Use SYSREF to reference count struct vnode. v_usecount is now
[dragonfly.git] / sys / vfs / procfs / procfs_subr.c
CommitLineData
984263bc
MD
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 $
3c37c940 40 * $DragonFly: src/sys/vfs/procfs/procfs_subr.c,v 1.17 2007/05/06 19:23:35 dillon Exp $
984263bc
MD
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
1f2de5d4 51#include <vfs/procfs/procfs.h>
984263bc 52
3446c007
MD
53#define PFS_HSIZE 256
54#define PFS_HMASK (PFS_HSIZE - 1)
55
56static struct pfsnode *pfshead[PFS_HSIZE];
984263bc
MD
57static int pfsvplock;
58
3446c007
MD
59#define PFSHASH(pid) &pfshead[(pid) & PFS_HMASK]
60
984263bc 61/*
3446c007
MD
62 * Allocate a pfsnode/vnode pair. If no error occurs the returned vnode
63 * will be referenced and exclusively locked.
984263bc 64 *
3446c007
MD
65 * The pid, pfs_type, and mount point uniquely identify a pfsnode.
66 * The mount point is needed because someone might mount this filesystem
984263bc
MD
67 * twice.
68 *
3446c007
MD
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.
984263bc 72 *
3446c007
MD
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.
984263bc 78 *
3446c007
MD
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.
984263bc
MD
82 */
83int
ac424f9b 84procfs_allocvp(struct mount *mp, struct vnode **vpp, long pid, pfstype pfs_type)
984263bc 85{
984263bc
MD
86 struct pfsnode *pfs;
87 struct vnode *vp;
88 struct pfsnode **pp;
89 int error;
90
3446c007 91 pp = PFSHASH(pid);
984263bc 92loop:
3446c007
MD
93 for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
94 if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type &&
95 PFSTOV(pfs)->v_mount == mp) {
96 vp = PFSTOV(pfs);
87de5057 97 if (vget(vp, LK_EXCLUSIVE))
5fd012e0 98 goto loop;
3446c007
MD
99
100 /*
101 * Make sure the vnode is still in the cache after
102 * getting the interlock to avoid racing a free.
103 */
104 for (pfs = *pp; pfs; pfs = pfs->pfs_next) {
105 if (PFSTOV(pfs) == vp &&
106 pfs->pfs_pid == pid &&
107 pfs->pfs_type == pfs_type &&
108 PFSTOV(pfs)->v_mount == mp) {
109 break;
110 }
111 }
5fd012e0
MD
112 if (pfs == NULL || PFSTOV(pfs) != vp) {
113 vput(vp);
3446c007
MD
114 goto loop;
115
116 }
984263bc
MD
117 *vpp = vp;
118 return (0);
119 }
120 }
121
122 /*
123 * otherwise lock the vp list while we call getnewvnode
124 * since that can block.
125 */
126 if (pfsvplock & PROCFS_LOCKED) {
127 pfsvplock |= PROCFS_WANT;
377d4740 128 (void) tsleep((caddr_t) &pfsvplock, 0, "pfsavp", 0);
984263bc
MD
129 goto loop;
130 }
131 pfsvplock |= PROCFS_LOCKED;
132
133 /*
134 * Do the MALLOC before the getnewvnode since doing so afterward
135 * might cause a bogus v_data pointer to get dereferenced
136 * elsewhere if MALLOC should block.
5fd012e0
MD
137 *
138 * XXX this may not matter anymore since getnewvnode now returns
139 * a VX locked vnode.
984263bc
MD
140 */
141 MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
142
6ddb7618 143 error = getnewvnode(VT_PROCFS, mp, vpp, 0, 0);
3446c007 144 if (error) {
efda3bd0 145 kfree(pfs, M_TEMP);
984263bc
MD
146 goto out;
147 }
148 vp = *vpp;
149
150 vp->v_data = pfs;
151
152 pfs->pfs_next = 0;
153 pfs->pfs_pid = (pid_t) pid;
154 pfs->pfs_type = pfs_type;
155 pfs->pfs_vnode = vp;
156 pfs->pfs_flags = 0;
157 pfs->pfs_lockowner = 0;
158 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
159
160 switch (pfs_type) {
161 case Proot: /* /proc = dr-xr-xr-x */
162 pfs->pfs_mode = (VREAD|VEXEC) |
163 (VREAD|VEXEC) >> 3 |
164 (VREAD|VEXEC) >> 6;
165 vp->v_type = VDIR;
166 vp->v_flag = VROOT;
167 break;
168
169 case Pcurproc: /* /proc/curproc = lr--r--r-- */
170 pfs->pfs_mode = (VREAD) |
171 (VREAD >> 3) |
172 (VREAD >> 6);
173 vp->v_type = VLNK;
174 break;
175
176 case Pproc:
177 pfs->pfs_mode = (VREAD|VEXEC) |
178 (VREAD|VEXEC) >> 3 |
179 (VREAD|VEXEC) >> 6;
180 vp->v_type = VDIR;
181 break;
182
183 case Pfile:
184 pfs->pfs_mode = (VREAD|VEXEC) |
185 (VREAD|VEXEC) >> 3 |
186 (VREAD|VEXEC) >> 6;
187 vp->v_type = VLNK;
188 break;
189
190 case Pmem:
191 pfs->pfs_mode = (VREAD|VWRITE);
192 vp->v_type = VREG;
193 break;
194
195 case Pregs:
196 case Pfpregs:
197 case Pdbregs:
198 pfs->pfs_mode = (VREAD|VWRITE);
199 vp->v_type = VREG;
200 break;
201
202 case Pctl:
203 case Pnote:
204 case Pnotepg:
205 pfs->pfs_mode = (VWRITE);
206 vp->v_type = VREG;
207 break;
208
209 case Ptype:
210 case Pmap:
211 case Pstatus:
212 case Pcmdline:
213 case Prlimit:
214 pfs->pfs_mode = (VREAD) |
215 (VREAD >> 3) |
216 (VREAD >> 6);
217 vp->v_type = VREG;
218 break;
219
220 default:
221 panic("procfs_allocvp");
222 }
223
224 /* add to procfs vnode list */
3446c007 225 pfs->pfs_next = *pp;
984263bc
MD
226 *pp = pfs;
227
228out:
229 pfsvplock &= ~PROCFS_LOCKED;
230
231 if (pfsvplock & PROCFS_WANT) {
232 pfsvplock &= ~PROCFS_WANT;
233 wakeup((caddr_t) &pfsvplock);
234 }
235
236 return (error);
237}
238
239int
ac424f9b 240procfs_freevp(struct vnode *vp)
984263bc
MD
241{
242 struct pfsnode **pfspp;
3446c007 243 struct pfsnode *pfs;
984263bc 244
3446c007
MD
245 pfs = VTOPFS(vp);
246 vp->v_data = NULL;
984263bc 247
3446c007
MD
248 pfspp = PFSHASH(pfs->pfs_pid);
249 while (*pfspp != pfs && *pfspp)
250 pfspp = &(*pfspp)->pfs_next;
251 KKASSERT(*pfspp);
252 *pfspp = pfs->pfs_next;
253 pfs->pfs_next = NULL;
efda3bd0 254 kfree(pfs, M_TEMP);
984263bc
MD
255 return (0);
256}
257
258int
dadab5e9 259procfs_rw(struct vop_read_args *ap)
984263bc
MD
260{
261 struct vnode *vp = ap->a_vp;
262 struct uio *uio = ap->a_uio;
dadab5e9
MD
263 struct thread *curtd = uio->uio_td;
264 struct proc *curp;
984263bc
MD
265 struct pfsnode *pfs = VTOPFS(vp);
266 struct proc *p;
c7e98b2f 267 struct lwp *lp;
984263bc
MD
268 int rtval;
269
dadab5e9
MD
270 if (curtd == NULL)
271 return (EINVAL);
272 if ((curp = curtd->td_proc) == NULL) /* XXX */
273 return (EINVAL);
274
984263bc
MD
275 p = PFIND(pfs->pfs_pid);
276 if (p == NULL)
277 return (EINVAL);
278 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE)
279 return (EACCES);
c7e98b2f
SS
280 /* XXX lwp */
281 lp = FIRST_LWP_IN_PROC(p);
984263bc
MD
282
283 while (pfs->pfs_lockowner) {
377d4740 284 tsleep(&pfs->pfs_lockowner, 0, "pfslck", 0);
984263bc
MD
285 }
286 pfs->pfs_lockowner = curproc->p_pid;
287
288 switch (pfs->pfs_type) {
289 case Pnote:
290 case Pnotepg:
c7e98b2f 291 rtval = procfs_donote(curp, lp, pfs, uio);
984263bc
MD
292 break;
293
294 case Pregs:
c7e98b2f 295 rtval = procfs_doregs(curp, lp, pfs, uio);
984263bc
MD
296 break;
297
298 case Pfpregs:
c7e98b2f 299 rtval = procfs_dofpregs(curp, lp, pfs, uio);
984263bc
MD
300 break;
301
302 case Pdbregs:
c7e98b2f 303 rtval = procfs_dodbregs(curp, lp, pfs, uio);
984263bc
MD
304 break;
305
306 case Pctl:
c7e98b2f 307 rtval = procfs_doctl(curp, lp, pfs, uio);
984263bc
MD
308 break;
309
310 case Pstatus:
c7e98b2f 311 rtval = procfs_dostatus(curp, lp, pfs, uio);
984263bc
MD
312 break;
313
314 case Pmap:
c7e98b2f 315 rtval = procfs_domap(curp, lp, pfs, uio);
984263bc
MD
316 break;
317
318 case Pmem:
c7e98b2f 319 rtval = procfs_domem(curp, lp, pfs, uio);
984263bc
MD
320 break;
321
322 case Ptype:
c7e98b2f 323 rtval = procfs_dotype(curp, lp, pfs, uio);
984263bc
MD
324 break;
325
326 case Pcmdline:
c7e98b2f 327 rtval = procfs_docmdline(curp, lp, pfs, uio);
984263bc
MD
328 break;
329
330 case Prlimit:
c7e98b2f 331 rtval = procfs_dorlimit(curp, lp, pfs, uio);
984263bc
MD
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 */
355int
ac424f9b 356vfs_getuserstr(struct uio *uio, char *buf, int *buflenp)
984263bc
MD
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
387vfs_namemap_t *
ac424f9b 388vfs_findname(vfs_namemap_t *nm, char *buf, int buflen)
984263bc
MD
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
398void
dadab5e9 399procfs_exit(struct thread *td)
984263bc
MD
400{
401 struct pfsnode *pfs;
5fd012e0 402 struct vnode *vp;
dadab5e9
MD
403 pid_t pid;
404
405 KKASSERT(td->td_proc);
406 pid = td->td_proc->p_pid;
984263bc
MD
407
408 /*
409 * The reason for this loop is not obvious -- basicly,
410 * procfs_freevp(), which is called via vgone() (eventually),
411 * removes the specified procfs node from the pfshead list.
412 * It does this by *pfsp = pfs->pfs_next, meaning that it
413 * overwrites the node. So when we do pfs = pfs->next, we
414 * end up skipping the node that replaces the one that was
415 * vgone'd. Since it may have been the last one on the list,
416 * it may also have been set to null -- but *our* pfs pointer,
417 * here, doesn't see this. So the loop starts from the beginning
418 * again.
419 *
420 * This is not a for() loop because the final event
421 * would be "pfs = pfs->pfs_next"; in the case where
422 * pfs is set to pfshead again, that would mean that
423 * pfshead is skipped over.
424 *
425 */
3446c007
MD
426again:
427 pfs = *PFSHASH(pid);
984263bc
MD
428 while (pfs) {
429 if (pfs->pfs_pid == pid) {
5fd012e0 430 vp = PFSTOV(pfs);
e3332475 431 vx_lock(vp);
3c37c940 432 vgone_vxlocked(vp);
e3332475 433 vx_unlock(vp);
3446c007
MD
434 goto again;
435 }
436 pfs = pfs->pfs_next;
984263bc
MD
437 }
438}
3446c007 439