xref: /netbsd-src/lib/libperfuse/ops.c (revision f89f6560d453f5e37386cc7938c072d2f528b9fa)
1 /*  $NetBSD: ops.c,v 1.83 2015/02/15 20:21:29 manu Exp $ */
2 
3 /*-
4  *  Copyright (c) 2010-2011 Emmanuel Dreyfus. All rights reserved.
5  *
6  *  Redistribution and use in source and binary forms, with or without
7  *  modification, are permitted provided that the following conditions
8  *  are met:
9  *  1. Redistributions of source code must retain the above copyright
10  *     notice, this list of conditions and the following disclaimer.
11  *  2. Redistributions in binary form must reproduce the above copyright
12  *     notice, this list of conditions and the following disclaimer in the
13  *     documentation and/or other materials provided with the distribution.
14  *
15  *  THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
16  *  ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
17  *  TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
18  *  PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
19  *  BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
20  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
21  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
22  *  INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
23  *  CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
24  *  ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
25  *  POSSIBILITY OF SUCH DAMAGE.
26  */
27 
28 #include <stdio.h>
29 #include <unistd.h>
30 #include <stdlib.h>
31 #include <libgen.h>
32 #include <errno.h>
33 #include <err.h>
34 #include <sysexits.h>
35 #include <syslog.h>
36 #include <puffs.h>
37 #include <sys/socket.h>
38 #include <sys/socket.h>
39 #include <sys/extattr.h>
40 #include <sys/time.h>
41 #include <machine/vmparam.h>
42 
43 #include "perfuse_priv.h"
44 #include "fuse.h"
45 
46 extern int perfuse_diagflags;
47 
48 #if 0
49 static void print_node(const char *, puffs_cookie_t);
50 #endif
51 #ifdef PUFFS_KFLAG_CACHE_FS_TTL
52 static void perfuse_newinfo_setttl(struct puffs_newinfo *,
53     struct puffs_node *, struct fuse_entry_out *, struct fuse_attr_out *);
54 #endif /* PUFFS_KFLAG_CACHE_FS_TTL */
55 static int xchg_msg(struct puffs_usermount *, puffs_cookie_t,
56     perfuse_msg_t *, size_t, enum perfuse_xchg_pb_reply);
57 static int mode_access(puffs_cookie_t, const struct puffs_cred *, mode_t);
58 static int sticky_access(puffs_cookie_t, struct puffs_node *,
59     const struct puffs_cred *);
60 static void fuse_attr_to_vap(struct perfuse_state *,
61     struct vattr *, struct fuse_attr *);
62 static int node_lookup_common(struct puffs_usermount *, puffs_cookie_t,
63     struct puffs_newinfo *, const char *, const struct puffs_cred *,
64     struct puffs_node **);
65 static int node_mk_common(struct puffs_usermount *, puffs_cookie_t,
66     struct puffs_newinfo *, const struct puffs_cn *pcn, perfuse_msg_t *);
67 static uint64_t readdir_last_cookie(struct fuse_dirent *, size_t);
68 static ssize_t fuse_to_dirent(struct puffs_usermount *, puffs_cookie_t,
69     struct fuse_dirent *, size_t);
70 static void readdir_buffered(puffs_cookie_t, struct dirent *, off_t *,
71     size_t *);
72 static void node_ref(puffs_cookie_t);
73 static void node_rele(puffs_cookie_t);
74 static void requeue_request(struct puffs_usermount *,
75     puffs_cookie_t opc, enum perfuse_qtype);
76 static int dequeue_requests(puffs_cookie_t opc, enum perfuse_qtype, int);
77 #define DEQUEUE_ALL 0
78 
79 /*
80  *  From <sys/vnode>, inside #ifdef _KERNEL section
81  */
82 #define IO_SYNC		(0x40|IO_DSYNC)
83 #define IO_DSYNC	0x00200
84 #define IO_DIRECT	0x02000
85 
86 /*
87  *  From <fcntl>, inside #ifdef _KERNEL section
88  */
89 #define F_WAIT		0x010
90 #define F_FLOCK		0x020
91 #define OFLAGS(fflags)  ((fflags) - 1)
92 
93 /*
94  * Borrowed from src/sys/kern/vfs_subr.c and src/sys/sys/vnode.h
95  */
96 const enum vtype iftovt_tab[16] = {
97 	VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
98         VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD,
99 };
100 const int vttoif_tab[9] = {
101 	0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK,
102         S_IFSOCK, S_IFIFO, S_IFMT,
103 };
104 
105 #define IFTOVT(mode) (iftovt_tab[((mode) & S_IFMT) >> 12])
106 #define VTTOIF(indx) (vttoif_tab[(int)(indx)])
107 
108 #if 0
109 static void
110 print_node(const char *func, puffs_cookie_t opc)
111 {
112 	struct puffs_node *pn;
113 	struct perfuse_node_data *pnd;
114 	struct vattr *vap;
115 
116 	pn = (struct puffs_node *)opc;
117 	pnd = PERFUSE_NODE_DATA(opc);
118 	vap = &pn->pn_va;
119 
120 	printf("%s: \"%s\", opc = %p, nodeid = 0x%"PRIx64" ino = %"PRIu64"\n",
121 	       func, pnd->pnd_name, opc, pnd->pnd_nodeid, vap->va_fileid);
122 
123 	return;
124 }
125 #endif /* PERFUSE_DEBUG */
126 
127 int
128 perfuse_node_close_common(struct puffs_usermount *pu, puffs_cookie_t opc,
129 	int mode)
130 {
131 	struct perfuse_state *ps;
132 	perfuse_msg_t *pm;
133 	int op;
134 	uint64_t fh;
135 	struct fuse_release_in *fri;
136 	struct perfuse_node_data *pnd;
137 	struct puffs_node *pn;
138 	int error;
139 
140 	ps = puffs_getspecific(pu);
141 	pn = (struct puffs_node *)opc;
142 	pnd = PERFUSE_NODE_DATA(pn);
143 
144 	if (puffs_pn_getvap(pn)->va_type == VDIR) {
145 		op = FUSE_RELEASEDIR;
146 		mode = FREAD;
147 	} else {
148 		op = FUSE_RELEASE;
149 	}
150 
151 	/*
152 	 * Destroy the filehandle before sending the
153 	 * request to the FUSE filesystem, otherwise
154 	 * we may get a second close() while we wait
155 	 * for the reply, and we would end up closing
156 	 * the same fh twice instead of closng both.
157 	 */
158 	fh = perfuse_get_fh(opc, mode);
159 	perfuse_destroy_fh(pn, fh);
160 
161 	/*
162 	 * release_flags may be set to FUSE_RELEASE_FLUSH
163 	 * to flush locks. lock_owner must be set in that case
164 	 *
165 	 * ps_new_msg() is called with NULL creds, which will
166 	 * be interpreted as FUSE superuser. We come here from the
167 	 * inactive method, which provides no creds, but obviously
168 	 * runs with kernel privilege.
169 	 */
170 	pm = ps->ps_new_msg(pu, opc, op, sizeof(*fri), NULL);
171 	fri = GET_INPAYLOAD(ps, pm, fuse_release_in);
172 	fri->fh = fh;
173 	fri->flags = 0;
174 	fri->release_flags = 0;
175 	fri->lock_owner = pnd->pnd_lock_owner;
176 	fri->flags = (fri->lock_owner != 0) ? FUSE_RELEASE_FLUSH : 0;
177 
178 #ifdef PERFUSE_DEBUG
179 	if (perfuse_diagflags & PDF_FH)
180 		DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", fh = 0x%"PRIx64"\n",
181 			 __func__, (void *)opc, pnd->pnd_nodeid, fri->fh);
182 #endif
183 
184 	if ((error = xchg_msg(pu, opc, pm,
185 			      NO_PAYLOAD_REPLY_LEN, wait_reply)) != 0)
186 		DERRX(EX_SOFTWARE, "%s: freed fh = 0x%"PRIx64" but filesystem "
187 		      "returned error = %d", __func__, fh, error);
188 
189 	ps->ps_destroy_msg(pm);
190 
191 	return 0;
192 }
193 
194 static int
195 xchg_msg(struct puffs_usermount *pu, puffs_cookie_t opc, perfuse_msg_t *pm,
196 	size_t len, enum perfuse_xchg_pb_reply wait)
197 {
198 	struct perfuse_state *ps;
199 	struct perfuse_node_data *pnd;
200 	struct perfuse_trace *pt = NULL;
201 	int error;
202 
203 	ps = puffs_getspecific(pu);
204 	pnd = NULL;
205 	if ((struct puffs_node *)opc != NULL)
206 		pnd = PERFUSE_NODE_DATA(opc);
207 
208 #ifdef PERFUSE_DEBUG
209 	if ((perfuse_diagflags & PDF_FILENAME) && (opc != 0))
210 		DPRINTF("file = \"%s\", ino = %"PRIu64" flags = 0x%x\n",
211 			perfuse_node_path(ps, opc),
212 			((struct puffs_node *)opc)->pn_va.va_fileid,
213 			PERFUSE_NODE_DATA(opc)->pnd_flags);
214 #endif
215 	ps->ps_xchgcount++;
216 	if (pnd)
217 		pnd->pnd_inxchg++;
218 
219 	/*
220 	 * Record FUSE call start if requested
221 	 */
222 	if (perfuse_diagflags & PDF_TRACE)
223 		pt = perfuse_trace_begin(ps, opc, pm);
224 
225 	/*
226 	 * Do actual FUSE exchange
227 	 */
228 	if ((error = ps->ps_xchg_msg(pu, pm, len, wait)) != 0)
229 		ps->ps_destroy_msg(pm);
230 
231 	/*
232 	 * Record FUSE call end if requested
233 	 */
234 	if (pt != NULL)
235 		perfuse_trace_end(ps, pt, error);
236 
237 	ps->ps_xchgcount--;
238 	if (pnd) {
239 		pnd->pnd_inxchg--;
240 		(void)dequeue_requests(opc, PCQ_AFTERXCHG, DEQUEUE_ALL);
241 	}
242 
243 	return error;
244 }
245 
246 static int
247 mode_access(puffs_cookie_t opc, const struct puffs_cred *pcr, mode_t mode)
248 {
249 	struct puffs_node *pn;
250 	struct vattr *va;
251 
252 	/*
253 	 * pcr is NULL for self open through fsync or readdir.
254 	 * In both case, access control is useless, as it was
255 	 * done before, at open time.
256 	 */
257 	if (pcr == NULL)
258 		return 0;
259 
260 	pn = (struct puffs_node *)opc;
261 	va = puffs_pn_getvap(pn);
262 	return puffs_access(va->va_type, va->va_mode,
263 			    va->va_uid, va->va_gid,
264 			    mode, pcr);
265 }
266 
267 static int
268 sticky_access(puffs_cookie_t opc, struct puffs_node *targ,
269 	      const struct puffs_cred *pcr)
270 {
271 	uid_t uid;
272 	int sticky, owner, parent_owner;
273 
274 	/*
275 	 * This covers the case where the kernel requests a DELETE
276 	 * or RENAME on its own, and where puffs_cred_getuid would
277 	 * return -1. While such a situation should not happen,
278 	 * we allow it here.
279 	 *
280 	 * This also allows root to tamper with other users' files
281 	 * that have the sticky bit.
282 	 */
283 	if (puffs_cred_isjuggernaut(pcr))
284 		return 0;
285 
286 	if (puffs_cred_getuid(pcr, &uid) != 0)
287 		DERRX(EX_SOFTWARE, "puffs_cred_getuid fails in %s", __func__);
288 
289 	sticky = puffs_pn_getvap(opc)->va_mode & S_ISTXT;
290 	owner = puffs_pn_getvap(targ)->va_uid == uid;
291 	parent_owner = puffs_pn_getvap(opc)->va_uid == uid;
292 
293 	if (sticky && !owner && !parent_owner)
294 		return EPERM;
295 
296 	return 0;
297 }
298 
299 
300 static void
301 fuse_attr_to_vap(struct perfuse_state *ps, struct vattr *vap,
302 	struct fuse_attr *fa)
303 {
304 	vap->va_type = IFTOVT(fa->mode);
305 	vap->va_mode = fa->mode & ALLPERMS;
306 	vap->va_nlink = fa->nlink;
307 	vap->va_uid = fa->uid;
308 	vap->va_gid = fa->gid;
309 	vap->va_fsid = (long)ps->ps_fsid;
310 	vap->va_fileid = fa->ino;
311 	vap->va_size = fa->size;
312 	vap->va_blocksize = fa->blksize;
313 	vap->va_atime.tv_sec = (time_t)fa->atime;
314 	vap->va_atime.tv_nsec = (long) fa->atimensec;
315 	vap->va_mtime.tv_sec = (time_t)fa->mtime;
316 	vap->va_mtime.tv_nsec = (long)fa->mtimensec;
317 	vap->va_ctime.tv_sec = (time_t)fa->ctime;
318 	vap->va_ctime.tv_nsec = (long)fa->ctimensec;
319 	vap->va_birthtime.tv_sec = 0;
320 	vap->va_birthtime.tv_nsec = 0;
321 	vap->va_gen = 0;
322 	vap->va_flags = 0;
323 	vap->va_rdev = fa->rdev;
324 	vap->va_bytes = fa->blocks * S_BLKSIZE;
325 	vap->va_filerev = (u_quad_t)PUFFS_VNOVAL;
326 	vap->va_vaflags = 0;
327 
328 	if (vap->va_blocksize == 0)
329 		vap->va_blocksize = DEV_BSIZE;
330 
331 	if (vap->va_size == (size_t)PUFFS_VNOVAL) /* XXX */
332 		vap->va_size = 0;
333 
334 	return;
335 }
336 
337 #ifdef PUFFS_KFLAG_CACHE_FS_TTL
338 static void
339 perfuse_newinfo_setttl(struct puffs_newinfo *pni,
340     struct puffs_node *pn, struct fuse_entry_out *feo,
341     struct fuse_attr_out *fao)
342 {
343 #ifdef PERFUSE_DEBUG
344 	if ((feo == NULL) && (fao == NULL))
345 		DERRX(EX_SOFTWARE, "%s: feo and fao NULL", __func__);
346 
347 	if ((feo != NULL) && (fao != NULL))
348 		DERRX(EX_SOFTWARE, "%s: feo and fao != NULL", __func__);
349 #endif /* PERFUSE_DEBUG */
350 
351 	if (fao != NULL) {
352 		struct timespec va_ttl;
353 
354 		va_ttl.tv_sec = fao->attr_valid;
355 		va_ttl.tv_nsec = fao->attr_valid_nsec;
356 
357 		puffs_newinfo_setvattl(pni, &va_ttl);
358 	}
359 
360 	if (feo != NULL) {
361 		struct timespec va_ttl;
362 		struct timespec cn_ttl;
363 		struct timespec now;
364 		struct perfuse_node_data *pnd = PERFUSE_NODE_DATA(pn);
365 
366 		va_ttl.tv_sec = feo->attr_valid;
367 		va_ttl.tv_nsec = feo->attr_valid_nsec;
368 		cn_ttl.tv_sec = feo->entry_valid;
369 		cn_ttl.tv_nsec = feo->entry_valid_nsec;
370 
371 		puffs_newinfo_setvattl(pni, &va_ttl);
372 		puffs_newinfo_setcnttl(pni, &cn_ttl);
373 
374 		if (clock_gettime(CLOCK_REALTIME, &now) != 0)
375 			DERR(EX_OSERR, "clock_gettime failed");
376 
377                 timespecadd(&now, &cn_ttl, &pnd->pnd_cn_expire);
378 	}
379 
380 	return;
381 }
382 #endif /* PUFFS_KFLAG_CACHE_FS_TTL */
383 
384 static int
385 node_lookup_common(struct puffs_usermount *pu, puffs_cookie_t opc,
386 	struct puffs_newinfo *pni, const char *path,
387 	const struct puffs_cred *pcr, struct puffs_node **pnp)
388 {
389 	struct perfuse_state *ps;
390 	struct perfuse_node_data *oldpnd;
391 	perfuse_msg_t *pm;
392 	struct fuse_entry_out *feo;
393 	struct puffs_node *pn;
394 	size_t len;
395 	int error;
396 
397 	/*
398 	 * Prevent further lookups if the parent was removed
399 	 */
400 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
401 		return ESTALE;
402 
403 	if (pnp == NULL)
404 		DERRX(EX_SOFTWARE, "pnp must be != NULL");
405 
406 	ps = puffs_getspecific(pu);
407 
408 #ifdef PERFUSE_DEBUG
409 	if (perfuse_diagflags & PDF_FILENAME)
410 		DPRINTF("%s: opc = %p, file = \"%s\" looking up \"%s\"\n",
411 			__func__, (void *)opc,
412 			perfuse_node_path(ps, opc), path);
413 
414 	if (strcmp(path, ".") == 0)
415 		DERRX(EX_SOFTWARE, "unexpected dot-lookup");
416 
417 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_RECLAIMED)
418 		DERRX(EX_SOFTWARE,
419 		      "looking up reclaimed node opc = %p, name = \"%s\"",
420 		      opc, path);
421 
422 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_INVALID)
423 		DERRX(EX_SOFTWARE,
424 		      "looking up freed node opc = %p, name = \"%s\"",
425 		      opc, path);
426 #endif /* PERFUSE_DEBUG */
427 
428 	len = strlen(path) + 1;
429 	pm = ps->ps_new_msg(pu, opc, FUSE_LOOKUP, len, pcr);
430 	(void)strlcpy(_GET_INPAYLOAD(ps, pm, char *), path, len);
431 
432 	if ((error = xchg_msg(pu, opc, pm, sizeof(*feo), wait_reply)) != 0)
433 		return error;
434 
435 	feo = GET_OUTPAYLOAD(ps, pm, fuse_entry_out);
436 
437 	/*
438 	 * Starting with ABI 7.4, inode number 0 means ENOENT,
439 	 * with entry_valid / entry_valid_nsec giving negative
440 	 * cache timeout (which we do not implement yet).
441 	 */
442 	if (feo->attr.ino == 0) {
443 		ps->ps_destroy_msg(pm);
444 		return ENOENT;
445 	}
446 
447 	/*
448 	 * Check for a known node, not reclaimed, with another name.
449 	 * It may have been moved, or we can lookup ../
450 	 */
451 	if (((oldpnd = perfuse_node_bynodeid(ps, feo->nodeid)) != NULL) &&
452 	    !(oldpnd->pnd_flags & PND_RECLAIMED)) {
453 		/*
454 		 * Save the new node name if not ..
455 		 */
456 		if (strncmp(path, "..", len) != 0)
457 			(void)strlcpy(oldpnd->pnd_name,
458 				      path, MAXPATHLEN);
459 		pn = oldpnd->pnd_pn;
460 
461 	} else {
462 		pn = perfuse_new_pn(pu, path, opc);
463 		PERFUSE_NODE_DATA(pn)->pnd_nodeid = feo->nodeid;
464 		perfuse_node_cache(ps, pn);
465 	}
466 
467 #ifdef PERFUSE_DEBUG
468 	if (PERFUSE_NODE_DATA(pn)->pnd_flags & PND_RECLAIMED)
469 		DERRX(EX_SOFTWARE,
470 		      "reclaimed in lookup opc = %p, name = \"%s\", ck = %p",
471 		      opc, path, pn);
472 
473 	if (PERFUSE_NODE_DATA(pn)->pnd_flags & PND_INVALID)
474 		DERRX(EX_SOFTWARE,
475 		      "freed in lookup opc = %p, name = \"%s\", ck = %p",
476 		      opc, path, pn);
477 #endif /* PERFUSE_DEBUG */
478 
479 	fuse_attr_to_vap(ps, &pn->pn_va, &feo->attr);
480 	pn->pn_va.va_gen = (u_long)(feo->generation);
481 	PERFUSE_NODE_DATA(pn)->pnd_fuse_nlookup++;
482 
483 	*pnp = pn;
484 
485 #ifdef PERFUSE_DEBUG
486 	if (perfuse_diagflags & PDF_FILENAME)
487 		DPRINTF("%s: opc = %p, looked up opc = %p, "
488 			"nodeid = 0x%"PRIx64" file = \"%s\"\n", __func__,
489 			(void *)opc, pn, feo->nodeid, path);
490 #endif
491 
492 	if (pni != NULL) {
493 #ifdef PUFFS_KFLAG_CACHE_FS_TTL
494 		puffs_newinfo_setva(pni, &pn->pn_va);
495 		perfuse_newinfo_setttl(pni, pn, feo, NULL);
496 #endif /* PUFFS_KFLAG_CACHE_FS_TTL */
497 		puffs_newinfo_setcookie(pni, pn);
498 		puffs_newinfo_setvtype(pni, pn->pn_va.va_type);
499 		puffs_newinfo_setsize(pni, (voff_t)pn->pn_va.va_size);
500 		puffs_newinfo_setrdev(pni, pn->pn_va.va_rdev);
501 	}
502 
503 	if (PERFUSE_NODE_DATA(pn)->pnd_flags & PND_NODELEAK) {
504 		PERFUSE_NODE_DATA(pn)->pnd_flags &= ~PND_NODELEAK;
505 		ps->ps_nodeleakcount--;
506 	}
507 
508 	ps->ps_destroy_msg(pm);
509 
510 	return 0;
511 }
512 
513 
514 /*
515  * Common code for methods that create objects:
516  * perfuse_node_mkdir
517  * perfuse_node_mknod
518  * perfuse_node_symlink
519  */
520 static int
521 node_mk_common(struct puffs_usermount *pu, puffs_cookie_t opc,
522 	struct puffs_newinfo *pni, const struct puffs_cn *pcn,
523 	perfuse_msg_t *pm)
524 {
525 	struct perfuse_state *ps;
526 	struct puffs_node *pn;
527 	struct fuse_entry_out *feo;
528 	int error;
529 
530 	ps =  puffs_getspecific(pu);
531 
532 	if ((error = xchg_msg(pu, opc, pm, sizeof(*feo), wait_reply)) != 0)
533 		return error;
534 
535 	feo = GET_OUTPAYLOAD(ps, pm, fuse_entry_out);
536 	if (feo->nodeid == PERFUSE_UNKNOWN_NODEID)
537 		DERRX(EX_SOFTWARE, "%s: no nodeid", __func__);
538 
539 	pn = perfuse_new_pn(pu, pcn->pcn_name, opc);
540 	PERFUSE_NODE_DATA(pn)->pnd_nodeid = feo->nodeid;
541 	PERFUSE_NODE_DATA(pn)->pnd_puffs_nlookup++;
542 	perfuse_node_cache(ps, pn);
543 
544 	fuse_attr_to_vap(ps, &pn->pn_va, &feo->attr);
545 	pn->pn_va.va_gen = (u_long)(feo->generation);
546 
547 	puffs_newinfo_setcookie(pni, pn);
548 #ifdef PUFFS_KFLAG_CACHE_FS_TTL
549 	puffs_newinfo_setva(pni, &pn->pn_va);
550 	perfuse_newinfo_setttl(pni, pn, feo, NULL);
551 #endif /* PUFFS_KFLAG_CACHE_FS_TTL */
552 
553 
554 #ifdef PERFUSE_DEBUG
555 	if (perfuse_diagflags & PDF_FILENAME)
556 		DPRINTF("%s: opc = %p, file = \"%s\", flags = 0x%x "
557 			"nodeid = 0x%"PRIx64"\n",
558 			__func__, (void *)pn, pcn->pcn_name,
559 			PERFUSE_NODE_DATA(pn)->pnd_flags, feo->nodeid);
560 #endif
561 	ps->ps_destroy_msg(pm);
562 
563 	/* Parents is now dirty */
564 	PERFUSE_NODE_DATA(opc)->pnd_flags |= PND_DIRTY;
565 
566 	return 0;
567 }
568 
569 static uint64_t
570 readdir_last_cookie(struct fuse_dirent *fd, size_t fd_len)
571 {
572 	size_t len;
573 	size_t seen = 0;
574 	char *ndp;
575 
576 	do {
577 		len = FUSE_DIRENT_ALIGN(sizeof(*fd) + fd->namelen);
578 		seen += len;
579 
580 		if (seen >= fd_len)
581 			break;
582 
583 		ndp = (char *)(void *)fd + (size_t)len;
584 		fd = (struct fuse_dirent *)(void *)ndp;
585 	} while (1 /* CONSTCOND */);
586 
587 	return fd->off;
588 }
589 
590 static ssize_t
591 fuse_to_dirent(struct puffs_usermount *pu, puffs_cookie_t opc,
592 	struct fuse_dirent *fd, size_t fd_len)
593 {
594 	struct dirent *dents;
595 	size_t dents_len;
596 	ssize_t written;
597 	uint64_t fd_offset;
598 	struct fuse_dirent *fd_base;
599 	size_t len;
600 
601 	fd_base = fd;
602 	fd_offset = 0;
603 	written = 0;
604 	dents = PERFUSE_NODE_DATA(opc)->pnd_dirent;
605 	dents_len = (size_t)PERFUSE_NODE_DATA(opc)->pnd_dirent_len;
606 
607 	do {
608 		char *ndp;
609 		size_t reclen;
610 		char name[MAXPATHLEN];
611 
612 		reclen = _DIRENT_RECLEN(dents, fd->namelen);
613 
614 		/*
615 		 * Check we do not overflow the output buffer
616 		 * struct fuse_dirent is bigger than struct dirent,
617 		 * so we should always use fd_len and never reallocate
618 		 * later.
619 		 * If we have to reallocate,try to double the buffer
620 		 * each time so that we do not have to do it too often.
621 		 */
622 		if (written + reclen > dents_len) {
623 			if (dents_len == 0)
624 				dents_len = fd_len;
625 			else
626 				dents_len =
627 				   MAX(2 * dents_len, written + reclen);
628 
629 			dents = PERFUSE_NODE_DATA(opc)->pnd_dirent;
630 			if ((dents = realloc(dents, dents_len)) == NULL)
631 				DERR(EX_OSERR, "%s: malloc failed", __func__);
632 
633 			PERFUSE_NODE_DATA(opc)->pnd_dirent = dents;
634 			PERFUSE_NODE_DATA(opc)->pnd_dirent_len = dents_len;
635 
636 			/*
637 			 * (void *) for delint
638 			 */
639 			ndp = (char *)(void *)dents + written;
640 			dents = (struct dirent *)(void *)ndp;
641 		}
642 
643 		strncpy(name, fd->name, fd->namelen);
644 		name[fd->namelen] = '\0';
645 
646 		/*
647 		 * Filesystem was mounted without -o use_ino
648 		 * Perform a lookup to find it.
649 		 */
650 		if (fd->ino == PERFUSE_UNKNOWN_INO) {
651 			struct puffs_node *pn;
652 			struct perfuse_node_data *pnd = PERFUSE_NODE_DATA(opc);
653 
654 			/*
655 			 * Avoid breaking out of fs
656 			 * by lookup to .. on root
657 			 */
658 			if ((strcmp(name, "..") == 0) &&
659 			    (pnd->pnd_nodeid == FUSE_ROOT_ID)) {
660 				fd->ino = FUSE_ROOT_ID;
661 			} else {
662 				int error;
663 
664 				error = node_lookup_common(pu, opc, NULL,
665 							   name, NULL, &pn);
666 				if (error != 0) {
667 					DWARNX("node_lookup_common %s "
668 					       "failed: %d", name, error);
669 				} else {
670 					fd->ino = pn->pn_va.va_fileid;
671 					(void)perfuse_node_reclaim(pu, pn);
672 				}
673 			}
674 		}
675 
676 		dents->d_fileno = fd->ino;
677 		dents->d_reclen = (unsigned short)reclen;
678 		dents->d_namlen = fd->namelen;
679 		dents->d_type = fd->type;
680 		strlcpy(dents->d_name, name, fd->namelen + 1);
681 
682 #ifdef PERFUSE_DEBUG
683 		if (perfuse_diagflags & PDF_READDIR)
684 			DPRINTF("%s: translated \"%s\" ino = %"PRIu64"\n",
685 				__func__, dents->d_name, dents->d_fileno);
686 #endif
687 
688 		dents = _DIRENT_NEXT(dents);
689 		written += reclen;
690 
691 		/*
692 		 * Move to the next record.
693 		 * fd->off is not the offset, it is an opaque cookie
694 		 * given by the filesystem to keep state across multiple
695 		 * readdir() operation.
696 		 * Use record alignement instead.
697 		 */
698 		len = FUSE_DIRENT_ALIGN(sizeof(*fd) + fd->namelen);
699 #ifdef PERFUSE_DEBUG
700 		if (perfuse_diagflags & PDF_READDIR)
701 			DPRINTF("%s: record at %"PRId64"/0x%"PRIx64" "
702 				"length = %zd/0x%zx. "
703 				"next record at %"PRId64"/0x%"PRIx64" "
704 				"max %zd/0x%zx\n",
705 				__func__, fd_offset, fd_offset, len, len,
706 				fd_offset + len, fd_offset + len,
707 				fd_len, fd_len);
708 #endif
709 		fd_offset += len;
710 
711 		/*
712 		 * Check if next record is still within the packet
713 		 * If it is not, we reached the end of the buffer.
714 		 */
715 		if (fd_offset >= fd_len)
716 			break;
717 
718 		/*
719 		 * (void *) for delint
720 		 */
721 		ndp = (char *)(void *)fd_base + (size_t)fd_offset;
722 		fd = (struct fuse_dirent *)(void *)ndp;
723 
724 	} while (1 /* CONSTCOND */);
725 
726 	/*
727 	 * Adjust the dirent output length
728 	 */
729 	if (written != -1)
730 		PERFUSE_NODE_DATA(opc)->pnd_dirent_len = written;
731 
732 	return written;
733 }
734 
735 static void
736 readdir_buffered(puffs_cookie_t opc, struct dirent *dent, off_t *readoff,
737 	size_t *reslen)
738 {
739 	struct dirent *fromdent;
740 	struct perfuse_node_data *pnd;
741 	char *ndp;
742 
743 	pnd = PERFUSE_NODE_DATA(opc);
744 
745 	while (*readoff < pnd->pnd_dirent_len) {
746 		/*
747 		 * (void *) for delint
748 		 */
749 		ndp = (char *)(void *)pnd->pnd_dirent + (size_t)*readoff;
750 		fromdent = (struct dirent *)(void *)ndp;
751 
752 		if (*reslen < _DIRENT_SIZE(fromdent))
753 			break;
754 
755 		memcpy(dent, fromdent, _DIRENT_SIZE(fromdent));
756 		*readoff += _DIRENT_SIZE(fromdent);
757 		*reslen -= _DIRENT_SIZE(fromdent);
758 
759 		dent = _DIRENT_NEXT(dent);
760 	}
761 
762 #ifdef PERFUSE_DEBUG
763 	if (perfuse_diagflags & PDF_READDIR)
764 		DPRINTF("%s: readoff = %"PRId64",  "
765 			"pnd->pnd_dirent_len = %"PRId64"\n",
766 			__func__, *readoff, pnd->pnd_dirent_len);
767 #endif
768 	if (*readoff >=  pnd->pnd_dirent_len) {
769 		free(pnd->pnd_dirent);
770 		pnd->pnd_dirent = NULL;
771 		pnd->pnd_dirent_len = 0;
772 	}
773 
774 	return;
775 }
776 
777 
778 static void
779 node_ref(puffs_cookie_t opc)
780 {
781 	struct perfuse_node_data *pnd = PERFUSE_NODE_DATA(opc);
782 
783 #ifdef PERFUSE_DEBUG
784 	if (pnd->pnd_flags & PND_INVALID)
785 		DERRX(EX_SOFTWARE, "Use of freed node opc = %p", opc);
786 #endif /* PERFUSE_DEBUG */
787 
788 	pnd->pnd_ref++;
789 	return;
790 }
791 
792 static void
793 node_rele(puffs_cookie_t opc)
794 {
795 	struct perfuse_node_data *pnd = PERFUSE_NODE_DATA(opc);
796 
797 #ifdef PERFUSE_DEBUG
798 	if (pnd->pnd_flags & PND_INVALID)
799 		DERRX(EX_SOFTWARE, "Use of freed node opc = %p", opc);
800 #endif /* PERFUSE_DEBUG */
801 
802 	pnd->pnd_ref--;
803 
804 	if (pnd->pnd_ref == 0)
805 		(void)dequeue_requests(opc, PCQ_REF, DEQUEUE_ALL);
806 
807 	return;
808 }
809 
810 static void
811 requeue_request(struct puffs_usermount *pu, puffs_cookie_t opc,
812 	enum perfuse_qtype type)
813 {
814 	struct perfuse_cc_queue pcq;
815 	struct perfuse_node_data *pnd;
816 
817 	pnd = PERFUSE_NODE_DATA(opc);
818 	pcq.pcq_type = type;
819 	pcq.pcq_cc = puffs_cc_getcc(pu);
820 	TAILQ_INSERT_TAIL(&pnd->pnd_pcq, &pcq, pcq_next);
821 
822 #ifdef PERFUSE_DEBUG
823 	if (perfuse_diagflags & PDF_REQUEUE)
824 		DPRINTF("%s: REQUEUE opc = %p, pcc = %p (%s)\n",
825 		        __func__, (void *)opc, pcq.pcq_cc,
826 			perfuse_qtypestr[type]);
827 #endif
828 
829 	puffs_cc_yield(pcq.pcq_cc);
830 	TAILQ_REMOVE(&pnd->pnd_pcq, &pcq, pcq_next);
831 
832 #ifdef PERFUSE_DEBUG
833 	if (perfuse_diagflags & PDF_REQUEUE)
834 		DPRINTF("%s: RESUME opc = %p, pcc = %p (%s)\n",
835 		        __func__, (void *)opc, pcq.pcq_cc,
836 			perfuse_qtypestr[type]);
837 #endif
838 
839 	return;
840 }
841 
842 static int
843 dequeue_requests(puffs_cookie_t opc, enum perfuse_qtype type, int max)
844 {
845 	struct perfuse_cc_queue *pcq;
846 	struct perfuse_node_data *pnd;
847 	int dequeued;
848 
849 	pnd = PERFUSE_NODE_DATA(opc);
850 	dequeued = 0;
851 	TAILQ_FOREACH(pcq, &pnd->pnd_pcq, pcq_next) {
852 		if (pcq->pcq_type != type)
853 			continue;
854 
855 #ifdef PERFUSE_DEBUG
856 		if (perfuse_diagflags & PDF_REQUEUE)
857 			DPRINTF("%s: SCHEDULE opc = %p, pcc = %p (%s)\n",
858 				__func__, (void *)opc, pcq->pcq_cc,
859 				 perfuse_qtypestr[type]);
860 #endif
861 		puffs_cc_schedule(pcq->pcq_cc);
862 
863 		if (++dequeued == max)
864 			break;
865 	}
866 
867 #ifdef PERFUSE_DEBUG
868 	if (perfuse_diagflags & PDF_REQUEUE)
869 		DPRINTF("%s: DONE  opc = %p\n", __func__, (void *)opc);
870 #endif
871 
872 	return dequeued;
873 }
874 
875 void
876 perfuse_fs_init(struct puffs_usermount *pu)
877 {
878 	struct perfuse_state *ps;
879 	perfuse_msg_t *pm;
880 	struct fuse_init_in *fii;
881 	struct fuse_init_out *fio;
882 	int error;
883 
884 	ps = puffs_getspecific(pu);
885 
886         if (puffs_mount(pu, ps->ps_target, ps->ps_mountflags, ps->ps_root) != 0)
887                 DERR(EX_OSERR, "%s: puffs_mount failed", __func__);
888 
889 	/*
890 	 * Linux 2.6.34.1 sends theses flags:
891 	 * FUSE_ASYNC_READ | FUSE_POSIX_LOCKS | FUSE_ATOMIC_O_TRUNC
892 	 * FUSE_EXPORT_SUPPORT | FUSE_BIG_WRITES | FUSE_DONT_MASK
893 	 *
894 	 * Linux also sets max_readahead at 32 pages (128 kB)
895 	 *
896 	 * ps_new_msg() is called with NULL creds, which will
897 	 * be interpreted as FUSE superuser.
898 	 */
899 	pm = ps->ps_new_msg(pu, 0, FUSE_INIT, sizeof(*fii), NULL);
900 	fii = GET_INPAYLOAD(ps, pm, fuse_init_in);
901 	fii->major = FUSE_KERNEL_VERSION;
902 	fii->minor = FUSE_KERNEL_MINOR_VERSION;
903 	fii->max_readahead = (unsigned int)(32 * sysconf(_SC_PAGESIZE));
904 	fii->flags = (FUSE_ASYNC_READ|FUSE_POSIX_LOCKS|FUSE_ATOMIC_O_TRUNC);
905 
906 	if ((error = xchg_msg(pu, 0, pm, sizeof(*fio), wait_reply)) != 0)
907 		DERRX(EX_SOFTWARE, "init message exchange failed (%d)", error);
908 
909 	fio = GET_OUTPAYLOAD(ps, pm, fuse_init_out);
910 	ps->ps_max_readahead = fio->max_readahead;
911 	ps->ps_max_write = fio->max_write;
912 
913 	ps->ps_destroy_msg(pm);
914 
915 	return;
916 }
917 
918 int
919 perfuse_fs_unmount(struct puffs_usermount *pu, int flags)
920 {
921 	perfuse_msg_t *pm;
922 	struct perfuse_state *ps;
923 	puffs_cookie_t opc;
924 	int error;
925 
926 	ps = puffs_getspecific(pu);
927 	opc = (puffs_cookie_t)puffs_getroot(pu);
928 
929 	/*
930 	 * ps_new_msg() is called with NULL creds, which will
931 	 * be interpreted as FUSE superuser.
932 	 */
933 	pm = ps->ps_new_msg(pu, opc, FUSE_DESTROY, 0, NULL);
934 
935 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0){
936 		DWARN("unmount %s", ps->ps_target);
937 		if (!(flags & MNT_FORCE))
938 			return error;
939 		else
940 			error = 0;
941 	} else {
942 		ps->ps_destroy_msg(pm);
943 	}
944 
945 	ps->ps_umount(pu);
946 
947 	if (perfuse_diagflags & PDF_MISC)
948 		DPRINTF("%s unmounted, exit\n", ps->ps_target);
949 
950 	return 0;
951 }
952 
953 int
954 perfuse_fs_statvfs(struct puffs_usermount *pu, struct statvfs *svfsb)
955 {
956 	struct perfuse_state *ps;
957 	perfuse_msg_t *pm;
958 	puffs_cookie_t opc;
959 	struct fuse_statfs_out *fso;
960 	int error;
961 
962 	ps = puffs_getspecific(pu);
963 	opc = (puffs_cookie_t)puffs_getroot(pu);
964 
965 	/*
966 	 * ps_new_msg() is called with NULL creds, which will
967 	 * be interpreted as FUSE superuser.
968 	 */
969 	pm = ps->ps_new_msg(pu, opc, FUSE_STATFS, 0, NULL);
970 
971 	if ((error = xchg_msg(pu, opc, pm, sizeof(*fso), wait_reply)) != 0)
972 		return error;
973 
974 	fso = GET_OUTPAYLOAD(ps, pm, fuse_statfs_out);
975 	svfsb->f_flag = ps->ps_mountflags;
976 	svfsb->f_bsize = fso->st.bsize;
977 	svfsb->f_frsize = fso->st.frsize;
978 	svfsb->f_iosize = ((struct puffs_node *)opc)->pn_va.va_blocksize;
979 	svfsb->f_blocks = fso->st.blocks;
980 	svfsb->f_bfree = fso->st.bfree;
981 	svfsb->f_bavail = fso->st.bavail;
982 	svfsb->f_bresvd = fso->st.bfree - fso->st.bavail;
983 	svfsb->f_files = fso->st.files;
984 	svfsb->f_ffree = fso->st.ffree;
985 	svfsb->f_favail = fso->st.ffree;/* files not reserved for root */
986 	svfsb->f_fresvd = 0;		/* files reserved for root */
987 
988 	svfsb->f_syncreads = ps->ps_syncreads;
989 	svfsb->f_syncwrites = ps->ps_syncwrites;
990 
991 	svfsb->f_asyncreads = ps->ps_asyncreads;
992 	svfsb->f_asyncwrites = ps->ps_asyncwrites;
993 
994 	(void)memcpy(&svfsb->f_fsidx, &ps->ps_fsid, sizeof(ps->ps_fsid));
995 	svfsb->f_fsid = (unsigned long)ps->ps_fsid;
996 	svfsb->f_namemax = MAXPATHLEN;	/* XXX */
997 	svfsb->f_owner = ps->ps_owner_uid;
998 
999 	(void)strlcpy(svfsb->f_mntonname, ps->ps_target, _VFS_NAMELEN);
1000 
1001 	if (ps->ps_filesystemtype != NULL)
1002 		(void)strlcpy(svfsb->f_fstypename,
1003 			      ps->ps_filesystemtype, _VFS_NAMELEN);
1004 	else
1005 		(void)strlcpy(svfsb->f_fstypename, "fuse", _VFS_NAMELEN);
1006 
1007 	if (ps->ps_source != NULL)
1008 		strlcpy(svfsb->f_mntfromname, ps->ps_source, _VFS_NAMELEN);
1009 	else
1010 		strlcpy(svfsb->f_mntfromname, _PATH_FUSE, _VFS_NAMELEN);
1011 
1012 	ps->ps_destroy_msg(pm);
1013 
1014 	return 0;
1015 }
1016 
1017 int
1018 perfuse_fs_sync(struct puffs_usermount *pu, int waitfor,
1019 	const struct puffs_cred *pcr)
1020 {
1021 	/*
1022 	 * FUSE does not seem to have a FS sync callback.
1023 	 * Maybe do not even register this callback
1024 	 */
1025 	return puffs_fsnop_sync(pu, waitfor, pcr);
1026 }
1027 
1028 /* ARGSUSED0 */
1029 int
1030 perfuse_fs_fhtonode(struct puffs_usermount *pu, void *fid, size_t fidsize,
1031 	struct puffs_newinfo *pni)
1032 {
1033 	DERRX(EX_SOFTWARE, "%s: UNIMPLEMENTED (FATAL)", __func__);
1034 	return 0;
1035 }
1036 
1037 /* ARGSUSED0 */
1038 int
1039 perfuse_fs_nodetofh(struct puffs_usermount *pu, puffs_cookie_t cookie,
1040 	void *fid, size_t *fidsize)
1041 {
1042 	DERRX(EX_SOFTWARE, "%s: UNIMPLEMENTED (FATAL)", __func__);
1043 	return 0;
1044 }
1045 
1046 #if 0
1047 /* ARGSUSED0 */
1048 void
1049 perfuse_fs_extattrctl(struct puffs_usermount *pu, int cmd,
1050 	puffs_cookie_t *cookie, int flags, int namespace, const char *attrname)
1051 {
1052 	DERRX(EX_SOFTWARE, "%s: UNIMPLEMENTED (FATAL)", __func__);
1053 	return 0;
1054 }
1055 #endif /* 0 */
1056 
1057 /* ARGSUSED0 */
1058 void
1059 perfuse_fs_suspend(struct puffs_usermount *pu, int status)
1060 {
1061 	return;
1062 }
1063 
1064 
1065 int
1066 perfuse_node_lookup(struct puffs_usermount *pu, puffs_cookie_t opc,
1067 	struct puffs_newinfo *pni, const struct puffs_cn *pcn)
1068 {
1069 	struct perfuse_state *ps;
1070 	struct puffs_node *pn;
1071 	mode_t mode;
1072 	int error;
1073 
1074 	ps = puffs_getspecific(pu);
1075 	node_ref(opc);
1076 
1077 	/*
1078 	 * Check permissions
1079 	 */
1080 	switch(pcn->pcn_nameiop) {
1081 	case NAMEI_DELETE: /* FALLTHROUGH */
1082 	case NAMEI_RENAME: /* FALLTHROUGH */
1083 	case NAMEI_CREATE:
1084 		if (pcn->pcn_flags & NAMEI_ISLASTCN)
1085 			mode = PUFFS_VEXEC|PUFFS_VWRITE;
1086 		else
1087 			mode = PUFFS_VEXEC;
1088 		break;
1089 	case NAMEI_LOOKUP: /* FALLTHROUGH */
1090 	default:
1091 		mode = PUFFS_VEXEC;
1092 		break;
1093 	}
1094 
1095 	if ((error = mode_access(opc, pcn->pcn_cred, mode)) != 0)
1096 		goto out;
1097 
1098 	error = node_lookup_common(pu, (puffs_cookie_t)opc, pni,
1099 				   pcn->pcn_name, pcn->pcn_cred, &pn);
1100 
1101 	if (error != 0)
1102 		goto out;
1103 
1104 	/*
1105 	 * Kernel would kill us if the filesystem returned the parent
1106 	 * itself. If we want to live, hide that!
1107 	 */
1108 	if ((opc == (puffs_cookie_t)pn) && (strcmp(pcn->pcn_name, ".") != 0)) {
1109 		DERRX(EX_SOFTWARE, "lookup \"%s\" in \"%s\" returned parent",
1110 		      pcn->pcn_name, perfuse_node_path(ps, opc));
1111 		/* NOTREACHED */
1112 		error = ESTALE;
1113 		goto out;
1114 	}
1115 
1116 	/*
1117 	 * Removed node
1118 	 */
1119 	if (PERFUSE_NODE_DATA(pn)->pnd_flags & PND_REMOVED) {
1120 		error = ENOENT;
1121 		goto out;
1122 	}
1123 
1124 	/*
1125 	 * Check for sticky bit. Unfortunately there is no way to
1126 	 * do this before creating the puffs_node, since we require
1127 	 * this operation to get the node owner.
1128 	 */
1129 	switch (pcn->pcn_nameiop) {
1130 	case NAMEI_DELETE: /* FALLTHROUGH */
1131 	case NAMEI_RENAME:
1132 		error = sticky_access(opc, pn, pcn->pcn_cred);
1133 		if (error != 0) {
1134 			(void)perfuse_node_reclaim(pu, pn);
1135 			goto out;
1136 		}
1137 		break;
1138 	default:
1139 		break;
1140 	}
1141 
1142 	PERFUSE_NODE_DATA(pn)->pnd_puffs_nlookup++;
1143 
1144 	error = 0;
1145 
1146 out:
1147 	node_rele(opc);
1148 	return error;
1149 }
1150 
1151 int
1152 perfuse_node_create(struct puffs_usermount *pu, puffs_cookie_t opc,
1153 	struct puffs_newinfo *pni, const struct puffs_cn *pcn,
1154 	const struct vattr *vap)
1155 {
1156 	perfuse_msg_t *pm;
1157 	struct perfuse_state *ps;
1158 	struct fuse_create_in *fci;
1159 	struct fuse_entry_out *feo;
1160 	struct fuse_open_out *foo;
1161 	struct puffs_node *pn;
1162 	const char *name;
1163 	size_t namelen;
1164 	size_t len;
1165 	int error;
1166 
1167 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
1168 		return ENOENT;
1169 
1170 	node_ref(opc);
1171 
1172 	/*
1173 	 * If create is unimplemented: Check that it does not
1174 	 * already exists, and if not, do mknod and open
1175 	 */
1176 	ps = puffs_getspecific(pu);
1177 	if (ps->ps_flags & PS_NO_CREAT) {
1178 		error = node_lookup_common(pu, opc, NULL, pcn->pcn_name,
1179 					   pcn->pcn_cred, &pn);
1180 		if (error == 0)	{
1181 			(void)perfuse_node_reclaim(pu, pn);
1182 			error = EEXIST;
1183 			goto out;
1184 		}
1185 
1186 		error = perfuse_node_mknod(pu, opc, pni, pcn, vap);
1187 		if (error != 0)
1188 			goto out;
1189 
1190 		error = node_lookup_common(pu, opc, NULL, pcn->pcn_name,
1191 					   pcn->pcn_cred, &pn);
1192 		if (error != 0)
1193 			goto out;
1194 
1195 		/*
1196 		 * FUSE does the open at create time, while
1197 		 * NetBSD will open in a subsequent operation.
1198 		 * We need to open now, in order to retain FUSE
1199 		 * semantics. The calling process will not get
1200 		 * a file descriptor before the kernel sends
1201 		 * the open operation.
1202 		 */
1203 		error = perfuse_node_open(pu, (puffs_cookie_t)pn,
1204 					  FWRITE, pcn->pcn_cred);
1205 		goto out;
1206 	}
1207 
1208 	name = pcn->pcn_name;
1209 	namelen = pcn->pcn_namelen + 1;
1210 	len = sizeof(*fci) + namelen;
1211 
1212 	/*
1213 	 * flags should use O_WRONLY instead of O_RDWR, but it
1214 	 * breaks when the caller tries to read from file.
1215 	 *
1216 	 * mode must contain file type (ie: S_IFREG), use VTTOIF(vap->va_type)
1217 	 */
1218 	pm = ps->ps_new_msg(pu, opc, FUSE_CREATE, len, pcn->pcn_cred);
1219 	fci = GET_INPAYLOAD(ps, pm, fuse_create_in);
1220 	fci->flags = O_CREAT | O_TRUNC | O_RDWR;
1221 	fci->mode = vap->va_mode | VTTOIF(vap->va_type);
1222 	fci->umask = 0; 	/* Seems unused by libfuse */
1223 	(void)strlcpy((char*)(void *)(fci + 1), name, namelen);
1224 
1225 	len = sizeof(*feo) + sizeof(*foo);
1226 	if ((error = xchg_msg(pu, opc, pm, len, wait_reply)) != 0) {
1227 		/*
1228 		 * create is unimplmented, remember it for later,
1229 		 * and start over using mknod and open instead.
1230 		 */
1231 		if (error == ENOSYS) {
1232 			ps->ps_flags |= PS_NO_CREAT;
1233 			error = perfuse_node_create(pu, opc, pni, pcn, vap);
1234 		}
1235 
1236 		goto out;
1237 	}
1238 
1239 	feo = GET_OUTPAYLOAD(ps, pm, fuse_entry_out);
1240 	foo = (struct fuse_open_out *)(void *)(feo + 1);
1241 	if (feo->nodeid == PERFUSE_UNKNOWN_NODEID)
1242 		DERRX(EX_SOFTWARE, "%s: no nodeid", __func__);
1243 
1244 	/*
1245 	 * Save the file handle and inode in node private data
1246 	 * so that we can reuse it later
1247 	 */
1248 	pn = perfuse_new_pn(pu, name, opc);
1249 	perfuse_new_fh((puffs_cookie_t)pn, foo->fh, FWRITE);
1250 	PERFUSE_NODE_DATA(pn)->pnd_nodeid = feo->nodeid;
1251 	PERFUSE_NODE_DATA(pn)->pnd_puffs_nlookup++;
1252 	perfuse_node_cache(ps, pn);
1253 
1254 	fuse_attr_to_vap(ps, &pn->pn_va, &feo->attr);
1255 	pn->pn_va.va_gen = (u_long)(feo->generation);
1256 
1257 	puffs_newinfo_setcookie(pni, pn);
1258 #ifdef PUFFS_KFLAG_CACHE_FS_TTL
1259 	puffs_newinfo_setva(pni, &pn->pn_va);
1260 	perfuse_newinfo_setttl(pni, pn, feo, NULL);
1261 #endif /* PUFFS_KFLAG_CACHE_FS_TTL */
1262 
1263 #ifdef PERFUSE_DEBUG
1264 	if (perfuse_diagflags & (PDF_FH|PDF_FILENAME))
1265 		DPRINTF("%s: opc = %p, file = \"%s\", flags = 0x%x "
1266 			"nodeid = 0x%"PRIx64", wfh = 0x%"PRIx64"\n",
1267 			__func__, (void *)pn, pcn->pcn_name,
1268 			PERFUSE_NODE_DATA(pn)->pnd_flags, feo->nodeid,
1269 			foo->fh);
1270 #endif
1271 
1272 	ps->ps_destroy_msg(pm);
1273 	error = 0;
1274 
1275 out:
1276 	node_rele(opc);
1277 	return error;
1278 }
1279 
1280 
1281 int
1282 perfuse_node_mknod(struct puffs_usermount *pu, puffs_cookie_t opc,
1283 	struct puffs_newinfo *pni, const struct puffs_cn *pcn,
1284 	const struct vattr *vap)
1285 {
1286 	struct perfuse_state *ps;
1287 	perfuse_msg_t *pm;
1288 	struct fuse_mknod_in *fmi;
1289 	const char* path;
1290 	size_t len;
1291 	int error;
1292 
1293 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
1294 		return ENOENT;
1295 
1296 	node_ref(opc);
1297 
1298 	/*
1299 	 * Only superuser can mknod objects other than
1300 	 * directories, files, socks, fifo and links.
1301 	 *
1302 	 * Create an object require -WX permission in the parent directory
1303 	 */
1304 	switch (vap->va_type) {
1305 	case VDIR:	/* FALLTHROUGH */
1306 	case VREG:	/* FALLTHROUGH */
1307 	case VFIFO:	/* FALLTHROUGH */
1308 	case VSOCK:
1309 		break;
1310 	default:	/* VNON, VBLK, VCHR, VBAD */
1311 		if (!puffs_cred_isjuggernaut(pcn->pcn_cred)) {
1312 			error = EPERM;
1313 			goto out;
1314 		}
1315 		break;
1316 	}
1317 
1318 
1319 	ps = puffs_getspecific(pu);
1320 	path = pcn->pcn_name;
1321 	len = sizeof(*fmi) + pcn->pcn_namelen + 1;
1322 
1323 	/*
1324 	 * mode must contain file type (ie: S_IFREG), use VTTOIF(vap->va_type)
1325 	 */
1326 	pm = ps->ps_new_msg(pu, opc, FUSE_MKNOD, len, pcn->pcn_cred);
1327 	fmi = GET_INPAYLOAD(ps, pm, fuse_mknod_in);
1328 	fmi->mode = vap->va_mode | VTTOIF(vap->va_type);
1329 	fmi->rdev = (uint32_t)vap->va_rdev;
1330 	fmi->umask = 0; 	/* Seems unused bu libfuse */
1331 	(void)strlcpy((char *)(void *)(fmi + 1), path, len - sizeof(*fmi));
1332 
1333 	error = node_mk_common(pu, opc, pni, pcn, pm);
1334 
1335 out:
1336 	node_rele(opc);
1337 	return error;
1338 }
1339 
1340 
1341 int
1342 perfuse_node_open(struct puffs_usermount *pu, puffs_cookie_t opc, int mode,
1343 	const struct puffs_cred *pcr)
1344 {
1345 	return perfuse_node_open2(pu, opc, mode, pcr, NULL);
1346 }
1347 
1348 int
1349 perfuse_node_open2(struct puffs_usermount *pu, puffs_cookie_t opc, int mode,
1350 	const struct puffs_cred *pcr, int *oflags)
1351 {
1352 	struct perfuse_state *ps;
1353 	struct perfuse_node_data *pnd;
1354 	perfuse_msg_t *pm;
1355 	mode_t fmode;
1356 	int op;
1357 	struct fuse_open_in *foi;
1358 	struct fuse_open_out *foo;
1359 	struct puffs_node *pn;
1360 	int error;
1361 
1362 	ps = puffs_getspecific(pu);
1363 	pn = (struct puffs_node *)opc;
1364 	pnd = PERFUSE_NODE_DATA(opc);
1365 	error = 0;
1366 
1367 	if (pnd->pnd_flags & PND_REMOVED)
1368 		return ENOENT;
1369 
1370 	node_ref(opc);
1371 
1372 	if (puffs_pn_getvap(pn)->va_type == VDIR)
1373 		op = FUSE_OPENDIR;
1374 	else
1375 		op = FUSE_OPEN;
1376 
1377 	/*
1378 	 * libfuse docs says
1379 	 * - O_CREAT and O_EXCL should never be set.
1380 	 * - O_TRUNC may be used if mount option atomic_o_trunc is used XXX
1381 	 *
1382 	 * O_APPEND makes no sense since FUSE always sends
1383 	 * the file offset for write operations. If the
1384 	 * filesystem uses pwrite(), O_APPEND would cause
1385 	 * the offset to be ignored and cause file corruption.
1386 	 */
1387 	mode &= ~(O_CREAT|O_EXCL|O_APPEND);
1388 
1389 	/*
1390 	 * Do not open twice, and do not reopen for reading
1391 	 * if we already have write handle.
1392 	 */
1393 	switch (mode & (FREAD|FWRITE)) {
1394 	case FREAD:
1395 		if (pnd->pnd_flags & (PND_RFH|PND_WFH))
1396 			goto out;
1397 		break;
1398 	case FWRITE:
1399 		if (pnd->pnd_flags & PND_WFH)
1400 			goto out;
1401 		break;
1402 	case FREAD|FWRITE:
1403 		if (pnd->pnd_flags & PND_WFH)
1404 			goto out;
1405 
1406 		/*
1407 		 * Corner case: if already open for reading (PND_RFH)
1408 		 * and re-opening FREAD|FWRITE, we need to reopen,
1409 		 * but only for writing. Note the change on mode
1410 		 * will only affect perfuse_new_fh()
1411 		 */
1412 		if (pnd->pnd_flags & PND_RFH)
1413 			mode &= ~FREAD;
1414 		break;
1415 	default:
1416 		DWARNX("open without either FREAD nor FWRITE");
1417 		error = EPERM;
1418 		goto out;
1419 	}
1420 
1421 	/*
1422 	 * Queue open on a node so that we do not open
1423 	 * twice. This would be better with read and
1424 	 * write distinguished.
1425 	 */
1426 	while (pnd->pnd_flags & PND_INOPEN)
1427 		requeue_request(pu, opc, PCQ_OPEN);
1428 	pnd->pnd_flags |= PND_INOPEN;
1429 
1430 	/*
1431 	 * Convert PUFFS mode to FUSE mode: convert FREAD/FWRITE
1432 	 * to O_RDONLY/O_WRONLY while perserving the other options.
1433 	 */
1434 	fmode = mode & ~(FREAD|FWRITE);
1435 	fmode |= (mode & FWRITE) ? O_RDWR : O_RDONLY;
1436 
1437 	pm = ps->ps_new_msg(pu, opc, op, sizeof(*foi), pcr);
1438 	foi = GET_INPAYLOAD(ps, pm, fuse_open_in);
1439 	foi->flags = fmode;
1440 	foi->unused = 0;
1441 
1442 	if ((error = xchg_msg(pu, opc, pm, sizeof(*foo), wait_reply)) != 0)
1443 		goto out;
1444 
1445 	foo = GET_OUTPAYLOAD(ps, pm, fuse_open_out);
1446 
1447 	/*
1448 	 * Save the file handle in node private data
1449 	 * so that we can reuse it later
1450 	 */
1451 	perfuse_new_fh(opc, foo->fh, mode);
1452 
1453 	/*
1454 	 * Set direct I/O if the filesystems forces it
1455 	 */
1456 	if ((foo->open_flags & FUSE_FOPEN_DIRECT_IO) && (oflags != NULL))
1457 		*oflags |= PUFFS_OPEN_IO_DIRECT;
1458 
1459 #ifdef PERFUSE_DEBUG
1460 	if (perfuse_diagflags & (PDF_FH|PDF_FILENAME))
1461 		DPRINTF("%s: opc = %p, file = \"%s\", "
1462 			"nodeid = 0x%"PRIx64", %s%sfh = 0x%"PRIx64"\n",
1463 			__func__, (void *)opc, perfuse_node_path(ps, opc),
1464 			pnd->pnd_nodeid, mode & FREAD ? "r" : "",
1465 			mode & FWRITE ? "w" : "", foo->fh);
1466 #endif
1467 
1468 	ps->ps_destroy_msg(pm);
1469 out:
1470 
1471 	pnd->pnd_flags &= ~PND_INOPEN;
1472 	(void)dequeue_requests(opc, PCQ_OPEN, DEQUEUE_ALL);
1473 
1474 	node_rele(opc);
1475 	return error;
1476 }
1477 
1478 /* ARGSUSED0 */
1479 int
1480 perfuse_node_close(struct puffs_usermount *pu, puffs_cookie_t opc, int flags,
1481 	const struct puffs_cred *pcr)
1482 {
1483 	struct perfuse_node_data *pnd;
1484 
1485 	pnd = PERFUSE_NODE_DATA(opc);
1486 
1487 	if (!(pnd->pnd_flags & PND_OPEN))
1488 		return EBADF;
1489 
1490 	/*
1491 	 * Actual close is postponed at inactive time.
1492 	 */
1493 	return 0;
1494 }
1495 
1496 int
1497 perfuse_node_access(struct puffs_usermount *pu, puffs_cookie_t opc, int mode,
1498 	const struct puffs_cred *pcr)
1499 {
1500 	perfuse_msg_t *pm;
1501 	struct perfuse_state *ps;
1502 	struct fuse_access_in *fai;
1503 	int error;
1504 
1505 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
1506 		return ENOENT;
1507 
1508 	node_ref(opc);
1509 
1510 	/*
1511 	 * If we previously detected the filesystem does not
1512 	 * implement access(), short-circuit the call and skip
1513 	 * to libpuffs access() emulation.
1514 	 */
1515 	ps = puffs_getspecific(pu);
1516 	if (ps->ps_flags & PS_NO_ACCESS) {
1517 		const struct vattr *vap;
1518 
1519 		vap = puffs_pn_getvap((struct puffs_node *)opc);
1520 
1521 		error = puffs_access(IFTOVT(vap->va_mode),
1522 				     vap->va_mode & ACCESSPERMS,
1523 				     vap->va_uid, vap->va_gid,
1524 				     (mode_t)mode, pcr);
1525 		goto out;
1526 	}
1527 
1528 	/*
1529 	 * Plain access call
1530 	 */
1531 	pm = ps->ps_new_msg(pu, opc, FUSE_ACCESS, sizeof(*fai), pcr);
1532 	fai = GET_INPAYLOAD(ps, pm, fuse_access_in);
1533 	fai->mask = 0;
1534 	fai->mask |= (mode & PUFFS_VREAD) ? R_OK : 0;
1535 	fai->mask |= (mode & PUFFS_VWRITE) ? W_OK : 0;
1536 	fai->mask |= (mode & PUFFS_VEXEC) ? X_OK : 0;
1537 
1538 	error = xchg_msg(pu, opc, pm, NO_PAYLOAD_REPLY_LEN, wait_reply);
1539 
1540 	ps->ps_destroy_msg(pm);
1541 
1542 	/*
1543 	 * If unimplemented, start over with emulation
1544 	 */
1545 	if (error == ENOSYS) {
1546 		ps->ps_flags |= PS_NO_ACCESS;
1547 		error = perfuse_node_access(pu, opc, mode, pcr);
1548 	}
1549 
1550 out:
1551 	node_rele(opc);
1552 	return error;
1553 }
1554 
1555 int
1556 perfuse_node_getattr(struct puffs_usermount *pu, puffs_cookie_t opc,
1557 	struct vattr *vap, const struct puffs_cred *pcr)
1558 {
1559 	return perfuse_node_getattr_ttl(pu, opc, vap, pcr, NULL);
1560 }
1561 
1562 int
1563 perfuse_node_getattr_ttl(struct puffs_usermount *pu, puffs_cookie_t opc,
1564 	struct vattr *vap, const struct puffs_cred *pcr,
1565 	struct timespec *va_ttl)
1566 {
1567 	perfuse_msg_t *pm = NULL;
1568 	struct perfuse_state *ps;
1569 	struct perfuse_node_data *pnd = PERFUSE_NODE_DATA(opc);
1570 	struct fuse_getattr_in *fgi;
1571 	struct fuse_attr_out *fao;
1572 	int error = 0;
1573 
1574 	if ((pnd->pnd_flags & PND_REMOVED) && !(pnd->pnd_flags & PND_OPEN))
1575 		return ENOENT;
1576 
1577 	node_ref(opc);
1578 
1579 	/*
1580 	 * Serialize size access, see comment in perfuse_node_setattr().
1581 	 */
1582 	while (pnd->pnd_flags & PND_INRESIZE)
1583 		requeue_request(pu, opc, PCQ_RESIZE);
1584 	pnd->pnd_flags |= PND_INRESIZE;
1585 
1586 	ps = puffs_getspecific(pu);
1587 
1588 	/*
1589 	 * FUSE_GETATTR_FH must be set in fgi->flags
1590 	 * if we use for fgi->fh
1591 	 */
1592 	pm = ps->ps_new_msg(pu, opc, FUSE_GETATTR, sizeof(*fgi), pcr);
1593 	fgi = GET_INPAYLOAD(ps, pm, fuse_getattr_in);
1594 	fgi->getattr_flags = 0;
1595 	fgi->dummy = 0;
1596 	fgi->fh = 0;
1597 
1598 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_OPEN) {
1599 		fgi->fh = perfuse_get_fh(opc, FREAD);
1600 		fgi->getattr_flags |= FUSE_GETATTR_FH;
1601 	}
1602 
1603 #ifdef PERFUSE_DEBUG
1604 	if (perfuse_diagflags & PDF_RESIZE)
1605 		DPRINTF(">> %s %p %" PRIu64 "\n", __func__, (void *)opc,
1606 		    vap->va_size);
1607 #endif
1608 
1609 	if ((error = xchg_msg(pu, opc, pm, sizeof(*fao), wait_reply)) != 0)
1610 		goto out;
1611 
1612 	fao = GET_OUTPAYLOAD(ps, pm, fuse_attr_out);
1613 
1614 #ifdef PERFUSE_DEBUG
1615 	if (perfuse_diagflags & PDF_RESIZE)
1616 		DPRINTF("<< %s %p %" PRIu64 " -> %" PRIu64 "\n", __func__,
1617 		    (void *)opc, vap->va_size, fao->attr.size);
1618 #endif
1619 
1620 	/*
1621 	 * We set birthtime, flags, filerev,vaflags to 0.
1622 	 * This seems the best bet, since the information is
1623 	 * not available from filesystem.
1624 	 */
1625 	fuse_attr_to_vap(ps, vap, &fao->attr);
1626 
1627 	if (va_ttl != NULL) {
1628 		va_ttl->tv_sec = fao->attr_valid;
1629 		va_ttl->tv_nsec = fao->attr_valid_nsec;
1630 	}
1631 
1632 	ps->ps_destroy_msg(pm);
1633 	error = 0;
1634 out:
1635 
1636 	pnd->pnd_flags &= ~PND_INRESIZE;
1637 	(void)dequeue_requests(opc, PCQ_RESIZE, DEQUEUE_ALL);
1638 
1639 	node_rele(opc);
1640 	return error;
1641 }
1642 
1643 int
1644 perfuse_node_setattr(struct puffs_usermount *pu, puffs_cookie_t opc,
1645 	const struct vattr *vap, const struct puffs_cred *pcr)
1646 {
1647 	return perfuse_node_setattr_ttl(pu, opc,
1648 					__UNCONST(vap), pcr, NULL, 0);
1649 }
1650 
1651 int
1652 perfuse_node_setattr_ttl(struct puffs_usermount *pu, puffs_cookie_t opc,
1653 	struct vattr *vap, const struct puffs_cred *pcr,
1654 	struct timespec *va_ttl, int xflag)
1655 {
1656 	perfuse_msg_t *pm;
1657 	uint64_t fh;
1658 	struct perfuse_state *ps;
1659 	struct perfuse_node_data *pnd;
1660 	struct fuse_setattr_in *fsi;
1661 	struct fuse_attr_out *fao;
1662 	struct vattr *old_va;
1663 	enum perfuse_xchg_pb_reply reply;
1664 	int error;
1665 #ifdef PERFUSE_DEBUG
1666 	struct vattr *old_vap;
1667 	int resize_debug = 0;
1668 #endif
1669 	ps = puffs_getspecific(pu);
1670 	pnd = PERFUSE_NODE_DATA(opc);
1671 
1672 	/*
1673 	 * The only operation we can do once the file is removed
1674 	 * is to resize it, and we can do it only if it is open.
1675 	 * Do not even send the operation to the filesystem: the
1676 	 * file is not there anymore.
1677 	 */
1678 	if (pnd->pnd_flags & PND_REMOVED) {
1679 		if (!(pnd->pnd_flags & PND_OPEN))
1680 			return ENOENT;
1681 
1682 		return 0;
1683 	}
1684 
1685 	old_va = puffs_pn_getvap((struct puffs_node *)opc);
1686 
1687 	/*
1688 	 * Check for permission to change size
1689 	 * It is always allowed if we already have a write file handle
1690 	 */
1691 	if ((vap->va_size != (u_quad_t)PUFFS_VNOVAL) &&
1692 	    !(pnd->pnd_flags & PND_WFH) &&
1693 	    (error = mode_access(opc, pcr, PUFFS_VWRITE)) != 0)
1694 		return error;
1695 
1696 	/*
1697 	 * Check for permission to change dates
1698 	 */
1699 	if (((vap->va_atime.tv_sec != (time_t)PUFFS_VNOVAL) ||
1700 	     (vap->va_mtime.tv_sec != (time_t)PUFFS_VNOVAL)) &&
1701 	    (puffs_access_times(old_va->va_uid, old_va->va_gid,
1702 				old_va->va_mode, 0, pcr) != 0))
1703 		return EPERM;
1704 
1705 	/*
1706 	 * Check for permission to change owner and group
1707 	 */
1708 	if (((vap->va_uid != (uid_t)PUFFS_VNOVAL) ||
1709 	     (vap->va_gid != (gid_t)PUFFS_VNOVAL)) &&
1710 	    (puffs_access_chown(old_va->va_uid, old_va->va_gid,
1711 				vap->va_uid, vap->va_gid, pcr)) != 0)
1712 		return EPERM;
1713 
1714 	/*
1715 	 * Check for sticky bit on non-directory by non root user
1716 	 */
1717 	if ((vap->va_mode != (mode_t)PUFFS_VNOVAL) &&
1718 	    (vap->va_mode & S_ISTXT) && (old_va->va_type != VDIR) &&
1719 	    !puffs_cred_isjuggernaut(pcr))
1720 		return EFTYPE;
1721 
1722 	/*
1723 	 * Check for permission to change permissions
1724 	 */
1725 	if ((vap->va_mode != (mode_t)PUFFS_VNOVAL) &&
1726 	    (puffs_access_chmod(old_va->va_uid, old_va->va_gid,
1727 				old_va->va_type, vap->va_mode, pcr)) != 0)
1728 		return EPERM;
1729 
1730 	node_ref(opc);
1731 
1732 	if (pnd->pnd_flags & PND_WFH)
1733 		fh = perfuse_get_fh(opc, FWRITE);
1734 	else
1735 		fh = FUSE_UNKNOWN_FH;
1736 
1737 	/*
1738 	 * fchmod() sets mode and fh, and it may carry
1739 	 * a resize as well. That may break if the
1740 	 * filesystem does chmod then resize, and fails
1741 	 * because it does not have permission anymore.
1742 	 * We work this around by splitting into two setattr.
1743 	 */
1744 	if ((vap->va_size != (u_quad_t)PUFFS_VNOVAL) &&
1745 	    (vap->va_mode != (mode_t)PUFFS_VNOVAL) &&
1746 	    (fh != FUSE_UNKNOWN_FH)) {
1747 		struct vattr resize_va;
1748 
1749 		(void)memcpy(&resize_va, vap, sizeof(resize_va));
1750 		resize_va.va_mode = (mode_t)PUFFS_VNOVAL;
1751 		if ((error = perfuse_node_setattr_ttl(pu, opc, &resize_va,
1752 						      pcr, va_ttl, xflag)) != 0)
1753 			goto out2;
1754 
1755 		vap->va_size = (u_quad_t)PUFFS_VNOVAL;
1756 	}
1757 
1758 	pm = ps->ps_new_msg(pu, opc, FUSE_SETATTR, sizeof(*fsi), pcr);
1759 	fsi = GET_INPAYLOAD(ps, pm, fuse_setattr_in);
1760 	fsi->valid = 0;
1761 
1762 	/*
1763 	 * Get a fh if the node is open for writing
1764 	 */
1765 	if (fh != FUSE_UNKNOWN_FH) {
1766 		fsi->fh = fh;
1767 		fsi->valid |= FUSE_FATTR_FH;
1768 	}
1769 
1770 
1771 	if (vap->va_size != (u_quad_t)PUFFS_VNOVAL) {
1772 		fsi->size = vap->va_size;
1773 		fsi->valid |= FUSE_FATTR_SIZE;
1774 
1775 		/*
1776 		 * Serialize anything that can touch file size
1777 		 * to avoid reordered GETATTR and SETATTR.
1778 		 * Out of order SETATTR can report stale size,
1779 		 * which will cause the kernel to truncate the file.
1780 		 * XXX Probably useless now we have a lock on GETATTR
1781 		 */
1782 		while (pnd->pnd_flags & PND_INRESIZE)
1783 			requeue_request(pu, opc, PCQ_RESIZE);
1784 		pnd->pnd_flags |= PND_INRESIZE;
1785 	}
1786 
1787 	/*
1788  	 * When not sending a time field, still fill with
1789 	 * current value, as the filesystem may just reset
1790 	 * the field to Epoch even if fsi->valid bit is
1791 	 * not set (GlusterFS does that).
1792  	 */
1793 	if (vap->va_atime.tv_sec != (time_t)PUFFS_VNOVAL) {
1794 		fsi->atime = vap->va_atime.tv_sec;
1795 		fsi->atimensec = (uint32_t)vap->va_atime.tv_nsec;
1796 		fsi->valid |= FUSE_FATTR_ATIME;
1797 	} else {
1798 		fsi->atime = old_va->va_atime.tv_sec;
1799 		fsi->atimensec = (uint32_t)old_va->va_atime.tv_nsec;
1800 	}
1801 
1802 	if (vap->va_mtime.tv_sec != (time_t)PUFFS_VNOVAL) {
1803 		fsi->mtime = vap->va_mtime.tv_sec;
1804 		fsi->mtimensec = (uint32_t)vap->va_mtime.tv_nsec;
1805 		fsi->valid |= FUSE_FATTR_MTIME;
1806 	} else {
1807 		fsi->mtime = old_va->va_mtime.tv_sec;
1808 		fsi->mtimensec = (uint32_t)old_va->va_mtime.tv_nsec;
1809 	}
1810 
1811 	if (vap->va_mode != (mode_t)PUFFS_VNOVAL) {
1812 		fsi->mode = vap->va_mode;
1813 		fsi->valid |= FUSE_FATTR_MODE;
1814 	}
1815 
1816 	if (vap->va_uid != (uid_t)PUFFS_VNOVAL) {
1817 		fsi->uid = vap->va_uid;
1818 		fsi->valid |= FUSE_FATTR_UID;
1819 	}
1820 
1821 	if (vap->va_gid != (gid_t)PUFFS_VNOVAL) {
1822 		fsi->gid = vap->va_gid;
1823 		fsi->valid |= FUSE_FATTR_GID;
1824 	}
1825 
1826 	if (pnd->pnd_lock_owner != 0) {
1827 		fsi->lock_owner = pnd->pnd_lock_owner;
1828 		fsi->valid |= FUSE_FATTR_LOCKOWNER;
1829 	}
1830 
1831 #ifndef PUFFS_KFLAG_NOFLUSH_META
1832 	/*
1833 	 * ftruncate() sends only va_size, and metadata cache
1834 	 * flush adds va_atime and va_mtime. Some FUSE
1835 	 * filesystems will attempt to detect ftruncate by
1836 	 * checking for FATTR_SIZE being set without
1837 	 * FATTR_UID|FATTR_GID|FATTR_ATIME|FATTR_MTIME|FATTR_MODE
1838 	 *
1839 	 * Try to adapt and remove FATTR_ATIME|FATTR_MTIME
1840 	 * if we suspect a ftruncate().
1841 	 */
1842 	if ((vap->va_size != (u_quad_t)PUFFS_VNOVAL) &&
1843 	    ((vap->va_mode == (mode_t)PUFFS_VNOVAL) &&
1844 	     (vap->va_uid == (uid_t)PUFFS_VNOVAL) &&
1845 	     (vap->va_gid == (gid_t)PUFFS_VNOVAL))) {
1846 		fsi->atime = 0;
1847 		fsi->atimensec = 0;
1848 		fsi->mtime = 0;
1849 		fsi->mtimensec = 0;
1850 		fsi->valid &= ~(FUSE_FATTR_ATIME|FUSE_FATTR_MTIME);
1851 	}
1852 
1853 	/*
1854 	 * If only atime is changed, discard the operation: it
1855 	 * happens after read, and in that case the filesystem
1856 	 * already updaed atime. NB: utimes() also change mtime.
1857 	 */
1858 	if (fsi->valid == FUSE_FATTR_ATIME)
1859 		fsi->valid &= ~FUSE_FATTR_ATIME;
1860 #endif /* PUFFS_KFLAG_NOFLUSH_META */
1861 
1862 	/*
1863 	 * If nothing remain, discard the operation.
1864 	 */
1865 	if (!(fsi->valid & (FUSE_FATTR_SIZE|FUSE_FATTR_ATIME|FUSE_FATTR_MTIME|
1866 			    FUSE_FATTR_MODE|FUSE_FATTR_UID|FUSE_FATTR_GID))) {
1867 		error = 0;
1868 		ps->ps_destroy_msg(pm);
1869 		goto out;
1870 	}
1871 
1872 #ifdef PERFUSE_DEBUG
1873 	old_vap = puffs_pn_getvap((struct puffs_node *)opc);
1874 
1875 	if ((perfuse_diagflags & PDF_RESIZE) &&
1876 	    (old_vap->va_size != (u_quad_t)PUFFS_VNOVAL)) {
1877 		resize_debug = 1;
1878 
1879 		DPRINTF(">> %s %p %" PRIu64 " -> %" PRIu64 "\n", __func__,
1880 		    (void *)opc,
1881 		    puffs_pn_getvap((struct puffs_node *)opc)->va_size,
1882 		    fsi->size);
1883 	}
1884 #endif
1885 
1886 	/*
1887 	 * Do not honour FAF when changing size. How do
1888 	 * you want such a thing to work?
1889 	 */
1890 	reply = wait_reply;
1891 #ifdef PUFFS_SETATTR_FAF
1892 	if ((xflag & PUFFS_SETATTR_FAF) && !(fsi->valid & FUSE_FATTR_SIZE))
1893 		reply = no_reply;
1894 #endif
1895 	if ((error = xchg_msg(pu, opc, pm, sizeof(*fao), reply)) != 0)
1896 		goto out;
1897 
1898 	if (reply == no_reply)
1899 		goto out;
1900 
1901 	/*
1902 	 * Copy back the new values
1903 	 */
1904 	fao = GET_OUTPAYLOAD(ps, pm, fuse_attr_out);
1905 
1906 #ifdef PERFUSE_DEBUG
1907 	if (resize_debug)
1908 		DPRINTF("<< %s %p %" PRIu64 " -> %" PRIu64 "\n", __func__,
1909 		    (void *)opc, old_vap->va_size, fao->attr.size);
1910 #endif
1911 
1912 	fuse_attr_to_vap(ps, old_va, &fao->attr);
1913 
1914 	if (va_ttl != NULL) {
1915 		va_ttl->tv_sec = fao->attr_valid;
1916 		va_ttl->tv_nsec = fao->attr_valid_nsec;
1917 		(void)memcpy(vap, old_va, sizeof(*vap));
1918 	}
1919 
1920 	ps->ps_destroy_msg(pm);
1921 	error = 0;
1922 
1923 out:
1924 	if (pnd->pnd_flags & PND_INRESIZE) {
1925 		pnd->pnd_flags &= ~PND_INRESIZE;
1926 		(void)dequeue_requests(opc, PCQ_RESIZE, DEQUEUE_ALL);
1927 	}
1928 
1929 out2:
1930 	node_rele(opc);
1931 	return error;
1932 }
1933 
1934 int
1935 perfuse_node_poll(struct puffs_usermount *pu, puffs_cookie_t opc, int *events)
1936 {
1937 	struct perfuse_state *ps;
1938 	perfuse_msg_t *pm;
1939 	struct fuse_poll_in *fpi;
1940 	struct fuse_poll_out *fpo;
1941 	int error;
1942 
1943 	node_ref(opc);
1944 	ps = puffs_getspecific(pu);
1945 	/*
1946 	 * kh is set if FUSE_POLL_SCHEDULE_NOTIFY is set.
1947 	 *
1948 	 * XXX ps_new_msg() is called with NULL creds, which will
1949 	 * be interpreted as FUSE superuser. We have no way to
1950 	 * know the requesting process' credential, but since poll
1951 	 * is supposed to operate on a file that has been open,
1952 	 * permission should have already been checked at open time.
1953 	 * That still may breaks on filesystems that provides odd
1954 	 * semantics.
1955  	 */
1956 	pm = ps->ps_new_msg(pu, opc, FUSE_POLL, sizeof(*fpi), NULL);
1957 	fpi = GET_INPAYLOAD(ps, pm, fuse_poll_in);
1958 	fpi->fh = perfuse_get_fh(opc, FREAD);
1959 	fpi->kh = 0;
1960 	fpi->flags = 0;
1961 
1962 #ifdef PERFUSE_DEBUG
1963 	if (perfuse_diagflags & PDF_FH)
1964 		DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", "
1965 			"fh = 0x%"PRIx64"\n", __func__, (void *)opc,
1966 			PERFUSE_NODE_DATA(opc)->pnd_nodeid, fpi->fh);
1967 #endif
1968 	if ((error = xchg_msg(pu, opc, pm, sizeof(*fpo), wait_reply)) != 0)
1969 		goto out;
1970 
1971 	fpo = GET_OUTPAYLOAD(ps, pm, fuse_poll_out);
1972 	*events = fpo->revents;
1973 
1974 	ps->ps_destroy_msg(pm);
1975 	error = 0;
1976 
1977 out:
1978 	node_rele(opc);
1979 	return error;
1980 }
1981 
1982 /* ARGSUSED2 */
1983 int
1984 perfuse_node_fsync(struct puffs_usermount *pu, puffs_cookie_t opc,
1985 	const struct puffs_cred *pcr, int flags, off_t offlo, off_t offhi)
1986 {
1987 	int op;
1988 	perfuse_msg_t *pm;
1989 	struct perfuse_state *ps;
1990 	struct perfuse_node_data *pnd;
1991 	struct fuse_fsync_in *ffi;
1992 	uint64_t fh;
1993 	int error = 0;
1994 
1995 	pm = NULL;
1996 	ps = puffs_getspecific(pu);
1997 	pnd = PERFUSE_NODE_DATA(opc);
1998 
1999 	/*
2000 	 * No need to sync a removed node
2001 	 */
2002 	if (pnd->pnd_flags & PND_REMOVED)
2003 		return 0;
2004 
2005 	/*
2006 	 * We do not sync closed files. They have been
2007 	 * sync at inactive time already.
2008 	 */
2009 	if (!(pnd->pnd_flags & PND_OPEN))
2010 		return 0;
2011 
2012 	node_ref(opc);
2013 
2014 	if (puffs_pn_getvap((struct puffs_node *)opc)->va_type == VDIR)
2015 		op = FUSE_FSYNCDIR;
2016 	else 		/* VREG but also other types such as VLNK */
2017 		op = FUSE_FSYNC;
2018 
2019 	/*
2020 	 * Do not sync if there are no change to sync
2021 	 * XXX remove that test on files if we implement mmap
2022 	 */
2023 #ifdef PERFUSE_DEBUG
2024 	if (perfuse_diagflags & PDF_SYNC)
2025 		DPRINTF("%s: TEST opc = %p, file = \"%s\" is %sdirty\n",
2026 			__func__, (void*)opc, perfuse_node_path(ps, opc),
2027 			pnd->pnd_flags & PND_DIRTY ? "" : "not ");
2028 #endif
2029 	if (!(pnd->pnd_flags & PND_DIRTY))
2030 		goto out;
2031 
2032 	/*
2033 	 * It seems NetBSD can call fsync without open first
2034 	 * glusterfs complain in such a situation:
2035 	 * "FSYNC() ERR => -1 (Invalid argument)"
2036 	 * The file will be closed at inactive time.
2037 	 *
2038 	 * We open the directory for reading in order to sync.
2039 	 * This sounds rather counterintuitive, but it works.
2040 	 */
2041 	if (!(pnd->pnd_flags & PND_WFH)) {
2042 		if ((error = perfuse_node_open(pu, opc, FREAD, pcr)) != 0)
2043 			goto out;
2044 	}
2045 
2046 	if (op == FUSE_FSYNCDIR)
2047 		fh = perfuse_get_fh(opc, FREAD);
2048 	else
2049 		fh = perfuse_get_fh(opc, FWRITE);
2050 
2051 	/*
2052 	 * If fsync_flags  is set, meta data should not be flushed.
2053 	 */
2054 	pm = ps->ps_new_msg(pu, opc, op, sizeof(*ffi), pcr);
2055 	ffi = GET_INPAYLOAD(ps, pm, fuse_fsync_in);
2056 	ffi->fh = fh;
2057 	ffi->fsync_flags = (flags & FFILESYNC) ? 0 : 1;
2058 
2059 #ifdef PERFUSE_DEBUG
2060 	if (perfuse_diagflags & PDF_FH)
2061 		DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", fh = 0x%"PRIx64"\n",
2062 			__func__, (void *)opc,
2063 			PERFUSE_NODE_DATA(opc)->pnd_nodeid, ffi->fh);
2064 #endif
2065 
2066 	if ((error = xchg_msg(pu, opc, pm,
2067 			      NO_PAYLOAD_REPLY_LEN, wait_reply)) != 0)
2068 		goto out;
2069 
2070 	/*
2071 	 * No reply beyond fuse_out_header: nothing to do on success
2072 	 * just clear the dirty flag
2073 	 */
2074 	pnd->pnd_flags &= ~PND_DIRTY;
2075 
2076 #ifdef PERFUSE_DEBUG
2077 	if (perfuse_diagflags & PDF_SYNC)
2078 		DPRINTF("%s: CLEAR opc = %p, file = \"%s\"\n",
2079 			__func__, (void*)opc, perfuse_node_path(ps, opc));
2080 #endif
2081 
2082 	ps->ps_destroy_msg(pm);
2083 	error = 0;
2084 
2085 out:
2086 	/*
2087 	 * ENOSYS is not returned to kernel,
2088 	 */
2089 	if (error == ENOSYS)
2090 		error = 0;
2091 
2092 	node_rele(opc);
2093 	return error;
2094 }
2095 
2096 int
2097 perfuse_node_remove(struct puffs_usermount *pu, puffs_cookie_t opc,
2098 	puffs_cookie_t targ, const struct puffs_cn *pcn)
2099 {
2100 	struct perfuse_state *ps;
2101 	struct perfuse_node_data *pnd;
2102 	perfuse_msg_t *pm;
2103 	char *path;
2104 	const char *name;
2105 	size_t len;
2106 	int error;
2107 
2108 	pnd = PERFUSE_NODE_DATA(opc);
2109 
2110 	if ((pnd->pnd_flags & PND_REMOVED) ||
2111 	    (PERFUSE_NODE_DATA(targ)->pnd_flags & PND_REMOVED))
2112 		return ENOENT;
2113 
2114 #ifdef PERFUSE_DEBUG
2115 	if (targ == NULL)
2116 		DERRX(EX_SOFTWARE, "%s: targ is NULL", __func__);
2117 
2118 	if (perfuse_diagflags & (PDF_FH|PDF_FILENAME))
2119 		DPRINTF("%s: opc = %p, remove opc = %p, file = \"%s\"\n",
2120 			__func__, (void *)opc, (void *)targ, pcn->pcn_name);
2121 #endif
2122 	node_ref(opc);
2123 	node_ref(targ);
2124 
2125 	/*
2126 	 * Await for all operations on the deleted node to drain,
2127 	 * as the filesystem may be confused to have it deleted
2128 	 * during a getattr
2129 	 */
2130 	while (PERFUSE_NODE_DATA(targ)->pnd_inxchg)
2131 		requeue_request(pu, targ, PCQ_AFTERXCHG);
2132 
2133 	ps = puffs_getspecific(pu);
2134 	pnd = PERFUSE_NODE_DATA(opc);
2135 	name = pcn->pcn_name;
2136 	len = pcn->pcn_namelen + 1;
2137 
2138 	pm = ps->ps_new_msg(pu, opc, FUSE_UNLINK, len, pcn->pcn_cred);
2139 	path = _GET_INPAYLOAD(ps, pm, char *);
2140 	(void)strlcpy(path, name, len);
2141 
2142 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
2143 		goto out;
2144 
2145 	perfuse_cache_flush(targ);
2146 	PERFUSE_NODE_DATA(targ)->pnd_flags |= PND_REMOVED;
2147 
2148 	if (!(PERFUSE_NODE_DATA(targ)->pnd_flags & PND_OPEN))
2149 		puffs_setback(puffs_cc_getcc(pu), PUFFS_SETBACK_NOREF_N2);
2150 
2151 	/*
2152 	 * The parent directory needs a sync
2153 	 */
2154 	PERFUSE_NODE_DATA(opc)->pnd_flags |= PND_DIRTY;
2155 
2156 #ifdef PERFUSE_DEBUG
2157 	if (perfuse_diagflags & PDF_FILENAME)
2158 		DPRINTF("%s: remove nodeid = 0x%"PRIx64" file = \"%s\"\n",
2159 			__func__, PERFUSE_NODE_DATA(targ)->pnd_nodeid,
2160 			pcn->pcn_name);
2161 #endif
2162 	ps->ps_destroy_msg(pm);
2163 	error = 0;
2164 
2165 out:
2166 	node_rele(opc);
2167 	node_rele(targ);
2168 	return error;
2169 }
2170 
2171 int
2172 perfuse_node_link(struct puffs_usermount *pu, puffs_cookie_t opc,
2173 	puffs_cookie_t targ, const struct puffs_cn *pcn)
2174 {
2175 	struct perfuse_state *ps;
2176 	perfuse_msg_t *pm;
2177 	const char *name;
2178 	size_t len;
2179 	struct puffs_node *pn;
2180 	struct fuse_link_in *fli;
2181 	int error;
2182 
2183 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
2184 		return ENOENT;
2185 
2186 	node_ref(opc);
2187 	node_ref(targ);
2188 	ps = puffs_getspecific(pu);
2189 	pn = (struct puffs_node *)targ;
2190 	name = pcn->pcn_name;
2191 	len =  sizeof(*fli) + pcn->pcn_namelen + 1;
2192 
2193 	pm = ps->ps_new_msg(pu, opc, FUSE_LINK, len, pcn->pcn_cred);
2194 	fli = GET_INPAYLOAD(ps, pm, fuse_link_in);
2195 	fli->oldnodeid = PERFUSE_NODE_DATA(pn)->pnd_nodeid;
2196 	(void)strlcpy((char *)(void *)(fli + 1), name, len - sizeof(*fli));
2197 
2198 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
2199 		goto out;
2200 
2201 	ps->ps_destroy_msg(pm);
2202 	error = 0;
2203 
2204 out:
2205 	node_rele(opc);
2206 	node_rele(targ);
2207 	return error;
2208 }
2209 
2210 int
2211 perfuse_node_rename(struct puffs_usermount *pu, puffs_cookie_t opc,
2212 	puffs_cookie_t src, const struct puffs_cn *pcn_src,
2213 	puffs_cookie_t targ_dir, puffs_cookie_t targ,
2214 	const struct puffs_cn *pcn_targ)
2215 {
2216 	struct perfuse_state *ps;
2217 	struct perfuse_node_data *dstdir_pnd;
2218 	perfuse_msg_t *pm;
2219 	struct fuse_rename_in *fri;
2220 	const char *newname;
2221 	const char *oldname;
2222 	char *np;
2223 	int error;
2224 	size_t len;
2225 	size_t newname_len;
2226 	size_t oldname_len;
2227 
2228 	if ((PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED) ||
2229 	    (PERFUSE_NODE_DATA(src)->pnd_flags & PND_REMOVED) ||
2230 	    (PERFUSE_NODE_DATA(targ_dir)->pnd_flags & PND_REMOVED))
2231 		return ENOENT;
2232 
2233 	node_ref(opc);
2234 	node_ref(src);
2235 
2236 	/*
2237 	 * Await for all operations on the deleted node to drain,
2238 	 * as the filesystem may be confused to have it deleted
2239 	 * during a getattr
2240 	 */
2241 	if ((struct puffs_node *)targ != NULL) {
2242 		node_ref(targ);
2243 		while (PERFUSE_NODE_DATA(targ)->pnd_inxchg)
2244 			requeue_request(pu, targ, PCQ_AFTERXCHG);
2245 	} else {
2246 		while (PERFUSE_NODE_DATA(src)->pnd_inxchg)
2247 			requeue_request(pu, src, PCQ_AFTERXCHG);
2248 	}
2249 
2250 	ps = puffs_getspecific(pu);
2251 	newname =  pcn_targ->pcn_name;
2252 	newname_len = pcn_targ->pcn_namelen + 1;
2253 	oldname =  pcn_src->pcn_name;
2254 	oldname_len = pcn_src->pcn_namelen + 1;
2255 
2256 	len = sizeof(*fri) + oldname_len + newname_len;
2257 	pm = ps->ps_new_msg(pu, opc, FUSE_RENAME, len, pcn_targ->pcn_cred);
2258 	fri = GET_INPAYLOAD(ps, pm, fuse_rename_in);
2259 	fri->newdir = PERFUSE_NODE_DATA(targ_dir)->pnd_nodeid;
2260 	np = (char *)(void *)(fri + 1);
2261 	(void)strlcpy(np, oldname, oldname_len);
2262 	np += oldname_len;
2263 	(void)strlcpy(np, newname, newname_len);
2264 
2265 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
2266 		goto out;
2267 
2268 
2269 	/*
2270 	 * Record new parent nodeid
2271 	 */
2272 	dstdir_pnd = PERFUSE_NODE_DATA(targ_dir);
2273 	PERFUSE_NODE_DATA(src)->pnd_parent_nodeid = dstdir_pnd->pnd_nodeid;
2274 
2275 	if (opc != targ_dir)
2276 		dstdir_pnd->pnd_flags |= PND_DIRTY;
2277 
2278 	if (strcmp(newname, "..") != 0)
2279 		(void)strlcpy(PERFUSE_NODE_DATA(src)->pnd_name,
2280 		    newname, MAXPATHLEN);
2281 	else
2282 		PERFUSE_NODE_DATA(src)->pnd_name[0] = 0; /* forget name */
2283 
2284 	PERFUSE_NODE_DATA(opc)->pnd_flags |= PND_DIRTY;
2285 
2286 	if ((struct puffs_node *)targ != NULL) {
2287 		perfuse_cache_flush(targ);
2288 		PERFUSE_NODE_DATA(targ)->pnd_flags |= PND_REMOVED;
2289 	}
2290 
2291 #ifdef PERFUSE_DEBUG
2292 	if (perfuse_diagflags & PDF_FILENAME)
2293 		DPRINTF("%s: nodeid = 0x%"PRIx64" file = \"%s\" renamed \"%s\" "
2294 			"nodeid = 0x%"PRIx64" -> nodeid = 0x%"PRIx64" \"%s\"\n",
2295 	 		__func__, PERFUSE_NODE_DATA(src)->pnd_nodeid,
2296 			pcn_src->pcn_name, pcn_targ->pcn_name,
2297 			PERFUSE_NODE_DATA(opc)->pnd_nodeid,
2298 			PERFUSE_NODE_DATA(targ_dir)->pnd_nodeid,
2299 			perfuse_node_path(ps, targ_dir));
2300 #endif
2301 
2302 	ps->ps_destroy_msg(pm);
2303 	error = 0;
2304 
2305 out:
2306 	node_rele(opc);
2307 	node_rele(src);
2308 	if ((struct puffs_node *)targ != NULL)
2309 		node_rele(targ);
2310 
2311 	return error;
2312 }
2313 
2314 int
2315 perfuse_node_mkdir(struct puffs_usermount *pu, puffs_cookie_t opc,
2316 	struct puffs_newinfo *pni, const struct puffs_cn *pcn,
2317 	const struct vattr *vap)
2318 {
2319 	struct perfuse_state *ps;
2320 	perfuse_msg_t *pm;
2321 	struct fuse_mkdir_in *fmi;
2322 	const char *path;
2323 	size_t len;
2324 	int error;
2325 
2326 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
2327 		return ENOENT;
2328 
2329 	node_ref(opc);
2330 	ps = puffs_getspecific(pu);
2331 	path = pcn->pcn_name;
2332 	len = sizeof(*fmi) + pcn->pcn_namelen + 1;
2333 
2334 	pm = ps->ps_new_msg(pu, opc, FUSE_MKDIR, len, pcn->pcn_cred);
2335 	fmi = GET_INPAYLOAD(ps, pm, fuse_mkdir_in);
2336 	fmi->mode = vap->va_mode;
2337 	fmi->umask = 0; 	/* Seems unused by libfuse? */
2338 	(void)strlcpy((char *)(void *)(fmi + 1), path, len - sizeof(*fmi));
2339 
2340 	error = node_mk_common(pu, opc, pni, pcn, pm);
2341 
2342 	node_rele(opc);
2343 	return error;
2344 }
2345 
2346 
2347 int
2348 perfuse_node_rmdir(struct puffs_usermount *pu, puffs_cookie_t opc,
2349 	puffs_cookie_t targ, const struct puffs_cn *pcn)
2350 {
2351 	struct perfuse_state *ps;
2352 	struct perfuse_node_data *pnd;
2353 	perfuse_msg_t *pm;
2354 	char *path;
2355 	const char *name;
2356 	size_t len;
2357 	int error;
2358 
2359 	pnd = PERFUSE_NODE_DATA(opc);
2360 
2361 	if ((pnd->pnd_flags & PND_REMOVED) ||
2362 	    (PERFUSE_NODE_DATA(targ)->pnd_flags & PND_REMOVED))
2363 		return ENOENT;
2364 
2365 	/*
2366 	 * Attempt to rmdir dir/.. shoud raise ENOTEMPTY
2367 	 */
2368 	if (PERFUSE_NODE_DATA(targ)->pnd_nodeid == pnd->pnd_parent_nodeid)
2369 		return ENOTEMPTY;
2370 
2371 	node_ref(opc);
2372 	node_ref(targ);
2373 
2374 	/*
2375 	 * Await for all operations on the deleted node to drain,
2376 	 * as the filesystem may be confused to have it deleted
2377 	 * during a getattr
2378 	 */
2379 	while (PERFUSE_NODE_DATA(targ)->pnd_inxchg)
2380 		requeue_request(pu, targ, PCQ_AFTERXCHG);
2381 
2382 	ps = puffs_getspecific(pu);
2383 	name = pcn->pcn_name;
2384 	len = pcn->pcn_namelen + 1;
2385 
2386 	pm = ps->ps_new_msg(pu, opc, FUSE_RMDIR, len, pcn->pcn_cred);
2387 	path = _GET_INPAYLOAD(ps, pm, char *);
2388 	(void)strlcpy(path, name, len);
2389 
2390 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
2391 		goto out;
2392 
2393 	perfuse_cache_flush(targ);
2394 	PERFUSE_NODE_DATA(targ)->pnd_flags |= PND_REMOVED;
2395 
2396 	if (!(PERFUSE_NODE_DATA(targ)->pnd_flags & PND_OPEN))
2397 		puffs_setback(puffs_cc_getcc(pu), PUFFS_SETBACK_NOREF_N2);
2398 
2399 	/*
2400 	 * The parent directory needs a sync
2401 	 */
2402 	PERFUSE_NODE_DATA(opc)->pnd_flags |= PND_DIRTY;
2403 
2404 #ifdef PERFUSE_DEBUG
2405 	if (perfuse_diagflags & PDF_FILENAME)
2406 		DPRINTF("%s: remove nodeid = 0x%"PRIx64" file = \"%s\"\n",
2407 			__func__, PERFUSE_NODE_DATA(targ)->pnd_nodeid,
2408 			perfuse_node_path(ps, targ));
2409 #endif
2410 	ps->ps_destroy_msg(pm);
2411 	error = 0;
2412 
2413 out:
2414 	node_rele(opc);
2415 	node_rele(targ);
2416 	return error;
2417 }
2418 
2419 /* vap is unused */
2420 /* ARGSUSED4 */
2421 int
2422 perfuse_node_symlink(struct puffs_usermount *pu, puffs_cookie_t opc,
2423 	struct puffs_newinfo *pni, const struct puffs_cn *pcn_src,
2424 	const struct vattr *vap, const char *link_target)
2425 {
2426 	struct perfuse_state *ps;
2427 	perfuse_msg_t *pm;
2428 	char *np;
2429 	const char *path;
2430 	size_t path_len;
2431 	size_t linkname_len;
2432 	size_t len;
2433 	int error;
2434 
2435 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
2436 		return ENOENT;
2437 
2438 	node_ref(opc);
2439 	ps = puffs_getspecific(pu);
2440 	path = pcn_src->pcn_name;
2441 	path_len = pcn_src->pcn_namelen + 1;
2442 	linkname_len = strlen(link_target) + 1;
2443 	len = path_len + linkname_len;
2444 
2445 	pm = ps->ps_new_msg(pu, opc, FUSE_SYMLINK, len, pcn_src->pcn_cred);
2446 	np = _GET_INPAYLOAD(ps, pm, char *);
2447 	(void)strlcpy(np, path, path_len);
2448 	np += path_len;
2449 	(void)strlcpy(np, link_target, linkname_len);
2450 
2451 	error = node_mk_common(pu, opc, pni, pcn_src, pm);
2452 
2453 	node_rele(opc);
2454 	return error;
2455 }
2456 
2457 /* ARGSUSED4 */
2458 int
2459 perfuse_node_readdir(struct puffs_usermount *pu, puffs_cookie_t opc,
2460 	struct dirent *dent, off_t *readoff, size_t *reslen,
2461 	const struct puffs_cred *pcr, int *eofflag, off_t *cookies,
2462 	size_t *ncookies)
2463 {
2464 	perfuse_msg_t *pm;
2465 	uint64_t fh;
2466 	struct perfuse_state *ps;
2467 	struct perfuse_node_data *pnd;
2468 	struct fuse_read_in *fri;
2469 	struct fuse_out_header *foh;
2470 	struct fuse_dirent *fd;
2471 	size_t foh_len;
2472 	int error;
2473 	size_t fd_maxlen;
2474 
2475 	error = 0;
2476 	node_ref(opc);
2477 	ps = puffs_getspecific(pu);
2478 
2479 	/*
2480 	 * readdir state is kept at node level, and several readdir
2481 	 * requests can be issued at the same time on the same node.
2482 	 * We need to queue requests so that only one is in readdir
2483 	 * code at the same time.
2484 	 */
2485 	pnd = PERFUSE_NODE_DATA(opc);
2486 	while (pnd->pnd_flags & PND_INREADDIR)
2487 		requeue_request(pu, opc, PCQ_READDIR);
2488 	pnd->pnd_flags |= PND_INREADDIR;
2489 
2490 #ifdef PERFUSE_DEBUG
2491 	if (perfuse_diagflags & PDF_READDIR)
2492 		DPRINTF("%s: READDIR opc = %p enter critical section\n",
2493 			__func__, (void *)opc);
2494 #endif
2495 	/*
2496 	 * Re-initialize pnd->pnd_fd_cookie on the first readdir for a node
2497 	 */
2498 	if (*readoff == 0)
2499 		pnd->pnd_fd_cookie = 0;
2500 
2501 	/*
2502 	 * Do we already have the data bufered?
2503 	 */
2504 	if (pnd->pnd_dirent != NULL)
2505 		goto out;
2506 	pnd->pnd_dirent_len = 0;
2507 
2508 	/*
2509 	 * It seems NetBSD can call readdir without open first
2510 	 * libfuse will crash if it is done that way, hence open first.
2511 	 */
2512 	if (!(pnd->pnd_flags & PND_OPEN)) {
2513 		if ((error = perfuse_node_open(pu, opc, FREAD, pcr)) != 0)
2514 			goto out;
2515 	}
2516 
2517 	fh = perfuse_get_fh(opc, FREAD);
2518 
2519 #ifdef PERFUSE_DEBUG
2520 	if (perfuse_diagflags & PDF_FH)
2521 		DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", "
2522 			"rfh = 0x%"PRIx64"\n", __func__, (void *)opc,
2523 			PERFUSE_NODE_DATA(opc)->pnd_nodeid, fh);
2524 #endif
2525 
2526 	pnd->pnd_all_fd = NULL;
2527 	pnd->pnd_all_fd_len = 0;
2528 	fd_maxlen = ps->ps_max_readahead - sizeof(*foh);
2529 
2530 	do {
2531 		size_t fd_len;
2532 		char *afdp;
2533 
2534 		pm = ps->ps_new_msg(pu, opc, FUSE_READDIR, sizeof(*fri), pcr);
2535 
2536 		/*
2537 		 * read_flags, lock_owner and flags are unused in libfuse
2538 		 */
2539 		fri = GET_INPAYLOAD(ps, pm, fuse_read_in);
2540 		fri->fh = fh;
2541 		fri->offset = pnd->pnd_fd_cookie;
2542 		fri->size = (uint32_t)fd_maxlen;
2543 		fri->read_flags = 0;
2544 		fri->lock_owner = 0;
2545 		fri->flags = 0;
2546 
2547 		if ((error = xchg_msg(pu, opc, pm,
2548 				      UNSPEC_REPLY_LEN, wait_reply)) != 0)
2549 			goto out;
2550 
2551 		/*
2552 		 * There are many puffs_framebufs calls later,
2553 		 * therefore foh will not be valid for a long time.
2554 		 * Just get the length and forget it.
2555 		 */
2556 		foh = GET_OUTHDR(ps, pm);
2557 		foh_len = foh->len;
2558 
2559 		/*
2560 		 * Empty read: we reached the end of the buffer.
2561 		 */
2562 		if (foh_len == sizeof(*foh)) {
2563 			ps->ps_destroy_msg(pm);
2564 			*eofflag = 1;
2565 			break;
2566 		}
2567 
2568 		/*
2569 		 * Check for corrupted message.
2570 		 */
2571 		if (foh_len < sizeof(*foh) + sizeof(*fd)) {
2572 			ps->ps_destroy_msg(pm);
2573 			DWARNX("readdir reply too short");
2574 			error = EIO;
2575 			goto out;
2576 		}
2577 
2578 
2579 		fd = GET_OUTPAYLOAD(ps, pm, fuse_dirent);
2580 		fd_len = foh_len - sizeof(*foh);
2581 
2582 		pnd->pnd_all_fd = realloc(pnd->pnd_all_fd,
2583 					  pnd->pnd_all_fd_len + fd_len);
2584 		if (pnd->pnd_all_fd  == NULL)
2585 			DERR(EX_OSERR, "%s: malloc failed", __func__);
2586 
2587 		afdp = (char *)(void *)pnd->pnd_all_fd + pnd->pnd_all_fd_len;
2588 		(void)memcpy(afdp, fd, fd_len);
2589 
2590 		pnd->pnd_all_fd_len += fd_len;
2591 
2592 		/*
2593 		 * The fd->off field is used as a cookie for
2594 		 * resuming the next readdir() where this one was left.
2595 	 	 */
2596 		pnd->pnd_fd_cookie = readdir_last_cookie(fd, fd_len);
2597 
2598 		ps->ps_destroy_msg(pm);
2599 	} while (1 /* CONSTCOND */);
2600 
2601 	if (pnd->pnd_all_fd != NULL) {
2602 		if (fuse_to_dirent(pu, opc, pnd->pnd_all_fd,
2603 				   pnd->pnd_all_fd_len) == -1)
2604 			error = EIO;
2605 	}
2606 
2607 out:
2608 	if (pnd->pnd_all_fd != NULL) {
2609 		free(pnd->pnd_all_fd);
2610 		pnd->pnd_all_fd = NULL;
2611 		pnd->pnd_all_fd_len = 0;
2612 	}
2613 
2614 	if (error == 0)
2615 		readdir_buffered(opc, dent, readoff, reslen);
2616 
2617 	/*
2618 	 * Schedule queued readdir requests
2619 	 */
2620 	pnd->pnd_flags &= ~PND_INREADDIR;
2621 	(void)dequeue_requests(opc, PCQ_READDIR, DEQUEUE_ALL);
2622 
2623 #ifdef PERFUSE_DEBUG
2624 	if (perfuse_diagflags & PDF_READDIR)
2625 		DPRINTF("%s: READDIR opc = %p exit critical section\n",
2626 			__func__, (void *)opc);
2627 #endif
2628 
2629 	node_rele(opc);
2630 	return error;
2631 }
2632 
2633 int
2634 perfuse_node_readlink(struct puffs_usermount *pu, puffs_cookie_t opc,
2635 	const struct puffs_cred *pcr, char *linkname, size_t *linklen)
2636 {
2637 	struct perfuse_state *ps;
2638 	perfuse_msg_t *pm;
2639 	int error;
2640 	size_t len;
2641 	struct fuse_out_header *foh;
2642 
2643 	if (PERFUSE_NODE_DATA(opc)->pnd_flags & PND_REMOVED)
2644 		return ENOENT;
2645 
2646 	node_ref(opc);
2647 	ps = puffs_getspecific(pu);
2648 
2649 	pm = ps->ps_new_msg(pu, opc, FUSE_READLINK, 0, pcr);
2650 
2651 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
2652 		goto out;
2653 
2654 	foh = GET_OUTHDR(ps, pm);
2655 	len = foh->len - sizeof(*foh);
2656 	if (len > *linklen)
2657 		DERRX(EX_PROTOCOL, "path len = %zd too long", len);
2658 	if (len == 0)
2659 		DERRX(EX_PROTOCOL, "path len = %zd too short", len);
2660 
2661 	(void)memcpy(linkname, _GET_OUTPAYLOAD(ps, pm, char *), len);
2662 
2663 	/*
2664 	 * FUSE filesystems return a NUL terminated string, we
2665 	 * do not want the trailing \0
2666 	 */
2667 	while (len > 0 && linkname[len - 1] == '\0')
2668 		len--;
2669 
2670 	*linklen = len;
2671 
2672 	ps->ps_destroy_msg(pm);
2673 	error = 0;
2674 
2675 out:
2676 	node_rele(opc);
2677 	return error;
2678 }
2679 
2680 int
2681 perfuse_node_reclaim(struct puffs_usermount *pu, puffs_cookie_t opc)
2682 {
2683 	struct perfuse_state *ps;
2684 	perfuse_msg_t *pm;
2685 	struct perfuse_node_data *pnd;
2686 	struct fuse_forget_in *ffi;
2687 	int nlookup;
2688 	struct timespec now;
2689 
2690 	if (opc == 0)
2691 		return 0;
2692 
2693 	ps = puffs_getspecific(pu);
2694 	pnd = PERFUSE_NODE_DATA(opc);
2695 
2696 	/*
2697 	 * Never forget the root.
2698 	 */
2699 	if (pnd->pnd_nodeid == FUSE_ROOT_ID)
2700 		return 0;
2701 
2702 	/*
2703 	 * There is a race condition between reclaim and lookup.
2704 	 * When looking up an already known node, the kernel cannot
2705 	 * hold a reference on the result until it gets the PUFFS
2706 	 * reply. It mayy therefore reclaim the node after the
2707 	 * userland looked it up, and before it gets the reply.
2708 	 * On rely, the kernel re-creates the node, but at that
2709 	 * time the node has been reclaimed in userland.
2710 	 *
2711 	 * In order to avoid this, we refuse reclaiming nodes that
2712 	 * are too young since the last lookup - and that we do
2713 	 * not have removed on our own, of course.
2714 	 */
2715 	if (clock_gettime(CLOCK_REALTIME, &now) != 0)
2716 		DERR(EX_OSERR, "clock_gettime failed");
2717 
2718 	if (timespeccmp(&pnd->pnd_cn_expire, &now, >) &&
2719 	    !(pnd->pnd_flags & PND_REMOVED)) {
2720 		if (!(pnd->pnd_flags & PND_NODELEAK)) {
2721 			ps->ps_nodeleakcount++;
2722 			pnd->pnd_flags |= PND_NODELEAK;
2723 		}
2724 		DWARNX("possible leaked node:: opc = %p \"%s\"",
2725 		       opc, pnd->pnd_name);
2726 		return 0;
2727 	}
2728 
2729 	node_ref(opc);
2730 	pnd->pnd_flags |= PND_RECLAIMED;
2731 	pnd->pnd_puffs_nlookup--;
2732 	nlookup = pnd->pnd_puffs_nlookup;
2733 
2734 #ifdef PERFUSE_DEBUG
2735 	if (perfuse_diagflags & PDF_RECLAIM)
2736 		DPRINTF("%s (nodeid %"PRId64") reclaimed\n",
2737 			perfuse_node_path(ps, opc), pnd->pnd_nodeid);
2738 #endif
2739 
2740 #ifdef PERFUSE_DEBUG
2741 	if (perfuse_diagflags & PDF_RECLAIM)
2742 		DPRINTF("%s (nodeid %"PRId64") is %sreclaimed, nlookup = %d "
2743 			"%s%s%s%s, pending ops:%s%s%s\n",
2744 		        perfuse_node_path(ps, opc), pnd->pnd_nodeid,
2745 		        pnd->pnd_flags & PND_RECLAIMED ? "" : "not ",
2746 			pnd->pnd_puffs_nlookup,
2747 			pnd->pnd_flags & PND_OPEN ? "open " : "not open",
2748 			pnd->pnd_flags & PND_RFH ? "r" : "",
2749 			pnd->pnd_flags & PND_WFH ? "w" : "",
2750 			pnd->pnd_flags & PND_BUSY ? "" : " none",
2751 			pnd->pnd_flags & PND_INREADDIR ? " readdir" : "",
2752 			pnd->pnd_flags & PND_INWRITE ? " write" : "",
2753 			pnd->pnd_flags & PND_INOPEN ? " open" : "");
2754 #endif
2755 	/*
2756 	 * Make sure it is not looked up again
2757 	 */
2758 	if (!(pnd->pnd_flags & PND_REMOVED))
2759 		perfuse_cache_flush(opc);
2760 
2761 	/*
2762 	 * Purge any activity on the node, while checking
2763 	 * that it remains eligible for a reclaim.
2764 	 */
2765 	while (pnd->pnd_ref > 1)
2766 		requeue_request(pu, opc, PCQ_REF);
2767 
2768 	/*
2769 	 * reclaim cancel?
2770 	 */
2771 	if (pnd->pnd_puffs_nlookup > nlookup) {
2772 		pnd->pnd_flags &= ~PND_RECLAIMED;
2773 		perfuse_node_cache(ps, opc);
2774 		node_rele(opc);
2775 		return 0;
2776 	}
2777 
2778 
2779 #ifdef PERFUSE_DEBUG
2780 	if ((pnd->pnd_flags & PND_OPEN) ||
2781 	       !TAILQ_EMPTY(&pnd->pnd_pcq))
2782 		DERRX(EX_SOFTWARE, "%s: opc = %p \"%s\": still open",
2783 		      __func__, opc, pnd->pnd_name);
2784 
2785 	if ((pnd->pnd_flags & PND_BUSY) ||
2786 	       !TAILQ_EMPTY(&pnd->pnd_pcq))
2787 		DERRX(EX_SOFTWARE, "%s: opc = %p: queued operations",
2788 		      __func__, opc);
2789 
2790 	if (pnd->pnd_inxchg != 0)
2791 		DERRX(EX_SOFTWARE, "%s: opc = %p: ongoing operations",
2792 		      __func__, opc);
2793 #endif
2794 
2795 	/*
2796 	 * Send the FORGET message
2797 	 *
2798 	 * ps_new_msg() is called with NULL creds, which will
2799 	 * be interpreted as FUSE superuser. This is obviously
2800 	 * fine since we operate with kernel creds here.
2801 	 */
2802 	pm = ps->ps_new_msg(pu, opc, FUSE_FORGET,
2803 		      sizeof(*ffi), NULL);
2804 	ffi = GET_INPAYLOAD(ps, pm, fuse_forget_in);
2805 	ffi->nlookup = pnd->pnd_fuse_nlookup;
2806 
2807 	/*
2808 	 * No reply is expected, pm is freed in xchg_msg
2809 	 */
2810 	(void)xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, no_reply);
2811 
2812 	perfuse_destroy_pn(pu, opc);
2813 
2814 	return 0;
2815 }
2816 
2817 int
2818 perfuse_node_inactive(struct puffs_usermount *pu, puffs_cookie_t opc)
2819 {
2820 	struct perfuse_node_data *pnd;
2821 	int error;
2822 
2823 	if (opc == 0)
2824 		return 0;
2825 
2826 	pnd = PERFUSE_NODE_DATA(opc);
2827 	if (!(pnd->pnd_flags & (PND_OPEN|PND_REMOVED)))
2828 		return 0;
2829 
2830 	node_ref(opc);
2831 
2832 	/*
2833 	 * Make sure all operation are finished
2834 	 * There can be an ongoing write. Other
2835 	 * operation wait for all data before
2836 	 * the close/inactive.
2837 	 */
2838 	while (pnd->pnd_flags & PND_INWRITE)
2839 		requeue_request(pu, opc, PCQ_AFTERWRITE);
2840 
2841 	/*
2842 	 * The inactive operation may be cancelled,
2843 	 * If no open is in progress, set PND_INOPEN
2844 	 * so that a new open will be queued.
2845 	 */
2846 	if (pnd->pnd_flags & PND_INOPEN)
2847 		goto out;
2848 
2849 	pnd->pnd_flags |= PND_INOPEN;
2850 
2851 	/*
2852 	 * Sync data
2853 	 */
2854 	if (pnd->pnd_flags & PND_DIRTY) {
2855 		if ((error = perfuse_node_fsync(pu, opc, NULL, 0, 0, 0)) != 0)
2856 			DWARN("%s: perfuse_node_fsync failed error = %d",
2857 			      __func__, error);
2858 	}
2859 
2860 
2861 	/*
2862 	 * Close handles
2863 	 */
2864 	if (pnd->pnd_flags & PND_WFH) {
2865 		if ((error = perfuse_node_close_common(pu, opc, FWRITE)) != 0)
2866 			DWARN("%s: close write FH failed error = %d",
2867 			      __func__, error);
2868 	}
2869 
2870 	if (pnd->pnd_flags & PND_RFH) {
2871 		if ((error = perfuse_node_close_common(pu, opc, FREAD)) != 0)
2872 			DWARN("%s: close read FH failed error = %d",
2873 			      __func__, error);
2874 	}
2875 
2876 	/*
2877 	 * This will cause a reclaim to be sent
2878 	 */
2879 	if (pnd->pnd_flags & PND_REMOVED)
2880 		puffs_setback(puffs_cc_getcc(pu), PUFFS_SETBACK_NOREF_N1);
2881 
2882 	/*
2883 	 * Schedule awaiting operations
2884 	 */
2885 	pnd->pnd_flags &= ~PND_INOPEN;
2886 	(void)dequeue_requests(opc, PCQ_OPEN, DEQUEUE_ALL);
2887 
2888 	/*
2889 	 * errors are ignored, since the kernel ignores the return code.
2890 	 */
2891 out:
2892 	node_rele(opc);
2893 	return 0;
2894 }
2895 
2896 
2897 /* ARGSUSED0 */
2898 int
2899 perfuse_node_print(struct puffs_usermount *pu, puffs_cookie_t opc)
2900 {
2901 	DERRX(EX_SOFTWARE, "%s: UNIMPLEMENTED (FATAL)", __func__);
2902 	return 0;
2903 }
2904 
2905 int
2906 perfuse_node_pathconf(struct puffs_usermount *pu, puffs_cookie_t opc,
2907 	int name, register_t *retval)
2908 {
2909 	perfuse_msg_t *pm;
2910 	struct perfuse_state *ps;
2911 	struct fuse_statfs_out *fso;
2912 	int error = 0;
2913 
2914 	/*
2915 	 * Static values copied from UFS
2916 	 * in src/sys/ufs/ufs/ufs_vnops.c
2917 	 */
2918 	switch (name) {
2919 	case _PC_LINK_MAX:
2920 		*retval = LINK_MAX;
2921 		break;
2922 	case _PC_PATH_MAX:
2923 		*retval = PATH_MAX;
2924 		break;
2925 	case _PC_PIPE_BUF:
2926 		*retval = PIPE_BUF;
2927 		break;
2928 	case _PC_CHOWN_RESTRICTED:
2929 		*retval = 1;
2930 		break;
2931 	case _PC_NO_TRUNC:
2932 		*retval = 1;
2933 		break;
2934 	case _PC_SYNC_IO:
2935 		*retval = 1;
2936 		break;
2937 	case _PC_FILESIZEBITS:
2938 		*retval = 42;
2939 		break;
2940 	case _PC_SYMLINK_MAX:
2941 		*retval = MAXPATHLEN;
2942 		break;
2943 	case _PC_2_SYMLINKS:
2944 		*retval = 1;
2945 		break;
2946 	case _PC_NAME_MAX:
2947 		ps = puffs_getspecific(pu);
2948 		pm = ps->ps_new_msg(pu, opc, FUSE_STATFS, 0, NULL);
2949 
2950 		error = xchg_msg(pu, opc, pm, sizeof(*fso), wait_reply);
2951 		if (error != 0)
2952 			return error;
2953 
2954 		fso = GET_OUTPAYLOAD(ps, pm, fuse_statfs_out);
2955 		*retval = fso->st.namelen;
2956 
2957 		ps->ps_destroy_msg(pm);
2958 
2959 		break;
2960 	default:
2961 		DWARN("Unimplemented pathconf for name = %d", name);
2962 		error = ENOSYS;
2963 		break;
2964 	}
2965 
2966 	return error;
2967 }
2968 
2969 int
2970 perfuse_node_advlock(struct puffs_usermount *pu, puffs_cookie_t opc,
2971 	void *id, int op, struct flock *fl, int flags)
2972 {
2973 	struct perfuse_state *ps;
2974 	int fop;
2975 	perfuse_msg_t *pm;
2976 	uint64_t fh;
2977 	struct fuse_lk_in *fli;
2978 	struct fuse_out_header *foh;
2979 	struct fuse_lk_out *flo;
2980 	uint32_t owner;
2981 	size_t len;
2982 	int error;
2983 
2984 	node_ref(opc);
2985 
2986 	/*
2987 	 * Make sure we do have a filehandle, as the FUSE filesystem
2988 	 * expect one. E.g.: if we provide none, GlusterFS logs an error
2989 	 * "0-glusterfs-fuse: xl is NULL"
2990 	 *
2991 	 * We need the read file handle if the file is open read only,
2992 	 * in order to support shared locks on read-only files.
2993 	 * NB: The kernel always sends advlock for read-only
2994 	 * files at exit time when the process used lock, see
2995 	 * sys_exit -> exit1 -> fd_free -> fd_close -> VOP_ADVLOCK
2996 	 */
2997 	if ((fh = perfuse_get_fh(opc, FREAD)) == FUSE_UNKNOWN_FH) {
2998 		error = EBADF;
2999 		goto out;
3000 	}
3001 
3002 	ps = puffs_getspecific(pu);
3003 
3004 	if (op == F_GETLK)
3005 		fop = FUSE_GETLK;
3006 	else
3007 		fop = (flags & F_WAIT) ? FUSE_SETLKW : FUSE_SETLK;
3008 
3009 	/*
3010 	 * XXX ps_new_msg() is called with NULL creds, which will
3011 	 * be interpreted as FUSE superuser. We have no way to
3012 	 * know the requesting process' credential, but since advlock()
3013 	 * is supposed to operate on a file that has been open(),
3014 	 * permission should have already been checked at open() time.
3015 	 */
3016 	pm = ps->ps_new_msg(pu, opc, fop, sizeof(*fli), NULL);
3017 	fli = GET_INPAYLOAD(ps, pm, fuse_lk_in);
3018 	fli->fh = fh;
3019 	fli->owner = (uint64_t)(vaddr_t)id;
3020 	fli->lk.start = fl->l_start;
3021 	fli->lk.end = fl->l_start + fl->l_len;
3022 	fli->lk.type = fl->l_type;
3023 	fli->lk.pid = fl->l_pid;
3024 	fli->lk_flags = (flags & F_FLOCK) ? FUSE_LK_FLOCK : 0;
3025 
3026 	owner = (uint32_t)(vaddr_t)id;
3027 
3028 #ifdef PERFUSE_DEBUG
3029 	if (perfuse_diagflags & PDF_FH)
3030 		DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", fh = 0x%"PRIx64"\n",
3031 			__func__, (void *)opc,
3032 			PERFUSE_NODE_DATA(opc)->pnd_nodeid, fli->fh);
3033 #endif
3034 
3035 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
3036 		goto out;
3037 
3038 	foh = GET_OUTHDR(ps, pm);
3039 	len = foh->len - sizeof(*foh);
3040 
3041 	/*
3042 	 * Save or clear the lock
3043 	 */
3044 	switch (op) {
3045 	case F_GETLK:
3046 		if (len != sizeof(*flo))
3047 			DERRX(EX_SOFTWARE,
3048 			      "%s: Unexpected lock reply len %zd",
3049 			      __func__, len);
3050 
3051 		flo = GET_OUTPAYLOAD(ps, pm, fuse_lk_out);
3052 		fl->l_start = flo->lk.start;
3053 		fl->l_len = flo->lk.end - flo->lk.start;
3054 		fl->l_pid = flo->lk.pid;
3055 		fl->l_type = flo->lk.type;
3056 		fl->l_whence = SEEK_SET;	/* libfuse hardcodes it */
3057 
3058 		PERFUSE_NODE_DATA(opc)->pnd_lock_owner = flo->lk.pid;
3059 		break;
3060 	case F_UNLCK:
3061 		owner = 0;
3062 		/* FALLTHROUGH */
3063 	case F_SETLK:
3064 		/* FALLTHROUGH */
3065 	case F_SETLKW:
3066 		if (error != 0)
3067 			PERFUSE_NODE_DATA(opc)->pnd_lock_owner = owner;
3068 
3069 		if (len != 0)
3070 			DERRX(EX_SOFTWARE,
3071 			      "%s: Unexpected unlock reply len %zd",
3072 			      __func__, len);
3073 
3074 		break;
3075 	default:
3076 		DERRX(EX_SOFTWARE, "%s: Unexpected op %d", __func__, op);
3077 		break;
3078 	}
3079 
3080 	ps->ps_destroy_msg(pm);
3081 	error = 0;
3082 
3083 out:
3084 	node_rele(opc);
3085 	return error;
3086 }
3087 
3088 int
3089 perfuse_node_read(struct puffs_usermount *pu, puffs_cookie_t opc, uint8_t *buf,
3090 	off_t offset, size_t *resid, const struct puffs_cred *pcr, int ioflag)
3091 {
3092 	struct perfuse_state *ps;
3093 	struct perfuse_node_data *pnd;
3094 	const struct vattr *vap;
3095 	perfuse_msg_t *pm;
3096 	struct fuse_read_in *fri;
3097 	struct fuse_out_header *foh;
3098 	size_t readen;
3099 	int error;
3100 
3101 	ps = puffs_getspecific(pu);
3102 	pnd = PERFUSE_NODE_DATA(opc);
3103 	vap = puffs_pn_getvap((struct puffs_node *)opc);
3104 
3105 	/*
3106 	 * NetBSD turns that into a getdents(2) output
3107 	 * We just do a EISDIR as this feature is of little use.
3108 	 */
3109 	if (vap->va_type == VDIR)
3110 		return EISDIR;
3111 
3112 	do {
3113 		size_t max_read;
3114 
3115 		max_read = ps->ps_max_readahead - sizeof(*foh);
3116 		/*
3117 		 * flags may be set to FUSE_READ_LOCKOWNER
3118 		 * if lock_owner is provided.
3119 		 */
3120 		pm = ps->ps_new_msg(pu, opc, FUSE_READ, sizeof(*fri), pcr);
3121 		fri = GET_INPAYLOAD(ps, pm, fuse_read_in);
3122 		fri->fh = perfuse_get_fh(opc, FREAD);
3123 		fri->offset = offset;
3124 		fri->size = (uint32_t)MIN(*resid, max_read);
3125 		fri->read_flags = 0; /* XXX Unused by libfuse? */
3126 		fri->lock_owner = pnd->pnd_lock_owner;
3127 		fri->flags = 0;
3128 		fri->flags |= (fri->lock_owner != 0) ? FUSE_READ_LOCKOWNER : 0;
3129 
3130 #ifdef PERFUSE_DEBUG
3131 	if (perfuse_diagflags & PDF_FH)
3132 		DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", fh = 0x%"PRIx64"\n",
3133 			__func__, (void *)opc, pnd->pnd_nodeid, fri->fh);
3134 #endif
3135 		error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply);
3136 		if (error  != 0)
3137 			return error;
3138 
3139 		foh = GET_OUTHDR(ps, pm);
3140 		readen = foh->len - sizeof(*foh);
3141 
3142 #ifdef PERFUSE_DEBUG
3143 		if (readen > *resid)
3144 			DERRX(EX_SOFTWARE, "%s: Unexpected big read %zd",
3145 			      __func__, readen);
3146 #endif
3147 
3148 		(void)memcpy(buf,  _GET_OUTPAYLOAD(ps, pm, char *), readen);
3149 
3150 		buf += readen;
3151 		offset += readen;
3152 		*resid -= readen;
3153 
3154 		ps->ps_destroy_msg(pm);
3155 	} while ((*resid != 0) && (readen != 0));
3156 
3157 	if (ioflag & (IO_SYNC|IO_DSYNC))
3158 		ps->ps_syncreads++;
3159 	else
3160 		ps->ps_asyncreads++;
3161 
3162 	return 0;
3163 }
3164 
3165 int
3166 perfuse_node_write(struct puffs_usermount *pu, puffs_cookie_t opc,
3167 	uint8_t *buf, off_t offset, size_t *resid,
3168 	const struct puffs_cred *pcr, int ioflag)
3169 {
3170 	return perfuse_node_write2(pu, opc, buf, offset, resid, pcr, ioflag, 0);
3171 }
3172 
3173 /* ARGSUSED7 */
3174 int
3175 perfuse_node_write2(struct puffs_usermount *pu, puffs_cookie_t opc,
3176 	uint8_t *buf, off_t offset, size_t *resid,
3177 	const struct puffs_cred *pcr, int ioflag, int xflag)
3178 {
3179 	struct perfuse_state *ps;
3180 	struct perfuse_node_data *pnd;
3181 	struct vattr *vap;
3182 	perfuse_msg_t *pm;
3183 	struct fuse_write_in *fwi;
3184 	struct fuse_write_out *fwo;
3185 	size_t data_len;
3186 	size_t payload_len;
3187 	size_t written;
3188 	int inresize;
3189 	int error;
3190 
3191 	ps = puffs_getspecific(pu);
3192 	pnd = PERFUSE_NODE_DATA(opc);
3193 	vap = puffs_pn_getvap((struct puffs_node *)opc);
3194 	written = 0;
3195 	inresize = 0;
3196 	error = 0;
3197 
3198 	if (vap->va_type == VDIR)
3199 		return EISDIR;
3200 
3201 	node_ref(opc);
3202 
3203 	/*
3204 	 * We need to queue write requests in order to avoid
3205 	 * dequeueing PCQ_AFTERWRITE when there are pending writes.
3206 	 */
3207 	while (pnd->pnd_flags & PND_INWRITE)
3208 		requeue_request(pu, opc, PCQ_WRITE);
3209 	pnd->pnd_flags |= PND_INWRITE;
3210 
3211 	/*
3212 	 * append flag: re-read the file size so that
3213 	 * we get the latest value.
3214 	 */
3215 	if (ioflag & PUFFS_IO_APPEND) {
3216 		if ((error = perfuse_node_getattr(pu, opc, vap, pcr)) != 0)
3217 			goto out;
3218 
3219 		offset = vap->va_size;
3220 	}
3221 
3222 	/*
3223 	 * Serialize size access, see comment in perfuse_node_setattr().
3224 	 */
3225 	if ((u_quad_t)offset + *resid > vap->va_size) {
3226 		while (pnd->pnd_flags & PND_INRESIZE)
3227 			requeue_request(pu, opc, PCQ_RESIZE);
3228 		pnd->pnd_flags |= PND_INRESIZE;
3229 		inresize = 1;
3230 	}
3231 
3232 #ifdef PERFUSE_DEBUG
3233 	if (perfuse_diagflags & PDF_RESIZE)
3234 		DPRINTF(">> %s %p %" PRIu64 "\n", __func__,
3235 			(void *)opc, vap->va_size);
3236 #endif
3237 
3238 	do {
3239 		size_t max_write;
3240 		/*
3241 		 * There is a writepage flag when data
3242 		 * is aligned to page size. Use it for
3243 		 * everything but the data after the last
3244 		 * page boundary.
3245 		 */
3246 		max_write = ps->ps_max_write - sizeof(*fwi);
3247 
3248 		data_len = MIN(*resid, max_write);
3249 		if (data_len > (size_t)sysconf(_SC_PAGESIZE))
3250 			data_len = data_len & ~(sysconf(_SC_PAGESIZE) - 1);
3251 
3252 		payload_len = data_len + sizeof(*fwi);
3253 
3254 		/*
3255 		 * flags may be set to FUSE_WRITE_CACHE (XXX usage?)
3256 		 * or FUSE_WRITE_LOCKOWNER, if lock_owner is provided.
3257 		 * write_flags is set to 1 for writepage.
3258 		 */
3259 		pm = ps->ps_new_msg(pu, opc, FUSE_WRITE, payload_len, pcr);
3260 		fwi = GET_INPAYLOAD(ps, pm, fuse_write_in);
3261 		fwi->fh = perfuse_get_fh(opc, FWRITE);
3262 		fwi->offset = offset;
3263 		fwi->size = (uint32_t)data_len;
3264 		fwi->write_flags = (fwi->size % sysconf(_SC_PAGESIZE)) ? 0 : 1;
3265 		fwi->lock_owner = pnd->pnd_lock_owner;
3266 		fwi->flags = 0;
3267 		fwi->flags |= (fwi->lock_owner != 0) ? FUSE_WRITE_LOCKOWNER : 0;
3268 		fwi->flags |= (ioflag & IO_DIRECT) ? 0 : FUSE_WRITE_CACHE;
3269 		(void)memcpy((fwi + 1), buf, data_len);
3270 
3271 
3272 #ifdef PERFUSE_DEBUG
3273 		if (perfuse_diagflags & PDF_FH)
3274 			DPRINTF("%s: opc = %p, nodeid = 0x%"PRIx64", "
3275 				"fh = 0x%"PRIx64"\n", __func__,
3276 				(void *)opc, pnd->pnd_nodeid, fwi->fh);
3277 #endif
3278 		if ((error = xchg_msg(pu, opc, pm,
3279 				      sizeof(*fwo), wait_reply)) != 0)
3280 			goto out;
3281 
3282 		fwo = GET_OUTPAYLOAD(ps, pm, fuse_write_out);
3283 		written = fwo->size;
3284 		ps->ps_destroy_msg(pm);
3285 
3286 #ifdef PERFUSE_DEBUG
3287 		if (written > *resid)
3288 			DERRX(EX_SOFTWARE, "%s: Unexpected big write %zd",
3289 			      __func__, written);
3290 #endif
3291 		*resid -= written;
3292 		offset += written;
3293 		buf += written;
3294 
3295 	} while (*resid != 0);
3296 
3297 	/*
3298 	 * puffs_ops(3) says
3299 	 *  "everything must be written or an error will be generated"
3300 	 */
3301 	if (*resid != 0)
3302 		error = EFBIG;
3303 
3304 out:
3305 #ifdef PERFUSE_DEBUG
3306 	if (perfuse_diagflags & PDF_RESIZE) {
3307 		if (offset > (off_t)vap->va_size)
3308 			DPRINTF("<< %s %p %" PRIu64 " -> %lld\n", __func__,
3309 				(void *)opc, vap->va_size, (long long)offset);
3310 		else
3311 			DPRINTF("<< %s %p \n", __func__, (void *)opc);
3312 	}
3313 #endif
3314 
3315 	/*
3316 	 * Update file size if we wrote beyond the end
3317 	 */
3318 	if (offset > (off_t)vap->va_size)
3319 		vap->va_size = offset;
3320 
3321 	/*
3322 	 * Statistics
3323 	 */
3324 	if (ioflag & (IO_SYNC|IO_DSYNC))
3325 		ps->ps_syncwrites++;
3326 	else
3327 		ps->ps_asyncwrites++;
3328 
3329 	/*
3330 	 * Remember to sync the file
3331 	 */
3332 	pnd->pnd_flags |= PND_DIRTY;
3333 
3334 #ifdef PERFUSE_DEBUG
3335 	if (perfuse_diagflags & PDF_SYNC)
3336 		DPRINTF("%s: DIRTY opc = %p, file = \"%s\"\n",
3337 			__func__, (void*)opc, perfuse_node_path(ps, opc));
3338 #endif
3339 
3340 	if (inresize) {
3341 #ifdef PERFUSE_DEBUG
3342 		if (!(pnd->pnd_flags & PND_INRESIZE))
3343 			DERRX(EX_SOFTWARE, "file write grow without resize");
3344 #endif
3345 		pnd->pnd_flags &= ~PND_INRESIZE;
3346 		(void)dequeue_requests(opc, PCQ_RESIZE, DEQUEUE_ALL);
3347 	}
3348 
3349 	/*
3350 	 * VOP_PUTPAGE causes FAF write where kernel does not
3351 	 * check operation result. At least warn if it failed.
3352 	 */
3353 #ifdef PUFFS_WRITE_FAF
3354 	if (error && (xflag & PUFFS_WRITE_FAF))
3355 		DWARN("Data loss caused by FAF write failed on \"%s\"",
3356 		      pnd->pnd_name);
3357 #endif /* PUFFS_WRITE_FAF */
3358 
3359 	/*
3360 	 * If there are no more queued write, we can resume
3361 	 * an operation awaiting write completion.
3362 	 */
3363 	pnd->pnd_flags &= ~PND_INWRITE;
3364 	if (dequeue_requests(opc, PCQ_WRITE, 1) == 0)
3365 		(void)dequeue_requests(opc, PCQ_AFTERWRITE, DEQUEUE_ALL);
3366 
3367 	node_rele(opc);
3368 	return error;
3369 }
3370 
3371 /* ARGSUSED0 */
3372 void
3373 perfuse_cache_write(struct puffs_usermount *pu, puffs_cookie_t opc, size_t size,
3374 	struct puffs_cacherun *runs)
3375 {
3376 	return;
3377 }
3378 
3379 /* ARGSUSED4 */
3380 int
3381 perfuse_node_getextattr(struct puffs_usermount *pu, puffs_cookie_t opc,
3382 	int attrns, const char *attrname, size_t *attrsize, uint8_t *attr,
3383 	size_t *resid, const struct puffs_cred *pcr)
3384 {
3385 	struct perfuse_state *ps;
3386 	char fuse_attrname[LINUX_XATTR_NAME_MAX + 1];
3387 	perfuse_msg_t *pm;
3388 	struct fuse_getxattr_in *fgi;
3389 	struct fuse_getxattr_out *fgo;
3390 	struct fuse_out_header *foh;
3391 	size_t attrnamelen;
3392 	size_t len;
3393 	char *np;
3394 	int error;
3395 
3396 	/* system namespace attrs are not accessible to non root users */
3397 	if (attrns == EXTATTR_NAMESPACE_SYSTEM && !puffs_cred_isjuggernaut(pcr))
3398 		return EPERM;
3399 
3400 	node_ref(opc);
3401 	ps = puffs_getspecific(pu);
3402 	attrname = perfuse_native_ns(attrns, attrname, fuse_attrname);
3403 	attrnamelen = strlen(attrname) + 1;
3404 	len = sizeof(*fgi) + attrnamelen;
3405 
3406 	pm = ps->ps_new_msg(pu, opc, FUSE_GETXATTR, len, pcr);
3407 	fgi = GET_INPAYLOAD(ps, pm, fuse_getxattr_in);
3408 	fgi->size = (unsigned int)((resid != NULL) ? *resid : 0);
3409 	np = (char *)(void *)(fgi + 1);
3410 	(void)strlcpy(np, attrname, attrnamelen);
3411 
3412 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
3413 		goto out;
3414 
3415 	/*
3416 	 * We just get fuse_getattr_out with list size if we requested
3417 	 * a null size.
3418 	 */
3419 	if (resid == NULL) {
3420 		fgo = GET_OUTPAYLOAD(ps, pm, fuse_getxattr_out);
3421 
3422 		if (attrsize != NULL)
3423 			*attrsize = fgo->size;
3424 
3425 		ps->ps_destroy_msg(pm);
3426 		error = 0;
3427 		goto out;
3428 	}
3429 
3430 	/*
3431 	 * And with a non null requested size, we get the list just
3432 	 * after the header
3433 	 */
3434 	foh = GET_OUTHDR(ps, pm);
3435 	np = (char *)(void *)(foh + 1);
3436 	len = foh->len - sizeof(*foh);
3437 
3438 	if (attrsize != NULL)
3439 		*attrsize = len;
3440 
3441 	if (resid != NULL) {
3442 		if (*resid < len) {
3443 			error = ERANGE;
3444 			ps->ps_destroy_msg(pm);
3445 			goto out;
3446 		}
3447 
3448 		(void)memcpy(attr, np, len);
3449 		*resid -= len;
3450 	}
3451 
3452 	ps->ps_destroy_msg(pm);
3453 	error = 0;
3454 
3455 out:
3456 	node_rele(opc);
3457 	return error;
3458 }
3459 
3460 int
3461 perfuse_node_setextattr(struct puffs_usermount *pu, puffs_cookie_t opc,
3462 	int attrns, const char *attrname, uint8_t *attr, size_t *resid,
3463 	const struct puffs_cred *pcr)
3464 {
3465 	struct perfuse_state *ps;
3466 	char fuse_attrname[LINUX_XATTR_NAME_MAX + 1];
3467 	perfuse_msg_t *pm;
3468 	struct fuse_setxattr_in *fsi;
3469 	size_t attrnamelen;
3470 	size_t datalen;
3471 	size_t len;
3472 	char *np;
3473 	int error;
3474 
3475 	/* system namespace attrs are not accessible to non root users */
3476 	if (attrns == EXTATTR_NAMESPACE_SYSTEM && !puffs_cred_isjuggernaut(pcr))
3477 		return EPERM;
3478 
3479 	node_ref(opc);
3480 	ps = puffs_getspecific(pu);
3481 	attrname = perfuse_native_ns(attrns, attrname, fuse_attrname);
3482 	attrnamelen = strlen(attrname) + 1;
3483 
3484 	datalen = (resid != NULL) ? *resid : 0;
3485 	len = sizeof(*fsi) + attrnamelen + datalen;
3486 
3487 	pm = ps->ps_new_msg(pu, opc, FUSE_SETXATTR, len, pcr);
3488 	fsi = GET_INPAYLOAD(ps, pm, fuse_setxattr_in);
3489 	fsi->size = (unsigned int)datalen;
3490 	fsi->flags = 0;
3491 	np = (char *)(void *)(fsi + 1);
3492 	(void)strlcpy(np, attrname, attrnamelen);
3493 	np += attrnamelen;
3494 	if (datalen)
3495 		(void)memcpy(np, (char *)attr, datalen);
3496 
3497 	if ((error = xchg_msg(pu, opc, pm,
3498 			      NO_PAYLOAD_REPLY_LEN, wait_reply)) != 0)
3499 		goto out;
3500 
3501 	ps->ps_destroy_msg(pm);
3502 	if (resid)
3503 		*resid = 0;
3504 	error = 0;
3505 
3506 out:
3507 	node_rele(opc);
3508 	return error;
3509 }
3510 
3511 /* ARGSUSED2 */
3512 int
3513 perfuse_node_listextattr(struct puffs_usermount *pu, puffs_cookie_t opc,
3514 	int attrns, size_t *attrsize, uint8_t *attrs, size_t *resid, int flag,
3515 	const struct puffs_cred *pcr)
3516 {
3517 	struct perfuse_state *ps;
3518 	perfuse_msg_t *pm;
3519 	struct fuse_getxattr_in *fgi;
3520 	struct fuse_getxattr_out *fgo;
3521 	struct fuse_out_header *foh;
3522 	char *np;
3523 	size_t len, puffs_len, i, attrlen, outlen;
3524 	int error;
3525 
3526 	/* system namespace attrs are not accessible to non root users */
3527 	if (attrns == EXTATTR_NAMESPACE_SYSTEM && !puffs_cred_isjuggernaut(pcr))
3528 		return EPERM;
3529 
3530 	node_ref(opc);
3531 
3532 	ps = puffs_getspecific(pu);
3533 	len = sizeof(*fgi);
3534 
3535 	pm = ps->ps_new_msg(pu, opc, FUSE_LISTXATTR, len, pcr);
3536 	fgi = GET_INPAYLOAD(ps, pm, fuse_getxattr_in);
3537 	if (resid != NULL)
3538 		fgi->size = (unsigned int)*resid;
3539 	else
3540 		fgi->size = 0;
3541 
3542 	if ((error = xchg_msg(pu, opc, pm, UNSPEC_REPLY_LEN, wait_reply)) != 0)
3543 		goto out;
3544 
3545 	/*
3546 	 * We just get fuse_getattr_out with list size if we requested
3547 	 * a null size.
3548 	 */
3549 	if (resid == NULL) {
3550 		fgo = GET_OUTPAYLOAD(ps, pm, fuse_getxattr_out);
3551 
3552 		if (attrsize != NULL)
3553 			*attrsize = fgo->size;
3554 
3555 		ps->ps_destroy_msg(pm);
3556 
3557 		error = 0;
3558 		goto out;
3559 	}
3560 
3561 	/*
3562 	 * And with a non null requested size, we get the list just
3563 	 * after the header
3564 	 */
3565 	foh = GET_OUTHDR(ps, pm);
3566 	np = (char *)(void *)(foh + 1);
3567 	puffs_len = foh->len - sizeof(*foh);
3568 
3569 	if (attrsize != NULL)
3570 		*attrsize = puffs_len;
3571 
3572 	if (attrs != NULL) {
3573 		if (*resid < puffs_len) {
3574 			error = ERANGE;
3575 			ps->ps_destroy_msg(pm);
3576 			goto out;
3577 		}
3578 
3579 		outlen = 0;
3580 
3581 		for (i = 0; i < puffs_len; i += attrlen + 1) {
3582 			attrlen = strlen(np + i);
3583 
3584 			/*
3585 			 * Filter attributes per namespace
3586 			 */
3587 			if (!perfuse_ns_match(attrns, np + i))
3588 				continue;
3589 
3590 #ifdef PUFFS_EXTATTR_LIST_LENPREFIX
3591 			/*
3592 			 * Convert the FUSE reply to length prefixed strings
3593 			 * if this is what the kernel wants.
3594 			 */
3595 			if (flag & PUFFS_EXTATTR_LIST_LENPREFIX) {
3596 				(void)memcpy(attrs + outlen + 1,
3597 					     np + i, attrlen);
3598 				*(attrs + outlen) = (uint8_t)attrlen;
3599 			} else
3600 #endif /* PUFFS_EXTATTR_LIST_LENPREFIX */
3601 			(void)memcpy(attrs + outlen, np + i, attrlen + 1);
3602 			outlen += attrlen + 1;
3603 		}
3604 
3605 		*resid -= outlen;
3606 	}
3607 
3608 	ps->ps_destroy_msg(pm);
3609 	error = 0;
3610 
3611 out:
3612 	node_rele(opc);
3613 	return error;
3614 }
3615 
3616 int
3617 perfuse_node_deleteextattr(struct puffs_usermount *pu, puffs_cookie_t opc,
3618 	int attrns, const char *attrname, const struct puffs_cred *pcr)
3619 {
3620 	struct perfuse_state *ps;
3621 	char fuse_attrname[LINUX_XATTR_NAME_MAX + 1];
3622 	perfuse_msg_t *pm;
3623 	size_t attrnamelen;
3624 	char *np;
3625 	int error;
3626 
3627 	/* system namespace attrs are not accessible to non root users */
3628 	if (attrns == EXTATTR_NAMESPACE_SYSTEM && !puffs_cred_isjuggernaut(pcr))
3629 		return EPERM;
3630 
3631 	node_ref(opc);
3632 
3633 	ps = puffs_getspecific(pu);
3634 	attrname = perfuse_native_ns(attrns, attrname, fuse_attrname);
3635 	attrnamelen = strlen(attrname) + 1;
3636 
3637 	pm = ps->ps_new_msg(pu, opc, FUSE_REMOVEXATTR, attrnamelen, pcr);
3638 	np = _GET_INPAYLOAD(ps, pm, char *);
3639 	(void)strlcpy(np, attrname, attrnamelen);
3640 
3641 	error = xchg_msg(pu, opc, pm, NO_PAYLOAD_REPLY_LEN, wait_reply);
3642 	if (error != 0)
3643 		goto out;
3644 
3645 	ps->ps_destroy_msg(pm);
3646 
3647 out:
3648 	node_rele(opc);
3649 	return error;
3650 }
3651 
3652 int
3653 perfuse_node_fallocate(struct puffs_usermount *pu, puffs_cookie_t opc,
3654 	off_t off, off_t len)
3655 {
3656 	struct perfuse_state *ps;
3657 	perfuse_msg_t *pm;
3658 	struct fuse_fallocate_in *fai;
3659 	int error;
3660 
3661 	ps = puffs_getspecific(pu);
3662 	if (ps->ps_flags & PS_NO_FALLOCATE)
3663 		return EOPNOTSUPP;
3664 
3665 	node_ref(opc);
3666 
3667 	pm = ps->ps_new_msg(pu, opc, FUSE_FALLOCATE, sizeof(*fai), NULL);
3668 
3669 	fai = GET_INPAYLOAD(ps, pm, fuse_fallocate_in);
3670 	fai->fh = perfuse_get_fh(opc, FWRITE);
3671 	fai->offset = off;
3672 	fai->length = len;
3673 	fai->mode = 0;
3674 
3675 	error = xchg_msg(pu, opc, pm, NO_PAYLOAD_REPLY_LEN, wait_reply);
3676 	if (error == EOPNOTSUPP || error == ENOSYS) {
3677 		ps->ps_flags |= PS_NO_FALLOCATE;
3678 		error = EOPNOTSUPP;
3679 	}
3680 	if (error != 0)
3681 		goto out;
3682 
3683 	ps->ps_destroy_msg(pm);
3684 
3685 out:
3686 	node_rele(opc);
3687 	return error;
3688 }
3689