Still a problem (was: ATA write-dma interrupt was seen but timeout fired LBA=53346288)

On Sunday 29 August 2004 11:33 pm, Ari Suutari wrote:
> >> ATA write-dma interrupt was seen but timeout fired LBA=53346288
> >> ATA write-dma interrupt was seen but timeout fired LBA=53346288
> >> ATA write-dma interrupt was seen but taskqueue stalled LBA=53346288
> >
> >    I updated my old Compaq 1500c from 4.10 to -current last week.
> >    I'm seeing similar messages after resume and the system hangs
> > eventually.
> >    Otherwise the system seems to work ok.
>
>     Updated to recent current (28.8) seems to fix this. I haven't seen any
>     ATA interrupt/timeout messages after update. I assume that these
>     changes are not on RELENG_5 branch yet, since it looked like it
>     still suffers from the problems above.

This is still a problem on 5.3-BETA3... immediately after the taskqueue 
failure I get dropped into db with the panic

panic: Duplicate free of item 0xc12e7e76 from zone 0xc11cb000(g_bio)

When I first tried this I installed the apache package, and on boot it would 
panic before I could log in. I just did a minimal install, and I can log in 
and do stuff for a few seconds before panic.

5.2.1-R installs and boots fine on this system. As noted above, -CURRENT also 
would probably work, although I'd like to run RELENG_5.

On a side note, I'm appreciating the irony that I've run oh-so-unstable 
-CURRENT without a single problem for 2 years, and suddenly when 5-CURRENT is 
frozen for transition to -STABLE, I can't even boot it. (Probably because in 
-CURRENT fixes happen very fast, but even very critical fixes are taking a 
long time to make it to RELENG_5 these days...) Why are we even releasing 
BETAs when there are huge problems like this that have fixes in HEAD?

-David

-- 
+++ Divide By Cucumber Error. Please
Reinstall Universe And Reboot. +++