lwproc.c revision 1.14 
1/*      $NetBSD: lwproc.c,v 1.14 2011/02/10 13:31:30 pooka Exp $	*/
2
3/*
4 * Copyright (c) 2010, 2011 Antti Kantee.  All Rights Reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS
16 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
17 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
18 * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
21 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28#include <sys/cdefs.h>
29__KERNEL_RCSID(0, "$NetBSD: lwproc.c,v 1.14 2011/02/10 13:31:30 pooka Exp $");
30
31#include <sys/param.h>
32#include <sys/atomic.h>
33#include <sys/filedesc.h>
34#include <sys/kauth.h>
35#include <sys/kmem.h>
36#include <sys/lwp.h>
37#include <sys/pool.h>
38#include <sys/proc.h>
39#include <sys/queue.h>
40#include <sys/resourcevar.h>
41#include <sys/uidinfo.h>
42
43#include <rump/rumpuser.h>
44
45#include "rump_private.h"
46
47static void
48lwproc_proc_free(struct proc *p)
49{
50	kauth_cred_t cred;
51
52	mutex_enter(proc_lock);
53
54	KASSERT(p->p_nlwps == 0);
55	KASSERT(LIST_EMPTY(&p->p_lwps));
56	KASSERT(p->p_stat == SACTIVE || p->p_stat == SDYING ||
57	    p->p_stat == SDEAD);
58
59	LIST_REMOVE(p, p_list);
60	LIST_REMOVE(p, p_sibling);
61	proc_free_pid(p->p_pid); /* decrements nprocs */
62	proc_leavepgrp(p); /* releases proc_lock */
63
64	cred = p->p_cred;
65	chgproccnt(kauth_cred_getuid(cred), -1);
66	if (rump_proc_vfs_release)
67		rump_proc_vfs_release(p);
68
69	limfree(p->p_limit);
70	pstatsfree(p->p_stats);
71	kauth_cred_free(p->p_cred);
72	proc_finispecific(p);
73
74	mutex_obj_free(p->p_lock);
75	mutex_destroy(&p->p_stmutex);
76	mutex_destroy(&p->p_auxlock);
77	rw_destroy(&p->p_reflock);
78	cv_destroy(&p->p_waitcv);
79	cv_destroy(&p->p_lwpcv);
80
81	/* non-kernel vmspaces are not shared */
82	if (!RUMP_LOCALPROC_P(p)) {
83		KASSERT(p->p_vmspace->vm_refcnt == 1);
84		kmem_free(p->p_vmspace, sizeof(*p->p_vmspace));
85	}
86
87	proc_free_mem(p);
88}
89
90/*
91 * Allocate a new process.  Mostly mimic fork by
92 * copying the properties of the parent.  However, there are some
93 * differences.  For example, we never share the fd table.
94 *
95 * Switch to the new lwp and return a pointer to it.
96 */
97static struct proc *
98lwproc_newproc(struct proc *parent, int flags)
99{
100	uid_t uid = kauth_cred_getuid(parent->p_cred);
101	struct proc *p;
102
103	/* maxproc not enforced */
104	atomic_inc_uint(&nprocs);
105
106	/* allocate process */
107	p = proc_alloc();
108	memset(&p->p_startzero, 0,
109	    offsetof(struct proc, p_endzero)
110	      - offsetof(struct proc, p_startzero));
111	memcpy(&p->p_startcopy, &parent->p_startcopy,
112	    offsetof(struct proc, p_endcopy)
113	      - offsetof(struct proc, p_startcopy));
114
115	p->p_stats = pstatscopy(parent->p_stats);
116
117	p->p_vmspace = vmspace_kernel();
118	p->p_emul = &emul_netbsd;
119	if (*parent->p_comm)
120		strcpy(p->p_comm, parent->p_comm);
121	else
122		strcpy(p->p_comm, "rumproc");
123
124	if ((flags & RUMP_RFCFDG) == 0)
125		KASSERT(parent == curproc);
126	if (flags & RUMP_RFFDG)
127		p->p_fd = fd_copy();
128	else if (flags & RUMP_RFCFDG)
129		p->p_fd = fd_init(NULL);
130	else
131		fd_share(p);
132
133	lim_addref(parent->p_limit);
134	p->p_limit = parent->p_limit;
135
136	LIST_INIT(&p->p_lwps);
137	LIST_INIT(&p->p_children);
138
139	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
140	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_NONE);
141	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
142	rw_init(&p->p_reflock);
143	cv_init(&p->p_waitcv, "pwait");
144	cv_init(&p->p_lwpcv, "plwp");
145
146	p->p_pptr = parent;
147	p->p_ppid = parent->p_pid;
148	p->p_stat = SACTIVE;
149
150	kauth_proc_fork(parent, p);
151
152	/* initialize cwd in rump kernels with vfs */
153	if (rump_proc_vfs_init)
154		rump_proc_vfs_init(p);
155
156	chgproccnt(uid, 1); /* not enforced */
157
158	/* publish proc various proc lists */
159	mutex_enter(proc_lock);
160	LIST_INSERT_HEAD(&allproc, p, p_list);
161	LIST_INSERT_HEAD(&parent->p_children, p, p_sibling);
162	LIST_INSERT_AFTER(parent, p, p_pglist);
163	mutex_exit(proc_lock);
164
165	return p;
166}
167
168static void
169lwproc_freelwp(struct lwp *l)
170{
171	struct proc *p;
172	bool freeproc;
173
174	p = l->l_proc;
175	mutex_enter(p->p_lock);
176
177	/* XXX: l_refcnt */
178	KASSERT(l->l_flag & LW_WEXIT);
179	KASSERT(l->l_refcnt == 0);
180
181	/* ok, zero references, continue with nuke */
182	LIST_REMOVE(l, l_sibling);
183	KASSERT(p->p_nlwps >= 1);
184	if (--p->p_nlwps == 0) {
185		KASSERT(p != &proc0);
186		p->p_stat = SDEAD;
187	}
188	freeproc = p->p_nlwps == 0;
189	cv_broadcast(&p->p_lwpcv); /* nobody sleeps on this in rump? */
190	kauth_cred_free(l->l_cred);
191	mutex_exit(p->p_lock);
192
193	mutex_enter(proc_lock);
194	LIST_REMOVE(l, l_list);
195	mutex_exit(proc_lock);
196
197	if (l->l_name)
198		kmem_free(l->l_name, MAXCOMLEN);
199	lwp_finispecific(l);
200
201	kmem_free(l, sizeof(*l));
202
203	if (p->p_stat == SDEAD)
204		lwproc_proc_free(p);
205}
206
207extern kmutex_t unruntime_lock;
208
209/*
210 * called with p_lock held, releases lock before return
211 */
212static void
213lwproc_makelwp(struct proc *p, struct lwp *l, bool doswitch, bool procmake)
214{
215
216	p->p_nlwps++;
217	l->l_refcnt = 1;
218	l->l_proc = p;
219
220	l->l_lid = p->p_nlwpid++;
221	LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling);
222	mutex_exit(p->p_lock);
223
224	lwp_update_creds(l);
225
226	l->l_fd = p->p_fd;
227	l->l_cpu = rump_cpu;
228	l->l_target_cpu = rump_cpu; /* Initial target CPU always the same */
229	l->l_stat = LSRUN;
230	l->l_mutex = &unruntime_lock;
231	TAILQ_INIT(&l->l_ld_locks);
232
233	lwp_initspecific(l);
234
235	if (doswitch) {
236		rump_lwproc_switch(l);
237	}
238
239	/* filedesc already has refcount 1 when process is created */
240	if (!procmake) {
241		fd_hold(l);
242	}
243
244	mutex_enter(proc_lock);
245	LIST_INSERT_HEAD(&alllwp, l, l_list);
246	mutex_exit(proc_lock);
247}
248
249struct lwp *
250rump__lwproc_alloclwp(struct proc *p)
251{
252	struct lwp *l;
253	bool newproc = false;
254
255	if (p == NULL) {
256		p = lwproc_newproc(&proc0, 0);
257		newproc = true;
258	}
259
260	l = kmem_zalloc(sizeof(*l), KM_SLEEP);
261
262	mutex_enter(p->p_lock);
263	lwproc_makelwp(p, l, false, newproc);
264
265	return l;
266}
267
268int
269rump_lwproc_newlwp(pid_t pid)
270{
271	struct proc *p;
272	struct lwp *l;
273
274	l = kmem_zalloc(sizeof(*l), KM_SLEEP);
275	mutex_enter(proc_lock);
276	p = proc_find_raw(pid);
277	if (p == NULL) {
278		mutex_exit(proc_lock);
279		kmem_free(l, sizeof(*l));
280		return ESRCH;
281	}
282	mutex_enter(p->p_lock);
283	mutex_exit(proc_lock);
284	lwproc_makelwp(p, l, true, false);
285
286	return 0;
287}
288
289int
290rump_lwproc_rfork(int flags)
291{
292	struct proc *p;
293	struct lwp *l;
294
295	if (flags & ~(RUMP_RFFDG|RUMP_RFCFDG) ||
296	    (~flags & (RUMP_RFFDG|RUMP_RFCFDG)) == 0)
297		return EINVAL;
298
299	p = lwproc_newproc(curproc, flags);
300	l = kmem_zalloc(sizeof(*l), KM_SLEEP);
301	mutex_enter(p->p_lock);
302	lwproc_makelwp(p, l, true, true);
303
304	return 0;
305}
306
307/*
308 * Switch to a new process/thread.  Release previous one if
309 * deemed to be exiting.  This is considered a slow path for
310 * rump kernel entry.
311 */
312void
313rump_lwproc_switch(struct lwp *newlwp)
314{
315	struct lwp *l = curlwp;
316
317	KASSERT(!(l->l_flag & LW_WEXIT) || newlwp);
318
319	if (__predict_false(newlwp && (newlwp->l_pflag & LP_RUNNING)))
320		panic("lwp %p (%d:%d) already running",
321		    newlwp, newlwp->l_proc->p_pid, newlwp->l_lid);
322
323	if (newlwp == NULL) {
324		l->l_pflag &= ~LP_RUNNING;
325		l->l_flag |= LW_RUMP_CLEAR;
326		return;
327	}
328
329	/* fd_free() must be called from curlwp context.  talk about ugh */
330	if (l->l_flag & LW_WEXIT) {
331		fd_free();
332	}
333
334	rumpuser_set_curlwp(NULL);
335
336	newlwp->l_cpu = newlwp->l_target_cpu = l->l_cpu;
337	newlwp->l_mutex = l->l_mutex;
338	newlwp->l_pflag |= LP_RUNNING;
339
340	rumpuser_set_curlwp(newlwp);
341
342	/*
343	 * Check if the thread should get a signal.  This is
344	 * mostly to satisfy the "record" rump sigmodel.
345	 */
346	mutex_enter(newlwp->l_proc->p_lock);
347	if (sigispending(newlwp, 0)) {
348		newlwp->l_flag |= LW_PENDSIG;
349	}
350	mutex_exit(newlwp->l_proc->p_lock);
351
352	l->l_mutex = &unruntime_lock;
353	l->l_pflag &= ~LP_RUNNING;
354	l->l_flag &= ~LW_PENDSIG;
355	l->l_stat = LSRUN;
356
357	if (l->l_flag & LW_WEXIT) {
358		lwproc_freelwp(l);
359	}
360}
361
362void
363rump_lwproc_releaselwp(void)
364{
365	struct proc *p;
366	struct lwp *l = curlwp;
367
368	if (l->l_refcnt == 0 && l->l_flag & LW_WEXIT)
369		panic("releasing non-pertinent lwp");
370
371	p = l->l_proc;
372	mutex_enter(p->p_lock);
373	KASSERT(l->l_refcnt != 0);
374	l->l_refcnt--;
375	mutex_exit(p->p_lock);
376	l->l_flag |= LW_WEXIT; /* will be released when unscheduled */
377}
378
379struct lwp *
380rump_lwproc_curlwp(void)
381{
382	struct lwp *l = curlwp;
383
384	if (l->l_flag & LW_WEXIT)
385		return NULL;
386	return l;
387}
388