Re: 2.6.16.x CPUFREQ / SpeedStep-Centrino: couldn't enable Enchanced SpeedStep

[Dominik Brodowski <linux@dominikbrodowski.net>]

Hi,

On Tue, Jun 20, 2006 at 01:32:01PM -0700, Ben Kevan wrote:
> speedstep-centrino: P1 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P2 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P3 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P4 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P5 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P6 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P7 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P8 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P9 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P10 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P11 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P12 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P13 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P14 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: P15 has larger frequency (65535) than P0 (600), skipping
> speedstep-centrino: adding state 0 with frequency 600000 and control value 0610
> speedstep-centrino: adding state 1 with frequency 0 and control value ffff
> speedstep-centrino: adding state 2 with frequency 0 and control value ffff
> speedstep-centrino: adding state 3 with frequency 0 and control value ffff
> speedstep-centrino: adding state 4 with frequency 0 and control value ffff
> speedstep-centrino: adding state 5 with frequency 0 and control value ffff
> speedstep-centrino: adding state 6 with frequency 0 and control value ffff
> speedstep-centrino: adding state 7 with frequency 0 and control value ffff
> speedstep-centrino: adding state 8 with frequency 0 and control value ffff
> speedstep-centrino: adding state 9 with frequency 0 and control value ffff
> speedstep-centrino: adding state 10 with frequency 0 and control value ffff
> speedstep-centrino: adding state 11 with frequency 0 and control value ffff
> speedstep-centrino: adding state 12 with frequency 0 and control value ffff
> speedstep-centrino: adding state 13 with frequency 0 and control value ffff
> speedstep-centrino: adding state 14 with frequency 0 and control value ffff
> speedstep-centrino: adding state 15 with frequency 0 and control value ffff

As ACPI initialization succeeded, we didn't fall back to use the table. Does
this patch help?


[PATCH] speedstep_centrino: only use ACPI if it offers more than one state

Fall back to hard-coded tables in speedstep_centrino if ACPI only tells us
about one valid state.

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>

diff --git a/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c b/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c
index ce54ff1..5818d24 100644
--- a/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c
+++ b/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c
@@ -365,6 +365,7 @@ static int centrino_cpu_init_acpi(struct
 	unsigned long			cur_freq;
 	int				result = 0, i;
 	unsigned int			cpu = policy->cpu;
+	unsigned int			states = 0;
 
 	/* register with ACPI core */
 	if (acpi_processor_register_performance(&p, cpu)) {
@@ -452,6 +453,13 @@ static int centrino_cpu_init_acpi(struct
 
 		if (cur_freq == centrino_model[cpu]->op_points[i].frequency)
 			p.state = i;
+
+		states++;
+	}
+
+	if (states <= 1) {
+		result = -ENODEV;
+		goto err_kfree_all;
 	}
 
 	/* notify BIOS that we exist */