subr_syscall.c revision 210138 
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: head/sys/kern/subr_trap.c 210138 2010-07-15 20:24:37Z jhb $");
46
47#include "opt_ktrace.h"
48#include "opt_kdtrace.h"
49#include "opt_sched.h"
50
51#include <sys/param.h>
52#include <sys/bus.h>
53#include <sys/kernel.h>
54#include <sys/lock.h>
55#include <sys/mutex.h>
56#include <sys/pmckern.h>
57#include <sys/proc.h>
58#include <sys/ktr.h>
59#include <sys/pioctl.h>
60#include <sys/ptrace.h>
61#include <sys/resourcevar.h>
62#include <sys/sched.h>
63#include <sys/signalvar.h>
64#include <sys/syscall.h>
65#include <sys/syscallsubr.h>
66#include <sys/sysent.h>
67#include <sys/systm.h>
68#include <sys/vmmeter.h>
69#ifdef KTRACE
70#include <sys/uio.h>
71#include <sys/ktrace.h>
72#endif
73#include <security/audit/audit.h>
74
75#include <machine/cpu.h>
76
77#ifdef XEN
78#include <vm/vm.h>
79#include <vm/vm_param.h>
80#include <vm/pmap.h>
81#endif
82
83#include <security/mac/mac_framework.h>
84
85/*
86 * Define the code needed before returning to user mode, for trap and
87 * syscall.
88 */
89void
90userret(struct thread *td, struct trapframe *frame)
91{
92	struct proc *p = td->td_proc;
93
94	CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
95            td->td_name);
96#if 0
97#ifdef DIAGNOSTIC
98	/* Check that we called signotify() enough. */
99	PROC_LOCK(p);
100	thread_lock(td);
101	if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 ||
102	    (td->td_flags & TDF_ASTPENDING) == 0))
103		printf("failed to set signal flags properly for ast()\n");
104	thread_unlock(td);
105	PROC_UNLOCK(p);
106#endif
107#endif
108#ifdef KTRACE
109	KTRUSERRET(td);
110#endif
111	/*
112	 * If this thread tickled GEOM, we need to wait for the giggling to
113	 * stop before we return to userland
114	 */
115	if (td->td_pflags & TDP_GEOM)
116		g_waitidle();
117
118	/*
119	 * Charge system time if profiling.
120	 */
121	if (p->p_flag & P_PROFIL) {
122		addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
123	}
124	/*
125	 * Let the scheduler adjust our priority etc.
126	 */
127	sched_userret(td);
128	KASSERT(td->td_locks == 0,
129	    ("userret: Returning with %d locks held.", td->td_locks));
130#ifdef XEN
131	PT_UPDATES_FLUSH();
132#endif
133}
134
135/*
136 * Process an asynchronous software trap.
137 * This is relatively easy.
138 * This function will return with preemption disabled.
139 */
140void
141ast(struct trapframe *framep)
142{
143	struct thread *td;
144	struct proc *p;
145	int flags;
146	int sig;
147
148	td = curthread;
149	p = td->td_proc;
150
151	CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid,
152            p->p_comm);
153	KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode"));
154	WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
155	mtx_assert(&Giant, MA_NOTOWNED);
156	THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
157	td->td_frame = framep;
158	td->td_pticks = 0;
159
160	/*
161	 * This updates the td_flag's for the checks below in one
162	 * "atomic" operation with turning off the astpending flag.
163	 * If another AST is triggered while we are handling the
164	 * AST's saved in flags, the astpending flag will be set and
165	 * ast() will be called again.
166	 */
167	thread_lock(td);
168	flags = td->td_flags;
169	td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK |
170	    TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND);
171	thread_unlock(td);
172	PCPU_INC(cnt.v_trap);
173
174	if (td->td_ucred != p->p_ucred)
175		cred_update_thread(td);
176	if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) {
177		addupc_task(td, td->td_profil_addr, td->td_profil_ticks);
178		td->td_profil_ticks = 0;
179		td->td_pflags &= ~TDP_OWEUPC;
180	}
181	if (flags & TDF_ALRMPEND) {
182		PROC_LOCK(p);
183		psignal(p, SIGVTALRM);
184		PROC_UNLOCK(p);
185	}
186	if (flags & TDF_PROFPEND) {
187		PROC_LOCK(p);
188		psignal(p, SIGPROF);
189		PROC_UNLOCK(p);
190	}
191#ifdef MAC
192	if (flags & TDF_MACPEND)
193		mac_thread_userret(td);
194#endif
195	if (flags & TDF_NEEDRESCHED) {
196#ifdef KTRACE
197		if (KTRPOINT(td, KTR_CSW))
198			ktrcsw(1, 1);
199#endif
200		thread_lock(td);
201		sched_prio(td, td->td_user_pri);
202		mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL);
203		thread_unlock(td);
204#ifdef KTRACE
205		if (KTRPOINT(td, KTR_CSW))
206			ktrcsw(0, 1);
207#endif
208	}
209
210	/*
211	 * Check for signals. Unlocked reads of p_pendingcnt or
212	 * p_siglist might cause process-directed signal to be handled
213	 * later.
214	 */
215	if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 ||
216	    !SIGISEMPTY(p->p_siglist)) {
217		PROC_LOCK(p);
218		mtx_lock(&p->p_sigacts->ps_mtx);
219		while ((sig = cursig(td, SIG_STOP_ALLOWED)) != 0)
220			postsig(sig);
221		mtx_unlock(&p->p_sigacts->ps_mtx);
222		PROC_UNLOCK(p);
223	}
224	/*
225	 * We need to check to see if we have to exit or wait due to a
226	 * single threading requirement or some other STOP condition.
227	 */
228	if (flags & TDF_NEEDSUSPCHK) {
229		PROC_LOCK(p);
230		thread_suspend_check(0);
231		PROC_UNLOCK(p);
232	}
233
234	if (td->td_pflags & TDP_OLDMASK) {
235		td->td_pflags &= ~TDP_OLDMASK;
236		kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0);
237	}
238
239	userret(td, framep);
240	mtx_assert(&Giant, MA_NOTOWNED);
241}
242
243#ifdef HAVE_SYSCALL_ARGS_DEF
244const char *
245syscallname(struct proc *p, u_int code)
246{
247	static const char unknown[] = "unknown";
248	struct sysentvec *sv;
249
250	sv = p->p_sysent;
251	if (sv->sv_syscallnames == NULL || code >= sv->sv_size)
252		return (unknown);
253	return (sv->sv_syscallnames[code]);
254}
255
256int
257syscallenter(struct thread *td, struct syscall_args *sa)
258{
259	struct proc *p;
260	int error, traced;
261
262	PCPU_INC(cnt.v_syscall);
263	p = td->td_proc;
264
265	td->td_pticks = 0;
266	if (td->td_ucred != p->p_ucred)
267		cred_update_thread(td);
268	if (p->p_flag & P_TRACED) {
269		traced = 1;
270		PROC_LOCK(p);
271		td->td_dbgflags &= ~TDB_USERWR;
272		td->td_dbgflags |= TDB_SCE;
273		PROC_UNLOCK(p);
274	} else
275		traced = 0;
276	error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
277#ifdef KTRACE
278	if (KTRPOINT(td, KTR_SYSCALL))
279		ktrsyscall(sa->code, sa->narg, sa->args);
280#endif
281
282	CTR6(KTR_SYSC,
283"syscall: td=%p pid %d %s (%#lx, %#lx, %#lx)",
284	    td, td->td_proc->p_pid, syscallname(p, sa->code),
285	    sa->args[0], sa->args[1], sa->args[2]);
286
287	if (error == 0) {
288		STOPEVENT(p, S_SCE, sa->narg);
289		PTRACESTOP_SC(p, td, S_PT_SCE);
290		if (td->td_dbgflags & TDB_USERWR) {
291			/*
292			 * Reread syscall number and arguments if
293			 * debugger modified registers or memory.
294			 */
295			error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
296#ifdef KTRACE
297			if (KTRPOINT(td, KTR_SYSCALL))
298				ktrsyscall(sa->code, sa->narg, sa->args);
299#endif
300			if (error != 0)
301				goto retval;
302		}
303		error = syscall_thread_enter(td, sa->callp);
304		if (error != 0)
305			goto retval;
306
307#ifdef KDTRACE_HOOKS
308		/*
309		 * If the systrace module has registered it's probe
310		 * callback and if there is a probe active for the
311		 * syscall 'entry', process the probe.
312		 */
313		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
314			(*systrace_probe_func)(sa->callp->sy_entry, sa->code,
315			    sa->callp, sa->args);
316#endif
317
318		AUDIT_SYSCALL_ENTER(sa->code, td);
319		error = (sa->callp->sy_call)(td, sa->args);
320		AUDIT_SYSCALL_EXIT(error, td);
321
322		/* Save the latest error return value. */
323		td->td_errno = error;
324
325#ifdef KDTRACE_HOOKS
326		/*
327		 * If the systrace module has registered it's probe
328		 * callback and if there is a probe active for the
329		 * syscall 'return', process the probe.
330		 */
331		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
332			(*systrace_probe_func)(sa->callp->sy_return, sa->code,
333			    sa->callp, sa->args);
334#endif
335		syscall_thread_exit(td, sa->callp);
336		CTR4(KTR_SYSC, "syscall: p=%p error=%d return %#lx %#lx",
337		    p, error, td->td_retval[0], td->td_retval[1]);
338	}
339 retval:
340	if (traced) {
341		PROC_LOCK(p);
342		td->td_dbgflags &= ~TDB_SCE;
343		PROC_UNLOCK(p);
344	}
345	(p->p_sysent->sv_set_syscall_retval)(td, error);
346	return (error);
347}
348
349void
350syscallret(struct thread *td, int error, struct syscall_args *sa __unused)
351{
352	struct proc *p;
353	int traced;
354
355	p = td->td_proc;
356
357	/*
358	 * Check for misbehavior.
359	 */
360	WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
361	    syscallname(p, sa->code));
362	KASSERT(td->td_critnest == 0,
363	    ("System call %s returning in a critical section",
364	    syscallname(p, sa->code)));
365	KASSERT(td->td_locks == 0,
366	    ("System call %s returning with %d locks held",
367	     syscallname(p, sa->code), td->td_locks));
368
369	/*
370	 * Handle reschedule and other end-of-syscall issues
371	 */
372	userret(td, td->td_frame);
373
374	CTR4(KTR_SYSC, "syscall %s exit thread %p pid %d proc %s",
375	    syscallname(p, sa->code), td, td->td_proc->p_pid, td->td_name);
376
377#ifdef KTRACE
378	if (KTRPOINT(td, KTR_SYSRET))
379		ktrsysret(sa->code, error, td->td_retval[0]);
380#endif
381
382	if (p->p_flag & P_TRACED) {
383		traced = 1;
384		PROC_LOCK(p);
385		td->td_dbgflags |= TDB_SCX;
386		PROC_UNLOCK(p);
387	} else
388		traced = 0;
389	/*
390	 * This works because errno is findable through the
391	 * register set.  If we ever support an emulation where this
392	 * is not the case, this code will need to be revisited.
393	 */
394	STOPEVENT(p, S_SCX, sa->code);
395	PTRACESTOP_SC(p, td, S_PT_SCX);
396	if (traced || (td->td_dbgflags & TDB_EXEC) != 0) {
397		PROC_LOCK(p);
398		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC);
399		PROC_UNLOCK(p);
400	}
401}
402#endif /* HAVE_SYSCALL_ARGS_DEF */
403