[PATCH] futex: make FUTEX_OP_OPARG_SHIFT unsigned

[PATCH] futex: make FUTEX_OP_OPARG_SHIFT unsigned

[Jiri Slaby]

FUTEX_OP_OPARG_SHIFT is shifted left by 28 in futex_atomic_op_inuser and
other places. Since FUTEX_OP_OPARG_SHIFT is 8, we actually do (8 << 28).
But given both are ints, we perform operation with undefined behaviour
about which UBSAN reports:
UBSAN: Undefined behaviour in arch/x86/include/asm/futex.h:53:13
shift exponent 776 is too large for 32-bit type 'int'
CPU: 1 PID: 27469 Comm: trinity-c1 Tainted: G            EL     4.4.26-0-default #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.9.1-0-gb3ef39f-prebuilt.qemu-project.org 04/01/2014
 0000000000000000 ffff880553ad7908 ffffffff81cbcc16 0000000041b58ab3
 ffffffff82f515a3 ffffffff81cbcb16 ffff880553ad7930 ffff880553ad78d0
 ffffffff837d7d60 0000000000000000 0000000000000308 ffffffff837d7d62
Call Trace:
 [<ffffffff81d9ec50>] ? __ubsan_handle_shift_out_of_bounds+0x29c/0x300
 [<ffffffff81d9e9b4>] ? __ubsan_handle_load_invalid_value+0x1ba/0x1ba
 [<ffffffff814202a9>] ? get_futex_key+0x4f9/0xf80
 [<ffffffff81422456>] ? futex_wake_op+0x636/0xfc0
 [<ffffffff8141fdb0>] ? futex_lock_pi_atomic+0x2a0/0x2a0
 [<ffffffff81422b09>] ? futex_wake_op+0xce9/0xfc0
 [<ffffffff81421e20>] ? futex_wait_restart+0x240/0x240
 [<ffffffff814266ae>] ? do_futex+0x54e/0xb60
 [<ffffffff81426df8>] ? SyS_futex+0x138/0x2b0

Fix that by having FUTEX_OP_OPARG_SHIFT unsigned.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
---
 include/uapi/linux/futex.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h
index 0b1f716373c7..954a23b18e68 100644
--- a/include/uapi/linux/futex.h
+++ b/include/uapi/linux/futex.h
@@ -130,7 +130,7 @@ struct robust_list_head {
 #define FUTEX_OP_ANDN		3	/* *(int *)UADDR2 &= ~OPARG; */
 #define FUTEX_OP_XOR		4	/* *(int *)UADDR2 ^= OPARG; */
 
-#define FUTEX_OP_OPARG_SHIFT	8	/* Use (1 << OPARG) instead of OPARG.  */
+#define FUTEX_OP_OPARG_SHIFT	8U	/* Use (1 << OPARG) instead of OPARG.  */
 
 #define FUTEX_OP_CMP_EQ		0	/* if (oldval == CMPARG) wake */
 #define FUTEX_OP_CMP_NE		1	/* if (oldval != CMPARG) wake */
-- 
2.10.1

Re: [PATCH] futex: make FUTEX_OP_OPARG_SHIFT unsigned

[Thomas Gleixner]

On Mon, 24 Oct 2016, Jiri Slaby wrote:

> FUTEX_OP_OPARG_SHIFT is shifted left by 28 in futex_atomic_op_inuser and
> other places. Since FUTEX_OP_OPARG_SHIFT is 8, we actually do (8 << 28).
> But given both are ints, we perform operation with undefined behaviour
> about which UBSAN reports:
> UBSAN: Undefined behaviour in arch/x86/include/asm/futex.h:53:13
> shift exponent 776 is too large for 32-bit type 'int'

shift exponent 776 ????

Thanks,
	tglx