1 // SPDX-License-Identifier: GPL-2.0
2
3 #include "bcachefs.h"
4 #include "disk_groups.h"
5 #include "opts.h"
6 #include "replicas.h"
7 #include "sb-members.h"
8 #include "super-io.h"
9
10 #define x(t, n, ...) [n] = #t,
11 static const char * const bch2_iops_measurements[] = {
12 BCH_IOPS_MEASUREMENTS()
13 NULL
14 };
15
16 char * const bch2_member_error_strs[] = {
17 BCH_MEMBER_ERROR_TYPES()
18 NULL
19 };
20 #undef x
21
22 /* Code for bch_sb_field_members_v1: */
23
bch2_members_v2_get_mut(struct bch_sb * sb,int i)24 struct bch_member *bch2_members_v2_get_mut(struct bch_sb *sb, int i)
25 {
26 return __bch2_members_v2_get_mut(bch2_sb_field_get(sb, members_v2), i);
27 }
28
members_v2_get(struct bch_sb_field_members_v2 * mi,int i)29 static struct bch_member members_v2_get(struct bch_sb_field_members_v2 *mi, int i)
30 {
31 struct bch_member ret, *p = __bch2_members_v2_get_mut(mi, i);
32 memset(&ret, 0, sizeof(ret));
33 memcpy(&ret, p, min_t(size_t, le16_to_cpu(mi->member_bytes), sizeof(ret)));
34 return ret;
35 }
36
members_v1_get_mut(struct bch_sb_field_members_v1 * mi,int i)37 static struct bch_member *members_v1_get_mut(struct bch_sb_field_members_v1 *mi, int i)
38 {
39 return (void *) mi->_members + (i * BCH_MEMBER_V1_BYTES);
40 }
41
members_v1_get(struct bch_sb_field_members_v1 * mi,int i)42 static struct bch_member members_v1_get(struct bch_sb_field_members_v1 *mi, int i)
43 {
44 struct bch_member ret, *p = members_v1_get_mut(mi, i);
45 memset(&ret, 0, sizeof(ret));
46 memcpy(&ret, p, min_t(size_t, BCH_MEMBER_V1_BYTES, sizeof(ret)));
47 return ret;
48 }
49
bch2_sb_member_get(struct bch_sb * sb,int i)50 struct bch_member bch2_sb_member_get(struct bch_sb *sb, int i)
51 {
52 struct bch_sb_field_members_v2 *mi2 = bch2_sb_field_get(sb, members_v2);
53 if (mi2)
54 return members_v2_get(mi2, i);
55 struct bch_sb_field_members_v1 *mi1 = bch2_sb_field_get(sb, members_v1);
56 return members_v1_get(mi1, i);
57 }
58
sb_members_v2_resize_entries(struct bch_fs * c)59 static int sb_members_v2_resize_entries(struct bch_fs *c)
60 {
61 struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
62
63 if (le16_to_cpu(mi->member_bytes) < sizeof(struct bch_member)) {
64 unsigned u64s = DIV_ROUND_UP((sizeof(*mi) + sizeof(mi->_members[0]) *
65 c->disk_sb.sb->nr_devices), 8);
66
67 mi = bch2_sb_field_resize(&c->disk_sb, members_v2, u64s);
68 if (!mi)
69 return -BCH_ERR_ENOSPC_sb_members_v2;
70
71 for (int i = c->disk_sb.sb->nr_devices - 1; i >= 0; --i) {
72 void *dst = (void *) mi->_members + (i * sizeof(struct bch_member));
73 memmove(dst, __bch2_members_v2_get_mut(mi, i), le16_to_cpu(mi->member_bytes));
74 memset(dst + le16_to_cpu(mi->member_bytes),
75 0, (sizeof(struct bch_member) - le16_to_cpu(mi->member_bytes)));
76 }
77 mi->member_bytes = cpu_to_le16(sizeof(struct bch_member));
78 }
79 return 0;
80 }
81
bch2_sb_members_v2_init(struct bch_fs * c)82 int bch2_sb_members_v2_init(struct bch_fs *c)
83 {
84 struct bch_sb_field_members_v1 *mi1;
85 struct bch_sb_field_members_v2 *mi2;
86
87 if (!bch2_sb_field_get(c->disk_sb.sb, members_v2)) {
88 mi2 = bch2_sb_field_resize(&c->disk_sb, members_v2,
89 DIV_ROUND_UP(sizeof(*mi2) +
90 sizeof(struct bch_member) * c->sb.nr_devices,
91 sizeof(u64)));
92 mi1 = bch2_sb_field_get(c->disk_sb.sb, members_v1);
93 memcpy(&mi2->_members[0], &mi1->_members[0],
94 BCH_MEMBER_V1_BYTES * c->sb.nr_devices);
95 memset(&mi2->pad[0], 0, sizeof(mi2->pad));
96 mi2->member_bytes = cpu_to_le16(BCH_MEMBER_V1_BYTES);
97 }
98
99 return sb_members_v2_resize_entries(c);
100 }
101
bch2_sb_members_cpy_v2_v1(struct bch_sb_handle * disk_sb)102 int bch2_sb_members_cpy_v2_v1(struct bch_sb_handle *disk_sb)
103 {
104 struct bch_sb_field_members_v1 *mi1;
105 struct bch_sb_field_members_v2 *mi2;
106
107 mi1 = bch2_sb_field_resize(disk_sb, members_v1,
108 DIV_ROUND_UP(sizeof(*mi1) + BCH_MEMBER_V1_BYTES *
109 disk_sb->sb->nr_devices, sizeof(u64)));
110 if (!mi1)
111 return -BCH_ERR_ENOSPC_sb_members;
112
113 mi2 = bch2_sb_field_get(disk_sb->sb, members_v2);
114
115 for (unsigned i = 0; i < disk_sb->sb->nr_devices; i++)
116 memcpy(members_v1_get_mut(mi1, i), __bch2_members_v2_get_mut(mi2, i), BCH_MEMBER_V1_BYTES);
117
118 return 0;
119 }
120
validate_member(struct printbuf * err,struct bch_member m,struct bch_sb * sb,int i)121 static int validate_member(struct printbuf *err,
122 struct bch_member m,
123 struct bch_sb *sb,
124 int i)
125 {
126 if (le64_to_cpu(m.nbuckets) > LONG_MAX) {
127 prt_printf(err, "device %u: too many buckets (got %llu, max %lu)",
128 i, le64_to_cpu(m.nbuckets), LONG_MAX);
129 return -BCH_ERR_invalid_sb_members;
130 }
131
132 if (le64_to_cpu(m.nbuckets) -
133 le16_to_cpu(m.first_bucket) < BCH_MIN_NR_NBUCKETS) {
134 prt_printf(err, "device %u: not enough buckets (got %llu, max %u)",
135 i, le64_to_cpu(m.nbuckets), BCH_MIN_NR_NBUCKETS);
136 return -BCH_ERR_invalid_sb_members;
137 }
138
139 if (le16_to_cpu(m.bucket_size) <
140 le16_to_cpu(sb->block_size)) {
141 prt_printf(err, "device %u: bucket size %u smaller than block size %u",
142 i, le16_to_cpu(m.bucket_size), le16_to_cpu(sb->block_size));
143 return -BCH_ERR_invalid_sb_members;
144 }
145
146 if (le16_to_cpu(m.bucket_size) <
147 BCH_SB_BTREE_NODE_SIZE(sb)) {
148 prt_printf(err, "device %u: bucket size %u smaller than btree node size %llu",
149 i, le16_to_cpu(m.bucket_size), BCH_SB_BTREE_NODE_SIZE(sb));
150 return -BCH_ERR_invalid_sb_members;
151 }
152
153 return 0;
154 }
155
member_to_text(struct printbuf * out,struct bch_member m,struct bch_sb_field_disk_groups * gi,struct bch_sb * sb,int i)156 static void member_to_text(struct printbuf *out,
157 struct bch_member m,
158 struct bch_sb_field_disk_groups *gi,
159 struct bch_sb *sb,
160 int i)
161 {
162 unsigned data_have = bch2_sb_dev_has_data(sb, i);
163 u64 bucket_size = le16_to_cpu(m.bucket_size);
164 u64 device_size = le64_to_cpu(m.nbuckets) * bucket_size;
165
166 if (!bch2_member_exists(&m))
167 return;
168
169 prt_printf(out, "Device:");
170 prt_tab(out);
171 prt_printf(out, "%u", i);
172 prt_newline(out);
173
174 printbuf_indent_add(out, 2);
175
176 prt_printf(out, "Label:");
177 prt_tab(out);
178 if (BCH_MEMBER_GROUP(&m)) {
179 unsigned idx = BCH_MEMBER_GROUP(&m) - 1;
180
181 if (idx < disk_groups_nr(gi))
182 prt_printf(out, "%s (%u)",
183 gi->entries[idx].label, idx);
184 else
185 prt_printf(out, "(bad disk labels section)");
186 } else {
187 prt_printf(out, "(none)");
188 }
189 prt_newline(out);
190
191 prt_printf(out, "UUID:");
192 prt_tab(out);
193 pr_uuid(out, m.uuid.b);
194 prt_newline(out);
195
196 prt_printf(out, "Size:");
197 prt_tab(out);
198 prt_units_u64(out, device_size << 9);
199 prt_newline(out);
200
201 for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
202 prt_printf(out, "%s errors:", bch2_member_error_strs[i]);
203 prt_tab(out);
204 prt_u64(out, le64_to_cpu(m.errors[i]));
205 prt_newline(out);
206 }
207
208 for (unsigned i = 0; i < BCH_IOPS_NR; i++) {
209 prt_printf(out, "%s iops:", bch2_iops_measurements[i]);
210 prt_tab(out);
211 prt_printf(out, "%u", le32_to_cpu(m.iops[i]));
212 prt_newline(out);
213 }
214
215 prt_printf(out, "Bucket size:");
216 prt_tab(out);
217 prt_units_u64(out, bucket_size << 9);
218 prt_newline(out);
219
220 prt_printf(out, "First bucket:");
221 prt_tab(out);
222 prt_printf(out, "%u", le16_to_cpu(m.first_bucket));
223 prt_newline(out);
224
225 prt_printf(out, "Buckets:");
226 prt_tab(out);
227 prt_printf(out, "%llu", le64_to_cpu(m.nbuckets));
228 prt_newline(out);
229
230 prt_printf(out, "Last mount:");
231 prt_tab(out);
232 if (m.last_mount)
233 bch2_prt_datetime(out, le64_to_cpu(m.last_mount));
234 else
235 prt_printf(out, "(never)");
236 prt_newline(out);
237
238 prt_printf(out, "Last superblock write:");
239 prt_tab(out);
240 prt_u64(out, le64_to_cpu(m.seq));
241 prt_newline(out);
242
243 prt_printf(out, "State:");
244 prt_tab(out);
245 prt_printf(out, "%s",
246 BCH_MEMBER_STATE(&m) < BCH_MEMBER_STATE_NR
247 ? bch2_member_states[BCH_MEMBER_STATE(&m)]
248 : "unknown");
249 prt_newline(out);
250
251 prt_printf(out, "Data allowed:");
252 prt_tab(out);
253 if (BCH_MEMBER_DATA_ALLOWED(&m))
254 prt_bitflags(out, __bch2_data_types, BCH_MEMBER_DATA_ALLOWED(&m));
255 else
256 prt_printf(out, "(none)");
257 prt_newline(out);
258
259 prt_printf(out, "Has data:");
260 prt_tab(out);
261 if (data_have)
262 prt_bitflags(out, __bch2_data_types, data_have);
263 else
264 prt_printf(out, "(none)");
265 prt_newline(out);
266
267 prt_str(out, "Durability:");
268 prt_tab(out);
269 prt_printf(out, "%llu", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);
270 prt_newline(out);
271
272 prt_printf(out, "Discard:");
273 prt_tab(out);
274 prt_printf(out, "%llu", BCH_MEMBER_DISCARD(&m));
275 prt_newline(out);
276
277 prt_printf(out, "Freespace initialized:");
278 prt_tab(out);
279 prt_printf(out, "%llu", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
280 prt_newline(out);
281
282 printbuf_indent_sub(out, 2);
283 }
284
bch2_sb_members_v1_validate(struct bch_sb * sb,struct bch_sb_field * f,struct printbuf * err)285 static int bch2_sb_members_v1_validate(struct bch_sb *sb,
286 struct bch_sb_field *f,
287 struct printbuf *err)
288 {
289 struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
290 unsigned i;
291
292 if ((void *) members_v1_get_mut(mi, sb->nr_devices) > vstruct_end(&mi->field)) {
293 prt_printf(err, "too many devices for section size");
294 return -BCH_ERR_invalid_sb_members;
295 }
296
297 for (i = 0; i < sb->nr_devices; i++) {
298 struct bch_member m = members_v1_get(mi, i);
299
300 int ret = validate_member(err, m, sb, i);
301 if (ret)
302 return ret;
303 }
304
305 return 0;
306 }
307
bch2_sb_members_v1_to_text(struct printbuf * out,struct bch_sb * sb,struct bch_sb_field * f)308 static void bch2_sb_members_v1_to_text(struct printbuf *out, struct bch_sb *sb,
309 struct bch_sb_field *f)
310 {
311 struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
312 struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
313 unsigned i;
314
315 for (i = 0; i < sb->nr_devices; i++)
316 member_to_text(out, members_v1_get(mi, i), gi, sb, i);
317 }
318
319 const struct bch_sb_field_ops bch_sb_field_ops_members_v1 = {
320 .validate = bch2_sb_members_v1_validate,
321 .to_text = bch2_sb_members_v1_to_text,
322 };
323
bch2_sb_members_v2_to_text(struct printbuf * out,struct bch_sb * sb,struct bch_sb_field * f)324 static void bch2_sb_members_v2_to_text(struct printbuf *out, struct bch_sb *sb,
325 struct bch_sb_field *f)
326 {
327 struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
328 struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
329 unsigned i;
330
331 for (i = 0; i < sb->nr_devices; i++)
332 member_to_text(out, members_v2_get(mi, i), gi, sb, i);
333 }
334
bch2_sb_members_v2_validate(struct bch_sb * sb,struct bch_sb_field * f,struct printbuf * err)335 static int bch2_sb_members_v2_validate(struct bch_sb *sb,
336 struct bch_sb_field *f,
337 struct printbuf *err)
338 {
339 struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
340 size_t mi_bytes = (void *) __bch2_members_v2_get_mut(mi, sb->nr_devices) -
341 (void *) mi;
342
343 if (mi_bytes > vstruct_bytes(&mi->field)) {
344 prt_printf(err, "section too small (%zu > %zu)",
345 mi_bytes, vstruct_bytes(&mi->field));
346 return -BCH_ERR_invalid_sb_members;
347 }
348
349 for (unsigned i = 0; i < sb->nr_devices; i++) {
350 int ret = validate_member(err, members_v2_get(mi, i), sb, i);
351 if (ret)
352 return ret;
353 }
354
355 return 0;
356 }
357
358 const struct bch_sb_field_ops bch_sb_field_ops_members_v2 = {
359 .validate = bch2_sb_members_v2_validate,
360 .to_text = bch2_sb_members_v2_to_text,
361 };
362
bch2_sb_members_from_cpu(struct bch_fs * c)363 void bch2_sb_members_from_cpu(struct bch_fs *c)
364 {
365 struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
366
367 rcu_read_lock();
368 for_each_member_device_rcu(c, ca, NULL) {
369 struct bch_member *m = __bch2_members_v2_get_mut(mi, ca->dev_idx);
370
371 for (unsigned e = 0; e < BCH_MEMBER_ERROR_NR; e++)
372 m->errors[e] = cpu_to_le64(atomic64_read(&ca->errors[e]));
373 }
374 rcu_read_unlock();
375 }
376
bch2_dev_io_errors_to_text(struct printbuf * out,struct bch_dev * ca)377 void bch2_dev_io_errors_to_text(struct printbuf *out, struct bch_dev *ca)
378 {
379 struct bch_fs *c = ca->fs;
380 struct bch_member m;
381
382 mutex_lock(&ca->fs->sb_lock);
383 m = bch2_sb_member_get(c->disk_sb.sb, ca->dev_idx);
384 mutex_unlock(&ca->fs->sb_lock);
385
386 printbuf_tabstop_push(out, 12);
387
388 prt_str(out, "IO errors since filesystem creation");
389 prt_newline(out);
390
391 printbuf_indent_add(out, 2);
392 for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
393 prt_printf(out, "%s:", bch2_member_error_strs[i]);
394 prt_tab(out);
395 prt_u64(out, atomic64_read(&ca->errors[i]));
396 prt_newline(out);
397 }
398 printbuf_indent_sub(out, 2);
399
400 prt_str(out, "IO errors since ");
401 bch2_pr_time_units(out, (ktime_get_real_seconds() - le64_to_cpu(m.errors_reset_time)) * NSEC_PER_SEC);
402 prt_str(out, " ago");
403 prt_newline(out);
404
405 printbuf_indent_add(out, 2);
406 for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
407 prt_printf(out, "%s:", bch2_member_error_strs[i]);
408 prt_tab(out);
409 prt_u64(out, atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
410 prt_newline(out);
411 }
412 printbuf_indent_sub(out, 2);
413 }
414
bch2_dev_errors_reset(struct bch_dev * ca)415 void bch2_dev_errors_reset(struct bch_dev *ca)
416 {
417 struct bch_fs *c = ca->fs;
418 struct bch_member *m;
419
420 mutex_lock(&c->sb_lock);
421 m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
422 for (unsigned i = 0; i < ARRAY_SIZE(m->errors_at_reset); i++)
423 m->errors_at_reset[i] = cpu_to_le64(atomic64_read(&ca->errors[i]));
424 m->errors_reset_time = cpu_to_le64(ktime_get_real_seconds());
425
426 bch2_write_super(c);
427 mutex_unlock(&c->sb_lock);
428 }
429