[PATCH] cpufreq: fix pending powernow timer stuck condition

AMD recently discovered that on some hardware, there is a race condition
possible when a C-state change request goes onto the bus at the same
time as a P-state change request.

Both requests happen, but the southbridge hardware only acknowledges the
C-state change.  The PowerNow! driver is then stuck in a loop, waiting
for the P-state change acknowledgement.  The driver eventually times
out, but can no longer perform P-state changes.

It turns out the solution is to resend the P-state change, which the
southbridge will acknowledge normally.

Thanks to Johannes Winkelmann for reporting this and testing the fix.

Signed-off-by: Mark Langsdorf <mark.langsdorf@amd.com>
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

1 files changed, 19 insertions, 11 deletions

diff --git a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
index ab6e0611303d..ec3a9e335aa3 100644
--- a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
+++ b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c
@@ -44,7 +44,7 @@
 
 #define PFX "powernow-k8: "
 #define BFX PFX "BIOS error: "
-#define VERSION "version 1.50.3"
+#define VERSION "version 1.50.4"
 #include "powernow-k8.h"
 
 /* serialize freq changes  */
@@ -111,8 +111,8 @@ static int query_current_values_with_pending_wait(struct powernow_k8_data *data)
         u32 i = 0;
 
         do {
-                if (i++ > 0x1000000) {
-                        printk(KERN_ERR PFX "detected change pending stuck\n");
+                if (i++ > 10000) {
+                        dprintk("detected change pending stuck\n");
                         return 1;
                 }
                 rdmsr(MSR_FIDVID_STATUS, lo, hi);
@@ -159,6 +159,7 @@ static int write_new_fid(struct powernow_k8_data *data, u32 fid)
 {
         u32 lo;
         u32 savevid = data->currvid;
+        u32 i = 0;
 
         if ((fid & INVALID_FID_MASK) || (data->currvid & INVALID_VID_MASK)) {
                 printk(KERN_ERR PFX "internal error - overflow on fid write\n");
@@ -170,10 +171,13 @@ static int write_new_fid(struct powernow_k8_data *data, u32 fid)
         dprintk("writing fid 0x%x, lo 0x%x, hi 0x%x\n",
                 fid, lo, data->plllock * PLL_LOCK_CONVERSION);
 
-        wrmsr(MSR_FIDVID_CTL, lo, data->plllock * PLL_LOCK_CONVERSION);
-
-        if (query_current_values_with_pending_wait(data))
-                return 1;
+        do {
+                wrmsr(MSR_FIDVID_CTL, lo, data->plllock * PLL_LOCK_CONVERSION);
+                if (i++ > 100) {
+                        printk(KERN_ERR PFX "internal error - pending bit very stuck - no further pstate changes possible\n");
+                        retrun 1;
+                }                       
+        } while (query_current_values_with_pending_wait(data));
 
         count_off_irt(data);
 
@@ -197,6 +201,7 @@ static int write_new_vid(struct powernow_k8_data *data, u32 vid)
 {
         u32 lo;
         u32 savefid = data->currfid;
+        int i = 0;
 
         if ((data->currfid & INVALID_FID_MASK) || (vid & INVALID_VID_MASK)) {
                 printk(KERN_ERR PFX "internal error - overflow on vid write\n");
@@ -208,10 +213,13 @@ static int write_new_vid(struct powernow_k8_data *data, u32 vid)
         dprintk("writing vid 0x%x, lo 0x%x, hi 0x%x\n",
                 vid, lo, STOP_GRANT_5NS);
 
-        wrmsr(MSR_FIDVID_CTL, lo, STOP_GRANT_5NS);
-
-        if (query_current_values_with_pending_wait(data))
-                return 1;
+        do {
+                wrmsr(MSR_FIDVID_CTL, lo, STOP_GRANT_5NS);
+                if (i++ > 100) {
+                        printk(KERN_ERR PFX "internal error - pending bit very stuck - no further pstate changes possible\n");
+                        return 1;
+                }
+        } while (query_current_values_with_pending_wait(data));
 
         if (savefid != data->currfid) {
                 printk(KERN_ERR PFX "fid changed on vid trans, old 0x%x new 0x%x\n",