xref: /freebsd/lib/libufs/cgroup.c (revision ba3c1f5972d7b90feb6e6da47905ff2757e0fe57)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2003 Juli Mallett.  All rights reserved.
5  *
6  * This software was written by Juli Mallett <jmallett@FreeBSD.org> for the
7  * FreeBSD project.  Redistribution and use in source and binary forms, with
8  * or without modification, are permitted provided that the following
9  * conditions are met:
10  *
11  * 1. Redistribution of source code must retain the above copyright notice,
12  *    this list of conditions and the following disclaimer.
13  * 2. Redistribution in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
19  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
20  * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
21  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
22  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
23  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
25  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
26  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27  * POSSIBILITY OF SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/mount.h>
35 #include <sys/disklabel.h>
36 #include <sys/stat.h>
37 
38 #include <ufs/ufs/extattr.h>
39 #include <ufs/ufs/quota.h>
40 #include <ufs/ufs/ufsmount.h>
41 #include <ufs/ufs/dinode.h>
42 #include <ufs/ffs/fs.h>
43 
44 #include <errno.h>
45 #include <fcntl.h>
46 #include <stdio.h>
47 #include <stdlib.h>
48 #include <string.h>
49 #include <unistd.h>
50 
51 #include <libufs.h>
52 
53 ufs2_daddr_t
54 cgballoc(struct uufsd *disk)
55 {
56 	u_int8_t *blksfree;
57 	struct cg *cgp;
58 	struct fs *fs;
59 	long bno;
60 
61 	fs = &disk->d_fs;
62 	cgp = &disk->d_cg;
63 	blksfree = cg_blksfree(cgp);
64 	for (bno = 0; bno < fs->fs_fpg / fs->fs_frag; bno++)
65 		if (ffs_isblock(fs, blksfree, bno))
66 			goto gotit;
67 	return (0);
68 gotit:
69 	fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--;
70 	ffs_clrblock(fs, blksfree, (long)bno);
71 	ffs_clusteracct(fs, cgp, bno, -1);
72 	cgp->cg_cs.cs_nbfree--;
73 	fs->fs_cstotal.cs_nbfree--;
74 	fs->fs_fmod = 1;
75 	return (cgbase(fs, cgp->cg_cgx) + blkstofrags(fs, bno));
76 }
77 
78 int
79 cgbfree(struct uufsd *disk, ufs2_daddr_t bno, long size)
80 {
81 	u_int8_t *blksfree;
82 	struct fs *fs;
83 	struct cg *cgp;
84 	ufs1_daddr_t fragno, cgbno;
85 	int i, cg, blk, frags, bbase;
86 
87 	fs = &disk->d_fs;
88 	cg = dtog(fs, bno);
89 	if (cgread1(disk, cg) != 1)
90 		return (-1);
91 	cgp = &disk->d_cg;
92 	cgbno = dtogd(fs, bno);
93 	blksfree = cg_blksfree(cgp);
94 	if (size == fs->fs_bsize) {
95 		fragno = fragstoblks(fs, cgbno);
96 		ffs_setblock(fs, blksfree, fragno);
97 		ffs_clusteracct(fs, cgp, fragno, 1);
98 		cgp->cg_cs.cs_nbfree++;
99 		fs->fs_cstotal.cs_nbfree++;
100 		fs->fs_cs(fs, cg).cs_nbfree++;
101 	} else {
102 		bbase = cgbno - fragnum(fs, cgbno);
103 		/*
104 		 * decrement the counts associated with the old frags
105 		 */
106 		blk = blkmap(fs, blksfree, bbase);
107 		ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
108 		/*
109 		 * deallocate the fragment
110 		 */
111 		frags = numfrags(fs, size);
112 		for (i = 0; i < frags; i++)
113 			setbit(blksfree, cgbno + i);
114 		cgp->cg_cs.cs_nffree += i;
115 		fs->fs_cstotal.cs_nffree += i;
116 		fs->fs_cs(fs, cg).cs_nffree += i;
117 		/*
118 		 * add back in counts associated with the new frags
119 		 */
120 		blk = blkmap(fs, blksfree, bbase);
121 		ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
122 		/*
123 		 * if a complete block has been reassembled, account for it
124 		 */
125 		fragno = fragstoblks(fs, bbase);
126 		if (ffs_isblock(fs, blksfree, fragno)) {
127 			cgp->cg_cs.cs_nffree -= fs->fs_frag;
128 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
129 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
130 			ffs_clusteracct(fs, cgp, fragno, 1);
131 			cgp->cg_cs.cs_nbfree++;
132 			fs->fs_cstotal.cs_nbfree++;
133 			fs->fs_cs(fs, cg).cs_nbfree++;
134 		}
135 	}
136 	return cgwrite(disk);
137 }
138 
139 ino_t
140 cgialloc(struct uufsd *disk)
141 {
142 	struct ufs2_dinode *dp2;
143 	u_int8_t *inosused;
144 	struct cg *cgp;
145 	struct fs *fs;
146 	ino_t ino;
147 	int i;
148 
149 	fs = &disk->d_fs;
150 	cgp = &disk->d_cg;
151 	inosused = cg_inosused(cgp);
152 	for (ino = 0; ino < fs->fs_ipg; ino++)
153 		if (isclr(inosused, ino))
154 			goto gotit;
155 	return (0);
156 gotit:
157 	if (fs->fs_magic == FS_UFS2_MAGIC &&
158 	    ino + INOPB(fs) > cgp->cg_initediblk &&
159 	    cgp->cg_initediblk < cgp->cg_niblk) {
160 		char block[MAXBSIZE];
161 		bzero(block, (int)fs->fs_bsize);
162 		dp2 = (struct ufs2_dinode *)&block;
163 		for (i = 0; i < INOPB(fs); i++) {
164 			dp2->di_gen = arc4random();
165 			dp2++;
166 		}
167 		if (bwrite(disk, ino_to_fsba(fs,
168 		    cgp->cg_cgx * fs->fs_ipg + cgp->cg_initediblk),
169 		    block, fs->fs_bsize))
170 			return (0);
171 		cgp->cg_initediblk += INOPB(fs);
172 	}
173 
174 	setbit(inosused, ino);
175 	cgp->cg_irotor = ino;
176 	cgp->cg_cs.cs_nifree--;
177 	fs->fs_cstotal.cs_nifree--;
178 	fs->fs_cs(fs, cgp->cg_cgx).cs_nifree--;
179 	fs->fs_fmod = 1;
180 
181 	return (ino + (cgp->cg_cgx * fs->fs_ipg));
182 }
183 
184 int
185 cgread(struct uufsd *disk)
186 {
187 
188 	if (disk->d_ccg >= disk->d_fs.fs_ncg)
189 		return (0);
190 	return (cgread1(disk, disk->d_ccg++));
191 }
192 
193 /* Short read/write error messages from cgget()/cgput() */
194 static const char *failmsg;
195 
196 int
197 cgread1(struct uufsd *disk, int c)
198 {
199 
200 	if (cgget(disk->d_fd, &disk->d_fs, c, &disk->d_cg) == 0) {
201 		disk->d_lcg = c;
202 		return (1);
203 	}
204 	ERROR(disk, NULL);
205 	if (failmsg != NULL) {
206 		ERROR(disk, failmsg);
207 		return (-1);
208 	}
209 	switch (errno) {
210 	case EINTEGRITY:
211 		ERROR(disk, "cylinder group checks failed");
212 		break;
213 	case EIO:
214 		ERROR(disk, "read error from block device");
215 		break;
216 	default:
217 		ERROR(disk, strerror(errno));
218 		break;
219 	}
220 	return (-1);
221 }
222 
223 int
224 cgget(int devfd, struct fs *fs, int cg, struct cg *cgp)
225 {
226 	uint32_t cghash, calchash;
227 	size_t cnt;
228 
229 	failmsg = NULL;
230 	if ((cnt = pread(devfd, cgp, fs->fs_cgsize,
231 	    fsbtodb(fs, cgtod(fs, cg)) * (fs->fs_fsize / fsbtodb(fs,1)))) < 0)
232 		return (-1);
233 	if (cnt == 0) {
234 		failmsg = "end of file from block device";
235 		errno = EIO;
236 		return (-1);
237 	}
238 	if (cnt != fs->fs_cgsize) {
239 		failmsg = "short read from block device";
240 		errno = EIO;
241 		return (-1);
242 	}
243 	calchash = cgp->cg_ckhash;
244 	if ((fs->fs_metackhash & CK_CYLGRP) != 0) {
245 		cghash = cgp->cg_ckhash;
246 		cgp->cg_ckhash = 0;
247 		calchash = calculate_crc32c(~0L, (void *)cgp, fs->fs_cgsize);
248 		cgp->cg_ckhash = cghash;
249 	}
250 	if (cgp->cg_ckhash != calchash || !cg_chkmagic(cgp) ||
251 	    cgp->cg_cgx != cg) {
252 		errno = EINTEGRITY;
253 		return (-1);
254 	}
255 	return (0);
256 }
257 
258 int
259 cgwrite(struct uufsd *disk)
260 {
261 
262 	return (cgwrite1(disk, disk->d_cg.cg_cgx));
263 }
264 
265 int
266 cgwrite1(struct uufsd *disk, int cg)
267 {
268 	static char errmsg[BUFSIZ];
269 
270 	if (cg == disk->d_cg.cg_cgx) {
271 		if (ufs_disk_write(disk) == -1) {
272 			ERROR(disk, "failed to open disk for writing");
273 			return (-1);
274 		}
275 		if (cgput(disk->d_fd, &disk->d_fs, &disk->d_cg) == 0)
276 			return (0);
277 		ERROR(disk, NULL);
278 		if (failmsg != NULL) {
279 			ERROR(disk, failmsg);
280 			return (-1);
281 		}
282 		switch (errno) {
283 		case EIO:
284 			ERROR(disk, "unable to write cylinder group");
285 			break;
286 		default:
287 			ERROR(disk, strerror(errno));
288 			break;
289 		}
290 		return (-1);
291 	}
292 	snprintf(errmsg, BUFSIZ, "Cylinder group %d in buffer does not match "
293 	    "the cylinder group %d that cgwrite1 requested",
294 	    disk->d_cg.cg_cgx, cg);
295 	ERROR(disk, errmsg);
296 	errno = EDOOFUS;
297 	return (-1);
298 }
299 
300 int
301 cgput(int devfd, struct fs *fs, struct cg *cgp)
302 {
303 	size_t cnt;
304 
305 	if ((fs->fs_metackhash & CK_CYLGRP) != 0) {
306 		cgp->cg_ckhash = 0;
307 		cgp->cg_ckhash =
308 		    calculate_crc32c(~0L, (void *)cgp, fs->fs_cgsize);
309 	}
310 	failmsg = NULL;
311 	if ((cnt = pwrite(devfd, cgp, fs->fs_cgsize,
312 	    fsbtodb(fs, cgtod(fs, cgp->cg_cgx)) *
313 	    (fs->fs_fsize / fsbtodb(fs,1)))) < 0)
314 		return (-1);
315 	if (cnt != fs->fs_cgsize) {
316 		failmsg = "short write to block device";
317 		return (-1);
318 	}
319 	return (0);
320 }
321