Re: kern/36824 (CF card reading creates 97% interrupt storm)

["George Michaelson" <ggm%pobox.com@localhost>]

ok. I installed current as of last night (24th May) and this is still a problem.

the symptoms I see are, 99% to 100% interrupt shown in top while doing
a CP from an MS-DOS camera DCIM dirtree off the CF
card, into a FFS1 /var/tmp.

once the first copy is done, its in filecache and subsequent copies
are almost immediate. but if you umount/mount the CF device and
repeat, the interrupt count goes right back up to 99% during the copy.

X becomes un-responsive, window start (eg twm -> new xterm) is
sluggish in the extreme, cursor disappears, but a top window running
continues to update and shows the high interrupt rate.

I don't want to over-state this as a major problem btw. it appears now
to be very contained to doing active copy off the device, and it stops
as soon as the filebuf cache is full. More recent kernels may well
have reduced the impact of this.

the dmesg on the card is:

May 24 20:37:22  /netbsd: wdc2 at pcmcia1 function 0: <SanDisk, SDP, 5/3 0.6>
May 24 20:37:22  /netbsd: wdc2: i/o mapped mode
May 24 20:37:22  /netbsd: atabus2 at wdc2 channel 0
May 24 20:37:22  /netbsd: wd1 at atabus2 drive 0
May 24 20:37:22  /netbsd: : <SanDisk SDCFH-2048>
May 24 20:37:22  /netbsd: wd1: drive supports 4-sector PIO transfers,
LBA addressing
May 24 20:37:22  /netbsd: wd1: 1953 MB, 3970 cyl, 16 head, 63 sec, 512
bytes/sect x 4001760 sectors
May 24 20:37:22  /netbsd: wd1: drive supports PIO mode 4

there are no kernel msgs or logs about problems. Its just about user
responsiveness, and the interrupt count shown by top.

any more useful info or debug I can do?

-George

On Fri, May 23, 2008 at 6:55 AM,  <joerg%netbsd.org@localhost> wrote:
> Synopsis: CF card reading creates 97% interrupt storm
>
> Responsible-Changed-From-To: kern-bug-people->joerg
> Responsible-Changed-By: joerg%NetBSD.org@localhost
> Responsible-Changed-When: Thu, 22 May 2008 20:55:33 +0000
> Responsible-Changed-Why:
> Sounds like an interrupt issue.
>
>
> State-Changed-From-To: open->feedback
> State-Changed-By: joerg%NetBSD.org@localhost
> State-Changed-When: Thu, 22 May 2008 20:55:33 +0000
> State-Changed-Why:
> Can you please update to newer current and see if the problem
> still exists? I've fixed an important bug that could result in
> cbb not getting an interupt with the results as you have.
>
>
>
>