From ae77930060338226a4377d3b93580c43b5ec82ae Mon Sep 17 00:00:00 2001 From: Mel Gorman Date: Wed, 6 Aug 2014 14:19:18 +0100 Subject: cpuidle: menu: Use shifts when calculating averages where possible We use do_div even though the divisor will usually be a power-of-two unless there are unusual outliers. Use shifts where possible Signed-off-by: Mel Gorman Signed-off-by: Rafael J. Wysocki --- drivers/cpuidle/governors/menu.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c index c3732fa..c36e1ea 100644 --- a/drivers/cpuidle/governors/menu.c +++ b/drivers/cpuidle/governors/menu.c @@ -31,7 +31,8 @@ * The default values do not overflow. */ #define BUCKETS 12 -#define INTERVALS 8 +#define INTERVAL_SHIFT 3 +#define INTERVALS (1UL << INTERVAL_SHIFT) #define RESOLUTION 1024 #define DECAY 8 #define MAX_INTERESTING 50000 @@ -227,7 +228,10 @@ again: max = value; } } - do_div(avg, divisor); + if (divisor == INTERVALS) + avg >>= INTERVAL_SHIFT; + else + do_div(avg, divisor); /* Then try to determine standard deviation */ stddev = 0; @@ -238,7 +242,11 @@ again: stddev += diff * diff; } } - do_div(stddev, divisor); + if (divisor == INTERVALS) + stddev >>= INTERVAL_SHIFT; + else + do_div(stddev, divisor); + /* * The typical interval is obtained when standard deviation is small * or standard deviation is small compared to the average interval. -- cgit v1.1 From 107d4f4601a1408d04a5b54ffba507c92c235f58 Mon Sep 17 00:00:00 2001 From: Mel Gorman Date: Wed, 6 Aug 2014 14:19:19 +0100 Subject: cpuidle: menu: Use ktime_to_us instead of reinventing the wheel The ktime_to_us implementation is slightly better than the one implemented in menu.c. Use it Signed-off-by: Mel Gorman Acked-by: Daniel Lezcano Signed-off-by: Rafael J. Wysocki --- drivers/cpuidle/governors/menu.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c index c36e1ea..ba6df60 100644 --- a/drivers/cpuidle/governors/menu.c +++ b/drivers/cpuidle/governors/menu.c @@ -296,7 +296,6 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) int latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY); int i; unsigned int interactivity_req; - struct timespec t; if (data->needs_update) { menu_update(drv, dev); @@ -310,9 +309,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) return 0; /* determine the expected residency time, round up */ - t = ktime_to_timespec(tick_nohz_get_sleep_length()); - data->next_timer_us = - t.tv_sec * USEC_PER_SEC + t.tv_nsec / NSEC_PER_USEC; + data->next_timer_us = ktime_to_us(tick_nohz_get_sleep_length()); data->bucket = which_bucket(data->next_timer_us); -- cgit v1.1 From 64b4ca5cb6e1a9f577588db5765dc996ddf595e1 Mon Sep 17 00:00:00 2001 From: Mel Gorman Date: Wed, 6 Aug 2014 14:19:20 +0100 Subject: cpuidle: menu: Call nr_iowait_cpu less times menu_select() via inline functions calls nr_iowait_cpu() twice as much as necessary. Signed-off-by: Mel Gorman Signed-off-by: Rafael J. Wysocki --- drivers/cpuidle/governors/menu.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c index ba6df60..f55d826 100644 --- a/drivers/cpuidle/governors/menu.c +++ b/drivers/cpuidle/governors/menu.c @@ -142,7 +142,7 @@ static int get_loadavg(void) return LOAD_INT(this) * 10 + LOAD_FRAC(this) / 10; } -static inline int which_bucket(unsigned int duration) +static inline int which_bucket(unsigned int duration, unsigned long nr_iowaiters) { int bucket = 0; @@ -152,7 +152,7 @@ static inline int which_bucket(unsigned int duration) * This allows us to calculate * E(duration)|iowait */ - if (nr_iowait_cpu(smp_processor_id())) + if (nr_iowaiters) bucket = BUCKETS/2; if (duration < 10) @@ -175,7 +175,7 @@ static inline int which_bucket(unsigned int duration) * to be, the higher this multiplier, and thus the higher * the barrier to go to an expensive C state. */ -static inline int performance_multiplier(void) +static inline int performance_multiplier(unsigned long nr_iowaiters) { int mult = 1; @@ -184,7 +184,7 @@ static inline int performance_multiplier(void) mult += 2 * get_loadavg(); /* for IO wait tasks (per cpu!) we add 5x each */ - mult += 10 * nr_iowait_cpu(smp_processor_id()); + mult += 10 * nr_iowaiters; return mult; } @@ -296,6 +296,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) int latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY); int i; unsigned int interactivity_req; + unsigned long nr_iowaiters; if (data->needs_update) { menu_update(drv, dev); @@ -311,8 +312,8 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) /* determine the expected residency time, round up */ data->next_timer_us = ktime_to_us(tick_nohz_get_sleep_length()); - - data->bucket = which_bucket(data->next_timer_us); + nr_iowaiters = nr_iowait_cpu(smp_processor_id()); + data->bucket = which_bucket(data->next_timer_us, nr_iowaiters); /* * Force the result of multiplication to be 64 bits even if both @@ -330,7 +331,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) * duration / latency ratio. Adjust the latency limit if * necessary. */ - interactivity_req = data->predicted_us / performance_multiplier(); + interactivity_req = data->predicted_us / performance_multiplier(nr_iowaiters); if (latency_req > interactivity_req) latency_req = interactivity_req; -- cgit v1.1 From 372ba8cb46b271a7662b92cbefedee56725f6bd0 Mon Sep 17 00:00:00 2001 From: Mel Gorman Date: Wed, 6 Aug 2014 14:19:21 +0100 Subject: cpuidle: menu: Lookup CPU runqueues less The menu governer makes separate lookups of the CPU runqueue to get load and number of IO waiters but it can be done with a single lookup. Signed-off-by: Mel Gorman Signed-off-by: Rafael J. Wysocki --- drivers/cpuidle/governors/menu.c | 17 +++++++---------- include/linux/sched.h | 3 +-- kernel/sched/core.c | 7 +++++++ kernel/sched/proc.c | 7 ------- 4 files changed, 15 insertions(+), 19 deletions(-) diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c index f55d826..2770274 100644 --- a/drivers/cpuidle/governors/menu.c +++ b/drivers/cpuidle/governors/menu.c @@ -134,12 +134,9 @@ struct menu_device { #define LOAD_INT(x) ((x) >> FSHIFT) #define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100) -static int get_loadavg(void) +static inline int get_loadavg(unsigned long load) { - unsigned long this = this_cpu_load(); - - - return LOAD_INT(this) * 10 + LOAD_FRAC(this) / 10; + return LOAD_INT(load) * 10 + LOAD_FRAC(load) / 10; } static inline int which_bucket(unsigned int duration, unsigned long nr_iowaiters) @@ -175,13 +172,13 @@ static inline int which_bucket(unsigned int duration, unsigned long nr_iowaiters * to be, the higher this multiplier, and thus the higher * the barrier to go to an expensive C state. */ -static inline int performance_multiplier(unsigned long nr_iowaiters) +static inline int performance_multiplier(unsigned long nr_iowaiters, unsigned long load) { int mult = 1; /* for higher loadavg, we are more reluctant */ - mult += 2 * get_loadavg(); + mult += 2 * get_loadavg(load); /* for IO wait tasks (per cpu!) we add 5x each */ mult += 10 * nr_iowaiters; @@ -296,7 +293,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) int latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY); int i; unsigned int interactivity_req; - unsigned long nr_iowaiters; + unsigned long nr_iowaiters, cpu_load; if (data->needs_update) { menu_update(drv, dev); @@ -312,7 +309,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) /* determine the expected residency time, round up */ data->next_timer_us = ktime_to_us(tick_nohz_get_sleep_length()); - nr_iowaiters = nr_iowait_cpu(smp_processor_id()); + get_iowait_load(&nr_iowaiters, &cpu_load); data->bucket = which_bucket(data->next_timer_us, nr_iowaiters); /* @@ -331,7 +328,7 @@ static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev) * duration / latency ratio. Adjust the latency limit if * necessary. */ - interactivity_req = data->predicted_us / performance_multiplier(nr_iowaiters); + interactivity_req = data->predicted_us / performance_multiplier(nr_iowaiters, cpu_load); if (latency_req > interactivity_req) latency_req = interactivity_req; diff --git a/include/linux/sched.h b/include/linux/sched.h index 306f4f0..641bd95 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -168,8 +168,7 @@ extern int nr_processes(void); extern unsigned long nr_running(void); extern unsigned long nr_iowait(void); extern unsigned long nr_iowait_cpu(int cpu); -extern unsigned long this_cpu_load(void); - +extern void get_iowait_load(unsigned long *nr_waiters, unsigned long *load); extern void calc_global_load(unsigned long ticks); extern void update_cpu_load_nohz(void); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 3bdf01b..863ef1d 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2385,6 +2385,13 @@ unsigned long nr_iowait_cpu(int cpu) return atomic_read(&this->nr_iowait); } +void get_iowait_load(unsigned long *nr_waiters, unsigned long *load) +{ + struct rq *this = this_rq(); + *nr_waiters = atomic_read(&this->nr_iowait); + *load = this->cpu_load[0]; +} + #ifdef CONFIG_SMP /* diff --git a/kernel/sched/proc.c b/kernel/sched/proc.c index 16f5a30..8ecd552 100644 --- a/kernel/sched/proc.c +++ b/kernel/sched/proc.c @@ -8,13 +8,6 @@ #include "sched.h" -unsigned long this_cpu_load(void) -{ - struct rq *this = this_rq(); - return this->cpu_load[0]; -} - - /* * Global load-average calculations * -- cgit v1.1 From ea9c167d93e6e2f4697f5061756f26356793cd95 Mon Sep 17 00:00:00 2001 From: Lan Tianyu Date: Thu, 31 Jul 2014 22:28:50 +0800 Subject: ACPI / processor: Make acpi_cpu_soft_notify() process CPU FROZEN events CPU hotplug happens during S2RAM and CPU notify event will be CPU_XXX_FROZEN. acpi_cpu_soft_notify() ignores to check these events. This also may make acpi_cpu_soft_notify() fall into sleep during CPU_DYING/STARTING_FROZEN events which don't allow callbacks to sleep. This patch is to fix it. Signed-off-by: Lan Tianyu Signed-off-by: Rafael J. Wysocki --- drivers/acpi/processor_driver.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/acpi/processor_driver.c b/drivers/acpi/processor_driver.c index 4fcbd67..d9f7158 100644 --- a/drivers/acpi/processor_driver.c +++ b/drivers/acpi/processor_driver.c @@ -120,6 +120,7 @@ static int acpi_cpu_soft_notify(struct notifier_block *nfb, unsigned int cpu = (unsigned long)hcpu; struct acpi_processor *pr = per_cpu(processors, cpu); struct acpi_device *device; + action &= ~CPU_TASKS_FROZEN; /* * CPU_STARTING and CPU_DYING must not sleep. Return here since -- cgit v1.1 From 84c91b7ae07c62cf6dee7fde3277f4be21331f85 Mon Sep 17 00:00:00 2001 From: "Lee, Chun-Yi" Date: Mon, 4 Aug 2014 23:23:21 +0800 Subject: PM / hibernate: avoid unsafe pages in e820 reserved regions When the machine doesn't well handle the e820 persistent when hibernate resuming, then it may cause page fault when writing image to snapshot buffer: [ 17.929495] BUG: unable to handle kernel paging request at ffff880069d4f000 [ 17.933469] IP: [] load_image_lzo+0x810/0xe40 [ 17.933469] PGD 2194067 PUD 77ffff067 PMD 2197067 PTE 0 [ 17.933469] Oops: 0002 [#1] SMP ... The ffff880069d4f000 page is in e820 reserved region of resume boot kernel: [ 0.000000] BIOS-e820: [mem 0x0000000069d4f000-0x0000000069e12fff] reserved ... [ 0.000000] PM: Registered nosave memory: [mem 0x69d4f000-0x69e12fff] So snapshot.c mark the pfn to forbidden pages map. But, this page is also in the memory bitmap in snapshot image because it's an original page used by image kernel, so it will also mark as an unsafe(free) page in prepare_image(). That means the page in e820 when resuming mark as "forbidden" and "free", it causes get_buffer() treat it as an allocated unsafe page. Then snapshot_write_next() return this page to load_image, load_image writing content to this address, but this page didn't really allocated . So, we got page fault. Although the root cause is from BIOS, I think aggressive check and significant message in kernel will better then a page fault for issue tracking, especially when serial console unavailable. This patch adds code in mark_unsafe_pages() for check does free pages in nosave region. If so, then it print message and return fault to stop whole S4 resume process: [ 8.166004] PM: Image loading progress: 0% [ 8.658717] PM: 0x6796c000 in e820 nosave region: [mem 0x6796c000-0x6796cfff] [ 8.918737] PM: Read 2511940 kbytes in 1.04 seconds (2415.32 MB/s) [ 8.926633] PM: Error -14 resuming [ 8.933534] PM: Failed to load hibernation image, recovering. Reviewed-by: Takashi Iwai Acked-by: Pavel Machek Signed-off-by: Lee, Chun-Yi [rjw: Subject] Signed-off-by: Rafael J. Wysocki --- kernel/power/snapshot.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/kernel/power/snapshot.c b/kernel/power/snapshot.c index 4fc5c32..c4b8093 100644 --- a/kernel/power/snapshot.c +++ b/kernel/power/snapshot.c @@ -954,6 +954,25 @@ static void mark_nosave_pages(struct memory_bitmap *bm) } } +static bool is_nosave_page(unsigned long pfn) +{ + struct nosave_region *region; + + list_for_each_entry(region, &nosave_regions, list) { + if (pfn >= region->start_pfn && pfn < region->end_pfn) { + pr_err("PM: %#010llx in e820 nosave region: " + "[mem %#010llx-%#010llx]\n", + (unsigned long long) pfn << PAGE_SHIFT, + (unsigned long long) region->start_pfn << PAGE_SHIFT, + ((unsigned long long) region->end_pfn << PAGE_SHIFT) + - 1); + return true; + } + } + + return false; +} + /** * create_basic_memory_bitmaps - create bitmaps needed for marking page * frames that should not be saved and free page frames. The pointers @@ -2015,7 +2034,7 @@ static int mark_unsafe_pages(struct memory_bitmap *bm) do { pfn = memory_bm_next_pfn(bm); if (likely(pfn != BM_END_OF_MAP)) { - if (likely(pfn_valid(pfn))) + if (likely(pfn_valid(pfn)) && !is_nosave_page(pfn)) swsusp_set_page_free(pfn_to_page(pfn)); else return -EFAULT; -- cgit v1.1 From 18360d6ed9877d81aa7b3ab811e7e14f3fb0ac61 Mon Sep 17 00:00:00 2001 From: Himangi Saraogi Date: Sun, 3 Aug 2014 15:45:40 +0530 Subject: cpufreq: integrator: Use set_cpus_allowed_ptr Several years ago there was an effort to convert all uses of set_cpus_allowed to use set_cpus_allowed_ptr with the goal of eventually removing the current definition of set_cpus_allowed and renaming set_cpus_allowed_ptr as set_cpus_allowed (https://lkml.org/lkml/2010/3/26/59). This is another step in this direction. The Coccinelle semantic patch that makes this change is as follows: // @@ expression E1,E2; @@ - set_cpus_allowed(E1, cpumask_of_cpu(E2)) + set_cpus_allowed_ptr(E1, cpumask_of(E2)) @@ expression E; identifier I; @@ - set_cpus_allowed(E, I) + set_cpus_allowed_ptr(E, &I) // Signed-off-by: Himangi Saraogi Acked-by: Julia Lawall Acked-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/integrator-cpufreq.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/cpufreq/integrator-cpufreq.c b/drivers/cpufreq/integrator-cpufreq.c index e5122f1..c132052 100644 --- a/drivers/cpufreq/integrator-cpufreq.c +++ b/drivers/cpufreq/integrator-cpufreq.c @@ -92,7 +92,7 @@ static int integrator_set_target(struct cpufreq_policy *policy, * Bind to the specified CPU. When this call returns, * we should be running on the right CPU. */ - set_cpus_allowed(current, cpumask_of_cpu(cpu)); + set_cpus_allowed_ptr(current, cpumask_of(cpu)); BUG_ON(cpu != smp_processor_id()); /* get current setting */ @@ -118,7 +118,7 @@ static int integrator_set_target(struct cpufreq_policy *policy, freqs.new = icst_hz(&cclk_params, vco) / 1000; if (freqs.old == freqs.new) { - set_cpus_allowed(current, cpus_allowed); + set_cpus_allowed_ptr(current, &cpus_allowed); return 0; } @@ -141,7 +141,7 @@ static int integrator_set_target(struct cpufreq_policy *policy, /* * Restore the CPUs allowed mask. */ - set_cpus_allowed(current, cpus_allowed); + set_cpus_allowed_ptr(current, &cpus_allowed); cpufreq_freq_transition_end(policy, &freqs, 0); @@ -157,7 +157,7 @@ static unsigned int integrator_get(unsigned int cpu) cpus_allowed = current->cpus_allowed; - set_cpus_allowed(current, cpumask_of_cpu(cpu)); + set_cpus_allowed_ptr(current, cpumask_of(cpu)); BUG_ON(cpu != smp_processor_id()); /* detect memory etc. */ @@ -173,7 +173,7 @@ static unsigned int integrator_get(unsigned int cpu) current_freq = icst_hz(&cclk_params, vco) / 1000; /* current freq */ - set_cpus_allowed(current, cpus_allowed); + set_cpus_allowed_ptr(current, &cpus_allowed); return current_freq; } -- cgit v1.1 From 713a3fa6b71578cae3af820be10dd9e0702f3280 Mon Sep 17 00:00:00 2001 From: Markus Pargmann Date: Mon, 4 Aug 2014 14:48:03 +0200 Subject: cpufreq: cpu0: Do not print error message when deferring -EPROBE_DEFER is no real error. We are just waiting unti the necessary components are ready. The driver core infrastructure will also print an appropriate info message. This patch changes the error message to a debug message. Signed-off-by: Markus Pargmann Acked-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/cpufreq-cpu0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/cpufreq/cpufreq-cpu0.c b/drivers/cpufreq/cpufreq-cpu0.c index 86beda9..0d2172b 100644 --- a/drivers/cpufreq/cpufreq-cpu0.c +++ b/drivers/cpufreq/cpufreq-cpu0.c @@ -137,7 +137,7 @@ static int cpu0_cpufreq_probe(struct platform_device *pdev) * not yet registered, we should try defering probe. */ if (PTR_ERR(cpu_reg) == -EPROBE_DEFER) { - dev_err(cpu_dev, "cpu0 regulator not ready, retry\n"); + dev_dbg(cpu_dev, "cpu0 regulator not ready, retry\n"); ret = -EPROBE_DEFER; goto out_put_node; } -- cgit v1.1 From 3c5445ce3a0c6d6935911212b735772af5115517 Mon Sep 17 00:00:00 2001 From: Stephen Boyd Date: Wed, 6 Aug 2014 22:04:58 +0200 Subject: cpufreq: OPP: Avoid sleeping while atomic We allocate the cpufreq table after calling rcu_read_lock(), which disables preemption. This causes scheduling while atomic warnings. Use GFP_ATOMIC instead of GFP_KERNEL and update for kcalloc while we're here. BUG: sleeping function called from invalid context at mm/slub.c:1246 in_atomic(): 0, irqs_disabled(): 0, pid: 80, name: modprobe 5 locks held by modprobe/80: #0: (&dev->mutex){......}, at: [] __driver_attach+0x48/0x98 #1: (&dev->mutex){......}, at: [] __driver_attach+0x58/0x98 #2: (subsys mutex#5){+.+.+.}, at: [] subsys_interface_register+0x38/0xc8 #3: (cpufreq_rwsem){.+.+.+}, at: [] __cpufreq_add_dev.isra.22+0x84/0x92c #4: (rcu_read_lock){......}, at: [] dev_pm_opp_init_cpufreq_table+0x18/0x10c Preemption disabled at:[< (null)>] (null) CPU: 2 PID: 80 Comm: modprobe Not tainted 3.16.0-rc3-next-20140701-00035-g286857f216aa-dirty #217 [] (unwind_backtrace) from [] (show_stack+0x10/0x14) [] (show_stack) from [] (dump_stack+0x70/0xbc) [] (dump_stack) from [] (__kmalloc+0x124/0x250) [] (__kmalloc) from [] (dev_pm_opp_init_cpufreq_table+0x3c/0x10c) [] (dev_pm_opp_init_cpufreq_table) from [] (cpufreq_init+0x48/0x378 [cpufreq_generic]) [] (cpufreq_init [cpufreq_generic]) from [] (__cpufreq_add_dev.isra.22+0x200/0x92c) [] (__cpufreq_add_dev.isra.22) from [] (subsys_interface_register+0x84/0xc8) [] (subsys_interface_register) from [] (cpufreq_register_driver+0x108/0x2d8) [] (cpufreq_register_driver) from [] (generic_cpufreq_probe+0x50/0x74 [cpufreq_generic]) [] (generic_cpufreq_probe [cpufreq_generic]) from [] (platform_drv_probe+0x18/0x48) [] (platform_drv_probe) from [] (driver_probe_device+0x128/0x370) [] (driver_probe_device) from [] (__driver_attach+0x94/0x98) [] (__driver_attach) from [] (bus_for_each_dev+0x54/0x88) [] (bus_for_each_dev) from [] (bus_add_driver+0xe8/0x204) [] (bus_add_driver) from [] (driver_register+0x78/0xf4) [] (driver_register) from [] (do_one_initcall+0xac/0x1d8) [] (do_one_initcall) from [] (load_module+0x190c/0x21e8) [] (load_module) from [] (SyS_init_module+0xa4/0x110) [] (SyS_init_module) from [] (ret_fast_syscall+0x0/0x48) Fixes: a0dd7b79657b (PM / OPP: Move cpufreq specific OPP functions out of generic OPP library) Signed-off-by: Stephen Boyd Acked-by: Viresh Kumar Cc: 3.16+ # 3.16+ Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/cpufreq_opp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/cpufreq/cpufreq_opp.c b/drivers/cpufreq/cpufreq_opp.c index c0c6f4a..f7a32d2 100644 --- a/drivers/cpufreq/cpufreq_opp.c +++ b/drivers/cpufreq/cpufreq_opp.c @@ -60,7 +60,7 @@ int dev_pm_opp_init_cpufreq_table(struct device *dev, goto out; } - freq_table = kzalloc(sizeof(*freq_table) * (max_opps + 1), GFP_KERNEL); + freq_table = kcalloc(sizeof(*freq_table), (max_opps + 1), GFP_ATOMIC); if (!freq_table) { ret = -ENOMEM; goto out; -- cgit v1.1 From dee1592638ab7ea35a32179b73f9284dead49c03 Mon Sep 17 00:00:00 2001 From: Tang Chen Date: Fri, 8 Aug 2014 10:30:45 +0800 Subject: ACPI / hotplug: Check scan handlers in acpi_scan_hot_remove() When ACPI_HOTPLUG_MEMORY is not configured, memory_device_handler.attach is not set. In acpi_scan_attach_handler(), the acpi_device->handler will not be initialized. In acpi_scan_hot_remove(), it doesn't check if acpi_device->handler is NULL. If we do memory hot-remove without ACPI_HOTPLUG_MEMORY configured, the kernel will panic. BUG: unable to handle kernel NULL pointer dereference at 0000000000000088 IP: [] acpi_device_hotplug+0x1d7/0x4c4 PGD 0 Oops: 0000 [#1] SMP Modules linked in: sd_mod(E) sr_mod(E) cdrom(E) crc_t10dif(E) crct10dif_common(E) ata_piix(E) libata(E) CPU: 0 PID: 41 Comm: kworker/u2:1 Tainted: G E 3.16.0-rc7--3.16-rc7-tangchen+ #20 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.7.5-0-ge51488c-20140602_164612-nilsson.home.kraxel.org 04/01/2014 Workqueue: kacpi_hotplug acpi_hotplug_work_fn task: ffff8800182436c0 ti: ffff880018254000 task.ti: ffff880018254000 RIP: 0010:[] [] acpi_device_hotplug+0x1d7/0x4c4 RSP: 0000:ffff880018257da8 EFLAGS: 00000246 RAX: 0000000000000000 RBX: ffff88001cd8d800 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff88001e40e6f8 RDI: 0000000000000246 RBP: ffff880018257df0 R08: 0000000000000096 R09: 00000000000011a0 R10: 63735f6970636120 R11: 725f746f685f6e61 R12: 0000000000000003 R13: ffff88001cc1c400 R14: ffff88001e062028 R15: 0000000000000040 FS: 0000000000000000(0000) GS:ffff88001e400000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 0000000000000088 CR3: 000000001a9a2000 CR4: 00000000000006f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 0000000000000000 DR7: 0000000000000000 Stack: 00000000523cab58 ffff88001cd8d9f8 ffff88001852d480 00000000523cab58 ffff88001852d480 ffff880018221e40 ffff88001cc1c400 ffff88001cce2d00 0000000000000040 ffff880018257e08 ffffffff813dc31d ffff88001852d480 Call Trace: [] acpi_hotplug_work_fn+0x1e/0x29 [] process_one_work+0x17b/0x460 [] worker_thread+0x11d/0x5b0 [] ? rescuer_thread+0x3a0/0x3a0 [] kthread+0xe1/0x100 [] ? kthread_create_on_node+0x1a0/0x1a0 [] ret_from_fork+0x7c/0xb0 [] ? kthread_create_on_node+0x1a0/0x1a0 This patch fixes this problem by checking if acpi_device->handler is NULL in acpi_scan_hot_remove(). Fixes: d22ddcbc4fb7 (ACPI / hotplug: Add demand_offline hotplug profile flag) Signed-off-by: Tang Chen Cc: 3.14+ # 3.14+ [rjw: Subject] Signed-off-by: Rafael J. Wysocki --- drivers/acpi/scan.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c index 3ffc930..9e4422b 100644 --- a/drivers/acpi/scan.c +++ b/drivers/acpi/scan.c @@ -353,7 +353,8 @@ static int acpi_scan_hot_remove(struct acpi_device *device) unsigned long long sta; acpi_status status; - if (device->handler->hotplug.demand_offline && !acpi_force_hot_remove) { + if (device->handler && device->handler->hotplug.demand_offline + && !acpi_force_hot_remove) { if (!acpi_scan_is_offline(device, true)) return -EBUSY; } else { -- cgit v1.1 From a5f30eba78969624ed8e86db81a8dd91736f79f3 Mon Sep 17 00:00:00 2001 From: Hans Wennborg Date: Wed, 6 Aug 2014 22:52:23 -0700 Subject: cpufreq: speedstep-smi: fix decimal printf specifiers The prefix suggests the number should be printed in hex, so use the %x specifier to do that. Also, these are 32-bit values, so drop the l characters. Signed-off-by: Hans Wennborg Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/speedstep-smi.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/cpufreq/speedstep-smi.c b/drivers/cpufreq/speedstep-smi.c index 8635eec..5fc96d5 100644 --- a/drivers/cpufreq/speedstep-smi.c +++ b/drivers/cpufreq/speedstep-smi.c @@ -324,8 +324,8 @@ static int __init speedstep_init(void) return -ENODEV; } - pr_debug("signature:0x%.8ulx, command:0x%.8ulx, " - "event:0x%.8ulx, perf_level:0x%.8ulx.\n", + pr_debug("signature:0x%.8x, command:0x%.8x, " + "event:0x%.8x, perf_level:0x%.8x.\n", ist_info.signature, ist_info.command, ist_info.event, ist_info.perf_level); -- cgit v1.1 From 39c8bbaf67b157017929703a5eea7e83525b147c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Uwe=20Kleine-K=C3=B6nig?= Date: Fri, 8 Aug 2014 08:56:30 +0200 Subject: cpufreq: arm_big_little: fix module license spec MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Having no license specification in a module taints the kernel during load with: arm_big_little: module license 'unspecified' taints kernel. and also the linker doesn't allow it to make use of GPL-exported symbols which in this case also results in errors like: arm_big_little: Unknown symbol cpufreq_register_driver (err 0) . The header of the driver specifies a GPL v2 license, so note that accordingly. While at it also add a description and an author and fix the license in a companion file to explicit v2. Reported-by: Andreas Schwab Acked-by: Viresh Kumar Signed-off-by: Uwe Kleine-König Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/arm_big_little.c | 5 +++++ drivers/cpufreq/arm_big_little_dt.c | 2 +- 2 files changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/cpufreq/arm_big_little.c b/drivers/cpufreq/arm_big_little.c index 1f4d4e3..a46c223 100644 --- a/drivers/cpufreq/arm_big_little.c +++ b/drivers/cpufreq/arm_big_little.c @@ -24,6 +24,7 @@ #include #include #include +#include #include #include #include @@ -593,3 +594,7 @@ void bL_cpufreq_unregister(struct cpufreq_arm_bL_ops *ops) arm_bL_ops = NULL; } EXPORT_SYMBOL_GPL(bL_cpufreq_unregister); + +MODULE_AUTHOR("Viresh Kumar "); +MODULE_DESCRIPTION("Generic ARM big LITTLE cpufreq driver"); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/cpufreq/arm_big_little_dt.c b/drivers/cpufreq/arm_big_little_dt.c index 8d9d591..4550f69 100644 --- a/drivers/cpufreq/arm_big_little_dt.c +++ b/drivers/cpufreq/arm_big_little_dt.c @@ -114,4 +114,4 @@ module_platform_driver(generic_bL_platdrv); MODULE_AUTHOR("Viresh Kumar "); MODULE_DESCRIPTION("Generic ARM big LITTLE cpufreq driver via DT"); -MODULE_LICENSE("GPL"); +MODULE_LICENSE("GPL v2"); -- cgit v1.1 From b8432c6fc12bdf4a6921e1640d3ead23fcd04c10 Mon Sep 17 00:00:00 2001 From: Todd E Brandt Date: Fri, 8 Aug 2014 09:11:59 -0700 Subject: PM / tools: analyze_suspend.py: update to v3.0 Update of analyze_suspend.py to v3.0 New features include back-2-back suspend testing, device filters to reduce the html size, the inclusion of device_prepare and device_complete callbacks, a usb topography list, and the ability to control USB device autosuspend. UI upgrades include a device detail window and mini-timeline, the addition of a suspend_prepare and resume_complete phase to the timeline which includes the associated device callbacks, automatic highlight of related callbacks, and general color and name changes for better reability. The new version relies on two trace point patches that are already in the kernel: enable_trace_events_suspend_resume.patch enable_trace_events_device_pm_callback.patch It has legacy support for older kernels without these trace events, but when available the tool processes the ftrace output alone (dmesg has been deprecated as a tool input, and is only gathered for convenience). Link: https://01.org/suspendresume/downloads/analyzesuspend-v3.0 Signed-off-by: Todd Brandt Signed-off-by: Rafael J. Wysocki --- scripts/analyze_suspend.py | 3817 ++++++++++++++++++++++++++++++++++---------- 1 file changed, 2981 insertions(+), 836 deletions(-) diff --git a/scripts/analyze_suspend.py b/scripts/analyze_suspend.py index 4f2cc12..93e1fd4 100755 --- a/scripts/analyze_suspend.py +++ b/scripts/analyze_suspend.py @@ -36,146 +36,392 @@ # CONFIG_FUNCTION_TRACER=y # CONFIG_FUNCTION_GRAPH_TRACER=y # +# For kernel versions older than 3.15: # The following additional kernel parameters are required: # (e.g. in file /etc/default/grub) # GRUB_CMDLINE_LINUX_DEFAULT="... initcall_debug log_buf_len=16M ..." # +# ----------------- LIBRARIES -------------------- + import sys import time import os import string import re -import array import platform -import datetime +from datetime import datetime import struct -# -- classes -- +# ----------------- CLASSES -------------------- +# Class: SystemValues +# Description: +# A global, single-instance container used to +# store system values and test parameters class SystemValues: - testdir = "." - tpath = "/sys/kernel/debug/tracing/" - mempath = "/dev/mem" - powerfile = "/sys/power/state" - suspendmode = "mem" - prefix = "test" - teststamp = "" - dmesgfile = "" - ftracefile = "" - htmlfile = "" + version = 3.0 + verbose = False + testdir = '.' + tpath = '/sys/kernel/debug/tracing/' + fpdtpath = '/sys/firmware/acpi/tables/FPDT' + epath = '/sys/kernel/debug/tracing/events/power/' + traceevents = [ + 'suspend_resume', + 'device_pm_callback_end', + 'device_pm_callback_start' + ] + modename = { + 'freeze': 'Suspend-To-Idle (S0)', + 'standby': 'Power-On Suspend (S1)', + 'mem': 'Suspend-to-RAM (S3)', + 'disk': 'Suspend-to-disk (S4)' + } + mempath = '/dev/mem' + powerfile = '/sys/power/state' + suspendmode = 'mem' + hostname = 'localhost' + prefix = 'test' + teststamp = '' + dmesgfile = '' + ftracefile = '' + htmlfile = '' rtcwake = False + rtcwaketime = 10 + rtcpath = '' + android = False + adb = 'adb' + devicefilter = [] + stamp = 0 + execcount = 1 + x2delay = 0 + usecallgraph = False + usetraceevents = False + usetraceeventsonly = False + notestrun = False + altdevname = dict() + postresumetime = 0 + tracertypefmt = '# tracer: (?P.*)' + firmwarefmt = '# fwsuspend (?P[0-9]*) fwresume (?P[0-9]*)$' + postresumefmt = '# post resume time (?P[0-9]*)$' + stampfmt = '# suspend-(?P[0-9]{2})(?P[0-9]{2})(?P[0-9]{2})-'+\ + '(?P[0-9]{2})(?P[0-9]{2})(?P[0-9]{2})'+\ + ' (?P.*) (?P.*) (?P.*)$' + def __init__(self): + self.hostname = platform.node() + if(self.hostname == ''): + self.hostname = 'localhost' + rtc = "rtc0" + if os.path.exists('/dev/rtc'): + rtc = os.readlink('/dev/rtc') + rtc = '/sys/class/rtc/'+rtc + if os.path.exists(rtc) and os.path.exists(rtc+'/date') and \ + os.path.exists(rtc+'/time') and os.path.exists(rtc+'/wakealarm'): + self.rtcpath = rtc def setOutputFile(self): - if((self.htmlfile == "") and (self.dmesgfile != "")): - m = re.match(r"(?P.*)_dmesg\.txt$", self.dmesgfile) + if((self.htmlfile == '') and (self.dmesgfile != '')): + m = re.match('(?P.*)_dmesg\.txt$', self.dmesgfile) if(m): - self.htmlfile = m.group("name")+".html" - if((self.htmlfile == "") and (self.ftracefile != "")): - m = re.match(r"(?P.*)_ftrace\.txt$", self.ftracefile) + self.htmlfile = m.group('name')+'.html' + if((self.htmlfile == '') and (self.ftracefile != '')): + m = re.match('(?P.*)_ftrace\.txt$', self.ftracefile) if(m): - self.htmlfile = m.group("name")+".html" - if(self.htmlfile == ""): - self.htmlfile = "output.html" - def initTestOutput(self): - hostname = platform.node() - if(hostname != ""): - self.prefix = hostname - v = os.popen("cat /proc/version").read().strip() - kver = string.split(v)[2] - self.testdir = os.popen("date \"+suspend-%m%d%y-%H%M%S\"").read().strip() - self.teststamp = "# "+self.testdir+" "+self.prefix+" "+self.suspendmode+" "+kver - self.dmesgfile = self.testdir+"/"+self.prefix+"_"+self.suspendmode+"_dmesg.txt" - self.ftracefile = self.testdir+"/"+self.prefix+"_"+self.suspendmode+"_ftrace.txt" - self.htmlfile = self.testdir+"/"+self.prefix+"_"+self.suspendmode+".html" + self.htmlfile = m.group('name')+'.html' + if(self.htmlfile == ''): + self.htmlfile = 'output.html' + def initTestOutput(self, subdir): + if(not self.android): + self.prefix = self.hostname + v = open('/proc/version', 'r').read().strip() + kver = string.split(v)[2] + else: + self.prefix = 'android' + v = os.popen(self.adb+' shell cat /proc/version').read().strip() + kver = string.split(v)[2] + testtime = datetime.now().strftime('suspend-%m%d%y-%H%M%S') + if(subdir != "."): + self.testdir = subdir+"/"+testtime + else: + self.testdir = testtime + self.teststamp = \ + '# '+testtime+' '+self.prefix+' '+self.suspendmode+' '+kver + self.dmesgfile = \ + self.testdir+'/'+self.prefix+'_'+self.suspendmode+'_dmesg.txt' + self.ftracefile = \ + self.testdir+'/'+self.prefix+'_'+self.suspendmode+'_ftrace.txt' + self.htmlfile = \ + self.testdir+'/'+self.prefix+'_'+self.suspendmode+'.html' os.mkdir(self.testdir) + def setDeviceFilter(self, devnames): + self.devicefilter = string.split(devnames) + def rtcWakeAlarm(self): + os.system('echo 0 > '+self.rtcpath+'/wakealarm') + outD = open(self.rtcpath+'/date', 'r').read().strip() + outT = open(self.rtcpath+'/time', 'r').read().strip() + mD = re.match('^(?P[0-9]*)-(?P[0-9]*)-(?P[0-9]*)', outD) + mT = re.match('^(?P[0-9]*):(?P[0-9]*):(?P[0-9]*)', outT) + if(mD and mT): + # get the current time from hardware + utcoffset = int((datetime.now() - datetime.utcnow()).total_seconds()) + dt = datetime(\ + int(mD.group('y')), int(mD.group('m')), int(mD.group('d')), + int(mT.group('h')), int(mT.group('m')), int(mT.group('s'))) + nowtime = int(dt.strftime('%s')) + utcoffset + else: + # if hardware time fails, use the software time + nowtime = int(datetime.now().strftime('%s')) + alarm = nowtime + self.rtcwaketime + os.system('echo %d > %s/wakealarm' % (alarm, self.rtcpath)) +sysvals = SystemValues() + +# Class: DeviceNode +# Description: +# A container used to create a device hierachy, with a single root node +# and a tree of child nodes. Used by Data.deviceTopology() +class DeviceNode: + name = '' + children = 0 + depth = 0 + def __init__(self, nodename, nodedepth): + self.name = nodename + self.children = [] + self.depth = nodedepth + +# Class: Data +# Description: +# The primary container for suspend/resume test data. There is one for +# each test run. The data is organized into a cronological hierarchy: +# Data.dmesg { +# root structure, started as dmesg & ftrace, but now only ftrace +# contents: times for suspend start/end, resume start/end, fwdata +# phases { +# 10 sequential, non-overlapping phases of S/R +# contents: times for phase start/end, order/color data for html +# devlist { +# device callback or action list for this phase +# device { +# a single device callback or generic action +# contents: start/stop times, pid/cpu/driver info +# parents/children, html id for timeline/callgraph +# optionally includes an ftrace callgraph +# optionally includes intradev trace events +# } +# } +# } +# } +# class Data: - altdevname = dict() - usedmesg = False - useftrace = False - notestrun = False - verbose = False - phases = [] - dmesg = {} # root data structure - start = 0.0 - end = 0.0 - stamp = {'time': "", 'host': "", 'mode': ""} - id = 0 - tSuspended = 0.0 - fwValid = False - fwSuspend = 0 - fwResume = 0 - def initialize(self): - self.dmesg = { # dmesg log data - 'suspend_general': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "#CCFFCC", 'order': 0}, - 'suspend_early': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "green", 'order': 1}, + dmesg = {} # root data structure + phases = [] # ordered list of phases + start = 0.0 # test start + end = 0.0 # test end + tSuspended = 0.0 # low-level suspend start + tResumed = 0.0 # low-level resume start + tLow = 0.0 # time spent in low-level suspend (standby/freeze) + fwValid = False # is firmware data available + fwSuspend = 0 # time spent in firmware suspend + fwResume = 0 # time spent in firmware resume + dmesgtext = [] # dmesg text file in memory + testnumber = 0 + idstr = '' + html_device_id = 0 + stamp = 0 + outfile = '' + def __init__(self, num): + idchar = 'abcdefghijklmnopqrstuvwxyz' + self.testnumber = num + self.idstr = idchar[num] + self.dmesgtext = [] + self.phases = [] + self.dmesg = { # fixed list of 10 phases + 'suspend_prepare': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#CCFFCC', 'order': 0}, + 'suspend': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#88FF88', 'order': 1}, + 'suspend_late': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#00AA00', 'order': 2}, 'suspend_noirq': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "#00FFFF", 'order': 2}, - 'suspend_cpu': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "blue", 'order': 3}, - 'resume_cpu': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "red", 'order': 4}, + 'row': 0, 'color': '#008888', 'order': 3}, + 'suspend_machine': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#0000FF', 'order': 4}, + 'resume_machine': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#FF0000', 'order': 5}, 'resume_noirq': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "orange", 'order': 5}, + 'row': 0, 'color': '#FF9900', 'order': 6}, 'resume_early': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "yellow", 'order': 6}, - 'resume_general': {'list': dict(), 'start': -1.0, 'end': -1.0, - 'row': 0, 'color': "#FFFFCC", 'order': 7} + 'row': 0, 'color': '#FFCC00', 'order': 7}, + 'resume': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#FFFF88', 'order': 8}, + 'resume_complete': {'list': dict(), 'start': -1.0, 'end': -1.0, + 'row': 0, 'color': '#FFFFCC', 'order': 9} } self.phases = self.sortedPhases() - def normalizeTime(self): - tSus = tRes = self.tSuspended - if self.fwValid: - tSus -= -self.fwSuspend / 1000000000.0 - tRes -= self.fwResume / 1000000000.0 - self.tSuspended = 0.0 - self.start -= tSus - self.end -= tRes + def getStart(self): + return self.dmesg[self.phases[0]]['start'] + def setStart(self, time): + self.start = time + self.dmesg[self.phases[0]]['start'] = time + def getEnd(self): + return self.dmesg[self.phases[-1]]['end'] + def setEnd(self, time): + self.end = time + self.dmesg[self.phases[-1]]['end'] = time + def isTraceEventOutsideDeviceCalls(self, pid, time): + for phase in self.phases: + list = self.dmesg[phase]['list'] + for dev in list: + d = list[dev] + if(d['pid'] == pid and time >= d['start'] and + time <= d['end']): + return False + return True + def addIntraDevTraceEvent(self, action, name, pid, time): + if(action == 'mutex_lock_try'): + color = 'red' + elif(action == 'mutex_lock_pass'): + color = 'green' + elif(action == 'mutex_unlock'): + color = 'blue' + else: + # create separate colors based on the name + v1 = len(name)*10 % 256 + v2 = string.count(name, 'e')*100 % 256 + v3 = ord(name[0])*20 % 256 + color = '#%06X' % ((v1*0x10000) + (v2*0x100) + v3) + for phase in self.phases: + list = self.dmesg[phase]['list'] + for dev in list: + d = list[dev] + if(d['pid'] == pid and time >= d['start'] and + time <= d['end']): + e = TraceEvent(action, name, color, time) + if('traceevents' not in d): + d['traceevents'] = [] + d['traceevents'].append(e) + return d + break + return 0 + def capIntraDevTraceEvent(self, action, name, pid, time): + for phase in self.phases: + list = self.dmesg[phase]['list'] + for dev in list: + d = list[dev] + if(d['pid'] == pid and time >= d['start'] and + time <= d['end']): + if('traceevents' not in d): + return + for e in d['traceevents']: + if(e.action == action and + e.name == name and not e.ready): + e.length = time - e.time + e.ready = True + break + return + def trimTimeVal(self, t, t0, dT, left): + if left: + if(t > t0): + if(t - dT < t0): + return t0 + return t - dT + else: + return t + else: + if(t < t0 + dT): + if(t > t0): + return t0 + dT + return t + dT + else: + return t + def trimTime(self, t0, dT, left): + self.tSuspended = self.trimTimeVal(self.tSuspended, t0, dT, left) + self.tResumed = self.trimTimeVal(self.tResumed, t0, dT, left) + self.start = self.trimTimeVal(self.start, t0, dT, left) + self.end = self.trimTimeVal(self.end, t0, dT, left) for phase in self.phases: - zero = tRes - if "suspend" in phase: - zero = tSus p = self.dmesg[phase] - p['start'] -= zero - p['end'] -= zero + p['start'] = self.trimTimeVal(p['start'], t0, dT, left) + p['end'] = self.trimTimeVal(p['end'], t0, dT, left) list = p['list'] for name in list: d = list[name] - d['start'] -= zero - d['end'] -= zero + d['start'] = self.trimTimeVal(d['start'], t0, dT, left) + d['end'] = self.trimTimeVal(d['end'], t0, dT, left) if('ftrace' in d): cg = d['ftrace'] - cg.start -= zero - cg.end -= zero + cg.start = self.trimTimeVal(cg.start, t0, dT, left) + cg.end = self.trimTimeVal(cg.end, t0, dT, left) for line in cg.list: - line.time -= zero - if self.fwValid: - fws = -self.fwSuspend / 1000000000.0 - fwr = self.fwResume / 1000000000.0 - list = dict() - self.id += 1 - devid = "dc%d" % self.id - list["firmware-suspend"] = \ - {'start': fws, 'end': 0, 'pid': 0, 'par': "", - 'length': -fws, 'row': 0, 'id': devid }; - self.id += 1 - devid = "dc%d" % self.id - list["firmware-resume"] = \ - {'start': 0, 'end': fwr, 'pid': 0, 'par': "", - 'length': fwr, 'row': 0, 'id': devid }; - self.dmesg['BIOS'] = \ - {'list': list, 'start': fws, 'end': fwr, - 'row': 0, 'color': "purple", 'order': 4} - self.dmesg['resume_cpu']['order'] += 1 - self.dmesg['resume_noirq']['order'] += 1 - self.dmesg['resume_early']['order'] += 1 - self.dmesg['resume_general']['order'] += 1 - self.phases = self.sortedPhases() - def vprint(self, msg): - if(self.verbose): - print(msg) + line.time = self.trimTimeVal(line.time, t0, dT, left) + if('traceevents' in d): + for e in d['traceevents']: + e.time = self.trimTimeVal(e.time, t0, dT, left) + def normalizeTime(self, tZero): + # first trim out any standby or freeze clock time + if(self.tSuspended != self.tResumed): + if(self.tResumed > tZero): + self.trimTime(self.tSuspended, \ + self.tResumed-self.tSuspended, True) + else: + self.trimTime(self.tSuspended, \ + self.tResumed-self.tSuspended, False) + # shift the timeline so that tZero is the new 0 + self.tSuspended -= tZero + self.tResumed -= tZero + self.start -= tZero + self.end -= tZero + for phase in self.phases: + p = self.dmesg[phase] + p['start'] -= tZero + p['end'] -= tZero + list = p['list'] + for name in list: + d = list[name] + d['start'] -= tZero + d['end'] -= tZero + if('ftrace' in d): + cg = d['ftrace'] + cg.start -= tZero + cg.end -= tZero + for line in cg.list: + line.time -= tZero + if('traceevents' in d): + for e in d['traceevents']: + e.time -= tZero + def newPhaseWithSingleAction(self, phasename, devname, start, end, color): + for phase in self.phases: + self.dmesg[phase]['order'] += 1 + self.html_device_id += 1 + devid = '%s%d' % (self.idstr, self.html_device_id) + list = dict() + list[devname] = \ + {'start': start, 'end': end, 'pid': 0, 'par': '', + 'length': (end-start), 'row': 0, 'id': devid, 'drv': '' }; + self.dmesg[phasename] = \ + {'list': list, 'start': start, 'end': end, + 'row': 0, 'color': color, 'order': 0} + self.phases = self.sortedPhases() + def newPhase(self, phasename, start, end, color, order): + if(order < 0): + order = len(self.phases) + for phase in self.phases[order:]: + self.dmesg[phase]['order'] += 1 + if(order > 0): + p = self.phases[order-1] + self.dmesg[p]['end'] = start + if(order < len(self.phases)): + p = self.phases[order] + self.dmesg[p]['start'] = end + list = dict() + self.dmesg[phasename] = \ + {'list': list, 'start': start, 'end': end, + 'row': 0, 'color': color, 'order': order} + self.phases = self.sortedPhases() + def setPhase(self, phase, ktime, isbegin): + if(isbegin): + self.dmesg[phase]['start'] = ktime + else: + self.dmesg[phase]['end'] = ktime def dmesgSortVal(self, phase): return self.dmesg[phase]['order'] def sortedPhases(self): @@ -197,59 +443,180 @@ class Data: dev = phaselist[devname] if(dev['end'] < 0): dev['end'] = end - self.vprint("%s (%s): callback didn't return" % (devname, phase)) + vprint('%s (%s): callback didnt return' % (devname, phase)) + def deviceFilter(self, devicefilter): + # remove all by the relatives of the filter devnames + filter = [] + for phase in self.phases: + list = self.dmesg[phase]['list'] + for name in devicefilter: + dev = name + while(dev in list): + if(dev not in filter): + filter.append(dev) + dev = list[dev]['par'] + children = self.deviceDescendants(name, phase) + for dev in children: + if(dev not in filter): + filter.append(dev) + for phase in self.phases: + list = self.dmesg[phase]['list'] + rmlist = [] + for name in list: + pid = list[name]['pid'] + if(name not in filter and pid >= 0): + rmlist.append(name) + for name in rmlist: + del list[name] def fixupInitcallsThatDidntReturn(self): # if any calls never returned, clip them at system resume end for phase in self.phases: - self.fixupInitcalls(phase, self.dmesg['resume_general']['end']) - if(phase == "resume_general"): - break - def newAction(self, phase, name, pid, parent, start, end): - self.id += 1 - devid = "dc%d" % self.id + self.fixupInitcalls(phase, self.getEnd()) + def newActionGlobal(self, name, start, end): + # which phase is this device callback or action "in" + targetphase = "none" + overlap = 0.0 + for phase in self.phases: + pstart = self.dmesg[phase]['start'] + pend = self.dmesg[phase]['end'] + o = max(0, min(end, pend) - max(start, pstart)) + if(o > overlap): + targetphase = phase + overlap = o + if targetphase in self.phases: + self.newAction(targetphase, name, -1, '', start, end, '') + return True + return False + def newAction(self, phase, name, pid, parent, start, end, drv): + # new device callback for a specific phase + self.html_device_id += 1 + devid = '%s%d' % (self.idstr, self.html_device_id) list = self.dmesg[phase]['list'] length = -1.0 if(start >= 0 and end >= 0): length = end - start list[name] = {'start': start, 'end': end, 'pid': pid, 'par': parent, - 'length': length, 'row': 0, 'id': devid } + 'length': length, 'row': 0, 'id': devid, 'drv': drv } def deviceIDs(self, devlist, phase): idlist = [] - for p in self.phases: - if(p[0] != phase[0]): - continue - list = data.dmesg[p]['list'] - for devname in list: - if devname in devlist: - idlist.append(list[devname]['id']) + list = self.dmesg[phase]['list'] + for devname in list: + if devname in devlist: + idlist.append(list[devname]['id']) return idlist def deviceParentID(self, devname, phase): - pdev = "" - pdevid = "" - for p in self.phases: - if(p[0] != phase[0]): - continue - list = data.dmesg[p]['list'] - if devname in list: - pdev = list[devname]['par'] - for p in self.phases: - if(p[0] != phase[0]): - continue - list = data.dmesg[p]['list'] - if pdev in list: - return list[pdev]['id'] + pdev = '' + pdevid = '' + list = self.dmesg[phase]['list'] + if devname in list: + pdev = list[devname]['par'] + if pdev in list: + return list[pdev]['id'] return pdev - def deviceChildrenIDs(self, devname, phase): + def deviceChildren(self, devname, phase): devlist = [] - for p in self.phases: - if(p[0] != phase[0]): - continue - list = data.dmesg[p]['list'] - for child in list: - if(list[child]['par'] == devname): - devlist.append(child) + list = self.dmesg[phase]['list'] + for child in list: + if(list[child]['par'] == devname): + devlist.append(child) + return devlist + def deviceDescendants(self, devname, phase): + children = self.deviceChildren(devname, phase) + family = children + for child in children: + family += self.deviceDescendants(child, phase) + return family + def deviceChildrenIDs(self, devname, phase): + devlist = self.deviceChildren(devname, phase) return self.deviceIDs(devlist, phase) - + def printDetails(self): + vprint(' test start: %f' % self.start) + for phase in self.phases: + dc = len(self.dmesg[phase]['list']) + vprint(' %16s: %f - %f (%d devices)' % (phase, \ + self.dmesg[phase]['start'], self.dmesg[phase]['end'], dc)) + vprint(' test end: %f' % self.end) + def masterTopology(self, name, list, depth): + node = DeviceNode(name, depth) + for cname in list: + clist = self.deviceChildren(cname, 'resume') + cnode = self.masterTopology(cname, clist, depth+1) + node.children.append(cnode) + return node + def printTopology(self, node): + html = '' + if node.name: + info = '' + drv = '' + for phase in self.phases: + list = self.dmesg[phase]['list'] + if node.name in list: + s = list[node.name]['start'] + e = list[node.name]['end'] + if list[node.name]['drv']: + drv = ' {'+list[node.name]['drv']+'}' + info += ('
  • %s: %.3fms
    • ' % (phase, (e-s)*1000)) + html += '
  • '+node.name+drv+'' + if info: + html += '
      '+info+'
    ' + html += '
    • ' + if len(node.children) > 0: + html += '
      ' + for cnode in node.children: + html += self.printTopology(cnode) + html += '
    ' + return html + def rootDeviceList(self): + # list of devices graphed + real = [] + for phase in self.dmesg: + list = self.dmesg[phase]['list'] + for dev in list: + if list[dev]['pid'] >= 0 and dev not in real: + real.append(dev) + # list of top-most root devices + rootlist = [] + for phase in self.dmesg: + list = self.dmesg[phase]['list'] + for dev in list: + pdev = list[dev]['par'] + if(re.match('[0-9]*-[0-9]*\.[0-9]*[\.0-9]*\:[\.0-9]*$', pdev)): + continue + if pdev and pdev not in real and pdev not in rootlist: + rootlist.append(pdev) + return rootlist + def deviceTopology(self): + rootlist = self.rootDeviceList() + master = self.masterTopology('', rootlist, 0) + return self.printTopology(master) + +# Class: TraceEvent +# Description: +# A container for trace event data found in the ftrace file +class TraceEvent: + ready = False + name = '' + time = 0.0 + color = '#FFFFFF' + length = 0.0 + action = '' + def __init__(self, a, n, c, t): + self.action = a + self.name = n + self.color = c + self.time = t + +# Class: FTraceLine +# Description: +# A container for a single line of ftrace data. There are six basic types: +# callgraph line: +# call: " dpm_run_callback() {" +# return: " }" +# leaf: " dpm_run_callback();" +# trace event: +# tracing_mark_write: SUSPEND START or RESUME COMPLETE +# suspend_resume: phase or custom exec block data +# device_pm_callback: device callback info class FTraceLine: time = 0.0 length = 0.0 @@ -257,20 +624,33 @@ class FTraceLine: freturn = False fevent = False depth = 0 - name = "" + name = '' + type = '' def __init__(self, t, m, d): self.time = float(t) - # check to see if this is a trace event - em = re.match(r"^ *\/\* *(?P.*) \*\/ *$", m) - if(em): - self.name = em.group("msg") + # is this a trace event + if(d == 'traceevent' or re.match('^ *\/\* *(?P.*) \*\/ *$', m)): + if(d == 'traceevent'): + # nop format trace event + msg = m + else: + # function_graph format trace event + em = re.match('^ *\/\* *(?P.*) \*\/ *$', m) + msg = em.group('msg') + + emm = re.match('^(?P.*?): (?P.*)', msg) + if(emm): + self.name = emm.group('msg') + self.type = emm.group('call') + else: + self.name = msg self.fevent = True return # convert the duration to seconds if(d): self.length = float(d)/1000000 # the indentation determines the depth - match = re.match(r"^(?P *)(?P.*)$", m) + match = re.match('^(?P *)(?P.*)$', m) if(not match): return self.depth = self.getDepth(match.group('d')) @@ -280,7 +660,7 @@ class FTraceLine: self.freturn = True if(len(m) > 1): # includes comment with function name - match = re.match(r"^} *\/\* *(?P.*) *\*\/$", m) + match = re.match('^} *\/\* *(?P.*) *\*\/$', m) if(match): self.name = match.group('n') # function call @@ -288,13 +668,13 @@ class FTraceLine: self.fcall = True # function call with children if(m[-1] == '{'): - match = re.match(r"^(?P.*) *\(.*", m) + match = re.match('^(?P.*) *\(.*', m) if(match): self.name = match.group('n') # function call with no children (leaf) elif(m[-1] == ';'): self.freturn = True - match = re.match(r"^(?P.*) *\(.*", m) + match = re.match('^(?P.*) *\(.*', m) if(match): self.name = match.group('n') # something else (possibly a trace marker) @@ -302,7 +682,23 @@ class FTraceLine: self.name = m def getDepth(self, str): return len(str)/2 + def debugPrint(self, dev): + if(self.freturn and self.fcall): + print('%s -- %f (%02d): %s(); (%.3f us)' % (dev, self.time, \ + self.depth, self.name, self.length*1000000)) + elif(self.freturn): + print('%s -- %f (%02d): %s} (%.3f us)' % (dev, self.time, \ + self.depth, self.name, self.length*1000000)) + else: + print('%s -- %f (%02d): %s() { (%.3f us)' % (dev, self.time, \ + self.depth, self.name, self.length*1000000)) +# Class: FTraceCallGraph +# Description: +# A container for the ftrace callgraph of a single recursive function. +# This can be a dpm_run_callback, dpm_prepare, or dpm_complete callgraph +# Each instance is tied to a single device in a single phase, and is +# comprised of an ordered list of FTraceLine objects class FTraceCallGraph: start = -1.0 end = -1.0 @@ -327,24 +723,53 @@ class FTraceCallGraph: if(not self.invalid): self.setDepth(line) if(line.depth == 0 and line.freturn): + if(self.start < 0): + self.start = line.time self.end = line.time self.list.append(line) return True if(self.invalid): return False if(len(self.list) >= 1000000 or self.depth < 0): - first = self.list[0] - self.list = [] - self.list.append(first) - self.invalid = True - id = "task %s cpu %s" % (match.group("pid"), match.group("cpu")) - window = "(%f - %f)" % (self.start, line.time) - data.vprint("Too much data for "+id+" "+window+", ignoring this callback") - return False + if(len(self.list) > 0): + first = self.list[0] + self.list = [] + self.list.append(first) + self.invalid = True + if(not match): + return False + id = 'task %s cpu %s' % (match.group('pid'), match.group('cpu')) + window = '(%f - %f)' % (self.start, line.time) + if(self.depth < 0): + print('Too much data for '+id+\ + ' (buffer overflow), ignoring this callback') + else: + print('Too much data for '+id+\ + ' '+window+', ignoring this callback') + return False self.list.append(line) if(self.start < 0): self.start = line.time return False + def slice(self, t0, tN): + minicg = FTraceCallGraph() + count = -1 + firstdepth = 0 + for l in self.list: + if(l.time < t0 or l.time > tN): + continue + if(count < 0): + if(not l.fcall or l.name == 'dev_driver_string'): + continue + firstdepth = l.depth + count = 0 + l.depth -= firstdepth + minicg.addLine(l, 0) + if((count == 0 and l.freturn and l.fcall) or + (count > 0 and l.depth <= 0)): + break + count += 1 + return minicg def sanityCheck(self): stack = dict() cnt = 0 @@ -353,7 +778,7 @@ class FTraceCallGraph: stack[l.depth] = l cnt += 1 elif(l.freturn and not l.fcall): - if(not stack[l.depth]): + if(l.depth not in stack): return False stack[l.depth].length = l.length stack[l.depth] = 0 @@ -363,40 +788,51 @@ class FTraceCallGraph: return True return False def debugPrint(self, filename): - if(filename == "stdout"): - print("[%f - %f]") % (self.start, self.end) + if(filename == 'stdout'): + print('[%f - %f]') % (self.start, self.end) for l in self.list: if(l.freturn and l.fcall): - print("%f (%02d): %s(); (%.3f us)" % (l.time, l.depth, l.name, l.length*1000000)) + print('%f (%02d): %s(); (%.3f us)' % (l.time, \ + l.depth, l.name, l.length*1000000)) elif(l.freturn): - print("%f (%02d): %s} (%.3f us)" % (l.time, l.depth, l.name, l.length*1000000)) + print('%f (%02d): %s} (%.3f us)' % (l.time, \ + l.depth, l.name, l.length*1000000)) else: - print("%f (%02d): %s() { (%.3f us)" % (l.time, l.depth, l.name, l.length*1000000)) - print(" ") + print('%f (%02d): %s() { (%.3f us)' % (l.time, \ + l.depth, l.name, l.length*1000000)) + print(' ') else: fp = open(filename, 'w') print(filename) for l in self.list: if(l.freturn and l.fcall): - fp.write("%f (%02d): %s(); (%.3f us)\n" % (l.time, l.depth, l.name, l.length*1000000)) + fp.write('%f (%02d): %s(); (%.3f us)\n' % (l.time, \ + l.depth, l.name, l.length*1000000)) elif(l.freturn): - fp.write("%f (%02d): %s} (%.3f us)\n" % (l.time, l.depth, l.name, l.length*1000000)) + fp.write('%f (%02d): %s} (%.3f us)\n' % (l.time, \ + l.depth, l.name, l.length*1000000)) else: - fp.write("%f (%02d): %s() { (%.3f us)\n" % (l.time, l.depth, l.name, l.length*1000000)) + fp.write('%f (%02d): %s() { (%.3f us)\n' % (l.time, \ + l.depth, l.name, l.length*1000000)) fp.close() +# Class: Timeline +# Description: +# A container for a suspend/resume html timeline. In older versions +# of the script there were multiple timelines, but in the latest +# there is only one. class Timeline: html = {} - scaleH = 0.0 # height of the timescale row as a percent of the timeline height + scaleH = 0.0 # height of the row as a percent of the timeline height rowH = 0.0 # height of each row in percent of the timeline height row_height_pixels = 30 maxrows = 0 height = 0 def __init__(self): self.html = { - 'timeline': "", - 'legend': "", - 'scale': "" + 'timeline': '', + 'legend': '', + 'scale': '' } def setRows(self, rows): self.maxrows = int(rows) @@ -407,104 +843,261 @@ class Timeline: r = 1.0 self.rowH = (100.0 - self.scaleH)/r -# -- global objects -- +# Class: TestRun +# Description: +# A container for a suspend/resume test run. This is necessary as +# there could be more than one, and they need to be separate. +class TestRun: + ftrace_line_fmt_fg = \ + '^ *(?P