You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

258 lines
7.5 KiB

#!/usr/bin/env python
# @lint-avoid-python-3-compatibility-imports
#
# opensnoop Trace open() syscalls.
# For Linux, uses BCC, eBPF. Embedded C.
#
# USAGE: opensnoop [-h] [-T] [-x] [-p PID] [-d DURATION] [-t TID] [-n NAME]
#
# Copyright (c) 2015 Brendan Gregg.
# Licensed under the Apache License, Version 2.0 (the "License")
#
# 17-Sep-2015 Brendan Gregg Created this.
# 29-Apr-2016 Allan McAleavy Updated for BPF_PERF_OUTPUT.
# 08-Oct-2016 Dina Goldshtein Support filtering by PID and TID.
# 28-Dec-2018 Tim Douglas Print flags argument, enable filtering
# 06-Jan-2019 Takuma Kume Support filtering by UID
from __future__ import print_function
from bcc import ArgString, BPF
from bcc.utils import printb
import argparse
import ctypes as ct
from datetime import datetime, timedelta
import os
# arguments
examples = """examples:
./opensnoop # trace all open() syscalls
./opensnoop -T # include timestamps
./opensnoop -U # include UID
./opensnoop -x # only show failed opens
./opensnoop -p 181 # only trace PID 181
./opensnoop -t 123 # only trace TID 123
./opensnoop -u 1000 # only trace UID 1000
./opensnoop -d 10 # trace for 10 seconds only
./opensnoop -n main # only print process names containing "main"
./opensnoop -e # show extended fields
./opensnoop -f O_WRONLY -f O_RDWR # only print calls for writing
"""
parser = argparse.ArgumentParser(
description="Trace open() syscalls",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=examples)
parser.add_argument("-T", "--timestamp", action="store_true",
help="include timestamp on output")
parser.add_argument("-U", "--print-uid", action="store_true",
help="print UID column")
parser.add_argument("-x", "--failed", action="store_true",
help="only show failed opens")
parser.add_argument("-p", "--pid",
help="trace this PID only")
parser.add_argument("-t", "--tid",
help="trace this TID only")
parser.add_argument("-u", "--uid",
help="trace this UID only")
parser.add_argument("-d", "--duration",
help="total duration of trace in seconds")
parser.add_argument("-n", "--name",
type=ArgString,
help="only print process names containing this name")
parser.add_argument("--ebpf", action="store_true",
help=argparse.SUPPRESS)
parser.add_argument("-e", "--extended_fields", action="store_true",
help="show extended fields")
parser.add_argument("-f", "--flag_filter", action="append",
help="filter on flags argument (e.g., O_WRONLY)")
args = parser.parse_args()
debug = 0
if args.duration:
args.duration = timedelta(seconds=int(args.duration))
flag_filter_mask = 0
for flag in args.flag_filter or []:
if not flag.startswith('O_'):
exit("Bad flag: %s" % flag)
try:
flag_filter_mask |= getattr(os, flag)
except AttributeError:
exit("Bad flag: %s" % flag)
# define BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#include <uapi/linux/limits.h>
#include <linux/sched.h>
struct val_t {
u64 id;
char comm[TASK_COMM_LEN];
const char *fname;
int flags; // EXTENDED_STRUCT_MEMBER
};
struct data_t {
u64 id;
u64 ts;
u32 uid;
int ret;
char comm[TASK_COMM_LEN];
char fname[NAME_MAX];
int flags; // EXTENDED_STRUCT_MEMBER
};
BPF_HASH(infotmp, u64, struct val_t);
BPF_PERF_OUTPUT(events);
int trace_entry(struct pt_regs *ctx, int dfd, const char __user *filename, int flags)
{
struct val_t val = {};
u64 id = bpf_get_current_pid_tgid();
u32 pid = id >> 32; // PID is higher part
u32 tid = id; // Cast and get the lower part
u32 uid = bpf_get_current_uid_gid();
PID_TID_FILTER
UID_FILTER
FLAGS_FILTER
if (bpf_get_current_comm(&val.comm, sizeof(val.comm)) == 0) {
val.id = id;
val.fname = filename;
val.flags = flags; // EXTENDED_STRUCT_MEMBER
infotmp.update(&id, &val);
}
return 0;
};
int trace_return(struct pt_regs *ctx)
{
u64 id = bpf_get_current_pid_tgid();
struct val_t *valp;
struct data_t data = {};
u64 tsp = bpf_ktime_get_ns();
valp = infotmp.lookup(&id);
if (valp == 0) {
// missed entry
return 0;
}
bpf_probe_read(&data.comm, sizeof(data.comm), valp->comm);
bpf_probe_read(&data.fname, sizeof(data.fname), (void *)valp->fname);
data.id = valp->id;
data.ts = tsp / 1000;
data.uid = bpf_get_current_uid_gid();
data.flags = valp->flags; // EXTENDED_STRUCT_MEMBER
data.ret = PT_REGS_RC(ctx);
events.perf_submit(ctx, &data, sizeof(data));
infotmp.delete(&id);
return 0;
}
"""
if args.tid: # TID trumps PID
bpf_text = bpf_text.replace('PID_TID_FILTER',
'if (tid != %s) { return 0; }' % args.tid)
elif args.pid:
bpf_text = bpf_text.replace('PID_TID_FILTER',
'if (pid != %s) { return 0; }' % args.pid)
else:
bpf_text = bpf_text.replace('PID_TID_FILTER', '')
if args.uid:
bpf_text = bpf_text.replace('UID_FILTER',
'if (uid != %s) { return 0; }' % args.uid)
else:
bpf_text = bpf_text.replace('UID_FILTER', '')
if args.flag_filter:
bpf_text = bpf_text.replace('FLAGS_FILTER',
'if (!(flags & %d)) { return 0; }' % flag_filter_mask)
else:
bpf_text = bpf_text.replace('FLAGS_FILTER', '')
if not (args.extended_fields or args.flag_filter):
bpf_text = '\n'.join(x for x in bpf_text.split('\n')
if 'EXTENDED_STRUCT_MEMBER' not in x)
if debug or args.ebpf:
print(bpf_text)
if args.ebpf:
exit()
# initialize BPF
b = BPF(text=bpf_text)
b.attach_kprobe(event="do_sys_open", fn_name="trace_entry")
b.attach_kretprobe(event="do_sys_open", fn_name="trace_return")
TASK_COMM_LEN = 16 # linux/sched.h
NAME_MAX = 255 # linux/limits.h
class Data(ct.Structure):
_fields_ = [
("id", ct.c_ulonglong),
("ts", ct.c_ulonglong),
("uid", ct.c_uint32),
("ret", ct.c_int),
("comm", ct.c_char * TASK_COMM_LEN),
("fname", ct.c_char * NAME_MAX),
("flags", ct.c_int),
]
initial_ts = 0
# header
if args.timestamp:
print("%-14s" % ("TIME(s)"), end="")
if args.print_uid:
print("%-6s" % ("UID"), end="")
print("%-6s %-16s %4s %3s " %
("TID" if args.tid else "PID", "COMM", "FD", "ERR"), end="")
if args.extended_fields:
print("%-9s" % ("FLAGS"), end="")
print("PATH")
# process event
def print_event(cpu, data, size):
event = ct.cast(data, ct.POINTER(Data)).contents
global initial_ts
# split return value into FD and errno columns
if event.ret >= 0:
fd_s = event.ret
err = 0
else:
fd_s = -1
err = - event.ret
if not initial_ts:
initial_ts = event.ts
if args.failed and (event.ret >= 0):
return
if args.name and bytes(args.name) not in event.comm:
return
if args.timestamp:
delta = event.ts - initial_ts
print("%-14.9f" % (float(delta) / 1000000), end="")
if args.print_uid:
print("%-6d" % event.uid, end="")
print("%-6d %-16s %4d %3d " %
(event.id & 0xffffffff if args.tid else event.id >> 32,
event.comm.decode('utf-8', 'replace'), fd_s, err), end="")
if args.extended_fields:
print("%08o " % event.flags, end="")
printb(b'%s' % event.fname)
# loop with callback to print_event
b["events"].open_perf_buffer(print_event, page_cnt=64)
start_time = datetime.now()
while not args.duration or datetime.now() - start_time < args.duration:
try:
b.perf_buffer_poll()
except KeyboardInterrupt:
exit()