Skip to content

Latest commit

 

History

History
2517 lines (1951 loc) · 72 KB

reference_guide.md

File metadata and controls

2517 lines (1951 loc) · 72 KB
Raw

bpftrace Reference Guide

For a reference summary, see the README.md for the sections on Probe types and Builtins.

This is a work in progress. If something is missing, check the bpftrace source to see if these docs are just out of date. And if you find something, please file an issue or pull request to update these docs. Also, please keep these docs as terse as possible to maintain it's brevity (inspired by the 6-page awk summary from page 106 of v7vol2b.pdf). Leave longer examples and discussion to other files in /docs, the /tools/*_examples.txt files, or blog posts and other articles.

Contents

Terminology

Term Description
BPF Berkely Packet Filter: a kernel technology originally developed for optimizing the processing of packet filters (eg, tcpdump expressions)
eBPF Enhanced BPF: a kernel technology that extends BPF so that it can execute more generic programs on any events, such as the bpftrace programs listed below. It makes use of the BPF sandboxed virtual machine environment. Also note that eBPF is often just referred to as BPF.
probe An instrumentation point in software or hardware, that generates events that can execute bpftrace programs.
static tracing Hard-coded instrumentation points in code. Since these are fixed, they may be provided as part of a stable API, and documented.
dynamic tracing Also known as dynamic instrumentation, this is a technology that can instrument any software event, such as function calls and returns, by live modification of instruction text. Target software usually does not need special capabilities to support dynamic tracing, other than a symbol table that bpftrace can read. Since this instruments all software text, it is not considered a stable API, and the target functions may not be documented outside of their source code.
tracepoints A Linux kernel technology for providing static tracing.
kprobes A Linux kernel technology for providing dynamic tracing of kernel functions.
uprobes A Linux kernel technology for providing dynamic tracing of user-level functions.
USDT User Statically-Defined Tracing: static tracing points for user-level software. Some applications support USDT.
BPF map A BPF memory object, which is used by bpftrace to create many higher-level objects.

Usage

Command line usage is summarized by bpftrace without options:

# bpftrace
USAGE:
    bpftrace [options] filename
    bpftrace [options] -e 'program'

OPTIONS:
    -B MODE        output buffering mode ('line', 'full', or 'none')
    -d             debug info dry run
    -dd            verbose debug info dry run
    -e 'program'   execute this program
    -h             show this help message
    -I DIR         add the specified DIR to the search path for include files.
    --include FILE adds an implicit #include which is read before the source file is preprocessed.
    -l [search]    list probes
    -p PID         enable USDT probes on PID
    -c 'CMD'       run CMD and enable USDT probes on resulting process
    -v             verbose messages
    --version      bpftrace version

ENVIRONMENT:
    BPFTRACE_STRLEN           [default: 64] bytes on BPF stack per str()
    BPFTRACE_NO_CPP_DEMANGLE  [default: 0] disable C++ symbol demangling
    BPFTRACE_MAP_KEYS_MAX     [default: 4096] max keys in a map
    BPFTRACE_MAX_PROBES       [default: 512] max number of probes bpftrace can attach to

EXAMPLES:
bpftrace -l '*sleep*'
    list probes containing "sleep"
bpftrace -e 'kprobe:do_nanosleep { printf("PID %d sleeping...\n", pid); }'
    trace processes calling sleep
bpftrace -e 'tracepoint:raw_syscalls:sys_enter { @[comm] = count(); }'
    count syscalls by process name

1. Hello World

The most basic example of a bpftrace program:

# bpftrace -e 'BEGIN { printf("Hello, World!\n"); }'
Attaching 1 probe...
Hello, World!
^C

The syntax to this program will be explained in the Language section. In this section, we'll cover tool usage.

A program will continue running until Ctrl-C is hit, or an exit() function is called. When a program exits, all populated maps are printed: this behavior, and maps, are explained in later sections.

2. -e 'program': One-Liners

The -e option allows a program to be specified, and is a way to construct one-liners:

# bpftrace -e 'tracepoint:syscalls:sys_enter_nanosleep { printf("%s is sleeping.\n", comm); }'
Attaching 1 probe...
iscsid is sleeping.
irqbalance is sleeping.
iscsid is sleeping.
iscsid is sleeping.
[...]

This example is printing when processes call the nanosleep syscall. Again, the syntax of the program will be explained in the Language section.

3. filename: Program Files

Programs saved as files are often called scripts, and can be executed by specifying their file name. We'll often use a .bt file extension, short for bpftrace, but the extension is ignored.

For example, listing the sleepers.bt file using cat -n (which enumerates the output lines):

# cat -n sleepers.bt
     1	tracepoint:syscalls:sys_enter_nanosleep
     2	{
     3		printf("%s is sleeping.\n", comm);
     4	}

Running sleepers.bt:

# bpftrace sleepers.bt
Attaching 1 probe...
iscsid is sleeping.
iscsid is sleeping.
[...]

It can also be made executable to run stand-alone. Start by adding an interpreter line at the top (#!) with either the path to your installed bpftrace (/usr/local/bin is the default) or the path to env (usually just /usr/bin/env) followed by bpftrace (so it will find bpftrace in your $PATH):

     1	#!/usr/local/bin/bpftrace
     1	#!/usr/bin/env bpftrace
     2
     3	tracepoint:syscalls:sys_enter_nanosleep
     4	{
     5	        printf("%s is sleeping.\n", comm);
     6	}

Then make it executable:

# chmod 755 sleepers.bt
# ./sleepers.bt
Attaching 1 probe...
iscsid is sleeping.
iscsid is sleeping.
[...]

4. -l: Listing Probes

Probes from the tracepoint and kprobe libraries can be listed with -l.

# bpftrace -l | more
tracepoint:xfs:xfs_attr_list_sf
tracepoint:xfs:xfs_attr_list_sf_all
tracepoint:xfs:xfs_attr_list_leaf
tracepoint:xfs:xfs_attr_list_leaf_end
[...]
# bpftrace -l | wc -l
46260

Other libraries generate probes dynamically, such as uprobe, and require specific ways to determine available probes. See the later Probes sections.

Search terms can be added:

# bpftrace -l '*nanosleep*'
tracepoint:syscalls:sys_enter_clock_nanosleep
tracepoint:syscalls:sys_exit_clock_nanosleep
tracepoint:syscalls:sys_enter_nanosleep
tracepoint:syscalls:sys_exit_nanosleep
kprobe:nanosleep_copyout
kprobe:hrtimer_nanosleep
[...]

The -v option when listing tracepoints will show their arguments for use from the args builtin. For example:

# bpftrace -lv tracepoint:syscalls:sys_enter_open
tracepoint:syscalls:sys_enter_open
    int __syscall_nr;
    const char * filename;
    int flags;
    umode_t mode;

5. -d: Debug Output

The -d option produces debug output, and does not run the program. This is mostly useful for debugging issues with bpftrace itself. You can also use -dd to produce a more verbose debug output, which will also print unoptimized IR.

If you are an end-user of bpftrace, you should not normally need the -d or -v options, and you can skip to the Language section.

# bpftrace -d -e 'tracepoint:syscalls:sys_enter_nanosleep { printf("%s is sleeping.\n", comm); }'
Program
 tracepoint:syscalls:sys_enter_nanosleep
  call: printf
   string: %s is sleeping.\n
   builtin: comm
[...]

The output begins with Program and then an abstract syntax tree (AST) representation of the program.

Continued:

[...]
%printf_t = type { i64, [16 x i8] }
[...]
define i64 @"tracepoint:syscalls:sys_enter_nanosleep"(i8*) local_unnamed_addr section "s_tracepoint:syscalls:sys_enter_nanosleep" {
entry:
  %comm = alloca [16 x i8], align 1
  %printf_args = alloca %printf_t, align 8
  %1 = bitcast %printf_t* %printf_args to i8*
  call void @llvm.lifetime.start.p0i8(i64 -1, i8* nonnull %1)
  %2 = getelementptr inbounds [16 x i8], [16 x i8]* %comm, i64 0, i64 0
  %3 = bitcast %printf_t* %printf_args to i8*
  call void @llvm.memset.p0i8.i64(i8* nonnull %3, i8 0, i64 24, i32 8, i1 false)
  call void @llvm.lifetime.start.p0i8(i64 -1, i8* nonnull %2)
  call void @llvm.memset.p0i8.i64(i8* nonnull %2, i8 0, i64 16, i32 1, i1 false)
  %get_comm = call i64 inttoptr (i64 16 to i64 (i8*, i64)*)([16 x i8]* nonnull %comm, i64 16)
  %4 = getelementptr inbounds %printf_t, %printf_t* %printf_args, i64 0, i32 1, i64 0
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull %4, i8* nonnull %2, i64 16, i32 1, i1 false)
  %pseudo = call i64 @llvm.bpf.pseudo(i64 1, i64 1)
  %get_cpu_id = call i64 inttoptr (i64 8 to i64 ()*)()
  %perf_event_output = call i64 inttoptr (i64 25 to i64 (i8*, i8*, i64, i8*, i64)*)(i8* %0, i64 %pseudo, i64 %get_cpu_id, %printf_t* nonnull %printf_args, i64 24)
  call void @llvm.lifetime.end.p0i8(i64 -1, i8* nonnull %1)
  ret i64 0
[...]

This section shows the llvm intermediate representation (IR) assembly, which is then compiled into BPF.

6. -v: Verbose Output

The -v option prints more information about the program as it is run:

# bpftrace -v -e 'tracepoint:syscalls:sys_enter_nanosleep { printf("%s is sleeping.\n", comm); }'
Attaching 1 probe...

Bytecode:
0: (bf) r6 = r1
1: (b7) r1 = 0
2: (7b) *(u64 *)(r10 -24) = r1
3: (7b) *(u64 *)(r10 -32) = r1
4: (7b) *(u64 *)(r10 -40) = r1
5: (7b) *(u64 *)(r10 -8) = r1
6: (7b) *(u64 *)(r10 -16) = r1
7: (bf) r1 = r10
8: (07) r1 += -16
9: (b7) r2 = 16
10: (85) call bpf_get_current_comm#16
11: (79) r1 = *(u64 *)(r10 -16)
12: (7b) *(u64 *)(r10 -32) = r1
13: (79) r1 = *(u64 *)(r10 -8)
14: (7b) *(u64 *)(r10 -24) = r1
15: (18) r7 = 0xffff9044e65f1000
17: (85) call bpf_get_smp_processor_id#8
18: (bf) r4 = r10
19: (07) r4 += -40
20: (bf) r1 = r6
21: (bf) r2 = r7
22: (bf) r3 = r0
23: (b7) r5 = 24
24: (85) call bpf_perf_event_output#25
25: (b7) r0 = 0
26: (95) exit
processed 26 insns (limit 131072), stack depth 40

Attaching tracepoint:syscalls:sys_enter_nanosleep
Running...
iscsid is sleeping.
iscsid is sleeping.
[...]

This includes Bytecode: and then the eBPF bytecode after it was compiled from the llvm assembly.

7. Preprocessor Options

The -I option can be used to add directories to the list of directories that bpftrace uses to look for headers. Can be defined multiple times.

# cat program.bt
#include <foo.h>

BEGIN { @ = FOO }

# bpftrace program.bt
definitions.h:1:10: fatal error: 'foo.h' file not found

# /tmp/include
foo.h

# bpftrace -I /tmp/include program.bt
Attaching 1 probe...

The --include option can be used to include headers by default. Can be defined multiple times. Headers are included in the order they are defined, and they are included before any other include in the program being executed.

# bpftrace --include linux/path.h --include linux/dcache.h -e 'kprobe:vfs_open { printf("open path: %s\n", str(((path *)arg0)->dentry->d_name.name)); }'
Attaching 1 probe...
open path: .com.google.Chrome.ASsbu2
open path: .com.google.Chrome.gimc10
open path: .com.google.Chrome.R1234s

8. Other Options

The --version option prints the bpftrace version:

# bpftrace --version
bpftrace v0.8-90-g585e-dirty

9. Environment Variables

9.1 BPFTRACE_STRLEN

Default: 64

Number of bytes allocated on the BPF stack for the string returned by str().

Make this larger if you wish to read bigger strings with str().

Beware that the BPF stack is small (512 bytes), and that you pay the toll again inside printf() (whilst it composes a perf event output buffer). So in practice you can only grow this to about 200 bytes.

Support for even larger strings is being discussed.

9.2 BPFTRACE_NO_CPP_DEMANGLE

Default: 0

C++ symbol demangling in userspace stack traces is enabled by default.

This feature can be turned off by setting the value of this environment variable to 1.

9.3 BPFTRACE_MAP_KEYS_MAX

Default: 4096

This is the maximum number of keys that can be stored in a map. Increasing the value will consume more memory and increase startup times. There are some cases where you will want to: for example, sampling stack traces, recording timestamps for each page, etc.

9.4 BPFTRACE_MAX_PROBES

Default: 512

This is the maximum number of probes that bpftrace can attach to. Increasing the value will consume more memory, increase startup times and can incur high performance overhead or even freeze or crash the system.

10. Clang Environment Variables

bpftrace parses header files using libclang, the C interface to Clang. Thus environment variables affecting the clang toolchain can be used. For example, if header files are included from a non-default directory, the CPATH or C_INCLUDE_PATH environment variables can be set to allow clang to locate the files. See clang documentation for more information on these environment variables and their usage.

Language

1. {...}: Action Blocks

Syntax: probe[,probe,...] /filter/ { action }

A bpftrace program can have multiple action blocks. The filter is optional.

Example:

# bpftrace -e 'kprobe:do_sys_open { printf("opening: %s\n", str(arg1)); }'
Attaching 1 probe...
opening: /proc/cpuinfo
opening: /proc/stat
opening: /proc/diskstats
opening: /proc/stat
opening: /proc/vmstat
[...]

This is a one-liner invocation of bpftrace. The probe is kprobe:do_sys_open. When that probe "fires" (the instrumentation event occurred) the action will be executed, which consists of a print() statement. Explanations of the probe and action are in the sections that follow.

2. /.../: Filtering

Syntax: /filter/

Filters (also known as predicates) can be added after probe names. The probe still fires, but it will skip the action unless the filter is true.

Examples:

# bpftrace -e 'kprobe:vfs_read /arg2 < 16/ { printf("small read: %d byte buffer\n", arg2); }'
Attaching 1 probe...
small read: 8 byte buffer
small read: 8 byte buffer
small read: 8 byte buffer
small read: 8 byte buffer
small read: 8 byte buffer
small read: 12 byte buffer
^C

# bpftrace -e 'kprobe:vfs_read /comm == "bash"/ { printf("read by %s\n", comm); }'
Attaching 1 probe...
read by bash
read by bash
read by bash
read by bash
^C

3. //, /*: Comments

Syntax

// single-line comment

/*
 * multi-line comment
 */

These can be used in bpftrace scripts to document your code.

4. ->: C Struct Navigation

tracepoint example:

# bpftrace -e 'tracepoint:syscalls:sys_enter_openat { printf("%s %s\n", comm, str(args->filename)); }'
Attaching 1 probe...
snmpd /proc/diskstats
snmpd /proc/stat
snmpd /proc/vmstat
[...]

This is returning the filename member from the args struct, which for tracepoint probes contains the tracepoint arguments. See the Static Tracing, Kernel-Level Arguments section for the contents of this struct.

kprobe example:

# cat path.bt
#include <linux/path.h>
#include <linux/dcache.h>

kprobe:vfs_open
{
	printf("open path: %s\n", str(((struct path *)arg0)->dentry->d_name.name));
}

# bpftrace path.bt
Attaching 1 probe...
open path: dev
open path: if_inet6
open path: retrans_time_ms
[...]

This uses dynamic tracing of the vfs_open() kernel function, via the short script path.bt. Some kernel headers needed to be included to understand the path and dentry structs.

5. struct: Struct Declaration

Example:

// from fs/namei.c:
struct nameidata {
        struct path     path;
        struct qstr     last;
        // [...]
};

You can define your own structs when needed. In some cases, kernel structs are not declared in the kernel headers package, and are declared manually in bpftrace tools (or partial structs are: enough to reach the member to dereference).

6. ? :: ternary operators

Example:

# bpftrace -e 'tracepoint:syscalls:sys_exit_read { @error[args->ret < 0 ? - args->ret : 0] = count(); }'
Attaching 1 probe...
^C

@error[11]: 24
@error[0]: 78

7. if () {...} else {...}: if-else statements

Example:

# bpftrace -e 'tracepoint:syscalls:sys_enter_read { @reads = count(); if (args->count > 1024) { @large = count(); } }'
Attaching 1 probe...
^C
@large: 72

@reads: 80

8. unroll () {...}: Unroll

Example:

# bpftrace -e 'kprobe:do_nanosleep { $i = 1; unroll(5) { printf("i: %d\n", $i); $i = $i + 1; } }'
Attaching 1 probe...
i: 1
i: 2
i: 3
i: 4
i: 5
^C

9. ++ and --: Increment operators

++ and -- can be used to conveniently increment or decrement counters in maps or variables.

Note that maps will be implictly declared and initalized to 0 if not already declared or defined. Scratch variables must be initalized before using these operators.

Example - variable:

bpftrace -e 'BEGIN { $x = 0; $x++; $x++; printf("x: %d\n", $x); }'
Attaching 1 probe...
x: 2
^C

Example - map:

bpftrace -e 'k:vfs_read { @++ }'
Attaching 1 probe...
^C

@: 12807

Example - map with key:

# bpftrace -e 'k:vfs_read { @[probe]++ }'
Attaching 1 probe...
^C

@[kprobe:vfs_read]: 13369

10. []: Array Access

You may access one-dimensional constant arrays with the array acccess operator [].

Example:

# bpftrace -e 'struct MyStruct { int y[4]; } uprobe:./testprogs/array_access:test_struct { $s = (struct MyStruct *) arg0; @x = $s->y[0]; exit(); }'
Attaching 1 probe...

@x: 1

Probes

  • kprobe - kernel function start
  • kretprobe - kernel function return
  • uprobe - user-level function start
  • uretprobe - user-level function return
  • tracepoint - kernel static tracepoints
  • usdt - user-level static tracepoints
  • profile - timed sampling
  • interval - timed output
  • software - kernel software events
  • hardware - processor-level events

Some probe types allow wildcards to match multiple probes, eg, kprobe:vfs_*. You may also specify multiple attach points for an action block using a comma separated list.

1. kprobe/kretprobe: Dynamic Tracing, Kernel-Level

Syntax:

kprobe:function_name
kretprobe:function_name

These use kprobes (a Linux kernel capability). kprobe instruments the beginning of a function's execution, and kretprobe instruments the end (its return).

Examples:

# bpftrace -e 'kprobe:do_nanosleep { printf("sleep by %d\n", tid); }'
Attaching 1 probe...
sleep by 1396
sleep by 3669
sleep by 1396
sleep by 27662
sleep by 3669
^C

2. kprobe/kretprobe: Dynamic Tracing, Kernel-Level Arguments

Syntax:

kprobe: arg0, arg1, ..., argN
kretprobe: retval

Arguments can be accessed via these variables names. arg0 is the first argument and can only be accessed with a kprobe. retval is the return value for the instrumented function, and can only be accessed on kretprobe.

Examples:

# bpftrace -e 'kprobe:do_sys_open { printf("opening: %s\n", str(arg1)); }'
Attaching 1 probe...
opening: /proc/cpuinfo
opening: /proc/stat
opening: /proc/diskstats
opening: /proc/stat
opening: /proc/vmstat
[...]

# bpftrace -e 'kprobe:do_sys_open { printf("open flags: %d\n", arg2); }'
Attaching 1 probe...
open flags: 557056
open flags: 32768
open flags: 32768
open flags: 32768
[...]

# bpftrace -e 'kretprobe:do_sys_open { printf("returned: %d\n", retval); }'
Attaching 1 probe...
returned: 8
returned: 21
returned: -2
returned: 21
[...]

As an example of struct arguments:

# cat path.bt
#include <linux/path.h>
#include <linux/dcache.h>

kprobe:vfs_open
{
	printf("open path: %s\n", str(((struct path *)arg0)->dentry->d_name.name));
}

# bpftrace path.bt
Attaching 1 probe...
open path: dev
open path: if_inet6
open path: retrans_time_ms
[...]

Here arg0 was casted as a (struct path *), since that is the first argument to vfs_open(). The struct support is currently the same as bcc, and based on available kernel headers. This means that many, but not all, structs will be available, and you may need to manually define some structs. In the future, bpftrace will use the newer Linux BTF support so that all kernel structs are always available.

3. uprobe/uretprobe: Dynamic Tracing, User-Level

Syntax:

uprobe:library_name:function_name
uretprobe:library_name:function_name

These use uprobes (a Linux kernel capability). uprobe instruments the beginning of a user-level function's execution, and uretprobe instruments the end (its return).

To list available uprobes, you can use any program to list the text segment symbols from a binary, such as objdump and nm. For example:

# objdump -tT /bin/bash | grep readline
00000000007003f8 g    DO .bss	0000000000000004  Base        rl_readline_state
0000000000499e00 g    DF .text	00000000000001c5  Base        readline_internal_char
00000000004993d0 g    DF .text	0000000000000126  Base        readline_internal_setup
000000000046d400 g    DF .text	000000000000004b  Base        posix_readline_initialize
000000000049a520 g    DF .text	0000000000000081  Base        readline
[...]

This has listed various functions containing "readline" from /bin/bash. These can be instrumented using uprobe and uretprobe.

Examples:

# bpftrace -e 'uretprobe:/bin/bash:readline { printf("read a line\n"); }'
Attaching 1 probe...
read a line
read a line
read a line
^C

While tracing, this has caught a few executions of the readline() function in /bin/bash. This example is continued in the next section.

4. uprobe/uretprobe: Dynamic Tracing, User-Level Arguments

Syntax:

uprobe: arg0, arg1, ..., argN
uretprobe: retval

Arguments can be accessed via these variables names. arg0 is the first argument, and can only be accessed with a uprobe. retval is the return value for the instrumented function, and can only be accessed on uretprobe.

Examples:

# bpftrace -e 'uprobe:/bin/bash:readline { printf("arg0: %d\n", arg0); }'
Attaching 1 probe...
arg0: 19755784
arg0: 19755016
arg0: 19755784
^C

What does arg0 of readline() in /bin/bash contain? I don't know. I'd need to look at the bash source code to find out what its arguments were.

# bpftrace -e 'uprobe:/lib/x86_64-linux-gnu/libc-2.23.so:fopen { printf("fopen: %s\n", str(arg0)); }'
Attaching 1 probe...
fopen: /proc/filesystems
fopen: /usr/share/locale/locale.alias
fopen: /proc/self/mountinfo
^C

In this case, I know that the first argument of libc fopen() is the pathname (see the fopen(3) man page), so I've traced it using a uprobe. Adjust the path to libc to match your system (it may not be libc-2.23.so). A str() call is necessary to turn the char * pointer to a string, as explained in a later section.

# bpftrace -e 'uretprobe:/bin/bash:readline { printf("readline: \"%s\"\n", str(retval)); }'
Attaching 1 probe...
readline: "echo hi"
readline: "ls -l"
readline: "date"
readline: "uname -r"
^C

Back to the bash readline() example: after checking the source code, I saw that the return value was the string read. So I can use a uretprobe and the retval variable to see the read string.

5. tracepoint: Static Tracing, Kernel-Level

Syntax: tracepoint:name

These use tracepoints (a Linux kernel capability).

# bpftrace -e 'tracepoint:block:block_rq_insert { printf("block I/O created by %d\n", tid); }'
Attaching 1 probe...
block I/O created by 28922
block I/O created by 3949
block I/O created by 883
block I/O created by 28941
block I/O created by 28941
block I/O created by 28941
[...]

6. tracepoint: Static Tracing, Kernel-Level Arguments

Example:

# bpftrace -e 'tracepoint:syscalls:sys_enter_openat { printf("%s %s\n", comm, str(args->filename)); }'
Attaching 1 probe...
irqbalance /proc/interrupts
irqbalance /proc/stat
snmpd /proc/diskstats
snmpd /proc/stat
snmpd /proc/vmstat
snmpd /proc/net/dev
[...]

The available members for each tracepoint can be listed from their /format file in /sys. For example:

# cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_open/format
name: sys_enter_openat
ID: 608
format:
	field:unsigned short common_type;	offset:0;	size:2;	signed:0;
	field:unsigned char common_flags;	offset:2;	size:1;	signed:0;
	field:unsigned char common_preempt_count;	offset:3;	size:1;	signed:0;
	field:int common_pid;	offset:4;	size:4;	signed:1;

	field:int __syscall_nr;	offset:8;	size:4;	signed:1;
	field:int dfd;	offset:16;	size:8;	signed:0;
	field:const char * filename;	offset:24;	size:8;	signed:0;
	field:int flags;	offset:32;	size:8;	signed:0;
	field:umode_t mode;	offset:40;	size:8;	signed:0;

print fmt: "dfd: 0x%08lx, filename: 0x%08lx, flags: 0x%08lx, mode: 0x%08lx", ((unsigned long)(REC->dfd)), ((unsigned long)(REC->filename)), ((unsigned long)(REC->flags)), ((unsigned long)(REC->mode))

Apart from the filename member, we can also print flags, mode, and more. After the "common" members listed first, the members are specific to the tracepoint.

7. usdt: Static Tracing, User-Level

Syntax:

usdt:binary_path:probe_name
usdt:binary_path:[probe_namespace]:probe_name
usdt:library_path:probe_name
usdt:library_path:[probe_namespace]:probe_name

Where the probe_namespace is optional, and will default to the basename of the binary or library path.

Examples:

# bpftrace -e 'usdt:/root/tick:loop { printf("hi\n"); }'
Attaching 1 probe...
hi
hi
hi
hi
hi
^C

The basename of a path will be used for the namespace of a probe. If it doesn't match, the probe won't be found. In this example, the function name loop is in the namespace tick. If we rename the binary to tock, it won't be found:

mv /root/tick /root/tock
bpftrace -e 'usdt:/root/tock:loop { printf("hi\n"); }'
Attaching 1 probe...
Error finding location for probe: usdt:/root/tock:loop

The probe namespace can be manually specified, between the path and probe function name. This allows for the probe to be found, regardless of the name of the binary:

bpftrace -e 'usdt:/root/tock:tick:loop { printf("hi\n"); }'

8. usdt: Static Tracing, User-Level Arguments

Examples:

# bpftrace -e 'usdt:/root/tick:loop { printf("%s: %d\n", str(arg0), arg1); }'
my string: 1
my string: 2
my string: 3
my string: 4
my string: 5
^C

# bpftrace -e 'usdt:/root/tick:loop /arg1 > 2/ { printf("%s: %d\n", str(arg0), arg1); }'
my string: 3
my string: 4
my string: 5
my string: 6
^C

9. profile: Timed Sampling Events

Syntax:

profile:hz:rate
profile:s:rate
profile:ms:rate
profile:us:rate

These operating using perf_events (a Linux kernel facility), which is also used by the perf command).

Examples:

# bpftrace -e 'profile:hz:99 { @[tid] = count(); }'
Attaching 1 probe...
^C

@[32586]: 98
@[0]: 579

10. interval: Timed Output

Syntax:

interval:ms:rate
interval:s:rate

This fires on one CPU only, and can be used for generating per-interval output.

Example:

# bpftrace -e 'tracepoint:raw_syscalls:sys_enter { @syscalls = count(); }
    interval:s:1 { print(@syscalls); clear(@syscalls); }'
Attaching 2 probes...
@syscalls: 1263
@syscalls: 731
@syscalls: 891
@syscalls: 1195
@syscalls: 1154
@syscalls: 1635
@syscalls: 1208
[...]

This prints the rate of syscalls per second.

11. software: Pre-defined Software Events

Syntax:

software:event_name:count
software:event_name:

These are the pre-defined software events provided by the Linux kernel, as commonly traced via the perf utility. They are similar to tracepoints, but there is only about a dozen of these, and they are documented in the perf_event_open(2) man page. The event names are:

  • cpu-clock or cpu
  • task-clock
  • page-faults or faults
  • context-switches or cs
  • cpu-migrations
  • minor-faults
  • major-faults
  • alignment-faults
  • emulation-faults
  • dummy
  • bpf-output

The count is the trigger for the probe, which will fire once for every count events. If the count is not provided, a default is used.

Examples:

# bpftrace -e 'software:faults:100 { @[comm] = count(); }'
Attaching 1 probe...
^C

@[ls]: 1
@[pager]: 2
@[locale]: 2
@[preconv]: 2
@[sh]: 3
@[tbl]: 3
@[bash]: 4
@[groff]: 5
@[grotty]: 7
@[sleep]: 9
@[nroff]: 12
@[troff]: 18
@[man]: 97

This roughly counts who is causing page faults, by sampling the process name for every one in one hundred faults.

12. hardware: Pre-defined Hardware Events

Syntax:

hardware:event_name:count
hardware:event_name:

These are the pre-defined hardware events provided by the Linux kernel, as commonly traced by the perf utility. They are implemented using performance monitoring counters (PMCs): hardware resources on the processor. There are about ten of these, and they are documented in the perf_event_open(2) man page. The event names are:

  • cpu-cycles or cycles
  • instructions
  • cache-references
  • cache-misses
  • branch-instructions or branches
  • branch-misses
  • bus-cycles
  • frontend-stalls
  • backend-stalls
  • ref-cycles

The count is the trigger for the probe, which will fire once for every count events. If the count is not provided, a default is used.

Examples:

bpftrace -e 'hardware:cache-misses:1000000 { @[pid] = count(); }'

That would fire once for every 1000000 cache misses. This usually indicates the last level cache (LLC).

13. BEGIN/END: Built-in events

Syntax:

BEGIN
END

These are special built-in events provided by the bpftrace runtime. BEGIN is triggered before all other probes are attached. END is triggered after all other probes are detached.

14. watchpoint: Memory watchpoints

WARNING: this feature is experimental and may be subject to interface changes.

Syntax:

watchpoint::hex_address:length:mode

These are memory watchpoints provided by the kernel. Whenever a memory address is written to (w), read from (r), or executed (x), the kernel can generate an event. Note that a pid (-p) or a command (-c) must be provided to bpftrace. Also note you may not monitor for execution while monitoring read or write.

Examples:

bpftrace -e 'watchpoint::0x10000000:8:rw { printf("hit!\n"); }' -c ~/binary

Variables

1. Builtins

  • pid - Process ID (kernel tgid)
  • tid - Thread ID (kernel pid)
  • uid - User ID
  • gid - Group ID
  • nsecs - Nanosecond timestamp
  • elapsed - Nanosecond timestamp since bpftrace initialization
  • cpu - Processor ID
  • comm - Process name
  • kstack - Kernel stack trace
  • ustack - User stack trace
  • arg0, arg1, ..., argN. - Arguments to the traced function
  • retval - Return value from traced function
  • func - Name of the traced function
  • probe - Full name of the probe
  • curtask - Current task struct as a u64
  • rand - Random number as a u32
  • cgroup - Cgroup ID of the current process
  • $1, $2, ..., $N, $#. - Positional parameters for the bpftrace program

Many of these are discussed in other sections (use search).

2. @, $: Basic variables

Syntax:

@global_name
@thread_local_variable_name[tid]
$scratch_name

bpftrace supports global & per-thread variables (via BPF maps), and scratch variables.

Examples:

2.1. Global

Syntax: @name

For example, @start:

# bpftrace -e 'BEGIN { @start = nsecs; }
    kprobe:do_nanosleep /@start != 0/ { printf("at %d ms: sleep\n", (nsecs - @start) / 1000000); }'
Attaching 2 probes...
at 437 ms: sleep
at 647 ms: sleep
at 1098 ms: sleep
at 1438 ms: sleep
at 1648 ms: sleep
^C

@start: 4064438886907216

2.2. Per-Thread:

These can be implemented as an associative array keyed on the thread ID. For example, @start[tid]:

# bpftrace -e 'kprobe:do_nanosleep { @start[tid] = nsecs; }
    kretprobe:do_nanosleep /@start[tid] != 0/ { printf("slept for %d ms\n", (nsecs - @start[tid]) / 1000000); delete(@start[tid]); }'
Attaching 2 probes...
slept for 1000 ms
slept for 1000 ms
slept for 1000 ms
slept for 1009 ms
slept for 2002 ms
[...]

2.3. Scratch:

Syntax: $name

For example, $delta:

# bpftrace -e 'kprobe:do_nanosleep { @start[tid] = nsecs; }
    kretprobe:do_nanosleep /@start[tid] != 0/ { $delta = nsecs - @start[tid]; printf("slept for %d ms\n", $delta / 1000000); delete(@start[tid]); }'
Attaching 2 probes...
slept for 1000 ms
slept for 1000 ms
slept for 1000 ms

3. @[]: Associative Arrays

Syntax: @associative_array_name[key_name] = value

These are implemented using BPF maps.

For example, @start[tid]:

# bpftrace -e 'kprobe:do_nanosleep { @start[tid] = nsecs; }
    kretprobe:do_nanosleep /@start[tid] != 0/ { printf("slept for %d ms\n", (nsecs - @start[tid]) / 1000000); delete(@start[tid]); }'
Attaching 2 probes...
slept for 1000 ms
slept for 1000 ms
slept for 1000 ms
[...]

4. count(): Frequency Counting

This is provided by the count() function: see the Count section.

5. hist(), lhist(): Histograms

These are provided by the hist() and lhist() functions. See the Log2 Histogram and Linear Histogram sections.

6. nsecs: Timestamps and Time Deltas

Syntax: nsecs

These are implemented using bpf_ktime_get_ns().

Examples:

# bpftrace -e 'BEGIN { @start = nsecs; }
    kprobe:do_nanosleep /@start != 0/ { printf("at %d ms: sleep\n", (nsecs - @start) / 1000000); }'
Attaching 2 probes...
at 437 ms: sleep
at 647 ms: sleep
at 1098 ms: sleep
at 1438 ms: sleep
^C

7. kstack: Stack Traces, Kernel

Syntax: kstack

This builtin is an alias to kstack().

Examples:

# bpftrace -e 'kprobe:ip_output { @[kstack] = count(); }'
Attaching 1 probe...
[...]
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
    tcp_release_cb+225
    release_sock+64
    tcp_sendmsg+49
    sock_sendmsg+48
    sock_write_iter+135
    __vfs_write+247
    vfs_write+179
    sys_write+82
    entry_SYSCALL_64_fastpath+30
]: 1708
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
    __tcp_push_pending_frames+45
    tcp_sendmsg_locked+2637
    tcp_sendmsg+39
    sock_sendmsg+48
    sock_write_iter+135
    __vfs_write+247
    vfs_write+179
    sys_write+82
    entry_SYSCALL_64_fastpath+30
]: 9048
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
    tcp_tasklet_func+348
    tasklet_action+241
    __do_softirq+239
    irq_exit+174
    do_IRQ+74
    ret_from_intr+0
    cpuidle_enter_state+159
    do_idle+389
    cpu_startup_entry+111
    start_secondary+398
    secondary_startup_64+165
]: 11430

8. ustack: Stack Traces, User

Syntax: ustack

This builtin is an alias to ustack().

Examples:

# bpftrace -e 'kprobe:do_sys_open /comm == "bash"/ { @[ustack] = count(); }'
Attaching 1 probe...
^C

@[
    __open_nocancel+65
    command_word_completion_function+3604
    rl_completion_matches+370
    bash_default_completion+540
    attempt_shell_completion+2092
    gen_completion_matches+82
    rl_complete_internal+288
    rl_complete+145
    _rl_dispatch_subseq+647
    _rl_dispatch+44
    readline_internal_char+479
    readline_internal_charloop+22
    readline_internal+23
    readline+91
    yy_readline_get+152
    yy_readline_get+429
    yy_getc+13
    shell_getc+469
    read_token+251
    yylex+192
    yyparse+777
    parse_command+126
    read_command+207
    reader_loop+391
    main+2409
    __libc_start_main+231
    0x61ce258d4c544155
]: 9
@[
    __open_nocancel+65
    command_word_completion_function+3604
    rl_completion_matches+370
    bash_default_completion+540
    attempt_shell_completion+2092
    gen_completion_matches+82
    rl_complete_internal+288
    rl_complete+89
    _rl_dispatch_subseq+647
    _rl_dispatch+44
    readline_internal_char+479
    readline_internal_charloop+22
    readline_internal+23
    readline+91
    yy_readline_get+152
    yy_readline_get+429
    yy_getc+13
    shell_getc+469
    read_token+251
    yylex+192
    yyparse+777
    parse_command+126
    read_command+207
    reader_loop+391
    main+2409
    __libc_start_main+231
    0x61ce258d4c544155
]: 18

Note that for this example to work, bash had to be recompiled with frame pointers.

9. $1, ..., $N, $#: Positional Parameters

Syntax: $1, $2, ..., $N, $#

These are the positional parameters to the bpftrace program, also referred to as command line arguments. If the parameter is numeric (entirely digits), it can be used as a number. If it is non-numeric, it must be used as a string in the str() call. If a parameter is used that was not provided, it will default to zero for numeric context, and "" for string context.

$# returns the number of positional arguments supplied.

This allows scripts to be written that use basic arguments to change their behavior. If you develop a script that requires more complex argument processing, it may be better suited for bcc instead, which supports Python's argparse and completely custom argument processing.

One-liner example:

# bpftrace -e 'BEGIN { printf("I got %d, %s (%d args)\n", $1, str($2), $#); }' 42 "hello"
Attaching 1 probe...
I got 42, hello (2 args)

Script example, bsize.d:

#!/usr/local/bin/bpftrace

BEGIN
{
	printf("Tracing block I/O sizes > %d bytes\n", $1);
}

tracepoint:block:block_rq_issue
/args->bytes > $1/
{
	@ = hist(args->bytes);
}

When run with a 65536 argument:

# ./bsize.bt 65536
Attaching 2 probes...
Tracing block I/O sizes > 65536 bytes
^C

@:
[512K, 1M)             1 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|

It has passed the argument in as $1, and used it as a filter.

With no arguments, $1 defaults to zero:

# ./bsize.bt
Attaching 2 probes...
Tracing block I/O sizes > 0 bytes
^C

@:
[4K, 8K)             115 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[8K, 16K)             35 |@@@@@@@@@@@@@@@                                     |
[16K, 32K)             5 |@@                                                  |
[32K, 64K)             3 |@                                                   |
[64K, 128K)            1 |                                                    |
[128K, 256K)           0 |                                                    |
[256K, 512K)           0 |                                                    |
[512K, 1M)             1 |                                                    |

Functions

1. Builtins

  • printf(char *fmt, ...) - Print formatted
  • time(char *fmt) - Print formatted time
  • join(char *arr[] [, char *delim]) - Print the array
  • str(char *s [, int length]) - Returns the string pointed to by s
  • ksym(void *p) - Resolve kernel address
  • usym(void *p) - Resolve user space address
  • kaddr(char *name) - Resolve kernel symbol name
  • uaddr(char *name) - Resolve user-level symbol name
  • reg(char *name) - Returns the value stored in the named register
  • system(char *fmt) - Execute shell command
  • exit() - Quit bpftrace
  • cgroupid(char *path) - Resolve cgroup ID
  • kstack([StackMode mode, ][int level]) - Kernel stack trace
  • ustack([StackMode mode, ][int level]) - User stack trace
  • ntop([int af, ]int|char[4|16] addr) - Convert IP address data to text
  • cat(char *filename) - Print file content

Some of these are asynchronous: the kernel queues the event, but some time later (milliseconds) it is processed in user-space. The asynchronous actions are: printf(), time(), and join(). Both ksym() and usym(), as well as the variables kstack and ustack, record addresses synchronously, but then do symbol translation asynchronously.

A selection of these are discussed in the following sections.

2. printf(): Printing

Syntax: printf(fmt, args)

This behaves like printf() from C and other languages, with a limited set of format characters. Example:

# bpftrace -e 'tracepoint:syscalls:sys_enter_execve { printf("%s called %s\n", comm, str(args->filename)); }'
Attaching 1 probe...
bash called /bin/ls
bash called /usr/bin/man
man called /apps/nflx-bash-utils/bin/preconv
man called /usr/local/sbin/preconv
man called /usr/local/bin/preconv
man called /usr/sbin/preconv
man called /usr/bin/preconv
man called /apps/nflx-bash-utils/bin/tbl
[...]

3. time(): Time

Syntax: time(fmt)

This prints the current time using the format string supported by libc strftime(3).

# bpftrace -e 'kprobe:do_nanosleep { time("%H:%M:%S\n"); }'
07:11:03
07:11:09
^C

If a format string is not provided, it defaults to "%H:%M:%S\n".

4. join(): Join

Syntax: join(char *arr[] [, char *delim])

This joins the array of strings with a space character, and prints it out, separated by delimiters. The default delimiter, if none is provided, is the space character. This current version does not return a string, so it cannot be used as an argument in printf(). Example:

# bpftrace -e 'tracepoint:syscalls:sys_enter_execve { join(args->argv); }'
Attaching 1 probe...
ls --color=auto
man ls
preconv -e UTF-8
preconv -e UTF-8
preconv -e UTF-8
preconv -e UTF-8
preconv -e UTF-8
tbl
[...]

# bpftrace -e 'tracepoint:syscalls:sys_enter_execve { join(args->argv, ","); }'
Attaching 1 probe...
ls,--color=auto
man,ls
preconv,-e,UTF-8
preconv,-e,UTF-8
preconv,-e,UTF-8
preconv,-e,UTF-8
preconv,-e,UTF-8
tbl
[...]

5. str(): Strings

Syntax: str(char *s [, int length])

Returns the string pointed to by s. length can be used to limit the size of the read, and/or introduce a null-terminator. By default, the string will have size 64 bytes (tuneable using env var BPFTRACE_STRLEN).

Examples:

We can take the args->filename of sys_enter_execve (a const char *filename), and read the string to which it points. This string can be provided as an argument to printf():

# bpftrace -e 'tracepoint:syscalls:sys_enter_execve { printf("%s called %s\n", comm, str(args->filename)); }'
Attaching 1 probe...
bash called /bin/ls
bash called /usr/bin/man
man called /apps/nflx-bash-utils/bin/preconv
man called /usr/local/sbin/preconv
man called /usr/local/bin/preconv
man called /usr/sbin/preconv
man called /usr/bin/preconv
man called /apps/nflx-bash-utils/bin/tbl
[...]

We can trace strings that are displayed in a bash shell. Some length tuning is employed, because:

  • sys_enter_write()'s args->buf does not point to null-terminated strings
    • we use the length parameter to limit how many bytes to read of the pointed-to string
  • sys_enter_write()'s args->buf contains messages larger than 64 bytes
    • we increase BPFTRACE_STRLEN to accommodate the large messages
# BPFTRACE_STRLEN=200 bpftrace -e 'tracepoint:syscalls:sys_enter_write /pid == 23506/ { printf("<%s>\n", str(args->buf, args->count)); }'
# type pwd into terminal 23506
<p>
<w>
<d>
# press enter in terminal 23506
<
>
</home/anon
>
<anon@anon-VirtualBox:~$ >

6. ksym(): Symbol resolution, kernel-level

Syntax: ksym(addr)

Examples:

# bpftrace -e 'kprobe:do_nanosleep { printf("%s\n", ksym(reg("ip"))); }'
Attaching 1 probe...
do_nanosleep
do_nanosleep

7. usym(): Symbol resolution, user-level

Syntax: usym(addr)

Examples:

# bpftrace -e 'uprobe:/bin/bash:readline { printf("%s\n", usym(reg("ip"))); }'
Attaching 1 probe...
readline
readline
readline
^C

8. kaddr(): Address resolution, kernel-level

Syntax: kaddr(char *name)

Examples:

# bpftrace -e 'BEGIN { printf("%s\n", str(*kaddr("usbcore_name"))); }'
Attaching 1 probe...
usbcore
^C

This is printing the usbcore_name string from drivers/usb/core/usb.c:

const char *usbcore_name = "usbcore";

9. uaddr(): Address resolution, user-level

Syntax: uaddr(char *name)

Examples:

# bpftrace -e 'uprobe:/bin/bash:readline { printf("PS1: %s\n", str(*uaddr("ps1_prompt"))); }'
Attaching 1 probe...
PS1: \[\e[34;1m\]\u@\h:\w>\[\e[0m\]
PS1: \[\e[34;1m\]\u@\h:\w>\[\e[0m\]
^C

This is printing the ps1_prompt string from /bin/bash, whenever a readline() function is executed.

10. reg(): Registers

Syntax: reg(char *name)

Examples:

# bpftrace -e 'kprobe:tcp_sendmsg { @[ksym(reg("ip"))] = count(); }'
Attaching 1 probe...
^C

@[tcp_sendmsg]: 7

See src/arch/x86_64.cpp for the register name list.

11. system(): System

Syntax: system(fmt)

This runs the provided command at the shell. For example:

# bpftrace --unsafe -e 'kprobe:do_nanosleep { system("ps -p %d\n", pid); }'
Attaching 1 probe...
  PID TTY          TIME CMD
 1339 ?        00:00:15 iscsid
  PID TTY          TIME CMD
 1339 ?        00:00:15 iscsid
  PID TTY          TIME CMD
 1518 ?        00:01:07 irqbalance
  PID TTY          TIME CMD
 1339 ?        00:00:15 iscsid
^C

This can be useful to execute commands or a shell script when an instrumented event happens.

Note this is an unsafe function. To use it, bpftrace must be run with --unsafe.

12. exit(): Exit

Syntax: exit()

This exits bpftrace, and can be combined with an interval probe to record statistics for a certain duration. Example:

# bpftrace -e 'kprobe:do_sys_open { @opens = count(); } interval:s:1 { exit(); }'
Attaching 2 probes...
@opens: 119

13. cgroupid(): Resolve cgroup ID

Syntax: cgroupid(char *path)

This returns a cgroup ID of a specific cgroup, and can be combined with the cgroup builtin to filter the tasks that belong to the specific cgroup, for example:

# bpftrace -e 'tracepoint:syscalls:sys_enter_openat /cgroup == cgroupid("/sys/fs/cgroup/unified/mycg")/ { printf("%s\n", str(args->filename)); }':
Attaching 1 probe...
/etc/ld.so.cache
/lib64/libc.so.6
/usr/lib/locale/locale-archive
/etc/shadow
^C

And in other terminal:

# echo $$ > /sys/fs/cgroup/unified/mycg/cgroup.procs
# cat /etc/shadow

14. ntop(): Convert IP address data to text

Syntax: ntop([int af, ]int|char[4|16] addr)

This returns the string representation of an IPv4 or IPv6 address. ntop will infer the address type (IPv4 or IPv6) based on the addr type and size. If an integer or char[4] is given, ntop assumes IPv4, if a char[16] is given, ntop assumes IPv6. You can also pass the address type explicitly as the first parameter.

Examples:

A simple example of ntop with an ipv4 hex-encoded literal:

bpftrace -e 'BEGIN { printf("%s\n", ntop(0x0100007f));}'
127.0.0.1
^C

Same example as before, but passing the address type explicitly to ntop:

bpftrace -e '#include <linux/socket.h>
BEGIN { printf("%s\n", ntop(AF_INET, 0x0100007f));}'
127.0.0.1
^C

A less trivial example of this usage, tracing tcp state changes, and printing the destination IPv6 address:

bpftrace -e 'tracepoint:tcp:tcp_set_state { printf("%s\n", ntop(args->daddr_v6)) }'
Attaching 1 probe...
::ffff:216.58.194.164
::ffff:216.58.194.164
::ffff:216.58.194.164
::ffff:216.58.194.164
::ffff:216.58.194.164
^C

And initiate a connection to this (or any) address in another terminal:

curl www.google.com

15. kstack(): Stack Traces, Kernel

Syntax: kstack([StackMode mode, ][int limit])

These are implemented using BPF stack maps.

Examples:

# bpftrace -e 'kprobe:ip_output { @[kstack()] = count(); }'
Attaching 1 probe...
[...]
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
    tcp_release_cb+225
    release_sock+64
    tcp_sendmsg+49
    sock_sendmsg+48
    sock_write_iter+135
    __vfs_write+247
    vfs_write+179
    sys_write+82
    entry_SYSCALL_64_fastpath+30
]: 1708
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
    __tcp_push_pending_frames+45
    tcp_sendmsg_locked+2637
    tcp_sendmsg+39
    sock_sendmsg+48
    sock_write_iter+135
    __vfs_write+247
    vfs_write+179
    sys_write+82
    entry_SYSCALL_64_fastpath+30
]: 9048
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
    tcp_tasklet_func+348
    tasklet_action+241
    __do_softirq+239
    irq_exit+174
    do_IRQ+74
    ret_from_intr+0
    cpuidle_enter_state+159
    do_idle+389
    cpu_startup_entry+111
    start_secondary+398
    secondary_startup_64+165
]: 11430

Sampling only three frames from the stack (limit = 3):

# bpftrace -e 'kprobe:ip_output { @[kstack(3)] = count(); }'
Attaching 1 probe...
[...]
@[
    ip_output+1
    tcp_transmit_skb+1308
    tcp_write_xmit+482
]: 22186

You can also choose a different output format. Available formats are bpftrace and perf:

# bpftrace -e 'kprobe:do_mmap { @[kstack(perf)] = count(); }'
Attaching 1 probe...
[...]
@[
	ffffffffb4019501 do_mmap+1
	ffffffffb401700a sys_mmap_pgoff+266
	ffffffffb3e334eb sys_mmap+27
	ffffffffb3e03ae3 do_syscall_64+115
	ffffffffb4800081 entry_SYSCALL_64_after_hwframe+61

]: 22186

It's also possible to use a different output format and limit the number of frames:

# bpftrace -e 'kprobe:do_mmap { @[kstack(perf, 3)] = count(); }'
Attaching 1 probe...
[...]
@[
	ffffffffb4019501 do_mmap+1
	ffffffffb401700a sys_mmap_pgoff+266
	ffffffffb3e334eb sys_mmap+27

]: 22186

16. ustack(): Stack Traces, User

Syntax: ustack([StackMode mode, ][int limit])

These are implemented using BPF stack maps.

Examples:

# bpftrace -e 'kprobe:do_sys_open /comm == "bash"/ { @[ustack()] = count(); }'
Attaching 1 probe...
^C

@[
    __open_nocancel+65
    command_word_completion_function+3604
    rl_completion_matches+370
    bash_default_completion+540
    attempt_shell_completion+2092
    gen_completion_matches+82
    rl_complete_internal+288
    rl_complete+145
    _rl_dispatch_subseq+647
    _rl_dispatch+44
    readline_internal_char+479
    readline_internal_charloop+22
    readline_internal+23
    readline+91
    yy_readline_get+152
    yy_readline_get+429
    yy_getc+13
    shell_getc+469
    read_token+251
    yylex+192
    yyparse+777
    parse_command+126
    read_command+207
    reader_loop+391
    main+2409
    __libc_start_main+231
    0x61ce258d4c544155
]: 9
@[
    __open_nocancel+65
    command_word_completion_function+3604
    rl_completion_matches+370
    bash_default_completion+540
    attempt_shell_completion+2092
    gen_completion_matches+82
    rl_complete_internal+288
    rl_complete+89
    _rl_dispatch_subseq+647
    _rl_dispatch+44
    readline_internal_char+479
    readline_internal_charloop+22
    readline_internal+23
    readline+91
    yy_readline_get+152
    yy_readline_get+429
    yy_getc+13
    shell_getc+469
    read_token+251
    yylex+192
    yyparse+777
    parse_command+126
    read_command+207
    reader_loop+391
    main+2409
    __libc_start_main+231
    0x61ce258d4c544155
]: 18

Sampling only three frames from the stack (limit = 6):

# bpftrace -e 'kprobe:do_sys_open /comm == "bash"/ { @[ustack(6)] = count(); }'
Attaching 1 probe...
^C

@[
    __open_nocancel+65
    command_word_completion_function+3604
    rl_completion_matches+370
    bash_default_completion+540
    attempt_shell_completion+2092
    gen_completion_matches+82
]: 27

You can also choose a different output format. Available formats are bpftrace and perf:

# bpftrace -e 'uprobe:bash:readline { printf("%s\n", ustack(perf)); }'
Attaching 1 probe...

	5649feec4090 readline+0 (/home/mmarchini/bash/bash/bash)
	5649fee2bfa6 yy_readline_get+451 (/home/mmarchini/bash/bash/bash)
	5649fee2bdc6 yy_getc+13 (/home/mmarchini/bash/bash/bash)
	5649fee2cd36 shell_getc+469 (/home/mmarchini/bash/bash/bash)
	5649fee2e527 read_token+251 (/home/mmarchini/bash/bash/bash)
	5649fee2d9e2 yylex+192 (/home/mmarchini/bash/bash/bash)
	5649fee286fd yyparse+777 (/home/mmarchini/bash/bash/bash)
	5649fee27dd6 parse_command+54 (/home/mmarchini/bash/bash/bash)

It's also possible to use a different output format and limit the number of frames:

# bpftrace -e 'uprobe:bash:readline { printf("%s\n", ustack(perf, 3)); }'
Attaching 1 probe...

	5649feec4090 readline+0 (/home/mmarchini/bash/bash/bash)
	5649fee2bfa6 yy_readline_get+451 (/home/mmarchini/bash/bash/bash)
	5649fee2bdc6 yy_getc+13 (/home/mmarchini/bash/bash/bash)

Note that for these examples to work, bash had to be recompiled with frame pointers.

17. cat(): Print file content

Syntax: cat(filename)

This prints the file content. For example:

# bpftrace -e 't:syscalls:sys_enter_execve { printf("%s ", str(args->filename)); cat("/proc/loadavg"); }'
Attaching 1 probe...
/usr/libexec/grepconf.sh 3.18 2.90 2.94 2/977 30138
/usr/bin/grep 3.18 2.90 2.94 4/978 30139
/usr/bin/flatpak 3.18 2.90 2.94 2/980 30143
/usr/bin/grep 3.18 2.90 2.94 3/977 30144
/usr/bin/sed 3.18 2.90 2.94 7/978 30146
/usr/bin/tclsh 3.18 2.90 2.94 5/978 30150
/usr/bin/manpath 3.18 2.90 2.94 2/978 30152
/bin/ps 3.18 2.90 2.94 2/979 30155
^C

The cat() builtin also supports a format string as argument:

./bpftrace -e 'tracepoint:syscalls:sys_enter_sendmsg { printf("%s => ", comm); cat("/proc/%d/cmdline", pid); printf("\n") }'
Attaching 1 probe...
Gecko_IOThread => /usr/lib64/firefox/firefox
Gecko_IOThread => /usr/lib64/firefox/firefox
Gecko_IOThread => /usr/lib64/firefox/firefox
Gecko_IOThread => /usr/lib64/firefox/firefox
Gecko_IOThread => /usr/lib64/firefox/firefox
Gecko_IOThread => /usr/lib64/firefox/firefox
Gecko_IOThread => /usr/lib64/firefox/firefox
^C

Map Functions

Maps are special BPF data types that can be used to store counts, statistics, and histograms. They are also used for some variable types as discussed in the previous section, whenever @ is used: globals, per thread variables, and associative arrays.

When bpftrace exits, all maps are printed. For example (the count() function is covered in the sections that follow):

# bpftrace -e 'kprobe:vfs_read { @[comm] = count(); }'
Attaching 1 probe...
^C

@[systemd]: 6
@[vi]: 7
@[sshd]: 16
@[snmpd]: 321
@[snmp-pass]: 374

The map was printed after the Ctrl-C to end the program. If you use maps that you do not wish to be automatically printed on exit, you can add an END block that clears the maps. For example:

END
{
	clear(@start);
}

1. Builtins

  • count() - Count the number of times this function is called
  • sum(int n) - Sum the value
  • avg(int n) - Average the value
  • min(int n) - Record the minimum value seen
  • max(int n) - Record the maximum value seen
  • stats(int n) - Return the count, average, and total for this value
  • hist(int n) - Produce a log2 histogram of values of n
  • lhist(int n, int min, int max, int step) - Produce a linear histogram of values of n
  • delete(@x[key]) - Delete the map element passed in as an argument
  • print(@x[, top [, div]]) - Print the map, optionally the top entries only and with a divisor
  • clear(@x) - Delete all keys from the map
  • zero(@x) - Set all map values to zero

Some of these are asynchronous: the kernel queues the event, but some time later (milliseconds) it is processed in user-space. The asynchronous actions are: print(), clear(), and zero().

2. count(): Count

Syntax: @counter_name[optional_keys] = count()

This is implemented using a BPF map.

For example, @reads:

# bpftrace -e 'kprobe:vfs_read { @reads = count();  }'
Attaching 1 probe...
^C

@reads: 119

That shows there were 119 calls to vfs_read() while tracing.

This next example includes the comm variable as a key, so that the value is broken down by each process name. For example, @reads[comm]:

# bpftrace -e 'kprobe:vfs_read { @reads[comm] = count(); }'
Attaching 1 probe...
^C

@reads[sleep]: 4
@reads[bash]: 5
@reads[ls]: 7
@reads[snmp-pass]: 8
@reads[snmpd]: 14
@reads[sshd]: 14

3. sum(): Sum

Syntax: @counter_name[optional_keys] = sum(value)

This is implemented using a BPF map.

For example, @bytes[comm]:

# bpftrace -e 'kprobe:vfs_read { @bytes[comm] = sum(arg2); }'
Attaching 1 probe...
^C

@bytes[bash]: 7
@bytes[sleep]: 4160
@bytes[ls]: 6208
@bytes[snmpd]: 20480
@bytes[snmp-pass]: 65536
@bytes[sshd]: 262144

That is summing requested bytes via the vfs_read() kernel function, which is one of two possible entry points for the read syscall. To see actual bytes read:

# bpftrace -e 'kretprobe:vfs_read /retval > 0/ { @bytes[comm] = sum(retval); }'
Attaching 1 probe...
^C

@bytes[bash]: 5
@bytes[sshd]: 1135
@bytes[systemd-journal]: 1699
@bytes[sleep]: 2496
@bytes[ls]: 4583
@bytes[snmpd]: 35549
@bytes[snmp-pass]: 55681

Now a filter is used to ensure the return value was positive before it is used in the sum(). The return value may be negative in cases of error, as is the case with other functions. Remember this whenever using sum() on a retval.

4. avg(): Average

Syntax: @counter_name[optional_keys] = avg(value)

This is implemented using a BPF map.

For example, @bytes[comm]:

# bpftrace -e 'kprobe:vfs_read { @bytes[comm] = avg(arg2); }'
Attaching 1 probe...
^C

@bytes[bash]: 1
@bytes[sleep]: 832
@bytes[ls]: 886
@bytes[snmpd]: 1706
@bytes[snmp-pass]: 8192
@bytes[sshd]: 16384

This is averaging the requested read size.

5. min(): Minimum

Syntax: @counter_name[optional_keys] = min(value)

This is implemented using a BPF map.

For example, @bytes[comm]:

# bpftrace -e 'kprobe:vfs_read { @bytes[comm] = min(arg2); }'
Attaching 1 probe...
^C

@bytes[bash]: 1
@bytes[systemd-journal]: 8
@bytes[snmpd]: 64
@bytes[ls]: 832
@bytes[sleep]: 832
@bytes[snmp-pass]: 8192
@bytes[sshd]: 16384

This shows the minimum value seen.

6. max(): Maximum

Syntax: @counter_name[optional_keys] = max(value)

This is implemented using a BPF map.

For example, @bytes[comm]:

# bpftrace -e 'kprobe:vfs_read { @bytes[comm] = max(arg2); }'
Attaching 1 probe...
^C

@bytes[bash]: 1
@bytes[systemd-journal]: 8
@bytes[sleep]: 832
@bytes[ls]: 1024
@bytes[snmpd]: 4096
@bytes[snmp-pass]: 8192
@bytes[sshd]: 16384

This shows the maximum value seen.

7. stats(): Stats

Syntax: @counter_name[optional_keys] = stats(value)

This is implemented using a BPF map.

For example, @bytes[comm]:

# bpftrace -e 'kprobe:vfs_read { @bytes[comm] = stats(arg2); }'
Attaching 1 probe...
^C

@bytes[bash]: count 7, average 1, total 7
@bytes[sleep]: count 5, average 832, total 4160
@bytes[ls]: count 7, average 886, total 6208
@bytes[snmpd]: count 18, average 1706, total 30718
@bytes[snmp-pass]: count 12, average 8192, total 98304
@bytes[sshd]: count 15, average 16384, total 245760

This stats() function returns three statistics: the count of events, the average for the argument value, and the total of the argument value. This is similar to using count(), avg(), and sum().

8. hist(): Log2 Histogram

Syntax:

@histogram_name[optional_key] = hist(value)

This is implemented using a BPF map.

Examples:

8.1. Power-Of-2:

# bpftrace -e 'kretprobe:vfs_read { @bytes = hist(retval); }'
Attaching 1 probe...
^C

@bytes:
(..., 0)             117 |@@@@@@@@@@@@                                        |
[0]                    5 |                                                    |
[1]                  325 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@                  |
[2, 4)                 6 |                                                    |
[4, 8)                 3 |                                                    |
[8, 16)              495 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[16, 32)              35 |@@@                                                 |
[32, 64)              25 |@@                                                  |
[64, 128)             21 |@@                                                  |
[128, 256)             1 |                                                    |
[256, 512)             3 |                                                    |
[512, 1K)              2 |                                                    |
[1K, 2K)               1 |                                                    |
[2K, 4K)               2 |                                                    |

8.2. Power-Of-2 By Key:

# bpftrace -e 'kretprobe:do_sys_open { @bytes[comm] = hist(retval); }'
Attaching 1 probe...
^C

@bytes[snmp-pass]:
[4, 8)                 6 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|

@bytes[ls]:
[2, 4)                 9 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|

@bytes[snmpd]:
[1]                    1 |@@@@                                                |
[2, 4)                 0 |                                                    |
[4, 8)                 0 |                                                    |
[8, 16)                4 |@@@@@@@@@@@@@@@@@@                                  |
[16, 32)              11 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|

@bytes[irqbalance]:
(..., 0)              15 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@               |
[0]                    0 |                                                    |
[1]                    0 |                                                    |
[2, 4)                 0 |                                                    |
[4, 8)                21 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|

9. lhist(): Linear Histogram

Syntax:

@histogram_name[optional_key] = lhist(value, min, max, step)

This is implemented using a BPF map. min must be non-negative.

Examples:

# bpftrace -e 'kretprobe:vfs_read { @bytes = lhist(retval, 0, 10000, 1000); }'
Attaching 1 probe...
^C

@bytes:
[0, 1000)            480 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[1000, 2000)          49 |@@@@@                                               |
[2000, 3000)          12 |@                                                   |
[3000, 4000)          39 |@@@@                                                |
[4000, 5000)         267 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@                        |

10. print(): Print Map

Syntax: print(@map [, top [, divisor]])

The print() function will print a map, similar to the automatic printing when bpftrace ends. Two optional arguments can be provided: a top number, so that only the top number of entries are printed, and a divisor, which divides the value. A couple of examples will explain their use.

As an example of top, tracing the top 5 syscalls via kprobe:SyS_*:

# bpftrace -e 'kprobe:vfs_* { @[func] = count(); } END { print(@, 5); clear(@); }'
Attaching 54 probes...
^C@[vfs_getattr]: 91
@[vfs_getattr_nosec]: 92
@[vfs_statx_fd]: 135
@[vfs_open]: 188
@[vfs_read]: 405

The final clear() is used to prevent printing the map automatically on exit.

As an example of divisor, summing total time in vfs_read() by process name as milliseconds:

# bpftrace -e 'kprobe:vfs_read { @start[tid] = nsecs; } kretprobe:vfs_read /@start[tid]/ { @ms[pid] = sum(nsecs - @start[tid]); delete(@start[tid]); } END { print(@ms, 0, 1000000); clear(@ms); clear(@start); }'

This one-liner sums the vfs_read() durations as nanoseconds, and then does the division to milliseconds when printing. Without this capability, should one try to divide to milliseconds when summing (eg, sum((nsecs - @start[tid]) / 1000000)), the value would often be rounded to zero, and not accumulate as it should.

Output

1. printf(): Per-Event Output

Syntax: printf(char *format, arguments)

Per-event details can be printed using print().

Examples:

# bpftrace -e 'kprobe:do_nanosleep { printf("sleep by %d\n", tid); }'
Attaching 1 probe...
sleep by 3669
sleep by 1396
sleep by 3669
sleep by 1396
[...]

2. interval: Interval output

Syntax: interval:s:duration_seconds

Examples:

# bpftrace -e 'kprobe:do_sys_open { @opens = @opens + 1; } interval:s:1 { printf("opens/sec: %d\n", @opens); @opens = 0; }'
Attaching 2 probes...
opens/sec: 16
opens/sec: 2
opens/sec: 3
opens/sec: 15
opens/sec: 8
opens/sec: 2
^C

@opens: 2

3. hist(), print(): Histogram Printing

Declared histograms are automatically printed out on program termination. See 5. Histograms for declarations.

Examples:

# bpftrace -e 'kretprobe:vfs_read { @bytes = hist(retval); }'
Attaching 1 probe...
^C

@bytes:
(..., 0)             117 |@@@@@@@@@@@@                                        |
[0]                    5 |                                                    |
[1]                  325 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@                  |
[2, 4)                 6 |                                                    |
[4, 8)                 3 |                                                    |
[8, 16)              495 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[16, 32)              35 |@@@                                                 |
[32, 64)              25 |@@                                                  |
[64, 128)             21 |@@                                                  |
[128, 256)             1 |                                                    |
[256, 512)             3 |                                                    |
[512, 1K)              2 |                                                    |
[1K, 2K)               1 |                                                    |
[2K, 4K)               2 |                                                    |

Histograms can also be printed on-demand, using the print() function. Eg:

# bpftrace -e 'kretprobe:vfs_read { @bytes = hist(retval); } interval:s:1 { print(@bytes); clear(@bytes); }'

[...]

Advanced Tools

bpftrace can be used to create some powerful one-liners and some simple tools. For complex tools, which may involve command line options, positional parameters, argument processing, and customized output, consider switching to bcc. bcc provides Python (and other) front-ends, enabling usage of all the other Python libraries (including argparse), as well as a direct control of the kernel BPF program. The down side is that bcc is much more verbose and laborious to program. Together, bpftrace and bcc are complimentary.

An expected development path would be exploration with bpftrace one-liners, then and ad hoc scripting with bpftrace, then finally, when needed, advanced tooling with bcc.

As an example of bpftrace vs bcc differences, the bpftrace xfsdist.bt tool also exists in bcc as xfsdist.py. Both measure the same functions and produce the same summary of information. However, the bcc version supports various arguments:

# ./xfsdist.py -h
usage: xfsdist.py [-h] [-T] [-m] [-p PID] [interval] [count]

Summarize XFS operation latency

positional arguments:
  interval            output interval, in seconds
  count               number of outputs

optional arguments:
  -h, --help          show this help message and exit
  -T, --notimestamp   don't include timestamp on interval output
  -m, --milliseconds  output in milliseconds
  -p PID, --pid PID   trace this PID only

examples:
    ./xfsdist            # show operation latency as a histogram
    ./xfsdist -p 181     # trace PID 181 only
    ./xfsdist 1 10       # print 1 second summaries, 10 times
    ./xfsdist -m 5       # 5s summaries, milliseconds

The bcc version is 131 lines of code. The bptrace version is 22.

Errors

1. Looks like the BPF stack limit of 512 bytes is exceeded

BPF programs that operate on many data items may hit this limit. There are a number of things you can try to stay within the limit:

  1. Find ways to reduce the size of the data used in the program. Eg, avoid strings if they are unnecessary: use pid instead of comm. Use fewer map keys.
  2. Split your program over multiple probes.
  3. Check the status of the BPF stack limit in Linux (it may be increased in the future, maybe as a tuneabe).
  4. (advanced): Run -d and examine the LLVM IR, and look for ways to optimize src/ast/codegen_llvm.cpp.

2. Kernel headers not found

bpftrace requires kernel headers for certain features, which are searched for by default in:

/lib/modules/$(uname -r)

The default search directory can be overridden using the environment variable BPFTRACE_KERNEL_SOURCE.