Attachment 174384 Details for Bug 299891

[patch] Ignore bogus CPU frequency values wrongly exported by ACPI layer early

acpi_powernow_workaround_wrong_data.patch (text/plain), 2.58 KB, created by Thomas Renninger on 2007-09-24 15:20:00 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Thomas Renninger

Created: 2007-09-24 15:20:00 UTC

Size: 2.58 KB

patch

obsolete

>From: Thomas Renninger <trenn@suse.de>
>Subject: Ignore bogus CPU freq values wrongly exported by acpi to powernow-k8
>Patch-mainline: not yet
>References: #299891 and #297119
>
>This is in fact an ACPI sublayer bug. The amount of packages is ignored and
>invalid (e.g. 0xFFFF/0xFFFFFFFF for frequency/control) info is passed to the
>powernow-k8 driver.
>This only affects BIOSes which fill up frequency info with these dummy values
>so that they do not need to allocate the memory for different processor
>frequency info dynamically.
>Known BIOSes which show this issue currently are:
>Gigabyte GA-K8N (Athlon) (Bug #297119) -> detects the invalid freqs and does
>                                          not show bad side effects, with this
>                                          patch it's still safer...
>MSI K8N Master2FAR (Opteron) (Bug #299891)
>                    -> Opteron and Athlon differ in Freq initialisation, here
>                       here HW is addressed with wrong info and the machine
>                       powers off.
>
>As this is a general ACPI bug, other non-AMD/powernow-k8 systems may also
>be affected, but it's too late for fiddling in ACPI parser and
>this is the safest thing to do for 10.3.
>                               
>---
> arch/i386/kernel/cpu/cpufreq/powernow-k8.c |   14 ++++++++++++++
> 1 file changed, 14 insertions(+)
>
>Index: linux-2.6.22/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
>===================================================================
>--- linux-2.6.22.orig/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
>+++ linux-2.6.22/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
>@@ -846,6 +846,13 @@ static int fill_powernow_table_pstate(st
> 		u32 fid;
> 		u32 did;
> 
>+		if (data->acpi_data.states[i].core_frequency == 0xFFFF ||
>+		    data->acpi_data.states[i].control == 0xFFFFFFFF) {
>+			printk(KERN_INFO PFX "Dummy frequency entry hit - ignoring\n");
>+			powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
>+			continue;
>+		}
>+		
> 		index = data->acpi_data.states[i].control & HW_PSTATE_MASK;
> 		if (index > MAX_HW_PSTATE) {
> 			printk(KERN_ERR PFX "invalid pstate %d - bad value %d.\n", i, index);
>@@ -886,6 +893,13 @@ static int fill_powernow_table_fidvid(st
> 		u32 fid;
> 		u32 vid;
> 
>+		if (data->acpi_data.states[i].core_frequency == 0xFFFF ||
>+		    data->acpi_data.states[i].control == 0xFFFFFFFF) {
>+			printk(KERN_INFO PFX "Dummy frequency entry hit - ignoring\n");
>+			powernow_table[i].frequency = CPUFREQ_ENTRY_INVALID;
>+			continue;
>+		}
>+
> 		if (data->exttype) {
> 			fid = data->acpi_data.states[i].status & EXT_FID_MASK;
> 			vid = (data->acpi_data.states[i].status >> VID_SHIFT) & EXT_VID_MASK;

Actions: View | Diff

Attachments on bug 299891: 157786 | 157787 | 158046 | 164149 | 164150 | 173419 | 173480 | 173544 | 173793 | 174384 | 175114 | 175489