xref: /openbmc/linux/fs/xfs/xfs_itable.c (revision 7ce05074)
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_inode	*ip;		/* incore inode pointer */
64 	struct inode		*inode;
65 	struct xfs_bulkstat	*buf = bc->buf;
66 	int			error = -EINVAL;
67 
68 	if (xfs_internal_inum(mp, ino))
69 		goto out_advance;
70 
71 	error = xfs_iget(mp, tp, ino,
72 			 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
73 			 XFS_ILOCK_SHARED, &ip);
74 	if (error == -ENOENT || error == -EINVAL)
75 		goto out_advance;
76 	if (error)
77 		goto out;
78 
79 	ASSERT(ip != NULL);
80 	ASSERT(ip->i_imap.im_blkno != 0);
81 	inode = VFS_I(ip);
82 
83 	/* xfs_iget returns the following without needing
84 	 * further change.
85 	 */
86 	buf->bs_projectid = ip->i_projid;
87 	buf->bs_ino = ino;
88 	buf->bs_uid = from_kuid(sb_userns, i_uid_into_mnt(mnt_userns, inode));
89 	buf->bs_gid = from_kgid(sb_userns, i_gid_into_mnt(mnt_userns, inode));
90 	buf->bs_size = ip->i_disk_size;
91 
92 	buf->bs_nlink = inode->i_nlink;
93 	buf->bs_atime = inode->i_atime.tv_sec;
94 	buf->bs_atime_nsec = inode->i_atime.tv_nsec;
95 	buf->bs_mtime = inode->i_mtime.tv_sec;
96 	buf->bs_mtime_nsec = inode->i_mtime.tv_nsec;
97 	buf->bs_ctime = inode->i_ctime.tv_sec;
98 	buf->bs_ctime_nsec = inode->i_ctime.tv_nsec;
99 	buf->bs_gen = inode->i_generation;
100 	buf->bs_mode = inode->i_mode;
101 
102 	buf->bs_xflags = xfs_ip2xflags(ip);
103 	buf->bs_extsize_blks = ip->i_extsize;
104 	buf->bs_extents = xfs_ifork_nextents(&ip->i_df);
105 	xfs_bulkstat_health(ip, buf);
106 	buf->bs_aextents = xfs_ifork_nextents(ip->i_afp);
107 	buf->bs_forkoff = XFS_IFORK_BOFF(ip);
108 	buf->bs_version = XFS_BULKSTAT_VERSION_V5;
109 
110 	if (xfs_sb_version_has_v3inode(&mp->m_sb)) {
111 		buf->bs_btime = ip->i_crtime.tv_sec;
112 		buf->bs_btime_nsec = ip->i_crtime.tv_nsec;
113 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
114 			buf->bs_cowextsize_blks = ip->i_cowextsize;
115 	}
116 
117 	switch (ip->i_df.if_format) {
118 	case XFS_DINODE_FMT_DEV:
119 		buf->bs_rdev = sysv_encode_dev(inode->i_rdev);
120 		buf->bs_blksize = BLKDEV_IOSIZE;
121 		buf->bs_blocks = 0;
122 		break;
123 	case XFS_DINODE_FMT_LOCAL:
124 		buf->bs_rdev = 0;
125 		buf->bs_blksize = mp->m_sb.sb_blocksize;
126 		buf->bs_blocks = 0;
127 		break;
128 	case XFS_DINODE_FMT_EXTENTS:
129 	case XFS_DINODE_FMT_BTREE:
130 		buf->bs_rdev = 0;
131 		buf->bs_blksize = mp->m_sb.sb_blocksize;
132 		buf->bs_blocks = ip->i_nblocks + ip->i_delayed_blks;
133 		break;
134 	}
135 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
136 	xfs_irele(ip);
137 
138 	error = bc->formatter(bc->breq, buf);
139 	if (error == -ECANCELED)
140 		goto out_advance;
141 	if (error)
142 		goto out;
143 
144 out_advance:
145 	/*
146 	 * Advance the cursor to the inode that comes after the one we just
147 	 * looked at.  We want the caller to move along if the bulkstat
148 	 * information was copied successfully; if we tried to grab the inode
149 	 * but it's no longer allocated; or if it's internal metadata.
150 	 */
151 	bc->breq->startino = ino + 1;
152 out:
153 	return error;
154 }
155 
156 /* Bulkstat a single inode. */
157 int
158 xfs_bulkstat_one(
159 	struct xfs_ibulk	*breq,
160 	bulkstat_one_fmt_pf	formatter)
161 {
162 	struct xfs_bstat_chunk	bc = {
163 		.formatter	= formatter,
164 		.breq		= breq,
165 	};
166 	int			error;
167 
168 	if (breq->mnt_userns != &init_user_ns) {
169 		xfs_warn_ratelimited(breq->mp,
170 			"bulkstat not supported inside of idmapped mounts.");
171 		return -EINVAL;
172 	}
173 
174 	ASSERT(breq->icount == 1);
175 
176 	bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
177 			KM_MAYFAIL);
178 	if (!bc.buf)
179 		return -ENOMEM;
180 
181 	error = xfs_bulkstat_one_int(breq->mp, breq->mnt_userns, NULL,
182 				     breq->startino, &bc);
183 
184 	kmem_free(bc.buf);
185 
186 	/*
187 	 * If we reported one inode to userspace then we abort because we hit
188 	 * the end of the buffer.  Don't leak that back to userspace.
189 	 */
190 	if (error == -ECANCELED)
191 		error = 0;
192 
193 	return error;
194 }
195 
196 static int
197 xfs_bulkstat_iwalk(
198 	struct xfs_mount	*mp,
199 	struct xfs_trans	*tp,
200 	xfs_ino_t		ino,
201 	void			*data)
202 {
203 	struct xfs_bstat_chunk	*bc = data;
204 	int			error;
205 
206 	error = xfs_bulkstat_one_int(mp, bc->breq->mnt_userns, tp, ino, data);
207 	/* bulkstat just skips over missing inodes */
208 	if (error == -ENOENT || error == -EINVAL)
209 		return 0;
210 	return error;
211 }
212 
213 /*
214  * Check the incoming lastino parameter.
215  *
216  * We allow any inode value that could map to physical space inside the
217  * filesystem because if there are no inodes there, bulkstat moves on to the
218  * next chunk.  In other words, the magic agino value of zero takes us to the
219  * first chunk in the AG, and an agino value past the end of the AG takes us to
220  * the first chunk in the next AG.
221  *
222  * Therefore we can end early if the requested inode is beyond the end of the
223  * filesystem or doesn't map properly.
224  */
225 static inline bool
226 xfs_bulkstat_already_done(
227 	struct xfs_mount	*mp,
228 	xfs_ino_t		startino)
229 {
230 	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, startino);
231 	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, startino);
232 
233 	return agno >= mp->m_sb.sb_agcount ||
234 	       startino != XFS_AGINO_TO_INO(mp, agno, agino);
235 }
236 
237 /* Return stat information in bulk (by-inode) for the filesystem. */
238 int
239 xfs_bulkstat(
240 	struct xfs_ibulk	*breq,
241 	bulkstat_one_fmt_pf	formatter)
242 {
243 	struct xfs_bstat_chunk	bc = {
244 		.formatter	= formatter,
245 		.breq		= breq,
246 	};
247 	int			error;
248 
249 	if (breq->mnt_userns != &init_user_ns) {
250 		xfs_warn_ratelimited(breq->mp,
251 			"bulkstat not supported inside of idmapped mounts.");
252 		return -EINVAL;
253 	}
254 	if (xfs_bulkstat_already_done(breq->mp, breq->startino))
255 		return 0;
256 
257 	bc.buf = kmem_zalloc(sizeof(struct xfs_bulkstat),
258 			KM_MAYFAIL);
259 	if (!bc.buf)
260 		return -ENOMEM;
261 
262 	error = xfs_iwalk(breq->mp, NULL, breq->startino, breq->flags,
263 			xfs_bulkstat_iwalk, breq->icount, &bc);
264 
265 	kmem_free(bc.buf);
266 
267 	/*
268 	 * We found some inodes, so clear the error status and return them.
269 	 * The lastino pointer will point directly at the inode that triggered
270 	 * any error that occurred, so on the next call the error will be
271 	 * triggered again and propagated to userspace as there will be no
272 	 * formatted inodes in the buffer.
273 	 */
274 	if (breq->ocount > 0)
275 		error = 0;
276 
277 	return error;
278 }
279 
280 /* Convert bulkstat (v5) to bstat (v1). */
281 void
282 xfs_bulkstat_to_bstat(
283 	struct xfs_mount		*mp,
284 	struct xfs_bstat		*bs1,
285 	const struct xfs_bulkstat	*bstat)
286 {
287 	/* memset is needed here because of padding holes in the structure. */
288 	memset(bs1, 0, sizeof(struct xfs_bstat));
289 	bs1->bs_ino = bstat->bs_ino;
290 	bs1->bs_mode = bstat->bs_mode;
291 	bs1->bs_nlink = bstat->bs_nlink;
292 	bs1->bs_uid = bstat->bs_uid;
293 	bs1->bs_gid = bstat->bs_gid;
294 	bs1->bs_rdev = bstat->bs_rdev;
295 	bs1->bs_blksize = bstat->bs_blksize;
296 	bs1->bs_size = bstat->bs_size;
297 	bs1->bs_atime.tv_sec = bstat->bs_atime;
298 	bs1->bs_mtime.tv_sec = bstat->bs_mtime;
299 	bs1->bs_ctime.tv_sec = bstat->bs_ctime;
300 	bs1->bs_atime.tv_nsec = bstat->bs_atime_nsec;
301 	bs1->bs_mtime.tv_nsec = bstat->bs_mtime_nsec;
302 	bs1->bs_ctime.tv_nsec = bstat->bs_ctime_nsec;
303 	bs1->bs_blocks = bstat->bs_blocks;
304 	bs1->bs_xflags = bstat->bs_xflags;
305 	bs1->bs_extsize = XFS_FSB_TO_B(mp, bstat->bs_extsize_blks);
306 	bs1->bs_extents = bstat->bs_extents;
307 	bs1->bs_gen = bstat->bs_gen;
308 	bs1->bs_projid_lo = bstat->bs_projectid & 0xFFFF;
309 	bs1->bs_forkoff = bstat->bs_forkoff;
310 	bs1->bs_projid_hi = bstat->bs_projectid >> 16;
311 	bs1->bs_sick = bstat->bs_sick;
312 	bs1->bs_checked = bstat->bs_checked;
313 	bs1->bs_cowextsize = XFS_FSB_TO_B(mp, bstat->bs_cowextsize_blks);
314 	bs1->bs_dmevmask = 0;
315 	bs1->bs_dmstate = 0;
316 	bs1->bs_aextents = bstat->bs_aextents;
317 }
318 
319 struct xfs_inumbers_chunk {
320 	inumbers_fmt_pf		formatter;
321 	struct xfs_ibulk	*breq;
322 };
323 
324 /*
325  * INUMBERS
326  * ========
327  * This is how we export inode btree records to userspace, so that XFS tools
328  * can figure out where inodes are allocated.
329  */
330 
331 /*
332  * Format the inode group structure and report it somewhere.
333  *
334  * Similar to xfs_bulkstat_one_int, lastino is the inode cursor as we walk
335  * through the filesystem so we move it forward unless there was a runtime
336  * error.  If the formatter tells us the buffer is now full we also move the
337  * cursor forward and abort the walk.
338  */
339 STATIC int
340 xfs_inumbers_walk(
341 	struct xfs_mount	*mp,
342 	struct xfs_trans	*tp,
343 	xfs_agnumber_t		agno,
344 	const struct xfs_inobt_rec_incore *irec,
345 	void			*data)
346 {
347 	struct xfs_inumbers	inogrp = {
348 		.xi_startino	= XFS_AGINO_TO_INO(mp, agno, irec->ir_startino),
349 		.xi_alloccount	= irec->ir_count - irec->ir_freecount,
350 		.xi_allocmask	= ~irec->ir_free,
351 		.xi_version	= XFS_INUMBERS_VERSION_V5,
352 	};
353 	struct xfs_inumbers_chunk *ic = data;
354 	int			error;
355 
356 	error = ic->formatter(ic->breq, &inogrp);
357 	if (error && error != -ECANCELED)
358 		return error;
359 
360 	ic->breq->startino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino) +
361 			XFS_INODES_PER_CHUNK;
362 	return error;
363 }
364 
365 /*
366  * Return inode number table for the filesystem.
367  */
368 int
369 xfs_inumbers(
370 	struct xfs_ibulk	*breq,
371 	inumbers_fmt_pf		formatter)
372 {
373 	struct xfs_inumbers_chunk ic = {
374 		.formatter	= formatter,
375 		.breq		= breq,
376 	};
377 	int			error = 0;
378 
379 	if (xfs_bulkstat_already_done(breq->mp, breq->startino))
380 		return 0;
381 
382 	error = xfs_inobt_walk(breq->mp, NULL, breq->startino, breq->flags,
383 			xfs_inumbers_walk, breq->icount, &ic);
384 
385 	/*
386 	 * We found some inode groups, so clear the error status and return
387 	 * them.  The lastino pointer will point directly at the inode that
388 	 * triggered any error that occurred, so on the next call the error
389 	 * will be triggered again and propagated to userspace as there will be
390 	 * no formatted inode groups in the buffer.
391 	 */
392 	if (breq->ocount > 0)
393 		error = 0;
394 
395 	return error;
396 }
397 
398 /* Convert an inumbers (v5) struct to a inogrp (v1) struct. */
399 void
400 xfs_inumbers_to_inogrp(
401 	struct xfs_inogrp		*ig1,
402 	const struct xfs_inumbers	*ig)
403 {
404 	/* memset is needed here because of padding holes in the structure. */
405 	memset(ig1, 0, sizeof(struct xfs_inogrp));
406 	ig1->xi_startino = ig->xi_startino;
407 	ig1->xi_alloccount = ig->xi_alloccount;
408 	ig1->xi_allocmask = ig->xi_allocmask;
409 }
410