?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
linprocfs - Introduce /proc/mounts
[dragonfly.git] / sys / kern / sys_process.c
blob10ba639fa535136df586c6ee5a2771ebb5534059
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
53 #include <sys/thread2.h>
54 #include <sys/mplock2.h>
55 #include <sys/spinlock2.h>
56
57 /* use the equivalent procfs code */
58 #if 0
59 static int
60 pread (struct proc *procp, unsigned int addr, unsigned int *retval) {
61 int rv;
62 vm_map_t map, tmap;
63 vm_object_t object;
64 vm_offset_t kva = 0;
65 int page_offset; /* offset into page */
66 vm_offset_t pageno; /* page number */
67 vm_map_entry_t out_entry;
68 vm_prot_t out_prot;
69 boolean_t wired;
70 vm_pindex_t pindex;
71
72 /* Map page into kernel space */
73
74 map = &procp->p_vmspace->vm_map;
75
76 page_offset = addr - trunc_page(addr);
77 pageno = trunc_page(addr);
78
79 tmap = map;
80 rv = vm_map_lookup (&tmap, pageno, VM_PROT_READ, &out_entry,
81 &object, &pindex, &out_prot, &wired);
82
83 if (rv != KERN_SUCCESS)
84 return EINVAL;
85
86 vm_map_lookup_done (tmap, out_entry, 0);
87
88 /* Find space in kernel_map for the page we're interested in */
89 rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex),
90 &kva,
91 PAGE_SIZE, PAGE_SIZE,
92 0, VM_MAPTYPE_NORMAL,
93 VM_PROT_ALL, VM_PROT_ALL,
94 0);
95
96 if (!rv) {
97 vm_object_reference (object);
98
99 rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0);
100 if (!rv) {
101 *retval = 0;
102 bcopy ((caddr_t)kva + page_offset,
103 retval, sizeof *retval);
104 }
105 vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE);
106 }
107
108 return rv;
109 }
110
111 static int
112 pwrite (struct proc *procp, unsigned int addr, unsigned int datum) {
113 int rv;
114 vm_map_t map, tmap;
115 vm_object_t object;
116 vm_offset_t kva = 0;
117 int page_offset; /* offset into page */
118 vm_offset_t pageno; /* page number */
119 vm_map_entry_t out_entry;
120 vm_prot_t out_prot;
121 boolean_t wired;
122 vm_pindex_t pindex;
123 boolean_t fix_prot = 0;
124
125 /* Map page into kernel space */
126
127 map = &procp->p_vmspace->vm_map;
128
129 page_offset = addr - trunc_page(addr);
130 pageno = trunc_page(addr);
131
132 /*
133 * Check the permissions for the area we're interested in.
134 */
135
136 if (vm_map_check_protection (map, pageno, pageno + PAGE_SIZE,
137 VM_PROT_WRITE) == FALSE) {
138 /*
139 * If the page was not writable, we make it so.
140 * XXX It is possible a page may *not* be read/executable,
141 * if a process changes that!
142 */
143 fix_prot = 1;
144 /* The page isn't writable, so let's try making it so... */
145 if ((rv = vm_map_protect (map, pageno, pageno + PAGE_SIZE,
146 VM_PROT_ALL, 0)) != KERN_SUCCESS)
147 return EFAULT; /* I guess... */
148 }
149
150 /*
151 * Now we need to get the page. out_entry, out_prot, wired, and
152 * single_use aren't used. One would think the vm code would be
153 * a *bit* nicer... We use tmap because vm_map_lookup() can
154 * change the map argument.
155 */
156
157 tmap = map;
158 rv = vm_map_lookup (&tmap, pageno, VM_PROT_WRITE, &out_entry,
159 &object, &pindex, &out_prot, &wired);
160 if (rv != KERN_SUCCESS) {
161 return EINVAL;
162 }
163
164 /*
165 * Okay, we've got the page. Let's release tmap.
166 */
167
168 vm_map_lookup_done (tmap, out_entry, 0);
169
170 /*
171 * Fault the page in...
172 */
173
174 rv = vm_fault(map, pageno, VM_PROT_WRITE|VM_PROT_READ, FALSE);
175 if (rv != KERN_SUCCESS)
176 return EFAULT;
177
178 /* Find space in kernel_map for the page we're interested in */
179 rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex),
180 &kva,
181 PAGE_SIZE, PAGE_SIZE,
182 0, VM_MAPTYPE_NORMAL,
183 VM_PROT_ALL, VM_PROT_ALL,
184 0);
185 if (!rv) {
186 vm_object_reference (object);
187
188 rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0);
189 if (!rv) {
190 bcopy (&datum, (caddr_t)kva + page_offset, sizeof datum);
191 }
192 vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE);
193 }
194
195 if (fix_prot)
196 vm_map_protect (map, pageno, pageno + PAGE_SIZE,
197 VM_PROT_READ|VM_PROT_EXECUTE, 0);
198 return rv;
199 }
200 #endif
201
202 /*
203 * Process debugging system call.
204 *
205 * MPALMOSTSAFE
206 */
207 int
208 sys_ptrace(struct ptrace_args *uap)
209 {
210 struct proc *p = curproc;
211
212 /*
213 * XXX this obfuscation is to reduce stack usage, but the register
214 * structs may be too large to put on the stack anyway.
215 */
216 union {
217 struct ptrace_io_desc piod;
218 struct dbreg dbreg;
219 struct fpreg fpreg;
220 struct reg reg;
221 } r;
222 void *addr;
223 int error = 0;
224
225 addr = &r;
226 switch (uap->req) {
227 case PT_GETREGS:
228 case PT_GETFPREGS:
229 #ifdef PT_GETDBREGS
230 case PT_GETDBREGS:
231 #endif
232 break;
233 case PT_SETREGS:
234 error = copyin(uap->addr, &r.reg, sizeof r.reg);
235 break;
236 case PT_SETFPREGS:
237 error = copyin(uap->addr, &r.fpreg, sizeof r.fpreg);
238 break;
239 #ifdef PT_SETDBREGS
240 case PT_SETDBREGS:
241 error = copyin(uap->addr, &r.dbreg, sizeof r.dbreg);
242 break;
243 #endif
244 case PT_IO:
245 error = copyin(uap->addr, &r.piod, sizeof r.piod);
246 break;
247 default:
248 addr = uap->addr;
249 }
250 if (error)
251 return (error);
252
253 get_mplock();
254 error = kern_ptrace(p, uap->req, uap->pid, addr, uap->data,
255 &uap->sysmsg_result);
256 rel_mplock();
257 if (error)
258 return (error);
259
260 switch (uap->req) {
261 case PT_IO:
262 (void)copyout(&r.piod, uap->addr, sizeof r.piod);
263 break;
264 case PT_GETREGS:
265 error = copyout(&r.reg, uap->addr, sizeof r.reg);
266 break;
267 case PT_GETFPREGS:
268 error = copyout(&r.fpreg, uap->addr, sizeof r.fpreg);
269 break;
270 #ifdef PT_GETDBREGS
271 case PT_GETDBREGS:
272 error = copyout(&r.dbreg, uap->addr, sizeof r.dbreg);
273 break;
274 #endif
275 }
276
277 return (error);
278 }
279
280 int
281 kern_ptrace(struct proc *curp, int req, pid_t pid, void *addr, int data, int *res)
282 {
283 struct proc *p, *pp;
284 struct lwp *lp;
285 struct iovec iov;
286 struct uio uio;
287 struct ptrace_io_desc *piod;
288 int error = 0;
289 int write, tmp;
290
291 write = 0;
292 if (req == PT_TRACE_ME) {
293 p = curp;
294 } else {
295 if ((p = pfind(pid)) == NULL)
296 return ESRCH;
297 }
298 if (!PRISON_CHECK(curp->p_ucred, p->p_ucred))
299 return (ESRCH);
300
301 /* Can't trace a process that's currently exec'ing. */
302 if ((p->p_flag & P_INEXEC) != 0)
303 return EAGAIN;
304
305 /*
306 * Permissions check
307 */
308 switch (req) {
309 case PT_TRACE_ME:
310 /* Always legal. */
311 break;
312
313 case PT_ATTACH:
314 /* Self */
315 if (p->p_pid == curp->p_pid)
316 return EINVAL;
317
318 /* Already traced */
319 if (p->p_flag & P_TRACED)
320 return EBUSY;
321
322 if (curp->p_flag & P_TRACED)
323 for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr)
324 if (pp == p)
325 return (EINVAL);
326
327 /* not owned by you, has done setuid (unless you're root) */
328 if ((p->p_ucred->cr_ruid != curp->p_ucred->cr_ruid) ||
329 (p->p_flag & P_SUGID)) {
330 if ((error = priv_check_cred(curp->p_ucred, PRIV_ROOT, 0)) != 0)
331 return error;
332 }
333
334 /* can't trace init when securelevel > 0 */
335 if (securelevel > 0 && p->p_pid == 1)
336 return EPERM;
337
338 /* OK */
339 break;
340
341 case PT_READ_I:
342 case PT_READ_D:
343 case PT_WRITE_I:
344 case PT_WRITE_D:
345 case PT_IO:
346 case PT_CONTINUE:
347 case PT_KILL:
348 case PT_STEP:
349 case PT_DETACH:
350 #ifdef PT_GETREGS
351 case PT_GETREGS:
352 #endif
353 #ifdef PT_SETREGS
354 case PT_SETREGS:
355 #endif
356 #ifdef PT_GETFPREGS
357 case PT_GETFPREGS:
358 #endif
359 #ifdef PT_SETFPREGS
360 case PT_SETFPREGS:
361 #endif
362 #ifdef PT_GETDBREGS
363 case PT_GETDBREGS:
364 #endif
365 #ifdef PT_SETDBREGS
366 case PT_SETDBREGS:
367 #endif
368 /* not being traced... */
369 if ((p->p_flag & P_TRACED) == 0)
370 return EPERM;
371
372 /* not being traced by YOU */
373 if (p->p_pptr != curp)
374 return EBUSY;
375
376 /* not currently stopped */
377 if (p->p_stat != SSTOP ||
378 (p->p_flag & P_WAITED) == 0) {
379 return EBUSY;
380 }
381
382 /* OK */
383 break;
384
385 default:
386 return EINVAL;
387 }
388
389 /* XXX lwp */
390 lp = FIRST_LWP_IN_PROC(p);
391 #ifdef FIX_SSTEP
392 /*
393 * Single step fixup ala procfs
394 */
395 FIX_SSTEP(lp);
396 #endif
397
398 /*
399 * Actually do the requests
400 */
401
402 *res = 0;
403
404 switch (req) {
405 case PT_TRACE_ME:
406 /* set my trace flag and "owner" so it can read/write me */
407 p->p_flag |= P_TRACED;
408 p->p_oppid = p->p_pptr->p_pid;
409 return 0;
410
411 case PT_ATTACH:
412 /* security check done above */
413 p->p_flag |= P_TRACED;
414 p->p_oppid = p->p_pptr->p_pid;
415 if (p->p_pptr != curp)
416 proc_reparent(p, curp);
417 data = SIGSTOP;
418 goto sendsig; /* in PT_CONTINUE below */
419
420 case PT_STEP:
421 case PT_CONTINUE:
422 case PT_DETACH:
423 /* Zero means do not send any signal */
424 if (data < 0 || data > _SIG_MAXSIG)
425 return EINVAL;
426
427 LWPHOLD(lp);
428
429 if (req == PT_STEP) {
430 if ((error = ptrace_single_step (lp))) {
431 LWPRELE(lp);
432 return error;
433 }
434 }
435
436 if (addr != (void *)1) {
437 if ((error = ptrace_set_pc (lp,
438 (u_long)(uintfptr_t)addr))) {
439 LWPRELE(lp);
440 return error;
441 }
442 }
443 LWPRELE(lp);
444
445 if (req == PT_DETACH) {
446 /* reset process parent */
447 if (p->p_oppid != p->p_pptr->p_pid) {
448 struct proc *pp;
449
450 pp = pfind(p->p_oppid);
451 proc_reparent(p, pp ? pp : initproc);
452 }
453
454 p->p_flag &= ~(P_TRACED | P_WAITED);
455 p->p_oppid = 0;
456
457 /* should we send SIGCHLD? */
458 }
459
460 sendsig:
461 /*
462 * Deliver or queue signal. If the process is stopped
463 * force it to be SACTIVE again.
464 */
465 crit_enter();
466 if (p->p_stat == SSTOP) {
467 p->p_xstat = data;
468 lp->lwp_flag |= LWP_BREAKTSLEEP;
469 proc_unstop(p);
470 } else if (data) {
471 ksignal(p, data);
472 }
473 crit_exit();
474 return 0;
475
476 case PT_WRITE_I:
477 case PT_WRITE_D:
478 write = 1;
479 /* fallthrough */
480 case PT_READ_I:
481 case PT_READ_D:
482 /*
483 * NOTE! uio_offset represents the offset in the target
484 * process. The iov is in the current process (the guy
485 * making the ptrace call) so uio_td must be the current
486 * process (though for a SYSSPACE transfer it doesn't
487 * really matter).
488 */
489 tmp = 0;
490 /* write = 0 set above */
491 iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp;
492 iov.iov_len = sizeof(int);
493 uio.uio_iov = &iov;
494 uio.uio_iovcnt = 1;
495 uio.uio_offset = (off_t)(uintptr_t)addr;
496 uio.uio_resid = sizeof(int);
497 uio.uio_segflg = UIO_SYSSPACE;
498 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
499 uio.uio_td = curthread;
500 error = procfs_domem(curp, lp, NULL, &uio);
501 if (uio.uio_resid != 0) {
502 /*
503 * XXX procfs_domem() doesn't currently return ENOSPC,
504 * so I think write() can bogusly return 0.
505 * XXX what happens for short writes? We don't want
506 * to write partial data.
507 * XXX procfs_domem() returns EPERM for other invalid
508 * addresses. Convert this to EINVAL. Does this
509 * clobber returns of EPERM for other reasons?
510 */
511 if (error == 0 || error == ENOSPC || error == EPERM)
512 error = EINVAL; /* EOF */
513 }
514 if (!write)
515 *res = tmp;
516 return (error);
517
518 case PT_IO:
519 /*
520 * NOTE! uio_offset represents the offset in the target
521 * process. The iov is in the current process (the guy
522 * making the ptrace call) so uio_td must be the current
523 * process.
524 */
525 piod = addr;
526 iov.iov_base = piod->piod_addr;
527 iov.iov_len = piod->piod_len;
528 uio.uio_iov = &iov;
529 uio.uio_iovcnt = 1;
530 uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs;
531 uio.uio_resid = piod->piod_len;
532 uio.uio_segflg = UIO_USERSPACE;
533 uio.uio_td = curthread;
534 switch (piod->piod_op) {
535 case PIOD_READ_D:
536 case PIOD_READ_I:
537 uio.uio_rw = UIO_READ;
538 break;
539 case PIOD_WRITE_D:
540 case PIOD_WRITE_I:
541 uio.uio_rw = UIO_WRITE;
542 break;
543 default:
544 return (EINVAL);
545 }
546 error = procfs_domem(curp, lp, NULL, &uio);
547 piod->piod_len -= uio.uio_resid;
548 return (error);
549
550 case PT_KILL:
551 data = SIGKILL;
552 goto sendsig; /* in PT_CONTINUE above */
553
554 #ifdef PT_SETREGS
555 case PT_SETREGS:
556 write = 1;
557 /* fallthrough */
558 #endif /* PT_SETREGS */
559 #ifdef PT_GETREGS
560 case PT_GETREGS:
561 /* write = 0 above */
562 #endif /* PT_SETREGS */
563 #if defined(PT_SETREGS) || defined(PT_GETREGS)
564 if (!procfs_validregs(lp)) /* no P_SYSTEM procs please */
565 return EINVAL;
566 else {
567 iov.iov_base = addr;
568 iov.iov_len = sizeof(struct reg);
569 uio.uio_iov = &iov;
570 uio.uio_iovcnt = 1;
571 uio.uio_offset = 0;
572 uio.uio_resid = sizeof(struct reg);
573 uio.uio_segflg = UIO_SYSSPACE;
574 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
575 uio.uio_td = curthread;
576 return (procfs_doregs(curp, lp, NULL, &uio));
577 }
578 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */
579
580 #ifdef PT_SETFPREGS
581 case PT_SETFPREGS:
582 write = 1;
583 /* fallthrough */
584 #endif /* PT_SETFPREGS */
585 #ifdef PT_GETFPREGS
586 case PT_GETFPREGS:
587 /* write = 0 above */
588 #endif /* PT_SETFPREGS */
589 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
590 if (!procfs_validfpregs(lp)) /* no P_SYSTEM procs please */
591 return EINVAL;
592 else {
593 iov.iov_base = addr;
594 iov.iov_len = sizeof(struct fpreg);
595 uio.uio_iov = &iov;
596 uio.uio_iovcnt = 1;
597 uio.uio_offset = 0;
598 uio.uio_resid = sizeof(struct fpreg);
599 uio.uio_segflg = UIO_SYSSPACE;
600 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
601 uio.uio_td = curthread;
602 return (procfs_dofpregs(curp, lp, NULL, &uio));
603 }
604 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */
605
606 #ifdef PT_SETDBREGS
607 case PT_SETDBREGS:
608 write = 1;
609 /* fallthrough */
610 #endif /* PT_SETDBREGS */
611 #ifdef PT_GETDBREGS
612 case PT_GETDBREGS:
613 /* write = 0 above */
614 #endif /* PT_SETDBREGS */
615 #if defined(PT_SETDBREGS) || defined(PT_GETDBREGS)
616 if (!procfs_validdbregs(lp)) /* no P_SYSTEM procs please */
617 return EINVAL;
618 else {
619 iov.iov_base = addr;
620 iov.iov_len = sizeof(struct dbreg);
621 uio.uio_iov = &iov;
622 uio.uio_iovcnt = 1;
623 uio.uio_offset = 0;
624 uio.uio_resid = sizeof(struct dbreg);
625 uio.uio_segflg = UIO_SYSSPACE;
626 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
627 uio.uio_td = curthread;
628 return (procfs_dodbregs(curp, lp, NULL, &uio));
629 }
630 #endif /* defined(PT_SETDBREGS) || defined(PT_GETDBREGS) */
631
632 default:
633 break;
634 }
635
636 return 0;
637 }
638
639 int
640 trace_req(struct proc *p)
641 {
642 return 1;
643 }
644
645 /*
646 * stopevent()
647 *
648 * Stop a process because of a procfs event. Stay stopped until p->p_step
649 * is cleared (cleared by PIOCCONT in procfs).
650 *
651 * MPSAFE
652 */
653 void
654 stopevent(struct proc *p, unsigned int event, unsigned int val)
655 {
656 /*
657 * Set event info. Recheck p_stops in case we are
658 * racing a close() on procfs.
659 */
660 spin_lock_wr(&p->p_spin);
661 if ((p->p_stops & event) == 0) {
662 spin_unlock_wr(&p->p_spin);
663 return;
664 }
665 p->p_xstat = val;
666 p->p_stype = event;
667 p->p_step = 1;
668 tsleep_interlock(&p->p_step, 0);
669 spin_unlock_wr(&p->p_spin);
670
671 /*
672 * Wakeup any PIOCWAITing procs and wait for p_step to
673 * be cleared.
674 */
675 for (;;) {
676 wakeup(&p->p_stype);
677 tsleep(&p->p_step, PINTERLOCKED, "stopevent", 0);
678 spin_lock_wr(&p->p_spin);
679 if (p->p_step == 0) {
680 spin_unlock_wr(&p->p_spin);
681 break;
682 }
683 tsleep_interlock(&p->p_step, 0);
684 spin_unlock_wr(&p->p_spin);
685 }
686 }
687
DragonFly BSD