xref: /freebsd/sys/fs/fuse/fuse_vnops.c (revision 5fccbf313a04e7f533dbd7313cb9ac51d37f175d)
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 
1985fe58019SAttilio Rao int	fuse_lookup_cache_enable = 1;
1995fe58019SAttilio Rao 
200123af6ecSAlan Somers SYSCTL_INT(_vfs_fusefs, OID_AUTO, lookup_cache_enable, CTLFLAG_RW,
201bd4cb2a4SConrad Meyer     &fuse_lookup_cache_enable, 0, "if non-zero, enable lookup cache");
2025fe58019SAttilio Rao 
2035fe58019SAttilio Rao /*
2045fe58019SAttilio Rao  * XXX: This feature is highly experimental and can bring to instabilities,
2055fe58019SAttilio Rao  * needs revisiting before to be enabled by default.
2065fe58019SAttilio Rao  */
2075fe58019SAttilio Rao static int fuse_reclaim_revoked = 0;
2085fe58019SAttilio Rao 
209123af6ecSAlan Somers SYSCTL_INT(_vfs_fusefs, OID_AUTO, reclaim_revoked, CTLFLAG_RW,
2105fe58019SAttilio Rao     &fuse_reclaim_revoked, 0, "");
2115fe58019SAttilio Rao 
212756a5412SGleb Smirnoff uma_zone_t fuse_pbuf_zone;
2135fe58019SAttilio Rao 
2145fe58019SAttilio Rao #define fuse_vm_page_lock(m)		vm_page_lock((m));
2155fe58019SAttilio Rao #define fuse_vm_page_unlock(m)		vm_page_unlock((m));
2165fe58019SAttilio Rao #define fuse_vm_page_lock_queues()	((void)0)
2175fe58019SAttilio Rao #define fuse_vm_page_unlock_queues()	((void)0)
2185fe58019SAttilio Rao 
2195fe58019SAttilio Rao /*
2205fe58019SAttilio Rao     struct vnop_access_args {
2215fe58019SAttilio Rao 	struct vnode *a_vp;
2225fe58019SAttilio Rao #if VOP_ACCESS_TAKES_ACCMODE_T
2235fe58019SAttilio Rao 	accmode_t a_accmode;
2245fe58019SAttilio Rao #else
2255fe58019SAttilio Rao 	int a_mode;
2265fe58019SAttilio Rao #endif
2275fe58019SAttilio Rao 	struct ucred *a_cred;
2285fe58019SAttilio Rao 	struct thread *a_td;
2295fe58019SAttilio Rao     };
2305fe58019SAttilio Rao */
2315fe58019SAttilio Rao static int
2325fe58019SAttilio Rao fuse_vnop_access(struct vop_access_args *ap)
2335fe58019SAttilio Rao {
2345fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
2355fe58019SAttilio Rao 	int accmode = ap->a_accmode;
2365fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
2375fe58019SAttilio Rao 
2385fe58019SAttilio Rao 	struct fuse_access_param facp;
2395fe58019SAttilio Rao 	struct fuse_data *data = fuse_get_mpdata(vnode_mount(vp));
2405fe58019SAttilio Rao 
2415fe58019SAttilio Rao 	int err;
2425fe58019SAttilio Rao 
2435fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
2445fe58019SAttilio Rao 		if (vnode_isvroot(vp)) {
2455fe58019SAttilio Rao 			return 0;
2465fe58019SAttilio Rao 		}
2475fe58019SAttilio Rao 		return ENXIO;
2485fe58019SAttilio Rao 	}
2495fe58019SAttilio Rao 	if (!(data->dataflags & FSESS_INITED)) {
2505fe58019SAttilio Rao 		if (vnode_isvroot(vp)) {
251cc426dd3SMateusz Guzik 			if (priv_check_cred(cred, PRIV_VFS_ADMIN) ||
2525fe58019SAttilio Rao 			    (fuse_match_cred(data->daemoncred, cred) == 0)) {
2535fe58019SAttilio Rao 				return 0;
2545fe58019SAttilio Rao 			}
2555fe58019SAttilio Rao 		}
2565fe58019SAttilio Rao 		return EBADF;
2575fe58019SAttilio Rao 	}
2585fe58019SAttilio Rao 	if (vnode_islnk(vp)) {
2595fe58019SAttilio Rao 		return 0;
2605fe58019SAttilio Rao 	}
2615fe58019SAttilio Rao 	bzero(&facp, sizeof(facp));
2625fe58019SAttilio Rao 
2635fe58019SAttilio Rao 	err = fuse_internal_access(vp, accmode, &facp, ap->a_td, ap->a_cred);
2645fe58019SAttilio Rao 	return err;
2655fe58019SAttilio Rao }
2665fe58019SAttilio Rao 
2675fe58019SAttilio Rao /*
2685fe58019SAttilio Rao     struct vnop_close_args {
2695fe58019SAttilio Rao 	struct vnode *a_vp;
2705fe58019SAttilio Rao 	int  a_fflag;
2715fe58019SAttilio Rao 	struct ucred *a_cred;
2725fe58019SAttilio Rao 	struct thread *a_td;
2735fe58019SAttilio Rao     };
2745fe58019SAttilio Rao */
2755fe58019SAttilio Rao static int
2765fe58019SAttilio Rao fuse_vnop_close(struct vop_close_args *ap)
2775fe58019SAttilio Rao {
2785fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
2795fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
2805fe58019SAttilio Rao 	int fflag = ap->a_fflag;
2815fe58019SAttilio Rao 	fufh_type_t fufh_type;
2825fe58019SAttilio Rao 
2835fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
2845fe58019SAttilio Rao 		return 0;
2855fe58019SAttilio Rao 	}
2865fe58019SAttilio Rao 	if (vnode_isdir(vp)) {
2875fe58019SAttilio Rao 		if (fuse_filehandle_valid(vp, FUFH_RDONLY)) {
2885fe58019SAttilio Rao 			fuse_filehandle_close(vp, FUFH_RDONLY, NULL, cred);
2895fe58019SAttilio Rao 		}
2905fe58019SAttilio Rao 		return 0;
2915fe58019SAttilio Rao 	}
2925fe58019SAttilio Rao 	if (fflag & IO_NDELAY) {
2935fe58019SAttilio Rao 		return 0;
2945fe58019SAttilio Rao 	}
2955fe58019SAttilio Rao 	fufh_type = fuse_filehandle_xlate_from_fflags(fflag);
2965fe58019SAttilio Rao 
2975fe58019SAttilio Rao 	if (!fuse_filehandle_valid(vp, fufh_type)) {
2985fe58019SAttilio Rao 		int i;
2995fe58019SAttilio Rao 
3005fe58019SAttilio Rao 		for (i = 0; i < FUFH_MAXTYPE; i++)
3015fe58019SAttilio Rao 			if (fuse_filehandle_valid(vp, i))
3025fe58019SAttilio Rao 				break;
3035fe58019SAttilio Rao 		if (i == FUFH_MAXTYPE)
3045fe58019SAttilio Rao 			panic("FUSE: fufh type %d found to be invalid in close"
3055fe58019SAttilio Rao 			      " (fflag=0x%x)\n",
3065fe58019SAttilio Rao 			      fufh_type, fflag);
3075fe58019SAttilio Rao 	}
3085fe58019SAttilio Rao 	if ((VTOFUD(vp)->flag & FN_SIZECHANGE) != 0) {
3095fe58019SAttilio Rao 		fuse_vnode_savesize(vp, cred);
3105fe58019SAttilio Rao 	}
3115fe58019SAttilio Rao 	return 0;
3125fe58019SAttilio Rao }
3135fe58019SAttilio Rao 
31419ef317dSAlan Somers static void
31519ef317dSAlan Somers fdisp_make_mknod_for_fallback(
31619ef317dSAlan Somers 	struct fuse_dispatcher *fdip,
31719ef317dSAlan Somers 	struct componentname *cnp,
31819ef317dSAlan Somers 	struct vnode *dvp,
31919ef317dSAlan Somers 	uint64_t parentnid,
32019ef317dSAlan Somers 	struct thread *td,
32119ef317dSAlan Somers 	struct ucred *cred,
32219ef317dSAlan Somers 	mode_t mode,
32319ef317dSAlan Somers 	enum fuse_opcode *op)
32419ef317dSAlan Somers {
32519ef317dSAlan Somers 	struct fuse_mknod_in *fmni;
32619ef317dSAlan Somers 
32719ef317dSAlan Somers 	fdisp_init(fdip, sizeof(*fmni) + cnp->cn_namelen + 1);
32819ef317dSAlan Somers 	*op = FUSE_MKNOD;
32919ef317dSAlan Somers 	fdisp_make(fdip, *op, vnode_mount(dvp), parentnid, td, cred);
33019ef317dSAlan Somers 	fmni = fdip->indata;
33119ef317dSAlan Somers 	fmni->mode = mode;
33219ef317dSAlan Somers 	fmni->rdev = 0;
33319ef317dSAlan Somers 	memcpy((char *)fdip->indata + sizeof(*fmni), cnp->cn_nameptr,
33419ef317dSAlan Somers 	    cnp->cn_namelen);
33519ef317dSAlan Somers 	((char *)fdip->indata)[sizeof(*fmni) + cnp->cn_namelen] = '\0';
33619ef317dSAlan Somers }
3375fe58019SAttilio Rao /*
3385fe58019SAttilio Rao     struct vnop_create_args {
3395fe58019SAttilio Rao 	struct vnode *a_dvp;
3405fe58019SAttilio Rao 	struct vnode **a_vpp;
3415fe58019SAttilio Rao 	struct componentname *a_cnp;
3425fe58019SAttilio Rao 	struct vattr *a_vap;
3435fe58019SAttilio Rao     };
3445fe58019SAttilio Rao */
3455fe58019SAttilio Rao static int
3465fe58019SAttilio Rao fuse_vnop_create(struct vop_create_args *ap)
3475fe58019SAttilio Rao {
3485fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
3495fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
3505fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
3515fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
3525fe58019SAttilio Rao 	struct thread *td = cnp->cn_thread;
3535fe58019SAttilio Rao 	struct ucred *cred = cnp->cn_cred;
3545fe58019SAttilio Rao 
3555fe58019SAttilio Rao 	struct fuse_open_in *foi;
3565fe58019SAttilio Rao 	struct fuse_entry_out *feo;
35719ef317dSAlan Somers 	struct fuse_open_out *foo;
35819ef317dSAlan Somers 	struct fuse_dispatcher fdi, fdi2;
3595fe58019SAttilio Rao 	struct fuse_dispatcher *fdip = &fdi;
36019ef317dSAlan Somers 	struct fuse_dispatcher *fdip2 = NULL;
3615fe58019SAttilio Rao 
3625fe58019SAttilio Rao 	int err;
3635fe58019SAttilio Rao 
3645fe58019SAttilio Rao 	struct mount *mp = vnode_mount(dvp);
3655fe58019SAttilio Rao 	uint64_t parentnid = VTOFUD(dvp)->nid;
3665fe58019SAttilio Rao 	mode_t mode = MAKEIMODE(vap->va_type, vap->va_mode);
36719ef317dSAlan Somers 	enum fuse_opcode op;
3685fe58019SAttilio Rao 
3695fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
3705fe58019SAttilio Rao 		return ENXIO;
3715fe58019SAttilio Rao 	}
3725fe58019SAttilio Rao 	bzero(&fdi, sizeof(fdi));
3735fe58019SAttilio Rao 
37419ef317dSAlan Somers 	if ((vap->va_type != VREG))
375372b97d0SRick Macklem 		return (EINVAL);
3765fe58019SAttilio Rao 
3775fe58019SAttilio Rao 	if (!fsess_isimpl(mp, FUSE_CREATE)) {
37819ef317dSAlan Somers 		/* Fallback to FUSE_MKNOD/FUSE_OPEN */
37919ef317dSAlan Somers 		fdisp_make_mknod_for_fallback(fdip, cnp, dvp, parentnid, td,
38019ef317dSAlan Somers 			cred, mode, &op);
38119ef317dSAlan Somers 	} else {
38219ef317dSAlan Somers 		/* Use FUSE_CREATE */
38319ef317dSAlan Somers 		op = FUSE_CREATE;
38419ef317dSAlan Somers 		fdisp_init(fdip, sizeof(*foi) + cnp->cn_namelen + 1);
38519ef317dSAlan Somers 		fdisp_make(fdip, op, vnode_mount(dvp), parentnid, td, cred);
3865fe58019SAttilio Rao 		foi = fdip->indata;
3875fe58019SAttilio Rao 		foi->mode = mode;
3885fe58019SAttilio Rao 		foi->flags = O_CREAT | O_RDWR;
3895fe58019SAttilio Rao 		memcpy((char *)fdip->indata + sizeof(*foi), cnp->cn_nameptr,
3905fe58019SAttilio Rao 		    cnp->cn_namelen);
3915fe58019SAttilio Rao 		((char *)fdip->indata)[sizeof(*foi) + cnp->cn_namelen] = '\0';
39219ef317dSAlan Somers 	}
3935fe58019SAttilio Rao 
3945fe58019SAttilio Rao 	err = fdisp_wait_answ(fdip);
3955fe58019SAttilio Rao 
3966de3b00dSAttilio Rao 	if (err) {
39719ef317dSAlan Somers 		if (err == ENOSYS && op == FUSE_CREATE) {
3985fe58019SAttilio Rao 			fsess_set_notimpl(mp, FUSE_CREATE);
39919ef317dSAlan Somers 			fdisp_make_mknod_for_fallback(fdip, cnp, dvp,
40019ef317dSAlan Somers 				parentnid, td, cred, mode, &op);
40119ef317dSAlan Somers 			err = fdisp_wait_answ(fdip);
40219ef317dSAlan Somers 		}
40319ef317dSAlan Somers 		if (err)
4045fe58019SAttilio Rao 			goto out;
4055fe58019SAttilio Rao 	}
406372b97d0SRick Macklem 
4075fe58019SAttilio Rao 	feo = fdip->answ;
4085fe58019SAttilio Rao 
4095fe58019SAttilio Rao 	if ((err = fuse_internal_checkentry(feo, VREG))) {
4105fe58019SAttilio Rao 		goto out;
4115fe58019SAttilio Rao 	}
41219ef317dSAlan Somers 
41319ef317dSAlan Somers 	if (op == FUSE_CREATE) {
41419ef317dSAlan Somers 		foo = (struct fuse_open_out*)(feo + 1);
41519ef317dSAlan Somers 	} else {
41619ef317dSAlan Somers 		/* Issue a separate FUSE_OPEN */
41719ef317dSAlan Somers 		fdip2 = &fdi2;
41819ef317dSAlan Somers 		fdisp_init(fdip2, sizeof(*foi));
41919ef317dSAlan Somers 		fdisp_make(fdip2, FUSE_OPEN, vnode_mount(dvp), feo->nodeid, td,
42019ef317dSAlan Somers 			cred);
42119ef317dSAlan Somers 		foi = fdip2->indata;
42219ef317dSAlan Somers 		foi->mode = mode;
42319ef317dSAlan Somers 		foi->flags = O_RDWR;
42419ef317dSAlan Somers 		err = fdisp_wait_answ(fdip2);
42519ef317dSAlan Somers 		if (err)
42619ef317dSAlan Somers 			goto out;
42719ef317dSAlan Somers 		foo = fdip2->answ;
42819ef317dSAlan Somers 	}
42909176f09SConrad Meyer 	err = fuse_vnode_get(mp, feo, feo->nodeid, dvp, vpp, cnp, VREG);
4305fe58019SAttilio Rao 	if (err) {
4315fe58019SAttilio Rao 		struct fuse_release_in *fri;
4325fe58019SAttilio Rao 		uint64_t nodeid = feo->nodeid;
43319ef317dSAlan Somers 		uint64_t fh_id = foo->fh;
4345fe58019SAttilio Rao 
4355fe58019SAttilio Rao 		fdisp_init(fdip, sizeof(*fri));
4365fe58019SAttilio Rao 		fdisp_make(fdip, FUSE_RELEASE, mp, nodeid, td, cred);
4375fe58019SAttilio Rao 		fri = fdip->indata;
4385fe58019SAttilio Rao 		fri->fh = fh_id;
439e0bec057SAlan Somers 		fri->flags = fuse_filehandle_xlate_to_oflags(FUFH_RDWR);
4405fe58019SAttilio Rao 		fuse_insert_callback(fdip->tick, fuse_internal_forget_callback);
4415fe58019SAttilio Rao 		fuse_insert_message(fdip->tick);
44219ef317dSAlan Somers 		goto out;
4435fe58019SAttilio Rao 	}
4445fe58019SAttilio Rao 	ASSERT_VOP_ELOCKED(*vpp, "fuse_vnop_create");
4455fe58019SAttilio Rao 
44619ef317dSAlan Somers 	fuse_filehandle_init(*vpp, FUFH_RDWR, NULL, foo->fh);
44719ef317dSAlan Somers 	fuse_vnode_open(*vpp, foo->open_flags, td);
4485fe58019SAttilio Rao 	cache_purge_negative(dvp);
4495fe58019SAttilio Rao 
4505fe58019SAttilio Rao out:
45119ef317dSAlan Somers 	if (fdip2)
45219ef317dSAlan Somers 		fdisp_destroy(fdip2);
4535fe58019SAttilio Rao 	fdisp_destroy(fdip);
4545fe58019SAttilio Rao 	return err;
4555fe58019SAttilio Rao }
4565fe58019SAttilio Rao 
4575fe58019SAttilio Rao /*
458915012e0SAlan Somers     struct vnop_fdatasync_args {
459915012e0SAlan Somers 	struct vop_generic_args a_gen;
460915012e0SAlan Somers 	struct vnode * a_vp;
461915012e0SAlan Somers 	struct thread * a_td;
462915012e0SAlan Somers     };
4635fe58019SAttilio Rao */
464915012e0SAlan Somers static int
465915012e0SAlan Somers fuse_vnop_fdatasync(struct vop_fdatasync_args *ap)
466915012e0SAlan Somers {
467915012e0SAlan Somers 	struct vnode *vp = ap->a_vp;
468915012e0SAlan Somers 	struct thread *td = ap->a_td;
469915012e0SAlan Somers 	int waitfor = MNT_WAIT;
470915012e0SAlan Somers 
471915012e0SAlan Somers 	int err = 0;
472915012e0SAlan Somers 
473915012e0SAlan Somers 	if (fuse_isdeadfs(vp)) {
474915012e0SAlan Somers 		return 0;
475915012e0SAlan Somers 	}
476915012e0SAlan Somers 	if ((err = vop_stdfdatasync_buf(ap)))
477915012e0SAlan Somers 		return err;
478915012e0SAlan Somers 
479915012e0SAlan Somers 	return fuse_internal_fsync(vp, td, waitfor, true);
480915012e0SAlan Somers }
4815fe58019SAttilio Rao 
4825fe58019SAttilio Rao /*
4835fe58019SAttilio Rao     struct vnop_fsync_args {
484915012e0SAlan Somers 	struct vop_generic_args a_gen;
4855fe58019SAttilio Rao 	struct vnode * a_vp;
4865fe58019SAttilio Rao 	int  a_waitfor;
4875fe58019SAttilio Rao 	struct thread * a_td;
4885fe58019SAttilio Rao     };
4895fe58019SAttilio Rao */
4905fe58019SAttilio Rao static int
4915fe58019SAttilio Rao fuse_vnop_fsync(struct vop_fsync_args *ap)
4925fe58019SAttilio Rao {
4935fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
4945fe58019SAttilio Rao 	struct thread *td = ap->a_td;
49590612f3cSAlan Somers 	int waitfor = ap->a_waitfor;
496915012e0SAlan Somers 	int err = 0;
4975fe58019SAttilio Rao 
4985fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
4995fe58019SAttilio Rao 		return 0;
5005fe58019SAttilio Rao 	}
5015fe58019SAttilio Rao 	if ((err = vop_stdfsync(ap)))
5025fe58019SAttilio Rao 		return err;
5035fe58019SAttilio Rao 
504915012e0SAlan Somers 	return fuse_internal_fsync(vp, td, waitfor, false);
5055fe58019SAttilio Rao }
5065fe58019SAttilio Rao 
5075fe58019SAttilio Rao /*
5085fe58019SAttilio Rao     struct vnop_getattr_args {
5095fe58019SAttilio Rao 	struct vnode *a_vp;
5105fe58019SAttilio Rao 	struct vattr *a_vap;
5115fe58019SAttilio Rao 	struct ucred *a_cred;
5125fe58019SAttilio Rao 	struct thread *a_td;
5135fe58019SAttilio Rao     };
5145fe58019SAttilio Rao */
5155fe58019SAttilio Rao static int
5165fe58019SAttilio Rao fuse_vnop_getattr(struct vop_getattr_args *ap)
5175fe58019SAttilio Rao {
5185fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
5195fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
5205fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
5215fe58019SAttilio Rao 	struct thread *td = curthread;
5225fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(vp);
5237e4844f7SAlan Somers 	struct fuse_attr_out *fao;
5245fe58019SAttilio Rao 
5255fe58019SAttilio Rao 	int err = 0;
5265fe58019SAttilio Rao 	int dataflags;
5275fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
5285fe58019SAttilio Rao 
5295fe58019SAttilio Rao 	dataflags = fuse_get_mpdata(vnode_mount(vp))->dataflags;
5305fe58019SAttilio Rao 
5315fe58019SAttilio Rao 	/* Note that we are not bailing out on a dead file system just yet. */
5325fe58019SAttilio Rao 
5335fe58019SAttilio Rao 	if (!(dataflags & FSESS_INITED)) {
5345fe58019SAttilio Rao 		if (!vnode_isvroot(vp)) {
5355fe58019SAttilio Rao 			fdata_set_dead(fuse_get_mpdata(vnode_mount(vp)));
5365fe58019SAttilio Rao 			err = ENOTCONN;
5375fe58019SAttilio Rao 			return err;
5385fe58019SAttilio Rao 		} else {
5395fe58019SAttilio Rao 			goto fake;
5405fe58019SAttilio Rao 		}
5415fe58019SAttilio Rao 	}
5425fe58019SAttilio Rao 	fdisp_init(&fdi, 0);
5435fe58019SAttilio Rao 	if ((err = fdisp_simple_putget_vp(&fdi, FUSE_GETATTR, vp, td, cred))) {
5445fe58019SAttilio Rao 		if ((err == ENOTCONN) && vnode_isvroot(vp)) {
545cf169498SAlan Somers 			/* see comment in fuse_vfsop_statfs() */
5465fe58019SAttilio Rao 			fdisp_destroy(&fdi);
5475fe58019SAttilio Rao 			goto fake;
5485fe58019SAttilio Rao 		}
5495fe58019SAttilio Rao 		if (err == ENOENT) {
5505fe58019SAttilio Rao 			fuse_internal_vnode_disappear(vp);
5515fe58019SAttilio Rao 		}
5525fe58019SAttilio Rao 		goto out;
5535fe58019SAttilio Rao 	}
55478a7722fSConrad Meyer 
5557e4844f7SAlan Somers 	fao = (struct fuse_attr_out *)fdi.answ;
5567e4844f7SAlan Somers 	fuse_internal_cache_attrs(vp, &fao->attr, fao->attr_valid,
5577e4844f7SAlan Somers 		fao->attr_valid_nsec, vap);
558d9454fabSAttilio Rao 	if (vap->va_type != vnode_vtype(vp)) {
559d9454fabSAttilio Rao 		fuse_internal_vnode_disappear(vp);
560d9454fabSAttilio Rao 		err = ENOENT;
561d9454fabSAttilio Rao 		goto out;
562d9454fabSAttilio Rao 	}
5635fe58019SAttilio Rao 	if ((fvdat->flag & FN_SIZECHANGE) != 0)
5645fe58019SAttilio Rao 		vap->va_size = fvdat->filesize;
5655fe58019SAttilio Rao 
5665fe58019SAttilio Rao 	if (vnode_isreg(vp) && (fvdat->flag & FN_SIZECHANGE) == 0) {
5675fe58019SAttilio Rao 		/*
5685fe58019SAttilio Rao 	         * This is for those cases when the file size changed without us
5695fe58019SAttilio Rao 	         * knowing, and we want to catch up.
5705fe58019SAttilio Rao 	         */
5715fe58019SAttilio Rao 		off_t new_filesize = ((struct fuse_attr_out *)
5725fe58019SAttilio Rao 				      fdi.answ)->attr.size;
5735fe58019SAttilio Rao 
5745fe58019SAttilio Rao 		if (fvdat->filesize != new_filesize) {
5755fe58019SAttilio Rao 			fuse_vnode_setsize(vp, cred, new_filesize);
576194e691aSConrad Meyer 			fvdat->flag &= ~FN_SIZECHANGE;
5775fe58019SAttilio Rao 		}
5785fe58019SAttilio Rao 	}
5795fe58019SAttilio Rao 
5805fe58019SAttilio Rao out:
5815fe58019SAttilio Rao 	fdisp_destroy(&fdi);
5825fe58019SAttilio Rao 	return err;
5835fe58019SAttilio Rao 
5845fe58019SAttilio Rao fake:
5855fe58019SAttilio Rao 	bzero(vap, sizeof(*vap));
5865fe58019SAttilio Rao 	vap->va_type = vnode_vtype(vp);
5875fe58019SAttilio Rao 
5885fe58019SAttilio Rao 	return 0;
5895fe58019SAttilio Rao }
5905fe58019SAttilio Rao 
5915fe58019SAttilio Rao /*
5925fe58019SAttilio Rao     struct vnop_inactive_args {
5935fe58019SAttilio Rao 	struct vnode *a_vp;
5945fe58019SAttilio Rao 	struct thread *a_td;
5955fe58019SAttilio Rao     };
5965fe58019SAttilio Rao */
5975fe58019SAttilio Rao static int
5985fe58019SAttilio Rao fuse_vnop_inactive(struct vop_inactive_args *ap)
5995fe58019SAttilio Rao {
6005fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
6015fe58019SAttilio Rao 	struct thread *td = ap->a_td;
6025fe58019SAttilio Rao 
6035fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(vp);
6045fe58019SAttilio Rao 	struct fuse_filehandle *fufh = NULL;
6055fe58019SAttilio Rao 
6065fe58019SAttilio Rao 	int type, need_flush = 1;
6075fe58019SAttilio Rao 
6085fe58019SAttilio Rao 	for (type = 0; type < FUFH_MAXTYPE; type++) {
6095fe58019SAttilio Rao 		fufh = &(fvdat->fufh[type]);
6105fe58019SAttilio Rao 		if (FUFH_IS_VALID(fufh)) {
6115fe58019SAttilio Rao 			if (need_flush && vp->v_type == VREG) {
6125fe58019SAttilio Rao 				if ((VTOFUD(vp)->flag & FN_SIZECHANGE) != 0) {
6135fe58019SAttilio Rao 					fuse_vnode_savesize(vp, NULL);
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 			}
6225fe58019SAttilio Rao 			fuse_filehandle_close(vp, type, td, NULL);
6235fe58019SAttilio Rao 		}
6245fe58019SAttilio Rao 	}
6255fe58019SAttilio Rao 
6265fe58019SAttilio Rao 	if ((fvdat->flag & FN_REVOKED) != 0 && fuse_reclaim_revoked) {
6275fe58019SAttilio Rao 		vrecycle(vp);
6285fe58019SAttilio Rao 	}
6295fe58019SAttilio Rao 	return 0;
6305fe58019SAttilio Rao }
6315fe58019SAttilio Rao 
6325fe58019SAttilio Rao /*
6335fe58019SAttilio Rao     struct vnop_link_args {
6345fe58019SAttilio Rao 	struct vnode *a_tdvp;
6355fe58019SAttilio Rao 	struct vnode *a_vp;
6365fe58019SAttilio Rao 	struct componentname *a_cnp;
6375fe58019SAttilio Rao     };
6385fe58019SAttilio Rao */
6395fe58019SAttilio Rao static int
6405fe58019SAttilio Rao fuse_vnop_link(struct vop_link_args *ap)
6415fe58019SAttilio Rao {
6425fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
6435fe58019SAttilio Rao 	struct vnode *tdvp = ap->a_tdvp;
6445fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
6455fe58019SAttilio Rao 
6465fe58019SAttilio Rao 	struct vattr *vap = VTOVA(vp);
6475fe58019SAttilio Rao 
6485fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
6495fe58019SAttilio Rao 	struct fuse_entry_out *feo;
6505fe58019SAttilio Rao 	struct fuse_link_in fli;
6515fe58019SAttilio Rao 
6525fe58019SAttilio Rao 	int err;
6535fe58019SAttilio Rao 
6545fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
6555fe58019SAttilio Rao 		return ENXIO;
6565fe58019SAttilio Rao 	}
6575fe58019SAttilio Rao 	if (vnode_mount(tdvp) != vnode_mount(vp)) {
6585fe58019SAttilio Rao 		return EXDEV;
6595fe58019SAttilio Rao 	}
66078a7722fSConrad Meyer 
66178a7722fSConrad Meyer 	/*
66278a7722fSConrad Meyer 	 * This is a seatbelt check to protect naive userspace filesystems from
66378a7722fSConrad Meyer 	 * themselves and the limitations of the FUSE IPC protocol.  If a
66478a7722fSConrad Meyer 	 * filesystem does not allow attribute caching, assume it is capable of
66578a7722fSConrad Meyer 	 * validating that nlink does not overflow.
66678a7722fSConrad Meyer 	 */
66778a7722fSConrad Meyer 	if (vap != NULL && vap->va_nlink >= FUSE_LINK_MAX)
6685fe58019SAttilio Rao 		return EMLINK;
6695fe58019SAttilio Rao 	fli.oldnodeid = VTOI(vp);
6705fe58019SAttilio Rao 
6715fe58019SAttilio Rao 	fdisp_init(&fdi, 0);
6725fe58019SAttilio Rao 	fuse_internal_newentry_makerequest(vnode_mount(tdvp), VTOI(tdvp), cnp,
6735fe58019SAttilio Rao 	    FUSE_LINK, &fli, sizeof(fli), &fdi);
6745fe58019SAttilio Rao 	if ((err = fdisp_wait_answ(&fdi))) {
6755fe58019SAttilio Rao 		goto out;
6765fe58019SAttilio Rao 	}
6775fe58019SAttilio Rao 	feo = fdi.answ;
6785fe58019SAttilio Rao 
6795fe58019SAttilio Rao 	err = fuse_internal_checkentry(feo, vnode_vtype(vp));
6805fe58019SAttilio Rao out:
6815fe58019SAttilio Rao 	fdisp_destroy(&fdi);
6825fe58019SAttilio Rao 	return err;
6835fe58019SAttilio Rao }
6845fe58019SAttilio Rao 
6855fe58019SAttilio Rao /*
6865fe58019SAttilio Rao     struct vnop_lookup_args {
6875fe58019SAttilio Rao 	struct vnodeop_desc *a_desc;
6885fe58019SAttilio Rao 	struct vnode *a_dvp;
6895fe58019SAttilio Rao 	struct vnode **a_vpp;
6905fe58019SAttilio Rao 	struct componentname *a_cnp;
6915fe58019SAttilio Rao     };
6925fe58019SAttilio Rao */
6935fe58019SAttilio Rao int
6945fe58019SAttilio Rao fuse_vnop_lookup(struct vop_lookup_args *ap)
6955fe58019SAttilio Rao {
6965fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
6975fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
6985fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
6995fe58019SAttilio Rao 	struct thread *td = cnp->cn_thread;
7005fe58019SAttilio Rao 	struct ucred *cred = cnp->cn_cred;
7015fe58019SAttilio Rao 
7025fe58019SAttilio Rao 	int nameiop = cnp->cn_nameiop;
7035fe58019SAttilio Rao 	int flags = cnp->cn_flags;
7045fe58019SAttilio Rao 	int wantparent = flags & (LOCKPARENT | WANTPARENT);
7055fe58019SAttilio Rao 	int islastcn = flags & ISLASTCN;
7065fe58019SAttilio Rao 	struct mount *mp = vnode_mount(dvp);
7075fe58019SAttilio Rao 
7085fe58019SAttilio Rao 	int err = 0;
7095fe58019SAttilio Rao 	int lookup_err = 0;
7105fe58019SAttilio Rao 	struct vnode *vp = NULL;
7115fe58019SAttilio Rao 
7125fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
7135fe58019SAttilio Rao 	enum fuse_opcode op;
7145fe58019SAttilio Rao 
7155fe58019SAttilio Rao 	uint64_t nid;
7165fe58019SAttilio Rao 	struct fuse_access_param facp;
7175fe58019SAttilio Rao 
7185fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
7195fe58019SAttilio Rao 		*vpp = NULL;
7205fe58019SAttilio Rao 		return ENXIO;
7215fe58019SAttilio Rao 	}
7225fe58019SAttilio Rao 	if (!vnode_isdir(dvp)) {
7235fe58019SAttilio Rao 		return ENOTDIR;
7245fe58019SAttilio Rao 	}
7255fe58019SAttilio Rao 	if (islastcn && vfs_isrdonly(mp) && (nameiop != LOOKUP)) {
7265fe58019SAttilio Rao 		return EROFS;
7275fe58019SAttilio Rao 	}
7285fe58019SAttilio Rao 	/*
7295fe58019SAttilio Rao 	 * We do access check prior to doing anything else only in the case
7305fe58019SAttilio Rao 	 * when we are at fs root (we'd like to say, "we are at the first
7315fe58019SAttilio Rao 	 * component", but that's not exactly the same... nevermind).
7325fe58019SAttilio Rao 	 * See further comments at further access checks.
7335fe58019SAttilio Rao 	 */
7345fe58019SAttilio Rao 
7355fe58019SAttilio Rao 	bzero(&facp, sizeof(facp));
7365fe58019SAttilio Rao 	if (vnode_isvroot(dvp)) {	/* early permission check hack */
7375fe58019SAttilio Rao 		if ((err = fuse_internal_access(dvp, VEXEC, &facp, td, cred))) {
7385fe58019SAttilio Rao 			return err;
7395fe58019SAttilio Rao 		}
7405fe58019SAttilio Rao 	}
7415fe58019SAttilio Rao 	if (flags & ISDOTDOT) {
7425fe58019SAttilio Rao 		nid = VTOFUD(dvp)->parent_nid;
7435fe58019SAttilio Rao 		if (nid == 0) {
7445fe58019SAttilio Rao 			return ENOENT;
7455fe58019SAttilio Rao 		}
7465fe58019SAttilio Rao 		fdisp_init(&fdi, 0);
7475fe58019SAttilio Rao 		op = FUSE_GETATTR;
7485fe58019SAttilio Rao 		goto calldaemon;
7495fe58019SAttilio Rao 	} else if (cnp->cn_namelen == 1 && *(cnp->cn_nameptr) == '.') {
7505fe58019SAttilio Rao 		nid = VTOI(dvp);
7515fe58019SAttilio Rao 		fdisp_init(&fdi, 0);
7525fe58019SAttilio Rao 		op = FUSE_GETATTR;
7535fe58019SAttilio Rao 		goto calldaemon;
7545fe58019SAttilio Rao 	} else if (fuse_lookup_cache_enable) {
7555fe58019SAttilio Rao 		err = cache_lookup(dvp, vpp, cnp, NULL, NULL);
7565fe58019SAttilio Rao 		switch (err) {
7575fe58019SAttilio Rao 
7585fe58019SAttilio Rao 		case -1:		/* positive match */
7595fe58019SAttilio Rao 			atomic_add_acq_long(&fuse_lookup_cache_hits, 1);
7605fe58019SAttilio Rao 			return 0;
7615fe58019SAttilio Rao 
7625fe58019SAttilio Rao 		case 0:		/* no match in cache */
7635fe58019SAttilio Rao 			atomic_add_acq_long(&fuse_lookup_cache_misses, 1);
7645fe58019SAttilio Rao 			break;
7655fe58019SAttilio Rao 
7665fe58019SAttilio Rao 		case ENOENT:		/* negative match */
7675fe58019SAttilio Rao 			/* fall through */
7685fe58019SAttilio Rao 		default:
7695fe58019SAttilio Rao 			return err;
7705fe58019SAttilio Rao 		}
7715fe58019SAttilio Rao 	}
7725fe58019SAttilio Rao 	nid = VTOI(dvp);
7735fe58019SAttilio Rao 	fdisp_init(&fdi, cnp->cn_namelen + 1);
7745fe58019SAttilio Rao 	op = FUSE_LOOKUP;
7755fe58019SAttilio Rao 
7765fe58019SAttilio Rao calldaemon:
7775fe58019SAttilio Rao 	fdisp_make(&fdi, op, mp, nid, td, cred);
7785fe58019SAttilio Rao 
7795fe58019SAttilio Rao 	if (op == FUSE_LOOKUP) {
7805fe58019SAttilio Rao 		memcpy(fdi.indata, cnp->cn_nameptr, cnp->cn_namelen);
7815fe58019SAttilio Rao 		((char *)fdi.indata)[cnp->cn_namelen] = '\0';
7825fe58019SAttilio Rao 	}
7835fe58019SAttilio Rao 	lookup_err = fdisp_wait_answ(&fdi);
7845fe58019SAttilio Rao 
7855fe58019SAttilio Rao 	if ((op == FUSE_LOOKUP) && !lookup_err) {	/* lookup call succeeded */
7865fe58019SAttilio Rao 		nid = ((struct fuse_entry_out *)fdi.answ)->nodeid;
7875fe58019SAttilio Rao 		if (!nid) {
7885fe58019SAttilio Rao 			/*
7895fe58019SAttilio Rao 	                 * zero nodeid is the same as "not found",
7905fe58019SAttilio Rao 	                 * but it's also cacheable (which we keep
7915fe58019SAttilio Rao 	                 * keep on doing not as of writing this)
7925fe58019SAttilio Rao 	                 */
7936248288eSAlan Somers 			fdi.answ_stat = ENOENT;
7945fe58019SAttilio Rao 			lookup_err = ENOENT;
7955fe58019SAttilio Rao 		} else if (nid == FUSE_ROOT_ID) {
7965fe58019SAttilio Rao 			lookup_err = EINVAL;
7975fe58019SAttilio Rao 		}
7985fe58019SAttilio Rao 	}
7995fe58019SAttilio Rao 	if (lookup_err &&
8005fe58019SAttilio Rao 	    (!fdi.answ_stat || lookup_err != ENOENT || op != FUSE_LOOKUP)) {
8015fe58019SAttilio Rao 		fdisp_destroy(&fdi);
8025fe58019SAttilio Rao 		return lookup_err;
8035fe58019SAttilio Rao 	}
8045fe58019SAttilio Rao 	/* lookup_err, if non-zero, must be ENOENT at this point */
8055fe58019SAttilio Rao 
8065fe58019SAttilio Rao 	if (lookup_err) {
8075fe58019SAttilio Rao 
8085fe58019SAttilio Rao 		if ((nameiop == CREATE || nameiop == RENAME) && islastcn
8095fe58019SAttilio Rao 		     /* && directory dvp has not been removed */ ) {
8105fe58019SAttilio Rao 
8115fe58019SAttilio Rao 			if (vfs_isrdonly(mp)) {
8125fe58019SAttilio Rao 				err = EROFS;
8135fe58019SAttilio Rao 				goto out;
8145fe58019SAttilio Rao 			}
8155fe58019SAttilio Rao #if 0 /* THINK_ABOUT_THIS */
8165fe58019SAttilio Rao 			if ((err = fuse_internal_access(dvp, VWRITE, cred, td, &facp))) {
8175fe58019SAttilio Rao 				goto out;
8185fe58019SAttilio Rao 			}
8195fe58019SAttilio Rao #endif
8205fe58019SAttilio Rao 
8215fe58019SAttilio Rao 			/*
8225fe58019SAttilio Rao 	                 * Possibly record the position of a slot in the
8235fe58019SAttilio Rao 	                 * directory large enough for the new component name.
8245fe58019SAttilio Rao 	                 * This can be recorded in the vnode private data for
8255fe58019SAttilio Rao 	                 * dvp. Set the SAVENAME flag to hold onto the
8265fe58019SAttilio Rao 	                 * pathname for use later in VOP_CREATE or VOP_RENAME.
8275fe58019SAttilio Rao 	                 */
8285fe58019SAttilio Rao 			cnp->cn_flags |= SAVENAME;
8295fe58019SAttilio Rao 
8305fe58019SAttilio Rao 			err = EJUSTRETURN;
8315fe58019SAttilio Rao 			goto out;
8325fe58019SAttilio Rao 		}
8335fe58019SAttilio Rao 		/* Consider inserting name into cache. */
8345fe58019SAttilio Rao 
8355fe58019SAttilio Rao 		/*
8365fe58019SAttilio Rao 	         * No we can't use negative caching, as the fs
8375fe58019SAttilio Rao 	         * changes are out of our control.
8385fe58019SAttilio Rao 	         * False positives' falseness turns out just as things
8395fe58019SAttilio Rao 	         * go by, but false negatives' falseness doesn't.
8405fe58019SAttilio Rao 	         * (and aiding the caching mechanism with extra control
8415fe58019SAttilio Rao 	         * mechanisms comes quite close to beating the whole purpose
8425fe58019SAttilio Rao 	         * caching...)
8435fe58019SAttilio Rao 	         */
8445fe58019SAttilio Rao #if 0
8456c21f6edSKonstantin Belousov 		if ((cnp->cn_flags & MAKEENTRY) != 0) {
846cf169498SAlan Somers 			SDT_PROBE2(fuse, , vnops, trace, 1,
847cf169498SAlan Somers 				"inserting NULL into cache");
8485fe58019SAttilio Rao 			cache_enter(dvp, NULL, cnp);
8495fe58019SAttilio Rao 		}
8505fe58019SAttilio Rao #endif
8515fe58019SAttilio Rao 		err = ENOENT;
8525fe58019SAttilio Rao 		goto out;
8535fe58019SAttilio Rao 
8545fe58019SAttilio Rao 	} else {
8555fe58019SAttilio Rao 
8565fe58019SAttilio Rao 		/* !lookup_err */
8575fe58019SAttilio Rao 
8585fe58019SAttilio Rao 		struct fuse_entry_out *feo = NULL;
8595fe58019SAttilio Rao 		struct fuse_attr *fattr = NULL;
8605fe58019SAttilio Rao 
8615fe58019SAttilio Rao 		if (op == FUSE_GETATTR) {
8625fe58019SAttilio Rao 			fattr = &((struct fuse_attr_out *)fdi.answ)->attr;
8635fe58019SAttilio Rao 		} else {
8645fe58019SAttilio Rao 			feo = (struct fuse_entry_out *)fdi.answ;
8655fe58019SAttilio Rao 			fattr = &(feo->attr);
8665fe58019SAttilio Rao 		}
8675fe58019SAttilio Rao 
8685fe58019SAttilio Rao 		/*
8695fe58019SAttilio Rao 	         * If deleting, and at end of pathname, return parameters
8705fe58019SAttilio Rao 	         * which can be used to remove file.  If the wantparent flag
8715fe58019SAttilio Rao 	         * isn't set, we return only the directory, otherwise we go on
8725fe58019SAttilio Rao 	         * and lock the inode, being careful with ".".
8735fe58019SAttilio Rao 	         */
8745fe58019SAttilio Rao 		if (nameiop == DELETE && islastcn) {
8755fe58019SAttilio Rao 			/*
8765fe58019SAttilio Rao 	                 * Check for write access on directory.
8775fe58019SAttilio Rao 	                 */
8785fe58019SAttilio Rao 			facp.xuid = fattr->uid;
8795fe58019SAttilio Rao 			facp.facc_flags |= FACCESS_STICKY;
8805fe58019SAttilio Rao 			err = fuse_internal_access(dvp, VWRITE, &facp, td, cred);
8815fe58019SAttilio Rao 			facp.facc_flags &= ~FACCESS_XQUERIES;
8825fe58019SAttilio Rao 
8835fe58019SAttilio Rao 			if (err) {
8845fe58019SAttilio Rao 				goto out;
8855fe58019SAttilio Rao 			}
8865fe58019SAttilio Rao 			if (nid == VTOI(dvp)) {
8875fe58019SAttilio Rao 				vref(dvp);
8885fe58019SAttilio Rao 				*vpp = dvp;
8895fe58019SAttilio Rao 			} else {
89009176f09SConrad Meyer 				err = fuse_vnode_get(dvp->v_mount, feo, nid,
89109176f09SConrad Meyer 				    dvp, &vp, cnp, IFTOVT(fattr->mode));
8925fe58019SAttilio Rao 				if (err)
8935fe58019SAttilio Rao 					goto out;
8945fe58019SAttilio Rao 				*vpp = vp;
8955fe58019SAttilio Rao 			}
8965fe58019SAttilio Rao 
8975fe58019SAttilio Rao 			/*
8985fe58019SAttilio Rao 			 * Save the name for use in VOP_RMDIR and VOP_REMOVE
8995fe58019SAttilio Rao 			 * later.
9005fe58019SAttilio Rao 			 */
9015fe58019SAttilio Rao 			cnp->cn_flags |= SAVENAME;
9025fe58019SAttilio Rao 			goto out;
9035fe58019SAttilio Rao 
9045fe58019SAttilio Rao 		}
9055fe58019SAttilio Rao 		/*
9065fe58019SAttilio Rao 	         * If rewriting (RENAME), return the inode and the
9075fe58019SAttilio Rao 	         * information required to rewrite the present directory
9085fe58019SAttilio Rao 	         * Must get inode of directory entry to verify it's a
9095fe58019SAttilio Rao 	         * regular file, or empty directory.
9105fe58019SAttilio Rao 	         */
9115fe58019SAttilio Rao 		if (nameiop == RENAME && wantparent && islastcn) {
9125fe58019SAttilio Rao 
9135fe58019SAttilio Rao #if 0 /* THINK_ABOUT_THIS */
9145fe58019SAttilio Rao 			if ((err = fuse_internal_access(dvp, VWRITE, cred, td, &facp))) {
9155fe58019SAttilio Rao 				goto out;
9165fe58019SAttilio Rao 			}
9175fe58019SAttilio Rao #endif
9185fe58019SAttilio Rao 
9195fe58019SAttilio Rao 			/*
9205fe58019SAttilio Rao 	                 * Check for "."
9215fe58019SAttilio Rao 	                 */
9225fe58019SAttilio Rao 			if (nid == VTOI(dvp)) {
9235fe58019SAttilio Rao 				err = EISDIR;
9245fe58019SAttilio Rao 				goto out;
9255fe58019SAttilio Rao 			}
92609176f09SConrad Meyer 			err = fuse_vnode_get(vnode_mount(dvp), feo, nid, dvp,
92709176f09SConrad Meyer 			    &vp, cnp, IFTOVT(fattr->mode));
9285fe58019SAttilio Rao 			if (err) {
9295fe58019SAttilio Rao 				goto out;
9305fe58019SAttilio Rao 			}
9315fe58019SAttilio Rao 			*vpp = vp;
9325fe58019SAttilio Rao 			/*
9335fe58019SAttilio Rao 	                 * Save the name for use in VOP_RENAME later.
9345fe58019SAttilio Rao 	                 */
9355fe58019SAttilio Rao 			cnp->cn_flags |= SAVENAME;
9365fe58019SAttilio Rao 
9375fe58019SAttilio Rao 			goto out;
9385fe58019SAttilio Rao 		}
9395fe58019SAttilio Rao 		if (flags & ISDOTDOT) {
9405fe58019SAttilio Rao 			struct mount *mp;
9415fe58019SAttilio Rao 			int ltype;
9425fe58019SAttilio Rao 
9435fe58019SAttilio Rao 			/*
9445fe58019SAttilio Rao 			 * Expanded copy of vn_vget_ino() so that
9455fe58019SAttilio Rao 			 * fuse_vnode_get() can be used.
9465fe58019SAttilio Rao 			 */
9475fe58019SAttilio Rao 			mp = dvp->v_mount;
9485fe58019SAttilio Rao 			ltype = VOP_ISLOCKED(dvp);
9495fe58019SAttilio Rao 			err = vfs_busy(mp, MBF_NOWAIT);
9505fe58019SAttilio Rao 			if (err != 0) {
9515fe58019SAttilio Rao 				vfs_ref(mp);
9525fe58019SAttilio Rao 				VOP_UNLOCK(dvp, 0);
9535fe58019SAttilio Rao 				err = vfs_busy(mp, 0);
9545fe58019SAttilio Rao 				vn_lock(dvp, ltype | LK_RETRY);
9555fe58019SAttilio Rao 				vfs_rel(mp);
9565fe58019SAttilio Rao 				if (err)
9575fe58019SAttilio Rao 					goto out;
9585fe58019SAttilio Rao 				if ((dvp->v_iflag & VI_DOOMED) != 0) {
9595fe58019SAttilio Rao 					err = ENOENT;
9605fe58019SAttilio Rao 					vfs_unbusy(mp);
9615fe58019SAttilio Rao 					goto out;
9625fe58019SAttilio Rao 				}
9635fe58019SAttilio Rao 			}
9645fe58019SAttilio Rao 			VOP_UNLOCK(dvp, 0);
96509176f09SConrad Meyer 			err = fuse_vnode_get(vnode_mount(dvp), feo, nid, NULL,
96609176f09SConrad Meyer 			    &vp, cnp, IFTOVT(fattr->mode));
9675fe58019SAttilio Rao 			vfs_unbusy(mp);
9685fe58019SAttilio Rao 			vn_lock(dvp, ltype | LK_RETRY);
9695fe58019SAttilio Rao 			if ((dvp->v_iflag & VI_DOOMED) != 0) {
9705fe58019SAttilio Rao 				if (err == 0)
9715fe58019SAttilio Rao 					vput(vp);
9725fe58019SAttilio Rao 				err = ENOENT;
9735fe58019SAttilio Rao 			}
9745fe58019SAttilio Rao 			if (err)
9755fe58019SAttilio Rao 				goto out;
9765fe58019SAttilio Rao 			*vpp = vp;
9775fe58019SAttilio Rao 		} else if (nid == VTOI(dvp)) {
9785fe58019SAttilio Rao 			vref(dvp);
9795fe58019SAttilio Rao 			*vpp = dvp;
9805fe58019SAttilio Rao 		} else {
9813c324b94SConrad Meyer 			struct fuse_vnode_data *fvdat;
9823c324b94SConrad Meyer 
98309176f09SConrad Meyer 			err = fuse_vnode_get(vnode_mount(dvp), feo, nid, dvp,
98409176f09SConrad Meyer 			    &vp, cnp, IFTOVT(fattr->mode));
9855fe58019SAttilio Rao 			if (err) {
9865fe58019SAttilio Rao 				goto out;
9875fe58019SAttilio Rao 			}
9885fe58019SAttilio Rao 			fuse_vnode_setparent(vp, dvp);
9893c324b94SConrad Meyer 
9903c324b94SConrad Meyer 			/*
9913c324b94SConrad Meyer 			 * In the case where we are looking up a FUSE node
9923c324b94SConrad Meyer 			 * represented by an existing cached vnode, and the
9933c324b94SConrad Meyer 			 * true size reported by FUSE_LOOKUP doesn't match
9943c324b94SConrad Meyer 			 * the vnode's cached size, fix the vnode cache to
9953c324b94SConrad Meyer 			 * match the real object size.
9963c324b94SConrad Meyer 			 *
9973c324b94SConrad Meyer 			 * This can occur via FUSE distributed filesystems,
9983c324b94SConrad Meyer 			 * irregular files, etc.
9993c324b94SConrad Meyer 			 */
10003c324b94SConrad Meyer 			fvdat = VTOFUD(vp);
10013c324b94SConrad Meyer 			if (vnode_isreg(vp) &&
10023c324b94SConrad Meyer 			    fattr->size != fvdat->filesize) {
10033c324b94SConrad Meyer 				/*
10043c324b94SConrad Meyer 				 * The FN_SIZECHANGE flag reflects a dirty
10053c324b94SConrad Meyer 				 * append.  If userspace lets us know our cache
10063c324b94SConrad Meyer 				 * is invalid, that write was lost.  (Dirty
10073c324b94SConrad Meyer 				 * writes that do not cause append are also
10083c324b94SConrad Meyer 				 * lost, but we don't detect them here.)
10093c324b94SConrad Meyer 				 *
10103c324b94SConrad Meyer 				 * XXX: Maybe disable WB caching on this mount.
10113c324b94SConrad Meyer 				 */
10123c324b94SConrad Meyer 				if (fvdat->flag & FN_SIZECHANGE)
10133c324b94SConrad Meyer 					printf("%s: WB cache incoherent on "
10143c324b94SConrad Meyer 					    "%s!\n", __func__,
10153c324b94SConrad Meyer 					    vnode_mount(vp)->mnt_stat.f_mntonname);
10163c324b94SConrad Meyer 
10173c324b94SConrad Meyer 				(void)fuse_vnode_setsize(vp, cred, fattr->size);
10183c324b94SConrad Meyer 				fvdat->flag &= ~FN_SIZECHANGE;
10193c324b94SConrad Meyer 			}
10205fe58019SAttilio Rao 			*vpp = vp;
10215fe58019SAttilio Rao 		}
10225fe58019SAttilio Rao 
10235fe58019SAttilio Rao 		if (op == FUSE_GETATTR) {
10247e4844f7SAlan Somers 			struct fuse_attr_out *fao =
10257e4844f7SAlan Somers 				(struct fuse_attr_out*)fdi.answ;
10267e4844f7SAlan Somers 			fuse_internal_cache_attrs(*vpp,
10277e4844f7SAlan Somers 				&fao->attr, fao->attr_valid,
10287e4844f7SAlan Somers 				fao->attr_valid_nsec, NULL);
10295fe58019SAttilio Rao 		} else {
10307e4844f7SAlan Somers 			struct fuse_entry_out *feo =
10317e4844f7SAlan Somers 				(struct fuse_entry_out*)fdi.answ;
10327e4844f7SAlan Somers 			fuse_internal_cache_attrs(*vpp,
10337e4844f7SAlan Somers 				&feo->attr, feo->attr_valid,
10347e4844f7SAlan Somers 				feo->attr_valid_nsec, NULL);
10355fe58019SAttilio Rao 		}
10365fe58019SAttilio Rao 
10375fe58019SAttilio Rao 		/* Insert name into cache if appropriate. */
10385fe58019SAttilio Rao 
10395fe58019SAttilio Rao 		/*
10405fe58019SAttilio Rao 	         * Nooo, caching is evil. With caching, we can't avoid stale
10415fe58019SAttilio Rao 	         * information taking over the playground (cached info is not
10425fe58019SAttilio Rao 	         * just positive/negative, it does have qualitative aspects,
10435fe58019SAttilio Rao 	         * too). And a (VOP/FUSE)_GETATTR is always thrown anyway, when
10445fe58019SAttilio Rao 	         * walking down along cached path components, and that's not
10455fe58019SAttilio Rao 	         * any cheaper than FUSE_LOOKUP. This might change with
10465fe58019SAttilio Rao 	         * implementing kernel side attr caching, but... In Linux,
10475fe58019SAttilio Rao 	         * lookup results are not cached, and the daemon is bombarded
10485fe58019SAttilio Rao 	         * with FUSE_LOOKUPS on and on. This shows that by design, the
10495fe58019SAttilio Rao 	         * daemon is expected to handle frequent lookup queries
10505fe58019SAttilio Rao 	         * efficiently, do its caching in userspace, and so on.
10515fe58019SAttilio Rao 	         *
10525fe58019SAttilio Rao 	         * So just leave the name cache alone.
10535fe58019SAttilio Rao 	         */
10545fe58019SAttilio Rao 
10555fe58019SAttilio Rao 		/*
10565fe58019SAttilio Rao 	         * Well, now I know, Linux caches lookups, but with a
10575fe58019SAttilio Rao 	         * timeout... So it's the same thing as attribute caching:
10585fe58019SAttilio Rao 	         * we can deal with it when implement timeouts.
10595fe58019SAttilio Rao 	         */
10605fe58019SAttilio Rao #if 0
10615fe58019SAttilio Rao 		if (cnp->cn_flags & MAKEENTRY) {
10625fe58019SAttilio Rao 			cache_enter(dvp, *vpp, cnp);
10635fe58019SAttilio Rao 		}
10645fe58019SAttilio Rao #endif
10655fe58019SAttilio Rao 	}
10665fe58019SAttilio Rao out:
10675fe58019SAttilio Rao 	if (!lookup_err) {
10685fe58019SAttilio Rao 
10695fe58019SAttilio Rao 		/* No lookup error; need to clean up. */
10705fe58019SAttilio Rao 
10715fe58019SAttilio Rao 		if (err) {		/* Found inode; exit with no vnode. */
10725fe58019SAttilio Rao 			if (op == FUSE_LOOKUP) {
10735fe58019SAttilio Rao 				fuse_internal_forget_send(vnode_mount(dvp), td, cred,
10745fe58019SAttilio Rao 				    nid, 1);
10755fe58019SAttilio Rao 			}
10765fe58019SAttilio Rao 			fdisp_destroy(&fdi);
10775fe58019SAttilio Rao 			return err;
10785fe58019SAttilio Rao 		} else {
10795fe58019SAttilio Rao #ifndef NO_EARLY_PERM_CHECK_HACK
10805fe58019SAttilio Rao 			if (!islastcn) {
10815fe58019SAttilio Rao 				/*
10825fe58019SAttilio Rao 				 * We have the attributes of the next item
10835fe58019SAttilio Rao 				 * *now*, and it's a fact, and we do not
10845fe58019SAttilio Rao 				 * have to do extra work for it (ie, beg the
10855fe58019SAttilio Rao 				 * daemon), and it neither depends on such
10865fe58019SAttilio Rao 				 * accidental things like attr caching. So
10875fe58019SAttilio Rao 				 * the big idea: check credentials *now*,
10885fe58019SAttilio Rao 				 * not at the beginning of the next call to
10895fe58019SAttilio Rao 				 * lookup.
10905fe58019SAttilio Rao 				 *
10915fe58019SAttilio Rao 				 * The first item of the lookup chain (fs root)
10925fe58019SAttilio Rao 				 * won't be checked then here, of course, as
10935fe58019SAttilio Rao 				 * its never "the next". But go and see that
10945fe58019SAttilio Rao 				 * the root is taken care about at the very
10955fe58019SAttilio Rao 				 * beginning of this function.
10965fe58019SAttilio Rao 				 *
10975fe58019SAttilio Rao 				 * Now, given we want to do the access check
10985fe58019SAttilio Rao 				 * this way, one might ask: so then why not
10995fe58019SAttilio Rao 				 * do the access check just after fetching
11005fe58019SAttilio Rao 				 * the inode and its attributes from the
11015fe58019SAttilio Rao 				 * daemon? Why bother with producing the
11025fe58019SAttilio Rao 				 * corresponding vnode at all if something
11035fe58019SAttilio Rao 				 * is not OK? We know what's the deal as
11045fe58019SAttilio Rao 				 * soon as we get those attrs... There is
11055fe58019SAttilio Rao 				 * one bit of info though not given us by
11065fe58019SAttilio Rao 				 * the daemon: whether his response is
1107b3a15dddSPedro F. Giffuni 				 * authoritative or not... His response should
11085fe58019SAttilio Rao 				 * be ignored if something is mounted over
11095fe58019SAttilio Rao 				 * the dir in question. But that can be
11105fe58019SAttilio Rao 				 * known only by having the vnode...
11115fe58019SAttilio Rao 				 */
11125fe58019SAttilio Rao 				int tmpvtype = vnode_vtype(*vpp);
11135fe58019SAttilio Rao 
11145fe58019SAttilio Rao 				bzero(&facp, sizeof(facp));
11155fe58019SAttilio Rao 				/*the early perm check hack */
11165fe58019SAttilio Rao 				    facp.facc_flags |= FACCESS_VA_VALID;
11175fe58019SAttilio Rao 
11185fe58019SAttilio Rao 				if ((tmpvtype != VDIR) && (tmpvtype != VLNK)) {
11195fe58019SAttilio Rao 					err = ENOTDIR;
11205fe58019SAttilio Rao 				}
11215fe58019SAttilio Rao 				if (!err && !vnode_mountedhere(*vpp)) {
11225fe58019SAttilio Rao 					err = fuse_internal_access(*vpp, VEXEC, &facp, td, cred);
11235fe58019SAttilio Rao 				}
11245fe58019SAttilio Rao 				if (err) {
11255fe58019SAttilio Rao 					if (tmpvtype == VLNK)
1126cf169498SAlan Somers 						SDT_PROBE2(fuse, , vnops, trace,
1127cf169498SAlan Somers 						    1, "weird, permission "
1128cf169498SAlan Somers 						    "error with a symlink?");
11295fe58019SAttilio Rao 					vput(*vpp);
11305fe58019SAttilio Rao 					*vpp = NULL;
11315fe58019SAttilio Rao 				}
11325fe58019SAttilio Rao 			}
11335fe58019SAttilio Rao #endif
11345fe58019SAttilio Rao 		}
11355fe58019SAttilio Rao 	}
11365fe58019SAttilio Rao 	fdisp_destroy(&fdi);
11375fe58019SAttilio Rao 
11385fe58019SAttilio Rao 	return err;
11395fe58019SAttilio Rao }
11405fe58019SAttilio Rao 
11415fe58019SAttilio Rao /*
11425fe58019SAttilio Rao     struct vnop_mkdir_args {
11435fe58019SAttilio Rao 	struct vnode *a_dvp;
11445fe58019SAttilio Rao 	struct vnode **a_vpp;
11455fe58019SAttilio Rao 	struct componentname *a_cnp;
11465fe58019SAttilio Rao 	struct vattr *a_vap;
11475fe58019SAttilio Rao     };
11485fe58019SAttilio Rao */
11495fe58019SAttilio Rao static int
11505fe58019SAttilio Rao fuse_vnop_mkdir(struct vop_mkdir_args *ap)
11515fe58019SAttilio Rao {
11525fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
11535fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
11545fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
11555fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
11565fe58019SAttilio Rao 
11575fe58019SAttilio Rao 	struct fuse_mkdir_in fmdi;
11585fe58019SAttilio Rao 
11595fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
11605fe58019SAttilio Rao 		return ENXIO;
11615fe58019SAttilio Rao 	}
11625fe58019SAttilio Rao 	fmdi.mode = MAKEIMODE(vap->va_type, vap->va_mode);
11635fe58019SAttilio Rao 
1164d9454fabSAttilio Rao 	return (fuse_internal_newentry(dvp, vpp, cnp, FUSE_MKDIR, &fmdi,
1165d9454fabSAttilio Rao 	    sizeof(fmdi), VDIR));
11665fe58019SAttilio Rao }
11675fe58019SAttilio Rao 
11685fe58019SAttilio Rao /*
11695fe58019SAttilio Rao     struct vnop_mknod_args {
11705fe58019SAttilio Rao 	struct vnode *a_dvp;
11715fe58019SAttilio Rao 	struct vnode **a_vpp;
11725fe58019SAttilio Rao 	struct componentname *a_cnp;
11735fe58019SAttilio Rao 	struct vattr *a_vap;
11745fe58019SAttilio Rao     };
11755fe58019SAttilio Rao */
11765fe58019SAttilio Rao static int
11775fe58019SAttilio Rao fuse_vnop_mknod(struct vop_mknod_args *ap)
11785fe58019SAttilio Rao {
11795fe58019SAttilio Rao 
1180bf4d7084SAlan Somers 	struct vnode *dvp = ap->a_dvp;
1181bf4d7084SAlan Somers 	struct vnode **vpp = ap->a_vpp;
1182bf4d7084SAlan Somers 	struct componentname *cnp = ap->a_cnp;
1183bf4d7084SAlan Somers 	struct vattr *vap = ap->a_vap;
1184bf4d7084SAlan Somers 	struct fuse_mknod_in fmni;
11855fe58019SAttilio Rao 
1186bf4d7084SAlan Somers 	if (fuse_isdeadfs(dvp))
1187bf4d7084SAlan Somers 		return ENXIO;
1188bf4d7084SAlan Somers 
1189bf4d7084SAlan Somers 	fmni.mode = MAKEIMODE(vap->va_type, vap->va_mode);
1190bf4d7084SAlan Somers 	fmni.rdev = vap->va_rdev;
1191bf4d7084SAlan Somers 	return (fuse_internal_newentry(dvp, vpp, cnp, FUSE_MKNOD, &fmni,
1192bf4d7084SAlan Somers 	    sizeof(fmni), vap->va_type));
1193bf4d7084SAlan Somers }
11945fe58019SAttilio Rao 
11955fe58019SAttilio Rao /*
11965fe58019SAttilio Rao     struct vnop_open_args {
11975fe58019SAttilio Rao 	struct vnode *a_vp;
11985fe58019SAttilio Rao 	int  a_mode;
11995fe58019SAttilio Rao 	struct ucred *a_cred;
12005fe58019SAttilio Rao 	struct thread *a_td;
12015fe58019SAttilio Rao 	int a_fdidx; / struct file *a_fp;
12025fe58019SAttilio Rao     };
12035fe58019SAttilio Rao */
12045fe58019SAttilio Rao static int
12055fe58019SAttilio Rao fuse_vnop_open(struct vop_open_args *ap)
12065fe58019SAttilio Rao {
12075fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
12085fe58019SAttilio Rao 	int mode = ap->a_mode;
12095fe58019SAttilio Rao 	struct thread *td = ap->a_td;
12105fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
12115fe58019SAttilio Rao 
12125fe58019SAttilio Rao 	fufh_type_t fufh_type;
12135fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat;
12145fe58019SAttilio Rao 
1215bf4d7084SAlan Somers 	if (fuse_isdeadfs(vp))
12165fe58019SAttilio Rao 		return ENXIO;
1217bf4d7084SAlan Somers 	if (vp->v_type == VCHR || vp->v_type == VBLK || vp->v_type == VFIFO)
1218bf4d7084SAlan Somers 		return (EOPNOTSUPP);
1219e7df9886SConrad Meyer 	if ((mode & (FREAD | FWRITE)) == 0)
1220e7df9886SConrad Meyer 		return EINVAL;
1221e7df9886SConrad Meyer 
12225fe58019SAttilio Rao 	fvdat = VTOFUD(vp);
12235fe58019SAttilio Rao 
12245fe58019SAttilio Rao 	if (vnode_isdir(vp)) {
12255fe58019SAttilio Rao 		fufh_type = FUFH_RDONLY;
12265fe58019SAttilio Rao 	} else {
12275fe58019SAttilio Rao 		fufh_type = fuse_filehandle_xlate_from_fflags(mode);
12285fe58019SAttilio Rao 	}
12295fe58019SAttilio Rao 
1230e6e24456SRick Macklem 	if (fuse_filehandle_validrw(vp, fufh_type) != FUFH_INVALID) {
1231*5fccbf31SAlan Somers 		fuse_vnode_open(vp, 0, td);
12325fe58019SAttilio Rao 		return 0;
12335fe58019SAttilio Rao 	}
12345fe58019SAttilio Rao 
123519ef317dSAlan Somers 	return fuse_filehandle_open(vp, fufh_type, NULL, td, cred);
12365fe58019SAttilio Rao }
12375fe58019SAttilio Rao 
1238746c92e0SJohn Baldwin static int
1239746c92e0SJohn Baldwin fuse_vnop_pathconf(struct vop_pathconf_args *ap)
1240746c92e0SJohn Baldwin {
1241746c92e0SJohn Baldwin 
1242746c92e0SJohn Baldwin 	switch (ap->a_name) {
1243746c92e0SJohn Baldwin 	case _PC_FILESIZEBITS:
1244746c92e0SJohn Baldwin 		*ap->a_retval = 64;
1245746c92e0SJohn Baldwin 		return (0);
1246599afe53SJohn Baldwin 	case _PC_NAME_MAX:
1247599afe53SJohn Baldwin 		*ap->a_retval = NAME_MAX;
1248599afe53SJohn Baldwin 		return (0);
1249599afe53SJohn Baldwin 	case _PC_LINK_MAX:
1250f83f3d79SJohn Baldwin 		*ap->a_retval = MIN(LONG_MAX, FUSE_LINK_MAX);
1251599afe53SJohn Baldwin 		return (0);
1252746c92e0SJohn Baldwin 	case _PC_SYMLINK_MAX:
1253746c92e0SJohn Baldwin 		*ap->a_retval = MAXPATHLEN;
1254746c92e0SJohn Baldwin 		return (0);
1255746c92e0SJohn Baldwin 	case _PC_NO_TRUNC:
1256746c92e0SJohn Baldwin 		*ap->a_retval = 1;
1257746c92e0SJohn Baldwin 		return (0);
1258746c92e0SJohn Baldwin 	default:
1259746c92e0SJohn Baldwin 		return (vop_stdpathconf(ap));
1260746c92e0SJohn Baldwin 	}
1261746c92e0SJohn Baldwin }
1262746c92e0SJohn Baldwin 
12635fe58019SAttilio Rao /*
12645fe58019SAttilio Rao     struct vnop_read_args {
12655fe58019SAttilio Rao 	struct vnode *a_vp;
12665fe58019SAttilio Rao 	struct uio *a_uio;
12675fe58019SAttilio Rao 	int  a_ioflag;
12685fe58019SAttilio Rao 	struct ucred *a_cred;
12695fe58019SAttilio Rao     };
12705fe58019SAttilio Rao */
12715fe58019SAttilio Rao static int
12725fe58019SAttilio Rao fuse_vnop_read(struct vop_read_args *ap)
12735fe58019SAttilio Rao {
12745fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
12755fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
12765fe58019SAttilio Rao 	int ioflag = ap->a_ioflag;
12775fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
12785fe58019SAttilio Rao 
12795fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
12805fe58019SAttilio Rao 		return ENXIO;
12815fe58019SAttilio Rao 	}
1282ead063e0SEdward Tomasz Napierala 
1283ead063e0SEdward Tomasz Napierala 	if (VTOFUD(vp)->flag & FN_DIRECTIO) {
1284ead063e0SEdward Tomasz Napierala 		ioflag |= IO_DIRECT;
1285ead063e0SEdward Tomasz Napierala 	}
1286ead063e0SEdward Tomasz Napierala 
12875fe58019SAttilio Rao 	return fuse_io_dispatch(vp, uio, ioflag, cred);
12885fe58019SAttilio Rao }
12895fe58019SAttilio Rao 
12905fe58019SAttilio Rao /*
12915fe58019SAttilio Rao     struct vnop_readdir_args {
12925fe58019SAttilio Rao 	struct vnode *a_vp;
12935fe58019SAttilio Rao 	struct uio *a_uio;
12945fe58019SAttilio Rao 	struct ucred *a_cred;
12955fe58019SAttilio Rao 	int *a_eofflag;
12965fe58019SAttilio Rao 	int *ncookies;
12975fe58019SAttilio Rao 	u_long **a_cookies;
12985fe58019SAttilio Rao     };
12995fe58019SAttilio Rao */
13005fe58019SAttilio Rao static int
13015fe58019SAttilio Rao fuse_vnop_readdir(struct vop_readdir_args *ap)
13025fe58019SAttilio Rao {
13035fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
13045fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
13055fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
13065fe58019SAttilio Rao 
13075fe58019SAttilio Rao 	struct fuse_filehandle *fufh = NULL;
13085fe58019SAttilio Rao 	struct fuse_iov cookediov;
13095fe58019SAttilio Rao 
13105fe58019SAttilio Rao 	int err = 0;
13115fe58019SAttilio Rao 	int freefufh = 0;
13125fe58019SAttilio Rao 
13135fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
13145fe58019SAttilio Rao 		return ENXIO;
13155fe58019SAttilio Rao 	}
13165fe58019SAttilio Rao 	if (				/* XXXIP ((uio_iovcnt(uio) > 1)) || */
13175fe58019SAttilio Rao 	    (uio_resid(uio) < sizeof(struct dirent))) {
13185fe58019SAttilio Rao 		return EINVAL;
13195fe58019SAttilio Rao 	}
13205fe58019SAttilio Rao 
13215fe58019SAttilio Rao 	if (!fuse_filehandle_valid(vp, FUFH_RDONLY)) {
1322cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1,
1323cf169498SAlan Somers 			"calling readdir() before open()");
13245fe58019SAttilio Rao 		err = fuse_filehandle_open(vp, FUFH_RDONLY, &fufh, NULL, cred);
13255fe58019SAttilio Rao 		freefufh = 1;
13265fe58019SAttilio Rao 	} else {
13275fe58019SAttilio Rao 		err = fuse_filehandle_get(vp, FUFH_RDONLY, &fufh);
13285fe58019SAttilio Rao 	}
13295fe58019SAttilio Rao 	if (err) {
13305fe58019SAttilio Rao 		return (err);
13315fe58019SAttilio Rao 	}
13325fe58019SAttilio Rao #define DIRCOOKEDSIZE FUSE_DIRENT_ALIGN(FUSE_NAME_OFFSET + MAXNAMLEN + 1)
13335fe58019SAttilio Rao 	fiov_init(&cookediov, DIRCOOKEDSIZE);
13345fe58019SAttilio Rao 
13355fe58019SAttilio Rao 	err = fuse_internal_readdir(vp, uio, fufh, &cookediov);
13365fe58019SAttilio Rao 
13375fe58019SAttilio Rao 	fiov_teardown(&cookediov);
13385fe58019SAttilio Rao 	if (freefufh) {
13395fe58019SAttilio Rao 		fuse_filehandle_close(vp, FUFH_RDONLY, NULL, cred);
13405fe58019SAttilio Rao 	}
13415fe58019SAttilio Rao 	return err;
13425fe58019SAttilio Rao }
13435fe58019SAttilio Rao 
13445fe58019SAttilio Rao /*
13455fe58019SAttilio Rao     struct vnop_readlink_args {
13465fe58019SAttilio Rao 	struct vnode *a_vp;
13475fe58019SAttilio Rao 	struct uio *a_uio;
13485fe58019SAttilio Rao 	struct ucred *a_cred;
13495fe58019SAttilio Rao     };
13505fe58019SAttilio Rao */
13515fe58019SAttilio Rao static int
13525fe58019SAttilio Rao fuse_vnop_readlink(struct vop_readlink_args *ap)
13535fe58019SAttilio Rao {
13545fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
13555fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
13565fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
13575fe58019SAttilio Rao 
13585fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
13595fe58019SAttilio Rao 	int err;
13605fe58019SAttilio Rao 
13615fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
13625fe58019SAttilio Rao 		return ENXIO;
13635fe58019SAttilio Rao 	}
13645fe58019SAttilio Rao 	if (!vnode_islnk(vp)) {
13655fe58019SAttilio Rao 		return EINVAL;
13665fe58019SAttilio Rao 	}
13675fe58019SAttilio Rao 	fdisp_init(&fdi, 0);
13685fe58019SAttilio Rao 	err = fdisp_simple_putget_vp(&fdi, FUSE_READLINK, vp, curthread, cred);
13695fe58019SAttilio Rao 	if (err) {
13705fe58019SAttilio Rao 		goto out;
13715fe58019SAttilio Rao 	}
13725fe58019SAttilio Rao 	if (((char *)fdi.answ)[0] == '/' &&
13735fe58019SAttilio Rao 	    fuse_get_mpdata(vnode_mount(vp))->dataflags & FSESS_PUSH_SYMLINKS_IN) {
13745fe58019SAttilio Rao 		char *mpth = vnode_mount(vp)->mnt_stat.f_mntonname;
13755fe58019SAttilio Rao 
13765fe58019SAttilio Rao 		err = uiomove(mpth, strlen(mpth), uio);
13775fe58019SAttilio Rao 	}
13785fe58019SAttilio Rao 	if (!err) {
13795fe58019SAttilio Rao 		err = uiomove(fdi.answ, fdi.iosize, uio);
13805fe58019SAttilio Rao 	}
13815fe58019SAttilio Rao out:
13825fe58019SAttilio Rao 	fdisp_destroy(&fdi);
13835fe58019SAttilio Rao 	return err;
13845fe58019SAttilio Rao }
13855fe58019SAttilio Rao 
13865fe58019SAttilio Rao /*
13875fe58019SAttilio Rao     struct vnop_reclaim_args {
13885fe58019SAttilio Rao 	struct vnode *a_vp;
13895fe58019SAttilio Rao 	struct thread *a_td;
13905fe58019SAttilio Rao     };
13915fe58019SAttilio Rao */
13925fe58019SAttilio Rao static int
13935fe58019SAttilio Rao fuse_vnop_reclaim(struct vop_reclaim_args *ap)
13945fe58019SAttilio Rao {
13955fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
13965fe58019SAttilio Rao 	struct thread *td = ap->a_td;
13975fe58019SAttilio Rao 
13985fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(vp);
13995fe58019SAttilio Rao 	struct fuse_filehandle *fufh = NULL;
14005fe58019SAttilio Rao 
14015fe58019SAttilio Rao 	int type;
14025fe58019SAttilio Rao 
14035fe58019SAttilio Rao 	if (!fvdat) {
14045fe58019SAttilio Rao 		panic("FUSE: no vnode data during recycling");
14055fe58019SAttilio Rao 	}
14065fe58019SAttilio Rao 	for (type = 0; type < FUFH_MAXTYPE; type++) {
14075fe58019SAttilio Rao 		fufh = &(fvdat->fufh[type]);
14085fe58019SAttilio Rao 		if (FUFH_IS_VALID(fufh)) {
14095fe58019SAttilio Rao 			printf("FUSE: vnode being reclaimed but fufh (type=%d) is valid",
14105fe58019SAttilio Rao 			    type);
14115fe58019SAttilio Rao 			fuse_filehandle_close(vp, type, td, NULL);
14125fe58019SAttilio Rao 		}
14135fe58019SAttilio Rao 	}
14145fe58019SAttilio Rao 
14155fe58019SAttilio Rao 	if ((!fuse_isdeadfs(vp)) && (fvdat->nlookup)) {
14165fe58019SAttilio Rao 		fuse_internal_forget_send(vnode_mount(vp), td, NULL, VTOI(vp),
14175fe58019SAttilio Rao 		    fvdat->nlookup);
14185fe58019SAttilio Rao 	}
14195fe58019SAttilio Rao 	fuse_vnode_setparent(vp, NULL);
14205fe58019SAttilio Rao 	cache_purge(vp);
14215fe58019SAttilio Rao 	vfs_hash_remove(vp);
14225fe58019SAttilio Rao 	vnode_destroy_vobject(vp);
14235fe58019SAttilio Rao 	fuse_vnode_destroy(vp);
14245fe58019SAttilio Rao 
14255fe58019SAttilio Rao 	return 0;
14265fe58019SAttilio Rao }
14275fe58019SAttilio Rao 
14285fe58019SAttilio Rao /*
14295fe58019SAttilio Rao     struct vnop_remove_args {
14305fe58019SAttilio Rao 	struct vnode *a_dvp;
14315fe58019SAttilio Rao 	struct vnode *a_vp;
14325fe58019SAttilio Rao 	struct componentname *a_cnp;
14335fe58019SAttilio Rao     };
14345fe58019SAttilio Rao */
14355fe58019SAttilio Rao static int
14365fe58019SAttilio Rao fuse_vnop_remove(struct vop_remove_args *ap)
14375fe58019SAttilio Rao {
14385fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
14395fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
14405fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
14415fe58019SAttilio Rao 
14425fe58019SAttilio Rao 	int err;
14435fe58019SAttilio Rao 
14445fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
14455fe58019SAttilio Rao 		return ENXIO;
14465fe58019SAttilio Rao 	}
14475fe58019SAttilio Rao 	if (vnode_isdir(vp)) {
14485fe58019SAttilio Rao 		return EPERM;
14495fe58019SAttilio Rao 	}
14505fe58019SAttilio Rao 	cache_purge(vp);
14515fe58019SAttilio Rao 
14525fe58019SAttilio Rao 	err = fuse_internal_remove(dvp, vp, cnp, FUSE_UNLINK);
14535fe58019SAttilio Rao 
1454d9454fabSAttilio Rao 	if (err == 0)
14555fe58019SAttilio Rao 		fuse_internal_vnode_disappear(vp);
14565fe58019SAttilio Rao 	return err;
14575fe58019SAttilio Rao }
14585fe58019SAttilio Rao 
14595fe58019SAttilio Rao /*
14605fe58019SAttilio Rao     struct vnop_rename_args {
14615fe58019SAttilio Rao 	struct vnode *a_fdvp;
14625fe58019SAttilio Rao 	struct vnode *a_fvp;
14635fe58019SAttilio Rao 	struct componentname *a_fcnp;
14645fe58019SAttilio Rao 	struct vnode *a_tdvp;
14655fe58019SAttilio Rao 	struct vnode *a_tvp;
14665fe58019SAttilio Rao 	struct componentname *a_tcnp;
14675fe58019SAttilio Rao     };
14685fe58019SAttilio Rao */
14695fe58019SAttilio Rao static int
14705fe58019SAttilio Rao fuse_vnop_rename(struct vop_rename_args *ap)
14715fe58019SAttilio Rao {
14725fe58019SAttilio Rao 	struct vnode *fdvp = ap->a_fdvp;
14735fe58019SAttilio Rao 	struct vnode *fvp = ap->a_fvp;
14745fe58019SAttilio Rao 	struct componentname *fcnp = ap->a_fcnp;
14755fe58019SAttilio Rao 	struct vnode *tdvp = ap->a_tdvp;
14765fe58019SAttilio Rao 	struct vnode *tvp = ap->a_tvp;
14775fe58019SAttilio Rao 	struct componentname *tcnp = ap->a_tcnp;
14785fe58019SAttilio Rao 	struct fuse_data *data;
14795fe58019SAttilio Rao 
14805fe58019SAttilio Rao 	int err = 0;
14815fe58019SAttilio Rao 
14825fe58019SAttilio Rao 	if (fuse_isdeadfs(fdvp)) {
14835fe58019SAttilio Rao 		return ENXIO;
14845fe58019SAttilio Rao 	}
14855fe58019SAttilio Rao 	if (fvp->v_mount != tdvp->v_mount ||
14865fe58019SAttilio Rao 	    (tvp && fvp->v_mount != tvp->v_mount)) {
1487cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1, "cross-device rename");
14885fe58019SAttilio Rao 		err = EXDEV;
14895fe58019SAttilio Rao 		goto out;
14905fe58019SAttilio Rao 	}
14915fe58019SAttilio Rao 	cache_purge(fvp);
14925fe58019SAttilio Rao 
14935fe58019SAttilio Rao 	/*
14945fe58019SAttilio Rao 	 * FUSE library is expected to check if target directory is not
14955fe58019SAttilio Rao 	 * under the source directory in the file system tree.
14965fe58019SAttilio Rao 	 * Linux performs this check at VFS level.
14975fe58019SAttilio Rao 	 */
14985fe58019SAttilio Rao 	data = fuse_get_mpdata(vnode_mount(tdvp));
14995fe58019SAttilio Rao 	sx_xlock(&data->rename_lock);
15005fe58019SAttilio Rao 	err = fuse_internal_rename(fdvp, fcnp, tdvp, tcnp);
15015fe58019SAttilio Rao 	if (err == 0) {
1502d9454fabSAttilio Rao 		if (tdvp != fdvp)
15035fe58019SAttilio Rao 			fuse_vnode_setparent(fvp, tdvp);
15045fe58019SAttilio Rao 		if (tvp != NULL)
15055fe58019SAttilio Rao 			fuse_vnode_setparent(tvp, NULL);
15065fe58019SAttilio Rao 	}
15075fe58019SAttilio Rao 	sx_unlock(&data->rename_lock);
15085fe58019SAttilio Rao 
15095fe58019SAttilio Rao 	if (tvp != NULL && tvp != fvp) {
15105fe58019SAttilio Rao 		cache_purge(tvp);
15115fe58019SAttilio Rao 	}
15125fe58019SAttilio Rao 	if (vnode_isdir(fvp)) {
15135fe58019SAttilio Rao 		if ((tvp != NULL) && vnode_isdir(tvp)) {
15145fe58019SAttilio Rao 			cache_purge(tdvp);
15155fe58019SAttilio Rao 		}
15165fe58019SAttilio Rao 		cache_purge(fdvp);
15175fe58019SAttilio Rao 	}
15185fe58019SAttilio Rao out:
15195fe58019SAttilio Rao 	if (tdvp == tvp) {
15205fe58019SAttilio Rao 		vrele(tdvp);
15215fe58019SAttilio Rao 	} else {
15225fe58019SAttilio Rao 		vput(tdvp);
15235fe58019SAttilio Rao 	}
15245fe58019SAttilio Rao 	if (tvp != NULL) {
15255fe58019SAttilio Rao 		vput(tvp);
15265fe58019SAttilio Rao 	}
15275fe58019SAttilio Rao 	vrele(fdvp);
15285fe58019SAttilio Rao 	vrele(fvp);
15295fe58019SAttilio Rao 
15305fe58019SAttilio Rao 	return err;
15315fe58019SAttilio Rao }
15325fe58019SAttilio Rao 
15335fe58019SAttilio Rao /*
15345fe58019SAttilio Rao     struct vnop_rmdir_args {
15355fe58019SAttilio Rao 	    struct vnode *a_dvp;
15365fe58019SAttilio Rao 	    struct vnode *a_vp;
15375fe58019SAttilio Rao 	    struct componentname *a_cnp;
15385fe58019SAttilio Rao     } *ap;
15395fe58019SAttilio Rao */
15405fe58019SAttilio Rao static int
15415fe58019SAttilio Rao fuse_vnop_rmdir(struct vop_rmdir_args *ap)
15425fe58019SAttilio Rao {
15435fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
15445fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
15455fe58019SAttilio Rao 
15465fe58019SAttilio Rao 	int err;
15475fe58019SAttilio Rao 
15485fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
15495fe58019SAttilio Rao 		return ENXIO;
15505fe58019SAttilio Rao 	}
15515fe58019SAttilio Rao 	if (VTOFUD(vp) == VTOFUD(dvp)) {
15525fe58019SAttilio Rao 		return EINVAL;
15535fe58019SAttilio Rao 	}
15545fe58019SAttilio Rao 	err = fuse_internal_remove(dvp, vp, ap->a_cnp, FUSE_RMDIR);
15555fe58019SAttilio Rao 
1556d9454fabSAttilio Rao 	if (err == 0)
15575fe58019SAttilio Rao 		fuse_internal_vnode_disappear(vp);
15585fe58019SAttilio Rao 	return err;
15595fe58019SAttilio Rao }
15605fe58019SAttilio Rao 
15615fe58019SAttilio Rao /*
15625fe58019SAttilio Rao     struct vnop_setattr_args {
15635fe58019SAttilio Rao 	struct vnode *a_vp;
15645fe58019SAttilio Rao 	struct vattr *a_vap;
15655fe58019SAttilio Rao 	struct ucred *a_cred;
15665fe58019SAttilio Rao 	struct thread *a_td;
15675fe58019SAttilio Rao     };
15685fe58019SAttilio Rao */
15695fe58019SAttilio Rao static int
15705fe58019SAttilio Rao fuse_vnop_setattr(struct vop_setattr_args *ap)
15715fe58019SAttilio Rao {
15725fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
15735fe58019SAttilio Rao 	struct vattr *vap = ap->a_vap;
15745fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
15755fe58019SAttilio Rao 	struct thread *td = curthread;
15765fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
15775fe58019SAttilio Rao 	struct fuse_setattr_in *fsai;
15785fe58019SAttilio Rao 	struct fuse_access_param facp;
15795fe58019SAttilio Rao 
15805fe58019SAttilio Rao 	int err = 0;
15815fe58019SAttilio Rao 	enum vtype vtyp;
15825fe58019SAttilio Rao 	int sizechanged = 0;
15835fe58019SAttilio Rao 	uint64_t newsize = 0;
15845fe58019SAttilio Rao 
15855fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
15865fe58019SAttilio Rao 		return ENXIO;
15875fe58019SAttilio Rao 	}
15885fe58019SAttilio Rao 	fdisp_init(&fdi, sizeof(*fsai));
15895fe58019SAttilio Rao 	fdisp_make_vp(&fdi, FUSE_SETATTR, vp, td, cred);
15905fe58019SAttilio Rao 	fsai = fdi.indata;
15915fe58019SAttilio Rao 	fsai->valid = 0;
15925fe58019SAttilio Rao 
15935fe58019SAttilio Rao 	bzero(&facp, sizeof(facp));
15945fe58019SAttilio Rao 
15955fe58019SAttilio Rao 	facp.xuid = vap->va_uid;
15965fe58019SAttilio Rao 	facp.xgid = vap->va_gid;
15975fe58019SAttilio Rao 
15985fe58019SAttilio Rao 	if (vap->va_uid != (uid_t)VNOVAL) {
15995fe58019SAttilio Rao 		facp.facc_flags |= FACCESS_CHOWN;
16005fe58019SAttilio Rao 		fsai->uid = vap->va_uid;
16015fe58019SAttilio Rao 		fsai->valid |= FATTR_UID;
16025fe58019SAttilio Rao 	}
16035fe58019SAttilio Rao 	if (vap->va_gid != (gid_t)VNOVAL) {
16045fe58019SAttilio Rao 		facp.facc_flags |= FACCESS_CHOWN;
16055fe58019SAttilio Rao 		fsai->gid = vap->va_gid;
16065fe58019SAttilio Rao 		fsai->valid |= FATTR_GID;
16075fe58019SAttilio Rao 	}
16085fe58019SAttilio Rao 	if (vap->va_size != VNOVAL) {
16095fe58019SAttilio Rao 
16105fe58019SAttilio Rao 		struct fuse_filehandle *fufh = NULL;
16115fe58019SAttilio Rao 
16125fe58019SAttilio Rao 		/*Truncate to a new value. */
16135fe58019SAttilio Rao 		    fsai->size = vap->va_size;
16145fe58019SAttilio Rao 		sizechanged = 1;
16155fe58019SAttilio Rao 		newsize = vap->va_size;
16165fe58019SAttilio Rao 		fsai->valid |= FATTR_SIZE;
16175fe58019SAttilio Rao 
16185fe58019SAttilio Rao 		fuse_filehandle_getrw(vp, FUFH_WRONLY, &fufh);
16195fe58019SAttilio Rao 		if (fufh) {
16205fe58019SAttilio Rao 			fsai->fh = fufh->fh_id;
16215fe58019SAttilio Rao 			fsai->valid |= FATTR_FH;
16225fe58019SAttilio Rao 		}
16235fe58019SAttilio Rao 	}
16245fe58019SAttilio Rao 	if (vap->va_atime.tv_sec != VNOVAL) {
16255fe58019SAttilio Rao 		fsai->atime = vap->va_atime.tv_sec;
16265fe58019SAttilio Rao 		fsai->atimensec = vap->va_atime.tv_nsec;
16275fe58019SAttilio Rao 		fsai->valid |= FATTR_ATIME;
16285fe58019SAttilio Rao 	}
16295fe58019SAttilio Rao 	if (vap->va_mtime.tv_sec != VNOVAL) {
16305fe58019SAttilio Rao 		fsai->mtime = vap->va_mtime.tv_sec;
16315fe58019SAttilio Rao 		fsai->mtimensec = vap->va_mtime.tv_nsec;
16325fe58019SAttilio Rao 		fsai->valid |= FATTR_MTIME;
16335fe58019SAttilio Rao 	}
16345fe58019SAttilio Rao 	if (vap->va_mode != (mode_t)VNOVAL) {
16355fe58019SAttilio Rao 		fsai->mode = vap->va_mode & ALLPERMS;
16365fe58019SAttilio Rao 		fsai->valid |= FATTR_MODE;
16375fe58019SAttilio Rao 	}
16385fe58019SAttilio Rao 	if (!fsai->valid) {
16395fe58019SAttilio Rao 		goto out;
16405fe58019SAttilio Rao 	}
16415fe58019SAttilio Rao 	vtyp = vnode_vtype(vp);
16425fe58019SAttilio Rao 
16435fe58019SAttilio Rao 	if (fsai->valid & FATTR_SIZE && vtyp == VDIR) {
16445fe58019SAttilio Rao 		err = EISDIR;
16455fe58019SAttilio Rao 		goto out;
16465fe58019SAttilio Rao 	}
16475fe58019SAttilio Rao 	if (vfs_isrdonly(vnode_mount(vp)) && (fsai->valid & ~FATTR_SIZE || vtyp == VREG)) {
16485fe58019SAttilio Rao 		err = EROFS;
16495fe58019SAttilio Rao 		goto out;
16505fe58019SAttilio Rao 	}
16515fe58019SAttilio Rao 
1652d9454fabSAttilio Rao 	if ((err = fdisp_wait_answ(&fdi)))
16535fe58019SAttilio Rao 		goto out;
16545fe58019SAttilio Rao 	vtyp = IFTOVT(((struct fuse_attr_out *)fdi.answ)->attr.mode);
16555fe58019SAttilio Rao 
16565fe58019SAttilio Rao 	if (vnode_vtype(vp) != vtyp) {
16575fe58019SAttilio Rao 		if (vnode_vtype(vp) == VNON && vtyp != VNON) {
1658cf169498SAlan Somers 			SDT_PROBE2(fuse, , vnops, trace, 1, "FUSE: Dang! "
1659cf169498SAlan Somers 				"vnode_vtype is VNON and vtype isn't.");
16605fe58019SAttilio Rao 		} else {
16615fe58019SAttilio Rao 			/*
16625fe58019SAttilio Rao 	                 * STALE vnode, ditch
16635fe58019SAttilio Rao 	                 *
1664cf169498SAlan Somers 			 * The vnode has changed its type "behind our back".
1665cf169498SAlan Somers 			 * There's nothing really we can do, so let us just
1666cf169498SAlan Somers 			 * force an internal revocation and tell the caller to
1667cf169498SAlan Somers 			 * try again, if interested.
16685fe58019SAttilio Rao 	                 */
16695fe58019SAttilio Rao 			fuse_internal_vnode_disappear(vp);
16705fe58019SAttilio Rao 			err = EAGAIN;
16715fe58019SAttilio Rao 		}
16725fe58019SAttilio Rao 	}
16737e4844f7SAlan Somers 	if (err == 0) {
16747e4844f7SAlan Somers 		struct fuse_attr_out *fao = (struct fuse_attr_out*)fdi.answ;
16757e4844f7SAlan Somers 		fuse_internal_cache_attrs(vp, &fao->attr, fao->attr_valid,
16767e4844f7SAlan Somers 			fao->attr_valid_nsec, NULL);
16777e4844f7SAlan Somers 	}
1678194e691aSConrad Meyer 
16795fe58019SAttilio Rao out:
16805fe58019SAttilio Rao 	fdisp_destroy(&fdi);
16815fe58019SAttilio Rao 	if (!err && sizechanged) {
16825fe58019SAttilio Rao 		fuse_vnode_setsize(vp, cred, newsize);
16835fe58019SAttilio Rao 		VTOFUD(vp)->flag &= ~FN_SIZECHANGE;
16845fe58019SAttilio Rao 	}
16855fe58019SAttilio Rao 	return err;
16865fe58019SAttilio Rao }
16875fe58019SAttilio Rao 
16885fe58019SAttilio Rao /*
16895fe58019SAttilio Rao     struct vnop_strategy_args {
16905fe58019SAttilio Rao 	struct vnode *a_vp;
16915fe58019SAttilio Rao 	struct buf *a_bp;
16925fe58019SAttilio Rao     };
16935fe58019SAttilio Rao */
16945fe58019SAttilio Rao static int
16955fe58019SAttilio Rao fuse_vnop_strategy(struct vop_strategy_args *ap)
16965fe58019SAttilio Rao {
16975fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
16985fe58019SAttilio Rao 	struct buf *bp = ap->a_bp;
16995fe58019SAttilio Rao 
17005fe58019SAttilio Rao 	if (!vp || fuse_isdeadfs(vp)) {
17015fe58019SAttilio Rao 		bp->b_ioflags |= BIO_ERROR;
17025fe58019SAttilio Rao 		bp->b_error = ENXIO;
17035fe58019SAttilio Rao 		bufdone(bp);
170498852a32SAlan Somers 		return 0;
17055fe58019SAttilio Rao 	}
1706f203d173SAlan Somers 	if (bp->b_iocmd == BIO_WRITE) {
1707f203d173SAlan Somers 		int err;
17085fe58019SAttilio Rao 
1709f203d173SAlan Somers 		err = fuse_vnode_refreshsize(vp, NOCRED);
1710f203d173SAlan Somers 		if (err) {
1711f203d173SAlan Somers 			bp->b_ioflags |= BIO_ERROR;
1712f203d173SAlan Somers 			bp->b_error = err;
171398852a32SAlan Somers 			bufdone(bp);
1714f203d173SAlan Somers 			return 0;
1715f203d173SAlan Somers 		}
1716f203d173SAlan Somers 	}
17175fe58019SAttilio Rao 
17185fe58019SAttilio Rao 	/*
1719f203d173SAlan Somers 	 * VOP_STRATEGY always returns zero and signals error via bp->b_ioflags.
1720f203d173SAlan Somers 	 * fuse_io_strategy sets bp's error fields
17215fe58019SAttilio Rao 	 */
1722f203d173SAlan Somers 	(void)fuse_io_strategy(vp, bp);
1723f203d173SAlan Somers 
17245fe58019SAttilio Rao 	return 0;
17255fe58019SAttilio Rao }
17265fe58019SAttilio Rao 
17275fe58019SAttilio Rao 
17285fe58019SAttilio Rao /*
17295fe58019SAttilio Rao     struct vnop_symlink_args {
17305fe58019SAttilio Rao 	struct vnode *a_dvp;
17315fe58019SAttilio Rao 	struct vnode **a_vpp;
17325fe58019SAttilio Rao 	struct componentname *a_cnp;
17335fe58019SAttilio Rao 	struct vattr *a_vap;
17345fe58019SAttilio Rao 	char *a_target;
17355fe58019SAttilio Rao     };
17365fe58019SAttilio Rao */
17375fe58019SAttilio Rao static int
17385fe58019SAttilio Rao fuse_vnop_symlink(struct vop_symlink_args *ap)
17395fe58019SAttilio Rao {
17405fe58019SAttilio Rao 	struct vnode *dvp = ap->a_dvp;
17415fe58019SAttilio Rao 	struct vnode **vpp = ap->a_vpp;
17425fe58019SAttilio Rao 	struct componentname *cnp = ap->a_cnp;
17431493c2eeSBrooks Davis 	const char *target = ap->a_target;
17445fe58019SAttilio Rao 
17455fe58019SAttilio Rao 	struct fuse_dispatcher fdi;
17465fe58019SAttilio Rao 
17475fe58019SAttilio Rao 	int err;
17485fe58019SAttilio Rao 	size_t len;
17495fe58019SAttilio Rao 
17505fe58019SAttilio Rao 	if (fuse_isdeadfs(dvp)) {
17515fe58019SAttilio Rao 		return ENXIO;
17525fe58019SAttilio Rao 	}
17535fe58019SAttilio Rao 	/*
17545fe58019SAttilio Rao 	 * Unlike the other creator type calls, here we have to create a message
17555fe58019SAttilio Rao 	 * where the name of the new entry comes first, and the data describing
17565fe58019SAttilio Rao 	 * the entry comes second.
17575fe58019SAttilio Rao 	 * Hence we can't rely on our handy fuse_internal_newentry() routine,
17585fe58019SAttilio Rao 	 * but put together the message manually and just call the core part.
17595fe58019SAttilio Rao 	 */
17605fe58019SAttilio Rao 
17615fe58019SAttilio Rao 	len = strlen(target) + 1;
17625fe58019SAttilio Rao 	fdisp_init(&fdi, len + cnp->cn_namelen + 1);
17635fe58019SAttilio Rao 	fdisp_make_vp(&fdi, FUSE_SYMLINK, dvp, curthread, NULL);
17645fe58019SAttilio Rao 
17655fe58019SAttilio Rao 	memcpy(fdi.indata, cnp->cn_nameptr, cnp->cn_namelen);
17665fe58019SAttilio Rao 	((char *)fdi.indata)[cnp->cn_namelen] = '\0';
17675fe58019SAttilio Rao 	memcpy((char *)fdi.indata + cnp->cn_namelen + 1, target, len);
17685fe58019SAttilio Rao 
17695fe58019SAttilio Rao 	err = fuse_internal_newentry_core(dvp, vpp, cnp, VLNK, &fdi);
17705fe58019SAttilio Rao 	fdisp_destroy(&fdi);
17715fe58019SAttilio Rao 	return err;
17725fe58019SAttilio Rao }
17735fe58019SAttilio Rao 
17745fe58019SAttilio Rao /*
17755fe58019SAttilio Rao     struct vnop_write_args {
17765fe58019SAttilio Rao 	struct vnode *a_vp;
17775fe58019SAttilio Rao 	struct uio *a_uio;
17785fe58019SAttilio Rao 	int  a_ioflag;
17795fe58019SAttilio Rao 	struct ucred *a_cred;
17805fe58019SAttilio Rao     };
17815fe58019SAttilio Rao */
17825fe58019SAttilio Rao static int
17835fe58019SAttilio Rao fuse_vnop_write(struct vop_write_args *ap)
17845fe58019SAttilio Rao {
17855fe58019SAttilio Rao 	struct vnode *vp = ap->a_vp;
17865fe58019SAttilio Rao 	struct uio *uio = ap->a_uio;
17875fe58019SAttilio Rao 	int ioflag = ap->a_ioflag;
17885fe58019SAttilio Rao 	struct ucred *cred = ap->a_cred;
1789f203d173SAlan Somers 	int err;
17905fe58019SAttilio Rao 
17915fe58019SAttilio Rao 	if (fuse_isdeadfs(vp)) {
17925fe58019SAttilio Rao 		return ENXIO;
17935fe58019SAttilio Rao 	}
1794f203d173SAlan Somers 	err = fuse_vnode_refreshsize(vp, cred);
1795f203d173SAlan Somers 	if (err)
1796f203d173SAlan Somers 		return err;
17975fe58019SAttilio Rao 
1798ead063e0SEdward Tomasz Napierala 	if (VTOFUD(vp)->flag & FN_DIRECTIO) {
1799ead063e0SEdward Tomasz Napierala 		ioflag |= IO_DIRECT;
1800ead063e0SEdward Tomasz Napierala 	}
1801ead063e0SEdward Tomasz Napierala 
18025fe58019SAttilio Rao 	return fuse_io_dispatch(vp, uio, ioflag, cred);
18035fe58019SAttilio Rao }
18045fe58019SAttilio Rao 
1805cf169498SAlan Somers SDT_PROBE_DEFINE1(fuse, , vnops, vnop_getpages_error, "int");
18065fe58019SAttilio Rao /*
18075fe58019SAttilio Rao     struct vnop_getpages_args {
18085fe58019SAttilio Rao 	struct vnode *a_vp;
18095fe58019SAttilio Rao 	vm_page_t *a_m;
18105fe58019SAttilio Rao 	int a_count;
18115fe58019SAttilio Rao 	int a_reqpage;
18125fe58019SAttilio Rao     };
18135fe58019SAttilio Rao */
18145fe58019SAttilio Rao static int
18155fe58019SAttilio Rao fuse_vnop_getpages(struct vop_getpages_args *ap)
18165fe58019SAttilio Rao {
18175fe58019SAttilio Rao 	int i, error, nextoff, size, toff, count, npages;
18185fe58019SAttilio Rao 	struct uio uio;
18195fe58019SAttilio Rao 	struct iovec iov;
18205fe58019SAttilio Rao 	vm_offset_t kva;
18215fe58019SAttilio Rao 	struct buf *bp;
18225fe58019SAttilio Rao 	struct vnode *vp;
18235fe58019SAttilio Rao 	struct thread *td;
18245fe58019SAttilio Rao 	struct ucred *cred;
18255fe58019SAttilio Rao 	vm_page_t *pages;
18265fe58019SAttilio Rao 
18275fe58019SAttilio Rao 	vp = ap->a_vp;
18285fe58019SAttilio Rao 	KASSERT(vp->v_object, ("objectless vp passed to getpages"));
18295fe58019SAttilio Rao 	td = curthread;			/* XXX */
18305fe58019SAttilio Rao 	cred = curthread->td_ucred;	/* XXX */
18315fe58019SAttilio Rao 	pages = ap->a_m;
1832f17f88d3SGleb Smirnoff 	npages = ap->a_count;
18335fe58019SAttilio Rao 
18345fe58019SAttilio Rao 	if (!fsess_opt_mmap(vnode_mount(vp))) {
1835cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1,
1836cf169498SAlan Somers 			"called on non-cacheable vnode??\n");
18375fe58019SAttilio Rao 		return (VM_PAGER_ERROR);
18385fe58019SAttilio Rao 	}
18395fe58019SAttilio Rao 
18405fe58019SAttilio Rao 	/*
1841b0cd2017SGleb Smirnoff 	 * If the last page is partially valid, just return it and allow
1842b0cd2017SGleb Smirnoff 	 * the pager to zero-out the blanks.  Partially valid pages can
1843b0cd2017SGleb Smirnoff 	 * only occur at the file EOF.
1844b0cd2017SGleb Smirnoff 	 *
1845b0cd2017SGleb Smirnoff 	 * XXXGL: is that true for FUSE, which is a local filesystem,
1846b0cd2017SGleb Smirnoff 	 * but still somewhat disconnected from the kernel?
18475fe58019SAttilio Rao 	 */
184889f6b863SAttilio Rao 	VM_OBJECT_WLOCK(vp->v_object);
1849f17f88d3SGleb Smirnoff 	if (pages[npages - 1]->valid != 0 && --npages == 0)
1850f17f88d3SGleb Smirnoff 		goto out;
185189f6b863SAttilio Rao 	VM_OBJECT_WUNLOCK(vp->v_object);
18525fe58019SAttilio Rao 
18535fe58019SAttilio Rao 	/*
18545fe58019SAttilio Rao 	 * We use only the kva address for the buffer, but this is extremely
1855b3a15dddSPedro F. Giffuni 	 * convenient and fast.
18565fe58019SAttilio Rao 	 */
1857756a5412SGleb Smirnoff 	bp = uma_zalloc(fuse_pbuf_zone, M_WAITOK);
18585fe58019SAttilio Rao 
18595fe58019SAttilio Rao 	kva = (vm_offset_t)bp->b_data;
18605fe58019SAttilio Rao 	pmap_qenter(kva, pages, npages);
186183c9dea1SGleb Smirnoff 	VM_CNT_INC(v_vnodein);
186283c9dea1SGleb Smirnoff 	VM_CNT_ADD(v_vnodepgsin, npages);
18635fe58019SAttilio Rao 
1864f17f88d3SGleb Smirnoff 	count = npages << PAGE_SHIFT;
18655fe58019SAttilio Rao 	iov.iov_base = (caddr_t)kva;
18665fe58019SAttilio Rao 	iov.iov_len = count;
18675fe58019SAttilio Rao 	uio.uio_iov = &iov;
18685fe58019SAttilio Rao 	uio.uio_iovcnt = 1;
18695fe58019SAttilio Rao 	uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
18705fe58019SAttilio Rao 	uio.uio_resid = count;
18715fe58019SAttilio Rao 	uio.uio_segflg = UIO_SYSSPACE;
18725fe58019SAttilio Rao 	uio.uio_rw = UIO_READ;
18735fe58019SAttilio Rao 	uio.uio_td = td;
18745fe58019SAttilio Rao 
18755fe58019SAttilio Rao 	error = fuse_io_dispatch(vp, &uio, IO_DIRECT, cred);
18765fe58019SAttilio Rao 	pmap_qremove(kva, npages);
18775fe58019SAttilio Rao 
1878756a5412SGleb Smirnoff 	uma_zfree(fuse_pbuf_zone, bp);
18795fe58019SAttilio Rao 
18805fe58019SAttilio Rao 	if (error && (uio.uio_resid == count)) {
1881cf169498SAlan Somers 		SDT_PROBE1(fuse, , vnops, vnop_getpages_error, error);
18825fe58019SAttilio Rao 		return VM_PAGER_ERROR;
18835fe58019SAttilio Rao 	}
18845fe58019SAttilio Rao 	/*
18855fe58019SAttilio Rao 	 * Calculate the number of bytes read and validate only that number
18865fe58019SAttilio Rao 	 * of bytes.  Note that due to pending writes, size may be 0.  This
18875fe58019SAttilio Rao 	 * does not mean that the remaining data is invalid!
18885fe58019SAttilio Rao 	 */
18895fe58019SAttilio Rao 
18905fe58019SAttilio Rao 	size = count - uio.uio_resid;
189189f6b863SAttilio Rao 	VM_OBJECT_WLOCK(vp->v_object);
18925fe58019SAttilio Rao 	fuse_vm_page_lock_queues();
18935fe58019SAttilio Rao 	for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
18945fe58019SAttilio Rao 		vm_page_t m;
18955fe58019SAttilio Rao 
18965fe58019SAttilio Rao 		nextoff = toff + PAGE_SIZE;
18975fe58019SAttilio Rao 		m = pages[i];
18985fe58019SAttilio Rao 
18995fe58019SAttilio Rao 		if (nextoff <= size) {
19005fe58019SAttilio Rao 			/*
19015fe58019SAttilio Rao 			 * Read operation filled an entire page
19025fe58019SAttilio Rao 			 */
19035fe58019SAttilio Rao 			m->valid = VM_PAGE_BITS_ALL;
19045fe58019SAttilio Rao 			KASSERT(m->dirty == 0,
19055fe58019SAttilio Rao 			    ("fuse_getpages: page %p is dirty", m));
19065fe58019SAttilio Rao 		} else if (size > toff) {
19075fe58019SAttilio Rao 			/*
19085fe58019SAttilio Rao 			 * Read operation filled a partial page.
19095fe58019SAttilio Rao 			 */
19105fe58019SAttilio Rao 			m->valid = 0;
19115fe58019SAttilio Rao 			vm_page_set_valid_range(m, 0, size - toff);
19125fe58019SAttilio Rao 			KASSERT(m->dirty == 0,
19135fe58019SAttilio Rao 			    ("fuse_getpages: page %p is dirty", m));
19145fe58019SAttilio Rao 		} else {
19155fe58019SAttilio Rao 			/*
1916b3a15dddSPedro F. Giffuni 			 * Read operation was short.  If no error occurred
19175fe58019SAttilio Rao 			 * we may have hit a zero-fill section.   We simply
19185fe58019SAttilio Rao 			 * leave valid set to 0.
19195fe58019SAttilio Rao 			 */
19205fe58019SAttilio Rao 			;
19215fe58019SAttilio Rao 		}
19225fe58019SAttilio Rao 	}
19235fe58019SAttilio Rao 	fuse_vm_page_unlock_queues();
1924f17f88d3SGleb Smirnoff out:
192589f6b863SAttilio Rao 	VM_OBJECT_WUNLOCK(vp->v_object);
1926f17f88d3SGleb Smirnoff 	if (ap->a_rbehind)
1927f17f88d3SGleb Smirnoff 		*ap->a_rbehind = 0;
1928f17f88d3SGleb Smirnoff 	if (ap->a_rahead)
1929f17f88d3SGleb Smirnoff 		*ap->a_rahead = 0;
1930f17f88d3SGleb Smirnoff 	return (VM_PAGER_OK);
19315fe58019SAttilio Rao }
19325fe58019SAttilio Rao 
19335fe58019SAttilio Rao /*
19345fe58019SAttilio Rao     struct vnop_putpages_args {
19355fe58019SAttilio Rao 	struct vnode *a_vp;
19365fe58019SAttilio Rao 	vm_page_t *a_m;
19375fe58019SAttilio Rao 	int a_count;
19385fe58019SAttilio Rao 	int a_sync;
19395fe58019SAttilio Rao 	int *a_rtvals;
19405fe58019SAttilio Rao 	vm_ooffset_t a_offset;
19415fe58019SAttilio Rao     };
19425fe58019SAttilio Rao */
19435fe58019SAttilio Rao static int
19445fe58019SAttilio Rao fuse_vnop_putpages(struct vop_putpages_args *ap)
19455fe58019SAttilio Rao {
19465fe58019SAttilio Rao 	struct uio uio;
19475fe58019SAttilio Rao 	struct iovec iov;
19485fe58019SAttilio Rao 	vm_offset_t kva;
19495fe58019SAttilio Rao 	struct buf *bp;
19505fe58019SAttilio Rao 	int i, error, npages, count;
19515fe58019SAttilio Rao 	off_t offset;
19525fe58019SAttilio Rao 	int *rtvals;
19535fe58019SAttilio Rao 	struct vnode *vp;
19545fe58019SAttilio Rao 	struct thread *td;
19555fe58019SAttilio Rao 	struct ucred *cred;
19565fe58019SAttilio Rao 	vm_page_t *pages;
19575fe58019SAttilio Rao 	vm_ooffset_t fsize;
19585fe58019SAttilio Rao 
19595fe58019SAttilio Rao 	vp = ap->a_vp;
19605fe58019SAttilio Rao 	KASSERT(vp->v_object, ("objectless vp passed to putpages"));
19615fe58019SAttilio Rao 	fsize = vp->v_object->un_pager.vnp.vnp_size;
19625fe58019SAttilio Rao 	td = curthread;			/* XXX */
19635fe58019SAttilio Rao 	cred = curthread->td_ucred;	/* XXX */
19645fe58019SAttilio Rao 	pages = ap->a_m;
19655fe58019SAttilio Rao 	count = ap->a_count;
19665fe58019SAttilio Rao 	rtvals = ap->a_rtvals;
19675fe58019SAttilio Rao 	npages = btoc(count);
19685fe58019SAttilio Rao 	offset = IDX_TO_OFF(pages[0]->pindex);
19695fe58019SAttilio Rao 
19705fe58019SAttilio Rao 	if (!fsess_opt_mmap(vnode_mount(vp))) {
1971cf169498SAlan Somers 		SDT_PROBE2(fuse, , vnops, trace, 1,
1972cf169498SAlan Somers 			"called on non-cacheable vnode??\n");
19735fe58019SAttilio Rao 	}
19745fe58019SAttilio Rao 	for (i = 0; i < npages; i++)
19755fe58019SAttilio Rao 		rtvals[i] = VM_PAGER_AGAIN;
19765fe58019SAttilio Rao 
19775fe58019SAttilio Rao 	/*
19785fe58019SAttilio Rao 	 * When putting pages, do not extend file past EOF.
19795fe58019SAttilio Rao 	 */
19805fe58019SAttilio Rao 
19815fe58019SAttilio Rao 	if (offset + count > fsize) {
19825fe58019SAttilio Rao 		count = fsize - offset;
19835fe58019SAttilio Rao 		if (count < 0)
19845fe58019SAttilio Rao 			count = 0;
19855fe58019SAttilio Rao 	}
19865fe58019SAttilio Rao 	/*
19875fe58019SAttilio Rao 	 * We use only the kva address for the buffer, but this is extremely
1988b3a15dddSPedro F. Giffuni 	 * convenient and fast.
19895fe58019SAttilio Rao 	 */
1990756a5412SGleb Smirnoff 	bp = uma_zalloc(fuse_pbuf_zone, M_WAITOK);
19915fe58019SAttilio Rao 
19925fe58019SAttilio Rao 	kva = (vm_offset_t)bp->b_data;
19935fe58019SAttilio Rao 	pmap_qenter(kva, pages, npages);
199483c9dea1SGleb Smirnoff 	VM_CNT_INC(v_vnodeout);
199583c9dea1SGleb Smirnoff 	VM_CNT_ADD(v_vnodepgsout, count);
19965fe58019SAttilio Rao 
19975fe58019SAttilio Rao 	iov.iov_base = (caddr_t)kva;
19985fe58019SAttilio Rao 	iov.iov_len = count;
19995fe58019SAttilio Rao 	uio.uio_iov = &iov;
20005fe58019SAttilio Rao 	uio.uio_iovcnt = 1;
20015fe58019SAttilio Rao 	uio.uio_offset = offset;
20025fe58019SAttilio Rao 	uio.uio_resid = count;
20035fe58019SAttilio Rao 	uio.uio_segflg = UIO_SYSSPACE;
20045fe58019SAttilio Rao 	uio.uio_rw = UIO_WRITE;
20055fe58019SAttilio Rao 	uio.uio_td = td;
20065fe58019SAttilio Rao 
20075fe58019SAttilio Rao 	error = fuse_io_dispatch(vp, &uio, IO_DIRECT, cred);
20085fe58019SAttilio Rao 
20095fe58019SAttilio Rao 	pmap_qremove(kva, npages);
2010756a5412SGleb Smirnoff 	uma_zfree(fuse_pbuf_zone, bp);
20115fe58019SAttilio Rao 
20125fe58019SAttilio Rao 	if (!error) {
20135fe58019SAttilio Rao 		int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
20145fe58019SAttilio Rao 
20155fe58019SAttilio Rao 		for (i = 0; i < nwritten; i++) {
20165fe58019SAttilio Rao 			rtvals[i] = VM_PAGER_OK;
201789f6b863SAttilio Rao 			VM_OBJECT_WLOCK(pages[i]->object);
20185fe58019SAttilio Rao 			vm_page_undirty(pages[i]);
201989f6b863SAttilio Rao 			VM_OBJECT_WUNLOCK(pages[i]->object);
20205fe58019SAttilio Rao 		}
20215fe58019SAttilio Rao 	}
20225fe58019SAttilio Rao 	return rtvals[0];
20235fe58019SAttilio Rao }
20245fe58019SAttilio Rao 
202504660064SFedor Uporov static const char extattr_namespace_separator = '.';
202604660064SFedor Uporov 
202704660064SFedor Uporov /*
202804660064SFedor Uporov     struct vop_getextattr_args {
202904660064SFedor Uporov 	struct vop_generic_args a_gen;
203004660064SFedor Uporov 	struct vnode *a_vp;
203104660064SFedor Uporov 	int a_attrnamespace;
203204660064SFedor Uporov 	const char *a_name;
203304660064SFedor Uporov 	struct uio *a_uio;
203404660064SFedor Uporov 	size_t *a_size;
203504660064SFedor Uporov 	struct ucred *a_cred;
203604660064SFedor Uporov 	struct thread *a_td;
203704660064SFedor Uporov     };
203804660064SFedor Uporov */
203904660064SFedor Uporov static int
204004660064SFedor Uporov fuse_vnop_getextattr(struct vop_getextattr_args *ap)
204104660064SFedor Uporov {
204204660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
204304660064SFedor Uporov 	struct uio *uio = ap->a_uio;
204428f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
204504660064SFedor Uporov 	struct fuse_getxattr_in *get_xattr_in;
204604660064SFedor Uporov 	struct fuse_getxattr_out *get_xattr_out;
204704660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
204804660064SFedor Uporov 	struct thread *td = ap->a_td;
204904660064SFedor Uporov 	struct ucred *cred = ap->a_cred;
205028f4f623SFedor Uporov 	char *prefix;
205128f4f623SFedor Uporov 	char *attr_str;
205228f4f623SFedor Uporov 	size_t len;
205328f4f623SFedor Uporov 	int err;
205404660064SFedor Uporov 
205504660064SFedor Uporov 	if (fuse_isdeadfs(vp))
205628f4f623SFedor Uporov 		return (ENXIO);
205704660064SFedor Uporov 
205804660064SFedor Uporov 	/* Default to looking for user attributes. */
205904660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
206004660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
206104660064SFedor Uporov 	else
206204660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
206304660064SFedor Uporov 
206404660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) +
206504660064SFedor Uporov 	    strlen(ap->a_name) + 1;
206604660064SFedor Uporov 
206704660064SFedor Uporov 	fdisp_init(&fdi, len + sizeof(*get_xattr_in));
206804660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_GETXATTR, vp, td, cred);
206904660064SFedor Uporov 
207004660064SFedor Uporov 	get_xattr_in = fdi.indata;
207104660064SFedor Uporov 	/*
207204660064SFedor Uporov 	 * Check to see whether we're querying the available size or
207304660064SFedor Uporov 	 * issuing the actual request.  If we pass in 0, we get back struct
207404660064SFedor Uporov 	 * fuse_getxattr_out.  If we pass in a non-zero size, we get back
207504660064SFedor Uporov 	 * that much data, without the struct fuse_getxattr_out header.
207604660064SFedor Uporov 	 */
2077493b4a8cSFedor Uporov 	if (uio == NULL)
207804660064SFedor Uporov 		get_xattr_in->size = 0;
207904660064SFedor Uporov 	else
208004660064SFedor Uporov 		get_xattr_in->size = uio->uio_resid;
208104660064SFedor Uporov 
208204660064SFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*get_xattr_in);
208304660064SFedor Uporov 	snprintf(attr_str, len, "%s%c%s", prefix, extattr_namespace_separator,
208404660064SFedor Uporov 	    ap->a_name);
208504660064SFedor Uporov 
208604660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
208704660064SFedor Uporov 	if (err != 0) {
208804660064SFedor Uporov 		if (err == ENOSYS)
208904660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_GETXATTR);
209004660064SFedor Uporov 		goto out;
209104660064SFedor Uporov 	}
209204660064SFedor Uporov 
209304660064SFedor Uporov 	get_xattr_out = fdi.answ;
209404660064SFedor Uporov 
2095493b4a8cSFedor Uporov 	if (ap->a_size != NULL)
209604660064SFedor Uporov 		*ap->a_size = get_xattr_out->size;
2097493b4a8cSFedor Uporov 
2098493b4a8cSFedor Uporov 	if (uio != NULL)
209904660064SFedor Uporov 		err = uiomove(fdi.answ, fdi.iosize, uio);
210004660064SFedor Uporov 
210104660064SFedor Uporov out:
210204660064SFedor Uporov 	fdisp_destroy(&fdi);
210304660064SFedor Uporov 	return (err);
210404660064SFedor Uporov }
210504660064SFedor Uporov 
210604660064SFedor Uporov /*
210704660064SFedor Uporov     struct vop_setextattr_args {
210804660064SFedor Uporov 	struct vop_generic_args a_gen;
210904660064SFedor Uporov 	struct vnode *a_vp;
211004660064SFedor Uporov 	int a_attrnamespace;
211104660064SFedor Uporov 	const char *a_name;
211204660064SFedor Uporov 	struct uio *a_uio;
211304660064SFedor Uporov 	struct ucred *a_cred;
211404660064SFedor Uporov 	struct thread *a_td;
211504660064SFedor Uporov     };
211604660064SFedor Uporov */
211704660064SFedor Uporov static int
211804660064SFedor Uporov fuse_vnop_setextattr(struct vop_setextattr_args *ap)
211904660064SFedor Uporov {
212004660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
212104660064SFedor Uporov 	struct uio *uio = ap->a_uio;
212228f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
212304660064SFedor Uporov 	struct fuse_setxattr_in *set_xattr_in;
212404660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
212528f4f623SFedor Uporov 	struct thread *td = ap->a_td;
212628f4f623SFedor Uporov 	struct ucred *cred = ap->a_cred;
212704660064SFedor Uporov 	char *prefix;
212804660064SFedor Uporov 	size_t len;
212904660064SFedor Uporov 	char *attr_str;
213028f4f623SFedor Uporov 	int err;
213104660064SFedor Uporov 
213204660064SFedor Uporov 	if (fuse_isdeadfs(vp))
213328f4f623SFedor Uporov 		return (ENXIO);
213404660064SFedor Uporov 
213504660064SFedor Uporov 	/* Default to looking for user attributes. */
213604660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
213704660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
213804660064SFedor Uporov 	else
213904660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
214004660064SFedor Uporov 
214104660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) +
214204660064SFedor Uporov 	    strlen(ap->a_name) + 1;
214304660064SFedor Uporov 
214404660064SFedor Uporov 	fdisp_init(&fdi, len + sizeof(*set_xattr_in) + uio->uio_resid);
214504660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_SETXATTR, vp, td, cred);
214604660064SFedor Uporov 
214704660064SFedor Uporov 	set_xattr_in = fdi.indata;
214804660064SFedor Uporov 	set_xattr_in->size = uio->uio_resid;
214904660064SFedor Uporov 
215004660064SFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*set_xattr_in);
215104660064SFedor Uporov 	snprintf(attr_str, len, "%s%c%s", prefix, extattr_namespace_separator,
215204660064SFedor Uporov 	    ap->a_name);
215304660064SFedor Uporov 
215404660064SFedor Uporov 	err = uiomove((char *)fdi.indata + sizeof(*set_xattr_in) + len,
215504660064SFedor Uporov 	    uio->uio_resid, uio);
215604660064SFedor Uporov 	if (err != 0) {
215704660064SFedor Uporov 		goto out;
215804660064SFedor Uporov 	}
215904660064SFedor Uporov 
216004660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
216104660064SFedor Uporov 
216204660064SFedor Uporov 	if (err != 0) {
216304660064SFedor Uporov 		if (err == ENOSYS)
216404660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_SETXATTR);
216504660064SFedor Uporov 		goto out;
216604660064SFedor Uporov 	}
216704660064SFedor Uporov 
216804660064SFedor Uporov out:
216904660064SFedor Uporov 	fdisp_destroy(&fdi);
217004660064SFedor Uporov 	return (err);
217104660064SFedor Uporov }
217204660064SFedor Uporov 
217304660064SFedor Uporov /*
217404660064SFedor Uporov  * The Linux / FUSE extended attribute list is simply a collection of
217504660064SFedor Uporov  * NUL-terminated strings.  The FreeBSD extended attribute list is a single
217604660064SFedor Uporov  * byte length followed by a non-NUL terminated string.  So, this allows
217704660064SFedor Uporov  * conversion of the Linux / FUSE format to the FreeBSD format in place.
217804660064SFedor Uporov  * Linux attribute names are reported with the namespace as a prefix (e.g.
217904660064SFedor Uporov  * "user.attribute_name"), but in FreeBSD they are reported without the
218004660064SFedor Uporov  * namespace prefix (e.g. "attribute_name").  So, we're going from:
218104660064SFedor Uporov  *
218204660064SFedor Uporov  * user.attr_name1\0user.attr_name2\0
218304660064SFedor Uporov  *
218404660064SFedor Uporov  * to:
218504660064SFedor Uporov  *
218604660064SFedor Uporov  * <num>attr_name1<num>attr_name2
218704660064SFedor Uporov  *
218804660064SFedor Uporov  * Where "<num>" is a single byte number of characters in the attribute name.
218904660064SFedor Uporov  *
219004660064SFedor Uporov  * Args:
219104660064SFedor Uporov  * prefix - exattr namespace prefix string
219204660064SFedor Uporov  * list, list_len - input list with namespace prefixes
219304660064SFedor Uporov  * bsd_list, bsd_list_len - output list compatible with bsd vfs
219404660064SFedor Uporov  */
219504660064SFedor Uporov static int
219604660064SFedor Uporov fuse_xattrlist_convert(char *prefix, const char *list, int list_len,
219704660064SFedor Uporov     char *bsd_list, int *bsd_list_len)
219804660064SFedor Uporov {
219904660064SFedor Uporov 	int len, pos, dist_to_next, prefix_len;
220004660064SFedor Uporov 
220104660064SFedor Uporov 	pos = 0;
220204660064SFedor Uporov 	*bsd_list_len = 0;
220304660064SFedor Uporov 	prefix_len = strlen(prefix);
220404660064SFedor Uporov 
220504660064SFedor Uporov 	while (pos < list_len && list[pos] != '\0') {
220604660064SFedor Uporov 		dist_to_next = strlen(&list[pos]) + 1;
220704660064SFedor Uporov 		if (bcmp(&list[pos], prefix, prefix_len) == 0 &&
220804660064SFedor Uporov 		    list[pos + prefix_len] == extattr_namespace_separator) {
220904660064SFedor Uporov 			len = dist_to_next -
221004660064SFedor Uporov 			    (prefix_len + sizeof(extattr_namespace_separator)) - 1;
221104660064SFedor Uporov 			if (len >= EXTATTR_MAXNAMELEN)
221204660064SFedor Uporov 				return (ENAMETOOLONG);
221304660064SFedor Uporov 
221404660064SFedor Uporov 			bsd_list[*bsd_list_len] = len;
221504660064SFedor Uporov 			memcpy(&bsd_list[*bsd_list_len + 1],
221604660064SFedor Uporov 			    &list[pos + prefix_len +
221704660064SFedor Uporov 			    sizeof(extattr_namespace_separator)], len);
221804660064SFedor Uporov 
221904660064SFedor Uporov 			*bsd_list_len += len + 1;
222004660064SFedor Uporov 		}
222104660064SFedor Uporov 
222204660064SFedor Uporov 		pos += dist_to_next;
222304660064SFedor Uporov 	}
222404660064SFedor Uporov 
222504660064SFedor Uporov 	return (0);
222604660064SFedor Uporov }
222704660064SFedor Uporov 
222804660064SFedor Uporov /*
222904660064SFedor Uporov     struct vop_listextattr_args {
223004660064SFedor Uporov 	struct vop_generic_args a_gen;
223104660064SFedor Uporov 	struct vnode *a_vp;
223204660064SFedor Uporov 	int a_attrnamespace;
223304660064SFedor Uporov 	struct uio *a_uio;
223404660064SFedor Uporov 	size_t *a_size;
223504660064SFedor Uporov 	struct ucred *a_cred;
223604660064SFedor Uporov 	struct thread *a_td;
223704660064SFedor Uporov     };
223804660064SFedor Uporov */
223904660064SFedor Uporov static int
224004660064SFedor Uporov fuse_vnop_listextattr(struct vop_listextattr_args *ap)
224104660064SFedor Uporov {
224204660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
224304660064SFedor Uporov 	struct uio *uio = ap->a_uio;
224428f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
2245493b4a8cSFedor Uporov 	struct fuse_listxattr_in *list_xattr_in;
2246493b4a8cSFedor Uporov 	struct fuse_listxattr_out *list_xattr_out;
224704660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
224828f4f623SFedor Uporov 	struct thread *td = ap->a_td;
224928f4f623SFedor Uporov 	struct ucred *cred = ap->a_cred;
225004660064SFedor Uporov 	size_t len;
225104660064SFedor Uporov 	char *prefix;
225204660064SFedor Uporov 	char *attr_str;
225304660064SFedor Uporov 	char *bsd_list = NULL;
2254493b4a8cSFedor Uporov 	char *linux_list;
225504660064SFedor Uporov 	int bsd_list_len;
2256493b4a8cSFedor Uporov 	int linux_list_len;
225728f4f623SFedor Uporov 	int err;
225804660064SFedor Uporov 
225904660064SFedor Uporov 	if (fuse_isdeadfs(vp))
226028f4f623SFedor Uporov 		return (ENXIO);
226104660064SFedor Uporov 
226204660064SFedor Uporov 	/*
226304660064SFedor Uporov 	 * Add space for a NUL and the period separator if enabled.
226404660064SFedor Uporov 	 * Default to looking for user attributes.
226504660064SFedor Uporov 	 */
226604660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
226704660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
226804660064SFedor Uporov 	else
226904660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
227004660064SFedor Uporov 
227104660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) + 1;
227204660064SFedor Uporov 
2273493b4a8cSFedor Uporov 	fdisp_init(&fdi, sizeof(*list_xattr_in) + len);
227404660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_LISTXATTR, vp, td, cred);
227504660064SFedor Uporov 
2276493b4a8cSFedor Uporov 	/*
2277493b4a8cSFedor Uporov 	 * Retrieve Linux / FUSE compatible list size.
2278493b4a8cSFedor Uporov 	 */
2279493b4a8cSFedor Uporov 	list_xattr_in = fdi.indata;
2280493b4a8cSFedor Uporov 	list_xattr_in->size = 0;
2281493b4a8cSFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*list_xattr_in);
228204660064SFedor Uporov 	snprintf(attr_str, len, "%s%c", prefix, extattr_namespace_separator);
228304660064SFedor Uporov 
228404660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
228504660064SFedor Uporov 	if (err != 0) {
228604660064SFedor Uporov 		if (err == ENOSYS)
228704660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_LISTXATTR);
228804660064SFedor Uporov 		goto out;
228904660064SFedor Uporov 	}
229004660064SFedor Uporov 
2291493b4a8cSFedor Uporov 	list_xattr_out = fdi.answ;
2292493b4a8cSFedor Uporov 	linux_list_len = list_xattr_out->size;
2293493b4a8cSFedor Uporov 	if (linux_list_len == 0) {
2294493b4a8cSFedor Uporov 		if (ap->a_size != NULL)
2295493b4a8cSFedor Uporov 			*ap->a_size = linux_list_len;
229604660064SFedor Uporov 		goto out;
229704660064SFedor Uporov 	}
229804660064SFedor Uporov 
229904660064SFedor Uporov 	/*
2300493b4a8cSFedor Uporov 	 * Retrieve Linux / FUSE compatible list values.
2301493b4a8cSFedor Uporov 	 */
2302493b4a8cSFedor Uporov 	fdisp_make_vp(&fdi, FUSE_LISTXATTR, vp, td, cred);
2303493b4a8cSFedor Uporov 	list_xattr_in = fdi.indata;
2304493b4a8cSFedor Uporov 	list_xattr_in->size = linux_list_len + sizeof(*list_xattr_out);
2305493b4a8cSFedor Uporov 	attr_str = (char *)fdi.indata + sizeof(*list_xattr_in);
2306493b4a8cSFedor Uporov 	snprintf(attr_str, len, "%s%c", prefix, extattr_namespace_separator);
2307493b4a8cSFedor Uporov 
2308493b4a8cSFedor Uporov 	err = fdisp_wait_answ(&fdi);
2309493b4a8cSFedor Uporov 	if (err != 0)
2310493b4a8cSFedor Uporov 		goto out;
2311493b4a8cSFedor Uporov 
2312493b4a8cSFedor Uporov 	linux_list = fdi.answ;
2313493b4a8cSFedor Uporov 	linux_list_len = fdi.iosize;
2314493b4a8cSFedor Uporov 
2315493b4a8cSFedor Uporov 	/*
2316493b4a8cSFedor Uporov 	 * Retrieve the BSD compatible list values.
231704660064SFedor Uporov 	 * The Linux / FUSE attribute list format isn't the same
231804660064SFedor Uporov 	 * as FreeBSD's format. So we need to transform it into
231904660064SFedor Uporov 	 * FreeBSD's format before giving it to the user.
232004660064SFedor Uporov 	 */
2321493b4a8cSFedor Uporov 	bsd_list = malloc(linux_list_len, M_TEMP, M_WAITOK);
2322493b4a8cSFedor Uporov 	err = fuse_xattrlist_convert(prefix, linux_list, linux_list_len,
232304660064SFedor Uporov 	    bsd_list, &bsd_list_len);
232404660064SFedor Uporov 	if (err != 0)
232504660064SFedor Uporov 		goto out;
232604660064SFedor Uporov 
2327493b4a8cSFedor Uporov 	if (ap->a_size != NULL)
2328493b4a8cSFedor Uporov 		*ap->a_size = bsd_list_len;
2329493b4a8cSFedor Uporov 
2330493b4a8cSFedor Uporov 	if (uio != NULL)
233104660064SFedor Uporov 		err = uiomove(bsd_list, bsd_list_len, uio);
233204660064SFedor Uporov 
233304660064SFedor Uporov out:
233404660064SFedor Uporov 	free(bsd_list, M_TEMP);
233504660064SFedor Uporov 	fdisp_destroy(&fdi);
233604660064SFedor Uporov 	return (err);
233704660064SFedor Uporov }
233804660064SFedor Uporov 
233904660064SFedor Uporov /*
234004660064SFedor Uporov     struct vop_deleteextattr_args {
234104660064SFedor Uporov 	struct vop_generic_args a_gen;
234204660064SFedor Uporov 	struct vnode *a_vp;
234304660064SFedor Uporov 	int a_attrnamespace;
234404660064SFedor Uporov 	const char *a_name;
234504660064SFedor Uporov 	struct ucred *a_cred;
234604660064SFedor Uporov 	struct thread *a_td;
234704660064SFedor Uporov     };
234804660064SFedor Uporov */
234904660064SFedor Uporov static int
235004660064SFedor Uporov fuse_vnop_deleteextattr(struct vop_deleteextattr_args *ap)
235104660064SFedor Uporov {
235204660064SFedor Uporov 	struct vnode *vp = ap->a_vp;
235328f4f623SFedor Uporov 	struct fuse_dispatcher fdi;
235404660064SFedor Uporov 	struct mount *mp = vnode_mount(vp);
235528f4f623SFedor Uporov 	struct thread *td = ap->a_td;
235628f4f623SFedor Uporov 	struct ucred *cred = ap->a_cred;
235704660064SFedor Uporov 	char *prefix;
235804660064SFedor Uporov 	size_t len;
235904660064SFedor Uporov 	char *attr_str;
236004660064SFedor Uporov 	int err;
236104660064SFedor Uporov 
236204660064SFedor Uporov 	if (fuse_isdeadfs(vp))
236328f4f623SFedor Uporov 		return (ENXIO);
236404660064SFedor Uporov 
236504660064SFedor Uporov 	/* Default to looking for user attributes. */
236604660064SFedor Uporov 	if (ap->a_attrnamespace == EXTATTR_NAMESPACE_SYSTEM)
236704660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_SYSTEM_STRING;
236804660064SFedor Uporov 	else
236904660064SFedor Uporov 		prefix = EXTATTR_NAMESPACE_USER_STRING;
237004660064SFedor Uporov 
237104660064SFedor Uporov 	len = strlen(prefix) + sizeof(extattr_namespace_separator) +
237204660064SFedor Uporov 	    strlen(ap->a_name) + 1;
237304660064SFedor Uporov 
237404660064SFedor Uporov 	fdisp_init(&fdi, len);
237504660064SFedor Uporov 	fdisp_make_vp(&fdi, FUSE_REMOVEXATTR, vp, td, cred);
237604660064SFedor Uporov 
237704660064SFedor Uporov 	attr_str = fdi.indata;
237804660064SFedor Uporov 	snprintf(attr_str, len, "%s%c%s", prefix, extattr_namespace_separator,
237904660064SFedor Uporov 	    ap->a_name);
238004660064SFedor Uporov 
238104660064SFedor Uporov 	err = fdisp_wait_answ(&fdi);
238204660064SFedor Uporov 	if (err != 0) {
238304660064SFedor Uporov 		if (err == ENOSYS)
238404660064SFedor Uporov 			fsess_set_notimpl(mp, FUSE_REMOVEXATTR);
238504660064SFedor Uporov 	}
238604660064SFedor Uporov 
238704660064SFedor Uporov 	fdisp_destroy(&fdi);
238804660064SFedor Uporov 	return (err);
238904660064SFedor Uporov }
239004660064SFedor Uporov 
23915fe58019SAttilio Rao /*
23925fe58019SAttilio Rao     struct vnop_print_args {
23935fe58019SAttilio Rao 	struct vnode *a_vp;
23945fe58019SAttilio Rao     };
23955fe58019SAttilio Rao */
23965fe58019SAttilio Rao static int
23975fe58019SAttilio Rao fuse_vnop_print(struct vop_print_args *ap)
23985fe58019SAttilio Rao {
23995fe58019SAttilio Rao 	struct fuse_vnode_data *fvdat = VTOFUD(ap->a_vp);
24005fe58019SAttilio Rao 
24015fe58019SAttilio Rao 	printf("nodeid: %ju, parent nodeid: %ju, nlookup: %ju, flag: %#x\n",
24025fe58019SAttilio Rao 	    (uintmax_t)VTOILLU(ap->a_vp), (uintmax_t)fvdat->parent_nid,
24035fe58019SAttilio Rao 	    (uintmax_t)fvdat->nlookup,
24045fe58019SAttilio Rao 	    fvdat->flag);
24055fe58019SAttilio Rao 
24065fe58019SAttilio Rao 	return 0;
24075fe58019SAttilio Rao }
2408