Overhaul of the SMP code. Several portions of the SMP kernel support have

been made machine independent and various other adjustments have been made
to support Alpha SMP.

- It splits the per-process portions of hardclock() and statclock() off
  into hardclock_process() and statclock_process() respectively.  hardclock()
  and statclock() call the *_process() functions for the current process so
  that UP systems will run as before.  For SMP systems, it is simply necessary
  to ensure that all other processors execute the *_process() functions when the
  main clock functions are triggered on one CPU by an interrupt.  For the alpha
  4100, clock interrupts are delievered in a staggered broadcast fashion, so
  we simply call hardclock/statclock on the boot CPU and call the *_process()
  functions on the secondaries.  For x86, we call statclock and hardclock as
  usual and then call forward_hardclock/statclock in the MD code to send an IPI
  to cause the AP's to execute forwared_hardclock/statclock which then call the
  *_process() functions.
- forward_signal() and forward_roundrobin() have been reworked to be MI and to
  involve less hackery.  Now the cpu doing the forward sets any flags, etc. and
  sends a very simple IPI_AST to the other cpu(s).  AST IPIs now just basically
  return so that they can execute ast() and don't bother with setting the
  astpending or needresched flags themselves.  This also removes the loop in
  forward_signal() as sched_lock closes the race condition that the loop worked
  around.
- need_resched(), resched_wanted() and clear_resched() have been changed to take
  a process to act on rather than assuming curproc so that they can be used to
  implement forward_roundrobin() as described above.
- Various other SMP variables have been moved to a MI subr_smp.c and a new
  header sys/smp.h declares MI SMP variables and API's.   The IPI API's from
  machine/ipl.h have moved to machine/smp.h which is included by sys/smp.h.
- The globaldata_register() and globaldata_find() functions as well as the
  SLIST of globaldata structures has become MI and moved into subr_smp.c.
  Also, the globaldata list is only available if SMP support is compiled in.

Reviewed by:	jake, peter
Looked over by:	eivind

9 files changed, 37 insertions, 348 deletions

diff --git a/sys/ia64/ia64/db_interface.c b/sys/ia64/ia64/db_interface.c
index 3750189..7069f60 100644
--- a/sys/ia64/ia64/db_interface.c
+++ b/sys/ia64/ia64/db_interface.c
@@ -44,6 +44,7 @@
 #include <sys/reboot.h>
 #include <sys/systm.h>
 #include <sys/kernel.h>
+#include <sys/smp.h>
 #include <sys/cons.h>
 #include <sys/ktr.h>
 
@@ -51,7 +52,6 @@
 
 #include <machine/db_machdep.h>
 #include <machine/mutex.h>
-#include <machine/smp.h>
 
 #include <machine/inst.h>
 
diff --git a/sys/ia64/ia64/machdep.c b/sys/ia64/ia64/machdep.c
index a113c9b..9eac9ed 100644
--- a/sys/ia64/ia64/machdep.c
+++ b/sys/ia64/ia64/machdep.c
@@ -82,8 +82,6 @@ u_int32_t cycles_per_sec;
 int cold = 1;
 struct bootinfo_kernel bootinfo;
 
-struct cpuhead cpuhead;
-
 struct mtx sched_lock;
 struct mtx Giant;
 
@@ -579,12 +577,6 @@ ia64_init()
 	proc0.p_md.md_tf =
 	    (struct trapframe *)(proc0.p_addr->u_pcb.pcb_sp + 16);
 
-	/*
-	 * Record all cpus in a list.
-	 */
-	SLIST_INIT(&cpuhead);
-	SLIST_INSERT_HEAD(&cpuhead, GLOBALP, gd_allcpu);
-
 	/* Setup curproc so that mutexes work */
 	PCPU_SET(curproc, &proc0);
 	PCPU_SET(spinlocks, NULL);
@@ -1367,3 +1359,16 @@ ia64_fpstate_switch(struct proc *p)
 
 	p->p_md.md_flags |= MDP_FPUSED;
 }
+
+/*
+ * Initialise a struct globaldata.
+ */
+void
+globaldata_init(struct globaldata *globaldata, int cpuid, size_t sz)
+{
+	bzero(globaldata, sz);
+	globaldata->gd_cpuid = cpuid;
+#ifdef SMP
+	globaldata_register(globaldata);
+#endif
+}
diff --git a/sys/ia64/ia64/mp_machdep.c b/sys/ia64/ia64/mp_machdep.c
index db73f61..143963a 100644
--- a/sys/ia64/ia64/mp_machdep.c
+++ b/sys/ia64/ia64/mp_machdep.c
@@ -34,6 +34,7 @@
 #include <sys/malloc.h>
 #include <sys/mutex.h>
 #include <sys/kernel.h>
+#include <sys/smp.h>
 #include <sys/sysctl.h>
 
 #include <vm/vm.h>
@@ -42,7 +43,6 @@
 #include <sys/user.h>
 #include <sys/dkstat.h>
 
-#include <machine/smp.h>
 #include <machine/atomic.h>
 #include <machine/ipl.h>
 #include <machine/globaldata.h>
@@ -53,8 +53,6 @@
 #define CHECKSTATE_SYS	1
 #define CHECKSTATE_INTR	2
 
-volatile u_int		stopped_cpus;
-volatile u_int		started_cpus;
 volatile u_int		checkstate_probed_cpus;
 volatile u_int		checkstate_need_ast;
 volatile u_int		checkstate_pending_ast;
@@ -62,76 +60,27 @@ struct proc*		checkstate_curproc[MAXCPU];
 int			checkstate_cpustate[MAXCPU];
 u_long			checkstate_pc[MAXCPU];
 volatile u_int		resched_cpus;
-void (*cpustop_restartfunc) __P((void));
-int			mp_ncpus;
 
-int			smp_started;
 int			boot_cpu_id;
-u_int32_t		all_cpus;
-
-static struct globaldata	*cpuid_to_globaldata[MAXCPU];
-
-int smp_active = 0;	/* are the APs allowed to run? */
-SYSCTL_INT(_machdep, OID_AUTO, smp_active, CTLFLAG_RW, &smp_active, 0, "");
 
 /* Is forwarding of a interrupt to the CPU holding the ISR lock enabled ? */
 int forward_irq_enabled = 1;
 SYSCTL_INT(_machdep, OID_AUTO, forward_irq_enabled, CTLFLAG_RW,
 	   &forward_irq_enabled, 0, "");
 
-/* Enable forwarding of a signal to a process running on a different CPU */
-static int forward_signal_enabled = 1;
-SYSCTL_INT(_machdep, OID_AUTO, forward_signal_enabled, CTLFLAG_RW,
-	   &forward_signal_enabled, 0, "");
-
-/* Enable forwarding of roundrobin to all other cpus */
-static int forward_roundrobin_enabled = 1;
-SYSCTL_INT(_machdep, OID_AUTO, forward_roundrobin_enabled, CTLFLAG_RW,
-	   &forward_roundrobin_enabled, 0, "");
-
-/*
- * Initialise a struct globaldata.
- */
-void
-globaldata_init(struct globaldata *globaldata, int cpuid, size_t sz)
-{
-	bzero(globaldata, sz);
-	globaldata->gd_cpuid = cpuid;
-	globaldata->gd_other_cpus = all_cpus & ~(1 << cpuid);
-	cpuid_to_globaldata[cpuid] = globaldata;
-}
-
-struct globaldata *
-globaldata_find(int cpuid)
-{
-	return cpuid_to_globaldata[cpuid];
-}
-
-/* Other stuff */
-
-/* lock around the MP rendezvous */
-static struct mtx smp_rv_mtx;
-
-static void
-init_locks(void)
-{
-
-	mtx_init(&smp_rv_mtx, "smp_rendezvous", MTX_SPIN);
-}
-
-void
-mp_start()
+int
+cpu_mp_probe()
 {
-	init_locks();
+	return (0);
 }
 
 void
-mp_announce()
+cpu_mp_start()
 {
 }
 
 void
-smp_invltlb()
+cpu_mp_announce()
 {
 }
 
@@ -426,227 +375,6 @@ forward_hardclock(int pscnt)
 	}
 }
 
-void
-forward_signal(struct proc *p)
-{
-	int map;
-	int id;
-	int i;
-
-	/* Kludge. We don't yet have separate locks for the interrupts
-	 * and the kernel. This means that we cannot let the other processors
-	 * handle complex interrupts while inhibiting them from entering
-	 * the kernel in a non-interrupt context.
-	 *
-	 * What we can do, without changing the locking mechanisms yet,
-	 * is letting the other processors handle a very simple interrupt
-	 * (wich determines the processor states), and do the main
-	 * work ourself.
-	 */
-
-	CTR1(KTR_SMP, "forward_signal(%p)", p);
-
-	if (!smp_started || cold || panicstr)
-		return;
-	if (!forward_signal_enabled)
-		return;
-	while (1) {
-		if (p->p_stat != SRUN)
-			return;
-		id = p->p_oncpu;
-		if (id == 0xff)
-			return;
-		map = (1<<id);
-		checkstate_need_ast |= map;
-		ipi_selected(map, IPI_AST);
-		i = 0;
-		while ((checkstate_need_ast & map) != 0) {
-			/* spin */
-			i++;
-			if (i > 100000) { 
-#if 0
-				printf("forward_signal: dropped ast 0x%x\n",
-				       checkstate_need_ast & map);
-#endif
-				break;
-			}
-		}
-		if (id == p->p_oncpu)
-			return;
-	}
-}
-
-void
-forward_roundrobin(void)
-{
-	u_int map;
-	int i;
-
-	CTR0(KTR_SMP, "forward_roundrobin()");
-
-	if (!smp_started || cold || panicstr)
-		return;
-	if (!forward_roundrobin_enabled)
-		return;
-	resched_cpus |= PCPU_GET(other_cpus);
-	map = PCPU_GET(other_cpus) & ~stopped_cpus ;
-	ipi_selected(map, IPI_AST);
-	i = 0;
-	while ((checkstate_need_ast & map) != 0) {
-		/* spin */
-		i++;
-		if (i > 100000) {
-#if 0
-			printf("forward_roundrobin: dropped ast 0x%x\n",
-			       checkstate_need_ast & map);
-#endif
-			break;
-		}
-	}
-}
-
-/*
- * When called the executing CPU will send an IPI to all other CPUs
- *  requesting that they halt execution.
- *
- * Usually (but not necessarily) called with 'other_cpus' as its arg.
- *
- *  - Signals all CPUs in map to stop.
- *  - Waits for each to stop.
- *
- * Returns:
- *  -1: error
- *   0: NA
- *   1: ok
- *
- * XXX FIXME: this is not MP-safe, needs a lock to prevent multiple CPUs
- *            from executing at same time.
- */
-int
-stop_cpus(u_int map)
-{
-	int i;
-
-	if (!smp_started)
-		return 0;
-
-	CTR1(KTR_SMP, "stop_cpus(%x)", map);
-
-	/* send the stop IPI to all CPUs in map */
-	ipi_selected(map, IPI_STOP);
-	
-	i = 0;
-	while ((stopped_cpus & map) != map) {
-		/* spin */
-		i++;
-		if (i == 100000) {
-			printf("timeout stopping cpus\n");
-			break;
-		}
-		ia64_mf();
-	}
-
-	printf("stopped_cpus=%x\n", stopped_cpus);
-
-	return 1;
-}
-
-
-/*
- * Called by a CPU to restart stopped CPUs. 
- *
- * Usually (but not necessarily) called with 'stopped_cpus' as its arg.
- *
- *  - Signals all CPUs in map to restart.
- *  - Waits for each to restart.
- *
- * Returns:
- *  -1: error
- *   0: NA
- *   1: ok
- */
-int
-restart_cpus(u_int map)
-{
-	if (!smp_started)
-		return 0;
-
-	CTR1(KTR_SMP, "restart_cpus(%x)", map);
-
-	started_cpus = map;		/* signal other cpus to restart */
-	ia64_mf();
-
-	while ((stopped_cpus & map) != 0) /* wait for each to clear its bit */
-		ia64_mf();
-
-	return 1;
-}
-
-/*
- * All-CPU rendezvous.  CPUs are signalled, all execute the setup function 
- * (if specified), rendezvous, execute the action function (if specified),
- * rendezvous again, execute the teardown function (if specified), and then
- * resume.
- *
- * Note that the supplied external functions _must_ be reentrant and aware
- * that they are running in parallel and in an unknown lock context.
- */
-static void (*smp_rv_setup_func)(void *arg);
-static void (*smp_rv_action_func)(void *arg);
-static void (*smp_rv_teardown_func)(void *arg);
-static void *smp_rv_func_arg;
-static volatile int smp_rv_waiters[2];
-
-void
-smp_rendezvous_action(void)
-{
-	/* setup function */
-	if (smp_rv_setup_func != NULL)
-		smp_rv_setup_func(smp_rv_func_arg);
-	/* spin on entry rendezvous */
-	atomic_add_int(&smp_rv_waiters[0], 1);
-	while (smp_rv_waiters[0] < mp_ncpus)
-		;
-	/* action function */
-	if (smp_rv_action_func != NULL)
-		smp_rv_action_func(smp_rv_func_arg);
-	/* spin on exit rendezvous */
-	atomic_add_int(&smp_rv_waiters[1], 1);
-	while (smp_rv_waiters[1] < mp_ncpus)
-		;
-	/* teardown function */
-	if (smp_rv_teardown_func != NULL)
-		smp_rv_teardown_func(smp_rv_func_arg);
-}
-
-void
-smp_rendezvous(void (* setup_func)(void *), 
-	       void (* action_func)(void *),
-	       void (* teardown_func)(void *),
-	       void *arg)
-{
-
-	/* obtain rendezvous lock */
-	mtx_lock_spin(&smp_rv_mtx);
-
-	/* set static function pointers */
-	smp_rv_setup_func = setup_func;
-	smp_rv_action_func = action_func;
-	smp_rv_teardown_func = teardown_func;
-	smp_rv_func_arg = arg;
-	smp_rv_waiters[0] = 0;
-	smp_rv_waiters[1] = 0;
-
-	/* signal other processors, which will enter the IPI with interrupts off */
-	ipi_all_but_self(IPI_RENDEZVOUS);
-
-	/* call executor function */
-	smp_rendezvous_action();
-
-	/* release lock */
-	mtx_unlock_spin(&smp_rv_mtx);
-}
-
 /*
  * send an IPI to a set of cpus.
  */
@@ -661,7 +389,7 @@ ipi_selected(u_int32_t cpus, u_int64_t ipi)
 		int cpuid = ffs(cpus) - 1;
 		cpus &= ~(1 << cpuid);
 
-		globaldata = cpuid_to_globaldata[cpuid];
+		globaldata = globaldata_find(cpuid);
 		if (globaldata) {
 			atomic_set_64(&globaldata->gd_pending_ipis, ipi);
 			ia64_mf();
@@ -733,9 +461,6 @@ smp_handle_ipi(struct trapframe *frame)
 			CTR0(KTR_SMP, "IPI_AST");
 			atomic_clear_int(&checkstate_need_ast, 1<<cpuid);
 			atomic_set_int(&checkstate_pending_ast, 1<<cpuid);
-			if ((frame->tf_cr_ipsr & IA64_PSR_CPL)
-			    == IA64_PSR_CPL_USER)
-				ast(frame); /* XXX */
 			break;
 
 		case IPI_CHECKSTATE:
diff --git a/sys/ia64/ia64/trap.c b/sys/ia64/ia64/trap.c
index 8486877..8f9b258 100644
--- a/sys/ia64/ia64/trap.c
+++ b/sys/ia64/ia64/trap.c
@@ -41,6 +41,7 @@
 #include <sys/exec.h>
 #include <sys/lock.h>
 #include <sys/mutex.h>
+#include <sys/smp.h>
 #include <sys/vmmeter.h>
 #include <sys/sysent.h>
 #include <sys/syscall.h>
@@ -58,7 +59,6 @@
 #include <machine/reg.h>
 #include <machine/pal.h>
 #include <machine/fpu.h>
-#include <machine/smp.h>
 
 #ifdef KTRACE
 #include <sys/uio.h>
@@ -90,7 +90,7 @@ userret(register struct proc *p, struct trapframe *frame, u_quad_t oticks)
 
 	mtx_lock_spin(&sched_lock);
 	p->p_pri.pri_level = p->p_pri.pri_user;
-	if (resched_wanted()) {
+	if (resched_wanted(p)) {
 		/*
 		 * Since we are curproc, a clock interrupt could
 		 * change our priority without changing run queues
@@ -638,7 +638,7 @@ ast(framep)
 	 * acquiring and release mutexes in assembly is not fun.
 	 */
 	mtx_lock_spin(&sched_lock);
-	if (!(astpending(p) || resched_wanted())) {
+	if (!(astpending(p) || resched_wanted(p))) {
 		mtx_unlock_spin(&sched_lock);
 		return;
 	}
diff --git a/sys/ia64/include/cpu.h b/sys/ia64/include/cpu.h
index 2b15a48..61766d2 100644
--- a/sys/ia64/include/cpu.h
+++ b/sys/ia64/include/cpu.h
@@ -68,9 +68,6 @@ struct clockframe {
 
 #define	CLKF_USERMODE(framep)	TRAPF_USERMODE(&(framep)->cf_tf)
 #define	CLKF_PC(framep)		TRAPF_PC(&(framep)->cf_tf)
-#define	CLKF_BASEPRI(framep)						\
-	(((framep)->cf_tf.tf_cr_ipsr & IA64_PSR_I) == 0)
-#define	CLKF_INTR(framep)	(curproc->p_intr_nesting_level >= 2)
 
 /*
  * Give a profiling tick to the current process when the user profiling
diff --git a/sys/ia64/include/globaldata.h b/sys/ia64/include/globaldata.h
index a92f543..0ec1e19 100644
--- a/sys/ia64/include/globaldata.h
+++ b/sys/ia64/include/globaldata.h
@@ -66,11 +66,7 @@ struct globaldata {
 #endif
 };
 
-SLIST_HEAD(cpuhead, globaldata);
-extern struct cpuhead cpuhead;
-
 void globaldata_init(struct globaldata *pcpu, int cpuid, size_t sz);
-struct globaldata *globaldata_find(int cpuid);
 
 #endif	/* _KERNEL */
 
diff --git a/sys/ia64/include/ipl.h b/sys/ia64/include/ipl.h
index 8432299..cbdecbb 100644
--- a/sys/ia64/include/ipl.h
+++ b/sys/ia64/include/ipl.h
@@ -29,19 +29,4 @@
 #ifndef _MACHINE_IPL_H_
 #define	_MACHINE_IPL_H_
 
-/*
- * Interprocessor interrupts for SMP.
- */
-#define IPI_INVLTLB		0x0001
-#define IPI_RENDEZVOUS		0x0002
-#define IPI_AST			0x0004
-#define IPI_CHECKSTATE		0x0008
-#define IPI_STOP		0x0010
-
-void ipi_selected(u_int32_t cpus, u_int64_t ipi);
-void ipi_all(u_int64_t ipi);
-void ipi_all_but_self(u_int64_t ipi);
-void ipi_self(u_int64_t ipi);
-void smp_handle_ipi(struct trapframe *frame);
-
 #endif /* !_MACHINE_IPL_H_ */
diff --git a/sys/ia64/include/pcpu.h b/sys/ia64/include/pcpu.h
index a92f543..0ec1e19 100644
--- a/sys/ia64/include/pcpu.h
+++ b/sys/ia64/include/pcpu.h
@@ -66,11 +66,7 @@ struct globaldata {
 #endif
 };
 
-SLIST_HEAD(cpuhead, globaldata);
-extern struct cpuhead cpuhead;
-
 void globaldata_init(struct globaldata *pcpu, int cpuid, size_t sz);
-struct globaldata *globaldata_find(int cpuid);
 
 #endif	/* _KERNEL */
 
diff --git a/sys/ia64/include/smp.h b/sys/ia64/include/smp.h
index 65c85ef..e4f2091 100644
--- a/sys/ia64/include/smp.h
+++ b/sys/ia64/include/smp.h
@@ -6,43 +6,28 @@
 
 #ifdef _KERNEL
 
-#include <machine/mutex.h>
-#include <machine/ipl.h>
-#include <sys/ktr.h>
+/*
+ * Interprocessor interrupts for SMP.
+ */
+#define	IPI_INVLTLB		0x0001
+#define	IPI_RENDEZVOUS		0x0002
+#define	IPI_AST			0x0004
+#define	IPI_CHECKSTATE		0x0008
+#define	IPI_STOP		0x0010
 
 #ifndef LOCORE
 
-#define BETTER_CLOCK		/* unconditional on ia64 */
-
 /* global data in mp_machdep.c */
 extern volatile u_int		checkstate_probed_cpus;
 extern volatile u_int		checkstate_need_ast;
 extern volatile u_int		resched_cpus;
-extern void (*cpustop_restartfunc) __P((void));
-
-extern int			smp_active;
-extern int			mp_ncpus;
-extern u_int			all_cpus;
-extern u_int			started_cpus;
-extern u_int			stopped_cpus;
 
-/* functions in mp_machdep.c */
-void	mp_start(void);
-void	mp_announce(void);
-void	smp_invltlb(void);
-void	forward_statclock(int pscnt);
-void	forward_hardclock(int pscnt);
-void	forward_signal(struct proc *);
-void	forward_roundrobin(void);
-int	stop_cpus(u_int);
-int	restart_cpus(u_int);
-void	smp_rendezvous_action(void);
-void	smp_rendezvous(void (*)(void *), 
-		       void (*)(void *),
-		       void (*)(void *),
-		       void *arg);
+void	ipi_selected(u_int cpus, u_int ipi);
+void	ipi_all(u_int ipi);
+void	ipi_all_but_self(u_int ipi);
+void	ipi_self(u_int ipi);
 void	smp_init_secondary(void);
 
 #endif /* !LOCORE */
 #endif /* _KERNEL */
-#endif
+#endif /* !_MACHINE_SMP_H */