aboutsummaryrefslogtreecommitdiff
path: root/tools/filetop.py
blob: aec11a86c230d3b4f5701222eb36cd78331e3420 (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
#!/usr/bin/python
# @lint-avoid-python-3-compatibility-imports
#
# filetop  file reads and writes by process.
#          For Linux, uses BCC, eBPF.
#
# USAGE: filetop.py [-h] [-C] [-r MAXROWS] [interval] [count]
#
# This uses in-kernel eBPF maps to store per process summaries for efficiency.
#
# Copyright 2016 Netflix, Inc.
# Licensed under the Apache License, Version 2.0 (the "License")
#
# 06-Feb-2016   Brendan Gregg   Created this.

from __future__ import print_function
from bcc import BPF
from time import sleep, strftime
import argparse
from subprocess import call

# arguments
examples = """examples:
    ./filetop            # file I/O top, 1 second refresh
    ./filetop -C         # don't clear the screen
    ./filetop -p 181     # PID 181 only
    ./filetop 5          # 5 second summaries
    ./filetop 5 10       # 5 second summaries, 10 times only
"""
parser = argparse.ArgumentParser(
    description="File reads and writes by process",
    formatter_class=argparse.RawDescriptionHelpFormatter,
    epilog=examples)
parser.add_argument("-a", "--all-files", action="store_true",
    help="include non-regular file types (sockets, FIFOs, etc)")
parser.add_argument("-C", "--noclear", action="store_true",
    help="don't clear the screen")
parser.add_argument("-r", "--maxrows", default=20,
    help="maximum rows to print, default 20")
parser.add_argument("-s", "--sort", default="all",
    choices=["all", "reads", "writes", "rbytes", "wbytes"],
    help="sort column, default all")
parser.add_argument("-p", "--pid", type=int, metavar="PID", dest="tgid",
    help="trace this PID only")
parser.add_argument("interval", nargs="?", default=1,
    help="output interval, in seconds")
parser.add_argument("count", nargs="?", default=99999999,
    help="number of outputs")
parser.add_argument("--ebpf", action="store_true",
    help=argparse.SUPPRESS)
args = parser.parse_args()
interval = int(args.interval)
countdown = int(args.count)
maxrows = int(args.maxrows)
clear = not int(args.noclear)
debug = 0

# linux stats
loadavg = "/proc/loadavg"

# define BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#include <linux/blkdev.h>

// the key for the output summary
struct info_t {
    unsigned long inode;
    dev_t dev;
    dev_t rdev;
    u32 pid;
    u32 name_len;
    char comm[TASK_COMM_LEN];
    // de->d_name.name may point to de->d_iname so limit len accordingly
    char name[DNAME_INLINE_LEN];
    char type;
};

// the value of the output summary
struct val_t {
    u64 reads;
    u64 writes;
    u64 rbytes;
    u64 wbytes;
};

BPF_HASH(counts, struct info_t, struct val_t);

static int do_entry(struct pt_regs *ctx, struct file *file,
    char __user *buf, size_t count, int is_read)
{
    u32 tgid = bpf_get_current_pid_tgid() >> 32;
    if (TGID_FILTER)
        return 0;

    u32 pid = bpf_get_current_pid_tgid();

    // skip I/O lacking a filename
    struct dentry *de = file->f_path.dentry;
    int mode = file->f_inode->i_mode;
    struct qstr d_name = de->d_name;
    if (d_name.len == 0 || TYPE_FILTER)
        return 0;

    // store counts and sizes by pid & file
    struct info_t info = {
        .pid = pid,
        .inode = file->f_inode->i_ino,
        .dev = file->f_inode->i_sb->s_dev,
        .rdev = file->f_inode->i_rdev,
    };
    bpf_get_current_comm(&info.comm, sizeof(info.comm));
    info.name_len = d_name.len;
    bpf_probe_read_kernel(&info.name, sizeof(info.name), d_name.name);
    if (S_ISREG(mode)) {
        info.type = 'R';
    } else if (S_ISSOCK(mode)) {
        info.type = 'S';
    } else {
        info.type = 'O';
    }

    struct val_t *valp, zero = {};
    valp = counts.lookup_or_try_init(&info, &zero);
    if (valp) {
        if (is_read) {
            valp->reads++;
            valp->rbytes += count;
        } else {
            valp->writes++;
            valp->wbytes += count;
        }
    }

    return 0;
}

int trace_read_entry(struct pt_regs *ctx, struct file *file,
    char __user *buf, size_t count)
{
    return do_entry(ctx, file, buf, count, 1);
}

int trace_write_entry(struct pt_regs *ctx, struct file *file,
    char __user *buf, size_t count)
{
    return do_entry(ctx, file, buf, count, 0);
}

"""
if args.tgid:
    bpf_text = bpf_text.replace('TGID_FILTER', 'tgid != %d' % args.tgid)
else:
    bpf_text = bpf_text.replace('TGID_FILTER', '0')
if args.all_files:
    bpf_text = bpf_text.replace('TYPE_FILTER', '0')
else:
    bpf_text = bpf_text.replace('TYPE_FILTER', '!S_ISREG(mode)')

if debug or args.ebpf:
    print(bpf_text)
    if args.ebpf:
        exit()

# initialize BPF
b = BPF(text=bpf_text)
b.attach_kprobe(event="vfs_read", fn_name="trace_read_entry")
b.attach_kprobe(event="vfs_write", fn_name="trace_write_entry")

DNAME_INLINE_LEN = 32  # linux/dcache.h

print('Tracing... Output every %d secs. Hit Ctrl-C to end' % interval)

def sort_fn(counts):
    if args.sort == "all":
        return (counts[1].rbytes + counts[1].wbytes + counts[1].reads + counts[1].writes)
    else:
        return getattr(counts[1], args.sort)

# output
exiting = 0
while 1:
    try:
        sleep(interval)
    except KeyboardInterrupt:
        exiting = 1

    # header
    if clear:
        call("clear")
    else:
        print()
    with open(loadavg) as stats:
        print("%-8s loadavg: %s" % (strftime("%H:%M:%S"), stats.read()))
    print("%-7s %-16s %-6s %-6s %-7s %-7s %1s %s" % ("TID", "COMM",
        "READS", "WRITES", "R_Kb", "W_Kb", "T", "FILE"))

    # by-TID output
    counts = b.get_table("counts")
    line = 0
    for k, v in reversed(sorted(counts.items(),
                                key=sort_fn)):
        name = k.name.decode('utf-8', 'replace')
        if k.name_len > DNAME_INLINE_LEN:
            name = name[:-3] + "..."

        # print line
        print("%-7d %-16s %-6d %-6d %-7d %-7d %1s %s" % (k.pid,
            k.comm.decode('utf-8', 'replace'), v.reads, v.writes,
            v.rbytes / 1024, v.wbytes / 1024,
            k.type.decode('utf-8', 'replace'), name))

        line += 1
        if line >= maxrows:
            break
    counts.clear()

    countdown -= 1
    if exiting or countdown == 0:
        print("Detaching...")
        exit()