xref: /linux/fs/xfs/xfs_itable.c (revision d198b34f3855eee2571dda03eea75a09c7c31480)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_mount.h"
13 #include "xfs_inode.h"
14 #include "xfs_btree.h"
15 #include "xfs_ialloc.h"
16 #include "xfs_ialloc_btree.h"
17 #include "xfs_iwalk.h"
18 #include "xfs_itable.h"
19 #include "xfs_error.h"
20 #include "xfs_icache.h"
21 #include "xfs_health.h"
22 
23 /*
24  * Bulk Stat
25  * =========
26  *
27  * Use the inode walking functions to fill out struct xfs_bulkstat for every
28  * allocated inode, then pass the stat information to some externally provided
29  * iteration function.
30  */
31 
32 struct xfs_bstat_chunk {
33 	bulkstat_one_fmt_pf	formatter;
34 	struct xfs_ibulk	*breq;
35 	struct xfs_bulkstat	*buf;
36 };
37 
38 /*
39  * Fill out the bulkstat info for a single inode and report it somewhere.
40  *
41  * bc->breq->lastino is effectively the inode cursor as we walk through the
42  * filesystem.  Therefore, we update it any time we need to move the cursor
43  * forward, regardless of whether or not we're sending any bstat information
44  * back to userspace.  If the inode is internal metadata or, has been freed
45  * out from under us, we just simply keep going.
46  *
47  * However, if any other type of error happens we want to stop right where we
48  * are so that userspace will call back with exact number of the bad inode and
49  * we can send back an error code.
50  *
51  * Note that if the formatter tells us there's no space left in the buffer we
52  * move the cursor forward and abort the walk.
53  */
54 STATIC int
55 xfs_bulkstat_one_int(
56 	struct xfs_mount	*mp,
57 	struct xfs_trans	*tp,
58 	xfs_ino_t		ino,
59 	struct xfs_bstat_chunk	*bc)
60 {
61 	struct xfs_icdinode	*dic;		/* dinode core info pointer */
62 	struct xfs_inode	*ip;		/* incore inode pointer */
63 	struct inode		*inode;
64 	struct xfs_bulkstat	*buf = bc->buf;
65 	int			error = -EINVAL;
66 
67 	if (xfs_internal_inum(mp, ino))
68 		goto out_advance;
69 
70 	error = xfs_iget(mp, tp, ino,
71 			 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
72 			 XFS_ILOCK_SHARED, &ip);
73 	if (error == -ENOENT || error == -EINVAL)
74 		goto out_advance;
75 	if (error)
76 		goto out;
77 
78 	ASSERT(ip != NULL);
79 	ASSERT(ip->i_imap.im_blkno != 0);
80 	inode = VFS_I(ip);
81 
82 	dic = &ip->i_d;
83 
84 	/* xfs_iget returns the following without needing
85 	 * further change.
86 	 */
87 	buf->bs_projectid = ip->i_d.di_projid;
88 	buf->bs_ino = ino;
89 	buf->bs_uid = dic->di_uid;
90 	buf->bs_gid = dic->di_gid;
91 	buf->bs_size = dic->di_size;
92 
93 	buf->bs_nlink = inode->i_nlink;
94 	buf->bs_atime = inode->i_atime.tv_sec;
95 	buf->bs_atime_nsec = inode->i_atime.tv_nsec;
96 	buf->bs_mtime = inode->i_mtime.tv_sec;
97 	buf->bs_mtime_nsec = inode->i_mtime.tv_nsec;
98 	buf->bs_ctime = inode->i_ctime.tv_sec;
99 	buf->bs_ctime_nsec = inode->i_ctime.tv_nsec;
100 	buf->bs_btime = dic->di_crtime.tv_sec;
101 	buf->bs_btime_nsec = dic->di_crtime.tv_nsec;
102 	buf->bs_gen = inode->i_generation;
103 	buf->bs_mode = inode->i_mode;
104 
105 	buf->bs_xflags = xfs_ip2xflags(ip);
106 	buf->bs_extsize_blks = dic->di_extsize;
107 	buf->bs_extents = dic->di_nextents;
108 	xfs_bulkstat_health(ip, buf);
109 	buf->bs_aextents = dic->di_anextents;
110 	buf->bs_forkoff = XFS_IFORK_BOFF(ip);
111 	buf->bs_version = XFS_BULKSTAT_VERSION_V5;
112 
113 	if (dic->di_version == 3) {
114 		if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
115 			buf->bs_cowextsize_blks = dic->di_cowextsize;
116 	}
117 
118 	switch (dic->di_format) {
119 	case XFS_DINODE_FMT_DEV:
120 		buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
121 		buf->bs_blksize = BLKDEV_IOSIZE;
122 		buf->bs_blocks = 0;
123 		break;
124 	case XFS_DINODE_FMT_LOCAL:
125 		buf->bs_rdev = 0;
126 		buf->bs_blksize = mp->m_sb.sb_blocksize;
127 		buf->bs_blocks = 0;
128 		break;
129 	case XFS_DINODE_FMT_EXTENTS:
130 	case XFS_DINODE_FMT_BTREE:
131 		buf->bs_rdev = 0;
132 		buf->bs_blksize = mp->m_sb.sb_blocksize;
133 		buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
134 		break;
135 	}
136 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
137 	xfs_irele(ip);
138 
139 	error = bc->formatter(bc->breq, buf);
140 	if (error == -ECANCELED)
141 		goto out_advance;
142 	if (error)
143 		goto out;
144 
145 out_advance:
146 	/*
147 	 * Advance the cursor to the inode that comes after the one we just
148 	 * looked at.  We want the caller to move along if the bulkstat
149 	 * information was copied successfully; if we tried to grab the inode
150 	 * but it's no longer allocated; or if it's internal metadata.
151 	 */
152 	bc->breq->startino = ino + 1;
153 out:
154 	return error;
155 }
156 
157 /* Bulkstat a single inode. */
158 int
159 xfs_bulkstat_one(
160 	struct xfs_ibulk	*breq,
161 	bulkstat_one_fmt_pf	formatter)
162 {
163 	struct xfs_bstat_chunk	bc = {
164 		.formatter	= formatter,
165 		.breq		= breq,
166 	};
167 	int			error;
168 
169 	ASSERT(breq->icount == 1);
170 
171 	bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
172 			KM_MAYFAIL);
173 	if (!bc.buf)
174 		return -ENOMEM;
175 
176 	error = xfs_bulkstat_one_int(breq->mp, NULL, breq->startino, &bc);
177 
178 	kmem_free(bc.buf);
179 
180 	/*
181 	 * If we reported one inode to userspace then we abort because we hit
182 	 * the end of the buffer.  Don't leak that back to userspace.
183 	 */
184 	if (error == -ECANCELED)
185 		error = 0;
186 
187 	return error;
188 }
189 
190 static int
191 xfs_bulkstat_iwalk(
192 	struct xfs_mount	*mp,
193 	struct xfs_trans	*tp,
194 	xfs_ino_t		ino,
195 	void			*data)
196 {
197 	int			error;
198 
199 	error = xfs_bulkstat_one_int(mp, tp, ino, data);
200 	/* bulkstat just skips over missing inodes */
201 	if (error == -ENOENT || error == -EINVAL)
202 		return 0;
203 	return error;
204 }
205 
206 /*
207  * Check the incoming lastino parameter.
208  *
209  * We allow any inode value that could map to physical space inside the
210  * filesystem because if there are no inodes there, bulkstat moves on to the
211  * next chunk.  In other words, the magic agino value of zero takes us to the
212  * first chunk in the AG, and an agino value past the end of the AG takes us to
213  * the first chunk in the next AG.
214  *
215  * Therefore we can end early if the requested inode is beyond the end of the
216  * filesystem or doesn't map properly.
217  */
218 static inline bool
219 xfs_bulkstat_already_done(
220 	struct xfs_mount	*mp,
221 	xfs_ino_t		startino)
222 {
223 	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, startino);
224 	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, startino);
225 
226 	return agno >= mp->m_sb.sb_agcount ||
227 	       startino != XFS_AGINO_TO_INO(mp, agno, agino);
228 }
229 
230 /* Return stat information in bulk (by-inode) for the filesystem. */
231 int
232 xfs_bulkstat(
233 	struct xfs_ibulk	*breq,
234 	bulkstat_one_fmt_pf	formatter)
235 {
236 	struct xfs_bstat_chunk	bc = {
237 		.formatter	= formatter,
238 		.breq		= breq,
239 	};
240 	int			error;
241 
242 	if (xfs_bulkstat_already_done(breq->mp, breq->startino))
243 		return 0;
244 
245 	bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
246 			KM_MAYFAIL);
247 	if (!bc.buf)
248 		return -ENOMEM;
249 
250 	error = xfs_iwalk(breq->mp, NULL, breq->startino, breq->flags,
251 			xfs_bulkstat_iwalk, breq->icount, &bc);
252 
253 	kmem_free(bc.buf);
254 
255 	/*
256 	 * We found some inodes, so clear the error status and return them.
257 	 * The lastino pointer will point directly at the inode that triggered
258 	 * any error that occurred, so on the next call the error will be
259 	 * triggered again and propagated to userspace as there will be no
260 	 * formatted inodes in the buffer.
261 	 */
262 	if (breq->ocount > 0)
263 		error = 0;
264 
265 	return error;
266 }
267 
268 /* Convert bulkstat (v5) to bstat (v1). */
269 void
270 xfs_bulkstat_to_bstat(
271 	struct xfs_mount		*mp,
272 	struct xfs_bstat		*bs1,
273 	const struct xfs_bulkstat	*bstat)
274 {
275 	/* memset is needed here because of padding holes in the structure. */
276 	memset(bs1, 0, sizeof(struct xfs_bstat));
277 	bs1->bs_ino = bstat->bs_ino;
278 	bs1->bs_mode = bstat->bs_mode;
279 	bs1->bs_nlink = bstat->bs_nlink;
280 	bs1->bs_uid = bstat->bs_uid;
281 	bs1->bs_gid = bstat->bs_gid;
282 	bs1->bs_rdev = bstat->bs_rdev;
283 	bs1->bs_blksize = bstat->bs_blksize;
284 	bs1->bs_size = bstat->bs_size;
285 	bs1->bs_atime.tv_sec = bstat->bs_atime;
286 	bs1->bs_mtime.tv_sec = bstat->bs_mtime;
287 	bs1->bs_ctime.tv_sec = bstat->bs_ctime;
288 	bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
289 	bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
290 	bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
291 	bs1->bs_blocks = bstat->bs_blocks;
292 	bs1->bs_xflags = bstat->bs_xflags;
293 	bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
294 	bs1->bs_extents = bstat->bs_extents;
295 	bs1->bs_gen = bstat->bs_gen;
296 	bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
297 	bs1->bs_forkoff = bstat->bs_forkoff;
298 	bs1->bs_projid_hi = bstat->bs_projectid >> 16;
299 	bs1->bs_sick = bstat->bs_sick;
300 	bs1->bs_checked = bstat->bs_checked;
301 	bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
302 	bs1->bs_dmevmask = 0;
303 	bs1->bs_dmstate = 0;
304 	bs1->bs_aextents = bstat->bs_aextents;
305 }
306 
307 struct xfs_inumbers_chunk {
308 	inumbers_fmt_pf		formatter;
309 	struct xfs_ibulk	*breq;
310 };
311 
312 /*
313  * INUMBERS
314  * ========
315  * This is how we export inode btree records to userspace, so that XFS tools
316  * can figure out where inodes are allocated.
317  */
318 
319 /*
320  * Format the inode group structure and report it somewhere.
321  *
322  * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
323  * through the filesystem so we move it forward unless there was a runtime
324  * error.  If the formatter tells us the buffer is now full we also move the
325  * cursor forward and abort the walk.
326  */
327 STATIC int
328 xfs_inumbers_walk(
329 	struct xfs_mount	*mp,
330 	struct xfs_trans	*tp,
331 	xfs_agnumber_t		agno,
332 	const struct xfs_inobt_rec_incore *irec,
333 	void			*data)
334 {
335 	struct xfs_inumbers	inogrp = {
336 		.xi_startino	= XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
337 		.xi_alloccount	= irec->ir_count - irec->ir_freecount,
338 		.xi_allocmask	= ~irec->ir_free,
339 		.xi_version	= XFS_INUMBERS_VERSION_V5,
340 	};
341 	struct xfs_inumbers_chunk *ic = data;
342 	int			error;
343 
344 	error = ic->formatter(ic->breq, &inogrp);
345 	if (error && error != -ECANCELED)
346 		return error;
347 
348 	ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
349 			XFS_INODES_PER_CHUNK;
350 	return error;
351 }
352 
353 /*
354  * Return inode number table for the filesystem.
355  */
356 int
357 xfs_inumbers(
358 	struct xfs_ibulk	*breq,
359 	inumbers_fmt_pf		formatter)
360 {
361 	struct xfs_inumbers_chunk ic = {
362 		.formatter	= formatter,
363 		.breq		= breq,
364 	};
365 	int			error = 0;
366 
367 	if (xfs_bulkstat_already_done(breq->mp, breq->startino))
368 		return 0;
369 
370 	error = xfs_inobt_walk(breq->mp, NULL, breq->startino, breq->flags,
371 			xfs_inumbers_walk, breq->icount, &ic);
372 
373 	/*
374 	 * We found some inode groups, so clear the error status and return
375 	 * them.  The lastino pointer will point directly at the inode that
376 	 * triggered any error that occurred, so on the next call the error
377 	 * will be triggered again and propagated to userspace as there will be
378 	 * no formatted inode groups in the buffer.
379 	 */
380 	if (breq->ocount > 0)
381 		error = 0;
382 
383 	return error;
384 }
385 
386 /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
387 void
388 xfs_inumbers_to_inogrp(
389 	struct xfs_inogrp		*ig1,
390 	const struct xfs_inumbers	*ig)
391 {
392 	/* memset is needed here because of padding holes in the structure. */
393 	memset(ig1, 0, sizeof(struct xfs_inogrp));
394 	ig1->xi_startino = ig->xi_startino;
395 	ig1->xi_alloccount = ig->xi_alloccount;
396 	ig1->xi_allocmask = ig->xi_allocmask;
397 }
398