arm64: futex: Bound number of LDXR/STXR loops in FUTEX_WAKE_OP

Our futex implementation makes use of LDXR/STXR loops to perform atomic
updates to user memory from atomic context. This can lead to latency
problems if we end up spinning around the LL/SC sequence at the expense
of doing something useful.

Rework our futex atomic operations so that we return -EAGAIN if we fail
to update the futex word after 128 attempts. The core futex code will
reschedule if necessary and we'll try again later.

Cc: <stable@kernel.org>
Fixes: 6170a974 ("arm64: Atomic operations")
Signed-off-by: Will Deacon <will.deacon@arm.com>

arch/arm64/include/asm/futex.h

@@ -23,26 +23,34 @@
 
 #include <asm/errno.h>
 
+#define FUTEX_MAX_LOOPS	128 /* What's the largest number you can think of? */
+
 #define __futex_atomic_op(insn, ret, oldval, uaddr, tmp, oparg)		\
 do {									\
+	unsigned int loops = FUTEX_MAX_LOOPS;				\
+									\
 	uaccess_enable();						\
 	asm volatile(							\
 "	prfm	pstl1strm, %2\n"					\
 "1:	ldxr	%w1, %2\n"						\
 	insn "\n"							\
 "2:	stlxr	%w0, %w3, %2\n"						\
-"	cbnz	%w0, 1b\n"						\
-"	dmb	ish\n"							\
+"	cbz	%w0, 3f\n"						\
+"	sub	%w4, %w4, %w0\n"					\
+"	cbnz	%w4, 1b\n"						\
+"	mov	%w0, %w7\n"						\
 "3:\n"									\
+"	dmb	ish\n"							\
 "	.pushsection .fixup,\"ax\"\n"					\
 "	.align	2\n"							\
-"4:	mov	%w0, %w5\n"						\
+"4:	mov	%w0, %w6\n"						\
 "	b	3b\n"							\
 "	.popsection\n"							\
 	_ASM_EXTABLE(1b, 4b)						\
 	_ASM_EXTABLE(2b, 4b)						\
-	: "=&r" (ret), "=&r" (oldval), "+Q" (*uaddr), "=&r" (tmp)	\
-	: "r" (oparg), "Ir" (-EFAULT)					\
+	: "=&r" (ret), "=&r" (oldval), "+Q" (*uaddr), "=&r" (tmp),	\
+	  "+r" (loops)							\
+	: "r" (oparg), "Ir" (-EFAULT), "Ir" (-EAGAIN)			\
 	: "memory");							\
 	uaccess_disable();						\
 } while (0)
@@ -57,23 +65,23 @@ arch_futex_atomic_op_inuser(int op, int oparg, int *oval, u32 __user *_uaddr)
 
 	switch (op) {
 	case FUTEX_OP_SET:
-		__futex_atomic_op("mov	%w3, %w4",
+		__futex_atomic_op("mov	%w3, %w5",
 				  ret, oldval, uaddr, tmp, oparg);
 		break;
 	case FUTEX_OP_ADD:
-		__futex_atomic_op("add	%w3, %w1, %w4",
+		__futex_atomic_op("add	%w3, %w1, %w5",
 				  ret, oldval, uaddr, tmp, oparg);
 		break;
 	case FUTEX_OP_OR:
-		__futex_atomic_op("orr	%w3, %w1, %w4",
+		__futex_atomic_op("orr	%w3, %w1, %w5",
 				  ret, oldval, uaddr, tmp, oparg);
 		break;
 	case FUTEX_OP_ANDN:
-		__futex_atomic_op("and	%w3, %w1, %w4",
+		__futex_atomic_op("and	%w3, %w1, %w5",
 				  ret, oldval, uaddr, tmp, ~oparg);
 		break;
 	case FUTEX_OP_XOR:
-		__futex_atomic_op("eor	%w3, %w1, %w4",
+		__futex_atomic_op("eor	%w3, %w1, %w5",
 				  ret, oldval, uaddr, tmp, oparg);
 		break;
 	default:
@@ -93,6 +101,7 @@ futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *_uaddr,
 			      u32 oldval, u32 newval)
 {
 	int ret = 0;
+	unsigned int loops = FUTEX_MAX_LOOPS;
 	u32 val, tmp;
 	u32 __user *uaddr;
 
@@ -104,20 +113,24 @@ futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *_uaddr,
 	asm volatile("// futex_atomic_cmpxchg_inatomic\n"
 "	prfm	pstl1strm, %2\n"
 "1:	ldxr	%w1, %2\n"
-"	sub	%w3, %w1, %w4\n"
-"	cbnz	%w3, 3f\n"
-"2:	stlxr	%w3, %w5, %2\n"
-"	cbnz	%w3, 1b\n"
-"	dmb	ish\n"
+"	sub	%w3, %w1, %w5\n"
+"	cbnz	%w3, 4f\n"
+"2:	stlxr	%w3, %w6, %2\n"
+"	cbz	%w3, 3f\n"
+"	sub	%w4, %w4, %w3\n"
+"	cbnz	%w4, 1b\n"
+"	mov	%w0, %w8\n"
 "3:\n"
+"	dmb	ish\n"
+"4:\n"
 "	.pushsection .fixup,\"ax\"\n"
-"4:	mov	%w0, %w6\n"
-"	b	3b\n"
+"5:	mov	%w0, %w7\n"
+"	b	4b\n"
 "	.popsection\n"
-	_ASM_EXTABLE(1b, 4b)
-	_ASM_EXTABLE(2b, 4b)
-	: "+r" (ret), "=&r" (val), "+Q" (*uaddr), "=&r" (tmp)
-	: "r" (oldval), "r" (newval), "Ir" (-EFAULT)
+	_ASM_EXTABLE(1b, 5b)
+	_ASM_EXTABLE(2b, 5b)
+	: "+r" (ret), "=&r" (val), "+Q" (*uaddr), "=&r" (tmp), "+r" (loops)
+	: "r" (oldval), "r" (newval), "Ir" (-EFAULT), "Ir" (-EAGAIN)
 	: "memory");
 	uaccess_disable();