From 0793a61d4df8daeac6492dbf8d2f3e5713caae5e Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 4 Dec 2008 20:12:29 +0100 Subject: performance counters: core code Implement the core kernel bits of Performance Counters subsystem. The Linux Performance Counter subsystem provides an abstraction of performance counter hardware capabilities. It provides per task and per CPU counters, and it provides event capabilities on top of those. Performance counters are accessed via special file descriptors. There's one file descriptor per virtual counter used. The special file descriptor is opened via the perf_counter_open() system call: int perf_counter_open(u32 hw_event_type, u32 hw_event_period, u32 record_type, pid_t pid, int cpu); The syscall returns the new fd. The fd can be used via the normal VFS system calls: read() can be used to read the counter, fcntl() can be used to set the blocking mode, etc. Multiple counters can be kept open at a time, and the counters can be poll()ed. See more details in Documentation/perf-counters.txt. Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- include/linux/sched.h | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index 55e30d11447..4c530278391 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -71,6 +71,7 @@ struct sched_param { #include #include #include +#include #include #include #include @@ -1326,6 +1327,7 @@ struct task_struct { struct list_head pi_state_list; struct futex_pi_state *pi_state_cache; #endif + struct perf_counter_context perf_counter_ctx; #ifdef CONFIG_NUMA struct mempolicy *mempolicy; short il_next; @@ -2285,6 +2287,13 @@ static inline void inc_syscw(struct task_struct *tsk) #define TASK_SIZE_OF(tsk) TASK_SIZE #endif +/* + * Call the function if the target task is executing on a CPU right now: + */ +extern void task_oncpu_function_call(struct task_struct *p, + void (*func) (void *info), void *info); + + #ifdef CONFIG_MM_OWNER extern void mm_update_next_owner(struct mm_struct *mm); extern void mm_init_owner(struct mm_struct *mm, struct task_struct *p); -- cgit v1.2.3-70-g09d2 From 6c594c21fcb02c662f11c97be4d7d2b73060a205 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 14 Dec 2008 12:34:15 +0100 Subject: perfcounters: add task migrations counter Impact: add new feature, new sw counter Add a counter that counts the number of cross-CPU migrations a task is suffering. Signed-off-by: Ingo Molnar --- include/linux/perf_counter.h | 8 +++++--- include/linux/sched.h | 3 ++- kernel/perf_counter.c | 49 ++++++++++++++++++++++++++++++++++++++++++++ kernel/sched.c | 7 +++++-- 4 files changed, 61 insertions(+), 6 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index d2a16563415..f30486fc55d 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -42,6 +42,8 @@ enum hw_event_types { PERF_COUNT_BRANCH_INSTRUCTIONS = 4, PERF_COUNT_BRANCH_MISSES = 5, + PERF_HW_EVENTS_MAX = 6, + /* * Special "software" counters provided by the kernel, even if * the hardware does not support performance counters. These @@ -50,11 +52,11 @@ enum hw_event_types { */ PERF_COUNT_CPU_CLOCK = -1, PERF_COUNT_TASK_CLOCK = -2, - /* - * Future software events: - */ PERF_COUNT_PAGE_FAULTS = -3, PERF_COUNT_CONTEXT_SWITCHES = -4, + PERF_COUNT_CPU_MIGRATIONS = -5, + + PERF_SW_EVENTS_MIN = -6, }; /* diff --git a/include/linux/sched.h b/include/linux/sched.h index 4c530278391..2e15be8fc79 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1014,6 +1014,8 @@ struct sched_entity { u64 last_wakeup; u64 avg_overlap; + u64 nr_migrations; + #ifdef CONFIG_SCHEDSTATS u64 wait_start; u64 wait_max; @@ -1029,7 +1031,6 @@ struct sched_entity { u64 exec_max; u64 slice_max; - u64 nr_migrations; u64 nr_migrations_cold; u64 nr_failed_migrations_affine; u64 nr_failed_migrations_running; diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 09287091c52..fb11e351e44 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -936,6 +936,52 @@ static const struct hw_perf_counter_ops perf_ops_context_switches = { .hw_perf_counter_read = context_switches_perf_counter_read, }; +static inline u64 get_cpu_migrations(void) +{ + return current->se.nr_migrations; +} + +static void cpu_migrations_perf_counter_update(struct perf_counter *counter) +{ + u64 prev, now; + s64 delta; + + prev = atomic64_read(&counter->hw.prev_count); + now = get_cpu_migrations(); + + atomic64_set(&counter->hw.prev_count, now); + + delta = now - prev; + if (WARN_ON_ONCE(delta < 0)) + delta = 0; + + atomic64_add(delta, &counter->count); +} + +static void cpu_migrations_perf_counter_read(struct perf_counter *counter) +{ + cpu_migrations_perf_counter_update(counter); +} + +static void cpu_migrations_perf_counter_enable(struct perf_counter *counter) +{ + /* + * se.nr_migrations is a per-task value already, + * so we dont have to clear it on switch-in. + */ +} + +static void cpu_migrations_perf_counter_disable(struct perf_counter *counter) +{ + cpu_migrations_perf_counter_update(counter); +} + +static const struct hw_perf_counter_ops perf_ops_cpu_migrations = { + .hw_perf_counter_enable = cpu_migrations_perf_counter_enable, + .hw_perf_counter_disable = cpu_migrations_perf_counter_disable, + .hw_perf_counter_read = cpu_migrations_perf_counter_read, +}; + static const struct hw_perf_counter_ops * sw_perf_counter_init(struct perf_counter *counter) { @@ -951,6 +997,9 @@ sw_perf_counter_init(struct perf_counter *counter) case PERF_COUNT_CONTEXT_SWITCHES: hw_ops = &perf_ops_context_switches; break; + case PERF_COUNT_CPU_MIGRATIONS: + hw_ops = &perf_ops_cpu_migrations; + break; default: break; } diff --git a/kernel/sched.c b/kernel/sched.c index 5c3f4106314..382cfdb5e38 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -1852,12 +1852,14 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu) p->se.sleep_start -= clock_offset; if (p->se.block_start) p->se.block_start -= clock_offset; +#endif if (old_cpu != new_cpu) { - schedstat_inc(p, se.nr_migrations); + p->se.nr_migrations++; +#ifdef CONFIG_SCHEDSTATS if (task_hot(p, old_rq->clock, NULL)) schedstat_inc(p, se.nr_forced2_migrations); - } #endif + } p->se.vruntime -= old_cfsrq->min_vruntime - new_cfsrq->min_vruntime; @@ -2375,6 +2377,7 @@ static void __sched_fork(struct task_struct *p) p->se.exec_start = 0; p->se.sum_exec_runtime = 0; p->se.prev_sum_exec_runtime = 0; + p->se.nr_migrations = 0; p->se.last_wakeup = 0; p->se.avg_overlap = 0; -- cgit v1.2.3-70-g09d2 From 23a185ca8abbeef64b6ffc33059b1d630e43ec10 Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Mon, 9 Feb 2009 22:42:47 +1100 Subject: perf_counters: make software counters work as per-cpu counters Impact: kernel crash fix Yanmin Zhang reported that using a PERF_COUNT_TASK_CLOCK software counter as a per-cpu counter would reliably crash the system, because it calls __task_delta_exec with a null pointer. The page fault, context switch and cpu migration counters also won't function correctly as per-cpu counters since they reference the current task. This fixes the problem by redirecting the task_clock counter to the cpu_clock counter when used as a per-cpu counter, and by implementing per-cpu page fault, context switch and cpu migration counters. Along the way, this: - Initializes counter->ctx earlier, in perf_counter_alloc, so that sw_perf_counter_init can use it - Adds code to kernel/sched.c to count task migrations into each cpu, in rq->nr_migrations_in - Exports the per-cpu context switch and task migration counts via new functions added to kernel/sched.c - Makes sure that if sw_perf_counter_init fails, we don't try to initialize the counter as a hardware counter. Since the user has passed a negative, non-raw event type, they clearly don't intend for it to be interpreted as a hardware event. Reported-by: "Zhang Yanmin" Signed-off-by: Paul Mackerras Signed-off-by: Ingo Molnar --- include/linux/sched.h | 2 ++ kernel/perf_counter.c | 78 +++++++++++++++++++++++++++++---------------------- kernel/sched.c | 17 +++++++++++ 3 files changed, 64 insertions(+), 33 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index b85b10abf77..1e5f70062a9 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -137,6 +137,8 @@ extern unsigned long nr_running(void); extern unsigned long nr_uninterruptible(void); extern unsigned long nr_active(void); extern unsigned long nr_iowait(void); +extern u64 cpu_nr_switches(int cpu); +extern u64 cpu_nr_migrations(int cpu); struct seq_file; struct cfs_rq; diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index f27a7e9f3c4..544193cbc47 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -20,6 +20,8 @@ #include #include #include +#include +#include /* * Each CPU has a list of per CPU counters: @@ -502,7 +504,6 @@ perf_install_in_context(struct perf_counter_context *ctx, { struct task_struct *task = ctx->task; - counter->ctx = ctx; if (!task) { /* * Per cpu counters are installed via an smp call and @@ -1417,11 +1418,19 @@ static const struct hw_perf_counter_ops perf_ops_task_clock = { .read = task_clock_perf_counter_read, }; -static u64 get_page_faults(void) +#ifdef CONFIG_VM_EVENT_COUNTERS +#define cpu_page_faults() __get_cpu_var(vm_event_states).event[PGFAULT] +#else +#define cpu_page_faults() 0 +#endif + +static u64 get_page_faults(struct perf_counter *counter) { - struct task_struct *curr = current; + struct task_struct *curr = counter->ctx->task; - return curr->maj_flt + curr->min_flt; + if (curr) + return curr->maj_flt + curr->min_flt; + return cpu_page_faults(); } static void page_faults_perf_counter_update(struct perf_counter *counter) @@ -1430,7 +1439,7 @@ static void page_faults_perf_counter_update(struct perf_counter *counter) s64 delta; prev = atomic64_read(&counter->hw.prev_count); - now = get_page_faults(); + now = get_page_faults(counter); atomic64_set(&counter->hw.prev_count, now); @@ -1446,11 +1455,7 @@ static void page_faults_perf_counter_read(struct perf_counter *counter) static int page_faults_perf_counter_enable(struct perf_counter *counter) { - /* - * page-faults is a per-task value already, - * so we dont have to clear it on switch-in. - */ - + atomic64_set(&counter->hw.prev_count, get_page_faults(counter)); return 0; } @@ -1465,11 +1470,13 @@ static const struct hw_perf_counter_ops perf_ops_page_faults = { .read = page_faults_perf_counter_read, }; -static u64 get_context_switches(void) +static u64 get_context_switches(struct perf_counter *counter) { - struct task_struct *curr = current; + struct task_struct *curr = counter->ctx->task; - return curr->nvcsw + curr->nivcsw; + if (curr) + return curr->nvcsw + curr->nivcsw; + return cpu_nr_switches(smp_processor_id()); } static void context_switches_perf_counter_update(struct perf_counter *counter) @@ -1478,7 +1485,7 @@ static void context_switches_perf_counter_update(struct perf_counter *counter) s64 delta; prev = atomic64_read(&counter->hw.prev_count); - now = get_context_switches(); + now = get_context_switches(counter); atomic64_set(&counter->hw.prev_count, now); @@ -1494,11 +1501,7 @@ static void context_switches_perf_counter_read(struct perf_counter *counter) static int context_switches_perf_counter_enable(struct perf_counter *counter) { - /* - * ->nvcsw + curr->nivcsw is a per-task value already, - * so we dont have to clear it on switch-in. - */ - + atomic64_set(&counter->hw.prev_count, get_context_switches(counter)); return 0; } @@ -1513,9 +1516,13 @@ static const struct hw_perf_counter_ops perf_ops_context_switches = { .read = context_switches_perf_counter_read, }; -static inline u64 get_cpu_migrations(void) +static inline u64 get_cpu_migrations(struct perf_counter *counter) { - return current->se.nr_migrations; + struct task_struct *curr = counter->ctx->task; + + if (curr) + return curr->se.nr_migrations; + return cpu_nr_migrations(smp_processor_id()); } static void cpu_migrations_perf_counter_update(struct perf_counter *counter) @@ -1524,7 +1531,7 @@ static void cpu_migrations_perf_counter_update(struct perf_counter *counter) s64 delta; prev = atomic64_read(&counter->hw.prev_count); - now = get_cpu_migrations(); + now = get_cpu_migrations(counter); atomic64_set(&counter->hw.prev_count, now); @@ -1540,11 +1547,7 @@ static void cpu_migrations_perf_counter_read(struct perf_counter *counter) static int cpu_migrations_perf_counter_enable(struct perf_counter *counter) { - /* - * se.nr_migrations is a per-task value already, - * so we dont have to clear it on switch-in. - */ - + atomic64_set(&counter->hw.prev_count, get_cpu_migrations(counter)); return 0; } @@ -1569,7 +1572,14 @@ sw_perf_counter_init(struct perf_counter *counter) hw_ops = &perf_ops_cpu_clock; break; case PERF_COUNT_TASK_CLOCK: - hw_ops = &perf_ops_task_clock; + /* + * If the user instantiates this as a per-cpu counter, + * use the cpu_clock counter instead. + */ + if (counter->ctx->task) + hw_ops = &perf_ops_task_clock; + else + hw_ops = &perf_ops_cpu_clock; break; case PERF_COUNT_PAGE_FAULTS: hw_ops = &perf_ops_page_faults; @@ -1592,6 +1602,7 @@ sw_perf_counter_init(struct perf_counter *counter) static struct perf_counter * perf_counter_alloc(struct perf_counter_hw_event *hw_event, int cpu, + struct perf_counter_context *ctx, struct perf_counter *group_leader, gfp_t gfpflags) { @@ -1623,6 +1634,7 @@ perf_counter_alloc(struct perf_counter_hw_event *hw_event, counter->wakeup_pending = 0; counter->group_leader = group_leader; counter->hw_ops = NULL; + counter->ctx = ctx; counter->state = PERF_COUNTER_STATE_INACTIVE; if (hw_event->disabled) @@ -1631,7 +1643,7 @@ perf_counter_alloc(struct perf_counter_hw_event *hw_event, hw_ops = NULL; if (!hw_event->raw && hw_event->type < 0) hw_ops = sw_perf_counter_init(counter); - if (!hw_ops) + else hw_ops = hw_perf_counter_init(counter); if (!hw_ops) { @@ -1707,7 +1719,8 @@ sys_perf_counter_open(struct perf_counter_hw_event *hw_event_uptr __user, } ret = -EINVAL; - counter = perf_counter_alloc(&hw_event, cpu, group_leader, GFP_KERNEL); + counter = perf_counter_alloc(&hw_event, cpu, ctx, group_leader, + GFP_KERNEL); if (!counter) goto err_put_context; @@ -1777,15 +1790,14 @@ inherit_counter(struct perf_counter *parent_counter, parent_counter = parent_counter->parent; child_counter = perf_counter_alloc(&parent_counter->hw_event, - parent_counter->cpu, group_leader, - GFP_KERNEL); + parent_counter->cpu, child_ctx, + group_leader, GFP_KERNEL); if (!child_counter) return NULL; /* * Link it up in the child's context: */ - child_counter->ctx = child_ctx; child_counter->task = child; list_add_counter(child_counter, child_ctx); child_ctx->nr_counters++; diff --git a/kernel/sched.c b/kernel/sched.c index 8db1a4cf208..173768f142a 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -558,6 +558,7 @@ struct rq { struct load_weight load; unsigned long nr_load_updates; u64 nr_switches; + u64 nr_migrations_in; struct cfs_rq cfs; struct rt_rq rt; @@ -1908,6 +1909,7 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu) #endif if (old_cpu != new_cpu) { p->se.nr_migrations++; + new_rq->nr_migrations_in++; #ifdef CONFIG_SCHEDSTATS if (task_hot(p, old_rq->clock, NULL)) schedstat_inc(p, se.nr_forced2_migrations); @@ -2810,6 +2812,21 @@ unsigned long nr_active(void) return running + uninterruptible; } +/* + * Externally visible per-cpu scheduler statistics: + * cpu_nr_switches(cpu) - number of context switches on that cpu + * cpu_nr_migrations(cpu) - number of migrations into that cpu + */ +u64 cpu_nr_switches(int cpu) +{ + return cpu_rq(cpu)->nr_switches; +} + +u64 cpu_nr_migrations(int cpu) +{ + return cpu_rq(cpu)->nr_migrations_in; +} + /* * Update rq->cpu_load[] statistics. This function is usually called every * scheduler tick (TICK_NSEC). -- cgit v1.2.3-70-g09d2 From 4a0deca657f3dbb8a707b5dc8f173beec01e7ed2 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 19 Mar 2009 20:26:12 +0100 Subject: perf_counter: generic context switch event Impact: cleanup Use the generic software events for context switches. Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Steven Rostedt Orig-LKML-Reference: <20090319194233.283522645@chello.nl> Signed-off-by: Ingo Molnar --- include/linux/sched.h | 1 - kernel/perf_counter.c | 60 ++++----------------------------------------------- kernel/sched.c | 6 ------ 3 files changed, 4 insertions(+), 63 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index 75b2fc5306d..7ed41f7c5ac 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -138,7 +138,6 @@ extern unsigned long nr_running(void); extern unsigned long nr_uninterruptible(void); extern unsigned long nr_active(void); extern unsigned long nr_iowait(void); -extern u64 cpu_nr_switches(int cpu); extern u64 cpu_nr_migrations(int cpu); extern unsigned long get_parent_ip(unsigned long addr); diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 4c4e9eb37ab..99d5930f0a5 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -710,10 +710,13 @@ void perf_counter_task_sched_out(struct task_struct *task, int cpu) { struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu); struct perf_counter_context *ctx = &task->perf_counter_ctx; + struct pt_regs *regs; if (likely(!cpuctx->task_ctx)) return; + regs = task_pt_regs(task); + perf_swcounter_event(PERF_COUNT_CONTEXT_SWITCHES, 1, 1, regs); __perf_counter_sched_out(ctx, cpuctx); cpuctx->task_ctx = NULL; @@ -1667,58 +1670,6 @@ static const struct hw_perf_counter_ops perf_ops_task_clock = { .read = task_clock_perf_counter_read, }; -/* - * Software counter: context switches - */ - -static u64 get_context_switches(struct perf_counter *counter) -{ - struct task_struct *curr = counter->ctx->task; - - if (curr) - return curr->nvcsw + curr->nivcsw; - return cpu_nr_switches(smp_processor_id()); -} - -static void context_switches_perf_counter_update(struct perf_counter *counter) -{ - u64 prev, now; - s64 delta; - - prev = atomic64_read(&counter->hw.prev_count); - now = get_context_switches(counter); - - atomic64_set(&counter->hw.prev_count, now); - - delta = now - prev; - - atomic64_add(delta, &counter->count); -} - -static void context_switches_perf_counter_read(struct perf_counter *counter) -{ - context_switches_perf_counter_update(counter); -} - -static int context_switches_perf_counter_enable(struct perf_counter *counter) -{ - if (counter->prev_state <= PERF_COUNTER_STATE_OFF) - atomic64_set(&counter->hw.prev_count, - get_context_switches(counter)); - return 0; -} - -static void context_switches_perf_counter_disable(struct perf_counter *counter) -{ - context_switches_perf_counter_update(counter); -} - -static const struct hw_perf_counter_ops perf_ops_context_switches = { - .enable = context_switches_perf_counter_enable, - .disable = context_switches_perf_counter_disable, - .read = context_switches_perf_counter_read, -}; - /* * Software counter: cpu migrations */ @@ -1808,11 +1759,8 @@ sw_perf_counter_init(struct perf_counter *counter) case PERF_COUNT_PAGE_FAULTS: case PERF_COUNT_PAGE_FAULTS_MIN: case PERF_COUNT_PAGE_FAULTS_MAJ: - hw_ops = &perf_ops_generic; - break; case PERF_COUNT_CONTEXT_SWITCHES: - if (!counter->hw_event.exclude_kernel) - hw_ops = &perf_ops_context_switches; + hw_ops = &perf_ops_generic; break; case PERF_COUNT_CPU_MIGRATIONS: if (!counter->hw_event.exclude_kernel) diff --git a/kernel/sched.c b/kernel/sched.c index 39e70860216..f76e3c0188a 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2900,14 +2900,8 @@ unsigned long nr_active(void) /* * Externally visible per-cpu scheduler statistics: - * cpu_nr_switches(cpu) - number of context switches on that cpu * cpu_nr_migrations(cpu) - number of migrations into that cpu */ -u64 cpu_nr_switches(int cpu) -{ - return cpu_rq(cpu)->nr_switches; -} - u64 cpu_nr_migrations(int cpu) { return cpu_rq(cpu)->nr_migrations_in; -- cgit v1.2.3-70-g09d2 From a26b89f05d194413c7238e0bea071054f6b5d3c8 Mon Sep 17 00:00:00 2001 From: Markus Metzger Date: Fri, 3 Apr 2009 16:43:34 +0200 Subject: sched, hw-branch-tracer: add wait_task_context_switch() function to sched.h Add a function to wait until some other task has been switched out at least once. This differs from wait_task_inactive() subtly, in that the latter will wait until the task has left the CPU. Signed-off-by: Markus Metzger Cc: markus.t.metzger@gmail.com Cc: roland@redhat.com Cc: eranian@googlemail.com Cc: oleg@redhat.com Cc: juan.villacis@intel.com Cc: ak@linux.jf.intel.com LKML-Reference: <20090403144549.794157000@intel.com> Signed-off-by: Ingo Molnar --- include/linux/sched.h | 2 ++ kernel/sched.c | 43 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 45 insertions(+) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index b94f3541f67..a5b9a83065f 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1993,8 +1993,10 @@ extern void set_task_comm(struct task_struct *tsk, char *from); extern char *get_task_comm(char *to, struct task_struct *tsk); #ifdef CONFIG_SMP +extern void wait_task_context_switch(struct task_struct *p); extern unsigned long wait_task_inactive(struct task_struct *, long match_state); #else +static inline void wait_task_context_switch(struct task_struct *p) {} static inline unsigned long wait_task_inactive(struct task_struct *p, long match_state) { diff --git a/kernel/sched.c b/kernel/sched.c index 6cc1fd5d507..f91bc8141dc 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -2002,6 +2002,49 @@ migrate_task(struct task_struct *p, int dest_cpu, struct migration_req *req) return 1; } +/* + * wait_task_context_switch - wait for a thread to complete at least one + * context switch. + * + * @p must not be current. + */ +void wait_task_context_switch(struct task_struct *p) +{ + unsigned long nvcsw, nivcsw, flags; + int running; + struct rq *rq; + + nvcsw = p->nvcsw; + nivcsw = p->nivcsw; + for (;;) { + /* + * The runqueue is assigned before the actual context + * switch. We need to take the runqueue lock. + * + * We could check initially without the lock but it is + * very likely that we need to take the lock in every + * iteration. + */ + rq = task_rq_lock(p, &flags); + running = task_running(rq, p); + task_rq_unlock(rq, &flags); + + if (likely(!running)) + break; + /* + * The switch count is incremented before the actual + * context switch. We thus wait for two switches to be + * sure at least one completed. + */ + if ((p->nvcsw - nvcsw) > 1) + break; + if ((p->nivcsw - nivcsw) > 1) + break; + + cpu_relax(); + } +} + /* * wait_task_inactive - wait for a thread to unschedule. * -- cgit v1.2.3-70-g09d2 From e2b371f00a6f529f6362654239bdec8dcd510760 Mon Sep 17 00:00:00 2001 From: Markus Metzger Date: Fri, 3 Apr 2009 16:43:35 +0200 Subject: mm, x86, ptrace, bts: defer branch trace stopping When a ptraced task is unlinked, we need to stop branch tracing for that task. Since the unlink is called with interrupts disabled, and we need interrupts enabled to stop branch tracing, we defer the work. Collect all branch tracing related stuff in a branch tracing context. Reviewed-by: Oleg Nesterov Signed-off-by: Markus Metzger Cc: Andrew Morton Cc: Peter Zijlstra Cc: roland@redhat.com Cc: eranian@googlemail.com Cc: juan.villacis@intel.com Cc: ak@linux.jf.intel.com LKML-Reference: <20090403144550.712401000@intel.com> Signed-off-by: Ingo Molnar --- arch/x86/include/asm/processor.h | 4 - arch/x86/kernel/ptrace.c | 254 ++++++++++++++++++++++++++------------- include/linux/mm.h | 3 +- include/linux/sched.h | 9 +- mm/mlock.c | 13 +- 5 files changed, 179 insertions(+), 104 deletions(-) (limited to 'include/linux/sched.h') diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h index 34c52370f2f..2483807e06e 100644 --- a/arch/x86/include/asm/processor.h +++ b/arch/x86/include/asm/processor.h @@ -458,10 +458,6 @@ struct thread_struct { /* Debug Store context; see include/asm-x86/ds.h; goes into MSR_IA32_DS_AREA */ struct ds_context *ds_ctx; #endif /* CONFIG_X86_DS */ -#ifdef CONFIG_X86_PTRACE_BTS -/* the signal to send on a bts buffer overflow */ - unsigned int bts_ovfl_signal; -#endif /* CONFIG_X86_PTRACE_BTS */ }; static inline unsigned long native_get_debugreg(int regno) diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index fe9345c967d..7c21d1e8cae 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -22,6 +22,7 @@ #include #include #include +#include #include #include @@ -577,17 +578,119 @@ static int ioperm_get(struct task_struct *target, } #ifdef CONFIG_X86_PTRACE_BTS +/* + * A branch trace store context. + * + * Contexts may only be installed by ptrace_bts_config() and only for + * ptraced tasks. + * + * Contexts are destroyed when the tracee is detached from the tracer. + * The actual destruction work requires interrupts enabled, so the + * work is deferred and will be scheduled during __ptrace_unlink(). + * + * Contexts hold an additional task_struct reference on the traced + * task, as well as a reference on the tracer's mm. + * + * Ptrace already holds a task_struct for the duration of ptrace operations, + * but since destruction is deferred, it may be executed after both + * tracer and tracee exited. + */ +struct bts_context { + /* The branch trace handle. */ + struct bts_tracer *tracer; + + /* The buffer used to store the branch trace and its size. */ + void *buffer; + unsigned int size; + + /* The mm that paid for the above buffer. */ + struct mm_struct *mm; + + /* The task this context belongs to. */ + struct task_struct *task; + + /* The signal to send on a bts buffer overflow. */ + unsigned int bts_ovfl_signal; + + /* The work struct to destroy a context. */ + struct work_struct work; +}; + +static inline void alloc_bts_buffer(struct bts_context *context, + unsigned int size) +{ + void *buffer; + + buffer = alloc_locked_buffer(size); + if (buffer) { + context->buffer = buffer; + context->size = size; + context->mm = get_task_mm(current); + } +} + +static inline void free_bts_buffer(struct bts_context *context) +{ + if (!context->buffer) + return; + + kfree(context->buffer); + context->buffer = NULL; + + refund_locked_buffer_memory(context->mm, context->size); + context->size = 0; + + mmput(context->mm); + context->mm = NULL; +} + +static void free_bts_context_work(struct work_struct *w) +{ + struct bts_context *context; + + context = container_of(w, struct bts_context, work); + + ds_release_bts(context->tracer); + put_task_struct(context->task); + free_bts_buffer(context); + kfree(context); +} + +static inline void free_bts_context(struct bts_context *context) +{ + INIT_WORK(&context->work, free_bts_context_work); + schedule_work(&context->work); +} + +static inline struct bts_context *alloc_bts_context(struct task_struct *task) +{ + struct bts_context *context = kzalloc(sizeof(*context), GFP_KERNEL); + if (context) { + context->task = task; + task->bts = context; + + get_task_struct(task); + } + + return context; +} + static int ptrace_bts_read_record(struct task_struct *child, size_t index, struct bts_struct __user *out) { + struct bts_context *context; const struct bts_trace *trace; struct bts_struct bts; const unsigned char *at; int error; - trace = ds_read_bts(child->bts); + context = child->bts; + if (!context) + return -ESRCH; + + trace = ds_read_bts(context->tracer); if (!trace) - return -EPERM; + return -ESRCH; at = trace->ds.top - ((index + 1) * trace->ds.size); if ((void *)at < trace->ds.begin) @@ -596,7 +699,7 @@ static int ptrace_bts_read_record(struct task_struct *child, size_t index, if (!trace->read) return -EOPNOTSUPP; - error = trace->read(child->bts, at, &bts); + error = trace->read(context->tracer, at, &bts); if (error < 0) return error; @@ -610,13 +713,18 @@ static int ptrace_bts_drain(struct task_struct *child, long size, struct bts_struct __user *out) { + struct bts_context *context; const struct bts_trace *trace; const unsigned char *at; int error, drained = 0; - trace = ds_read_bts(child->bts); + context = child->bts; + if (!context) + return -ESRCH; + + trace = ds_read_bts(context->tracer); if (!trace) - return -EPERM; + return -ESRCH; if (!trace->read) return -EOPNOTSUPP; @@ -627,9 +735,8 @@ static int ptrace_bts_drain(struct task_struct *child, for (at = trace->ds.begin; (void *)at < trace->ds.top; out++, drained++, at += trace->ds.size) { struct bts_struct bts; - int error; - error = trace->read(child->bts, at, &bts); + error = trace->read(context->tracer, at, &bts); if (error < 0) return error; @@ -639,35 +746,18 @@ static int ptrace_bts_drain(struct task_struct *child, memset(trace->ds.begin, 0, trace->ds.n * trace->ds.size); - error = ds_reset_bts(child->bts); + error = ds_reset_bts(context->tracer); if (error < 0) return error; return drained; } -static int ptrace_bts_allocate_buffer(struct task_struct *child, size_t size) -{ - child->bts_buffer = alloc_locked_buffer(size); - if (!child->bts_buffer) - return -ENOMEM; - - child->bts_size = size; - - return 0; -} - -static void ptrace_bts_free_buffer(struct task_struct *child) -{ - free_locked_buffer(child->bts_buffer, child->bts_size); - child->bts_buffer = NULL; - child->bts_size = 0; -} - static int ptrace_bts_config(struct task_struct *child, long cfg_size, const struct ptrace_bts_config __user *ucfg) { + struct bts_context *context; struct ptrace_bts_config cfg; unsigned int flags = 0; @@ -677,28 +767,31 @@ static int ptrace_bts_config(struct task_struct *child, if (copy_from_user(&cfg, ucfg, sizeof(cfg))) return -EFAULT; - if (child->bts) { - ds_release_bts(child->bts); - child->bts = NULL; - } + context = child->bts; + if (!context) + context = alloc_bts_context(child); + if (!context) + return -ENOMEM; if (cfg.flags & PTRACE_BTS_O_SIGNAL) { if (!cfg.signal) return -EINVAL; - child->thread.bts_ovfl_signal = cfg.signal; return -EOPNOTSUPP; + context->bts_ovfl_signal = cfg.signal; } - if ((cfg.flags & PTRACE_BTS_O_ALLOC) && - (cfg.size != child->bts_size)) { - int error; + ds_release_bts(context->tracer); + context->tracer = NULL; - ptrace_bts_free_buffer(child); + if ((cfg.flags & PTRACE_BTS_O_ALLOC) && (cfg.size != context->size)) { + free_bts_buffer(context); + if (!cfg.size) + return 0; - error = ptrace_bts_allocate_buffer(child, cfg.size); - if (error < 0) - return error; + alloc_bts_buffer(context, cfg.size); + if (!context->buffer) + return -ENOMEM; } if (cfg.flags & PTRACE_BTS_O_TRACE) @@ -707,15 +800,13 @@ static int ptrace_bts_config(struct task_struct *child, if (cfg.flags & PTRACE_BTS_O_SCHED) flags |= BTS_TIMESTAMPS; - child->bts = ds_request_bts(child, child->bts_buffer, child->bts_size, - /* ovfl = */ NULL, /* th = */ (size_t)-1, - flags); - if (IS_ERR(child->bts)) { - int error = PTR_ERR(child->bts); - - ptrace_bts_free_buffer(child); - child->bts = NULL; + context->tracer = ds_request_bts(child, context->buffer, context->size, + NULL, (size_t)-1, flags); + if (unlikely(IS_ERR(context->tracer))) { + int error = PTR_ERR(context->tracer); + free_bts_buffer(context); + context->tracer = NULL; return error; } @@ -726,20 +817,25 @@ static int ptrace_bts_status(struct task_struct *child, long cfg_size, struct ptrace_bts_config __user *ucfg) { + struct bts_context *context; const struct bts_trace *trace; struct ptrace_bts_config cfg; + context = child->bts; + if (!context) + return -ESRCH; + if (cfg_size < sizeof(cfg)) return -EIO; - trace = ds_read_bts(child->bts); + trace = ds_read_bts(context->tracer); if (!trace) - return -EPERM; + return -ESRCH; memset(&cfg, 0, sizeof(cfg)); - cfg.size = trace->ds.end - trace->ds.begin; - cfg.signal = child->thread.bts_ovfl_signal; - cfg.bts_size = sizeof(struct bts_struct); + cfg.size = trace->ds.end - trace->ds.begin; + cfg.signal = context->bts_ovfl_signal; + cfg.bts_size = sizeof(struct bts_struct); if (cfg.signal) cfg.flags |= PTRACE_BTS_O_SIGNAL; @@ -758,67 +854,56 @@ static int ptrace_bts_status(struct task_struct *child, static int ptrace_bts_clear(struct task_struct *child) { + struct bts_context *context; const struct bts_trace *trace; - trace = ds_read_bts(child->bts); + context = child->bts; + if (!context) + return -ESRCH; + + trace = ds_read_bts(context->tracer); if (!trace) - return -EPERM; + return -ESRCH; memset(trace->ds.begin, 0, trace->ds.n * trace->ds.size); - return ds_reset_bts(child->bts); + return ds_reset_bts(context->tracer); } static int ptrace_bts_size(struct task_struct *child) { + struct bts_context *context; const struct bts_trace *trace; - trace = ds_read_bts(child->bts); + context = child->bts; + if (!context) + return -ESRCH; + + trace = ds_read_bts(context->tracer); if (!trace) - return -EPERM; + return -ESRCH; return (trace->ds.top - trace->ds.begin) / trace->ds.size; } -static void ptrace_bts_fork(struct task_struct *tsk) +static inline void ptrace_bts_fork(struct task_struct *tsk) { tsk->bts = NULL; - tsk->bts_buffer = NULL; - tsk->bts_size = 0; - tsk->thread.bts_ovfl_signal = 0; } -static void ptrace_bts_untrace(struct task_struct *child) +/* + * Called from __ptrace_unlink() after the child has been moved back + * to its original parent. + */ +static inline void ptrace_bts_untrace(struct task_struct *child) { if (unlikely(child->bts)) { - ds_release_bts(child->bts); + free_bts_context(child->bts); child->bts = NULL; - - /* We cannot update total_vm and locked_vm since - child's mm is already gone. But we can reclaim the - memory. */ - kfree(child->bts_buffer); - child->bts_buffer = NULL; - child->bts_size = 0; } } - -static void ptrace_bts_detach(struct task_struct *child) -{ - /* - * Ptrace_detach() races with ptrace_untrace() in case - * the child dies and is reaped by another thread. - * - * We only do the memory accounting at this point and - * leave the buffer deallocation and the bts tracer - * release to ptrace_bts_untrace() which will be called - * later on with tasklist_lock held. - */ - release_locked_buffer(child->bts_buffer, child->bts_size); -} #else static inline void ptrace_bts_fork(struct task_struct *tsk) {} -static inline void ptrace_bts_detach(struct task_struct *child) {} static inline void ptrace_bts_untrace(struct task_struct *child) {} #endif /* CONFIG_X86_PTRACE_BTS */ @@ -843,7 +928,6 @@ void ptrace_disable(struct task_struct *child) #ifdef TIF_SYSCALL_EMU clear_tsk_thread_flag(child, TIF_SYSCALL_EMU); #endif - ptrace_bts_detach(child); } #if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION diff --git a/include/linux/mm.h b/include/linux/mm.h index bff1f0d475c..64d8ed2538a 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -13,6 +13,7 @@ #include #include #include +#include struct mempolicy; struct anon_vma; @@ -1321,6 +1322,6 @@ void vmemmap_populate_print_last(void); extern void *alloc_locked_buffer(size_t size); extern void free_locked_buffer(void *buffer, size_t size); -extern void release_locked_buffer(void *buffer, size_t size); +extern void refund_locked_buffer_memory(struct mm_struct *mm, size_t size); #endif /* __KERNEL__ */ #endif /* _LINUX_MM_H */ diff --git a/include/linux/sched.h b/include/linux/sched.h index a5b9a83065f..52b8cd049c2 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -96,8 +96,8 @@ struct exec_domain; struct futex_pi_state; struct robust_list_head; struct bio; -struct bts_tracer; struct fs_struct; +struct bts_context; /* * List of flags we want to share for kernel threads, @@ -1210,12 +1210,7 @@ struct task_struct { * This is the tracer handle for the ptrace BTS extension. * This field actually belongs to the ptracer task. */ - struct bts_tracer *bts; - /* - * The buffer to hold the BTS data. - */ - void *bts_buffer; - size_t bts_size; + struct bts_context *bts; #endif /* CONFIG_X86_PTRACE_BTS */ /* PID/PID hash table linkage. */ diff --git a/mm/mlock.c b/mm/mlock.c index cbe9e0581b7..749383b442c 100644 --- a/mm/mlock.c +++ b/mm/mlock.c @@ -660,21 +660,20 @@ void *alloc_locked_buffer(size_t size) return buffer; } -void release_locked_buffer(void *buffer, size_t size) +void refund_locked_buffer_memory(struct mm_struct *mm, size_t size) { unsigned long pgsz = PAGE_ALIGN(size) >> PAGE_SHIFT; - down_write(¤t->mm->mmap_sem); + down_write(&mm->mmap_sem); - current->mm->total_vm -= pgsz; - current->mm->locked_vm -= pgsz; + mm->total_vm -= pgsz; + mm->locked_vm -= pgsz; - up_write(¤t->mm->mmap_sem); + up_write(&mm->mmap_sem); } void free_locked_buffer(void *buffer, size_t size) { - release_locked_buffer(buffer, size); - + refund_locked_buffer_memory(current->mm, size); kfree(buffer); } -- cgit v1.2.3-70-g09d2 From 0f4814065ff8c24ca8bfd75c9b73502be152c287 Mon Sep 17 00:00:00 2001 From: Markus Metzger Date: Fri, 3 Apr 2009 16:43:48 +0200 Subject: x86, ptrace: add bts context unconditionally Add the ptrace bts context field to task_struct unconditionally. Initialize the field directly in copy_process(). Remove all the unneeded functionality used to initialize that field. Signed-off-by: Markus Metzger Cc: roland@redhat.com Cc: eranian@googlemail.com Cc: oleg@redhat.com Cc: juan.villacis@intel.com Cc: ak@linux.jf.intel.com LKML-Reference: <20090403144603.292754000@intel.com> Signed-off-by: Ingo Molnar --- arch/x86/include/asm/ptrace.h | 9 ++++----- arch/x86/kernel/ptrace.c | 20 +------------------- include/linux/ptrace.h | 10 ---------- include/linux/sched.h | 2 -- kernel/fork.c | 4 ++-- kernel/ptrace.c | 10 ---------- 6 files changed, 7 insertions(+), 48 deletions(-) (limited to 'include/linux/sched.h') diff --git a/arch/x86/include/asm/ptrace.h b/arch/x86/include/asm/ptrace.h index e304b66abee..5cdd19f20b5 100644 --- a/arch/x86/include/asm/ptrace.h +++ b/arch/x86/include/asm/ptrace.h @@ -235,12 +235,11 @@ extern int do_get_thread_area(struct task_struct *p, int idx, extern int do_set_thread_area(struct task_struct *p, int idx, struct user_desc __user *info, int can_allocate); -extern void x86_ptrace_untrace(struct task_struct *); -extern void x86_ptrace_fork(struct task_struct *child, - unsigned long clone_flags); +#ifdef CONFIG_X86_PTRACE_BTS +extern void ptrace_bts_untrace(struct task_struct *tsk); -#define arch_ptrace_untrace(tsk) x86_ptrace_untrace(tsk) -#define arch_ptrace_fork(child, flags) x86_ptrace_fork(child, flags) +#define arch_ptrace_untrace(tsk) ptrace_bts_untrace(tsk) +#endif /* CONFIG_X86_PTRACE_BTS */ #endif /* __KERNEL__ */ diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index adbb24322d8..b32a8ee5338 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -887,37 +887,19 @@ static int ptrace_bts_size(struct task_struct *child) return (trace->ds.top - trace->ds.begin) / trace->ds.size; } -static inline void ptrace_bts_fork(struct task_struct *tsk) -{ - tsk->bts = NULL; -} - /* * Called from __ptrace_unlink() after the child has been moved back * to its original parent. */ -static inline void ptrace_bts_untrace(struct task_struct *child) +void ptrace_bts_untrace(struct task_struct *child) { if (unlikely(child->bts)) { free_bts_context(child->bts); child->bts = NULL; } } -#else -static inline void ptrace_bts_fork(struct task_struct *tsk) {} -static inline void ptrace_bts_untrace(struct task_struct *child) {} #endif /* CONFIG_X86_PTRACE_BTS */ -void x86_ptrace_fork(struct task_struct *child, unsigned long clone_flags) -{ - ptrace_bts_fork(child); -} - -void x86_ptrace_untrace(struct task_struct *child) -{ - ptrace_bts_untrace(child); -} - /* * Called by kernel/ptrace.c when detaching.. * diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h index 67c15653fc2..59e133d39d5 100644 --- a/include/linux/ptrace.h +++ b/include/linux/ptrace.h @@ -95,7 +95,6 @@ extern void __ptrace_link(struct task_struct *child, struct task_struct *new_parent); extern void __ptrace_unlink(struct task_struct *child); extern void exit_ptrace(struct task_struct *tracer); -extern void ptrace_fork(struct task_struct *task, unsigned long clone_flags); #define PTRACE_MODE_READ 1 #define PTRACE_MODE_ATTACH 2 /* Returns 0 on success, -errno on denial. */ @@ -327,15 +326,6 @@ static inline void user_enable_block_step(struct task_struct *task) #define arch_ptrace_untrace(task) do { } while (0) #endif -#ifndef arch_ptrace_fork -/* - * Do machine-specific work to initialize a new task. - * - * This is called from copy_process(). - */ -#define arch_ptrace_fork(child, clone_flags) do { } while (0) -#endif - extern int task_current_syscall(struct task_struct *target, long *callno, unsigned long args[6], unsigned int maxargs, unsigned long *sp, unsigned long *pc); diff --git a/include/linux/sched.h b/include/linux/sched.h index 52b8cd049c2..451186a22ef 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1205,13 +1205,11 @@ struct task_struct { struct list_head ptraced; struct list_head ptrace_entry; -#ifdef CONFIG_X86_PTRACE_BTS /* * This is the tracer handle for the ptrace BTS extension. * This field actually belongs to the ptracer task. */ struct bts_context *bts; -#endif /* CONFIG_X86_PTRACE_BTS */ /* PID/PID hash table linkage. */ struct pid_link pids[PIDTYPE_MAX]; diff --git a/kernel/fork.c b/kernel/fork.c index 660c2b8765b..69bde7a22e9 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1086,8 +1086,8 @@ static struct task_struct *copy_process(unsigned long clone_flags, #ifdef CONFIG_DEBUG_MUTEXES p->blocked_on = NULL; /* not blocked yet */ #endif - if (unlikely(current->ptrace)) - ptrace_fork(p, clone_flags); + + p->bts = NULL; /* Perform scheduler related setup. Assign this task to a CPU. */ sched_fork(p, clone_flags); diff --git a/kernel/ptrace.c b/kernel/ptrace.c index aaad0ec3419..321127d965c 100644 --- a/kernel/ptrace.c +++ b/kernel/ptrace.c @@ -26,16 +26,6 @@ #include -/* - * Initialize a new task whose father had been ptraced. - * - * Called from copy_process(). - */ -void ptrace_fork(struct task_struct *child, unsigned long clone_flags) -{ - arch_ptrace_fork(child, clone_flags); -} - /* * ptrace a task: make the debugger its new parent and * move it to the ptrace list. -- cgit v1.2.3-70-g09d2 From 05725f7eb4b8acb147c5fc7b91397b1f6bcab00d Mon Sep 17 00:00:00 2001 From: Jiri Pirko Date: Tue, 14 Apr 2009 20:17:16 +0200 Subject: rculist: use list_entry_rcu in places where it's appropriate Use previously introduced list_entry_rcu instead of an open-coded list_entry + rcu_dereference combination. Signed-off-by: Jiri Pirko Reviewed-by: Paul E. McKenney Cc: dipankar@in.ibm.com LKML-Reference: <20090414181715.GA3634@psychotron.englab.brq.redhat.com> Signed-off-by: Ingo Molnar --- include/linux/sched.h | 8 +++++--- ipc/sem.c | 4 ++-- security/integrity/ima/ima_fs.c | 4 ++-- security/smack/smackfs.c | 8 ++++---- 4 files changed, 13 insertions(+), 11 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index b4c38bc8049..886df41e745 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -77,6 +77,7 @@ struct sched_param { #include #include #include +#include #include #include @@ -2010,7 +2011,8 @@ static inline unsigned long wait_task_inactive(struct task_struct *p, } #endif -#define next_task(p) list_entry(rcu_dereference((p)->tasks.next), struct task_struct, tasks) +#define next_task(p) \ + list_entry_rcu((p)->tasks.next, struct task_struct, tasks) #define for_each_process(p) \ for (p = &init_task ; (p = next_task(p)) != &init_task ; ) @@ -2049,8 +2051,8 @@ int same_thread_group(struct task_struct *p1, struct task_struct *p2) static inline struct task_struct *next_thread(const struct task_struct *p) { - return list_entry(rcu_dereference(p->thread_group.next), - struct task_struct, thread_group); + return list_entry_rcu(p->thread_group.next, + struct task_struct, thread_group); } static inline int thread_group_empty(struct task_struct *p) diff --git a/ipc/sem.c b/ipc/sem.c index 16a2189e96f..87c2b641fd7 100644 --- a/ipc/sem.c +++ b/ipc/sem.c @@ -1290,8 +1290,8 @@ void exit_sem(struct task_struct *tsk) int i; rcu_read_lock(); - un = list_entry(rcu_dereference(ulp->list_proc.next), - struct sem_undo, list_proc); + un = list_entry_rcu(ulp->list_proc.next, + struct sem_undo, list_proc); if (&un->list_proc == &ulp->list_proc) semid = -1; else diff --git a/security/integrity/ima/ima_fs.c b/security/integrity/ima/ima_fs.c index ffbe259700b..510186f0b72 100644 --- a/security/integrity/ima/ima_fs.c +++ b/security/integrity/ima/ima_fs.c @@ -84,8 +84,8 @@ static void *ima_measurements_next(struct seq_file *m, void *v, loff_t *pos) * against concurrent list-extension */ rcu_read_lock(); - qe = list_entry(rcu_dereference(qe->later.next), - struct ima_queue_entry, later); + qe = list_entry_rcu(qe->later.next, + struct ima_queue_entry, later); rcu_read_unlock(); (*pos)++; diff --git a/security/smack/smackfs.c b/security/smack/smackfs.c index e03a7e19c73..11d2cb19d7a 100644 --- a/security/smack/smackfs.c +++ b/security/smack/smackfs.c @@ -734,8 +734,8 @@ static void smk_netlbladdr_insert(struct smk_netlbladdr *new) return; } - m = list_entry(rcu_dereference(smk_netlbladdr_list.next), - struct smk_netlbladdr, list); + m = list_entry_rcu(smk_netlbladdr_list.next, + struct smk_netlbladdr, list); /* the comparison '>' is a bit hacky, but works */ if (new->smk_mask.s_addr > m->smk_mask.s_addr) { @@ -748,8 +748,8 @@ static void smk_netlbladdr_insert(struct smk_netlbladdr *new) list_add_rcu(&new->list, &m->list); return; } - m_next = list_entry(rcu_dereference(m->list.next), - struct smk_netlbladdr, list); + m_next = list_entry_rcu(m->list.next, + struct smk_netlbladdr, list); if (new->smk_mask.s_addr > m_next->smk_mask.s_addr) { list_add_rcu(&new->list, &m->list); return; -- cgit v1.2.3-70-g09d2 From 261842b7c9099f56de2eb969c8ad65402d68e00e Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 16 Apr 2009 21:41:52 -0400 Subject: tracing: add same level recursion detection The tracing infrastructure allows for recursion. That is, an interrupt may interrupt the act of tracing an event, and that interrupt may very well perform its own trace. This is a recursive trace, and is fine to do. The problem arises when there is a bug, and the utility doing the trace calls something that recurses back into the tracer. This recursion is not caused by an external event like an interrupt, but by code that is not expected to recurse. The result could be a lockup. This patch adds a bitmask to the task structure that keeps track of the trace recursion. To find the interrupt depth, the following algorithm is used: level = hardirq_count() + softirq_count() + in_nmi; Here, level will be the depth of interrutps and softirqs, and even handles the nmi. Then the corresponding bit is set in the recursion bitmask. If the bit was already set, we know we had a recursion at the same level and we warn about it and fail the writing to the buffer. After the data has been committed to the buffer, we clear the bit. No atomics are needed. The only races are with interrupts and they reset the bitmask before returning anywy. [ Impact: detect same irq level trace recursion ] Signed-off-by: Steven Rostedt --- include/linux/ftrace.h | 7 +++++++ include/linux/init_task.h | 1 + include/linux/sched.h | 4 +++- kernel/trace/ring_buffer.c | 42 ++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 53 insertions(+), 1 deletion(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h index 97c83e1bc58..39b95c56587 100644 --- a/include/linux/ftrace.h +++ b/include/linux/ftrace.h @@ -488,8 +488,15 @@ static inline int test_tsk_trace_graph(struct task_struct *tsk) extern int ftrace_dump_on_oops; +#ifdef CONFIG_PREEMPT +#define INIT_TRACE_RECURSION .trace_recursion = 0, +#endif + #endif /* CONFIG_TRACING */ +#ifndef INIT_TRACE_RECURSION +#define INIT_TRACE_RECURSION +#endif #ifdef CONFIG_HW_BRANCH_TRACER diff --git a/include/linux/init_task.h b/include/linux/init_task.h index dcfb93337e9..6fc21852986 100644 --- a/include/linux/init_task.h +++ b/include/linux/init_task.h @@ -187,6 +187,7 @@ extern struct cred init_cred; INIT_TRACE_IRQFLAGS \ INIT_LOCKDEP \ INIT_FTRACE_GRAPH \ + INIT_TRACE_RECURSION \ } diff --git a/include/linux/sched.h b/include/linux/sched.h index b4c38bc8049..7ede5e49091 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1428,7 +1428,9 @@ struct task_struct { #ifdef CONFIG_TRACING /* state flags for use by tracers */ unsigned long trace; -#endif + /* bitmask of trace recursion */ + unsigned long trace_recursion; +#endif /* CONFIG_TRACING */ }; /* Future-safe accessor for struct task_struct's cpus_allowed. */ diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c index 84a6055f37c..b421b0ea911 100644 --- a/kernel/trace/ring_buffer.c +++ b/kernel/trace/ring_buffer.c @@ -1481,6 +1481,40 @@ rb_reserve_next_event(struct ring_buffer_per_cpu *cpu_buffer, return event; } +static int trace_irq_level(void) +{ + return hardirq_count() + softirq_count() + in_nmi(); +} + +static int trace_recursive_lock(void) +{ + int level; + + level = trace_irq_level(); + + if (unlikely(current->trace_recursion & (1 << level))) { + /* Disable all tracing before we do anything else */ + tracing_off_permanent(); + WARN_ON_ONCE(1); + return -1; + } + + current->trace_recursion |= 1 << level; + + return 0; +} + +static void trace_recursive_unlock(void) +{ + int level; + + level = trace_irq_level(); + + WARN_ON_ONCE(!current->trace_recursion & (1 << level)); + + current->trace_recursion &= ~(1 << level); +} + static DEFINE_PER_CPU(int, rb_need_resched); /** @@ -1514,6 +1548,9 @@ ring_buffer_lock_reserve(struct ring_buffer *buffer, unsigned long length) /* If we are tracing schedule, we don't want to recurse */ resched = ftrace_preempt_disable(); + if (trace_recursive_lock()) + goto out_nocheck; + cpu = raw_smp_processor_id(); if (!cpumask_test_cpu(cpu, buffer->cpumask)) @@ -1543,6 +1580,9 @@ ring_buffer_lock_reserve(struct ring_buffer *buffer, unsigned long length) return event; out: + trace_recursive_unlock(); + + out_nocheck: ftrace_preempt_enable(resched); return NULL; } @@ -1581,6 +1621,8 @@ int ring_buffer_unlock_commit(struct ring_buffer *buffer, rb_commit(cpu_buffer, event); + trace_recursive_unlock(); + /* * Only the last preempt count needs to restore preemption. */ -- cgit v1.2.3-70-g09d2 From 3bcac0263f0b45e67a64034ebcb69eb9abb742f4 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 29 Apr 2009 13:45:05 +0100 Subject: SELinux: Don't flush inherited SIGKILL during execve() Don't flush inherited SIGKILL during execve() in SELinux's post cred commit hook. This isn't really a security problem: if the SIGKILL came before the credentials were changed, then we were right to receive it at the time, and should honour it; if it came after the creds were changed, then we definitely should honour it; and in any case, all that will happen is that the process will be scrapped before it ever returns to userspace. Signed-off-by: David Howells Signed-off-by: Oleg Nesterov Signed-off-by: James Morris --- include/linux/sched.h | 1 + kernel/signal.c | 11 ++++++++--- security/selinux/hooks.c | 9 +++++---- 3 files changed, 14 insertions(+), 7 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index 1d19c025f9d..d3b787c7aef 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1875,6 +1875,7 @@ extern void sched_dead(struct task_struct *p); extern void proc_caches_init(void); extern void flush_signals(struct task_struct *); +extern void __flush_signals(struct task_struct *); extern void ignore_signals(struct task_struct *); extern void flush_signal_handlers(struct task_struct *, int force_default); extern int dequeue_signal(struct task_struct *tsk, sigset_t *mask, siginfo_t *info); diff --git a/kernel/signal.c b/kernel/signal.c index 1c8814481a1..f93efec14ff 100644 --- a/kernel/signal.c +++ b/kernel/signal.c @@ -238,14 +238,19 @@ void flush_sigqueue(struct sigpending *queue) /* * Flush all pending signals for a task. */ +void __flush_signals(struct task_struct *t) +{ + clear_tsk_thread_flag(t, TIF_SIGPENDING); + flush_sigqueue(&t->pending); + flush_sigqueue(&t->signal->shared_pending); +} + void flush_signals(struct task_struct *t) { unsigned long flags; spin_lock_irqsave(&t->sighand->siglock, flags); - clear_tsk_thread_flag(t, TIF_SIGPENDING); - flush_sigqueue(&t->pending); - flush_sigqueue(&t->signal->shared_pending); + __flush_signals(t); spin_unlock_irqrestore(&t->sighand->siglock, flags); } diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c index dd19ba81201..5a345115036 100644 --- a/security/selinux/hooks.c +++ b/security/selinux/hooks.c @@ -2394,11 +2394,12 @@ static void selinux_bprm_committed_creds(struct linux_binprm *bprm) memset(&itimer, 0, sizeof itimer); for (i = 0; i < 3; i++) do_setitimer(i, &itimer, NULL); - flush_signals(current); spin_lock_irq(¤t->sighand->siglock); - flush_signal_handlers(current, 1); - sigemptyset(¤t->blocked); - recalc_sigpending(); + if (!(current->signal->flags & SIGNAL_GROUP_EXIT)) { + __flush_signals(current); + flush_signal_handlers(current, 1); + sigemptyset(¤t->blocked); + } spin_unlock_irq(¤t->sighand->siglock); } -- cgit v1.2.3-70-g09d2 From 5e751e992f3fb08ba35e1ca8095ec8fbf9eda523 Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 8 May 2009 13:55:22 +0100 Subject: CRED: Rename cred_exec_mutex to reflect that it's a guard against ptrace Rename cred_exec_mutex to reflect that it's a guard against foreign intervention on a process's credential state, such as is made by ptrace(). The attachment of a debugger to a process affects execve()'s calculation of the new credential state - _and_ also setprocattr()'s calculation of that state. Signed-off-by: David Howells Signed-off-by: James Morris --- fs/compat.c | 6 +++--- fs/exec.c | 10 +++++----- include/linux/init_task.h | 4 ++-- include/linux/sched.h | 4 +++- kernel/cred.c | 4 ++-- kernel/ptrace.c | 9 +++++---- 6 files changed, 20 insertions(+), 17 deletions(-) (limited to 'include/linux/sched.h') diff --git a/fs/compat.c b/fs/compat.c index 681ed81e6be..bb2a9b2e817 100644 --- a/fs/compat.c +++ b/fs/compat.c @@ -1488,7 +1488,7 @@ int compat_do_execve(char * filename, if (!bprm) goto out_files; - retval = mutex_lock_interruptible(¤t->cred_exec_mutex); + retval = mutex_lock_interruptible(¤t->cred_guard_mutex); if (retval < 0) goto out_free; current->in_execve = 1; @@ -1550,7 +1550,7 @@ int compat_do_execve(char * filename, /* execve succeeded */ current->fs->in_exec = 0; current->in_execve = 0; - mutex_unlock(¤t->cred_exec_mutex); + mutex_unlock(¤t->cred_guard_mutex); acct_update_integrals(current); free_bprm(bprm); if (displaced) @@ -1573,7 +1573,7 @@ out_unmark: out_unlock: current->in_execve = 0; - mutex_unlock(¤t->cred_exec_mutex); + mutex_unlock(¤t->cred_guard_mutex); out_free: free_bprm(bprm); diff --git a/fs/exec.c b/fs/exec.c index 639177b0eea..998e856c307 100644 --- a/fs/exec.c +++ b/fs/exec.c @@ -1045,7 +1045,7 @@ void install_exec_creds(struct linux_binprm *bprm) commit_creds(bprm->cred); bprm->cred = NULL; - /* cred_exec_mutex must be held at least to this point to prevent + /* cred_guard_mutex must be held at least to this point to prevent * ptrace_attach() from altering our determination of the task's * credentials; any time after this it may be unlocked */ @@ -1055,7 +1055,7 @@ EXPORT_SYMBOL(install_exec_creds); /* * determine how safe it is to execute the proposed program - * - the caller must hold current->cred_exec_mutex to protect against + * - the caller must hold current->cred_guard_mutex to protect against * PTRACE_ATTACH */ int check_unsafe_exec(struct linux_binprm *bprm) @@ -1297,7 +1297,7 @@ int do_execve(char * filename, if (!bprm) goto out_files; - retval = mutex_lock_interruptible(¤t->cred_exec_mutex); + retval = mutex_lock_interruptible(¤t->cred_guard_mutex); if (retval < 0) goto out_free; current->in_execve = 1; @@ -1360,7 +1360,7 @@ int do_execve(char * filename, /* execve succeeded */ current->fs->in_exec = 0; current->in_execve = 0; - mutex_unlock(¤t->cred_exec_mutex); + mutex_unlock(¤t->cred_guard_mutex); acct_update_integrals(current); free_bprm(bprm); if (displaced) @@ -1383,7 +1383,7 @@ out_unmark: out_unlock: current->in_execve = 0; - mutex_unlock(¤t->cred_exec_mutex); + mutex_unlock(¤t->cred_guard_mutex); out_free: free_bprm(bprm); diff --git a/include/linux/init_task.h b/include/linux/init_task.h index d87247d2641..7f54ba94242 100644 --- a/include/linux/init_task.h +++ b/include/linux/init_task.h @@ -145,8 +145,8 @@ extern struct cred init_cred; .group_leader = &tsk, \ .real_cred = &init_cred, \ .cred = &init_cred, \ - .cred_exec_mutex = \ - __MUTEX_INITIALIZER(tsk.cred_exec_mutex), \ + .cred_guard_mutex = \ + __MUTEX_INITIALIZER(tsk.cred_guard_mutex), \ .comm = "swapper", \ .thread = INIT_THREAD, \ .fs = &init_fs, \ diff --git a/include/linux/sched.h b/include/linux/sched.h index 3fa82b353c9..5932ace2240 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1247,7 +1247,9 @@ struct task_struct { * credentials (COW) */ const struct cred *cred; /* effective (overridable) subjective task * credentials (COW) */ - struct mutex cred_exec_mutex; /* execve vs ptrace cred calculation mutex */ + struct mutex cred_guard_mutex; /* guard against foreign influences on + * credential calculations + * (notably. ptrace) */ char comm[TASK_COMM_LEN]; /* executable name excluding path - access with [gs]et_task_comm (which lock diff --git a/kernel/cred.c b/kernel/cred.c index 3a039189d70..1bb4d7e5d61 100644 --- a/kernel/cred.c +++ b/kernel/cred.c @@ -167,7 +167,7 @@ EXPORT_SYMBOL(prepare_creds); /* * Prepare credentials for current to perform an execve() - * - The caller must hold current->cred_exec_mutex + * - The caller must hold current->cred_guard_mutex */ struct cred *prepare_exec_creds(void) { @@ -276,7 +276,7 @@ int copy_creds(struct task_struct *p, unsigned long clone_flags) struct cred *new; int ret; - mutex_init(&p->cred_exec_mutex); + mutex_init(&p->cred_guard_mutex); if ( #ifdef CONFIG_KEYS diff --git a/kernel/ptrace.c b/kernel/ptrace.c index 0692ab5a0d6..27ac80298bf 100644 --- a/kernel/ptrace.c +++ b/kernel/ptrace.c @@ -185,10 +185,11 @@ int ptrace_attach(struct task_struct *task) if (same_thread_group(task, current)) goto out; - /* Protect exec's credential calculations against our interference; - * SUID, SGID and LSM creds get determined differently under ptrace. + /* Protect the target's credential calculations against our + * interference; SUID, SGID and LSM creds get determined differently + * under ptrace. */ - retval = mutex_lock_interruptible(&task->cred_exec_mutex); + retval = mutex_lock_interruptible(&task->cred_guard_mutex); if (retval < 0) goto out; @@ -232,7 +233,7 @@ repeat: bad: write_unlock_irqrestore(&tasklist_lock, flags); task_unlock(task); - mutex_unlock(&task->cred_exec_mutex); + mutex_unlock(&task->cred_guard_mutex); out: return retval; } -- cgit v1.2.3-70-g09d2 From cd1bb94b4a0531e8211a3774f17de831f8285f76 Mon Sep 17 00:00:00 2001 From: Arun R Bharadwaj Date: Thu, 16 Apr 2009 12:15:34 +0530 Subject: timers: /proc/sys sysctl hook to enable timer migration * Arun R Bharadwaj [2009-04-16 12:11:36]: This patch creates the /proc/sys sysctl interface at /proc/sys/kernel/timer_migration Timer migration is enabled by default. To disable timer migration, when CONFIG_SCHED_DEBUG = y, echo 0 > /proc/sys/kernel/timer_migration Signed-off-by: Arun R Bharadwaj Signed-off-by: Thomas Gleixner --- include/linux/sched.h | 1 + kernel/sched.c | 2 ++ kernel/sysctl.c | 8 ++++++++ 3 files changed, 11 insertions(+) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index b4c38bc8049..61850401040 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1766,6 +1766,7 @@ extern unsigned int sysctl_sched_child_runs_first; extern unsigned int sysctl_sched_features; extern unsigned int sysctl_sched_migration_cost; extern unsigned int sysctl_sched_nr_migrate; +extern unsigned int sysctl_timer_migration; int sched_nr_latency_handler(struct ctl_table *table, int write, struct file *file, void __user *buffer, size_t *length, diff --git a/kernel/sched.c b/kernel/sched.c index 9c5b4d3f97a..7f1dd56af86 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -8731,6 +8731,8 @@ void __init sched_init_smp(void) } #endif /* CONFIG_SMP */ +const_debug unsigned int sysctl_timer_migration = 1; + int in_sched_functions(unsigned long addr) { return in_lock_functions(addr) || diff --git a/kernel/sysctl.c b/kernel/sysctl.c index e3d2c7dd59b..b3ce5813730 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -324,6 +324,14 @@ static struct ctl_table kern_table[] = { .mode = 0644, .proc_handler = &proc_dointvec, }, + { + .ctl_name = CTL_UNNUMBERED, + .procname = "timer_migration", + .data = &sysctl_timer_migration, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = &proc_dointvec, + }, #endif { .ctl_name = CTL_UNNUMBERED, -- cgit v1.2.3-70-g09d2 From eea08f32adb3f97553d49a4f79a119833036000a Mon Sep 17 00:00:00 2001 From: Arun R Bharadwaj Date: Thu, 16 Apr 2009 12:16:41 +0530 Subject: timers: Logic to move non pinned timers * Arun R Bharadwaj [2009-04-16 12:11:36]: This patch migrates all non pinned timers and hrtimers to the current idle load balancer, from all the idle CPUs. Timers firing on busy CPUs are not migrated. While migrating hrtimers, care should be taken to check if migrating a hrtimer would result in a latency or not. So we compare the expiry of the hrtimer with the next timer interrupt on the target cpu and migrate the hrtimer only if it expires *after* the next interrupt on the target cpu. So, added a clockevents_get_next_event() helper function to return the next_event on the target cpu's clock_event_device. [ tglx: cleanups and simplifications ] Signed-off-by: Arun R Bharadwaj Signed-off-by: Thomas Gleixner --- include/linux/clockchips.h | 9 ++++++++ include/linux/sched.h | 12 +++++++++++ kernel/hrtimer.c | 51 ++++++++++++++++++++++++++++++++++++++++++++-- kernel/sched.c | 5 +++++ kernel/time/clockevents.c | 12 +++++++++++ kernel/timer.c | 17 +++++++++++++--- 6 files changed, 101 insertions(+), 5 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/clockchips.h b/include/linux/clockchips.h index 3a1dbba4d3a..20a100fe2b4 100644 --- a/include/linux/clockchips.h +++ b/include/linux/clockchips.h @@ -143,3 +143,12 @@ extern void clockevents_notify(unsigned long reason, void *arg); #endif #endif + +#ifdef CONFIG_GENERIC_CLOCKEVENTS +extern ktime_t clockevents_get_next_event(int cpu); +#else +static inline ktime_t clockevents_get_next_event(int cpu) +{ + return (ktime_t) { .tv64 = KTIME_MAX }; +} +#endif diff --git a/include/linux/sched.h b/include/linux/sched.h index 61850401040..311dec12397 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -257,6 +257,7 @@ extern void task_rq_unlock_wait(struct task_struct *p); extern cpumask_var_t nohz_cpu_mask; #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ) extern int select_nohz_load_balancer(int cpu); +extern int get_nohz_load_balancer(void); #else static inline int select_nohz_load_balancer(int cpu) { @@ -1772,6 +1773,17 @@ int sched_nr_latency_handler(struct ctl_table *table, int write, struct file *file, void __user *buffer, size_t *length, loff_t *ppos); #endif +#ifdef CONFIG_SCHED_DEBUG +static inline unsigned int get_sysctl_timer_migration(void) +{ + return sysctl_timer_migration; +} +#else +static inline unsigned int get_sysctl_timer_migration(void) +{ + return 1; +} +#endif extern unsigned int sysctl_sched_rt_period; extern int sysctl_sched_rt_runtime; diff --git a/kernel/hrtimer.c b/kernel/hrtimer.c index c71bcd54924..b675a67c9ac 100644 --- a/kernel/hrtimer.c +++ b/kernel/hrtimer.c @@ -43,6 +43,8 @@ #include #include #include +#include +#include #include @@ -198,8 +200,19 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base, { struct hrtimer_clock_base *new_base; struct hrtimer_cpu_base *new_cpu_base; + int cpu, preferred_cpu = -1; + + cpu = smp_processor_id(); +#if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) + if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) { + preferred_cpu = get_nohz_load_balancer(); + if (preferred_cpu >= 0) + cpu = preferred_cpu; + } +#endif - new_cpu_base = &__get_cpu_var(hrtimer_bases); +again: + new_cpu_base = &per_cpu(hrtimer_bases, cpu); new_base = &new_cpu_base->clock_base[base->index]; if (base != new_base) { @@ -219,6 +232,40 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base, timer->base = NULL; spin_unlock(&base->cpu_base->lock); spin_lock(&new_base->cpu_base->lock); + + /* Optimized away for NOHZ=n SMP=n */ + if (cpu == preferred_cpu) { + /* Calculate clock monotonic expiry time */ +#ifdef CONFIG_HIGH_RES_TIMERS + ktime_t expires = ktime_sub(hrtimer_get_expires(timer), + new_base->offset); +#else + ktime_t expires = hrtimer_get_expires(timer); +#endif + + /* + * Get the next event on target cpu from the + * clock events layer. + * This covers the highres=off nohz=on case as well. + */ + ktime_t next = clockevents_get_next_event(cpu); + + ktime_t delta = ktime_sub(expires, next); + + /* + * We do not migrate the timer when it is expiring + * before the next event on the target cpu because + * we cannot reprogram the target cpu hardware and + * we would cause it to fire late. + */ + if (delta.tv64 < 0) { + cpu = smp_processor_id(); + spin_unlock(&new_base->cpu_base->lock); + spin_lock(&base->cpu_base->lock); + timer->base = base; + goto again; + } + } timer->base = new_base; } return new_base; @@ -236,7 +283,7 @@ lock_hrtimer_base(const struct hrtimer *timer, unsigned long *flags) return base; } -# define switch_hrtimer_base(t, b) (b) +# define switch_hrtimer_base(t, b, p) (b) #endif /* !CONFIG_SMP */ diff --git a/kernel/sched.c b/kernel/sched.c index 7f1dd56af86..9fe3774a0fd 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -4244,6 +4244,11 @@ static struct { .load_balancer = ATOMIC_INIT(-1), }; +int get_nohz_load_balancer(void) +{ + return atomic_read(&nohz.load_balancer); +} + /* * This routine will try to nominate the ilb (idle load balancing) * owner among the cpus whose ticks are stopped. ilb owner will do the idle diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c index d13be216a79..ab20ded013b 100644 --- a/kernel/time/clockevents.c +++ b/kernel/time/clockevents.c @@ -18,6 +18,7 @@ #include #include #include +#include /* The registered clock event devices */ static LIST_HEAD(clockevent_devices); @@ -251,4 +252,15 @@ void clockevents_notify(unsigned long reason, void *arg) spin_unlock(&clockevents_lock); } EXPORT_SYMBOL_GPL(clockevents_notify); + +ktime_t clockevents_get_next_event(int cpu) +{ + struct tick_device *td; + struct clock_event_device *dev; + + td = &per_cpu(tick_cpu_device, cpu); + dev = td->evtdev; + + return dev->next_event; +} #endif diff --git a/kernel/timer.c b/kernel/timer.c index 3424dfd11d5..3f841db5edf 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include @@ -609,9 +610,7 @@ __mod_timer(struct timer_list *timer, unsigned long expires, { struct tvec_base *base, *new_base; unsigned long flags; - int ret; - - ret = 0; + int ret = 0 , cpu; timer_stats_timer_set_start_info(timer); BUG_ON(!timer->function); @@ -630,6 +629,18 @@ __mod_timer(struct timer_list *timer, unsigned long expires, new_base = __get_cpu_var(tvec_bases); + cpu = smp_processor_id(); + +#if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP) + if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) { + int preferred_cpu = get_nohz_load_balancer(); + + if (preferred_cpu >= 0) + cpu = preferred_cpu; + } +#endif + new_base = per_cpu(tvec_bases, cpu); + if (base != new_base) { /* * We are trying to schedule the timer on the local CPU. -- cgit v1.2.3-70-g09d2 From 789f90fcf6b0b54e655740e9396c954378542c79 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 15 May 2009 15:19:27 +0200 Subject: perf_counter: per user mlock gift Instead of a per-process mlock gift for perf-counters, use a per-user gift so that there is less of a DoS potential. [ Impact: allow less worst-case unprivileged memory consumption ] Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Corey Ashford Cc: Arnaldo Carvalho de Melo LKML-Reference: <20090515132018.496182835@chello.nl> Signed-off-by: Ingo Molnar --- include/linux/sched.h | 4 ++++ kernel/perf_counter.c | 22 +++++++++++++++------- 2 files changed, 19 insertions(+), 7 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index d1857580a13..ff59d123151 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -674,6 +674,10 @@ struct user_struct { struct work_struct work; #endif #endif + +#ifdef CONFIG_PERF_COUNTERS + atomic_long_t locked_vm; +#endif }; extern int uids_sysfs_init(void); diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 0173738dd54..93f4a0e4b87 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -45,7 +45,7 @@ static atomic_t nr_munmap_tracking __read_mostly; static atomic_t nr_comm_tracking __read_mostly; int sysctl_perf_counter_priv __read_mostly; /* do we need to be privileged */ -int sysctl_perf_counter_mlock __read_mostly = 128; /* 'free' kb per counter */ +int sysctl_perf_counter_mlock __read_mostly = 512; /* 'free' kb per user */ /* * Lock for (sysadmin-configurable) counter reservations: @@ -1522,6 +1522,9 @@ static void perf_mmap_close(struct vm_area_struct *vma) if (atomic_dec_and_mutex_lock(&counter->mmap_count, &counter->mmap_mutex)) { + struct user_struct *user = current_user(); + + atomic_long_sub(counter->data->nr_pages + 1, &user->locked_vm); vma->vm_mm->locked_vm -= counter->data->nr_locked; perf_mmap_data_free(counter); mutex_unlock(&counter->mmap_mutex); @@ -1537,11 +1540,13 @@ static struct vm_operations_struct perf_mmap_vmops = { static int perf_mmap(struct file *file, struct vm_area_struct *vma) { struct perf_counter *counter = file->private_data; + struct user_struct *user = current_user(); unsigned long vma_size; unsigned long nr_pages; + unsigned long user_locked, user_lock_limit; unsigned long locked, lock_limit; + long user_extra, extra; int ret = 0; - long extra; if (!(vma->vm_flags & VM_SHARED) || (vma->vm_flags & VM_WRITE)) return -EINVAL; @@ -1569,15 +1574,17 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) goto unlock; } - extra = nr_pages /* + 1 only account the data pages */; - extra -= sysctl_perf_counter_mlock >> (PAGE_SHIFT - 10); - if (extra < 0) - extra = 0; + user_extra = nr_pages + 1; + user_lock_limit = sysctl_perf_counter_mlock >> (PAGE_SHIFT - 10); + user_locked = atomic_long_read(&user->locked_vm) + user_extra; - locked = vma->vm_mm->locked_vm + extra; + extra = 0; + if (user_locked > user_lock_limit) + extra = user_locked - user_lock_limit; lock_limit = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur; lock_limit >>= PAGE_SHIFT; + locked = vma->vm_mm->locked_vm + extra; if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) { ret = -EPERM; @@ -1590,6 +1597,7 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma) goto unlock; atomic_set(&counter->mmap_count, 1); + atomic_long_add(user_extra, &user->locked_vm); vma->vm_mm->locked_vm += extra; counter->data->nr_locked = extra; unlock: -- cgit v1.2.3-70-g09d2 From dce48a84adf1806676319f6f480e30a6daa012f9 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 11 Apr 2009 10:43:41 +0200 Subject: sched, timers: move calc_load() to scheduler Dimitri Sivanich noticed that xtime_lock is held write locked across calc_load() which iterates over all online CPUs. That can cause long latencies for xtime_lock readers on large SMP systems. The load average calculation is an rough estimate anyway so there is no real need to protect the readers vs. the update. It's not a problem when the avenrun array is updated while a reader copies the values. Instead of iterating over all online CPUs let the scheduler_tick code update the number of active tasks shortly before the avenrun update happens. The avenrun update itself is handled by the CPU which calls do_timer(). [ Impact: reduce xtime_lock write locked section ] Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra --- include/linux/sched.h | 2 +- kernel/sched.c | 84 +++++++++++++++++++++++++++++++++++++++++------ kernel/sched_idletask.c | 3 +- kernel/time/timekeeping.c | 2 +- kernel/timer.c | 54 ++---------------------------- 5 files changed, 80 insertions(+), 65 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index b4c38bc8049..6eb4892efe4 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -135,8 +135,8 @@ DECLARE_PER_CPU(unsigned long, process_counts); extern int nr_processes(void); extern unsigned long nr_running(void); extern unsigned long nr_uninterruptible(void); -extern unsigned long nr_active(void); extern unsigned long nr_iowait(void); +extern void calc_global_load(void); extern unsigned long get_parent_ip(unsigned long addr); diff --git a/kernel/sched.c b/kernel/sched.c index 8908d190a34..f4eb88153bd 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -630,6 +630,10 @@ struct rq { struct list_head migration_queue; #endif + /* calc_load related fields */ + unsigned long calc_load_update; + long calc_load_active; + #ifdef CONFIG_SCHED_HRTICK #ifdef CONFIG_SMP int hrtick_csd_pending; @@ -1728,6 +1732,8 @@ static void cfs_rq_set_shares(struct cfs_rq *cfs_rq, unsigned long shares) } #endif +static void calc_load_account_active(struct rq *this_rq); + #include "sched_stats.h" #include "sched_idletask.c" #include "sched_fair.c" @@ -2856,19 +2862,57 @@ unsigned long nr_iowait(void) return sum; } -unsigned long nr_active(void) +/* Variables and functions for calc_load */ +static atomic_long_t calc_load_tasks; +static unsigned long calc_load_update; +unsigned long avenrun[3]; +EXPORT_SYMBOL(avenrun); + +static unsigned long +calc_load(unsigned long load, unsigned long exp, unsigned long active) { - unsigned long i, running = 0, uninterruptible = 0; + load *= exp; + load += active * (FIXED_1 - exp); + return load >> FSHIFT; +} - for_each_online_cpu(i) { - running += cpu_rq(i)->nr_running; - uninterruptible += cpu_rq(i)->nr_uninterruptible; - } +/* + * calc_load - update the avenrun load estimates 10 ticks after the + * CPUs have updated calc_load_tasks. + */ +void calc_global_load(void) +{ + unsigned long upd = calc_load_update + 10; + long active; + + if (time_before(jiffies, upd)) + return; - if (unlikely((long)uninterruptible < 0)) - uninterruptible = 0; + active = atomic_long_read(&calc_load_tasks); + active = active > 0 ? active * FIXED_1 : 0; - return running + uninterruptible; + avenrun[0] = calc_load(avenrun[0], EXP_1, active); + avenrun[1] = calc_load(avenrun[1], EXP_5, active); + avenrun[2] = calc_load(avenrun[2], EXP_15, active); + + calc_load_update += LOAD_FREQ; +} + +/* + * Either called from update_cpu_load() or from a cpu going idle + */ +static void calc_load_account_active(struct rq *this_rq) +{ + long nr_active, delta; + + nr_active = this_rq->nr_running; + nr_active += (long) this_rq->nr_uninterruptible; + + if (nr_active != this_rq->calc_load_active) { + delta = nr_active - this_rq->calc_load_active; + this_rq->calc_load_active = nr_active; + atomic_long_add(delta, &calc_load_tasks); + } } /* @@ -2899,6 +2943,11 @@ static void update_cpu_load(struct rq *this_rq) new_load += scale-1; this_rq->cpu_load[i] = (old_load*(scale-1) + new_load) >> i; } + + if (time_after_eq(jiffies, this_rq->calc_load_update)) { + this_rq->calc_load_update += LOAD_FREQ; + calc_load_account_active(this_rq); + } } #ifdef CONFIG_SMP @@ -7091,6 +7140,14 @@ static void migrate_dead_tasks(unsigned int dead_cpu) } } + +/* + * remove the tasks which were accounted by rq from calc_load_tasks. + */ +static void calc_global_load_remove(struct rq *rq) +{ + atomic_long_sub(rq->calc_load_active, &calc_load_tasks); +} #endif /* CONFIG_HOTPLUG_CPU */ #if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL) @@ -7325,6 +7382,8 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) /* Update our root-domain */ rq = cpu_rq(cpu); spin_lock_irqsave(&rq->lock, flags); + rq->calc_load_update = calc_load_update; + rq->calc_load_active = 0; if (rq->rd) { BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span)); @@ -7364,7 +7423,7 @@ migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu) cpuset_unlock(); migrate_nr_uninterruptible(rq); BUG_ON(rq->nr_running != 0); - + calc_global_load_remove(rq); /* * No need to migrate the tasks: it was best-effort if * they didn't take sched_hotcpu_mutex. Just wake up @@ -9059,6 +9118,8 @@ void __init sched_init(void) rq = cpu_rq(i); spin_lock_init(&rq->lock); rq->nr_running = 0; + rq->calc_load_active = 0; + rq->calc_load_update = jiffies + LOAD_FREQ; init_cfs_rq(&rq->cfs, rq); init_rt_rq(&rq->rt, rq); #ifdef CONFIG_FAIR_GROUP_SCHED @@ -9166,6 +9227,9 @@ void __init sched_init(void) * when this runqueue becomes "idle". */ init_idle(current, smp_processor_id()); + + calc_load_update = jiffies + LOAD_FREQ; + /* * During early bootup we pretend to be a normal task: */ diff --git a/kernel/sched_idletask.c b/kernel/sched_idletask.c index 8a21a2e28c1..499672c10cb 100644 --- a/kernel/sched_idletask.c +++ b/kernel/sched_idletask.c @@ -22,7 +22,8 @@ static void check_preempt_curr_idle(struct rq *rq, struct task_struct *p, int sy static struct task_struct *pick_next_task_idle(struct rq *rq) { schedstat_inc(rq, sched_goidle); - + /* adjust the active tasks as we might go into a long sleep */ + calc_load_account_active(rq); return rq->idle; } diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c index 687dff49f6e..52a8bf8931f 100644 --- a/kernel/time/timekeeping.c +++ b/kernel/time/timekeeping.c @@ -22,7 +22,7 @@ /* * This read-write spinlock protects us from races in SMP while - * playing with xtime and avenrun. + * playing with xtime. */ __cacheline_aligned_in_smp DEFINE_SEQLOCK(xtime_lock); diff --git a/kernel/timer.c b/kernel/timer.c index cffffad01c3..6a21d7af962 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -1122,47 +1122,6 @@ void update_process_times(int user_tick) run_posix_cpu_timers(p); } -/* - * Nr of active tasks - counted in fixed-point numbers - */ -static unsigned long count_active_tasks(void) -{ - return nr_active() * FIXED_1; -} - -/* - * Hmm.. Changed this, as the GNU make sources (load.c) seems to - * imply that avenrun[] is the standard name for this kind of thing. - * Nothing else seems to be standardized: the fractional size etc - * all seem to differ on different machines. - * - * Requires xtime_lock to access. - */ -unsigned long avenrun[3]; - -EXPORT_SYMBOL(avenrun); - -/* - * calc_load - given tick count, update the avenrun load estimates. - * This is called while holding a write_lock on xtime_lock. - */ -static inline void calc_load(unsigned long ticks) -{ - unsigned long active_tasks; /* fixed-point */ - static int count = LOAD_FREQ; - - count -= ticks; - if (unlikely(count < 0)) { - active_tasks = count_active_tasks(); - do { - CALC_LOAD(avenrun[0], EXP_1, active_tasks); - CALC_LOAD(avenrun[1], EXP_5, active_tasks); - CALC_LOAD(avenrun[2], EXP_15, active_tasks); - count += LOAD_FREQ; - } while (count < 0); - } -} - /* * This function runs timers and the timer-tq in bottom half context. */ @@ -1186,16 +1145,6 @@ void run_local_timers(void) softlockup_tick(); } -/* - * Called by the timer interrupt. xtime_lock must already be taken - * by the timer IRQ! - */ -static inline void update_times(unsigned long ticks) -{ - update_wall_time(); - calc_load(ticks); -} - /* * The 64-bit jiffies value is not atomic - you MUST NOT read it * without sampling the sequence number in xtime_lock. @@ -1205,7 +1154,8 @@ static inline void update_times(unsigned long ticks) void do_timer(unsigned long ticks) { jiffies_64 += ticks; - update_times(ticks); + update_wall_time(); + calc_global_load(); } #ifdef __ARCH_WANT_SYS_ALARM -- cgit v1.2.3-70-g09d2 From 2d02494f5a90f2e4b3c4c6acc85ec94674cdc431 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Sat, 2 May 2009 20:08:52 +0200 Subject: sched, timers: cleanup avenrun users avenrun is an rough estimate so we don't have to worry about consistency of the three avenrun values. Remove the xtime lock dependency and provide a function to scale the values. Cleanup the users. [ Impact: cleanup ] Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra --- fs/proc/loadavg.c | 18 ++++++------------ include/linux/sched.h | 1 + kernel/sched.c | 15 +++++++++++++++ kernel/timer.c | 32 ++++++-------------------------- 4 files changed, 28 insertions(+), 38 deletions(-) (limited to 'include/linux/sched.h') diff --git a/fs/proc/loadavg.c b/fs/proc/loadavg.c index 9bca39cf99e..1afa4dd4cae 100644 --- a/fs/proc/loadavg.c +++ b/fs/proc/loadavg.c @@ -12,20 +12,14 @@ static int loadavg_proc_show(struct seq_file *m, void *v) { - int a, b, c; - unsigned long seq; + unsigned long avnrun[3]; - do { - seq = read_seqbegin(&xtime_lock); - a = avenrun[0] + (FIXED_1/200); - b = avenrun[1] + (FIXED_1/200); - c = avenrun[2] + (FIXED_1/200); - } while (read_seqretry(&xtime_lock, seq)); + get_avenrun(avnrun, FIXED_1/200, 0); - seq_printf(m, "%d.%02d %d.%02d %d.%02d %ld/%d %d\n", - LOAD_INT(a), LOAD_FRAC(a), - LOAD_INT(b), LOAD_FRAC(b), - LOAD_INT(c), LOAD_FRAC(c), + seq_printf(m, "%lu.%02lu %lu.%02lu %lu.%02lu %ld/%d %d\n", + LOAD_INT(avnrun[0]), LOAD_FRAC(avnrun[0]), + LOAD_INT(avnrun[1]), LOAD_FRAC(avnrun[1]), + LOAD_INT(avnrun[2]), LOAD_FRAC(avnrun[2]), nr_running(), nr_threads, task_active_pid_ns(current)->last_pid); return 0; diff --git a/include/linux/sched.h b/include/linux/sched.h index 6eb4892efe4..de7b3b21777 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -116,6 +116,7 @@ struct fs_struct; * 11 bit fractions. */ extern unsigned long avenrun[]; /* Load averages */ +extern void get_avenrun(unsigned long *loads, unsigned long offset, int shift); #define FSHIFT 11 /* nr of bits of precision */ #define FIXED_1 (1<= 0) { - tp.tv_nsec = tp.tv_nsec - NSEC_PER_SEC; - tp.tv_sec++; - } - info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); + ktime_get_ts(&tp); + monotonic_to_bootbased(&tp); + info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0); - info->loads[0] = avenrun[0] << (SI_LOAD_SHIFT - FSHIFT); - info->loads[1] = avenrun[1] << (SI_LOAD_SHIFT - FSHIFT); - info->loads[2] = avenrun[2] << (SI_LOAD_SHIFT - FSHIFT); + get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT); - info->procs = nr_threads; - } while (read_seqretry(&xtime_lock, seq)); + info->procs = nr_threads; si_meminfo(info); si_swapinfo(info); -- cgit v1.2.3-70-g09d2 From 690cc3ffe33ac4a2857583c22d4c6244ae11684d Mon Sep 17 00:00:00 2001 From: "Eric W. Biederman" Date: Wed, 13 May 2009 16:55:10 +0000 Subject: syscall: Implement a convinience function restart_syscall Currently when we have a signal pending we have the functionality to restart that the current system call. There are other cases such as nasty lock ordering issues where it makes sense to have a simple fix that uses try lock and restarts the system call. Buying time to figure out how to rework the locking strategy. Signed-off-by: Eric W. Biederman Signed-off-by: David S. Miller --- include/linux/sched.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index b4c38bc8049..d853f6bb0ba 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -2178,6 +2178,12 @@ static inline int test_tsk_need_resched(struct task_struct *tsk) return unlikely(test_tsk_thread_flag(tsk,TIF_NEED_RESCHED)); } +static inline int restart_syscall(void) +{ + set_tsk_thread_flag(current, TIF_SIGPENDING); + return -ERESTARTNOINTR; +} + static inline int signal_pending(struct task_struct *p) { return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING)); -- cgit v1.2.3-70-g09d2 From 4200efd9acda4accf24640f1e77d24fdcdb524df Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 19 May 2009 09:22:19 +0200 Subject: sched: properly define the sched_group::cpumask and sched_domain::span fields Properly document the variable-size structure tricks we are doing wrt. struct sched_group and sched_domain, and use the field[0] GCC extension instead of defining a vla array. Dont use unions for this, as pointed out by Linus. [ Impact: cleanup, un-confuse Sparse and LLVM ] Reported-by: Jeff Garzik Acked-by: Linus Torvalds LKML-Reference: Signed-off-by: Ingo Molnar --- include/linux/sched.h | 25 ++++++++++++++++++++++--- kernel/sched.c | 5 +++-- 2 files changed, 25 insertions(+), 5 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index de7b3b21777..dbb1043e865 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -839,7 +839,17 @@ struct sched_group { */ u32 reciprocal_cpu_power; - unsigned long cpumask[]; + /* + * The CPUs this group covers. + * + * NOTE: this field is variable length. (Allocated dynamically + * by attaching extra space to the end of the structure, + * depending on how many CPUs the kernel has booted up with) + * + * It is also be embedded into static data structures at build + * time. (See 'struct static_sched_group' in kernel/sched.c) + */ + unsigned long cpumask[0]; }; static inline struct cpumask *sched_group_cpus(struct sched_group *sg) @@ -925,8 +935,17 @@ struct sched_domain { char *name; #endif - /* span of all CPUs in this domain */ - unsigned long span[]; + /* + * Span of all CPUs in this domain. + * + * NOTE: this field is variable length. (Allocated dynamically + * by attaching extra space to the end of the structure, + * depending on how many CPUs the kernel has booted up with) + * + * It is also be embedded into static data structures at build + * time. (See 'struct static_sched_domain' in kernel/sched.c) + */ + unsigned long span[0]; }; static inline struct cpumask *sched_domain_span(struct sched_domain *sd) diff --git a/kernel/sched.c b/kernel/sched.c index 497c09ba61e..228acae8821 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -7948,8 +7948,9 @@ int sched_smt_power_savings = 0, sched_mc_power_savings = 0; /* * The cpus mask in sched_group and sched_domain hangs off the end. - * FIXME: use cpumask_var_t or dynamic percpu alloc to avoid wasting space - * for nr_cpu_ids < CONFIG_NR_CPUS. + * + * ( See the the comments in include/linux/sched.h:struct sched_group + * and struct sched_domain. ) */ struct static_sched_group { struct sched_group sg; -- cgit v1.2.3-70-g09d2 From a63eaf34ae60bdb067a354cc8def2e8f4a01f5f4 Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Fri, 22 May 2009 14:17:31 +1000 Subject: perf_counter: Dynamically allocate tasks' perf_counter_context struct This replaces the struct perf_counter_context in the task_struct with a pointer to a dynamically allocated perf_counter_context struct. The main reason for doing is this is to allow us to transfer a perf_counter_context from one task to another when we do lazy PMU switching in a later patch. This has a few side-benefits: the task_struct becomes a little smaller, we save some memory because only tasks that have perf_counters attached get a perf_counter_context allocated for them, and we can remove the inclusion of in sched.h, meaning that we don't end up recompiling nearly everything whenever perf_counter.h changes. The perf_counter_context structures are reference-counted and freed when the last reference is dropped. A context can have references from its task and the counters on its task. Counters can outlive the task so it is possible that a context will be freed well after its task has exited. Contexts are allocated on fork if the parent had a context, or otherwise the first time that a per-task counter is created on a task. In the latter case, we set the context pointer in the task struct locklessly using an atomic compare-and-exchange operation in case we raced with some other task in creating a context for the subject task. This also removes the task pointer from the perf_counter struct. The task pointer was not used anywhere and would make it harder to move a context from one task to another. Anything that needed to know which task a counter was attached to was already using counter->ctx->task. The __perf_counter_init_context function moves up in perf_counter.c so that it can be called from find_get_context, and now initializes the refcount, but is otherwise unchanged. We were potentially calling list_del_counter twice: once from __perf_counter_exit_task when the task exits and once from __perf_counter_remove_from_context when the counter's fd gets closed. This adds a check in list_del_counter so it doesn't do anything if the counter has already been removed from the lists. Since perf_counter_task_sched_in doesn't do anything if the task doesn't have a context, and leaves cpuctx->task_ctx = NULL, this adds code to __perf_install_in_context to set cpuctx->task_ctx if necessary, i.e. in the case where the current task adds the first counter to itself and thus creates a context for itself. This also adds similar code to __perf_counter_enable to handle a similar situation which can arise when the counters have been disabled using prctl; that also leaves cpuctx->task_ctx = NULL. [ Impact: refactor counter context management to prepare for new feature ] Signed-off-by: Paul Mackerras Acked-by: Peter Zijlstra Cc: Corey Ashford Cc: Marcelo Tosatti Cc: Arnaldo Carvalho de Melo LKML-Reference: <18966.10075.781053.231153@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic/apic.c | 1 + include/linux/init_task.h | 13 --- include/linux/perf_counter.h | 4 +- include/linux/sched.h | 6 +- kernel/exit.c | 3 +- kernel/fork.c | 1 + kernel/perf_counter.c | 218 +++++++++++++++++++++++++++---------------- 7 files changed, 145 insertions(+), 101 deletions(-) (limited to 'include/linux/sched.h') diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c index e9021a90802..b4f64402a82 100644 --- a/arch/x86/kernel/apic/apic.c +++ b/arch/x86/kernel/apic/apic.c @@ -14,6 +14,7 @@ * Mikael Pettersson : PM converted to driver model. */ +#include #include #include #include diff --git a/include/linux/init_task.h b/include/linux/init_task.h index 503afaa0afa..d87247d2641 100644 --- a/include/linux/init_task.h +++ b/include/linux/init_task.h @@ -108,18 +108,6 @@ extern struct group_info init_groups; extern struct cred init_cred; -#ifdef CONFIG_PERF_COUNTERS -# define INIT_PERF_COUNTERS(tsk) \ - .perf_counter_ctx.counter_list = \ - LIST_HEAD_INIT(tsk.perf_counter_ctx.counter_list), \ - .perf_counter_ctx.event_list = \ - LIST_HEAD_INIT(tsk.perf_counter_ctx.event_list), \ - .perf_counter_ctx.lock = \ - __SPIN_LOCK_UNLOCKED(tsk.perf_counter_ctx.lock), -#else -# define INIT_PERF_COUNTERS(tsk) -#endif - /* * INIT_TASK is used to set up the first task table, touch at * your own risk!. Base=0, limit=0x1fffff (=2MB) @@ -183,7 +171,6 @@ extern struct cred init_cred; }, \ .dirties = INIT_PROP_LOCAL_SINGLE(dirties), \ INIT_IDS \ - INIT_PERF_COUNTERS(tsk) \ INIT_TRACE_IRQFLAGS \ INIT_LOCKDEP \ INIT_FTRACE_GRAPH \ diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index f612941ef46..07130900546 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -449,7 +449,6 @@ struct perf_counter { struct hw_perf_counter hw; struct perf_counter_context *ctx; - struct task_struct *task; struct file *filp; struct perf_counter *parent; @@ -498,7 +497,6 @@ struct perf_counter { * Used as a container for task counters and CPU counters as well: */ struct perf_counter_context { -#ifdef CONFIG_PERF_COUNTERS /* * Protect the states of the counters in the list, * nr_active, and the list: @@ -516,6 +514,7 @@ struct perf_counter_context { int nr_counters; int nr_active; int is_active; + atomic_t refcount; struct task_struct *task; /* @@ -523,7 +522,6 @@ struct perf_counter_context { */ u64 time; u64 timestamp; -#endif }; /** diff --git a/include/linux/sched.h b/include/linux/sched.h index ff59d123151..9714d450f41 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -71,7 +71,6 @@ struct sched_param { #include #include #include -#include #include #include #include @@ -99,6 +98,7 @@ struct robust_list_head; struct bio; struct bts_tracer; struct fs_struct; +struct perf_counter_context; /* * List of flags we want to share for kernel threads, @@ -1387,7 +1387,9 @@ struct task_struct { struct list_head pi_state_list; struct futex_pi_state *pi_state_cache; #endif - struct perf_counter_context perf_counter_ctx; +#ifdef CONFIG_PERF_COUNTERS + struct perf_counter_context *perf_counter_ctxp; +#endif #ifdef CONFIG_NUMA struct mempolicy *mempolicy; short il_next; diff --git a/kernel/exit.c b/kernel/exit.c index f9dfedd94af..99ad4063ee4 100644 --- a/kernel/exit.c +++ b/kernel/exit.c @@ -48,6 +48,7 @@ #include #include #include +#include #include #include @@ -159,7 +160,7 @@ static void delayed_put_task_struct(struct rcu_head *rhp) struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); #ifdef CONFIG_PERF_COUNTERS - WARN_ON_ONCE(!list_empty(&tsk->perf_counter_ctx.counter_list)); + WARN_ON_ONCE(tsk->perf_counter_ctxp); #endif trace_sched_process_free(tsk); put_task_struct(tsk); diff --git a/kernel/fork.c b/kernel/fork.c index d32fef4d38e..e72a09f5355 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -63,6 +63,7 @@ #include #include #include +#include #include #include diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 08584c16049..06ea3eae886 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -97,6 +97,17 @@ void perf_enable(void) hw_perf_enable(); } +static void get_ctx(struct perf_counter_context *ctx) +{ + atomic_inc(&ctx->refcount); +} + +static void put_ctx(struct perf_counter_context *ctx) +{ + if (atomic_dec_and_test(&ctx->refcount)) + kfree(ctx); +} + static void list_add_counter(struct perf_counter *counter, struct perf_counter_context *ctx) { @@ -118,11 +129,17 @@ list_add_counter(struct perf_counter *counter, struct perf_counter_context *ctx) ctx->nr_counters++; } +/* + * Remove a counter from the lists for its context. + * Must be called with counter->mutex and ctx->mutex held. + */ static void list_del_counter(struct perf_counter *counter, struct perf_counter_context *ctx) { struct perf_counter *sibling, *tmp; + if (list_empty(&counter->list_entry)) + return; ctx->nr_counters--; list_del_init(&counter->list_entry); @@ -216,8 +233,6 @@ static void __perf_counter_remove_from_context(void *info) counter_sched_out(counter, cpuctx, ctx); - counter->task = NULL; - list_del_counter(counter, ctx); if (!ctx->task) { @@ -279,7 +294,6 @@ retry: */ if (!list_empty(&counter->list_entry)) { list_del_counter(counter, ctx); - counter->task = NULL; } spin_unlock_irq(&ctx->lock); } @@ -568,11 +582,17 @@ static void __perf_install_in_context(void *info) * If this is a task context, we need to check whether it is * the current task context of this cpu. If not it has been * scheduled out before the smp call arrived. + * Or possibly this is the right context but it isn't + * on this cpu because it had no counters. */ - if (ctx->task && cpuctx->task_ctx != ctx) - return; + if (ctx->task && cpuctx->task_ctx != ctx) { + if (cpuctx->task_ctx || ctx->task != current) + return; + cpuctx->task_ctx = ctx; + } spin_lock_irqsave(&ctx->lock, flags); + ctx->is_active = 1; update_context_time(ctx); /* @@ -653,7 +673,6 @@ perf_install_in_context(struct perf_counter_context *ctx, return; } - counter->task = task; retry: task_oncpu_function_call(task, __perf_install_in_context, counter); @@ -693,10 +712,14 @@ static void __perf_counter_enable(void *info) * If this is a per-task counter, need to check whether this * counter's task is the current task on this cpu. */ - if (ctx->task && cpuctx->task_ctx != ctx) - return; + if (ctx->task && cpuctx->task_ctx != ctx) { + if (cpuctx->task_ctx || ctx->task != current) + return; + cpuctx->task_ctx = ctx; + } spin_lock_irqsave(&ctx->lock, flags); + ctx->is_active = 1; update_context_time(ctx); counter->prev_state = counter->state; @@ -852,10 +875,10 @@ void __perf_counter_sched_out(struct perf_counter_context *ctx, void perf_counter_task_sched_out(struct task_struct *task, int cpu) { struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu); - struct perf_counter_context *ctx = &task->perf_counter_ctx; + struct perf_counter_context *ctx = task->perf_counter_ctxp; struct pt_regs *regs; - if (likely(!cpuctx->task_ctx)) + if (likely(!ctx || !cpuctx->task_ctx)) return; update_context_time(ctx); @@ -871,6 +894,8 @@ static void __perf_counter_task_sched_out(struct perf_counter_context *ctx) { struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context); + if (!cpuctx->task_ctx) + return; __perf_counter_sched_out(ctx, cpuctx); cpuctx->task_ctx = NULL; } @@ -969,8 +994,10 @@ __perf_counter_sched_in(struct perf_counter_context *ctx, void perf_counter_task_sched_in(struct task_struct *task, int cpu) { struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu); - struct perf_counter_context *ctx = &task->perf_counter_ctx; + struct perf_counter_context *ctx = task->perf_counter_ctxp; + if (likely(!ctx)) + return; __perf_counter_sched_in(ctx, cpuctx, cpu); cpuctx->task_ctx = ctx; } @@ -985,11 +1012,11 @@ static void perf_counter_cpu_sched_in(struct perf_cpu_context *cpuctx, int cpu) int perf_counter_task_disable(void) { struct task_struct *curr = current; - struct perf_counter_context *ctx = &curr->perf_counter_ctx; + struct perf_counter_context *ctx = curr->perf_counter_ctxp; struct perf_counter *counter; unsigned long flags; - if (likely(!ctx->nr_counters)) + if (!ctx || !ctx->nr_counters) return 0; local_irq_save(flags); @@ -1020,12 +1047,12 @@ int perf_counter_task_disable(void) int perf_counter_task_enable(void) { struct task_struct *curr = current; - struct perf_counter_context *ctx = &curr->perf_counter_ctx; + struct perf_counter_context *ctx = curr->perf_counter_ctxp; struct perf_counter *counter; unsigned long flags; int cpu; - if (likely(!ctx->nr_counters)) + if (!ctx || !ctx->nr_counters) return 0; local_irq_save(flags); @@ -1128,19 +1155,23 @@ void perf_counter_task_tick(struct task_struct *curr, int cpu) return; cpuctx = &per_cpu(perf_cpu_context, cpu); - ctx = &curr->perf_counter_ctx; + ctx = curr->perf_counter_ctxp; perf_adjust_freq(&cpuctx->ctx); - perf_adjust_freq(ctx); + if (ctx) + perf_adjust_freq(ctx); perf_counter_cpu_sched_out(cpuctx); - __perf_counter_task_sched_out(ctx); + if (ctx) + __perf_counter_task_sched_out(ctx); rotate_ctx(&cpuctx->ctx); - rotate_ctx(ctx); + if (ctx) + rotate_ctx(ctx); perf_counter_cpu_sched_in(cpuctx, cpu); - perf_counter_task_sched_in(curr, cpu); + if (ctx) + perf_counter_task_sched_in(curr, cpu); } /* @@ -1176,6 +1207,22 @@ static u64 perf_counter_read(struct perf_counter *counter) return atomic64_read(&counter->count); } +/* + * Initialize the perf_counter context in a task_struct: + */ +static void +__perf_counter_init_context(struct perf_counter_context *ctx, + struct task_struct *task) +{ + memset(ctx, 0, sizeof(*ctx)); + spin_lock_init(&ctx->lock); + mutex_init(&ctx->mutex); + INIT_LIST_HEAD(&ctx->counter_list); + INIT_LIST_HEAD(&ctx->event_list); + atomic_set(&ctx->refcount, 1); + ctx->task = task; +} + static void put_context(struct perf_counter_context *ctx) { if (ctx->task) @@ -1186,6 +1233,7 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu) { struct perf_cpu_context *cpuctx; struct perf_counter_context *ctx; + struct perf_counter_context *tctx; struct task_struct *task; /* @@ -1225,15 +1273,36 @@ static struct perf_counter_context *find_get_context(pid_t pid, int cpu) if (!task) return ERR_PTR(-ESRCH); - ctx = &task->perf_counter_ctx; - ctx->task = task; - /* Reuse ptrace permission checks for now. */ if (!ptrace_may_access(task, PTRACE_MODE_READ)) { - put_context(ctx); + put_task_struct(task); return ERR_PTR(-EACCES); } + ctx = task->perf_counter_ctxp; + if (!ctx) { + ctx = kmalloc(sizeof(struct perf_counter_context), GFP_KERNEL); + if (!ctx) { + put_task_struct(task); + return ERR_PTR(-ENOMEM); + } + __perf_counter_init_context(ctx, task); + /* + * Make sure other cpus see correct values for *ctx + * once task->perf_counter_ctxp is visible to them. + */ + smp_wmb(); + tctx = cmpxchg(&task->perf_counter_ctxp, NULL, ctx); + if (tctx) { + /* + * We raced with some other task; use + * the context they set. + */ + kfree(ctx); + ctx = tctx; + } + } + return ctx; } @@ -1242,6 +1311,7 @@ static void free_counter_rcu(struct rcu_head *head) struct perf_counter *counter; counter = container_of(head, struct perf_counter, rcu_head); + put_ctx(counter->ctx); kfree(counter); } @@ -2247,7 +2317,7 @@ static void perf_counter_comm_event(struct perf_comm_event *comm_event) perf_counter_comm_ctx(&cpuctx->ctx, comm_event); put_cpu_var(perf_cpu_context); - perf_counter_comm_ctx(¤t->perf_counter_ctx, comm_event); + perf_counter_comm_ctx(current->perf_counter_ctxp, comm_event); } void perf_counter_comm(struct task_struct *task) @@ -2256,7 +2326,9 @@ void perf_counter_comm(struct task_struct *task) if (!atomic_read(&nr_comm_tracking)) return; - + if (!current->perf_counter_ctxp) + return; + comm_event = (struct perf_comm_event){ .task = task, .event = { @@ -2372,7 +2444,7 @@ got_name: perf_counter_mmap_ctx(&cpuctx->ctx, mmap_event); put_cpu_var(perf_cpu_context); - perf_counter_mmap_ctx(¤t->perf_counter_ctx, mmap_event); + perf_counter_mmap_ctx(current->perf_counter_ctxp, mmap_event); kfree(buf); } @@ -2384,6 +2456,8 @@ void perf_counter_mmap(unsigned long addr, unsigned long len, if (!atomic_read(&nr_mmap_tracking)) return; + if (!current->perf_counter_ctxp) + return; mmap_event = (struct perf_mmap_event){ .file = file, @@ -2985,6 +3059,7 @@ perf_counter_alloc(struct perf_counter_hw_event *hw_event, counter->group_leader = group_leader; counter->pmu = NULL; counter->ctx = ctx; + get_ctx(ctx); counter->state = PERF_COUNTER_STATE_INACTIVE; if (hw_event->disabled) @@ -3149,21 +3224,6 @@ err_put_context: goto out_fput; } -/* - * Initialize the perf_counter context in a task_struct: - */ -static void -__perf_counter_init_context(struct perf_counter_context *ctx, - struct task_struct *task) -{ - memset(ctx, 0, sizeof(*ctx)); - spin_lock_init(&ctx->lock); - mutex_init(&ctx->mutex); - INIT_LIST_HEAD(&ctx->counter_list); - INIT_LIST_HEAD(&ctx->event_list); - ctx->task = task; -} - /* * inherit a counter from parent task to child task: */ @@ -3195,7 +3255,6 @@ inherit_counter(struct perf_counter *parent_counter, /* * Link it up in the child's context: */ - child_counter->task = child; add_counter_to_ctx(child_counter, child_ctx); child_counter->parent = parent_counter; @@ -3294,40 +3353,15 @@ __perf_counter_exit_task(struct task_struct *child, struct perf_counter *parent_counter; /* - * If we do not self-reap then we have to wait for the - * child task to unschedule (it will happen for sure), - * so that its counter is at its final count. (This - * condition triggers rarely - child tasks usually get - * off their CPU before the parent has a chance to - * get this far into the reaping action) + * Protect against concurrent operations on child_counter + * due its fd getting closed, etc. */ - if (child != current) { - wait_task_inactive(child, 0); - update_counter_times(child_counter); - list_del_counter(child_counter, child_ctx); - } else { - struct perf_cpu_context *cpuctx; - unsigned long flags; - - /* - * Disable and unlink this counter. - * - * Be careful about zapping the list - IRQ/NMI context - * could still be processing it: - */ - local_irq_save(flags); - perf_disable(); - - cpuctx = &__get_cpu_var(perf_cpu_context); + mutex_lock(&child_counter->mutex); - group_sched_out(child_counter, cpuctx, child_ctx); - update_counter_times(child_counter); + update_counter_times(child_counter); + list_del_counter(child_counter, child_ctx); - list_del_counter(child_counter, child_ctx); - - perf_enable(); - local_irq_restore(flags); - } + mutex_unlock(&child_counter->mutex); parent_counter = child_counter->parent; /* @@ -3346,19 +3380,29 @@ __perf_counter_exit_task(struct task_struct *child, * * Note: we may be running in child context, but the PID is not hashed * anymore so new counters will not be added. + * (XXX not sure that is true when we get called from flush_old_exec. + * -- paulus) */ void perf_counter_exit_task(struct task_struct *child) { struct perf_counter *child_counter, *tmp; struct perf_counter_context *child_ctx; + unsigned long flags; WARN_ON_ONCE(child != current); - child_ctx = &child->perf_counter_ctx; + child_ctx = child->perf_counter_ctxp; - if (likely(!child_ctx->nr_counters)) + if (likely(!child_ctx)) return; + local_irq_save(flags); + __perf_counter_task_sched_out(child_ctx); + child->perf_counter_ctxp = NULL; + local_irq_restore(flags); + + mutex_lock(&child_ctx->mutex); + again: list_for_each_entry_safe(child_counter, tmp, &child_ctx->counter_list, list_entry) @@ -3371,6 +3415,10 @@ again: */ if (!list_empty(&child_ctx->counter_list)) goto again; + + mutex_unlock(&child_ctx->mutex); + + put_ctx(child_ctx); } /* @@ -3382,19 +3430,25 @@ void perf_counter_init_task(struct task_struct *child) struct perf_counter *counter; struct task_struct *parent = current; - child_ctx = &child->perf_counter_ctx; - parent_ctx = &parent->perf_counter_ctx; - - __perf_counter_init_context(child_ctx, child); + child->perf_counter_ctxp = NULL; /* * This is executed from the parent task context, so inherit - * counters that have been marked for cloning: + * counters that have been marked for cloning. + * First allocate and initialize a context for the child. */ - if (likely(!parent_ctx->nr_counters)) + child_ctx = kmalloc(sizeof(struct perf_counter_context), GFP_KERNEL); + if (!child_ctx) + return; + + parent_ctx = parent->perf_counter_ctxp; + if (likely(!parent_ctx || !parent_ctx->nr_counters)) return; + __perf_counter_init_context(child_ctx, child); + child->perf_counter_ctxp = child_ctx; + /* * Lock the parent list. No need to lock the child - not PID * hashed yet and not running, so nobody can access it. -- cgit v1.2.3-70-g09d2 From 082ff5a2767a0679ee543f14883adbafb631ffbe Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Sat, 23 May 2009 18:29:00 +0200 Subject: perf_counter: Change pctrl() behaviour Instead of en/dis-abling all counters acting on a particular task, en/dis- able all counters we created. [ v2: fix crash on first counter enable ] Signed-off-by: Peter Zijlstra Cc: Paul Mackerras Cc: Corey Ashford Cc: Arnaldo Carvalho de Melo Cc: John Kacur LKML-Reference: <20090523163012.916937244@chello.nl> Signed-off-by: Ingo Molnar --- include/linux/init_task.h | 10 +++++ include/linux/perf_counter.h | 3 ++ include/linux/sched.h | 2 + kernel/perf_counter.c | 87 ++++++++++++-------------------------------- 4 files changed, 39 insertions(+), 63 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/init_task.h b/include/linux/init_task.h index d87247d2641..353c0ac7723 100644 --- a/include/linux/init_task.h +++ b/include/linux/init_task.h @@ -108,6 +108,15 @@ extern struct group_info init_groups; extern struct cred init_cred; +#ifdef CONFIG_PERF_COUNTERS +# define INIT_PERF_COUNTERS(tsk) \ + .perf_counter_mutex = \ + __MUTEX_INITIALIZER(tsk.perf_counter_mutex), \ + .perf_counter_list = LIST_HEAD_INIT(tsk.perf_counter_list), +#else +# define INIT_PERF_COUNTERS(tsk) +#endif + /* * INIT_TASK is used to set up the first task table, touch at * your own risk!. Base=0, limit=0x1fffff (=2MB) @@ -171,6 +180,7 @@ extern struct cred init_cred; }, \ .dirties = INIT_PROP_LOCAL_SINGLE(dirties), \ INIT_IDS \ + INIT_PERF_COUNTERS(tsk) \ INIT_TRACE_IRQFLAGS \ INIT_LOCKDEP \ INIT_FTRACE_GRAPH \ diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index 4ab8050eb9e..4159ee5940f 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -469,6 +469,9 @@ struct perf_counter { int oncpu; int cpu; + struct list_head owner_entry; + struct task_struct *owner; + /* mmap bits */ struct mutex mmap_mutex; atomic_t mmap_count; diff --git a/include/linux/sched.h b/include/linux/sched.h index 9714d450f41..bc9326dcdde 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1389,6 +1389,8 @@ struct task_struct { #endif #ifdef CONFIG_PERF_COUNTERS struct perf_counter_context *perf_counter_ctxp; + struct mutex perf_counter_mutex; + struct list_head perf_counter_list; #endif #ifdef CONFIG_NUMA struct mempolicy *mempolicy; diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 0e97f896133..4c86a636976 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -1076,79 +1076,26 @@ static void perf_counter_cpu_sched_in(struct perf_cpu_context *cpuctx, int cpu) __perf_counter_sched_in(ctx, cpuctx, cpu); } -int perf_counter_task_disable(void) +int perf_counter_task_enable(void) { - struct task_struct *curr = current; - struct perf_counter_context *ctx = curr->perf_counter_ctxp; struct perf_counter *counter; - unsigned long flags; - - if (!ctx || !ctx->nr_counters) - return 0; - - local_irq_save(flags); - __perf_counter_task_sched_out(ctx); - - spin_lock(&ctx->lock); - - /* - * Disable all the counters: - */ - perf_disable(); - - list_for_each_entry(counter, &ctx->counter_list, list_entry) { - if (counter->state != PERF_COUNTER_STATE_ERROR) { - update_group_times(counter); - counter->state = PERF_COUNTER_STATE_OFF; - } - } - - perf_enable(); - - spin_unlock_irqrestore(&ctx->lock, flags); + mutex_lock(¤t->perf_counter_mutex); + list_for_each_entry(counter, ¤t->perf_counter_list, owner_entry) + perf_counter_enable(counter); + mutex_unlock(¤t->perf_counter_mutex); return 0; } -int perf_counter_task_enable(void) +int perf_counter_task_disable(void) { - struct task_struct *curr = current; - struct perf_counter_context *ctx = curr->perf_counter_ctxp; struct perf_counter *counter; - unsigned long flags; - int cpu; - - if (!ctx || !ctx->nr_counters) - return 0; - - local_irq_save(flags); - cpu = smp_processor_id(); - - __perf_counter_task_sched_out(ctx); - - spin_lock(&ctx->lock); - /* - * Disable all the counters: - */ - perf_disable(); - - list_for_each_entry(counter, &ctx->counter_list, list_entry) { - if (counter->state > PERF_COUNTER_STATE_OFF) - continue; - counter->state = PERF_COUNTER_STATE_INACTIVE; - counter->tstamp_enabled = - ctx->time - counter->total_time_enabled; - counter->hw_event.disabled = 0; - } - perf_enable(); - - spin_unlock(&ctx->lock); - - perf_counter_task_sched_in(curr, cpu); - - local_irq_restore(flags); + mutex_lock(¤t->perf_counter_mutex); + list_for_each_entry(counter, ¤t->perf_counter_list, owner_entry) + perf_counter_disable(counter); + mutex_unlock(¤t->perf_counter_mutex); return 0; } @@ -1416,6 +1363,11 @@ static int perf_release(struct inode *inode, struct file *file) perf_counter_remove_from_context(counter); mutex_unlock(&ctx->mutex); + mutex_lock(&counter->owner->perf_counter_mutex); + list_del_init(&counter->owner_entry); + mutex_unlock(&counter->owner->perf_counter_mutex); + put_task_struct(counter->owner); + free_counter(counter); put_context(ctx); @@ -3272,6 +3224,12 @@ SYSCALL_DEFINE5(perf_counter_open, perf_install_in_context(ctx, counter, cpu); mutex_unlock(&ctx->mutex); + counter->owner = current; + get_task_struct(current); + mutex_lock(¤t->perf_counter_mutex); + list_add_tail(&counter->owner_entry, ¤t->perf_counter_list); + mutex_unlock(¤t->perf_counter_mutex); + fput_light(counter_file, fput_needed2); out_fput: @@ -3488,6 +3446,9 @@ void perf_counter_init_task(struct task_struct *child) child->perf_counter_ctxp = NULL; + mutex_init(&child->perf_counter_mutex); + INIT_LIST_HEAD(&child->perf_counter_list); + /* * This is executed from the parent task context, so inherit * counters that have been marked for cloning. -- cgit v1.2.3-70-g09d2 From 3959214f971417f4162926ac52ad4cd042958caa Mon Sep 17 00:00:00 2001 From: Kay Sievers Date: Tue, 24 Mar 2009 15:43:30 +0100 Subject: sched: delayed cleanup of user_struct During bootup performance tracing we see repeated occurrences of /sys/kernel/uid/* events for the same uid, leading to a, in this case, rather pointless userspace processing for the same uid over and over. This is usually caused by tools which change their uid to "nobody", to run without privileges to read data supplied by untrusted users. This change delays the execution of the (already existing) scheduled work, to cleanup the uid after one second, so the allocated and announced uid can possibly be re-used by another process. This is the current behavior, where almost every invocation of a binary, which changes the uid, creates two events: $ read START < /sys/kernel/uevent_seqnum; \ for i in `seq 100`; do su --shell=/bin/true bin; done; \ read END < /sys/kernel/uevent_seqnum; \ echo $(($END - $START)) 178 With the delayed cleanup, we get only two events, and userspace finishes a bit faster too: $ read START < /sys/kernel/uevent_seqnum; \ for i in `seq 100`; do su --shell=/bin/true bin; done; \ read END < /sys/kernel/uevent_seqnum; \ echo $(($END - $START)) 1 Acked-by: Dhaval Giani Signed-off-by: Kay Sievers Signed-off-by: Greg Kroah-Hartman --- include/linux/sched.h | 2 +- kernel/user.c | 67 ++++++++++++++++++++++++++++++--------------------- 2 files changed, 40 insertions(+), 29 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index c900aa53007..7531b1c2820 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -674,7 +674,7 @@ struct user_struct { struct task_group *tg; #ifdef CONFIG_SYSFS struct kobject kobj; - struct work_struct work; + struct delayed_work work; #endif #endif diff --git a/kernel/user.c b/kernel/user.c index 850e0ba41c1..2c000e7132a 100644 --- a/kernel/user.c +++ b/kernel/user.c @@ -75,21 +75,6 @@ static void uid_hash_remove(struct user_struct *up) put_user_ns(up->user_ns); } -static struct user_struct *uid_hash_find(uid_t uid, struct hlist_head *hashent) -{ - struct user_struct *user; - struct hlist_node *h; - - hlist_for_each_entry(user, h, hashent, uidhash_node) { - if (user->uid == uid) { - atomic_inc(&user->__count); - return user; - } - } - - return NULL; -} - #ifdef CONFIG_USER_SCHED static void sched_destroy_user(struct user_struct *up) @@ -119,6 +104,23 @@ static int sched_create_user(struct user_struct *up) { return 0; } #if defined(CONFIG_USER_SCHED) && defined(CONFIG_SYSFS) +static struct user_struct *uid_hash_find(uid_t uid, struct hlist_head *hashent) +{ + struct user_struct *user; + struct hlist_node *h; + + hlist_for_each_entry(user, h, hashent, uidhash_node) { + if (user->uid == uid) { + /* possibly resurrect an "almost deleted" object */ + if (atomic_inc_return(&user->__count) == 1) + cancel_delayed_work(&user->work); + return user; + } + } + + return NULL; +} + static struct kset *uids_kset; /* represents the /sys/kernel/uids/ directory */ static DEFINE_MUTEX(uids_mutex); @@ -283,12 +285,12 @@ int __init uids_sysfs_init(void) return uids_user_create(&root_user); } -/* work function to remove sysfs directory for a user and free up +/* delayed work function to remove sysfs directory for a user and free up * corresponding structures. */ static void cleanup_user_struct(struct work_struct *w) { - struct user_struct *up = container_of(w, struct user_struct, work); + struct user_struct *up = container_of(w, struct user_struct, work.work); unsigned long flags; int remove_user = 0; @@ -297,15 +299,12 @@ static void cleanup_user_struct(struct work_struct *w) */ uids_mutex_lock(); - local_irq_save(flags); - - if (atomic_dec_and_lock(&up->__count, &uidhash_lock)) { + spin_lock_irqsave(&uidhash_lock, flags); + if (atomic_read(&up->__count) == 0) { uid_hash_remove(up); remove_user = 1; - spin_unlock_irqrestore(&uidhash_lock, flags); - } else { - local_irq_restore(flags); } + spin_unlock_irqrestore(&uidhash_lock, flags); if (!remove_user) goto done; @@ -331,16 +330,28 @@ done: */ static void free_user(struct user_struct *up, unsigned long flags) { - /* restore back the count */ - atomic_inc(&up->__count); spin_unlock_irqrestore(&uidhash_lock, flags); - - INIT_WORK(&up->work, cleanup_user_struct); - schedule_work(&up->work); + INIT_DELAYED_WORK(&up->work, cleanup_user_struct); + schedule_delayed_work(&up->work, msecs_to_jiffies(1000)); } #else /* CONFIG_USER_SCHED && CONFIG_SYSFS */ +static struct user_struct *uid_hash_find(uid_t uid, struct hlist_head *hashent) +{ + struct user_struct *user; + struct hlist_node *h; + + hlist_for_each_entry(user, h, hashent, uidhash_node) { + if (user->uid == uid) { + atomic_inc(&user->__count); + return user; + } + } + + return NULL; +} + int uids_sysfs_init(void) { return 0; } static inline int uids_user_create(struct user_struct *up) { return 0; } static inline void uids_mutex_lock(void) { } -- cgit v1.2.3-70-g09d2 From 58568d2a8215cb6f55caf2332017d7bdff954e1c Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 16 Jun 2009 15:31:49 -0700 Subject: cpuset,mm: update tasks' mems_allowed in time Fix allocating page cache/slab object on the unallowed node when memory spread is set by updating tasks' mems_allowed after its cpuset's mems is changed. In order to update tasks' mems_allowed in time, we must modify the code of memory policy. Because the memory policy is applied in the process's context originally. After applying this patch, one task directly manipulates anothers mems_allowed, and we use alloc_lock in the task_struct to protect mems_allowed and memory policy of the task. But in the fast path, we didn't use lock to protect them, because adding a lock may lead to performance regression. But if we don't add a lock,the task might see no nodes when changing cpuset's mems_allowed to some non-overlapping set. In order to avoid it, we set all new allowed nodes, then clear newly disallowed ones. [lee.schermerhorn@hp.com: The rework of mpol_new() to extract the adjusting of the node mask to apply cpuset and mpol flags "context" breaks set_mempolicy() and mbind() with MPOL_PREFERRED and a NULL nodemask--i.e., explicit local allocation. Fix this by adding the check for MPOL_PREFERRED and empty node mask to mpol_new_mpolicy(). Remove the now unneeded 'nodes = NULL' from mpol_new(). Note that mpol_new_mempolicy() is always called with a non-NULL 'nodes' parameter now that it has been removed from mpol_new(). Therefore, we don't need to test nodes for NULL before testing it for 'empty'. However, just to be extra paranoid, add a VM_BUG_ON() to verify this assumption.] [lee.schermerhorn@hp.com: I don't think the function name 'mpol_new_mempolicy' is descriptive enough to differentiate it from mpol_new(). This function applies cpuset set context, usually constraining nodes to those allowed by the cpuset. However, when the 'RELATIVE_NODES flag is set, it also translates the nodes. So I settled on 'mpol_set_nodemask()', because the comment block for mpol_new() mentions that we need to call this function to "set nodes". Some additional minor line length, whitespace and typo cleanup.] Signed-off-by: Miao Xie Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Christoph Lameter Cc: Paul Menage Cc: Nick Piggin Cc: Yasunori Goto Cc: Pekka Enberg Cc: David Rientjes Signed-off-by: Lee Schermerhorn Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/cpuset.h | 13 ++-- include/linux/sched.h | 8 +-- init/main.c | 6 +- kernel/cpuset.c | 184 +++++++++++++------------------------------------ kernel/kthread.c | 2 + mm/mempolicy.c | 143 +++++++++++++++++++++++++++----------- mm/page_alloc.c | 5 +- 7 files changed, 170 insertions(+), 191 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/cpuset.h b/include/linux/cpuset.h index 05ea1dd7d68..a5740fc4d04 100644 --- a/include/linux/cpuset.h +++ b/include/linux/cpuset.h @@ -18,7 +18,6 @@ extern int number_of_cpusets; /* How many cpusets are defined in system? */ -extern int cpuset_init_early(void); extern int cpuset_init(void); extern void cpuset_init_smp(void); extern void cpuset_cpus_allowed(struct task_struct *p, struct cpumask *mask); @@ -27,7 +26,6 @@ extern void cpuset_cpus_allowed_locked(struct task_struct *p, extern nodemask_t cpuset_mems_allowed(struct task_struct *p); #define cpuset_current_mems_allowed (current->mems_allowed) void cpuset_init_current_mems_allowed(void); -void cpuset_update_task_memory_state(void); int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask); extern int __cpuset_node_allowed_softwall(int node, gfp_t gfp_mask); @@ -92,9 +90,13 @@ extern void rebuild_sched_domains(void); extern void cpuset_print_task_mems_allowed(struct task_struct *p); +static inline void set_mems_allowed(nodemask_t nodemask) +{ + current->mems_allowed = nodemask; +} + #else /* !CONFIG_CPUSETS */ -static inline int cpuset_init_early(void) { return 0; } static inline int cpuset_init(void) { return 0; } static inline void cpuset_init_smp(void) {} @@ -116,7 +118,6 @@ static inline nodemask_t cpuset_mems_allowed(struct task_struct *p) #define cpuset_current_mems_allowed (node_states[N_HIGH_MEMORY]) static inline void cpuset_init_current_mems_allowed(void) {} -static inline void cpuset_update_task_memory_state(void) {} static inline int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask) { @@ -188,6 +189,10 @@ static inline void cpuset_print_task_mems_allowed(struct task_struct *p) { } +static inline void set_mems_allowed(nodemask_t nodemask) +{ +} + #endif /* !CONFIG_CPUSETS */ #endif /* _LINUX_CPUSET_H */ diff --git a/include/linux/sched.h b/include/linux/sched.h index c900aa53007..1048bf50540 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1318,7 +1318,8 @@ struct task_struct { /* Thread group tracking */ u32 parent_exec_id; u32 self_exec_id; -/* Protection of (de-)allocation: mm, files, fs, tty, keyrings */ +/* Protection of (de-)allocation: mm, files, fs, tty, keyrings, mems_allowed, + * mempolicy */ spinlock_t alloc_lock; #ifdef CONFIG_GENERIC_HARDIRQS @@ -1386,8 +1387,7 @@ struct task_struct { cputime_t acct_timexpd; /* stime + utime since last update */ #endif #ifdef CONFIG_CPUSETS - nodemask_t mems_allowed; - int cpuset_mems_generation; + nodemask_t mems_allowed; /* Protected by alloc_lock */ int cpuset_mem_spread_rotor; #endif #ifdef CONFIG_CGROUPS @@ -1410,7 +1410,7 @@ struct task_struct { struct list_head perf_counter_list; #endif #ifdef CONFIG_NUMA - struct mempolicy *mempolicy; + struct mempolicy *mempolicy; /* Protected by alloc_lock */ short il_next; #endif atomic_t fs_excl; /* holding fs exclusive resources */ diff --git a/init/main.c b/init/main.c index f6204f712e7..5e0d3f047ea 100644 --- a/init/main.c +++ b/init/main.c @@ -670,7 +670,6 @@ asmlinkage void __init start_kernel(void) initrd_start = 0; } #endif - cpuset_init_early(); page_cgroup_init(); enable_debug_pagealloc(); cpu_hotplug_init(); @@ -867,6 +866,11 @@ static noinline int init_post(void) static int __init kernel_init(void * unused) { lock_kernel(); + + /* + * init can allocate pages on any node + */ + set_mems_allowed(node_possible_map); /* * init can run on any cpu. */ diff --git a/kernel/cpuset.c b/kernel/cpuset.c index af5a83d5218..7e75a41bd50 100644 --- a/kernel/cpuset.c +++ b/kernel/cpuset.c @@ -97,12 +97,6 @@ struct cpuset { struct cpuset *parent; /* my parent */ - /* - * Copy of global cpuset_mems_generation as of the most - * recent time this cpuset changed its mems_allowed. - */ - int mems_generation; - struct fmeter fmeter; /* memory_pressure filter */ /* partition number for rebuild_sched_domains() */ @@ -176,27 +170,6 @@ static inline int is_spread_slab(const struct cpuset *cs) return test_bit(CS_SPREAD_SLAB, &cs->flags); } -/* - * Increment this integer everytime any cpuset changes its - * mems_allowed value. Users of cpusets can track this generation - * number, and avoid having to lock and reload mems_allowed unless - * the cpuset they're using changes generation. - * - * A single, global generation is needed because cpuset_attach_task() could - * reattach a task to a different cpuset, which must not have its - * generation numbers aliased with those of that tasks previous cpuset. - * - * Generations are needed for mems_allowed because one task cannot - * modify another's memory placement. So we must enable every task, - * on every visit to __alloc_pages(), to efficiently check whether - * its current->cpuset->mems_allowed has changed, requiring an update - * of its current->mems_allowed. - * - * Since writes to cpuset_mems_generation are guarded by the cgroup lock - * there is no need to mark it atomic. - */ -static int cpuset_mems_generation; - static struct cpuset top_cpuset = { .flags = ((1 << CS_CPU_EXCLUSIVE) | (1 << CS_MEM_EXCLUSIVE)), }; @@ -228,8 +201,9 @@ static struct cpuset top_cpuset = { * If a task is only holding callback_mutex, then it has read-only * access to cpusets. * - * The task_struct fields mems_allowed and mems_generation may only - * be accessed in the context of that task, so require no locks. + * Now, the task_struct fields mems_allowed and mempolicy may be changed + * by other task, we use alloc_lock in the task_struct fields to protect + * them. * * The cpuset_common_file_read() handlers only hold callback_mutex across * small pieces of code, such as when reading out possibly multi-word @@ -349,69 +323,6 @@ static void cpuset_update_task_spread_flag(struct cpuset *cs, tsk->flags &= ~PF_SPREAD_SLAB; } -/** - * cpuset_update_task_memory_state - update task memory placement - * - * If the current tasks cpusets mems_allowed changed behind our - * backs, update current->mems_allowed, mems_generation and task NUMA - * mempolicy to the new value. - * - * Task mempolicy is updated by rebinding it relative to the - * current->cpuset if a task has its memory placement changed. - * Do not call this routine if in_interrupt(). - * - * Call without callback_mutex or task_lock() held. May be - * called with or without cgroup_mutex held. Thanks in part to - * 'the_top_cpuset_hack', the task's cpuset pointer will never - * be NULL. This routine also might acquire callback_mutex during - * call. - * - * Reading current->cpuset->mems_generation doesn't need task_lock - * to guard the current->cpuset derefence, because it is guarded - * from concurrent freeing of current->cpuset using RCU. - * - * The rcu_dereference() is technically probably not needed, - * as I don't actually mind if I see a new cpuset pointer but - * an old value of mems_generation. However this really only - * matters on alpha systems using cpusets heavily. If I dropped - * that rcu_dereference(), it would save them a memory barrier. - * For all other arch's, rcu_dereference is a no-op anyway, and for - * alpha systems not using cpusets, another planned optimization, - * avoiding the rcu critical section for tasks in the root cpuset - * which is statically allocated, so can't vanish, will make this - * irrelevant. Better to use RCU as intended, than to engage in - * some cute trick to save a memory barrier that is impossible to - * test, for alpha systems using cpusets heavily, which might not - * even exist. - * - * This routine is needed to update the per-task mems_allowed data, - * within the tasks context, when it is trying to allocate memory - * (in various mm/mempolicy.c routines) and notices that some other - * task has been modifying its cpuset. - */ - -void cpuset_update_task_memory_state(void) -{ - int my_cpusets_mem_gen; - struct task_struct *tsk = current; - struct cpuset *cs; - - rcu_read_lock(); - my_cpusets_mem_gen = task_cs(tsk)->mems_generation; - rcu_read_unlock(); - - if (my_cpusets_mem_gen != tsk->cpuset_mems_generation) { - mutex_lock(&callback_mutex); - task_lock(tsk); - cs = task_cs(tsk); /* Maybe changed when task not locked */ - guarantee_online_mems(cs, &tsk->mems_allowed); - tsk->cpuset_mems_generation = cs->mems_generation; - task_unlock(tsk); - mutex_unlock(&callback_mutex); - mpol_rebind_task(tsk, &tsk->mems_allowed); - } -} - /* * is_cpuset_subset(p, q) - Is cpuset p a subset of cpuset q? * @@ -1017,14 +928,6 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs, * other task, the task_struct mems_allowed that we are hacking * is for our current task, which must allocate new pages for that * migrating memory region. - * - * We call cpuset_update_task_memory_state() before hacking - * our tasks mems_allowed, so that we are assured of being in - * sync with our tasks cpuset, and in particular, callbacks to - * cpuset_update_task_memory_state() from nested page allocations - * won't see any mismatch of our cpuset and task mems_generation - * values, so won't overwrite our hacked tasks mems_allowed - * nodemask. */ static void cpuset_migrate_mm(struct mm_struct *mm, const nodemask_t *from, @@ -1032,22 +935,37 @@ static void cpuset_migrate_mm(struct mm_struct *mm, const nodemask_t *from, { struct task_struct *tsk = current; - cpuset_update_task_memory_state(); - - mutex_lock(&callback_mutex); tsk->mems_allowed = *to; - mutex_unlock(&callback_mutex); do_migrate_pages(mm, from, to, MPOL_MF_MOVE_ALL); - mutex_lock(&callback_mutex); guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed); - mutex_unlock(&callback_mutex); } /* - * Rebind task's vmas to cpuset's new mems_allowed, and migrate pages to new - * nodes if memory_migrate flag is set. Called with cgroup_mutex held. + * cpuset_change_task_nodemask - change task's mems_allowed and mempolicy + * @tsk: the task to change + * @newmems: new nodes that the task will be set + * + * In order to avoid seeing no nodes if the old and new nodes are disjoint, + * we structure updates as setting all new allowed nodes, then clearing newly + * disallowed ones. + * + * Called with task's alloc_lock held + */ +static void cpuset_change_task_nodemask(struct task_struct *tsk, + nodemask_t *newmems) +{ + nodes_or(tsk->mems_allowed, tsk->mems_allowed, *newmems); + mpol_rebind_task(tsk, &tsk->mems_allowed); + mpol_rebind_task(tsk, newmems); + tsk->mems_allowed = *newmems; +} + +/* + * Update task's mems_allowed and rebind its mempolicy and vmas' mempolicy + * of it to cpuset's new mems_allowed, and migrate pages to new nodes if + * memory_migrate flag is set. Called with cgroup_mutex held. */ static void cpuset_change_nodemask(struct task_struct *p, struct cgroup_scanner *scan) @@ -1056,12 +974,19 @@ static void cpuset_change_nodemask(struct task_struct *p, struct cpuset *cs; int migrate; const nodemask_t *oldmem = scan->data; + nodemask_t newmems; + + cs = cgroup_cs(scan->cg); + guarantee_online_mems(cs, &newmems); + + task_lock(p); + cpuset_change_task_nodemask(p, &newmems); + task_unlock(p); mm = get_task_mm(p); if (!mm) return; - cs = cgroup_cs(scan->cg); migrate = is_memory_migrate(cs); mpol_rebind_mm(mm, &cs->mems_allowed); @@ -1114,10 +1039,10 @@ static void update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem, /* * Handle user request to change the 'mems' memory placement * of a cpuset. Needs to validate the request, update the - * cpusets mems_allowed and mems_generation, and for each - * task in the cpuset, rebind any vma mempolicies and if - * the cpuset is marked 'memory_migrate', migrate the tasks - * pages to the new memory. + * cpusets mems_allowed, and for each task in the cpuset, + * update mems_allowed and rebind task's mempolicy and any vma + * mempolicies and if the cpuset is marked 'memory_migrate', + * migrate the tasks pages to the new memory. * * Call with cgroup_mutex held. May take callback_mutex during call. * Will take tasklist_lock, scan tasklist for tasks in cpuset cs, @@ -1170,7 +1095,6 @@ static int update_nodemask(struct cpuset *cs, struct cpuset *trialcs, mutex_lock(&callback_mutex); cs->mems_allowed = trialcs->mems_allowed; - cs->mems_generation = cpuset_mems_generation++; mutex_unlock(&callback_mutex); update_tasks_nodemask(cs, &oldmem, &heap); @@ -1434,15 +1358,18 @@ static void cpuset_attach(struct cgroup_subsys *ss, if (cs == &top_cpuset) { cpumask_copy(cpus_attach, cpu_possible_mask); + to = node_possible_map; } else { - mutex_lock(&callback_mutex); guarantee_online_cpus(cs, cpus_attach); - mutex_unlock(&callback_mutex); + guarantee_online_mems(cs, &to); } err = set_cpus_allowed_ptr(tsk, cpus_attach); if (err) return; + task_lock(tsk); + cpuset_change_task_nodemask(tsk, &to); + task_unlock(tsk); cpuset_update_task_spread_flag(cs, tsk); from = oldcs->mems_allowed; @@ -1848,8 +1775,6 @@ static struct cgroup_subsys_state *cpuset_create( struct cpuset *parent; if (!cont->parent) { - /* This is early initialization for the top cgroup */ - top_cpuset.mems_generation = cpuset_mems_generation++; return &top_cpuset.css; } parent = cgroup_cs(cont->parent); @@ -1861,7 +1786,6 @@ static struct cgroup_subsys_state *cpuset_create( return ERR_PTR(-ENOMEM); } - cpuset_update_task_memory_state(); cs->flags = 0; if (is_spread_page(parent)) set_bit(CS_SPREAD_PAGE, &cs->flags); @@ -1870,7 +1794,6 @@ static struct cgroup_subsys_state *cpuset_create( set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags); cpumask_clear(cs->cpus_allowed); nodes_clear(cs->mems_allowed); - cs->mems_generation = cpuset_mems_generation++; fmeter_init(&cs->fmeter); cs->relax_domain_level = -1; @@ -1889,8 +1812,6 @@ static void cpuset_destroy(struct cgroup_subsys *ss, struct cgroup *cont) { struct cpuset *cs = cgroup_cs(cont); - cpuset_update_task_memory_state(); - if (is_sched_load_balance(cs)) update_flag(CS_SCHED_LOAD_BALANCE, cs, 0); @@ -1911,21 +1832,6 @@ struct cgroup_subsys cpuset_subsys = { .early_init = 1, }; -/* - * cpuset_init_early - just enough so that the calls to - * cpuset_update_task_memory_state() in early init code - * are harmless. - */ - -int __init cpuset_init_early(void) -{ - alloc_cpumask_var(&top_cpuset.cpus_allowed, GFP_NOWAIT); - - top_cpuset.mems_generation = cpuset_mems_generation++; - return 0; -} - - /** * cpuset_init - initialize cpusets at system boot * @@ -1936,11 +1842,13 @@ int __init cpuset_init(void) { int err = 0; + if (!alloc_cpumask_var(&top_cpuset.cpus_allowed, GFP_KERNEL)) + BUG(); + cpumask_setall(top_cpuset.cpus_allowed); nodes_setall(top_cpuset.mems_allowed); fmeter_init(&top_cpuset.fmeter); - top_cpuset.mems_generation = cpuset_mems_generation++; set_bit(CS_SCHED_LOAD_BALANCE, &top_cpuset.flags); top_cpuset.relax_domain_level = -1; diff --git a/kernel/kthread.c b/kernel/kthread.c index 41c88fe4050..7fa44133352 100644 --- a/kernel/kthread.c +++ b/kernel/kthread.c @@ -9,6 +9,7 @@ #include #include #include +#include #include #include #include @@ -236,6 +237,7 @@ int kthreadd(void *unused) ignore_signals(tsk); set_user_nice(tsk, KTHREAD_NICE_LEVEL); set_cpus_allowed_ptr(tsk, cpu_all_mask); + set_mems_allowed(node_possible_map); current->flags |= PF_NOFREEZE | PF_FREEZER_NOSIG; diff --git a/mm/mempolicy.c b/mm/mempolicy.c index 3eb4a6fdc04..46bdf9ddf2b 100644 --- a/mm/mempolicy.c +++ b/mm/mempolicy.c @@ -182,13 +182,54 @@ static int mpol_new_bind(struct mempolicy *pol, const nodemask_t *nodes) return 0; } -/* Create a new policy */ +/* + * mpol_set_nodemask is called after mpol_new() to set up the nodemask, if + * any, for the new policy. mpol_new() has already validated the nodes + * parameter with respect to the policy mode and flags. But, we need to + * handle an empty nodemask with MPOL_PREFERRED here. + * + * Must be called holding task's alloc_lock to protect task's mems_allowed + * and mempolicy. May also be called holding the mmap_semaphore for write. + */ +static int mpol_set_nodemask(struct mempolicy *pol, const nodemask_t *nodes) +{ + nodemask_t cpuset_context_nmask; + int ret; + + /* if mode is MPOL_DEFAULT, pol is NULL. This is right. */ + if (pol == NULL) + return 0; + + VM_BUG_ON(!nodes); + if (pol->mode == MPOL_PREFERRED && nodes_empty(*nodes)) + nodes = NULL; /* explicit local allocation */ + else { + if (pol->flags & MPOL_F_RELATIVE_NODES) + mpol_relative_nodemask(&cpuset_context_nmask, nodes, + &cpuset_current_mems_allowed); + else + nodes_and(cpuset_context_nmask, *nodes, + cpuset_current_mems_allowed); + if (mpol_store_user_nodemask(pol)) + pol->w.user_nodemask = *nodes; + else + pol->w.cpuset_mems_allowed = + cpuset_current_mems_allowed; + } + + ret = mpol_ops[pol->mode].create(pol, + nodes ? &cpuset_context_nmask : NULL); + return ret; +} + +/* + * This function just creates a new policy, does some check and simple + * initialization. You must invoke mpol_set_nodemask() to set nodes. + */ static struct mempolicy *mpol_new(unsigned short mode, unsigned short flags, nodemask_t *nodes) { struct mempolicy *policy; - nodemask_t cpuset_context_nmask; - int ret; pr_debug("setting mode %d flags %d nodes[0] %lx\n", mode, flags, nodes ? nodes_addr(*nodes)[0] : -1); @@ -210,7 +251,6 @@ static struct mempolicy *mpol_new(unsigned short mode, unsigned short flags, if (((flags & MPOL_F_STATIC_NODES) || (flags & MPOL_F_RELATIVE_NODES))) return ERR_PTR(-EINVAL); - nodes = NULL; /* flag local alloc */ } } else if (nodes_empty(*nodes)) return ERR_PTR(-EINVAL); @@ -221,30 +261,6 @@ static struct mempolicy *mpol_new(unsigned short mode, unsigned short flags, policy->mode = mode; policy->flags = flags; - if (nodes) { - /* - * cpuset related setup doesn't apply to local allocation - */ - cpuset_update_task_memory_state(); - if (flags & MPOL_F_RELATIVE_NODES) - mpol_relative_nodemask(&cpuset_context_nmask, nodes, - &cpuset_current_mems_allowed); - else - nodes_and(cpuset_context_nmask, *nodes, - cpuset_current_mems_allowed); - if (mpol_store_user_nodemask(policy)) - policy->w.user_nodemask = *nodes; - else - policy->w.cpuset_mems_allowed = - cpuset_mems_allowed(current); - } - - ret = mpol_ops[mode].create(policy, - nodes ? &cpuset_context_nmask : NULL); - if (ret < 0) { - kmem_cache_free(policy_cache, policy); - return ERR_PTR(ret); - } return policy; } @@ -324,6 +340,8 @@ static void mpol_rebind_policy(struct mempolicy *pol, /* * Wrapper for mpol_rebind_policy() that just requires task * pointer, and updates task mempolicy. + * + * Called with task's alloc_lock held. */ void mpol_rebind_task(struct task_struct *tsk, const nodemask_t *new) @@ -600,8 +618,9 @@ static void mpol_set_task_struct_flag(void) static long do_set_mempolicy(unsigned short mode, unsigned short flags, nodemask_t *nodes) { - struct mempolicy *new; + struct mempolicy *new, *old; struct mm_struct *mm = current->mm; + int ret; new = mpol_new(mode, flags, nodes); if (IS_ERR(new)) @@ -615,20 +634,33 @@ static long do_set_mempolicy(unsigned short mode, unsigned short flags, */ if (mm) down_write(&mm->mmap_sem); - mpol_put(current->mempolicy); + task_lock(current); + ret = mpol_set_nodemask(new, nodes); + if (ret) { + task_unlock(current); + if (mm) + up_write(&mm->mmap_sem); + mpol_put(new); + return ret; + } + old = current->mempolicy; current->mempolicy = new; mpol_set_task_struct_flag(); if (new && new->mode == MPOL_INTERLEAVE && nodes_weight(new->v.nodes)) current->il_next = first_node(new->v.nodes); + task_unlock(current); if (mm) up_write(&mm->mmap_sem); + mpol_put(old); return 0; } /* * Return nodemask for policy for get_mempolicy() query + * + * Called with task's alloc_lock held */ static void get_policy_nodemask(struct mempolicy *p, nodemask_t *nodes) { @@ -674,7 +706,6 @@ static long do_get_mempolicy(int *policy, nodemask_t *nmask, struct vm_area_struct *vma = NULL; struct mempolicy *pol = current->mempolicy; - cpuset_update_task_memory_state(); if (flags & ~(unsigned long)(MPOL_F_NODE|MPOL_F_ADDR|MPOL_F_MEMS_ALLOWED)) return -EINVAL; @@ -683,7 +714,9 @@ static long do_get_mempolicy(int *policy, nodemask_t *nmask, if (flags & (MPOL_F_NODE|MPOL_F_ADDR)) return -EINVAL; *policy = 0; /* just so it's initialized */ + task_lock(current); *nmask = cpuset_current_mems_allowed; + task_unlock(current); return 0; } @@ -738,8 +771,11 @@ static long do_get_mempolicy(int *policy, nodemask_t *nmask, } err = 0; - if (nmask) + if (nmask) { + task_lock(current); get_policy_nodemask(pol, nmask); + task_unlock(current); + } out: mpol_cond_put(pol); @@ -979,6 +1015,14 @@ static long do_mbind(unsigned long start, unsigned long len, return err; } down_write(&mm->mmap_sem); + task_lock(current); + err = mpol_set_nodemask(new, nmask); + task_unlock(current); + if (err) { + up_write(&mm->mmap_sem); + mpol_put(new); + return err; + } vma = check_range(mm, start, end, nmask, flags | MPOL_MF_INVERT, &pagelist); @@ -1545,8 +1589,6 @@ alloc_page_vma(gfp_t gfp, struct vm_area_struct *vma, unsigned long addr) struct mempolicy *pol = get_vma_policy(current, vma, addr); struct zonelist *zl; - cpuset_update_task_memory_state(); - if (unlikely(pol->mode == MPOL_INTERLEAVE)) { unsigned nid; @@ -1593,8 +1635,6 @@ struct page *alloc_pages_current(gfp_t gfp, unsigned order) { struct mempolicy *pol = current->mempolicy; - if ((gfp & __GFP_WAIT) && !in_interrupt()) - cpuset_update_task_memory_state(); if (!pol || in_interrupt() || (gfp & __GFP_THISNODE)) pol = &default_policy; @@ -1854,6 +1894,8 @@ restart: */ void mpol_shared_policy_init(struct shared_policy *sp, struct mempolicy *mpol) { + int ret; + sp->root = RB_ROOT; /* empty tree == default mempolicy */ spin_lock_init(&sp->lock); @@ -1863,9 +1905,19 @@ void mpol_shared_policy_init(struct shared_policy *sp, struct mempolicy *mpol) /* contextualize the tmpfs mount point mempolicy */ new = mpol_new(mpol->mode, mpol->flags, &mpol->w.user_nodemask); - mpol_put(mpol); /* drop our ref on sb mpol */ - if (IS_ERR(new)) + if (IS_ERR(new)) { + mpol_put(mpol); /* drop our ref on sb mpol */ return; /* no valid nodemask intersection */ + } + + task_lock(current); + ret = mpol_set_nodemask(new, &mpol->w.user_nodemask); + task_unlock(current); + mpol_put(mpol); /* drop our ref on sb mpol */ + if (ret) { + mpol_put(new); + return; + } /* Create pseudo-vma that contains just the policy */ memset(&pvma, 0, sizeof(struct vm_area_struct)); @@ -2086,8 +2138,19 @@ int mpol_parse_str(char *str, struct mempolicy **mpol, int no_context) new = mpol_new(mode, mode_flags, &nodes); if (IS_ERR(new)) err = 1; - else if (no_context) - new->w.user_nodemask = nodes; /* save for contextualization */ + else { + int ret; + + task_lock(current); + ret = mpol_set_nodemask(new, &nodes); + task_unlock(current); + if (ret) + err = 1; + else if (no_context) { + /* save for contextualization */ + new->w.user_nodemask = nodes; + } + } out: /* Restore string for error message */ diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 17d5f539a9a..7cc3179e359 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -1569,10 +1569,7 @@ nofail_alloc: /* We now go into synchronous reclaim */ cpuset_memory_pressure_bump(); - /* - * The task's cpuset might have expanded its set of allowable nodes - */ - cpuset_update_task_memory_state(); + p->flags |= PF_MEMALLOC; lockdep_set_current_reclaim_state(gfp_mask); -- cgit v1.2.3-70-g09d2 From 2ff05b2b4eac2e63d345fc731ea151a060247f53 Mon Sep 17 00:00:00 2001 From: David Rientjes Date: Tue, 16 Jun 2009 15:32:56 -0700 Subject: oom: move oom_adj value from task_struct to mm_struct The per-task oom_adj value is a characteristic of its mm more than the task itself since it's not possible to oom kill any thread that shares the mm. If a task were to be killed while attached to an mm that could not be freed because another thread were set to OOM_DISABLE, it would have needlessly been terminated since there is no potential for future memory freeing. This patch moves oomkilladj (now more appropriately named oom_adj) from struct task_struct to struct mm_struct. This requires task_lock() on a task to check its oom_adj value to protect against exec, but it's already necessary to take the lock when dereferencing the mm to find the total VM size for the badness heuristic. This fixes a livelock if the oom killer chooses a task and another thread sharing the same memory has an oom_adj value of OOM_DISABLE. This occurs because oom_kill_task() repeatedly returns 1 and refuses to kill the chosen task while select_bad_process() will repeatedly choose the same task during the next retry. Taking task_lock() in select_bad_process() to check for OOM_DISABLE and in oom_kill_task() to check for threads sharing the same memory will be removed in the next patch in this series where it will no longer be necessary. Writing to /proc/pid/oom_adj for a kthread will now return -EINVAL since these threads are immune from oom killing already. They simply report an oom_adj value of OOM_DISABLE. Cc: Nick Piggin Cc: Rik van Riel Cc: Mel Gorman Signed-off-by: David Rientjes Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- Documentation/filesystems/proc.txt | 15 ++++++++++----- fs/proc/base.c | 19 ++++++++++++++++--- include/linux/mm_types.h | 2 ++ include/linux/sched.h | 1 - mm/oom_kill.c | 34 ++++++++++++++++++++++------------ 5 files changed, 50 insertions(+), 21 deletions(-) (limited to 'include/linux/sched.h') diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt index cd8717a3627..ebff3c10a07 100644 --- a/Documentation/filesystems/proc.txt +++ b/Documentation/filesystems/proc.txt @@ -1003,11 +1003,13 @@ CHAPTER 3: PER-PROCESS PARAMETERS 3.1 /proc//oom_adj - Adjust the oom-killer score ------------------------------------------------------ -This file can be used to adjust the score used to select which processes -should be killed in an out-of-memory situation. Giving it a high score will -increase the likelihood of this process being killed by the oom-killer. Valid -values are in the range -16 to +15, plus the special value -17, which disables -oom-killing altogether for this process. +This file can be used to adjust the score used to select which processes should +be killed in an out-of-memory situation. The oom_adj value is a characteristic +of the task's mm, so all threads that share an mm with pid will have the same +oom_adj value. A high value will increase the likelihood of this process being +killed by the oom-killer. Valid values are in the range -16 to +15 as +explained below and a special value of -17, which disables oom-killing +altogether for threads sharing pid's mm. The process to be killed in an out-of-memory situation is selected among all others based on its badness score. This value equals the original memory size of the process @@ -1021,6 +1023,9 @@ the parent's score if they do not share the same memory. Thus forking servers are the prime candidates to be killed. Having only one 'hungry' child will make parent less preferable than the child. +/proc//oom_adj cannot be changed for kthreads since they are immune from +oom-killing already. + /proc//oom_score shows process' current badness score. The following heuristics are then applied: diff --git a/fs/proc/base.c b/fs/proc/base.c index 1539e630c47..3ce5ae9e3d2 100644 --- a/fs/proc/base.c +++ b/fs/proc/base.c @@ -1006,7 +1006,12 @@ static ssize_t oom_adjust_read(struct file *file, char __user *buf, if (!task) return -ESRCH; - oom_adjust = task->oomkilladj; + task_lock(task); + if (task->mm) + oom_adjust = task->mm->oom_adj; + else + oom_adjust = OOM_DISABLE; + task_unlock(task); put_task_struct(task); len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust); @@ -1035,11 +1040,19 @@ static ssize_t oom_adjust_write(struct file *file, const char __user *buf, task = get_proc_task(file->f_path.dentry->d_inode); if (!task) return -ESRCH; - if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) { + task_lock(task); + if (!task->mm) { + task_unlock(task); + put_task_struct(task); + return -EINVAL; + } + if (oom_adjust < task->mm->oom_adj && !capable(CAP_SYS_RESOURCE)) { + task_unlock(task); put_task_struct(task); return -EACCES; } - task->oomkilladj = oom_adjust; + task->mm->oom_adj = oom_adjust; + task_unlock(task); put_task_struct(task); if (end - buffer == 0) return -EIO; diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 0e80e26ecf2..f4408106fcb 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -232,6 +232,8 @@ struct mm_struct { unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ + s8 oom_adj; /* OOM kill score adjustment (bit shift) */ + cpumask_t cpu_vm_mask; /* Architecture-specific MM context */ diff --git a/include/linux/sched.h b/include/linux/sched.h index 1048bf50540..1bc6fae0c13 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1178,7 +1178,6 @@ struct task_struct { * a short time */ unsigned char fpu_counter; - s8 oomkilladj; /* OOM kill score adjustment (bit shift). */ #ifdef CONFIG_BLK_DEV_IO_TRACE unsigned int btrace_seq; #endif diff --git a/mm/oom_kill.c b/mm/oom_kill.c index a7b2460e922..b60913520ef 100644 --- a/mm/oom_kill.c +++ b/mm/oom_kill.c @@ -58,6 +58,7 @@ unsigned long badness(struct task_struct *p, unsigned long uptime) unsigned long points, cpu_time, run_time; struct mm_struct *mm; struct task_struct *child; + int oom_adj; task_lock(p); mm = p->mm; @@ -65,6 +66,7 @@ unsigned long badness(struct task_struct *p, unsigned long uptime) task_unlock(p); return 0; } + oom_adj = mm->oom_adj; /* * The memory size of the process is the basis for the badness. @@ -148,15 +150,15 @@ unsigned long badness(struct task_struct *p, unsigned long uptime) points /= 8; /* - * Adjust the score by oomkilladj. + * Adjust the score by oom_adj. */ - if (p->oomkilladj) { - if (p->oomkilladj > 0) { + if (oom_adj) { + if (oom_adj > 0) { if (!points) points = 1; - points <<= p->oomkilladj; + points <<= oom_adj; } else - points >>= -(p->oomkilladj); + points >>= -(oom_adj); } #ifdef DEBUG @@ -251,8 +253,12 @@ static struct task_struct *select_bad_process(unsigned long *ppoints, *ppoints = ULONG_MAX; } - if (p->oomkilladj == OOM_DISABLE) + task_lock(p); + if (p->mm && p->mm->oom_adj == OOM_DISABLE) { + task_unlock(p); continue; + } + task_unlock(p); points = badness(p, uptime.tv_sec); if (points > *ppoints || !chosen) { @@ -304,8 +310,7 @@ static void dump_tasks(const struct mem_cgroup *mem) } printk(KERN_INFO "[%5d] %5d %5d %8lu %8lu %3d %3d %s\n", p->pid, __task_cred(p)->uid, p->tgid, mm->total_vm, - get_mm_rss(mm), (int)task_cpu(p), p->oomkilladj, - p->comm); + get_mm_rss(mm), (int)task_cpu(p), mm->oom_adj, p->comm); task_unlock(p); } while_each_thread(g, p); } @@ -367,8 +372,12 @@ static int oom_kill_task(struct task_struct *p) * Don't kill the process if any threads are set to OOM_DISABLE */ do_each_thread(g, q) { - if (q->mm == mm && q->oomkilladj == OOM_DISABLE) + task_lock(q); + if (q->mm == mm && q->mm && q->mm->oom_adj == OOM_DISABLE) { + task_unlock(q); return 1; + } + task_unlock(q); } while_each_thread(g, q); __oom_kill_task(p, 1); @@ -393,10 +402,11 @@ static int oom_kill_process(struct task_struct *p, gfp_t gfp_mask, int order, struct task_struct *c; if (printk_ratelimit()) { - printk(KERN_WARNING "%s invoked oom-killer: " - "gfp_mask=0x%x, order=%d, oomkilladj=%d\n", - current->comm, gfp_mask, order, current->oomkilladj); task_lock(current); + printk(KERN_WARNING "%s invoked oom-killer: " + "gfp_mask=0x%x, order=%d, oom_adj=%d\n", + current->comm, gfp_mask, order, + current->mm ? current->mm->oom_adj : OOM_DISABLE); cpuset_print_task_mems_allowed(current); task_unlock(current); dump_stack(); -- cgit v1.2.3-70-g09d2 From 20ebcdda78a282d1d5266887ddf8a2d670182576 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Wed, 17 Jun 2009 16:27:16 -0700 Subject: memcg: remove unneeded forward declaration from sched.h This forward declaration seems pointless. Signed-off-by: Li Zefan Cc: Balbir Singh Acked-by: KAMEZAWA Hiroyuki Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/sched.h | 1 - 1 file changed, 1 deletion(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index 02042e7f219..d0342101756 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -92,7 +92,6 @@ struct sched_param { #include -struct mem_cgroup; struct exec_domain; struct futex_pi_state; struct robust_list_head; -- cgit v1.2.3-70-g09d2 From 17f98dcf6010a1cfd25d179fd0ce77d3dc2685c3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 17 Jun 2009 16:27:51 -0700 Subject: pids: clean up find_task_by_pid variants find_task_by_pid_type_ns is only used to implement find_task_by_vpid and find_task_by_pid_ns, but both of them pass PIDTYPE_PID as first argument. So just fold find_task_by_pid_type_ns into find_task_by_pid_ns and use find_task_by_pid_ns to implement find_task_by_vpid. While we're at it also remove the exports for find_task_by_pid_ns and find_task_by_vpid - we don't have any modular callers left as the only modular caller of he old pre pid namespace find_task_by_pid (gfs2) was switched to pid_task which operates on a struct pid pointer instead of a pid_t. Given the confusion about pid_t values vs namespace that's generally the better option anyway and I think we're better of restricting modules to do it that way. Signed-off-by: Christoph Hellwig Cc: Pavel Emelyanov Cc: "Eric W. Biederman" Cc: Ingo Molnar Cc: Oleg Nesterov Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/sched.h | 6 ------ kernel/pid.c | 17 +++-------------- 2 files changed, 3 insertions(+), 20 deletions(-) (limited to 'include/linux/sched.h') diff --git a/include/linux/sched.h b/include/linux/sched.h index d0342101756..4d075426988 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1878,9 +1878,6 @@ extern struct pid_namespace init_pid_ns; /* * find a task by one of its numerical ids * - * find_task_by_pid_type_ns(): - * it is the most generic call - it finds a task by all id, - * type and namespace specified * find_task_by_pid_ns(): * finds a task by its pid in the specified namespace * find_task_by_vpid(): @@ -1889,9 +1886,6 @@ extern struct pid_namespace init_pid_ns; * see also find_vpid() etc in include/linux/pid.h */ -extern struct task_struct *find_task_by_pid_type_ns(int type, int pid, - struct pid_namespace *ns); - extern struct task_struct *find_task_by_vpid(pid_t nr); extern struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns); diff --git a/kernel/pid.c b/kernel/pid.c index b2e5f78fd28..31310b5d3f5 100644 --- a/kernel/pid.c +++ b/kernel/pid.c @@ -378,26 +378,15 @@ EXPORT_SYMBOL(pid_task); /* * Must be called under rcu_read_lock() or with tasklist_lock read-held. */ -struct task_struct *find_task_by_pid_type_ns(int type, int nr, - struct pid_namespace *ns) +struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns) { - return pid_task(find_pid_ns(nr, ns), type); + return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID); } -EXPORT_SYMBOL(find_task_by_pid_type_ns); - struct task_struct *find_task_by_vpid(pid_t vnr) { - return find_task_by_pid_type_ns(PIDTYPE_PID, vnr, - current->nsproxy->pid_ns); -} -EXPORT_SYMBOL(find_task_by_vpid); - -struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns) -{ - return find_task_by_pid_type_ns(PIDTYPE_PID, nr, ns); + return find_task_by_pid_ns(vnr, current->nsproxy->pid_ns); } -EXPORT_SYMBOL(find_task_by_pid_ns); struct pid *get_task_pid(struct task_struct *task, enum pid_type type) { -- cgit v1.2.3-70-g09d2