On 11/26/25 17:16, John Paul Adrian Glaubitz wrote:
Hi Helge,
On Wed, 2025-11-26 at 12:31 +0100, Helge Deller wrote:
Like this (untested!) patch:
[PATCH] apparmor: Optimize table creation from possibly unaligned memory
Source blob may come from userspace and might be unaligned.
Try to optize the copying process by avoiding unaligned memory accesses.
Signed-off-by: Helge Deller <[email protected]>
diff --git a/security/apparmor/include/match.h
b/security/apparmor/include/match.h
index 1fbe82f5021b..225df6495c84 100644
--- a/security/apparmor/include/match.h
+++ b/security/apparmor/include/match.h
@@ -111,9 +111,14 @@ struct aa_dfa {
typeof(LEN) __i; \
TTYPE *__t = (TTYPE *) TABLE; \
BTYPE *__b = (BTYPE *) BLOB; \
- for (__i = 0; __i < LEN; __i++) { \
- __t[__i] = NTOHX(__b[__i]); \
- } \
+ BUILD_BUG_ON(sizeof(TTYPE) != sizeof(BTYPE)); \
+ /* copy to naturally aligned table address */ \
+ memcpy(__t, __b, (LEN) * sizeof(BTYPE)); \
+ /* convert from big-endian if necessary */ \
+ if (!IS_ENABLED(CONFIG_CPU_BIG_ENDIAN)) \
+ for (__i = 0; __i < LEN; __i++, __t++) { \
+ *__t = NTOHX(*__t); \
+ } \
} while (0)
static inline size_t table_size(size_t len, size_t el_size)
So, is this patch supposed to replace all the other proposed patches?
It just replaces the patch from John.
But please test the v2 patch I sent instead...
Helge
