xref: /freebsd/sys/kern/sys_process.c (revision 4a558355e5f3b4521cef56a6b705fa84be41dfa0)
1 /*
2  * Copyright (c) 1994, Sean Eric Fagan
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by Sean Eric Fagan.
16  * 4. The name of the author may not be used to endorse or promote products
17  *    derived from this software without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  * $FreeBSD$
32  */
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/sysproto.h>
37 #include <sys/proc.h>
38 #include <sys/vnode.h>
39 #include <sys/ptrace.h>
40 
41 #include <machine/reg.h>
42 #include <vm/vm.h>
43 #include <sys/lock.h>
44 #include <vm/pmap.h>
45 #include <vm/vm_map.h>
46 #include <vm/vm_page.h>
47 
48 #include <sys/user.h>
49 #include <miscfs/procfs/procfs.h>
50 
51 /* use the equivalent procfs code */
52 #if 0
53 static int
54 pread (struct proc *procp, unsigned int addr, unsigned int *retval) {
55 	int		rv;
56 	vm_map_t	map, tmap;
57 	vm_object_t	object;
58 	vm_offset_t	kva = 0;
59 	int		page_offset;	/* offset into page */
60 	vm_offset_t	pageno;		/* page number */
61 	vm_map_entry_t	out_entry;
62 	vm_prot_t	out_prot;
63 	boolean_t	wired;
64 	vm_pindex_t	pindex;
65 
66 	/* Map page into kernel space */
67 
68 	map = &procp->p_vmspace->vm_map;
69 
70 	page_offset = addr - trunc_page(addr);
71 	pageno = trunc_page(addr);
72 
73 	tmap = map;
74 	rv = vm_map_lookup (&tmap, pageno, VM_PROT_READ, &out_entry,
75 		&object, &pindex, &out_prot, &wired);
76 
77 	if (rv != KERN_SUCCESS)
78 		return EINVAL;
79 
80 	vm_map_lookup_done (tmap, out_entry);
81 
82 	/* Find space in kernel_map for the page we're interested in */
83 	rv = vm_map_find (kernel_map, object, IDX_TO_OFF(pindex),
84 		&kva, PAGE_SIZE, 0, VM_PROT_ALL, VM_PROT_ALL, 0);
85 
86 	if (!rv) {
87 		vm_object_reference (object);
88 
89 		rv = vm_map_pageable (kernel_map, kva, kva + PAGE_SIZE, 0);
90 		if (!rv) {
91 			*retval = 0;
92 			bcopy ((caddr_t)kva + page_offset,
93 			       retval, sizeof *retval);
94 		}
95 		vm_map_remove (kernel_map, kva, kva + PAGE_SIZE);
96 	}
97 
98 	return rv;
99 }
100 
101 static int
102 pwrite (struct proc *procp, unsigned int addr, unsigned int datum) {
103 	int		rv;
104 	vm_map_t	map, tmap;
105 	vm_object_t	object;
106 	vm_offset_t	kva = 0;
107 	int		page_offset;	/* offset into page */
108 	vm_offset_t	pageno;		/* page number */
109 	vm_map_entry_t	out_entry;
110 	vm_prot_t	out_prot;
111 	boolean_t	wired;
112 	vm_pindex_t	pindex;
113 	boolean_t	fix_prot = 0;
114 
115 	/* Map page into kernel space */
116 
117 	map = &procp->p_vmspace->vm_map;
118 
119 	page_offset = addr - trunc_page(addr);
120 	pageno = trunc_page(addr);
121 
122 	/*
123 	 * Check the permissions for the area we're interested in.
124 	 */
125 
126 	if (vm_map_check_protection (map, pageno, pageno + PAGE_SIZE,
127 		VM_PROT_WRITE) == FALSE) {
128 		/*
129 		 * If the page was not writable, we make it so.
130 		 * XXX It is possible a page may *not* be read/executable,
131 		 * if a process changes that!
132 		 */
133 		fix_prot = 1;
134 		/* The page isn't writable, so let's try making it so... */
135 		if ((rv = vm_map_protect (map, pageno, pageno + PAGE_SIZE,
136 			VM_PROT_ALL, 0)) != KERN_SUCCESS)
137 		  return EFAULT;	/* I guess... */
138 	}
139 
140 	/*
141 	 * Now we need to get the page.  out_entry, out_prot, wired, and
142 	 * single_use aren't used.  One would think the vm code would be
143 	 * a *bit* nicer...  We use tmap because vm_map_lookup() can
144 	 * change the map argument.
145 	 */
146 
147 	tmap = map;
148 	rv = vm_map_lookup (&tmap, pageno, VM_PROT_WRITE, &out_entry,
149 		&object, &pindex, &out_prot, &wired);
150 	if (rv != KERN_SUCCESS) {
151 		return EINVAL;
152 	}
153 
154 	/*
155 	 * Okay, we've got the page.  Let's release tmap.
156 	 */
157 
158 	vm_map_lookup_done (tmap, out_entry);
159 
160 	/*
161 	 * Fault the page in...
162 	 */
163 
164 	rv = vm_fault(map, pageno, VM_PROT_WRITE|VM_PROT_READ, FALSE);
165 	if (rv != KERN_SUCCESS)
166 		return EFAULT;
167 
168 	/* Find space in kernel_map for the page we're interested in */
169 	rv = vm_map_find (kernel_map, object, IDX_TO_OFF(pindex),
170 		&kva, PAGE_SIZE, 0,
171 		VM_PROT_ALL, VM_PROT_ALL, 0);
172 	if (!rv) {
173 		vm_object_reference (object);
174 
175 		rv = vm_map_pageable (kernel_map, kva, kva + PAGE_SIZE, 0);
176 		if (!rv) {
177 		  bcopy (&datum, (caddr_t)kva + page_offset, sizeof datum);
178 		}
179 		vm_map_remove (kernel_map, kva, kva + PAGE_SIZE);
180 	}
181 
182 	if (fix_prot)
183 		vm_map_protect (map, pageno, pageno + PAGE_SIZE,
184 			VM_PROT_READ|VM_PROT_EXECUTE, 0);
185 	return rv;
186 }
187 #endif
188 
189 /*
190  * Process debugging system call.
191  */
192 #ifndef _SYS_SYSPROTO_H_
193 struct ptrace_args {
194 	int	req;
195 	pid_t	pid;
196 	caddr_t	addr;
197 	int	data;
198 };
199 #endif
200 
201 int
202 ptrace(curp, uap)
203 	struct proc *curp;
204 	struct ptrace_args *uap;
205 {
206 	struct proc *p;
207 	struct iovec iov;
208 	struct uio uio;
209 	int error = 0;
210 	int write;
211 
212 	write = 0;
213 	if (uap->req == PT_TRACE_ME)
214 		p = curp;
215 	else {
216 		if ((p = pfind(uap->pid)) == NULL)
217 			return ESRCH;
218 	}
219 	if (p_can(curp, p, P_CAN_SEE, NULL))
220 		return (ESRCH);
221 
222 	/*
223 	 * Permissions check
224 	 */
225 	switch (uap->req) {
226 	case PT_TRACE_ME:
227 		/* Always legal. */
228 		break;
229 
230 	case PT_ATTACH:
231 		/* Self */
232 		if (p->p_pid == curp->p_pid)
233 			return EINVAL;
234 
235 		/* Already traced */
236 		PROC_LOCK(p);
237 		if (p->p_flag & P_TRACED) {
238 			PROC_UNLOCK(p);
239 			return EBUSY;
240 		}
241 		PROC_UNLOCK(p);
242 
243 		if ((error = p_can(curp, p, P_CAN_DEBUG, NULL)))
244 			return error;
245 
246 		/* OK */
247 		break;
248 
249 	case PT_READ_I:
250 	case PT_READ_D:
251 	case PT_READ_U:
252 	case PT_WRITE_I:
253 	case PT_WRITE_D:
254 	case PT_WRITE_U:
255 	case PT_CONTINUE:
256 	case PT_KILL:
257 	case PT_STEP:
258 	case PT_DETACH:
259 #ifdef PT_GETREGS
260 	case PT_GETREGS:
261 #endif
262 #ifdef PT_SETREGS
263 	case PT_SETREGS:
264 #endif
265 #ifdef PT_GETFPREGS
266 	case PT_GETFPREGS:
267 #endif
268 #ifdef PT_SETFPREGS
269 	case PT_SETFPREGS:
270 #endif
271 #ifdef PT_GETDBREGS
272 	case PT_GETDBREGS:
273 #endif
274 #ifdef PT_SETDBREGS
275 	case PT_SETDBREGS:
276 #endif
277 		/* not being traced... */
278 		PROC_LOCK(p);
279 		if ((p->p_flag & P_TRACED) == 0) {
280 			PROC_UNLOCK(p);
281 			return EPERM;
282 		}
283 
284 		/* not being traced by YOU */
285 		if (p->p_pptr != curp) {
286 			PROC_UNLOCK(p);
287 			return EBUSY;
288 		}
289 
290 		/* not currently stopped */
291 		mtx_lock_spin(&sched_lock);
292 		if (p->p_stat != SSTOP || (p->p_flag & P_WAITED) == 0) {
293 			mtx_unlock_spin(&sched_lock);
294 			PROC_UNLOCK(p);
295 			return EBUSY;
296 		}
297 		mtx_unlock_spin(&sched_lock);
298 		PROC_UNLOCK(p);
299 
300 		/* OK */
301 		break;
302 
303 	default:
304 		return EINVAL;
305 	}
306 
307 #ifdef FIX_SSTEP
308 	/*
309 	 * Single step fixup ala procfs
310 	 */
311 	FIX_SSTEP(p);
312 #endif
313 
314 	/*
315 	 * Actually do the requests
316 	 */
317 
318 	curp->p_retval[0] = 0;
319 
320 	switch (uap->req) {
321 	case PT_TRACE_ME:
322 		/* set my trace flag and "owner" so it can read/write me */
323 		PROCTREE_LOCK(PT_EXCLUSIVE);
324 		PROC_LOCK(p);
325 		p->p_flag |= P_TRACED;
326 		p->p_oppid = p->p_pptr->p_pid;
327 		PROC_UNLOCK(p);
328 		PROCTREE_LOCK(PT_RELEASE);
329 		return 0;
330 
331 	case PT_ATTACH:
332 		/* security check done above */
333 		PROCTREE_LOCK(PT_EXCLUSIVE);
334 		PROC_LOCK(p);
335 		p->p_flag |= P_TRACED;
336 		p->p_oppid = p->p_pptr->p_pid;
337 		if (p->p_pptr != curp)
338 			proc_reparent(p, curp);
339 		PROC_UNLOCK(p);
340 		PROCTREE_LOCK(PT_RELEASE);
341 		uap->data = SIGSTOP;
342 		goto sendsig;	/* in PT_CONTINUE below */
343 
344 	case PT_STEP:
345 	case PT_CONTINUE:
346 	case PT_DETACH:
347 		if ((uap->req != PT_STEP) && ((unsigned)uap->data >= NSIG))
348 			return EINVAL;
349 
350 		PHOLD(p);
351 
352 		if (uap->req == PT_STEP) {
353 			if ((error = ptrace_single_step (p))) {
354 				PRELE(p);
355 				return error;
356 			}
357 		}
358 
359 		if (uap->addr != (caddr_t)1) {
360 			fill_kinfo_proc (p, &p->p_addr->u_kproc);
361 			if ((error = ptrace_set_pc (p,
362 			    (u_long)(uintfptr_t)uap->addr))) {
363 				PRELE(p);
364 				return error;
365 			}
366 		}
367 		PRELE(p);
368 
369 		if (uap->req == PT_DETACH) {
370 			/* reset process parent */
371 			PROCTREE_LOCK(PT_EXCLUSIVE);
372 			if (p->p_oppid != p->p_pptr->p_pid) {
373 				struct proc *pp;
374 
375 				pp = pfind(p->p_oppid);
376 				PROC_LOCK(p);
377 				proc_reparent(p, pp ? pp : initproc);
378 			} else
379 				PROC_LOCK(p);
380 			p->p_flag &= ~(P_TRACED | P_WAITED);
381 			p->p_oppid = 0;
382 
383 			PROC_UNLOCK(p);
384 			PROCTREE_LOCK(PT_RELEASE);
385 
386 			/* should we send SIGCHLD? */
387 
388 		}
389 
390 	sendsig:
391 		/* deliver or queue signal */
392 		PROC_LOCK(p);
393 		mtx_lock_spin(&sched_lock);
394 		if (p->p_stat == SSTOP) {
395 			p->p_xstat = uap->data;
396 			setrunnable(p);
397 			mtx_unlock_spin(&sched_lock);
398 		} else {
399 			mtx_unlock_spin(&sched_lock);
400 			if (uap->data)
401 				psignal(p, uap->data);
402 
403 		}
404 		PROC_UNLOCK(p);
405 		return 0;
406 
407 	case PT_WRITE_I:
408 	case PT_WRITE_D:
409 		write = 1;
410 		/* fallthrough */
411 	case PT_READ_I:
412 	case PT_READ_D:
413 		/* write = 0 set above */
414 		iov.iov_base = write ? (caddr_t)&uap->data : (caddr_t)curp->p_retval;
415 		iov.iov_len = sizeof(int);
416 		uio.uio_iov = &iov;
417 		uio.uio_iovcnt = 1;
418 		uio.uio_offset = (off_t)(uintptr_t)uap->addr;
419 		uio.uio_resid = sizeof(int);
420 		uio.uio_segflg = UIO_SYSSPACE;	/* ie: the uap */
421 		uio.uio_rw = write ? UIO_WRITE : UIO_READ;
422 		uio.uio_procp = p;
423 		error = procfs_domem(curp, p, NULL, &uio);
424 		if (uio.uio_resid != 0) {
425 			/*
426 			 * XXX procfs_domem() doesn't currently return ENOSPC,
427 			 * so I think write() can bogusly return 0.
428 			 * XXX what happens for short writes?  We don't want
429 			 * to write partial data.
430 			 * XXX procfs_domem() returns EPERM for other invalid
431 			 * addresses.  Convert this to EINVAL.  Does this
432 			 * clobber returns of EPERM for other reasons?
433 			 */
434 			if (error == 0 || error == ENOSPC || error == EPERM)
435 				error = EINVAL;	/* EOF */
436 		}
437 		return (error);
438 
439 	case PT_READ_U:
440 		if ((uintptr_t)uap->addr > UPAGES * PAGE_SIZE - sizeof(int)) {
441 			return EFAULT;
442 		}
443 		if ((uintptr_t)uap->addr & (sizeof(int) - 1)) {
444 			return EFAULT;
445 		}
446 		if (ptrace_read_u_check(p,(vm_offset_t) uap->addr,
447 					sizeof(int))) {
448 			return EFAULT;
449 		}
450 		error = 0;
451 		PHOLD(p);	/* user had damn well better be incore! */
452 		mtx_lock_spin(&sched_lock);
453 		if (p->p_sflag & PS_INMEM) {
454 			mtx_unlock_spin(&sched_lock);
455 			fill_kinfo_proc (p, &p->p_addr->u_kproc);
456 			curp->p_retval[0] = *(int *)
457 			    ((uintptr_t)p->p_addr + (uintptr_t)uap->addr);
458 		} else {
459 			mtx_unlock_spin(&sched_lock);
460 			curp->p_retval[0] = 0;
461 			error = EFAULT;
462 		}
463 		PRELE(p);
464 		return error;
465 
466 	case PT_WRITE_U:
467 		PHOLD(p);	/* user had damn well better be incore! */
468 		mtx_lock_spin(&sched_lock);
469 		if (p->p_sflag & PS_INMEM) {
470 			mtx_unlock_spin(&sched_lock);
471 			fill_kinfo_proc (p, &p->p_addr->u_kproc);
472 			error = ptrace_write_u(p, (vm_offset_t)uap->addr, uap->data);
473 		} else {
474 			mtx_unlock_spin(&sched_lock);
475 			error = EFAULT;
476 		}
477 		PRELE(p);
478 		return error;
479 
480 	case PT_KILL:
481 		uap->data = SIGKILL;
482 		goto sendsig;	/* in PT_CONTINUE above */
483 
484 #ifdef PT_SETREGS
485 	case PT_SETREGS:
486 		write = 1;
487 		/* fallthrough */
488 #endif /* PT_SETREGS */
489 #ifdef PT_GETREGS
490 	case PT_GETREGS:
491 		/* write = 0 above */
492 #endif /* PT_SETREGS */
493 #if defined(PT_SETREGS) || defined(PT_GETREGS)
494 		if (!procfs_validregs(p))	/* no P_SYSTEM procs please */
495 			return EINVAL;
496 		else {
497 			iov.iov_base = uap->addr;
498 			iov.iov_len = sizeof(struct reg);
499 			uio.uio_iov = &iov;
500 			uio.uio_iovcnt = 1;
501 			uio.uio_offset = 0;
502 			uio.uio_resid = sizeof(struct reg);
503 			uio.uio_segflg = UIO_USERSPACE;
504 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
505 			uio.uio_procp = curp;
506 			return (procfs_doregs(curp, p, NULL, &uio));
507 		}
508 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */
509 
510 #ifdef PT_SETFPREGS
511 	case PT_SETFPREGS:
512 		write = 1;
513 		/* fallthrough */
514 #endif /* PT_SETFPREGS */
515 #ifdef PT_GETFPREGS
516 	case PT_GETFPREGS:
517 		/* write = 0 above */
518 #endif /* PT_SETFPREGS */
519 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
520 		if (!procfs_validfpregs(p))	/* no P_SYSTEM procs please */
521 			return EINVAL;
522 		else {
523 			iov.iov_base = uap->addr;
524 			iov.iov_len = sizeof(struct fpreg);
525 			uio.uio_iov = &iov;
526 			uio.uio_iovcnt = 1;
527 			uio.uio_offset = 0;
528 			uio.uio_resid = sizeof(struct fpreg);
529 			uio.uio_segflg = UIO_USERSPACE;
530 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
531 			uio.uio_procp = curp;
532 			return (procfs_dofpregs(curp, p, NULL, &uio));
533 		}
534 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */
535 
536 #ifdef PT_SETDBREGS
537 	case PT_SETDBREGS:
538 		write = 1;
539 		/* fallthrough */
540 #endif /* PT_SETDBREGS */
541 #ifdef PT_GETDBREGS
542 	case PT_GETDBREGS:
543 		/* write = 0 above */
544 #endif /* PT_SETDBREGS */
545 #if defined(PT_SETDBREGS) || defined(PT_GETDBREGS)
546 		if (!procfs_validdbregs(p))	/* no P_SYSTEM procs please */
547 			return EINVAL;
548 		else {
549 			iov.iov_base = uap->addr;
550 			iov.iov_len = sizeof(struct dbreg);
551 			uio.uio_iov = &iov;
552 			uio.uio_iovcnt = 1;
553 			uio.uio_offset = 0;
554 			uio.uio_resid = sizeof(struct dbreg);
555 			uio.uio_segflg = UIO_USERSPACE;
556 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
557 			uio.uio_procp = curp;
558 			return (procfs_dodbregs(curp, p, NULL, &uio));
559 		}
560 #endif /* defined(PT_SETDBREGS) || defined(PT_GETDBREGS) */
561 
562 	default:
563 		break;
564 	}
565 
566 	return 0;
567 }
568 
569 int
570 trace_req(p)
571 	struct proc *p;
572 {
573 	return 1;
574 }
575 
576 /*
577  * stopevent()
578  * Stop a process because of a procfs event;
579  * stay stopped until p->p_step is cleared
580  * (cleared by PIOCCONT in procfs).
581  *
582  * Must be called with the proc struct mutex held.
583  */
584 
585 void
586 stopevent(p, event, val)
587 	struct proc *p;
588 	unsigned int event;
589 	unsigned int val;
590 {
591 
592 	PROC_LOCK_ASSERT(p, MA_OWNED | MA_NOTRECURSED);
593 	p->p_step = 1;
594 
595 	do {
596 		p->p_xstat = val;
597 		p->p_stype = event;	/* Which event caused the stop? */
598 		wakeup(&p->p_stype);	/* Wake up any PIOCWAIT'ing procs */
599 		msleep(&p->p_step, &p->p_mtx, PWAIT, "stopevent", 0);
600 	} while (p->p_step);
601 }
602