xref: /netbsd-src/sys/kern/kern_idle.c (revision 04615d5639e70b4bfdaf3942e603ac11f4f45d74)
1 /*	$NetBSD: kern_idle.c,v 1.36 2024/03/01 04:32:38 mrg Exp $	*/
2 
3 /*-
4  * Copyright (c)2002, 2006, 2007 YAMAMOTO Takashi,
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 
31 __KERNEL_RCSID(0, "$NetBSD: kern_idle.c,v 1.36 2024/03/01 04:32:38 mrg Exp $");
32 
33 #include <sys/param.h>
34 #include <sys/cpu.h>
35 #include <sys/idle.h>
36 #include <sys/kthread.h>
37 #include <sys/lockdebug.h>
38 #include <sys/kmem.h>
39 #include <sys/proc.h>
40 #include <sys/atomic.h>
41 
42 #include <uvm/uvm.h>	/* uvm_idle */
43 
44 void
idle_loop(void * dummy)45 idle_loop(void *dummy)
46 {
47 	struct cpu_info *ci = curcpu();
48 	struct schedstate_percpu *spc;
49 	struct lwp *l = curlwp;
50 
51 	KASSERT(l->l_blcnt == 0);
52 
53 	lwp_lock(l);
54 	spc = &ci->ci_schedstate;
55 	KASSERT(lwp_locked(l, spc->spc_lwplock));
56 	kcpuset_atomic_set(kcpuset_running, cpu_index(ci));
57 	/* Update start time for this thread. */
58 	binuptime(&l->l_stime);
59 	spc->spc_flags |= SPCF_RUNNING;
60 	KASSERT((l->l_pflag & LP_RUNNING) != 0);
61 	l->l_stat = LSIDL;
62 	lwp_unlock(l);
63 
64 	/*
65 	 * Use spl0() here to ensure that we have the correct interrupt
66 	 * priority.  This may be the first thread running on the CPU,
67 	 * in which case we took an odd route to get here.
68 	 */
69 	spl0();
70 
71 	for (;;) {
72 		LOCKDEBUG_BARRIER(NULL, 0);
73 		KASSERT((l->l_flag & LW_IDLE) != 0);
74 		KASSERT(ci == curcpu());
75 		KASSERT(l == curlwp);
76 		KASSERT(CURCPU_IDLE_P());
77 		KASSERT(l->l_priority == PRI_IDLE);
78 		KASSERTMSG(l->l_nopreempt == 0, "lwp %p nopreempt %d",
79 		    l, l->l_nopreempt);
80 
81 		sched_idle();
82 		if (!sched_curcpu_runnable_p()) {
83 			if ((spc->spc_flags & SPCF_OFFLINE) == 0) {
84 				uvm_idle();
85 			}
86 			if (!sched_curcpu_runnable_p()) {
87 				cpu_idle();
88 				if (!sched_curcpu_runnable_p() &&
89 				    !ci->ci_want_resched) {
90 					continue;
91 				}
92 			}
93 		}
94 		KASSERT(l->l_mutex == l->l_cpu->ci_schedstate.spc_lwplock);
95 		lwp_lock(l);
96 		spc_lock(l->l_cpu);
97 		mi_switch(l);
98 		KASSERT(curlwp == l);
99 		KASSERT(l->l_stat == LSIDL);
100 	}
101 }
102 
103 int
create_idle_lwp(struct cpu_info * ci)104 create_idle_lwp(struct cpu_info *ci)
105 {
106 	lwp_t *l;
107 	int error;
108 
109 	KASSERT(ci->ci_data.cpu_idlelwp == NULL);
110 	error = kthread_create(PRI_IDLE, KTHREAD_MPSAFE | KTHREAD_IDLE,
111 	    ci, idle_loop, NULL, &l, "idle/%u", ci->ci_index);
112 	if (error != 0)
113 		panic("create_idle_lwp: error %d", error);
114 	lwp_lock(l);
115 	l->l_flag |= LW_IDLE;
116 	if (ci != lwp0.l_cpu) {
117 		/*
118 		 * For secondary CPUs, the idle LWP is the first to run, and
119 		 * it's directly entered from MD code without a trip through
120 		 * mi_switch().  Make the picture look good in case the CPU
121 		 * takes an interrupt before it calls idle_loop().
122 		 */
123 		l->l_stat = LSIDL;
124 		l->l_pflag |= LP_RUNNING;
125 		ci->ci_onproc = l;
126 	}
127 	lwp_unlock(l);
128 	ci->ci_data.cpu_idlelwp = l;
129 
130 	return error;
131 }
132