[v3,1/2] scs: switch to vmapped shadow stacks

Commit Message

Sami Tolvanen Nov. 30, 2020, 11:34 p.m. UTC

The kernel currently uses kmem_cache to allocate shadow call stacks,
which means an overflows may not be immediately detected and can
potentially result in another task's shadow stack to be overwritten.

This change switches SCS to use virtually mapped shadow stacks for
tasks, which increases shadow stack size to a full page and provides
more robust overflow detection, similarly to VMAP_STACK.

Signed-off-by: Sami Tolvanen <[e-mail]>
Acked-by: Will Deacon <[e-mail]>
---
 include/linux/scs.h | 12 ++++----
 kernel/scs.c        | 71 ++++++++++++++++++++++++++++++++++++++-------
 2 files changed, 66 insertions(+), 17 deletions(-)

Patch

diff --git a/include/linux/scs.h b/include/linux/scs.h
index 6dec390cf154..2a506c2a16f4 100644
--- a/include/linux/scs.h
+++ b/include/linux/scs.h
@@ -15,12 +15,8 @@ 
 
 #ifdef CONFIG_SHADOW_CALL_STACK
 
-/*
- * In testing, 1 KiB shadow stack size (i.e. 128 stack frames on a 64-bit
- * architecture) provided ~40% safety margin on stack usage while keeping
- * memory allocation overhead reasonable.
- */
-#define SCS_SIZE		SZ_1K
+#define SCS_ORDER		0
+#define SCS_SIZE		(PAGE_SIZE << SCS_ORDER)
 #define GFP_SCS			(GFP_KERNEL | __GFP_ZERO)
 
 /* An illegal pointer value to mark the end of the shadow stack. */
@@ -33,6 +29,8 @@ 
 #define task_scs(tsk)		(task_thread_info(tsk)->scs_base)
 #define task_scs_sp(tsk)	(task_thread_info(tsk)->scs_sp)
 
+void *scs_alloc(int node);
+void scs_free(void *s);
 void scs_init(void);
 int scs_prepare(struct task_struct *tsk, int node);
 void scs_release(struct task_struct *tsk);
@@ -61,6 +59,8 @@  static inline bool task_scs_end_corrupted(struct task_struct *tsk)
 
 #else /* CONFIG_SHADOW_CALL_STACK */
 
+static inline void *scs_alloc(int node) { return NULL; }
+static inline void scs_free(void *s) {}
 static inline void scs_init(void) {}
 static inline void scs_task_reset(struct task_struct *tsk) {}
 static inline int scs_prepare(struct task_struct *tsk, int node) { return 0; }
diff --git a/kernel/scs.c b/kernel/scs.c
index 4ff4a7ba0094..e2a71fc82fa0 100644
--- a/kernel/scs.c
+++ b/kernel/scs.c
@@ -5,26 +5,49 @@ 
  * Copyright (C) 2019 Google LLC
  */
 
+#include <linux/cpuhotplug.h>
 #include <linux/kasan.h>
 #include <linux/mm.h>
 #include <linux/scs.h>
-#include <linux/slab.h>
+#include <linux/vmalloc.h>
 #include <linux/vmstat.h>
 
-static struct kmem_cache *scs_cache;
-
 static void __scs_account(void *s, int account)
 {
-	struct page *scs_page = virt_to_page(s);
+	struct page *scs_page = vmalloc_to_page(s);
 
 	mod_node_page_state(page_pgdat(scs_page), NR_KERNEL_SCS_KB,
 			    account * (SCS_SIZE / SZ_1K));
 }
 
-static void *scs_alloc(int node)
+/* Matches NR_CACHED_STACKS for VMAP_STACK */
+#define NR_CACHED_SCS 2
+static DEFINE_PER_CPU(void *, scs_cache[NR_CACHED_SCS]);
+
+static void *__scs_alloc(int node)
 {
-	void *s = kmem_cache_alloc_node(scs_cache, GFP_SCS, node);
+	int i;
+	void *s;
+
+	for (i = 0; i < NR_CACHED_SCS; i++) {
+		s = this_cpu_xchg(scs_cache[i], NULL);
+		if (s) {
+			kasan_unpoison_vmalloc(s, SCS_SIZE);
+			memset(s, 0, SCS_SIZE);
+			return s;
+		}
+	}
+
+	return __vmalloc_node_range(SCS_SIZE, 1, VMALLOC_START, VMALLOC_END,
+				    GFP_SCS, PAGE_KERNEL, 0, node,
+				    __builtin_return_address(0));
+}
 
+void *scs_alloc(int node)
+{
+	void *s;
+
+	s = __scs_alloc(node);
 	if (!s)
 		return NULL;
 
@@ -34,21 +57,47 @@  static void *scs_alloc(int node)
 	 * Poison the allocation to catch unintentional accesses to
 	 * the shadow stack when KASAN is enabled.
 	 */
-	kasan_poison_object_data(scs_cache, s);
+	kasan_poison_vmalloc(s, SCS_SIZE);
 	__scs_account(s, 1);
 	return s;
 }
 
-static void scs_free(void *s)
+void scs_free(void *s)
 {
+	int i;
+
 	__scs_account(s, -1);
-	kasan_unpoison_object_data(scs_cache, s);
-	kmem_cache_free(scs_cache, s);
+
+	/*
+	 * We cannot sleep as this can be called in interrupt context,
+	 * so use this_cpu_cmpxchg to update the cache, and vfree_atomic
+	 * to free the stack.
+	 */
+
+	for (i = 0; i < NR_CACHED_SCS; i++)
+		if (this_cpu_cmpxchg(scs_cache[i], 0, s) == NULL)
+			return;
+
+	vfree_atomic(s);
+}
+
+static int scs_cleanup(unsigned int cpu)
+{
+	int i;
+	void **cache = per_cpu_ptr(scs_cache, cpu);
+
+	for (i = 0; i < NR_CACHED_SCS; i++) {
+		vfree(cache[i]);
+		cache[i] = NULL;
+	}
+
+	return 0;
 }
 
 void __init scs_init(void)
 {
-	scs_cache = kmem_cache_create("scs_cache", SCS_SIZE, 0, 0, NULL);
+	cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "scs:scs_cache", NULL,
+			  scs_cleanup);
 }
 
 int scs_prepare(struct task_struct *tsk, int node)