From 15316ba81aee6775d6079fb46c66c801989e7d10 Mon Sep 17 00:00:00 2001 From: Christoph Lameter Date: Sun, 8 Jan 2006 01:00:43 -0800 Subject: [PATCH] add schedule_on_each_cpu() swap migration's isolate_lru_page() currently uses an IPI to notify other processors that the lru caches need to be drained if the page cannot be found on the LRU. The IPI interrupt may interrupt a processor that is just processing lru requests and cause a race condition. This patch introduces a new function run_on_each_cpu() that uses the keventd() to run the LRU draining on each processor. Processors disable preemption when dealing the LRU caches (these are per processor) and thus executing LRU draining from another process is safe. Thanks to Lee Schermerhorn for finding this race condition. Signed-off-by: Christoph Lameter Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/workqueue.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) (limited to 'kernel/workqueue.c') diff --git a/kernel/workqueue.c b/kernel/workqueue.c index 2bd5aee1c736..62d47220696a 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -419,6 +419,25 @@ int schedule_delayed_work_on(int cpu, return ret; } +int schedule_on_each_cpu(void (*func) (void *info), void *info) +{ + int cpu; + struct work_struct *work; + + work = kmalloc(NR_CPUS * sizeof(struct work_struct), GFP_KERNEL); + + if (!work) + return -ENOMEM; + for_each_online_cpu(cpu) { + INIT_WORK(work + cpu, func, info); + __queue_work(per_cpu_ptr(keventd_wq->cpu_wq, cpu), + work + cpu); + } + flush_workqueue(keventd_wq); + kfree(work); + return 0; +} + void flush_scheduled_work(void) { flush_workqueue(keventd_wq); -- cgit From 676121fcb66c861804e38d94214fd5670a1ef595 Mon Sep 17 00:00:00 2001 From: Ben Collins Date: Sun, 8 Jan 2006 01:03:04 -0800 Subject: [PATCH] Unchecked alloc_percpu() return in __create_workqueue() __create_workqueue() not checking return of alloc_percpu() NULL dereference was possible. Signed-off-by: Ben Collins Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/workqueue.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'kernel/workqueue.c') diff --git a/kernel/workqueue.c b/kernel/workqueue.c index 62d47220696a..e72fb6478da6 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -315,6 +315,11 @@ struct workqueue_struct *__create_workqueue(const char *name, return NULL; wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct); + if (!wq->cpu_wq) { + kfree(wq); + return NULL; + } + wq->name = name; /* We don't need the distraction of CPUs appearing and vanishing. */ lock_cpu_hotplug(); -- cgit From f756d5e256059018d753f0ba79980ebeb87a1bc0 Mon Sep 17 00:00:00 2001 From: Nathan Lynch Date: Sun, 8 Jan 2006 01:05:12 -0800 Subject: [PATCH] fix workqueue oops during cpu offline Use first_cpu(cpu_possible_map) for the single-thread workqueue case. We used to hardcode 0, but that broke on systems where !cpu_possible(0) when workqueue_struct->cpu_workqueue_struct was changed from a static array to alloc_percpu. Commit id bce61dd49d6ba7799be2de17c772e4c701558f14 ("Fix hardcoded cpu=0 in workqueue for per_cpu_ptr() calls") fixed that for Ben's funky sparc64 system, but it regressed my Power5. Offlining cpu 0 oopses upon the next call to queue_work for a single-thread workqueue, because now we try to manipulate per_cpu_ptr(wq->cpu_wq, 1), which is uninitialized. So we need to establish an unchanging "slot" for single-thread workqueues which will have a valid percpu allocation. Since alloc_percpu keys off of cpu_possible_map, which must not change after initialization, make this slot == first_cpu(cpu_possible_map). Signed-off-by: Nathan Lynch Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/workqueue.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) (limited to 'kernel/workqueue.c') diff --git a/kernel/workqueue.c b/kernel/workqueue.c index e72fb6478da6..82c4fa70595c 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -29,7 +29,8 @@ #include /* - * The per-CPU workqueue (if single thread, we always use cpu 0's). + * The per-CPU workqueue (if single thread, we always use the first + * possible cpu). * * The sequence counters are for flush_scheduled_work(). It wants to wait * until until all currently-scheduled works are completed, but it doesn't @@ -69,6 +70,8 @@ struct workqueue_struct { static DEFINE_SPINLOCK(workqueue_lock); static LIST_HEAD(workqueues); +static int singlethread_cpu; + /* If it's single threaded, it isn't in the list of workqueues. */ static inline int is_single_threaded(struct workqueue_struct *wq) { @@ -102,7 +105,7 @@ int fastcall queue_work(struct workqueue_struct *wq, struct work_struct *work) if (!test_and_set_bit(0, &work->pending)) { if (unlikely(is_single_threaded(wq))) - cpu = any_online_cpu(cpu_online_map); + cpu = singlethread_cpu; BUG_ON(!list_empty(&work->entry)); __queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work); ret = 1; @@ -118,7 +121,7 @@ static void delayed_work_timer_fn(unsigned long __data) int cpu = smp_processor_id(); if (unlikely(is_single_threaded(wq))) - cpu = any_online_cpu(cpu_online_map); + cpu = singlethread_cpu; __queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work); } @@ -267,7 +270,7 @@ void fastcall flush_workqueue(struct workqueue_struct *wq) if (is_single_threaded(wq)) { /* Always use first cpu's area. */ - flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, any_online_cpu(cpu_online_map))); + flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, singlethread_cpu)); } else { int cpu; @@ -325,7 +328,7 @@ struct workqueue_struct *__create_workqueue(const char *name, lock_cpu_hotplug(); if (singlethread) { INIT_LIST_HEAD(&wq->list); - p = create_workqueue_thread(wq, any_online_cpu(cpu_online_map)); + p = create_workqueue_thread(wq, singlethread_cpu); if (!p) destroy = 1; else @@ -379,7 +382,7 @@ void destroy_workqueue(struct workqueue_struct *wq) /* We don't need the distraction of CPUs appearing and vanishing. */ lock_cpu_hotplug(); if (is_single_threaded(wq)) - cleanup_workqueue_thread(wq, any_online_cpu(cpu_online_map)); + cleanup_workqueue_thread(wq, singlethread_cpu); else { for_each_online_cpu(cpu) cleanup_workqueue_thread(wq, cpu); @@ -567,6 +570,7 @@ static int __devinit workqueue_cpu_callback(struct notifier_block *nfb, void init_workqueues(void) { + singlethread_cpu = first_cpu(cpu_possible_map); hotcpu_notifier(workqueue_cpu_callback, 0); keventd_wq = create_workqueue("events"); BUG_ON(!keventd_wq); -- cgit From 858119e159384308a5dde67776691a2ebf70df0f Mon Sep 17 00:00:00 2001 From: Arjan van de Ven Date: Sat, 14 Jan 2006 13:20:43 -0800 Subject: [PATCH] Unlinline a bunch of other functions Remove the "inline" keyword from a bunch of big functions in the kernel with the goal of shrinking it by 30kb to 40kb Signed-off-by: Arjan van de Ven Signed-off-by: Ingo Molnar Acked-by: Jeff Garzik Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- kernel/workqueue.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'kernel/workqueue.c') diff --git a/kernel/workqueue.c b/kernel/workqueue.c index 82c4fa70595c..b052e2c4c710 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -147,7 +147,7 @@ int fastcall queue_delayed_work(struct workqueue_struct *wq, return ret; } -static inline void run_workqueue(struct cpu_workqueue_struct *cwq) +static void run_workqueue(struct cpu_workqueue_struct *cwq) { unsigned long flags; -- cgit