Re: [PATCH] mm: kfence: Improve the performance of __kfence_alloc() and __kfence_free()
From: Peng Zhang
Date: Mon Apr 03 2023 - 07:11:33 EST
在 2023/4/3 17:21, Marco Elver 写道:
On Mon, 3 Apr 2023 at 08:28, Peng Zhang <zhangpeng.00@xxxxxxxxxxxxx> wrote:
In __kfence_alloc() and __kfence_free(), we will set and check canary.
Assuming that the size of the object is close to 0, nearly 4k memory
accesses are required because setting and checking canary is executed
byte by byte.
canary is now defined like this:
KFENCE_CANARY_PATTERN(addr) ((u8)0xaa ^ (u8)((unsigned long)(addr) & 0x7))
Observe that canary is only related to the lower three bits of the
address, so every 8 bytes of canary are the same. We can access 8-byte
canary each time instead of byte-by-byte, thereby optimizing nearly 4k
memory accesses to 4k/8 times.
Use the bcc tool funclatency to measure the latency of __kfence_alloc()
and __kfence_free(), the numbers (deleted the distribution of latency)
is posted below. Though different object sizes will have an impact on the
measurement, we ignore it for now and assume the average object size is
roughly equal.
Before playing patch:
__kfence_alloc:
avg = 5055 nsecs, total: 5515252 nsecs, count: 1091
__kfence_free:
avg = 5319 nsecs, total: 9735130 nsecs, count: 1830
After playing patch:
__kfence_alloc:
avg = 3597 nsecs, total: 6428491 nsecs, count: 1787
__kfence_free:
avg = 3046 nsecs, total: 3415390 nsecs, count: 1121
Seems like a nice improvement!
The numbers indicate that there is ~30% - ~40% performance improvement.
Signed-off-by: Peng Zhang <zhangpeng.00@xxxxxxxxxxxxx>
---
mm/kfence/core.c | 71 +++++++++++++++++++++++++++++++++-------------
mm/kfence/kfence.h | 10 ++++++-
mm/kfence/report.c | 2 +-
3 files changed, 62 insertions(+), 21 deletions(-)
diff --git a/mm/kfence/core.c b/mm/kfence/core.c
index 79c94ee55f97..0b1b1298c738 100644
--- a/mm/kfence/core.c
+++ b/mm/kfence/core.c
@@ -297,20 +297,13 @@ metadata_update_state(struct kfence_metadata *meta, enum kfence_object_state nex
WRITE_ONCE(meta->state, next);
}
-/* Write canary byte to @addr. */
-static inline bool set_canary_byte(u8 *addr)
-{
- *addr = KFENCE_CANARY_PATTERN(addr);
- return true;
-}
-
/* Check canary byte at @addr. */
static inline bool check_canary_byte(u8 *addr)
{
struct kfence_metadata *meta;
unsigned long flags;
- if (likely(*addr == KFENCE_CANARY_PATTERN(addr)))
+ if (likely(*addr == KFENCE_CANARY_PATTERN_U8(addr)))
return true;
atomic_long_inc(&counters[KFENCE_COUNTER_BUGS]);
@@ -323,11 +316,27 @@ static inline bool check_canary_byte(u8 *addr)
return false;
}
-/* __always_inline this to ensure we won't do an indirect call to fn. */
-static __always_inline void for_each_canary(const struct kfence_metadata *meta, bool (*fn)(u8 *))
+static inline void set_canary(const struct kfence_metadata *meta)
{
const unsigned long pageaddr = ALIGN_DOWN(meta->addr, PAGE_SIZE);
- unsigned long addr;
+ unsigned long addr = pageaddr;
+
+ /*
+ * The canary may be written to part of the object memory, but it does
+ * not affect it. The user should initialize the object before using it.
+ */
+ for (; addr < meta->addr; addr += sizeof(u64))
+ *((u64 *)addr) = KFENCE_CANARY_PATTERN_U64;
+
+ addr = ALIGN_DOWN(meta->addr + meta->size, sizeof(u64));
+ for (; addr - pageaddr < PAGE_SIZE; addr += sizeof(u64))
+ *((u64 *)addr) = KFENCE_CANARY_PATTERN_U64;
+}
+
+static inline void check_canary(const struct kfence_metadata *meta)
+{
+ const unsigned long pageaddr = ALIGN_DOWN(meta->addr, PAGE_SIZE);
+ unsigned long addr = pageaddr;
/*
* We'll iterate over each canary byte per-side until fn() returns
This comment is now out-of-date ("fn" no longer exists).
@@ -339,14 +348,38 @@ static __always_inline void for_each_canary(const struct kfence_metadata *meta,
*/
/* Apply to left of object. */
- for (addr = pageaddr; addr < meta->addr; addr++) {
- if (!fn((u8 *)addr))
+ for (; meta->addr - addr >= sizeof(u64); addr += sizeof(u64)) {
+ if (unlikely(*((u64 *)addr) != KFENCE_CANARY_PATTERN_U64))
break;
}
- /* Apply to right of object. */
- for (addr = meta->addr + meta->size; addr < pageaddr + PAGE_SIZE; addr++) {
- if (!fn((u8 *)addr))
+ /*
+ * If the canary is damaged in a certain 64 bytes, or the canay memory
"damaged" -> "corrupted"
"canay" -> "canary"
+ * cannot be completely covered by multiple consecutive 64 bytes, it
+ * needs to be checked one by one.
+ */
+ for (; addr < meta->addr; addr++) {
+ if (unlikely(!check_canary_byte((u8 *)addr)))
+ break;
+ }
+
+ /*
+ * Apply to right of object.
+ * For easier implementation, check from high address to low address.
+ */
+ addr = pageaddr + PAGE_SIZE - sizeof(u64);
+ for (; addr >= meta->addr + meta->size ; addr -= sizeof(u64)) {
+ if (unlikely(*((u64 *)addr) != KFENCE_CANARY_PATTERN_U64))
+ break;
+ }
+
+ /*
+ * Same as above, checking byte by byte, but here is the reverse of
+ * the above.
+ */
+ addr = addr + sizeof(u64) - 1;
+ for (; addr >= meta->addr + meta->size; addr--) {
The re-checking should forward-check i.e. not in reverse, otherwise
the report might not include some corrupted bytes that had in the
previous version been included. I think you need to check from low to
high address to start with above.
Yes, it's better to forward-check to avoid losing the corrupted bytes
which be used in report.
I will include all your suggestions in the next version of the patch.
Thanks.
+ if (unlikely(!check_canary_byte((u8 *)addr)))
break;
}
}
@@ -434,7 +467,7 @@ static void *kfence_guarded_alloc(struct kmem_cache *cache, size_t size, gfp_t g
#endif
/* Memory initialization. */
- for_each_canary(meta, set_canary_byte);
+ set_canary(meta);
/*
* We check slab_want_init_on_alloc() ourselves, rather than letting
@@ -495,7 +528,7 @@ static void kfence_guarded_free(void *addr, struct kfence_metadata *meta, bool z
alloc_covered_add(meta->alloc_stack_hash, -1);
/* Check canary bytes for memory corruption. */
- for_each_canary(meta, check_canary_byte);
+ check_canary(meta);
/*
* Clear memory if init-on-free is set. While we protect the page, the
@@ -751,7 +784,7 @@ static void kfence_check_all_canary(void)
struct kfence_metadata *meta = &kfence_metadata[i];
if (meta->state == KFENCE_OBJECT_ALLOCATED)
- for_each_canary(meta, check_canary_byte);
+ check_canary(meta);
}
}
diff --git a/mm/kfence/kfence.h b/mm/kfence/kfence.h
index 600f2e2431d6..2aafc46a4aaf 100644
--- a/mm/kfence/kfence.h
+++ b/mm/kfence/kfence.h
@@ -21,7 +21,15 @@
* lower 3 bits of the address, to detect memory corruptions with higher
* probability, where similar constants are used.
*/
-#define KFENCE_CANARY_PATTERN(addr) ((u8)0xaa ^ (u8)((unsigned long)(addr) & 0x7))
+#define KFENCE_CANARY_PATTERN_U8(addr) ((u8)0xaa ^ (u8)((unsigned long)(addr) & 0x7))
+
+/*
+ * Define a continuous 8-byte canary starting from a multiple of 8. The canary
+ * of each byte is only related to the lowest three bits of its address, so the
+ * canary of every 8 bytes is the same. 64-bit memory can be filled and checked
+ * at a time instead of byte by byte to improve performance.
+ */
+#define KFENCE_CANARY_PATTERN_U64 ((u64)0xaaaaaaaaaaaaaaaa ^ (u64)(0x0706050403020100))
/* Maximum stack depth for reports. */
#define KFENCE_STACK_DEPTH 64
diff --git a/mm/kfence/report.c b/mm/kfence/report.c
index 60205f1257ef..197430a5be4a 100644
--- a/mm/kfence/report.c
+++ b/mm/kfence/report.c
@@ -168,7 +168,7 @@ static void print_diff_canary(unsigned long address, size_t bytes_to_show,
pr_cont("[");
for (cur = (const u8 *)address; cur < end; cur++) {
- if (*cur == KFENCE_CANARY_PATTERN(cur))
+ if (*cur == KFENCE_CANARY_PATTERN_U8(cur))
pr_cont(" .");
else if (no_hash_pointers)
pr_cont(" 0x%02x", *cur);
--
2.20.1