On Tue, 4 Feb 2014 20:00:07 +0100
Oleg Nesterov <o...@redhat.com> wrote:
> On 02/04, Steven Rostedt wrote:
> >
> > Now to fix this we need to save the filename before calling
> > search_binary_handler(). But we don't want to save it if we are not
> > tracing. Why slow everyone else down?
>
> Yes, but it would be much simpler to dup filename unconditionally.
I was hoping not to add more work to exec, for the purpose if tracing
is enabled.
>
> Note also that in this case we can kill linux_binprm->tcomm[] and
> simplify filename_to_taskname().
>
> > This works, but is rather ugly.
>
> Yes ;)
>
> > Looking for any other suggestions here.
>
> Perhaps we can change flush_old_exec() to do
>
> if (!current->mm) {
> bprm->filename = kstrdup(bprm->filename);
> if (bprm->filename)
> bprm->filename_was_dupped = true; // for free_bprm()
> else
> bprm->filename = "//enomem";
> }
>
> This won't penalize the normal exec, and this should fix the problem
> afaics.
Don't forget to add the free. And in any case, we added a branch
conditional. Yes, I'm that anal.
>
> Perhaps, instead of "//enomem" flush_old_exec() should simply fail,
> in this case we can kill bprm->tcomm[] too.
We could just add "//enomem" to the tracepoint too, and not use tcomm.
Also, I'm working on a patch that gets rid of all the setup of the
static_key and reuses the tracepoint static key instead (suggested by
Peter Zijlstra).
Thus, the change in exec.c would just look like this:
if (tracepoint_enabled(sched_process_exec)) {
[...]
} else
ret = search_binary_handler(bprm);
That is, we can get rid of the sched_process_exec_key_reg and unreg
functions, and just use this and it would become the same thing.
This will be helpful in other places too (David, I'm talking to you ;-)
-- Steve
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
