bahman/scylla
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

partition_snapshot_row_cursor: fix a continuity loss in ensure_entry_in_latest() with reverse reads

Browse Source 
The FIXME comment claims that setting continity isn't very important in this
place, but in fact this is just wrong.

If two calls to read_from_underlying() get into a race, the one which finishes
later can call ensure_entry_in_latest() on a position which lies inside a
continuous interval in the newest version. If we don't take care to preserve
the total continuity of the version, this can punch a hole in the continuity of the
newest version, potentially reverting the affected interval to an older version.

Fix that.
This commit is contained in:
Michał Chojnowski
2023-10-31 21:02:19 +01:00
parent b5988fb389
commit 47299d6b06
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
partition_snapshot_row_cursor.hh
View File

@@ -637,11 +637,14 @@ public:
}();
rows_entry& re = *e;
if (_reversed) { // latest_i is not reliably a successor
// FIXME: set continuity when possible. Not that important since cache sets it anyway when populating.
re.set_continuous(false);
e->set_range_tombstone(range_tombstone_for_row());
rows_entry::tri_compare cmp(*_snp.schema());
auto res = rows.insert(std::move(e), cmp);
if (auto l = std::next(res.first); l != rows.end() && l->continuous()) {
re.set_continuous(true);
re.set_range_tombstone(l->range_tombstone());
}
if (res.second) {
_snp.tracker()->insert(re);
}