xref: /openbmc/linux/fs/xfs/xfs_itable.c (revision 32daa5d7)
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 user_namespace	*mnt_userns,
58 	struct xfs_trans	*tp,
59 	xfs_ino_t		ino,
60 	struct xfs_bstat_chunk	*bc)
61 {
62 	struct user_namespace	*sb_userns = mp->m_super->s_user_ns;
63 	struct xfs_icdinode	*dic;		/* dinode core info pointer */
64 	struct xfs_inode	*ip;		/* incore inode pointer */
65 	struct inode		*inode;
66 	struct xfs_bulkstat	*buf = bc->buf;
67 	int			error = -EINVAL;
68 
69 	if (xfs_internal_inum(mp, ino))
70 		goto out_advance;
71 
72 	error = xfs_iget(mp, tp, ino,
73 			 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
74 			 XFS_ILOCK_SHARED, &ip);
75 	if (error == -ENOENT || error == -EINVAL)
76 		goto out_advance;
77 	if (error)
78 		goto out;
79 
80 	ASSERT(ip != NULL);
81 	ASSERT(ip->i_imap.im_blkno != 0);
82 	inode = VFS_I(ip);
83 
84 	dic = &ip->i_d;
85 
86 	/* xfs_iget returns the following without needing
87 	 * further change.
88 	 */
89 	buf->bs_projectid = ip->i_d.di_projid;
90 	buf->bs_ino = ino;
91 	buf->bs_uid = from_kuid(sb_userns, i_uid_into_mnt(mnt_userns, inode));
92 	buf->bs_gid = from_kgid(sb_userns, i_gid_into_mnt(mnt_userns, inode));
93 	buf->bs_size = dic->di_size;
94 
95 	buf->bs_nlink = inode->i_nlink;
96 	buf->bs_atime = inode->i_atime.tv_sec;
97 	buf->bs_atime_nsec = inode->i_atime.tv_nsec;
98 	buf->bs_mtime = inode->i_mtime.tv_sec;
99 	buf->bs_mtime_nsec = inode->i_mtime.tv_nsec;
100 	buf->bs_ctime = inode->i_ctime.tv_sec;
101 	buf->bs_ctime_nsec = inode->i_ctime.tv_nsec;
102 	buf->bs_btime = dic->di_crtime.tv_sec;
103 	buf->bs_btime_nsec = dic->di_crtime.tv_nsec;
104 	buf->bs_gen = inode->i_generation;
105 	buf->bs_mode = inode->i_mode;
106 
107 	buf->bs_xflags = xfs_ip2xflags(ip);
108 	buf->bs_extsize_blks = dic->di_extsize;
109 	buf->bs_extents = xfs_ifork_nextents(&ip->i_df);
110 	xfs_bulkstat_health(ip, buf);
111 	buf->bs_aextents = xfs_ifork_nextents(ip->i_afp);
112 	buf->bs_forkoff = XFS_IFORK_BOFF(ip);
113 	buf->bs_version = XFS_BULKSTAT_VERSION_V5;
114 
115 	if (xfs_sb_version_has_v3inode(&mp->m_sb)) {
116 		if (dic->di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
117 			buf->bs_cowextsize_blks = dic->di_cowextsize;
118 	}
119 
120 	switch (ip->i_df.if_format) {
121 	case XFS_DINODE_FMT_DEV:
122 		buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
123 		buf->bs_blksize = BLKDEV_IOSIZE;
124 		buf->bs_blocks = 0;
125 		break;
126 	case XFS_DINODE_FMT_LOCAL:
127 		buf->bs_rdev = 0;
128 		buf->bs_blksize = mp->m_sb.sb_blocksize;
129 		buf->bs_blocks = 0;
130 		break;
131 	case XFS_DINODE_FMT_EXTENTS:
132 	case XFS_DINODE_FMT_BTREE:
133 		buf->bs_rdev = 0;
134 		buf->bs_blksize = mp->m_sb.sb_blocksize;
135 		buf->bs_blocks = dic->di_nblocks + ip->i_delayed_blks;
136 		break;
137 	}
138 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
139 	xfs_irele(ip);
140 
141 	error = bc->formatter(bc->breq, buf);
142 	if (error == -ECANCELED)
143 		goto out_advance;
144 	if (error)
145 		goto out;
146 
147 out_advance:
148 	/*
149 	 * Advance the cursor to the inode that comes after the one we just
150 	 * looked at.  We want the caller to move along if the bulkstat
151 	 * information was copied successfully; if we tried to grab the inode
152 	 * but it's no longer allocated; or if it's internal metadata.
153 	 */
154 	bc->breq->startino = ino + 1;
155 out:
156 	return error;
157 }
158 
159 /* Bulkstat a single inode. */
160 int
161 xfs_bulkstat_one(
162 	struct xfs_ibulk	*breq,
163 	bulkstat_one_fmt_pf	formatter)
164 {
165 	struct xfs_bstat_chunk	bc = {
166 		.formatter	= formatter,
167 		.breq		= breq,
168 	};
169 	int			error;
170 
171 	if (breq->mnt_userns != &init_user_ns) {
172 		xfs_warn_ratelimited(breq->mp,
173 			"bulkstat not supported inside of idmapped mounts.");
174 		return -EINVAL;
175 	}
176 
177 	ASSERT(breq->icount == 1);
178 
179 	bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
180 			KM_MAYFAIL);
181 	if (!bc.buf)
182 		return -ENOMEM;
183 
184 	error = xfs_bulkstat_one_int(breq->mp, breq->mnt_userns, NULL,
185 				     breq->startino, &bc);
186 
187 	kmem_free(bc.buf);
188 
189 	/*
190 	 * If we reported one inode to userspace then we abort because we hit
191 	 * the end of the buffer.  Don't leak that back to userspace.
192 	 */
193 	if (error == -ECANCELED)
194 		error = 0;
195 
196 	return error;
197 }
198 
199 static int
200 xfs_bulkstat_iwalk(
201 	struct xfs_mount	*mp,
202 	struct xfs_trans	*tp,
203 	xfs_ino_t		ino,
204 	void			*data)
205 {
206 	struct xfs_bstat_chunk	*bc = data;
207 	int			error;
208 
209 	error = xfs_bulkstat_one_int(mp, bc->breq->mnt_userns, tp, ino, data);
210 	/* bulkstat just skips over missing inodes */
211 	if (error == -ENOENT || error == -EINVAL)
212 		return 0;
213 	return error;
214 }
215 
216 /*
217  * Check the incoming lastino parameter.
218  *
219  * We allow any inode value that could map to physical space inside the
220  * filesystem because if there are no inodes there, bulkstat moves on to the
221  * next chunk.  In other words, the magic agino value of zero takes us to the
222  * first chunk in the AG, and an agino value past the end of the AG takes us to
223  * the first chunk in the next AG.
224  *
225  * Therefore we can end early if the requested inode is beyond the end of the
226  * filesystem or doesn't map properly.
227  */
228 static inline bool
229 xfs_bulkstat_already_done(
230 	struct xfs_mount	*mp,
231 	xfs_ino_t		startino)
232 {
233 	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, startino);
234 	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, startino);
235 
236 	return agno >= mp->m_sb.sb_agcount ||
237 	       startino != XFS_AGINO_TO_INO(mp, agno, agino);
238 }
239 
240 /* Return stat information in bulk (by-inode) for the filesystem. */
241 int
242 xfs_bulkstat(
243 	struct xfs_ibulk	*breq,
244 	bulkstat_one_fmt_pf	formatter)
245 {
246 	struct xfs_bstat_chunk	bc = {
247 		.formatter	= formatter,
248 		.breq		= breq,
249 	};
250 	int			error;
251 
252 	if (breq->mnt_userns != &init_user_ns) {
253 		xfs_warn_ratelimited(breq->mp,
254 			"bulkstat not supported inside of idmapped mounts.");
255 		return -EINVAL;
256 	}
257 	if (xfs_bulkstat_already_done(breq->mp, breq->startino))
258 		return 0;
259 
260 	bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
261 			KM_MAYFAIL);
262 	if (!bc.buf)
263 		return -ENOMEM;
264 
265 	error = xfs_iwalk(breq->mp, NULL, breq->startino, breq->flags,
266 			xfs_bulkstat_iwalk, breq->icount, &bc);
267 
268 	kmem_free(bc.buf);
269 
270 	/*
271 	 * We found some inodes, so clear the error status and return them.
272 	 * The lastino pointer will point directly at the inode that triggered
273 	 * any error that occurred, so on the next call the error will be
274 	 * triggered again and propagated to userspace as there will be no
275 	 * formatted inodes in the buffer.
276 	 */
277 	if (breq->ocount > 0)
278 		error = 0;
279 
280 	return error;
281 }
282 
283 /* Convert bulkstat (v5) to bstat (v1). */
284 void
285 xfs_bulkstat_to_bstat(
286 	struct xfs_mount		*mp,
287 	struct xfs_bstat		*bs1,
288 	const struct xfs_bulkstat	*bstat)
289 {
290 	/* memset is needed here because of padding holes in the structure. */
291 	memset(bs1, 0, sizeof(struct xfs_bstat));
292 	bs1->bs_ino = bstat->bs_ino;
293 	bs1->bs_mode = bstat->bs_mode;
294 	bs1->bs_nlink = bstat->bs_nlink;
295 	bs1->bs_uid = bstat->bs_uid;
296 	bs1->bs_gid = bstat->bs_gid;
297 	bs1->bs_rdev = bstat->bs_rdev;
298 	bs1->bs_blksize = bstat->bs_blksize;
299 	bs1->bs_size = bstat->bs_size;
300 	bs1->bs_atime.tv_sec = bstat->bs_atime;
301 	bs1->bs_mtime.tv_sec = bstat->bs_mtime;
302 	bs1->bs_ctime.tv_sec = bstat->bs_ctime;
303 	bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
304 	bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
305 	bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
306 	bs1->bs_blocks = bstat->bs_blocks;
307 	bs1->bs_xflags = bstat->bs_xflags;
308 	bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
309 	bs1->bs_extents = bstat->bs_extents;
310 	bs1->bs_gen = bstat->bs_gen;
311 	bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
312 	bs1->bs_forkoff = bstat->bs_forkoff;
313 	bs1->bs_projid_hi = bstat->bs_projectid >> 16;
314 	bs1->bs_sick = bstat->bs_sick;
315 	bs1->bs_checked = bstat->bs_checked;
316 	bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
317 	bs1->bs_dmevmask = 0;
318 	bs1->bs_dmstate = 0;
319 	bs1->bs_aextents = bstat->bs_aextents;
320 }
321 
322 struct xfs_inumbers_chunk {
323 	inumbers_fmt_pf		formatter;
324 	struct xfs_ibulk	*breq;
325 };
326 
327 /*
328  * INUMBERS
329  * ========
330  * This is how we export inode btree records to userspace, so that XFS tools
331  * can figure out where inodes are allocated.
332  */
333 
334 /*
335  * Format the inode group structure and report it somewhere.
336  *
337  * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
338  * through the filesystem so we move it forward unless there was a runtime
339  * error.  If the formatter tells us the buffer is now full we also move the
340  * cursor forward and abort the walk.
341  */
342 STATIC int
343 xfs_inumbers_walk(
344 	struct xfs_mount	*mp,
345 	struct xfs_trans	*tp,
346 	xfs_agnumber_t		agno,
347 	const struct xfs_inobt_rec_incore *irec,
348 	void			*data)
349 {
350 	struct xfs_inumbers	inogrp = {
351 		.xi_startino	= XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
352 		.xi_alloccount	= irec->ir_count - irec->ir_freecount,
353 		.xi_allocmask	= ~irec->ir_free,
354 		.xi_version	= XFS_INUMBERS_VERSION_V5,
355 	};
356 	struct xfs_inumbers_chunk *ic = data;
357 	int			error;
358 
359 	error = ic->formatter(ic->breq, &inogrp);
360 	if (error && error != -ECANCELED)
361 		return error;
362 
363 	ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
364 			XFS_INODES_PER_CHUNK;
365 	return error;
366 }
367 
368 /*
369  * Return inode number table for the filesystem.
370  */
371 int
372 xfs_inumbers(
373 	struct xfs_ibulk	*breq,
374 	inumbers_fmt_pf		formatter)
375 {
376 	struct xfs_inumbers_chunk ic = {
377 		.formatter	= formatter,
378 		.breq		= breq,
379 	};
380 	int			error = 0;
381 
382 	if (xfs_bulkstat_already_done(breq->mp, breq->startino))
383 		return 0;
384 
385 	error = xfs_inobt_walk(breq->mp, NULL, breq->startino, breq->flags,
386 			xfs_inumbers_walk, breq->icount, &ic);
387 
388 	/*
389 	 * We found some inode groups, so clear the error status and return
390 	 * them.  The lastino pointer will point directly at the inode that
391 	 * triggered any error that occurred, so on the next call the error
392 	 * will be triggered again and propagated to userspace as there will be
393 	 * no formatted inode groups in the buffer.
394 	 */
395 	if (breq->ocount > 0)
396 		error = 0;
397 
398 	return error;
399 }
400 
401 /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
402 void
403 xfs_inumbers_to_inogrp(
404 	struct xfs_inogrp		*ig1,
405 	const struct xfs_inumbers	*ig)
406 {
407 	/* memset is needed here because of padding holes in the structure. */
408 	memset(ig1, 0, sizeof(struct xfs_inogrp));
409 	ig1->xi_startino = ig->xi_startino;
410 	ig1->xi_alloccount = ig->xi_alloccount;
411 	ig1->xi_allocmask = ig->xi_allocmask;
412 }
413