> Machine is a PCI based pentium with an Adaptec 2940 and a wide seagate
> barracuda as the root drive. I've flashed the adaptec bios to 1.34 just in
> case that was the problem.
>
> The machine is still pingable and if I'm rlogged in, I can still execute
> commands from the buffer cache. The X process is hanging in `D' and the
> machine is almost fully idle.
>
> This is a UP machine. I've tried compiling with both egcs-1.0.2 and
> gcc-2.7.2.1, both UP and SMP with the same results.
>
> When the wedge occurs I can still access an ide based file-system, so I think
> the problem lies either in the scsi layer or the adaptec driver.
>
> ps -axln fails to work, but the kernel debugger says that the X process is in
> __wait_on_page().
>
> This has been happening since 2.1.94 with the aic7xxx-5.0.12-94 patches.
> Unpatched 2.1.94 works just fine, which is why I suspect the aic7xxx driver.
>
> However, I have 2 other machines (both SMP motherboards, one UP one dual)
> which don't exhibit this problem. They have on board aic7880 controllers
> rather than a 2940 card.
>
> I strongly suspect that a buffer completion gets lost in the adaptec layer
> under heavy load and will pursue this unless anyone has a better suggestion.
Unlikely, and should be fixed with the combination of my latest driver
(which I'm getting ready to upload but isn't on the ftp site yet) and
2.1.102. Most likely, it was a termination problem (fixed in 5.0.15) that
cause a reset condition to be needed. The reset condition then hung the
computer due to a problem in the mid level code (fixed in 2.1.102).
--Doug Ledford <dledford@dialnet.net> Opinions expressed are my own, but they should be everybody's.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu