Repository navigation
Expand file tree
/
Copy pathext4dist.py
More file actions
executable file
·203 lines (178 loc) · 5.4 KB
/
Copy pathext4dist.py
File metadata 8000 and controls
executable file
·203 lines (178 loc) · 5.4 KB
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
#!/usr/bin/python
# @lint-avoid-python-3-compatibility-imports
#
# ext4dist Summarize ext4 operation latency.
# For Linux, uses BCC, eBPF.
#
# USAGE: ext4dist [-h] [-T] [-m] [-p PID] [interval] [count]
#
# Copyright 2016 Netflix, Inc.
# Licensed under the Apache License, Version 2.0 (the "License")
#
# 12-Feb-2016 Brendan Gregg Created this.
from __future__ import print_function
from bcc import BPF
from time import sleep, strftime
import argparse
# symbols
kallsyms = "/proc/kallsyms"
# arguments
examples = """examples:
./ext4dist # show operation latency as a histogram
./ext4dist -p 181 # trace PID 181 only
./ext4dist 1 10 # print 1 second summaries, 10 times
./ext4dist -m 5 # 5s summaries, milliseconds
"""
parser = argparse.ArgumentParser(
description="Summarize ext4 operation latency",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=examples)
parser.add_argument("-T", "--notimestamp", action="store_true",
help="don't include timestamp on interval output")
parser.add_argument("-m", "--milliseconds", action="store_true",
help="output in milliseconds")
parser.add_argument("-p", "--pid",
help="trace this PID only")
parser.add_argument("interval", nargs="?",
help="output interval, in seconds")
parser.add_argument("count", nargs="?", default=99999999,
help="number of outputs")
args = parser.parse_args()
pid = args.pid
countdown = int(args.count)
if args.milliseconds:
factor = 1000000
label = "msecs"
else:
factor = 1000
label = "usecs"
if args.interval and int(args.interval) == 0:
print("ERROR: interval 0. Exiting.")
exit()
debug = 0
# define BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#include <linux/fs.h>
#include <linux/sched.h>
#define OP_NAME_LEN 8
typedef struct dist_key {
char op[OP_NAME_LEN];
u64 slot;
} dist_key_t;
BPF_HASH(start, u32);
BPF_HISTOGRAM(dist, dist_key_t);
// time operation
int trace_entry(struct pt_regs *ctx)
{
u32 pid = bpf_get_current_pid_tgid();
if (FILTER_PID)
return 0;
u64 ts = bpf_ktime_get_ns();
start.update(&pid, &ts);
return 0;
}
// The current ext4 (Linux 4.5) uses generic_file_read_iter(), instead of it's
// own function, for reads. So we need to trace that and then filter on ext4,
// which I do by checking file->f_op.
int trace_read_entry(struct pt_regs *ctx, struct kiocb *iocb)
{
u32 pid = bpf_get_current_pid_tgid();
if (FILTER_PID)
return 0;
// ext4 filter on file->f_op == ext4_file_operations
struct file *fp = iocb->ki_filp;
if ((u64)fp->f_op != EXT4_FILE_OPERATIONS)
return 0;
u64 ts = bpf_ktime_get_ns();
start.update(&pid, &ts);
return 0;
}
static int trace_return(struct pt_regs *ctx, const char *op)
{
u64 *tsp;
u32 pid = bpf_get_current_pid_tgid();
// fetch timestamp and calculate delta
tsp = start.lookup(&pid);
if (tsp == 0) {
return 0; // missed start or filtered
}
u64 delta = (bpf_ktime_get_ns() - *tsp) / FACTOR;
// store as histogram
dist_key_t key = {.slot = bpf_log2l(delta)};
__builtin_memcpy(&key.op, op, sizeof(key.op));
dist.increment(key);
start.delete(&pid);
return 0;
}
int trace_read_return(struct pt_regs *ctx)
{
char *op = "read";
return trace_return(ctx, op);
}
int trace_write_return(struct pt_regs *ctx)
{
char *op = "write";
return trace_return(ctx, op);
}
int trace_open_return(struct pt_regs *ctx)
{
char *op = "open";
return trace_return(ctx, op);
}
int trace_fsync_return(struct pt_regs *ctx)
{
char *op = "fsync";
return trace_return(ctx, op);
}
"""
# code replacements
with open(kallsyms) as syms:
ops = ''
for line in syms:
(addr, size, name) = line.rstrip().split(" ", 2)
if name == "ext4_file_operations":
ops = "0x" + addr
break
if ops == '':
print("ERROR: no ext4_file_operations in /proc/kallsyms. Exiting.")
exit()
bpf_text = bpf_text.replace('EXT4_FILE_OPERATIONS', ops)
bpf_text = bpf_text.replace('FACTOR', str(factor))
if args.pid:
bpf_text = bpf_text.replace('FILTER_PID', 'pid != %s' % pid)
else:
bpf_text = bpf_text.replace('FILTER_PID', '0')
if debug:
print(bpf_text)
# load BPF program
b = BPF(text=bpf_text)
# Common file functions. See earlier comment about generic_file_read_iter().
b.attach_kprobe(event="generic_file_read_iter", fn_name="trace_read_entry")
b.attach_kprobe(event="ext4_file_write_iter", fn_name="trace_entry")
b.attach_kprobe(event="ext4_file_open", fn_name="trace_entry")
b.attach_kprobe(event="ext4_sync_file", fn_name="trace_entry")
b.attach_kretprobe(event="generic_file_read_iter", fn_name="trace_read_return")
b.attach_kretprobe(event="ext4_file_write_iter", fn_name="trace_write_return")
b.attach_kretprobe(event="ext4_file_open", fn_name="trace_open_return")
b.attach_kretprobe(event="ext4_sync_file", fn_name="trace_fsync_return")
print("Tracing ext4 operation latency... Hit Ctrl-C to end.")
# output
exiting = 0
dist = b.get_table("dist")
while (1):
try:
if args.interval:
sleep(int(args.interval))
else:
sleep(99999999)
except KeyboardInterrupt:
exiting = 1
print()
if args.interval and (not args.notimestamp):
print(strftime("%H:%M:%S:"))
dist.print_log2_hist(label, "operation")
dist.clear()
countdown -= 1
if exiting or countdown == 0:
exit()