Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263851AbUDSDKd (ORCPT ); Sun, 18 Apr 2004 23:10:33 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264272AbUDSDKd (ORCPT ); Sun, 18 Apr 2004 23:10:33 -0400 Received: from ausmtp01.au.ibm.com ([202.81.18.186]:58316 "EHLO ausmtp01.au.ibm.com") by vger.kernel.org with ESMTP id S263851AbUDSDKK (ORCPT ); Sun, 18 Apr 2004 23:10:10 -0400 Subject: [PATCH] create_singlethread_workqueue() From: "Rusty Russell (IBM)" To: lkml - Kernel Mailing List Cc: Linus Torvalds , Andrew Morton Content-Type: text/plain Organization: IBM Message-Id: <1082344185.29285.4.camel@bach> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Mon, 19 Apr 2004 13:09:45 +1000 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 8091 Lines: 232 I considered a more general solution, but in the end this was simplest: I can't currently see much point in anything other than 1 thread or 1 thread per-cpu. Name: Workqueues With Only A Single Thread, Not One Per CPU Status: Tested on 2.6.6-rc1-bk3 Workqueues are a great primitive for running things from user context from a completely clean environment. Unfortunately, they currently insist on creating one thread per CPU, which is overkill for many situations, so the more generic keventd workqueue is used for these. Recently deadlocks using keventd were demonstrated, showing that it is not suitable for all uses. 1) Clean up CPU iterators. Always a nice touch. 2) Add __create_workqueue() and create_singlethread_workqueue(), keeping source compatibility. 3) Put workqueues in workqueue list even if !CONFIG_HOTPLUG_CPU (means we need a lock to protect that list). Now we can tell if a wq is single-threaded using list_empty(&wq->list). 4) For single-threaded workqueues, override CPU in queue_work, delayed_work_timer_fn and flush_workqueue to be 0. flush_workqueue now does redundant passes for single-threaded workqueues, but the code remains simple. 5) Make create_workqueue_thread return the thread, so we can easily kthread_bind for multi-threaded workqueues. diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .13527-linux-2.6.6-rc1-bk2/include/linux/workqueue.h .13527-linux-2.6.6-rc1-bk2.updated/include/linux/workqueue.h diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .26663-linux-2.6.6-rc1-bk2/include/linux/workqueue.h .26663-linux-2.6.6-rc1-bk2.updated/include/linux/workqueue.h --- .26663-linux-2.6.6-rc1-bk2/include/linux/workqueue.h 2004-03-12 07:57:28.000000000 +1100 +++ .26663-linux-2.6.6-rc1-bk2.updated/include/linux/workqueue.h 2004-04-18 23:25:50.000000000 +1000 @@ -49,7 +50,11 @@ struct work_struct { init_timer(&(_work)->timer); \ } while (0) -extern struct workqueue_struct *create_workqueue(const char *name); +extern struct workqueue_struct *__create_workqueue(const char *name, + int singlethread); +#define create_workqueue(name) __create_workqueue((name), 0) +#define create_singlethread_workqueue(name) __create_workqueue((name), 1) + extern void destroy_workqueue(struct workqueue_struct *wq); extern int FASTCALL(queue_work(struct workqueue_struct *wq, struct work_struct *work)); diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .26663-linux-2.6.6-rc1-bk2/kernel/workqueue.c .26663-linux-2.6.6-rc1-bk2.updated/kernel/workqueue.c --- .26663-linux-2.6.6-rc1-bk2/kernel/workqueue.c 2004-04-05 09:04:48.000000000 +1000 +++ .26663-linux-2.6.6-rc1-bk2.updated/kernel/workqueue.c 2004-04-18 23:25:50.000000000 +1000 @@ -27,7 +27,7 @@ #include /* - * The per-CPU workqueue. + * The per-CPU workqueue (if single thread, we always use cpu 0's). * * The sequence counters are for flush_scheduled_work(). It wants to wait * until until all currently-scheduled works are completed, but it doesn't @@ -59,20 +59,19 @@ struct cpu_workqueue_struct { struct workqueue_struct { struct cpu_workqueue_struct cpu_wq[NR_CPUS]; const char *name; - struct list_head list; + struct list_head list; /* Empty if single thread */ }; -#ifdef CONFIG_HOTPLUG_CPU -/* All the workqueues on the system, for hotplug cpu to add/remove - threads to each one as cpus come/go. Protected by cpucontrol - sem. */ +/* All the per-cpu workqueues on the system, for hotplug cpu to add/remove + threads to each one as cpus come/go. */ +static spinlock_t workqueue_lock = SPIN_LOCK_UNLOCKED; static LIST_HEAD(workqueues); -#define add_workqueue(wq) list_add(&(wq)->list, &workqueues) -#define del_workqueue(wq) list_del(&(wq)->list) -#else -#define add_workqueue(wq) -#define del_workqueue(wq) -#endif /* CONFIG_HOTPLUG_CPU */ + +/* If it's single threaded, it isn't in the list of workqueues. */ +static inline int is_single_threaded(struct workqueue_struct *wq) +{ + return list_empty(&wq->list); +} /* Preempt must be disabled. */ static void __queue_work(struct cpu_workqueue_struct *cwq, @@ -100,6 +99,8 @@ int fastcall queue_work(struct workqueue int ret = 0, cpu = get_cpu(); if (!test_and_set_bit(0, &work->pending)) { + if (unlikely(is_single_threaded(wq))) + cpu = 0; BUG_ON(!list_empty(&work->entry)); __queue_work(wq->cpu_wq + cpu, work); ret = 1; @@ -112,8 +113,12 @@ static void delayed_work_timer_fn(unsign { struct work_struct *work = (struct work_struct *)__data; struct workqueue_struct *wq = work->wq_data; + int cpu = smp_processor_id(); - __queue_work(wq->cpu_wq + smp_processor_id(), work); + if (unlikely(is_single_threaded(wq))) + cpu = 0; + + __queue_work(wq->cpu_wq + cpu, work); } int fastcall queue_delayed_work(struct workqueue_struct *wq, @@ -234,12 +239,12 @@ void fastcall flush_workqueue(struct wor might_sleep(); lock_cpu_hotplug(); - for (cpu = 0; cpu < NR_CPUS; cpu++) { + for_each_online_cpu(cpu) { DEFINE_WAIT(wait); long sequence_needed; - if (!cpu_online(cpu)) - continue; + if (is_single_threaded(wq)) + cpu = 0; cwq = wq->cpu_wq + cpu; if (cwq->thread == current) { @@ -266,7 +271,8 @@ void fastcall flush_workqueue(struct wor unlock_cpu_hotplug(); } -static int create_workqueue_thread(struct workqueue_struct *wq, int cpu) +static struct task_struct *create_workqueue_thread(struct workqueue_struct *wq, + int cpu) { struct cpu_workqueue_struct *cwq = wq->cpu_wq + cpu; struct task_struct *p; @@ -282,16 +288,17 @@ static int create_workqueue_thread(struc p = kthread_create(worker_thread, cwq, "%s/%d", wq->name, cpu); if (IS_ERR(p)) - return PTR_ERR(p); + return NULL; cwq->thread = p; - kthread_bind(p, cpu); - return 0; + return p; } -struct workqueue_struct *create_workqueue(const char *name) +struct workqueue_struct *__create_workqueue(const char *name, + int singlethread) { int cpu, destroy = 0; struct workqueue_struct *wq; + struct task_struct *p; BUG_ON(strlen(name) > 10); @@ -301,15 +309,26 @@ struct workqueue_struct *create_workqueu wq->name = name; /* We don't need the distraction of CPUs appearing and vanishing. */ lock_cpu_hotplug(); - for (cpu = 0; cpu < NR_CPUS; cpu++) { - if (!cpu_online(cpu)) - continue; - if (create_workqueue_thread(wq, cpu) < 0) + if (singlethread) { + INIT_LIST_HEAD(&wq->list); + p = create_workqueue_thread(wq, 0); + if (!p) destroy = 1; else - wake_up_process(wq->cpu_wq[cpu].thread); + wake_up_process(p); + } else { + spin_lock(&workqueue_lock); + list_add(&wq->list, &workqueues); + spin_unlock_irq(&workqueue_lock); + for_each_online_cpu(cpu) { + p = create_workqueue_thread(wq, cpu); + if (p) { + kthread_bind(p, cpu); + wake_up_process(p); + } else + destroy = 1; + } } - add_workqueue(wq); /* * Was there any error during startup? If yes then clean up: @@ -347,11 +367,15 @@ void destroy_workqueue(struct workqueue_ /* We don't need the distraction of CPUs appearing and vanishing. */ lock_cpu_hotplug(); - for (cpu = 0; cpu < NR_CPUS; cpu++) { - if (cpu_online(cpu)) + if (is_single_threaded(wq)) + cleanup_workqueue_thread(wq, 0); + else { + for_each_online_cpu(cpu) cleanup_workqueue_thread(wq, cpu); + spin_lock(&workqueue_lock); + list_del(&wq->list); + spin_unlock_irq(&workqueue_lock); } - del_workqueue(wq); unlock_cpu_hotplug(); kfree(wq); } @@ -467,7 +491,7 @@ void init_workqueues(void) BUG_ON(!keventd_wq); } -EXPORT_SYMBOL_GPL(create_workqueue); +EXPORT_SYMBOL_GPL(__create_workqueue); EXPORT_SYMBOL_GPL(queue_work); EXPORT_SYMBOL_GPL(queue_delayed_work); EXPORT_SYMBOL_GPL(flush_workqueue); - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/