Re: [PATCH RT 7/9][RFC] [PATCH 7/9] cpu/rt: Rework cpu down for PREEMPT_RT
From: Steven Rostedt <rostedt@goodmis.org>
Date: 2012-03-02 15:37:01
Also in:
lkml
Subsystem:
cpu hotplug, the rest · Maintainers:
Thomas Gleixner, Peter Zijlstra, Linus Torvalds
On Fri, 2012-03-02 at 16:20 +0100, Thomas Gleixner wrote:
Already queued. I'd say #4 is a bug fix as well, though I fundamentally hate the ass backwards semantics of that new function.
It is a bug fix, but I don't like it either. Note, that the patch didn't update the pin_current_cpu() code, which would have to be done too (to be a full fix). But I was also thinking that as a work around, as we plan on changing this code anyway. Instead of adding a new API which is ass backwards, just encompass the cpu_hotplug.lock instead. Something like this: Not tested nor complied. Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Yes this is ugly, but at least we don't introduce an ass backwards semantic.
diff --git a/kernel/cpu.c b/kernel/cpu.c
index fa40834..c25b5ff 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c@@ -46,7 +46,12 @@ static int cpu_hotplug_disabled; static struct { struct task_struct *active_writer; +#ifdef CONFIG_PREEMPT_RT_FULL + /* Makes the lock keep the task's state */ + spinlock_t lock; +#else struct mutex lock; /* Synchronizes accesses to refcount, */ +#endif /* * Also blocks the new readers during * an ongoing cpu hotplug operation.
@@ -58,6 +63,14 @@ static struct { .refcount = 0, }; +#ifdef CONFIG_PREEMPT_RT_FULL +# define hotplug_lock() spin_lock(&cpu_hotplug.lock) +# define hotplug_unlock() spin_unlock(&cpu_hotplug.lock) +#else +# define hotplug_lock() mutex_lock(&cpu_hotplug.lock) +# define hotplug_lock() mutex_unlock(&cpu_hotplug.lock) +#endif + struct hotplug_pcp { struct task_struct *unplug; int refcount;
@@ -87,8 +100,8 @@ retry: return; } preempt_enable(); - mutex_lock(&cpu_hotplug.lock); - mutex_unlock(&cpu_hotplug.lock); + hotplug_lock(); + hotplug_unlock(); preempt_disable(); goto retry; }
@@ -161,9 +174,9 @@ void get_online_cpus(void) might_sleep(); if (cpu_hotplug.active_writer == current) return; - mutex_lock(&cpu_hotplug.lock); + hotplug_lock(); cpu_hotplug.refcount++; - mutex_unlock(&cpu_hotplug.lock); + hotplug_unlock(); } EXPORT_SYMBOL_GPL(get_online_cpus);
@@ -172,10 +185,10 @@ void put_online_cpus(void) { if (cpu_hotplug.active_writer == current) return; - mutex_lock(&cpu_hotplug.lock); + hotplug_lock(); if (!--cpu_hotplug.refcount && unlikely(cpu_hotplug.active_writer)) wake_up_process(cpu_hotplug.active_writer); - mutex_unlock(&cpu_hotplug.lock); + hotplug_unlock(); } EXPORT_SYMBOL_GPL(put_online_cpus);
@@ -207,11 +220,11 @@ static void cpu_hotplug_begin(void) cpu_hotplug.active_writer = current; for (;;) { - mutex_lock(&cpu_hotplug.lock); + hotplug_lock(); if (likely(!cpu_hotplug.refcount)) break; __set_current_state(TASK_UNINTERRUPTIBLE); - mutex_unlock(&cpu_hotplug.lock); + hotplug_unlock(); schedule(); } }
@@ -219,7 +232,7 @@ static void cpu_hotplug_begin(void) static void cpu_hotplug_done(void) { cpu_hotplug.active_writer = NULL; - mutex_unlock(&cpu_hotplug.lock); + hotplug_unlock(); } #else /* #if CONFIG_HOTPLUG_CPU */