Re: [PATCH v5 16/18] perf: enable reading of perf.data files from different ABI rev

[Stephane Eranian <eranian@google.com>]

On Tue, Feb 7, 2012 at 5:41 PM, David Ahern <dsahern@gmail.com> wrote:
> On 02/07/2012 08:50 AM, Stephane Eranian wrote:
>>>> +     if (sz == 0) {
>>>> +             /* assume ABI0 */
>>>> +             sz =  PERF_ATTR_SIZE_VER0;
>>>
>>> Shouldn't this be a failure? ie., problem with the file (or the
>>> swapping) since size can't be 0
>>>
>> size can be zero. In which case, it means ABI0 version.
>> See kernel/event/core.c:perf_copy_attr().
>
> ok
>
>>
>>
>>> And then for the following why not restrict sz to known, expected sizes
>>> -- using the PERF_ATTR_SIZE_VER defines introduced in patch 15?
>>>
>> Well, the current code solves the problem once and for all. Old tools
>> can still read new files and vice-versa. If you think that's a problem I
>> can simply bail out if sz > our_sz.
>
> My sensitivity on this is when endianness is broken it is a nightmare to
> find. You end up lacing the code with printfs trying to find which size
> field is going off the charts making the parsing of the file fail - or
> worse the sizes are slightly off and you get non-sense out.
>
> New commands should be able to read old files; old commands reading new
> files is a bit of a stretch in that the code has to be future-proofed.
> It seems like a reasonable requirement for data files to be examined by
> a command of the same vintage or newer as the command that wrote the file.
>
Fine then, I can simply strip that part of the code and return an
error is sz > our_sz.
How about that?