From cd7240c0b900eb6d690ccee088a6c9b46dae815a Mon Sep 17 00:00:00 2001 From: Suresh Siddha Date: Thu, 19 Aug 2010 17:03:38 -0700 Subject: x86, tsc, sched: Recompute cyc2ns_offset's during resume from sleep states TSC's get reset after suspend/resume (even on cpu's with invariant TSC which runs at a constant rate across ACPI P-, C- and T-states). And in some systems BIOS seem to reinit TSC to arbitrary large value (still sync'd across cpu's) during resume. This leads to a scenario of scheduler rq->clock (sched_clock_cpu()) less than rq->age_stamp (introduced in 2.6.32). This leads to a big value returned by scale_rt_power() and the resulting big group power set by the update_group_power() is causing improper load balancing between busy and idle cpu's after suspend/resume. This resulted in multi-threaded workloads (like kernel-compilation) go slower after suspend/resume cycle on core i5 laptops. Fix this by recomputing cyc2ns_offset's during resume, so that sched_clock() continues from the point where it was left off during suspend. Reported-by: Florian Pritz Signed-off-by: Suresh Siddha Cc: # [v2.6.32+] Signed-off-by: Peter Zijlstra LKML-Reference: <1282262618.2675.24.camel@sbsiddha-MOBL3.sc.intel.com> Signed-off-by: Ingo Molnar --- arch/x86/kernel/tsc.c | 38 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 38 insertions(+) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c index ce8e50239332..d632934cb638 100644 --- a/arch/x86/kernel/tsc.c +++ b/arch/x86/kernel/tsc.c @@ -626,6 +626,44 @@ static void set_cyc2ns_scale(unsigned long cpu_khz, int cpu) local_irq_restore(flags); } +static unsigned long long cyc2ns_suspend; + +void save_sched_clock_state(void) +{ + if (!sched_clock_stable) + return; + + cyc2ns_suspend = sched_clock(); +} + +/* + * Even on processors with invariant TSC, TSC gets reset in some the + * ACPI system sleep states. And in some systems BIOS seem to reinit TSC to + * arbitrary value (still sync'd across cpu's) during resume from such sleep + * states. To cope up with this, recompute the cyc2ns_offset for each cpu so + * that sched_clock() continues from the point where it was left off during + * suspend. + */ +void restore_sched_clock_state(void) +{ + unsigned long long offset; + unsigned long flags; + int cpu; + + if (!sched_clock_stable) + return; + + local_irq_save(flags); + + get_cpu_var(cyc2ns_offset) = 0; + offset = cyc2ns_suspend - sched_clock(); + + for_each_possible_cpu(cpu) + per_cpu(cyc2ns_offset, cpu) = offset; + + local_irq_restore(flags); +} + #ifdef CONFIG_CPU_FREQ /* Frequency scaling support. Adjust the TSC based timer when the cpu frequency -- cgit v1.2.2 From 51e3c1b558b31b11bf5fc66d3c6f5adacf3573f7 Mon Sep 17 00:00:00 2001 From: Sergey Senozhatsky Date: Fri, 20 Aug 2010 10:36:34 +0300 Subject: x86, hwmon: Fix unsafe smp_processor_id() in thermal_throttle_add_dev Fix BUG: using smp_processor_id() in preemptible thermal_throttle_add_dev. We know the cpu number when calling thermal_throttle_add_dev, so we can remove smp_processor_id call in thermal_throttle_add_dev by supplying the cpu number as argument. This should resolve kernel bugzilla 16615/16629. Signed-off-by: Sergey Senozhatsky LKML-Reference: <20100820073634.GB5209@swordfish.minsk.epam.com> Cc: Fenghua Yu Cc: Joerg Roedel Cc: Maciej Rutecki Signed-off-by: H. Peter Anvin --- arch/x86/kernel/cpu/mcheck/therm_throt.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/mcheck/therm_throt.c b/arch/x86/kernel/cpu/mcheck/therm_throt.c index c2a8b26d4fea..d9368eeda309 100644 --- a/arch/x86/kernel/cpu/mcheck/therm_throt.c +++ b/arch/x86/kernel/cpu/mcheck/therm_throt.c @@ -202,10 +202,11 @@ static int therm_throt_process(bool new_event, int event, int level) #ifdef CONFIG_SYSFS /* Add/Remove thermal_throttle interface for CPU device: */ -static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev) +static __cpuinit int thermal_throttle_add_dev(struct sys_device *sys_dev, + unsigned int cpu) { int err; - struct cpuinfo_x86 *c = &cpu_data(smp_processor_id()); + struct cpuinfo_x86 *c = &cpu_data(cpu); err = sysfs_create_group(&sys_dev->kobj, &thermal_attr_group); if (err) @@ -251,7 +252,7 @@ thermal_throttle_cpu_callback(struct notifier_block *nfb, case CPU_UP_PREPARE: case CPU_UP_PREPARE_FROZEN: mutex_lock(&therm_cpu_lock); - err = thermal_throttle_add_dev(sys_dev); + err = thermal_throttle_add_dev(sys_dev, cpu); mutex_unlock(&therm_cpu_lock); WARN_ON(err); break; @@ -287,7 +288,7 @@ static __init int thermal_throttle_init_device(void) #endif /* connect live CPUs to sysfs */ for_each_online_cpu(cpu) { - err = thermal_throttle_add_dev(get_cpu_sysdev(cpu)); + err = thermal_throttle_add_dev(get_cpu_sysdev(cpu), cpu); WARN_ON(err); } #ifdef CONFIG_HOTPLUG_CPU -- cgit v1.2.2 From b7d460897739e02f186425b7276e3fdb1595cea7 Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Tue, 24 Aug 2010 22:44:12 -0700 Subject: x86, mm: Fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline rc2 kernel crashes when booting second cpu on this CONFIG_VMSPLIT_2G_OPT laptop: whereas cloning from kernel to low mappings pgd range does need to limit by both KERNEL_PGD_PTRS and KERNEL_PGD_BOUNDARY, cloning kernel pgd range itself must not be limited by the smaller KERNEL_PGD_BOUNDARY. Signed-off-by: Hugh Dickins LKML-Reference: Signed-off-by: H. Peter Anvin --- arch/x86/kernel/trampoline.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/trampoline.c b/arch/x86/kernel/trampoline.c index a874495b3673..e2a595257390 100644 --- a/arch/x86/kernel/trampoline.c +++ b/arch/x86/kernel/trampoline.c @@ -45,8 +45,7 @@ void __init setup_trampoline_page_table(void) /* Copy kernel address range */ clone_pgd_range(trampoline_pg_dir + KERNEL_PGD_BOUNDARY, swapper_pg_dir + KERNEL_PGD_BOUNDARY, - min_t(unsigned long, KERNEL_PGD_PTRS, - KERNEL_PGD_BOUNDARY)); + KERNEL_PGD_PTRS); /* Initialize low mappings */ clone_pgd_range(trampoline_pg_dir, -- cgit v1.2.2 From 1389298f7d14f179c2fa4b4b343079bd56082c70 Mon Sep 17 00:00:00 2001 From: Andreas Herrmann Date: Fri, 27 Aug 2010 11:20:07 +0200 Subject: x86, mcheck: Avoid duplicate sysfs links/files for thresholding banks kobject_add_internal failed for threshold_bank2 with -EEXIST, don't try to register things with the same name in the same directory: Pid: 1, comm: swapper Tainted: G W 2.6.31 #1 Call Trace: [] ? kobject_add_internal+0x156/0x180 [] ? kobject_add+0x66/0x6b [] ? kobject_init+0x42/0x82 [] ? kobject_create_and_add+0x34/0x63 [] ? threshold_create_bank+0x14f/0x259 [] ? mce_create_device+0x8d/0x1b8 [] ? threshold_init_device+0x3f/0x80 [] ? threshold_init_device+0x0/0x80 [] ? do_one_initcall+0x4f/0x143 [] ? kernel_init+0x14c/0x1a2 [] ? child_rip+0xa/0x20 [] ? kernel_init+0x0/0x1a2 [] ? child_rip+0x0/0x20 kobject_create_and_add: kobject_add error: -17 (Probably the for_each_cpu loop should be entirely removed.) Signed-off-by: Andreas Herrmann LKML-Reference: <20100827092006.GB5348@loge.amd.com> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/mcheck/mce_amd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd.c b/arch/x86/kernel/cpu/mcheck/mce_amd.c index 224392d8fe8c..5e975298fa81 100644 --- a/arch/x86/kernel/cpu/mcheck/mce_amd.c +++ b/arch/x86/kernel/cpu/mcheck/mce_amd.c @@ -530,7 +530,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) err = -ENOMEM; goto out; } - if (!alloc_cpumask_var(&b->cpus, GFP_KERNEL)) { + if (!zalloc_cpumask_var(&b->cpus, GFP_KERNEL)) { kfree(b); err = -ENOMEM; goto out; @@ -543,7 +543,7 @@ static __cpuinit int threshold_create_bank(unsigned int cpu, unsigned int bank) #ifndef CONFIG_SMP cpumask_setall(b->cpus); #else - cpumask_copy(b->cpus, c->llc_shared_map); + cpumask_set_cpu(cpu, b->cpus); #endif per_cpu(threshold_banks, cpu)[bank] = b; -- cgit v1.2.2 From 5ee5e97ee9bca919af11c562beeaf61741ad33f1 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 10 Sep 2010 22:32:53 +0200 Subject: x86, tsc: Fix a preemption leak in restore_sched_clock_state() A real life genuine preemption leak.. Reported-and-tested-by: Jeff Chua Signed-off-by: Peter Zijlstra Acked-by: Suresh Siddha Signed-off-by: Linus Torvalds --- arch/x86/kernel/tsc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/x86/kernel') diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c index d632934cb638..26a863a9c2a8 100644 --- a/arch/x86/kernel/tsc.c +++ b/arch/x86/kernel/tsc.c @@ -655,7 +655,7 @@ void restore_sched_clock_state(void) local_irq_save(flags); - get_cpu_var(cyc2ns_offset) = 0; + __get_cpu_var(cyc2ns_offset) = 0; offset = cyc2ns_suspend - sched_clock(); for_each_possible_cpu(cpu) -- cgit v1.2.2