Message-ID: <20180515214145.4c6b31bc@cakuba.netronome.com>
Date:   Tue, 15 May 2018 21:41:45 -0700
From:   Jakub Kicinski <jakub.kicinski@...ronome.com>
To:     Yonghong Song <yhs@...com>
Cc:     <peterz@...radead.org>, <ast@...com>, <daniel@...earbox.net>,
        <netdev@...r.kernel.org>, <kernel-team@...com>,
        Quentin Monnet <quentin.monnet@...ronome.com>
Subject: Re: [PATCH bpf-next 7/7] tools/bpftool: add perf subcommand

On Tue, 15 May 2018 16:45:21 -0700, Yonghong Song wrote:
> The new command "bpftool perf [show]" will traverse
> all processes under /proc, and if any fd is associated
> with a perf event, it will print out related perf event
> information.
> 
> Below is an example to show the results using bcc commands.
> Running the following 4 bcc commands:
>   kprobe:     trace.py '__x64_sys_nanosleep'
>   kretprobe:  trace.py 'r::__x64_sys_nanosleep'
>   tracepoint: trace.py 't:syscalls:sys_enter_nanosleep'
>   uprobe:     trace.py 'p:/home/yhs/a.out:main'
> 
> The bpftool command line and result:
> 
>   $ bpftool perf
>   21711: prog_id 5 kprobe func __x64_sys_write offset 0
>   21765: prog_id 7 kretprobe func __x64_sys_nanosleep offset 0
>   21767: prog_id 8 tracepoint sys_enter_nanosleep
>   21800: prog_id 9 uprobe filename /home/yhs/a.out offset 1159
> 
>   $ bpftool -j perf
>   {"pid":21711,"prog_id":5,"prog_info":"kprobe","func":"__x64_sys_write","offset":0}, \
>   {"pid":21765,"prog_id":7,"prog_info":"kretprobe","func":"__x64_sys_nanosleep","offset":0}, \
>   {"pid":21767,"prog_id":8,"prog_info":"tracepoint","tracepoint":"sys_enter_nanosleep"}, \
>   {"pid":21800,"prog_id":9,"prog_info":"uprobe","filename":"/home/yhs/a.out","offset":1159}

You need to wrap the objects inside an array, so

	if (json_output)
		jsonw_start_array(json_wtr);
	nftw();
	if (json_output)
		jsonw_end_array(json_wtr);

otherwise output will not be a valid JSON.  To validate JSON try:

$ bpftool -j perf | python -m json.tool

>   $ bpftool prog
>   5: kprobe  name probe___x64_sys  tag e495a0c82f2c7a8d  gpl
> 	  loaded_at 2018-05-15T04:46:37-0700  uid 0
> 	  xlated 200B  not jited  memlock 4096B  map_ids 4
>   7: kprobe  name probe___x64_sys  tag f2fdee479a503abf  gpl
> 	  loaded_at 2018-05-15T04:48:32-0700  uid 0
> 	  xlated 200B  not jited  memlock 4096B  map_ids 7
>   8: tracepoint  name tracepoint__sys  tag 5390badef2395fcf  gpl
> 	  loaded_at 2018-05-15T04:48:48-0700  uid 0
> 	  xlated 200B  not jited  memlock 4096B  map_ids 8
>   9: kprobe  name probe_main_1  tag 0a87bdc2e2953b6d  gpl
> 	  loaded_at 2018-05-15T04:49:52-0700  uid 0
> 	  xlated 200B  not jited  memlock 4096B  map_ids 9
> 
>   $ ps ax | grep "python ./trace.py"
>   21711 pts/0    T      0:03 python ./trace.py __x64_sys_write
>   21765 pts/0    S+     0:00 python ./trace.py r::__x64_sys_nanosleep
>   21767 pts/2    S+     0:00 python ./trace.py t:syscalls:sys_enter_nanosleep
>   21800 pts/3    S+     0:00 python ./trace.py p:/home/yhs/a.out:main
>   22374 pts/1    S+     0:00 grep --color=auto python ./trace.py
> 
> Signed-off-by: Yonghong Song <yhs@...com>
> ---
>  tools/bpf/bpftool/main.c |   3 +-
>  tools/bpf/bpftool/main.h |   1 +
>  tools/bpf/bpftool/perf.c | 188 +++++++++++++++++++++++++++++++++++++++++++++++

Would you be able to also extend the Documentation/ and bash
completions?

>  3 files changed, 191 insertions(+), 1 deletion(-)
>  create mode 100644 tools/bpf/bpftool/perf.c
> 
> diff --git a/tools/bpf/bpftool/main.c b/tools/bpf/bpftool/main.c
> index 1ec852d..eea7f14 100644
> --- a/tools/bpf/bpftool/main.c
> +++ b/tools/bpf/bpftool/main.c
> @@ -87,7 +87,7 @@ static int do_help(int argc, char **argv)
>  		"       %s batch file FILE\n"
>  		"       %s version\n"
>  		"\n"
> -		"       OBJECT := { prog | map | cgroup }\n"
> +		"       OBJECT := { prog | map | cgroup | perf }\n"
>  		"       " HELP_SPEC_OPTIONS "\n"
>  		"",
>  		bin_name, bin_name, bin_name);
> @@ -216,6 +216,7 @@ static const struct cmd cmds[] = {
>  	{ "prog",	do_prog },
>  	{ "map",	do_map },
>  	{ "cgroup",	do_cgroup },
> +	{ "perf",	do_perf },
>  	{ "version",	do_version },
>  	{ 0 }
>  };
> diff --git a/tools/bpf/bpftool/main.h b/tools/bpf/bpftool/main.h
> index 6173cd9..63fdb31 100644
> --- a/tools/bpf/bpftool/main.h
> +++ b/tools/bpf/bpftool/main.h
> @@ -119,6 +119,7 @@ int do_prog(int argc, char **arg);
>  int do_map(int argc, char **arg);
>  int do_event_pipe(int argc, char **argv);
>  int do_cgroup(int argc, char **arg);
> +int do_perf(int argc, char **arg);
>  
>  int prog_parse_fd(int *argc, char ***argv);
>  int map_parse_fd_and_info(int *argc, char ***argv, void *info, __u32 *info_len);
> diff --git a/tools/bpf/bpftool/perf.c b/tools/bpf/bpftool/perf.c
> new file mode 100644
> index 0000000..6d676e4
> --- /dev/null
> +++ b/tools/bpf/bpftool/perf.c
> @@ -0,0 +1,188 @@
> +// SPDX-License-Identifier: GPL-2.0+
> +// Copyright (C) 2018 Facebook
> +// Author: Yonghong Song <yhs@...com>
> +
> +#define _GNU_SOURCE
> +#include <fcntl.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <sys/stat.h>
> +#include <sys/types.h>
> +#include <unistd.h>
> +#include <ftw.h>
> +
> +#include <bpf.h>
> +
> +#include "main.h"
> +
> +static void print_perf_json(int pid, __u32 prog_id, __u32 prog_info,
> +			    char *buf, __u64 probe_offset, __u64 probe_addr)
> +{
> +	jsonw_start_object(json_wtr);
> +	jsonw_int_field(json_wtr, "pid", pid);
> +	jsonw_uint_field(json_wtr, "prog_id", prog_id);
> +	switch (prog_info) {
> +	case BPF_PERF_INFO_TP_NAME:
> +		jsonw_string_field(json_wtr, "prog_info", "tracepoint");
> +		jsonw_string_field(json_wtr, "tracepoint", buf);
> +		break;
> +	case BPF_PERF_INFO_KPROBE:
> +		jsonw_string_field(json_wtr, "prog_info", "kprobe");
> +		if (buf[0] != '\0') {
> +			jsonw_string_field(json_wtr, "func", buf);
> +			jsonw_lluint_field(json_wtr, "offset", probe_offset);
> +		} else {
> +			jsonw_lluint_field(json_wtr, "addr", probe_addr);
> +		}
> +		break;
> +	case BPF_PERF_INFO_KRETPROBE:
> +		jsonw_string_field(json_wtr, "prog_info", "kretprobe");
> +		if (buf[0] != '\0') {
> +			jsonw_string_field(json_wtr, "func", buf);
> +			jsonw_lluint_field(json_wtr, "offset", probe_offset);
> +		} else {
> +			jsonw_lluint_field(json_wtr, "addr", probe_addr);
> +		}
> +		break;
> +	case BPF_PERF_INFO_UPROBE:
> +		jsonw_string_field(json_wtr, "prog_info", "uprobe");
> +		jsonw_string_field(json_wtr, "filename", buf);
> +		jsonw_lluint_field(json_wtr, "offset", probe_offset);
> +		break;
> +	case BPF_PERF_INFO_URETPROBE:
> +		jsonw_string_field(json_wtr, "prog_info", "uretprobe");
> +		jsonw_string_field(json_wtr, "filename", buf);
> +		jsonw_lluint_field(json_wtr, "offset", probe_offset);
> +		break;
> +	}
> +	jsonw_end_object(json_wtr);
> +}
> +
> +static void print_perf_plain(int pid, __u32 prog_id, __u32 prog_info,
> +			    char *buf, __u64 probe_offset, __u64 probe_addr)
> +{
> +	printf("%d: prog_id %u ", pid, prog_id);

nit: for consistency with prog and map listings consider using double
spaces after prog_id (i.e. between fields).  Not a hard requirement,
though, perhaps I'm the only one who finds that more readable :)

> +	switch (prog_info) {
> +	case BPF_PERF_INFO_TP_NAME:
> +		printf("tracepoint %s\n", buf);
> +		break;
> +	case BPF_PERF_INFO_KPROBE:
> +		if (buf[0] != '\0')
> +			printf("kprobe func %s offset %llu\n", buf,
> +			       probe_offset);
> +		else
> +			printf("kprobe addr %llu\n", probe_addr);
> +		break;
> +	case BPF_PERF_INFO_KRETPROBE:
> +		if (buf[0] != '\0')
> +			printf("kretprobe func %s offset %llu\n", buf,
> +			       probe_offset);
> +		else
> +			printf("kretprobe addr %llu\n", probe_addr);
> +		break;
> +	case BPF_PERF_INFO_UPROBE:
> +		printf("uprobe filename %s offset %llu\n", buf, probe_offset);
> +		break;
> +	case BPF_PERF_INFO_URETPROBE:
> +		printf("uretprobe filename %s offset %llu\n", buf,
> +		       probe_offset);
> +		break;
> +	}
> +}
> +
> +static int show_proc(const char *fpath, const struct stat *sb,
> +		     int tflag, struct FTW *ftwbuf)
> +{
> +	__u64 probe_offset, probe_addr;
> +	__u32 prog_id, prog_info;
> +	int err, pid = 0, fd = 0;
> +	const char *pch;
> +	char buf[4096];
> +
> +	/* prefix always /proc */
> +	pch = fpath + 5;
> +	if (*pch == '\0')
> +		return 0;
> +
> +	/* pid should be all numbers */
> +	pch++;
> +	while (*pch >= '0' && *pch <= '9') {

nit: isdigit()?  strtoul() with its endptr also an option.  That said
     the code is actually quite readable as is, so I'm not sure if it's
     worth complicating it.

> +		pid = pid * 10 + *pch - '0';
> +		pch++;
> +	}
> +	if (*pch == '\0')
> +		return 0;
> +	if (*pch != '/')
> +		return FTW_SKIP_SUBTREE;
> +
> +	/* check /proc/<pid>/fd directory */
> +	pch++;
> +	if (*pch == '\0' || *pch != 'f')
> +		return FTW_SKIP_SUBTREE;

but == '\0' implies != 'f'

> +	pch++;
> +	if (*pch == '\0' || *pch != 'd')
> +		return FTW_SKIP_SUBTREE;

nit: possibly just:
     if (strncmp(pch, "fd", 2))
          return FTW_SKIP_SUBTREE;
     pch += 2;

> +	pch++;
> +	if (*pch == '\0')
> +		return 0;
> +	if (*pch != '/')
> +		return FTW_SKIP_SUBTREE;
> +
> +	/* check /proc/<pid>/fd/<fd_num> */
> +	pch++;
> +	while (*pch >= '0' && *pch <= '9') {
> +		fd = fd * 10 + *pch - '0';
> +		pch++;
> +	}
> +	if (*pch != '\0')
> +		return FTW_SKIP_SUBTREE;
> +
> +	/* query (pid, fd) for potential perf events */
> +	err = bpf_trace_event_query(pid, fd, buf, sizeof(buf),
> +		&prog_id, &prog_info, &probe_offset, &probe_addr);

nit: continuation line not aligned with opening bracket

> +	if (err < 0)
> +		return 0;
> +
> +	if (json_output)
> +		print_perf_json(pid, prog_id, prog_info, buf, probe_offset,
> +				probe_addr);
> +	else
> +		print_perf_plain(pid, prog_id, prog_info, buf, probe_offset,
> +				 probe_addr);
> +
> +	return 0;
> +}
> +
> +static int do_show(int argc, char **argv)
> +{
> +	int nopenfd = 16;
> +	int flags = FTW_ACTIONRETVAL | FTW_PHYS;

nit: reverse christmas tree networking style if you don't mind

> +	if (nftw("/proc", show_proc, nopenfd, flags) == -1) {
> +		perror("nftw");

nit: p_err("%s", strerror(errno)); would also show up in JSON output

> +		return -1;
> +	}
> +
> +	return 0;
> +}
> +
> +static int do_help(int argc, char **argv)
> +{
> +	fprintf(stderr,
> +		"Usage: %s %s { show | help }\n"
> +		"",
> +		bin_name, argv[-2]);
> +
> +	return 0;
> +}
> +
> +static const struct cmd cmds[] = {
> +	{ "show",	do_show },

Other commands alias show and list, so could you add:

	{ "list",	do_show },

and list to help output?

> +	{ "help",	do_help },
> +	{ 0 }
> +};
> +
> +int do_perf(int argc, char **argv)
> +{
> +	return cmd_select(cmds, argc, argv, do_help);
> +}

Thanks a lot for adding bpftool support, and with JSON output included!

Powered by blists - more mailing lists