ISAR/libmdbx
3
0
Fork 0
mirror of https://github.com/isar/libmdbx.git synced 2025-01-02 02:24:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

mdbx: fix and refine mdbx_update_gc() to avoid infinite looping possibility (squashed).

Browse Source
This commit is contained in:
Leonid Yuriev 2021-10-03 02:12:56 +03:00
parent 6737d304a6
commit fe0ec8ceca
1 changed files with 19 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
src/core.c
View File

@ -8682,7 +8682,7 @@ retry_noaccount:
mdbx_pnl_check4assert(txn->tw.reclaimed_pglist,
txn->mt_next_pgno - MDBX_ENABLE_REFUND));
mdbx_tassert(txn, mdbx_dirtylist_check(txn));
if (unlikely(/* paranoia */ loop > ((MDBX_DEBUG > 0) ? 9 : 99))) {
if (unlikely(/* paranoia */ loop > ((MDBX_DEBUG > 0) ? 12 : 42))) {
mdbx_error("too more loops %u, bailout", loop);
rc = MDBX_PROBLEM;
goto bailout;
@ -8743,8 +8743,6 @@ retry_noaccount:
/* If using records from GC which we have not yet deleted,
* now delete them and any we reserved for tw.reclaimed_pglist. */
while (cleaned_gc_id <= txn->tw.last_reclaimed) {
gc_rid = cleaned_gc_id;
settled = 0;
rc = mdbx_cursor_first(&couple.outer, &key, NULL);
if (unlikely(rc != MDBX_SUCCESS)) {
if (rc == MDBX_NOTFOUND)
@ -8756,6 +8754,9 @@ retry_noaccount:
rc = MDBX_CORRUPTED;
goto bailout;
}
gc_rid = cleaned_gc_id;
settled = 0;
reused_gc_slot = 0;
cleaned_gc_id = unaligned_peek_u64(4, key.iov_base);
if (!MDBX_DISABLE_PAGECHECKS &&
unlikely(cleaned_gc_id < MIN_TXNID || cleaned_gc_id > MAX_TXNID)) {
@ -9121,6 +9122,7 @@ retry_noaccount:
} else if (rc != MDBX_NOTFOUND)
goto bailout;
txn->tw.last_reclaimed = gc_rid;
cleaned_gc_id = gc_rid + 1;
}
reservation_gc_id = gc_rid--;
mdbx_trace("%s: take @%" PRIaTXN " from head-gc-id", dbg_prefix_mode,
@ -9198,7 +9200,7 @@ retry_noaccount:
key.iov_len = sizeof(reservation_gc_id);
key.iov_base = &reservation_gc_id;
data.iov_len = (chunk + 1) * sizeof(pgno_t);
mdbx_trace("%s.reserve: %u [%u...%u] @%" PRIaTXN, dbg_prefix_mode, chunk,
mdbx_trace("%s.reserve: %u [%u...%u) @%" PRIaTXN, dbg_prefix_mode, chunk,
settled + 1, settled + chunk + 1, reservation_gc_id);
mdbx_prep_backlog(txn, &couple.outer, data.iov_len);
rc = mdbx_cursor_put(&couple.outer, &key, &data,
@ -9374,15 +9376,21 @@ retry_noaccount:
}
mdbx_tassert(txn, rc == MDBX_SUCCESS);
if (unlikely(txn->tw.loose_count != 0 ||
filled_gc_slot !=
(txn->tw.lifo_reclaimed
? (unsigned)MDBX_PNL_SIZE(txn->tw.lifo_reclaimed)
: 0))) {
mdbx_notice("** restart: reserve excess (filled-slot %u, loose-count %u)",
filled_gc_slot, txn->tw.loose_count);
if (unlikely(txn->tw.loose_count != 0)) {
mdbx_notice("** restart: got %u loose pages", txn->tw.loose_count);
goto retry;
}
if (unlikely(filled_gc_slot !=
(txn->tw.lifo_reclaimed
? (unsigned)MDBX_PNL_SIZE(txn->tw.lifo_reclaimed)
: 0))) {
const bool will_retry = loop < 9;
mdbx_notice("** %s: reserve excess (filled-slot %u, loop %u)",
will_retry ? "restart" : "ignore", filled_gc_slot, loop);
if (will_retry)
goto retry;
}
mdbx_tassert(txn,
txn->tw.lifo_reclaimed == NULL ||