From e7bdd7a531320eb4a4a8160afbe0c7cc98ac7187 Mon Sep 17 00:00:00 2001 From: "Langsdorf, Mark" Date: Thu, 8 Jun 2006 10:30:17 -0500 Subject: [CPUFREQ] Clarify powernow-k8 cpu_family statements This patch clarifies the meaning of the cpu_family if statements in the hw pstate driver patch for powernow-k8 Signed-off-by: Mark Langsdorf Signed-off-by: Dave Jones --- arch/i386/kernel/cpu/cpufreq/powernow-k8.c | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c index b4277f58f40..d8be4b01488 100644 --- a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c +++ b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c @@ -120,7 +120,7 @@ static int pending_bit_stuck(void) { u32 lo, hi; - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) return 0; rdmsr(MSR_FIDVID_STATUS, lo, hi); @@ -136,7 +136,7 @@ static int query_current_values_with_pending_wait(struct powernow_k8_data *data) u32 lo, hi; u32 i = 0; - if (cpu_family) { + if (cpu_family == CPU_HW_PSTATE) { rdmsr(MSR_PSTATE_STATUS, lo, hi); i = lo & HW_PSTATE_MASK; rdmsr(MSR_PSTATE_DEF_BASE + i, lo, hi); @@ -598,7 +598,7 @@ static void print_basics(struct powernow_k8_data *data) int j; for (j = 0; j < data->numps; j++) { if (data->powernow_table[j].frequency != CPUFREQ_ENTRY_INVALID) { - if (cpu_family) { + if (cpu_family == CPU_HW_PSTATE) { printk(KERN_INFO PFX " %d : fid 0x%x gid 0x%x (%d MHz)\n", j, (data->powernow_table[j].index & 0xff00) >> 8, (data->powernow_table[j].index & 0xff0000) >> 16, data->powernow_table[j].frequency/1000); @@ -758,7 +758,7 @@ static int find_psb_table(struct powernow_k8_data *data) #ifdef CONFIG_X86_POWERNOW_K8_ACPI static void powernow_k8_acpi_pst_values(struct powernow_k8_data *data, unsigned int index) { - if (!data->acpi_data.state_count || cpu_family) + if (!data->acpi_data.state_count || (cpu_family == CPU_HW_PSTATE)) return; data->irt = (data->acpi_data.states[index].control >> IRT_SHIFT) & IRT_MASK; @@ -801,7 +801,7 @@ static int powernow_k8_cpu_init_acpi(struct powernow_k8_data *data) goto err_out; } - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) ret_val = fill_powernow_table_pstate(data, powernow_table); else ret_val = fill_powernow_table_fidvid(data, powernow_table); @@ -1082,7 +1082,7 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi if (query_current_values_with_pending_wait(data)) goto err_out; - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) dprintk("targ: curr fid 0x%x, did 0x%x\n", data->currfid, data->currvid); else { @@ -1103,7 +1103,7 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi powernow_k8_acpi_pst_values(data, newstate); - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) ret = transition_frequency_pstate(data, newstate); else ret = transition_frequency_fidvid(data, newstate); @@ -1115,7 +1115,7 @@ static int powernowk8_target(struct cpufreq_policy *pol, unsigned targfreq, unsi } mutex_unlock(&fidvid_mutex); - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) pol->cur = find_khz_freq_from_fiddid(data->currfid, data->currdid); else pol->cur = find_khz_freq_from_fid(data->currfid); @@ -1197,14 +1197,14 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol) if (query_current_values_with_pending_wait(data)) goto err_out; - if (!cpu_family) + if (cpu_family == CPU_OPTERON) fidvid_msr_init(); /* run on any CPU again */ set_cpus_allowed(current, oldmask); pol->governor = CPUFREQ_DEFAULT_GOVERNOR; - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) pol->cpus = cpumask_of_cpu(pol->cpu); else pol->cpus = cpu_core_map[pol->cpu]; @@ -1215,7 +1215,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol) pol->cpuinfo.transition_latency = (((data->rvo + 8) * data->vstable * VST_UNITS_20US) + (3 * (1 << data->irt) * 10)) * 1000; - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) pol->cur = find_khz_freq_from_fiddid(data->currfid, data->currdid); else pol->cur = find_khz_freq_from_fid(data->currfid); @@ -1232,7 +1232,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol) cpufreq_frequency_table_get_attr(data->powernow_table, pol->cpu); - if (cpu_family) + if (cpu_family == CPU_HW_PSTATE) dprintk("cpu_init done, current fid 0x%x, did 0x%x\n", data->currfid, data->currdid); else -- cgit v1.2.3-70-g09d2 From 6cad647da228486f36a9794137ad459e39b02590 Mon Sep 17 00:00:00 2001 From: "Langsdorf, Mark" Date: Thu, 8 Jun 2006 10:33:19 -0500 Subject: [CPUFREQ] correct powernow-k8 fid/vid masks for extended parts The fid/vid masks for parts using the extended parts are slightly incorrect and can result in incorrect fid/vid codes being applied. No instances of this problem have been reported in the field but it could be a problem with future parts. Signed-off-by: Mark Langsdorf Signed-off-by: Dave Jones --- arch/i386/kernel/cpu/cpufreq/powernow-k8.c | 4 ++-- arch/i386/kernel/cpu/cpufreq/powernow-k8.h | 4 +++- 2 files changed, 5 insertions(+), 3 deletions(-) diff --git a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c index d8be4b01488..756d0a333ed 100644 --- a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c +++ b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c @@ -885,8 +885,8 @@ static int fill_powernow_table_fidvid(struct powernow_k8_data *data, struct cpuf u32 vid; if (data->exttype) { - fid = data->acpi_data.states[i].status & FID_MASK; - vid = (data->acpi_data.states[i].status >> VID_SHIFT) & VID_MASK; + fid = data->acpi_data.states[i].status & EXT_FID_MASK; + vid = (data->acpi_data.states[i].status >> VID_SHIFT) & EXT_VID_MASK; } else { fid = data->acpi_data.states[i].control & FID_MASK; vid = (data->acpi_data.states[i].control >> VID_SHIFT) & VID_MASK; diff --git a/arch/i386/kernel/cpu/cpufreq/powernow-k8.h b/arch/i386/kernel/cpu/cpufreq/powernow-k8.h index bf8ad9e43da..0fb2a3001ba 100644 --- a/arch/i386/kernel/cpu/cpufreq/powernow-k8.h +++ b/arch/i386/kernel/cpu/cpufreq/powernow-k8.h @@ -169,7 +169,9 @@ struct powernow_k8_data { #define MVS_MASK 3 #define VST_MASK 0x7f #define VID_MASK 0x1f -#define FID_MASK 0x3f +#define FID_MASK 0x1f +#define EXT_VID_MASK 0x3f +#define EXT_FID_MASK 0x3f /* -- cgit v1.2.3-70-g09d2 From 491b07c98f2ac75f1a4370af76ae2403a4c579f5 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 21 Jun 2006 13:15:48 -0700 Subject: [PATCH] redirect speedstep-centrino maintainer mail to cpufreq list I haven't really maintained this driver for a while, and I'm not keeping up with the latest in Intel power management. I get a steady stream of mail which I don't really do anything useful with; the cpufreq list seems like a better destination, unless someone wants to get the mail directly. Also clean up a couple of ancient comments which don't really apply anymore (as far as I know, nobody has ever damaged a CPU with this driver). Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Dave Jones --- arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) diff --git a/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c b/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c index ce54ff12c15..f1a82c5de1b 100644 --- a/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c +++ b/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c @@ -2,19 +2,15 @@ * cpufreq driver for Enhanced SpeedStep, as found in Intel's Pentium * M (part of the Centrino chipset). * + * Since the original Pentium M, most new Intel CPUs support Enhanced + * SpeedStep. + * * Despite the "SpeedStep" in the name, this is almost entirely unlike * traditional SpeedStep. * * Modelled on speedstep.c * * Copyright (C) 2003 Jeremy Fitzhardinge - * - * WARNING WARNING WARNING - * - * This driver manipulates the PERF_CTL MSR, which is only somewhat - * documented. While it seems to work on my laptop, it has not been - * tested anywhere else, and it may not work for you, do strange - * things or simply crash. */ #include @@ -36,7 +32,7 @@ #include #define PFX "speedstep-centrino: " -#define MAINTAINER "Jeremy Fitzhardinge " +#define MAINTAINER "cpufreq@lists.linux.org.uk" #define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_DRIVER, "speedstep-centrino", msg) -- cgit v1.2.3-70-g09d2 From 9ed059e1551bf36092215b965838502ac21f42e4 Mon Sep 17 00:00:00 2001 From: Randy Dunlap Date: Tue, 20 Jun 2006 22:32:56 -0700 Subject: [CPUFREQ] Fix powernow-k8 SMP kernel on UP hardware bug. Fix powernow-k8 doesn't load bug. Reference: https://launchpad.net/distros/ubuntu/+source/linux-source-2.6.15/+bug/35145 Signed-off-by: Ben Collins Signed-off-by: Dave Jones --- arch/i386/kernel/cpu/cpufreq/powernow-k8.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c index 756d0a333ed..2d649167255 100644 --- a/arch/i386/kernel/cpu/cpufreq/powernow-k8.c +++ b/arch/i386/kernel/cpu/cpufreq/powernow-k8.c @@ -1163,7 +1163,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol) * Use the PSB BIOS structure. This is only availabe on * an UP version, and is deprecated by AMD. */ - if ((num_online_cpus() != 1) || (num_possible_cpus() != 1)) { + if (num_online_cpus() != 1) { printk(KERN_ERR PFX "MP systems not supported by PSB BIOS structure\n"); kfree(data); return -ENODEV; -- cgit v1.2.3-70-g09d2 From 4ec223d02f4d5f5a3129edc0e3d22550d6ac8a32 Mon Sep 17 00:00:00 2001 From: Venkatesh Pallipadi Date: Wed, 21 Jun 2006 15:18:34 -0700 Subject: [CPUFREQ] Fix ondemand vs suspend deadlock Rootcaused the bug to a deadlock in cpufreq and ondemand. Due to non-existent ordering between cpu_hotplug lock and dbs_mutex. Basically a race condition between cpu_down() and do_dbs_timer(). cpu_down() flow: * cpu_down() call for CPU 1 * Takes hot plug lock * Calls pre down notifier * cpufreq notifier handler calls cpufreq_driver_target() which takes cpu_hotplug lock again. OK as cpu_hotplug lock is recursive in same process context * CPU 1 goes down * Calls post down notifier * cpufreq notifier handler calls ondemand event stop which takes dbs_mutex So, cpu_hotplug lock is taken before dbs_mutex in this flow. do_dbs_timer is triggerred by a periodic timer event. It first takes dbs_mutex and then takes cpu_hotplug lock in cpufreq_driver_target(). Note the reverse order here compared to above. So, if this timer event happens at right moment during cpu_down, system will deadlok. Attached patch fixes the issue for both ondemand and conservative. Signed-off-by: Venkatesh Pallipadi Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq_conservative.c | 12 ++++++++++++ drivers/cpufreq/cpufreq_ondemand.c | 12 ++++++++++++ 2 files changed, 24 insertions(+) diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c index e07a35487bd..8878a154ed4 100644 --- a/drivers/cpufreq/cpufreq_conservative.c +++ b/drivers/cpufreq/cpufreq_conservative.c @@ -72,6 +72,14 @@ static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info); static unsigned int dbs_enable; /* number of CPUs using this policy */ +/* + * DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug + * lock and dbs_mutex. cpu_hotplug lock should always be held before + * dbs_mutex. If any function that can potentially take cpu_hotplug lock + * (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then + * cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock + * is recursive for the same process. -Venki + */ static DEFINE_MUTEX (dbs_mutex); static DECLARE_WORK (dbs_work, do_dbs_timer, NULL); @@ -414,12 +422,14 @@ static void dbs_check_cpu(int cpu) static void do_dbs_timer(void *data) { int i; + lock_cpu_hotplug(); mutex_lock(&dbs_mutex); for_each_online_cpu(i) dbs_check_cpu(i); schedule_delayed_work(&dbs_work, usecs_to_jiffies(dbs_tuners_ins.sampling_rate)); mutex_unlock(&dbs_mutex); + unlock_cpu_hotplug(); } static inline void dbs_timer_init(void) @@ -514,6 +524,7 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, break; case CPUFREQ_GOV_LIMITS: + lock_cpu_hotplug(); mutex_lock(&dbs_mutex); if (policy->max < this_dbs_info->cur_policy->cur) __cpufreq_driver_target( @@ -524,6 +535,7 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, this_dbs_info->cur_policy, policy->min, CPUFREQ_RELATION_L); mutex_unlock(&dbs_mutex); + unlock_cpu_hotplug(); break; } return 0; diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c index 3e6ffcaa5af..4d308410b60 100644 --- a/drivers/cpufreq/cpufreq_ondemand.c +++ b/drivers/cpufreq/cpufreq_ondemand.c @@ -71,6 +71,14 @@ static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info); static unsigned int dbs_enable; /* number of CPUs using this policy */ +/* + * DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug + * lock and dbs_mutex. cpu_hotplug lock should always be held before + * dbs_mutex. If any function that can potentially take cpu_hotplug lock + * (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then + * cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock + * is recursive for the same process. -Venki + */ static DEFINE_MUTEX (dbs_mutex); static DECLARE_WORK (dbs_work, do_dbs_timer, NULL); @@ -363,12 +371,14 @@ static void dbs_check_cpu(int cpu) static void do_dbs_timer(void *data) { int i; + lock_cpu_hotplug(); mutex_lock(&dbs_mutex); for_each_online_cpu(i) dbs_check_cpu(i); queue_delayed_work(dbs_workq, &dbs_work, usecs_to_jiffies(dbs_tuners_ins.sampling_rate)); mutex_unlock(&dbs_mutex); + unlock_cpu_hotplug(); } static inline void dbs_timer_init(void) @@ -469,6 +479,7 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, break; case CPUFREQ_GOV_LIMITS: + lock_cpu_hotplug(); mutex_lock(&dbs_mutex); if (policy->max < this_dbs_info->cur_policy->cur) __cpufreq_driver_target( @@ -479,6 +490,7 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, this_dbs_info->cur_policy, policy->min, CPUFREQ_RELATION_L); mutex_unlock(&dbs_mutex); + unlock_cpu_hotplug(); break; } return 0; -- cgit v1.2.3-70-g09d2