* [PATCH 8/8] speedstep-smi: reduce failures
@ 2003-11-19 18:34 Dominik Brodowski
0 siblings, 0 replies; only message in thread
From: Dominik Brodowski @ 2003-11-19 18:34 UTC (permalink / raw)
To: cpufreq, davej
Transitions initiated by speedstep-smi currently fail often if system load
is high. Bruno Ducrot and I have found out that the failures disappear if
DMA is disabled. As disabling (U)DMA isn't something nice, here's a
workaround which (almost) always works: retrying the SMI call up to five
times, and waiting some time before doing so. Please note that this will
completely mess up real-time-critical tasks or dynamic scaling. But
speedstep-smi (1st generation of speedstep) wasn't meant to be used
for that anyway.
arch/i386/kernel/cpu/cpufreq/speedstep-smi.c | 32 +++++++++++++++++++--------
1 files changed, 23 insertions(+), 9 deletions(-)
diff -ruN linux-original/arch/i386/kernel/cpu/cpufreq/speedstep-smi.c linux/arch/i386/kernel/cpu/cpufreq/speedstep-smi.c
--- linux-original/arch/i386/kernel/cpu/cpufreq/speedstep-smi.c 2003-11-19 17:22:33.657417080 +0100
+++ linux/arch/i386/kernel/cpu/cpufreq/speedstep-smi.c 2003-11-19 19:14:44.153226192 +0100
@@ -19,6 +19,7 @@
#include <linux/cpufreq.h>
#include <linux/pci.h>
#include <linux/slab.h>
+#include <linux/delay.h>
#include <asm/ist.h>
#include "speedstep-lib.h"
@@ -51,6 +52,10 @@
#define SET_SPEEDSTEP_STATE 2
#define GET_SPEEDSTEP_FREQS 4
+/* how often shall the SMI call be tried if it failed, e.g. because
+ * of DMA activity going on? */
+#define SMI_TRIES 5
+
/* DEBUG
* Define it if you want verbose debug output, e.g. for bug reporting
*/
@@ -140,10 +145,11 @@
*/
static void speedstep_set_state (unsigned int state, unsigned int notify)
{
- unsigned int old_state, result, command, new_state;
+ unsigned int old_state, result = 0, command, new_state;
unsigned long flags;
struct cpufreq_freqs freqs;
unsigned int function=SET_SPEEDSTEP_STATE;
+ unsigned int retry = 0;
if (state > 0x1)
return;
@@ -163,20 +169,28 @@
local_irq_save(flags);
command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
- __asm__ __volatile__(
- "movl $0, %%edi\n"
- "out %%al, (%%dx)\n"
- : "=b" (new_state), "=D" (result)
- : "a" (command), "b" (function), "c" (state), "d" (smi_port), "S" (0)
- );
+
+ do {
+ if (retry) {
+ dprintk(KERN_INFO "cpufreq: retry %u, previous result %u\n", retry, result);
+ mdelay(retry * 50);
+ }
+ retry++;
+ __asm__ __volatile__(
+ "movl $0, %%edi\n"
+ "out %%al, (%%dx)\n"
+ : "=b" (new_state), "=D" (result)
+ : "a" (command), "b" (function), "c" (state), "d" (smi_port), "S" (0)
+ );
+ } while ((new_state != state) && (retry <= SMI_TRIES));
/* enable IRQs */
local_irq_restore(flags);
if (new_state == state) {
- dprintk(KERN_INFO "cpufreq: change to %u MHz succeeded\n", (freqs.new / 1000));
+ dprintk(KERN_INFO "cpufreq: change to %u MHz succeeded after %u tries with result %u\n", (freqs.new / 1000), retry, result);
} else {
- printk(KERN_ERR "cpufreq: change failed\n");
+ printk(KERN_ERR "cpufreq: change failed with new_state %u and result %u\n", new_state, result);
}
if (notify)
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2003-11-19 18:34 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2003-11-19 18:34 [PATCH 8/8] speedstep-smi: reduce failures Dominik Brodowski
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox