xref: /freebsd/sys/fs/fuse/fuse_vnops.c (revision 3f2c630c74d2cc9f78b5c92ae48fb3646f12fe96)
151369649SPedro F. Giffuni /*-
251369649SPedro F. Giffuni  * SPDX-License-Identifier: BSD-3-Clause
351369649SPedro F. Giffuni  *
45fe58019SAttilio Rao  * Copyright (c) 2007-2009 Google Inc. and Amit Singh
55fe58019SAttilio Rao  * All rights reserved.
65fe58019SAttilio Rao  *
75fe58019SAttilio Rao  * Redistribution and use in source and binary forms, with or without
85fe58019SAttilio Rao  * modification, are permitted provided that the following conditions are
95fe58019SAttilio Rao  * met:
105fe58019SAttilio Rao  *
115fe58019SAttilio Rao  * * Redistributions of source code must retain the above copyright
125fe58019SAttilio Rao  *   notice, this list of conditions and the following disclaimer.
135fe58019SAttilio Rao  * * Redistributions in binary form must reproduce the above
145fe58019SAttilio Rao  *   copyright notice, this list of conditions and the following disclaimer
155fe58019SAttilio Rao  *   in the documentation and/or other materials provided with the
165fe58019SAttilio Rao  *   distribution.
175fe58019SAttilio Rao  * * Neither the name of Google Inc. nor the names of its
185fe58019SAttilio Rao  *   contributors may be used to endorse or promote products derived from
195fe58019SAttilio Rao  *   this software without specific prior written permission.
205fe58019SAttilio Rao  *
215fe58019SAttilio Rao  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
225fe58019SAttilio Rao  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
235fe58019SAttilio Rao  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
245fe58019SAttilio Rao  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
255fe58019SAttilio Rao  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
265fe58019SAttilio Rao  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
275fe58019SAttilio Rao  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
285fe58019SAttilio Rao  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
295fe58019SAttilio Rao  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
305fe58019SAttilio Rao  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
315fe58019SAttilio Rao  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
325fe58019SAttilio Rao  *
335fe58019SAttilio Rao  * Copyright (C) 2005 Csaba Henk.
345fe58019SAttilio Rao  * All rights reserved.
355fe58019SAttilio Rao  *
365fe58019SAttilio Rao  * Redistribution and use in source and binary forms, with or without
375fe58019SAttilio Rao  * modification, are permitted provided that the following conditions
385fe58019SAttilio Rao  * are met:
395fe58019SAttilio Rao  * 1. Redistributions of source code must retain the above copyright
405fe58019SAttilio Rao  *    notice, this list of conditions and the following disclaimer.
415fe58019SAttilio Rao  * 2. Redistributions in binary form must reproduce the above copyright
425fe58019SAttilio Rao  *    notice, this list of conditions and the following disclaimer in the
435fe58019SAttilio Rao  *    documentation and/or other materials provided with the distribution.
445fe58019SAttilio Rao  *
455fe58019SAttilio Rao  * THIS SOFTWARE IS PROVIDED BY AUTHOR AND CONTRIBUTORS ``AS IS'' AND
465fe58019SAttilio Rao  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
475fe58019SAttilio Rao  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
485fe58019SAttilio Rao  * ARE DISCLAIMED.  IN NO EVENT SHALL AUTHOR OR CONTRIBUTORS BE LIABLE
495fe58019SAttilio Rao  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
505fe58019SAttilio Rao  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
515fe58019SAttilio Rao  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
525fe58019SAttilio Rao  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
535fe58019SAttilio Rao  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
545fe58019SAttilio Rao  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
555fe58019SAttilio Rao  * SUCH DAMAGE.
565fe58019SAttilio Rao  */
575fe58019SAttilio Rao 
585fe58019SAttilio Rao #include <sys/cdefs.h>
595fe58019SAttilio Rao __FBSDID("$FreeBSD$");
605fe58019SAttilio Rao 
61cf169498SAlan Somers #include <sys/param.h>
625fe58019SAttilio Rao #include <sys/module.h>
635fe58019SAttilio Rao #include <sys/systm.h>
645fe58019SAttilio Rao #include <sys/errno.h>
655fe58019SAttilio Rao #include <sys/kernel.h>
665fe58019SAttilio Rao #include <sys/conf.h>
675fe58019SAttilio Rao #include <sys/uio.h>
685fe58019SAttilio Rao #include <sys/malloc.h>
695fe58019SAttilio Rao #include <sys/queue.h>
705fe58019SAttilio Rao #include <sys/lock.h>
7189f6b863SAttilio Rao #include <sys/rwlock.h>
725fe58019SAttilio Rao #include <sys/sx.h>
735fe58019SAttilio Rao #include <sys/proc.h>
745fe58019SAttilio Rao #include <sys/mount.h>
755fe58019SAttilio Rao #include <sys/vnode.h>
765fe58019SAttilio Rao #include <sys/namei.h>
7704660064SFedor Uporov #include <sys/extattr.h>
785fe58019SAttilio Rao #include <sys/stat.h>
795fe58019SAttilio Rao #include <sys/unistd.h>
805fe58019SAttilio Rao #include <sys/filedesc.h>
815fe58019SAttilio Rao #include <sys/file.h>
825fe58019SAttilio Rao #include <sys/fcntl.h>
835fe58019SAttilio Rao #include <sys/dirent.h>
845fe58019SAttilio Rao #include <sys/bio.h>
855fe58019SAttilio Rao #include <sys/buf.h>
865fe58019SAttilio Rao #include <sys/sysctl.h>
87ca148cdaSGleb Smirnoff #include <sys/vmmeter.h>
885fe58019SAttilio Rao 
895fe58019SAttilio Rao #include <vm/vm.h>
905fe58019SAttilio Rao #include <vm/vm_extern.h>
915fe58019SAttilio Rao #include <vm/pmap.h>
925fe58019SAttilio Rao #include <vm/vm_map.h>
935fe58019SAttilio Rao #include <vm/vm_page.h>
945fe58019SAttilio Rao #include <vm/vm_param.h>
955fe58019SAttilio Rao #include <vm/vm_object.h>
965fe58019SAttilio Rao #include <vm/vm_pager.h>
975fe58019SAttilio Rao #include <vm/vnode_pager.h>
985fe58019SAttilio Rao #include <vm/vm_object.h>
995fe58019SAttilio Rao 
1005fe58019SAttilio Rao #include "fuse.h"
1015fe58019SAttilio Rao #include "fuse_file.h"
1025fe58019SAttilio Rao #include "fuse_internal.h"
1035fe58019SAttilio Rao #include "fuse_ipc.h"
1045fe58019SAttilio Rao #include "fuse_node.h"
1055fe58019SAttilio Rao #include "fuse_io.h"
1065fe58019SAttilio Rao 
1075fe58019SAttilio Rao #include <sys/priv.h>
1085fe58019SAttilio Rao 
109fd2749f2SAlan Somers /* Maximum number of hardlinks to a single FUSE file */
110fd2749f2SAlan Somers #define FUSE_LINK_MAX                      UINT32_MAX
111fd2749f2SAlan Somers 
112cf169498SAlan Somers SDT_PROVIDER_DECLARE(fuse);
113cf169498SAlan Somers /*
114cf169498SAlan Somers  * Fuse trace probe:
115cf169498SAlan Somers  * arg0: verbosity.  Higher numbers give more verbose messages
116cf169498SAlan Somers  * arg1: Textual message
117cf169498SAlan Somers  */
118cf169498SAlan Somers SDT_PROBE_DEFINE2(fuse, , vnops, trace, "int", "char*");
1195fe58019SAttilio Rao 
1205fe58019SAttilio Rao /* vnode ops */
1215fe58019SAttilio Rao static vop_access_t fuse_vnop_access;
1225fe58019SAttilio Rao static vop_close_t fuse_vnop_close;
1235fe58019SAttilio Rao static vop_create_t fuse_vnop_create;
12404660064SFedor Uporov static vop_deleteextattr_t fuse_vnop_deleteextattr;
125915012e0SAlan Somers static vop_fdatasync_t fuse_vnop_fdatasync;
1265fe58019SAttilio Rao static vop_fsync_t fuse_vnop_fsync;
1275fe58019SAttilio Rao static vop_getattr_t fuse_vnop_getattr;
12804660064SFedor Uporov static vop_getextattr_t fuse_vnop_getextattr;
1295fe58019SAttilio Rao static vop_inactive_t fuse_vnop_inactive;
1305fe58019SAttilio Rao static vop_link_t fuse_vnop_link;
13104660064SFedor Uporov static vop_listextattr_t fuse_vnop_listextattr;
1325fe58019SAttilio Rao static vop_lookup_t fuse_vnop_lookup;
1335fe58019SAttilio Rao static vop_mkdir_t fuse_vnop_mkdir;
1345fe58019SAttilio Rao static vop_mknod_t fuse_vnop_mknod;
1355fe58019SAttilio Rao static vop_open_t fuse_vnop_open;
136746c92e0SJohn Baldwin static vop_pathconf_t fuse_vnop_pathconf;
1375fe58019SAttilio Rao static vop_read_t fuse_vnop_read;
1385fe58019SAttilio Rao static vop_readdir_t fuse_vnop_readdir;
1395fe58019SAttilio Rao static vop_readlink_t fuse_vnop_readlink;
1405fe58019SAttilio Rao static vop_reclaim_t fuse_vnop_reclaim;
1415fe58019SAttilio Rao static vop_remove_t fuse_vnop_remove;
1425fe58019SAttilio Rao static vop_rename_t fuse_vnop_rename;
1435fe58019SAttilio Rao static vop_rmdir_t fuse_vnop_rmdir;
1445fe58019SAttilio Rao static vop_setattr_t fuse_vnop_setattr;
14504660064SFedor Uporov static vop_setextattr_t fuse_vnop_setextattr;
1465fe58019SAttilio Rao static vop_strategy_t fuse_vnop_strategy;
1475fe58019SAttilio Rao static vop_symlink_t fuse_vnop_symlink;
1485fe58019SAttilio Rao static vop_write_t fuse_vnop_write;
1495fe58019SAttilio Rao static vop_getpages_t fuse_vnop_getpages;
1505fe58019SAttilio Rao static vop_putpages_t fuse_vnop_putpages;
1515fe58019SAttilio Rao static vop_print_t fuse_vnop_print;
1525fe58019SAttilio Rao 
1535fe58019SAttilio Rao struct vop_vector fuse_vnops = {
1545fe58019SAttilio Rao 	.vop_default = &default_vnodeops,
1555fe58019SAttilio Rao 	.vop_access = fuse_vnop_access,
1565fe58019SAttilio Rao 	.vop_close = fuse_vnop_close,
1575fe58019SAttilio Rao 	.vop_create = fuse_vnop_create,
15804660064SFedor Uporov 	.vop_deleteextattr = fuse_vnop_deleteextattr,
1595fe58019SAttilio Rao 	.vop_fsync = fuse_vnop_fsync,
160915012e0SAlan Somers 	.vop_fdatasync = fuse_vnop_fdatasync,
1615fe58019SAttilio Rao 	.vop_getattr = fuse_vnop_getattr,
16204660064SFedor Uporov 	.vop_getextattr = fuse_vnop_getextattr,
1635fe58019SAttilio Rao 	.vop_inactive = fuse_vnop_inactive,
1645fe58019SAttilio Rao 	.vop_link = fuse_vnop_link,
16504660064SFedor Uporov 	.vop_listextattr = fuse_vnop_listextattr,
1665fe58019SAttilio Rao 	.vop_lookup = fuse_vnop_lookup,
1675fe58019SAttilio Rao 	.vop_mkdir = fuse_vnop_mkdir,
1685fe58019SAttilio Rao 	.vop_mknod = fuse_vnop_mknod,
1695fe58019SAttilio Rao 	.vop_open = fuse_vnop_open,
170746c92e0SJohn Baldwin 	.vop_pathconf = fuse_vnop_pathconf,
1715fe58019SAttilio Rao 	.vop_read = fuse_vnop_read,
1725fe58019SAttilio Rao 	.vop_readdir = fuse_vnop_readdir,
1735fe58019SAttilio Rao 	.vop_readlink = fuse_vnop_readlink,
1745fe58019SAttilio Rao 	.vop_reclaim = fuse_vnop_reclaim,
1755fe58019SAttilio Rao 	.vop_remove = fuse_vnop_remove,
1765fe58019SAttilio Rao 	.vop_rename = fuse_vnop_rename,
1775fe58019SAttilio Rao 	.vop_rmdir = fuse_vnop_rmdir,
1785fe58019SAttilio Rao 	.vop_setattr = fuse_vnop_setattr,
17904660064SFedor Uporov 	.vop_setextattr = fuse_vnop_setextattr,
1805fe58019SAttilio Rao 	.vop_strategy = fuse_vnop_strategy,
1815fe58019SAttilio Rao 	.vop_symlink = fuse_vnop_symlink,
1825fe58019SAttilio Rao 	.vop_write = fuse_vnop_write,
1835fe58019SAttilio Rao 	.vop_getpages = fuse_vnop_getpages,
1845fe58019SAttilio Rao 	.vop_putpages = fuse_vnop_putpages,
1855fe58019SAttilio Rao 	.vop_print = fuse_vnop_print,
1865fe58019SAttilio Rao };
1875fe58019SAttilio Rao 
1885fe58019SAttilio Rao static u_long fuse_lookup_cache_hits = 0;
1895fe58019SAttilio Rao 
190123af6ecSAlan Somers SYSCTL_ULONG(_vfs_fusefs, OID_AUTO, lookup_cache_hits, CTLFLAG_RD,
191bd4cb2a4SConrad Meyer     &fuse_lookup_cache_hits, 0, "number of positive cache hits in lookup");
1925fe58019SAttilio Rao 
1935fe58019SAttilio Rao static u_long fuse_lookup_cache_misses = 0;
1945fe58019SAttilio Rao 
195123af6ecSAlan Somers SYSCTL_ULONG(_vfs_fusefs, OID_AUTO, lookup_cache_misses, CTLFLAG_RD,
196bd4cb2a4SConrad Meyer     &fuse_lookup_cache_misses, 0, "number of cache misses in lookup");
1975fe58019SAttilio Rao 
198*3f2c630cSAlan Somers int	fuse_lookup_cache_expire = 0;
1995fe58019SAttilio Rao 
200*3f2c630cSAlan Somers SYSCTL_INT(_vfs_fusefs, OID_AUTO, lookup_cache_expire, CTLFLAG_RW,
201*3f2c630cSAlan Somers     &fuse_lookup_cache_expire, 0,
202*3f2c630cSAlan Somers     "if non-zero, expire fuse lookup cache entries at the proper time");
2035fe58019SAttilio Rao 
2045fe58019SAttilio Rao /*
2055fe58019SAttilio Rao  * XXX: This feature is highly experimental and can bring to instabilities,
2065fe58019SAttilio Rao  * needs revisiting before to be enabled by default.
2075fe58019SAttilio Rao  */
2085fe58019SAttilio Rao static int fuse_reclaim_revoked = 0;
2095fe58019SAttilio Rao 
210123af6ecSAlan Somers SYSCTL_INT(_vfs_fusefs, OID_AUTO, reclaim_revoked, CTLFLAG_RW,
2115fe58019SAttilio Rao     &fuse_reclaim_revoked, 0, "");
2125fe58019SAttilio Rao 
213756a5412SGleb Smirnoff uma_zone_t fuse_pbuf_zone;
2145fe58019SAttilio Rao 
2155fe58019SAttilio Rao #define fuse_vm_page_lock(m)		vm_page_lock((m));
2165fe58019SAttilio Rao #define fuse_vm_page_unlock(m)		vm_page_unlock((m));
2175fe58019SAttilio Rao #define fuse_vm_page_lock_queues()	((void)0)
2185fe58019SAttilio Rao #define fuse_vm_page_unlock_queues()	((void)0)
2195fe58019SAttilio Rao 
220f8d4af10SAlan Somers /* Get a filehandle for a directory */
221f8d4af10SAlan Somers static int
222f8d4af10SAlan Somers fuse_filehandle_get_dir(struct vnode *vp, struct fuse_filehandle **fufhp,
223f8d4af10SAlan Somers 	struct ucred *cred, pid_t pid)
224f8d4af10SAlan Somers {
2259f10f423SAlan Somers 	if (fuse_filehandle_get(vp, FREAD, fufhp, cred, pid) == 0)
226f8d4af10SAlan Somers 		return 0;
2279f10f423SAlan Somers 	return fuse_filehandle_get(vp, FEXEC, fufhp, cred, pid);
2289f10f423SAlan Somers }
2299f10f423SAlan Somers 
2309f10f423SAlan Somers /* Send FUSE_FLUSH for this vnode */
2319f10f423SAlan Somers static int
2329f10f423SAlan Somers fuse_flush(struct vnode *vp, struct ucred *cred, pid_t pid, int fflag)
2339f10f423SAlan Somers {
2349f10f423SAlan Somers 	struct fuse_flush_in *ffi;
2359f10f423SAlan Somers 	struct fuse_filehandle *fufh;
2369f10f423SAlan Somers 	struct fuse_dispatcher fdi;
2379f10f423SAlan Somers 	struct thread *td = curthread;
2389f10f423SAlan Somers 	struct mount *mp = vnode_mount(vp);
2399f10f423SAlan Somers 	int err;
2409f10f423SAlan Somers 
2419f10f423SAlan Somers 	if (!fsess_isimpl(vnode_mount(vp), FUSE_FLUSH))
2429f10f423SAlan Somers 		return 0;
2439f10f423SAlan Somers 
2449f10f423SAlan Somers 	err = fuse_filehandle_get(vp, fflag, &fufh, cred, pid);
2459f10f423SAlan Somers 	if (err)
2469f10f423SAlan Somers 		return err;
2479f10f423SAlan Somers 
2489f10f423SAlan Somers 	fdisp_init(&fdi, sizeof(*ffi));
2499f10f423SAlan Somers 	fdisp_make_vp(&fdi, FUSE_FLUSH, vp, td, cred);
2509f10f423SAlan Somers 	ffi = fdi.indata;
2519f10f423SAlan Somers 	ffi->fh = fufh->fh_id;
2529f10f423SAlan Somers 
2539f10f423SAlan Somers 	err = fdisp_wait_answ(&fdi);
2549f10f423SAlan Somers 	if (err == ENOSYS) {
2559f10f423SAlan Somers 		fsess_set_notimpl(mp, FUSE_FLUSH);
2569f10f423SAlan Somers 		err = 0;
2579f10f423SAlan Somers 	}
2589f10f423SAlan Somers 	fdisp_destroy(&fdi);
2599f10f423SAlan Somers 	return err;
260f8d4af10SAlan Somers }
261f8d4af10SAlan Somers 
2625fe58019SAttilio Rao /*
2635fe58019SAttilio Rao     struct vnop_access_args {
2645fe58019SAttilio Rao 	struct vnode *a_vp;
2655fe58019SAttilio Rao #if VOP_ACCESS_TAKES_ACCMODE_T
2665fe58019SAttilio Rao 	accmode_t a_accmode;
2675fe58019SAttilio Rao #else
2685fe58019SAttilio Rao 	int a_mode;
2695fe58019SAttilio Rao #endif
2705fe58019SAttilio Rao 	struct ucred *a_cred;
2715fe58019SAttilio Rao 	struct thread *a_td;
2725fe58019SAttilio Rao     };
2735fe58019SAttilio Rao */
2745fe58019SAttilio Rao static int
2755fe58019SAttilio Rao fuse_vnop_access(struct vop_access_args *ap)
2765fe58019SAttilio Rao {
2775fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
2785fe58019SAttilio Rao 	int accmode = ap->a_accmode;
2795fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
2805fe58019SAttilio Rao 
2815fe58019SAttilio Rao 	struct fuse_access_param facp;
2825fe58019SAttilio Rao 	struct fuse_data *data = fuse_get_mpdata(vnode_mount(vp));
2835fe58019SAttilio Rao 
2845fe58019SAttilio Rao 	int err;
2855fe58019SAttilio Rao 
2865fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
2875fe58019SAttilio Rao 		if (vnode_isvroot(vp)) {
2885fe58019SAttilio Rao 			return 0;
2895fe58019SAttilio Rao 		}
2905fe58019SAttilio Rao 		return ENXIO;
2915fe58019SAttilio Rao 	}
2925fe58019SAttilio Rao 	if (!(data->dataflags & FSESS_INITED)) {
2935fe58019SAttilio Rao 		if (vnode_isvroot(vp)) {
294cc426dd3SMateusz Guzik 			if (priv_check_cred(cred, PRIV_VFS_ADMIN) ||
2955fe58019SAttilio Rao 			    (fuse_match_cred(data->daemoncred, cred) == 0)) {
2965fe58019SAttilio Rao 				return 0;
2975fe58019SAttilio Rao 			}
2985fe58019SAttilio Rao 		}
2995fe58019SAttilio Rao 		return EBADF;
3005fe58019SAttilio Rao 	}
3015fe58019SAttilio Rao 	if (vnode_islnk(vp)) {
3025fe58019SAttilio Rao 		return 0;
3035fe58019SAttilio Rao 	}
3045fe58019SAttilio Rao 	bzero(&facp, sizeof(facp));
3055fe58019SAttilio Rao 
3065fe58019SAttilio Rao 	err = fuse_internal_access(vp, accmode, &facp, ap->a_td, ap->a_cred);
3075fe58019SAttilio Rao 	return err;
3085fe58019SAttilio Rao }
3095fe58019SAttilio Rao 
3105fe58019SAttilio Rao /*
3119f10f423SAlan Somers     struct vop_close_args {
3125fe58019SAttilio Rao 	struct vnode *a_vp;
3135fe58019SAttilio Rao 	int  a_fflag;
3145fe58019SAttilio Rao 	struct ucred *a_cred;
3155fe58019SAttilio Rao 	struct thread *a_td;
3165fe58019SAttilio Rao     };
3175fe58019SAttilio Rao */
3185fe58019SAttilio Rao static int
3195fe58019SAttilio Rao fuse_vnop_close(struct vop_close_args *ap)
3205fe58019SAttilio Rao {
3215fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
3225fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
3235fe58019SAttilio Rao 	int fflag = ap->a_fflag;
324f8d4af10SAlan Somers 	struct thread *td = ap->a_td;
325f8d4af10SAlan Somers 	pid_t pid = td->td_proc->p_pid;
3269f10f423SAlan Somers 	int err = 0;
3275fe58019SAttilio Rao 
32835cf0e7eSAlan Somers 	if (fuse_isdeadfs(vp))
3295fe58019SAttilio Rao 		return 0;
33035cf0e7eSAlan Somers 	if (vnode_isdir(vp))
33135cf0e7eSAlan Somers 		return 0;
33235cf0e7eSAlan Somers 	if (fflag & IO_NDELAY)
33335cf0e7eSAlan Somers 		return 0;
3345ec10aa5SAlan Somers 
3359f10f423SAlan Somers 	err = fuse_flush(vp, cred, pid, fflag);
3365ec10aa5SAlan Somers 	/* TODO: close the file handle, if we're sure it's no longer used */
3375fe58019SAttilio Rao 	if ((VTOFUD(vp)->flag & FN_SIZECHANGE) != 0) {
338f8d4af10SAlan Somers 		fuse_vnode_savesize(vp, cred, td->td_proc->p_pid);
3395fe58019SAttilio Rao 	}
3409f10f423SAlan Somers 	return err;
3415fe58019SAttilio Rao }
3425fe58019SAttilio Rao 
34319ef317dSAlan Somers static void
34419ef317dSAlan Somers fdisp_make_mknod_for_fallback(
34519ef317dSAlan Somers 	struct fuse_dispatcher *fdip,
34619ef317dSAlan Somers 	struct componentname *cnp,
34719ef317dSAlan Somers 	struct vnode *dvp,
34819ef317dSAlan Somers 	uint64_t parentnid,
34919ef317dSAlan Somers 	struct thread *td,
35019ef317dSAlan Somers 	struct ucred *cred,
35119ef317dSAlan Somers 	mode_t mode,
35219ef317dSAlan Somers 	enum fuse_opcode *op)
35319ef317dSAlan Somers {
35419ef317dSAlan Somers 	struct fuse_mknod_in *fmni;
35519ef317dSAlan Somers 
35619ef317dSAlan Somers 	fdisp_init(fdip, sizeof(*fmni) + cnp->cn_namelen + 1);
35719ef317dSAlan Somers 	*op = FUSE_MKNOD;
35819ef317dSAlan Somers 	fdisp_make(fdip, *op, vnode_mount(dvp), parentnid, td, cred);
35919ef317dSAlan Somers 	fmni = fdip->indata;
36019ef317dSAlan Somers 	fmni->mode = mode;
36119ef317dSAlan Somers 	fmni->rdev = 0;
36219ef317dSAlan Somers 	memcpy((char *)fdip->indata + sizeof(*fmni), cnp->cn_nameptr,
36319ef317dSAlan Somers 	    cnp->cn_namelen);
36419ef317dSAlan Somers 	((char *)fdip->indata)[sizeof(*fmni) + cnp->cn_namelen] = '\0';
36519ef317dSAlan Somers }
3665fe58019SAttilio Rao /*
3675fe58019SAttilio Rao     struct vnop_create_args {
3685fe58019SAttilio Rao 	struct vnode *a_dvp;
3695fe58019SAttilio Rao 	struct vnode **a_vpp;
3705fe58019SAttilio Rao 	struct componentname *a_cnp;
3715fe58019SAttilio Rao 	struct vattr *a_vap;
3725fe58019SAttilio Rao     };
3735fe58019SAttilio Rao */
3745fe58019SAttilio Rao static int
3755fe58019SAttilio Rao fuse_vnop_create(struct vop_create_args *ap)
3765fe58019SAttilio Rao {
3775fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
3785fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
3795fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
3805fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
3815fe58019SAttilio Rao 	struct thread *td = cnp->cn_thread;
3825fe58019SAttilio Rao 	struct ucred *cred = cnp->cn_cred;
3835fe58019SAttilio Rao 
3845fe58019SAttilio Rao 	struct fuse_open_in *foi;
3855fe58019SAttilio Rao 	struct fuse_entry_out *feo;
38619ef317dSAlan Somers 	struct fuse_open_out *foo;
38719ef317dSAlan Somers 	struct fuse_dispatcher fdi, fdi2;
3885fe58019SAttilio Rao 	struct fuse_dispatcher *fdip = &fdi;
38919ef317dSAlan Somers 	struct fuse_dispatcher *fdip2 = NULL;
3905fe58019SAttilio Rao 
3915fe58019SAttilio Rao 	int err;
3925fe58019SAttilio Rao 
3935fe58019SAttilio Rao 	struct mount *mp = vnode_mount(dvp);
3945fe58019SAttilio Rao 	uint64_t parentnid = VTOFUD(dvp)->nid;
3955fe58019SAttilio Rao 	mode_t mode = MAKEIMODE(vap->va_type, vap->va_mode);
39619ef317dSAlan Somers 	enum fuse_opcode op;
3979e444871SAlan Somers 	int flags;
3989e444871SAlan Somers 
3999e444871SAlan Somers 	/*
4009e444871SAlan Somers 	 * VOP_CREATE doesn't tell us the open(2) flags, so we guess.  Only a
4019e444871SAlan Somers 	 * writable mode makes sense, and we might as well include readability
4029e444871SAlan Somers 	 * too.
4039e444871SAlan Somers 	 */
4049e444871SAlan Somers 	flags = O_RDWR;
4055fe58019SAttilio Rao 
4065fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
4075fe58019SAttilio Rao 		return ENXIO;
4085fe58019SAttilio Rao 	}
4095fe58019SAttilio Rao 	bzero(&fdi, sizeof(fdi));
4105fe58019SAttilio Rao 
41119ef317dSAlan Somers 	if ((vap->va_type != VREG))
412372b97d0SRick Macklem 		return (EINVAL);
4135fe58019SAttilio Rao 
4145fe58019SAttilio Rao 	if (!fsess_isimpl(mp, FUSE_CREATE)) {
41519ef317dSAlan Somers 		/* Fallback to FUSE_MKNOD/FUSE_OPEN */
41619ef317dSAlan Somers 		fdisp_make_mknod_for_fallback(fdip, cnp, dvp, parentnid, td,
41719ef317dSAlan Somers 			cred, mode, &op);
41819ef317dSAlan Somers 	} else {
41919ef317dSAlan Somers 		/* Use FUSE_CREATE */
42019ef317dSAlan Somers 		op = FUSE_CREATE;
42119ef317dSAlan Somers 		fdisp_init(fdip, sizeof(*foi) + cnp->cn_namelen + 1);
42219ef317dSAlan Somers 		fdisp_make(fdip, op, vnode_mount(dvp), parentnid, td, cred);
4235fe58019SAttilio Rao 		foi = fdip->indata;
4245fe58019SAttilio Rao 		foi->mode = mode;
4259e444871SAlan Somers 		foi->flags = O_CREAT | flags;
4265fe58019SAttilio Rao 		memcpy((char *)fdip->indata + sizeof(*foi), cnp->cn_nameptr,
4275fe58019SAttilio Rao 		    cnp->cn_namelen);
4285fe58019SAttilio Rao 		((char *)fdip->indata)[sizeof(*foi) + cnp->cn_namelen] = '\0';
42919ef317dSAlan Somers 	}
4305fe58019SAttilio Rao 
4315fe58019SAttilio Rao 	err = fdisp_wait_answ(fdip);
4325fe58019SAttilio Rao 
4336de3b00dSAttilio Rao 	if (err) {
43419ef317dSAlan Somers 		if (err == ENOSYS && op == FUSE_CREATE) {
4355fe58019SAttilio Rao 			fsess_set_notimpl(mp, FUSE_CREATE);
43619ef317dSAlan Somers 			fdisp_make_mknod_for_fallback(fdip, cnp, dvp,
43719ef317dSAlan Somers 				parentnid, td, cred, mode, &op);
43819ef317dSAlan Somers 			err = fdisp_wait_answ(fdip);
43919ef317dSAlan Somers 		}
44019ef317dSAlan Somers 		if (err)
4415fe58019SAttilio Rao 			goto out;
4425fe58019SAttilio Rao 	}
443372b97d0SRick Macklem 
4445fe58019SAttilio Rao 	feo = fdip->answ;
4455fe58019SAttilio Rao 
4465fe58019SAttilio Rao 	if ((err = fuse_internal_checkentry(feo, VREG))) {
4475fe58019SAttilio Rao 		goto out;
4485fe58019SAttilio Rao 	}
44919ef317dSAlan Somers 
45019ef317dSAlan Somers 	if (op == FUSE_CREATE) {
45119ef317dSAlan Somers 		foo = (struct fuse_open_out*)(feo + 1);
45219ef317dSAlan Somers 	} else {
45319ef317dSAlan Somers 		/* Issue a separate FUSE_OPEN */
45419ef317dSAlan Somers 		fdip2 = &fdi2;
45519ef317dSAlan Somers 		fdisp_init(fdip2, sizeof(*foi));
45619ef317dSAlan Somers 		fdisp_make(fdip2, FUSE_OPEN, vnode_mount(dvp), feo->nodeid, td,
45719ef317dSAlan Somers 			cred);
45819ef317dSAlan Somers 		foi = fdip2->indata;
45919ef317dSAlan Somers 		foi->mode = mode;
4609e444871SAlan Somers 		foi->flags = flags;
46119ef317dSAlan Somers 		err = fdisp_wait_answ(fdip2);
46219ef317dSAlan Somers 		if (err)
46319ef317dSAlan Somers 			goto out;
46419ef317dSAlan Somers 		foo = fdip2->answ;
46519ef317dSAlan Somers 	}
46609176f09SConrad Meyer 	err = fuse_vnode_get(mp, feo, feo->nodeid, dvp, vpp, cnp, VREG);
4675fe58019SAttilio Rao 	if (err) {
4685fe58019SAttilio Rao 		struct fuse_release_in *fri;
4695fe58019SAttilio Rao 		uint64_t nodeid = feo->nodeid;
47019ef317dSAlan Somers 		uint64_t fh_id = foo->fh;
4715fe58019SAttilio Rao 
4725fe58019SAttilio Rao 		fdisp_init(fdip, sizeof(*fri));
4735fe58019SAttilio Rao 		fdisp_make(fdip, FUSE_RELEASE, mp, nodeid, td, cred);
4745fe58019SAttilio Rao 		fri = fdip->indata;
4755fe58019SAttilio Rao 		fri->fh = fh_id;
4769e444871SAlan Somers 		fri->flags = flags;
4775fe58019SAttilio Rao 		fuse_insert_callback(fdip->tick, fuse_internal_forget_callback);
4785fe58019SAttilio Rao 		fuse_insert_message(fdip->tick);
47919ef317dSAlan Somers 		goto out;
4805fe58019SAttilio Rao 	}
4815fe58019SAttilio Rao 	ASSERT_VOP_ELOCKED(*vpp, "fuse_vnop_create");
482cad67791SAlan Somers 	fuse_internal_cache_attrs(*vpp, &feo->attr, feo->attr_valid,
483cad67791SAlan Somers 		feo->attr_valid_nsec, NULL);
4845fe58019SAttilio Rao 
485a7e81cb3SAlan Somers 	fuse_filehandle_init(*vpp, FUFH_RDWR, NULL, td, cred, foo);
48619ef317dSAlan Somers 	fuse_vnode_open(*vpp, foo->open_flags, td);
4875fe58019SAttilio Rao 	cache_purge_negative(dvp);
4885fe58019SAttilio Rao 
4895fe58019SAttilio Rao out:
49019ef317dSAlan Somers 	if (fdip2)
49119ef317dSAlan Somers 		fdisp_destroy(fdip2);
4925fe58019SAttilio Rao 	fdisp_destroy(fdip);
4935fe58019SAttilio Rao 	return err;
4945fe58019SAttilio Rao }
4955fe58019SAttilio Rao 
4965fe58019SAttilio Rao /*
497915012e0SAlan Somers     struct vnop_fdatasync_args {
498915012e0SAlan Somers 	struct vop_generic_args a_gen;
499915012e0SAlan Somers 	struct vnode * a_vp;
500915012e0SAlan Somers 	struct thread * a_td;
501915012e0SAlan Somers     };
5025fe58019SAttilio Rao */
503915012e0SAlan Somers static int
504915012e0SAlan Somers fuse_vnop_fdatasync(struct vop_fdatasync_args *ap)
505915012e0SAlan Somers {
506915012e0SAlan Somers 	struct vnode *vp = ap->a_vp;
507915012e0SAlan Somers 	struct thread *td = ap->a_td;
508915012e0SAlan Somers 	int waitfor = MNT_WAIT;
509915012e0SAlan Somers 
510915012e0SAlan Somers 	int err = 0;
511915012e0SAlan Somers 
512915012e0SAlan Somers 	if (fuse_isdeadfs(vp)) {
513915012e0SAlan Somers 		return 0;
514915012e0SAlan Somers 	}
515915012e0SAlan Somers 	if ((err = vop_stdfdatasync_buf(ap)))
516915012e0SAlan Somers 		return err;
517915012e0SAlan Somers 
518915012e0SAlan Somers 	return fuse_internal_fsync(vp, td, waitfor, true);
519915012e0SAlan Somers }
5205fe58019SAttilio Rao 
5215fe58019SAttilio Rao /*
5225fe58019SAttilio Rao     struct vnop_fsync_args {
523915012e0SAlan Somers 	struct vop_generic_args a_gen;
5245fe58019SAttilio Rao 	struct vnode * a_vp;
5255fe58019SAttilio Rao 	int  a_waitfor;
5265fe58019SAttilio Rao 	struct thread * a_td;
5275fe58019SAttilio Rao     };
5285fe58019SAttilio Rao */
5295fe58019SAttilio Rao static int
5305fe58019SAttilio Rao fuse_vnop_fsync(struct vop_fsync_args *ap)
5315fe58019SAttilio Rao {
5325fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
5335fe58019SAttilio Rao 	struct thread *td = ap->a_td;
53490612f3cSAlan Somers 	int waitfor = ap->a_waitfor;
535915012e0SAlan Somers 	int err = 0;
5365fe58019SAttilio Rao 
5375fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
5385fe58019SAttilio Rao 		return 0;
5395fe58019SAttilio Rao 	}
5405fe58019SAttilio Rao 	if ((err = vop_stdfsync(ap)))
5415fe58019SAttilio Rao 		return err;
5425fe58019SAttilio Rao 
543915012e0SAlan Somers 	return fuse_internal_fsync(vp, td, waitfor, false);
5445fe58019SAttilio Rao }
5455fe58019SAttilio Rao 
5465fe58019SAttilio Rao /*
5475fe58019SAttilio Rao     struct vnop_getattr_args {
5485fe58019SAttilio Rao 	struct vnode *a_vp;
5495fe58019SAttilio Rao 	struct vattr *a_vap;
5505fe58019SAttilio Rao 	struct ucred *a_cred;
5515fe58019SAttilio Rao 	struct thread *a_td;
5525fe58019SAttilio Rao     };
5535fe58019SAttilio Rao */
5545fe58019SAttilio Rao static int
5555fe58019SAttilio Rao fuse_vnop_getattr(struct vop_getattr_args *ap)
5565fe58019SAttilio Rao {
5575fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
5585fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
5595fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
5605fe58019SAttilio Rao 	struct thread *td = curthread;
5615fe58019SAttilio Rao 
5625fe58019SAttilio Rao 	int err = 0;
5635fe58019SAttilio Rao 	int dataflags;
5645fe58019SAttilio Rao 
5655fe58019SAttilio Rao 	dataflags = fuse_get_mpdata(vnode_mount(vp))->dataflags;
5665fe58019SAttilio Rao 
5675fe58019SAttilio Rao 	/* Note that we are not bailing out on a dead file system just yet. */
5685fe58019SAttilio Rao 
5695fe58019SAttilio Rao 	if (!(dataflags & FSESS_INITED)) {
5705fe58019SAttilio Rao 		if (!vnode_isvroot(vp)) {
5715fe58019SAttilio Rao 			fdata_set_dead(fuse_get_mpdata(vnode_mount(vp)));
5725fe58019SAttilio Rao 			err = ENOTCONN;
5735fe58019SAttilio Rao 			return err;
5745fe58019SAttilio Rao 		} else {
5755fe58019SAttilio Rao 			goto fake;
5765fe58019SAttilio Rao 		}
5775fe58019SAttilio Rao 	}
578cad67791SAlan Somers 	err = fuse_internal_getattr(vp, vap, cred, td);
579cad67791SAlan Somers 	if (err == ENOTCONN && vnode_isvroot(vp)) {
580cf169498SAlan Somers 		/* see comment in fuse_vfsop_statfs() */
5815fe58019SAttilio Rao 		goto fake;
582cad67791SAlan Somers 	} else {
5835fe58019SAttilio Rao 		return err;
584cad67791SAlan Somers 	}
5855fe58019SAttilio Rao 
5865fe58019SAttilio Rao fake:
5875fe58019SAttilio Rao 	bzero(vap, sizeof(*vap));
5885fe58019SAttilio Rao 	vap->va_type = vnode_vtype(vp);
5895fe58019SAttilio Rao 
5905fe58019SAttilio Rao 	return 0;
5915fe58019SAttilio Rao }
5925fe58019SAttilio Rao 
5935fe58019SAttilio Rao /*
5945fe58019SAttilio Rao     struct vnop_inactive_args {
5955fe58019SAttilio Rao 	struct vnode *a_vp;
5965fe58019SAttilio Rao 	struct thread *a_td;
5975fe58019SAttilio Rao     };
5985fe58019SAttilio Rao */
5995fe58019SAttilio Rao static int
6005fe58019SAttilio Rao fuse_vnop_inactive(struct vop_inactive_args *ap)
6015fe58019SAttilio Rao {
6025fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
6035fe58019SAttilio Rao 	struct thread *td = ap->a_td;
6045fe58019SAttilio Rao 
6055fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(vp);
6065ec10aa5SAlan Somers 	struct fuse_filehandle *fufh, *fufh_tmp;
6075fe58019SAttilio Rao 
6085ec10aa5SAlan Somers 	int need_flush = 1;
6095fe58019SAttilio Rao 
6105ec10aa5SAlan Somers 	LIST_FOREACH_SAFE(fufh, &fvdat->handles, next, fufh_tmp) {
6115fe58019SAttilio Rao 		if (need_flush && vp->v_type == VREG) {
6125fe58019SAttilio Rao 			if ((VTOFUD(vp)->flag & FN_SIZECHANGE) != 0) {
613f8d4af10SAlan Somers 				fuse_vnode_savesize(vp, NULL, 0);
6145fe58019SAttilio Rao 			}
6155fe58019SAttilio Rao 			if (fuse_data_cache_invalidate ||
6165fe58019SAttilio Rao 			    (fvdat->flag & FN_REVOKED) != 0)
6175fe58019SAttilio Rao 				fuse_io_invalbuf(vp, td);
6185fe58019SAttilio Rao 			else
6195fe58019SAttilio Rao 				fuse_io_flushbuf(vp, MNT_WAIT, td);
6205fe58019SAttilio Rao 			need_flush = 0;
6215fe58019SAttilio Rao 		}
6225ec10aa5SAlan Somers 		fuse_filehandle_close(vp, fufh, td, NULL);
6235fe58019SAttilio Rao 	}
6245fe58019SAttilio Rao 
6255fe58019SAttilio Rao 	if ((fvdat->flag & FN_REVOKED) != 0 && fuse_reclaim_revoked) {
6265fe58019SAttilio Rao 		vrecycle(vp);
6275fe58019SAttilio Rao 	}
6285fe58019SAttilio Rao 	return 0;
6295fe58019SAttilio Rao }
6305fe58019SAttilio Rao 
6315fe58019SAttilio Rao /*
6325fe58019SAttilio Rao     struct vnop_link_args {
6335fe58019SAttilio Rao 	struct vnode *a_tdvp;
6345fe58019SAttilio Rao 	struct vnode *a_vp;
6355fe58019SAttilio Rao 	struct componentname *a_cnp;
6365fe58019SAttilio Rao     };
6375fe58019SAttilio Rao */
6385fe58019SAttilio Rao static int
6395fe58019SAttilio Rao fuse_vnop_link(struct vop_link_args *ap)
6405fe58019SAttilio Rao {
6415fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
6425fe58019SAttilio Rao 	struct vnode *tdvp = ap->a_tdvp;
6435fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
6445fe58019SAttilio Rao 
6455fe58019SAttilio Rao 	struct vattr *vap = VTOVA(vp);
6465fe58019SAttilio Rao 
6475fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
6485fe58019SAttilio Rao 	struct fuse_entry_out *feo;
6495fe58019SAttilio Rao 	struct fuse_link_in fli;
6505fe58019SAttilio Rao 
6515fe58019SAttilio Rao 	int err;
6525fe58019SAttilio Rao 
6535fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
6545fe58019SAttilio Rao 		return ENXIO;
6555fe58019SAttilio Rao 	}
6565fe58019SAttilio Rao 	if (vnode_mount(tdvp) != vnode_mount(vp)) {
6575fe58019SAttilio Rao 		return EXDEV;
6585fe58019SAttilio Rao 	}
65978a7722fSConrad Meyer 
66078a7722fSConrad Meyer 	/*
66178a7722fSConrad Meyer 	 * This is a seatbelt check to protect naive userspace filesystems from
66278a7722fSConrad Meyer 	 * themselves and the limitations of the FUSE IPC protocol.  If a
66378a7722fSConrad Meyer 	 * filesystem does not allow attribute caching, assume it is capable of
66478a7722fSConrad Meyer 	 * validating that nlink does not overflow.
66578a7722fSConrad Meyer 	 */
66678a7722fSConrad Meyer 	if (vap != NULL && vap->va_nlink >= FUSE_LINK_MAX)
6675fe58019SAttilio Rao 		return EMLINK;
6685fe58019SAttilio Rao 	fli.oldnodeid = VTOI(vp);
6695fe58019SAttilio Rao 
6705fe58019SAttilio Rao 	fdisp_init(&fdi, 0);
6715fe58019SAttilio Rao 	fuse_internal_newentry_makerequest(vnode_mount(tdvp), VTOI(tdvp), cnp,
6725fe58019SAttilio Rao 	    FUSE_LINK, &fli, sizeof(fli), &fdi);
6735fe58019SAttilio Rao 	if ((err = fdisp_wait_answ(&fdi))) {
6745fe58019SAttilio Rao 		goto out;
6755fe58019SAttilio Rao 	}
6765fe58019SAttilio Rao 	feo = fdi.answ;
6775fe58019SAttilio Rao 
6785fe58019SAttilio Rao 	err = fuse_internal_checkentry(feo, vnode_vtype(vp));
6795fe58019SAttilio Rao out:
6805fe58019SAttilio Rao 	fdisp_destroy(&fdi);
6815fe58019SAttilio Rao 	return err;
6825fe58019SAttilio Rao }
6835fe58019SAttilio Rao 
6845fe58019SAttilio Rao /*
6855fe58019SAttilio Rao     struct vnop_lookup_args {
6865fe58019SAttilio Rao 	struct vnodeop_desc *a_desc;
6875fe58019SAttilio Rao 	struct vnode *a_dvp;
6885fe58019SAttilio Rao 	struct vnode **a_vpp;
6895fe58019SAttilio Rao 	struct componentname *a_cnp;
6905fe58019SAttilio Rao     };
6915fe58019SAttilio Rao */
6925fe58019SAttilio Rao int
6935fe58019SAttilio Rao fuse_vnop_lookup(struct vop_lookup_args *ap)
6945fe58019SAttilio Rao {
6955fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
6965fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
6975fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
6985fe58019SAttilio Rao 	struct thread *td = cnp->cn_thread;
6995fe58019SAttilio Rao 	struct ucred *cred = cnp->cn_cred;
7005fe58019SAttilio Rao 
7015fe58019SAttilio Rao 	int nameiop = cnp->cn_nameiop;
7025fe58019SAttilio Rao 	int flags = cnp->cn_flags;
7035fe58019SAttilio Rao 	int wantparent = flags & (LOCKPARENT | WANTPARENT);
7045fe58019SAttilio Rao 	int islastcn = flags & ISLASTCN;
7055fe58019SAttilio Rao 	struct mount *mp = vnode_mount(dvp);
7065fe58019SAttilio Rao 
7075fe58019SAttilio Rao 	int err = 0;
7085fe58019SAttilio Rao 	int lookup_err = 0;
7095fe58019SAttilio Rao 	struct vnode *vp = NULL;
7105fe58019SAttilio Rao 
7115fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
7125fe58019SAttilio Rao 	enum fuse_opcode op;
7135fe58019SAttilio Rao 
7145fe58019SAttilio Rao 	uint64_t nid;
7155fe58019SAttilio Rao 	struct fuse_access_param facp;
7165fe58019SAttilio Rao 
7175fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
7185fe58019SAttilio Rao 		*vpp = NULL;
7195fe58019SAttilio Rao 		return ENXIO;
7205fe58019SAttilio Rao 	}
7215fe58019SAttilio Rao 	if (!vnode_isdir(dvp)) {
7225fe58019SAttilio Rao 		return ENOTDIR;
7235fe58019SAttilio Rao 	}
7245fe58019SAttilio Rao 	if (islastcn && vfs_isrdonly(mp) && (nameiop != LOOKUP)) {
7255fe58019SAttilio Rao 		return EROFS;
7265fe58019SAttilio Rao 	}
7275fe58019SAttilio Rao 	/*
7285fe58019SAttilio Rao 	 * We do access check prior to doing anything else only in the case
7295fe58019SAttilio Rao 	 * when we are at fs root (we'd like to say, "we are at the first
7305fe58019SAttilio Rao 	 * component", but that's not exactly the same... nevermind).
7315fe58019SAttilio Rao 	 * See further comments at further access checks.
7325fe58019SAttilio Rao 	 */
7335fe58019SAttilio Rao 
7345fe58019SAttilio Rao 	bzero(&facp, sizeof(facp));
7355fe58019SAttilio Rao 	if (vnode_isvroot(dvp)) {	/* early permission check hack */
7365fe58019SAttilio Rao 		if ((err = fuse_internal_access(dvp, VEXEC, &facp, td, cred))) {
7375fe58019SAttilio Rao 			return err;
7385fe58019SAttilio Rao 		}
7395fe58019SAttilio Rao 	}
7405fe58019SAttilio Rao 	if (flags & ISDOTDOT) {
7415fe58019SAttilio Rao 		nid = VTOFUD(dvp)->parent_nid;
7425fe58019SAttilio Rao 		if (nid == 0) {
7435fe58019SAttilio Rao 			return ENOENT;
7445fe58019SAttilio Rao 		}
7455fe58019SAttilio Rao 		fdisp_init(&fdi, 0);
7465fe58019SAttilio Rao 		op = FUSE_GETATTR;
7475fe58019SAttilio Rao 		goto calldaemon;
7485fe58019SAttilio Rao 	} else if (cnp->cn_namelen == 1 && *(cnp->cn_nameptr) == '.') {
7495fe58019SAttilio Rao 		nid = VTOI(dvp);
7505fe58019SAttilio Rao 		fdisp_init(&fdi, 0);
7515fe58019SAttilio Rao 		op = FUSE_GETATTR;
7525fe58019SAttilio Rao 		goto calldaemon;
753*3f2c630cSAlan Somers 	} else if (fuse_lookup_cache_expire) {
754*3f2c630cSAlan Somers 		struct timespec now, timeout;
755*3f2c630cSAlan Somers 
756*3f2c630cSAlan Somers 		err = cache_lookup(dvp, vpp, cnp, &timeout, NULL);
757*3f2c630cSAlan Somers 		switch (err) {
758*3f2c630cSAlan Somers 
759*3f2c630cSAlan Somers 		case -1:		/* positive match */
760*3f2c630cSAlan Somers 			getnanouptime(&now);
761*3f2c630cSAlan Somers 			if (timespeccmp(&timeout, &now, <=)) {
762*3f2c630cSAlan Somers 				atomic_add_acq_long(&fuse_lookup_cache_hits, 1);
763*3f2c630cSAlan Somers 			} else {
764*3f2c630cSAlan Somers 				/* Cache timeout */
765*3f2c630cSAlan Somers 				atomic_add_acq_long(&fuse_lookup_cache_misses,
766*3f2c630cSAlan Somers 					1);
767*3f2c630cSAlan Somers 				cache_purge(*vpp);
768*3f2c630cSAlan Somers 				break;
769*3f2c630cSAlan Somers 			}
770*3f2c630cSAlan Somers 			return 0;
771*3f2c630cSAlan Somers 
772*3f2c630cSAlan Somers 		case 0:		/* no match in cache */
773*3f2c630cSAlan Somers 			atomic_add_acq_long(&fuse_lookup_cache_misses, 1);
774*3f2c630cSAlan Somers 			break;
775*3f2c630cSAlan Somers 
776*3f2c630cSAlan Somers 		case ENOENT:		/* negative match */
777*3f2c630cSAlan Somers 			getnanouptime(&now);
778*3f2c630cSAlan Somers 			if (timespeccmp(&timeout, &now, >)) {
779*3f2c630cSAlan Somers 				/* Cache timeout */
780*3f2c630cSAlan Somers 				printf("Purging vnode %p name=%s\n", *vpp,
781*3f2c630cSAlan Somers 					cnp->cn_nameptr);
782*3f2c630cSAlan Somers 				fuse_internal_vnode_disappear(*vpp);
783*3f2c630cSAlan Somers 				break;
784*3f2c630cSAlan Somers 			}
785*3f2c630cSAlan Somers 			/* fall through */
786*3f2c630cSAlan Somers 		default:
787*3f2c630cSAlan Somers 			return err;
788*3f2c630cSAlan Somers 		}
789*3f2c630cSAlan Somers 	} else {
7905fe58019SAttilio Rao 		err = cache_lookup(dvp, vpp, cnp, NULL, NULL);
7915fe58019SAttilio Rao 		switch (err) {
7925fe58019SAttilio Rao 
7935fe58019SAttilio Rao 		case -1:		/* positive match */
7945fe58019SAttilio Rao 			atomic_add_acq_long(&fuse_lookup_cache_hits, 1);
7955fe58019SAttilio Rao 			return 0;
7965fe58019SAttilio Rao 
7975fe58019SAttilio Rao 		case 0:		/* no match in cache */
7985fe58019SAttilio Rao 			atomic_add_acq_long(&fuse_lookup_cache_misses, 1);
7995fe58019SAttilio Rao 			break;
8005fe58019SAttilio Rao 
8015fe58019SAttilio Rao 		case ENOENT:		/* negative match */
8025fe58019SAttilio Rao 			/* fall through */
8035fe58019SAttilio Rao 		default:
8045fe58019SAttilio Rao 			return err;
8055fe58019SAttilio Rao 		}
8065fe58019SAttilio Rao 	}
8075fe58019SAttilio Rao 	nid = VTOI(dvp);
8085fe58019SAttilio Rao 	fdisp_init(&fdi, cnp->cn_namelen + 1);
8095fe58019SAttilio Rao 	op = FUSE_LOOKUP;
8105fe58019SAttilio Rao 
8115fe58019SAttilio Rao calldaemon:
8125fe58019SAttilio Rao 	fdisp_make(&fdi, op, mp, nid, td, cred);
8135fe58019SAttilio Rao 
8145fe58019SAttilio Rao 	if (op == FUSE_LOOKUP) {
8155fe58019SAttilio Rao 		memcpy(fdi.indata, cnp->cn_nameptr, cnp->cn_namelen);
8165fe58019SAttilio Rao 		((char *)fdi.indata)[cnp->cn_namelen] = '\0';
8175fe58019SAttilio Rao 	}
8185fe58019SAttilio Rao 	lookup_err = fdisp_wait_answ(&fdi);
8195fe58019SAttilio Rao 
8205fe58019SAttilio Rao 	if ((op == FUSE_LOOKUP) && !lookup_err) {	/* lookup call succeeded */
8215fe58019SAttilio Rao 		nid = ((struct fuse_entry_out *)fdi.answ)->nodeid;
8225fe58019SAttilio Rao 		if (!nid) {
8235fe58019SAttilio Rao 			/*
8245fe58019SAttilio Rao 	                 * zero nodeid is the same as "not found",
8255fe58019SAttilio Rao 	                 * but it's also cacheable (which we keep
8265fe58019SAttilio Rao 	                 * keep on doing not as of writing this)
8275fe58019SAttilio Rao 	                 */
8286248288eSAlan Somers 			fdi.answ_stat = ENOENT;
8295fe58019SAttilio Rao 			lookup_err = ENOENT;
8305fe58019SAttilio Rao 		} else if (nid == FUSE_ROOT_ID) {
8315fe58019SAttilio Rao 			lookup_err = EINVAL;
8325fe58019SAttilio Rao 		}
8335fe58019SAttilio Rao 	}
8345fe58019SAttilio Rao 	if (lookup_err &&
8355fe58019SAttilio Rao 	    (!fdi.answ_stat || lookup_err != ENOENT || op != FUSE_LOOKUP)) {
8365fe58019SAttilio Rao 		fdisp_destroy(&fdi);
8375fe58019SAttilio Rao 		return lookup_err;
8385fe58019SAttilio Rao 	}
8395fe58019SAttilio Rao 	/* lookup_err, if non-zero, must be ENOENT at this point */
8405fe58019SAttilio Rao 
8415fe58019SAttilio Rao 	if (lookup_err) {
8425fe58019SAttilio Rao 
8435fe58019SAttilio Rao 		if ((nameiop == CREATE || nameiop == RENAME) && islastcn
8445fe58019SAttilio Rao 		     /* && directory dvp has not been removed */ ) {
8455fe58019SAttilio Rao 
8465fe58019SAttilio Rao 			if (vfs_isrdonly(mp)) {
8475fe58019SAttilio Rao 				err = EROFS;
8485fe58019SAttilio Rao 				goto out;
8495fe58019SAttilio Rao 			}
8505fe58019SAttilio Rao #if 0 /* THINK_ABOUT_THIS */
8515fe58019SAttilio Rao 			if ((err = fuse_internal_access(dvp, VWRITE, cred, td, &facp))) {
8525fe58019SAttilio Rao 				goto out;
8535fe58019SAttilio Rao 			}
8545fe58019SAttilio Rao #endif
8555fe58019SAttilio Rao 
8565fe58019SAttilio Rao 			/*
8575fe58019SAttilio Rao 	                 * Possibly record the position of a slot in the
8585fe58019SAttilio Rao 	                 * directory large enough for the new component name.
8595fe58019SAttilio Rao 	                 * This can be recorded in the vnode private data for
8605fe58019SAttilio Rao 	                 * dvp. Set the SAVENAME flag to hold onto the
8615fe58019SAttilio Rao 	                 * pathname for use later in VOP_CREATE or VOP_RENAME.
8625fe58019SAttilio Rao 	                 */
8635fe58019SAttilio Rao 			cnp->cn_flags |= SAVENAME;
8645fe58019SAttilio Rao 
8655fe58019SAttilio Rao 			err = EJUSTRETURN;
8665fe58019SAttilio Rao 			goto out;
8675fe58019SAttilio Rao 		}
8685fe58019SAttilio Rao 		/* Consider inserting name into cache. */
8695fe58019SAttilio Rao 
8705fe58019SAttilio Rao 		/*
8715fe58019SAttilio Rao 	         * No we can't use negative caching, as the fs
8725fe58019SAttilio Rao 	         * changes are out of our control.
8735fe58019SAttilio Rao 	         * False positives' falseness turns out just as things
8745fe58019SAttilio Rao 	         * go by, but false negatives' falseness doesn't.
8755fe58019SAttilio Rao 	         * (and aiding the caching mechanism with extra control
8765fe58019SAttilio Rao 	         * mechanisms comes quite close to beating the whole purpose
8775fe58019SAttilio Rao 	         * caching...)
8785fe58019SAttilio Rao 	         */
8795fe58019SAttilio Rao #if 0
8806c21f6edSKonstantin Belousov 		if ((cnp->cn_flags & MAKEENTRY) != 0) {
881cf169498SAlan Somers 			SDT_PROBE2(fuse, , vnops, trace, 1,
882cf169498SAlan Somers 				"inserting NULL into cache");
8835fe58019SAttilio Rao 			cache_enter(dvp, NULL, cnp);
8845fe58019SAttilio Rao 		}
8855fe58019SAttilio Rao #endif
8865fe58019SAttilio Rao 		err = ENOENT;
8875fe58019SAttilio Rao 		goto out;
8885fe58019SAttilio Rao 
8895fe58019SAttilio Rao 	} else {
8905fe58019SAttilio Rao 
8915fe58019SAttilio Rao 		/* !lookup_err */
8925fe58019SAttilio Rao 
8935fe58019SAttilio Rao 		struct fuse_entry_out *feo = NULL;
8945fe58019SAttilio Rao 		struct fuse_attr *fattr = NULL;
8955fe58019SAttilio Rao 
8965fe58019SAttilio Rao 		if (op == FUSE_GETATTR) {
8975fe58019SAttilio Rao 			fattr = &((struct fuse_attr_out *)fdi.answ)->attr;
8985fe58019SAttilio Rao 		} else {
8995fe58019SAttilio Rao 			feo = (struct fuse_entry_out *)fdi.answ;
9005fe58019SAttilio Rao 			fattr = &(feo->attr);
9015fe58019SAttilio Rao 		}
9025fe58019SAttilio Rao 
9035fe58019SAttilio Rao 		/*
9045fe58019SAttilio Rao 	         * If deleting, and at end of pathname, return parameters
9055fe58019SAttilio Rao 	         * which can be used to remove file.  If the wantparent flag
9065fe58019SAttilio Rao 	         * isn't set, we return only the directory, otherwise we go on
9075fe58019SAttilio Rao 	         * and lock the inode, being careful with ".".
9085fe58019SAttilio Rao 	         */
9095fe58019SAttilio Rao 		if (nameiop == DELETE && islastcn) {
9105fe58019SAttilio Rao 			/*
9115fe58019SAttilio Rao 	                 * Check for write access on directory.
9125fe58019SAttilio Rao 	                 */
9135fe58019SAttilio Rao 			facp.xuid = fattr->uid;
9145fe58019SAttilio Rao 			facp.facc_flags |= FACCESS_STICKY;
9155fe58019SAttilio Rao 			err = fuse_internal_access(dvp, VWRITE, &facp, td, cred);
9165fe58019SAttilio Rao 			facp.facc_flags &= ~FACCESS_XQUERIES;
9175fe58019SAttilio Rao 
9185fe58019SAttilio Rao 			if (err) {
9195fe58019SAttilio Rao 				goto out;
9205fe58019SAttilio Rao 			}
9215fe58019SAttilio Rao 			if (nid == VTOI(dvp)) {
9225fe58019SAttilio Rao 				vref(dvp);
9235fe58019SAttilio Rao 				*vpp = dvp;
9245fe58019SAttilio Rao 			} else {
92509176f09SConrad Meyer 				err = fuse_vnode_get(dvp->v_mount, feo, nid,
92609176f09SConrad Meyer 				    dvp, &vp, cnp, IFTOVT(fattr->mode));
9275fe58019SAttilio Rao 				if (err)
9285fe58019SAttilio Rao 					goto out;
9295fe58019SAttilio Rao 				*vpp = vp;
9305fe58019SAttilio Rao 			}
9315fe58019SAttilio Rao 
9325fe58019SAttilio Rao 			/*
9335fe58019SAttilio Rao 			 * Save the name for use in VOP_RMDIR and VOP_REMOVE
9345fe58019SAttilio Rao 			 * later.
9355fe58019SAttilio Rao 			 */
9365fe58019SAttilio Rao 			cnp->cn_flags |= SAVENAME;
9375fe58019SAttilio Rao 			goto out;
9385fe58019SAttilio Rao 
9395fe58019SAttilio Rao 		}
9405fe58019SAttilio Rao 		/*
9415fe58019SAttilio Rao 	         * If rewriting (RENAME), return the inode and the
9425fe58019SAttilio Rao 	         * information required to rewrite the present directory
9435fe58019SAttilio Rao 	         * Must get inode of directory entry to verify it's a
9445fe58019SAttilio Rao 	         * regular file, or empty directory.
9455fe58019SAttilio Rao 	         */
9465fe58019SAttilio Rao 		if (nameiop == RENAME && wantparent && islastcn) {
9475fe58019SAttilio Rao 
9485fe58019SAttilio Rao #if 0 /* THINK_ABOUT_THIS */
9495fe58019SAttilio Rao 			if ((err = fuse_internal_access(dvp, VWRITE, cred, td, &facp))) {
9505fe58019SAttilio Rao 				goto out;
9515fe58019SAttilio Rao 			}
9525fe58019SAttilio Rao #endif
9535fe58019SAttilio Rao 
9545fe58019SAttilio Rao 			/*
9555fe58019SAttilio Rao 	                 * Check for "."
9565fe58019SAttilio Rao 	                 */
9575fe58019SAttilio Rao 			if (nid == VTOI(dvp)) {
9585fe58019SAttilio Rao 				err = EISDIR;
9595fe58019SAttilio Rao 				goto out;
9605fe58019SAttilio Rao 			}
96109176f09SConrad Meyer 			err = fuse_vnode_get(vnode_mount(dvp), feo, nid, dvp,
96209176f09SConrad Meyer 			    &vp, cnp, IFTOVT(fattr->mode));
9635fe58019SAttilio Rao 			if (err) {
9645fe58019SAttilio Rao 				goto out;
9655fe58019SAttilio Rao 			}
9665fe58019SAttilio Rao 			*vpp = vp;
9675fe58019SAttilio Rao 			/*
9685fe58019SAttilio Rao 	                 * Save the name for use in VOP_RENAME later.
9695fe58019SAttilio Rao 	                 */
9705fe58019SAttilio Rao 			cnp->cn_flags |= SAVENAME;
9715fe58019SAttilio Rao 
9725fe58019SAttilio Rao 			goto out;
9735fe58019SAttilio Rao 		}
9745fe58019SAttilio Rao 		if (flags & ISDOTDOT) {
9755fe58019SAttilio Rao 			struct mount *mp;
9765fe58019SAttilio Rao 			int ltype;
9775fe58019SAttilio Rao 
9785fe58019SAttilio Rao 			/*
9795fe58019SAttilio Rao 			 * Expanded copy of vn_vget_ino() so that
9805fe58019SAttilio Rao 			 * fuse_vnode_get() can be used.
9815fe58019SAttilio Rao 			 */
9825fe58019SAttilio Rao 			mp = dvp->v_mount;
9835fe58019SAttilio Rao 			ltype = VOP_ISLOCKED(dvp);
9845fe58019SAttilio Rao 			err = vfs_busy(mp, MBF_NOWAIT);
9855fe58019SAttilio Rao 			if (err != 0) {
9865fe58019SAttilio Rao 				vfs_ref(mp);
9875fe58019SAttilio Rao 				VOP_UNLOCK(dvp, 0);
9885fe58019SAttilio Rao 				err = vfs_busy(mp, 0);
9895fe58019SAttilio Rao 				vn_lock(dvp, ltype | LK_RETRY);
9905fe58019SAttilio Rao 				vfs_rel(mp);
9915fe58019SAttilio Rao 				if (err)
9925fe58019SAttilio Rao 					goto out;
9935fe58019SAttilio Rao 				if ((dvp->v_iflag & VI_DOOMED) != 0) {
9945fe58019SAttilio Rao 					err = ENOENT;
9955fe58019SAttilio Rao 					vfs_unbusy(mp);
9965fe58019SAttilio Rao 					goto out;
9975fe58019SAttilio Rao 				}
9985fe58019SAttilio Rao 			}
9995fe58019SAttilio Rao 			VOP_UNLOCK(dvp, 0);
100009176f09SConrad Meyer 			err = fuse_vnode_get(vnode_mount(dvp), feo, nid, NULL,
100109176f09SConrad Meyer 			    &vp, cnp, IFTOVT(fattr->mode));
10025fe58019SAttilio Rao 			vfs_unbusy(mp);
10035fe58019SAttilio Rao 			vn_lock(dvp, ltype | LK_RETRY);
10045fe58019SAttilio Rao 			if ((dvp->v_iflag & VI_DOOMED) != 0) {
10055fe58019SAttilio Rao 				if (err == 0)
10065fe58019SAttilio Rao 					vput(vp);
10075fe58019SAttilio Rao 				err = ENOENT;
10085fe58019SAttilio Rao 			}
10095fe58019SAttilio Rao 			if (err)
10105fe58019SAttilio Rao 				goto out;
10115fe58019SAttilio Rao 			*vpp = vp;
10125fe58019SAttilio Rao 		} else if (nid == VTOI(dvp)) {
10135fe58019SAttilio Rao 			vref(dvp);
10145fe58019SAttilio Rao 			*vpp = dvp;
10155fe58019SAttilio Rao 		} else {
10163c324b94SConrad Meyer 			struct fuse_vnode_data *fvdat;
10173c324b94SConrad Meyer 
101809176f09SConrad Meyer 			err = fuse_vnode_get(vnode_mount(dvp), feo, nid, dvp,
101909176f09SConrad Meyer 			    &vp, cnp, IFTOVT(fattr->mode));
10205fe58019SAttilio Rao 			if (err) {
10215fe58019SAttilio Rao 				goto out;
10225fe58019SAttilio Rao 			}
10235fe58019SAttilio Rao 			fuse_vnode_setparent(vp, dvp);
10243c324b94SConrad Meyer 
10253c324b94SConrad Meyer 			/*
10263c324b94SConrad Meyer 			 * In the case where we are looking up a FUSE node
10273c324b94SConrad Meyer 			 * represented by an existing cached vnode, and the
10283c324b94SConrad Meyer 			 * true size reported by FUSE_LOOKUP doesn't match
10293c324b94SConrad Meyer 			 * the vnode's cached size, fix the vnode cache to
10303c324b94SConrad Meyer 			 * match the real object size.
10313c324b94SConrad Meyer 			 *
10323c324b94SConrad Meyer 			 * This can occur via FUSE distributed filesystems,
10333c324b94SConrad Meyer 			 * irregular files, etc.
10343c324b94SConrad Meyer 			 */
10353c324b94SConrad Meyer 			fvdat = VTOFUD(vp);
10363c324b94SConrad Meyer 			if (vnode_isreg(vp) &&
10373c324b94SConrad Meyer 			    fattr->size != fvdat->filesize) {
10383c324b94SConrad Meyer 				/*
10393c324b94SConrad Meyer 				 * The FN_SIZECHANGE flag reflects a dirty
10403c324b94SConrad Meyer 				 * append.  If userspace lets us know our cache
10413c324b94SConrad Meyer 				 * is invalid, that write was lost.  (Dirty
10423c324b94SConrad Meyer 				 * writes that do not cause append are also
10433c324b94SConrad Meyer 				 * lost, but we don't detect them here.)
10443c324b94SConrad Meyer 				 *
10453c324b94SConrad Meyer 				 * XXX: Maybe disable WB caching on this mount.
10463c324b94SConrad Meyer 				 */
10473c324b94SConrad Meyer 				if (fvdat->flag & FN_SIZECHANGE)
10483c324b94SConrad Meyer 					printf("%s: WB cache incoherent on "
10493c324b94SConrad Meyer 					    "%s!\n", __func__,
10503c324b94SConrad Meyer 					    vnode_mount(vp)->mnt_stat.f_mntonname);
10513c324b94SConrad Meyer 
10523c324b94SConrad Meyer 				(void)fuse_vnode_setsize(vp, cred, fattr->size);
10533c324b94SConrad Meyer 				fvdat->flag &= ~FN_SIZECHANGE;
10543c324b94SConrad Meyer 			}
10555fe58019SAttilio Rao 			*vpp = vp;
10565fe58019SAttilio Rao 		}
10575fe58019SAttilio Rao 
10585fe58019SAttilio Rao 		if (op == FUSE_GETATTR) {
10597e4844f7SAlan Somers 			struct fuse_attr_out *fao =
10607e4844f7SAlan Somers 				(struct fuse_attr_out*)fdi.answ;
10617e4844f7SAlan Somers 			fuse_internal_cache_attrs(*vpp,
10627e4844f7SAlan Somers 				&fao->attr, fao->attr_valid,
10637e4844f7SAlan Somers 				fao->attr_valid_nsec, NULL);
10645fe58019SAttilio Rao 		} else {
10657e4844f7SAlan Somers 			struct fuse_entry_out *feo =
10667e4844f7SAlan Somers 				(struct fuse_entry_out*)fdi.answ;
10677e4844f7SAlan Somers 			fuse_internal_cache_attrs(*vpp,
10687e4844f7SAlan Somers 				&feo->attr, feo->attr_valid,
10697e4844f7SAlan Somers 				feo->attr_valid_nsec, NULL);
10705fe58019SAttilio Rao 		}
10715fe58019SAttilio Rao 
10725fe58019SAttilio Rao 		/* Insert name into cache if appropriate. */
10735fe58019SAttilio Rao 
10745fe58019SAttilio Rao 		/*
10755fe58019SAttilio Rao 	         * Nooo, caching is evil. With caching, we can't avoid stale
10765fe58019SAttilio Rao 	         * information taking over the playground (cached info is not
10775fe58019SAttilio Rao 	         * just positive/negative, it does have qualitative aspects,
10785fe58019SAttilio Rao 	         * too). And a (VOP/FUSE)_GETATTR is always thrown anyway, when
10795fe58019SAttilio Rao 	         * walking down along cached path components, and that's not
10805fe58019SAttilio Rao 	         * any cheaper than FUSE_LOOKUP. This might change with
10815fe58019SAttilio Rao 	         * implementing kernel side attr caching, but... In Linux,
10825fe58019SAttilio Rao 	         * lookup results are not cached, and the daemon is bombarded
10835fe58019SAttilio Rao 	         * with FUSE_LOOKUPS on and on. This shows that by design, the
10845fe58019SAttilio Rao 	         * daemon is expected to handle frequent lookup queries
10855fe58019SAttilio Rao 	         * efficiently, do its caching in userspace, and so on.
10865fe58019SAttilio Rao 	         *
10875fe58019SAttilio Rao 	         * So just leave the name cache alone.
10885fe58019SAttilio Rao 	         */
10895fe58019SAttilio Rao 
10905fe58019SAttilio Rao 		/*
10915fe58019SAttilio Rao 	         * Well, now I know, Linux caches lookups, but with a
10925fe58019SAttilio Rao 	         * timeout... So it's the same thing as attribute caching:
10935fe58019SAttilio Rao 	         * we can deal with it when implement timeouts.
10945fe58019SAttilio Rao 	         */
10955fe58019SAttilio Rao #if 0
10965fe58019SAttilio Rao 		if (cnp->cn_flags & MAKEENTRY) {
10975fe58019SAttilio Rao 			cache_enter(dvp, *vpp, cnp);
10985fe58019SAttilio Rao 		}
10995fe58019SAttilio Rao #endif
11005fe58019SAttilio Rao 	}
11015fe58019SAttilio Rao out:
11025fe58019SAttilio Rao 	if (!lookup_err) {
11035fe58019SAttilio Rao 
11045fe58019SAttilio Rao 		/* No lookup error; need to clean up. */
11055fe58019SAttilio Rao 
11065fe58019SAttilio Rao 		if (err) {		/* Found inode; exit with no vnode. */
11075fe58019SAttilio Rao 			if (op == FUSE_LOOKUP) {
11085fe58019SAttilio Rao 				fuse_internal_forget_send(vnode_mount(dvp), td, cred,
11095fe58019SAttilio Rao 				    nid, 1);
11105fe58019SAttilio Rao 			}
11115fe58019SAttilio Rao 			fdisp_destroy(&fdi);
11125fe58019SAttilio Rao 			return err;
11135fe58019SAttilio Rao 		} else {
11145fe58019SAttilio Rao #ifndef NO_EARLY_PERM_CHECK_HACK
11155fe58019SAttilio Rao 			if (!islastcn) {
11165fe58019SAttilio Rao 				/*
11175fe58019SAttilio Rao 				 * We have the attributes of the next item
11185fe58019SAttilio Rao 				 * *now*, and it's a fact, and we do not
11195fe58019SAttilio Rao 				 * have to do extra work for it (ie, beg the
11205fe58019SAttilio Rao 				 * daemon), and it neither depends on such
11215fe58019SAttilio Rao 				 * accidental things like attr caching. So
11225fe58019SAttilio Rao 				 * the big idea: check credentials *now*,
11235fe58019SAttilio Rao 				 * not at the beginning of the next call to
11245fe58019SAttilio Rao 				 * lookup.
11255fe58019SAttilio Rao 				 *
11265fe58019SAttilio Rao 				 * The first item of the lookup chain (fs root)
11275fe58019SAttilio Rao 				 * won't be checked then here, of course, as
11285fe58019SAttilio Rao 				 * its never "the next". But go and see that
11295fe58019SAttilio Rao 				 * the root is taken care about at the very
11305fe58019SAttilio Rao 				 * beginning of this function.
11315fe58019SAttilio Rao 				 *
11325fe58019SAttilio Rao 				 * Now, given we want to do the access check
11335fe58019SAttilio Rao 				 * this way, one might ask: so then why not
11345fe58019SAttilio Rao 				 * do the access check just after fetching
11355fe58019SAttilio Rao 				 * the inode and its attributes from the
11365fe58019SAttilio Rao 				 * daemon? Why bother with producing the
11375fe58019SAttilio Rao 				 * corresponding vnode at all if something
11385fe58019SAttilio Rao 				 * is not OK? We know what's the deal as
11395fe58019SAttilio Rao 				 * soon as we get those attrs... There is
11405fe58019SAttilio Rao 				 * one bit of info though not given us by
11415fe58019SAttilio Rao 				 * the daemon: whether his response is
1142b3a15dddSPedro F. Giffuni 				 * authoritative or not... His response should
11435fe58019SAttilio Rao 				 * be ignored if something is mounted over
11445fe58019SAttilio Rao 				 * the dir in question. But that can be
11455fe58019SAttilio Rao 				 * known only by having the vnode...
11465fe58019SAttilio Rao 				 */
11475fe58019SAttilio Rao 				int tmpvtype = vnode_vtype(*vpp);
11485fe58019SAttilio Rao 
11495fe58019SAttilio Rao 				bzero(&facp, sizeof(facp));
11505fe58019SAttilio Rao 				/*the early perm check hack */
11515fe58019SAttilio Rao 				    facp.facc_flags |= FACCESS_VA_VALID;
11525fe58019SAttilio Rao 
11535fe58019SAttilio Rao 				if ((tmpvtype != VDIR) && (tmpvtype != VLNK)) {
11545fe58019SAttilio Rao 					err = ENOTDIR;
11555fe58019SAttilio Rao 				}
11565fe58019SAttilio Rao 				if (!err && !vnode_mountedhere(*vpp)) {
11575fe58019SAttilio Rao 					err = fuse_internal_access(*vpp, VEXEC, &facp, td, cred);
11585fe58019SAttilio Rao 				}
11595fe58019SAttilio Rao 				if (err) {
11605fe58019SAttilio Rao 					if (tmpvtype == VLNK)
1161cf169498SAlan Somers 						SDT_PROBE2(fuse, , vnops, trace,
1162cf169498SAlan Somers 						    1, "weird, permission "
1163cf169498SAlan Somers 						    "error with a symlink?");
11645fe58019SAttilio Rao 					vput(*vpp);
11655fe58019SAttilio Rao 					*vpp = NULL;
11665fe58019SAttilio Rao 				}
11675fe58019SAttilio Rao 			}
11685fe58019SAttilio Rao #endif
11695fe58019SAttilio Rao 		}
11705fe58019SAttilio Rao 	}
11715fe58019SAttilio Rao 	fdisp_destroy(&fdi);
11725fe58019SAttilio Rao 
11735fe58019SAttilio Rao 	return err;
11745fe58019SAttilio Rao }
11755fe58019SAttilio Rao 
11765fe58019SAttilio Rao /*
11775fe58019SAttilio Rao     struct vnop_mkdir_args {
11785fe58019SAttilio Rao 	struct vnode *a_dvp;
11795fe58019SAttilio Rao 	struct vnode **a_vpp;
11805fe58019SAttilio Rao 	struct componentname *a_cnp;
11815fe58019SAttilio Rao 	struct vattr *a_vap;
11825fe58019SAttilio Rao     };
11835fe58019SAttilio Rao */
11845fe58019SAttilio Rao static int
11855fe58019SAttilio Rao fuse_vnop_mkdir(struct vop_mkdir_args *ap)
11865fe58019SAttilio Rao {
11875fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
11885fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
11895fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
11905fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
11915fe58019SAttilio Rao 
11925fe58019SAttilio Rao 	struct fuse_mkdir_in fmdi;
11935fe58019SAttilio Rao 
11945fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
11955fe58019SAttilio Rao 		return ENXIO;
11965fe58019SAttilio Rao 	}
11975fe58019SAttilio Rao 	fmdi.mode = MAKEIMODE(vap->va_type, vap->va_mode);
11985fe58019SAttilio Rao 
1199d9454fabSAttilio Rao 	return (fuse_internal_newentry(dvp, vpp, cnp, FUSE_MKDIR, &fmdi,
1200d9454fabSAttilio Rao 	    sizeof(fmdi), VDIR));
12015fe58019SAttilio Rao }
12025fe58019SAttilio Rao 
12035fe58019SAttilio Rao /*
12045fe58019SAttilio Rao     struct vnop_mknod_args {
12055fe58019SAttilio Rao 	struct vnode *a_dvp;
12065fe58019SAttilio Rao 	struct vnode **a_vpp;
12075fe58019SAttilio Rao 	struct componentname *a_cnp;
12085fe58019SAttilio Rao 	struct vattr *a_vap;
12095fe58019SAttilio Rao     };
12105fe58019SAttilio Rao */
12115fe58019SAttilio Rao static int
12125fe58019SAttilio Rao fuse_vnop_mknod(struct vop_mknod_args *ap)
12135fe58019SAttilio Rao {
12145fe58019SAttilio Rao 
1215bf4d7084SAlan Somers 	struct vnode *dvp = ap->a_dvp;
1216bf4d7084SAlan Somers 	struct vnode **vpp = ap->a_vpp;
1217bf4d7084SAlan Somers 	struct componentname *cnp = ap->a_cnp;
1218bf4d7084SAlan Somers 	struct vattr *vap = ap->a_vap;
1219bf4d7084SAlan Somers 	struct fuse_mknod_in fmni;
12205fe58019SAttilio Rao 
1221bf4d7084SAlan Somers 	if (fuse_isdeadfs(dvp))
1222bf4d7084SAlan Somers 		return ENXIO;
1223bf4d7084SAlan Somers 
1224bf4d7084SAlan Somers 	fmni.mode = MAKEIMODE(vap->va_type, vap->va_mode);
1225bf4d7084SAlan Somers 	fmni.rdev = vap->va_rdev;
1226bf4d7084SAlan Somers 	return (fuse_internal_newentry(dvp, vpp, cnp, FUSE_MKNOD, &fmni,
1227bf4d7084SAlan Somers 	    sizeof(fmni), vap->va_type));
1228bf4d7084SAlan Somers }
12295fe58019SAttilio Rao 
12305fe58019SAttilio Rao /*
12315fe58019SAttilio Rao     struct vnop_open_args {
12325fe58019SAttilio Rao 	struct vnode *a_vp;
12335fe58019SAttilio Rao 	int  a_mode;
12345fe58019SAttilio Rao 	struct ucred *a_cred;
12355fe58019SAttilio Rao 	struct thread *a_td;
12365fe58019SAttilio Rao 	int a_fdidx; / struct file *a_fp;
12375fe58019SAttilio Rao     };
12385fe58019SAttilio Rao */
12395fe58019SAttilio Rao static int
12405fe58019SAttilio Rao fuse_vnop_open(struct vop_open_args *ap)
12415fe58019SAttilio Rao {
12425fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
12439e444871SAlan Somers 	int a_mode = ap->a_mode;
12445fe58019SAttilio Rao 	struct thread *td = ap->a_td;
12455fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
1246f8d4af10SAlan Somers 	pid_t pid = td->td_proc->p_pid;
12475fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat;
12485fe58019SAttilio Rao 
1249bf4d7084SAlan Somers 	if (fuse_isdeadfs(vp))
12505fe58019SAttilio Rao 		return ENXIO;
1251bf4d7084SAlan Somers 	if (vp->v_type == VCHR || vp->v_type == VBLK || vp->v_type == VFIFO)
1252bf4d7084SAlan Somers 		return (EOPNOTSUPP);
12539e444871SAlan Somers 	if ((a_mode & (FREAD | FWRITE | FEXEC)) == 0)
1254e7df9886SConrad Meyer 		return EINVAL;
1255e7df9886SConrad Meyer 
12565fe58019SAttilio Rao 	fvdat = VTOFUD(vp);
12575fe58019SAttilio Rao 
12589e444871SAlan Somers 	if (fuse_filehandle_validrw(vp, a_mode, cred, pid)) {
12595fccbf31SAlan Somers 		fuse_vnode_open(vp, 0, td);
12605fe58019SAttilio Rao 		return 0;
12615fe58019SAttilio Rao 	}
12625fe58019SAttilio Rao 
12639e444871SAlan Somers 	return fuse_filehandle_open(vp, a_mode, NULL, td, cred);
12645fe58019SAttilio Rao }
12655fe58019SAttilio Rao 
1266746c92e0SJohn Baldwin static int
1267746c92e0SJohn Baldwin fuse_vnop_pathconf(struct vop_pathconf_args *ap)
1268746c92e0SJohn Baldwin {
1269746c92e0SJohn Baldwin 
1270746c92e0SJohn Baldwin 	switch (ap->a_name) {
1271746c92e0SJohn Baldwin 	case _PC_FILESIZEBITS:
1272746c92e0SJohn Baldwin 		*ap->a_retval = 64;
1273746c92e0SJohn Baldwin 		return (0);
1274599afe53SJohn Baldwin 	case _PC_NAME_MAX:
1275599afe53SJohn Baldwin 		*ap->a_retval = NAME_MAX;
1276599afe53SJohn Baldwin 		return (0);
1277599afe53SJohn Baldwin 	case _PC_LINK_MAX:
1278f83f3d79SJohn Baldwin 		*ap->a_retval = MIN(LONG_MAX, FUSE_LINK_MAX);
1279599afe53SJohn Baldwin 		return (0);
1280746c92e0SJohn Baldwin 	case _PC_SYMLINK_MAX:
1281746c92e0SJohn Baldwin 		*ap->a_retval = MAXPATHLEN;
1282746c92e0SJohn Baldwin 		return (0);
1283746c92e0SJohn Baldwin 	case _PC_NO_TRUNC:
1284746c92e0SJohn Baldwin 		*ap->a_retval = 1;
1285746c92e0SJohn Baldwin 		return (0);
1286746c92e0SJohn Baldwin 	default:
1287746c92e0SJohn Baldwin 		return (vop_stdpathconf(ap));
1288746c92e0SJohn Baldwin 	}
1289746c92e0SJohn Baldwin }
1290746c92e0SJohn Baldwin 
12915fe58019SAttilio Rao /*
12925fe58019SAttilio Rao     struct vnop_read_args {
12935fe58019SAttilio Rao 	struct vnode *a_vp;
12945fe58019SAttilio Rao 	struct uio *a_uio;
12955fe58019SAttilio Rao 	int  a_ioflag;
12965fe58019SAttilio Rao 	struct ucred *a_cred;
12975fe58019SAttilio Rao     };
12985fe58019SAttilio Rao */
12995fe58019SAttilio Rao static int
13005fe58019SAttilio Rao fuse_vnop_read(struct vop_read_args *ap)
13015fe58019SAttilio Rao {
13025fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
13035fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
13045fe58019SAttilio Rao 	int ioflag = ap->a_ioflag;
13055fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
1306f8d4af10SAlan Somers 	pid_t pid = curthread->td_proc->p_pid;
13075fe58019SAttilio Rao 
13085fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
13095fe58019SAttilio Rao 		return ENXIO;
13105fe58019SAttilio Rao 	}
1311ead063e0SEdward Tomasz Napierala 
1312ead063e0SEdward Tomasz Napierala 	if (VTOFUD(vp)->flag & FN_DIRECTIO) {
1313ead063e0SEdward Tomasz Napierala 		ioflag |= IO_DIRECT;
1314ead063e0SEdward Tomasz Napierala 	}
1315ead063e0SEdward Tomasz Napierala 
1316f8d4af10SAlan Somers 	return fuse_io_dispatch(vp, uio, ioflag, cred, pid);
13175fe58019SAttilio Rao }
13185fe58019SAttilio Rao 
13195fe58019SAttilio Rao /*
13205fe58019SAttilio Rao     struct vnop_readdir_args {
13215fe58019SAttilio Rao 	struct vnode *a_vp;
13225fe58019SAttilio Rao 	struct uio *a_uio;
13235fe58019SAttilio Rao 	struct ucred *a_cred;
13245fe58019SAttilio Rao 	int *a_eofflag;
13255fe58019SAttilio Rao 	int *ncookies;
13265fe58019SAttilio Rao 	u_long **a_cookies;
13275fe58019SAttilio Rao     };
13285fe58019SAttilio Rao */
13295fe58019SAttilio Rao static int
13305fe58019SAttilio Rao fuse_vnop_readdir(struct vop_readdir_args *ap)
13315fe58019SAttilio Rao {
13325fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
13335fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
13345fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
13355fe58019SAttilio Rao 	struct fuse_filehandle *fufh = NULL;
13365fe58019SAttilio Rao 	struct fuse_iov cookediov;
13375fe58019SAttilio Rao 	int err = 0;
1338f8d4af10SAlan Somers 	pid_t pid = curthread->td_proc->p_pid;
13395fe58019SAttilio Rao 
13405fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
13415fe58019SAttilio Rao 		return ENXIO;
13425fe58019SAttilio Rao 	}
13435fe58019SAttilio Rao 	if (				/* XXXIP ((uio_iovcnt(uio) > 1)) || */
13445fe58019SAttilio Rao 	    (uio_resid(uio) < sizeof(struct dirent))) {
13455fe58019SAttilio Rao 		return EINVAL;
13465fe58019SAttilio Rao 	}
13475fe58019SAttilio Rao 
134835cf0e7eSAlan Somers 	err = fuse_filehandle_get_dir(vp, &fufh, cred, pid);
134935cf0e7eSAlan Somers 	if (err)
13505fe58019SAttilio Rao 		return (err);
13515fe58019SAttilio Rao #define DIRCOOKEDSIZE FUSE_DIRENT_ALIGN(FUSE_NAME_OFFSET + MAXNAMLEN + 1)
13525fe58019SAttilio Rao 	fiov_init(&cookediov, DIRCOOKEDSIZE);
13535fe58019SAttilio Rao 
13545fe58019SAttilio Rao 	err = fuse_internal_readdir(vp, uio, fufh, &cookediov);
13555fe58019SAttilio Rao 
13565fe58019SAttilio Rao 	fiov_teardown(&cookediov);
13575ec10aa5SAlan Somers 
13585fe58019SAttilio Rao 	return err;
13595fe58019SAttilio Rao }
13605fe58019SAttilio Rao 
13615fe58019SAttilio Rao /*
13625fe58019SAttilio Rao     struct vnop_readlink_args {
13635fe58019SAttilio Rao 	struct vnode *a_vp;
13645fe58019SAttilio Rao 	struct uio *a_uio;
13655fe58019SAttilio Rao 	struct ucred *a_cred;
13665fe58019SAttilio Rao     };
13675fe58019SAttilio Rao */
13685fe58019SAttilio Rao static int
13695fe58019SAttilio Rao fuse_vnop_readlink(struct vop_readlink_args *ap)
13705fe58019SAttilio Rao {
13715fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
13725fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
13735fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
13745fe58019SAttilio Rao 
13755fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
13765fe58019SAttilio Rao 	int err;
13775fe58019SAttilio Rao 
13785fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
13795fe58019SAttilio Rao 		return ENXIO;
13805fe58019SAttilio Rao 	}
13815fe58019SAttilio Rao 	if (!vnode_islnk(vp)) {
13825fe58019SAttilio Rao 		return EINVAL;
13835fe58019SAttilio Rao 	}
13845fe58019SAttilio Rao 	fdisp_init(&fdi, 0);
13855fe58019SAttilio Rao 	err = fdisp_simple_putget_vp(&fdi, FUSE_READLINK, vp, curthread, cred);
13865fe58019SAttilio Rao 	if (err) {
13875fe58019SAttilio Rao 		goto out;
13885fe58019SAttilio Rao 	}
13895fe58019SAttilio Rao 	if (((char *)fdi.answ)[0] == '/' &&
13905fe58019SAttilio Rao 	    fuse_get_mpdata(vnode_mount(vp))->dataflags & FSESS_PUSH_SYMLINKS_IN) {
13915fe58019SAttilio Rao 		char *mpth = vnode_mount(vp)->mnt_stat.f_mntonname;
13925fe58019SAttilio Rao 
13935fe58019SAttilio Rao 		err = uiomove(mpth, strlen(mpth), uio);
13945fe58019SAttilio Rao 	}
13955fe58019SAttilio Rao 	if (!err) {
13965fe58019SAttilio Rao 		err = uiomove(fdi.answ, fdi.iosize, uio);
13975fe58019SAttilio Rao 	}
13985fe58019SAttilio Rao out:
13995fe58019SAttilio Rao 	fdisp_destroy(&fdi);
14005fe58019SAttilio Rao 	return err;
14015fe58019SAttilio Rao }
14025fe58019SAttilio Rao 
14035fe58019SAttilio Rao /*
14045fe58019SAttilio Rao     struct vnop_reclaim_args {
14055fe58019SAttilio Rao 	struct vnode *a_vp;
14065fe58019SAttilio Rao 	struct thread *a_td;
14075fe58019SAttilio Rao     };
14085fe58019SAttilio Rao */
14095fe58019SAttilio Rao static int
14105fe58019SAttilio Rao fuse_vnop_reclaim(struct vop_reclaim_args *ap)
14115fe58019SAttilio Rao {
14125fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
14135fe58019SAttilio Rao 	struct thread *td = ap->a_td;
14145fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(vp);
14155ec10aa5SAlan Somers 	struct fuse_filehandle *fufh, *fufh_tmp;
14165fe58019SAttilio Rao 
14175fe58019SAttilio Rao 	if (!fvdat) {
14185fe58019SAttilio Rao 		panic("FUSE: no vnode data during recycling");
14195fe58019SAttilio Rao 	}
14205ec10aa5SAlan Somers 	LIST_FOREACH_SAFE(fufh, &fvdat->handles, next, fufh_tmp) {
14215ec10aa5SAlan Somers 		printf("FUSE: vnode being reclaimed with open fufh "
14229e444871SAlan Somers 			"(type=%#x)", fufh->fufh_type);
14235ec10aa5SAlan Somers 		fuse_filehandle_close(vp, fufh, td, NULL);
14245fe58019SAttilio Rao 	}
14255fe58019SAttilio Rao 
14265fe58019SAttilio Rao 	if ((!fuse_isdeadfs(vp)) && (fvdat->nlookup)) {
14275fe58019SAttilio Rao 		fuse_internal_forget_send(vnode_mount(vp), td, NULL, VTOI(vp),
14285fe58019SAttilio Rao 		    fvdat->nlookup);
14295fe58019SAttilio Rao 	}
14305fe58019SAttilio Rao 	fuse_vnode_setparent(vp, NULL);
14315fe58019SAttilio Rao 	cache_purge(vp);
14325fe58019SAttilio Rao 	vfs_hash_remove(vp);
14335fe58019SAttilio Rao 	vnode_destroy_vobject(vp);
14345fe58019SAttilio Rao 	fuse_vnode_destroy(vp);
14355fe58019SAttilio Rao 
14365fe58019SAttilio Rao 	return 0;
14375fe58019SAttilio Rao }
14385fe58019SAttilio Rao 
14395fe58019SAttilio Rao /*
14405fe58019SAttilio Rao     struct vnop_remove_args {
14415fe58019SAttilio Rao 	struct vnode *a_dvp;
14425fe58019SAttilio Rao 	struct vnode *a_vp;
14435fe58019SAttilio Rao 	struct componentname *a_cnp;
14445fe58019SAttilio Rao     };
14455fe58019SAttilio Rao */
14465fe58019SAttilio Rao static int
14475fe58019SAttilio Rao fuse_vnop_remove(struct vop_remove_args *ap)
14485fe58019SAttilio Rao {
14495fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
14505fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
14515fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
14525fe58019SAttilio Rao 
14535fe58019SAttilio Rao 	int err;
14545fe58019SAttilio Rao 
14555fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
14565fe58019SAttilio Rao 		return ENXIO;
14575fe58019SAttilio Rao 	}
14585fe58019SAttilio Rao 	if (vnode_isdir(vp)) {
14595fe58019SAttilio Rao 		return EPERM;
14605fe58019SAttilio Rao 	}
14615fe58019SAttilio Rao 	cache_purge(vp);
14625fe58019SAttilio Rao 
14635fe58019SAttilio Rao 	err = fuse_internal_remove(dvp, vp, cnp, FUSE_UNLINK);
14645fe58019SAttilio Rao 
1465d9454fabSAttilio Rao 	if (err == 0)
14665fe58019SAttilio Rao 		fuse_internal_vnode_disappear(vp);
14675fe58019SAttilio Rao 	return err;
14685fe58019SAttilio Rao }
14695fe58019SAttilio Rao 
14705fe58019SAttilio Rao /*
14715fe58019SAttilio Rao     struct vnop_rename_args {
14725fe58019SAttilio Rao 	struct vnode *a_fdvp;
14735fe58019SAttilio Rao 	struct vnode *a_fvp;
14745fe58019SAttilio Rao 	struct componentname *a_fcnp;
14755fe58019SAttilio Rao 	struct vnode *a_tdvp;
14765fe58019SAttilio Rao 	struct vnode *a_tvp;
14775fe58019SAttilio Rao 	struct componentname *a_tcnp;
14785fe58019SAttilio Rao     };
14795fe58019SAttilio Rao */
14805fe58019SAttilio Rao static int
14815fe58019SAttilio Rao fuse_vnop_rename(struct vop_rename_args *ap)
14825fe58019SAttilio Rao {
14835fe58019SAttilio Rao 	struct vnode *fdvp = ap->a_fdvp;
14845fe58019SAttilio Rao 	struct vnode *fvp = ap->a_fvp;
14855fe58019SAttilio Rao 	struct componentname *fcnp = ap->a_fcnp;
14865fe58019SAttilio Rao 	struct vnode *tdvp = ap->a_tdvp;
14875fe58019SAttilio Rao 	struct vnode *tvp = ap->a_tvp;
14885fe58019SAttilio Rao 	struct componentname *tcnp = ap->a_tcnp;
14895fe58019SAttilio Rao 	struct fuse_data *data;
14905fe58019SAttilio Rao 
14915fe58019SAttilio Rao 	int err = 0;
14925fe58019SAttilio Rao 
14935fe58019SAttilio Rao 	if (fuse_isdeadfs(fdvp)) {
14945fe58019SAttilio Rao 		return ENXIO;
14955fe58019SAttilio Rao 	}
14965fe58019SAttilio Rao 	if (fvp->v_mount != tdvp->v_mount ||
14975fe58019SAttilio Rao 	    (tvp && fvp->v_mount != tvp->v_mount)) {
1498cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1, "cross-device rename");
14995fe58019SAttilio Rao 		err = EXDEV;
15005fe58019SAttilio Rao 		goto out;
15015fe58019SAttilio Rao 	}
15025fe58019SAttilio Rao 	cache_purge(fvp);
15035fe58019SAttilio Rao 
15045fe58019SAttilio Rao 	/*
15055fe58019SAttilio Rao 	 * FUSE library is expected to check if target directory is not
15065fe58019SAttilio Rao 	 * under the source directory in the file system tree.
15075fe58019SAttilio Rao 	 * Linux performs this check at VFS level.
15085fe58019SAttilio Rao 	 */
15095fe58019SAttilio Rao 	data = fuse_get_mpdata(vnode_mount(tdvp));
15105fe58019SAttilio Rao 	sx_xlock(&data->rename_lock);
15115fe58019SAttilio Rao 	err = fuse_internal_rename(fdvp, fcnp, tdvp, tcnp);
15125fe58019SAttilio Rao 	if (err == 0) {
1513d9454fabSAttilio Rao 		if (tdvp != fdvp)
15145fe58019SAttilio Rao 			fuse_vnode_setparent(fvp, tdvp);
15155fe58019SAttilio Rao 		if (tvp != NULL)
15165fe58019SAttilio Rao 			fuse_vnode_setparent(tvp, NULL);
15175fe58019SAttilio Rao 	}
15185fe58019SAttilio Rao 	sx_unlock(&data->rename_lock);
15195fe58019SAttilio Rao 
15205fe58019SAttilio Rao 	if (tvp != NULL && tvp != fvp) {
15215fe58019SAttilio Rao 		cache_purge(tvp);
15225fe58019SAttilio Rao 	}
15235fe58019SAttilio Rao 	if (vnode_isdir(fvp)) {
15245fe58019SAttilio Rao 		if ((tvp != NULL) && vnode_isdir(tvp)) {
15255fe58019SAttilio Rao 			cache_purge(tdvp);
15265fe58019SAttilio Rao 		}
15275fe58019SAttilio Rao 		cache_purge(fdvp);
15285fe58019SAttilio Rao 	}
15295fe58019SAttilio Rao out:
15305fe58019SAttilio Rao 	if (tdvp == tvp) {
15315fe58019SAttilio Rao 		vrele(tdvp);
15325fe58019SAttilio Rao 	} else {
15335fe58019SAttilio Rao 		vput(tdvp);
15345fe58019SAttilio Rao 	}
15355fe58019SAttilio Rao 	if (tvp != NULL) {
15365fe58019SAttilio Rao 		vput(tvp);
15375fe58019SAttilio Rao 	}
15385fe58019SAttilio Rao 	vrele(fdvp);
15395fe58019SAttilio Rao 	vrele(fvp);
15405fe58019SAttilio Rao 
15415fe58019SAttilio Rao 	return err;
15425fe58019SAttilio Rao }
15435fe58019SAttilio Rao 
15445fe58019SAttilio Rao /*
15455fe58019SAttilio Rao     struct vnop_rmdir_args {
15465fe58019SAttilio Rao 	    struct vnode *a_dvp;
15475fe58019SAttilio Rao 	    struct vnode *a_vp;
15485fe58019SAttilio Rao 	    struct componentname *a_cnp;
15495fe58019SAttilio Rao     } *ap;
15505fe58019SAttilio Rao */
15515fe58019SAttilio Rao static int
15525fe58019SAttilio Rao fuse_vnop_rmdir(struct vop_rmdir_args *ap)
15535fe58019SAttilio Rao {
15545fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
15555fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
15565fe58019SAttilio Rao 
15575fe58019SAttilio Rao 	int err;
15585fe58019SAttilio Rao 
15595fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
15605fe58019SAttilio Rao 		return ENXIO;
15615fe58019SAttilio Rao 	}
15625fe58019SAttilio Rao 	if (VTOFUD(vp) == VTOFUD(dvp)) {
15635fe58019SAttilio Rao 		return EINVAL;
15645fe58019SAttilio Rao 	}
15655fe58019SAttilio Rao 	err = fuse_internal_remove(dvp, vp, ap->a_cnp, FUSE_RMDIR);
15665fe58019SAttilio Rao 
1567d9454fabSAttilio Rao 	if (err == 0)
15685fe58019SAttilio Rao 		fuse_internal_vnode_disappear(vp);
15695fe58019SAttilio Rao 	return err;
15705fe58019SAttilio Rao }
15715fe58019SAttilio Rao 
15725fe58019SAttilio Rao /*
15735fe58019SAttilio Rao     struct vnop_setattr_args {
15745fe58019SAttilio Rao 	struct vnode *a_vp;
15755fe58019SAttilio Rao 	struct vattr *a_vap;
15765fe58019SAttilio Rao 	struct ucred *a_cred;
15775fe58019SAttilio Rao 	struct thread *a_td;
15785fe58019SAttilio Rao     };
15795fe58019SAttilio Rao */
15805fe58019SAttilio Rao static int
15815fe58019SAttilio Rao fuse_vnop_setattr(struct vop_setattr_args *ap)
15825fe58019SAttilio Rao {
15835fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
15845fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
15855fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
15865fe58019SAttilio Rao 	struct thread *td = curthread;
15875fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
15885fe58019SAttilio Rao 	struct fuse_setattr_in *fsai;
15895fe58019SAttilio Rao 	struct fuse_access_param facp;
1590f8d4af10SAlan Somers 	pid_t pid = td->td_proc->p_pid;
15915fe58019SAttilio Rao 
15925fe58019SAttilio Rao 	int err = 0;
15935fe58019SAttilio Rao 	enum vtype vtyp;
15945fe58019SAttilio Rao 	int sizechanged = 0;
15955fe58019SAttilio Rao 	uint64_t newsize = 0;
15965fe58019SAttilio Rao 
15975fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
15985fe58019SAttilio Rao 		return ENXIO;
15995fe58019SAttilio Rao 	}
16005fe58019SAttilio Rao 	fdisp_init(&fdi, sizeof(*fsai));
16015fe58019SAttilio Rao 	fdisp_make_vp(&fdi, FUSE_SETATTR, vp, td, cred);
16025fe58019SAttilio Rao 	fsai = fdi.indata;
16035fe58019SAttilio Rao 	fsai->valid = 0;
16045fe58019SAttilio Rao 
16055fe58019SAttilio Rao 	bzero(&facp, sizeof(facp));
16065fe58019SAttilio Rao 
16075fe58019SAttilio Rao 	facp.xuid = vap->va_uid;
16085fe58019SAttilio Rao 	facp.xgid = vap->va_gid;
16095fe58019SAttilio Rao 
16105fe58019SAttilio Rao 	if (vap->va_uid != (uid_t)VNOVAL) {
16115fe58019SAttilio Rao 		facp.facc_flags |= FACCESS_CHOWN;
16125fe58019SAttilio Rao 		fsai->uid = vap->va_uid;
16135fe58019SAttilio Rao 		fsai->valid |= FATTR_UID;
16145fe58019SAttilio Rao 	}
16155fe58019SAttilio Rao 	if (vap->va_gid != (gid_t)VNOVAL) {
16165fe58019SAttilio Rao 		facp.facc_flags |= FACCESS_CHOWN;
16175fe58019SAttilio Rao 		fsai->gid = vap->va_gid;
16185fe58019SAttilio Rao 		fsai->valid |= FATTR_GID;
16195fe58019SAttilio Rao 	}
16205fe58019SAttilio Rao 	if (vap->va_size != VNOVAL) {
16215fe58019SAttilio Rao 
16225fe58019SAttilio Rao 		struct fuse_filehandle *fufh = NULL;
16235fe58019SAttilio Rao 
16245fe58019SAttilio Rao 		/*Truncate to a new value. */
16255fe58019SAttilio Rao 		    fsai->size = vap->va_size;
16265fe58019SAttilio Rao 		sizechanged = 1;
16275fe58019SAttilio Rao 		newsize = vap->va_size;
16285fe58019SAttilio Rao 		fsai->valid |= FATTR_SIZE;
16295fe58019SAttilio Rao 
16309f10f423SAlan Somers 		fuse_filehandle_getrw(vp, FWRITE, &fufh, cred, pid);
16315fe58019SAttilio Rao 		if (fufh) {
16325fe58019SAttilio Rao 			fsai->fh = fufh->fh_id;
16335fe58019SAttilio Rao 			fsai->valid |= FATTR_FH;
16345fe58019SAttilio Rao 		}
16355fe58019SAttilio Rao 	}
16365fe58019SAttilio Rao 	if (vap->va_atime.tv_sec != VNOVAL) {
16375fe58019SAttilio Rao 		fsai->atime = vap->va_atime.tv_sec;
16385fe58019SAttilio Rao 		fsai->atimensec = vap->va_atime.tv_nsec;
16395fe58019SAttilio Rao 		fsai->valid |= FATTR_ATIME;
16405fe58019SAttilio Rao 	}
16415fe58019SAttilio Rao 	if (vap->va_mtime.tv_sec != VNOVAL) {
16425fe58019SAttilio Rao 		fsai->mtime = vap->va_mtime.tv_sec;
16435fe58019SAttilio Rao 		fsai->mtimensec = vap->va_mtime.tv_nsec;
16445fe58019SAttilio Rao 		fsai->valid |= FATTR_MTIME;
16455fe58019SAttilio Rao 	}
16465fe58019SAttilio Rao 	if (vap->va_mode != (mode_t)VNOVAL) {
16475fe58019SAttilio Rao 		fsai->mode = vap->va_mode & ALLPERMS;
16485fe58019SAttilio Rao 		fsai->valid |= FATTR_MODE;
16495fe58019SAttilio Rao 	}
16505fe58019SAttilio Rao 	if (!fsai->valid) {
16515fe58019SAttilio Rao 		goto out;
16525fe58019SAttilio Rao 	}
16535fe58019SAttilio Rao 	vtyp = vnode_vtype(vp);
16545fe58019SAttilio Rao 
16555fe58019SAttilio Rao 	if (fsai->valid & FATTR_SIZE && vtyp == VDIR) {
16565fe58019SAttilio Rao 		err = EISDIR;
16575fe58019SAttilio Rao 		goto out;
16585fe58019SAttilio Rao 	}
16595fe58019SAttilio Rao 	if (vfs_isrdonly(vnode_mount(vp)) && (fsai->valid & ~FATTR_SIZE || vtyp == VREG)) {
16605fe58019SAttilio Rao 		err = EROFS;
16615fe58019SAttilio Rao 		goto out;
16625fe58019SAttilio Rao 	}
16635fe58019SAttilio Rao 
1664d9454fabSAttilio Rao 	if ((err = fdisp_wait_answ(&fdi)))
16655fe58019SAttilio Rao 		goto out;
16665fe58019SAttilio Rao 	vtyp = IFTOVT(((struct fuse_attr_out *)fdi.answ)->attr.mode);
16675fe58019SAttilio Rao 
16685fe58019SAttilio Rao 	if (vnode_vtype(vp) != vtyp) {
16695fe58019SAttilio Rao 		if (vnode_vtype(vp) == VNON && vtyp != VNON) {
1670cf169498SAlan Somers 			SDT_PROBE2(fuse, , vnops, trace, 1, "FUSE: Dang! "
1671cf169498SAlan Somers 				"vnode_vtype is VNON and vtype isn't.");
16725fe58019SAttilio Rao 		} else {
16735fe58019SAttilio Rao 			/*
16745fe58019SAttilio Rao 	                 * STALE vnode, ditch
16755fe58019SAttilio Rao 	                 *
1676cf169498SAlan Somers 			 * The vnode has changed its type "behind our back".
1677cf169498SAlan Somers 			 * There's nothing really we can do, so let us just
1678cf169498SAlan Somers 			 * force an internal revocation and tell the caller to
1679cf169498SAlan Somers 			 * try again, if interested.
16805fe58019SAttilio Rao 	                 */
16815fe58019SAttilio Rao 			fuse_internal_vnode_disappear(vp);
16825fe58019SAttilio Rao 			err = EAGAIN;
16835fe58019SAttilio Rao 		}
16845fe58019SAttilio Rao 	}
16857e4844f7SAlan Somers 	if (err == 0) {
16867e4844f7SAlan Somers 		struct fuse_attr_out *fao = (struct fuse_attr_out*)fdi.answ;
16877e4844f7SAlan Somers 		fuse_internal_cache_attrs(vp, &fao->attr, fao->attr_valid,
16887e4844f7SAlan Somers 			fao->attr_valid_nsec, NULL);
16897e4844f7SAlan Somers 	}
1690194e691aSConrad Meyer 
16915fe58019SAttilio Rao out:
16925fe58019SAttilio Rao 	fdisp_destroy(&fdi);
16935fe58019SAttilio Rao 	if (!err && sizechanged) {
16945fe58019SAttilio Rao 		fuse_vnode_setsize(vp, cred, newsize);
16955fe58019SAttilio Rao 		VTOFUD(vp)->flag &= ~FN_SIZECHANGE;
16965fe58019SAttilio Rao 	}
16975fe58019SAttilio Rao 	return err;
16985fe58019SAttilio Rao }
16995fe58019SAttilio Rao 
17005fe58019SAttilio Rao /*
17015fe58019SAttilio Rao     struct vnop_strategy_args {
17025fe58019SAttilio Rao 	struct vnode *a_vp;
17035fe58019SAttilio Rao 	struct buf *a_bp;
17045fe58019SAttilio Rao     };
17055fe58019SAttilio Rao */
17065fe58019SAttilio Rao static int
17075fe58019SAttilio Rao fuse_vnop_strategy(struct vop_strategy_args *ap)
17085fe58019SAttilio Rao {
17095fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
17105fe58019SAttilio Rao 	struct buf *bp = ap->a_bp;
17115fe58019SAttilio Rao 
17125fe58019SAttilio Rao 	if (!vp || fuse_isdeadfs(vp)) {
17135fe58019SAttilio Rao 		bp->b_ioflags |= BIO_ERROR;
17145fe58019SAttilio Rao 		bp->b_error = ENXIO;
17155fe58019SAttilio Rao 		bufdone(bp);
171698852a32SAlan Somers 		return 0;
17175fe58019SAttilio Rao 	}
1718f203d173SAlan Somers 	if (bp->b_iocmd == BIO_WRITE) {
1719f203d173SAlan Somers 		int err;
17205fe58019SAttilio Rao 
1721f203d173SAlan Somers 		err = fuse_vnode_refreshsize(vp, NOCRED);
1722f203d173SAlan Somers 		if (err) {
1723f203d173SAlan Somers 			bp->b_ioflags |= BIO_ERROR;
1724f203d173SAlan Somers 			bp->b_error = err;
172598852a32SAlan Somers 			bufdone(bp);
1726f203d173SAlan Somers 			return 0;
1727f203d173SAlan Somers 		}
1728f203d173SAlan Somers 	}
17295fe58019SAttilio Rao 
17305fe58019SAttilio Rao 	/*
1731f203d173SAlan Somers 	 * VOP_STRATEGY always returns zero and signals error via bp->b_ioflags.
1732f203d173SAlan Somers 	 * fuse_io_strategy sets bp's error fields
17335fe58019SAttilio Rao 	 */
1734f203d173SAlan Somers 	(void)fuse_io_strategy(vp, bp);
1735f203d173SAlan Somers 
17365fe58019SAttilio Rao 	return 0;
17375fe58019SAttilio Rao }
17385fe58019SAttilio Rao 
17395fe58019SAttilio Rao 
17405fe58019SAttilio Rao /*
17415fe58019SAttilio Rao     struct vnop_symlink_args {
17425fe58019SAttilio Rao 	struct vnode *a_dvp;
17435fe58019SAttilio Rao 	struct vnode **a_vpp;
17445fe58019SAttilio Rao 	struct componentname *a_cnp;
17455fe58019SAttilio Rao 	struct vattr *a_vap;
17465fe58019SAttilio Rao 	char *a_target;
17475fe58019SAttilio Rao     };
17485fe58019SAttilio Rao */
17495fe58019SAttilio Rao static int
17505fe58019SAttilio Rao fuse_vnop_symlink(struct vop_symlink_args *ap)
17515fe58019SAttilio Rao {
17525fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
17535fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
17545fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
17551493c2eeSBrooks Davis 	const char *target = ap->a_target;
17565fe58019SAttilio Rao 
17575fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
17585fe58019SAttilio Rao 
17595fe58019SAttilio Rao 	int err;
17605fe58019SAttilio Rao 	size_t len;
17615fe58019SAttilio Rao 
17625fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
17635fe58019SAttilio Rao 		return ENXIO;
17645fe58019SAttilio Rao 	}
17655fe58019SAttilio Rao 	/*
17665fe58019SAttilio Rao 	 * Unlike the other creator type calls, here we have to create a message
17675fe58019SAttilio Rao 	 * where the name of the new entry comes first, and the data describing
17685fe58019SAttilio Rao 	 * the entry comes second.
17695fe58019SAttilio Rao 	 * Hence we can't rely on our handy fuse_internal_newentry() routine,
17705fe58019SAttilio Rao 	 * but put together the message manually and just call the core part.
17715fe58019SAttilio Rao 	 */
17725fe58019SAttilio Rao 
17735fe58019SAttilio Rao 	len = strlen(target) + 1;
17745fe58019SAttilio Rao 	fdisp_init(&fdi, len + cnp->cn_namelen + 1);
17755fe58019SAttilio Rao 	fdisp_make_vp(&fdi, FUSE_SYMLINK, dvp, curthread, NULL);
17765fe58019SAttilio Rao 
17775fe58019SAttilio Rao 	memcpy(fdi.indata, cnp->cn_nameptr, cnp->cn_namelen);
17785fe58019SAttilio Rao 	((char *)fdi.indata)[cnp->cn_namelen] = '\0';
17795fe58019SAttilio Rao 	memcpy((char *)fdi.indata + cnp->cn_namelen + 1, target, len);
17805fe58019SAttilio Rao 
17815fe58019SAttilio Rao 	err = fuse_internal_newentry_core(dvp, vpp, cnp, VLNK, &fdi);
17825fe58019SAttilio Rao 	fdisp_destroy(&fdi);
17835fe58019SAttilio Rao 	return err;
17845fe58019SAttilio Rao }
17855fe58019SAttilio Rao 
17865fe58019SAttilio Rao /*
17875fe58019SAttilio Rao     struct vnop_write_args {
17885fe58019SAttilio Rao 	struct vnode *a_vp;
17895fe58019SAttilio Rao 	struct uio *a_uio;
17905fe58019SAttilio Rao 	int  a_ioflag;
17915fe58019SAttilio Rao 	struct ucred *a_cred;
17925fe58019SAttilio Rao     };
17935fe58019SAttilio Rao */
17945fe58019SAttilio Rao static int
17955fe58019SAttilio Rao fuse_vnop_write(struct vop_write_args *ap)
17965fe58019SAttilio Rao {
17975fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
17985fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
17995fe58019SAttilio Rao 	int ioflag = ap->a_ioflag;
18005fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
1801f8d4af10SAlan Somers 	pid_t pid = curthread->td_proc->p_pid;
1802f203d173SAlan Somers 	int err;
18035fe58019SAttilio Rao 
18045fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
18055fe58019SAttilio Rao 		return ENXIO;
18065fe58019SAttilio Rao 	}
1807f203d173SAlan Somers 	err = fuse_vnode_refreshsize(vp, cred);
1808f203d173SAlan Somers 	if (err)
1809f203d173SAlan Somers 		return err;
18105fe58019SAttilio Rao 
1811ead063e0SEdward Tomasz Napierala 	if (VTOFUD(vp)->flag & FN_DIRECTIO) {
1812ead063e0SEdward Tomasz Napierala 		ioflag |= IO_DIRECT;
1813ead063e0SEdward Tomasz Napierala 	}
1814ead063e0SEdward Tomasz Napierala 
1815f8d4af10SAlan Somers 	return fuse_io_dispatch(vp, uio, ioflag, cred, pid);
18165fe58019SAttilio Rao }
18175fe58019SAttilio Rao 
1818cf169498SAlan Somers SDT_PROBE_DEFINE1(fuse, , vnops, vnop_getpages_error, "int");
18195fe58019SAttilio Rao /*
18205fe58019SAttilio Rao     struct vnop_getpages_args {
18215fe58019SAttilio Rao 	struct vnode *a_vp;
18225fe58019SAttilio Rao 	vm_page_t *a_m;
18235fe58019SAttilio Rao 	int a_count;
18245fe58019SAttilio Rao 	int a_reqpage;
18255fe58019SAttilio Rao     };
18265fe58019SAttilio Rao */
18275fe58019SAttilio Rao static int
18285fe58019SAttilio Rao fuse_vnop_getpages(struct vop_getpages_args *ap)
18295fe58019SAttilio Rao {
18305fe58019SAttilio Rao 	int i, error, nextoff, size, toff, count, npages;
18315fe58019SAttilio Rao 	struct uio uio;
18325fe58019SAttilio Rao 	struct iovec iov;
18335fe58019SAttilio Rao 	vm_offset_t kva;
18345fe58019SAttilio Rao 	struct buf *bp;
18355fe58019SAttilio Rao 	struct vnode *vp;
18365fe58019SAttilio Rao 	struct thread *td;
18375fe58019SAttilio Rao 	struct ucred *cred;
18385fe58019SAttilio Rao 	vm_page_t *pages;
1839f8d4af10SAlan Somers 	pid_t pid = curthread->td_proc->p_pid;
18405fe58019SAttilio Rao 
18415fe58019SAttilio Rao 	vp = ap->a_vp;
18425fe58019SAttilio Rao 	KASSERT(vp->v_object, ("objectless vp passed to getpages"));
18435fe58019SAttilio Rao 	td = curthread;			/* XXX */
18445fe58019SAttilio Rao 	cred = curthread->td_ucred;	/* XXX */
18455fe58019SAttilio Rao 	pages = ap->a_m;
1846f17f88d3SGleb Smirnoff 	npages = ap->a_count;
18475fe58019SAttilio Rao 
18485fe58019SAttilio Rao 	if (!fsess_opt_mmap(vnode_mount(vp))) {
1849cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1,
1850cf169498SAlan Somers 			"called on non-cacheable vnode??\n");
18515fe58019SAttilio Rao 		return (VM_PAGER_ERROR);
18525fe58019SAttilio Rao 	}
18535fe58019SAttilio Rao 
18545fe58019SAttilio Rao 	/*
1855b0cd2017SGleb Smirnoff 	 * If the last page is partially valid, just return it and allow
1856b0cd2017SGleb Smirnoff 	 * the pager to zero-out the blanks.  Partially valid pages can
1857b0cd2017SGleb Smirnoff 	 * only occur at the file EOF.
1858b0cd2017SGleb Smirnoff 	 *
1859b0cd2017SGleb Smirnoff 	 * XXXGL: is that true for FUSE, which is a local filesystem,
1860b0cd2017SGleb Smirnoff 	 * but still somewhat disconnected from the kernel?
18615fe58019SAttilio Rao 	 */
186289f6b863SAttilio Rao 	VM_OBJECT_WLOCK(vp->v_object);
1863f17f88d3SGleb Smirnoff 	if (pages[npages - 1]->valid != 0 && --npages == 0)
1864f17f88d3SGleb Smirnoff 		goto out;
186589f6b863SAttilio Rao 	VM_OBJECT_WUNLOCK(vp->v_object);
18665fe58019SAttilio Rao 
18675fe58019SAttilio Rao 	/*
18685fe58019SAttilio Rao 	 * We use only the kva address for the buffer, but this is extremely
1869b3a15dddSPedro F. Giffuni 	 * convenient and fast.
18705fe58019SAttilio Rao 	 */
1871756a5412SGleb Smirnoff 	bp = uma_zalloc(fuse_pbuf_zone, M_WAITOK);
18725fe58019SAttilio Rao 
18735fe58019SAttilio Rao 	kva = (vm_offset_t)bp->b_data;
18745fe58019SAttilio Rao 	pmap_qenter(kva, pages, npages);
187583c9dea1SGleb Smirnoff 	VM_CNT_INC(v_vnodein);
187683c9dea1SGleb Smirnoff 	VM_CNT_ADD(v_vnodepgsin, npages);
18775fe58019SAttilio Rao 
1878f17f88d3SGleb Smirnoff 	count = npages << PAGE_SHIFT;
18795fe58019SAttilio Rao 	iov.iov_base = (caddr_t)kva;
18805fe58019SAttilio Rao 	iov.iov_len = count;
18815fe58019SAttilio Rao 	uio.uio_iov = &iov;
18825fe58019SAttilio Rao 	uio.uio_iovcnt = 1;
18835fe58019SAttilio Rao 	uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
18845fe58019SAttilio Rao 	uio.uio_resid = count;
18855fe58019SAttilio Rao 	uio.uio_segflg = UIO_SYSSPACE;
18865fe58019SAttilio Rao 	uio.uio_rw = UIO_READ;
18875fe58019SAttilio Rao 	uio.uio_td = td;
18885fe58019SAttilio Rao 
1889f8d4af10SAlan Somers 	error = fuse_io_dispatch(vp, &uio, IO_DIRECT, cred, pid);
18905fe58019SAttilio Rao 	pmap_qremove(kva, npages);
18915fe58019SAttilio Rao 
1892756a5412SGleb Smirnoff 	uma_zfree(fuse_pbuf_zone, bp);
18935fe58019SAttilio Rao 
18945fe58019SAttilio Rao 	if (error && (uio.uio_resid == count)) {
1895cf169498SAlan Somers 		SDT_PROBE1(fuse, , vnops, vnop_getpages_error, error);
18965fe58019SAttilio Rao 		return VM_PAGER_ERROR;
18975fe58019SAttilio Rao 	}
18985fe58019SAttilio Rao 	/*
18995fe58019SAttilio Rao 	 * Calculate the number of bytes read and validate only that number
19005fe58019SAttilio Rao 	 * of bytes.  Note that due to pending writes, size may be 0.  This
19015fe58019SAttilio Rao 	 * does not mean that the remaining data is invalid!
19025fe58019SAttilio Rao 	 */
19035fe58019SAttilio Rao 
19045fe58019SAttilio Rao 	size = count - uio.uio_resid;
190589f6b863SAttilio Rao 	VM_OBJECT_WLOCK(vp->v_object);
19065fe58019SAttilio Rao 	fuse_vm_page_lock_queues();
19075fe58019SAttilio Rao 	for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
19085fe58019SAttilio Rao 		vm_page_t m;
19095fe58019SAttilio Rao 
19105fe58019SAttilio Rao 		nextoff = toff + PAGE_SIZE;
19115fe58019SAttilio Rao 		m = pages[i];
19125fe58019SAttilio Rao 
19135fe58019SAttilio Rao 		if (nextoff <= size) {
19145fe58019SAttilio Rao 			/*
19155fe58019SAttilio Rao 			 * Read operation filled an entire page
19165fe58019SAttilio Rao 			 */
19175fe58019SAttilio Rao 			m->valid = VM_PAGE_BITS_ALL;
19185fe58019SAttilio Rao 			KASSERT(m->dirty == 0,
19195fe58019SAttilio Rao 			    ("fuse_getpages: page %p is dirty", m));
19205fe58019SAttilio Rao 		} else if (size > toff) {
19215fe58019SAttilio Rao 			/*
19225fe58019SAttilio Rao 			 * Read operation filled a partial page.
19235fe58019SAttilio Rao 			 */
19245fe58019SAttilio Rao 			m->valid = 0;
19255fe58019SAttilio Rao 			vm_page_set_valid_range(m, 0, size - toff);
19265fe58019SAttilio Rao 			KASSERT(m->dirty == 0,
19275fe58019SAttilio Rao 			    ("fuse_getpages: page %p is dirty", m));
19285fe58019SAttilio Rao 		} else {
19295fe58019SAttilio Rao 			/*
1930b3a15dddSPedro F. Giffuni 			 * Read operation was short.  If no error occurred
19315fe58019SAttilio Rao 			 * we may have hit a zero-fill section.   We simply
19325fe58019SAttilio Rao 			 * leave valid set to 0.
19335fe58019SAttilio Rao 			 */
19345fe58019SAttilio Rao 			;
19355fe58019SAttilio Rao 		}
19365fe58019SAttilio Rao 	}
19375fe58019SAttilio Rao 	fuse_vm_page_unlock_queues();
1938f17f88d3SGleb Smirnoff out:
193989f6b863SAttilio Rao 	VM_OBJECT_WUNLOCK(vp->v_object);
1940f17f88d3SGleb Smirnoff 	if (ap->a_rbehind)
1941f17f88d3SGleb Smirnoff 		*ap->a_rbehind = 0;
1942f17f88d3SGleb Smirnoff 	if (ap->a_rahead)
1943f17f88d3SGleb Smirnoff 		*ap->a_rahead = 0;
1944f17f88d3SGleb Smirnoff 	return (VM_PAGER_OK);
19455fe58019SAttilio Rao }
19465fe58019SAttilio Rao 
19475fe58019SAttilio Rao /*
19485fe58019SAttilio Rao     struct vnop_putpages_args {
19495fe58019SAttilio Rao 	struct vnode *a_vp;
19505fe58019SAttilio Rao 	vm_page_t *a_m;
19515fe58019SAttilio Rao 	int a_count;
19525fe58019SAttilio Rao 	int a_sync;
19535fe58019SAttilio Rao 	int *a_rtvals;
19545fe58019SAttilio Rao 	vm_ooffset_t a_offset;
19555fe58019SAttilio Rao     };
19565fe58019SAttilio Rao */
19575fe58019SAttilio Rao static int
19585fe58019SAttilio Rao fuse_vnop_putpages(struct vop_putpages_args *ap)
19595fe58019SAttilio Rao {
19605fe58019SAttilio Rao 	struct uio uio;
19615fe58019SAttilio Rao 	struct iovec iov;
19625fe58019SAttilio Rao 	vm_offset_t kva;
19635fe58019SAttilio Rao 	struct buf *bp;
19645fe58019SAttilio Rao 	int i, error, npages, count;
19655fe58019SAttilio Rao 	off_t offset;
19665fe58019SAttilio Rao 	int *rtvals;
19675fe58019SAttilio Rao 	struct vnode *vp;
19685fe58019SAttilio Rao 	struct thread *td;
19695fe58019SAttilio Rao 	struct ucred *cred;
19705fe58019SAttilio Rao 	vm_page_t *pages;
19715fe58019SAttilio Rao 	vm_ooffset_t fsize;
1972f8d4af10SAlan Somers 	pid_t pid = curthread->td_proc->p_pid;
19735fe58019SAttilio Rao 
19745fe58019SAttilio Rao 	vp = ap->a_vp;
19755fe58019SAttilio Rao 	KASSERT(vp->v_object, ("objectless vp passed to putpages"));
19765fe58019SAttilio Rao 	fsize = vp->v_object->un_pager.vnp.vnp_size;
19775fe58019SAttilio Rao 	td = curthread;			/* XXX */
19785fe58019SAttilio Rao 	cred = curthread->td_ucred;	/* XXX */
19795fe58019SAttilio Rao 	pages = ap->a_m;
19805fe58019SAttilio Rao 	count = ap->a_count;
19815fe58019SAttilio Rao 	rtvals = ap->a_rtvals;
19825fe58019SAttilio Rao 	npages = btoc(count);
19835fe58019SAttilio Rao 	offset = IDX_TO_OFF(pages[0]->pindex);
19845fe58019SAttilio Rao 
19855fe58019SAttilio Rao 	if (!fsess_opt_mmap(vnode_mount(vp))) {
1986cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1,
1987cf169498SAlan Somers 			"called on non-cacheable vnode??\n");
19885fe58019SAttilio Rao 	}
19895fe58019SAttilio Rao 	for (i = 0; i < npages; i++)
19905fe58019SAttilio Rao 		rtvals[i] = VM_PAGER_AGAIN;
19915fe58019SAttilio Rao 
19925fe58019SAttilio Rao 	/*
19935fe58019SAttilio Rao 	 * When putting pages, do not extend file past EOF.
19945fe58019SAttilio Rao 	 */
19955fe58019SAttilio Rao 
19965fe58019SAttilio Rao 	if (offset + count > fsize) {
19975fe58019SAttilio Rao 		count = fsize - offset;
19985fe58019SAttilio Rao 		if (count < 0)
19995fe58019SAttilio Rao 			count = 0;
20005fe58019SAttilio Rao 	}
20015fe58019SAttilio Rao 	/*
20025fe58019SAttilio Rao 	 * We use only the kva address for the buffer, but this is extremely
2003b3a15dddSPedro F. Giffuni 	 * convenient and fast.
20045fe58019SAttilio Rao 	 */
2005756a5412SGleb Smirnoff 	bp = uma_zalloc(fuse_pbuf_zone, M_WAITOK);
20065fe58019SAttilio Rao 
20075fe58019SAttilio Rao 	kva = (vm_offset_t)bp->b_data;
20085fe58019SAttilio Rao 	pmap_qenter(kva, pages, npages);
200983c9dea1SGleb Smirnoff 	VM_CNT_INC(v_vnodeout);
201083c9dea1SGleb Smirnoff 	VM_CNT_ADD(v_vnodepgsout, count);
20115fe58019SAttilio Rao 
20125fe58019SAttilio Rao 	iov.iov_base = (caddr_t)kva;
20135fe58019SAttilio Rao 	iov.iov_len = count;
20145fe58019SAttilio Rao 	uio.uio_iov = &iov;
20155fe58019SAttilio Rao 	uio.uio_iovcnt = 1;
20165fe58019SAttilio Rao 	uio.uio_offset = offset;
20175fe58019SAttilio Rao 	uio.uio_resid = count;
20185fe58019SAttilio Rao 	uio.uio_segflg = UIO_SYSSPACE;
20195fe58019SAttilio Rao 	uio.uio_rw = UIO_WRITE;
20205fe58019SAttilio Rao 	uio.uio_td = td;
20215fe58019SAttilio Rao 
2022f8d4af10SAlan Somers 	error = fuse_io_dispatch(vp, &uio, IO_DIRECT, cred, pid);
20235fe58019SAttilio Rao 
20245fe58019SAttilio Rao 	pmap_qremove(kva, npages);
2025756a5412SGleb Smirnoff 	uma_zfree(fuse_pbuf_zone, bp);
20265fe58019SAttilio Rao 
20275fe58019SAttilio Rao 	if (!error) {
20285fe58019SAttilio Rao 		int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
20295fe58019SAttilio Rao 
20305fe58019SAttilio Rao 		for (i = 0; i < nwritten; i++) {
20315fe58019SAttilio Rao 			rtvals[i] = VM_PAGER_OK;
203289f6b863SAttilio Rao 			VM_OBJECT_WLOCK(pages[i]->object);
20335fe58019SAttilio Rao 			vm_page_undirty(pages[i]);
203489f6b863SAttilio Rao 			VM_OBJECT_WUNLOCK(pages[i]->object);
20355fe58019SAttilio Rao 		}
20365fe58019SAttilio Rao 	}
20375fe58019SAttilio Rao 	return rtvals[0];
20385fe58019SAttilio Rao }
20395fe58019SAttilio Rao 
204004660064SFedor Uporov static const char extattr_namespace_separator = '.';
204104660064SFedor Uporov 
204204660064SFedor Uporov /*
204304660064SFedor Uporov     struct vop_getextattr_args {
204404660064SFedor Uporov 	struct vop_generic_args a_gen;
204504660064SFedor Uporov 	struct vnode *a_vp;
204604660064SFedor Uporov 	int a_attrnamespace;
204704660064SFedor Uporov 	const char *a_name;
204804660064SFedor Uporov 	struct uio *a_uio;
204904660064SFedor Uporov 	size_t *a_size;
205004660064SFedor Uporov 	struct ucred *a_cred;
205104660064SFedor Uporov 	struct thread *a_td;
205204660064SFedor Uporov     };
205304660064SFedor Uporov */
205404660064SFedor Uporov static int
205504660064SFedor Uporov fuse_vnop_getextattr(struct vop_getextattr_args *ap)
205604660064SFedor Uporov {
205704660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
205804660064SFedor Uporov 	struct uio *uio = ap->a_uio;
205928f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
206004660064SFedor Uporov 	struct fuse_getxattr_in *get_xattr_in;
206104660064SFedor Uporov 	struct fuse_getxattr_out *get_xattr_out;
206204660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
206304660064SFedor Uporov 	struct thread *td = ap->a_td;
206404660064SFedor Uporov 	struct ucred *cred = ap->a_cred;
206528f4f623SFedor Uporov 	char *prefix;
206628f4f623SFedor Uporov 	char *attr_str;
206728f4f623SFedor Uporov 	size_t len;
206828f4f623SFedor Uporov 	int err;
206904660064SFedor Uporov 
207004660064SFedor Uporov 	if (fuse_isdeadfs(vp))
207128f4f623SFedor Uporov 		return (ENXIO);
207204660064SFedor Uporov 
207304660064SFedor Uporov 	/* Default to looking for user attributes. */
207404660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
207504660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
207604660064SFedor Uporov 	else
207704660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
207804660064SFedor Uporov 
207904660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) +
208004660064SFedor Uporov 	    strlen(ap->a_name) + 1;
208104660064SFedor Uporov 
208204660064SFedor Uporov 	fdisp_init(&fdi, len + sizeof(*get_xattr_in));
208304660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_GETXATTR, vp, td, cred);
208404660064SFedor Uporov 
208504660064SFedor Uporov 	get_xattr_in = fdi.indata;
208604660064SFedor Uporov 	/*
208704660064SFedor Uporov 	 * Check to see whether we're querying the available size or
208804660064SFedor Uporov 	 * issuing the actual request.  If we pass in 0, we get back struct
208904660064SFedor Uporov 	 * fuse_getxattr_out.  If we pass in a non-zero size, we get back
209004660064SFedor Uporov 	 * that much data, without the struct fuse_getxattr_out header.
209104660064SFedor Uporov 	 */
2092493b4a8cSFedor Uporov 	if (uio == NULL)
209304660064SFedor Uporov 		get_xattr_in->size = 0;
209404660064SFedor Uporov 	else
209504660064SFedor Uporov 		get_xattr_in->size = uio->uio_resid;
209604660064SFedor Uporov 
209704660064SFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*get_xattr_in);
209804660064SFedor Uporov 	snprintf(attr_str, len, "%s%c%s", prefix, extattr_namespace_separator,
209904660064SFedor Uporov 	    ap->a_name);
210004660064SFedor Uporov 
210104660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
210204660064SFedor Uporov 	if (err != 0) {
210304660064SFedor Uporov 		if (err == ENOSYS)
210404660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_GETXATTR);
210504660064SFedor Uporov 		goto out;
210604660064SFedor Uporov 	}
210704660064SFedor Uporov 
210804660064SFedor Uporov 	get_xattr_out = fdi.answ;
210904660064SFedor Uporov 
2110493b4a8cSFedor Uporov 	if (ap->a_size != NULL)
211104660064SFedor Uporov 		*ap->a_size = get_xattr_out->size;
2112493b4a8cSFedor Uporov 
2113493b4a8cSFedor Uporov 	if (uio != NULL)
211404660064SFedor Uporov 		err = uiomove(fdi.answ, fdi.iosize, uio);
211504660064SFedor Uporov 
211604660064SFedor Uporov out:
211704660064SFedor Uporov 	fdisp_destroy(&fdi);
211804660064SFedor Uporov 	return (err);
211904660064SFedor Uporov }
212004660064SFedor Uporov 
212104660064SFedor Uporov /*
212204660064SFedor Uporov     struct vop_setextattr_args {
212304660064SFedor Uporov 	struct vop_generic_args a_gen;
212404660064SFedor Uporov 	struct vnode *a_vp;
212504660064SFedor Uporov 	int a_attrnamespace;
212604660064SFedor Uporov 	const char *a_name;
212704660064SFedor Uporov 	struct uio *a_uio;
212804660064SFedor Uporov 	struct ucred *a_cred;
212904660064SFedor Uporov 	struct thread *a_td;
213004660064SFedor Uporov     };
213104660064SFedor Uporov */
213204660064SFedor Uporov static int
213304660064SFedor Uporov fuse_vnop_setextattr(struct vop_setextattr_args *ap)
213404660064SFedor Uporov {
213504660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
213604660064SFedor Uporov 	struct uio *uio = ap->a_uio;
213728f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
213804660064SFedor Uporov 	struct fuse_setxattr_in *set_xattr_in;
213904660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
214028f4f623SFedor Uporov 	struct thread *td = ap->a_td;
214128f4f623SFedor Uporov 	struct ucred *cred = ap->a_cred;
214204660064SFedor Uporov 	char *prefix;
214304660064SFedor Uporov 	size_t len;
214404660064SFedor Uporov 	char *attr_str;
214528f4f623SFedor Uporov 	int err;
214604660064SFedor Uporov 
214704660064SFedor Uporov 	if (fuse_isdeadfs(vp))
214828f4f623SFedor Uporov 		return (ENXIO);
214904660064SFedor Uporov 
215004660064SFedor Uporov 	/* Default to looking for user attributes. */
215104660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
215204660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
215304660064SFedor Uporov 	else
215404660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
215504660064SFedor Uporov 
215604660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) +
215704660064SFedor Uporov 	    strlen(ap->a_name) + 1;
215804660064SFedor Uporov 
215904660064SFedor Uporov 	fdisp_init(&fdi, len + sizeof(*set_xattr_in) + uio->uio_resid);
216004660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_SETXATTR, vp, td, cred);
216104660064SFedor Uporov 
216204660064SFedor Uporov 	set_xattr_in = fdi.indata;
216304660064SFedor Uporov 	set_xattr_in->size = uio->uio_resid;
216404660064SFedor Uporov 
216504660064SFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*set_xattr_in);
216604660064SFedor Uporov 	snprintf(attr_str, len, "%s%c%s", prefix, extattr_namespace_separator,
216704660064SFedor Uporov 	    ap->a_name);
216804660064SFedor Uporov 
216904660064SFedor Uporov 	err = uiomove((char *)fdi.indata + sizeof(*set_xattr_in) + len,
217004660064SFedor Uporov 	    uio->uio_resid, uio);
217104660064SFedor Uporov 	if (err != 0) {
217204660064SFedor Uporov 		goto out;
217304660064SFedor Uporov 	}
217404660064SFedor Uporov 
217504660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
217604660064SFedor Uporov 
217704660064SFedor Uporov 	if (err != 0) {
217804660064SFedor Uporov 		if (err == ENOSYS)
217904660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_SETXATTR);
218004660064SFedor Uporov 		goto out;
218104660064SFedor Uporov 	}
218204660064SFedor Uporov 
218304660064SFedor Uporov out:
218404660064SFedor Uporov 	fdisp_destroy(&fdi);
218504660064SFedor Uporov 	return (err);
218604660064SFedor Uporov }
218704660064SFedor Uporov 
218804660064SFedor Uporov /*
218904660064SFedor Uporov  * The Linux / FUSE extended attribute list is simply a collection of
219004660064SFedor Uporov  * NUL-terminated strings.  The FreeBSD extended attribute list is a single
219104660064SFedor Uporov  * byte length followed by a non-NUL terminated string.  So, this allows
219204660064SFedor Uporov  * conversion of the Linux / FUSE format to the FreeBSD format in place.
219304660064SFedor Uporov  * Linux attribute names are reported with the namespace as a prefix (e.g.
219404660064SFedor Uporov  * "user.attribute_name"), but in FreeBSD they are reported without the
219504660064SFedor Uporov  * namespace prefix (e.g. "attribute_name").  So, we're going from:
219604660064SFedor Uporov  *
219704660064SFedor Uporov  * user.attr_name1\0user.attr_name2\0
219804660064SFedor Uporov  *
219904660064SFedor Uporov  * to:
220004660064SFedor Uporov  *
220104660064SFedor Uporov  * <num>attr_name1<num>attr_name2
220204660064SFedor Uporov  *
220304660064SFedor Uporov  * Where "<num>" is a single byte number of characters in the attribute name.
220404660064SFedor Uporov  *
220504660064SFedor Uporov  * Args:
220604660064SFedor Uporov  * prefix - exattr namespace prefix string
220704660064SFedor Uporov  * list, list_len - input list with namespace prefixes
220804660064SFedor Uporov  * bsd_list, bsd_list_len - output list compatible with bsd vfs
220904660064SFedor Uporov  */
221004660064SFedor Uporov static int
221104660064SFedor Uporov fuse_xattrlist_convert(char *prefix, const char *list, int list_len,
221204660064SFedor Uporov     char *bsd_list, int *bsd_list_len)
221304660064SFedor Uporov {
221404660064SFedor Uporov 	int len, pos, dist_to_next, prefix_len;
221504660064SFedor Uporov 
221604660064SFedor Uporov 	pos = 0;
221704660064SFedor Uporov 	*bsd_list_len = 0;
221804660064SFedor Uporov 	prefix_len = strlen(prefix);
221904660064SFedor Uporov 
222004660064SFedor Uporov 	while (pos < list_len && list[pos] != '\0') {
222104660064SFedor Uporov 		dist_to_next = strlen(&list[pos]) + 1;
222204660064SFedor Uporov 		if (bcmp(&list[pos], prefix, prefix_len) == 0 &&
222304660064SFedor Uporov 		    list[pos + prefix_len] == extattr_namespace_separator) {
222404660064SFedor Uporov 			len = dist_to_next -
222504660064SFedor Uporov 			    (prefix_len + sizeof(extattr_namespace_separator)) - 1;
222604660064SFedor Uporov 			if (len >= EXTATTR_MAXNAMELEN)
222704660064SFedor Uporov 				return (ENAMETOOLONG);
222804660064SFedor Uporov 
222904660064SFedor Uporov 			bsd_list[*bsd_list_len] = len;
223004660064SFedor Uporov 			memcpy(&bsd_list[*bsd_list_len + 1],
223104660064SFedor Uporov 			    &list[pos + prefix_len +
223204660064SFedor Uporov 			    sizeof(extattr_namespace_separator)], len);
223304660064SFedor Uporov 
223404660064SFedor Uporov 			*bsd_list_len += len + 1;
223504660064SFedor Uporov 		}
223604660064SFedor Uporov 
223704660064SFedor Uporov 		pos += dist_to_next;
223804660064SFedor Uporov 	}
223904660064SFedor Uporov 
224004660064SFedor Uporov 	return (0);
224104660064SFedor Uporov }
224204660064SFedor Uporov 
224304660064SFedor Uporov /*
224404660064SFedor Uporov     struct vop_listextattr_args {
224504660064SFedor Uporov 	struct vop_generic_args a_gen;
224604660064SFedor Uporov 	struct vnode *a_vp;
224704660064SFedor Uporov 	int a_attrnamespace;
224804660064SFedor Uporov 	struct uio *a_uio;
224904660064SFedor Uporov 	size_t *a_size;
225004660064SFedor Uporov 	struct ucred *a_cred;
225104660064SFedor Uporov 	struct thread *a_td;
225204660064SFedor Uporov     };
225304660064SFedor Uporov */
225404660064SFedor Uporov static int
225504660064SFedor Uporov fuse_vnop_listextattr(struct vop_listextattr_args *ap)
225604660064SFedor Uporov {
225704660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
225804660064SFedor Uporov 	struct uio *uio = ap->a_uio;
225928f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
2260493b4a8cSFedor Uporov 	struct fuse_listxattr_in *list_xattr_in;
2261493b4a8cSFedor Uporov 	struct fuse_listxattr_out *list_xattr_out;
226204660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
226328f4f623SFedor Uporov 	struct thread *td = ap->a_td;
226428f4f623SFedor Uporov 	struct ucred *cred = ap->a_cred;
226504660064SFedor Uporov 	size_t len;
226604660064SFedor Uporov 	char *prefix;
226704660064SFedor Uporov 	char *attr_str;
226804660064SFedor Uporov 	char *bsd_list = NULL;
2269493b4a8cSFedor Uporov 	char *linux_list;
227004660064SFedor Uporov 	int bsd_list_len;
2271493b4a8cSFedor Uporov 	int linux_list_len;
227228f4f623SFedor Uporov 	int err;
227304660064SFedor Uporov 
227404660064SFedor Uporov 	if (fuse_isdeadfs(vp))
227528f4f623SFedor Uporov 		return (ENXIO);
227604660064SFedor Uporov 
227704660064SFedor Uporov 	/*
227804660064SFedor Uporov 	 * Add space for a NUL and the period separator if enabled.
227904660064SFedor Uporov 	 * Default to looking for user attributes.
228004660064SFedor Uporov 	 */
228104660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
228204660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
228304660064SFedor Uporov 	else
228404660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
228504660064SFedor Uporov 
228604660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) + 1;
228704660064SFedor Uporov 
2288493b4a8cSFedor Uporov 	fdisp_init(&fdi, sizeof(*list_xattr_in) + len);
228904660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_LISTXATTR, vp, td, cred);
229004660064SFedor Uporov 
2291493b4a8cSFedor Uporov 	/*
2292493b4a8cSFedor Uporov 	 * Retrieve Linux / FUSE compatible list size.
2293493b4a8cSFedor Uporov 	 */
2294493b4a8cSFedor Uporov 	list_xattr_in = fdi.indata;
2295493b4a8cSFedor Uporov 	list_xattr_in->size = 0;
2296493b4a8cSFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*list_xattr_in);
229704660064SFedor Uporov 	snprintf(attr_str, len, "%s%c", prefix, extattr_namespace_separator);
229804660064SFedor Uporov 
229904660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
230004660064SFedor Uporov 	if (err != 0) {
230104660064SFedor Uporov 		if (err == ENOSYS)
230204660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_LISTXATTR);
230304660064SFedor Uporov 		goto out;
230404660064SFedor Uporov 	}
230504660064SFedor Uporov 
2306493b4a8cSFedor Uporov 	list_xattr_out = fdi.answ;
2307493b4a8cSFedor Uporov 	linux_list_len = list_xattr_out->size;
2308493b4a8cSFedor Uporov 	if (linux_list_len == 0) {
2309493b4a8cSFedor Uporov 		if (ap->a_size != NULL)
2310493b4a8cSFedor Uporov 			*ap->a_size = linux_list_len;
231104660064SFedor Uporov 		goto out;
231204660064SFedor Uporov 	}
231304660064SFedor Uporov 
231404660064SFedor Uporov 	/*
2315493b4a8cSFedor Uporov 	 * Retrieve Linux / FUSE compatible list values.
2316493b4a8cSFedor Uporov 	 */
231712292a99SAlan Somers 	fdisp_refresh_vp(&fdi, FUSE_LISTXATTR, vp, td, cred);
2318493b4a8cSFedor Uporov 	list_xattr_in = fdi.indata;
2319493b4a8cSFedor Uporov 	list_xattr_in->size = linux_list_len + sizeof(*list_xattr_out);
232012292a99SAlan Somers 	list_xattr_in->flags = 0;
2321493b4a8cSFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*list_xattr_in);
2322493b4a8cSFedor Uporov 	snprintf(attr_str, len, "%s%c", prefix, extattr_namespace_separator);
2323493b4a8cSFedor Uporov 
2324493b4a8cSFedor Uporov 	err = fdisp_wait_answ(&fdi);
2325493b4a8cSFedor Uporov 	if (err != 0)
2326493b4a8cSFedor Uporov 		goto out;
2327493b4a8cSFedor Uporov 
2328493b4a8cSFedor Uporov 	linux_list = fdi.answ;
2329493b4a8cSFedor Uporov 	linux_list_len = fdi.iosize;
2330493b4a8cSFedor Uporov 
2331493b4a8cSFedor Uporov 	/*
2332493b4a8cSFedor Uporov 	 * Retrieve the BSD compatible list values.
233304660064SFedor Uporov 	 * The Linux / FUSE attribute list format isn't the same
233404660064SFedor Uporov 	 * as FreeBSD's format. So we need to transform it into
233504660064SFedor Uporov 	 * FreeBSD's format before giving it to the user.
233604660064SFedor Uporov 	 */
2337493b4a8cSFedor Uporov 	bsd_list = malloc(linux_list_len, M_TEMP, M_WAITOK);
2338493b4a8cSFedor Uporov 	err = fuse_xattrlist_convert(prefix, linux_list, linux_list_len,
233904660064SFedor Uporov 	    bsd_list, &bsd_list_len);
234004660064SFedor Uporov 	if (err != 0)
234104660064SFedor Uporov 		goto out;
234204660064SFedor Uporov 
2343493b4a8cSFedor Uporov 	if (ap->a_size != NULL)
2344493b4a8cSFedor Uporov 		*ap->a_size = bsd_list_len;
2345493b4a8cSFedor Uporov 
2346493b4a8cSFedor Uporov 	if (uio != NULL)
234704660064SFedor Uporov 		err = uiomove(bsd_list, bsd_list_len, uio);
234804660064SFedor Uporov 
234904660064SFedor Uporov out:
235004660064SFedor Uporov 	free(bsd_list, M_TEMP);
235104660064SFedor Uporov 	fdisp_destroy(&fdi);
235204660064SFedor Uporov 	return (err);
235304660064SFedor Uporov }
235404660064SFedor Uporov 
235504660064SFedor Uporov /*
235604660064SFedor Uporov     struct vop_deleteextattr_args {
235704660064SFedor Uporov 	struct vop_generic_args a_gen;
235804660064SFedor Uporov 	struct vnode *a_vp;
235904660064SFedor Uporov 	int a_attrnamespace;
236004660064SFedor Uporov 	const char *a_name;
236104660064SFedor Uporov 	struct ucred *a_cred;
236204660064SFedor Uporov 	struct thread *a_td;
236304660064SFedor Uporov     };
236404660064SFedor Uporov */
236504660064SFedor Uporov static int
236604660064SFedor Uporov fuse_vnop_deleteextattr(struct vop_deleteextattr_args *ap)
236704660064SFedor Uporov {
236804660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
236928f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
237004660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
237128f4f623SFedor Uporov 	struct thread *td = ap->a_td;
237228f4f623SFedor Uporov 	struct ucred *cred = ap->a_cred;
237304660064SFedor Uporov 	char *prefix;
237404660064SFedor Uporov 	size_t len;
237504660064SFedor Uporov 	char *attr_str;
237604660064SFedor Uporov 	int err;
237704660064SFedor Uporov 
237804660064SFedor Uporov 	if (fuse_isdeadfs(vp))
237928f4f623SFedor Uporov 		return (ENXIO);
238004660064SFedor Uporov 
238104660064SFedor Uporov 	/* Default to looking for user attributes. */
238204660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
238304660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
238404660064SFedor Uporov 	else
238504660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
238604660064SFedor Uporov 
238704660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) +
238804660064SFedor Uporov 	    strlen(ap->a_name) + 1;
238904660064SFedor Uporov 
239004660064SFedor Uporov 	fdisp_init(&fdi, len);
239104660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_REMOVEXATTR, vp, td, cred);
239204660064SFedor Uporov 
239304660064SFedor Uporov 	attr_str = fdi.indata;
239404660064SFedor Uporov 	snprintf(attr_str, len, "%s%c%s", prefix, extattr_namespace_separator,
239504660064SFedor Uporov 	    ap->a_name);
239604660064SFedor Uporov 
239704660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
239804660064SFedor Uporov 	if (err != 0) {
239904660064SFedor Uporov 		if (err == ENOSYS)
240004660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_REMOVEXATTR);
240104660064SFedor Uporov 	}
240204660064SFedor Uporov 
240304660064SFedor Uporov 	fdisp_destroy(&fdi);
240404660064SFedor Uporov 	return (err);
240504660064SFedor Uporov }
240604660064SFedor Uporov 
24075fe58019SAttilio Rao /*
24085fe58019SAttilio Rao     struct vnop_print_args {
24095fe58019SAttilio Rao 	struct vnode *a_vp;
24105fe58019SAttilio Rao     };
24115fe58019SAttilio Rao */
24125fe58019SAttilio Rao static int
24135fe58019SAttilio Rao fuse_vnop_print(struct vop_print_args *ap)
24145fe58019SAttilio Rao {
24155fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(ap->a_vp);
24165fe58019SAttilio Rao 
24175fe58019SAttilio Rao 	printf("nodeid: %ju, parent nodeid: %ju, nlookup: %ju, flag: %#x\n",
24185fe58019SAttilio Rao 	    (uintmax_t)VTOILLU(ap->a_vp), (uintmax_t)fvdat->parent_nid,
24195fe58019SAttilio Rao 	    (uintmax_t)fvdat->nlookup,
24205fe58019SAttilio Rao 	    fvdat->flag);
24215fe58019SAttilio Rao 
24225fe58019SAttilio Rao 	return 0;
24235fe58019SAttilio Rao }
2424