xref: /netbsd-src/sys/miscfs/procfs/procfs_subr.c (revision aaf4ece63a859a04e37cf3a7229b5fab0157cc06)
1 /*	$NetBSD: procfs_subr.c,v 1.67 2005/12/11 12:24:51 christos 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 1.67 2005/12/11 12:24:51 christos 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 PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
250 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
251 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
252 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
253 		vp->v_type = VREG;
254 		break;
255 
256 #ifdef __HAVE_PROCFS_MACHDEP
257 	PROCFS_MACHDEP_NODETYPE_CASES
258 		procfs_machdep_allocvp(vp);
259 		break;
260 #endif
261 
262 	default:
263 		panic("procfs_allocvp");
264 	}
265 
266 	procfs_hashins(pfs);
267 	uvm_vnp_setsize(vp, 0);
268 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
269 
270 	*vpp = vp;
271 	return (0);
272 
273  bad:
274 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
275 	FREE(pfs, M_TEMP);
276 	ungetnewvnode(vp);
277 	return (error);
278 }
279 
280 int
281 procfs_freevp(vp)
282 	struct vnode *vp;
283 {
284 	struct pfsnode *pfs = VTOPFS(vp);
285 
286 	procfs_hashrem(pfs);
287 
288 	FREE(vp->v_data, M_TEMP);
289 	vp->v_data = 0;
290 	return (0);
291 }
292 
293 int
294 procfs_rw(v)
295 	void *v;
296 {
297 	struct vop_read_args *ap = v;
298 	struct vnode *vp = ap->a_vp;
299 	struct uio *uio = ap->a_uio;
300 	struct lwp *curl;
301 	struct lwp *l;
302 	struct pfsnode *pfs = VTOPFS(vp);
303 	struct proc *p;
304 
305 	if (uio->uio_offset < 0)
306 		return EINVAL;
307 	p = PFIND(pfs->pfs_pid);
308 	if (p == 0)
309 		return ESRCH;
310 	/*
311 	 * Do not allow init to be modified while in secure mode; it
312 	 * could be duped into changing the security level.
313 	 */
314 	if (uio->uio_rw == UIO_WRITE && p == initproc && securelevel > -1)
315 		return EPERM;
316 
317 	curl = uio->uio_lwp;
318 
319 	/* XXX NJWLWP
320 	 * The entire procfs interface needs work to be useful to
321 	 * a process with multiple LWPs. For the moment, we'll
322 	 * just kluge this and fail on others.
323 	 */
324 	l = proc_representative_lwp(p);
325 
326 	switch (pfs->pfs_type) {
327 	case PFSnote:
328 	case PFSnotepg:
329 		return (procfs_donote(curl, p, pfs, uio));
330 
331 	case PFSregs:
332 		return (procfs_doregs(curl, l, pfs, uio));
333 
334 	case PFSfpregs:
335 		return (procfs_dofpregs(curl, l, pfs, uio));
336 
337 	case PFSctl:
338 		return (procfs_doctl(curl, l, pfs, uio));
339 
340 	case PFSstatus:
341 		return (procfs_dostatus(curl, l, pfs, uio));
342 
343 	case PFSstat:
344 		return (procfs_do_pid_stat(curl, l, pfs, uio));
345 
346 	case PFSmap:
347 		return (procfs_domap(curl, p, pfs, uio, 0));
348 
349 	case PFSmaps:
350 		return (procfs_domap(curl, p, pfs, uio, 1));
351 
352 	case PFSmem:
353 		return (procfs_domem(curl, l, pfs, uio));
354 
355 	case PFScmdline:
356 		return (procfs_docmdline(curl, p, pfs, uio));
357 
358 	case PFSmeminfo:
359 		return (procfs_domeminfo(curl, p, pfs, uio));
360 
361 	case PFScpuinfo:
362 		return (procfs_docpuinfo(curl, p, pfs, uio));
363 
364 	case PFSfd:
365 		return (procfs_dofd(curl, p, pfs, uio));
366 
367 	case PFSuptime:
368 		return (procfs_douptime(curl, p, pfs, uio));
369 
370 	case PFSmounts:
371 		return (procfs_domounts(curl, p, pfs, uio));
372 
373 #ifdef __HAVE_PROCFS_MACHDEP
374 	PROCFS_MACHDEP_NODETYPE_CASES
375 		return (procfs_machdep_rw(curl, l, pfs, uio));
376 #endif
377 
378 	default:
379 		return (EOPNOTSUPP);
380 	}
381 }
382 
383 /*
384  * Get a string from userland into (bf).  Strip a trailing
385  * nl character (to allow easy access from the shell).
386  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
387  * will automatically add a nul char at the end.
388  *
389  * Returns 0 on success or the following errors
390  *
391  * EINVAL:    file offset is non-zero.
392  * EMSGSIZE:  message is longer than kernel buffer
393  * EFAULT:    user i/o buffer is not addressable
394  */
395 int
396 vfs_getuserstr(uio, bf, buflenp)
397 	struct uio *uio;
398 	char *bf;
399 	int *buflenp;
400 {
401 	int xlen;
402 	int error;
403 
404 	if (uio->uio_offset != 0)
405 		return (EINVAL);
406 
407 	xlen = *buflenp;
408 
409 	/* must be able to read the whole string in one go */
410 	if (xlen < uio->uio_resid)
411 		return (EMSGSIZE);
412 	xlen = uio->uio_resid;
413 
414 	if ((error = uiomove(bf, xlen, uio)) != 0)
415 		return (error);
416 
417 	/* allow multiple writes without seeks */
418 	uio->uio_offset = 0;
419 
420 	/* cleanup string and remove trailing newline */
421 	bf[xlen] = '\0';
422 	xlen = strlen(bf);
423 	if (xlen > 0 && bf[xlen-1] == '\n')
424 		bf[--xlen] = '\0';
425 	*buflenp = xlen;
426 
427 	return (0);
428 }
429 
430 const vfs_namemap_t *
431 vfs_findname(nm, bf, buflen)
432 	const vfs_namemap_t *nm;
433 	const char *bf;
434 	int buflen;
435 {
436 
437 	for (; nm->nm_name; nm++)
438 		if (memcmp(bf, nm->nm_name, buflen+1) == 0)
439 			return (nm);
440 
441 	return (0);
442 }
443 
444 /*
445  * Initialize pfsnode hash table.
446  */
447 void
448 procfs_hashinit()
449 {
450 	lockinit(&pfs_hashlock, PINOD, "pfs_hashlock", 0, 0);
451 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT,
452 	    M_WAITOK, &pfs_ihash);
453 	simple_lock_init(&pfs_hash_slock);
454 }
455 
456 void
457 procfs_hashreinit()
458 {
459 	struct pfsnode *pp;
460 	struct pfs_hashhead *oldhash, *hash;
461 	u_long i, oldmask, mask, val;
462 
463 	hash = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT, M_WAITOK,
464 	    &mask);
465 
466 	simple_lock(&pfs_hash_slock);
467 	oldhash = pfs_hashtbl;
468 	oldmask = pfs_ihash;
469 	pfs_hashtbl = hash;
470 	pfs_ihash = mask;
471 	for (i = 0; i <= oldmask; i++) {
472 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
473 			LIST_REMOVE(pp, pfs_hash);
474 			val = PFSPIDHASH(pp->pfs_pid);
475 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
476 		}
477 	}
478 	simple_unlock(&pfs_hash_slock);
479 	hashdone(oldhash, M_UFSMNT);
480 }
481 
482 /*
483  * Free pfsnode hash table.
484  */
485 void
486 procfs_hashdone()
487 {
488 	hashdone(pfs_hashtbl, M_UFSMNT);
489 }
490 
491 struct vnode *
492 procfs_hashget(pid, type, fd, mp)
493 	pid_t pid;
494 	pfstype type;
495 	int fd;
496 	struct mount *mp;
497 {
498 	struct pfs_hashhead *ppp;
499 	struct pfsnode *pp;
500 	struct vnode *vp;
501 
502 loop:
503 	simple_lock(&pfs_hash_slock);
504 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
505 	LIST_FOREACH(pp, ppp, pfs_hash) {
506 		vp = PFSTOV(pp);
507 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
508 		    pp->pfs_fd == fd && vp->v_mount == mp) {
509 			simple_lock(&vp->v_interlock);
510 			simple_unlock(&pfs_hash_slock);
511 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
512 				goto loop;
513 			return (vp);
514 		}
515 	}
516 	simple_unlock(&pfs_hash_slock);
517 	return (NULL);
518 }
519 
520 /*
521  * Insert the pfsnode into the hash table and lock it.
522  */
523 void
524 procfs_hashins(pp)
525 	struct pfsnode *pp;
526 {
527 	struct pfs_hashhead *ppp;
528 
529 	/* lock the pfsnode, then put it on the appropriate hash list */
530 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, (struct simplelock *)0);
531 
532 	simple_lock(&pfs_hash_slock);
533 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
534 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
535 	simple_unlock(&pfs_hash_slock);
536 }
537 
538 /*
539  * Remove the pfsnode from the hash table.
540  */
541 void
542 procfs_hashrem(pp)
543 	struct pfsnode *pp;
544 {
545 	simple_lock(&pfs_hash_slock);
546 	LIST_REMOVE(pp, pfs_hash);
547 	simple_unlock(&pfs_hash_slock);
548 }
549 
550 void
551 procfs_revoke_vnodes(p, arg)
552 	struct proc *p;
553 	void *arg;
554 {
555 	struct pfsnode *pfs, *pnext;
556 	struct vnode *vp;
557 	struct mount *mp = (struct mount *)arg;
558 	struct pfs_hashhead *ppp;
559 
560 	if (!(p->p_flag & P_SUGID))
561 		return;
562 
563 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
564 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
565 		vp = PFSTOV(pfs);
566 		pnext = LIST_NEXT(pfs, pfs_hash);
567 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
568 		    vp->v_mount == mp)
569 			VOP_REVOKE(vp, REVOKEALL);
570 	}
571 }
572 
573 int
574 procfs_getfp(pfs, pown, fp)
575 	struct pfsnode *pfs;
576 	struct proc **pown;
577 	struct file **fp;
578 {
579 	struct proc *p = PFIND(pfs->pfs_pid);
580 
581 	if (p == NULL)
582 		return ESRCH;
583 
584 	if (pfs->pfs_fd == -1)
585 		return EINVAL;
586 
587 	if ((*fp = fd_getfile(p->p_fd, pfs->pfs_fd)) == NULL)
588 		return EBADF;
589 
590 	*pown = p;
591 	return 0;
592 }
593