aboutsummaryrefslogtreecommitdiff
path: root/kernel/fork.c
diff options
context:
space:
mode:
authorAndy Lutomirski <luto@kernel.org>2016-09-15 22:45:49 -0700
committerIngo Molnar <mingo@kernel.org>2016-09-16 09:18:54 +0200
commitac496bf48d97f2503eaa353996a4dd5e4383eaf0 (patch)
treed57eb8f649cc8268a3d3595f04bbe4bab72264ef /kernel/fork.c
parent68f24b08ee892d47bdef925d676e1ae1ccc316f8 (diff)
downloadlinux-linaro-stable-ac496bf48d97f2503eaa353996a4dd5e4383eaf0.tar.gz
fork: Optimize task creation by caching two thread stacks per CPU if CONFIG_VMAP_STACK=y
vmalloc() is a bit slow, and pounding vmalloc()/vfree() will eventually force a global TLB flush. To reduce pressure on them, if CONFIG_VMAP_STACK=y, cache two thread stacks per CPU. This will let us quickly allocate a hopefully cache-hot, TLB-hot stack under heavy forking workloads (shell script style). On my silly pthread_create() benchmark, it saves about 2 ┬Ás per pthread_create()+join() with CONFIG_VMAP_STACK=y. Signed-off-by: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Jann Horn <jann@thejh.net> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/94811d8e3994b2e962f88866290017d498eb069c.1474003868.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
Diffstat (limited to 'kernel/fork.c')
-rw-r--r--kernel/fork.c62
1 files changed, 53 insertions, 9 deletions
diff --git a/kernel/fork.c b/kernel/fork.c
index 5dd0a516626d..c060c7e7c247 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -159,15 +159,41 @@ void __weak arch_release_thread_stack(unsigned long *stack)
* kmemcache based allocator.
*/
# if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
+
+#ifdef CONFIG_VMAP_STACK
+/*
+ * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
+ * flush. Try to minimize the number of calls by caching stacks.
+ */
+#define NR_CACHED_STACKS 2
+static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
+#endif
+
static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
{
#ifdef CONFIG_VMAP_STACK
- void *stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
- VMALLOC_START, VMALLOC_END,
- THREADINFO_GFP | __GFP_HIGHMEM,
- PAGE_KERNEL,
- 0, node,
- __builtin_return_address(0));
+ void *stack;
+ int i;
+
+ local_irq_disable();
+ for (i = 0; i < NR_CACHED_STACKS; i++) {
+ struct vm_struct *s = this_cpu_read(cached_stacks[i]);
+
+ if (!s)
+ continue;
+ this_cpu_write(cached_stacks[i], NULL);
+
+ tsk->stack_vm_area = s;
+ local_irq_enable();
+ return s->addr;
+ }
+ local_irq_enable();
+
+ stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
+ VMALLOC_START, VMALLOC_END,
+ THREADINFO_GFP | __GFP_HIGHMEM,
+ PAGE_KERNEL,
+ 0, node, __builtin_return_address(0));
/*
* We can't call find_vm_area() in interrupt context, and
@@ -187,10 +213,28 @@ static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
static inline void free_thread_stack(struct task_struct *tsk)
{
- if (task_stack_vm_area(tsk))
+#ifdef CONFIG_VMAP_STACK
+ if (task_stack_vm_area(tsk)) {
+ unsigned long flags;
+ int i;
+
+ local_irq_save(flags);
+ for (i = 0; i < NR_CACHED_STACKS; i++) {
+ if (this_cpu_read(cached_stacks[i]))
+ continue;
+
+ this_cpu_write(cached_stacks[i], tsk->stack_vm_area);
+ local_irq_restore(flags);
+ return;
+ }
+ local_irq_restore(flags);
+
vfree(tsk->stack);
- else
- __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
+ return;
+ }
+#endif
+
+ __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
}
# else
static struct kmem_cache *thread_stack_cache;