Add a KPI for the delay while spinning on a spin lock.

Replace a call to DELAY(1) with a new cpu_lock_delay() KPI.  Currently
cpu_lock_delay() is defined to DELAY(1) on all platforms.  However,
platforms with a DELAY() implementation that uses spin locks should
implement a custom cpu_lock_delay() doesn't use locks.

Reviewed by:	kib
MFC after:	3 days

sys/amd64/include/cpu.h

@@ -50,6 +50,7 @@
 #define	cpu_getstack(td)		((td)->td_frame->tf_rsp)
 #define	cpu_setstack(td, ap)		((td)->td_frame->tf_rsp = (ap))
 #define	cpu_spinwait()			ia32_pause()
+#define	cpu_lock_delay()		DELAY(1)
 
 #define	TRAPF_USERMODE(framep) \
 	(ISPL((framep)->tf_cs) == SEL_UPL)

sys/arm/include/cpu.h

@@ -61,6 +61,7 @@ get_cyclecount(void)
 #define cpu_getstack(td)	((td)->td_frame->tf_usr_sp)
 #define cpu_setstack(td, sp)	((td)->td_frame->tf_usr_sp = (sp))
 #define cpu_spinwait()		/* nothing */
+#define	cpu_lock_delay()	DELAY(1)
 
 #define ARM_NVEC		8
 #define ARM_VEC_ALL		0xffffffff

sys/arm64/include/cpu.h

@@ -51,6 +51,7 @@
 #define	cpu_getstack(td)	((td)->td_frame->tf_sp)
 #define	cpu_setstack(td, sp)	((td)->td_frame->tf_sp = (sp))
 #define	cpu_spinwait()		__asm __volatile("yield" ::: "memory")
+#define	cpu_lock_delay()	DELAY(1)
 
 /* Extract CPU affinity levels 0-3 */
 #define	CPU_AFF0(mpidr)	(u_int)(((mpidr) >> 0) & 0xff)

sys/i386/include/cpu.h

@@ -50,6 +50,7 @@
 #define	cpu_getstack(td)		((td)->td_frame->tf_esp)
 #define	cpu_setstack(td, ap)		((td)->td_frame->tf_esp = (ap))
 #define	cpu_spinwait()			ia32_pause()
+#define	cpu_lock_delay()		DELAY(1)
 
 #define	TRAPF_USERMODE(framep) \
 	((ISPL((framep)->tf_cs) == SEL_UPL) || ((framep)->tf_eflags & PSL_VM))

sys/kern/kern_mutex.c

@@ -1206,7 +1206,7 @@ _mtx_lock_indefinite_check(struct mtx *m, struct lock_delay_arg *ldap)
 
 	ldap->spin_cnt++;
 	if (ldap->spin_cnt < 60000000 || kdb_active || panicstr != NULL)
-		DELAY(1);
+		cpu_lock_delay();
 	else {
 		td = mtx_owner(m);
 

sys/mips/include/cpu.h

@@ -71,6 +71,7 @@
 #define	cpu_getstack(td)	((td)->td_frame->sp)
 #define	cpu_setstack(td, nsp)	((td)->td_frame->sp = (nsp))
 #define	cpu_spinwait()		/* nothing */
+#define	cpu_lock_delay()	DELAY(1)
 
 /*
  * A machine-independent interface to the CPU's counter.

sys/powerpc/include/cpu.h

@@ -128,6 +128,7 @@ get_cyclecount(void)
 
 #define	cpu_getstack(td)	((td)->td_frame->fixreg[1])
 #define	cpu_spinwait()		__asm __volatile("or 27,27,27") /* yield */
+#define	cpu_lock_delay()	DELAY(1)
 
 extern char btext[];
 extern char etext[];

sys/riscv/include/cpu.h

@@ -46,6 +46,7 @@
 #define	cpu_getstack(td)	((td)->td_frame->tf_sp)
 #define	cpu_setstack(td, sp)	((td)->td_frame->tf_sp = (sp))
 #define	cpu_spinwait()		/* nothing */
+#define	cpu_lock_delay()	DELAY(1)
 
 #ifdef _KERNEL
 

sys/sparc64/include/cpu.h

@@ -48,6 +48,7 @@
 #define	cpu_getstack(td)	((td)->td_frame->tf_sp)
 #define	cpu_setstack(td, sp)	((td)->td_frame->tf_sp = (sp))
 #define	cpu_spinwait()		/* nothing */
+#define	cpu_lock_delay()	DELAY(1)
 
 #ifdef _KERNEL