[PATCH] perf report: Fix segfault with '-F phys_daddr'

[PATCH] perf report: Fix segfault with '-F phys_daddr'

[Ravi Bangoria]

If perf.data file is not recorded with mem-info, adding 'phys_daddr'
to output field in perf report results in segfault. Fix that.

Before:
  $ ./perf record ls
  $ ./perf report -F +phys_daddr
  Segmentation fault (core dumped)

After:
  $ ./perf report -F +phys_daddr
  Samples: 11  of event 'cycles:u', Event count (approx.): 1485821
  Overhead  Data Physical Address  Command  Shared Object  Symbol
    22.57%  [.] 0000000000000000   ls       libc-2.29.so   [.] __strcoll_l
    21.87%  [.] 0000000000000000   ls       ld-2.29.so     [.] _dl_relocate_object
    ...

Fixes: 8780fb25ab06 ("perf sort: Add sort option for physical address")
Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
---
 tools/perf/util/sort.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 43d1d410854a..c2430676e569 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -1413,7 +1413,7 @@ static int hist_entry__phys_daddr_snprintf(struct hist_entry *he, char *bf,
 	size_t ret = 0;
 	size_t len = BITS_PER_LONG / 4;
 
-	addr = he->mem_info->daddr.phys_addr;
+	addr = he->mem_info ? he->mem_info->daddr.phys_addr : 0;
 
 	ret += repsep_snprintf(bf + ret, size - ret, "[%c] ", he->level);
 
-- 
2.21.0

Re: [PATCH] perf report: Fix segfault with '-F phys_daddr'

[Arnaldo Carvalho de Melo]

Em Tue, Nov 12, 2019 at 11:19:46AM +0530, Ravi Bangoria escreveu:
> If perf.data file is not recorded with mem-info, adding 'phys_daddr'
> to output field in perf report results in segfault. Fix that.
> 
> Before:
>   $ ./perf record ls
>   $ ./perf report -F +phys_daddr
>   Segmentation fault (core dumped)
> 
> After:
>   $ ./perf report -F +phys_daddr
>   Samples: 11  of event 'cycles:u', Event count (approx.): 1485821
>   Overhead  Data Physical Address  Command  Shared Object  Symbol
>     22.57%  [.] 0000000000000000   ls       libc-2.29.so   [.] __strcoll_l
>     21.87%  [.] 0000000000000000   ls       ld-2.29.so     [.] _dl_relocate_object
>     ...

Shouldn't we instead just bail out and state that this isn't possible
and leave the user wondering why what was asked isn't presented?

- Arnaldo
 
> Fixes: 8780fb25ab06 ("perf sort: Add sort option for physical address")
> Signed-off-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
> ---
>  tools/perf/util/sort.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> index 43d1d410854a..c2430676e569 100644
> --- a/tools/perf/util/sort.c
> +++ b/tools/perf/util/sort.c
> @@ -1413,7 +1413,7 @@ static int hist_entry__phys_daddr_snprintf(struct hist_entry *he, char *bf,
>  	size_t ret = 0;
>  	size_t len = BITS_PER_LONG / 4;
>  
> -	addr = he->mem_info->daddr.phys_addr;
> +	addr = he->mem_info ? he->mem_info->daddr.phys_addr : 0;
>  
>  	ret += repsep_snprintf(bf + ret, size - ret, "[%c] ", he->level);
>  
> -- 
> 2.21.0

-- 

- Arnaldo

Re: [PATCH] perf report: Fix segfault with '-F phys_daddr'

[Ravi Bangoria]

On 11/12/19 4:34 PM, Arnaldo Carvalho de Melo wrote:
> Em Tue, Nov 12, 2019 at 11:19:46AM +0530, Ravi Bangoria escreveu:
>> If perf.data file is not recorded with mem-info, adding 'phys_daddr'
>> to output field in perf report results in segfault. Fix that.
>>
>> Before:
>>    $ ./perf record ls
>>    $ ./perf report -F +phys_daddr
>>    Segmentation fault (core dumped)
>>
>> After:
>>    $ ./perf report -F +phys_daddr
>>    Samples: 11  of event 'cycles:u', Event count (approx.): 1485821
>>    Overhead  Data Physical Address  Command  Shared Object  Symbol
>>      22.57%  [.] 0000000000000000   ls       libc-2.29.so   [.] __strcoll_l
>>      21.87%  [.] 0000000000000000   ls       ld-2.29.so     [.] _dl_relocate_object
>>      ...
> 
> Shouldn't we instead just bail out and state that this isn't possible
> and leave the user wondering why what was asked isn't presented?

You mean popup with something like "phys_daddr is not available in perf.data"
and also don't show that column in perf report?

Re: [PATCH] perf report: Fix segfault with '-F phys_daddr'

[Arnaldo Carvalho de Melo]

Em Tue, Nov 12, 2019 at 07:28:06PM +0530, Ravi Bangoria escreveu:
> 
> 
> On 11/12/19 4:34 PM, Arnaldo Carvalho de Melo wrote:
> > Em Tue, Nov 12, 2019 at 11:19:46AM +0530, Ravi Bangoria escreveu:
> > > If perf.data file is not recorded with mem-info, adding 'phys_daddr'
> > > to output field in perf report results in segfault. Fix that.
> > > 
> > > Before:
> > >    $ ./perf record ls
> > >    $ ./perf report -F +phys_daddr
> > >    Segmentation fault (core dumped)
> > > 
> > > After:
> > >    $ ./perf report -F +phys_daddr
> > >    Samples: 11  of event 'cycles:u', Event count (approx.): 1485821
> > >    Overhead  Data Physical Address  Command  Shared Object  Symbol
> > >      22.57%  [.] 0000000000000000   ls       libc-2.29.so   [.] __strcoll_l
> > >      21.87%  [.] 0000000000000000   ls       ld-2.29.so     [.] _dl_relocate_object
> > >      ...
> > 
> > Shouldn't we instead just bail out and state that this isn't possible
> > and leave the user wondering why what was asked isn't presented?
> 
> You mean popup with something like "phys_daddr is not available in perf.data"
> and also don't show that column in perf report?

Just bail out completely, something like:


   $ ./perf report -F +phys_daddr
   "phys_daddr" is not available in perf.data, use 'record -e some,thing' to have it.

- Arnaldo

Re: [PATCH] perf report: Fix segfault with '-F phys_daddr'

[Ravi Bangoria]

On 11/12/19 7:48 PM, Arnaldo Carvalho de Melo wrote:
> Em Tue, Nov 12, 2019 at 07:28:06PM +0530, Ravi Bangoria escreveu:
>>
>>
>> On 11/12/19 4:34 PM, Arnaldo Carvalho de Melo wrote:
>>> Em Tue, Nov 12, 2019 at 11:19:46AM +0530, Ravi Bangoria escreveu:
>>>> If perf.data file is not recorded with mem-info, adding 'phys_daddr'
>>>> to output field in perf report results in segfault. Fix that.
>>>>
>>>> Before:
>>>>     $ ./perf record ls
>>>>     $ ./perf report -F +phys_daddr
>>>>     Segmentation fault (core dumped)
>>>>
>>>> After:
>>>>     $ ./perf report -F +phys_daddr
>>>>     Samples: 11  of event 'cycles:u', Event count (approx.): 1485821
>>>>     Overhead  Data Physical Address  Command  Shared Object  Symbol
>>>>       22.57%  [.] 0000000000000000   ls       libc-2.29.so   [.] __strcoll_l
>>>>       21.87%  [.] 0000000000000000   ls       ld-2.29.so     [.] _dl_relocate_object
>>>>       ...
>>>
>>> Shouldn't we instead just bail out and state that this isn't possible
>>> and leave the user wondering why what was asked isn't presented?
>>
>> You mean popup with something like "phys_daddr is not available in perf.data"
>> and also don't show that column in perf report?
> 
> Just bail out completely, something like:
> 
> 
>     $ ./perf report -F +phys_daddr
>     "phys_daddr" is not available in perf.data, use 'record -e some,thing' to have it.

Sure. Thanks for the suggestion. Will re-spin.

Ravi