Merge centrino speedstep duplicate patches

arch/arm/mach-integrator/cpu.c

@@ -172,7 +172,7 @@ static int integrator_cpufreq_init(struct cpufreq_policy *policy)
 	policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
 	policy->cpuinfo.max_freq = 160000;
 	policy->cpuinfo.min_freq = 12000;
-	policy->cpuinfo.transition_latency = 1000; /* 1 ms, assumed */
+	policy->cpuinfo.transition_latency = 1000000; /* 1 ms, assumed */
 	policy->cur = policy->min = policy->max =
 		icst525_khz(&cclk_params, vco); /* current freq */
 

arch/i386/kernel/cpu/cpufreq/acpi.c

@@ -578,8 +578,8 @@ acpi_cpufreq_cpu_init (
 	/* detect transition latency */
 	policy->cpuinfo.transition_latency = 0;
 	for (i=0;i<perf->state_count;i++) {
-		if (perf->states[i].transition_latency > policy->cpuinfo.transition_latency)
-			policy->cpuinfo.transition_latency = perf->states[i].transition_latency;
+		if ((perf->states[i].transition_latency * 1000) > policy->cpuinfo.transition_latency)
+			policy->cpuinfo.transition_latency = perf->states[i].transition_latency * 1000;
 	}
 	policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
 	policy->cur = perf->states[pr->limit.state.px].core_frequency * 1000;

arch/i386/kernel/cpu/cpufreq/longhaul.c

@@ -170,6 +170,9 @@ static void longhaul_setstate (unsigned int clock_ratio_index)
  * between that value multiplied by possible FSBs and cpu_mhz which
  * was calculated at boot time. Really ugly, but no other way to do this.
  */
+
+#define ROUNDING	0xf
+
 static int _guess (int guess, int maxmult)
 {
 	int target;
@@ -177,16 +180,20 @@ static int _guess (int guess, int maxmult)
 	target = ((maxmult/10)*guess);
 	if (maxmult%10 != 0)
 		target += (guess/2);
-	target &= ~0xf;
+	target += ROUNDING/2;
+	target &= ~ROUNDING;
 	return target;
 }
 
 static int guess_fsb(int maxmult)
 {
-	int speed = (cpu_khz/1000) & ~0xf;
+	int speed = (cpu_khz/1000);
 	int i;
 	int speeds[3] = { 66, 100, 133 };
 
+	speed += ROUNDING/2;
+	speed &= ~ROUNDING;
+
 	for (i=0; i<3; i++) {
 		if (_guess(speeds[i],maxmult) == speed)
 			return speeds[i];

arch/i386/kernel/cpu/cpufreq/p4-clockmod.c

@@ -48,8 +48,7 @@ enum {
 
 
 static int has_N44_O17_errata[NR_CPUS];
-static int stock_freq;
-
+static unsigned int stock_freq;
 
 static int cpufreq_p4_setdc(unsigned int cpu, unsigned int newstate)
 {
@@ -175,6 +174,54 @@ static int cpufreq_p4_verify(struct cpufreq_policy *policy)
 	return cpufreq_frequency_table_verify(policy, &p4clockmod_table[0]);
 }
 
+/* copied from speedstep_lib, made SMP-compatible */
+static unsigned int cpufreq_p4_get_frequency(struct cpuinfo_x86 *c)
+{
+	u32 msr_lo, msr_hi, mult;
+	unsigned int fsb = 0;
+
+	if (c->x86 != 0xF) {
+		printk(KERN_DEBUG PFX "Unknown P4. Please send an e-mail to <linux@brodo.de>\n");
+		return 0;
+	}
+
+	rdmsr(0x2c, msr_lo, msr_hi);
+
+	/* printk(KERN_DEBUG PFX "P4 - MSR_EBC_FREQUENCY_ID: 0x%x 0x%x\n", msr_lo, msr_hi); */
+	/* decode the FSB: see IA-32 Intel (C) Architecture Software 
+	 * Developer's Manual, Volume 3: System Prgramming Guide,
+	 * revision #12 in Table B-1: MSRs in the Pentium 4 and
+	 * Intel Xeon Processors, on page B-4 and B-5.
+	 */
+	if (c->x86_model < 2)
+		fsb = 100 * 1000;
+	else {
+		u8 fsb_code = (msr_lo >> 16) & 0x7;
+		switch (fsb_code) {
+		case 0:
+			fsb = 100 * 1000;
+			break;
+		case 1:
+			fsb = 13333 * 10;
+			break;
+		case 2:
+			fsb = 200 * 1000;
+			break;
+		}
+	}
+
+	if (!fsb) {
+		printk(KERN_DEBUG PFX "couldn't detect FSB speed. Please send an e-mail to <linux@brodo.de>\n");
+		printk(KERN_DEBUG PFX "P4 - MSR_EBC_FREQUENCY_ID: 0x%x 0x%x\n", msr_lo, msr_hi);
+	}
+
+	/* Multiplier. */
+	mult = msr_lo >> 24;
+
+	return (fsb * mult);
+}
+
+ 
 
 static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
 {
@@ -192,15 +239,10 @@ static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
 		has_N44_O17_errata[policy->cpu] = 1;
 	}
 	
-	/* get frequency */
-	if (!stock_freq) {
-		if (cpu_khz)
-			stock_freq = cpu_khz;
-		else {
-			printk(KERN_INFO PFX "unknown core frequency - please use module parameter 'stock_freq'\n");
+	/* get max frequency */
+	stock_freq = cpufreq_p4_get_frequency(c);
+	if (!stock_freq)
 		return -EINVAL;
-		}
-	}
 
 	/* table init */
 	for (i=1; (p4clockmod_table[i].frequency != CPUFREQ_TABLE_END); i++) {
@@ -213,7 +255,7 @@ static int cpufreq_p4_cpu_init(struct cpufreq_policy *policy)
 	
 	/* cpuinfo and default policy values */
 	policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
-	policy->cpuinfo.transition_latency = 1000;
+	policy->cpuinfo.transition_latency = 1000000; /* assumed */
 	policy->cur = stock_freq;
 
 	return cpufreq_frequency_table_cpuinfo(policy, &p4clockmod_table[0]);
@@ -269,8 +311,6 @@ static void __exit cpufreq_p4_exit(void)
 }
 
 
-MODULE_PARM(stock_freq, "i");
-
 MODULE_AUTHOR ("Zwane Mwaikambo <zwane@commfireservices.com>");
 MODULE_DESCRIPTION ("cpufreq driver for Pentium(TM) 4/Xeon(TM)");
 MODULE_LICENSE ("GPL");

arch/i386/kernel/cpu/cpufreq/powernow-k7.c

@@ -337,7 +337,8 @@ static int powernow_decode_bios (int maxfid, int startvid)
 				}
 			}
 			printk (KERN_INFO PFX "No PST tables match this cpuid (0x%x)\n", etuple);
-			printk ("This is indicative of a broken BIOS. Email davej@redhat.com\n");
+			printk (KERN_INFO PFX "This is indicative of a broken BIOS.\n");
+			printk (KERN_INFO PFX "See http://www.codemonkey.org.uk/projects/cpufreq/powernow-k7.shtml\n");
 			return -EINVAL;
 		}
 		p++;
@@ -386,7 +387,11 @@ static int __init powernow_cpu_init (struct cpufreq_policy *policy)
 				minimum_speed, maximum_speed);
 
 	policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
-	policy->cpuinfo.transition_latency = latency;
+
+	/* latency is in 10 ns (look for SGTC above) for each VID
+	 * and FID transition, so multiply that value with 20 */
+	policy->cpuinfo.transition_latency = latency * 20;
+
 	policy->cur = maximum_speed;
 
 	return cpufreq_frequency_table_cpuinfo(policy, powernow_table);

arch/i386/kernel/cpu/cpufreq/powernow-k8.c

@@ -687,11 +687,13 @@ find_psb_table(void)
 			if (ppst[j].vid < rvo) {	/* vid+rvo >= 0 */
 				printk(KERN_ERR BFX
 				       "0 vid exceeded with pstate %d\n", j);
+				kfree(ppst);
 				return -ENODEV;
 			}
 			if (ppst[j].vid < maxvid+rvo) { /* vid+rvo >= maxvid */
 				printk(KERN_ERR BFX
 				       "maxvid exceeded with pstate %d\n", j);
+				kfree(ppst);
 				return -ENODEV;
 			}
 		}
@@ -706,7 +708,7 @@ find_psb_table(void)
 
 		for (j = 0; j < numps; j++)
 			if ((ppst[j].fid==currfid) && (ppst[j].vid==currvid))
-				return (0);
+				return 0;
 
 		printk(KERN_ERR BFX "currfid/vid do not match PST, ignoring\n");
 		return 0;
@@ -935,10 +937,7 @@ powernowk8_verify(struct cpufreq_policy *pol)
 		return -ENODEV;
 	}
 
-#warning pol->policy is in undefined state here
-	res = find_match(&targ, &min, &max,
-			 pol->policy == CPUFREQ_POLICY_POWERSAVE ?
-			 SEARCH_DOWN : SEARCH_UP, 0, 0);
+	res = find_match(&targ, &min, &max, SEARCH_DOWN, 0, 0);
 	if (!res) {
 		pol->min = min * 1000;
 		pol->max = max * 1000;
@@ -957,9 +956,10 @@ powernowk8_cpu_init(struct cpufreq_policy *pol)
 
 	pol->governor = CPUFREQ_DEFAULT_GOVERNOR;
 
-	/* Take a crude guess here. */
-	pol->cpuinfo.transition_latency = ((rvo + 8) * vstable * VST_UNITS_20US)
-	    + (3 * (1 << irt) * 10);
+	/* Take a crude guess here. 
+	 * That guess was in microseconds, so multply with 1000 */
+	pol->cpuinfo.transition_latency = (((rvo + 8) * vstable * VST_UNITS_20US)
+	    + (3 * (1 << irt) * 10)) * 1000;
 
 	if (query_current_values_with_pending_wait())
 		return -EIO;

arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c

@@ -75,13 +75,13 @@ static struct cpufreq_frequency_table op_900[] =
 
 /* Ultra Low Voltage Intel Pentium M processor 1000MHz */
 static struct cpufreq_frequency_table op_1000[] =
-  {
+{
 	OP(600,  844),
 	OP(800,  972),
 	OP(900,  988),
 	OP(1000, 1004),
 	{ .frequency = CPUFREQ_TABLE_END }
-  };
+};
 
 /* Low Voltage Intel Pentium M processor 1.10GHz */
 static struct cpufreq_frequency_table op_1100[] =

arch/i386/kernel/cpu/cpufreq/speedstep-lib.c

@@ -106,14 +106,45 @@ static unsigned int pentium3_get_frequency (unsigned int processor)
 
 static unsigned int pentium4_get_frequency(void)
 {
-	u32 msr_lo, msr_hi;
+	struct cpuinfo_x86 *c = &boot_cpu_data;
+	u32 msr_lo, msr_hi, mult;
+	unsigned int fsb = 0;
 
 	rdmsr(0x2c, msr_lo, msr_hi);
 
 	dprintk(KERN_DEBUG "speedstep-lib: P4 - MSR_EBC_FREQUENCY_ID: 0x%x 0x%x\n", msr_lo, msr_hi);
 
-	msr_lo >>= 24;
-	return (msr_lo * 100000);
+	/* decode the FSB: see IA-32 Intel (C) Architecture Software 
+	 * Developer's Manual, Volume 3: System Prgramming Guide,
+	 * revision #12 in Table B-1: MSRs in the Pentium 4 and
+	 * Intel Xeon Processors, on page B-4 and B-5.
+	 */
+	if (c->x86_model < 2)
+		fsb = 100 * 1000;
+	else {
+		u8 fsb_code = (msr_lo >> 16) & 0x7;
+		switch (fsb_code) {
+		case 0:
+			fsb = 100 * 1000;
+			break;
+		case 1:
+			fsb = 13333 * 10;
+			break;
+		case 2:
+			fsb = 200 * 1000;
+			break;
+		}
+	}
+
+	if (!fsb)
+		printk(KERN_DEBUG "speedstep-lib: couldn't detect FSB speed. Please send an e-mail to <linux@brodo.de>\n");
+
+	/* Multiplier. */
+	mult = msr_lo >> 24;
+
+	dprintk(KERN_DEBUG "speedstep-lib: P4 - FSB %u kHz; Multiplier %u\n", fsb, mult);
+
+	return (fsb * mult);
 }
 
  

arch/i386/kernel/cpu/cpufreq/speedstep-lib.h

@@ -15,7 +15,7 @@
 #define SPEEDSTEP_PROCESSOR_PIII_C_EARLY	0x00000001  /* Coppermine core */
 #define SPEEDSTEP_PROCESSOR_PIII_C		0x00000002  /* Coppermine core */
 #define SPEEDSTEP_PROCESSOR_PIII_T 		0x00000003  /* Tualatin core */
-#define SPEEDSTEP_PROCESSOR_P4M			0x00000004  /* P4-M with 100 MHz FSB */
+#define SPEEDSTEP_PROCESSOR_P4M			0x00000004  /* P4-M  */
 
 /* speedstep states -- only two of them */
 

arch/i386/kernel/cpu/cpufreq/speedstep-smi.c

@@ -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"
@@ -51,10 +52,14 @@ static struct cpufreq_frequency_table speedstep_freqs[] = {
 #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
  */
-#define SPEEDSTEP_DEBUG
+//#define SPEEDSTEP_DEBUG
 
 #ifdef SPEEDSTEP_DEBUG
 #define dprintk(msg...) printk(msg)
@@ -85,6 +90,9 @@ static int speedstep_smi_ownership (void)
 
 /**
  * speedstep_smi_get_freqs - get SpeedStep preferred & current freq.
+ * Only available on later SpeedStep-enabled systems, returns false results or
+ * even hangs [cf. bugme.osdl.org # 1422] on earlier systems. Empirical testing
+ * shows that the latter occurs if !(ist_info.event & 0xFFFF).
  *
  */
 static int speedstep_smi_get_freqs (unsigned int *low, unsigned int *high)
@@ -93,6 +101,9 @@ static int speedstep_smi_get_freqs (unsigned int *low, unsigned int *high)
 	u32 state=0;
 	u32 function = GET_SPEEDSTEP_FREQS;
 
+	if (!(ist_info.event & 0xFFFF))
+		return -ENODEV;
+
 	command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
 
 	__asm__ __volatile__("movl $0, %%edi\n"
@@ -134,10 +145,11 @@ static int speedstep_get_state (void)
  */
 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;
@@ -157,20 +169,28 @@ static void speedstep_set_state (unsigned int state, unsigned int notify)
 	local_irq_save(flags);
 
 	command = (smi_sig & 0xffffff00) | (smi_cmd & 0xff);
+
+	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 succeded\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)
@@ -225,9 +245,10 @@ static int speedstep_cpu_init(struct cpufreq_policy *policy)
 		return -ENODEV;
 
 	result = speedstep_smi_ownership();
-
-	if (result)
+	if (result) {
 		dprintk(KERN_INFO "cpufreq: fails an aquiring ownership of a SMI interface.\n");
+		return -EINVAL;
+	}
 
 	/* detect low and high frequency */
 	result = speedstep_smi_get_freqs(&speedstep_freqs[SPEEDSTEP_LOW].frequency,
@@ -286,6 +307,7 @@ static struct cpufreq_driver speedstep_driver = {
 	.target 	= speedstep_target,
 	.init		= speedstep_cpu_init,
 	.resume		= speedstep_resume,
+	.owner		= THIS_MODULE,
 };
 
 /**

arch/i386/kernel/timers/timer_tsc.c

@@ -317,9 +317,17 @@ static void mark_offset_tsc_hpet(void)
 }
 #endif
 
+
 #ifdef CONFIG_CPU_FREQ
+/* If the CPU frequency is scaled, TSC-based delays will need a different
+ * loops_per_jiffy value to function properly. An exception to this
+ * are modern Intel Pentium 4 processors, where the TSC runs at a constant
+ * speed independent of frequency scaling. 
+ */
+
 static unsigned int  ref_freq = 0;
 static unsigned long loops_per_jiffy_ref = 0;
+static unsigned int  variable_tsc = 1;
 
 #ifndef CONFIG_SMP
 static unsigned long fast_gettimeoffset_ref = 0;
@@ -344,13 +352,16 @@ time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
 
 	if ((val == CPUFREQ_PRECHANGE  && freq->old < freq->new) ||
 	    (val == CPUFREQ_POSTCHANGE && freq->old > freq->new)) {
+		if (variable_tsc)
 			cpu_data[freq->cpu].loops_per_jiffy = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new);
 #ifndef CONFIG_SMP
 		if (use_tsc) {
-			fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_ref, freq->new, ref_freq);
 			cpu_khz = cpufreq_scale(cpu_khz_ref, ref_freq, freq->new);
+			if (variable_tsc) {
+				fast_gettimeoffset_quotient = cpufreq_scale(fast_gettimeoffset_ref, freq->new, ref_freq);
 				set_cyc2ns_scale(cpu_khz/1000);
 			}
+		}
 #endif
 	}
 	write_sequnlock_irq(&xtime_lock);
@@ -361,6 +372,18 @@ time_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
 static struct notifier_block time_cpufreq_notifier_block = {
 	.notifier_call	= time_cpufreq_notifier
 };
+
+
+static int __init cpufreq_tsc(void)
+{
+	/* P4 and above CPU TSC freq doesn't change when CPU frequency changes*/
+	if ((boot_cpu_data.x86 >= 15) && (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL))
+		variable_tsc = 0;
+
+	return cpufreq_register_notifier(&time_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
+}
+core_initcall(cpufreq_tsc);
+
 #endif 
 
 
@@ -404,10 +427,6 @@ static int __init init_tsc(char* override)
  	 *	moaned if you have the only one in the world - you fix it!
  	 */
 
-#ifdef CONFIG_CPU_FREQ
-	cpufreq_register_notifier(&time_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
-#endif
-
 	count2 = LATCH; /* initialize counter for mark_offset_tsc() */
 
 	if (cpu_has_tsc) {

drivers/cpufreq/Kconfig

@@ -74,7 +74,7 @@ config CPU_FREQ_24_API
 	help
 	  This enables the /proc/sys/cpu/ sysctl interface for controlling
 	  the CPUFreq,"userspace" governor. This is the same interface
-	  as known from the.4.-kernel patches for CPUFreq, and offers
+	  as known from the 2.4.-kernel patches for CPUFreq, and offers
 	  the same functionality as long as "userspace" is the
 	  selected governor for the specified CPU.
 	

include/linux/cpufreq.h

@@ -57,7 +57,7 @@ struct cpufreq_governor;
 struct cpufreq_cpuinfo {
 	unsigned int		max_freq;
 	unsigned int		min_freq;
-	unsigned int		transition_latency; /* in 10^(-9) s */
+	unsigned int		transition_latency; /* in 10^(-9) s = nanoseconds */
 };
 
 struct cpufreq_real_policy {
@@ -231,17 +231,18 @@ int cpufreq_update_policy(unsigned int cpu);
 /* the proc_intf.c needs this */
 int cpufreq_parse_governor (char *str_governor, unsigned int *policy, struct cpufreq_governor **governor);
 
-#if defined(CONFIG_CPU_FREQ_GOV_USERSPACE) || defined(CONFIG_CPU_FREQ_GOV_USERSPACE_MODULE)
+
 /*********************************************************************
  *                      CPUFREQ USERSPACE GOVERNOR                   *
  *********************************************************************/
 int cpufreq_gov_userspace_init(void);
 
+#ifdef CONFIG_CPU_FREQ_24_API
+
 int cpufreq_setmax(unsigned int cpu);
 int cpufreq_set(unsigned int kHz, unsigned int cpu);
 unsigned int cpufreq_get(unsigned int cpu);
 
-#ifdef CONFIG_CPU_FREQ_24_API
 
 /* /proc/sys/cpu */
 enum {
@@ -289,8 +290,6 @@ enum {
 
 #endif /* CONFIG_CPU_FREQ_24_API */
 
-#endif /* CONFIG_CPU_FREQ_GOV_USERSPACE */
-
 
 /*********************************************************************
  *                       CPUFREQ DEFAULT GOVERNOR                    *
@@ -305,6 +304,7 @@ extern struct cpufreq_governor cpufreq_gov_userspace;
 #define CPUFREQ_DEFAULT_GOVERNOR	&cpufreq_gov_userspace
 #endif
 
+
 /*********************************************************************
  *                     FREQUENCY TABLE HELPERS                       *
  *********************************************************************/
@@ -318,7 +318,6 @@ struct cpufreq_frequency_table {
 				    * order */
 };
 
-#if defined(CONFIG_CPU_FREQ_TABLE) || defined(CONFIG_CPU_FREQ_TABLE_MODULE)
 int cpufreq_frequency_table_cpuinfo(struct cpufreq_policy *policy,
 				    struct cpufreq_frequency_table *table);
 
@@ -340,5 +339,4 @@ void cpufreq_frequency_table_get_attr(struct cpufreq_frequency_table *table,
 void cpufreq_frequency_table_put_attr(unsigned int cpu);
 
 
-#endif /* CONFIG_CPU_FREQ_TABLE */
 #endif /* _LINUX_CPUFREQ_H */