VFS messaging/interfacing work stage 1/99. This stage replaces the old
[dragonfly.git] / sys / vfs / umapfs / umap_vnops.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1992, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software donated to Berkeley by
6 * the UCLA Ficus project.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)umap_vnops.c 8.6 (Berkeley) 5/22/95
37 * $FreeBSD: src/sys/miscfs/umapfs/umap_vnops.c,v 1.30 1999/08/30 07:08:04 bde Exp $
2d3e977e 38 * $DragonFly: src/sys/vfs/umapfs/Attic/umap_vnops.c,v 1.9 2004/08/13 17:51:14 dillon Exp $
984263bc
MD
39 */
40
41/*
42 * Umap Layer
43 */
44
45#include <sys/param.h>
46#include <sys/systm.h>
47#include <sys/kernel.h>
48#include <sys/sysctl.h>
49#include <sys/vnode.h>
50#include <sys/mount.h>
dadab5e9 51#include <sys/proc.h>
984263bc
MD
52#include <sys/namei.h>
53#include <sys/malloc.h>
54#include <sys/buf.h>
1f2de5d4
MD
55#include "umap.h"
56#include <vfs/nullfs/null.h>
984263bc
MD
57
58static int umap_bug_bypass = 0; /* for debugging: enables bypass printf'ing */
59SYSCTL_INT(_debug, OID_AUTO, umapfs_bug_bypass, CTLFLAG_RW,
60 &umap_bug_bypass, 0, "");
61
a6ee311a
RG
62static int umap_bypass (struct vop_generic_args *ap);
63static int umap_getattr (struct vop_getattr_args *ap);
64static int umap_inactive (struct vop_inactive_args *ap);
65static int umap_lock (struct vop_lock_args *ap);
66static int umap_print (struct vop_print_args *ap);
67static int umap_reclaim (struct vop_reclaim_args *ap);
68static int umap_rename (struct vop_rename_args *ap);
69static int umap_unlock (struct vop_unlock_args *ap);
984263bc
MD
70
71/*
72 * This is the 10-Apr-92 bypass routine.
73 * See null_vnops.c:null_bypass for more details.
0af0c3d9
CP
74 *
75 * umap_bypass(struct vnodeop_desc *a_desc, ...)
984263bc
MD
76 */
77static int
0af0c3d9 78umap_bypass(struct vop_generic_args *ap)
984263bc
MD
79{
80 struct ucred **credpp = 0, *credp = 0;
81 struct ucred *savecredp = 0, *savecompcredp = 0;
82 struct ucred *compcredp = 0;
83 struct vnode **this_vp_p;
84 int error;
85 struct vnode *old_vps[VDESC_MAX_VPS];
86 struct vnode *vp1 = 0;
87 struct vnode **vps_p[VDESC_MAX_VPS];
88 struct vnode ***vppp;
89 struct vnodeop_desc *descp = ap->a_desc;
90 int reles, i;
91 struct componentname **compnamepp = 0;
92
93 if (umap_bug_bypass)
94 printf ("umap_bypass: %s\n", descp->vdesc_name);
95
96#ifdef DIAGNOSTIC
97 /*
98 * We require at least one vp.
99 */
100 if (descp->vdesc_vp_offsets == NULL ||
101 descp->vdesc_vp_offsets[0] == VDESC_NO_OFFSET)
102 panic ("umap_bypass: no vp's in map");
103#endif
104
105 /*
106 * Map the vnodes going in.
107 * Later, we'll invoke the operation based on
108 * the first mapped vnode's operation vector.
109 */
110 reles = descp->vdesc_flags;
111 for (i = 0; i < VDESC_MAX_VPS; reles >>= 1, i++) {
112 if (descp->vdesc_vp_offsets[i] == VDESC_NO_OFFSET)
113 break; /* bail out at end of list */
114 vps_p[i] = this_vp_p =
115 VOPARG_OFFSETTO(struct vnode**, descp->vdesc_vp_offsets[i], ap);
116
117 if (i == 0) {
118 vp1 = *vps_p[0];
119 }
120
121 /*
122 * We're not guaranteed that any but the first vnode
123 * are of our type. Check for and don't map any
124 * that aren't. (Must map first vp or vclean fails.)
125 */
126
2d3e977e 127 if (i && (*this_vp_p)->v_vops != umap_vnode_vops) {
984263bc
MD
128 old_vps[i] = NULL;
129 } else {
130 old_vps[i] = *this_vp_p;
131 *(vps_p[i]) = UMAPVPTOLOWERVP(*this_vp_p);
132 if (reles & 1)
597aea93 133 vref(*this_vp_p);
984263bc
MD
134 }
135
136 }
137
138 /*
139 * Fix the credentials. (That's the purpose of this layer.)
140 */
141
142 if (descp->vdesc_cred_offset != VDESC_NO_OFFSET) {
143
144 credpp = VOPARG_OFFSETTO(struct ucred**,
145 descp->vdesc_cred_offset, ap);
146
147 /* Save old values */
148
149 savecredp = (*credpp);
150 if (savecredp != NOCRED)
151 (*credpp) = crdup(savecredp);
152 credp = *credpp;
153
154 if (umap_bug_bypass && credp->cr_uid != 0)
155 printf("umap_bypass: user was %lu, group %lu\n",
156 (u_long)credp->cr_uid, (u_long)credp->cr_gid);
157
158 /* Map all ids in the credential structure. */
159
160 umap_mapids(vp1->v_mount, credp);
161
162 if (umap_bug_bypass && credp->cr_uid != 0)
163 printf("umap_bypass: user now %lu, group %lu\n",
164 (u_long)credp->cr_uid, (u_long)credp->cr_gid);
165 }
166
167 /* BSD often keeps a credential in the componentname structure
168 * for speed. If there is one, it better get mapped, too.
169 */
170
171 if (descp->vdesc_componentname_offset != VDESC_NO_OFFSET) {
172
173 compnamepp = VOPARG_OFFSETTO(struct componentname**,
174 descp->vdesc_componentname_offset, ap);
175
176 compcredp = (*compnamepp)->cn_cred;
177 savecompcredp = compcredp;
178 if (savecompcredp != NOCRED)
179 (*compnamepp)->cn_cred = crdup(savecompcredp);
180 compcredp = (*compnamepp)->cn_cred;
181
182 if (umap_bug_bypass && compcredp->cr_uid != 0)
183 printf(
184 "umap_bypass: component credit user was %lu, group %lu\n",
185 (u_long)compcredp->cr_uid,
186 (u_long)compcredp->cr_gid);
187
188 /* Map all ids in the credential structure. */
189
190 umap_mapids(vp1->v_mount, compcredp);
191
192 if (umap_bug_bypass && compcredp->cr_uid != 0)
193 printf(
194 "umap_bypass: component credit user now %lu, group %lu\n",
195 (u_long)compcredp->cr_uid,
196 (u_long)compcredp->cr_gid);
197 }
198
199 /*
200 * Call the operation on the lower layer
201 * with the modified argument structure.
202 */
203 error = VCALL(*(vps_p[0]), descp->vdesc_offset, ap);
204
205 /*
206 * Maintain the illusion of call-by-value
207 * by restoring vnodes in the argument structure
208 * to their original value.
209 */
210 reles = descp->vdesc_flags;
211 for (i = 0; i < VDESC_MAX_VPS; reles >>= 1, i++) {
212 if (descp->vdesc_vp_offsets[i] == VDESC_NO_OFFSET)
213 break; /* bail out at end of list */
214 if (old_vps[i]) {
215 *(vps_p[i]) = old_vps[i];
216 if (reles & 1)
217 vrele(*(vps_p[i]));
218 };
219 };
220
221 /*
222 * Map the possible out-going vpp
223 * (Assumes that the lower layer always returns
597aea93 224 * a vref'ed vpp unless it gets an error.)
984263bc
MD
225 */
226 if (descp->vdesc_vpp_offset != VDESC_NO_OFFSET &&
227 !(descp->vdesc_flags & VDESC_NOMAP_VPP) &&
228 !error) {
229 if (descp->vdesc_flags & VDESC_VPP_WILLRELE)
230 goto out;
231 vppp = VOPARG_OFFSETTO(struct vnode***,
232 descp->vdesc_vpp_offset, ap);
233 if (*vppp)
234 error = umap_node_create(old_vps[0]->v_mount, **vppp, *vppp);
235 };
236
237 out:
238 /*
239 * Free duplicate cred structure and restore old one.
240 */
241 if (descp->vdesc_cred_offset != VDESC_NO_OFFSET) {
242 if (umap_bug_bypass && credp && credp->cr_uid != 0)
243 printf("umap_bypass: returning-user was %lu\n",
244 (u_long)credp->cr_uid);
245
246 if (savecredp != NOCRED) {
247 crfree(credp);
248 (*credpp) = savecredp;
249 if (umap_bug_bypass && credpp && (*credpp)->cr_uid != 0)
250 printf(
251 "umap_bypass: returning-user now %lu\n\n",
252 (u_long)(*credpp)->cr_uid);
253 }
254 }
255
256 if (descp->vdesc_componentname_offset != VDESC_NO_OFFSET) {
257 if (umap_bug_bypass && compcredp && compcredp->cr_uid != 0)
258 printf(
259 "umap_bypass: returning-component-user was %lu\n",
260 (u_long)compcredp->cr_uid);
261
262 if (savecompcredp != NOCRED) {
263 crfree(compcredp);
264 (*compnamepp)->cn_cred = savecompcredp;
265 if (umap_bug_bypass && credpp && (*credpp)->cr_uid != 0)
266 printf(
267 "umap_bypass: returning-component-user now %lu\n",
268 (u_long)compcredp->cr_uid);
269 }
270 }
271
272 return (error);
273}
274
275
276/*
277 * We handle getattr to change the fsid.
0af0c3d9
CP
278 *
279 * umap_getattr(struct vnode *a_vp, struct vattr *a_vap, struct ucred *a_cred,
280 * struct thread *a_td)
984263bc
MD
281 */
282static int
0af0c3d9 283umap_getattr(struct vop_getattr_args *ap)
984263bc
MD
284{
285 short uid, gid;
286 int error, tmpid, nentries, gnentries;
287 u_long (*mapdata)[2], (*gmapdata)[2];
288 struct vnode **vp1p;
2d3e977e 289 struct vnodeop_desc *descp = ap->a_head.a_desc;
984263bc 290
2d3e977e 291 error = umap_bypass(&ap->a_head);
984263bc
MD
292 if (error)
293 return (error);
294
295 /*
296 * Umap needs to map the uid and gid returned by a stat
297 * into the proper values for this site. This involves
298 * finding the returned uid in the mapping information,
299 * translating it into the uid on the other end,
300 * and filling in the proper field in the vattr
301 * structure pointed to by ap->a_vap. The group
302 * is easier, since currently all groups will be
303 * translate to the NULLGROUP.
304 */
305
306 /* Find entry in map */
307
308 uid = ap->a_vap->va_uid;
309 gid = ap->a_vap->va_gid;
310 if (umap_bug_bypass)
311 printf("umap_getattr: mapped uid = %d, mapped gid = %d\n", uid,
312 gid);
313
314 vp1p = VOPARG_OFFSETTO(struct vnode**, descp->vdesc_vp_offsets[0], ap);
315 nentries = MOUNTTOUMAPMOUNT((*vp1p)->v_mount)->info_nentries;
316 mapdata = (MOUNTTOUMAPMOUNT((*vp1p)->v_mount)->info_mapdata);
317 gnentries = MOUNTTOUMAPMOUNT((*vp1p)->v_mount)->info_gnentries;
318 gmapdata = (MOUNTTOUMAPMOUNT((*vp1p)->v_mount)->info_gmapdata);
319
320 /* Reverse map the uid for the vnode. Since it's a reverse
321 map, we can't use umap_mapids() to do it. */
322
323 tmpid = umap_reverse_findid(uid, mapdata, nentries);
324
325 if (tmpid != -1) {
326
327 ap->a_vap->va_uid = (uid_t) tmpid;
328 if (umap_bug_bypass)
329 printf("umap_getattr: original uid = %d\n", uid);
330 } else
331 ap->a_vap->va_uid = (uid_t) NOBODY;
332
333 /* Reverse map the gid for the vnode. */
334
335 tmpid = umap_reverse_findid(gid, gmapdata, gnentries);
336
337 if (tmpid != -1) {
338
339 ap->a_vap->va_gid = (gid_t) tmpid;
340 if (umap_bug_bypass)
341 printf("umap_getattr: original gid = %d\n", gid);
342 } else
343 ap->a_vap->va_gid = (gid_t) NULLGROUP;
344
345 return (0);
346}
347
348/*
349 * We need to process our own vnode lock and then clear the
350 * interlock flag as it applies only to our vnode, not the
351 * vnodes below us on the stack.
0af0c3d9
CP
352 *
353 * umap_lock(struct vnode *a_vp, lwkt_tokref_t a_vlock, int a_flags,
354 * struct thread *a_td)
984263bc
MD
355 */
356static int
0af0c3d9 357umap_lock(struct vop_lock_args *ap)
984263bc 358{
984263bc
MD
359 vop_nolock(ap);
360 if ((ap->a_flags & LK_TYPE_MASK) == LK_DRAIN)
361 return (0);
362 ap->a_flags &= ~LK_INTERLOCK;
2d3e977e 363 return (null_bypass(&ap->a_head));
984263bc
MD
364}
365
366/*
367 * We need to process our own vnode unlock and then clear the
368 * interlock flag as it applies only to our vnode, not the
369 * vnodes below us on the stack.
0af0c3d9
CP
370 *
371 * umap_unlock(struct vnode *a_vp, int a_flags, struct thread *a_td)
984263bc
MD
372 */
373int
0af0c3d9 374umap_unlock(struct vop_unlock_args *ap)
984263bc
MD
375{
376 vop_nounlock(ap);
377 ap->a_flags &= ~LK_INTERLOCK;
2d3e977e 378 return (null_bypass(&ap->a_head));
984263bc
MD
379}
380
0af0c3d9
CP
381/*
382 * umap_inactive(struct vnode *a_vp, struct thread *a_td)
383 */
984263bc 384static int
0af0c3d9 385umap_inactive(struct vop_inactive_args *ap)
984263bc
MD
386{
387 struct vnode *vp = ap->a_vp;
388 struct umap_node *xp = VTOUMAP(vp);
389 struct vnode *lowervp = xp->umap_lowervp;
390 /*
391 * Do nothing (and _don't_ bypass).
392 * Wait to vrele lowervp until reclaim,
393 * so that until then our umap_node is in the
394 * cache and reusable.
395 *
396 */
dadab5e9 397 VOP_INACTIVE(lowervp, ap->a_td);
41a01a4d 398 VOP_UNLOCK(ap->a_vp, NULL, 0, ap->a_td);
984263bc
MD
399 return (0);
400}
401
0af0c3d9
CP
402/*
403 * umap_reclaim(struct vnode *a_vp)
404 */
984263bc 405static int
0af0c3d9 406umap_reclaim(struct vop_reclaim_args *ap)
984263bc
MD
407{
408 struct vnode *vp = ap->a_vp;
409 struct umap_node *xp = VTOUMAP(vp);
410 struct vnode *lowervp = xp->umap_lowervp;
411
412 /* After this assignment, this node will not be re-used. */
413 xp->umap_lowervp = NULL;
414 LIST_REMOVE(xp, umap_hash);
415 FREE(vp->v_data, M_TEMP);
416 vp->v_data = NULL;
417 vrele(lowervp);
418 return (0);
419}
420
0af0c3d9
CP
421/*
422 * umap_print(struct vop_print_args *ap)
423 */
984263bc 424static int
0af0c3d9 425umap_print(struct vop_print_args *ap)
984263bc
MD
426{
427 struct vnode *vp = ap->a_vp;
428 printf("\ttag VT_UMAPFS, vp=%p, lowervp=%p\n", vp, UMAPVPTOLOWERVP(vp));
429 return (0);
430}
431
0af0c3d9
CP
432/*
433 * umap_rename(struct vnode *a_fdvp, struct vnode *a_fvp,
434 * struct componentname *a_fcnp, struct vnode *a_tdvp,
435 * struct vnode *a_tvp, struct componentname *a_tcnp)
436 */
984263bc 437static int
0af0c3d9 438umap_rename(struct vop_rename_args *ap)
984263bc
MD
439{
440 int error;
441 struct componentname *compnamep;
442 struct ucred *compcredp, *savecompcredp;
443 struct vnode *vp;
444
445 /*
446 * Rename is irregular, having two componentname structures.
447 * We need to map the cre in the second structure,
448 * and then bypass takes care of the rest.
449 */
450
451 vp = ap->a_fdvp;
452 compnamep = ap->a_tcnp;
453 compcredp = compnamep->cn_cred;
454
455 savecompcredp = compcredp;
456 compcredp = compnamep->cn_cred = crdup(savecompcredp);
457
458 if (umap_bug_bypass && compcredp->cr_uid != 0)
459 printf(
460 "umap_rename: rename component credit user was %lu, group %lu\n",
461 (u_long)compcredp->cr_uid, (u_long)compcredp->cr_gid);
462
463 /* Map all ids in the credential structure. */
464
465 umap_mapids(vp->v_mount, compcredp);
466
467 if (umap_bug_bypass && compcredp->cr_uid != 0)
468 printf(
469 "umap_rename: rename component credit user now %lu, group %lu\n",
470 (u_long)compcredp->cr_uid, (u_long)compcredp->cr_gid);
471
2d3e977e 472 error = umap_bypass(&ap->a_head);
984263bc
MD
473
474 /* Restore the additional mapped componentname cred structure. */
475
476 crfree(compcredp);
477 compnamep->cn_cred = savecompcredp;
478
479 return error;
480}
481
482/*
483 * Global vfs data structures
484 */
485/*
486 * XXX - strategy, bwrite are hand coded currently. They should
487 * go away with a merged buffer/block cache.
488 *
489 */
2d3e977e 490struct vop_ops *umap_vnode_vops;
984263bc 491static struct vnodeopv_entry_desc umap_vnodeop_entries[] = {
2d3e977e
MD
492 { &vop_default_desc, (void *) umap_bypass },
493 { &vop_getattr_desc, (void *) umap_getattr },
494 { &vop_inactive_desc, (void *) umap_inactive },
495 { &vop_lock_desc, (void *) umap_lock },
496 { &vop_print_desc, (void *) umap_print },
497 { &vop_reclaim_desc, (void *) umap_reclaim },
498 { &vop_rename_desc, (void *) umap_rename },
499 { &vop_unlock_desc, (void *) umap_unlock },
984263bc
MD
500 { NULL, NULL }
501};
502static struct vnodeopv_desc umap_vnodeop_opv_desc =
2d3e977e 503 { &umap_vnode_vops, umap_vnodeop_entries };
984263bc
MD
504
505VNODEOP_SET(umap_vnodeop_opv_desc);