Re: [PATCH] Fix argument checking in sched_setaffinity
From: Paul Jackson
Date: Sat Sep 04 2004 - 22:50:15 EST
Linus wrote:
> /* We just assume that 8k CPU's aren't going to happen */
SGI doesn't so assume ;).
> but it's just a lot easier to do the "getaffinity" thing - if it fails,
> you can double the size of your buffer and try again. O(log(n)) rather
> than O(n) ;)
I agree. That's what my cpumask sizing loop does.
Well ... did.
Now it reads /sys/devices/system/node/node0/cpumap and computes the
size of the cpumask as an arithmetic function of the number of bytes
read (the ascii format uses 9 chars for each 32 bits of mask).
Either way works ...
My nodemask sizing code loops on get_mempolicy() calls of increasing
size, until they stop failing -EINVAL.
> Well, historically we _have_ required sizes to match.
I'm not sure what history you're looking at here, Linus.
Last weeks sys_sched_setaffinity didn't seem to require matching size,
only that user size is >= kernel size. The kernel ignored the extra
user bits.
For nodemask_t, well let me just say the mbind/mempolicy calls are different.
If we want to go in the direction of requiring sizes to match in the
'set' calls, then instead of this weeks changes to sys_sched_setaffinity
allowing user size < kernel size, shouldn't we be going the other way,
and tightening the check in kernel/sched.c:sys_sched_setaffinity(), from
what it was a week ago:
if (len < sizeof(new_mask))
return -EINVAL;
to:
if (len != sizeof(new_mask))
return -EINVAL;
Or at least reverting this last weeks changes back to the '<' check?
> I don't know how to sanely expose the damn things
How about:
$ cd /proc/sys/kernel
$ head sizeof*
==> sizeof_cpumask <==
64
==> sizeof_nodemask <==
32
--
I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <pj@xxxxxxx> 1.650.933.1373
-
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/