Re: Sluggishness in 2.6.7 caused by IDE stack

[Kenneth Johansson]

On Sat, 2004-09-25 at 16:11 +0200, Kenneth Johansson wrote:
> On Sat, 2004-09-25 at 07:26 -0500, Scott A Crosby wrote:
> > Much of the CPU time was spent in system mode. I setup a quick
> > oprofile, which blamed the function task_no_data_intr, but an
> > opannotate reports confusing results, possibly from interrupts? dmesg
> > reported nothing interesting.
> > 
> > 
> > Scott
> > 
> > 
> > *** vmstat output ***
> > 
> > procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> >  1  1 410496   8520  14224 1253716    0    0  2180    16  409   633 10 43  0 48
> >  7  1 410496   6920  14236 1255380    0    0  1668    20  589   832  9 18  0 73
> >  3  0 410496   8264  14232 1253988    0    0  2436    80  334   438 18 82  0  0
> >  0  1 410496   8384  14244 1253860    0    0  1932     0  464   706 13 35  0 52
> >  5  1 410496   8056  14264 1253872  328    0  2280     0  717   965 14 20  0 65
> >  8  1 410496   7352  14268 1254640    0    0  2692     0  351   485 17 83  0  0
> >  2  2 410496   6968  14264 1255036   32    0  2336     0  332   522 17 83  0  0
> >  5  0 410496   8568  14276 1253384    0    0  2192    12  464   794 19 33  0 48
> 
> I think you have the same problem as me. The interrupt rate drops under
> 1000 during use of the DVD and that is strange as the HZ is 1000 and
> that should be the lowest possible value unless I misunderstood
> something.  

I did some more testing and it turns out that I was not using the via
IDE driver but some generic code. I had the IDE driver as a module and
it was loaded but it did only drive the two unused ports the generic
driver had control over the DVD drive. I removed CONFIG_IDE_GENERIC and
compiled the via driver into the kernel and everything work OK now. 

I wonder why the generic code blocks interrupts but now I'm not that
interested in this problem anymore :)

Attachment: signature.asc
Description: This is a digitally signed message part