xref: /netbsd-src/sys/miscfs/procfs/procfs_subr.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: procfs_subr.c,v 1.62 2004/09/20 17:53:08 jdolecek 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 1.62 2004/09/20 17:53:08 jdolecek 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 __P((struct pfsnode *));
92 void procfs_hashrem __P((struct pfsnode *));
93 struct vnode *procfs_hashget __P((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 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
173 		vp->v_type = VLNK;
174 		break;
175 
176 	case PFSproc:	/* /proc/N = dr-xr-xr-x */
177 	case PFSfd:
178 		if (fd == -1) {	/* /proc/N/fd = dr-xr-xr-x */
179 			pfs->pfs_mode = S_IRUSR|S_IXUSR;
180 			vp->v_type = VDIR;
181 		} else {	/* /proc/N/fd/M = [ps-]rw------- */
182 			struct file *fp;
183 			struct vnode *vxp;
184 			struct proc *pown;
185 
186 			/* XXX can procfs_getfp() ever fail here? */
187 			if ((error = procfs_getfp(pfs, &pown, &fp)) != 0)
188 				goto bad;
189 			FILE_USE(fp);
190 
191 			pfs->pfs_mode = S_IRUSR|S_IWUSR;
192 			switch (fp->f_type) {
193 			case DTYPE_VNODE:
194 				vxp = (struct vnode *)fp->f_data;
195 
196 				/*
197 				 * We make symlinks for directories
198 				 * to avoid cycles.
199 				 */
200 				if (vxp->v_type == VDIR)
201 					goto symlink;
202 				vp->v_type = vxp->v_type;
203 				break;
204 			case DTYPE_PIPE:
205 				vp->v_type = VFIFO;
206 				break;
207 			case DTYPE_SOCKET:
208 				vp->v_type = VSOCK;
209 				break;
210 			case DTYPE_KQUEUE:
211 			case DTYPE_MISC:
212 			symlink:
213 				pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|
214 				    S_IXGRP|S_IROTH|S_IXOTH;
215 				vp->v_type = VLNK;
216 				break;
217 			default:
218 				error = EOPNOTSUPP;
219 				FILE_UNUSE(fp, pown);
220 				goto bad;
221 			}
222 			FILE_UNUSE(fp, pown);
223 		}
224 		break;
225 
226 	case PFSfile:	/* /proc/N/file = -rw------- */
227 	case PFSmem:	/* /proc/N/mem = -rw------- */
228 	case PFSregs:	/* /proc/N/regs = -rw------- */
229 	case PFSfpregs:	/* /proc/N/fpregs = -rw------- */
230 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
231 		vp->v_type = VREG;
232 		break;
233 
234 	case PFSctl:	/* /proc/N/ctl = --w------ */
235 	case PFSnote:	/* /proc/N/note = --w------ */
236 	case PFSnotepg:	/* /proc/N/notepg = --w------ */
237 		pfs->pfs_mode = S_IWUSR;
238 		vp->v_type = VREG;
239 		break;
240 
241 	case PFSmap:	/* /proc/N/map = -r--r--r-- */
242 	case PFSmaps:	/* /proc/N/maps = -r--r--r-- */
243 	case PFSstatus:	/* /proc/N/status = -r--r--r-- */
244 	case PFSstat:	/* /proc/N/stat = -r--r--r-- */
245 	case PFScmdline:	/* /proc/N/cmdline = -r--r--r-- */
246 	case PFSmeminfo:	/* /proc/meminfo = -r--r--r-- */
247 	case PFScpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
248 	case PFSuptime:	/* /proc/uptime = -r--r--r-- */
249 	case PFSmounts:	/* /proc/mounts = -r--r--r-- */
250 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
251 		vp->v_type = VREG;
252 		break;
253 
254 #ifdef __HAVE_PROCFS_MACHDEP
255 	PROCFS_MACHDEP_NODETYPE_CASES
256 		procfs_machdep_allocvp(vp);
257 		break;
258 #endif
259 
260 	default:
261 		panic("procfs_allocvp");
262 	}
263 
264 	procfs_hashins(pfs);
265 	uvm_vnp_setsize(vp, 0);
266 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
267 
268 	*vpp = vp;
269 	return (0);
270 
271  bad:
272 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
273 	FREE(pfs, M_TEMP);
274 	ungetnewvnode(vp);
275 	return (error);
276 }
277 
278 int
279 procfs_freevp(vp)
280 	struct vnode *vp;
281 {
282 	struct pfsnode *pfs = VTOPFS(vp);
283 
284 	procfs_hashrem(pfs);
285 
286 	FREE(vp->v_data, M_TEMP);
287 	vp->v_data = 0;
288 	return (0);
289 }
290 
291 int
292 procfs_rw(v)
293 	void *v;
294 {
295 	struct vop_read_args *ap = v;
296 	struct vnode *vp = ap->a_vp;
297 	struct uio *uio = ap->a_uio;
298 	struct proc *curp = uio->uio_procp;
299 	struct pfsnode *pfs = VTOPFS(vp);
300 	struct lwp *l;
301 	struct proc *p;
302 
303 	if (uio->uio_offset < 0)
304 		return EINVAL;
305 	p = PFIND(pfs->pfs_pid);
306 	if (p == 0)
307 		return ESRCH;
308 	/*
309 	 * Do not allow init to be modified while in secure mode; it
310 	 * could be duped into changing the security level.
311 	 */
312 	if (uio->uio_rw == UIO_WRITE && p == initproc && securelevel > -1)
313 		return EPERM;
314 
315 	/* XXX NJWLWP
316 	 * The entire procfs interface needs work to be useful to
317 	 * a process with multiple LWPs. For the moment, we'll
318 	 * just kluge this and fail on others.
319 	 */
320 	l = proc_representative_lwp(p);
321 
322 	switch (pfs->pfs_type) {
323 	case PFSnote:
324 	case PFSnotepg:
325 		return (procfs_donote(curp, p, pfs, uio));
326 
327 	case PFSregs:
328 		return (procfs_doregs(curp, l, pfs, uio));
329 
330 	case PFSfpregs:
331 		return (procfs_dofpregs(curp, l, pfs, uio));
332 
333 	case PFSctl:
334 		return (procfs_doctl(curp, l, pfs, uio));
335 
336 	case PFSstatus:
337 		return (procfs_dostatus(curp, l, pfs, uio));
338 
339 	case PFSstat:
340 		return (procfs_do_pid_stat(curp, l, pfs, uio));
341 
342 	case PFSmap:
343 		return (procfs_domap(curp, p, pfs, uio, 0));
344 
345 	case PFSmaps:
346 		return (procfs_domap(curp, p, pfs, uio, 1));
347 
348 	case PFSmem:
349 		return (procfs_domem(curp, p, pfs, uio));
350 
351 	case PFScmdline:
352 		return (procfs_docmdline(curp, p, pfs, uio));
353 
354 	case PFSmeminfo:
355 		return (procfs_domeminfo(curp, p, pfs, uio));
356 
357 	case PFScpuinfo:
358 		return (procfs_docpuinfo(curp, p, pfs, uio));
359 
360 	case PFSfd:
361 		return (procfs_dofd(curp, p, pfs, uio));
362 
363 	case PFSuptime:
364 		return (procfs_douptime(curp, p, pfs, uio));
365 
366 	case PFSmounts:
367 		return (procfs_domounts(curp, p, pfs, uio));
368 
369 #ifdef __HAVE_PROCFS_MACHDEP
370 	PROCFS_MACHDEP_NODETYPE_CASES
371 		return (procfs_machdep_rw(curp, l, pfs, uio));
372 #endif
373 
374 	default:
375 		return (EOPNOTSUPP);
376 	}
377 }
378 
379 /*
380  * Get a string from userland into (buf).  Strip a trailing
381  * nl character (to allow easy access from the shell).
382  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
383  * will automatically add a nul char at the end.
384  *
385  * Returns 0 on success or the following errors
386  *
387  * EINVAL:    file offset is non-zero.
388  * EMSGSIZE:  message is longer than kernel buffer
389  * EFAULT:    user i/o buffer is not addressable
390  */
391 int
392 vfs_getuserstr(uio, buf, buflenp)
393 	struct uio *uio;
394 	char *buf;
395 	int *buflenp;
396 {
397 	int xlen;
398 	int error;
399 
400 	if (uio->uio_offset != 0)
401 		return (EINVAL);
402 
403 	xlen = *buflenp;
404 
405 	/* must be able to read the whole string in one go */
406 	if (xlen < uio->uio_resid)
407 		return (EMSGSIZE);
408 	xlen = uio->uio_resid;
409 
410 	if ((error = uiomove(buf, xlen, uio)) != 0)
411 		return (error);
412 
413 	/* allow multiple writes without seeks */
414 	uio->uio_offset = 0;
415 
416 	/* cleanup string and remove trailing newline */
417 	buf[xlen] = '\0';
418 	xlen = strlen(buf);
419 	if (xlen > 0 && buf[xlen-1] == '\n')
420 		buf[--xlen] = '\0';
421 	*buflenp = xlen;
422 
423 	return (0);
424 }
425 
426 const vfs_namemap_t *
427 vfs_findname(nm, buf, buflen)
428 	const vfs_namemap_t *nm;
429 	const char *buf;
430 	int buflen;
431 {
432 
433 	for (; nm->nm_name; nm++)
434 		if (memcmp(buf, nm->nm_name, buflen+1) == 0)
435 			return (nm);
436 
437 	return (0);
438 }
439 
440 /*
441  * Initialize pfsnode hash table.
442  */
443 void
444 procfs_hashinit()
445 {
446 	lockinit(&pfs_hashlock, PINOD, "pfs_hashlock", 0, 0);
447 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT,
448 	    M_WAITOK, &pfs_ihash);
449 	simple_lock_init(&pfs_hash_slock);
450 }
451 
452 void
453 procfs_hashreinit()
454 {
455 	struct pfsnode *pp;
456 	struct pfs_hashhead *oldhash, *hash;
457 	u_long i, oldmask, mask, val;
458 
459 	hash = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT, M_WAITOK,
460 	    &mask);
461 
462 	simple_lock(&pfs_hash_slock);
463 	oldhash = pfs_hashtbl;
464 	oldmask = pfs_ihash;
465 	pfs_hashtbl = hash;
466 	pfs_ihash = mask;
467 	for (i = 0; i <= oldmask; i++) {
468 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
469 			LIST_REMOVE(pp, pfs_hash);
470 			val = PFSPIDHASH(pp->pfs_pid);
471 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
472 		}
473 	}
474 	simple_unlock(&pfs_hash_slock);
475 	hashdone(oldhash, M_UFSMNT);
476 }
477 
478 /*
479  * Free pfsnode hash table.
480  */
481 void
482 procfs_hashdone()
483 {
484 	hashdone(pfs_hashtbl, M_UFSMNT);
485 }
486 
487 struct vnode *
488 procfs_hashget(pid, type, fd, mp)
489 	pid_t pid;
490 	pfstype type;
491 	int fd;
492 	struct mount *mp;
493 {
494 	struct pfs_hashhead *ppp;
495 	struct pfsnode *pp;
496 	struct vnode *vp;
497 
498 loop:
499 	simple_lock(&pfs_hash_slock);
500 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
501 	LIST_FOREACH(pp, ppp, pfs_hash) {
502 		vp = PFSTOV(pp);
503 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
504 		    pp->pfs_fd == fd && vp->v_mount == mp) {
505 			simple_lock(&vp->v_interlock);
506 			simple_unlock(&pfs_hash_slock);
507 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
508 				goto loop;
509 			return (vp);
510 		}
511 	}
512 	simple_unlock(&pfs_hash_slock);
513 	return (NULL);
514 }
515 
516 /*
517  * Insert the pfsnode into the hash table and lock it.
518  */
519 void
520 procfs_hashins(pp)
521 	struct pfsnode *pp;
522 {
523 	struct pfs_hashhead *ppp;
524 
525 	/* lock the pfsnode, then put it on the appropriate hash list */
526 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, (struct simplelock *)0);
527 
528 	simple_lock(&pfs_hash_slock);
529 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
530 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
531 	simple_unlock(&pfs_hash_slock);
532 }
533 
534 /*
535  * Remove the pfsnode from the hash table.
536  */
537 void
538 procfs_hashrem(pp)
539 	struct pfsnode *pp;
540 {
541 	simple_lock(&pfs_hash_slock);
542 	LIST_REMOVE(pp, pfs_hash);
543 	simple_unlock(&pfs_hash_slock);
544 }
545 
546 void
547 procfs_revoke_vnodes(p, arg)
548 	struct proc *p;
549 	void *arg;
550 {
551 	struct pfsnode *pfs, *pnext;
552 	struct vnode *vp;
553 	struct mount *mp = (struct mount *)arg;
554 	struct pfs_hashhead *ppp;
555 
556 	if (!(p->p_flag & P_SUGID))
557 		return;
558 
559 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
560 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
561 		vp = PFSTOV(pfs);
562 		pnext = LIST_NEXT(pfs, pfs_hash);
563 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
564 		    vp->v_mount == mp)
565 			VOP_REVOKE(vp, REVOKEALL);
566 	}
567 }
568 
569 int
570 procfs_getfp(pfs, pown, fp)
571 	struct pfsnode *pfs;
572 	struct proc **pown;
573 	struct file **fp;
574 {
575 	struct proc *p = PFIND(pfs->pfs_pid);
576 
577 	if (p == NULL)
578 		return ESRCH;
579 
580 	if (pfs->pfs_fd == -1)
581 		return EINVAL;
582 
583 	if ((*fp = fd_getfile(p->p_fd, pfs->pfs_fd)) == NULL)
584 		return EBADF;
585 
586 	*pown = p;
587 	return 0;
588 }
589