xref: /freebsd/sys/kern/subr_trap.c (revision 8d20be1e22095c27faf8fe8b2f0d089739cc742e)
1 /*-
2  * Copyright (C) 1994, David Greenman
3  * Copyright (c) 1990, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  * Copyright (c) 2007 The FreeBSD Foundation
6  *
7  * This code is derived from software contributed to Berkeley by
8  * the University of Utah, and William Jolitz.
9  *
10  * Portions of this software were developed by A. Joseph Koshy under
11  * sponsorship from the FreeBSD Foundation and Google, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. All advertising materials mentioning features or use of this software
22  *    must display the following acknowledgement:
23  *	This product includes software developed by the University of
24  *	California, Berkeley and its contributors.
25  * 4. Neither the name of the University nor the names of its contributors
26  *    may be used to endorse or promote products derived from this software
27  *    without specific prior written permission.
28  *
29  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39  * SUCH DAMAGE.
40  *
41  *	from: @(#)trap.c	7.4 (Berkeley) 5/13/91
42  */
43 
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
46 
47 #include "opt_hwpmc_hooks.h"
48 #include "opt_ktrace.h"
49 #include "opt_kdtrace.h"
50 #include "opt_sched.h"
51 
52 #include <sys/param.h>
53 #include <sys/bus.h>
54 #include <sys/capability.h>
55 #include <sys/kernel.h>
56 #include <sys/lock.h>
57 #include <sys/mutex.h>
58 #include <sys/pmckern.h>
59 #include <sys/proc.h>
60 #include <sys/ktr.h>
61 #include <sys/pioctl.h>
62 #include <sys/ptrace.h>
63 #include <sys/resourcevar.h>
64 #include <sys/sched.h>
65 #include <sys/signalvar.h>
66 #include <sys/syscall.h>
67 #include <sys/syscallsubr.h>
68 #include <sys/sysent.h>
69 #include <sys/systm.h>
70 #include <sys/vmmeter.h>
71 #ifdef KTRACE
72 #include <sys/uio.h>
73 #include <sys/ktrace.h>
74 #endif
75 #include <security/audit/audit.h>
76 
77 #include <machine/cpu.h>
78 
79 #ifdef VIMAGE
80 #include <net/vnet.h>
81 #endif
82 
83 #ifdef XEN
84 #include <vm/vm.h>
85 #include <vm/vm_param.h>
86 #include <vm/pmap.h>
87 #endif
88 
89 #ifdef	HWPMC_HOOKS
90 #include <sys/pmckern.h>
91 #endif
92 
93 #include <security/mac/mac_framework.h>
94 
95 /*
96  * Define the code needed before returning to user mode, for trap and
97  * syscall.
98  */
99 void
100 userret(struct thread *td, struct trapframe *frame)
101 {
102 	struct proc *p = td->td_proc;
103 
104 	CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
105             td->td_name);
106 	KASSERT((p->p_flag & P_WEXIT) == 0,
107 	    ("Exiting process returns to usermode"));
108 #if 0
109 #ifdef DIAGNOSTIC
110 	/* Check that we called signotify() enough. */
111 	PROC_LOCK(p);
112 	thread_lock(td);
113 	if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 ||
114 	    (td->td_flags & TDF_ASTPENDING) == 0))
115 		printf("failed to set signal flags properly for ast()\n");
116 	thread_unlock(td);
117 	PROC_UNLOCK(p);
118 #endif
119 #endif
120 #ifdef KTRACE
121 	KTRUSERRET(td);
122 #endif
123 	/*
124 	 * If this thread tickled GEOM, we need to wait for the giggling to
125 	 * stop before we return to userland
126 	 */
127 	if (td->td_pflags & TDP_GEOM)
128 		g_waitidle();
129 
130 	/*
131 	 * Charge system time if profiling.
132 	 */
133 	if (p->p_flag & P_PROFIL)
134 		addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
135 	/*
136 	 * Let the scheduler adjust our priority etc.
137 	 */
138 	sched_userret(td);
139 #ifdef XEN
140 	PT_UPDATES_FLUSH();
141 #endif
142 
143 	/*
144 	 * Check for misbehavior.
145 	 *
146 	 * In case there is a callchain tracing ongoing because of
147 	 * hwpmc(4), skip the scheduler pinning check.
148 	 * hwpmc(4) subsystem, infact, will collect callchain informations
149 	 * at ast() checkpoint, which is past userret().
150 	 */
151 	WITNESS_WARN(WARN_PANIC, NULL, "userret: returning");
152 	KASSERT(td->td_critnest == 0,
153 	    ("userret: Returning in a critical section"));
154 	KASSERT(td->td_locks == 0,
155 	    ("userret: Returning with %d locks held", td->td_locks));
156 	KASSERT((td->td_pflags & TDP_NOFAULTING) == 0,
157 	    ("userret: Returning with pagefaults disabled"));
158 	KASSERT(td->td_no_sleeping == 0,
159 	    ("userret: Returning with sleep disabled"));
160 	KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0,
161 	    ("userret: Returning with with pinned thread"));
162 	KASSERT(td->td_vp_reserv == 0,
163 	    ("userret: Returning while holding vnode reservation"));
164 	KASSERT((td->td_flags & TDF_SBDRY) == 0,
165 	    ("userret: Returning with stop signals deferred"));
166 #ifdef VIMAGE
167 	/* Unfortunately td_vnet_lpush needs VNET_DEBUG. */
168 	VNET_ASSERT(curvnet == NULL,
169 	    ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s",
170 	    __func__, td, p->p_pid, td->td_name, curvnet,
171 	    (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A"));
172 #endif
173 #ifdef	RACCT
174 	PROC_LOCK(p);
175 	while (p->p_throttled == 1)
176 		msleep(p->p_racct, &p->p_mtx, 0, "racct", 0);
177 	PROC_UNLOCK(p);
178 #endif
179 }
180 
181 /*
182  * Process an asynchronous software trap.
183  * This is relatively easy.
184  * This function will return with preemption disabled.
185  */
186 void
187 ast(struct trapframe *framep)
188 {
189 	struct thread *td;
190 	struct proc *p;
191 	int flags;
192 	int sig;
193 
194 	td = curthread;
195 	p = td->td_proc;
196 
197 	CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid,
198             p->p_comm);
199 	KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode"));
200 	WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
201 	mtx_assert(&Giant, MA_NOTOWNED);
202 	THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
203 	td->td_frame = framep;
204 	td->td_pticks = 0;
205 
206 	/*
207 	 * This updates the td_flag's for the checks below in one
208 	 * "atomic" operation with turning off the astpending flag.
209 	 * If another AST is triggered while we are handling the
210 	 * AST's saved in flags, the astpending flag will be set and
211 	 * ast() will be called again.
212 	 */
213 	thread_lock(td);
214 	flags = td->td_flags;
215 	td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK |
216 	    TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND);
217 	thread_unlock(td);
218 	PCPU_INC(cnt.v_trap);
219 
220 	if (td->td_ucred != p->p_ucred)
221 		cred_update_thread(td);
222 	if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) {
223 		addupc_task(td, td->td_profil_addr, td->td_profil_ticks);
224 		td->td_profil_ticks = 0;
225 		td->td_pflags &= ~TDP_OWEUPC;
226 	}
227 #ifdef HWPMC_HOOKS
228 	/* Handle Software PMC callchain capture. */
229 	if (PMC_IS_PENDING_CALLCHAIN(td))
230 		PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep);
231 #endif
232 	if (flags & TDF_ALRMPEND) {
233 		PROC_LOCK(p);
234 		kern_psignal(p, SIGVTALRM);
235 		PROC_UNLOCK(p);
236 	}
237 	if (flags & TDF_PROFPEND) {
238 		PROC_LOCK(p);
239 		kern_psignal(p, SIGPROF);
240 		PROC_UNLOCK(p);
241 	}
242 #ifdef MAC
243 	if (flags & TDF_MACPEND)
244 		mac_thread_userret(td);
245 #endif
246 	if (flags & TDF_NEEDRESCHED) {
247 #ifdef KTRACE
248 		if (KTRPOINT(td, KTR_CSW))
249 			ktrcsw(1, 1, __func__);
250 #endif
251 		thread_lock(td);
252 		sched_prio(td, td->td_user_pri);
253 		mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL);
254 		thread_unlock(td);
255 #ifdef KTRACE
256 		if (KTRPOINT(td, KTR_CSW))
257 			ktrcsw(0, 1, __func__);
258 #endif
259 	}
260 
261 	/*
262 	 * Check for signals. Unlocked reads of p_pendingcnt or
263 	 * p_siglist might cause process-directed signal to be handled
264 	 * later.
265 	 */
266 	if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 ||
267 	    !SIGISEMPTY(p->p_siglist)) {
268 		PROC_LOCK(p);
269 		mtx_lock(&p->p_sigacts->ps_mtx);
270 		while ((sig = cursig(td)) != 0)
271 			postsig(sig);
272 		mtx_unlock(&p->p_sigacts->ps_mtx);
273 		PROC_UNLOCK(p);
274 	}
275 	/*
276 	 * We need to check to see if we have to exit or wait due to a
277 	 * single threading requirement or some other STOP condition.
278 	 */
279 	if (flags & TDF_NEEDSUSPCHK) {
280 		PROC_LOCK(p);
281 		thread_suspend_check(0);
282 		PROC_UNLOCK(p);
283 	}
284 
285 	if (td->td_pflags & TDP_OLDMASK) {
286 		td->td_pflags &= ~TDP_OLDMASK;
287 		kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0);
288 	}
289 
290 	userret(td, framep);
291 }
292 
293 const char *
294 syscallname(struct proc *p, u_int code)
295 {
296 	static const char unknown[] = "unknown";
297 	struct sysentvec *sv;
298 
299 	sv = p->p_sysent;
300 	if (sv->sv_syscallnames == NULL || code >= sv->sv_size)
301 		return (unknown);
302 	return (sv->sv_syscallnames[code]);
303 }
304