subr_syscall.c revision 213236 
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 213236 2010-09-28 01:36:01Z emaste $");
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	 * Let the scheduler adjust our priority etc.
125	 */
126	sched_userret(td);
127	KASSERT(td->td_locks == 0,
128	    ("userret: Returning with %d locks held.", td->td_locks));
129#ifdef XEN
130	PT_UPDATES_FLUSH();
131#endif
132}
133
134/*
135 * Process an asynchronous software trap.
136 * This is relatively easy.
137 * This function will return with preemption disabled.
138 */
139void
140ast(struct trapframe *framep)
141{
142	struct thread *td;
143	struct proc *p;
144	int flags;
145	int sig;
146
147	td = curthread;
148	p = td->td_proc;
149
150	CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid,
151            p->p_comm);
152	KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode"));
153	WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
154	mtx_assert(&Giant, MA_NOTOWNED);
155	THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
156	td->td_frame = framep;
157	td->td_pticks = 0;
158
159	/*
160	 * This updates the td_flag's for the checks below in one
161	 * "atomic" operation with turning off the astpending flag.
162	 * If another AST is triggered while we are handling the
163	 * AST's saved in flags, the astpending flag will be set and
164	 * ast() will be called again.
165	 */
166	thread_lock(td);
167	flags = td->td_flags;
168	td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK |
169	    TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND);
170	thread_unlock(td);
171	PCPU_INC(cnt.v_trap);
172
173	if (td->td_ucred != p->p_ucred)
174		cred_update_thread(td);
175	if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) {
176		addupc_task(td, td->td_profil_addr, td->td_profil_ticks);
177		td->td_profil_ticks = 0;
178		td->td_pflags &= ~TDP_OWEUPC;
179	}
180	if (flags & TDF_ALRMPEND) {
181		PROC_LOCK(p);
182		psignal(p, SIGVTALRM);
183		PROC_UNLOCK(p);
184	}
185	if (flags & TDF_PROFPEND) {
186		PROC_LOCK(p);
187		psignal(p, SIGPROF);
188		PROC_UNLOCK(p);
189	}
190#ifdef MAC
191	if (flags & TDF_MACPEND)
192		mac_thread_userret(td);
193#endif
194	if (flags & TDF_NEEDRESCHED) {
195#ifdef KTRACE
196		if (KTRPOINT(td, KTR_CSW))
197			ktrcsw(1, 1);
198#endif
199		thread_lock(td);
200		sched_prio(td, td->td_user_pri);
201		mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL);
202		thread_unlock(td);
203#ifdef KTRACE
204		if (KTRPOINT(td, KTR_CSW))
205			ktrcsw(0, 1);
206#endif
207	}
208
209	/*
210	 * Check for signals. Unlocked reads of p_pendingcnt or
211	 * p_siglist might cause process-directed signal to be handled
212	 * later.
213	 */
214	if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 ||
215	    !SIGISEMPTY(p->p_siglist)) {
216		PROC_LOCK(p);
217		mtx_lock(&p->p_sigacts->ps_mtx);
218		while ((sig = cursig(td, SIG_STOP_ALLOWED)) != 0)
219			postsig(sig);
220		mtx_unlock(&p->p_sigacts->ps_mtx);
221		PROC_UNLOCK(p);
222	}
223	/*
224	 * We need to check to see if we have to exit or wait due to a
225	 * single threading requirement or some other STOP condition.
226	 */
227	if (flags & TDF_NEEDSUSPCHK) {
228		PROC_LOCK(p);
229		thread_suspend_check(0);
230		PROC_UNLOCK(p);
231	}
232
233	if (td->td_pflags & TDP_OLDMASK) {
234		td->td_pflags &= ~TDP_OLDMASK;
235		kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0);
236	}
237
238	userret(td, framep);
239	mtx_assert(&Giant, MA_NOTOWNED);
240}
241
242#ifdef HAVE_SYSCALL_ARGS_DEF
243const char *
244syscallname(struct proc *p, u_int code)
245{
246	static const char unknown[] = "unknown";
247	struct sysentvec *sv;
248
249	sv = p->p_sysent;
250	if (sv->sv_syscallnames == NULL || code >= sv->sv_size)
251		return (unknown);
252	return (sv->sv_syscallnames[code]);
253}
254
255int
256syscallenter(struct thread *td, struct syscall_args *sa)
257{
258	struct proc *p;
259	int error, traced;
260
261	PCPU_INC(cnt.v_syscall);
262	p = td->td_proc;
263
264	td->td_pticks = 0;
265	if (td->td_ucred != p->p_ucred)
266		cred_update_thread(td);
267	if (p->p_flag & P_TRACED) {
268		traced = 1;
269		PROC_LOCK(p);
270		td->td_dbgflags &= ~TDB_USERWR;
271		td->td_dbgflags |= TDB_SCE;
272		PROC_UNLOCK(p);
273	} else
274		traced = 0;
275	error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
276#ifdef KTRACE
277	if (KTRPOINT(td, KTR_SYSCALL))
278		ktrsyscall(sa->code, sa->narg, sa->args);
279#endif
280
281	CTR6(KTR_SYSC,
282"syscall: td=%p pid %d %s (%#lx, %#lx, %#lx)",
283	    td, td->td_proc->p_pid, syscallname(p, sa->code),
284	    sa->args[0], sa->args[1], sa->args[2]);
285
286	if (error == 0) {
287		STOPEVENT(p, S_SCE, sa->narg);
288		PTRACESTOP_SC(p, td, S_PT_SCE);
289		if (td->td_dbgflags & TDB_USERWR) {
290			/*
291			 * Reread syscall number and arguments if
292			 * debugger modified registers or memory.
293			 */
294			error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
295#ifdef KTRACE
296			if (KTRPOINT(td, KTR_SYSCALL))
297				ktrsyscall(sa->code, sa->narg, sa->args);
298#endif
299			if (error != 0)
300				goto retval;
301		}
302		error = syscall_thread_enter(td, sa->callp);
303		if (error != 0)
304			goto retval;
305
306#ifdef KDTRACE_HOOKS
307		/*
308		 * If the systrace module has registered it's probe
309		 * callback and if there is a probe active for the
310		 * syscall 'entry', process the probe.
311		 */
312		if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
313			(*systrace_probe_func)(sa->callp->sy_entry, sa->code,
314			    sa->callp, sa->args, 0);
315#endif
316
317		AUDIT_SYSCALL_ENTER(sa->code, td);
318		error = (sa->callp->sy_call)(td, sa->args);
319		AUDIT_SYSCALL_EXIT(error, td);
320
321		/* Save the latest error return value. */
322		td->td_errno = error;
323
324#ifdef KDTRACE_HOOKS
325		/*
326		 * If the systrace module has registered it's probe
327		 * callback and if there is a probe active for the
328		 * syscall 'return', process the probe.
329		 */
330		if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
331			(*systrace_probe_func)(sa->callp->sy_return, sa->code,
332			    sa->callp, NULL, (error) ? -1 : td->td_retval[0]);
333#endif
334		syscall_thread_exit(td, sa->callp);
335		CTR4(KTR_SYSC, "syscall: p=%p error=%d return %#lx %#lx",
336		    p, error, td->td_retval[0], td->td_retval[1]);
337	}
338 retval:
339	if (traced) {
340		PROC_LOCK(p);
341		td->td_dbgflags &= ~TDB_SCE;
342		PROC_UNLOCK(p);
343	}
344	(p->p_sysent->sv_set_syscall_retval)(td, error);
345	return (error);
346}
347
348void
349syscallret(struct thread *td, int error, struct syscall_args *sa __unused)
350{
351	struct proc *p;
352	int traced;
353
354	p = td->td_proc;
355
356	/*
357	 * Check for misbehavior.
358	 */
359	WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
360	    syscallname(p, sa->code));
361	KASSERT(td->td_critnest == 0,
362	    ("System call %s returning in a critical section",
363	    syscallname(p, sa->code)));
364	KASSERT(td->td_locks == 0,
365	    ("System call %s returning with %d locks held",
366	     syscallname(p, sa->code), td->td_locks));
367
368	/*
369	 * Handle reschedule and other end-of-syscall issues
370	 */
371	userret(td, td->td_frame);
372
373	CTR4(KTR_SYSC, "syscall %s exit thread %p pid %d proc %s",
374	    syscallname(p, sa->code), td, td->td_proc->p_pid, td->td_name);
375
376#ifdef KTRACE
377	if (KTRPOINT(td, KTR_SYSRET))
378		ktrsysret(sa->code, error, td->td_retval[0]);
379#endif
380
381	if (p->p_flag & P_TRACED) {
382		traced = 1;
383		PROC_LOCK(p);
384		td->td_dbgflags |= TDB_SCX;
385		PROC_UNLOCK(p);
386	} else
387		traced = 0;
388	/*
389	 * This works because errno is findable through the
390	 * register set.  If we ever support an emulation where this
391	 * is not the case, this code will need to be revisited.
392	 */
393	STOPEVENT(p, S_SCX, sa->code);
394	PTRACESTOP_SC(p, td, S_PT_SCX);
395	if (traced || (td->td_dbgflags & TDB_EXEC) != 0) {
396		PROC_LOCK(p);
397		td->td_dbgflags &= ~(TDB_SCX | TDB_EXEC);
398		PROC_UNLOCK(p);
399	}
400}
401#endif /* HAVE_SYSCALL_ARGS_DEF */
402