xref: /netbsd-src/sys/miscfs/procfs/procfs_subr.c (revision b7ae68fde0d8ef1c03714e8bbb1ee7c6118ea93b)
1 /*	$NetBSD: procfs_subr.c,v 1.69 2006/09/20 08:09:05 manu Exp $	*/
2 
3 /*
4  * Copyright (c) 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Jan-Simon Pendry.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
35  */
36 
37 /*
38  * Copyright (c) 1994 Christopher G. Demetriou.  All rights reserved.
39  * Copyright (c) 1993 Jan-Simon Pendry
40  *
41  * This code is derived from software contributed to Berkeley by
42  * Jan-Simon Pendry.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. All advertising materials mentioning features or use of this software
53  *    must display the following acknowledgement:
54  *	This product includes software developed by the University of
55  *	California, Berkeley and its contributors.
56  * 4. Neither the name of the University nor the names of its contributors
57  *    may be used to endorse or promote products derived from this software
58  *    without specific prior written permission.
59  *
60  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
61  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
62  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
63  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
64  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
65  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
66  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
67  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
68  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
69  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70  * SUCH DAMAGE.
71  *
72  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
73  */
74 
75 #include <sys/cdefs.h>
76 __KERNEL_RCSID(0, "$NetBSD: procfs_subr.c,v 1.69 2006/09/20 08:09:05 manu Exp $");
77 
78 #include <sys/param.h>
79 #include <sys/systm.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/proc.h>
83 #include <sys/vnode.h>
84 #include <sys/malloc.h>
85 #include <sys/stat.h>
86 #include <sys/file.h>
87 #include <sys/filedesc.h>
88 
89 #include <miscfs/procfs/procfs.h>
90 
91 void procfs_hashins(struct pfsnode *);
92 void procfs_hashrem(struct pfsnode *);
93 struct vnode *procfs_hashget(pid_t, pfstype, int, struct mount *);
94 
95 LIST_HEAD(pfs_hashhead, pfsnode) *pfs_hashtbl;
96 u_long	pfs_ihash;	/* size of hash table - 1 */
97 #define PFSPIDHASH(pid)	((pid) & pfs_ihash)
98 
99 struct lock pfs_hashlock;
100 struct simplelock pfs_hash_slock;
101 
102 #define	ISSET(t, f)	((t) & (f))
103 
104 /*
105  * allocate a pfsnode/vnode pair.  the vnode is
106  * referenced, and locked.
107  *
108  * the pid, pfs_type, and mount point uniquely
109  * identify a pfsnode.  the mount point is needed
110  * because someone might mount this filesystem
111  * twice.
112  *
113  * all pfsnodes are maintained on a singly-linked
114  * list.  new nodes are only allocated when they cannot
115  * be found on this list.  entries on the list are
116  * removed when the vfs reclaim entry is called.
117  *
118  * a single lock is kept for the entire list.  this is
119  * needed because the getnewvnode() function can block
120  * waiting for a vnode to become free, in which case there
121  * may be more than one process trying to get the same
122  * vnode.  this lock is only taken if we are going to
123  * call getnewvnode, since the kernel itself is single-threaded.
124  *
125  * if an entry is found on the list, then call vget() to
126  * take a reference.  this is done because there may be
127  * zero references to it and so it needs to removed from
128  * the vnode free list.
129  */
130 int
131 procfs_allocvp(mp, vpp, pid, pfs_type, fd)
132 	struct mount *mp;
133 	struct vnode **vpp;
134 	pid_t pid;
135 	pfstype pfs_type;
136 	int fd;
137 {
138 	struct pfsnode *pfs;
139 	struct vnode *vp;
140 	int error;
141 
142 	do {
143 		if ((*vpp = procfs_hashget(pid, pfs_type, fd, mp)) != NULL)
144 			return (0);
145 	} while (lockmgr(&pfs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
146 
147 	if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, &vp)) != 0) {
148 		*vpp = NULL;
149 		lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
150 		return (error);
151 	}
152 
153 	MALLOC(pfs, void *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
154 	vp->v_data = pfs;
155 
156 	pfs->pfs_pid = pid;
157 	pfs->pfs_type = pfs_type;
158 	pfs->pfs_vnode = vp;
159 	pfs->pfs_flags = 0;
160 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type, fd);
161 	pfs->pfs_fd = fd;
162 
163 	switch (pfs_type) {
164 	case PFSroot:	/* /proc = dr-xr-xr-x */
165 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
166 		vp->v_type = VDIR;
167 		vp->v_flag = VROOT;
168 		break;
169 
170 	case PFScurproc:	/* /proc/curproc = lr-xr-xr-x */
171 	case PFSself:	/* /proc/self    = lr-xr-xr-x */
172 	case PFScwd:	/* /proc/N/cwd = lr-xr-xr-x */
173 	case PFSchroot:	/* /proc/N/chroot = lr-xr-xr-x */
174 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
175 		vp->v_type = VLNK;
176 		break;
177 
178 	case PFSproc:	/* /proc/N = dr-xr-xr-x */
179 	case PFSfd:
180 		if (fd == -1) {	/* /proc/N/fd = dr-xr-xr-x */
181 			pfs->pfs_mode = S_IRUSR|S_IXUSR;
182 			vp->v_type = VDIR;
183 		} else {	/* /proc/N/fd/M = [ps-]rw------- */
184 			struct file *fp;
185 			struct vnode *vxp;
186 			struct proc *pown;
187 
188 			/* XXX can procfs_getfp() ever fail here? */
189 			if ((error = procfs_getfp(pfs, &pown, &fp)) != 0)
190 				goto bad;
191 			FILE_USE(fp);
192 
193 			pfs->pfs_mode = S_IRUSR|S_IWUSR;
194 			switch (fp->f_type) {
195 			case DTYPE_VNODE:
196 				vxp = (struct vnode *)fp->f_data;
197 
198 				/*
199 				 * We make symlinks for directories
200 				 * to avoid cycles.
201 				 */
202 				if (vxp->v_type == VDIR)
203 					goto symlink;
204 				vp->v_type = vxp->v_type;
205 				break;
206 			case DTYPE_PIPE:
207 				vp->v_type = VFIFO;
208 				break;
209 			case DTYPE_SOCKET:
210 				vp->v_type = VSOCK;
211 				break;
212 			case DTYPE_KQUEUE:
213 			case DTYPE_MISC:
214 			symlink:
215 				pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|
216 				    S_IXGRP|S_IROTH|S_IXOTH;
217 				vp->v_type = VLNK;
218 				break;
219 			default:
220 				error = EOPNOTSUPP;
221 				FILE_UNUSE(fp, proc_representative_lwp(pown));
222 				goto bad;
223 			}
224 			FILE_UNUSE(fp, proc_representative_lwp(pown));
225 		}
226 		break;
227 
228 	case PFSfile:	/* /proc/N/file = -rw------- */
229 	case PFSmem:	/* /proc/N/mem = -rw------- */
230 	case PFSregs:	/* /proc/N/regs = -rw------- */
231 	case PFSfpregs:	/* /proc/N/fpregs = -rw------- */
232 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
233 		vp->v_type = VREG;
234 		break;
235 
236 	case PFSctl:	/* /proc/N/ctl = --w------ */
237 	case PFSnote:	/* /proc/N/note = --w------ */
238 	case PFSnotepg:	/* /proc/N/notepg = --w------ */
239 		pfs->pfs_mode = S_IWUSR;
240 		vp->v_type = VREG;
241 		break;
242 
243 	case PFSmap:	/* /proc/N/map = -r--r--r-- */
244 	case PFSmaps:	/* /proc/N/maps = -r--r--r-- */
245 	case PFSstatus:	/* /proc/N/status = -r--r--r-- */
246 	case PFSstat:	/* /proc/N/stat = -r--r--r-- */
247 	case PFScmdline:	/* /proc/N/cmdline = -r--r--r-- */
248 	case PFSmeminfo:	/* /proc/meminfo = -r--r--r-- */
249 	case PFSdevices:	/* /proc/devices = -r--r--r-- */
250 	case PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
251 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
252 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
253 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
254 		vp->v_type = VREG;
255 		break;
256 
257 #ifdef __HAVE_PROCFS_MACHDEP
258 	PROCFS_MACHDEP_NODETYPE_CASES
259 		procfs_machdep_allocvp(vp);
260 		break;
261 #endif
262 
263 	default:
264 		panic("procfs_allocvp");
265 	}
266 
267 	procfs_hashins(pfs);
268 	uvm_vnp_setsize(vp, 0);
269 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
270 
271 	*vpp = vp;
272 	return (0);
273 
274  bad:
275 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
276 	FREE(pfs, M_TEMP);
277 	ungetnewvnode(vp);
278 	return (error);
279 }
280 
281 int
282 procfs_freevp(vp)
283 	struct vnode *vp;
284 {
285 	struct pfsnode *pfs = VTOPFS(vp);
286 
287 	procfs_hashrem(pfs);
288 
289 	FREE(vp->v_data, M_TEMP);
290 	vp->v_data = 0;
291 	return (0);
292 }
293 
294 int
295 procfs_rw(v)
296 	void *v;
297 {
298 	struct vop_read_args *ap = v;
299 	struct vnode *vp = ap->a_vp;
300 	struct uio *uio = ap->a_uio;
301 	struct lwp *curl;
302 	struct lwp *l;
303 	struct pfsnode *pfs = VTOPFS(vp);
304 	struct proc *p;
305 
306 	if (uio->uio_offset < 0)
307 		return EINVAL;
308 	p = PFIND(pfs->pfs_pid);
309 	if (p == 0)
310 		return ESRCH;
311 	/*
312 	 * Do not allow init to be modified while in secure mode; it
313 	 * could be duped into changing the security level.
314 	 */
315 	if (uio->uio_rw == UIO_WRITE && p == initproc && securelevel > -1)
316 		return EPERM;
317 
318 	curl = curlwp;
319 
320 	/* XXX NJWLWP
321 	 * The entire procfs interface needs work to be useful to
322 	 * a process with multiple LWPs. For the moment, we'll
323 	 * just kluge this and fail on others.
324 	 */
325 	l = proc_representative_lwp(p);
326 
327 	switch (pfs->pfs_type) {
328 	case PFSnote:
329 	case PFSnotepg:
330 		return (procfs_donote(curl, p, pfs, uio));
331 
332 	case PFSregs:
333 		return (procfs_doregs(curl, l, pfs, uio));
334 
335 	case PFSfpregs:
336 		return (procfs_dofpregs(curl, l, pfs, uio));
337 
338 	case PFSctl:
339 		return (procfs_doctl(curl, l, pfs, uio));
340 
341 	case PFSstatus:
342 		return (procfs_dostatus(curl, l, pfs, uio));
343 
344 	case PFSstat:
345 		return (procfs_do_pid_stat(curl, l, pfs, uio));
346 
347 	case PFSmap:
348 		return (procfs_domap(curl, p, pfs, uio, 0));
349 
350 	case PFSmaps:
351 		return (procfs_domap(curl, p, pfs, uio, 1));
352 
353 	case PFSmem:
354 		return (procfs_domem(curl, l, pfs, uio));
355 
356 	case PFScmdline:
357 		return (procfs_docmdline(curl, p, pfs, uio));
358 
359 	case PFSmeminfo:
360 		return (procfs_domeminfo(curl, p, pfs, uio));
361 
362 	case PFSdevices:
363 		return (procfs_dodevices(curl, p, pfs, uio));
364 
365 	case PFScpuinfo:
366 		return (procfs_docpuinfo(curl, p, pfs, uio));
367 
368 	case PFSfd:
369 		return (procfs_dofd(curl, p, pfs, uio));
370 
371 	case PFSuptime:
372 		return (procfs_douptime(curl, p, pfs, uio));
373 
374 	case PFSmounts:
375 		return (procfs_domounts(curl, p, pfs, uio));
376 
377 #ifdef __HAVE_PROCFS_MACHDEP
378 	PROCFS_MACHDEP_NODETYPE_CASES
379 		return (procfs_machdep_rw(curl, l, pfs, uio));
380 #endif
381 
382 	default:
383 		return (EOPNOTSUPP);
384 	}
385 }
386 
387 /*
388  * Get a string from userland into (bf).  Strip a trailing
389  * nl character (to allow easy access from the shell).
390  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
391  * will automatically add a nul char at the end.
392  *
393  * Returns 0 on success or the following errors
394  *
395  * EINVAL:    file offset is non-zero.
396  * EMSGSIZE:  message is longer than kernel buffer
397  * EFAULT:    user i/o buffer is not addressable
398  */
399 int
400 vfs_getuserstr(uio, bf, buflenp)
401 	struct uio *uio;
402 	char *bf;
403 	int *buflenp;
404 {
405 	int xlen;
406 	int error;
407 
408 	if (uio->uio_offset != 0)
409 		return (EINVAL);
410 
411 	xlen = *buflenp;
412 
413 	/* must be able to read the whole string in one go */
414 	if (xlen < uio->uio_resid)
415 		return (EMSGSIZE);
416 	xlen = uio->uio_resid;
417 
418 	if ((error = uiomove(bf, xlen, uio)) != 0)
419 		return (error);
420 
421 	/* allow multiple writes without seeks */
422 	uio->uio_offset = 0;
423 
424 	/* cleanup string and remove trailing newline */
425 	bf[xlen] = '\0';
426 	xlen = strlen(bf);
427 	if (xlen > 0 && bf[xlen-1] == '\n')
428 		bf[--xlen] = '\0';
429 	*buflenp = xlen;
430 
431 	return (0);
432 }
433 
434 const vfs_namemap_t *
435 vfs_findname(nm, bf, buflen)
436 	const vfs_namemap_t *nm;
437 	const char *bf;
438 	int buflen;
439 {
440 
441 	for (; nm->nm_name; nm++)
442 		if (memcmp(bf, nm->nm_name, buflen+1) == 0)
443 			return (nm);
444 
445 	return (0);
446 }
447 
448 /*
449  * Initialize pfsnode hash table.
450  */
451 void
452 procfs_hashinit()
453 {
454 	lockinit(&pfs_hashlock, PINOD, "pfs_hashlock", 0, 0);
455 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT,
456 	    M_WAITOK, &pfs_ihash);
457 	simple_lock_init(&pfs_hash_slock);
458 }
459 
460 void
461 procfs_hashreinit()
462 {
463 	struct pfsnode *pp;
464 	struct pfs_hashhead *oldhash, *hash;
465 	u_long i, oldmask, mask, val;
466 
467 	hash = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT, M_WAITOK,
468 	    &mask);
469 
470 	simple_lock(&pfs_hash_slock);
471 	oldhash = pfs_hashtbl;
472 	oldmask = pfs_ihash;
473 	pfs_hashtbl = hash;
474 	pfs_ihash = mask;
475 	for (i = 0; i <= oldmask; i++) {
476 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
477 			LIST_REMOVE(pp, pfs_hash);
478 			val = PFSPIDHASH(pp->pfs_pid);
479 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
480 		}
481 	}
482 	simple_unlock(&pfs_hash_slock);
483 	hashdone(oldhash, M_UFSMNT);
484 }
485 
486 /*
487  * Free pfsnode hash table.
488  */
489 void
490 procfs_hashdone()
491 {
492 	hashdone(pfs_hashtbl, M_UFSMNT);
493 }
494 
495 struct vnode *
496 procfs_hashget(pid, type, fd, mp)
497 	pid_t pid;
498 	pfstype type;
499 	int fd;
500 	struct mount *mp;
501 {
502 	struct pfs_hashhead *ppp;
503 	struct pfsnode *pp;
504 	struct vnode *vp;
505 
506 loop:
507 	simple_lock(&pfs_hash_slock);
508 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
509 	LIST_FOREACH(pp, ppp, pfs_hash) {
510 		vp = PFSTOV(pp);
511 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
512 		    pp->pfs_fd == fd && vp->v_mount == mp) {
513 			simple_lock(&vp->v_interlock);
514 			simple_unlock(&pfs_hash_slock);
515 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
516 				goto loop;
517 			return (vp);
518 		}
519 	}
520 	simple_unlock(&pfs_hash_slock);
521 	return (NULL);
522 }
523 
524 /*
525  * Insert the pfsnode into the hash table and lock it.
526  */
527 void
528 procfs_hashins(pp)
529 	struct pfsnode *pp;
530 {
531 	struct pfs_hashhead *ppp;
532 
533 	/* lock the pfsnode, then put it on the appropriate hash list */
534 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, (struct simplelock *)0);
535 
536 	simple_lock(&pfs_hash_slock);
537 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
538 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
539 	simple_unlock(&pfs_hash_slock);
540 }
541 
542 /*
543  * Remove the pfsnode from the hash table.
544  */
545 void
546 procfs_hashrem(pp)
547 	struct pfsnode *pp;
548 {
549 	simple_lock(&pfs_hash_slock);
550 	LIST_REMOVE(pp, pfs_hash);
551 	simple_unlock(&pfs_hash_slock);
552 }
553 
554 void
555 procfs_revoke_vnodes(p, arg)
556 	struct proc *p;
557 	void *arg;
558 {
559 	struct pfsnode *pfs, *pnext;
560 	struct vnode *vp;
561 	struct mount *mp = (struct mount *)arg;
562 	struct pfs_hashhead *ppp;
563 
564 	if (!(p->p_flag & P_SUGID))
565 		return;
566 
567 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
568 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
569 		vp = PFSTOV(pfs);
570 		pnext = LIST_NEXT(pfs, pfs_hash);
571 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
572 		    vp->v_mount == mp)
573 			VOP_REVOKE(vp, REVOKEALL);
574 	}
575 }
576 
577 int
578 procfs_getfp(pfs, pown, fp)
579 	struct pfsnode *pfs;
580 	struct proc **pown;
581 	struct file **fp;
582 {
583 	struct proc *p = PFIND(pfs->pfs_pid);
584 
585 	if (p == NULL)
586 		return ESRCH;
587 
588 	if (pfs->pfs_fd == -1)
589 		return EINVAL;
590 
591 	if ((*fp = fd_getfile(p->p_fd, pfs->pfs_fd)) == NULL)
592 		return EBADF;
593 
594 	*pown = p;
595 	return 0;
596 }
597