Re: [patch 1/3] basic accounting over taskstats

[Shailabh Nagar]

Jay Lan wrote:
> Shailabh Nagar wrote:
>
> > Andrew Morton wrote:
> >
> > > Or we remove this field altogether, perhaps.  The same info is available
> > > from /proc/pid/stat anyway.  Is it really needed?
> >
> > Gathering this data in userspace from /proc might be
> > difficult esp. for short-lived tasks.
>
>
> This is a serious concern. I think increasing TS_COMM_LEN
> to 32 would be a good solution.
>
> > Also, /proc may not be mounted ? I'd heard somewhere that
> > some sysadmins don't install /proc for security reasons.
> > Don't know how far thats true.
> >
> > Several other fields, totalling 58 bytes, added by the CSA
> > patches are also duplicated in /proc/pid/stat. But all of them
> > could change in value during the lifetime of a task so I'm
> > guessing its not useful to get them from /proc
> > even if some kind of userspace polling of the value was
> > possible.
>
>
> The same concern above applies to here, doesn't it?

Yes. For some of the fields, pid/ppid/nice/sched you may not
really care about whether you get the value present sometime
during the lifetime vs. value at the time task exited, but for
others like utime/stime/minflt/majflt, getting the last value
is likely to matter. Either way, since there's no guarantee
that you can poll a short-lived task fast enough to get any value,
exporting the value from the kernel at exit seems to be the only safe
way.

--Shailabh

> Regards,
>  - jay
>
>
> > But if there is a way, it would sure save a lot of payload
> > sent over taskstats !
> >
> > "duplicate" fields from CSA:
> > +    __u8    ac_nice;        /* task_nice */
> > +    char    ac_comm[TS_COMM_LEN];    /* Command name */
> > +    __u8    ac_sched;        /* Scheduling discipline */
> > +    __u32    ac_pid;            /* Process ID */
> > +    __u32    ac_ppid;        /* Parent process ID */
> > +    __u64    ac_utime;        /* User CPU time [usec] */
> > +    __u64    ac_stime;        /* SYstem CPU time [usec] */
> > +    __u64    ac_minflt;        /* Minor Page Fault */
> > +    __u64    ac_majflt;        /* Major Page Fault */

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