procfs_vnops.c revision 1.24 1 1.24 cgd /* $NetBSD: procfs_vnops.c,v 1.24 1994/06/29 06:35:00 cgd Exp $ */
2 1.24 cgd
3 1.1 pk /*
4 1.9 cgd * Copyright (c) 1993 Jan-Simon Pendry
5 1.21 mycroft * Copyright (c) 1993
6 1.21 mycroft * The Regents of the University of California. All rights reserved.
7 1.2 pk *
8 1.9 cgd * This code is derived from software contributed to Berkeley by
9 1.9 cgd * Jan-Simon Pendry.
10 1.9 cgd *
11 1.2 pk * Redistribution and use in source and binary forms, with or without
12 1.2 pk * modification, are permitted provided that the following conditions
13 1.2 pk * are met:
14 1.2 pk * 1. Redistributions of source code must retain the above copyright
15 1.2 pk * notice, this list of conditions and the following disclaimer.
16 1.2 pk * 2. Redistributions in binary form must reproduce the above copyright
17 1.2 pk * notice, this list of conditions and the following disclaimer in the
18 1.2 pk * documentation and/or other materials provided with the distribution.
19 1.2 pk * 3. All advertising materials mentioning features or use of this software
20 1.2 pk * must display the following acknowledgement:
21 1.9 cgd * This product includes software developed by the University of
22 1.9 cgd * California, Berkeley and its contributors.
23 1.9 cgd * 4. Neither the name of the University nor the names of its contributors
24 1.9 cgd * may be used to endorse or promote products derived from this software
25 1.9 cgd * without specific prior written permission.
26 1.9 cgd *
27 1.9 cgd * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 1.9 cgd * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 1.9 cgd * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 1.9 cgd * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 1.9 cgd * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 1.9 cgd * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 1.9 cgd * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 1.9 cgd * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 1.9 cgd * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 1.9 cgd * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 1.9 cgd * SUCH DAMAGE.
38 1.2 pk *
39 1.24 cgd * @(#)procfs_vnops.c 8.8 (Berkeley) 6/15/94
40 1.2 pk */
41 1.2 pk
42 1.2 pk /*
43 1.9 cgd * procfs vnode interface
44 1.1 pk */
45 1.1 pk
46 1.8 mycroft #include <sys/param.h>
47 1.8 mycroft #include <sys/systm.h>
48 1.8 mycroft #include <sys/time.h>
49 1.8 mycroft #include <sys/kernel.h>
50 1.8 mycroft #include <sys/file.h>
51 1.8 mycroft #include <sys/proc.h>
52 1.8 mycroft #include <sys/vnode.h>
53 1.8 mycroft #include <sys/namei.h>
54 1.9 cgd #include <sys/malloc.h>
55 1.21 mycroft #include <sys/dirent.h>
56 1.8 mycroft #include <sys/resourcevar.h>
57 1.14 cgd #include <sys/ptrace.h>
58 1.21 mycroft #include <vm/vm.h> /* for PAGE_SIZE */
59 1.21 mycroft #include <machine/reg.h>
60 1.9 cgd #include <miscfs/procfs/procfs.h>
61 1.11 cgd
62 1.9 cgd /*
63 1.9 cgd * Vnode Operations.
64 1.9 cgd *
65 1.9 cgd */
66 1.1 pk
67 1.1 pk /*
68 1.9 cgd * This is a list of the valid names in the
69 1.9 cgd * process-specific sub-directories. It is
70 1.9 cgd * used in procfs_lookup and procfs_readdir
71 1.9 cgd */
72 1.9 cgd static struct pfsnames {
73 1.21 mycroft u_char d_type;
74 1.21 mycroft u_char d_namlen;
75 1.9 cgd char d_name[PROCFS_NAMELEN];
76 1.9 cgd pfstype d_pfstype;
77 1.21 mycroft int (*d_valid) __P((struct proc *p));
78 1.9 cgd } procent[] = {
79 1.9 cgd #define N(s) sizeof(s)-1, s
80 1.9 cgd /* namlen, nam, type */
81 1.21 mycroft { DT_DIR, N("."), Pproc, NULL },
82 1.21 mycroft { DT_DIR, N(".."), Proot, NULL },
83 1.21 mycroft { DT_REG, N("file"), Pfile, procfs_validfile },
84 1.21 mycroft { DT_REG, N("mem"), Pmem, NULL },
85 1.21 mycroft { DT_REG, N("regs"), Pregs, procfs_validregs },
86 1.21 mycroft { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs },
87 1.21 mycroft { DT_REG, N("ctl"), Pctl, NULL },
88 1.21 mycroft { DT_REG, N("status"), Pstatus, NULL },
89 1.21 mycroft { DT_REG, N("note"), Pnote, NULL },
90 1.21 mycroft { DT_REG, N("notepg"), Pnotepg, NULL },
91 1.9 cgd #undef N
92 1.1 pk };
93 1.9 cgd #define Nprocent (sizeof(procent)/sizeof(procent[0]))
94 1.9 cgd
95 1.9 cgd static pid_t atopid __P((const char *, u_int));
96 1.1 pk
97 1.1 pk /*
98 1.9 cgd * set things up for doing i/o on
99 1.9 cgd * the pfsnode (vp). (vp) is locked
100 1.9 cgd * on entry, and should be left locked
101 1.9 cgd * on exit.
102 1.1 pk *
103 1.9 cgd * for procfs we don't need to do anything
104 1.9 cgd * in particular for i/o. all that is done
105 1.9 cgd * is to support exclusive open on process
106 1.9 cgd * memory images.
107 1.1 pk */
108 1.21 mycroft procfs_open(ap)
109 1.22 mycroft struct vop_open_args /* {
110 1.22 mycroft struct vnode *a_vp;
111 1.22 mycroft int a_mode;
112 1.22 mycroft struct ucred *a_cred;
113 1.22 mycroft struct proc *a_p;
114 1.22 mycroft } */ *ap;
115 1.1 pk {
116 1.21 mycroft struct pfsnode *pfs = VTOPFS(ap->a_vp);
117 1.9 cgd
118 1.9 cgd switch (pfs->pfs_type) {
119 1.9 cgd case Pmem:
120 1.9 cgd if (PFIND(pfs->pfs_pid) == 0)
121 1.9 cgd return (ENOENT); /* was ESRCH, jsp */
122 1.9 cgd
123 1.21 mycroft if ((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL) ||
124 1.21 mycroft (pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))
125 1.9 cgd return (EBUSY);
126 1.1 pk
127 1.21 mycroft if (ap->a_mode & FWRITE)
128 1.21 mycroft pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL);
129 1.1 pk
130 1.9 cgd return (0);
131 1.1 pk
132 1.9 cgd default:
133 1.9 cgd break;
134 1.9 cgd }
135 1.1 pk
136 1.9 cgd return (0);
137 1.1 pk }
138 1.1 pk
139 1.1 pk /*
140 1.9 cgd * close the pfsnode (vp) after doing i/o.
141 1.9 cgd * (vp) is not locked on entry or exit.
142 1.9 cgd *
143 1.9 cgd * nothing to do for procfs other than undo
144 1.9 cgd * any exclusive open flag (see _open above).
145 1.1 pk */
146 1.21 mycroft procfs_close(ap)
147 1.22 mycroft struct vop_close_args /* {
148 1.22 mycroft struct vnode *a_vp;
149 1.22 mycroft int a_fflag;
150 1.22 mycroft struct ucred *a_cred;
151 1.22 mycroft struct proc *a_p;
152 1.22 mycroft } */ *ap;
153 1.1 pk {
154 1.21 mycroft struct pfsnode *pfs = VTOPFS(ap->a_vp);
155 1.1 pk
156 1.9 cgd switch (pfs->pfs_type) {
157 1.9 cgd case Pmem:
158 1.21 mycroft if ((ap->a_fflag & FWRITE) && (pfs->pfs_flags & O_EXCL))
159 1.9 cgd pfs->pfs_flags &= ~(FWRITE|O_EXCL);
160 1.9 cgd break;
161 1.9 cgd }
162 1.1 pk
163 1.1 pk return (0);
164 1.1 pk }
165 1.1 pk
166 1.1 pk /*
167 1.9 cgd * do an ioctl operation on pfsnode (vp).
168 1.9 cgd * (vp) is not locked on entry or exit.
169 1.1 pk */
170 1.21 mycroft procfs_ioctl(ap)
171 1.22 mycroft struct vop_ioctl_args /* {
172 1.22 mycroft struct vnode *a_vp;
173 1.22 mycroft int a_command;
174 1.22 mycroft caddr_t a_data;
175 1.22 mycroft int a_fflag;
176 1.22 mycroft struct ucred *a_cred;
177 1.22 mycroft struct proc *a_p;
178 1.22 mycroft } */ *ap;
179 1.1 pk {
180 1.1 pk
181 1.9 cgd return (ENOTTY);
182 1.1 pk }
183 1.1 pk
184 1.1 pk /*
185 1.9 cgd * do block mapping for pfsnode (vp).
186 1.9 cgd * since we don't use the buffer cache
187 1.9 cgd * for procfs this function should never
188 1.9 cgd * be called. in any case, it's not clear
189 1.9 cgd * what part of the kernel ever makes use
190 1.9 cgd * of this function. for sanity, this is the
191 1.9 cgd * usual no-op bmap, although returning
192 1.9 cgd * (EIO) would be a reasonable alternative.
193 1.1 pk */
194 1.21 mycroft procfs_bmap(ap)
195 1.22 mycroft struct vop_bmap_args /* {
196 1.22 mycroft struct vnode *a_vp;
197 1.22 mycroft daddr_t a_bn;
198 1.22 mycroft struct vnode **a_vpp;
199 1.22 mycroft daddr_t *a_bnp;
200 1.22 mycroft } */ *ap;
201 1.1 pk {
202 1.1 pk
203 1.21 mycroft if (ap->a_vpp != NULL)
204 1.21 mycroft *ap->a_vpp = ap->a_vp;
205 1.21 mycroft if (ap->a_bnp != NULL)
206 1.21 mycroft *ap->a_bnp = ap->a_bn;
207 1.1 pk return (0);
208 1.1 pk }
209 1.1 pk
210 1.1 pk /*
211 1.9 cgd * _inactive is called when the pfsnode
212 1.9 cgd * is vrele'd and the reference count goes
213 1.9 cgd * to zero. (vp) will be on the vnode free
214 1.9 cgd * list, so to get it back vget() must be
215 1.9 cgd * used.
216 1.9 cgd *
217 1.9 cgd * for procfs, check if the process is still
218 1.9 cgd * alive and if it isn't then just throw away
219 1.9 cgd * the vnode by calling vgone(). this may
220 1.9 cgd * be overkill and a waste of time since the
221 1.9 cgd * chances are that the process will still be
222 1.9 cgd * there and PFIND is not free.
223 1.9 cgd *
224 1.9 cgd * (vp) is not locked on entry or exit.
225 1.1 pk */
226 1.21 mycroft procfs_inactive(ap)
227 1.22 mycroft struct vop_inactive_args /* {
228 1.22 mycroft struct vnode *a_vp;
229 1.22 mycroft } */ *ap;
230 1.1 pk {
231 1.21 mycroft struct pfsnode *pfs = VTOPFS(ap->a_vp);
232 1.1 pk
233 1.9 cgd if (PFIND(pfs->pfs_pid) == 0)
234 1.21 mycroft vgone(ap->a_vp);
235 1.5 pk
236 1.9 cgd return (0);
237 1.1 pk }
238 1.1 pk
239 1.1 pk /*
240 1.9 cgd * _reclaim is called when getnewvnode()
241 1.9 cgd * wants to make use of an entry on the vnode
242 1.9 cgd * free list. at this time the filesystem needs
243 1.9 cgd * to free any private data and remove the node
244 1.9 cgd * from any private lists.
245 1.1 pk */
246 1.21 mycroft procfs_reclaim(ap)
247 1.22 mycroft struct vop_reclaim_args /* {
248 1.22 mycroft struct vnode *a_vp;
249 1.22 mycroft } */ *ap;
250 1.21 mycroft {
251 1.21 mycroft
252 1.21 mycroft return (procfs_freevp(ap->a_vp));
253 1.21 mycroft }
254 1.21 mycroft
255 1.21 mycroft /*
256 1.21 mycroft * Return POSIX pathconf information applicable to special devices.
257 1.21 mycroft */
258 1.21 mycroft procfs_pathconf(ap)
259 1.21 mycroft struct vop_pathconf_args /* {
260 1.21 mycroft struct vnode *a_vp;
261 1.21 mycroft int a_name;
262 1.21 mycroft int *a_retval;
263 1.21 mycroft } */ *ap;
264 1.1 pk {
265 1.1 pk
266 1.21 mycroft switch (ap->a_name) {
267 1.21 mycroft case _PC_LINK_MAX:
268 1.21 mycroft *ap->a_retval = LINK_MAX;
269 1.21 mycroft return (0);
270 1.21 mycroft case _PC_MAX_CANON:
271 1.21 mycroft *ap->a_retval = MAX_CANON;
272 1.21 mycroft return (0);
273 1.21 mycroft case _PC_MAX_INPUT:
274 1.21 mycroft *ap->a_retval = MAX_INPUT;
275 1.21 mycroft return (0);
276 1.21 mycroft case _PC_PIPE_BUF:
277 1.21 mycroft *ap->a_retval = PIPE_BUF;
278 1.21 mycroft return (0);
279 1.21 mycroft case _PC_CHOWN_RESTRICTED:
280 1.21 mycroft *ap->a_retval = 1;
281 1.21 mycroft return (0);
282 1.21 mycroft case _PC_VDISABLE:
283 1.21 mycroft *ap->a_retval = _POSIX_VDISABLE;
284 1.21 mycroft return (0);
285 1.21 mycroft default:
286 1.21 mycroft return (EINVAL);
287 1.21 mycroft }
288 1.21 mycroft /* NOTREACHED */
289 1.1 pk }
290 1.1 pk
291 1.1 pk /*
292 1.9 cgd * _print is used for debugging.
293 1.9 cgd * just print a readable description
294 1.9 cgd * of (vp).
295 1.1 pk */
296 1.21 mycroft procfs_print(ap)
297 1.22 mycroft struct vop_print_args /* {
298 1.22 mycroft struct vnode *a_vp;
299 1.22 mycroft } */ *ap;
300 1.1 pk {
301 1.21 mycroft struct pfsnode *pfs = VTOPFS(ap->a_vp);
302 1.5 pk
303 1.21 mycroft printf("tag VT_PROCFS, type %s, pid %d, mode %x, flags %x\n",
304 1.21 mycroft pfs->pfs_type, pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags);
305 1.1 pk }
306 1.1 pk
307 1.1 pk /*
308 1.9 cgd * _abortop is called when operations such as
309 1.9 cgd * rename and create fail. this entry is responsible
310 1.9 cgd * for undoing any side-effects caused by the lookup.
311 1.9 cgd * this will always include freeing the pathname buffer.
312 1.1 pk */
313 1.21 mycroft procfs_abortop(ap)
314 1.22 mycroft struct vop_abortop_args /* {
315 1.22 mycroft struct vnode *a_dvp;
316 1.22 mycroft struct componentname *a_cnp;
317 1.22 mycroft } */ *ap;
318 1.1 pk {
319 1.9 cgd
320 1.21 mycroft if ((ap->a_cnp->cn_flags & (HASBUF | SAVESTART)) == HASBUF)
321 1.21 mycroft FREE(ap->a_cnp->cn_pnbuf, M_NAMEI);
322 1.9 cgd return (0);
323 1.1 pk }
324 1.1 pk
325 1.1 pk /*
326 1.9 cgd * generic entry point for unsupported operations
327 1.1 pk */
328 1.9 cgd procfs_badop()
329 1.1 pk {
330 1.1 pk
331 1.9 cgd return (EIO);
332 1.1 pk }
333 1.1 pk
334 1.1 pk /*
335 1.9 cgd * Invent attributes for pfsnode (vp) and store
336 1.9 cgd * them in (vap).
337 1.9 cgd * Directories lengths are returned as zero since
338 1.9 cgd * any real length would require the genuine size
339 1.9 cgd * to be computed, and nothing cares anyway.
340 1.9 cgd *
341 1.9 cgd * this is relatively minimal for procfs.
342 1.1 pk */
343 1.21 mycroft procfs_getattr(ap)
344 1.22 mycroft struct vop_getattr_args /* {
345 1.22 mycroft struct vnode *a_vp;
346 1.22 mycroft struct vattr *a_vap;
347 1.22 mycroft struct ucred *a_cred;
348 1.22 mycroft struct proc *a_p;
349 1.22 mycroft } */ *ap;
350 1.1 pk {
351 1.21 mycroft struct pfsnode *pfs = VTOPFS(ap->a_vp);
352 1.21 mycroft struct vattr *vap = ap->a_vap;
353 1.1 pk struct proc *procp;
354 1.9 cgd int error;
355 1.1 pk
356 1.21 mycroft /* first check the process still exists */
357 1.21 mycroft switch (pfs->pfs_type) {
358 1.21 mycroft case Proot:
359 1.21 mycroft case Pcurproc:
360 1.21 mycroft procp = 0;
361 1.21 mycroft break;
362 1.21 mycroft
363 1.21 mycroft default:
364 1.21 mycroft procp = PFIND(pfs->pfs_pid);
365 1.21 mycroft if (procp == 0)
366 1.21 mycroft return (ENOENT);
367 1.21 mycroft }
368 1.21 mycroft
369 1.21 mycroft error = 0;
370 1.21 mycroft
371 1.9 cgd /* start by zeroing out the attributes */
372 1.1 pk VATTR_NULL(vap);
373 1.9 cgd
374 1.9 cgd /* next do all the common fields */
375 1.21 mycroft vap->va_type = ap->a_vp->v_type;
376 1.9 cgd vap->va_mode = pfs->pfs_mode;
377 1.9 cgd vap->va_fileid = pfs->pfs_fileno;
378 1.9 cgd vap->va_flags = 0;
379 1.19 cgd vap->va_blocksize = PAGE_SIZE;
380 1.9 cgd vap->va_bytes = vap->va_size = 0;
381 1.9 cgd
382 1.9 cgd /*
383 1.9 cgd * Make all times be current TOD.
384 1.9 cgd * It would be possible to get the process start
385 1.9 cgd * time from the p_stat structure, but there's
386 1.9 cgd * no "file creation" time stamp anyway, and the
387 1.9 cgd * p_stat structure is not addressible if u. gets
388 1.9 cgd * swapped out for that process.
389 1.21 mycroft *
390 1.21 mycroft * XXX
391 1.21 mycroft * Note that microtime() returns a timeval, not a timespec.
392 1.9 cgd */
393 1.9 cgd microtime(&vap->va_ctime);
394 1.9 cgd vap->va_atime = vap->va_mtime = vap->va_ctime;
395 1.9 cgd
396 1.21 mycroft /*
397 1.21 mycroft * If the process has exercised some setuid or setgid
398 1.21 mycroft * privilege, then rip away read/write permission so
399 1.21 mycroft * that only root can gain access.
400 1.21 mycroft */
401 1.21 mycroft switch (pfs->pfs_type) {
402 1.21 mycroft case Pmem:
403 1.21 mycroft case Pregs:
404 1.21 mycroft case Pfpregs:
405 1.21 mycroft if (procp->p_flag & P_SUGID)
406 1.21 mycroft vap->va_mode &= ~((VREAD|VWRITE)|
407 1.21 mycroft ((VREAD|VWRITE)>>3)|
408 1.21 mycroft ((VREAD|VWRITE)>>6));
409 1.21 mycroft case Pctl:
410 1.21 mycroft case Pstatus:
411 1.21 mycroft case Pnote:
412 1.21 mycroft case Pnotepg:
413 1.12 ws vap->va_nlink = 1;
414 1.12 ws vap->va_uid = procp->p_ucred->cr_uid;
415 1.12 ws vap->va_gid = procp->p_ucred->cr_gid;
416 1.21 mycroft break;
417 1.12 ws }
418 1.12 ws
419 1.9 cgd /*
420 1.9 cgd * now do the object specific fields
421 1.9 cgd *
422 1.9 cgd * The size could be set from struct reg, but it's hardly
423 1.9 cgd * worth the trouble, and it puts some (potentially) machine
424 1.9 cgd * dependent data into this machine-independent code. If it
425 1.9 cgd * becomes important then this function should break out into
426 1.9 cgd * a per-file stat function in the corresponding .c file.
427 1.9 cgd */
428 1.1 pk
429 1.9 cgd switch (pfs->pfs_type) {
430 1.9 cgd case Proot:
431 1.21 mycroft /*
432 1.21 mycroft * Set nlink to 1 to tell fts(3) we don't actually know.
433 1.21 mycroft */
434 1.21 mycroft vap->va_nlink = 1;
435 1.21 mycroft vap->va_uid = 0;
436 1.21 mycroft vap->va_gid = 0;
437 1.21 mycroft vap->va_size = vap->va_bytes = DEV_BSIZE;
438 1.21 mycroft break;
439 1.21 mycroft
440 1.21 mycroft case Pcurproc: {
441 1.21 mycroft char buf[16]; /* should be enough */
442 1.21 mycroft vap->va_nlink = 1;
443 1.21 mycroft vap->va_uid = 0;
444 1.21 mycroft vap->va_gid = 0;
445 1.21 mycroft vap->va_size = vap->va_bytes =
446 1.21 mycroft sprintf(buf, "%ld", (long)curproc->p_pid);
447 1.9 cgd break;
448 1.21 mycroft }
449 1.9 cgd
450 1.9 cgd case Pproc:
451 1.9 cgd vap->va_nlink = 2;
452 1.21 mycroft vap->va_uid = procp->p_ucred->cr_uid;
453 1.21 mycroft vap->va_gid = procp->p_ucred->cr_gid;
454 1.21 mycroft vap->va_size = vap->va_bytes = DEV_BSIZE;
455 1.21 mycroft break;
456 1.21 mycroft
457 1.21 mycroft case Pfile:
458 1.21 mycroft error = EOPNOTSUPP;
459 1.9 cgd break;
460 1.9 cgd
461 1.9 cgd case Pmem:
462 1.9 cgd vap->va_bytes = vap->va_size =
463 1.9 cgd ctob(procp->p_vmspace->vm_tsize +
464 1.9 cgd procp->p_vmspace->vm_dsize +
465 1.9 cgd procp->p_vmspace->vm_ssize);
466 1.9 cgd break;
467 1.9 cgd
468 1.18 cgd #if defined(PT_GETREGS) || defined(PT_SETREGS)
469 1.9 cgd case Pregs:
470 1.11 cgd vap->va_bytes = vap->va_size = sizeof(struct reg);
471 1.18 cgd break;
472 1.14 cgd #endif
473 1.14 cgd
474 1.18 cgd #if defined(PT_GETFPREGS) || defined(PT_SETFPREGS)
475 1.14 cgd case Pfpregs:
476 1.14 cgd vap->va_bytes = vap->va_size = sizeof(struct fpreg);
477 1.18 cgd break;
478 1.14 cgd #endif
479 1.12 ws
480 1.21 mycroft case Pctl:
481 1.12 ws case Pstatus:
482 1.9 cgd case Pnote:
483 1.9 cgd case Pnotepg:
484 1.9 cgd break;
485 1.1 pk
486 1.9 cgd default:
487 1.21 mycroft panic("procfs_getattr");
488 1.1 pk }
489 1.1 pk
490 1.9 cgd return (error);
491 1.1 pk }
492 1.1 pk
493 1.21 mycroft procfs_setattr(ap)
494 1.22 mycroft struct vop_setattr_args /* {
495 1.22 mycroft struct vnode *a_vp;
496 1.22 mycroft struct vattr *a_vap;
497 1.22 mycroft struct ucred *a_cred;
498 1.22 mycroft struct proc *a_p;
499 1.22 mycroft } */ *ap;
500 1.5 pk {
501 1.5 pk /*
502 1.9 cgd * just fake out attribute setting
503 1.9 cgd * it's not good to generate an error
504 1.9 cgd * return, otherwise things like creat()
505 1.9 cgd * will fail when they try to set the
506 1.9 cgd * file length to 0. worse, this means
507 1.9 cgd * that echo $note > /proc/$pid/note will fail.
508 1.5 pk */
509 1.5 pk
510 1.9 cgd return (0);
511 1.5 pk }
512 1.5 pk
513 1.9 cgd /*
514 1.9 cgd * implement access checking.
515 1.9 cgd *
516 1.9 cgd * something very similar to this code is duplicated
517 1.9 cgd * throughout the 4bsd kernel and should be moved
518 1.9 cgd * into kern/vfs_subr.c sometime.
519 1.9 cgd *
520 1.9 cgd * actually, the check for super-user is slightly
521 1.9 cgd * broken since it will allow read access to write-only
522 1.9 cgd * objects. this doesn't cause any particular trouble
523 1.9 cgd * but does mean that the i/o entry points need to check
524 1.9 cgd * that the operation really does make sense.
525 1.9 cgd */
526 1.21 mycroft procfs_access(ap)
527 1.22 mycroft struct vop_access_args /* {
528 1.22 mycroft struct vnode *a_vp;
529 1.22 mycroft int a_mode;
530 1.22 mycroft struct ucred *a_cred;
531 1.22 mycroft struct proc *a_p;
532 1.22 mycroft } */ *ap;
533 1.1 pk {
534 1.9 cgd struct vattr *vap;
535 1.1 pk struct vattr vattr;
536 1.1 pk int error;
537 1.1 pk
538 1.1 pk /*
539 1.1 pk * If you're the super-user,
540 1.1 pk * you always get access.
541 1.1 pk */
542 1.21 mycroft if (ap->a_cred->cr_uid == 0)
543 1.1 pk return (0);
544 1.21 mycroft
545 1.1 pk vap = &vattr;
546 1.21 mycroft if (error = VOP_GETATTR(ap->a_vp, vap, ap->a_cred, ap->a_p))
547 1.1 pk return (error);
548 1.9 cgd
549 1.1 pk /*
550 1.1 pk * Access check is based on only one of owner, group, public.
551 1.1 pk * If not owner, then check group. If not a member of the
552 1.1 pk * group, then check public access.
553 1.1 pk */
554 1.21 mycroft if (ap->a_cred->cr_uid != vap->va_uid) {
555 1.9 cgd gid_t *gp;
556 1.9 cgd int i;
557 1.9 cgd
558 1.21 mycroft ap->a_mode >>= 3;
559 1.21 mycroft gp = ap->a_cred->cr_groups;
560 1.21 mycroft for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++)
561 1.1 pk if (vap->va_gid == *gp)
562 1.1 pk goto found;
563 1.21 mycroft ap->a_mode >>= 3;
564 1.1 pk found:
565 1.1 pk ;
566 1.1 pk }
567 1.9 cgd
568 1.21 mycroft if ((vap->va_mode & ap->a_mode) == ap->a_mode)
569 1.1 pk return (0);
570 1.9 cgd
571 1.1 pk return (EACCES);
572 1.1 pk }
573 1.1 pk
574 1.1 pk /*
575 1.9 cgd * lookup. this is incredibly complicated in the
576 1.9 cgd * general case, however for most pseudo-filesystems
577 1.9 cgd * very little needs to be done.
578 1.9 cgd *
579 1.9 cgd * unless you want to get a migraine, just make sure your
580 1.9 cgd * filesystem doesn't do any locking of its own. otherwise
581 1.9 cgd * read and inwardly digest ufs_lookup().
582 1.9 cgd */
583 1.21 mycroft procfs_lookup(ap)
584 1.22 mycroft struct vop_lookup_args /* {
585 1.22 mycroft struct vnode * a_dvp;
586 1.22 mycroft struct vnode ** a_vpp;
587 1.22 mycroft struct componentname * a_cnp;
588 1.22 mycroft } */ *ap;
589 1.1 pk {
590 1.21 mycroft struct componentname *cnp = ap->a_cnp;
591 1.21 mycroft struct vnode **vpp = ap->a_vpp;
592 1.21 mycroft struct vnode *dvp = ap->a_dvp;
593 1.21 mycroft char *pname = cnp->cn_nameptr;
594 1.1 pk pid_t pid;
595 1.1 pk struct vnode *nvp;
596 1.9 cgd struct pfsnode *pfs;
597 1.1 pk struct proc *procp;
598 1.9 cgd pfstype pfs_type;
599 1.9 cgd int i;
600 1.1 pk
601 1.21 mycroft if (cnp->cn_namelen == 1 && *pname == '.') {
602 1.21 mycroft *vpp = dvp;
603 1.9 cgd VREF(dvp);
604 1.21 mycroft /*VOP_LOCK(dvp);*/
605 1.9 cgd return (0);
606 1.9 cgd }
607 1.1 pk
608 1.21 mycroft *vpp = NULL;
609 1.1 pk
610 1.9 cgd pfs = VTOPFS(dvp);
611 1.9 cgd switch (pfs->pfs_type) {
612 1.9 cgd case Proot:
613 1.21 mycroft if (cnp->cn_flags & ISDOTDOT)
614 1.9 cgd return (EIO);
615 1.9 cgd
616 1.21 mycroft if (CNEQ(cnp, "curproc", 7))
617 1.21 mycroft return (procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc));
618 1.12 ws
619 1.21 mycroft pid = atopid(pname, cnp->cn_namelen);
620 1.9 cgd if (pid == NO_PID)
621 1.9 cgd return (ENOENT);
622 1.9 cgd
623 1.9 cgd procp = PFIND(pid);
624 1.9 cgd if (procp == 0)
625 1.9 cgd return (ENOENT);
626 1.9 cgd
627 1.21 mycroft return (procfs_allocvp(dvp->v_mount, vpp, pid, Pproc));
628 1.9 cgd
629 1.9 cgd case Pproc:
630 1.23 mycroft if (cnp->cn_flags & ISDOTDOT)
631 1.23 mycroft return (procfs_root(dvp->v_mount, vpp));
632 1.1 pk
633 1.9 cgd procp = PFIND(pfs->pfs_pid);
634 1.9 cgd if (procp == 0)
635 1.9 cgd return (ENOENT);
636 1.9 cgd
637 1.9 cgd for (i = 0; i < Nprocent; i++) {
638 1.9 cgd struct pfsnames *dp = &procent[i];
639 1.9 cgd
640 1.21 mycroft if (cnp->cn_namelen == dp->d_namlen &&
641 1.18 cgd bcmp(pname, dp->d_name, dp->d_namlen) == 0 &&
642 1.18 cgd (dp->d_valid == NULL || (*dp->d_valid)(procp))) {
643 1.9 cgd pfs_type = dp->d_pfstype;
644 1.9 cgd goto found;
645 1.9 cgd }
646 1.9 cgd }
647 1.9 cgd return (ENOENT);
648 1.1 pk
649 1.9 cgd found:
650 1.9 cgd if (pfs_type == Pfile) {
651 1.9 cgd nvp = procfs_findtextvp(procp);
652 1.23 mycroft /* We already checked that it exists. */
653 1.21 mycroft VREF(nvp);
654 1.21 mycroft VOP_LOCK(nvp);
655 1.21 mycroft *vpp = nvp;
656 1.21 mycroft return (0);
657 1.21 mycroft }
658 1.1 pk
659 1.21 mycroft return (procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid,
660 1.21 mycroft pfs_type));
661 1.1 pk
662 1.9 cgd default:
663 1.9 cgd return (ENOTDIR);
664 1.1 pk }
665 1.1 pk }
666 1.1 pk
667 1.18 cgd int
668 1.21 mycroft procfs_validfile(p)
669 1.21 mycroft struct proc *p;
670 1.18 cgd {
671 1.21 mycroft
672 1.21 mycroft return (procfs_findtextvp(p) != NULLVP);
673 1.18 cgd }
674 1.18 cgd
675 1.9 cgd /*
676 1.9 cgd * readdir returns directory entries from pfsnode (vp).
677 1.9 cgd *
678 1.9 cgd * the strategy here with procfs is to generate a single
679 1.9 cgd * directory entry at a time (struct pfsdent) and then
680 1.9 cgd * copy that out to userland using uiomove. a more efficent
681 1.9 cgd * though more complex implementation, would try to minimize
682 1.9 cgd * the number of calls to uiomove(). for procfs, this is
683 1.9 cgd * hardly worth the added code complexity.
684 1.9 cgd *
685 1.9 cgd * this should just be done through read()
686 1.9 cgd */
687 1.21 mycroft procfs_readdir(ap)
688 1.22 mycroft struct vop_readdir_args /* {
689 1.22 mycroft struct vnode *a_vp;
690 1.22 mycroft struct uio *a_uio;
691 1.22 mycroft struct ucred *a_cred;
692 1.22 mycroft int *a_eofflag;
693 1.22 mycroft u_long *a_cookies;
694 1.22 mycroft int a_ncookies;
695 1.22 mycroft } */ *ap;
696 1.1 pk {
697 1.21 mycroft struct uio *uio = ap->a_uio;
698 1.9 cgd struct pfsdent d;
699 1.9 cgd struct pfsdent *dp = &d;
700 1.9 cgd struct pfsnode *pfs;
701 1.9 cgd int error;
702 1.9 cgd int count;
703 1.9 cgd int i;
704 1.22 mycroft
705 1.22 mycroft /*
706 1.22 mycroft * We don't allow exporting procfs mounts, and currently local
707 1.22 mycroft * requests do not need cookies.
708 1.22 mycroft */
709 1.22 mycroft if (ap->a_ncookies)
710 1.22 mycroft panic("procfs_readdir: not hungry");
711 1.9 cgd
712 1.21 mycroft pfs = VTOPFS(ap->a_vp);
713 1.1 pk
714 1.9 cgd if (uio->uio_resid < UIO_MX)
715 1.9 cgd return (EINVAL);
716 1.9 cgd if (uio->uio_offset & (UIO_MX-1))
717 1.9 cgd return (EINVAL);
718 1.9 cgd if (uio->uio_offset < 0)
719 1.1 pk return (EINVAL);
720 1.1 pk
721 1.9 cgd error = 0;
722 1.1 pk count = 0;
723 1.9 cgd i = uio->uio_offset / UIO_MX;
724 1.9 cgd
725 1.9 cgd switch (pfs->pfs_type) {
726 1.9 cgd /*
727 1.9 cgd * this is for the process-specific sub-directories.
728 1.9 cgd * all that is needed to is copy out all the entries
729 1.9 cgd * from the procent[] table (top of this file).
730 1.9 cgd */
731 1.9 cgd case Pproc: {
732 1.21 mycroft pid_t pid = pfs->pfs_pid;
733 1.21 mycroft struct pfsnames *dt;
734 1.21 mycroft
735 1.21 mycroft for (dt = &procent[i]; i < Nprocent && uio->uio_resid >= UIO_MX;
736 1.21 mycroft dt++, i++) {
737 1.21 mycroft struct proc *p = PFIND(pid);
738 1.9 cgd
739 1.21 mycroft if (p == NULL)
740 1.9 cgd break;
741 1.9 cgd
742 1.21 mycroft if (dt->d_valid && (*dt->d_valid)(p) == 0)
743 1.21 mycroft continue;
744 1.21 mycroft
745 1.9 cgd dp->d_reclen = UIO_MX;
746 1.12 ws dp->d_fileno = PROCFS_FILENO(pid, dt->d_pfstype);
747 1.21 mycroft dp->d_namlen = dt->d_namlen;
748 1.21 mycroft bcopy(dt->d_name, dp->d_name, dt->d_namlen + 1);
749 1.21 mycroft dp->d_type = dt->d_type;
750 1.15 ws
751 1.21 mycroft if (error = uiomove((caddr_t)dp, UIO_MX, uio))
752 1.9 cgd break;
753 1.6 ws }
754 1.9 cgd
755 1.9 cgd break;
756 1.9 cgd
757 1.9 cgd }
758 1.9 cgd
759 1.9 cgd /*
760 1.9 cgd * this is for the root of the procfs filesystem
761 1.9 cgd * what is needed is a special entry for "curproc"
762 1.9 cgd * followed by an entry for each process on allproc
763 1.9 cgd #ifdef PROCFS_ZOMBIE
764 1.9 cgd * and zombproc.
765 1.9 cgd #endif
766 1.9 cgd */
767 1.9 cgd
768 1.9 cgd case Proot: {
769 1.9 cgd #ifdef PROCFS_ZOMBIE
770 1.9 cgd int doingzomb = 0;
771 1.9 cgd #endif
772 1.21 mycroft int pcnt = 0;
773 1.21 mycroft volatile struct proc *p = allproc;
774 1.9 cgd
775 1.21 mycroft again:
776 1.21 mycroft for (; p && uio->uio_resid >= UIO_MX; i++, pcnt++) {
777 1.9 cgd bzero((char *) dp, UIO_MX);
778 1.9 cgd dp->d_reclen = UIO_MX;
779 1.9 cgd
780 1.9 cgd switch (i) {
781 1.12 ws case 0: /* `.' */
782 1.12 ws case 1: /* `..' */
783 1.12 ws dp->d_fileno = PROCFS_FILENO(0, Proot);
784 1.12 ws dp->d_namlen = i + 1;
785 1.21 mycroft bcopy("..", dp->d_name, dp->d_namlen);
786 1.21 mycroft dp->d_name[i + 1] = '\0';
787 1.21 mycroft dp->d_type = DT_DIR;
788 1.12 ws break;
789 1.21 mycroft
790 1.12 ws case 2:
791 1.21 mycroft dp->d_fileno = PROCFS_FILENO(0, Pcurproc);
792 1.9 cgd dp->d_namlen = 7;
793 1.21 mycroft bcopy("curproc", dp->d_name, 8);
794 1.21 mycroft dp->d_type = DT_LNK;
795 1.9 cgd break;
796 1.9 cgd
797 1.9 cgd default:
798 1.21 mycroft while (pcnt < i) {
799 1.21 mycroft pcnt++;
800 1.21 mycroft p = p->p_next;
801 1.21 mycroft if (!p)
802 1.21 mycroft goto done;
803 1.9 cgd }
804 1.21 mycroft dp->d_fileno = PROCFS_FILENO(p->p_pid, Pproc);
805 1.21 mycroft dp->d_namlen = sprintf(dp->d_name, "%ld",
806 1.21 mycroft (long)p->p_pid);
807 1.21 mycroft dp->d_type = DT_REG;
808 1.20 cgd p = p->p_next;
809 1.9 cgd break;
810 1.1 pk }
811 1.21 mycroft
812 1.21 mycroft if (error = uiomove((caddr_t)dp, UIO_MX, uio))
813 1.9 cgd break;
814 1.1 pk }
815 1.21 mycroft done:
816 1.21 mycroft
817 1.21 mycroft #ifdef PROCFS_ZOMBIE
818 1.21 mycroft if (p == 0 && doingzomb == 0) {
819 1.21 mycroft doingzomb = 1;
820 1.21 mycroft p = zombproc;
821 1.21 mycroft goto again;
822 1.21 mycroft }
823 1.21 mycroft #endif
824 1.1 pk
825 1.9 cgd break;
826 1.9 cgd
827 1.9 cgd }
828 1.9 cgd
829 1.9 cgd default:
830 1.9 cgd error = ENOTDIR;
831 1.9 cgd break;
832 1.1 pk }
833 1.9 cgd
834 1.9 cgd uio->uio_offset = i * UIO_MX;
835 1.1 pk
836 1.9 cgd return (error);
837 1.12 ws }
838 1.12 ws
839 1.12 ws /*
840 1.21 mycroft * readlink reads the link of `curproc'
841 1.12 ws */
842 1.21 mycroft procfs_readlink(ap)
843 1.21 mycroft struct vop_readlink_args *ap;
844 1.12 ws {
845 1.21 mycroft struct uio *uio = ap->a_uio;
846 1.12 ws char buf[16]; /* should be enough */
847 1.12 ws int len;
848 1.12 ws
849 1.21 mycroft if (VTOPFS(ap->a_vp)->pfs_fileno != PROCFS_FILENO(0, Pcurproc))
850 1.21 mycroft return (EINVAL);
851 1.12 ws
852 1.21 mycroft len = sprintf(buf, "%ld", (long)curproc->p_pid);
853 1.12 ws
854 1.21 mycroft return (uiomove((caddr_t)buf, len, ap->a_uio));
855 1.1 pk }
856 1.1 pk
857 1.1 pk /*
858 1.9 cgd * convert decimal ascii to pid_t
859 1.1 pk */
860 1.9 cgd static pid_t
861 1.9 cgd atopid(b, len)
862 1.9 cgd const char *b;
863 1.9 cgd u_int len;
864 1.1 pk {
865 1.9 cgd pid_t p = 0;
866 1.1 pk
867 1.1 pk while (len--) {
868 1.9 cgd char c = *b++;
869 1.1 pk if (c < '0' || c > '9')
870 1.9 cgd return (NO_PID);
871 1.9 cgd p = 10 * p + (c - '0');
872 1.9 cgd if (p > PID_MAX)
873 1.9 cgd return (NO_PID);
874 1.1 pk }
875 1.9 cgd
876 1.9 cgd return (p);
877 1.1 pk }
878 1.9 cgd
879 1.9 cgd /*
880 1.9 cgd * procfs vnode operations.
881 1.9 cgd */
882 1.21 mycroft int (**procfs_vnodeop_p)();
883 1.21 mycroft struct vnodeopv_entry_desc procfs_vnodeop_entries[] = {
884 1.21 mycroft { &vop_default_desc, vn_default_error },
885 1.21 mycroft { &vop_lookup_desc, procfs_lookup }, /* lookup */
886 1.21 mycroft { &vop_create_desc, procfs_create }, /* create */
887 1.21 mycroft { &vop_mknod_desc, procfs_mknod }, /* mknod */
888 1.21 mycroft { &vop_open_desc, procfs_open }, /* open */
889 1.21 mycroft { &vop_close_desc, procfs_close }, /* close */
890 1.21 mycroft { &vop_access_desc, procfs_access }, /* access */
891 1.21 mycroft { &vop_getattr_desc, procfs_getattr }, /* getattr */
892 1.21 mycroft { &vop_setattr_desc, procfs_setattr }, /* setattr */
893 1.21 mycroft { &vop_read_desc, procfs_read }, /* read */
894 1.21 mycroft { &vop_write_desc, procfs_write }, /* write */
895 1.21 mycroft { &vop_ioctl_desc, procfs_ioctl }, /* ioctl */
896 1.21 mycroft { &vop_select_desc, procfs_select }, /* select */
897 1.21 mycroft { &vop_mmap_desc, procfs_mmap }, /* mmap */
898 1.21 mycroft { &vop_fsync_desc, procfs_fsync }, /* fsync */
899 1.21 mycroft { &vop_seek_desc, procfs_seek }, /* seek */
900 1.21 mycroft { &vop_remove_desc, procfs_remove }, /* remove */
901 1.21 mycroft { &vop_link_desc, procfs_link }, /* link */
902 1.21 mycroft { &vop_rename_desc, procfs_rename }, /* rename */
903 1.21 mycroft { &vop_mkdir_desc, procfs_mkdir }, /* mkdir */
904 1.21 mycroft { &vop_rmdir_desc, procfs_rmdir }, /* rmdir */
905 1.21 mycroft { &vop_symlink_desc, procfs_symlink }, /* symlink */
906 1.21 mycroft { &vop_readdir_desc, procfs_readdir }, /* readdir */
907 1.21 mycroft { &vop_readlink_desc, procfs_readlink }, /* readlink */
908 1.21 mycroft { &vop_abortop_desc, procfs_abortop }, /* abortop */
909 1.21 mycroft { &vop_inactive_desc, procfs_inactive }, /* inactive */
910 1.21 mycroft { &vop_reclaim_desc, procfs_reclaim }, /* reclaim */
911 1.21 mycroft { &vop_lock_desc, procfs_lock }, /* lock */
912 1.21 mycroft { &vop_unlock_desc, procfs_unlock }, /* unlock */
913 1.21 mycroft { &vop_bmap_desc, procfs_bmap }, /* bmap */
914 1.21 mycroft { &vop_strategy_desc, procfs_strategy }, /* strategy */
915 1.21 mycroft { &vop_print_desc, procfs_print }, /* print */
916 1.21 mycroft { &vop_islocked_desc, procfs_islocked }, /* islocked */
917 1.21 mycroft { &vop_pathconf_desc, procfs_pathconf }, /* pathconf */
918 1.21 mycroft { &vop_advlock_desc, procfs_advlock }, /* advlock */
919 1.21 mycroft { &vop_blkatoff_desc, procfs_blkatoff }, /* blkatoff */
920 1.21 mycroft { &vop_valloc_desc, procfs_valloc }, /* valloc */
921 1.21 mycroft { &vop_vfree_desc, procfs_vfree }, /* vfree */
922 1.21 mycroft { &vop_truncate_desc, procfs_truncate }, /* truncate */
923 1.21 mycroft { &vop_update_desc, procfs_update }, /* update */
924 1.21 mycroft { (struct vnodeop_desc*)NULL, (int(*)())NULL }
925 1.9 cgd };
926 1.21 mycroft struct vnodeopv_desc procfs_vnodeop_opv_desc =
927 1.21 mycroft { &procfs_vnodeop_p, procfs_vnodeop_entries };
928