1 .. SPDX-License-Identifier: GPL-2.0
2 .. _xfs_self_describing_metadata:
3
4 ============================
5 XFS Self Describing Metadata
6 ============================
7
8 Introduction
9 ============
10
11 The largest scalability problem facing XFS is not one of algorithmic
12 scalability, but of verification of the filesystem structure. Scalabilty of the
13 structures and indexes on disk and the algorithms for iterating them are
14 adequate for supporting PB scale filesystems with billions of inodes, however it
15 is this very scalability that causes the verification problem.
16
17 Almost all metadata on XFS is dynamically allocated. The only fixed location
18 metadata is the allocation group headers (SB, AGF, AGFL and AGI), while all
19 other metadata structures need to be discovered by walking the filesystem
20 structure in different ways. While this is already done by userspace tools for
21 validating and repairing the structure, there are limits to what they can
22 verify, and this in turn limits the supportable size of an XFS filesystem.
23
24 For example, it is entirely possible to manually use xfs_db and a bit of
25 scripting to analyse the structure of a 100TB filesystem when trying to
26 determine the root cause of a corruption problem, but it is still mainly a
27 manual task of verifying that things like single bit errors or misplaced writes
28 weren't the ultimate cause of a corruption event. It may take a few hours to a
29 few days to perform such forensic analysis, so for at this scale root cause
30 analysis is entirely possible.
31
32 However, if we scale the filesystem up to 1PB, we now have 10x as much metadata
33 to analyse and so that analysis blows out towards weeks/months of forensic work.
34 Most of the analysis work is slow and tedious, so as the amount of analysis goes
35 up, the more likely that the cause will be lost in the noise. Hence the primary
36 concern for supporting PB scale filesystems is minimising the time and effort
37 required for basic forensic analysis of the filesystem structure.
38
39
40 Self Describing Metadata
41 ========================
42
43 One of the problems with the current metadata format is that apart from the
44 magic number in the metadata block, we have no other way of identifying what it
45 is supposed to be. We can't even identify if it is the right place. Put simply,
46 you can't look at a single metadata block in isolation and say "yes, it is
47 supposed to be there and the contents are valid".
48
49 Hence most of the time spent on forensic analysis is spent doing basic
50 verification of metadata values, looking for values that are in range (and hence
51 not detected by automated verification checks) but are not correct. Finding and
52 understanding how things like cross linked block lists (e.g. sibling
53 pointers in a btree end up with loops in them) are the key to understanding what
54 went wrong, but it is impossible to tell what order the blocks were linked into
55 each other or written to disk after the fact.
56
57 Hence we need to record more information into the metadata to allow us to
58 quickly determine if the metadata is intact and can be ignored for the purpose
59 of analysis. We can't protect against every possible type of error, but we can
60 ensure that common types of errors are easily detectable. Hence the concept of
61 self describing metadata.
62
63 The first, fundamental requirement of self describing metadata is that the
64 metadata object contains some form of unique identifier in a well known
65 location. This allows us to identify the expected contents of the block and
66 hence parse and verify the metadata object. IF we can't independently identify
67 the type of metadata in the object, then the metadata doesn't describe itself
68 very well at all!
69
70 Luckily, almost all XFS metadata has magic numbers embedded already - only the
71 AGFL, remote symlinks and remote attribute blocks do not contain identifying
72 magic numbers. Hence we can change the on-disk format of all these objects to
73 add more identifying information and detect this simply by changing the magic
74 numbers in the metadata objects. That is, if it has the current magic number,
75 the metadata isn't self identifying. If it contains a new magic number, it is
76 self identifying and we can do much more expansive automated verification of the
77 metadata object at runtime, during forensic analysis or repair.
78
79 As a primary concern, self describing metadata needs some form of overall
80 integrity checking. We cannot trust the metadata if we cannot verify that it has
81 not been changed as a result of external influences. Hence we need some form of
82 integrity check, and this is done by adding CRC32c validation to the metadata
83 block. If we can verify the block contains the metadata it was intended to
84 contain, a large amount of the manual verification work can be skipped.
85
86 CRC32c was selected as metadata cannot be more than 64k in length in XFS and
87 hence a 32 bit CRC is more than sufficient to detect multi-bit errors in
88 metadata blocks. CRC32c is also now hardware accelerated on common CPUs so it is
89 fast. So while CRC32c is not the strongest of possible integrity checks that
90 could be used, it is more than sufficient for our needs and has relatively
91 little overhead. Adding support for larger integrity fields and/or algorithms
92 does really provide any extra value over CRC32c, but it does add a lot of
93 complexity and so there is no provision for changing the integrity checking
94 mechanism.
95
96 Self describing metadata needs to contain enough information so that the
97 metadata block can be verified as being in the correct place without needing to
98 look at any other metadata. This means it needs to contain location information.
99 Just adding a block number to the metadata is not sufficient to protect against
100 mis-directed writes - a write might be misdirected to the wrong LUN and so be
101 written to the "correct block" of the wrong filesystem. Hence location
102 information must contain a filesystem identifier as well as a block number.
103
104 Another key information point in forensic analysis is knowing who the metadata
105 block belongs to. We already know the type, the location, that it is valid
106 and/or corrupted, and how long ago that it was last modified. Knowing the owner
107 of the block is important as it allows us to find other related metadata to
108 determine the scope of the corruption. For example, if we have a extent btree
109 object, we don't know what inode it belongs to and hence have to walk the entire
110 filesystem to find the owner of the block. Worse, the corruption could mean that
111 no owner can be found (i.e. it's an orphan block), and so without an owner field
112 in the metadata we have no idea of the scope of the corruption. If we have an
113 owner field in the metadata object, we can immediately do top down validation to
114 determine the scope of the problem.
115
116 Different types of metadata have different owner identifiers. For example,
117 directory, attribute and extent tree blocks are all owned by an inode, while
118 freespace btree blocks are owned by an allocation group. Hence the size and
119 contents of the owner field are determined by the type of metadata object we are
120 looking at. The owner information can also identify misplaced writes (e.g.
121 freespace btree block written to the wrong AG).
122
123 Self describing metadata also needs to contain some indication of when it was
124 written to the filesystem. One of the key information points when doing forensic
125 analysis is how recently the block was modified. Correlation of set of corrupted
126 metadata blocks based on modification times is important as it can indicate
127 whether the corruptions are related, whether there's been multiple corruption
128 events that lead to the eventual failure, and even whether there are corruptions
129 present that the run-time verification is not detecting.
130
131 For example, we can determine whether a metadata object is supposed to be free
132 space or still allocated if it is still referenced by its owner by looking at
133 when the free space btree block that contains the block was last written
134 compared to when the metadata object itself was last written. If the free space
135 block is more recent than the object and the object's owner, then there is a
136 very good chance that the block should have been removed from the owner.
137
138 To provide this "written timestamp", each metadata block gets the Log Sequence
139 Number (LSN) of the most recent transaction it was modified on written into it.
140 This number will always increase over the life of the filesystem, and the only
141 thing that resets it is running xfs_repair on the filesystem. Further, by use of
142 the LSN we can tell if the corrupted metadata all belonged to the same log
143 checkpoint and hence have some idea of how much modification occurred between
144 the first and last instance of corrupt metadata on disk and, further, how much
145 modification occurred between the corruption being written and when it was
146 detected.
147
148 Runtime Validation
149 ==================
150
151 Validation of self-describing metadata takes place at runtime in two places:
152
153 - immediately after a successful read from disk
154 - immediately prior to write IO submission
155
156 The verification is completely stateless - it is done independently of the
157 modification process, and seeks only to check that the metadata is what it says
158 it is and that the metadata fields are within bounds and internally consistent.
159 As such, we cannot catch all types of corruption that can occur within a block
160 as there may be certain limitations that operational state enforces of the
161 metadata, or there may be corruption of interblock relationships (e.g. corrupted
162 sibling pointer lists). Hence we still need stateful checking in the main code
163 body, but in general most of the per-field validation is handled by the
164 verifiers.
165
166 For read verification, the caller needs to specify the expected type of metadata
167 that it should see, and the IO completion process verifies that the metadata
168 object matches what was expected. If the verification process fails, then it
169 marks the object being read as EFSCORRUPTED. The caller needs to catch this
170 error (same as for IO errors), and if it needs to take special action due to a
171 verification error it can do so by catching the EFSCORRUPTED error value. If we
172 need more discrimination of error type at higher levels, we can define new
173 error numbers for different errors as necessary.
174
175 The first step in read verification is checking the magic number and determining
176 whether CRC validating is necessary. If it is, the CRC32c is calculated and
177 compared against the value stored in the object itself. Once this is validated,
178 further checks are made against the location information, followed by extensive
179 object specific metadata validation. If any of these checks fail, then the
180 buffer is considered corrupt and the EFSCORRUPTED error is set appropriately.
181
182 Write verification is the opposite of the read verification - first the object
183 is extensively verified and if it is OK we then update the LSN from the last
184 modification made to the object, After this, we calculate the CRC and insert it
185 into the object. Once this is done the write IO is allowed to continue. If any
186 error occurs during this process, the buffer is again marked with a EFSCORRUPTED
187 error for the higher layers to catch.
188
189 Structures
190 ==========
191
192 A typical on-disk structure needs to contain the following information::
193
194 struct xfs_ondisk_hdr {
195 __be32 magic; /* magic number */
196 __be32 crc; /* CRC, not logged */
197 uuid_t uuid; /* filesystem identifier */
198 __be64 owner; /* parent object */
199 __be64 blkno; /* location on disk */
200 __be64 lsn; /* last modification in log, not logged */
201 };
202
203 Depending on the metadata, this information may be part of a header structure
204 separate to the metadata contents, or may be distributed through an existing
205 structure. The latter occurs with metadata that already contains some of this
206 information, such as the superblock and AG headers.
207
208 Other metadata may have different formats for the information, but the same
209 level of information is generally provided. For example:
210
211 - short btree blocks have a 32 bit owner (ag number) and a 32 bit block
212 number for location. The two of these combined provide the same
213 information as @owner and @blkno in eh above structure, but using 8
214 bytes less space on disk.
215
216 - directory/attribute node blocks have a 16 bit magic number, and the
217 header that contains the magic number has other information in it as
218 well. hence the additional metadata headers change the overall format
219 of the metadata.
220
221 A typical buffer read verifier is structured as follows::
222
223 #define XFS_FOO_CRC_OFF offsetof(struct xfs_ondisk_hdr, crc)
224
225 static void
226 xfs_foo_read_verify(
227 struct xfs_buf *bp)
228 {
229 struct xfs_mount *mp = bp->b_mount;
230
231 if ((xfs_sb_version_hascrc(&mp->m_sb) &&
232 !xfs_verify_cksum(bp->b_addr, BBTOB(bp->b_length),
233 XFS_FOO_CRC_OFF)) ||
234 !xfs_foo_verify(bp)) {
235 XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr);
236 xfs_buf_ioerror(bp, EFSCORRUPTED);
237 }
238 }
239
240 The code ensures that the CRC is only checked if the filesystem has CRCs enabled
241 by checking the superblock of the feature bit, and then if the CRC verifies OK
242 (or is not needed) it verifies the actual contents of the block.
243
244 The verifier function will take a couple of different forms, depending on
245 whether the magic number can be used to determine the format of the block. In
246 the case it can't, the code is structured as follows::
247
248 static bool
249 xfs_foo_verify(
250 struct xfs_buf *bp)
251 {
252 struct xfs_mount *mp = bp->b_mount;
253 struct xfs_ondisk_hdr *hdr = bp->b_addr;
254
255 if (hdr->magic != cpu_to_be32(XFS_FOO_MAGIC))
256 return false;
257
258 if (!xfs_sb_version_hascrc(&mp->m_sb)) {
259 if (!uuid_equal(&hdr->uuid, &mp->m_sb.sb_uuid))
260 return false;
261 if (bp->b_bn != be64_to_cpu(hdr->blkno))
262 return false;
263 if (hdr->owner == 0)
264 return false;
265 }
266
267 /* object specific verification checks here */
268
269 return true;
270 }
271
272 If there are different magic numbers for the different formats, the verifier
273 will look like::
274
275 static bool
276 xfs_foo_verify(
277 struct xfs_buf *bp)
278 {
279 struct xfs_mount *mp = bp->b_mount;
280 struct xfs_ondisk_hdr *hdr = bp->b_addr;
281
282 if (hdr->magic == cpu_to_be32(XFS_FOO_CRC_MAGIC)) {
283 if (!uuid_equal(&hdr->uuid, &mp->m_sb.sb_uuid))
284 return false;
285 if (bp->b_bn != be64_to_cpu(hdr->blkno))
286 return false;
287 if (hdr->owner == 0)
288 return false;
289 } else if (hdr->magic != cpu_to_be32(XFS_FOO_MAGIC))
290 return false;
291
292 /* object specific verification checks here */
293
294 return true;
295 }
296
297 Write verifiers are very similar to the read verifiers, they just do things in
298 the opposite order to the read verifiers. A typical write verifier::
299
300 static void
301 xfs_foo_write_verify(
302 struct xfs_buf *bp)
303 {
304 struct xfs_mount *mp = bp->b_mount;
305 struct xfs_buf_log_item *bip = bp->b_fspriv;
306
307 if (!xfs_foo_verify(bp)) {
308 XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr);
309 xfs_buf_ioerror(bp, EFSCORRUPTED);
310 return;
311 }
312
313 if (!xfs_sb_version_hascrc(&mp->m_sb))
314 return;
315
316
317 if (bip) {
318 struct xfs_ondisk_hdr *hdr = bp->b_addr;
319 hdr->lsn = cpu_to_be64(bip->bli_item.li_lsn);
320 }
321 xfs_update_cksum(bp->b_addr, BBTOB(bp->b_length), XFS_FOO_CRC_OFF);
322 }
323
324 This will verify the internal structure of the metadata before we go any
325 further, detecting corruptions that have occurred as the metadata has been
326 modified in memory. If the metadata verifies OK, and CRCs are enabled, we then
327 update the LSN field (when it was last modified) and calculate the CRC on the
328 metadata. Once this is done, we can issue the IO.
329
330 Inodes and Dquots
331 =================
332
333 Inodes and dquots are special snowflakes. They have per-object CRC and
334 self-identifiers, but they are packed so that there are multiple objects per
335 buffer. Hence we do not use per-buffer verifiers to do the work of per-object
336 verification and CRC calculations. The per-buffer verifiers simply perform basic
337 identification of the buffer - that they contain inodes or dquots, and that
338 there are magic numbers in all the expected spots. All further CRC and
339 verification checks are done when each inode is read from or written back to the
340 buffer.
341
342 The structure of the verifiers and the identifiers checks is very similar to the
343 buffer code described above. The only difference is where they are called. For
344 example, inode read verification is done in xfs_inode_from_disk() when the inode
345 is first read out of the buffer and the struct xfs_inode is instantiated. The
346 inode is already extensively verified during writeback in xfs_iflush_int, so the
347 only addition here is to add the LSN and CRC to the inode as it is copied back
348 into the buffer.
349
350 XXX: inode unlinked list modification doesn't recalculate the inode CRC! None of
351 the unlinked list modifications check or update CRCs, neither during unlink nor
352 log recovery. So, it's gone unnoticed until now. This won't matter immediately -
353 repair will probably complain about it - but it needs to be fixed.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.