Is pci=nocrs bad?

[npaust]

I'm using a snazzy Dell Precision T3500 workstation and having some trouble. Basically, the 2.6.32 kernels worked perfectly fine but the 2.6.34 kernels fail to boot because of some issue with the SATA controller. There are numerous threads here in the gentoo forums and on the web talking about it, seems to be an issue with the ICH10 controller.

Anyway, the solution is to just add pci=nocrs to the kernel options and everything boots without a problem.

The question is this: will "pci=nocrs" cause any problems? I have to assume that using crs was chosen as a default for a reason. Should I be sticking with 2.6.32 kernels or just use the workaround for the 2.6.34 kernels and be happy?

[krinn]

stick with 2.6.32, an old kernel lack newer hardware support, this shouldn't be a problem as it already run on yours, and an old kernel might cause trouble with hal/udev... but for that you need a really older kernel than that one, and when it will be time to change, latest kernel will certainly have your issue corrected.

[roarinelk]

"pci=nocrs" is used to discard ACPI resource information for various devices. This was default
behavior before 2.6.34 (IIRC), but since then some improvements were made to resource handling
and "pci=use_crs" is now default for very recent BIOSes (I believe BIOS dates within the last
2 years).

If your BIOS has a recent date and you run into problems with pci=use_crs then report it to
LKML and/or create a kernel bugzilla entry so they can be fixed.