[PATCH bpf v2] tools/bpftool: fix a bug in bpftool perf

[PATCH bpf v2] tools/bpftool: fix a bug in bpftool perf

[Yonghong Song]

Commit b04df400c302 ("tools/bpftool: add perf subcommand")
introduced bpftool subcommand perf to query bpf program
kuprobe and tracepoint attachments.

The perf subcommand will first test whether bpf subcommand
BPF_TASK_FD_QUERY is supported in kernel or not. It does it
by opening a file with argv[0] and feeds the file descriptor
and current task pid to the kernel for querying.

Such an approach won't work if the argv[0] cannot be opened
successfully in the current directory. This is especially
true when bpftool is accessible through PATH env variable.
The error below reflects the open failure for file argv[0]
at home directory.

  [yhs@localhost ~]$ which bpftool
  /usr/local/sbin/bpftool
  [yhs@localhost ~]$ bpftool perf
  Error: perf_query_support: No such file or directory

To fix the issue, let us open root directory ("/")
which exists in every linux system. With the fix, the
error message will correctly reflect the permission issue.

  [yhs@localhost ~]$ which bpftool
  /usr/local/sbin/bpftool
  [yhs@localhost ~]$ bpftool perf
  Error: perf_query_support: Operation not permitted
  HINT: non root or kernel doesn't support TASK_FD_QUERY

Fixes: b04df400c302 ("tools/bpftool: add perf subcommand")
Reported-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
---
 tools/bpf/bpftool/perf.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

Changelogs:
 v1 -> v2:
   . remove '\n' in the p_err format string in order to
     have valid json output.

diff --git a/tools/bpf/bpftool/perf.c b/tools/bpf/bpftool/perf.c
index ac6b1a12c9b7..239715aa6fb9 100644
--- a/tools/bpf/bpftool/perf.c
+++ b/tools/bpf/bpftool/perf.c
@@ -29,9 +29,10 @@ static bool has_perf_query_support(void)
 	if (perf_query_supported)
 		goto out;
 
-	fd = open(bin_name, O_RDONLY);
-	if (fd < 0) {
-		p_err("perf_query_support: %s", strerror(errno));
+	fd = open("/", O_RDONLY);
+	if (fd > 0) {
+		p_err("perf_query_support: cannot open directory \"/\" (%s)",
+		      strerror(errno));
 		goto out;
 	}
 
-- 
2.14.3

Re: [PATCH bpf v2] tools/bpftool: fix a bug in bpftool perf

[Jakub Kicinski]

On Mon, 11 Jun 2018 19:01:08 -0700, Yonghong Song wrote:
> Commit b04df400c302 ("tools/bpftool: add perf subcommand")
> introduced bpftool subcommand perf to query bpf program
> kuprobe and tracepoint attachments.
> 
> The perf subcommand will first test whether bpf subcommand
> BPF_TASK_FD_QUERY is supported in kernel or not. It does it
> by opening a file with argv[0] and feeds the file descriptor
> and current task pid to the kernel for querying.
> 
> Such an approach won't work if the argv[0] cannot be opened
> successfully in the current directory. This is especially
> true when bpftool is accessible through PATH env variable.
> The error below reflects the open failure for file argv[0]
> at home directory.
> 
>   [yhs@localhost ~]$ which bpftool
>   /usr/local/sbin/bpftool
>   [yhs@localhost ~]$ bpftool perf
>   Error: perf_query_support: No such file or directory
> 
> To fix the issue, let us open root directory ("/")
> which exists in every linux system. With the fix, the
> error message will correctly reflect the permission issue.
> 
>   [yhs@localhost ~]$ which bpftool
>   /usr/local/sbin/bpftool
>   [yhs@localhost ~]$ bpftool perf
>   Error: perf_query_support: Operation not permitted
>   HINT: non root or kernel doesn't support TASK_FD_QUERY
> 
> Fixes: b04df400c302 ("tools/bpftool: add perf subcommand")
> Reported-by: Alexei Starovoitov <ast@kernel.org>
> Signed-off-by: Yonghong Song <yhs@fb.com>
> ---
>  tools/bpf/bpftool/perf.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> Changelogs:
>  v1 -> v2:
>    . remove '\n' in the p_err format string in order to
>      have valid json output.
> 
> diff --git a/tools/bpf/bpftool/perf.c b/tools/bpf/bpftool/perf.c
> index ac6b1a12c9b7..239715aa6fb9 100644
> --- a/tools/bpf/bpftool/perf.c
> +++ b/tools/bpf/bpftool/perf.c
> @@ -29,9 +29,10 @@ static bool has_perf_query_support(void)
>  	if (perf_query_supported)
>  		goto out;
>  
> -	fd = open(bin_name, O_RDONLY);
> -	if (fd < 0) {
> -		p_err("perf_query_support: %s", strerror(errno));
> +	fd = open("/", O_RDONLY);
> +	if (fd > 0) {

Looking at this again, why change the comparison from less than to
greater than 0?

> +		p_err("perf_query_support: cannot open directory \"/\" (%s)",
> +		      strerror(errno));
>  		goto out;
>  	}
>  

Re: [PATCH bpf v2] tools/bpftool: fix a bug in bpftool perf

[Yonghong Song]

On 6/11/18 7:42 PM, Jakub Kicinski wrote:
> On Mon, 11 Jun 2018 19:01:08 -0700, Yonghong Song wrote:
>> Commit b04df400c302 ("tools/bpftool: add perf subcommand")
>> introduced bpftool subcommand perf to query bpf program
>> kuprobe and tracepoint attachments.
>>
>> The perf subcommand will first test whether bpf subcommand
>> BPF_TASK_FD_QUERY is supported in kernel or not. It does it
>> by opening a file with argv[0] and feeds the file descriptor
>> and current task pid to the kernel for querying.
>>
>> Such an approach won't work if the argv[0] cannot be opened
>> successfully in the current directory. This is especially
>> true when bpftool is accessible through PATH env variable.
>> The error below reflects the open failure for file argv[0]
>> at home directory.
>>
>>    [yhs@localhost ~]$ which bpftool
>>    /usr/local/sbin/bpftool
>>    [yhs@localhost ~]$ bpftool perf
>>    Error: perf_query_support: No such file or directory
>>
>> To fix the issue, let us open root directory ("/")
>> which exists in every linux system. With the fix, the
>> error message will correctly reflect the permission issue.
>>
>>    [yhs@localhost ~]$ which bpftool
>>    /usr/local/sbin/bpftool
>>    [yhs@localhost ~]$ bpftool perf
>>    Error: perf_query_support: Operation not permitted
>>    HINT: non root or kernel doesn't support TASK_FD_QUERY
>>
>> Fixes: b04df400c302 ("tools/bpftool: add perf subcommand")
>> Reported-by: Alexei Starovoitov <ast@kernel.org>
>> Signed-off-by: Yonghong Song <yhs@fb.com>
>> ---
>>   tools/bpf/bpftool/perf.c | 7 ++++---
>>   1 file changed, 4 insertions(+), 3 deletions(-)
>>
>> Changelogs:
>>   v1 -> v2:
>>     . remove '\n' in the p_err format string in order to
>>       have valid json output.
>>
>> diff --git a/tools/bpf/bpftool/perf.c b/tools/bpf/bpftool/perf.c
>> index ac6b1a12c9b7..239715aa6fb9 100644
>> --- a/tools/bpf/bpftool/perf.c
>> +++ b/tools/bpf/bpftool/perf.c
>> @@ -29,9 +29,10 @@ static bool has_perf_query_support(void)
>>   	if (perf_query_supported)
>>   		goto out;
>>   
>> -	fd = open(bin_name, O_RDONLY);
>> -	if (fd < 0) {
>> -		p_err("perf_query_support: %s", strerror(errno));
>> +	fd = open("/", O_RDONLY);
>> +	if (fd > 0) {
> 
> Looking at this again, why change the comparison from less than to
> greater than 0?

Totally my fault. I altered the condition to test p_err output since
I cannot trigger it normally, but forgot to change it back.
Will send yet another revision :-(

>> +		p_err("perf_query_support: cannot open directory \"/\" (%s)",
>> +		      strerror(errno));
>>   		goto out;
>>   	}
>>   
>