s390/spinlock: do not yield to a CPU in udelay/mdelay

It does not make sense to try to relinquish the time slice with diag 0x9c
to a CPU in a state that does not allow to schedule the CPU. The scenario
where this can happen is a CPU waiting in udelay/mdelay while holding a
spin-lock.

Add a CIF bit to tag a CPU in enabled wait and use it to detect that the
yield of a CPU will not be successful and skip the diagnose call.
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/processor.h

@@ -18,12 +18,14 @@
 #define CIF_NOHZ_DELAY		2	/* delay HZ disable for a tick */
 #define CIF_FPU			3	/* restore FPU registers */
 #define CIF_IGNORE_IRQ		4	/* ignore interrupt (for udelay) */
+#define CIF_ENABLED_WAIT	5	/* in enabled wait state */
 
 #define _CIF_MCCK_PENDING	_BITUL(CIF_MCCK_PENDING)
 #define _CIF_ASCE		_BITUL(CIF_ASCE)
 #define _CIF_NOHZ_DELAY		_BITUL(CIF_NOHZ_DELAY)
 #define _CIF_FPU		_BITUL(CIF_FPU)
 #define _CIF_IGNORE_IRQ		_BITUL(CIF_IGNORE_IRQ)
+#define _CIF_ENABLED_WAIT	_BITUL(CIF_ENABLED_WAIT)
 
 #ifndef __ASSEMBLY__
 
@@ -52,6 +54,16 @@ static inline int test_cpu_flag(int flag)
 	return !!(S390_lowcore.cpu_flags & (1UL << flag));
 }
 
+/*
+ * Test CIF flag of another CPU. The caller needs to ensure that
+ * CPU hotplug can not happen, e.g. by disabling preemption.
+ */
+static inline int test_cpu_flag_of(int flag, int cpu)
+{
+	struct _lowcore *lc = lowcore_ptr[cpu];
+	return !!(lc->cpu_flags & (1UL << flag));
+}
+
 #define arch_needs_cpu() test_cpu_flag(CIF_NOHZ_DELAY)
 
 /*

arch/s390/kernel/entry.S

@@ -764,6 +764,7 @@ ENTRY(psw_idle)
 	.insn	rsy,0xeb0000000017,%r1,5,__SF_EMPTY+16(%r15)
 .Lpsw_idle_stcctm:
 #endif
+	oi	__LC_CPU_FLAGS+7,_CIF_ENABLED_WAIT
 	STCK	__CLOCK_IDLE_ENTER(%r2)
 	stpt	__TIMER_IDLE_ENTER(%r2)
 .Lpsw_idle_lpsw:
@@ -1146,6 +1147,7 @@ cleanup_critical:
 	.quad	.Lio_done - 4
 
 .Lcleanup_idle:
+	ni	__LC_CPU_FLAGS+7,255-_CIF_ENABLED_WAIT
 	# copy interrupt clock & cpu timer
 	mvc	__CLOCK_IDLE_EXIT(8,%r2),__LC_INT_CLOCK
 	mvc	__TIMER_IDLE_EXIT(8,%r2),__LC_ASYNC_ENTER_TIMER

arch/s390/lib/spinlock.c

@@ -37,6 +37,15 @@ static inline void _raw_compare_and_delay(unsigned int *lock, unsigned int old)
 	asm(".insn rsy,0xeb0000000022,%0,0,%1" : : "d" (old), "Q" (*lock));
 }
 
+static inline int cpu_is_preempted(int cpu)
+{
+	if (test_cpu_flag_of(CIF_ENABLED_WAIT, cpu))
+		return 0;
+	if (smp_vcpu_scheduled(cpu))
+		return 0;
+	return 1;
+}
+
 void arch_spin_lock_wait(arch_spinlock_t *lp)
 {
 	unsigned int cpu = SPINLOCK_LOCKVAL;
@@ -53,7 +62,7 @@ void arch_spin_lock_wait(arch_spinlock_t *lp)
 			continue;
 		}
 		/* First iteration: check if the lock owner is running. */
-		if (first_diag && !smp_vcpu_scheduled(~owner)) {
+		if (first_diag && cpu_is_preempted(~owner)) {
 			smp_yield_cpu(~owner);
 			first_diag = 0;
 			continue;
@@ -72,7 +81,7 @@ void arch_spin_lock_wait(arch_spinlock_t *lp)
 		 * yield the CPU unconditionally. For LPAR rely on the
 		 * sense running status.
 		 */
-		if (!MACHINE_IS_LPAR || !smp_vcpu_scheduled(~owner)) {
+		if (!MACHINE_IS_LPAR || cpu_is_preempted(~owner)) {
 			smp_yield_cpu(~owner);
 			first_diag = 0;
 		}
@@ -98,7 +107,7 @@ void arch_spin_lock_wait_flags(arch_spinlock_t *lp, unsigned long flags)
 			local_irq_restore(flags);
 		}
 		/* Check if the lock owner is running. */
-		if (first_diag && !smp_vcpu_scheduled(~owner)) {
+		if (first_diag && cpu_is_preempted(~owner)) {
 			smp_yield_cpu(~owner);
 			first_diag = 0;
 			continue;
@@ -117,7 +126,7 @@ void arch_spin_lock_wait_flags(arch_spinlock_t *lp, unsigned long flags)
 		 * yield the CPU unconditionally. For LPAR rely on the
 		 * sense running status.
 		 */
-		if (!MACHINE_IS_LPAR || !smp_vcpu_scheduled(~owner)) {
+		if (!MACHINE_IS_LPAR || cpu_is_preempted(~owner)) {
 			smp_yield_cpu(~owner);
 			first_diag = 0;
 		}
@@ -155,7 +164,7 @@ void _raw_read_lock_wait(arch_rwlock_t *rw)
 	owner = 0;
 	while (1) {
 		if (count-- <= 0) {
-			if (owner && !smp_vcpu_scheduled(~owner))
+			if (owner && cpu_is_preempted(~owner))
 				smp_yield_cpu(~owner);
 			count = spin_retry;
 		}
@@ -201,7 +210,7 @@ void _raw_write_lock_wait(arch_rwlock_t *rw, unsigned int prev)
 	owner = 0;
 	while (1) {
 		if (count-- <= 0) {
-			if (owner && !smp_vcpu_scheduled(~owner))
+			if (owner && cpu_is_preempted(~owner))
 				smp_yield_cpu(~owner);
 			count = spin_retry;
 		}
@@ -231,7 +240,7 @@ void _raw_write_lock_wait(arch_rwlock_t *rw)
 	owner = 0;
 	while (1) {
 		if (count-- <= 0) {
-			if (owner && !smp_vcpu_scheduled(~owner))
+			if (owner && cpu_is_preempted(~owner))
 				smp_yield_cpu(~owner);
 			count = spin_retry;
 		}
@@ -275,7 +284,7 @@ void arch_lock_relax(unsigned int cpu)
 {
 	if (!cpu)
 		return;
-	if (MACHINE_IS_LPAR && smp_vcpu_scheduled(~cpu))
+	if (MACHINE_IS_LPAR && !cpu_is_preempted(~cpu))
 		return;
 	smp_yield_cpu(~cpu);
 }