OSDN Git Service

lib/ubsan: remove null-pointer checks
authorAndrey Ryabinin <aryabinin@virtuozzo.com>
Sat, 11 Aug 2018 00:23:03 +0000 (17:23 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 11 Aug 2018 03:19:58 +0000 (20:19 -0700)
With gcc-8 fsanitize=null become very noisy.  GCC started to complain
about things like &a->b, where 'a' is NULL pointer.  There is no NULL
dereference, we just calculate address to struct member.  It's
technically undefined behavior so UBSAN is correct to report it.  But as
long as there is no real NULL-dereference, I think, we should be fine.

-fno-delete-null-pointer-checks compiler flag should protect us from any
consequences.  So let's just no use -fsanitize=null as it's not useful
for us.  If there is a real NULL-deref we will see crash.  Even if
userspace mapped something at NULL (root can do this), with things like
SMAP should catch the issue.

Link: http://lkml.kernel.org/r/20180802153209.813-1-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/parisc/Kconfig
arch/s390/Kconfig
lib/Kconfig.ubsan
scripts/Makefile.ubsan

index 46f656b..e7705dd 100644 (file)
@@ -11,7 +11,6 @@ config PARISC
        select ARCH_HAS_ELF_RANDOMIZE
        select ARCH_HAS_STRICT_KERNEL_RWX
        select ARCH_HAS_UBSAN_SANITIZE_ALL
-       select ARCH_WANTS_UBSAN_NO_NULL
        select ARCH_SUPPORTS_MEMORY_FAILURE
        select RTC_CLASS
        select RTC_DRV_GENERIC
index 8a1863d..4fe5b2a 100644 (file)
@@ -106,7 +106,6 @@ config S390
        select ARCH_USE_BUILTIN_BSWAP
        select ARCH_USE_CMPXCHG_LOCKREF
        select ARCH_WANTS_DYNAMIC_TASK_STRUCT
-       select ARCH_WANTS_UBSAN_NO_NULL
        select ARCH_WANT_IPC_PARSE_VERSION
        select BUILDTIME_EXTABLE_SORT
        select CLONE_BACKWARDS2
index 19d42ea..98fa559 100644 (file)
@@ -1,9 +1,6 @@
 config ARCH_HAS_UBSAN_SANITIZE_ALL
        bool
 
-config ARCH_WANTS_UBSAN_NO_NULL
-       def_bool n
-
 config UBSAN
        bool "Undefined behaviour sanity checker"
        help
@@ -39,14 +36,6 @@ config UBSAN_ALIGNMENT
          Enabling this option on architectures that support unaligned
          accesses may produce a lot of false positives.
 
-config UBSAN_NULL
-       bool "Enable checking of null pointers"
-       depends on UBSAN
-       default y if !ARCH_WANTS_UBSAN_NO_NULL
-       help
-         This option enables detection of memory accesses via a
-         null pointer.
-
 config TEST_UBSAN
        tristate "Module for testing for undefined behavior detection"
        depends on m && UBSAN
index b593b36..38b2b48 100644 (file)
@@ -14,10 +14,6 @@ ifdef CONFIG_UBSAN_ALIGNMENT
       CFLAGS_UBSAN += $(call cc-option, -fsanitize=alignment)
 endif
 
-ifdef CONFIG_UBSAN_NULL
-      CFLAGS_UBSAN += $(call cc-option, -fsanitize=null)
-endif
-
       # -fsanitize=* options makes GCC less smart than usual and
       # increase number of 'maybe-uninitialized false-positives
       CFLAGS_UBSAN += $(call cc-option, -Wno-maybe-uninitialized)