xref: /freebsd/sys/kern/subr_trap.c (revision 8ef24a0d4b28fe230e20637f56869cc4148cd2ca)
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_sched.h"
50 
51 #include <sys/param.h>
52 #include <sys/bus.h>
53 #include <sys/capsicum.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/mutex.h>
57 #include <sys/pmckern.h>
58 #include <sys/proc.h>
59 #include <sys/ktr.h>
60 #include <sys/pioctl.h>
61 #include <sys/ptrace.h>
62 #include <sys/racct.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	HWPMC_HOOKS
84 #include <sys/pmckern.h>
85 #endif
86 
87 #include <security/mac/mac_framework.h>
88 
89 void (*softdep_ast_cleanup)(void);
90 
91 /*
92  * Define the code needed before returning to user mode, for trap and
93  * syscall.
94  */
95 void
96 userret(struct thread *td, struct trapframe *frame)
97 {
98 	struct proc *p = td->td_proc;
99 
100 	CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
101             td->td_name);
102 	KASSERT((p->p_flag & P_WEXIT) == 0,
103 	    ("Exiting process returns to usermode"));
104 #if 0
105 #ifdef DIAGNOSTIC
106 	/* Check that we called signotify() enough. */
107 	PROC_LOCK(p);
108 	thread_lock(td);
109 	if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 ||
110 	    (td->td_flags & TDF_ASTPENDING) == 0))
111 		printf("failed to set signal flags properly for ast()\n");
112 	thread_unlock(td);
113 	PROC_UNLOCK(p);
114 #endif
115 #endif
116 #ifdef KTRACE
117 	KTRUSERRET(td);
118 #endif
119 	if (softdep_ast_cleanup != NULL)
120 		softdep_ast_cleanup();
121 
122 	/*
123 	 * If this thread tickled GEOM, we need to wait for the giggling to
124 	 * stop before we return to userland
125 	 */
126 	if (td->td_pflags & TDP_GEOM)
127 		g_waitidle();
128 
129 	/*
130 	 * Charge system time if profiling.
131 	 */
132 	if (p->p_flag & P_PROFIL)
133 		addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
134 	/*
135 	 * Let the scheduler adjust our priority etc.
136 	 */
137 	sched_userret(td);
138 
139 	/*
140 	 * Check for misbehavior.
141 	 *
142 	 * In case there is a callchain tracing ongoing because of
143 	 * hwpmc(4), skip the scheduler pinning check.
144 	 * hwpmc(4) subsystem, infact, will collect callchain informations
145 	 * at ast() checkpoint, which is past userret().
146 	 */
147 	WITNESS_WARN(WARN_PANIC, NULL, "userret: returning");
148 	KASSERT(td->td_critnest == 0,
149 	    ("userret: Returning in a critical section"));
150 	KASSERT(td->td_locks == 0,
151 	    ("userret: Returning with %d locks held", td->td_locks));
152 	KASSERT(td->td_rw_rlocks == 0,
153 	    ("userret: Returning with %d rwlocks held in read mode",
154 	    td->td_rw_rlocks));
155 	KASSERT((td->td_pflags & TDP_NOFAULTING) == 0,
156 	    ("userret: Returning with pagefaults disabled"));
157 	KASSERT(td->td_no_sleeping == 0,
158 	    ("userret: Returning with sleep disabled"));
159 	KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0,
160 	    ("userret: Returning with with pinned thread"));
161 	KASSERT(td->td_vp_reserv == 0,
162 	    ("userret: Returning while holding vnode reservation"));
163 	KASSERT((td->td_flags & TDF_SBDRY) == 0,
164 	    ("userret: Returning with stop signals deferred"));
165 	KASSERT(td->td_su == NULL,
166 	    ("userret: Returning with SU cleanup request not handled"));
167 #ifdef VIMAGE
168 	/* Unfortunately td_vnet_lpush needs VNET_DEBUG. */
169 	VNET_ASSERT(curvnet == NULL,
170 	    ("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s",
171 	    __func__, td, p->p_pid, td->td_name, curvnet,
172 	    (td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A"));
173 #endif
174 #ifdef RACCT
175 	if (racct_enable && p->p_throttled != 0) {
176 		PROC_LOCK(p);
177 		while (p->p_throttled != 0) {
178 			msleep(p->p_racct, &p->p_mtx, 0, "racct",
179 			    p->p_throttled < 0 ? 0 : p->p_throttled);
180 			if (p->p_throttled > 0)
181 				p->p_throttled = 0;
182 		}
183 		PROC_UNLOCK(p);
184 	}
185 #endif
186 }
187 
188 /*
189  * Process an asynchronous software trap.
190  * This is relatively easy.
191  * This function will return with preemption disabled.
192  */
193 void
194 ast(struct trapframe *framep)
195 {
196 	struct thread *td;
197 	struct proc *p;
198 	int flags;
199 	int sig;
200 
201 	td = curthread;
202 	p = td->td_proc;
203 
204 	CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid,
205             p->p_comm);
206 	KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode"));
207 	WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
208 	mtx_assert(&Giant, MA_NOTOWNED);
209 	THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
210 	td->td_frame = framep;
211 	td->td_pticks = 0;
212 
213 	/*
214 	 * This updates the td_flag's for the checks below in one
215 	 * "atomic" operation with turning off the astpending flag.
216 	 * If another AST is triggered while we are handling the
217 	 * AST's saved in flags, the astpending flag will be set and
218 	 * ast() will be called again.
219 	 */
220 	thread_lock(td);
221 	flags = td->td_flags;
222 	td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK |
223 	    TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND);
224 	thread_unlock(td);
225 	PCPU_INC(cnt.v_trap);
226 
227 	if (td->td_cowgen != p->p_cowgen)
228 		thread_cow_update(td);
229 	if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) {
230 		addupc_task(td, td->td_profil_addr, td->td_profil_ticks);
231 		td->td_profil_ticks = 0;
232 		td->td_pflags &= ~TDP_OWEUPC;
233 	}
234 #ifdef HWPMC_HOOKS
235 	/* Handle Software PMC callchain capture. */
236 	if (PMC_IS_PENDING_CALLCHAIN(td))
237 		PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep);
238 #endif
239 	if (flags & TDF_ALRMPEND) {
240 		PROC_LOCK(p);
241 		kern_psignal(p, SIGVTALRM);
242 		PROC_UNLOCK(p);
243 	}
244 	if (flags & TDF_PROFPEND) {
245 		PROC_LOCK(p);
246 		kern_psignal(p, SIGPROF);
247 		PROC_UNLOCK(p);
248 	}
249 #ifdef MAC
250 	if (flags & TDF_MACPEND)
251 		mac_thread_userret(td);
252 #endif
253 	if (flags & TDF_NEEDRESCHED) {
254 #ifdef KTRACE
255 		if (KTRPOINT(td, KTR_CSW))
256 			ktrcsw(1, 1, __func__);
257 #endif
258 		thread_lock(td);
259 		sched_prio(td, td->td_user_pri);
260 		mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL);
261 		thread_unlock(td);
262 #ifdef KTRACE
263 		if (KTRPOINT(td, KTR_CSW))
264 			ktrcsw(0, 1, __func__);
265 #endif
266 	}
267 
268 	/*
269 	 * Check for signals. Unlocked reads of p_pendingcnt or
270 	 * p_siglist might cause process-directed signal to be handled
271 	 * later.
272 	 */
273 	if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 ||
274 	    !SIGISEMPTY(p->p_siglist)) {
275 		PROC_LOCK(p);
276 		mtx_lock(&p->p_sigacts->ps_mtx);
277 		while ((sig = cursig(td)) != 0)
278 			postsig(sig);
279 		mtx_unlock(&p->p_sigacts->ps_mtx);
280 		PROC_UNLOCK(p);
281 	}
282 	/*
283 	 * We need to check to see if we have to exit or wait due to a
284 	 * single threading requirement or some other STOP condition.
285 	 */
286 	if (flags & TDF_NEEDSUSPCHK) {
287 		PROC_LOCK(p);
288 		thread_suspend_check(0);
289 		PROC_UNLOCK(p);
290 	}
291 
292 	if (td->td_pflags & TDP_OLDMASK) {
293 		td->td_pflags &= ~TDP_OLDMASK;
294 		kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0);
295 	}
296 
297 	userret(td, framep);
298 }
299 
300 const char *
301 syscallname(struct proc *p, u_int code)
302 {
303 	static const char unknown[] = "unknown";
304 	struct sysentvec *sv;
305 
306 	sv = p->p_sysent;
307 	if (sv->sv_syscallnames == NULL || code >= sv->sv_size)
308 		return (unknown);
309 	return (sv->sv_syscallnames[code]);
310 }
311