1 // SPDX-License-Identifier: GPL-2.0
3 * Assorted bcachefs debug code
5 * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
6 * Copyright 2012 Google, Inc.
10 #include "bkey_methods.h"
11 #include "btree_cache.h"
13 #include "btree_iter.h"
14 #include "btree_update.h"
24 #include <linux/console.h>
25 #include <linux/debugfs.h>
26 #include <linux/module.h>
27 #include <linux/random.h>
28 #include <linux/seq_file.h>
30 static struct dentry *bch_debug;
32 static bool bch2_btree_verify_replica(struct bch_fs *c, struct btree *b,
33 struct extent_ptr_decoded pick)
35 struct btree *v = c->verify_data;
36 struct btree_node *n_ondisk = c->verify_ondisk;
37 struct btree_node *n_sorted = c->verify_data->data;
38 struct bset *sorted, *inmemory = &b->data->keys;
39 struct bch_dev *ca = bch_dev_bkey_exists(c, pick.ptr.dev);
43 if (!bch2_dev_get_ioref(ca, READ))
46 bio = bio_alloc_bioset(GFP_NOIO,
47 buf_pages(n_sorted, btree_bytes(c)),
49 bio_set_dev(bio, ca->disk_sb.bdev);
50 bio->bi_opf = REQ_OP_READ|REQ_META;
51 bio->bi_iter.bi_sector = pick.ptr.offset;
52 bch2_bio_map(bio, n_sorted, btree_bytes(c));
57 percpu_ref_put(&ca->io_ref);
59 memcpy(n_ondisk, n_sorted, btree_bytes(c));
62 if (bch2_btree_node_read_done(c, ca, v, false))
65 n_sorted = c->verify_data->data;
66 sorted = &n_sorted->keys;
68 if (inmemory->u64s != sorted->u64s ||
69 memcmp(inmemory->start,
71 vstruct_end(inmemory) - (void *) inmemory->start)) {
72 unsigned offset = 0, sectors;
78 printk(KERN_ERR "*** in memory:\n");
79 bch2_dump_bset(c, b, inmemory, 0);
81 printk(KERN_ERR "*** read back in:\n");
82 bch2_dump_bset(c, v, sorted, 0);
84 while (offset < v->written) {
87 sectors = vstruct_blocks(n_ondisk, c->block_bits) <<
90 struct btree_node_entry *bne =
91 (void *) n_ondisk + (offset << 9);
94 sectors = vstruct_blocks(bne, c->block_bits) <<
98 printk(KERN_ERR "*** on disk block %u:\n", offset);
99 bch2_dump_bset(c, b, i, offset);
104 for (j = 0; j < le16_to_cpu(inmemory->u64s); j++)
105 if (inmemory->_data[j] != sorted->_data[j])
109 bch_err(c, "verify failed at key %u", j);
114 if (v->written != b->written) {
115 bch_err(c, "written wrong: expected %u, got %u",
116 b->written, v->written);
123 void __bch2_btree_verify(struct bch_fs *c, struct btree *b)
125 struct bkey_ptrs_c ptrs;
126 struct extent_ptr_decoded p;
127 const union bch_extent_entry *entry;
129 struct bset *inmemory = &b->data->keys;
130 struct bkey_packed *k;
133 if (c->opts.nochanges)
136 bch2_btree_node_io_lock(b);
137 mutex_lock(&c->verify_lock);
139 if (!c->verify_ondisk) {
140 c->verify_ondisk = kvpmalloc(btree_bytes(c), GFP_KERNEL);
141 if (!c->verify_ondisk)
145 if (!c->verify_data) {
146 c->verify_data = __bch2_btree_node_mem_alloc(c);
150 list_del_init(&c->verify_data->list);
153 BUG_ON(b->nsets != 1);
155 for (k = inmemory->start; k != vstruct_last(inmemory); k = bkey_next(k))
156 if (k->type == KEY_TYPE_btree_ptr_v2) {
157 struct bch_btree_ptr_v2 *v = (void *) bkeyp_val(&b->format, k);
162 bkey_copy(&v->key, &b->key);
163 v->c.level = b->c.level;
164 v->c.btree_id = b->c.btree_id;
165 bch2_btree_keys_init(v);
167 ptrs = bch2_bkey_ptrs_c(bkey_i_to_s_c(&b->key));
168 bkey_for_each_ptr_decode(&b->key.k, ptrs, p, entry)
169 failed |= bch2_btree_verify_replica(c, b, p);
172 struct printbuf buf = PRINTBUF;
174 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key));
175 bch2_fs_fatal_error(c, "btree node verify failed for : %s\n", buf.buf);
179 mutex_unlock(&c->verify_lock);
180 bch2_btree_node_io_unlock(b);
183 #ifdef CONFIG_DEBUG_FS
185 /* XXX: bch_fs refcounting */
195 char __user *ubuf; /* destination user buffer */
196 size_t size; /* size of requested read */
197 ssize_t ret; /* bytes read so far */
200 static int flush_buf(struct dump_iter *i)
203 size_t bytes = min_t(size_t, i->buf.pos, i->size);
204 int err = copy_to_user(i->ubuf, i->buf.buf, bytes);
213 memmove(i->buf.buf, i->buf.buf + bytes, i->buf.pos);
219 static int bch2_dump_open(struct inode *inode, struct file *file)
221 struct btree_debug *bd = inode->i_private;
224 i = kzalloc(sizeof(struct dump_iter), GFP_KERNEL);
228 file->private_data = i;
231 i->c = container_of(bd, struct bch_fs, btree_debug[bd->id]);
238 static int bch2_dump_release(struct inode *inode, struct file *file)
240 struct dump_iter *i = file->private_data;
242 printbuf_exit(&i->buf);
247 static ssize_t bch2_read_btree(struct file *file, char __user *buf,
248 size_t size, loff_t *ppos)
250 struct dump_iter *i = file->private_data;
251 struct btree_trans trans;
252 struct btree_iter iter;
267 bch2_trans_init(&trans, i->c, 0, 0);
269 bch2_trans_iter_init(&trans, &iter, i->id, i->from,
271 BTREE_ITER_ALL_SNAPSHOTS);
272 k = bch2_btree_iter_peek(&iter);
274 while (k.k && !(err = bkey_err(k))) {
275 bch2_bkey_val_to_text(&i->buf, i->c, k);
276 pr_char(&i->buf, '\n');
278 k = bch2_btree_iter_next(&iter);
288 bch2_trans_iter_exit(&trans, &iter);
290 bch2_trans_exit(&trans);
292 return err < 0 ? err : i->ret;
295 static const struct file_operations btree_debug_ops = {
296 .owner = THIS_MODULE,
297 .open = bch2_dump_open,
298 .release = bch2_dump_release,
299 .read = bch2_read_btree,
302 static ssize_t bch2_read_btree_formats(struct file *file, char __user *buf,
303 size_t size, loff_t *ppos)
305 struct dump_iter *i = file->private_data;
306 struct btree_trans trans;
307 struct btree_iter iter;
319 if (!i->size || !bpos_cmp(SPOS_MAX, i->from))
322 bch2_trans_init(&trans, i->c, 0, 0);
324 for_each_btree_node(&trans, iter, i->id, i->from, 0, b, err) {
325 bch2_btree_node_to_text(&i->buf, i->c, b);
331 * can't easily correctly restart a btree node traversal across
334 i->from = bpos_cmp(SPOS_MAX, b->key.k.p)
335 ? bpos_successor(b->key.k.p)
341 bch2_trans_iter_exit(&trans, &iter);
343 bch2_trans_exit(&trans);
345 return err < 0 ? err : i->ret;
348 static const struct file_operations btree_format_debug_ops = {
349 .owner = THIS_MODULE,
350 .open = bch2_dump_open,
351 .release = bch2_dump_release,
352 .read = bch2_read_btree_formats,
355 static ssize_t bch2_read_bfloat_failed(struct file *file, char __user *buf,
356 size_t size, loff_t *ppos)
358 struct dump_iter *i = file->private_data;
359 struct btree_trans trans;
360 struct btree_iter iter;
362 struct btree *prev_node = NULL;
376 bch2_trans_init(&trans, i->c, 0, 0);
378 bch2_trans_iter_init(&trans, &iter, i->id, i->from,
380 BTREE_ITER_ALL_SNAPSHOTS);
382 while ((k = bch2_btree_iter_peek(&iter)).k &&
383 !(err = bkey_err(k))) {
384 struct btree_path_level *l = &iter.path->l[0];
385 struct bkey_packed *_k =
386 bch2_btree_node_iter_peek(&l->iter, l->b);
388 if (l->b != prev_node) {
389 bch2_btree_node_to_text(&i->buf, i->c, l->b);
396 bch2_bfloat_to_text(&i->buf, l->b, _k);
401 bch2_btree_iter_advance(&iter);
411 bch2_trans_iter_exit(&trans, &iter);
413 bch2_trans_exit(&trans);
415 return err < 0 ? err : i->ret;
418 static const struct file_operations bfloat_failed_debug_ops = {
419 .owner = THIS_MODULE,
420 .open = bch2_dump_open,
421 .release = bch2_dump_release,
422 .read = bch2_read_bfloat_failed,
425 static void bch2_cached_btree_node_to_text(struct printbuf *out, struct bch_fs *c,
428 out->tabstops[0] = 32;
430 pr_buf(out, "%px btree=%s l=%u ",
432 bch2_btree_ids[b->c.btree_id],
436 pr_indent_push(out, 2);
438 bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(&b->key));
441 pr_buf(out, "flags: ");
443 bch2_flags_to_text(out, bch2_btree_node_flags, b->flags);
446 pr_buf(out, "written:");
448 pr_buf(out, "%u", b->written);
451 pr_buf(out, "writes blocked:");
453 pr_buf(out, "%u", !list_empty_careful(&b->write_blocked));
456 pr_buf(out, "will make reachable:");
458 pr_buf(out, "%lx", b->will_make_reachable);
461 pr_buf(out, "journal pin %px:", &b->writes[0].journal);
463 pr_buf(out, "%llu", b->writes[0].journal.seq);
466 pr_buf(out, "journal pin %px:", &b->writes[1].journal);
468 pr_buf(out, "%llu", b->writes[1].journal.seq);
471 pr_indent_pop(out, 2);
474 static ssize_t bch2_cached_btree_nodes_read(struct file *file, char __user *buf,
475 size_t size, loff_t *ppos)
477 struct dump_iter *i = file->private_data;
478 struct bch_fs *c = i->c;
487 struct bucket_table *tbl;
488 struct rhash_head *pos;
500 tbl = rht_dereference_rcu(c->btree_cache.table.tbl,
501 &c->btree_cache.table);
502 if (i->iter < tbl->size) {
503 rht_for_each_entry_rcu(b, pos, tbl, i->iter, hash)
504 bch2_cached_btree_node_to_text(&i->buf, c, b);
513 if (i->buf.allocation_failure)
519 static const struct file_operations cached_btree_nodes_ops = {
520 .owner = THIS_MODULE,
521 .open = bch2_dump_open,
522 .release = bch2_dump_release,
523 .read = bch2_cached_btree_nodes_read,
526 static ssize_t bch2_journal_pins_read(struct file *file, char __user *buf,
527 size_t size, loff_t *ppos)
529 struct dump_iter *i = file->private_data;
530 struct bch_fs *c = i->c;
546 done = bch2_journal_seq_pins_to_text(&i->buf, &c->journal, &i->iter);
550 if (i->buf.allocation_failure)
556 static const struct file_operations journal_pins_ops = {
557 .owner = THIS_MODULE,
558 .open = bch2_dump_open,
559 .release = bch2_dump_release,
560 .read = bch2_journal_pins_read,
563 void bch2_fs_debug_exit(struct bch_fs *c)
565 if (!IS_ERR_OR_NULL(c->fs_debug_dir))
566 debugfs_remove_recursive(c->fs_debug_dir);
569 void bch2_fs_debug_init(struct bch_fs *c)
571 struct btree_debug *bd;
574 if (IS_ERR_OR_NULL(bch_debug))
577 snprintf(name, sizeof(name), "%pU", c->sb.user_uuid.b);
578 c->fs_debug_dir = debugfs_create_dir(name, bch_debug);
579 if (IS_ERR_OR_NULL(c->fs_debug_dir))
582 debugfs_create_file("cached_btree_nodes", 0400, c->fs_debug_dir,
583 c->btree_debug, &cached_btree_nodes_ops);
585 debugfs_create_file("journal_pins", 0400, c->fs_debug_dir,
586 c->btree_debug, &journal_pins_ops);
588 c->btree_debug_dir = debugfs_create_dir("btrees", c->fs_debug_dir);
589 if (IS_ERR_OR_NULL(c->btree_debug_dir))
592 for (bd = c->btree_debug;
593 bd < c->btree_debug + ARRAY_SIZE(c->btree_debug);
595 bd->id = bd - c->btree_debug;
596 debugfs_create_file(bch2_btree_ids[bd->id],
597 0400, c->btree_debug_dir, bd,
600 snprintf(name, sizeof(name), "%s-formats",
601 bch2_btree_ids[bd->id]);
603 debugfs_create_file(name, 0400, c->btree_debug_dir, bd,
604 &btree_format_debug_ops);
606 snprintf(name, sizeof(name), "%s-bfloat-failed",
607 bch2_btree_ids[bd->id]);
609 debugfs_create_file(name, 0400, c->btree_debug_dir, bd,
610 &bfloat_failed_debug_ops);
616 void bch2_debug_exit(void)
618 if (!IS_ERR_OR_NULL(bch_debug))
619 debugfs_remove_recursive(bch_debug);
622 int __init bch2_debug_init(void)
626 bch_debug = debugfs_create_dir("bcachefs", NULL);