Virtually mapped stack have two bonuses: it eats order-0 pages and
adds guard page at the end. But it slightly slower if system have
plenty free high-order pages.
This patch adds option to use virtually bapped stack as fallback for
atomic allocation of traditional high-order page.
Signed-off-by: Konstantin Khlebnikov <khlebni...@yandex-team.ru>
---
arch/Kconfig | 14 ++++++++++++++
kernel/fork.c | 11 +++++++++++
2 files changed, 25 insertions(+)
diff --git a/arch/Kconfig b/arch/Kconfig
index 400b9e1b2f27..c181ab263e7f 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -904,6 +904,20 @@ config VMAP_STACK
the stack to map directly to the KASAN shadow map using a formula
that is incorrect if the stack is in vmalloc space.
+config VMAP_STACK_AS_FALLBACK
+ default n
+ bool "Use a virtually-mapped stack as fallback for directly-mapped"
+ depends on VMAP_STACK
+ help
+ With this option kernel first tries to allocate directly-mapped stack
+ without calling direct memory reclaim and fallback to vmap stack.
+
+ Allocation of directly mapped stack faster than vmap if system a lot
+ of free memory and much slower if all memory is used or fragmented.
+
+ This option neutralize stack overflow protection but allows to
+ achieve best performance for syscalls fork() and clone().
+
config ARCH_OPTIONAL_KERNEL_RWX
def_bool n
diff --git a/kernel/fork.c b/kernel/fork.c
index 457c9151f3c8..cc61a083954d 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -207,6 +207,17 @@ static unsigned long *alloc_thread_stack_node(struct
task_struct *tsk, int node)
struct vm_struct *stack;
int i;
+#ifdef CONFIG_VMAP_STACK_AS_FALLBACK
+ struct page *page;
+
+ page = alloc_pages_node(node, THREADINFO_GFP & ~__GFP_DIRECT_RECLAIM,
+ THREAD_SIZE_ORDER);
+ if (page) {
+ tsk->stack_vm_area = NULL;
+ return page_address(page);
+ }
+#endif
+
for (i = 0; i < NR_CACHED_STACKS; i++) {
stack = this_cpu_xchg(cached_stacks[i], NULL);
if (!stack)
