xref: /dflybsd-src/sys/vfs/procfs/procfs_subr.c (revision 48d201a5a8c1dab4aa7166b0812594c101fc43c3)
1 /*
2  * Copyright (c) 1993 Jan-Simon Pendry
3  * Copyright (c) 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Jan-Simon Pendry.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by the University of
20  *	California, Berkeley and its contributors.
21  * 4. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)procfs_subr.c	8.6 (Berkeley) 5/14/95
38  *
39  * $FreeBSD: src/sys/miscfs/procfs/procfs_subr.c,v 1.26.2.3 2002/02/18 21:28:04 des Exp $
40  * $DragonFly: src/sys/vfs/procfs/procfs_subr.c,v 1.10 2004/08/28 19:02:27 dillon Exp $
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/sysctl.h>
46 #include <sys/proc.h>
47 #include <sys/mount.h>
48 #include <sys/vnode.h>
49 #include <sys/malloc.h>
50 
51 #include <vfs/procfs/procfs.h>
52 
53 #define PFS_HSIZE	256
54 #define PFS_HMASK	(PFS_HSIZE - 1)
55 
56 static struct pfsnode *pfshead[PFS_HSIZE];
57 static int pfsvplock;
58 
59 #define PFSHASH(pid)	&pfshead[(pid) & PFS_HMASK]
60 
61 /*
62  * Allocate a pfsnode/vnode pair.  If no error occurs the returned vnode
63  * will be referenced and exclusively locked.
64  *
65  * The pid, pfs_type, and mount point uniquely identify a pfsnode.
66  * The mount point is needed because someone might mount this filesystem
67  * twice.
68  *
69  * All pfsnodes are maintained on a singly-linked list.  new nodes are
70  * only allocated when they cannot be found on this list.  entries on
71  * the list are removed when the vfs reclaim entry is called.
72  *
73  * A single lock is kept for the entire list.  this is needed because the
74  * getnewvnode() function can block waiting for a vnode to become free,
75  * in which case there may be more than one process trying to get the same
76  * vnode.  this lock is only taken if we are going to call getnewvnode,
77  * since the kernel itself is single-threaded.
78  *
79  * If an entry is found on the list, then call vget() to take a reference
80  * and obtain the lock.  This will properly re-reference the vnode if it
81  * had gotten onto the free list.
82  */
83 int
84 procfs_allocvp(struct mount *mp, struct vnode **vpp, long pid, pfstype pfs_type)
85 {
86 	struct thread *td = curthread;	/* XXX */
87 	struct pfsnode *pfs;
88 	struct vnode *vp;
89 	struct pfsnode **pp;
90 	int error;
91 	lwkt_tokref vlock;
92 
93 	pp = PFSHASH(pid);
94 loop:
95 	for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
96 		if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type &&
97 		    PFSTOV(pfs)->v_mount == mp) {
98 			vp = PFSTOV(pfs);
99 			lwkt_gettoken(&vlock, vp->v_interlock);
100 
101 			/*
102 			 * Make sure the vnode is still in the cache after
103 			 * getting the interlock to avoid racing a free.
104 			 */
105 			for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
106 				if (PFSTOV(pfs) == vp &&
107 				    pfs->pfs_pid == pid &&
108 				    pfs->pfs_type == pfs_type &&
109 				    PFSTOV(pfs)->v_mount == mp) {
110 					break;
111 				}
112 			}
113 			if (pfs == NULL) {
114 				lwkt_reltoken(&vlock);
115 				goto loop;
116 
117 			}
118 			if (vget(vp, &vlock, LK_EXCLUSIVE | LK_INTERLOCK, td))
119 				goto loop;
120 			*vpp = vp;
121 			return (0);
122 		}
123 	}
124 
125 	/*
126 	 * otherwise lock the vp list while we call getnewvnode
127 	 * since that can block.
128 	 */
129 	if (pfsvplock & PROCFS_LOCKED) {
130 		pfsvplock |= PROCFS_WANT;
131 		(void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0);
132 		goto loop;
133 	}
134 	pfsvplock |= PROCFS_LOCKED;
135 
136 	/*
137 	 * Do the MALLOC before the getnewvnode since doing so afterward
138 	 * might cause a bogus v_data pointer to get dereferenced
139 	 * elsewhere if MALLOC should block.
140 	 */
141 	MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
142 
143 	error = getnewvnode(VT_PROCFS, mp, mp->mnt_vn_ops, vpp, 0, 0);
144 	if (error) {
145 		free(pfs, M_TEMP);
146 		goto out;
147 	}
148 	vp = *vpp;
149 
150 	if (lockmgr(&vp->v_lock, LK_EXCLUSIVE, NULL, td))
151 		panic("procfs_allocvp: unexpected lock vailure");
152 
153 	vp->v_data = pfs;
154 
155 	pfs->pfs_next = 0;
156 	pfs->pfs_pid = (pid_t) pid;
157 	pfs->pfs_type = pfs_type;
158 	pfs->pfs_vnode = vp;
159 	pfs->pfs_flags = 0;
160 	pfs->pfs_lockowner = 0;
161 	pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
162 
163 	switch (pfs_type) {
164 	case Proot:	/* /proc = dr-xr-xr-x */
165 		pfs->pfs_mode = (VREAD|VEXEC) |
166 				(VREAD|VEXEC) >> 3 |
167 				(VREAD|VEXEC) >> 6;
168 		vp->v_type = VDIR;
169 		vp->v_flag = VROOT;
170 		break;
171 
172 	case Pcurproc:	/* /proc/curproc = lr--r--r-- */
173 		pfs->pfs_mode = (VREAD) |
174 				(VREAD >> 3) |
175 				(VREAD >> 6);
176 		vp->v_type = VLNK;
177 		break;
178 
179 	case Pproc:
180 		pfs->pfs_mode = (VREAD|VEXEC) |
181 				(VREAD|VEXEC) >> 3 |
182 				(VREAD|VEXEC) >> 6;
183 		vp->v_type = VDIR;
184 		break;
185 
186 	case Pfile:
187 		pfs->pfs_mode = (VREAD|VEXEC) |
188 				(VREAD|VEXEC) >> 3 |
189 				(VREAD|VEXEC) >> 6;
190 		vp->v_type = VLNK;
191 		break;
192 
193 	case Pmem:
194 		pfs->pfs_mode = (VREAD|VWRITE);
195 		vp->v_type = VREG;
196 		break;
197 
198 	case Pregs:
199 	case Pfpregs:
200 	case Pdbregs:
201 		pfs->pfs_mode = (VREAD|VWRITE);
202 		vp->v_type = VREG;
203 		break;
204 
205 	case Pctl:
206 	case Pnote:
207 	case Pnotepg:
208 		pfs->pfs_mode = (VWRITE);
209 		vp->v_type = VREG;
210 		break;
211 
212 	case Ptype:
213 	case Pmap:
214 	case Pstatus:
215 	case Pcmdline:
216 	case Prlimit:
217 		pfs->pfs_mode = (VREAD) |
218 				(VREAD >> 3) |
219 				(VREAD >> 6);
220 		vp->v_type = VREG;
221 		break;
222 
223 	default:
224 		panic("procfs_allocvp");
225 	}
226 
227 	/* add to procfs vnode list */
228 	pfs->pfs_next = *pp;
229 	*pp = pfs;
230 
231 out:
232 	pfsvplock &= ~PROCFS_LOCKED;
233 
234 	if (pfsvplock & PROCFS_WANT) {
235 		pfsvplock &= ~PROCFS_WANT;
236 		wakeup((caddr_t) &pfsvplock);
237 	}
238 
239 	return (error);
240 }
241 
242 int
243 procfs_freevp(struct vnode *vp)
244 {
245 	struct pfsnode **pfspp;
246 	struct pfsnode *pfs;
247 
248 	pfs = VTOPFS(vp);
249 	vp->v_data = NULL;
250 
251 	pfspp = PFSHASH(pfs->pfs_pid);
252 	while (*pfspp != pfs && *pfspp)
253 		pfspp = &(*pfspp)->pfs_next;
254 	KKASSERT(*pfspp);
255 	*pfspp = pfs->pfs_next;
256 	pfs->pfs_next = NULL;
257 	free(pfs, M_TEMP);
258 	return (0);
259 }
260 
261 int
262 procfs_rw(struct vop_read_args *ap)
263 {
264 	struct vnode *vp = ap->a_vp;
265 	struct uio *uio = ap->a_uio;
266 	struct thread *curtd = uio->uio_td;
267 	struct proc *curp;
268 	struct pfsnode *pfs = VTOPFS(vp);
269 	struct proc *p;
270 	int rtval;
271 
272 	if (curtd == NULL)
273 		return (EINVAL);
274 	if ((curp = curtd->td_proc) == NULL)	/* XXX */
275 		return (EINVAL);
276 
277 	p = PFIND(pfs->pfs_pid);
278 	if (p == NULL)
279 		return (EINVAL);
280 	if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE)
281 		return (EACCES);
282 
283 	while (pfs->pfs_lockowner) {
284 		tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0);
285 	}
286 	pfs->pfs_lockowner = curproc->p_pid;
287 
288 	switch (pfs->pfs_type) {
289 	case Pnote:
290 	case Pnotepg:
291 		rtval = procfs_donote(curp, p, pfs, uio);
292 		break;
293 
294 	case Pregs:
295 		rtval = procfs_doregs(curp, p, pfs, uio);
296 		break;
297 
298 	case Pfpregs:
299 		rtval = procfs_dofpregs(curp, p, pfs, uio);
300 		break;
301 
302         case Pdbregs:
303                 rtval = procfs_dodbregs(curp, p, pfs, uio);
304                 break;
305 
306 	case Pctl:
307 		rtval = procfs_doctl(curp, p, pfs, uio);
308 		break;
309 
310 	case Pstatus:
311 		rtval = procfs_dostatus(curp, p, pfs, uio);
312 		break;
313 
314 	case Pmap:
315 		rtval = procfs_domap(curp, p, pfs, uio);
316 		break;
317 
318 	case Pmem:
319 		rtval = procfs_domem(curp, p, pfs, uio);
320 		break;
321 
322 	case Ptype:
323 		rtval = procfs_dotype(curp, p, pfs, uio);
324 		break;
325 
326 	case Pcmdline:
327 		rtval = procfs_docmdline(curp, p, pfs, uio);
328 		break;
329 
330 	case Prlimit:
331 		rtval = procfs_dorlimit(curp, p, pfs, uio);
332 		break;
333 
334 	default:
335 		rtval = EOPNOTSUPP;
336 		break;
337 	}
338 	pfs->pfs_lockowner = 0;
339 	wakeup(&pfs->pfs_lockowner);
340 	return rtval;
341 }
342 
343 /*
344  * Get a string from userland into (buf).  Strip a trailing
345  * nl character (to allow easy access from the shell).
346  * The buffer should be *buflenp + 1 chars long.  vfs_getuserstr
347  * will automatically add a nul char at the end.
348  *
349  * Returns 0 on success or the following errors
350  *
351  * EINVAL:    file offset is non-zero.
352  * EMSGSIZE:  message is longer than kernel buffer
353  * EFAULT:    user i/o buffer is not addressable
354  */
355 int
356 vfs_getuserstr(struct uio *uio, char *buf, int *buflenp)
357 {
358 	int xlen;
359 	int error;
360 
361 	if (uio->uio_offset != 0)
362 		return (EINVAL);
363 
364 	xlen = *buflenp;
365 
366 	/* must be able to read the whole string in one go */
367 	if (xlen < uio->uio_resid)
368 		return (EMSGSIZE);
369 	xlen = uio->uio_resid;
370 
371 	if ((error = uiomove(buf, xlen, uio)) != 0)
372 		return (error);
373 
374 	/* allow multiple writes without seeks */
375 	uio->uio_offset = 0;
376 
377 	/* cleanup string and remove trailing newline */
378 	buf[xlen] = '\0';
379 	xlen = strlen(buf);
380 	if (xlen > 0 && buf[xlen-1] == '\n')
381 		buf[--xlen] = '\0';
382 	*buflenp = xlen;
383 
384 	return (0);
385 }
386 
387 vfs_namemap_t *
388 vfs_findname(vfs_namemap_t *nm, char *buf, int buflen)
389 {
390 
391 	for (; nm->nm_name; nm++)
392 		if (bcmp(buf, nm->nm_name, buflen+1) == 0)
393 			return (nm);
394 
395 	return (0);
396 }
397 
398 void
399 procfs_exit(struct thread *td)
400 {
401 	struct pfsnode *pfs;
402 	pid_t pid;
403 
404 	KKASSERT(td->td_proc);
405 	pid = td->td_proc->p_pid;
406 
407 	/*
408 	 * The reason for this loop is not obvious -- basicly,
409 	 * procfs_freevp(), which is called via vgone() (eventually),
410 	 * removes the specified procfs node from the pfshead list.
411 	 * It does this by *pfsp = pfs->pfs_next, meaning that it
412 	 * overwrites the node.  So when we do pfs = pfs->next, we
413 	 * end up skipping the node that replaces the one that was
414 	 * vgone'd.  Since it may have been the last one on the list,
415 	 * it may also have been set to null -- but *our* pfs pointer,
416 	 * here, doesn't see this.  So the loop starts from the beginning
417 	 * again.
418 	 *
419 	 * This is not a for() loop because the final event
420 	 * would be "pfs = pfs->pfs_next"; in the case where
421 	 * pfs is set to pfshead again, that would mean that
422 	 * pfshead is skipped over.
423 	 *
424 	 */
425 again:
426 	pfs = *PFSHASH(pid);
427 	while (pfs) {
428 		if (pfs->pfs_pid == pid) {
429 			vgone(PFSTOV(pfs));
430 			goto again;
431 		}
432 		pfs = pfs->pfs_next;
433 	}
434 }
435 
436