Sun Jan 4 19:20:32 2009 UTC ()
fix diagnostic printfs.


(christos)
diff -r1.106.8.7 -r1.106.8.8 src/sys/ufs/ffs/ffs_alloc.c

cvs diff -r1.106.8.7 -r1.106.8.8 src/sys/ufs/ffs/ffs_alloc.c (switch to unified diff)

--- src/sys/ufs/ffs/ffs_alloc.c 2008/12/30 19:30:31 1.106.8.7
+++ src/sys/ufs/ffs/ffs_alloc.c 2009/01/04 19:20:32 1.106.8.8
@@ -1,1316 +1,1318 @@ @@ -1,1316 +1,1318 @@
1/* $NetBSD: ffs_alloc.c,v 1.106.8.7 2008/12/30 19:30:31 christos Exp $ */ 1/* $NetBSD: ffs_alloc.c,v 1.106.8.8 2009/01/04 19:20:32 christos Exp $ */
2 2
3/*- 3/*-
4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 4 * Copyright (c) 2008 The NetBSD Foundation, Inc.
5 * All rights reserved. 5 * All rights reserved.
6 * 6 *
7 * This code is derived from software contributed to The NetBSD Foundation 7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Wasabi Systems, Inc. 8 * by Wasabi Systems, Inc.
9 * 9 *
10 * Redistribution and use in source and binary forms, with or without 10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions 11 * modification, are permitted provided that the following conditions
12 * are met: 12 * are met:
13 * 1. Redistributions of source code must retain the above copyright 13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer. 14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright 15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the 16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution. 17 * documentation and/or other materials provided with the distribution.
18 * 18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE. 29 * POSSIBILITY OF SUCH DAMAGE.
30 */ 30 */
31 31
32/* 32/*
33 * Copyright (c) 2002 Networks Associates Technology, Inc. 33 * Copyright (c) 2002 Networks Associates Technology, Inc.
34 * All rights reserved. 34 * All rights reserved.
35 * 35 *
36 * This software was developed for the FreeBSD Project by Marshall 36 * This software was developed for the FreeBSD Project by Marshall
37 * Kirk McKusick and Network Associates Laboratories, the Security 37 * Kirk McKusick and Network Associates Laboratories, the Security
38 * Research Division of Network Associates, Inc. under DARPA/SPAWAR 38 * Research Division of Network Associates, Inc. under DARPA/SPAWAR
39 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS 39 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
40 * research program 40 * research program
41 * 41 *
42 * Copyright (c) 1982, 1986, 1989, 1993 42 * Copyright (c) 1982, 1986, 1989, 1993
43 * The Regents of the University of California. All rights reserved. 43 * The Regents of the University of California. All rights reserved.
44 * 44 *
45 * Redistribution and use in source and binary forms, with or without 45 * Redistribution and use in source and binary forms, with or without
46 * modification, are permitted provided that the following conditions 46 * modification, are permitted provided that the following conditions
47 * are met: 47 * are met:
48 * 1. Redistributions of source code must retain the above copyright 48 * 1. Redistributions of source code must retain the above copyright
49 * notice, this list of conditions and the following disclaimer. 49 * notice, this list of conditions and the following disclaimer.
50 * 2. Redistributions in binary form must reproduce the above copyright 50 * 2. Redistributions in binary form must reproduce the above copyright
51 * notice, this list of conditions and the following disclaimer in the 51 * notice, this list of conditions and the following disclaimer in the
52 * documentation and/or other materials provided with the distribution. 52 * documentation and/or other materials provided with the distribution.
53 * 3. Neither the name of the University nor the names of its contributors 53 * 3. Neither the name of the University nor the names of its contributors
54 * may be used to endorse or promote products derived from this software 54 * may be used to endorse or promote products derived from this software
55 * without specific prior written permission. 55 * without specific prior written permission.
56 * 56 *
57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67 * SUCH DAMAGE. 67 * SUCH DAMAGE.
68 * 68 *
69 * @(#)ffs_alloc.c 8.19 (Berkeley) 7/13/95 69 * @(#)ffs_alloc.c 8.19 (Berkeley) 7/13/95
70 */ 70 */
71 71
72#include <sys/cdefs.h> 72#include <sys/cdefs.h>
73__KERNEL_RCSID(0, "$NetBSD: ffs_alloc.c,v 1.106.8.7 2008/12/30 19:30:31 christos Exp $"); 73__KERNEL_RCSID(0, "$NetBSD: ffs_alloc.c,v 1.106.8.8 2009/01/04 19:20:32 christos Exp $");
74 74
75#if defined(_KERNEL_OPT) 75#if defined(_KERNEL_OPT)
76#include "opt_ffs.h" 76#include "opt_ffs.h"
77#include "opt_quota.h" 77#include "opt_quota.h"
78#endif 78#endif
79 79
80#include <sys/param.h> 80#include <sys/param.h>
81#include <sys/systm.h> 81#include <sys/systm.h>
82#include <sys/buf.h> 82#include <sys/buf.h>
83#include <sys/fstrans.h> 83#include <sys/fstrans.h>
84#include <sys/kauth.h> 84#include <sys/kauth.h>
85#include <sys/kernel.h> 85#include <sys/kernel.h>
86#include <sys/mount.h> 86#include <sys/mount.h>
87#include <sys/proc.h> 87#include <sys/proc.h>
88#include <sys/syslog.h> 88#include <sys/syslog.h>
89#include <sys/vnode.h> 89#include <sys/vnode.h>
90#include <sys/wapbl.h> 90#include <sys/wapbl.h>
91 91
92#include <miscfs/specfs/specdev.h> 92#include <miscfs/specfs/specdev.h>
93#include <ufs/ufs/quota.h> 93#include <ufs/ufs/quota.h>
94#include <ufs/ufs/ufsmount.h> 94#include <ufs/ufs/ufsmount.h>
95#include <ufs/ufs/inode.h> 95#include <ufs/ufs/inode.h>
96#include <ufs/ufs/ufs_extern.h> 96#include <ufs/ufs/ufs_extern.h>
97#include <ufs/ufs/ufs_bswap.h> 97#include <ufs/ufs/ufs_bswap.h>
98#include <ufs/ufs/ufs_wapbl.h> 98#include <ufs/ufs/ufs_wapbl.h>
99 99
100#include <ufs/ffs/fs.h> 100#include <ufs/ffs/fs.h>
101#include <ufs/ffs/ffs_extern.h> 101#include <ufs/ffs/ffs_extern.h>
102 102
103static daddr_t ffs_alloccg(struct inode *, int, daddr_t, int, int); 103static daddr_t ffs_alloccg(struct inode *, int, daddr_t, int, int);
104static daddr_t ffs_alloccgblk(struct inode *, struct buf *, daddr_t, int); 104static daddr_t ffs_alloccgblk(struct inode *, struct buf *, daddr_t, int);
105static ino_t ffs_dirpref(struct inode *); 105static ino_t ffs_dirpref(struct inode *);
106static daddr_t ffs_fragextend(struct inode *, int, daddr_t, int, int); 106static daddr_t ffs_fragextend(struct inode *, int, daddr_t, int, int);
107static void ffs_fserr(struct fs *, u_int, const char *); 107static void ffs_fserr(struct fs *, u_int, const char *);
108static daddr_t ffs_hashalloc(struct inode *, int, daddr_t, int, int, 108static daddr_t ffs_hashalloc(struct inode *, int, daddr_t, int, int,
109 daddr_t (*)(struct inode *, int, daddr_t, int, int)); 109 daddr_t (*)(struct inode *, int, daddr_t, int, int));
110static daddr_t ffs_nodealloccg(struct inode *, int, daddr_t, int, int); 110static daddr_t ffs_nodealloccg(struct inode *, int, daddr_t, int, int);
111static int32_t ffs_mapsearch(struct fs *, struct cg *, 111static int32_t ffs_mapsearch(struct fs *, struct cg *,
112 daddr_t, int); 112 daddr_t, int);
113static void ffs_blkfree_common(struct ufsmount *, struct fs *, dev_t, struct buf *, 113static void ffs_blkfree_common(struct ufsmount *, struct fs *, dev_t, struct buf *,
114 daddr_t, long, bool); 114 daddr_t, long, bool);
115static void ffs_freefile_common(struct ufsmount *, struct fs *, dev_t, struct buf *, ino_t, 115static void ffs_freefile_common(struct ufsmount *, struct fs *, dev_t, struct buf *, ino_t,
116 int, bool); 116 int, bool);
117 117
118/* if 1, changes in optimalization strategy are logged */ 118/* if 1, changes in optimalization strategy are logged */
119int ffs_log_changeopt = 0; 119int ffs_log_changeopt = 0;
120 120
121/* in ffs_tables.c */ 121/* in ffs_tables.c */
122extern const int inside[], around[]; 122extern const int inside[], around[];
123extern const u_char * const fragtbl[]; 123extern const u_char * const fragtbl[];
124 124
125/* Basic consistency check for block allocations */ 125/* Basic consistency check for block allocations */
126static int 126static int
127ffs_check_bad_allocation(const char *func, struct fs *fs, daddr_t bno, 127ffs_check_bad_allocation(const char *func, struct fs *fs, daddr_t bno,
128 long size, dev_t dev, ino_t inum) 128 long size, dev_t dev, ino_t inum)
129{ 129{
130 if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 || 130 if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 ||
131 fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) { 131 fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) {
132 printf("dev = 0x%llx, bno = %" PRId64 " bsize = %d, " 132 printf("dev = 0x%llx, bno = %" PRId64 " bsize = %d, "
133 "size = %ld, fs = %s\n", 133 "size = %ld, fs = %s\n",
134 (long long)dev, bno, fs->fs_bsize, size, fs->fs_fsmnt); 134 (long long)dev, bno, fs->fs_bsize, size, fs->fs_fsmnt);
135 panic("%s: bad size", func); 135 panic("%s: bad size", func);
136 } 136 }
137 137
138 if (bno >= fs->fs_size) { 138 if (bno >= fs->fs_size) {
139 printf("bad block %" PRId64 ", ino %llu\n", bno, 139 printf("bad block %" PRId64 ", ino %llu\n", bno,
140 (unsigned long long)inum); 140 (unsigned long long)inum);
141 ffs_fserr(fs, inum, "bad block"); 141 ffs_fserr(fs, inum, "bad block");
142 return EINVAL; 142 return EINVAL;
143 } 143 }
144 return 0; 144 return 0;
145} 145}
146 146
147/* 147/*
148 * Allocate a block in the file system. 148 * Allocate a block in the file system.
149 * 149 *
150 * The size of the requested block is given, which must be some 150 * The size of the requested block is given, which must be some
151 * multiple of fs_fsize and <= fs_bsize. 151 * multiple of fs_fsize and <= fs_bsize.
152 * A preference may be optionally specified. If a preference is given 152 * A preference may be optionally specified. If a preference is given
153 * the following hierarchy is used to allocate a block: 153 * the following hierarchy is used to allocate a block:
154 * 1) allocate the requested block. 154 * 1) allocate the requested block.
155 * 2) allocate a rotationally optimal block in the same cylinder. 155 * 2) allocate a rotationally optimal block in the same cylinder.
156 * 3) allocate a block in the same cylinder group. 156 * 3) allocate a block in the same cylinder group.
157 * 4) quadradically rehash into other cylinder groups, until an 157 * 4) quadradically rehash into other cylinder groups, until an
158 * available block is located. 158 * available block is located.
159 * If no block preference is given the following hierarchy is used 159 * If no block preference is given the following hierarchy is used
160 * to allocate a block: 160 * to allocate a block:
161 * 1) allocate a block in the cylinder group that contains the 161 * 1) allocate a block in the cylinder group that contains the
162 * inode for the file. 162 * inode for the file.
163 * 2) quadradically rehash into other cylinder groups, until an 163 * 2) quadradically rehash into other cylinder groups, until an
164 * available block is located. 164 * available block is located.
165 * 165 *
166 * => called with um_lock held 166 * => called with um_lock held
167 * => releases um_lock before returning 167 * => releases um_lock before returning
168 */ 168 */
169int 169int
170ffs_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size, int flags, 170ffs_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size, int flags,
171 kauth_cred_t cred, daddr_t *bnp) 171 kauth_cred_t cred, daddr_t *bnp)
172{ 172{
173 struct ufsmount *ump; 173 struct ufsmount *ump;
174 struct fs *fs; 174 struct fs *fs;
175 daddr_t bno; 175 daddr_t bno;
176 int cg; 176 int cg;
177#ifdef QUOTA 177#ifdef QUOTA
178 int error; 178 int error;
179#endif 179#endif
180 180
181 fs = ip->i_fs; 181 fs = ip->i_fs;
182 ump = ip->i_ump; 182 ump = ip->i_ump;
183 183
184 KASSERT(mutex_owned(&ump->um_lock)); 184 KASSERT(mutex_owned(&ump->um_lock));
185 185
186#ifdef UVM_PAGE_TRKOWN 186#ifdef UVM_PAGE_TRKOWN
187 if (ITOV(ip)->v_type == VREG && 187 if (ITOV(ip)->v_type == VREG &&
188 lblktosize(fs, (voff_t)lbn) < round_page(ITOV(ip)->v_size)) { 188 lblktosize(fs, (voff_t)lbn) < round_page(ITOV(ip)->v_size)) {
189 struct vm_page *pg; 189 struct vm_page *pg;
190 struct uvm_object *uobj = &ITOV(ip)->v_uobj; 190 struct uvm_object *uobj = &ITOV(ip)->v_uobj;
191 voff_t off = trunc_page(lblktosize(fs, lbn)); 191 voff_t off = trunc_page(lblktosize(fs, lbn));
192 voff_t endoff = round_page(lblktosize(fs, lbn) + size); 192 voff_t endoff = round_page(lblktosize(fs, lbn) + size);
193 193
194 mutex_enter(&uobj->vmobjlock); 194 mutex_enter(&uobj->vmobjlock);
195 while (off < endoff) { 195 while (off < endoff) {
196 pg = uvm_pagelookup(uobj, off); 196 pg = uvm_pagelookup(uobj, off);
197 KASSERT(pg != NULL); 197 KASSERT(pg != NULL);
198 KASSERT(pg->owner == curproc->p_pid); 198 KASSERT(pg->owner == curproc->p_pid);
199 off += PAGE_SIZE; 199 off += PAGE_SIZE;
200 } 200 }
201 mutex_exit(&uobj->vmobjlock); 201 mutex_exit(&uobj->vmobjlock);
202 } 202 }
203#endif 203#endif
204 204
205 *bnp = 0; 205 *bnp = 0;
206#ifdef DIAGNOSTIC 206#ifdef DIAGNOSTIC
207 if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) { 207 if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
208 printf("dev = 0x%llx, bsize = %d, size = %d, fs = %s\n", 208 printf("dev = 0x%llx, bsize = %d, size = %d, fs = %s\n",
209 ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt); 209 (unsigned long long)ip->i_dev, fs->fs_bsize, size,
 210 fs->fs_fsmnt);
210 panic("ffs_alloc: bad size"); 211 panic("ffs_alloc: bad size");
211 } 212 }
212 if (cred == NOCRED) 213 if (cred == NOCRED)
213 panic("ffs_alloc: missing credential"); 214 panic("ffs_alloc: missing credential");
214#endif /* DIAGNOSTIC */ 215#endif /* DIAGNOSTIC */
215 if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0) 216 if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
216 goto nospace; 217 goto nospace;
217 if (freespace(fs, fs->fs_minfree) <= 0 && 218 if (freespace(fs, fs->fs_minfree) <= 0 &&
218 kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) != 0) 219 kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) != 0)
219 goto nospace; 220 goto nospace;
220#ifdef QUOTA 221#ifdef QUOTA
221 mutex_exit(&ump->um_lock); 222 mutex_exit(&ump->um_lock);
222 if ((error = chkdq(ip, btodb(size), cred, 0)) != 0) 223 if ((error = chkdq(ip, btodb(size), cred, 0)) != 0)
223 return (error); 224 return (error);
224 mutex_enter(&ump->um_lock); 225 mutex_enter(&ump->um_lock);
225#endif 226#endif
226 227
227 if (bpref >= fs->fs_size) 228 if (bpref >= fs->fs_size)
228 bpref = 0; 229 bpref = 0;
229 if (bpref == 0) 230 if (bpref == 0)
230 cg = ino_to_cg(fs, ip->i_number); 231 cg = ino_to_cg(fs, ip->i_number);
231 else 232 else
232 cg = dtog(fs, bpref); 233 cg = dtog(fs, bpref);
233 bno = ffs_hashalloc(ip, cg, bpref, size, flags, ffs_alloccg); 234 bno = ffs_hashalloc(ip, cg, bpref, size, flags, ffs_alloccg);
234 if (bno > 0) { 235 if (bno > 0) {
235 DIP_ADD(ip, blocks, btodb(size)); 236 DIP_ADD(ip, blocks, btodb(size));
236 ip->i_flag |= IN_CHANGE | IN_UPDATE; 237 ip->i_flag |= IN_CHANGE | IN_UPDATE;
237 *bnp = bno; 238 *bnp = bno;
238 return (0); 239 return (0);
239 } 240 }
240#ifdef QUOTA 241#ifdef QUOTA
241 /* 242 /*
242 * Restore user's disk quota because allocation failed. 243 * Restore user's disk quota because allocation failed.
243 */ 244 */
244 (void) chkdq(ip, -btodb(size), cred, FORCE); 245 (void) chkdq(ip, -btodb(size), cred, FORCE);
245#endif 246#endif
246 if (flags & B_CONTIG) { 247 if (flags & B_CONTIG) {
247 /* 248 /*
248 * XXX ump->um_lock handling is "suspect" at best. 249 * XXX ump->um_lock handling is "suspect" at best.
249 * For the case where ffs_hashalloc() fails early 250 * For the case where ffs_hashalloc() fails early
250 * in the B_CONTIG case we reach here with um_lock 251 * in the B_CONTIG case we reach here with um_lock
251 * already unlocked, so we can't release it again 252 * already unlocked, so we can't release it again
252 * like in the normal error path. See kern/39206. 253 * like in the normal error path. See kern/39206.
253 * 254 *
254 * 255 *
255 * Fail silently - it's up to our caller to report 256 * Fail silently - it's up to our caller to report
256 * errors. 257 * errors.
257 */ 258 */
258 return (ENOSPC); 259 return (ENOSPC);
259 } 260 }
260nospace: 261nospace:
261 mutex_exit(&ump->um_lock); 262 mutex_exit(&ump->um_lock);
262 ffs_fserr(fs, kauth_cred_geteuid(cred), "file system full"); 263 ffs_fserr(fs, kauth_cred_geteuid(cred), "file system full");
263 uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt); 264 uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
264 return (ENOSPC); 265 return (ENOSPC);
265} 266}
266 267
267/* 268/*
268 * Reallocate a fragment to a bigger size 269 * Reallocate a fragment to a bigger size
269 * 270 *
270 * The number and size of the old block is given, and a preference 271 * The number and size of the old block is given, and a preference
271 * and new size is also specified. The allocator attempts to extend 272 * and new size is also specified. The allocator attempts to extend
272 * the original block. Failing that, the regular block allocator is 273 * the original block. Failing that, the regular block allocator is
273 * invoked to get an appropriate block. 274 * invoked to get an appropriate block.
274 * 275 *
275 * => called with um_lock held 276 * => called with um_lock held
276 * => return with um_lock released 277 * => return with um_lock released
277 */ 278 */
278int 279int
279ffs_realloccg(struct inode *ip, daddr_t lbprev, daddr_t bpref, int osize, 280ffs_realloccg(struct inode *ip, daddr_t lbprev, daddr_t bpref, int osize,
280 int nsize, kauth_cred_t cred, struct buf **bpp, daddr_t *blknop) 281 int nsize, kauth_cred_t cred, struct buf **bpp, daddr_t *blknop)
281{ 282{
282 struct ufsmount *ump; 283 struct ufsmount *ump;
283 struct fs *fs; 284 struct fs *fs;
284 struct buf *bp; 285 struct buf *bp;
285 int cg, request, error; 286 int cg, request, error;
286 daddr_t bprev, bno; 287 daddr_t bprev, bno;
287 288
288 fs = ip->i_fs; 289 fs = ip->i_fs;
289 ump = ip->i_ump; 290 ump = ip->i_ump;
290 291
291 KASSERT(mutex_owned(&ump->um_lock)); 292 KASSERT(mutex_owned(&ump->um_lock));
292 293
293#ifdef UVM_PAGE_TRKOWN 294#ifdef UVM_PAGE_TRKOWN
294 if (ITOV(ip)->v_type == VREG) { 295 if (ITOV(ip)->v_type == VREG) {
295 struct vm_page *pg; 296 struct vm_page *pg;
296 struct uvm_object *uobj = &ITOV(ip)->v_uobj; 297 struct uvm_object *uobj = &ITOV(ip)->v_uobj;
297 voff_t off = trunc_page(lblktosize(fs, lbprev)); 298 voff_t off = trunc_page(lblktosize(fs, lbprev));
298 voff_t endoff = round_page(lblktosize(fs, lbprev) + osize); 299 voff_t endoff = round_page(lblktosize(fs, lbprev) + osize);
299 300
300 mutex_enter(&uobj->vmobjlock); 301 mutex_enter(&uobj->vmobjlock);
301 while (off < endoff) { 302 while (off < endoff) {
302 pg = uvm_pagelookup(uobj, off); 303 pg = uvm_pagelookup(uobj, off);
303 KASSERT(pg != NULL); 304 KASSERT(pg != NULL);
304 KASSERT(pg->owner == curproc->p_pid); 305 KASSERT(pg->owner == curproc->p_pid);
305 KASSERT((pg->flags & PG_CLEAN) == 0); 306 KASSERT((pg->flags & PG_CLEAN) == 0);
306 off += PAGE_SIZE; 307 off += PAGE_SIZE;
307 } 308 }
308 mutex_exit(&uobj->vmobjlock); 309 mutex_exit(&uobj->vmobjlock);
309 } 310 }
310#endif 311#endif
311 312
312#ifdef DIAGNOSTIC 313#ifdef DIAGNOSTIC
313 if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 || 314 if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
314 (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) { 315 (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
315 printf( 316 printf(
316 "dev = 0x%llx, bsize = %d, osize = %d, nsize = %d, fs = %s\n", 317 "dev = 0x%llx, bsize = %d, osize = %d, nsize = %d, fs = %s\n",
317 ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt); 318 (unsigned long long)ip->i_dev, fs->fs_bsize, osize, nsize,
 319 fs->fs_fsmnt);
318 panic("ffs_realloccg: bad size"); 320 panic("ffs_realloccg: bad size");
319 } 321 }
320 if (cred == NOCRED) 322 if (cred == NOCRED)
321 panic("ffs_realloccg: missing credential"); 323 panic("ffs_realloccg: missing credential");
322#endif /* DIAGNOSTIC */ 324#endif /* DIAGNOSTIC */
323 if (freespace(fs, fs->fs_minfree) <= 0 && 325 if (freespace(fs, fs->fs_minfree) <= 0 &&
324 kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) != 0) { 326 kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, NULL) != 0) {
325 mutex_exit(&ump->um_lock); 327 mutex_exit(&ump->um_lock);
326 goto nospace; 328 goto nospace;
327 } 329 }
328 if (fs->fs_magic == FS_UFS2_MAGIC) 330 if (fs->fs_magic == FS_UFS2_MAGIC)
329 bprev = ufs_rw64(ip->i_ffs2_db[lbprev], UFS_FSNEEDSWAP(fs)); 331 bprev = ufs_rw64(ip->i_ffs2_db[lbprev], UFS_FSNEEDSWAP(fs));
330 else 332 else
331 bprev = ufs_rw32(ip->i_ffs1_db[lbprev], UFS_FSNEEDSWAP(fs)); 333 bprev = ufs_rw32(ip->i_ffs1_db[lbprev], UFS_FSNEEDSWAP(fs));
332 334
333 if (bprev == 0) { 335 if (bprev == 0) {
334 printf("dev = 0x%llx, bsize = %d, bprev = %" PRId64 ", fs = %s\n", 336 printf("dev = 0x%llx, bsize = %d, bprev = %" PRId64 ", fs = %s\n",
335 (unsigned long long)ip->i_dev, fs->fs_bsize, bprev, 337 (unsigned long long)ip->i_dev, fs->fs_bsize, bprev,
336 fs->fs_fsmnt); 338 fs->fs_fsmnt);
337 panic("ffs_realloccg: bad bprev"); 339 panic("ffs_realloccg: bad bprev");
338 } 340 }
339 mutex_exit(&ump->um_lock); 341 mutex_exit(&ump->um_lock);
340 342
341 /* 343 /*
342 * Allocate the extra space in the buffer. 344 * Allocate the extra space in the buffer.
343 */ 345 */
344 if (bpp != NULL && 346 if (bpp != NULL &&
345 (error = bread(ITOV(ip), lbprev, osize, NOCRED, 0, &bp)) != 0) { 347 (error = bread(ITOV(ip), lbprev, osize, NOCRED, 0, &bp)) != 0) {
346 brelse(bp, 0); 348 brelse(bp, 0);
347 return (error); 349 return (error);
348 } 350 }
349#ifdef QUOTA 351#ifdef QUOTA
350 if ((error = chkdq(ip, btodb(nsize - osize), cred, 0)) != 0) { 352 if ((error = chkdq(ip, btodb(nsize - osize), cred, 0)) != 0) {
351 if (bpp != NULL) { 353 if (bpp != NULL) {
352 brelse(bp, 0); 354 brelse(bp, 0);
353 } 355 }
354 return (error); 356 return (error);
355 } 357 }
356#endif 358#endif
357 /* 359 /*
358 * Check for extension in the existing location. 360 * Check for extension in the existing location.
359 */ 361 */
360 cg = dtog(fs, bprev); 362 cg = dtog(fs, bprev);
361 mutex_enter(&ump->um_lock); 363 mutex_enter(&ump->um_lock);
362 if ((bno = ffs_fragextend(ip, cg, bprev, osize, nsize)) != 0) { 364 if ((bno = ffs_fragextend(ip, cg, bprev, osize, nsize)) != 0) {
363 DIP_ADD(ip, blocks, btodb(nsize - osize)); 365 DIP_ADD(ip, blocks, btodb(nsize - osize));
364 ip->i_flag |= IN_CHANGE | IN_UPDATE; 366 ip->i_flag |= IN_CHANGE | IN_UPDATE;
365 367
366 if (bpp != NULL) { 368 if (bpp != NULL) {
367 if (bp->b_blkno != fsbtodb(fs, bno)) 369 if (bp->b_blkno != fsbtodb(fs, bno))
368 panic("bad blockno"); 370 panic("bad blockno");
369 allocbuf(bp, nsize, 1); 371 allocbuf(bp, nsize, 1);
370 memset((char *)bp->b_data + osize, 0, nsize - osize); 372 memset((char *)bp->b_data + osize, 0, nsize - osize);
371 mutex_enter(bp->b_objlock); 373 mutex_enter(bp->b_objlock);
372 KASSERT(!cv_has_waiters(&bp->b_done)); 374 KASSERT(!cv_has_waiters(&bp->b_done));
373 bp->b_oflags |= BO_DONE; 375 bp->b_oflags |= BO_DONE;
374 mutex_exit(bp->b_objlock); 376 mutex_exit(bp->b_objlock);
375 *bpp = bp; 377 *bpp = bp;
376 } 378 }
377 if (blknop != NULL) { 379 if (blknop != NULL) {
378 *blknop = bno; 380 *blknop = bno;
379 } 381 }
380 return (0); 382 return (0);
381 } 383 }
382 /* 384 /*
383 * Allocate a new disk location. 385 * Allocate a new disk location.
384 */ 386 */
385 if (bpref >= fs->fs_size) 387 if (bpref >= fs->fs_size)
386 bpref = 0; 388 bpref = 0;
387 switch ((int)fs->fs_optim) { 389 switch ((int)fs->fs_optim) {
388 case FS_OPTSPACE: 390 case FS_OPTSPACE:
389 /* 391 /*
390 * Allocate an exact sized fragment. Although this makes 392 * Allocate an exact sized fragment. Although this makes
391 * best use of space, we will waste time relocating it if 393 * best use of space, we will waste time relocating it if
392 * the file continues to grow. If the fragmentation is 394 * the file continues to grow. If the fragmentation is
393 * less than half of the minimum free reserve, we choose 395 * less than half of the minimum free reserve, we choose
394 * to begin optimizing for time. 396 * to begin optimizing for time.
395 */ 397 */
396 request = nsize; 398 request = nsize;
397 if (fs->fs_minfree < 5 || 399 if (fs->fs_minfree < 5 ||
398 fs->fs_cstotal.cs_nffree > 400 fs->fs_cstotal.cs_nffree >
399 fs->fs_dsize * fs->fs_minfree / (2 * 100)) 401 fs->fs_dsize * fs->fs_minfree / (2 * 100))
400 break; 402 break;
401 403
402 if (ffs_log_changeopt) { 404 if (ffs_log_changeopt) {
403 log(LOG_NOTICE, 405 log(LOG_NOTICE,
404 "%s: optimization changed from SPACE to TIME\n", 406 "%s: optimization changed from SPACE to TIME\n",
405 fs->fs_fsmnt); 407 fs->fs_fsmnt);
406 } 408 }
407 409
408 fs->fs_optim = FS_OPTTIME; 410 fs->fs_optim = FS_OPTTIME;
409 break; 411 break;
410 case FS_OPTTIME: 412 case FS_OPTTIME:
411 /* 413 /*
412 * At this point we have discovered a file that is trying to 414 * At this point we have discovered a file that is trying to
413 * grow a small fragment to a larger fragment. To save time, 415 * grow a small fragment to a larger fragment. To save time,
414 * we allocate a full sized block, then free the unused portion. 416 * we allocate a full sized block, then free the unused portion.
415 * If the file continues to grow, the `ffs_fragextend' call 417 * If the file continues to grow, the `ffs_fragextend' call
416 * above will be able to grow it in place without further 418 * above will be able to grow it in place without further
417 * copying. If aberrant programs cause disk fragmentation to 419 * copying. If aberrant programs cause disk fragmentation to
418 * grow within 2% of the free reserve, we choose to begin 420 * grow within 2% of the free reserve, we choose to begin
419 * optimizing for space. 421 * optimizing for space.
420 */ 422 */
421 request = fs->fs_bsize; 423 request = fs->fs_bsize;
422 if (fs->fs_cstotal.cs_nffree < 424 if (fs->fs_cstotal.cs_nffree <
423 fs->fs_dsize * (fs->fs_minfree - 2) / 100) 425 fs->fs_dsize * (fs->fs_minfree - 2) / 100)
424 break; 426 break;
425 427
426 if (ffs_log_changeopt) { 428 if (ffs_log_changeopt) {
427 log(LOG_NOTICE, 429 log(LOG_NOTICE,
428 "%s: optimization changed from TIME to SPACE\n", 430 "%s: optimization changed from TIME to SPACE\n",
429 fs->fs_fsmnt); 431 fs->fs_fsmnt);
430 } 432 }
431 433
432 fs->fs_optim = FS_OPTSPACE; 434 fs->fs_optim = FS_OPTSPACE;
433 break; 435 break;
434 default: 436 default:
435 printf("dev = 0x%llx, optim = %d, fs = %s\n", 437 printf("dev = 0x%llx, optim = %d, fs = %s\n",
436 (unsigned long long)ip->i_dev, fs->fs_optim, fs->fs_fsmnt); 438 (unsigned long long)ip->i_dev, fs->fs_optim, fs->fs_fsmnt);
437 panic("ffs_realloccg: bad optim"); 439 panic("ffs_realloccg: bad optim");
438 /* NOTREACHED */ 440 /* NOTREACHED */
439 } 441 }
440 bno = ffs_hashalloc(ip, cg, bpref, request, 0, ffs_alloccg); 442 bno = ffs_hashalloc(ip, cg, bpref, request, 0, ffs_alloccg);
441 if (bno > 0) { 443 if (bno > 0) {
442 if (!DOINGSOFTDEP(ITOV(ip))) { 444 if (!DOINGSOFTDEP(ITOV(ip))) {
443 if ((ip->i_ump->um_mountp->mnt_wapbl) && 445 if ((ip->i_ump->um_mountp->mnt_wapbl) &&
444 (ITOV(ip)->v_type != VREG)) { 446 (ITOV(ip)->v_type != VREG)) {
445 UFS_WAPBL_REGISTER_DEALLOCATION( 447 UFS_WAPBL_REGISTER_DEALLOCATION(
446 ip->i_ump->um_mountp, fsbtodb(fs, bprev), 448 ip->i_ump->um_mountp, fsbtodb(fs, bprev),
447 osize); 449 osize);
448 } else 450 } else
449 ffs_blkfree(fs, ip->i_devvp, bprev, (long)osize, 451 ffs_blkfree(fs, ip->i_devvp, bprev, (long)osize,
450 ip->i_number); 452 ip->i_number);
451 } 453 }
452 if (nsize < request) { 454 if (nsize < request) {
453 if ((ip->i_ump->um_mountp->mnt_wapbl) && 455 if ((ip->i_ump->um_mountp->mnt_wapbl) &&
454 (ITOV(ip)->v_type != VREG)) { 456 (ITOV(ip)->v_type != VREG)) {
455 UFS_WAPBL_REGISTER_DEALLOCATION( 457 UFS_WAPBL_REGISTER_DEALLOCATION(
456 ip->i_ump->um_mountp, 458 ip->i_ump->um_mountp,
457 fsbtodb(fs, (bno + numfrags(fs, nsize))), 459 fsbtodb(fs, (bno + numfrags(fs, nsize))),
458 request - nsize); 460 request - nsize);
459 } else 461 } else
460 ffs_blkfree(fs, ip->i_devvp, 462 ffs_blkfree(fs, ip->i_devvp,
461 bno + numfrags(fs, nsize), 463 bno + numfrags(fs, nsize),
462 (long)(request - nsize), ip->i_number); 464 (long)(request - nsize), ip->i_number);
463 } 465 }
464 DIP_ADD(ip, blocks, btodb(nsize - osize)); 466 DIP_ADD(ip, blocks, btodb(nsize - osize));
465 ip->i_flag |= IN_CHANGE | IN_UPDATE; 467 ip->i_flag |= IN_CHANGE | IN_UPDATE;
466 if (bpp != NULL) { 468 if (bpp != NULL) {
467 bp->b_blkno = fsbtodb(fs, bno); 469 bp->b_blkno = fsbtodb(fs, bno);
468 allocbuf(bp, nsize, 1); 470 allocbuf(bp, nsize, 1);
469 memset((char *)bp->b_data + osize, 0, (u_int)nsize - osize); 471 memset((char *)bp->b_data + osize, 0, (u_int)nsize - osize);
470 mutex_enter(bp->b_objlock); 472 mutex_enter(bp->b_objlock);
471 KASSERT(!cv_has_waiters(&bp->b_done)); 473 KASSERT(!cv_has_waiters(&bp->b_done));
472 bp->b_oflags |= BO_DONE; 474 bp->b_oflags |= BO_DONE;
473 mutex_exit(bp->b_objlock); 475 mutex_exit(bp->b_objlock);
474 *bpp = bp; 476 *bpp = bp;
475 } 477 }
476 if (blknop != NULL) { 478 if (blknop != NULL) {
477 *blknop = bno; 479 *blknop = bno;
478 } 480 }
479 return (0); 481 return (0);
480 } 482 }
481 mutex_exit(&ump->um_lock); 483 mutex_exit(&ump->um_lock);
482 484
483#ifdef QUOTA 485#ifdef QUOTA
484 /* 486 /*
485 * Restore user's disk quota because allocation failed. 487 * Restore user's disk quota because allocation failed.
486 */ 488 */
487 (void) chkdq(ip, -btodb(nsize - osize), cred, FORCE); 489 (void) chkdq(ip, -btodb(nsize - osize), cred, FORCE);
488#endif 490#endif
489 if (bpp != NULL) { 491 if (bpp != NULL) {
490 brelse(bp, 0); 492 brelse(bp, 0);
491 } 493 }
492 494
493nospace: 495nospace:
494 /* 496 /*
495 * no space available 497 * no space available
496 */ 498 */
497 ffs_fserr(fs, kauth_cred_geteuid(cred), "file system full"); 499 ffs_fserr(fs, kauth_cred_geteuid(cred), "file system full");
498 uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt); 500 uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
499 return (ENOSPC); 501 return (ENOSPC);
500} 502}
501 503
502/* 504/*
503 * Allocate an inode in the file system. 505 * Allocate an inode in the file system.
504 * 506 *
505 * If allocating a directory, use ffs_dirpref to select the inode. 507 * If allocating a directory, use ffs_dirpref to select the inode.
506 * If allocating in a directory, the following hierarchy is followed: 508 * If allocating in a directory, the following hierarchy is followed:
507 * 1) allocate the preferred inode. 509 * 1) allocate the preferred inode.
508 * 2) allocate an inode in the same cylinder group. 510 * 2) allocate an inode in the same cylinder group.
509 * 3) quadradically rehash into other cylinder groups, until an 511 * 3) quadradically rehash into other cylinder groups, until an
510 * available inode is located. 512 * available inode is located.
511 * If no inode preference is given the following hierarchy is used 513 * If no inode preference is given the following hierarchy is used
512 * to allocate an inode: 514 * to allocate an inode:
513 * 1) allocate an inode in cylinder group 0. 515 * 1) allocate an inode in cylinder group 0.
514 * 2) quadradically rehash into other cylinder groups, until an 516 * 2) quadradically rehash into other cylinder groups, until an
515 * available inode is located. 517 * available inode is located.
516 * 518 *
517 * => um_lock not held upon entry or return 519 * => um_lock not held upon entry or return
518 */ 520 */
519int 521int
520ffs_valloc(struct vnode *pvp, int mode, kauth_cred_t cred, 522ffs_valloc(struct vnode *pvp, int mode, kauth_cred_t cred,
521 struct vnode **vpp) 523 struct vnode **vpp)
522{ 524{
523 struct ufsmount *ump; 525 struct ufsmount *ump;
524 struct inode *pip; 526 struct inode *pip;
525 struct fs *fs; 527 struct fs *fs;
526 struct inode *ip; 528 struct inode *ip;
527 struct timespec ts; 529 struct timespec ts;
528 ino_t ino, ipref; 530 ino_t ino, ipref;
529 int cg, error; 531 int cg, error;
530 532
531 UFS_WAPBL_JUNLOCK_ASSERT(pvp->v_mount); 533 UFS_WAPBL_JUNLOCK_ASSERT(pvp->v_mount);
532 534
533 *vpp = NULL; 535 *vpp = NULL;
534 pip = VTOI(pvp); 536 pip = VTOI(pvp);
535 fs = pip->i_fs; 537 fs = pip->i_fs;
536 ump = pip->i_ump; 538 ump = pip->i_ump;
537 539
538 error = UFS_WAPBL_BEGIN(pvp->v_mount); 540 error = UFS_WAPBL_BEGIN(pvp->v_mount);
539 if (error) { 541 if (error) {
540 return error; 542 return error;
541 } 543 }
542 mutex_enter(&ump->um_lock); 544 mutex_enter(&ump->um_lock);
543 if (fs->fs_cstotal.cs_nifree == 0) 545 if (fs->fs_cstotal.cs_nifree == 0)
544 goto noinodes; 546 goto noinodes;
545 547
546 if ((mode & IFMT) == IFDIR) 548 if ((mode & IFMT) == IFDIR)
547 ipref = ffs_dirpref(pip); 549 ipref = ffs_dirpref(pip);
548 else 550 else
549 ipref = pip->i_number; 551 ipref = pip->i_number;
550 if (ipref >= fs->fs_ncg * fs->fs_ipg) 552 if (ipref >= fs->fs_ncg * fs->fs_ipg)
551 ipref = 0; 553 ipref = 0;
552 cg = ino_to_cg(fs, ipref); 554 cg = ino_to_cg(fs, ipref);
553 /* 555 /*
554 * Track number of dirs created one after another 556 * Track number of dirs created one after another
555 * in a same cg without intervening by files. 557 * in a same cg without intervening by files.
556 */ 558 */
557 if ((mode & IFMT) == IFDIR) { 559 if ((mode & IFMT) == IFDIR) {
558 if (fs->fs_contigdirs[cg] < 255) 560 if (fs->fs_contigdirs[cg] < 255)
559 fs->fs_contigdirs[cg]++; 561 fs->fs_contigdirs[cg]++;
560 } else { 562 } else {
561 if (fs->fs_contigdirs[cg] > 0) 563 if (fs->fs_contigdirs[cg] > 0)
562 fs->fs_contigdirs[cg]--; 564 fs->fs_contigdirs[cg]--;
563 } 565 }
564 ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, 0, ffs_nodealloccg); 566 ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, 0, ffs_nodealloccg);
565 if (ino == 0) 567 if (ino == 0)
566 goto noinodes; 568 goto noinodes;
567 UFS_WAPBL_END(pvp->v_mount); 569 UFS_WAPBL_END(pvp->v_mount);
568 error = VFS_VGET(pvp->v_mount, ino, vpp); 570 error = VFS_VGET(pvp->v_mount, ino, vpp);
569 if (error) { 571 if (error) {
570 int err; 572 int err;
571 err = UFS_WAPBL_BEGIN(pvp->v_mount); 573 err = UFS_WAPBL_BEGIN(pvp->v_mount);
572 if (err == 0) 574 if (err == 0)
573 ffs_vfree(pvp, ino, mode); 575 ffs_vfree(pvp, ino, mode);
574 if (err == 0) 576 if (err == 0)
575 UFS_WAPBL_END(pvp->v_mount); 577 UFS_WAPBL_END(pvp->v_mount);
576 return (error); 578 return (error);
577 } 579 }
578 KASSERT((*vpp)->v_type == VNON); 580 KASSERT((*vpp)->v_type == VNON);
579 ip = VTOI(*vpp); 581 ip = VTOI(*vpp);
580 if (ip->i_mode) { 582 if (ip->i_mode) {
581#if 0 583#if 0
582 printf("mode = 0%o, inum = %d, fs = %s\n", 584 printf("mode = 0%o, inum = %d, fs = %s\n",
583 ip->i_mode, ip->i_number, fs->fs_fsmnt); 585 ip->i_mode, ip->i_number, fs->fs_fsmnt);
584#else 586#else
585 printf("dmode %x mode %x dgen %x gen %x\n", 587 printf("dmode %x mode %x dgen %x gen %x\n",
586 DIP(ip, mode), ip->i_mode, 588 DIP(ip, mode), ip->i_mode,
587 DIP(ip, gen), ip->i_gen); 589 DIP(ip, gen), ip->i_gen);
588 printf("size %llx blocks %llx\n", 590 printf("size %llx blocks %llx\n",
589 (long long)DIP(ip, size), (long long)DIP(ip, blocks)); 591 (long long)DIP(ip, size), (long long)DIP(ip, blocks));
590 printf("ino %llu ipref %llu\n", (unsigned long long)ino, 592 printf("ino %llu ipref %llu\n", (unsigned long long)ino,
591 (unsigned long long)ipref); 593 (unsigned long long)ipref);
592#if 0 594#if 0
593 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)), 595 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
594 (int)fs->fs_bsize, NOCRED, 0, &bp); 596 (int)fs->fs_bsize, NOCRED, 0, &bp);
595#endif 597#endif
596 598
597#endif 599#endif
598 panic("ffs_valloc: dup alloc"); 600 panic("ffs_valloc: dup alloc");
599 } 601 }
600 if (DIP(ip, blocks)) { /* XXX */ 602 if (DIP(ip, blocks)) { /* XXX */
601 printf("free inode %s/%llu had %" PRId64 " blocks\n", 603 printf("free inode %s/%llu had %" PRId64 " blocks\n",
602 fs->fs_fsmnt, (unsigned long long)ino, DIP(ip, blocks)); 604 fs->fs_fsmnt, (unsigned long long)ino, DIP(ip, blocks));
603 DIP_ASSIGN(ip, blocks, 0); 605 DIP_ASSIGN(ip, blocks, 0);
604 } 606 }
605 ip->i_flag &= ~IN_SPACECOUNTED; 607 ip->i_flag &= ~IN_SPACECOUNTED;
606 ip->i_flags = 0; 608 ip->i_flags = 0;
607 DIP_ASSIGN(ip, flags, 0); 609 DIP_ASSIGN(ip, flags, 0);
608 /* 610 /*
609 * Set up a new generation number for this inode. 611 * Set up a new generation number for this inode.
610 */ 612 */
611 ip->i_gen++; 613 ip->i_gen++;
612 DIP_ASSIGN(ip, gen, ip->i_gen); 614 DIP_ASSIGN(ip, gen, ip->i_gen);
613 if (fs->fs_magic == FS_UFS2_MAGIC) { 615 if (fs->fs_magic == FS_UFS2_MAGIC) {
614 vfs_timestamp(&ts); 616 vfs_timestamp(&ts);
615 ip->i_ffs2_birthtime = ts.tv_sec; 617 ip->i_ffs2_birthtime = ts.tv_sec;
616 ip->i_ffs2_birthnsec = ts.tv_nsec; 618 ip->i_ffs2_birthnsec = ts.tv_nsec;
617 } 619 }
618 return (0); 620 return (0);
619noinodes: 621noinodes:
620 mutex_exit(&ump->um_lock); 622 mutex_exit(&ump->um_lock);
621 UFS_WAPBL_END(pvp->v_mount); 623 UFS_WAPBL_END(pvp->v_mount);
622 ffs_fserr(fs, kauth_cred_geteuid(cred), "out of inodes"); 624 ffs_fserr(fs, kauth_cred_geteuid(cred), "out of inodes");
623 uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt); 625 uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
624 return (ENOSPC); 626 return (ENOSPC);
625} 627}
626 628
627/* 629/*
628 * Find a cylinder group in which to place a directory. 630 * Find a cylinder group in which to place a directory.
629 * 631 *
630 * The policy implemented by this algorithm is to allocate a 632 * The policy implemented by this algorithm is to allocate a
631 * directory inode in the same cylinder group as its parent 633 * directory inode in the same cylinder group as its parent
632 * directory, but also to reserve space for its files inodes 634 * directory, but also to reserve space for its files inodes
633 * and data. Restrict the number of directories which may be 635 * and data. Restrict the number of directories which may be
634 * allocated one after another in the same cylinder group 636 * allocated one after another in the same cylinder group
635 * without intervening allocation of files. 637 * without intervening allocation of files.
636 * 638 *
637 * If we allocate a first level directory then force allocation 639 * If we allocate a first level directory then force allocation
638 * in another cylinder group. 640 * in another cylinder group.
639 */ 641 */
640static ino_t 642static ino_t
641ffs_dirpref(struct inode *pip) 643ffs_dirpref(struct inode *pip)
642{ 644{
643 register struct fs *fs; 645 register struct fs *fs;
644 int cg, prefcg; 646 int cg, prefcg;
645 int64_t dirsize, cgsize, curdsz; 647 int64_t dirsize, cgsize, curdsz;
646 int avgifree, avgbfree, avgndir; 648 int avgifree, avgbfree, avgndir;
647 int minifree, minbfree, maxndir; 649 int minifree, minbfree, maxndir;
648 int mincg, minndir; 650 int mincg, minndir;
649 int maxcontigdirs; 651 int maxcontigdirs;
650 652
651 KASSERT(mutex_owned(&pip->i_ump->um_lock)); 653 KASSERT(mutex_owned(&pip->i_ump->um_lock));
652 654
653 fs = pip->i_fs; 655 fs = pip->i_fs;
654 656
655 avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg; 657 avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
656 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; 658 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
657 avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg; 659 avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg;
658 660
659 /* 661 /*
660 * Force allocation in another cg if creating a first level dir. 662 * Force allocation in another cg if creating a first level dir.
661 */ 663 */
662 if (ITOV(pip)->v_vflag & VV_ROOT) { 664 if (ITOV(pip)->v_vflag & VV_ROOT) {
663 prefcg = random() % fs->fs_ncg; 665 prefcg = random() % fs->fs_ncg;
664 mincg = prefcg; 666 mincg = prefcg;
665 minndir = fs->fs_ipg; 667 minndir = fs->fs_ipg;
666 for (cg = prefcg; cg < fs->fs_ncg; cg++) 668 for (cg = prefcg; cg < fs->fs_ncg; cg++)
667 if (fs->fs_cs(fs, cg).cs_ndir < minndir && 669 if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
668 fs->fs_cs(fs, cg).cs_nifree >= avgifree && 670 fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
669 fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { 671 fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
670 mincg = cg; 672 mincg = cg;
671 minndir = fs->fs_cs(fs, cg).cs_ndir; 673 minndir = fs->fs_cs(fs, cg).cs_ndir;
672 } 674 }
673 for (cg = 0; cg < prefcg; cg++) 675 for (cg = 0; cg < prefcg; cg++)
674 if (fs->fs_cs(fs, cg).cs_ndir < minndir && 676 if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
675 fs->fs_cs(fs, cg).cs_nifree >= avgifree && 677 fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
676 fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { 678 fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
677 mincg = cg; 679 mincg = cg;
678 minndir = fs->fs_cs(fs, cg).cs_ndir; 680 minndir = fs->fs_cs(fs, cg).cs_ndir;
679 } 681 }
680 return ((ino_t)(fs->fs_ipg * mincg)); 682 return ((ino_t)(fs->fs_ipg * mincg));
681 } 683 }
682 684
683 /* 685 /*
684 * Count various limits which used for 686 * Count various limits which used for
685 * optimal allocation of a directory inode. 687 * optimal allocation of a directory inode.
686 */ 688 */
687 maxndir = min(avgndir + fs->fs_ipg / 16, fs->fs_ipg); 689 maxndir = min(avgndir + fs->fs_ipg / 16, fs->fs_ipg);
688 minifree = avgifree - fs->fs_ipg / 4; 690 minifree = avgifree - fs->fs_ipg / 4;
689 if (minifree < 0) 691 if (minifree < 0)
690 minifree = 0; 692 minifree = 0;
691 minbfree = avgbfree - fragstoblks(fs, fs->fs_fpg) / 4; 693 minbfree = avgbfree - fragstoblks(fs, fs->fs_fpg) / 4;
692 if (minbfree < 0) 694 if (minbfree < 0)
693 minbfree = 0; 695 minbfree = 0;
694 cgsize = (int64_t)fs->fs_fsize * fs->fs_fpg; 696 cgsize = (int64_t)fs->fs_fsize * fs->fs_fpg;
695 dirsize = (int64_t)fs->fs_avgfilesize * fs->fs_avgfpdir; 697 dirsize = (int64_t)fs->fs_avgfilesize * fs->fs_avgfpdir;
696 if (avgndir != 0) { 698 if (avgndir != 0) {
697 curdsz = (cgsize - (int64_t)avgbfree * fs->fs_bsize) / avgndir; 699 curdsz = (cgsize - (int64_t)avgbfree * fs->fs_bsize) / avgndir;
698 if (dirsize < curdsz) 700 if (dirsize < curdsz)
699 dirsize = curdsz; 701 dirsize = curdsz;
700 } 702 }
701 if (cgsize < dirsize * 255) 703 if (cgsize < dirsize * 255)
702 maxcontigdirs = cgsize / dirsize; 704 maxcontigdirs = cgsize / dirsize;
703 else 705 else
704 maxcontigdirs = 255; 706 maxcontigdirs = 255;
705 if (fs->fs_avgfpdir > 0) 707 if (fs->fs_avgfpdir > 0)
706 maxcontigdirs = min(maxcontigdirs, 708 maxcontigdirs = min(maxcontigdirs,
707 fs->fs_ipg / fs->fs_avgfpdir); 709 fs->fs_ipg / fs->fs_avgfpdir);
708 if (maxcontigdirs == 0) 710 if (maxcontigdirs == 0)
709 maxcontigdirs = 1; 711 maxcontigdirs = 1;
710 712
711 /* 713 /*
712 * Limit number of dirs in one cg and reserve space for 714 * Limit number of dirs in one cg and reserve space for
713 * regular files, but only if we have no deficit in 715 * regular files, but only if we have no deficit in
714 * inodes or space. 716 * inodes or space.
715 */ 717 */
716 prefcg = ino_to_cg(fs, pip->i_number); 718 prefcg = ino_to_cg(fs, pip->i_number);
717 for (cg = prefcg; cg < fs->fs_ncg; cg++) 719 for (cg = prefcg; cg < fs->fs_ncg; cg++)
718 if (fs->fs_cs(fs, cg).cs_ndir < maxndir && 720 if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
719 fs->fs_cs(fs, cg).cs_nifree >= minifree && 721 fs->fs_cs(fs, cg).cs_nifree >= minifree &&
720 fs->fs_cs(fs, cg).cs_nbfree >= minbfree) { 722 fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
721 if (fs->fs_contigdirs[cg] < maxcontigdirs) 723 if (fs->fs_contigdirs[cg] < maxcontigdirs)
722 return ((ino_t)(fs->fs_ipg * cg)); 724 return ((ino_t)(fs->fs_ipg * cg));
723 } 725 }
724 for (cg = 0; cg < prefcg; cg++) 726 for (cg = 0; cg < prefcg; cg++)
725 if (fs->fs_cs(fs, cg).cs_ndir < maxndir && 727 if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
726 fs->fs_cs(fs, cg).cs_nifree >= minifree && 728 fs->fs_cs(fs, cg).cs_nifree >= minifree &&
727 fs->fs_cs(fs, cg).cs_nbfree >= minbfree) { 729 fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
728 if (fs->fs_contigdirs[cg] < maxcontigdirs) 730 if (fs->fs_contigdirs[cg] < maxcontigdirs)
729 return ((ino_t)(fs->fs_ipg * cg)); 731 return ((ino_t)(fs->fs_ipg * cg));
730 } 732 }
731 /* 733 /*
732 * This is a backstop when we are deficient in space. 734 * This is a backstop when we are deficient in space.
733 */ 735 */
734 for (cg = prefcg; cg < fs->fs_ncg; cg++) 736 for (cg = prefcg; cg < fs->fs_ncg; cg++)
735 if (fs->fs_cs(fs, cg).cs_nifree >= avgifree) 737 if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
736 return ((ino_t)(fs->fs_ipg * cg)); 738 return ((ino_t)(fs->fs_ipg * cg));
737 for (cg = 0; cg < prefcg; cg++) 739 for (cg = 0; cg < prefcg; cg++)
738 if (fs->fs_cs(fs, cg).cs_nifree >= avgifree) 740 if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
739 break; 741 break;
740 return ((ino_t)(fs->fs_ipg * cg)); 742 return ((ino_t)(fs->fs_ipg * cg));
741} 743}
742 744
743/* 745/*
744 * Select the desired position for the next block in a file. The file is 746 * Select the desired position for the next block in a file. The file is
745 * logically divided into sections. The first section is composed of the 747 * logically divided into sections. The first section is composed of the
746 * direct blocks. Each additional section contains fs_maxbpg blocks. 748 * direct blocks. Each additional section contains fs_maxbpg blocks.
747 * 749 *
748 * If no blocks have been allocated in the first section, the policy is to 750 * If no blocks have been allocated in the first section, the policy is to
749 * request a block in the same cylinder group as the inode that describes 751 * request a block in the same cylinder group as the inode that describes
750 * the file. If no blocks have been allocated in any other section, the 752 * the file. If no blocks have been allocated in any other section, the
751 * policy is to place the section in a cylinder group with a greater than 753 * policy is to place the section in a cylinder group with a greater than
752 * average number of free blocks. An appropriate cylinder group is found 754 * average number of free blocks. An appropriate cylinder group is found
753 * by using a rotor that sweeps the cylinder groups. When a new group of 755 * by using a rotor that sweeps the cylinder groups. When a new group of
754 * blocks is needed, the sweep begins in the cylinder group following the 756 * blocks is needed, the sweep begins in the cylinder group following the
755 * cylinder group from which the previous allocation was made. The sweep 757 * cylinder group from which the previous allocation was made. The sweep
756 * continues until a cylinder group with greater than the average number 758 * continues until a cylinder group with greater than the average number
757 * of free blocks is found. If the allocation is for the first block in an 759 * of free blocks is found. If the allocation is for the first block in an
758 * indirect block, the information on the previous allocation is unavailable; 760 * indirect block, the information on the previous allocation is unavailable;
759 * here a best guess is made based upon the logical block number being 761 * here a best guess is made based upon the logical block number being
760 * allocated. 762 * allocated.
761 * 763 *
762 * If a section is already partially allocated, the policy is to 764 * If a section is already partially allocated, the policy is to
763 * contiguously allocate fs_maxcontig blocks. The end of one of these 765 * contiguously allocate fs_maxcontig blocks. The end of one of these
764 * contiguous blocks and the beginning of the next is laid out 766 * contiguous blocks and the beginning of the next is laid out
765 * contigously if possible. 767 * contigously if possible.
766 * 768 *
767 * => um_lock held on entry and exit 769 * => um_lock held on entry and exit
768 */ 770 */
769daddr_t 771daddr_t
770ffs_blkpref_ufs1(struct inode *ip, daddr_t lbn, int indx, int flags, 772ffs_blkpref_ufs1(struct inode *ip, daddr_t lbn, int indx, int flags,
771 int32_t *bap /* XXX ondisk32 */) 773 int32_t *bap /* XXX ondisk32 */)
772{ 774{
773 struct fs *fs; 775 struct fs *fs;
774 int cg; 776 int cg;
775 int avgbfree, startcg; 777 int avgbfree, startcg;
776 778
777 KASSERT(mutex_owned(&ip->i_ump->um_lock)); 779 KASSERT(mutex_owned(&ip->i_ump->um_lock));
778 780
779 fs = ip->i_fs; 781 fs = ip->i_fs;
780 782
781 /* 783 /*
782 * If allocating a contiguous file with B_CONTIG, use the hints 784 * If allocating a contiguous file with B_CONTIG, use the hints
783 * in the inode extentions to return the desired block. 785 * in the inode extentions to return the desired block.
784 * 786 *
785 * For metadata (indirect blocks) return the address of where 787 * For metadata (indirect blocks) return the address of where
786 * the first indirect block resides - we'll scan for the next 788 * the first indirect block resides - we'll scan for the next
787 * available slot if we need to allocate more than one indirect 789 * available slot if we need to allocate more than one indirect
788 * block. For data, return the address of the actual block 790 * block. For data, return the address of the actual block
789 * relative to the address of the first data block. 791 * relative to the address of the first data block.
790 */ 792 */
791 if (flags & B_CONTIG) { 793 if (flags & B_CONTIG) {
792 KASSERT(ip->i_ffs_first_data_blk != 0); 794 KASSERT(ip->i_ffs_first_data_blk != 0);
793 KASSERT(ip->i_ffs_first_indir_blk != 0); 795 KASSERT(ip->i_ffs_first_indir_blk != 0);
794 if (flags & B_METAONLY) 796 if (flags & B_METAONLY)
795 return ip->i_ffs_first_indir_blk; 797 return ip->i_ffs_first_indir_blk;
796 else 798 else
797 return ip->i_ffs_first_data_blk + blkstofrags(fs, lbn); 799 return ip->i_ffs_first_data_blk + blkstofrags(fs, lbn);
798 } 800 }
799 801
800 if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) { 802 if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
801 if (lbn < NDADDR + NINDIR(fs)) { 803 if (lbn < NDADDR + NINDIR(fs)) {
802 cg = ino_to_cg(fs, ip->i_number); 804 cg = ino_to_cg(fs, ip->i_number);
803 return (cgbase(fs, cg) + fs->fs_frag); 805 return (cgbase(fs, cg) + fs->fs_frag);
804 } 806 }
805 /* 807 /*
806 * Find a cylinder with greater than average number of 808 * Find a cylinder with greater than average number of
807 * unused data blocks. 809 * unused data blocks.
808 */ 810 */
809 if (indx == 0 || bap[indx - 1] == 0) 811 if (indx == 0 || bap[indx - 1] == 0)
810 startcg = 812 startcg =
811 ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg; 813 ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
812 else 814 else
813 startcg = dtog(fs, 815 startcg = dtog(fs,
814 ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1); 816 ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
815 startcg %= fs->fs_ncg; 817 startcg %= fs->fs_ncg;
816 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; 818 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
817 for (cg = startcg; cg < fs->fs_ncg; cg++) 819 for (cg = startcg; cg < fs->fs_ncg; cg++)
818 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { 820 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
819 return (cgbase(fs, cg) + fs->fs_frag); 821 return (cgbase(fs, cg) + fs->fs_frag);
820 } 822 }
821 for (cg = 0; cg < startcg; cg++) 823 for (cg = 0; cg < startcg; cg++)
822 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { 824 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
823 return (cgbase(fs, cg) + fs->fs_frag); 825 return (cgbase(fs, cg) + fs->fs_frag);
824 } 826 }
825 return (0); 827 return (0);
826 } 828 }
827 /* 829 /*
828 * We just always try to lay things out contiguously. 830 * We just always try to lay things out contiguously.
829 */ 831 */
830 return ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag; 832 return ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
831} 833}
832 834
833daddr_t 835daddr_t
834ffs_blkpref_ufs2(struct inode *ip, daddr_t lbn, int indx, int flags, 836ffs_blkpref_ufs2(struct inode *ip, daddr_t lbn, int indx, int flags,
835 int64_t *bap) 837 int64_t *bap)
836{ 838{
837 struct fs *fs; 839 struct fs *fs;
838 int cg; 840 int cg;
839 int avgbfree, startcg; 841 int avgbfree, startcg;
840 842
841 KASSERT(mutex_owned(&ip->i_ump->um_lock)); 843 KASSERT(mutex_owned(&ip->i_ump->um_lock));
842 844
843 fs = ip->i_fs; 845 fs = ip->i_fs;
844 846
845 /* 847 /*
846 * If allocating a contiguous file with B_CONTIG, use the hints 848 * If allocating a contiguous file with B_CONTIG, use the hints
847 * in the inode extentions to return the desired block. 849 * in the inode extentions to return the desired block.
848 * 850 *
849 * For metadata (indirect blocks) return the address of where 851 * For metadata (indirect blocks) return the address of where
850 * the first indirect block resides - we'll scan for the next 852 * the first indirect block resides - we'll scan for the next
851 * available slot if we need to allocate more than one indirect 853 * available slot if we need to allocate more than one indirect
852 * block. For data, return the address of the actual block 854 * block. For data, return the address of the actual block
853 * relative to the address of the first data block. 855 * relative to the address of the first data block.
854 */ 856 */
855 if (flags & B_CONTIG) { 857 if (flags & B_CONTIG) {
856 KASSERT(ip->i_ffs_first_data_blk != 0); 858 KASSERT(ip->i_ffs_first_data_blk != 0);
857 KASSERT(ip->i_ffs_first_indir_blk != 0); 859 KASSERT(ip->i_ffs_first_indir_blk != 0);
858 if (flags & B_METAONLY) 860 if (flags & B_METAONLY)
859 return ip->i_ffs_first_indir_blk; 861 return ip->i_ffs_first_indir_blk;
860 else 862 else
861 return ip->i_ffs_first_data_blk + blkstofrags(fs, lbn); 863 return ip->i_ffs_first_data_blk + blkstofrags(fs, lbn);
862 } 864 }
863 865
864 if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) { 866 if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
865 if (lbn < NDADDR + NINDIR(fs)) { 867 if (lbn < NDADDR + NINDIR(fs)) {
866 cg = ino_to_cg(fs, ip->i_number); 868 cg = ino_to_cg(fs, ip->i_number);
867 return (cgbase(fs, cg) + fs->fs_frag); 869 return (cgbase(fs, cg) + fs->fs_frag);
868 } 870 }
869 /* 871 /*
870 * Find a cylinder with greater than average number of 872 * Find a cylinder with greater than average number of
871 * unused data blocks. 873 * unused data blocks.
872 */ 874 */
873 if (indx == 0 || bap[indx - 1] == 0) 875 if (indx == 0 || bap[indx - 1] == 0)
874 startcg = 876 startcg =
875 ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg; 877 ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
876 else 878 else
877 startcg = dtog(fs, 879 startcg = dtog(fs,
878 ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1); 880 ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
879 startcg %= fs->fs_ncg; 881 startcg %= fs->fs_ncg;
880 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; 882 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
881 for (cg = startcg; cg < fs->fs_ncg; cg++) 883 for (cg = startcg; cg < fs->fs_ncg; cg++)
882 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { 884 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
883 return (cgbase(fs, cg) + fs->fs_frag); 885 return (cgbase(fs, cg) + fs->fs_frag);
884 } 886 }
885 for (cg = 0; cg < startcg; cg++) 887 for (cg = 0; cg < startcg; cg++)
886 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { 888 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
887 return (cgbase(fs, cg) + fs->fs_frag); 889 return (cgbase(fs, cg) + fs->fs_frag);
888 } 890 }
889 return (0); 891 return (0);
890 } 892 }
891 /* 893 /*
892 * We just always try to lay things out contiguously. 894 * We just always try to lay things out contiguously.
893 */ 895 */
894 return ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag; 896 return ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
895} 897}
896 898
897 899
898/* 900/*
899 * Implement the cylinder overflow algorithm. 901 * Implement the cylinder overflow algorithm.
900 * 902 *
901 * The policy implemented by this algorithm is: 903 * The policy implemented by this algorithm is:
902 * 1) allocate the block in its requested cylinder group. 904 * 1) allocate the block in its requested cylinder group.
903 * 2) quadradically rehash on the cylinder group number. 905 * 2) quadradically rehash on the cylinder group number.
904 * 3) brute force search for a free block. 906 * 3) brute force search for a free block.
905 * 907 *
906 * => called with um_lock held 908 * => called with um_lock held
907 * => returns with um_lock released on success, held on failure 909 * => returns with um_lock released on success, held on failure
908 * (*allocator releases lock on success, retains lock on failure) 910 * (*allocator releases lock on success, retains lock on failure)
909 */ 911 */
910/*VARARGS5*/ 912/*VARARGS5*/
911static daddr_t 913static daddr_t
912ffs_hashalloc(struct inode *ip, int cg, daddr_t pref, 914ffs_hashalloc(struct inode *ip, int cg, daddr_t pref,
913 int size /* size for data blocks, mode for inodes */, 915 int size /* size for data blocks, mode for inodes */,
914 int flags, daddr_t (*allocator)(struct inode *, int, daddr_t, int, int)) 916 int flags, daddr_t (*allocator)(struct inode *, int, daddr_t, int, int))
915{ 917{
916 struct fs *fs; 918 struct fs *fs;
917 daddr_t result; 919 daddr_t result;
918 int i, icg = cg; 920 int i, icg = cg;
919 921
920 fs = ip->i_fs; 922 fs = ip->i_fs;
921 /* 923 /*
922 * 1: preferred cylinder group 924 * 1: preferred cylinder group
923 */ 925 */
924 result = (*allocator)(ip, cg, pref, size, flags); 926 result = (*allocator)(ip, cg, pref, size, flags);
925 if (result) 927 if (result)
926 return (result); 928 return (result);
927 929
928 if (flags & B_CONTIG) 930 if (flags & B_CONTIG)
929 return (result); 931 return (result);
930 /* 932 /*
931 * 2: quadratic rehash 933 * 2: quadratic rehash
932 */ 934 */
933 for (i = 1; i < fs->fs_ncg; i *= 2) { 935 for (i = 1; i < fs->fs_ncg; i *= 2) {
934 cg += i; 936 cg += i;
935 if (cg >= fs->fs_ncg) 937 if (cg >= fs->fs_ncg)
936 cg -= fs->fs_ncg; 938 cg -= fs->fs_ncg;
937 result = (*allocator)(ip, cg, 0, size, flags); 939 result = (*allocator)(ip, cg, 0, size, flags);
938 if (result) 940 if (result)
939 return (result); 941 return (result);
940 } 942 }
941 /* 943 /*
942 * 3: brute force search 944 * 3: brute force search
943 * Note that we start at i == 2, since 0 was checked initially, 945 * Note that we start at i == 2, since 0 was checked initially,
944 * and 1 is always checked in the quadratic rehash. 946 * and 1 is always checked in the quadratic rehash.
945 */ 947 */
946 cg = (icg + 2) % fs->fs_ncg; 948 cg = (icg + 2) % fs->fs_ncg;
947 for (i = 2; i < fs->fs_ncg; i++) { 949 for (i = 2; i < fs->fs_ncg; i++) {
948 result = (*allocator)(ip, cg, 0, size, flags); 950 result = (*allocator)(ip, cg, 0, size, flags);
949 if (result) 951 if (result)
950 return (result); 952 return (result);
951 cg++; 953 cg++;
952 if (cg == fs->fs_ncg) 954 if (cg == fs->fs_ncg)
953 cg = 0; 955 cg = 0;
954 } 956 }
955 return (0); 957 return (0);
956} 958}
957 959
958/* 960/*
959 * Determine whether a fragment can be extended. 961 * Determine whether a fragment can be extended.
960 * 962 *
961 * Check to see if the necessary fragments are available, and 963 * Check to see if the necessary fragments are available, and
962 * if they are, allocate them. 964 * if they are, allocate them.
963 * 965 *
964 * => called with um_lock held 966 * => called with um_lock held
965 * => returns with um_lock released on success, held on failure 967 * => returns with um_lock released on success, held on failure
966 */ 968 */
967static daddr_t 969static daddr_t
968ffs_fragextend(struct inode *ip, int cg, daddr_t bprev, int osize, int nsize) 970ffs_fragextend(struct inode *ip, int cg, daddr_t bprev, int osize, int nsize)
969{ 971{
970 struct ufsmount *ump; 972 struct ufsmount *ump;
971 struct fs *fs; 973 struct fs *fs;
972 struct cg *cgp; 974 struct cg *cgp;
973 struct buf *bp; 975 struct buf *bp;
974 daddr_t bno; 976 daddr_t bno;
975 int frags, bbase; 977 int frags, bbase;
976 int i, error; 978 int i, error;
977 u_int8_t *blksfree; 979 u_int8_t *blksfree;
978 980
979 fs = ip->i_fs; 981 fs = ip->i_fs;
980 ump = ip->i_ump; 982 ump = ip->i_ump;
981 983
982 KASSERT(mutex_owned(&ump->um_lock)); 984 KASSERT(mutex_owned(&ump->um_lock));
983 985
984 if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize)) 986 if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize))
985 return (0); 987 return (0);
986 frags = numfrags(fs, nsize); 988 frags = numfrags(fs, nsize);
987 bbase = fragnum(fs, bprev); 989 bbase = fragnum(fs, bprev);
988 if (bbase > fragnum(fs, (bprev + frags - 1))) { 990 if (bbase > fragnum(fs, (bprev + frags - 1))) {
989 /* cannot extend across a block boundary */ 991 /* cannot extend across a block boundary */
990 return (0); 992 return (0);
991 } 993 }
992 mutex_exit(&ump->um_lock); 994 mutex_exit(&ump->um_lock);
993 error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), 995 error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
994 (int)fs->fs_cgsize, NOCRED, B_MODIFY, &bp); 996 (int)fs->fs_cgsize, NOCRED, B_MODIFY, &bp);
995 if (error) 997 if (error)
996 goto fail; 998 goto fail;
997 cgp = (struct cg *)bp->b_data; 999 cgp = (struct cg *)bp->b_data;
998 if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) 1000 if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs)))
999 goto fail; 1001 goto fail;
1000 cgp->cg_old_time = ufs_rw32(time_second, UFS_FSNEEDSWAP(fs)); 1002 cgp->cg_old_time = ufs_rw32(time_second, UFS_FSNEEDSWAP(fs));
1001 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1003 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1002 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1004 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1003 cgp->cg_time = ufs_rw64(time_second, UFS_FSNEEDSWAP(fs)); 1005 cgp->cg_time = ufs_rw64(time_second, UFS_FSNEEDSWAP(fs));
1004 bno = dtogd(fs, bprev); 1006 bno = dtogd(fs, bprev);
1005 blksfree = cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)); 1007 blksfree = cg_blksfree(cgp, UFS_FSNEEDSWAP(fs));
1006 for (i = numfrags(fs, osize); i < frags; i++) 1008 for (i = numfrags(fs, osize); i < frags; i++)
1007 if (isclr(blksfree, bno + i)) 1009 if (isclr(blksfree, bno + i))
1008 goto fail; 1010 goto fail;
1009 /* 1011 /*
1010 * the current fragment can be extended 1012 * the current fragment can be extended
1011 * deduct the count on fragment being extended into 1013 * deduct the count on fragment being extended into
1012 * increase the count on the remaining fragment (if any) 1014 * increase the count on the remaining fragment (if any)
1013 * allocate the extended piece 1015 * allocate the extended piece
1014 */ 1016 */
1015 for (i = frags; i < fs->fs_frag - bbase; i++) 1017 for (i = frags; i < fs->fs_frag - bbase; i++)
1016 if (isclr(blksfree, bno + i)) 1018 if (isclr(blksfree, bno + i))
1017 break; 1019 break;
1018 ufs_add32(cgp->cg_frsum[i - numfrags(fs, osize)], -1, UFS_FSNEEDSWAP(fs)); 1020 ufs_add32(cgp->cg_frsum[i - numfrags(fs, osize)], -1, UFS_FSNEEDSWAP(fs));
1019 if (i != frags) 1021 if (i != frags)
1020 ufs_add32(cgp->cg_frsum[i - frags], 1, UFS_FSNEEDSWAP(fs)); 1022 ufs_add32(cgp->cg_frsum[i - frags], 1, UFS_FSNEEDSWAP(fs));
1021 mutex_enter(&ump->um_lock); 1023 mutex_enter(&ump->um_lock);
1022 for (i = numfrags(fs, osize); i < frags; i++) { 1024 for (i = numfrags(fs, osize); i < frags; i++) {
1023 clrbit(blksfree, bno + i); 1025 clrbit(blksfree, bno + i);
1024 ufs_add32(cgp->cg_cs.cs_nffree, -1, UFS_FSNEEDSWAP(fs)); 1026 ufs_add32(cgp->cg_cs.cs_nffree, -1, UFS_FSNEEDSWAP(fs));
1025 fs->fs_cstotal.cs_nffree--; 1027 fs->fs_cstotal.cs_nffree--;
1026 fs->fs_cs(fs, cg).cs_nffree--; 1028 fs->fs_cs(fs, cg).cs_nffree--;
1027 } 1029 }
1028 fs->fs_fmod = 1; 1030 fs->fs_fmod = 1;
1029 ACTIVECG_CLR(fs, cg); 1031 ACTIVECG_CLR(fs, cg);
1030 mutex_exit(&ump->um_lock); 1032 mutex_exit(&ump->um_lock);
1031 if (DOINGSOFTDEP(ITOV(ip))) 1033 if (DOINGSOFTDEP(ITOV(ip)))
1032 softdep_setup_blkmapdep(bp, fs, bprev); 1034 softdep_setup_blkmapdep(bp, fs, bprev);
1033 bdwrite(bp); 1035 bdwrite(bp);
1034 return (bprev); 1036 return (bprev);
1035 1037
1036 fail: 1038 fail:
1037 brelse(bp, 0); 1039 brelse(bp, 0);
1038 mutex_enter(&ump->um_lock); 1040 mutex_enter(&ump->um_lock);
1039 return (0); 1041 return (0);
1040} 1042}
1041 1043
1042/* 1044/*
1043 * Determine whether a block can be allocated. 1045 * Determine whether a block can be allocated.
1044 * 1046 *
1045 * Check to see if a block of the appropriate size is available, 1047 * Check to see if a block of the appropriate size is available,
1046 * and if it is, allocate it. 1048 * and if it is, allocate it.
1047 */ 1049 */
1048static daddr_t 1050static daddr_t
1049ffs_alloccg(struct inode *ip, int cg, daddr_t bpref, int size, int flags) 1051ffs_alloccg(struct inode *ip, int cg, daddr_t bpref, int size, int flags)
1050{ 1052{
1051 struct ufsmount *ump; 1053 struct ufsmount *ump;
1052 struct fs *fs = ip->i_fs; 1054 struct fs *fs = ip->i_fs;
1053 struct cg *cgp; 1055 struct cg *cgp;
1054 struct buf *bp; 1056 struct buf *bp;
1055 int32_t bno; 1057 int32_t bno;
1056 daddr_t blkno; 1058 daddr_t blkno;
1057 int error, frags, allocsiz, i; 1059 int error, frags, allocsiz, i;
1058 u_int8_t *blksfree; 1060 u_int8_t *blksfree;
1059#ifdef FFS_EI 1061#ifdef FFS_EI
1060 const int needswap = UFS_FSNEEDSWAP(fs); 1062 const int needswap = UFS_FSNEEDSWAP(fs);
1061#endif 1063#endif
1062 1064
1063 ump = ip->i_ump; 1065 ump = ip->i_ump;
1064 1066
1065 KASSERT(mutex_owned(&ump->um_lock)); 1067 KASSERT(mutex_owned(&ump->um_lock));
1066 1068
1067 if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize) 1069 if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
1068 return (0); 1070 return (0);
1069 mutex_exit(&ump->um_lock); 1071 mutex_exit(&ump->um_lock);
1070 error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), 1072 error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
1071 (int)fs->fs_cgsize, NOCRED, B_MODIFY, &bp); 1073 (int)fs->fs_cgsize, NOCRED, B_MODIFY, &bp);
1072 if (error) 1074 if (error)
1073 goto fail; 1075 goto fail;
1074 cgp = (struct cg *)bp->b_data; 1076 cgp = (struct cg *)bp->b_data;
1075 if (!cg_chkmagic(cgp, needswap) || 1077 if (!cg_chkmagic(cgp, needswap) ||
1076 (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) 1078 (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize))
1077 goto fail; 1079 goto fail;
1078 cgp->cg_old_time = ufs_rw32(time_second, needswap); 1080 cgp->cg_old_time = ufs_rw32(time_second, needswap);
1079 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1081 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1080 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1082 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1081 cgp->cg_time = ufs_rw64(time_second, needswap); 1083 cgp->cg_time = ufs_rw64(time_second, needswap);
1082 if (size == fs->fs_bsize) { 1084 if (size == fs->fs_bsize) {
1083 mutex_enter(&ump->um_lock); 1085 mutex_enter(&ump->um_lock);
1084 blkno = ffs_alloccgblk(ip, bp, bpref, flags); 1086 blkno = ffs_alloccgblk(ip, bp, bpref, flags);
1085 ACTIVECG_CLR(fs, cg); 1087 ACTIVECG_CLR(fs, cg);
1086 mutex_exit(&ump->um_lock); 1088 mutex_exit(&ump->um_lock);
1087 bdwrite(bp); 1089 bdwrite(bp);
1088 return (blkno); 1090 return (blkno);
1089 } 1091 }
1090 /* 1092 /*
1091 * check to see if any fragments are already available 1093 * check to see if any fragments are already available
1092 * allocsiz is the size which will be allocated, hacking 1094 * allocsiz is the size which will be allocated, hacking
1093 * it down to a smaller size if necessary 1095 * it down to a smaller size if necessary
1094 */ 1096 */
1095 blksfree = cg_blksfree(cgp, needswap); 1097 blksfree = cg_blksfree(cgp, needswap);
1096 frags = numfrags(fs, size); 1098 frags = numfrags(fs, size);
1097 for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++) 1099 for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
1098 if (cgp->cg_frsum[allocsiz] != 0) 1100 if (cgp->cg_frsum[allocsiz] != 0)
1099 break; 1101 break;
1100 if (allocsiz == fs->fs_frag) { 1102 if (allocsiz == fs->fs_frag) {
1101 /* 1103 /*
1102 * no fragments were available, so a block will be 1104 * no fragments were available, so a block will be
1103 * allocated, and hacked up 1105 * allocated, and hacked up
1104 */ 1106 */
1105 if (cgp->cg_cs.cs_nbfree == 0) 1107 if (cgp->cg_cs.cs_nbfree == 0)
1106 goto fail; 1108 goto fail;
1107 mutex_enter(&ump->um_lock); 1109 mutex_enter(&ump->um_lock);
1108 blkno = ffs_alloccgblk(ip, bp, bpref, flags); 1110 blkno = ffs_alloccgblk(ip, bp, bpref, flags);
1109 bno = dtogd(fs, blkno); 1111 bno = dtogd(fs, blkno);
1110 for (i = frags; i < fs->fs_frag; i++) 1112 for (i = frags; i < fs->fs_frag; i++)
1111 setbit(blksfree, bno + i); 1113 setbit(blksfree, bno + i);
1112 i = fs->fs_frag - frags; 1114 i = fs->fs_frag - frags;
1113 ufs_add32(cgp->cg_cs.cs_nffree, i, needswap); 1115 ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
1114 fs->fs_cstotal.cs_nffree += i; 1116 fs->fs_cstotal.cs_nffree += i;
1115 fs->fs_cs(fs, cg).cs_nffree += i; 1117 fs->fs_cs(fs, cg).cs_nffree += i;
1116 fs->fs_fmod = 1; 1118 fs->fs_fmod = 1;
1117 ufs_add32(cgp->cg_frsum[i], 1, needswap); 1119 ufs_add32(cgp->cg_frsum[i], 1, needswap);
1118 ACTIVECG_CLR(fs, cg); 1120 ACTIVECG_CLR(fs, cg);
1119 mutex_exit(&ump->um_lock); 1121 mutex_exit(&ump->um_lock);
1120 bdwrite(bp); 1122 bdwrite(bp);
1121 return (blkno); 1123 return (blkno);
1122 } 1124 }
1123 bno = ffs_mapsearch(fs, cgp, bpref, allocsiz); 1125 bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
1124#if 0 1126#if 0
1125 /* 1127 /*
1126 * XXX fvdl mapsearch will panic, and never return -1 1128 * XXX fvdl mapsearch will panic, and never return -1
1127 * also: returning NULL as daddr_t ? 1129 * also: returning NULL as daddr_t ?
1128 */ 1130 */
1129 if (bno < 0) 1131 if (bno < 0)
1130 goto fail; 1132 goto fail;
1131#endif 1133#endif
1132 for (i = 0; i < frags; i++) 1134 for (i = 0; i < frags; i++)
1133 clrbit(blksfree, bno + i); 1135 clrbit(blksfree, bno + i);
1134 mutex_enter(&ump->um_lock); 1136 mutex_enter(&ump->um_lock);
1135 ufs_add32(cgp->cg_cs.cs_nffree, -frags, needswap); 1137 ufs_add32(cgp->cg_cs.cs_nffree, -frags, needswap);
1136 fs->fs_cstotal.cs_nffree -= frags; 1138 fs->fs_cstotal.cs_nffree -= frags;
1137 fs->fs_cs(fs, cg).cs_nffree -= frags; 1139 fs->fs_cs(fs, cg).cs_nffree -= frags;
1138 fs->fs_fmod = 1; 1140 fs->fs_fmod = 1;
1139 ufs_add32(cgp->cg_frsum[allocsiz], -1, needswap); 1141 ufs_add32(cgp->cg_frsum[allocsiz], -1, needswap);
1140 if (frags != allocsiz) 1142 if (frags != allocsiz)
1141 ufs_add32(cgp->cg_frsum[allocsiz - frags], 1, needswap); 1143 ufs_add32(cgp->cg_frsum[allocsiz - frags], 1, needswap);
1142 blkno = cg * fs->fs_fpg + bno; 1144 blkno = cg * fs->fs_fpg + bno;
1143 ACTIVECG_CLR(fs, cg); 1145 ACTIVECG_CLR(fs, cg);
1144 mutex_exit(&ump->um_lock); 1146 mutex_exit(&ump->um_lock);
1145 if (DOINGSOFTDEP(ITOV(ip))) 1147 if (DOINGSOFTDEP(ITOV(ip)))
1146 softdep_setup_blkmapdep(bp, fs, blkno); 1148 softdep_setup_blkmapdep(bp, fs, blkno);
1147 bdwrite(bp); 1149 bdwrite(bp);
1148 return blkno; 1150 return blkno;
1149 1151
1150 fail: 1152 fail:
1151 brelse(bp, 0); 1153 brelse(bp, 0);
1152 mutex_enter(&ump->um_lock); 1154 mutex_enter(&ump->um_lock);
1153 return (0); 1155 return (0);
1154} 1156}
1155 1157
1156/* 1158/*
1157 * Allocate a block in a cylinder group. 1159 * Allocate a block in a cylinder group.
1158 * 1160 *
1159 * This algorithm implements the following policy: 1161 * This algorithm implements the following policy:
1160 * 1) allocate the requested block. 1162 * 1) allocate the requested block.
1161 * 2) allocate a rotationally optimal block in the same cylinder. 1163 * 2) allocate a rotationally optimal block in the same cylinder.
1162 * 3) allocate the next available block on the block rotor for the 1164 * 3) allocate the next available block on the block rotor for the
1163 * specified cylinder group. 1165 * specified cylinder group.
1164 * Note that this routine only allocates fs_bsize blocks; these 1166 * Note that this routine only allocates fs_bsize blocks; these
1165 * blocks may be fragmented by the routine that allocates them. 1167 * blocks may be fragmented by the routine that allocates them.
1166 */ 1168 */
1167static daddr_t 1169static daddr_t
1168ffs_alloccgblk(struct inode *ip, struct buf *bp, daddr_t bpref, int flags) 1170ffs_alloccgblk(struct inode *ip, struct buf *bp, daddr_t bpref, int flags)
1169{ 1171{
1170 struct ufsmount *ump; 1172 struct ufsmount *ump;
1171 struct fs *fs = ip->i_fs; 1173 struct fs *fs = ip->i_fs;
1172 struct cg *cgp; 1174 struct cg *cgp;
1173 daddr_t blkno; 1175 daddr_t blkno;
1174 int32_t bno; 1176 int32_t bno;
1175 u_int8_t *blksfree; 1177 u_int8_t *blksfree;
1176#ifdef FFS_EI 1178#ifdef FFS_EI
1177 const int needswap = UFS_FSNEEDSWAP(fs); 1179 const int needswap = UFS_FSNEEDSWAP(fs);
1178#endif 1180#endif
1179 1181
1180 ump = ip->i_ump; 1182 ump = ip->i_ump;
1181 1183
1182 KASSERT(mutex_owned(&ump->um_lock)); 1184 KASSERT(mutex_owned(&ump->um_lock));
1183 1185
1184 cgp = (struct cg *)bp->b_data; 1186 cgp = (struct cg *)bp->b_data;
1185 blksfree = cg_blksfree(cgp, needswap); 1187 blksfree = cg_blksfree(cgp, needswap);
1186 if (bpref == 0 || dtog(fs, bpref) != ufs_rw32(cgp->cg_cgx, needswap)) { 1188 if (bpref == 0 || dtog(fs, bpref) != ufs_rw32(cgp->cg_cgx, needswap)) {
1187 bpref = ufs_rw32(cgp->cg_rotor, needswap); 1189 bpref = ufs_rw32(cgp->cg_rotor, needswap);
1188 } else { 1190 } else {
1189 bpref = blknum(fs, bpref); 1191 bpref = blknum(fs, bpref);
1190 bno = dtogd(fs, bpref); 1192 bno = dtogd(fs, bpref);
1191 /* 1193 /*
1192 * if the requested block is available, use it 1194 * if the requested block is available, use it
1193 */ 1195 */
1194 if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno))) 1196 if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno)))
1195 goto gotit; 1197 goto gotit;
1196 /* 1198 /*
1197 * if the requested data block isn't available and we are 1199 * if the requested data block isn't available and we are
1198 * trying to allocate a contiguous file, return an error. 1200 * trying to allocate a contiguous file, return an error.
1199 */ 1201 */
1200 if ((flags & (B_CONTIG | B_METAONLY)) == B_CONTIG) 1202 if ((flags & (B_CONTIG | B_METAONLY)) == B_CONTIG)
1201 return (0); 1203 return (0);
1202 } 1204 }
1203 1205
1204 /* 1206 /*
1205 * Take the next available block in this cylinder group. 1207 * Take the next available block in this cylinder group.
1206 */ 1208 */
1207 bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag); 1209 bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
1208 if (bno < 0) 1210 if (bno < 0)
1209 return (0); 1211 return (0);
1210 cgp->cg_rotor = ufs_rw32(bno, needswap); 1212 cgp->cg_rotor = ufs_rw32(bno, needswap);
1211gotit: 1213gotit:
1212 blkno = fragstoblks(fs, bno); 1214 blkno = fragstoblks(fs, bno);
1213 ffs_clrblock(fs, blksfree, blkno); 1215 ffs_clrblock(fs, blksfree, blkno);
1214 ffs_clusteracct(fs, cgp, blkno, -1); 1216 ffs_clusteracct(fs, cgp, blkno, -1);
1215 ufs_add32(cgp->cg_cs.cs_nbfree, -1, needswap); 1217 ufs_add32(cgp->cg_cs.cs_nbfree, -1, needswap);
1216 fs->fs_cstotal.cs_nbfree--; 1218 fs->fs_cstotal.cs_nbfree--;
1217 fs->fs_cs(fs, ufs_rw32(cgp->cg_cgx, needswap)).cs_nbfree--; 1219 fs->fs_cs(fs, ufs_rw32(cgp->cg_cgx, needswap)).cs_nbfree--;
1218 if ((fs->fs_magic == FS_UFS1_MAGIC) && 1220 if ((fs->fs_magic == FS_UFS1_MAGIC) &&
1219 ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) { 1221 ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
1220 int cylno; 1222 int cylno;
1221 cylno = old_cbtocylno(fs, bno); 1223 cylno = old_cbtocylno(fs, bno);
1222 KASSERT(cylno >= 0); 1224 KASSERT(cylno >= 0);
1223 KASSERT(cylno < fs->fs_old_ncyl); 1225 KASSERT(cylno < fs->fs_old_ncyl);
1224 KASSERT(old_cbtorpos(fs, bno) >= 0); 1226 KASSERT(old_cbtorpos(fs, bno) >= 0);
1225 KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, bno) < fs->fs_old_nrpos); 1227 KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, bno) < fs->fs_old_nrpos);
1226 ufs_add16(old_cg_blks(fs, cgp, cylno, needswap)[old_cbtorpos(fs, bno)], -1, 1228 ufs_add16(old_cg_blks(fs, cgp, cylno, needswap)[old_cbtorpos(fs, bno)], -1,
1227 needswap); 1229 needswap);
1228 ufs_add32(old_cg_blktot(cgp, needswap)[cylno], -1, needswap); 1230 ufs_add32(old_cg_blktot(cgp, needswap)[cylno], -1, needswap);
1229 } 1231 }
1230 fs->fs_fmod = 1; 1232 fs->fs_fmod = 1;
1231 blkno = ufs_rw32(cgp->cg_cgx, needswap) * fs->fs_fpg + bno; 1233 blkno = ufs_rw32(cgp->cg_cgx, needswap) * fs->fs_fpg + bno;
1232 if (DOINGSOFTDEP(ITOV(ip))) { 1234 if (DOINGSOFTDEP(ITOV(ip))) {
1233 mutex_exit(&ump->um_lock); 1235 mutex_exit(&ump->um_lock);
1234 softdep_setup_blkmapdep(bp, fs, blkno); 1236 softdep_setup_blkmapdep(bp, fs, blkno);
1235 mutex_enter(&ump->um_lock); 1237 mutex_enter(&ump->um_lock);
1236 } 1238 }
1237 return (blkno); 1239 return (blkno);
1238} 1240}
1239 1241
1240/* 1242/*
1241 * Determine whether an inode can be allocated. 1243 * Determine whether an inode can be allocated.
1242 * 1244 *
1243 * Check to see if an inode is available, and if it is, 1245 * Check to see if an inode is available, and if it is,
1244 * allocate it using the following policy: 1246 * allocate it using the following policy:
1245 * 1) allocate the requested inode. 1247 * 1) allocate the requested inode.
1246 * 2) allocate the next available inode after the requested 1248 * 2) allocate the next available inode after the requested
1247 * inode in the specified cylinder group. 1249 * inode in the specified cylinder group.
1248 */ 1250 */
1249static daddr_t 1251static daddr_t
1250ffs_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode, int flags) 1252ffs_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode, int flags)
1251{ 1253{
1252 struct ufsmount *ump = ip->i_ump; 1254 struct ufsmount *ump = ip->i_ump;
1253 struct fs *fs = ip->i_fs; 1255 struct fs *fs = ip->i_fs;
1254 struct cg *cgp; 1256 struct cg *cgp;
1255 struct buf *bp, *ibp; 1257 struct buf *bp, *ibp;
1256 u_int8_t *inosused; 1258 u_int8_t *inosused;
1257 int error, start, len, loc, map, i; 1259 int error, start, len, loc, map, i;
1258 int32_t initediblk; 1260 int32_t initediblk;
1259 daddr_t nalloc; 1261 daddr_t nalloc;
1260 struct ufs2_dinode *dp2; 1262 struct ufs2_dinode *dp2;
1261#ifdef FFS_EI 1263#ifdef FFS_EI
1262 const int needswap = UFS_FSNEEDSWAP(fs); 1264 const int needswap = UFS_FSNEEDSWAP(fs);
1263#endif 1265#endif
1264 1266
1265 KASSERT(mutex_owned(&ump->um_lock)); 1267 KASSERT(mutex_owned(&ump->um_lock));
1266 UFS_WAPBL_JLOCK_ASSERT(ip->i_ump->um_mountp); 1268 UFS_WAPBL_JLOCK_ASSERT(ip->i_ump->um_mountp);
1267 1269
1268 if (fs->fs_cs(fs, cg).cs_nifree == 0) 1270 if (fs->fs_cs(fs, cg).cs_nifree == 0)
1269 return (0); 1271 return (0);
1270 mutex_exit(&ump->um_lock); 1272 mutex_exit(&ump->um_lock);
1271 ibp = NULL; 1273 ibp = NULL;
1272 initediblk = -1; 1274 initediblk = -1;
1273retry: 1275retry:
1274 error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), 1276 error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
1275 (int)fs->fs_cgsize, NOCRED, B_MODIFY, &bp); 1277 (int)fs->fs_cgsize, NOCRED, B_MODIFY, &bp);
1276 if (error) 1278 if (error)
1277 goto fail; 1279 goto fail;
1278 cgp = (struct cg *)bp->b_data; 1280 cgp = (struct cg *)bp->b_data;
1279 if (!cg_chkmagic(cgp, needswap) || cgp->cg_cs.cs_nifree == 0) 1281 if (!cg_chkmagic(cgp, needswap) || cgp->cg_cs.cs_nifree == 0)
1280 goto fail; 1282 goto fail;
1281 1283
1282 if (ibp != NULL && 1284 if (ibp != NULL &&
1283 initediblk != ufs_rw32(cgp->cg_initediblk, needswap)) { 1285 initediblk != ufs_rw32(cgp->cg_initediblk, needswap)) {
1284 /* Another thread allocated more inodes so we retry the test. */ 1286 /* Another thread allocated more inodes so we retry the test. */
1285 brelse(ibp, BC_INVAL); 1287 brelse(ibp, BC_INVAL);
1286 ibp = NULL; 1288 ibp = NULL;
1287 } 1289 }
1288 /* 1290 /*
1289 * Check to see if we need to initialize more inodes. 1291 * Check to see if we need to initialize more inodes.
1290 */ 1292 */
1291 if (fs->fs_magic == FS_UFS2_MAGIC && ibp == NULL) { 1293 if (fs->fs_magic == FS_UFS2_MAGIC && ibp == NULL) {
1292 initediblk = ufs_rw32(cgp->cg_initediblk, needswap); 1294 initediblk = ufs_rw32(cgp->cg_initediblk, needswap);
1293 nalloc = fs->fs_ipg - ufs_rw32(cgp->cg_cs.cs_nifree, needswap); 1295 nalloc = fs->fs_ipg - ufs_rw32(cgp->cg_cs.cs_nifree, needswap);
1294 if (nalloc + INOPB(fs) > initediblk && 1296 if (nalloc + INOPB(fs) > initediblk &&
1295 initediblk < ufs_rw32(cgp->cg_niblk, needswap)) { 1297 initediblk < ufs_rw32(cgp->cg_niblk, needswap)) {
1296 /* 1298 /*
1297 * We have to release the cg buffer here to prevent 1299 * We have to release the cg buffer here to prevent
1298 * a deadlock when reading the inode block will 1300 * a deadlock when reading the inode block will
1299 * run a copy-on-write that might use this cg. 1301 * run a copy-on-write that might use this cg.
1300 */ 1302 */
1301 brelse(bp, 0); 1303 brelse(bp, 0);
1302 bp = NULL; 1304 bp = NULL;
1303 error = ffs_getblk(ip->i_devvp, fsbtodb(fs, 1305 error = ffs_getblk(ip->i_devvp, fsbtodb(fs,
1304 ino_to_fsba(fs, cg * fs->fs_ipg + initediblk)), 1306 ino_to_fsba(fs, cg * fs->fs_ipg + initediblk)),
1305 FFS_NOBLK, fs->fs_bsize, false, &ibp); 1307 FFS_NOBLK, fs->fs_bsize, false, &ibp);
1306 if (error) 1308 if (error)
1307 goto fail; 1309 goto fail;
1308 goto retry; 1310 goto retry;
1309 } 1311 }
1310 } 1312 }
1311 1313
1312 cgp->cg_old_time = ufs_rw32(time_second, needswap); 1314 cgp->cg_old_time = ufs_rw32(time_second, needswap);
1313 if ((fs->fs_magic != FS_UFS1_MAGIC) || 1315 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1314 (fs->fs_old_flags & FS_FLAGS_UPDATED)) 1316 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1315 cgp->cg_time = ufs_rw64(time_second, needswap); 1317 cgp->cg_time = ufs_rw64(time_second, needswap);
1316 inosused = cg_inosused(cgp, needswap); 1318 inosused = cg_inosused(cgp, needswap);