On 09/29/11 13:16, Greg KH wrote: > On Thu, Sep 29, 2011 at 12:46:18PM -0700, Randy Dunlap wrote: >> On 09/29/11 11:50, Greg KH wrote: >>> On Thu, Sep 29, 2011 at 08:44:26PM +0200, Christian Schaubschläger wrote: >>>> >>>>>> Enabling ACPI is not an option for me in that case. >>>>> Why not? >>>> >>>> Well, the reason is the following: >>>> After starting the linux system I switch back the machine to real mode >>>> using the kexec system call (then I run some real/protected mode stuff). >>>> This works fine in most cases, but on some machines after kexec the >>>> machine is in a state where important things (eg. some bios interrupt >>>> calls) don't work any more. The reason for that is that the machine's >>>> interrupt controller/pci system/chipset is left in an 'unusable' state >>>> by the ACPI system... >>>> >>>> Disabling ACPI solved this issue on almost all machines; however, >>>> recently I had to deal with some machines with usb3, and on some of them >>>> I observed the behaviour I described before. >>>> >>>> Now I'm aware that this is not the typical usage scenario of a linux >>>> system, and I'm also aware that another option to get things working in >>>> real mode again would be trying to re-configure (or reset) the hardware >>>> after kexec; but this is not a trivial thing to do; for me it would be >>>> easiest to have xhci working without ACPI. That's why I was asking. >>> >>> Again, this isn't a xhci issue, ACPI is in charge of handling the PCI >>> irq routing here, so it is needed to have the pci devices work properly. >>> >>> I suggest you work with the Linux ACPI developers to solve your problem, >>> and that will then resolve the xhci issue as well. >> >> so (for arch/x86/Kconfig), this: >> >> config PCI >> bool "PCI support" >> default y >> select ARCH_SUPPORTS_MSI if (X86_LOCAL_APIC && X86_IO_APIC) >> >> should also say: >> select ACPI >> or >> depends on ACPI >> >> I hope not. > > No, but for this motherboard, it looks like yes, ACPI is required to get > PCI interrupt routing correct. It's been that way for a few years now > on a lot of different boards, nothing new here at all. > >> I do think it would be worthwhile to discuss this (xhci) problem >> with the kexec people. > > I think it would make more sense to discuss the acpi issue with the > kexec people... OK, it appears that one or the other of them is not being reset completely. Hopefully someone can help with that. -- ~Randy *** Remember to use Documentation/SubmitChecklist when testing your code *** -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html