X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fbtree_io.c;h=86415701b824076f10fc79e8bee220ed7075e328;hb=e773e86495534ede8efc0f3059cf0939c374a925;hp=93a548cd393d707039509ea1a441cdc27b66665e;hpb=f0334bbc1032b3be485767ec804ec868a8b1e7f0;p=bcachefs-tools-debian diff --git a/libbcachefs/btree_io.c b/libbcachefs/btree_io.c index 93a548c..8641570 100644 --- a/libbcachefs/btree_io.c +++ b/libbcachefs/btree_io.c @@ -103,7 +103,7 @@ static void btree_bounce_free(struct bch_fs *c, size_t size, if (used_mempool) mempool_free(p, &c->btree_bounce_pool); else - vpfree(p, size); + kvfree(p); } static void *btree_bounce_alloc(struct bch_fs *c, size_t size, @@ -112,10 +112,10 @@ static void *btree_bounce_alloc(struct bch_fs *c, size_t size, unsigned flags = memalloc_nofs_save(); void *p; - BUG_ON(size > btree_bytes(c)); + BUG_ON(size > c->opts.btree_node_size); *used_mempool = false; - p = vpmalloc(size, __GFP_NOWARN|GFP_NOWAIT); + p = kvmalloc(size, __GFP_NOWARN|GFP_NOWAIT); if (!p) { *used_mempool = true; p = mempool_alloc(&c->btree_bounce_pool, GFP_NOFS); @@ -174,8 +174,8 @@ static void bch2_sort_whiteouts(struct bch_fs *c, struct btree *b) ptrs = ptrs_end = ((void *) new_whiteouts + bytes); - for (k = unwritten_whiteouts_start(c, b); - k != unwritten_whiteouts_end(c, b); + for (k = unwritten_whiteouts_start(b); + k != unwritten_whiteouts_end(b); k = bkey_p_next(k)) *--ptrs = k; @@ -192,7 +192,7 @@ static void bch2_sort_whiteouts(struct bch_fs *c, struct btree *b) verify_no_dups(b, new_whiteouts, (void *) ((u64 *) new_whiteouts + b->whiteout_u64s)); - memcpy_u64s(unwritten_whiteouts_start(c, b), + memcpy_u64s(unwritten_whiteouts_start(b), new_whiteouts, b->whiteout_u64s); btree_bounce_free(c, bytes, used_mempool, new_whiteouts); @@ -313,7 +313,7 @@ static void btree_node_sort(struct bch_fs *c, struct btree *b, } bytes = sorting_entire_node - ? btree_bytes(c) + ? btree_buf_bytes(b) : __vstruct_bytes(struct btree_node, u64s); out = btree_bounce_alloc(c, bytes, &used_mempool); @@ -327,7 +327,7 @@ static void btree_node_sort(struct bch_fs *c, struct btree *b, BUG_ON(vstruct_end(&out->keys) > (void *) out + bytes); if (sorting_entire_node) - bch2_time_stats_update(&c->times[BCH_TIME_btree_node_sort], + time_stats_update(&c->times[BCH_TIME_btree_node_sort], start_time); /* Make sure we preserve bset journal_seq: */ @@ -338,7 +338,7 @@ static void btree_node_sort(struct bch_fs *c, struct btree *b, if (sorting_entire_node) { u64s = le16_to_cpu(out->keys.u64s); - BUG_ON(bytes != btree_bytes(c)); + BUG_ON(bytes != btree_buf_bytes(b)); /* * Our temporary buffer is the same size as the btree node's @@ -397,7 +397,7 @@ void bch2_btree_sort_into(struct bch_fs *c, &dst->format, true); - bch2_time_stats_update(&c->times[BCH_TIME_btree_node_sort], + time_stats_update(&c->times[BCH_TIME_btree_node_sort], start_time); set_btree_bset_end(dst, dst->set); @@ -502,7 +502,7 @@ void bch2_btree_init_next(struct btree_trans *trans, struct btree *b) bne = want_new_bset(c, b); if (bne) - bch2_bset_init_next(c, b, bne); + bch2_bset_init_next(b, bne); bch2_btree_build_aux_trees(b); @@ -524,7 +524,8 @@ static void btree_err_msg(struct printbuf *out, struct bch_fs *c, prt_printf(out, "at btree "); bch2_btree_pos_to_text(out, c, b); - prt_printf(out, "\n node offset %u", b->written); + prt_printf(out, "\n node offset %u/%u", + b->written, btree_ptr_sectors_written(&b->key)); if (i) prt_printf(out, " bset u64s %u", le16_to_cpu(i->u64s)); prt_str(out, ": "); @@ -580,8 +581,7 @@ static int __btree_err(int ret, break; case -BCH_ERR_btree_node_read_err_bad_node: bch2_print_string_as_lines(KERN_ERR, out.buf); - bch2_topology_error(c); - ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology) ?: -EIO; + ret = bch2_topology_error(c); break; case -BCH_ERR_btree_node_read_err_incompatible: bch2_print_string_as_lines(KERN_ERR, out.buf); @@ -830,6 +830,23 @@ static int bset_key_invalid(struct bch_fs *c, struct btree *b, (rw == WRITE ? bch2_bkey_val_invalid(c, k, READ, err) : 0); } +static bool __bkey_valid(struct bch_fs *c, struct btree *b, + struct bset *i, struct bkey_packed *k) +{ + if (bkey_p_next(k) > vstruct_last(i)) + return false; + + if (k->format > KEY_FORMAT_CURRENT) + return false; + + struct printbuf buf = PRINTBUF; + struct bkey tmp; + struct bkey_s u = __bkey_disassemble(b, k, &tmp); + bool ret = __bch2_bkey_invalid(c, u.s_c, btree_node_type(b), READ, &buf); + printbuf_exit(&buf); + return ret; +} + static int validate_bset_keys(struct bch_fs *c, struct btree *b, struct bset *i, int write, bool have_retry, bool *saw_error) @@ -845,6 +862,7 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b, k != vstruct_last(i);) { struct bkey_s u; struct bkey tmp; + unsigned next_good_key; if (btree_err_on(bkey_p_next(k) > vstruct_last(i), -BCH_ERR_btree_node_read_err_fixable, @@ -859,12 +877,8 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b, -BCH_ERR_btree_node_read_err_fixable, c, NULL, b, i, btree_node_bkey_bad_format, - "invalid bkey format %u", k->format)) { - i->u64s = cpu_to_le16(le16_to_cpu(i->u64s) - k->u64s); - memmove_u64s_down(k, bkey_p_next(k), - (u64 *) vstruct_end(i) - (u64 *) k); - continue; - } + "invalid bkey format %u", k->format)) + goto drop_this_key; /* XXX: validate k->u64s */ if (!write) @@ -885,11 +899,7 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b, c, NULL, b, i, btree_node_bad_bkey, "invalid bkey: %s", buf.buf); - - i->u64s = cpu_to_le16(le16_to_cpu(i->u64s) - k->u64s); - memmove_u64s_down(k, bkey_p_next(k), - (u64 *) vstruct_end(i) - (u64 *) k); - continue; + goto drop_this_key; } if (write) @@ -906,21 +916,45 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b, prt_printf(&buf, " > "); bch2_bkey_to_text(&buf, u.k); - bch2_dump_bset(c, b, i, 0); - if (btree_err(-BCH_ERR_btree_node_read_err_fixable, c, NULL, b, i, btree_node_bkey_out_of_order, - "%s", buf.buf)) { - i->u64s = cpu_to_le16(le16_to_cpu(i->u64s) - k->u64s); - memmove_u64s_down(k, bkey_p_next(k), - (u64 *) vstruct_end(i) - (u64 *) k); - continue; - } + "%s", buf.buf)) + goto drop_this_key; } prev = k; k = bkey_p_next(k); + continue; +drop_this_key: + next_good_key = k->u64s; + + if (!next_good_key || + (BSET_BIG_ENDIAN(i) == CPU_BIG_ENDIAN && + version >= bcachefs_metadata_version_snapshot)) { + /* + * only do scanning if bch2_bkey_compat() has nothing to + * do + */ + + if (!__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) { + for (next_good_key = 1; + next_good_key < (u64 *) vstruct_last(i) - (u64 *) k; + next_good_key++) + if (__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) + goto got_good_key; + + } + + /* + * didn't find a good key, have to truncate the rest of + * the bset + */ + next_good_key = (u64 *) vstruct_last(i) - (u64 *) k; + } +got_good_key: + le16_add_cpu(&i->u64s, -next_good_key); + memmove_u64s_down(k, bkey_p_next(k), (u64 *) vstruct_end(i) - (u64 *) k); } fsck_err: printbuf_exit(&buf); @@ -942,6 +976,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, unsigned ptr_written = btree_ptr_sectors_written(&b->key); struct printbuf buf = PRINTBUF; int ret = 0, retry_read = 0, write = READ; + u64 start_time = local_clock(); b->version_ondisk = U16_MAX; /* We might get called multiple times on read retry: */ @@ -1006,8 +1041,8 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, nonce = btree_nonce(i, b->written << 9); - csum_bad = bch2_crc_cmp(b->data->csum, - csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, b->data)); + struct bch_csum csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, b->data); + csum_bad = bch2_crc_cmp(b->data->csum, csum); if (csum_bad) bch2_io_error(ca, BCH_MEMBER_ERROR_checksum); @@ -1015,7 +1050,10 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, -BCH_ERR_btree_node_read_err_want_retry, c, ca, b, i, bset_bad_csum, - "invalid checksum"); + "%s", + (printbuf_reset(&buf), + bch2_csum_err_msg(&buf, BSET_CSUM_TYPE(i), b->data->csum, csum), + buf.buf)); ret = bset_encrypt(c, i, b->written << 9); if (bch2_fs_fatal_err_on(ret, c, @@ -1044,8 +1082,8 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, "unknown checksum type %llu", BSET_CSUM_TYPE(i)); nonce = btree_nonce(i, b->written << 9); - csum_bad = bch2_crc_cmp(bne->csum, - csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, bne)); + struct bch_csum csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, bne); + csum_bad = bch2_crc_cmp(bne->csum, csum); if (csum_bad) bch2_io_error(ca, BCH_MEMBER_ERROR_checksum); @@ -1053,7 +1091,10 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, -BCH_ERR_btree_node_read_err_want_retry, c, ca, b, i, bset_bad_csum, - "invalid checksum"); + "%s", + (printbuf_reset(&buf), + bch2_csum_err_msg(&buf, BSET_CSUM_TYPE(i), bne->csum, csum), + buf.buf)); ret = bset_encrypt(c, i, b->written << 9); if (bch2_fs_fatal_err_on(ret, c, @@ -1118,7 +1159,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, ptr_written, b->written); } else { for (bne = write_block(b); - bset_byte_offset(b, bne) < btree_bytes(c); + bset_byte_offset(b, bne) < btree_buf_bytes(b); bne = (void *) bne + block_bytes(c)) btree_err_on(bne->keys.seq == b->data->keys.seq && !bch2_journal_seq_is_blacklisted(c, @@ -1130,7 +1171,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, "found bset signature after last bset"); } - sorted = btree_bounce_alloc(c, btree_bytes(c), &used_mempool); + sorted = btree_bounce_alloc(c, btree_buf_bytes(b), &used_mempool); sorted->keys.u64s = 0; set_btree_bset(b, b->set, &b->data->keys); @@ -1146,7 +1187,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, BUG_ON(b->nr.live_u64s != u64s); - btree_bounce_free(c, btree_bytes(c), used_mempool, sorted); + btree_bounce_free(c, btree_buf_bytes(b), used_mempool, sorted); if (updated_range) bch2_btree_node_drop_keys_outside_node(b); @@ -1209,6 +1250,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca, out: mempool_free(iter, &c->fill_iter); printbuf_exit(&buf); + time_stats_update(&c->times[BCH_TIME_btree_node_read_done], start_time); return retry_read; fsck_err: if (ret == -BCH_ERR_btree_node_read_err_want_retry || @@ -1241,7 +1283,7 @@ static void btree_node_read_work(struct work_struct *work) rb->have_ioref = bch2_dev_get_ioref(ca, READ); bio_reset(bio, NULL, REQ_OP_READ|REQ_SYNC|REQ_META); bio->bi_iter.bi_sector = rb->pick.ptr.offset; - bio->bi_iter.bi_size = btree_bytes(c); + bio->bi_iter.bi_size = btree_buf_bytes(b); if (rb->have_ioref) { bio_set_dev(bio, ca->disk_sb.bdev); @@ -1280,7 +1322,7 @@ start: } } - bch2_time_stats_update(&c->times[BCH_TIME_btree_node_read], + time_stats_update(&c->times[BCH_TIME_btree_node_read], rb->start_time); bio_put(&rb->bio); @@ -1469,7 +1511,7 @@ fsck_err: } if (best >= 0) { - memcpy(b->data, ra->buf[best], btree_bytes(c)); + memcpy(b->data, ra->buf[best], btree_buf_bytes(b)); ret = bch2_btree_node_read_done(c, NULL, b, false, saw_error); } else { ret = -1; @@ -1535,7 +1577,7 @@ static int btree_node_read_all_replicas(struct bch_fs *c, struct btree *b, bool for (i = 0; i < ra->nr; i++) { ra->buf[i] = mempool_alloc(&c->btree_bounce_pool, GFP_NOFS); ra->bio[i] = bio_alloc_bioset(NULL, - buf_pages(ra->buf[i], btree_bytes(c)), + buf_pages(ra->buf[i], btree_buf_bytes(b)), REQ_OP_READ|REQ_SYNC|REQ_META, GFP_NOFS, &c->btree_bio); @@ -1555,7 +1597,7 @@ static int btree_node_read_all_replicas(struct bch_fs *c, struct btree *b, bool rb->pick = pick; rb->bio.bi_iter.bi_sector = pick.ptr.offset; rb->bio.bi_end_io = btree_node_read_all_replicas_endio; - bch2_bio_map(&rb->bio, ra->buf[i], btree_bytes(c)); + bch2_bio_map(&rb->bio, ra->buf[i], btree_buf_bytes(b)); if (rb->have_ioref) { this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_btree], @@ -1622,7 +1664,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b, ca = bch_dev_bkey_exists(c, pick.ptr.dev); bio = bio_alloc_bioset(NULL, - buf_pages(b->data, btree_bytes(c)), + buf_pages(b->data, btree_buf_bytes(b)), REQ_OP_READ|REQ_SYNC|REQ_META, GFP_NOFS, &c->btree_bio); @@ -1636,7 +1678,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b, INIT_WORK(&rb->work, btree_node_read_work); bio->bi_iter.bi_sector = pick.ptr.offset; bio->bi_end_io = btree_node_read_endio; - bch2_bio_map(bio, b->data, btree_bytes(c)); + bch2_bio_map(bio, b->data, btree_buf_bytes(b)); if (rb->have_ioref) { this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_btree], @@ -1645,7 +1687,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b, if (sync) { submit_bio_wait(bio); - + bch2_latency_acct(ca, rb->start_time, READ); btree_node_read_work(&rb->work); } else { submit_bio(bio); @@ -1694,7 +1736,7 @@ static int __bch2_btree_root_read(struct btree_trans *trans, enum btree_id id, list_move(&b->list, &c->btree_cache.freeable); mutex_unlock(&c->btree_cache.lock); - ret = -EIO; + ret = -BCH_ERR_btree_node_read_error; goto err; } @@ -1797,8 +1839,10 @@ static void btree_node_write_work(struct work_struct *work) bch2_bkey_drop_ptrs(bkey_i_to_s(&wbio->key), ptr, bch2_dev_list_has_dev(wbio->wbio.failed, ptr->dev)); - if (!bch2_bkey_nr_ptrs(bkey_i_to_s_c(&wbio->key))) + if (!bch2_bkey_nr_ptrs(bkey_i_to_s_c(&wbio->key))) { + ret = -BCH_ERR_btree_node_write_all_failed; goto err; + } if (wbio->wbio.first_btree_write) { if (wbio->wbio.failed.nr) { @@ -2029,8 +2073,8 @@ do_write: i->u64s = 0; sort_iter_add(&sort_iter.iter, - unwritten_whiteouts_start(c, b), - unwritten_whiteouts_end(c, b)); + unwritten_whiteouts_start(b), + unwritten_whiteouts_end(b)); SET_BSET_SEPARATE_WHITEOUTS(i, false); b->whiteout_u64s = 0; @@ -2206,7 +2250,7 @@ bool bch2_btree_post_write_cleanup(struct bch_fs *c, struct btree *b) bne = want_new_bset(c, b); if (bne) - bch2_bset_init_next(c, b, bne); + bch2_bset_init_next(b, bne); bch2_btree_build_aux_trees(b);