[PATCH mm 1/2] kfence: add option to use KFENCE without static keys

[Marco Elver]

For certain usecases, specifically where the sample interval is always
set to a very low value such as 1ms, it can make sense to use a dynamic
branch instead of static branches due to the overhead of toggling a
static branch.

Therefore, add a new Kconfig option to remove the static branches and
instead check kfence_allocation_gate if a KFENCE allocation should be
set up.

Suggested-by: Jörn Engel <joern@xxxxxxxxxxxxxxx>
Signed-off-by: Marco Elver <elver@xxxxxxxxxx>
---
 include/linux/kfence.h | 11 ++++++++++-
 lib/Kconfig.kfence     | 12 +++++++++++-
 mm/kfence/core.c       | 32 ++++++++++++++++++--------------
 3 files changed, 39 insertions(+), 16 deletions(-)

diff --git a/include/linux/kfence.h b/include/linux/kfence.h
index 76246889ecdb..dc86b69d3903 100644
--- a/include/linux/kfence.h
+++ b/include/linux/kfence.h
@@ -4,7 +4,6 @@
 #define _LINUX_KFENCE_H
 
 #include <linux/mm.h>
-#include <linux/static_key.h>
 #include <linux/types.h>
 
 #ifdef CONFIG_KFENCE
@@ -17,7 +16,13 @@
 #define KFENCE_POOL_SIZE ((CONFIG_KFENCE_NUM_OBJECTS + 1) * 2 * PAGE_SIZE)
 extern char *__kfence_pool;
 
+#ifdef CONFIG_KFENCE_STATIC_KEYS
+#include <linux/static_key.h>
 DECLARE_STATIC_KEY_FALSE(kfence_allocation_key);
+#else
+#include <linux/atomic.h>
+extern atomic_t kfence_allocation_gate;
+#endif
 
 /**
  * is_kfence_address() - check if an address belongs to KFENCE pool
@@ -104,7 +109,11 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags);
  */
 static __always_inline void *kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
 {
+#ifdef CONFIG_KFENCE_STATIC_KEYS
 	if (static_branch_unlikely(&kfence_allocation_key))
+#else
+	if (unlikely(!atomic_read(&kfence_allocation_gate)))
+#endif
 		return __kfence_alloc(s, size, flags);
 	return NULL;
 }
diff --git a/lib/Kconfig.kfence b/lib/Kconfig.kfence
index d3ea24fa30fc..78f50ccb3b45 100644
--- a/lib/Kconfig.kfence
+++ b/lib/Kconfig.kfence
@@ -6,7 +6,6 @@ config HAVE_ARCH_KFENCE
 menuconfig KFENCE
 	bool "KFENCE: low-overhead sampling-based memory safety error detector"
 	depends on HAVE_ARCH_KFENCE && (SLAB || SLUB)
-	depends on JUMP_LABEL # To ensure performance, require jump labels
 	select STACKTRACE
 	help
 	  KFENCE is a low-overhead sampling-based detector of heap out-of-bounds
@@ -25,6 +24,17 @@ menuconfig KFENCE
 
 if KFENCE
 
+config KFENCE_STATIC_KEYS
+	bool "Use static keys to set up allocations"
+	default y
+	depends on JUMP_LABEL # To ensure performance, require jump labels
+	help
+	  Use static keys (static branches) to set up KFENCE allocations. Using
+	  static keys is normally recommended, because it avoids a dynamic
+	  branch in the allocator's fast path. However, with very low sample
+	  intervals, or on systems that do not support jump labels, a dynamic
+	  branch may still be an acceptable performance trade-off.
+
 config KFENCE_SAMPLE_INTERVAL
 	int "Default sample interval in milliseconds"
 	default 100
diff --git a/mm/kfence/core.c b/mm/kfence/core.c
index f0816d5f5913..96a9a98e7453 100644
--- a/mm/kfence/core.c
+++ b/mm/kfence/core.c
@@ -88,11 +88,13 @@ struct kfence_metadata kfence_metadata[CONFIG_KFENCE_NUM_OBJECTS];
 static struct list_head kfence_freelist = LIST_HEAD_INIT(kfence_freelist);
 static DEFINE_RAW_SPINLOCK(kfence_freelist_lock); /* Lock protecting freelist. */
 
+#ifdef CONFIG_KFENCE_STATIC_KEYS
 /* The static key to set up a KFENCE allocation. */
 DEFINE_STATIC_KEY_FALSE(kfence_allocation_key);
+#endif
 
 /* Gates the allocation, ensuring only one succeeds in a given period. */
-static atomic_t allocation_gate = ATOMIC_INIT(1);
+atomic_t kfence_allocation_gate = ATOMIC_INIT(1);
 
 /* Statistics counters for debugfs. */
 enum kfence_counter_id {
@@ -583,29 +585,31 @@ late_initcall(kfence_debugfs_init);
 static struct delayed_work kfence_timer;
 static void toggle_allocation_gate(struct work_struct *work)
 {
-	unsigned long end_wait;
-
 	if (!READ_ONCE(kfence_enabled))
 		return;
 
 	/* Enable static key, and await allocation to happen. */
-	atomic_set(&allocation_gate, 0);
+	atomic_set(&kfence_allocation_gate, 0);
+#ifdef CONFIG_KFENCE_STATIC_KEYS
 	static_branch_enable(&kfence_allocation_key);
 	/*
 	 * Await an allocation. Timeout after 1 second, in case the kernel stops
 	 * doing allocations, to avoid stalling this worker task for too long.
 	 */
-	end_wait = jiffies + HZ;
-	do {
-		set_current_state(TASK_UNINTERRUPTIBLE);
-		if (atomic_read(&allocation_gate) != 0)
-			break;
-		schedule_timeout(1);
-	} while (time_before(jiffies, end_wait));
-	__set_current_state(TASK_RUNNING);
-
+	{
+		unsigned long end_wait = jiffies + HZ;
+
+		do {
+			set_current_state(TASK_UNINTERRUPTIBLE);
+			if (atomic_read(&kfence_allocation_gate) != 0)
+				break;
+			schedule_timeout(1);
+		} while (time_before(jiffies, end_wait));
+		__set_current_state(TASK_RUNNING);
+	}
 	/* Disable static key and reset timer. */
 	static_branch_disable(&kfence_allocation_key);
+#endif
 	schedule_delayed_work(&kfence_timer, msecs_to_jiffies(kfence_sample_interval));
 }
 static DECLARE_DELAYED_WORK(kfence_timer, toggle_allocation_gate);
@@ -711,7 +715,7 @@ void *__kfence_alloc(struct kmem_cache *s, size_t size, gfp_t flags)
 	 * sense to continue writing to it and pay the associated contention
 	 * cost, in case we have a large number of concurrent allocations.
 	 */
-	if (atomic_read(&allocation_gate) || atomic_inc_return(&allocation_gate) > 1)
+	if (atomic_read(&kfence_allocation_gate) || atomic_inc_return(&kfence_allocation_gate) > 1)
 		return NULL;
 
 	if (!READ_ONCE(kfence_enabled))
-- 
2.30.0.284.gd98b1dd5eaa7-goog