Re: [PATCH v4 15/16] perf tools: add feature header record to pipe-mode

[David Carrillo-Cisneros]

On Tue, Jun 13, 2017 at 11:54 AM, Jiri Olsa <jolsa@xxxxxxxxxx> wrote:
> On Mon, Jun 12, 2017 at 09:29:31PM -0700, David Carrillo-Cisneros wrote:
>
> SNIP
>
>>
>>  void perf_event__print_totals(void);
>> diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
>> index ddfaf157913d..6f6a54c15cb0 100644
>> --- a/tools/perf/util/header.c
>> +++ b/tools/perf/util/header.c
>> @@ -34,6 +34,7 @@
>>  #include "data.h"
>>  #include <api/fs/fs.h>
>>  #include "asm/bug.h"
>> +#include "tool.h"
>>
>>  #include "sane_ctype.h"
>>
>> @@ -97,11 +98,14 @@ static int __do_write_buf(struct feat_fd *ff,  const void *buf, size_t size)
>>
>>  retry:
>>       if (size > (ff->size - ff->offset)) {
>> -             addr = realloc(ff->buf, ff->size << 1);
>> +             size = ff->size << 1;
>> +             if (size > page_size)
>
> event size could be 0xffff - sizeof(struct perf_event_header)
> also the initial size is most likely > page_size anyway
>
> please put this into the patch that introduced __do_write_buf
>

Thank you for reviewing this. I just realized that I sent v5 without
properly addressing this issue. In v4 I tried to add a quick check
without properly understanding the issue. Why is 0xffff the maximum
size for an event? where is this constraint coming from?

Thanks,
David