Re: Add all thread stats for TASKSTATS_CMD_ATTR_TGID (v5)

[Oleg Nesterov <oleg@tv-sign.ru>]

On 09/20, Balbir Singh wrote:
>
> Andrew Morton wrote:
> > On Tue, 18 Sep 2007 00:23:39 +0200 Guillaume Chazarain <guichaz@yahoo.fr> wrote:
> > 
> >> TASKSTATS_CMD_ATTR_TGID used to return only the delay accounting stats, not
> >> the basic and extended accounting.  With this patch,
> >> TASKSTATS_CMD_ATTR_TGID also aggregates the accounting info for all threads
> >> of a thread group.  This makes TASKSTATS_CMD_ATTR_TGID usable in a similar
> >> fashion to TASKSTATS_CMD_ATTR_PID, for commands like iotop -P
> >> (http://guichaz.free.fr/misc/iotop.py).
> > 
> > This patch conflicts somewhat with
> > add-scaled-time-to-taskstats-based-process-accounting.patch
> > 
> > I fixed it up like this:
> > 
> > void bacct_add_tsk(struct taskstats *stats, struct task_struct *task)
> > {
> > 	if (task->flags & PF_SUPERPRIV)
> > 		stats->ac_flag |= ASU;
> > 	if (task->flags & PF_DUMPCORE)
> > 		stats->ac_flag |= ACORE;
> > 	if (task->flags & PF_SIGNALED)
> > 		stats->ac_flag |= AXSIG;
> > 	if (thread_group_leader(task) && (task->flags & PF_FORKNOEXEC))
> > 		/*
> > 		 * Threads are created by do_fork() and don't exec but not in
> > 		 * the AFORK sense, as the latter involves fork(2).
> > 		 */
> > 		stats->ac_flag |= AFORK;
> > 
> > 	stats->ac_utimescaled +=
> > 		cputime_to_msecs(task->utimescaled) * USEC_PER_MSEC;
> > 	stats->ac_stimescaled +=
> > 		cputime_to_msecs(task->stimescaled) * USEC_PER_MSEC;
> > 	stats->ac_utime	 += cputime_to_msecs(task->utime) * USEC_PER_MSEC;
> > 	stats->ac_stime	 += cputime_to_msecs(task->stime) * USEC_PER_MSEC;
> > 	stats->ac_minflt += task->min_flt;
> > 	stats->ac_majflt += task->maj_flt;
> > }
> > 
> > (note the s/=/+=/ in there) but it all needs reviewing and checking and
> > testing please.
> 
> Andrew,
> 
> Thanks for reviewing the patchset, this patch is on my review and test
> queue (which has gotten rather long of late). I'll test it further and
> get back.

I still think this version is very wrong. It makes the ->signal->stats
absolutely meaningless. Quoting myself:

	On 09/18, Guillaume Chazarain wrote:
	>
	> @@ -508,7 +543,7 @@ void taskstats_exit(struct task_struct *
	>    if (!stats)
	>            goto err;
	>
	> -  memcpy(stats, tsk->signal->stats, sizeof(*stats));
	> +  fill_tgid(tsk->pid, tsk, stats);

	No, no, this is wrong.

	tsk->signal->stats contains the accumulated info about the already exited
	threads, we shouldn't throw it out.

	Also, fill_tgid() doesn't make sense here, current is the last live sub-thread.

Oleg.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/