proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[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 $
dadab5e9 40 * $DragonFly: src/sys/vfs/procfs/procfs_subr.c,v 1.3 2003/06/25 03:56:00 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
51#include <miscfs/procfs/procfs.h>
52
53static struct pfsnode *pfshead;
54static int pfsvplock;
55
56/*
57 * allocate a pfsnode/vnode pair. the vnode is
58 * referenced, but not locked.
59 *
60 * the pid, pfs_type, and mount point uniquely
61 * identify a pfsnode. the mount point is needed
62 * because someone might mount this filesystem
63 * twice.
64 *
65 * all pfsnodes are maintained on a singly-linked
66 * list. new nodes are only allocated when they cannot
67 * be found on this list. entries on the list are
68 * removed when the vfs reclaim entry is called.
69 *
70 * a single lock is kept for the entire list. this is
71 * needed because the getnewvnode() function can block
72 * waiting for a vnode to become free, in which case there
73 * may be more than one process trying to get the same
74 * vnode. this lock is only taken if we are going to
75 * call getnewvnode, since the kernel itself is single-threaded.
76 *
77 * if an entry is found on the list, then call vget() to
78 * take a reference. this is done because there may be
79 * zero references to it and so it needs to removed from
80 * the vnode free list.
81 */
82int
83procfs_allocvp(mp, vpp, pid, pfs_type)
84 struct mount *mp;
85 struct vnode **vpp;
86 long pid;
87 pfstype pfs_type;
88{
dadab5e9 89 struct thread *td = curthread; /* XXX */
984263bc
MD
90 struct pfsnode *pfs;
91 struct vnode *vp;
92 struct pfsnode **pp;
93 int error;
94
95loop:
96 for (pfs = pfshead; pfs != 0; pfs = pfs->pfs_next) {
97 vp = PFSTOV(pfs);
98 if (pfs->pfs_pid == pid &&
99 pfs->pfs_type == pfs_type &&
100 vp->v_mount == mp) {
dadab5e9 101 if (vget(vp, 0, td))
984263bc
MD
102 goto loop;
103 *vpp = vp;
104 return (0);
105 }
106 }
107
108 /*
109 * otherwise lock the vp list while we call getnewvnode
110 * since that can block.
111 */
112 if (pfsvplock & PROCFS_LOCKED) {
113 pfsvplock |= PROCFS_WANT;
114 (void) tsleep((caddr_t) &pfsvplock, PINOD, "pfsavp", 0);
115 goto loop;
116 }
117 pfsvplock |= PROCFS_LOCKED;
118
119 /*
120 * Do the MALLOC before the getnewvnode since doing so afterward
121 * might cause a bogus v_data pointer to get dereferenced
122 * elsewhere if MALLOC should block.
123 */
124 MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK);
125
126 if ((error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, vpp)) != 0) {
127 FREE(pfs, M_TEMP);
128 goto out;
129 }
130 vp = *vpp;
131
132 vp->v_data = pfs;
133
134 pfs->pfs_next = 0;
135 pfs->pfs_pid = (pid_t) pid;
136 pfs->pfs_type = pfs_type;
137 pfs->pfs_vnode = vp;
138 pfs->pfs_flags = 0;
139 pfs->pfs_lockowner = 0;
140 pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type);
141
142 switch (pfs_type) {
143 case Proot: /* /proc = dr-xr-xr-x */
144 pfs->pfs_mode = (VREAD|VEXEC) |
145 (VREAD|VEXEC) >> 3 |
146 (VREAD|VEXEC) >> 6;
147 vp->v_type = VDIR;
148 vp->v_flag = VROOT;
149 break;
150
151 case Pcurproc: /* /proc/curproc = lr--r--r-- */
152 pfs->pfs_mode = (VREAD) |
153 (VREAD >> 3) |
154 (VREAD >> 6);
155 vp->v_type = VLNK;
156 break;
157
158 case Pproc:
159 pfs->pfs_mode = (VREAD|VEXEC) |
160 (VREAD|VEXEC) >> 3 |
161 (VREAD|VEXEC) >> 6;
162 vp->v_type = VDIR;
163 break;
164
165 case Pfile:
166 pfs->pfs_mode = (VREAD|VEXEC) |
167 (VREAD|VEXEC) >> 3 |
168 (VREAD|VEXEC) >> 6;
169 vp->v_type = VLNK;
170 break;
171
172 case Pmem:
173 pfs->pfs_mode = (VREAD|VWRITE);
174 vp->v_type = VREG;
175 break;
176
177 case Pregs:
178 case Pfpregs:
179 case Pdbregs:
180 pfs->pfs_mode = (VREAD|VWRITE);
181 vp->v_type = VREG;
182 break;
183
184 case Pctl:
185 case Pnote:
186 case Pnotepg:
187 pfs->pfs_mode = (VWRITE);
188 vp->v_type = VREG;
189 break;
190
191 case Ptype:
192 case Pmap:
193 case Pstatus:
194 case Pcmdline:
195 case Prlimit:
196 pfs->pfs_mode = (VREAD) |
197 (VREAD >> 3) |
198 (VREAD >> 6);
199 vp->v_type = VREG;
200 break;
201
202 default:
203 panic("procfs_allocvp");
204 }
205
206 /* add to procfs vnode list */
207 for (pp = &pfshead; *pp; pp = &(*pp)->pfs_next)
208 continue;
209 *pp = pfs;
210
211out:
212 pfsvplock &= ~PROCFS_LOCKED;
213
214 if (pfsvplock & PROCFS_WANT) {
215 pfsvplock &= ~PROCFS_WANT;
216 wakeup((caddr_t) &pfsvplock);
217 }
218
219 return (error);
220}
221
222int
223procfs_freevp(vp)
224 struct vnode *vp;
225{
226 struct pfsnode **pfspp;
227 struct pfsnode *pfs = VTOPFS(vp);
228
229 for (pfspp = &pfshead; *pfspp != 0; pfspp = &(*pfspp)->pfs_next) {
230 if (*pfspp == pfs) {
231 *pfspp = pfs->pfs_next;
232 break;
233 }
234 }
235
236 FREE(vp->v_data, M_TEMP);
237 vp->v_data = 0;
238 return (0);
239}
240
241int
dadab5e9 242procfs_rw(struct vop_read_args *ap)
984263bc
MD
243{
244 struct vnode *vp = ap->a_vp;
245 struct uio *uio = ap->a_uio;
dadab5e9
MD
246 struct thread *curtd = uio->uio_td;
247 struct proc *curp;
984263bc
MD
248 struct pfsnode *pfs = VTOPFS(vp);
249 struct proc *p;
250 int rtval;
251
dadab5e9
MD
252 if (curtd == NULL)
253 return (EINVAL);
254 if ((curp = curtd->td_proc) == NULL) /* XXX */
255 return (EINVAL);
256
984263bc
MD
257 p = PFIND(pfs->pfs_pid);
258 if (p == NULL)
259 return (EINVAL);
260 if (p->p_pid == 1 && securelevel > 0 && uio->uio_rw == UIO_WRITE)
261 return (EACCES);
262
263 while (pfs->pfs_lockowner) {
264 tsleep(&pfs->pfs_lockowner, PRIBIO, "pfslck", 0);
265 }
266 pfs->pfs_lockowner = curproc->p_pid;
267
268 switch (pfs->pfs_type) {
269 case Pnote:
270 case Pnotepg:
271 rtval = procfs_donote(curp, p, pfs, uio);
272 break;
273
274 case Pregs:
275 rtval = procfs_doregs(curp, p, pfs, uio);
276 break;
277
278 case Pfpregs:
279 rtval = procfs_dofpregs(curp, p, pfs, uio);
280 break;
281
282 case Pdbregs:
283 rtval = procfs_dodbregs(curp, p, pfs, uio);
284 break;
285
286 case Pctl:
287 rtval = procfs_doctl(curp, p, pfs, uio);
288 break;
289
290 case Pstatus:
291 rtval = procfs_dostatus(curp, p, pfs, uio);
292 break;
293
294 case Pmap:
295 rtval = procfs_domap(curp, p, pfs, uio);
296 break;
297
298 case Pmem:
299 rtval = procfs_domem(curp, p, pfs, uio);
300 break;
301
302 case Ptype:
303 rtval = procfs_dotype(curp, p, pfs, uio);
304 break;
305
306 case Pcmdline:
307 rtval = procfs_docmdline(curp, p, pfs, uio);
308 break;
309
310 case Prlimit:
311 rtval = procfs_dorlimit(curp, p, pfs, uio);
312 break;
313
314 default:
315 rtval = EOPNOTSUPP;
316 break;
317 }
318 pfs->pfs_lockowner = 0;
319 wakeup(&pfs->pfs_lockowner);
320 return rtval;
321}
322
323/*
324 * Get a string from userland into (buf). Strip a trailing
325 * nl character (to allow easy access from the shell).
326 * The buffer should be *buflenp + 1 chars long. vfs_getuserstr
327 * will automatically add a nul char at the end.
328 *
329 * Returns 0 on success or the following errors
330 *
331 * EINVAL: file offset is non-zero.
332 * EMSGSIZE: message is longer than kernel buffer
333 * EFAULT: user i/o buffer is not addressable
334 */
335int
336vfs_getuserstr(uio, buf, buflenp)
337 struct uio *uio;
338 char *buf;
339 int *buflenp;
340{
341 int xlen;
342 int error;
343
344 if (uio->uio_offset != 0)
345 return (EINVAL);
346
347 xlen = *buflenp;
348
349 /* must be able to read the whole string in one go */
350 if (xlen < uio->uio_resid)
351 return (EMSGSIZE);
352 xlen = uio->uio_resid;
353
354 if ((error = uiomove(buf, xlen, uio)) != 0)
355 return (error);
356
357 /* allow multiple writes without seeks */
358 uio->uio_offset = 0;
359
360 /* cleanup string and remove trailing newline */
361 buf[xlen] = '\0';
362 xlen = strlen(buf);
363 if (xlen > 0 && buf[xlen-1] == '\n')
364 buf[--xlen] = '\0';
365 *buflenp = xlen;
366
367 return (0);
368}
369
370vfs_namemap_t *
371vfs_findname(nm, buf, buflen)
372 vfs_namemap_t *nm;
373 char *buf;
374 int buflen;
375{
376
377 for (; nm->nm_name; nm++)
378 if (bcmp(buf, nm->nm_name, buflen+1) == 0)
379 return (nm);
380
381 return (0);
382}
383
384void
dadab5e9 385procfs_exit(struct thread *td)
984263bc
MD
386{
387 struct pfsnode *pfs;
dadab5e9
MD
388 pid_t pid;
389
390 KKASSERT(td->td_proc);
391 pid = td->td_proc->p_pid;
984263bc
MD
392
393 /*
394 * The reason for this loop is not obvious -- basicly,
395 * procfs_freevp(), which is called via vgone() (eventually),
396 * removes the specified procfs node from the pfshead list.
397 * It does this by *pfsp = pfs->pfs_next, meaning that it
398 * overwrites the node. So when we do pfs = pfs->next, we
399 * end up skipping the node that replaces the one that was
400 * vgone'd. Since it may have been the last one on the list,
401 * it may also have been set to null -- but *our* pfs pointer,
402 * here, doesn't see this. So the loop starts from the beginning
403 * again.
404 *
405 * This is not a for() loop because the final event
406 * would be "pfs = pfs->pfs_next"; in the case where
407 * pfs is set to pfshead again, that would mean that
408 * pfshead is skipped over.
409 *
410 */
411 pfs = pfshead;
412 while (pfs) {
413 if (pfs->pfs_pid == pid) {
414 vgone(PFSTOV(pfs));
415 pfs = pfshead;
416 } else
417 pfs = pfs->pfs_next;
418 }
419}