projects / postgresql.git / commitdiff
summary | shortlog | log | commit | commitdiff | tree
raw | patch | inline | side by side (parent: 706cbed)
MemoryContextCreate: assert parent is valid and different from node.
authorTom Lane <tgl@sss.pgh.pa.us>
Mon, 21 Apr 2025 15:34:36 +0000 (11:34 -0400)
committerTom Lane <tgl@sss.pgh.pa.us>
Mon, 21 Apr 2025 15:34:36 +0000 (11:34 -0400)
The case of "node == parent" might seem impossible, since we just
allocated the new node.  But it's possible if parent is a dangling
reference to a recently-deleted context.  In fact, given aset.c's
habit of recycling contexts, it's actually rather likely if that's so.
If we'd had this assertion before, it would have simplified debugging
a recently-identified walsender issue.

Reported-by: Anthonin Bonnefoy <anthonin.bonnefoy@datadoghq.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAO6_XqoJA7-_G6t7Uqe5nWF3nj+QBGn4F6Ptp=rUGDr0zo+KvA@mail.gmail.com

src/backend/utils/mmgr/mcxt.c patch | blob | blame | history

diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c
index 68f1cbcf9c28611399f9f34d27260a0627320c61..506f29029865c0590e02ff15b07f6f921826fae6 100644 (file)
--- a/src/backend/utils/mmgr/mcxt.c
+++ b/src/backend/utils/mmgr/mcxt.c
@@ -1181,6 +1181,10 @@ MemoryContextCreate(MemoryContext node,
    /* Creating new memory contexts is not allowed in a critical section */
    Assert(CritSectionCount == 0);
 
+   /* Validate parent, to help prevent crazy context linkages */
+   Assert(parent == NULL || MemoryContextIsValid(parent));
+   Assert(node != parent);
+
    /* Initialize all standard fields of memory context header */
    node->type = tag;
    node->isReset = true;
This is the main PostgreSQL git repository.