Messages in this thread |  | | | Date | Sat, 19 Nov 2011 23:03:58 -0200 | | From | Arnaldo Carvalho de Melo <> | | Subject | Re: [PATCH 2/5] perf tool: Initialize events IDs in a single function |
| |
Em Sat, Nov 19, 2011 at 08:36:21PM -0500, Jiri Olsa escreveu: > On Fri, Nov 18, 2011 at 12:22:55PM -0200, Arnaldo Carvalho de Melo wrote: > > Em Fri, Nov 18, 2011 at 02:46:42PM +0100, Jiri Olsa escreveu: > > > Currently events' IDs initialization is scattered among the > > > code. Adding function 'perf_evlist__init_ids' to allocate > > > and retrive events' IDs in one place.
> > I'm ok with this one, will merge it in my perf/core branch together with > > a fix for a bug I noticed in the original code and that remains with > > your changes: consider what happens if we do a second perf_evlist__mmap, > > we will call perf_evlist__id_add_fd again and thus bump evsel->ids++ > > over nr_threads * nr_cpus, overflowing it.
> is there a valid scenario for this function to be called more than once? > I'd think that just ensuring this is called once is enough
I don't know, I just don't want to preclude that, crystal balls are not that reliable :-)
- Arnaldo
|  |