lowlatency-2.2.14-B1 + 2.2.14aa7 fixes crash, but...

[William Montgomery]

The lowlatency-2.2.14-B1 patch applied to Andrea's 2.2.14aa7 kernel
no longer crashes on disk I/O during the rtc_latency test.
However, latency has increased to 1.4msec as reported by the tests.
This may be due to the fact that the patch does not apply cleanly
and my quick fixes were not complete.

Also, after 3 successful runs of rtc_latency, the test hung in the 
do_rtc_latency script while executing a sync command, the sync command
stayed in the "D" state ( uninterruptible sleep - IO).  Any ideas?

Wm

On Thu, 10 Feb 2000, Andrea Arcangeli wrote:

> On Thu, 10 Feb 2000, William Montgomery wrote:
> 
> >2.2.14.  I think I might try Andrea's 2.2.14aa6 kernel which I believe
> >has some MM bugfixes in it.  
> 
> (IIRC) none of my recent fixes cure oopses thus if something oopses you
> probably have a race in the lowlatency patch (I would guess you are
> sleeping a place were you can't sleep). Or maybe the patches trigger a
> real bug where some copy_user is called in a racy place or similar
> stuff. I have not checked it yet.
> 
> You are welcome to try out 2.2.14aa7, only be careful to _not_ compare
> 2.2.14 with 2.2.14aa7+lowlatency patch because I just have performance
> stuff in 2.2.14aa7.
> 
> Comparing 2.2.14aa7 and 2.2.14aa7+lowlatency is certainly ok instead.
> 
> Andrea
>