xref: /netbsd-src/sys/ufs/lfs/ulfs_inode.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: ulfs_inode.c,v 1.21 2017/10/28 00:37:13 pgoyette Exp $	*/
2 /*  from NetBSD: ufs_inode.c,v 1.95 2015/06/13 14:56:45 hannken Exp  */
3 
4 /*
5  * Copyright (c) 1991, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  * (c) UNIX System Laboratories, Inc.
8  * All or some portions of this file are derived from material licensed
9  * to the University of California by American Telephone and Telegraph
10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11  * the permission of UNIX System Laboratories, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)ufs_inode.c	8.9 (Berkeley) 5/14/95
38  */
39 
40 #include <sys/cdefs.h>
41 __KERNEL_RCSID(0, "$NetBSD: ulfs_inode.c,v 1.21 2017/10/28 00:37:13 pgoyette Exp $");
42 
43 #if defined(_KERNEL_OPT)
44 #include "opt_lfs.h"
45 #include "opt_quota.h"
46 #endif
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/proc.h>
51 #include <sys/vnode.h>
52 #include <sys/mount.h>
53 #include <sys/kernel.h>
54 #include <sys/namei.h>
55 #include <sys/kauth.h>
56 #include <sys/kmem.h>
57 
58 #include <ufs/lfs/lfs.h>
59 #include <ufs/lfs/lfs_accessors.h>
60 #include <ufs/lfs/lfs_extern.h>
61 
62 #include <ufs/lfs/ulfs_inode.h>
63 #include <ufs/lfs/ulfsmount.h>
64 #include <ufs/lfs/ulfs_extern.h>
65 #ifdef LFS_DIRHASH
66 #include <ufs/lfs/ulfs_dirhash.h>
67 #endif
68 #ifdef LFS_EXTATTR
69 #include <ufs/lfs/ulfs_extattr.h>
70 #endif
71 
72 #include <uvm/uvm.h>
73 
74 /*
75  * Last reference to an inode.  If necessary, write or delete it.
76  */
77 int
78 ulfs_inactive(void *v)
79 {
80 	struct vop_inactive_v2_args /* {
81 		struct vnode *a_vp;
82 		struct bool *a_recycle;
83 	} */ *ap = v;
84 	struct vnode *vp = ap->a_vp;
85 	struct inode *ip = VTOI(vp);
86 	mode_t mode;
87 	int error = 0;
88 
89 	/*
90 	 * Ignore inodes related to stale file handles.
91 	 */
92 	if (ip->i_mode == 0)
93 		goto out;
94 	if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
95 #ifdef LFS_EXTATTR
96 		ulfs_extattr_vnode_inactive(vp, curlwp);
97 #endif
98 		if (ip->i_size != 0) {
99 			error = lfs_truncate(vp, (off_t)0, 0, NOCRED);
100 		}
101 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
102 		(void)lfs_chkiq(ip, -1, NOCRED, 0);
103 #endif
104 		DIP_ASSIGN(ip, rdev, 0);
105 		mode = ip->i_mode;
106 		ip->i_mode = 0;
107 		ip->i_omode = mode;
108 		DIP_ASSIGN(ip, mode, 0);
109 		ip->i_state |= IN_CHANGE | IN_UPDATE;
110 		/*
111 		 * Defer final inode free and update to ulfs_reclaim().
112 		 */
113 	}
114 
115 	if (ip->i_state & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
116 		lfs_update(vp, NULL, NULL, 0);
117 	}
118 
119 out:
120 	/*
121 	 * If we are done with the inode, reclaim it
122 	 * so that it can be reused immediately.
123 	 */
124 	*ap->a_recycle = (ip->i_mode == 0);
125 
126 	return (error);
127 }
128 
129 /*
130  * Reclaim an inode so that it can be used for other purposes.
131  */
132 int
133 ulfs_reclaim(struct vnode *vp)
134 {
135 	struct inode *ip = VTOI(vp);
136 
137 	/* XXX: do we really need two of these? */
138 	/* note: originally the first was inside a wapbl txn */
139 	lfs_update(vp, NULL, NULL, UPDATE_CLOSE);
140 	lfs_update(vp, NULL, NULL, UPDATE_CLOSE);
141 
142 	if (ip->i_devvp) {
143 		vrele(ip->i_devvp);
144 		ip->i_devvp = 0;
145 	}
146 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
147 	ulfsquota_free(ip);
148 #endif
149 #ifdef LFS_DIRHASH
150 	if (ip->i_dirhash != NULL)
151 		ulfsdirhash_free(ip);
152 #endif
153 	return (0);
154 }
155 
156 /*
157  * allocate a range of blocks in a file.
158  * after this function returns, any page entirely contained within the range
159  * will map to invalid data and thus must be overwritten before it is made
160  * accessible to others.
161  */
162 
163 int
164 ulfs_balloc_range(struct vnode *vp, off_t off, off_t len, kauth_cred_t cred,
165     int flags)
166 {
167 	off_t neweof;	/* file size after the operation */
168 	off_t neweob;	/* offset next to the last block after the operation */
169 	off_t pagestart; /* starting offset of range covered by pgs */
170 	off_t eob;	/* offset next to allocated blocks */
171 	struct uvm_object *uobj;
172 	int i, delta, error, npages;
173 	int bshift = vp->v_mount->mnt_fs_bshift;
174 	int bsize = 1 << bshift;
175 	int ppb = MAX(bsize >> PAGE_SHIFT, 1);
176 	struct vm_page **pgs;
177 	size_t pgssize;
178 	UVMHIST_FUNC("ulfs_balloc_range"); UVMHIST_CALLED(ubchist);
179 	UVMHIST_LOG(ubchist, "vp %#jx off 0x%jx len 0x%jx u_size 0x%jx",
180 		    (uintptr_t)vp, off, len, vp->v_size);
181 
182 	neweof = MAX(vp->v_size, off + len);
183 	GOP_SIZE(vp, neweof, &neweob, 0);
184 
185 	error = 0;
186 	uobj = &vp->v_uobj;
187 
188 	/*
189 	 * read or create pages covering the range of the allocation and
190 	 * keep them locked until the new block is allocated, so there
191 	 * will be no window where the old contents of the new block are
192 	 * visible to racing threads.
193 	 */
194 
195 	pagestart = trunc_page(off) & ~(bsize - 1);
196 	npages = MIN(ppb, (round_page(neweob) - pagestart) >> PAGE_SHIFT);
197 	pgssize = npages * sizeof(struct vm_page *);
198 	pgs = kmem_zalloc(pgssize, KM_SLEEP);
199 
200 	/*
201 	 * adjust off to be block-aligned.
202 	 */
203 
204 	delta = off & (bsize - 1);
205 	off -= delta;
206 	len += delta;
207 
208 	genfs_node_wrlock(vp);
209 	mutex_enter(uobj->vmobjlock);
210 	error = VOP_GETPAGES(vp, pagestart, pgs, &npages, 0,
211 	    VM_PROT_WRITE, 0, PGO_SYNCIO | PGO_PASTEOF | PGO_NOBLOCKALLOC |
212 	    PGO_NOTIMESTAMP | PGO_GLOCKHELD);
213 	if (error) {
214 		genfs_node_unlock(vp);
215 		goto out;
216 	}
217 
218 	/*
219 	 * now allocate the range.
220 	 */
221 
222 	error = GOP_ALLOC(vp, off, len, flags, cred);
223 	genfs_node_unlock(vp);
224 
225 	/*
226 	 * if the allocation succeeded, clear PG_CLEAN on all the pages
227 	 * and clear PG_RDONLY on any pages that are now fully backed
228 	 * by disk blocks.  if the allocation failed, we do not invalidate
229 	 * the pages since they might have already existed and been dirty,
230 	 * in which case we need to keep them around.  if we created the pages,
231 	 * they will be clean and read-only, and leaving such pages
232 	 * in the cache won't cause any problems.
233 	 */
234 
235 	GOP_SIZE(vp, off + len, &eob, 0);
236 	mutex_enter(uobj->vmobjlock);
237 	mutex_enter(&uvm_pageqlock);
238 	for (i = 0; i < npages; i++) {
239 		KASSERT((pgs[i]->flags & PG_RELEASED) == 0);
240 		if (!error) {
241 			if (off <= pagestart + (i << PAGE_SHIFT) &&
242 			    pagestart + ((i + 1) << PAGE_SHIFT) <= eob) {
243 				pgs[i]->flags &= ~PG_RDONLY;
244 			}
245 			pgs[i]->flags &= ~PG_CLEAN;
246 		}
247 		uvm_pageactivate(pgs[i]);
248 	}
249 	mutex_exit(&uvm_pageqlock);
250 	uvm_page_unbusy(pgs, npages);
251 	mutex_exit(uobj->vmobjlock);
252 
253  out:
254  	kmem_free(pgs, pgssize);
255 	return error;
256 }
257