The value of object->pointer could be accessed concurrently as noticed
by KCSAN,
write to 0xffffb0ea683a7d50 of 4 bytes by task 23575 on cpu 12:
do_raw_spin_lock+0x114/0x200
debug_spin_lock_after at kernel/locking/spinlock_debug.c:91
(inlined by) do_raw_spin_lock at kernel/locking/spinlock_debug.c:115
_raw_spin_lock+0x40/0x50
__handle_mm_fault+0xa9e/0xd00
handle_mm_fault+0xfc/0x2f0
do_page_fault+0x263/0x6f9
page_fault+0x34/0x40
read to 0xffffb0ea683a7d50 of 4 bytes by task 839 on cpu 60:
crc32_le_base+0x67/0x350
crc32_le_base+0x67/0x350:
crc32_body at lib/crc32.c:106
(inlined by) crc32_le_generic at lib/crc32.c:179
(inlined by) crc32_le at lib/crc32.c:197
kmemleak_scan+0x528/0xd90
update_checksum at mm/kmemleak.c:1172
(inlined by) kmemleak_scan at mm/kmemleak.c:1497
kmemleak_scan_thread+0xcc/0xfa
kthread+0x1e0/0x200
ret_from_fork+0x27/0x50
write to 0xffff939bf07b95b8 of 4 bytes by interrupt on cpu 119:
__free_object+0x884/0xcb0
__free_object at lib/debugobjects.c:359
__debug_check_no_obj_freed+0x19d/0x370
debug_check_no_obj_freed+0x41/0x4b
slab_free_freelist_hook+0xfb/0x1c0
kmem_cache_free+0x10c/0x3a0
free_object_rcu+0x1ca/0x260
rcu_core+0x677/0xcc0
rcu_core_si+0x17/0x20
__do_softirq+0xd9/0x57c
run_ksoftirqd+0x29/0x50
smpboot_thread_fn+0x222/0x3f0
kthread+0x1e0/0x200
ret_from_fork+0x27/0x50
read to 0xffff939bf07b95b8 of 8 bytes by task 838 on cpu 109:
scan_block+0x69/0x190
scan_block at mm/kmemleak.c:1250
kmemleak_scan+0x249/0xd90
scan_large_block at mm/kmemleak.c:1309
(inlined by) kmemleak_scan at mm/kmemleak.c:1434
kmemleak_scan_thread+0xcc/0xfa
kthread+0x1e0/0x200
ret_from_fork+0x27/0x50
crc32() will dereference object->pointer. If a shattered value was
returned due to a data race, it will be corrected in the next scan.
scan_block() will dereference a range of addresses (e.g., percpu
sections) to search for valid pointers. Even if a data race heppens, it
will cause no issue because the code here does not care about the exact
value of a non-pointer. Thus, mark them as intentional data races using
the data_race() macro.
Signed-off-by: Qian Cai <[email protected]>
---
v2: add a missing annotation.
mm/kmemleak.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/mm/kmemleak.c b/mm/kmemleak.c
index 3a4259eeb5a0..aa6832432d6a 100644
--- a/mm/kmemleak.c
+++ b/mm/kmemleak.c
@@ -1169,7 +1169,12 @@ static bool update_checksum(struct kmemleak_object *object)
u32 old_csum = object->checksum;
kasan_disable_current();
- object->checksum = crc32(0, (void *)object->pointer, object->size);
+ /*
+ * crc32() will dereference object->pointer. If an unstable value was
+ * returned due to a data race, it will be corrected in the next scan.
+ */
+ object->checksum = data_race(crc32(0, (void *)object->pointer,
+ object->size));
kasan_enable_current();
return object->checksum != old_csum;
@@ -1243,7 +1248,7 @@ static void scan_block(void *_start, void *_end,
break;
kasan_disable_current();
- pointer = *ptr;
+ pointer = data_race(*ptr);
kasan_enable_current();
untagged_ptr = (unsigned long)kasan_reset_tag((void *)pointer);
--
1.8.3.1
On Thu, Feb 13, 2020 at 12:36:30PM -0500, Qian Cai wrote:
> The value of object->pointer could be accessed concurrently as noticed
> by KCSAN,
>
> write to 0xffffb0ea683a7d50 of 4 bytes by task 23575 on cpu 12:
> do_raw_spin_lock+0x114/0x200
> debug_spin_lock_after at kernel/locking/spinlock_debug.c:91
> (inlined by) do_raw_spin_lock at kernel/locking/spinlock_debug.c:115
> _raw_spin_lock+0x40/0x50
> __handle_mm_fault+0xa9e/0xd00
> handle_mm_fault+0xfc/0x2f0
> do_page_fault+0x263/0x6f9
> page_fault+0x34/0x40
>
> read to 0xffffb0ea683a7d50 of 4 bytes by task 839 on cpu 60:
> crc32_le_base+0x67/0x350
> crc32_le_base+0x67/0x350:
> crc32_body at lib/crc32.c:106
> (inlined by) crc32_le_generic at lib/crc32.c:179
> (inlined by) crc32_le at lib/crc32.c:197
> kmemleak_scan+0x528/0xd90
> update_checksum at mm/kmemleak.c:1172
> (inlined by) kmemleak_scan at mm/kmemleak.c:1497
> kmemleak_scan_thread+0xcc/0xfa
> kthread+0x1e0/0x200
> ret_from_fork+0x27/0x50
>
> write to 0xffff939bf07b95b8 of 4 bytes by interrupt on cpu 119:
> __free_object+0x884/0xcb0
> __free_object at lib/debugobjects.c:359
> __debug_check_no_obj_freed+0x19d/0x370
> debug_check_no_obj_freed+0x41/0x4b
> slab_free_freelist_hook+0xfb/0x1c0
> kmem_cache_free+0x10c/0x3a0
> free_object_rcu+0x1ca/0x260
> rcu_core+0x677/0xcc0
> rcu_core_si+0x17/0x20
> __do_softirq+0xd9/0x57c
> run_ksoftirqd+0x29/0x50
> smpboot_thread_fn+0x222/0x3f0
> kthread+0x1e0/0x200
> ret_from_fork+0x27/0x50
>
> read to 0xffff939bf07b95b8 of 8 bytes by task 838 on cpu 109:
> scan_block+0x69/0x190
> scan_block at mm/kmemleak.c:1250
> kmemleak_scan+0x249/0xd90
> scan_large_block at mm/kmemleak.c:1309
> (inlined by) kmemleak_scan at mm/kmemleak.c:1434
> kmemleak_scan_thread+0xcc/0xfa
> kthread+0x1e0/0x200
> ret_from_fork+0x27/0x50
>
> crc32() will dereference object->pointer. If a shattered value was
> returned due to a data race, it will be corrected in the next scan.
> scan_block() will dereference a range of addresses (e.g., percpu
> sections) to search for valid pointers. Even if a data race heppens, it
> will cause no issue because the code here does not care about the exact
> value of a non-pointer. Thus, mark them as intentional data races using
> the data_race() macro.
>
> Signed-off-by: Qian Cai <[email protected]>
Acked-by: Catalin Marinas <[email protected]>