APM BIOS code has a protective wrapper that runs it only on CPU zero. Thus,
no need to set APM BIOS segments in the GDT for other CPUs.
Signed-off-by: Zachary Amsden <[email protected]>
Index: linux-2.6.14-zach-work/arch/i386/kernel/apm.c
===================================================================
--- linux-2.6.14-zach-work.orig/arch/i386/kernel/apm.c 2005-11-07 10:17:45.000000000 -0800
+++ linux-2.6.14-zach-work/arch/i386/kernel/apm.c 2005-11-07 13:36:05.000000000 -0800
@@ -2170,8 +2170,8 @@ static struct dmi_system_id __initdata a
static int __init apm_init(void)
{
struct proc_dir_entry *apm_proc;
+ struct desc_struct *gdt;
int ret;
- int i;
dmi_check_system(apm_dmi_table);
@@ -2253,18 +2253,17 @@ static int __init apm_init(void)
* not restrict themselves to their claimed limit. When this happens,
* they will cause a segmentation violation in the kernel at boot time.
* Most BIOS's, however, will respect a 64k limit, so we use that.
+ *
+ * Note we only set APM segments on CPU zero, since we pin the APM
+ * code to that CPU.
*/
- for (i = 0; i < NR_CPUS; i++) {
- struct desc_struct *gdt = get_cpu_gdt_table(i);
- if (!gdt)
- continue;
- set_base(&gdt[APM_CS >> 3],
- __va((unsigned long)apm_info.bios.cseg << 4));
- set_base(&gdt[APM_CS_16 >> 3],
- __va((unsigned long)apm_info.bios.cseg_16 << 4));
- set_base(&gdt[APM_DS >> 3],
- __va((unsigned long)apm_info.bios.dseg << 4));
- }
+ gdt = get_cpu_gdt_table(0);
+ set_base(&gdt[APM_CS >> 3],
+ __va((unsigned long)apm_info.bios.cseg << 4));
+ set_base(&gdt[APM_CS_16 >> 3],
+ __va((unsigned long)apm_info.bios.cseg_16 << 4));
+ set_base(&gdt[APM_DS >> 3],
+ __va((unsigned long)apm_info.bios.dseg << 4));
apm_proc = create_proc_info_entry("apm", 0, NULL, apm_get_info);
if (apm_proc)
* Zachary Amsden <[email protected]> wrote:
> APM BIOS code has a protective wrapper that runs it only on CPU zero.
> Thus, no need to set APM BIOS segments in the GDT for other CPUs.
hm, do we want (need) to have that CPU#0 assumption forever?
Ingo
Ingo Molnar wrote:
>* Zachary Amsden <[email protected]> wrote:
>
>
>
>>APM BIOS code has a protective wrapper that runs it only on CPU zero.
>>Thus, no need to set APM BIOS segments in the GDT for other CPUs.
>>
>>
>
>hm, do we want (need) to have that CPU#0 assumption forever?
>
>
Can't hurt, and APM is largely obsolete because of ACPI, so I'm only
concerned with trimming and keeping adequate protection of the kernel
from APM code while maintaining correctness. I don't have a nice set of
old machines with enough wacky APM BIOSen to validate that unpinning the
CPU is ok.
Zach
On Maw, 2005-11-08 at 04:53 -0800, Zachary Amsden wrote:
> Can't hurt, and APM is largely obsolete because of ACPI, so I'm only
> concerned with trimming and keeping adequate protection of the kernel
> from APM code while maintaining correctness. I don't have a nice set of
> old machines with enough wacky APM BIOSen to validate that unpinning the
> CPU is ok.
A large number of SMP machines, probably the majority of APM based ones
require that APM calls occur on CPU#0. As I understand it from a BIOS
engineer involved in debugging that problem Redmond always does APM from
CPU #0 and may even guarantee it.
Alan
* Alan Cox <[email protected]> wrote:
> On Maw, 2005-11-08 at 04:53 -0800, Zachary Amsden wrote:
> > Can't hurt, and APM is largely obsolete because of ACPI, so I'm only
> > concerned with trimming and keeping adequate protection of the kernel
> > from APM code while maintaining correctness. I don't have a nice set of
> > old machines with enough wacky APM BIOSen to validate that unpinning the
> > CPU is ok.
>
> A large number of SMP machines, probably the majority of APM based
> ones require that APM calls occur on CPU#0. As I understand it from a
> BIOS engineer involved in debugging that problem Redmond always does
> APM from CPU #0 and may even guarantee it.
ok, then i'm all for making that more explicit - i.e. Zachary's patch is
the right one.
Ingo
Ingo Molnar wrote:
> * Zachary Amsden <[email protected]> wrote:
>
>
>>APM BIOS code has a protective wrapper that runs it only on CPU zero.
>>Thus, no need to set APM BIOS segments in the GDT for other CPUs.
>
> hm, do we want (need) to have that CPU#0 assumption forever?
>
APM BIOS should only ever run on the BSP, and I believe in Linux the BSP
is always 0. Since APM is obsolete, there is no future to consider.
-hpa