xref: /netbsd-src/sys/miscfs/procfs/procfs_subr.c (revision 27578b9aac214cc7796ead81dcc5427e79d5f2a0)
1 /*	$NetBSD: procfs_subr.c,v 1.38 2001/09/15 16:12:59 chs Exp $	*/
2 
3 /*
4  * Copyright (c) 1994 Christopher G. Demetriou.  All rights reserved.
5  * Copyright (c) 1993 Jan-Simon Pendry
6  * Copyright (c) 1993
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * This code is derived from software contributed to Berkeley by
10  * Jan-Simon Pendry.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 #include <sys/proc.h>
48 #include <sys/vnode.h>
49 #include <sys/malloc.h>
50 #include <sys/stat.h>
51 
52 #include <miscfs/procfs/procfs.h>
53 
54 void procfs_hashins __P((struct pfsnode *));
55 void procfs_hashrem __P((struct pfsnode *));
56 struct vnode *procfs_hashget __P((pid_t, pfstype, struct mount *));
57 
58 LIST_HEAD(pfs_hashhead, pfsnode) *pfs_hashtbl;
59 u_long	pfs_ihash;	/* size of hash table - 1 */
60 #define PFSPIDHASH(pid)	((pid) & pfs_ihash)
61 
62 struct lock pfs_hashlock;
63 struct simplelock pfs_hash_slock;
64 
65 #define	ISSET(t, f)	((t) & (f))
66 
67 /*
68  * allocate a pfsnode/vnode pair.  the vnode is
69  * referenced, and locked.
70  *
71  * the pid, pfs_type, and mount point uniquely
72  * identify a pfsnode.  the mount point is needed
73  * because someone might mount this filesystem
74  * twice.
75  *
76  * all pfsnodes are maintained on a singly-linked
77  * list.  new nodes are only allocated when they cannot
78  * be found on this list.  entries on the list are
79  * removed when the vfs reclaim entry is called.
80  *
81  * a single lock is kept for the entire list.  this is
82  * needed because the getnewvnode() function can block
83  * waiting for a vnode to become free, in which case there
84  * may be more than one process trying to get the same
85  * vnode.  this lock is only taken if we are going to
86  * call getnewvnode, since the kernel itself is single-threaded.
87  *
88  * if an entry is found on the list, then call vget() to
89  * take a reference.  this is done because there may be
90  * zero references to it and so it needs to removed from
91  * the vnode free list.
92  */
93 int
94 procfs_allocvp(mp, vpp, pid, pfs_type)
95 	struct mount *mp;
96 	struct vnode **vpp;
97 	long pid;
98 	pfstype pfs_type;
99 {
100 	struct pfsnode *pfs;
101 	struct vnode *vp;
102 	int error;
103 
104 	do {
105 		if ((*vpp = procfs_hashget(pid, pfs_type, mp)) != NULL)
106 			return (0);
107 	} while (lockmgr(&pfs_hashlock, LK_EXCLUSIVE|LK_SLEEPFAIL, 0));
108 
109 	if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, vpp)) != 0) {
110 		*vpp = NULL;
111 		lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
112 		return (error);
113 	}
114 	vp = *vpp;
115 
116 	MALLOC(pfs, void *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
117 	vp->v_data = pfs;
118 
119 	pfs->pfs_pid = (pid_t) pid;
120 	pfs->pfs_type = pfs_type;
121 	pfs->pfs_vnode = vp;
122 	pfs->pfs_flags = 0;
123 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
124 
125 	switch (pfs_type) {
126 	case Proot:	/* /proc = dr-xr-xr-x */
127 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
128 		vp->v_type = VDIR;
129 		vp->v_flag = VROOT;
130 		break;
131 
132 	case Pcurproc:	/* /proc/curproc = lr-xr-xr-x */
133 	case Pself:	/* /proc/self    = lr-xr-xr-x */
134 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
135 		vp->v_type = VLNK;
136 		break;
137 
138 	case Pproc:	/* /proc/N = dr-xr-xr-x */
139 		pfs->pfs_mode = S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH;
140 		vp->v_type = VDIR;
141 		break;
142 
143 	case Pfile:	/* /proc/N/file = -rw------- */
144 	case Pmem:	/* /proc/N/mem = -rw------- */
145 	case Pregs:	/* /proc/N/regs = -rw------- */
146 	case Pfpregs:	/* /proc/N/fpregs = -rw------- */
147 		pfs->pfs_mode = S_IRUSR|S_IWUSR;
148 		vp->v_type = VREG;
149 		break;
150 
151 	case Pctl:	/* /proc/N/ctl = --w------ */
152 	case Pnote:	/* /proc/N/note = --w------ */
153 	case Pnotepg:	/* /proc/N/notepg = --w------ */
154 		pfs->pfs_mode = S_IWUSR;
155 		vp->v_type = VREG;
156 		break;
157 
158 	case Pmap:	/* /proc/N/map = -r--r--r-- */
159 	case Pmaps:	/* /proc/N/maps = -r--r--r-- */
160 	case Pstatus:	/* /proc/N/status = -r--r--r-- */
161 	case Pcmdline:	/* /proc/N/cmdline = -r--r--r-- */
162 	case Pmeminfo:	/* /proc/meminfo = -r--r--r-- */
163 	case Pcpuinfo:	/* /proc/cpuinfo = -r--r--r-- */
164 		pfs->pfs_mode = S_IRUSR|S_IRGRP|S_IROTH;
165 		vp->v_type = VREG;
166 		break;
167 
168 	default:
169 		panic("procfs_allocvp");
170 	}
171 
172 	procfs_hashins(pfs);
173 	uvm_vnp_setsize(vp, 0);
174 	lockmgr(&pfs_hashlock, LK_RELEASE, NULL);
175 
176 	return (error);
177 }
178 
179 int
180 procfs_freevp(vp)
181 	struct vnode *vp;
182 {
183 	struct pfsnode *pfs = VTOPFS(vp);
184 
185 	procfs_hashrem(pfs);
186 
187 	FREE(vp->v_data, M_TEMP);
188 	vp->v_data = 0;
189 	return (0);
190 }
191 
192 int
193 procfs_rw(v)
194 	void *v;
195 {
196 	struct vop_read_args *ap = v;
197 	struct vnode *vp = ap->a_vp;
198 	struct uio *uio = ap->a_uio;
199 	struct proc *curp = uio->uio_procp;
200 	struct pfsnode *pfs = VTOPFS(vp);
201 	struct proc *p;
202 
203 	p = PFIND(pfs->pfs_pid);
204 	if (p == 0)
205 		return (EINVAL);
206 
207 	switch (pfs->pfs_type) {
208 	case Pregs:
209 	case Pfpregs:
210 	case Pmem:
211 		/*
212 		 * Do not allow init to be modified while in secure mode; it
213 		 * could be duped into changing the security level.
214 		 */
215 		if (uio->uio_rw == UIO_WRITE &&
216 		    p == initproc && securelevel > -1)
217 			return (EPERM);
218 		break;
219 
220 	default:
221 		break;
222 	}
223 
224 	switch (pfs->pfs_type) {
225 	case Pnote:
226 	case Pnotepg:
227 		return (procfs_donote(curp, p, pfs, uio));
228 
229 	case Pregs:
230 		return (procfs_doregs(curp, p, pfs, uio));
231 
232 	case Pfpregs:
233 		return (procfs_dofpregs(curp, p, pfs, uio));
234 
235 	case Pctl:
236 		return (procfs_doctl(curp, p, pfs, uio));
237 
238 	case Pstatus:
239 		return (procfs_dostatus(curp, p, pfs, uio));
240 
241 	case Pmap:
242 		return (procfs_domap(curp, p, pfs, uio, 0));
243 
244 	case Pmaps:
245 		return (procfs_domap(curp, p, pfs, uio, 1));
246 
247 	case Pmem:
248 		return (procfs_domem(curp, p, pfs, uio));
249 
250 	case Pcmdline:
251 		return (procfs_docmdline(curp, p, pfs, uio));
252 
253 	case Pmeminfo:
254 		return (procfs_domeminfo(curp, p, pfs, uio));
255 	case Pcpuinfo:
256 		return (procfs_docpuinfo(curp, p, pfs, uio));
257 
258 	default:
259 		return (EOPNOTSUPP);
260 	}
261 }
262 
263 /*
264  * Get a string from userland into (buf).  Strip a trailing
265  * nl character (to allow easy access from the shell).
266  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
267  * will automatically add a nul char at the end.
268  *
269  * Returns 0 on success or the following errors
270  *
271  * EINVAL:    file offset is non-zero.
272  * EMSGSIZE:  message is longer than kernel buffer
273  * EFAULT:    user i/o buffer is not addressable
274  */
275 int
276 vfs_getuserstr(uio, buf, buflenp)
277 	struct uio *uio;
278 	char *buf;
279 	int *buflenp;
280 {
281 	int xlen;
282 	int error;
283 
284 	if (uio->uio_offset != 0)
285 		return (EINVAL);
286 
287 	xlen = *buflenp;
288 
289 	/* must be able to read the whole string in one go */
290 	if (xlen < uio->uio_resid)
291 		return (EMSGSIZE);
292 	xlen = uio->uio_resid;
293 
294 	if ((error = uiomove(buf, xlen, uio)) != 0)
295 		return (error);
296 
297 	/* allow multiple writes without seeks */
298 	uio->uio_offset = 0;
299 
300 	/* cleanup string and remove trailing newline */
301 	buf[xlen] = '\0';
302 	xlen = strlen(buf);
303 	if (xlen > 0 && buf[xlen-1] == '\n')
304 		buf[--xlen] = '\0';
305 	*buflenp = xlen;
306 
307 	return (0);
308 }
309 
310 const vfs_namemap_t *
311 vfs_findname(nm, buf, buflen)
312 	const vfs_namemap_t *nm;
313 	const char *buf;
314 	int buflen;
315 {
316 
317 	for (; nm->nm_name; nm++)
318 		if (memcmp(buf, nm->nm_name, buflen+1) == 0)
319 			return (nm);
320 
321 	return (0);
322 }
323 
324 /*
325  * Initialize pfsnode hash table.
326  */
327 void
328 procfs_hashinit()
329 {
330 	lockinit(&pfs_hashlock, PINOD, "pfs_hashlock", 0, 0);
331 	pfs_hashtbl = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT,
332 	    M_WAITOK, &pfs_ihash);
333 	simple_lock_init(&pfs_hash_slock);
334 }
335 
336 void
337 procfs_hashreinit()
338 {
339 	struct pfsnode *pp;
340 	struct pfs_hashhead *oldhash, *hash;
341 	u_long oldmask, mask, val;
342 	int i;
343 
344 	hash = hashinit(desiredvnodes / 4, HASH_LIST, M_UFSMNT, M_WAITOK,
345 	    &mask);
346 
347 	simple_lock(&pfs_hash_slock);
348 	oldhash = pfs_hashtbl;
349 	oldmask = pfs_ihash;
350 	pfs_hashtbl = hash;
351 	pfs_ihash = mask;
352 	for (i = 0; i <= oldmask; i++) {
353 		while ((pp = LIST_FIRST(&oldhash[i])) != NULL) {
354 			LIST_REMOVE(pp, pfs_hash);
355 			val = PFSPIDHASH(pp->pfs_pid);
356 			LIST_INSERT_HEAD(&hash[val], pp, pfs_hash);
357 		}
358 	}
359 	simple_unlock(&pfs_hash_slock);
360 	hashdone(oldhash, M_UFSMNT);
361 }
362 
363 /*
364  * Free pfsnode hash table.
365  */
366 void
367 procfs_hashdone()
368 {
369 	hashdone(pfs_hashtbl, M_UFSMNT);
370 }
371 
372 struct vnode *
373 procfs_hashget(pid, type, mp)
374 	pid_t pid;
375 	pfstype type;
376 	struct mount *mp;
377 {
378 	struct pfs_hashhead *ppp;
379 	struct pfsnode *pp;
380 	struct vnode *vp;
381 
382 loop:
383 	simple_lock(&pfs_hash_slock);
384 	ppp = &pfs_hashtbl[PFSPIDHASH(pid)];
385 	LIST_FOREACH(pp, ppp, pfs_hash) {
386 		vp = PFSTOV(pp);
387 		if (pid == pp->pfs_pid && pp->pfs_type == type &&
388 		    vp->v_mount == mp) {
389 			simple_lock(&vp->v_interlock);
390 			simple_unlock(&pfs_hash_slock);
391 			if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK))
392 				goto loop;
393 			return (vp);
394 		}
395 	}
396 	simple_unlock(&pfs_hash_slock);
397 	return (NULL);
398 }
399 
400 /*
401  * Insert the pfsnode into the hash table and lock it.
402  */
403 void
404 procfs_hashins(pp)
405 	struct pfsnode *pp;
406 {
407 	struct pfs_hashhead *ppp;
408 
409 	/* lock the pfsnode, then put it on the appropriate hash list */
410 	lockmgr(&pp->pfs_vnode->v_lock, LK_EXCLUSIVE, (struct simplelock *)0);
411 
412 	simple_lock(&pfs_hash_slock);
413 	ppp = &pfs_hashtbl[PFSPIDHASH(pp->pfs_pid)];
414 	LIST_INSERT_HEAD(ppp, pp, pfs_hash);
415 	simple_unlock(&pfs_hash_slock);
416 }
417 
418 /*
419  * Remove the pfsnode from the hash table.
420  */
421 void
422 procfs_hashrem(pp)
423 	struct pfsnode *pp;
424 {
425 	simple_lock(&pfs_hash_slock);
426 	LIST_REMOVE(pp, pfs_hash);
427 	simple_unlock(&pfs_hash_slock);
428 }
429 
430 void
431 procfs_revoke_vnodes(p, arg)
432 	struct proc *p;
433 	void *arg;
434 {
435 	struct pfsnode *pfs, *pnext;
436 	struct vnode *vp;
437 	struct mount *mp = (struct mount *)arg;
438 	struct pfs_hashhead *ppp;
439 
440 	if (!(p->p_flag & P_SUGID))
441 		return;
442 
443 	ppp = &pfs_hashtbl[PFSPIDHASH(p->p_pid)];
444 	for (pfs = LIST_FIRST(ppp); pfs; pfs = pnext) {
445 		vp = PFSTOV(pfs);
446 		pnext = LIST_NEXT(pfs, pfs_hash);
447 		if (vp->v_usecount > 0 && pfs->pfs_pid == p->p_pid &&
448 		    vp->v_mount == mp)
449 			VOP_REVOKE(vp, REVOKEALL);
450 	}
451 }
452