Import dt(4) a driver and framework for Dynamic Profiling.

The design is fairly simple: events, in the form of descriptors on a
ring, are being produced in any kernel context and being consumed by
a userland process reading /dev/dt.

Code and hooks are all guarded under '#if NDT > 0' so this commit
shouldn't introduce any change as long as dt(4) is disable in GENERIC.

ok kettenis@, visa@, jasper@, deraadt@

4 files changed, 28 insertions, 4 deletions

diff --git a/sys/kern/kern_clock.c b/sys/kern/kern_clock.c
index 0a3679b0510..b6e0ca4f65a 100644
--- a/sys/kern/kern_clock.c
+++ b/sys/kern/kern_clock.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_clock.c,v 1.100 2019/11/02 16:56:17 cheloha Exp $	*/
+/*	$OpenBSD: kern_clock.c,v 1.101 2020/01/21 16:16:23 mpi Exp $	*/
 /*	$NetBSD: kern_clock.c,v 1.34 1996/06/09 04:51:03 briggs Exp $	*/
 
 /*-
@@ -55,6 +55,11 @@
 #include <sys/gmon.h>
 #endif
 
+#include "dt.h"
+#if NDT > 0
+#include <dev/dt/dtvar.h>
+#endif
+
 /*
  * Clock handling routines.
  *
@@ -168,6 +173,12 @@ hardclock(struct clockframe *frame)
 	if (--ci->ci_schedstate.spc_rrticks <= 0)
 		roundrobin(ci);
 
+#if NDT > 0
+	DT_ENTER(profile, NULL);
+	if (CPU_IS_PRIMARY(ci))
+		DT_ENTER(interval, NULL);
+#endif
+
 	/*
 	 * If we are not the primary CPU, we're not allowed to do
 	 * any more work.
diff --git a/sys/kern/kern_sched.c b/sys/kern/kern_sched.c
index 46de8fa7800..e25993099ee 100644
--- a/sys/kern/kern_sched.c
+++ b/sys/kern/kern_sched.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_sched.c,v 1.62 2019/11/04 18:06:03 visa Exp $	*/
+/*	$OpenBSD: kern_sched.c,v 1.63 2020/01/21 16:16:23 mpi Exp $	*/
 /*
  * Copyright (c) 2007, 2008 Artur Grabowski <art@openbsd.org>
  *
@@ -26,6 +26,7 @@
 #include <sys/mutex.h>
 #include <sys/task.h>
 #include <sys/smr.h>
+#include <sys/tracepoint.h>
 
 #include <uvm/uvm_extern.h>
 
@@ -261,6 +262,7 @@ setrunqueue(struct cpu_info *ci, struct proc *p, uint8_t prio)
 
 	spc = &p->p_cpu->ci_schedstate;
 	spc->spc_nrun++;
+	TRACEPOINT(sched, enqueue, p->p_tid, p->p_p->ps_pid);
 
 	TAILQ_INSERT_TAIL(&spc->spc_qs[queue], p, p_runq);
 	spc->spc_whichqs |= (1 << queue);
@@ -282,6 +284,7 @@ remrunqueue(struct proc *p)
 	SCHED_ASSERT_LOCKED();
 	spc = &p->p_cpu->ci_schedstate;
 	spc->spc_nrun--;
+	TRACEPOINT(sched, dequeue, p->p_tid, p->p_p->ps_pid);
 
 	TAILQ_REMOVE(&spc->spc_qs[queue], p, p_runq);
 	if (TAILQ_EMPTY(&spc->spc_qs[queue])) {
diff --git a/sys/kern/kern_synch.c b/sys/kern/kern_synch.c
index 02c733ad8f1..3ca58be881e 100644
--- a/sys/kern/kern_synch.c
+++ b/sys/kern/kern_synch.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_synch.c,v 1.159 2020/01/21 15:20:47 visa Exp $	*/
+/*	$OpenBSD: kern_synch.c,v 1.160 2020/01/21 16:16:23 mpi Exp $	*/
 /*	$NetBSD: kern_synch.c,v 1.37 1996/04/22 01:38:37 christos Exp $	*/
 
 /*
@@ -51,6 +51,8 @@
 #include <sys/refcnt.h>
 #include <sys/atomic.h>
 #include <sys/witness.h>
+#include <sys/tracepoint.h>
+
 #include <ddb/db_output.h>
 
 #include <machine/spinlock.h>
@@ -380,6 +382,8 @@ sleep_setup(struct sleep_state *sls, const volatile void *ident, int prio,
 
 	SCHED_LOCK(sls->sls_s);
 
+	TRACEPOINT(sched, sleep, NULL);
+
 	p->p_wchan = ident;
 	p->p_wmesg = wmesg;
 	p->p_slptime = 0;
@@ -552,6 +556,7 @@ unsleep(struct proc *p)
 	if (p->p_wchan != NULL) {
 		TAILQ_REMOVE(&slpque[LOOKUP(p->p_wchan)], p, p_runq);
 		p->p_wchan = NULL;
+		TRACEPOINT(sched, wakeup, p->p_tid, p->p_p->ps_pid);
 	}
 }
 
diff --git a/sys/kern/sched_bsd.c b/sys/kern/sched_bsd.c
index 3b8c4a80536..9172bc7e24e 100644
--- a/sys/kern/sched_bsd.c
+++ b/sys/kern/sched_bsd.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: sched_bsd.c,v 1.60 2019/12/11 07:30:09 guenther Exp $	*/
+/*	$OpenBSD: sched_bsd.c,v 1.61 2020/01/21 16:16:23 mpi Exp $	*/
 /*	$NetBSD: kern_synch.c,v 1.37 1996/04/22 01:38:37 christos Exp $	*/
 
 /*-
@@ -48,6 +48,7 @@
 #include <sys/sched.h>
 #include <sys/timeout.h>
 #include <sys/smr.h>
+#include <sys/tracepoint.h>
 
 #ifdef KTRACE
 #include <sys/ktrace.h>
@@ -392,8 +393,12 @@ mi_switch(void)
 
 	if (p != nextproc) {
 		uvmexp.swtch++;
+		TRACEPOINT(sched, off__cpu, nextproc->p_tid,
+		    nextproc->p_p->ps_pid);
 		cpu_switchto(p, nextproc);
+		TRACEPOINT(sched, on__cpu, NULL);
 	} else {
+		TRACEPOINT(sched, remain__cpu, NULL);
 		p->p_stat = SONPROC;
 	}