email thread between Bjorn and I: On Thursday 22 April 2010 01:42:01 pm R. Andrew Bailey wrote: > I noticed a couple of recent kernel bugs that you fielded, having to > do with achi controllers having issues at boot time. I think I've got > another example for you- this is a little Dell Optiplex 755 that > won't boot if the bios "SATA Operation" setting is configured for > "AHCI" and I don't append pci=nocrs. > > I've got /proc/iomem with the controller set to AHCI and pci=nocrs > below, but I couldn't get Bjorns debug patch to apply to 2.6.34-rc4, > was hoping maybe you had an updated patch I could try? > > Anyhow thanks in advance! I was about to open a bugzilla CR, but I > wasn't sure how to generate a boot log, or whether this might be a > reopening of an existing bug. Uh, oh. We're getting very late in the cycle, and we don't have much time to debug this. - These patches: https://patchwork.kernel.org/patch/90852/ https://patchwork.kernel.org/patch/93704/ are on their way upstream, but aren't there yet. Please test them first, just in case they fix the problem. If they do, we should go right out and buy some lottery tickets while our luck is good. But more likely, this is some new problem, and we'll have to debug it by collecting the information below. - The patch below applies to current upstream (c81eddb0e372). If you can apply it, set CONFIG_ACPI_DEBUG=y, and boot with "acpi.debug_level=0x00010000 acpi.debug_layer=0x00000100" for all the tests below, that would be great. - Boot with controller set to non-AHCI. This works, even though we're using _CRS. Right? Can you collect the dmesg log for this? - Boot with controller set to AHCI. This fails, right? Any log you can collect (e.g., boot with "ignore_loglevel" and use serial console, netconsole, digital video, photo, etc) would be helpful. - Boot with controller set to AHCI and "pci=nocrs". This works, right? I'd like to see the complete dmesg log.
Created attachment 26111 [details] dmesg with sata controller in ahci mode, acpi debug + patches, crs disabled
Created attachment 26112 [details] dmesg with sata controller in ata mode, acpi debug + patches, crs enabled
should be able to get you the abortive boot log this morning. My kingdom for a serial cable.
Created attachment 26113 [details] serial boot capture with ahci enabled, crs enabled
Thanks again for your help, Andy. I'm going to resolve this as a duplicate of https://bugzilla.kernel.org/show_bug.cgi?id=15744 so we can keep all the future discussion in one place. *** This bug has been marked as a duplicate of bug 15744 ***