barrier_var(size);
if (size > MAX_PACKET)
size = MAX_PACKET;
bpf_probe_read_kernel(event->pkt, size, skb->data);
events.ringbuf_submit(event, 0);
return 0;
}
int kprobe__ovs_packet_cmd_execute(struct pt_regs *ctx, struct sk_buff *skb)
{
uint64_t size;
if (skb == NULL || skb->data == NULL)
return 0;
struct event_t *event = init_event(EVENT_OVS_PKT_EXEC);
if (!event) {
return 1;
}
events.ringbuf_submit(event, 0);
return 0;
}
"""
#
# Event types
#
class EventType(StrEnum):
RECV_UPCALL = 'dpif_recv__recv_upcall'
DP_UPCALL = 'openvswitch__dp_upcall'
OP_FLOW_PUT = 'dpif_netlink_operate__op_flow_put'
OP_FLOW_EXECUTE = 'dpif_netlink_operate__op_flow_execute'
OVS_PKT_EXEC = 'ktrace__ovs_packet_cmd_execute'
def short_name(name, length=22):
if len(name) < length:
return name
return '..' + name[-(length - 2):]
def from_trace(trace_event):
if trace_event == 0:
return EventType.RECV_UPCALL
elif trace_event == 1:
return EventType.DP_UPCALL
elif trace_event == 2:
return EventType.OP_FLOW_PUT
elif trace_event == 3:
return EventType.OP_FLOW_EXECUTE
elif trace_event == 4:
return EventType.OVS_PKT_EXEC
raise ValueError
#
# Simple event class
#
class Event(object):
def __init__(self, ts, pid, comm, cpu, event_type):
self.ts = ts
self.pid = pid
self.comm = comm
self.cpu = cpu
self.event_type = event_type
def __str__(self):
return "[{:<22}] {:<16} {:8} [{:03}] {:18.9f}".format(
EventType.short_name(self.event_type),
self.comm,
self.pid,
self.cpu,
self.ts / 1000000000)
def __repr__(self):
more = ""
if self.__class__.__name__ != "Event":
more = ", ..."
return "{}({}, {}, {}, {}, {}{})".format(self.__class__.__name__,
self.ts, self.pid,
self.comm, self.cpu,
self.event_type, more)
def handle_event(event):
event = Event(event.ts, event.pid, event.comm.decode("utf-8"),
event.cpu, EventType.from_trace(event.event))
if not options.quiet:
print(event)
return event
def get_event_header_str():
return "{:<24} {:<16} {:>8} {:<3} {:<18} {}".format(
"EVENT", "COMM", "PID", "CPU", "TIME",
"EVENT DATA[dpif_name/dp_port/pkt_len/pkt_frag_len]")
#
# dp_upcall event class
#
class DpUpcall(Event):
def __init__(self, ts, pid, comm, cpu, dpif_name, port, pkt, pkt_len,
pkt_frag_len):
super(DpUpcall, self).__init__(ts, pid, comm, cpu, EventType.DP_UPCALL)
self.dpif_name = dpif_name
self.dp_port = get_dp_mapping(dpif_name, port)
if self.dp_port is None:
#
# As we only identify interfaces at startup, new interfaces could
# have been added, causing the lookup to fail. Just something to
# keep in mind when running this in a dynamic environment.
#
raise LookupError("Can't find datapath port mapping!")
self.pkt = pkt
self.pkt_len = pkt_len
self.pkt_frag_len = pkt_frag_len
def __str__(self):
return "[{:<22}] {:<16} {:8} [{:03}] {:18.9f}: " \
"{:<17} {:4} {:4} {:4}".format(self.event_type,
self.comm,
self.pid,
self.cpu,
self.ts / 1000000000,
self.dpif_name,
self.dp_port,
self.pkt_len,
self.pkt_frag_len)
def handle_event(event):
if event.pkt_size < options.packet_size:
pkt_len = event.pkt_size
else:
pkt_len = options.packet_size
pkt_data = bytes(event.pkt)[:pkt_len]
if len(pkt_data) <= 0 or event.pkt_size == 0:
return
try:
event = DpUpcall(event.ts, event.pid, event.comm.decode("utf-8"),
event.cpu, event.dpif_name.decode("utf-8"),
event.dev_name.decode("utf-8"),
pkt_data,
event.pkt_size,
event.pkt_frag_size)
except LookupError:
#
# If we can't do the port lookup, ignore this event.
#
return None
if not options.quiet:
print(event)
return event
#
# recv_upcall event class
#
class RecvUpcall(Event):
def __init__(self, ts, pid, comm, cpu, dpif_name, key, pkt, pkt_len):
super(RecvUpcall, self).__init__(ts, pid, comm, cpu,
EventType.RECV_UPCALL)
if dpif_name.startswith("system@"):
dpif_name = dpif_name[len("system@"):]
self.dpif_name = dpif_name
nla = RecvUpcall.decode_nlm(key, dump=False)
if "OVS_KEY_ATTR_IN_PORT" in nla:
self.dp_port = struct.unpack('=L', nla["OVS_KEY_ATTR_IN_PORT"])[0]
elif "OVS_KEY_ATTR_TUNNEL" in nla:
self.dp_port = DP_TUNNEL_PORT
else:
self.dp_port = RecvUpcall.get_system_dp_port(self.dpif_name)
if self.dp_port is None:
raise LookupError("Can't find RecvUpcall dp port mapping!")
self.pkt = pkt
self.pkt_len = pkt_len
def __str__(self):
return "[{:<22}] {:<16} {:8} [{:03}] {:18.9f}: {:<17} {:4} {:4}". \
format(
self.event_type,
self.comm,
self.pid,
self.cpu,
self.ts / 1000000000,
self.dpif_name,
self.dp_port,
self.pkt_len)
def get_system_dp_port(dpif_name):
dp_map = get_dp_mapping(dpif_name, "ovs-system", return_map=True)
if dpif_name not in dp_map:
return None
try:
return dp_map[dpif_name]["ovs-system"]
except KeyError:
return None
def decode_nlm(msg, indent=4, dump=True):
bytes_left = len(msg)
result = {}
while bytes_left:
if bytes_left < 4:
if dump:
print("{}WARN: decode truncated; can't read header".format(
' ' * indent))
break
nla_len, nla_type = struct.unpack("=HH", msg[:4])
if nla_len < 4:
if dump:
print("{}WARN: decode truncated; nla_len < 4".format(
' ' * indent))
break
nla_data = msg[4:nla_len]
trunc = ""
if nla_len > bytes_left:
trunc = "..."
nla_data = nla_data[:(bytes_left - 4)]
if RecvUpcall.get_ovs_key_attr_str(nla_type) == \
"OVS_KEY_ATTR_TUNNEL":
#
# If we have truncated tunnel information, we still would
# like to know. This is due to the special tunnel handling
# needed for port matching.
#
result[RecvUpcall.get_ovs_key_attr_str(nla_type)] = bytes()
else:
result[RecvUpcall.get_ovs_key_attr_str(nla_type)] = nla_data
if dump:
print("{}nla_len {}, nla_type {}[{}], data: {}{}".format(
' ' * indent, nla_len,
RecvUpcall.get_ovs_key_attr_str(nla_type),
nla_type,
"".join("{:02x} ".format(b) for b in nla_data), trunc))
if trunc != "":
if dump:
print("{}WARN: decode truncated; nla_len > msg_len[{}] ".
format(' ' * indent, bytes_left))
break
# Update next offset, but make sure it's aligned correctly.
next_offset = (nla_len + 3) & ~(3)
msg = msg[next_offset:]
bytes_left -= next_offset
return result
def get_ovs_key_attr_str(attr):
ovs_key_attr = ["OVS_KEY_ATTR_UNSPEC",
"OVS_KEY_ATTR_ENCAP",
"OVS_KEY_ATTR_PRIORITY",
"OVS_KEY_ATTR_IN_PORT",
"OVS_KEY_ATTR_ETHERNET",
"OVS_KEY_ATTR_VLAN",
"OVS_KEY_ATTR_ETHERTYPE",
"OVS_KEY_ATTR_IPV4",
"OVS_KEY_ATTR_IPV6",
"OVS_KEY_ATTR_TCP",
"OVS_KEY_ATTR_UDP",
"OVS_KEY_ATTR_ICMP",
"OVS_KEY_ATTR_ICMPV6",
"OVS_KEY_ATTR_ARP",
"OVS_KEY_ATTR_ND",
"OVS_KEY_ATTR_SKB_MARK",
"OVS_KEY_ATTR_TUNNEL",
"OVS_KEY_ATTR_SCTP",
"OVS_KEY_ATTR_TCP_FLAGS",
"OVS_KEY_ATTR_DP_HASH",
"OVS_KEY_ATTR_RECIRC_ID",
"OVS_KEY_ATTR_MPLS",
"OVS_KEY_ATTR_CT_STATE",
"OVS_KEY_ATTR_CT_ZONE",
"OVS_KEY_ATTR_CT_MARK",
"OVS_KEY_ATTR_CT_LABELS",
"OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4",
"OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6",
"OVS_KEY_ATTR_NSH"]
if attr < 0 or attr > len(ovs_key_attr):
return "<UNKNOWN>"
return ovs_key_attr[attr]
def handle_event(event):
#
# For us, only upcalls with a packet, flow_key, and upcall_type
# DPIF_UC_MISS are of interest.
#
if event.pkt_size <= 0 or event.key_size <= 0 or \
event.upcall_type != 0:
return
if event.key_size < options.flow_key_size:
key_len = event.key_size
else:
key_len = options.flow_key_size
if event.pkt_size < options.packet_size:
pkt_len = event.pkt_size
else:
pkt_len = options.packet_size
try:
event = RecvUpcall(event.ts, event.pid, event.comm.decode("utf-8"),
event.cpu, event.dpif_name.decode("utf-8"),
bytes(event.key)[:key_len],
bytes(event.pkt)[:pkt_len],
event.pkt_size)
except LookupError:
return None
if not options.quiet:
print(event)
return event
#
# op_flow_execute event class
#
class OpFlowExecute(Event):
def __init__(self, ts, pid, comm, cpu, pkt, pkt_len):
super(OpFlowExecute, self).__init__(ts, pid, comm, cpu,
EventType.OP_FLOW_EXECUTE)
self.pkt = pkt
self.pkt_len = pkt_len
def __str__(self):
return "[{:<22}] {:<16} {:8} [{:03}] {:18.9f}: " \
"{:<17} {:4} {:4}".format(EventType.short_name(self.event_type),
self.comm,
self.pid,
self.cpu,
self.ts / 1000000000,
"",
"",
self.pkt_len)
def handle_event(event):
if event.pkt_size < options.packet_size:
pkt_len = event.pkt_size
else:
pkt_len = options.packet_size
pkt_data = bytes(event.pkt)[:pkt_len]
if len(pkt_data) <= 0 or event.pkt_size == 0:
return
event = OpFlowExecute(event.ts, event.pid, event.comm.decode("utf-8"),
event.cpu, pkt_data, event.pkt_size)
if not options.quiet:
print(event)
return event
#
# get_dp_mapping()
#
def get_dp_mapping(dp, port, return_map=False, dp_map=None):
if options.unit_test:
return port
if dp_map is not None:
get_dp_mapping.dp_port_map_cache = dp_map
#
# Build a cache, so we do not have to execue the ovs command each time.
#
if not hasattr(get_dp_mapping, "dp_port_map_cache"):
try:
output = subprocess.check_output(['ovs-appctl', 'dpctl/show'],
encoding='utf8').split("\n")
except subprocess.CalledProcessError:
output = ""
pass
current_dp = None
get_dp_mapping.dp_port_map_cache = {}
for line in output:
match = re.match("^system@(.*):$", line)
if match is not None:
current_dp = match.group(1)
match = re.match("^ port ([0-9]+): ([^ /]*)", line)
if match is not None and current_dp is not None:
try:
get_dp_mapping.dp_port_map_cache[
current_dp][match.group(2)] = int(match.group(1))
except KeyError:
get_dp_mapping.dp_port_map_cache[current_dp] = \
{match.group(2): int(match.group(1))}
if return_map:
return get_dp_mapping.dp_port_map_cache
if dp not in get_dp_mapping.dp_port_map_cache or \
port not in get_dp_mapping.dp_port_map_cache[dp]:
return None
return get_dp_mapping.dp_port_map_cache[dp][port]
#
# event_to_dict()
#
def event_to_dict(event):
event_dict = {}
for field, _ in event._fields_:
if isinstance(getattr(event, field), (int, bytes)):
event_dict[field] = getattr(event, field)
else:
if (field == "key" and event.key_size == 0) or \
(field == "pkt" and event.pkt_size == 0):
data = bytes()
else:
data = bytes(getattr(event, field))
event_dict[field] = data
return event_dict
#
# receive_event_bcc()
#
def receive_event_bcc(ctx, data, size):
global events_received
events_received += 1
event = b['events'].event(data)
if export_file is not None:
export_file.write("event = {}\n".format(event_to_dict(event)))
receive_event(event)
#
# receive_event()
#
def receive_event(event):
global event_count
if event.event == 0:
trace_event = RecvUpcall.handle_event(event)
elif event.event == 1:
trace_event = DpUpcall.handle_event(event)
elif event.event == 2:
trace_event = Event.handle_event(event)
elif event.event == 3:
trace_event = OpFlowExecute.handle_event(event)
elif event.event == 4:
trace_event = Event.handle_event(event)
try:
event_count['total'][EventType.from_trace(event.event)] += 1
except KeyError:
event_count['total'][EventType.from_trace(event.event)] = 1
event_count['valid'][EventType.from_trace(event.event)] = 0
if trace_event is not None:
event_count['valid'][EventType.from_trace(event.event)] += 1
trace_data.append(trace_event)
#
# collect_event_sets()
#
def collect_event_sets(events, collect_stats=False, profile=False,
spinner=False):
t1_time = 0
def t1_start():
nonlocal t1_time
t1_time = process_time()
def t1_stop(description):
print("* PROFILING: {:<50}: {:.06f} seconds".format(
description, process_time() - t1_time))
warn_parcial_match = False
warn_frag = False
if profile:
t1_start()
#
# First let's create a dict of per handler thread events.
#
threads = {}
threads_result = {}
for idx, event in enumerate(events):
if event.event_type == EventType.DP_UPCALL:
continue
if event.pid not in threads:
threads[event.pid] = []
threads[event.pid].append([idx, event])
if profile:
t1_stop("Creating per thread dictionary")
t1_start()
#
# Now spit them in per upcall sets, but remember that
# RecvUpcall event can be batched.
#
batch_stats = []
for thread, items in threads.items():
thread_set = []
batch = []
ovs_pkt_exec_set = []
batching = True
collecting = 0
has_flow_put = False
has_flow_exec = False
def next_batch():
nonlocal batching, batch, collecting, has_flow_put, has_flow_exec
nonlocal ovs_pkt_exec_set, thread_set
if len(batch) > 0:
#
# If we are done with the batch, see if we need to match up
# any batched OVS_PKT_EXEC events.
#
for event in batch:
if len(ovs_pkt_exec_set) <= 0:
break
if any(isinstance(item,
OpFlowExecute) for item in event[2]):
event[2].append(ovs_pkt_exec_set.pop(0))
#
# Append the batch to the thread-specific set.
#
thread_set = thread_set + batch
if collect_stats:
batch_stats.append(len(batch))
batching = True
batch = []
ovs_pkt_exec_set = []
has_flow_put = False
has_flow_exec = False
collecting = 0
def next_batch_set():
nonlocal has_flow_put, has_flow_exec, collecting
has_flow_put = False
has_flow_exec = False
collecting += 1
for item in items:
idx, event = item
if batching:
if event.event_type == EventType.RECV_UPCALL:
batch.append(item + [[]])
elif len(batch) > 0:
batching = False
collecting = 0
else:
continue
if not batching:
if event.event_type == EventType.RECV_UPCALL:
next_batch()
batch.append(item + [[]])
else:
if event.event_type == EventType.OP_FLOW_PUT:
if has_flow_put:
next_batch_set()
if collecting >= len(batch):
next_batch()
continue
batch[collecting][2].append(item[1])
has_flow_put = True
elif event.event_type == EventType.OP_FLOW_EXECUTE:
if has_flow_exec:
next_batch_set()
if collecting >= len(batch):
next_batch()
continue
if (event.pkt_len == batch[collecting][1].pkt_len
and event.pkt == batch[collecting][1].pkt):
batch[collecting][2].append(item[1])
has_flow_put = True
has_flow_exec = True
else:
#
# If we end up here it could be that an upcall in a
# batch did not generate an EXECUTE and we are out
# of sync. Try to match it to the next batch entry.
#
next_idx = collecting + 1
while True:
if next_idx >= len(batch):
next_batch()
break
if (event.pkt_len == batch[next_idx][1].pkt_len
and event.pkt == batch[next_idx][1].pkt):
batch[next_idx][2] = batch[collecting][2]
batch[collecting][2] = []
collecting = next_idx
batch[collecting][2].append(item[1])
has_flow_put = True
has_flow_exec = True
break
next_idx += 1
elif event.event_type == EventType.OVS_PKT_EXEC:
#
# The OVS_PKT_EXEC might also be batched, so we keep
# them in a separate list and assign them to the
# correct set when completing the set.
#
ovs_pkt_exec_set.append(item[1])
continue
if collecting >= len(batch):
next_batch()
next_batch()
threads_result[thread] = thread_set
if profile:
t1_stop("Creating upcall sets")
t1_start()
#
# Move thread results from list to dictionary
#
thread_stats = {}
for thread, sets in threads_result.items():
if len(sets) > 0:
thread_stats[sets[0][1].comm] = len(sets)
threads_result[thread] = {}
for upcall in sets:
threads_result[thread][upcall[0]] = [upcall[1]] + upcall[2]
if profile:
t1_stop("Moving upcall list to dictionary")
t1_start()
if options.debug & 0x4000000 != 0:
print()
for thread, sets in threads_result.items():
for idx, idx_set in sets.items():
print("DBG: {}".format(idx_set))
#
# Create two lists on with DP_UPCALLs and RECV_UPCALLs
#
dp_upcall_list = []
recv_upcall_list = []
for idx, event in enumerate(events):
if event.event_type == EventType.DP_UPCALL:
dp_upcall_list.append([idx, event])
elif event.event_type == EventType.RECV_UPCALL:
recv_upcall_list.append([idx, event])
if profile:
t1_stop("Creating DP_UPCALL and RECV_UPCALL lists")
t1_start()
if options.debug & 0x4000000 != 0:
print()
for dp_upcall in dp_upcall_list:
print("DBG: {}".format(dp_upcall))
print()
for recv_upcall in recv_upcall_list:
print("DBG: {}".format(recv_upcall))
#
# Now find the matching DP_UPCALL and RECV_UPCALL events
#
event_sets = []
if spinner:
print()
with alive_bar(len(dp_upcall_list),
title="- Matching DP_UPCALLs to RECV_UPCALLs",
spinner=None, disable=not spinner) as bar:
for (idx, event) in dp_upcall_list:
remove_indexes = []
this_set = None
#
# TODO: This part needs some optimization, as it's slow in the
# PVP test scenario. This is because a lot of DP_UPCALLS
# will not have a matching RECV_UPCALL leading to walking
# the entire recv_upcall_list list.
#
# Probably some dictionary, but in the PVP scenario packets
# come from a limited set of ports, and the length is all the
# same. So we do need the key to be recv.dport +
# len(recv.pkt) + recv.pkt, however, the recv.pkt compare
# needs to happen on min(len(event.pkt), len(recv.pkt)).
#
for idx_in_list, (idx_recv, recv) in enumerate(recv_upcall_list):
match = False
if idx_recv < idx:
remove_indexes.append(idx_in_list)
continue
#
# If the RecvUpcall is a tunnel port, we can not map it to
# the correct tunnel. For now, we assume the first matching
# packet is the correct one. For more details see the OVS
# ukey_to_flow_netdev() function.
#
if (event.dp_port == recv.dp_port or
recv.dp_port == DP_TUNNEL_PORT) \
and event.pkt_len == recv.pkt_len:
compare_len = min(len(event.pkt), len(recv.pkt))
if len(event.pkt) != len(recv.pkt) \
and event.pkt_frag_len == 0:
warn_parcial_match = True
elif event.pkt_frag_len != 0:
warn_frag = True
compare_len = min(compare_len, event.pkt_frag_len)
if event.pkt[:compare_len] == recv.pkt[:compare_len]:
match = True
else:
#
# There are still some corner cases due to the fact
# the kernel dp_upcall tracepoint is hit before the