vfs_vnops.c revision 1.92 1 1.92 elad /* $NetBSD: vfs_vnops.c,v 1.92 2005/06/19 18:22:36 elad Exp $ */
2 1.12 cgd
3 1.10 cgd /*
4 1.11 mycroft * Copyright (c) 1982, 1986, 1989, 1993
5 1.11 mycroft * The Regents of the University of California. All rights reserved.
6 1.10 cgd * (c) UNIX System Laboratories, Inc.
7 1.10 cgd * All or some portions of this file are derived from material licensed
8 1.10 cgd * to the University of California by American Telephone and Telegraph
9 1.10 cgd * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 1.10 cgd * the permission of UNIX System Laboratories, Inc.
11 1.10 cgd *
12 1.10 cgd * Redistribution and use in source and binary forms, with or without
13 1.10 cgd * modification, are permitted provided that the following conditions
14 1.10 cgd * are met:
15 1.10 cgd * 1. Redistributions of source code must retain the above copyright
16 1.10 cgd * notice, this list of conditions and the following disclaimer.
17 1.10 cgd * 2. Redistributions in binary form must reproduce the above copyright
18 1.10 cgd * notice, this list of conditions and the following disclaimer in the
19 1.10 cgd * documentation and/or other materials provided with the distribution.
20 1.73 agc * 3. Neither the name of the University nor the names of its contributors
21 1.10 cgd * may be used to endorse or promote products derived from this software
22 1.10 cgd * without specific prior written permission.
23 1.10 cgd *
24 1.10 cgd * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 1.10 cgd * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 1.10 cgd * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 1.10 cgd * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 1.10 cgd * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 1.10 cgd * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 1.10 cgd * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 1.10 cgd * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 1.10 cgd * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 1.10 cgd * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 1.10 cgd * SUCH DAMAGE.
35 1.10 cgd *
36 1.28 fvdl * @(#)vfs_vnops.c 8.14 (Berkeley) 6/15/95
37 1.10 cgd */
38 1.52 lukem
39 1.52 lukem #include <sys/cdefs.h>
40 1.92 elad __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.92 2005/06/19 18:22:36 elad Exp $");
41 1.26 mrg
42 1.27 thorpej #include "fs_union.h"
43 1.10 cgd
44 1.10 cgd #include <sys/param.h>
45 1.10 cgd #include <sys/systm.h>
46 1.10 cgd #include <sys/kernel.h>
47 1.10 cgd #include <sys/file.h>
48 1.10 cgd #include <sys/stat.h>
49 1.10 cgd #include <sys/buf.h>
50 1.10 cgd #include <sys/proc.h>
51 1.77 hannken #include <sys/malloc.h>
52 1.10 cgd #include <sys/mount.h>
53 1.10 cgd #include <sys/namei.h>
54 1.10 cgd #include <sys/vnode.h>
55 1.10 cgd #include <sys/ioctl.h>
56 1.10 cgd #include <sys/tty.h>
57 1.21 mycroft #include <sys/poll.h>
58 1.11 mycroft
59 1.77 hannken #include <miscfs/specfs/specdev.h>
60 1.77 hannken
61 1.25 mrg #include <uvm/uvm_extern.h>
62 1.25 mrg
63 1.28 fvdl #ifdef UNION
64 1.65 jdolecek #include <fs/union/union.h>
65 1.28 fvdl #endif
66 1.64 jdolecek
67 1.66 jdolecek #if defined(LKM) || defined(UNION)
68 1.72 fvdl int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *);
69 1.66 jdolecek #endif
70 1.66 jdolecek
71 1.64 jdolecek #ifdef VERIFIED_EXEC
72 1.58 blymn #include <sys/verified_exec.h>
73 1.64 jdolecek #endif
74 1.28 fvdl
75 1.56 gmcgarry static int vn_read(struct file *fp, off_t *offset, struct uio *uio,
76 1.56 gmcgarry struct ucred *cred, int flags);
77 1.56 gmcgarry static int vn_write(struct file *fp, off_t *offset, struct uio *uio,
78 1.56 gmcgarry struct ucred *cred, int flags);
79 1.72 fvdl static int vn_closefile(struct file *fp, struct proc *p);
80 1.72 fvdl static int vn_poll(struct file *fp, int events, struct proc *p);
81 1.72 fvdl static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p);
82 1.72 fvdl static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p);
83 1.72 fvdl static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p);
84 1.48 jdolecek
85 1.83 christos const struct fileops vnops = {
86 1.48 jdolecek vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll,
87 1.57 jdolecek vn_statfile, vn_closefile, vn_kqfilter
88 1.48 jdolecek };
89 1.10 cgd
90 1.10 cgd /*
91 1.10 cgd * Common code for vnode open operations.
92 1.10 cgd * Check permissions, and call the VOP_OPEN or VOP_CREATE routine.
93 1.10 cgd */
94 1.20 christos int
95 1.89 thorpej vn_open(struct nameidata *ndp, int fmode, int cmode)
96 1.10 cgd {
97 1.91 elad struct vnode *vp = NULL; /* XXXGCC */
98 1.75 hannken struct mount *mp;
99 1.72 fvdl struct proc *p = ndp->ni_cnd.cn_proc;
100 1.72 fvdl struct ucred *cred = p->p_ucred;
101 1.19 mycroft struct vattr va;
102 1.10 cgd int error;
103 1.92 elad #ifdef VERIFIED_EXEC
104 1.92 elad struct veriexec_hash_entry *vhe = NULL;
105 1.92 elad #endif /* VERIFIED_EXEC */
106 1.75 hannken
107 1.75 hannken restart:
108 1.10 cgd if (fmode & O_CREAT) {
109 1.11 mycroft ndp->ni_cnd.cn_nameiop = CREATE;
110 1.11 mycroft ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
111 1.38 bouyer if ((fmode & O_EXCL) == 0 &&
112 1.59 christos ((fmode & O_NOFOLLOW) == 0))
113 1.11 mycroft ndp->ni_cnd.cn_flags |= FOLLOW;
114 1.20 christos if ((error = namei(ndp)) != 0)
115 1.10 cgd return (error);
116 1.10 cgd if (ndp->ni_vp == NULL) {
117 1.91 elad #ifdef VERIFIED_EXEC
118 1.91 elad /* Lockdown mode: Prevent creation of new files. */
119 1.91 elad if (veriexec_strict >= 3) {
120 1.91 elad VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
121 1.91 elad
122 1.91 elad printf("Veriexec: vn_open: Preventing "
123 1.91 elad "new file creation in %s.\n",
124 1.91 elad ndp->ni_dirp);
125 1.91 elad
126 1.91 elad error = EPERM;
127 1.91 elad goto bad;
128 1.91 elad }
129 1.91 elad #endif /* VERIFIED_EXEC */
130 1.91 elad
131 1.19 mycroft VATTR_NULL(&va);
132 1.19 mycroft va.va_type = VREG;
133 1.19 mycroft va.va_mode = cmode;
134 1.28 fvdl if (fmode & O_EXCL)
135 1.28 fvdl va.va_vaflags |= VA_EXCLUSIVE;
136 1.75 hannken if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) {
137 1.75 hannken VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
138 1.75 hannken vput(ndp->ni_dvp);
139 1.75 hannken if ((error = vn_start_write(NULL, &mp,
140 1.75 hannken V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0)
141 1.75 hannken return (error);
142 1.75 hannken goto restart;
143 1.75 hannken }
144 1.72 fvdl VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE);
145 1.20 christos error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
146 1.20 christos &ndp->ni_cnd, &va);
147 1.75 hannken vn_finished_write(mp, 0);
148 1.20 christos if (error)
149 1.10 cgd return (error);
150 1.10 cgd fmode &= ~O_TRUNC;
151 1.10 cgd vp = ndp->ni_vp;
152 1.10 cgd } else {
153 1.11 mycroft VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd);
154 1.10 cgd if (ndp->ni_dvp == ndp->ni_vp)
155 1.10 cgd vrele(ndp->ni_dvp);
156 1.10 cgd else
157 1.10 cgd vput(ndp->ni_dvp);
158 1.10 cgd ndp->ni_dvp = NULL;
159 1.10 cgd vp = ndp->ni_vp;
160 1.10 cgd if (fmode & O_EXCL) {
161 1.10 cgd error = EEXIST;
162 1.38 bouyer goto bad;
163 1.38 bouyer }
164 1.10 cgd fmode &= ~O_CREAT;
165 1.10 cgd }
166 1.10 cgd } else {
167 1.11 mycroft ndp->ni_cnd.cn_nameiop = LOOKUP;
168 1.74 cb ndp->ni_cnd.cn_flags = LOCKLEAF;
169 1.74 cb if ((fmode & O_NOFOLLOW) == 0)
170 1.74 cb ndp->ni_cnd.cn_flags |= FOLLOW;
171 1.20 christos if ((error = namei(ndp)) != 0)
172 1.10 cgd return (error);
173 1.10 cgd vp = ndp->ni_vp;
174 1.10 cgd }
175 1.10 cgd if (vp->v_type == VSOCK) {
176 1.10 cgd error = EOPNOTSUPP;
177 1.74 cb goto bad;
178 1.74 cb }
179 1.74 cb if (ndp->ni_vp->v_type == VLNK) {
180 1.74 cb error = EFTYPE;
181 1.10 cgd goto bad;
182 1.10 cgd }
183 1.58 blymn
184 1.58 blymn #ifdef VERIFIED_EXEC
185 1.72 fvdl if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0)
186 1.58 blymn goto bad;
187 1.58 blymn #endif
188 1.58 blymn
189 1.10 cgd if ((fmode & O_CREAT) == 0) {
190 1.58 blymn #ifdef VERIFIED_EXEC
191 1.91 elad /* XXX may need pathbuf instead */
192 1.91 elad if ((error = veriexec_verify(p, vp, &va, ndp->ni_dirp,
193 1.92 elad VERIEXEC_FILE, &vhe)) != 0)
194 1.87 blymn goto bad;
195 1.58 blymn #endif
196 1.91 elad
197 1.10 cgd if (fmode & FREAD) {
198 1.72 fvdl if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0)
199 1.10 cgd goto bad;
200 1.10 cgd }
201 1.87 blymn
202 1.10 cgd if (fmode & (FWRITE | O_TRUNC)) {
203 1.10 cgd if (vp->v_type == VDIR) {
204 1.10 cgd error = EISDIR;
205 1.10 cgd goto bad;
206 1.10 cgd }
207 1.20 christos if ((error = vn_writechk(vp)) != 0 ||
208 1.72 fvdl (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0)
209 1.10 cgd goto bad;
210 1.58 blymn #ifdef VERIFIED_EXEC
211 1.92 elad if (vhe != NULL) {
212 1.90 elad veriexec_report("Write access request.",
213 1.90 elad ndp->ni_dirp, &va, p,
214 1.90 elad REPORT_NOVERBOSE,
215 1.90 elad REPORT_ALARM,
216 1.90 elad REPORT_NOPANIC);
217 1.90 elad
218 1.91 elad /* IPS mode: Deny writing to monitored files. */
219 1.91 elad if (veriexec_strict >= 2) {
220 1.58 blymn error = EPERM;
221 1.58 blymn goto bad;
222 1.58 blymn } else {
223 1.92 elad vhe->status = FINGERPRINT_NOTEVAL;
224 1.58 blymn }
225 1.58 blymn }
226 1.58 blymn #endif
227 1.10 cgd }
228 1.10 cgd }
229 1.87 blymn
230 1.10 cgd if (fmode & O_TRUNC) {
231 1.28 fvdl VOP_UNLOCK(vp, 0); /* XXX */
232 1.75 hannken if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) {
233 1.75 hannken vput(vp);
234 1.75 hannken return (error);
235 1.75 hannken }
236 1.72 fvdl VOP_LEASE(vp, p, cred, LEASE_WRITE);
237 1.28 fvdl vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
238 1.19 mycroft VATTR_NULL(&va);
239 1.19 mycroft va.va_size = 0;
240 1.75 hannken error = VOP_SETATTR(vp, &va, cred, p);
241 1.75 hannken vn_finished_write(mp, 0);
242 1.75 hannken if (error != 0)
243 1.10 cgd goto bad;
244 1.10 cgd }
245 1.72 fvdl if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0)
246 1.10 cgd goto bad;
247 1.45 chs if (vp->v_type == VREG &&
248 1.45 chs uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) {
249 1.45 chs error = EIO;
250 1.45 chs goto bad;
251 1.45 chs }
252 1.10 cgd if (fmode & FWRITE)
253 1.10 cgd vp->v_writecount++;
254 1.45 chs
255 1.10 cgd return (0);
256 1.10 cgd bad:
257 1.10 cgd vput(vp);
258 1.10 cgd return (error);
259 1.10 cgd }
260 1.10 cgd
261 1.10 cgd /*
262 1.10 cgd * Check for write permissions on the specified vnode.
263 1.28 fvdl * Prototype text segments cannot be written.
264 1.10 cgd */
265 1.20 christos int
266 1.89 thorpej vn_writechk(struct vnode *vp)
267 1.10 cgd {
268 1.10 cgd
269 1.10 cgd /*
270 1.45 chs * If the vnode is in use as a process's text,
271 1.45 chs * we can't allow writing.
272 1.10 cgd */
273 1.45 chs if (vp->v_flag & VTEXT)
274 1.25 mrg return (ETXTBSY);
275 1.10 cgd return (0);
276 1.42 chs }
277 1.42 chs
278 1.42 chs /*
279 1.51 thorpej * Mark a vnode as having executable mappings.
280 1.42 chs */
281 1.42 chs void
282 1.89 thorpej vn_markexec(struct vnode *vp)
283 1.42 chs {
284 1.51 thorpej if ((vp->v_flag & VEXECMAP) == 0) {
285 1.53 chs uvmexp.filepages -= vp->v_uobj.uo_npages;
286 1.53 chs uvmexp.execpages += vp->v_uobj.uo_npages;
287 1.46 chs }
288 1.51 thorpej vp->v_flag |= VEXECMAP;
289 1.55 chs }
290 1.55 chs
291 1.55 chs /*
292 1.55 chs * Mark a vnode as being the text of a process.
293 1.55 chs * Fail if the vnode is currently writable.
294 1.55 chs */
295 1.55 chs int
296 1.89 thorpej vn_marktext(struct vnode *vp)
297 1.55 chs {
298 1.55 chs
299 1.55 chs if (vp->v_writecount != 0) {
300 1.55 chs KASSERT((vp->v_flag & VTEXT) == 0);
301 1.55 chs return (ETXTBSY);
302 1.55 chs }
303 1.55 chs vp->v_flag |= VTEXT;
304 1.55 chs vn_markexec(vp);
305 1.55 chs return (0);
306 1.10 cgd }
307 1.10 cgd
308 1.10 cgd /*
309 1.10 cgd * Vnode close call
310 1.32 wrstuden *
311 1.32 wrstuden * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node.
312 1.10 cgd */
313 1.20 christos int
314 1.89 thorpej vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p)
315 1.10 cgd {
316 1.10 cgd int error;
317 1.10 cgd
318 1.10 cgd if (flags & FWRITE)
319 1.10 cgd vp->v_writecount--;
320 1.32 wrstuden vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
321 1.72 fvdl error = VOP_CLOSE(vp, flags, cred, p);
322 1.32 wrstuden vput(vp);
323 1.10 cgd return (error);
324 1.10 cgd }
325 1.10 cgd
326 1.10 cgd /*
327 1.10 cgd * Package up an I/O request on a vnode into a uio and do it.
328 1.10 cgd */
329 1.20 christos int
330 1.89 thorpej vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset,
331 1.89 thorpej enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid,
332 1.89 thorpej struct proc *p)
333 1.10 cgd {
334 1.10 cgd struct uio auio;
335 1.10 cgd struct iovec aiov;
336 1.75 hannken struct mount *mp;
337 1.10 cgd int error;
338 1.10 cgd
339 1.50 chs if ((ioflg & IO_NODELOCKED) == 0) {
340 1.50 chs if (rw == UIO_READ) {
341 1.50 chs vn_lock(vp, LK_SHARED | LK_RETRY);
342 1.75 hannken } else /* UIO_WRITE */ {
343 1.75 hannken if (vp->v_type != VCHR &&
344 1.75 hannken (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH))
345 1.75 hannken != 0)
346 1.75 hannken return (error);
347 1.50 chs vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
348 1.50 chs }
349 1.50 chs }
350 1.10 cgd auio.uio_iov = &aiov;
351 1.10 cgd auio.uio_iovcnt = 1;
352 1.10 cgd aiov.iov_base = base;
353 1.10 cgd aiov.iov_len = len;
354 1.10 cgd auio.uio_resid = len;
355 1.10 cgd auio.uio_offset = offset;
356 1.10 cgd auio.uio_segflg = segflg;
357 1.10 cgd auio.uio_rw = rw;
358 1.72 fvdl auio.uio_procp = p;
359 1.11 mycroft if (rw == UIO_READ) {
360 1.10 cgd error = VOP_READ(vp, &auio, ioflg, cred);
361 1.11 mycroft } else {
362 1.10 cgd error = VOP_WRITE(vp, &auio, ioflg, cred);
363 1.11 mycroft }
364 1.10 cgd if (aresid)
365 1.10 cgd *aresid = auio.uio_resid;
366 1.10 cgd else
367 1.10 cgd if (auio.uio_resid && error == 0)
368 1.10 cgd error = EIO;
369 1.75 hannken if ((ioflg & IO_NODELOCKED) == 0) {
370 1.75 hannken if (rw == UIO_WRITE)
371 1.75 hannken vn_finished_write(mp, 0);
372 1.28 fvdl VOP_UNLOCK(vp, 0);
373 1.75 hannken }
374 1.10 cgd return (error);
375 1.23 fvdl }
376 1.23 fvdl
377 1.23 fvdl int
378 1.89 thorpej vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done,
379 1.89 thorpej struct proc *p, off_t **cookies, int *ncookies)
380 1.23 fvdl {
381 1.23 fvdl struct vnode *vp = (struct vnode *)fp->f_data;
382 1.23 fvdl struct iovec aiov;
383 1.23 fvdl struct uio auio;
384 1.23 fvdl int error, eofflag;
385 1.23 fvdl
386 1.23 fvdl unionread:
387 1.23 fvdl if (vp->v_type != VDIR)
388 1.23 fvdl return (EINVAL);
389 1.88 christos aiov.iov_base = bf;
390 1.23 fvdl aiov.iov_len = count;
391 1.23 fvdl auio.uio_iov = &aiov;
392 1.23 fvdl auio.uio_iovcnt = 1;
393 1.23 fvdl auio.uio_rw = UIO_READ;
394 1.23 fvdl auio.uio_segflg = segflg;
395 1.72 fvdl auio.uio_procp = p;
396 1.23 fvdl auio.uio_resid = count;
397 1.50 chs vn_lock(vp, LK_SHARED | LK_RETRY);
398 1.23 fvdl auio.uio_offset = fp->f_offset;
399 1.28 fvdl error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies,
400 1.23 fvdl ncookies);
401 1.23 fvdl fp->f_offset = auio.uio_offset;
402 1.28 fvdl VOP_UNLOCK(vp, 0);
403 1.23 fvdl if (error)
404 1.23 fvdl return (error);
405 1.23 fvdl
406 1.66 jdolecek #if defined(UNION) || defined(LKM)
407 1.66 jdolecek if (count == auio.uio_resid && vn_union_readdir_hook) {
408 1.66 jdolecek struct vnode *ovp = vp;
409 1.23 fvdl
410 1.72 fvdl error = (*vn_union_readdir_hook)(&vp, fp, p);
411 1.66 jdolecek if (error)
412 1.66 jdolecek return (error);
413 1.66 jdolecek if (vp != ovp)
414 1.23 fvdl goto unionread;
415 1.23 fvdl }
416 1.66 jdolecek #endif /* UNION || LKM */
417 1.23 fvdl
418 1.23 fvdl if (count == auio.uio_resid && (vp->v_flag & VROOT) &&
419 1.23 fvdl (vp->v_mount->mnt_flag & MNT_UNION)) {
420 1.23 fvdl struct vnode *tvp = vp;
421 1.23 fvdl vp = vp->v_mount->mnt_vnodecovered;
422 1.23 fvdl VREF(vp);
423 1.68 dsl fp->f_data = vp;
424 1.23 fvdl fp->f_offset = 0;
425 1.23 fvdl vrele(tvp);
426 1.23 fvdl goto unionread;
427 1.23 fvdl }
428 1.23 fvdl *done = count - auio.uio_resid;
429 1.23 fvdl return error;
430 1.10 cgd }
431 1.10 cgd
432 1.10 cgd /*
433 1.10 cgd * File table vnode read routine.
434 1.10 cgd */
435 1.56 gmcgarry static int
436 1.89 thorpej vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred,
437 1.89 thorpej int flags)
438 1.10 cgd {
439 1.28 fvdl struct vnode *vp = (struct vnode *)fp->f_data;
440 1.31 kleink int count, error, ioflag = 0;
441 1.10 cgd
442 1.72 fvdl VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ);
443 1.31 kleink if (fp->f_flag & FNONBLOCK)
444 1.31 kleink ioflag |= IO_NDELAY;
445 1.31 kleink if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC))
446 1.31 kleink ioflag |= IO_SYNC;
447 1.37 wrstuden if (fp->f_flag & FALTIO)
448 1.37 wrstuden ioflag |= IO_ALTSEMANTICS;
449 1.50 chs vn_lock(vp, LK_SHARED | LK_RETRY);
450 1.29 thorpej uio->uio_offset = *offset;
451 1.10 cgd count = uio->uio_resid;
452 1.31 kleink error = VOP_READ(vp, uio, ioflag, cred);
453 1.29 thorpej if (flags & FOF_UPDATE_OFFSET)
454 1.29 thorpej *offset += count - uio->uio_resid;
455 1.28 fvdl VOP_UNLOCK(vp, 0);
456 1.10 cgd return (error);
457 1.10 cgd }
458 1.10 cgd
459 1.10 cgd /*
460 1.10 cgd * File table vnode write routine.
461 1.10 cgd */
462 1.56 gmcgarry static int
463 1.89 thorpej vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred,
464 1.89 thorpej int flags)
465 1.10 cgd {
466 1.28 fvdl struct vnode *vp = (struct vnode *)fp->f_data;
467 1.75 hannken struct mount *mp;
468 1.17 mycroft int count, error, ioflag = IO_UNIT;
469 1.10 cgd
470 1.10 cgd if (vp->v_type == VREG && (fp->f_flag & O_APPEND))
471 1.10 cgd ioflag |= IO_APPEND;
472 1.10 cgd if (fp->f_flag & FNONBLOCK)
473 1.10 cgd ioflag |= IO_NDELAY;
474 1.24 thorpej if (fp->f_flag & FFSYNC ||
475 1.24 thorpej (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS)))
476 1.24 thorpej ioflag |= IO_SYNC;
477 1.31 kleink else if (fp->f_flag & FDSYNC)
478 1.31 kleink ioflag |= IO_DSYNC;
479 1.37 wrstuden if (fp->f_flag & FALTIO)
480 1.37 wrstuden ioflag |= IO_ALTSEMANTICS;
481 1.75 hannken mp = NULL;
482 1.75 hannken if (vp->v_type != VCHR &&
483 1.75 hannken (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0)
484 1.75 hannken return (error);
485 1.72 fvdl VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE);
486 1.28 fvdl vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
487 1.29 thorpej uio->uio_offset = *offset;
488 1.10 cgd count = uio->uio_resid;
489 1.10 cgd error = VOP_WRITE(vp, uio, ioflag, cred);
490 1.29 thorpej if (flags & FOF_UPDATE_OFFSET) {
491 1.29 thorpej if (ioflag & IO_APPEND)
492 1.29 thorpej *offset = uio->uio_offset;
493 1.29 thorpej else
494 1.29 thorpej *offset += count - uio->uio_resid;
495 1.29 thorpej }
496 1.28 fvdl VOP_UNLOCK(vp, 0);
497 1.75 hannken vn_finished_write(mp, 0);
498 1.10 cgd return (error);
499 1.10 cgd }
500 1.10 cgd
501 1.10 cgd /*
502 1.10 cgd * File table vnode stat routine.
503 1.10 cgd */
504 1.48 jdolecek static int
505 1.89 thorpej vn_statfile(struct file *fp, struct stat *sb, struct proc *p)
506 1.48 jdolecek {
507 1.48 jdolecek struct vnode *vp = (struct vnode *)fp->f_data;
508 1.48 jdolecek
509 1.72 fvdl return vn_stat(vp, sb, p);
510 1.48 jdolecek }
511 1.48 jdolecek
512 1.20 christos int
513 1.89 thorpej vn_stat(struct vnode *vp, struct stat *sb, struct proc *p)
514 1.10 cgd {
515 1.19 mycroft struct vattr va;
516 1.10 cgd int error;
517 1.35 wrstuden mode_t mode;
518 1.10 cgd
519 1.72 fvdl error = VOP_GETATTR(vp, &va, p->p_ucred, p);
520 1.10 cgd if (error)
521 1.10 cgd return (error);
522 1.10 cgd /*
523 1.10 cgd * Copy from vattr table
524 1.10 cgd */
525 1.19 mycroft sb->st_dev = va.va_fsid;
526 1.19 mycroft sb->st_ino = va.va_fileid;
527 1.19 mycroft mode = va.va_mode;
528 1.10 cgd switch (vp->v_type) {
529 1.10 cgd case VREG:
530 1.10 cgd mode |= S_IFREG;
531 1.10 cgd break;
532 1.10 cgd case VDIR:
533 1.10 cgd mode |= S_IFDIR;
534 1.10 cgd break;
535 1.10 cgd case VBLK:
536 1.10 cgd mode |= S_IFBLK;
537 1.10 cgd break;
538 1.10 cgd case VCHR:
539 1.10 cgd mode |= S_IFCHR;
540 1.10 cgd break;
541 1.10 cgd case VLNK:
542 1.10 cgd mode |= S_IFLNK;
543 1.10 cgd break;
544 1.10 cgd case VSOCK:
545 1.10 cgd mode |= S_IFSOCK;
546 1.10 cgd break;
547 1.10 cgd case VFIFO:
548 1.10 cgd mode |= S_IFIFO;
549 1.10 cgd break;
550 1.10 cgd default:
551 1.10 cgd return (EBADF);
552 1.10 cgd };
553 1.10 cgd sb->st_mode = mode;
554 1.19 mycroft sb->st_nlink = va.va_nlink;
555 1.19 mycroft sb->st_uid = va.va_uid;
556 1.19 mycroft sb->st_gid = va.va_gid;
557 1.19 mycroft sb->st_rdev = va.va_rdev;
558 1.19 mycroft sb->st_size = va.va_size;
559 1.19 mycroft sb->st_atimespec = va.va_atime;
560 1.19 mycroft sb->st_mtimespec = va.va_mtime;
561 1.19 mycroft sb->st_ctimespec = va.va_ctime;
562 1.69 fvdl sb->st_birthtimespec = va.va_birthtime;
563 1.19 mycroft sb->st_blksize = va.va_blocksize;
564 1.19 mycroft sb->st_flags = va.va_flags;
565 1.22 mycroft sb->st_gen = 0;
566 1.19 mycroft sb->st_blocks = va.va_bytes / S_BLKSIZE;
567 1.10 cgd return (0);
568 1.37 wrstuden }
569 1.37 wrstuden
570 1.37 wrstuden /*
571 1.37 wrstuden * File table vnode fcntl routine.
572 1.37 wrstuden */
573 1.56 gmcgarry static int
574 1.89 thorpej vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p)
575 1.37 wrstuden {
576 1.41 augustss struct vnode *vp = ((struct vnode *)fp->f_data);
577 1.37 wrstuden int error;
578 1.37 wrstuden
579 1.37 wrstuden vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
580 1.72 fvdl error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p);
581 1.37 wrstuden VOP_UNLOCK(vp, 0);
582 1.37 wrstuden return (error);
583 1.10 cgd }
584 1.10 cgd
585 1.10 cgd /*
586 1.10 cgd * File table vnode ioctl routine.
587 1.10 cgd */
588 1.56 gmcgarry static int
589 1.89 thorpej vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p)
590 1.10 cgd {
591 1.41 augustss struct vnode *vp = ((struct vnode *)fp->f_data);
592 1.10 cgd struct vattr vattr;
593 1.10 cgd int error;
594 1.10 cgd
595 1.10 cgd switch (vp->v_type) {
596 1.10 cgd
597 1.10 cgd case VREG:
598 1.10 cgd case VDIR:
599 1.10 cgd if (com == FIONREAD) {
600 1.72 fvdl error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
601 1.20 christos if (error)
602 1.10 cgd return (error);
603 1.10 cgd *(int *)data = vattr.va_size - fp->f_offset;
604 1.10 cgd return (0);
605 1.60 atatat }
606 1.82 christos if ((com == FIONWRITE) || (com == FIONSPACE)) {
607 1.81 wrstuden /*
608 1.81 wrstuden * Files don't have send queues, so there never
609 1.81 wrstuden * are any bytes in them, nor is there any
610 1.81 wrstuden * open space in them.
611 1.81 wrstuden */
612 1.81 wrstuden *(int *)data = 0;
613 1.81 wrstuden return (0);
614 1.81 wrstuden }
615 1.60 atatat if (com == FIOGETBMAP) {
616 1.60 atatat daddr_t *block;
617 1.60 atatat
618 1.62 atatat if (*(daddr_t *)data < 0)
619 1.62 atatat return (EINVAL);
620 1.60 atatat block = (daddr_t *)data;
621 1.60 atatat return (VOP_BMAP(vp, *block, NULL, block, NULL));
622 1.61 fvdl }
623 1.61 fvdl if (com == OFIOGETBMAP) {
624 1.61 fvdl daddr_t ibn, obn;
625 1.61 fvdl
626 1.62 atatat if (*(int32_t *)data < 0)
627 1.62 atatat return (EINVAL);
628 1.61 fvdl ibn = (daddr_t)*(int32_t *)data;
629 1.61 fvdl error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
630 1.61 fvdl *(int32_t *)data = (int32_t)obn;
631 1.61 fvdl return error;
632 1.10 cgd }
633 1.10 cgd if (com == FIONBIO || com == FIOASYNC) /* XXX */
634 1.10 cgd return (0); /* XXX */
635 1.10 cgd /* fall into ... */
636 1.10 cgd case VFIFO:
637 1.10 cgd case VCHR:
638 1.10 cgd case VBLK:
639 1.72 fvdl error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
640 1.10 cgd if (error == 0 && com == TIOCSCTTY) {
641 1.13 cgd if (p->p_session->s_ttyvp)
642 1.13 cgd vrele(p->p_session->s_ttyvp);
643 1.10 cgd p->p_session->s_ttyvp = vp;
644 1.10 cgd VREF(vp);
645 1.10 cgd }
646 1.10 cgd return (error);
647 1.63 perseant
648 1.63 perseant default:
649 1.63 perseant return (EPASSTHROUGH);
650 1.10 cgd }
651 1.10 cgd }
652 1.10 cgd
653 1.10 cgd /*
654 1.21 mycroft * File table vnode poll routine.
655 1.10 cgd */
656 1.56 gmcgarry static int
657 1.89 thorpej vn_poll(struct file *fp, int events, struct proc *p)
658 1.10 cgd {
659 1.10 cgd
660 1.72 fvdl return (VOP_POLL(((struct vnode *)fp->f_data), events, p));
661 1.57 jdolecek }
662 1.57 jdolecek
663 1.57 jdolecek /*
664 1.57 jdolecek * File table vnode kqfilter routine.
665 1.57 jdolecek */
666 1.57 jdolecek int
667 1.89 thorpej vn_kqfilter(struct file *fp, struct knote *kn)
668 1.57 jdolecek {
669 1.57 jdolecek
670 1.57 jdolecek return (VOP_KQFILTER((struct vnode *)fp->f_data, kn));
671 1.28 fvdl }
672 1.28 fvdl
673 1.28 fvdl /*
674 1.28 fvdl * Check that the vnode is still valid, and if so
675 1.28 fvdl * acquire requested lock.
676 1.28 fvdl */
677 1.28 fvdl int
678 1.89 thorpej vn_lock(struct vnode *vp, int flags)
679 1.28 fvdl {
680 1.28 fvdl int error;
681 1.34 sommerfe
682 1.84 yamt #if 0
683 1.80 yamt KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0
684 1.80 yamt || (vp->v_flag & VONWORKLST) != 0);
685 1.84 yamt #endif
686 1.80 yamt
687 1.28 fvdl do {
688 1.36 mycroft if ((flags & LK_INTERLOCK) == 0)
689 1.28 fvdl simple_lock(&vp->v_interlock);
690 1.28 fvdl if (vp->v_flag & VXLOCK) {
691 1.49 chs if (flags & LK_NOWAIT) {
692 1.49 chs simple_unlock(&vp->v_interlock);
693 1.49 chs return EBUSY;
694 1.49 chs }
695 1.28 fvdl vp->v_flag |= VXWANT;
696 1.49 chs ltsleep(vp, PINOD | PNORELOCK,
697 1.44 sommerfe "vn_lock", 0, &vp->v_interlock);
698 1.28 fvdl error = ENOENT;
699 1.28 fvdl } else {
700 1.79 yamt error = VOP_LOCK(vp,
701 1.79 yamt (flags & ~LK_RETRY) | LK_INTERLOCK);
702 1.49 chs if (error == 0 || error == EDEADLK || error == EBUSY)
703 1.28 fvdl return (error);
704 1.28 fvdl }
705 1.28 fvdl flags &= ~LK_INTERLOCK;
706 1.28 fvdl } while (flags & LK_RETRY);
707 1.28 fvdl return (error);
708 1.10 cgd }
709 1.10 cgd
710 1.10 cgd /*
711 1.10 cgd * File table vnode close routine.
712 1.10 cgd */
713 1.56 gmcgarry static int
714 1.89 thorpej vn_closefile(struct file *fp, struct proc *p)
715 1.10 cgd {
716 1.10 cgd
717 1.10 cgd return (vn_close(((struct vnode *)fp->f_data), fp->f_flag,
718 1.72 fvdl fp->f_cred, p));
719 1.39 fvdl }
720 1.39 fvdl
721 1.39 fvdl /*
722 1.39 fvdl * Enable LK_CANRECURSE on lock. Return prior status.
723 1.39 fvdl */
724 1.39 fvdl u_int
725 1.89 thorpej vn_setrecurse(struct vnode *vp)
726 1.39 fvdl {
727 1.39 fvdl struct lock *lkp = &vp->v_lock;
728 1.39 fvdl u_int retval = lkp->lk_flags & LK_CANRECURSE;
729 1.39 fvdl
730 1.39 fvdl lkp->lk_flags |= LK_CANRECURSE;
731 1.39 fvdl return retval;
732 1.39 fvdl }
733 1.39 fvdl
734 1.39 fvdl /*
735 1.39 fvdl * Called when done with locksetrecurse.
736 1.39 fvdl */
737 1.39 fvdl void
738 1.89 thorpej vn_restorerecurse(struct vnode *vp, u_int flags)
739 1.39 fvdl {
740 1.39 fvdl struct lock *lkp = &vp->v_lock;
741 1.39 fvdl
742 1.39 fvdl lkp->lk_flags &= ~LK_CANRECURSE;
743 1.39 fvdl lkp->lk_flags |= flags;
744 1.75 hannken }
745 1.77 hannken
746 1.77 hannken int
747 1.77 hannken vn_cow_establish(struct vnode *vp,
748 1.78 hannken int (*func)(void *, struct buf *), void *cookie)
749 1.77 hannken {
750 1.77 hannken int s;
751 1.77 hannken struct spec_cow_entry *e;
752 1.77 hannken
753 1.77 hannken MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry),
754 1.77 hannken M_DEVBUF, M_WAITOK);
755 1.77 hannken e->ce_func = func;
756 1.77 hannken e->ce_cookie = cookie;
757 1.77 hannken
758 1.77 hannken SPEC_COW_LOCK(vp->v_specinfo, s);
759 1.77 hannken vp->v_spec_cow_req++;
760 1.77 hannken while (vp->v_spec_cow_count > 0)
761 1.77 hannken ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0,
762 1.77 hannken &vp->v_spec_cow_slock);
763 1.77 hannken
764 1.77 hannken SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list);
765 1.77 hannken
766 1.77 hannken vp->v_spec_cow_req--;
767 1.77 hannken if (vp->v_spec_cow_req == 0)
768 1.77 hannken wakeup(&vp->v_spec_cow_req);
769 1.77 hannken SPEC_COW_UNLOCK(vp->v_specinfo, s);
770 1.77 hannken
771 1.77 hannken return 0;
772 1.77 hannken }
773 1.77 hannken
774 1.77 hannken int
775 1.77 hannken vn_cow_disestablish(struct vnode *vp,
776 1.78 hannken int (*func)(void *, struct buf *), void *cookie)
777 1.77 hannken {
778 1.77 hannken int s;
779 1.77 hannken struct spec_cow_entry *e;
780 1.77 hannken
781 1.77 hannken SPEC_COW_LOCK(vp->v_specinfo, s);
782 1.77 hannken vp->v_spec_cow_req++;
783 1.77 hannken while (vp->v_spec_cow_count > 0)
784 1.77 hannken ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0,
785 1.77 hannken &vp->v_spec_cow_slock);
786 1.77 hannken
787 1.77 hannken SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list)
788 1.77 hannken if (e->ce_func == func && e->ce_cookie == cookie) {
789 1.77 hannken SLIST_REMOVE(&vp->v_spec_cow_head, e,
790 1.77 hannken spec_cow_entry, ce_list);
791 1.77 hannken FREE(e, M_DEVBUF);
792 1.77 hannken break;
793 1.77 hannken }
794 1.77 hannken
795 1.77 hannken vp->v_spec_cow_req--;
796 1.77 hannken if (vp->v_spec_cow_req == 0)
797 1.77 hannken wakeup(&vp->v_spec_cow_req);
798 1.77 hannken SPEC_COW_UNLOCK(vp->v_specinfo, s);
799 1.77 hannken
800 1.77 hannken return e ? 0 : EINVAL;
801 1.77 hannken }
802 1.85 thorpej
803 1.85 thorpej /*
804 1.85 thorpej * Simplified in-kernel wrapper calls for extended attribute access.
805 1.85 thorpej * Both calls pass in a NULL credential, authorizing a "kernel" access.
806 1.85 thorpej * Set IO_NODELOCKED in ioflg if the vnode is already locked.
807 1.85 thorpej */
808 1.85 thorpej int
809 1.85 thorpej vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace,
810 1.88 christos const char *attrname, size_t *buflen, void *bf, struct proc *p)
811 1.85 thorpej {
812 1.85 thorpej struct uio auio;
813 1.85 thorpej struct iovec aiov;
814 1.85 thorpej int error;
815 1.85 thorpej
816 1.85 thorpej aiov.iov_len = *buflen;
817 1.88 christos aiov.iov_base = bf;
818 1.85 thorpej
819 1.85 thorpej auio.uio_iov = &aiov;
820 1.85 thorpej auio.uio_iovcnt = 1;
821 1.85 thorpej auio.uio_rw = UIO_READ;
822 1.85 thorpej auio.uio_segflg = UIO_SYSSPACE;
823 1.85 thorpej auio.uio_procp = p;
824 1.85 thorpej auio.uio_offset = 0;
825 1.85 thorpej auio.uio_resid = *buflen;
826 1.85 thorpej
827 1.85 thorpej if ((ioflg & IO_NODELOCKED) == 0)
828 1.85 thorpej vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
829 1.86 perry
830 1.85 thorpej error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL,
831 1.85 thorpej p);
832 1.86 perry
833 1.85 thorpej if ((ioflg & IO_NODELOCKED) == 0)
834 1.85 thorpej VOP_UNLOCK(vp, 0);
835 1.86 perry
836 1.85 thorpej if (error == 0)
837 1.85 thorpej *buflen = *buflen - auio.uio_resid;
838 1.86 perry
839 1.85 thorpej return (error);
840 1.85 thorpej }
841 1.85 thorpej
842 1.85 thorpej /*
843 1.85 thorpej * XXX Failure mode if partially written?
844 1.85 thorpej */
845 1.85 thorpej int
846 1.85 thorpej vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace,
847 1.88 christos const char *attrname, size_t buflen, const void *bf, struct proc *p)
848 1.85 thorpej {
849 1.85 thorpej struct uio auio;
850 1.85 thorpej struct iovec aiov;
851 1.85 thorpej struct mount *mp;
852 1.85 thorpej int error;
853 1.85 thorpej
854 1.85 thorpej aiov.iov_len = buflen;
855 1.88 christos aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */
856 1.85 thorpej
857 1.85 thorpej auio.uio_iov = &aiov;
858 1.85 thorpej auio.uio_iovcnt = 1;
859 1.85 thorpej auio.uio_rw = UIO_WRITE;
860 1.85 thorpej auio.uio_segflg = UIO_SYSSPACE;
861 1.85 thorpej auio.uio_procp = p;
862 1.85 thorpej auio.uio_offset = 0;
863 1.85 thorpej auio.uio_resid = buflen;
864 1.85 thorpej
865 1.85 thorpej if ((ioflg & IO_NODELOCKED) == 0) {
866 1.85 thorpej if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
867 1.85 thorpej return (error);
868 1.85 thorpej vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
869 1.85 thorpej }
870 1.85 thorpej
871 1.85 thorpej error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p);
872 1.85 thorpej
873 1.85 thorpej if ((ioflg & IO_NODELOCKED) == 0) {
874 1.85 thorpej vn_finished_write(mp, 0);
875 1.85 thorpej VOP_UNLOCK(vp, 0);
876 1.85 thorpej }
877 1.85 thorpej
878 1.85 thorpej return (error);
879 1.85 thorpej }
880 1.85 thorpej
881 1.85 thorpej int
882 1.85 thorpej vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace,
883 1.85 thorpej const char *attrname, struct proc *p)
884 1.85 thorpej {
885 1.85 thorpej struct mount *mp;
886 1.85 thorpej int error;
887 1.85 thorpej
888 1.85 thorpej if ((ioflg & IO_NODELOCKED) == 0) {
889 1.85 thorpej if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0)
890 1.85 thorpej return (error);
891 1.85 thorpej vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
892 1.85 thorpej }
893 1.85 thorpej
894 1.85 thorpej error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p);
895 1.85 thorpej if (error == EOPNOTSUPP)
896 1.85 thorpej error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL,
897 1.85 thorpej NULL, p);
898 1.86 perry
899 1.85 thorpej if ((ioflg & IO_NODELOCKED) == 0) {
900 1.85 thorpej vn_finished_write(mp, 0);
901 1.85 thorpej VOP_UNLOCK(vp, 0);
902 1.85 thorpej }
903 1.85 thorpej
904 1.85 thorpej return (error);
905 1.85 thorpej }
906