xref: /freebsd/sys/kern/sys_process.c (revision a8445737e740901f5f2c8d24c12ef7fc8b00134e)
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  *	$Id: sys_process.c,v 1.39 1998/07/15 04:43:49 bde Exp $
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 <vm/vm_prot.h>
44 #include <sys/lock.h>
45 #include <vm/pmap.h>
46 #include <vm/vm_map.h>
47 #include <vm/vm_page.h>
48 #include <vm/vm_extern.h>
49 
50 #include <sys/user.h>
51 #include <miscfs/procfs/procfs.h>
52 
53 /* use the equivalent procfs code */
54 #if 0
55 static int
56 pread (struct proc *procp, unsigned int addr, unsigned int *retval) {
57 	int		rv;
58 	vm_map_t	map, tmap;
59 	vm_object_t	object;
60 	vm_offset_t	kva = 0;
61 	int		page_offset;	/* offset into page */
62 	vm_offset_t	pageno;		/* page number */
63 	vm_map_entry_t	out_entry;
64 	vm_prot_t	out_prot;
65 	boolean_t	wired;
66 	vm_pindex_t	pindex;
67 
68 	/* Map page into kernel space */
69 
70 	map = &procp->p_vmspace->vm_map;
71 
72 	page_offset = addr - trunc_page(addr);
73 	pageno = trunc_page(addr);
74 
75 	tmap = map;
76 	rv = vm_map_lookup (&tmap, pageno, VM_PROT_READ, &out_entry,
77 		&object, &pindex, &out_prot, &wired);
78 
79 	if (rv != KERN_SUCCESS)
80 		return EINVAL;
81 
82 	vm_map_lookup_done (tmap, out_entry);
83 
84 	/* Find space in kernel_map for the page we're interested in */
85 	rv = vm_map_find (kernel_map, object, IDX_TO_OFF(pindex),
86 		&kva, PAGE_SIZE, 0, VM_PROT_ALL, VM_PROT_ALL, 0);
87 
88 	if (!rv) {
89 		vm_object_reference (object);
90 
91 		rv = vm_map_pageable (kernel_map, kva, kva + PAGE_SIZE, 0);
92 		if (!rv) {
93 			*retval = 0;
94 			bcopy ((caddr_t)kva + page_offset,
95 			       retval, sizeof *retval);
96 		}
97 		vm_map_remove (kernel_map, kva, kva + PAGE_SIZE);
98 	}
99 
100 	return rv;
101 }
102 
103 static int
104 pwrite (struct proc *procp, unsigned int addr, unsigned int datum) {
105 	int		rv;
106 	vm_map_t	map, tmap;
107 	vm_object_t	object;
108 	vm_offset_t	kva = 0;
109 	int		page_offset;	/* offset into page */
110 	vm_offset_t	pageno;		/* page number */
111 	vm_map_entry_t	out_entry;
112 	vm_prot_t	out_prot;
113 	boolean_t	wired;
114 	vm_pindex_t	pindex;
115 	boolean_t	fix_prot = 0;
116 
117 	/* Map page into kernel space */
118 
119 	map = &procp->p_vmspace->vm_map;
120 
121 	page_offset = addr - trunc_page(addr);
122 	pageno = trunc_page(addr);
123 
124 	/*
125 	 * Check the permissions for the area we're interested in.
126 	 */
127 
128 	if (vm_map_check_protection (map, pageno, pageno + PAGE_SIZE,
129 		VM_PROT_WRITE) == FALSE) {
130 		/*
131 		 * If the page was not writable, we make it so.
132 		 * XXX It is possible a page may *not* be read/executable,
133 		 * if a process changes that!
134 		 */
135 		fix_prot = 1;
136 		/* The page isn't writable, so let's try making it so... */
137 		if ((rv = vm_map_protect (map, pageno, pageno + PAGE_SIZE,
138 			VM_PROT_ALL, 0)) != KERN_SUCCESS)
139 		  return EFAULT;	/* I guess... */
140 	}
141 
142 	/*
143 	 * Now we need to get the page.  out_entry, out_prot, wired, and
144 	 * single_use aren't used.  One would think the vm code would be
145 	 * a *bit* nicer...  We use tmap because vm_map_lookup() can
146 	 * change the map argument.
147 	 */
148 
149 	tmap = map;
150 	rv = vm_map_lookup (&tmap, pageno, VM_PROT_WRITE, &out_entry,
151 		&object, &pindex, &out_prot, &wired);
152 	if (rv != KERN_SUCCESS) {
153 		return EINVAL;
154 	}
155 
156 	/*
157 	 * Okay, we've got the page.  Let's release tmap.
158 	 */
159 
160 	vm_map_lookup_done (tmap, out_entry);
161 
162 	/*
163 	 * Fault the page in...
164 	 */
165 
166 	rv = vm_fault(map, pageno, VM_PROT_WRITE|VM_PROT_READ, FALSE);
167 	if (rv != KERN_SUCCESS)
168 		return EFAULT;
169 
170 	/* Find space in kernel_map for the page we're interested in */
171 	rv = vm_map_find (kernel_map, object, IDX_TO_OFF(pindex),
172 		&kva, PAGE_SIZE, 0,
173 		VM_PROT_ALL, VM_PROT_ALL, 0);
174 	if (!rv) {
175 		vm_object_reference (object);
176 
177 		rv = vm_map_pageable (kernel_map, kva, kva + PAGE_SIZE, 0);
178 		if (!rv) {
179 		  bcopy (&datum, (caddr_t)kva + page_offset, sizeof datum);
180 		}
181 		vm_map_remove (kernel_map, kva, kva + PAGE_SIZE);
182 	}
183 
184 	if (fix_prot)
185 		vm_map_protect (map, pageno, pageno + PAGE_SIZE,
186 			VM_PROT_READ|VM_PROT_EXECUTE, 0);
187 	return rv;
188 }
189 #endif
190 
191 /*
192  * Process debugging system call.
193  */
194 #ifndef _SYS_SYSPROTO_H_
195 struct ptrace_args {
196 	int	req;
197 	pid_t	pid;
198 	caddr_t	addr;
199 	int	data;
200 };
201 #endif
202 
203 int
204 ptrace(curp, uap)
205 	struct proc *curp;
206 	struct ptrace_args *uap;
207 {
208 	struct proc *p;
209 	struct iovec iov;
210 	struct uio uio;
211 	int error = 0;
212 	int write;
213 	int s;
214 
215 	if (uap->req == PT_TRACE_ME)
216 		p = curp;
217 	else {
218 		if ((p = pfind(uap->pid)) == NULL)
219 			return ESRCH;
220 	}
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 		if (p->p_flag & P_TRACED)
237 			return EBUSY;
238 
239 		/* not owned by you, has done setuid (unless you're root) */
240 		if ((p->p_cred->p_ruid != curp->p_cred->p_ruid) ||
241 		     (p->p_flag & P_SUGID)) {
242 			if (error = suser(curp->p_ucred, &curp->p_acflag))
243 				return error;
244 		}
245 
246 		/* can't trace init when securelevel > 0 */
247 		if (securelevel > 0 && p->p_pid == 1)
248 			return EPERM;
249 
250 		/* OK */
251 		break;
252 
253 	case PT_READ_I:
254 	case PT_READ_D:
255 	case PT_READ_U:
256 	case PT_WRITE_I:
257 	case PT_WRITE_D:
258 	case PT_WRITE_U:
259 	case PT_CONTINUE:
260 	case PT_KILL:
261 	case PT_STEP:
262 	case PT_DETACH:
263 #ifdef PT_GETREGS
264 	case PT_GETREGS:
265 #endif
266 #ifdef PT_SETREGS
267 	case PT_SETREGS:
268 #endif
269 #ifdef PT_GETFPREGS
270 	case PT_GETFPREGS:
271 #endif
272 #ifdef PT_SETFPREGS
273 	case PT_SETFPREGS:
274 #endif
275 		/* not being traced... */
276 		if ((p->p_flag & P_TRACED) == 0)
277 			return EPERM;
278 
279 		/* not being traced by YOU */
280 		if (p->p_pptr != curp)
281 			return EBUSY;
282 
283 		/* not currently stopped */
284 		if (p->p_stat != SSTOP || (p->p_flag & P_WAITED) == 0)
285 			return EBUSY;
286 
287 		/* OK */
288 		break;
289 
290 	default:
291 		return EINVAL;
292 	}
293 
294 #ifdef FIX_SSTEP
295 	/*
296 	 * Single step fixup ala procfs
297 	 */
298 	FIX_SSTEP(p);
299 #endif
300 
301 	/*
302 	 * Actually do the requests
303 	 */
304 
305 	write = 0;
306 	curp->p_retval[0] = 0;
307 
308 	switch (uap->req) {
309 	case PT_TRACE_ME:
310 		/* set my trace flag and "owner" so it can read/write me */
311 		p->p_flag |= P_TRACED;
312 		p->p_oppid = p->p_pptr->p_pid;
313 		return 0;
314 
315 	case PT_ATTACH:
316 		/* security check done above */
317 		p->p_flag |= P_TRACED;
318 		p->p_oppid = p->p_pptr->p_pid;
319 		if (p->p_pptr != curp)
320 			proc_reparent(p, curp);
321 		uap->data = SIGSTOP;
322 		goto sendsig;	/* in PT_CONTINUE below */
323 
324 	case PT_STEP:
325 	case PT_CONTINUE:
326 	case PT_DETACH:
327 		if ((unsigned)uap->data >= NSIG)
328 			return EINVAL;
329 
330 		PHOLD(p);
331 
332 		if (uap->req == PT_STEP) {
333 			if ((error = ptrace_single_step (p))) {
334 				PRELE(p);
335 				return error;
336 			}
337 		}
338 
339 		if (uap->addr != (caddr_t)1) {
340 			fill_eproc (p, &p->p_addr->u_kproc.kp_eproc);
341 			if ((error = ptrace_set_pc (p,
342 			    (u_long)(uintfptr_t)uap->addr))) {
343 				PRELE(p);
344 				return error;
345 			}
346 		}
347 		PRELE(p);
348 
349 		if (uap->req == PT_DETACH) {
350 			/* reset process parent */
351 			if (p->p_oppid != p->p_pptr->p_pid) {
352 				struct proc *pp;
353 
354 				pp = pfind(p->p_oppid);
355 				proc_reparent(p, pp ? pp : initproc);
356 			}
357 
358 			p->p_flag &= ~(P_TRACED | P_WAITED);
359 			p->p_oppid = 0;
360 
361 			/* should we send SIGCHLD? */
362 
363 		}
364 
365 	sendsig:
366 		/* deliver or queue signal */
367 		s = splhigh();
368 		if (p->p_stat == SSTOP) {
369 			p->p_xstat = uap->data;
370 			setrunnable(p);
371 		} else if (uap->data) {
372 			psignal(p, uap->data);
373 		}
374 		splx(s);
375 		return 0;
376 
377 	case PT_WRITE_I:
378 	case PT_WRITE_D:
379 		write = 1;
380 		/* fallthrough */
381 	case PT_READ_I:
382 	case PT_READ_D:
383 		/* write = 0 set above */
384 		iov.iov_base = write ? (caddr_t)&uap->data : (caddr_t)curp->p_retval;
385 		iov.iov_len = sizeof(int);
386 		uio.uio_iov = &iov;
387 		uio.uio_iovcnt = 1;
388 		uio.uio_offset = (off_t)(uintptr_t)uap->addr;
389 		uio.uio_resid = sizeof(int);
390 		uio.uio_segflg = UIO_SYSSPACE;	/* ie: the uap */
391 		uio.uio_rw = write ? UIO_WRITE : UIO_READ;
392 		uio.uio_procp = p;
393 		error = procfs_domem(curp, p, NULL, &uio);
394 		if (uio.uio_resid != 0) {
395 			/*
396 			 * XXX procfs_domem() doesn't currently return ENOSPC,
397 			 * so I think write() can bogusly return 0.
398 			 * XXX what happens for short writes?  We don't want
399 			 * to write partial data.
400 			 * XXX procfs_domem() returns EPERM for other invalid
401 			 * addresses.  Convert this to EINVAL.  Does this
402 			 * clobber returns of EPERM for other reasons?
403 			 */
404 			if (error == 0 || error == ENOSPC || error == EPERM)
405 				error = EINVAL;	/* EOF */
406 		}
407 		return (error);
408 
409 	case PT_READ_U:
410 		if ((uintptr_t)uap->addr > UPAGES * PAGE_SIZE - sizeof(long)) {
411 			return EFAULT;
412 		}
413 		if (ptrace_read_u_check(p,(vm_offset_t) uap->addr,
414 					sizeof(long)) &&
415 		    !procfs_kmemaccess(curp)) {
416 			return EFAULT;
417 		}
418 		error = 0;
419 		PHOLD(p);	/* user had damn well better be incore! */
420 		if (p->p_flag & P_INMEM) {
421 			p->p_addr->u_kproc.kp_proc = *p;
422 			fill_eproc (p, &p->p_addr->u_kproc.kp_eproc);
423 			curp->p_retval[0] = *(int *)
424 			    ((uintptr_t)p->p_addr + (uintptr_t)uap->addr);
425 		} else {
426 			curp->p_retval[0] = 0;
427 			error = EFAULT;
428 		}
429 		PRELE(p);
430 		return error;
431 
432 	case PT_WRITE_U:
433 		PHOLD(p);	/* user had damn well better be incore! */
434 		if (p->p_flag & P_INMEM) {
435 			p->p_addr->u_kproc.kp_proc = *p;
436 			fill_eproc (p, &p->p_addr->u_kproc.kp_eproc);
437 			error = ptrace_write_u(p, (vm_offset_t)uap->addr, uap->data);
438 		} else {
439 			error = EFAULT;
440 		}
441 		PRELE(p);
442 		return error;
443 
444 	case PT_KILL:
445 		uap->data = SIGKILL;
446 		goto sendsig;	/* in PT_CONTINUE above */
447 
448 #ifdef PT_SETREGS
449 	case PT_SETREGS:
450 		write = 1;
451 		/* fallthrough */
452 #endif /* PT_SETREGS */
453 #ifdef PT_GETREGS
454 	case PT_GETREGS:
455 		/* write = 0 above */
456 #endif /* PT_SETREGS */
457 #if defined(PT_SETREGS) || defined(PT_GETREGS)
458 		if (!procfs_validregs(p))	/* no P_SYSTEM procs please */
459 			return EINVAL;
460 		else {
461 			iov.iov_base = uap->addr;
462 			iov.iov_len = sizeof(struct reg);
463 			uio.uio_iov = &iov;
464 			uio.uio_iovcnt = 1;
465 			uio.uio_offset = 0;
466 			uio.uio_resid = sizeof(struct reg);
467 			uio.uio_segflg = UIO_USERSPACE;
468 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
469 			uio.uio_procp = curp;
470 			return (procfs_doregs(curp, p, NULL, &uio));
471 		}
472 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */
473 
474 #ifdef PT_SETFPREGS
475 	case PT_SETFPREGS:
476 		write = 1;
477 		/* fallthrough */
478 #endif /* PT_SETFPREGS */
479 #ifdef PT_GETFPREGS
480 	case PT_GETFPREGS:
481 		/* write = 0 above */
482 #endif /* PT_SETFPREGS */
483 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
484 		if (!procfs_validfpregs(p))	/* no P_SYSTEM procs please */
485 			return EINVAL;
486 		else {
487 			iov.iov_base = uap->addr;
488 			iov.iov_len = sizeof(struct fpreg);
489 			uio.uio_iov = &iov;
490 			uio.uio_iovcnt = 1;
491 			uio.uio_offset = 0;
492 			uio.uio_resid = sizeof(struct fpreg);
493 			uio.uio_segflg = UIO_USERSPACE;
494 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
495 			uio.uio_procp = curp;
496 			return (procfs_dofpregs(curp, p, NULL, &uio));
497 		}
498 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */
499 
500 	default:
501 		break;
502 	}
503 
504 	return 0;
505 }
506 
507 int
508 trace_req(p)
509 	struct proc *p;
510 {
511 	return 1;
512 }
513 
514 /*
515  * stopevent()
516  * Stop a process because of a procfs event;
517  * stay stopped until p->p_step is cleared
518  * (cleared by PIOCCONT in procfs).
519  */
520 
521 void
522 stopevent(struct proc *p, unsigned int event, unsigned int val) {
523 	p->p_step = 1;
524 
525 	do {
526 		p->p_xstat = val;
527 		p->p_stype = event;	/* Which event caused the stop? */
528 		wakeup(&p->p_stype);	/* Wake up any PIOCWAIT'ing procs */
529 		tsleep(&p->p_step, PWAIT, "stopevent", 0);
530 	} while (p->p_step);
531 }
532