xref: /dflybsd-src/sys/kern/sys_process.c (revision 90ea502b8c5d21f908cedff6680ee2bc9e74ce74)
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: src/sys/kern/sys_process.c,v 1.51.2.6 2003/01/08 03:06:45 kan Exp $
32  * $DragonFly: src/sys/kern/sys_process.c,v 1.30 2007/02/19 01:14:23 corecode Exp $
33  */
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/sysproto.h>
38 #include <sys/proc.h>
39 #include <sys/priv.h>
40 #include <sys/vnode.h>
41 #include <sys/ptrace.h>
42 #include <sys/reg.h>
43 #include <sys/lock.h>
44 
45 #include <vm/vm.h>
46 #include <vm/pmap.h>
47 #include <vm/vm_map.h>
48 #include <vm/vm_page.h>
49 
50 #include <sys/user.h>
51 #include <vfs/procfs/procfs.h>
52 #include <sys/thread2.h>
53 
54 /* use the equivalent procfs code */
55 #if 0
56 static int
57 pread (struct proc *procp, unsigned int addr, unsigned int *retval) {
58 	int		rv;
59 	vm_map_t	map, tmap;
60 	vm_object_t	object;
61 	vm_offset_t	kva = 0;
62 	int		page_offset;	/* offset into page */
63 	vm_offset_t	pageno;		/* page number */
64 	vm_map_entry_t	out_entry;
65 	vm_prot_t	out_prot;
66 	boolean_t	wired;
67 	vm_pindex_t	pindex;
68 
69 	/* Map page into kernel space */
70 
71 	map = &procp->p_vmspace->vm_map;
72 
73 	page_offset = addr - trunc_page(addr);
74 	pageno = trunc_page(addr);
75 
76 	tmap = map;
77 	rv = vm_map_lookup (&tmap, pageno, VM_PROT_READ, &out_entry,
78 		&object, &pindex, &out_prot, &wired);
79 
80 	if (rv != KERN_SUCCESS)
81 		return EINVAL;
82 
83 	vm_map_lookup_done (tmap, out_entry, 0);
84 
85 	/* Find space in kernel_map for the page we're interested in */
86 	rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex),
87 			  &kva, PAGE_SIZE,
88 			  0,
89 			  VM_MAPTYPE_NORMAL,
90 			  VM_PROT_ALL, VM_PROT_ALL,
91 			  0);
92 
93 	if (!rv) {
94 		vm_object_reference (object);
95 
96 		rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0);
97 		if (!rv) {
98 			*retval = 0;
99 			bcopy ((caddr_t)kva + page_offset,
100 			       retval, sizeof *retval);
101 		}
102 		vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE);
103 	}
104 
105 	return rv;
106 }
107 
108 static int
109 pwrite (struct proc *procp, unsigned int addr, unsigned int datum) {
110 	int		rv;
111 	vm_map_t	map, tmap;
112 	vm_object_t	object;
113 	vm_offset_t	kva = 0;
114 	int		page_offset;	/* offset into page */
115 	vm_offset_t	pageno;		/* page number */
116 	vm_map_entry_t	out_entry;
117 	vm_prot_t	out_prot;
118 	boolean_t	wired;
119 	vm_pindex_t	pindex;
120 	boolean_t	fix_prot = 0;
121 
122 	/* Map page into kernel space */
123 
124 	map = &procp->p_vmspace->vm_map;
125 
126 	page_offset = addr - trunc_page(addr);
127 	pageno = trunc_page(addr);
128 
129 	/*
130 	 * Check the permissions for the area we're interested in.
131 	 */
132 
133 	if (vm_map_check_protection (map, pageno, pageno + PAGE_SIZE,
134 		VM_PROT_WRITE) == FALSE) {
135 		/*
136 		 * If the page was not writable, we make it so.
137 		 * XXX It is possible a page may *not* be read/executable,
138 		 * if a process changes that!
139 		 */
140 		fix_prot = 1;
141 		/* The page isn't writable, so let's try making it so... */
142 		if ((rv = vm_map_protect (map, pageno, pageno + PAGE_SIZE,
143 			VM_PROT_ALL, 0)) != KERN_SUCCESS)
144 		  return EFAULT;	/* I guess... */
145 	}
146 
147 	/*
148 	 * Now we need to get the page.  out_entry, out_prot, wired, and
149 	 * single_use aren't used.  One would think the vm code would be
150 	 * a *bit* nicer...  We use tmap because vm_map_lookup() can
151 	 * change the map argument.
152 	 */
153 
154 	tmap = map;
155 	rv = vm_map_lookup (&tmap, pageno, VM_PROT_WRITE, &out_entry,
156 		&object, &pindex, &out_prot, &wired);
157 	if (rv != KERN_SUCCESS) {
158 		return EINVAL;
159 	}
160 
161 	/*
162 	 * Okay, we've got the page.  Let's release tmap.
163 	 */
164 
165 	vm_map_lookup_done (tmap, out_entry, 0);
166 
167 	/*
168 	 * Fault the page in...
169 	 */
170 
171 	rv = vm_fault(map, pageno, VM_PROT_WRITE|VM_PROT_READ, FALSE);
172 	if (rv != KERN_SUCCESS)
173 		return EFAULT;
174 
175 	/* Find space in kernel_map for the page we're interested in */
176 	rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex),
177 			  &kva, PAGE_SIZE,
178 			  0,
179 			  VM_MAPTYPE_NORMAL,
180 			  VM_PROT_ALL, VM_PROT_ALL,
181 			  0);
182 	if (!rv) {
183 		vm_object_reference (object);
184 
185 		rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0);
186 		if (!rv) {
187 		  bcopy (&datum, (caddr_t)kva + page_offset, sizeof datum);
188 		}
189 		vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE);
190 	}
191 
192 	if (fix_prot)
193 		vm_map_protect (map, pageno, pageno + PAGE_SIZE,
194 			VM_PROT_READ|VM_PROT_EXECUTE, 0);
195 	return rv;
196 }
197 #endif
198 
199 /*
200  * Process debugging system call.
201  *
202  * MPALMOSTSAFE
203  */
204 int
205 sys_ptrace(struct ptrace_args *uap)
206 {
207 	struct proc *p = curproc;
208 
209 	/*
210 	 * XXX this obfuscation is to reduce stack usage, but the register
211 	 * structs may be too large to put on the stack anyway.
212 	 */
213 	union {
214 		struct ptrace_io_desc piod;
215 		struct dbreg dbreg;
216 		struct fpreg fpreg;
217 		struct reg reg;
218 	} r;
219 	void *addr;
220 	int error = 0;
221 
222 	addr = &r;
223 	switch (uap->req) {
224 	case PT_GETREGS:
225 	case PT_GETFPREGS:
226 #ifdef PT_GETDBREGS
227 	case PT_GETDBREGS:
228 #endif
229 		break;
230 	case PT_SETREGS:
231 		error = copyin(uap->addr, &r.reg, sizeof r.reg);
232 		break;
233 	case PT_SETFPREGS:
234 		error = copyin(uap->addr, &r.fpreg, sizeof r.fpreg);
235 		break;
236 #ifdef PT_SETDBREGS
237 	case PT_SETDBREGS:
238 		error = copyin(uap->addr, &r.dbreg, sizeof r.dbreg);
239 		break;
240 #endif
241 	case PT_IO:
242 		error = copyin(uap->addr, &r.piod, sizeof r.piod);
243 		break;
244 	default:
245 		addr = uap->addr;
246 	}
247 	if (error)
248 		return (error);
249 
250 	get_mplock();
251 	error = kern_ptrace(p, uap->req, uap->pid, addr, uap->data,
252 			&uap->sysmsg_result);
253 	rel_mplock();
254 	if (error)
255 		return (error);
256 
257 	switch (uap->req) {
258 	case PT_IO:
259 		(void)copyout(&r.piod, uap->addr, sizeof r.piod);
260 		break;
261 	case PT_GETREGS:
262 		error = copyout(&r.reg, uap->addr, sizeof r.reg);
263 		break;
264 	case PT_GETFPREGS:
265 		error = copyout(&r.fpreg, uap->addr, sizeof r.fpreg);
266 		break;
267 #ifdef PT_GETDBREGS
268 	case PT_GETDBREGS:
269 		error = copyout(&r.dbreg, uap->addr, sizeof r.dbreg);
270 		break;
271 #endif
272 	}
273 
274 	return (error);
275 }
276 
277 int
278 kern_ptrace(struct proc *curp, int req, pid_t pid, void *addr, int data, int *res)
279 {
280 	struct proc *p, *pp;
281 	struct lwp *lp;
282 	struct iovec iov;
283 	struct uio uio;
284 	struct ptrace_io_desc *piod;
285 	int error = 0;
286 	int write, tmp;
287 
288 	write = 0;
289 	if (req == PT_TRACE_ME) {
290 		p = curp;
291 	} else {
292 		if ((p = pfind(pid)) == NULL)
293 			return ESRCH;
294 	}
295 	if (!PRISON_CHECK(curp->p_ucred, p->p_ucred))
296 		return (ESRCH);
297 
298 	/* Can't trace a process that's currently exec'ing. */
299 	if ((p->p_flag & P_INEXEC) != 0)
300 		return EAGAIN;
301 
302 	/*
303 	 * Permissions check
304 	 */
305 	switch (req) {
306 	case PT_TRACE_ME:
307 		/* Always legal. */
308 		break;
309 
310 	case PT_ATTACH:
311 		/* Self */
312 		if (p->p_pid == curp->p_pid)
313 			return EINVAL;
314 
315 		/* Already traced */
316 		if (p->p_flag & P_TRACED)
317 			return EBUSY;
318 
319 		if (curp->p_flag & P_TRACED)
320 			for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr)
321 				if (pp == p)
322 					return (EINVAL);
323 
324 		/* not owned by you, has done setuid (unless you're root) */
325 		if ((p->p_ucred->cr_ruid != curp->p_ucred->cr_ruid) ||
326 		     (p->p_flag & P_SUGID)) {
327 			if ((error = priv_check_cred(curp->p_ucred, PRIV_ROOT, 0)) != 0)
328 				return error;
329 		}
330 
331 		/* can't trace init when securelevel > 0 */
332 		if (securelevel > 0 && p->p_pid == 1)
333 			return EPERM;
334 
335 		/* OK */
336 		break;
337 
338 	case PT_READ_I:
339 	case PT_READ_D:
340 	case PT_WRITE_I:
341 	case PT_WRITE_D:
342 	case PT_IO:
343 	case PT_CONTINUE:
344 	case PT_KILL:
345 	case PT_STEP:
346 	case PT_DETACH:
347 #ifdef PT_GETREGS
348 	case PT_GETREGS:
349 #endif
350 #ifdef PT_SETREGS
351 	case PT_SETREGS:
352 #endif
353 #ifdef PT_GETFPREGS
354 	case PT_GETFPREGS:
355 #endif
356 #ifdef PT_SETFPREGS
357 	case PT_SETFPREGS:
358 #endif
359 #ifdef PT_GETDBREGS
360 	case PT_GETDBREGS:
361 #endif
362 #ifdef PT_SETDBREGS
363 	case PT_SETDBREGS:
364 #endif
365 		/* not being traced... */
366 		if ((p->p_flag & P_TRACED) == 0)
367 			return EPERM;
368 
369 		/* not being traced by YOU */
370 		if (p->p_pptr != curp)
371 			return EBUSY;
372 
373 		/* not currently stopped */
374 		if (p->p_stat != SSTOP ||
375 		    (p->p_flag & P_WAITED) == 0) {
376 			return EBUSY;
377 		}
378 
379 		/* OK */
380 		break;
381 
382 	default:
383 		return EINVAL;
384 	}
385 
386 	/* XXX lwp */
387 	lp = FIRST_LWP_IN_PROC(p);
388 #ifdef FIX_SSTEP
389 	/*
390 	 * Single step fixup ala procfs
391 	 */
392 	FIX_SSTEP(lp);
393 #endif
394 
395 	/*
396 	 * Actually do the requests
397 	 */
398 
399 	*res = 0;
400 
401 	switch (req) {
402 	case PT_TRACE_ME:
403 		/* set my trace flag and "owner" so it can read/write me */
404 		p->p_flag |= P_TRACED;
405 		p->p_oppid = p->p_pptr->p_pid;
406 		return 0;
407 
408 	case PT_ATTACH:
409 		/* security check done above */
410 		p->p_flag |= P_TRACED;
411 		p->p_oppid = p->p_pptr->p_pid;
412 		if (p->p_pptr != curp)
413 			proc_reparent(p, curp);
414 		data = SIGSTOP;
415 		goto sendsig;	/* in PT_CONTINUE below */
416 
417 	case PT_STEP:
418 	case PT_CONTINUE:
419 	case PT_DETACH:
420 		/* Zero means do not send any signal */
421 		if (data < 0 || data > _SIG_MAXSIG)
422 			return EINVAL;
423 
424 		LWPHOLD(lp);
425 
426 		if (req == PT_STEP) {
427 			if ((error = ptrace_single_step (lp))) {
428 				LWPRELE(lp);
429 				return error;
430 			}
431 		}
432 
433 		if (addr != (void *)1) {
434 			if ((error = ptrace_set_pc (lp,
435 			    (u_long)(uintfptr_t)addr))) {
436 				LWPRELE(lp);
437 				return error;
438 			}
439 		}
440 		LWPRELE(lp);
441 
442 		if (req == PT_DETACH) {
443 			/* reset process parent */
444 			if (p->p_oppid != p->p_pptr->p_pid) {
445 				struct proc *pp;
446 
447 				pp = pfind(p->p_oppid);
448 				proc_reparent(p, pp ? pp : initproc);
449 			}
450 
451 			p->p_flag &= ~(P_TRACED | P_WAITED);
452 			p->p_oppid = 0;
453 
454 			/* should we send SIGCHLD? */
455 		}
456 
457 	sendsig:
458 		/*
459 		 * Deliver or queue signal.  If the process is stopped
460 		 * force it to be SACTIVE again.
461 		 */
462 		crit_enter();
463 		if (p->p_stat == SSTOP) {
464 			p->p_xstat = data;
465 			lp->lwp_flag |= LWP_BREAKTSLEEP;
466 			proc_unstop(p);
467 		} else if (data) {
468 			ksignal(p, data);
469 		}
470 		crit_exit();
471 		return 0;
472 
473 	case PT_WRITE_I:
474 	case PT_WRITE_D:
475 		write = 1;
476 		/* fallthrough */
477 	case PT_READ_I:
478 	case PT_READ_D:
479 		/*
480 		 * NOTE! uio_offset represents the offset in the target
481 		 * process.  The iov is in the current process (the guy
482 		 * making the ptrace call) so uio_td must be the current
483 		 * process (though for a SYSSPACE transfer it doesn't
484 		 * really matter).
485 		 */
486 		tmp = 0;
487 		/* write = 0 set above */
488 		iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp;
489 		iov.iov_len = sizeof(int);
490 		uio.uio_iov = &iov;
491 		uio.uio_iovcnt = 1;
492 		uio.uio_offset = (off_t)(uintptr_t)addr;
493 		uio.uio_resid = sizeof(int);
494 		uio.uio_segflg = UIO_SYSSPACE;
495 		uio.uio_rw = write ? UIO_WRITE : UIO_READ;
496 		uio.uio_td = curthread;
497 		error = procfs_domem(curp, lp, NULL, &uio);
498 		if (uio.uio_resid != 0) {
499 			/*
500 			 * XXX procfs_domem() doesn't currently return ENOSPC,
501 			 * so I think write() can bogusly return 0.
502 			 * XXX what happens for short writes?  We don't want
503 			 * to write partial data.
504 			 * XXX procfs_domem() returns EPERM for other invalid
505 			 * addresses.  Convert this to EINVAL.  Does this
506 			 * clobber returns of EPERM for other reasons?
507 			 */
508 			if (error == 0 || error == ENOSPC || error == EPERM)
509 				error = EINVAL;	/* EOF */
510 		}
511 		if (!write)
512 			*res = tmp;
513 		return (error);
514 
515 	case PT_IO:
516 		/*
517 		 * NOTE! uio_offset represents the offset in the target
518 		 * process.  The iov is in the current process (the guy
519 		 * making the ptrace call) so uio_td must be the current
520 		 * process.
521 		 */
522 		piod = addr;
523 		iov.iov_base = piod->piod_addr;
524 		iov.iov_len = piod->piod_len;
525 		uio.uio_iov = &iov;
526 		uio.uio_iovcnt = 1;
527 		uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs;
528 		uio.uio_resid = piod->piod_len;
529 		uio.uio_segflg = UIO_USERSPACE;
530 		uio.uio_td = curthread;
531 		switch (piod->piod_op) {
532 		case PIOD_READ_D:
533 		case PIOD_READ_I:
534 			uio.uio_rw = UIO_READ;
535 			break;
536 		case PIOD_WRITE_D:
537 		case PIOD_WRITE_I:
538 			uio.uio_rw = UIO_WRITE;
539 			break;
540 		default:
541 			return (EINVAL);
542 		}
543 		error = procfs_domem(curp, lp, NULL, &uio);
544 		piod->piod_len -= uio.uio_resid;
545 		return (error);
546 
547 	case PT_KILL:
548 		data = SIGKILL;
549 		goto sendsig;	/* in PT_CONTINUE above */
550 
551 #ifdef PT_SETREGS
552 	case PT_SETREGS:
553 		write = 1;
554 		/* fallthrough */
555 #endif /* PT_SETREGS */
556 #ifdef PT_GETREGS
557 	case PT_GETREGS:
558 		/* write = 0 above */
559 #endif /* PT_SETREGS */
560 #if defined(PT_SETREGS) || defined(PT_GETREGS)
561 		if (!procfs_validregs(lp))	/* no P_SYSTEM procs please */
562 			return EINVAL;
563 		else {
564 			iov.iov_base = addr;
565 			iov.iov_len = sizeof(struct reg);
566 			uio.uio_iov = &iov;
567 			uio.uio_iovcnt = 1;
568 			uio.uio_offset = 0;
569 			uio.uio_resid = sizeof(struct reg);
570 			uio.uio_segflg = UIO_SYSSPACE;
571 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
572 			uio.uio_td = curthread;
573 			return (procfs_doregs(curp, lp, NULL, &uio));
574 		}
575 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */
576 
577 #ifdef PT_SETFPREGS
578 	case PT_SETFPREGS:
579 		write = 1;
580 		/* fallthrough */
581 #endif /* PT_SETFPREGS */
582 #ifdef PT_GETFPREGS
583 	case PT_GETFPREGS:
584 		/* write = 0 above */
585 #endif /* PT_SETFPREGS */
586 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
587 		if (!procfs_validfpregs(lp))	/* no P_SYSTEM procs please */
588 			return EINVAL;
589 		else {
590 			iov.iov_base = addr;
591 			iov.iov_len = sizeof(struct fpreg);
592 			uio.uio_iov = &iov;
593 			uio.uio_iovcnt = 1;
594 			uio.uio_offset = 0;
595 			uio.uio_resid = sizeof(struct fpreg);
596 			uio.uio_segflg = UIO_SYSSPACE;
597 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
598 			uio.uio_td = curthread;
599 			return (procfs_dofpregs(curp, lp, NULL, &uio));
600 		}
601 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */
602 
603 #ifdef PT_SETDBREGS
604 	case PT_SETDBREGS:
605 		write = 1;
606 		/* fallthrough */
607 #endif /* PT_SETDBREGS */
608 #ifdef PT_GETDBREGS
609 	case PT_GETDBREGS:
610 		/* write = 0 above */
611 #endif /* PT_SETDBREGS */
612 #if defined(PT_SETDBREGS) || defined(PT_GETDBREGS)
613 		if (!procfs_validdbregs(lp))	/* no P_SYSTEM procs please */
614 			return EINVAL;
615 		else {
616 			iov.iov_base = addr;
617 			iov.iov_len = sizeof(struct dbreg);
618 			uio.uio_iov = &iov;
619 			uio.uio_iovcnt = 1;
620 			uio.uio_offset = 0;
621 			uio.uio_resid = sizeof(struct dbreg);
622 			uio.uio_segflg = UIO_SYSSPACE;
623 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
624 			uio.uio_td = curthread;
625 			return (procfs_dodbregs(curp, lp, NULL, &uio));
626 		}
627 #endif /* defined(PT_SETDBREGS) || defined(PT_GETDBREGS) */
628 
629 	default:
630 		break;
631 	}
632 
633 	return 0;
634 }
635 
636 int
637 trace_req(struct proc *p)
638 {
639 	return 1;
640 }
641 
642 /*
643  * stopevent()
644  *
645  * Stop a process because of a procfs event.  Stay stopped until p->p_step
646  * is cleared (cleared by PIOCCONT in procfs).
647  *
648  * MPSAFE
649  */
650 void
651 stopevent(struct proc *p, unsigned int event, unsigned int val)
652 {
653 	p->p_step = 1;
654 
655 	do {
656 		crit_enter();
657 		wakeup(&p->p_stype);	/* Wake up any PIOCWAIT'ing procs */
658 		p->p_xstat = val;
659 		p->p_stype = event;	/* Which event caused the stop? */
660 		tsleep(&p->p_step, 0, "stopevent", 0);
661 		crit_exit();
662 	} while (p->p_step);
663 }
664 
665