Re: [ck] Re: Preempt Threshold Measurements

[Andrew Morton]

William Lee Irwin III <wli@xxxxxxxxxxxxxx> wrote:
>
> On Mon, Jul 12, 2004 at 10:48:50PM -0400, Gabriel Devenyi wrote:
>  > Well I'm not particularly educated in kernel internals yet, here's some 
>  > reports from the system when its running.
>  > 6ms non-preemptible critical section violated 4 ms preempt threshold starting 
>  > at do_munmap+0xd2/0x140 and ending at do_munmap+0xeb/0x140
>  >  [<c014007b>] do_munmap+0xeb/0x140
>  >  [<c01163b0>] dec_preempt_count+0x110/0x120
>  >  [<c014007b>] do_munmap+0xeb/0x140
>  >  [<c014010f>] sys_munmap+0x3f/0x60
>  >  [<c0103ee1>] sysenter_past_esp+0x52/0x71
> 
>  Looks like ZAP_BLOCK_SIZE may be too large for you. Lowering that some
>  may "help" this. It's probably harmless, but try lowering that to half
>  of whatever it is now, or maybe 64*PAGE_SIZE. It may be worthwhile
>  to restructure how the preemption points are done in unmap_vmas() so
>  we don't end up in some kind of tuning nightmare.

Does that instrumentation patch have the cond_resched_lock() fixups?  If
not, this is a false positive.

The current setting of ZAP_BLOCK_SIZE is good for sub-500usec latencies on
a recentish CPU.

-
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/