Distribution: RHAS Hardware Environment: Hitachi HA8000/70 Software Environment: kernel.org kernel Problem Description: This hardware oopses on boot (due to a DSDT bug) if ACPI_BOOT is selected -- but ACPI_BOOT is forced to be enabled if SMP is on, even though the hardware works fine in SMP mode with ACPI_BOOT forcably turned off (and direct PCI bus configuration enabled). The kernel configuration tool should provide the ability to turn ACPI_BOOT off even on a SMP build. Steps to reproduce: Run "make menuconfig". Completely disable ACPI, enable SMP, select direct PCI bus configuration, exit and save, examine .config, notice that CONFIG_ACPI_BOOT is still enabled.
CONFIG_ACPI_BOOT is not a bug if acpi is disabled, since we must support HT. for BIOS error, _BBN error can't cause system oops, since you disabled ACPI. can you attach your config file? In my test, select ACPI_BOOT will not call any acpi bus related functions. I wonder why your does. thanks.
And for the BIOS error, could you please try the patch in Bug 1662. It should resolve your problem. thanks.
I'll grant that selecting CONFIG_ACPI_BOOT when SMP is enabled is a Good Thing by default (no doubt it wouldn't be in there if it weren't); the enhancement request I'm making is for a user-selectable option to force CONFIG_ACPI_BOOT to be disabled even when SMP is on, with appropriate help text warning the user about how disabling this on systems with HT support is a Bad Idea unless they really know what they're doing. And yes, the patch in bug#1662 does fix the oops on boot. (Incidentally, I've also put together a DSDT that boots correctly with all devices available).
At one point I had CONFIG_ACPI=n delete all ACPI code from the kernel. People complained that HT no longer worked b/c they didn't know that HT depended on ACPI. Marcelo mandated that HT should work even without CONFIG_ACPI or any explicit selection by the user in menuconfig. That is how CONFIG_ACPI_BOOT came to be; and 2.4 will not change. In 2.6, note that CONFIG_ACPI_BOOT doesn't depend on CONFIG_SMP, it depends on CONFIG_X86_HT. Maybe we can make CONFIG_X86_HT selectable and that would address the need in 2.6?
> Maybe we can make CONFIG_X86_HT selectable > and that would address the need in 2.6? WORKSFORME.
Created attachment 3565 [details] 2.6.8.1 patch The necessary fix is included in this Kconfig cleanup patch. let me know if you have any trouble with it.
the latest ACPI patches removes CONFIG_ACPI_BOOT altogether: http://ftp.kernel.org/pub/linux/kernel/people/lenb/acpi/patches/release/2.6.13/
CONFIG_ACPI_BOOT is gone as of 2.6.14-rc1 -- closing.