Fix potential integer handling issue in radixtree.h.
authorMasahiko Sawada <msawada@postgresql.org>
Mon, 25 Mar 2024 03:06:41 +0000 (12:06 +0900)
committerMasahiko Sawada <msawada@postgresql.org>
Mon, 25 Mar 2024 03:06:41 +0000 (12:06 +0900)
Coverity complained about the integer handling issue; if we start with
an arbitrary non-negative shift value, the loop may decrement it down
to something less than zero before exiting. This commit adds an
assertion to make sure the 'shift' is always 0 after the loop, and
uses 0 as the shift to get the key chunk in the following operation.

Introduced by ee1b30f12.

Reported-by: Tom Lane as per coverity
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/2089517.1711299216%40sss.pgh.pa.us

src/include/lib/radixtree.h

index b8ad51c14dd816d85d2c76fd48cb36665be24ee6..5fe89134a5ef349edcefb4a3a100e60bdb21f5ef 100644 (file)
@@ -1615,10 +1615,11 @@ RT_EXTEND_DOWN(RT_RADIX_TREE * tree, RT_PTR_ALLOC * parent_slot, uint64 key, int
                node = child;
                shift -= RT_SPAN;
        }
+       Assert(shift == 0);
 
        /* Reserve slot for the value. */
        n4 = (RT_NODE_4 *) node.local;
-       n4->chunks[0] = RT_GET_KEY_CHUNK(key, shift);
+       n4->chunks[0] = RT_GET_KEY_CHUNK(key, 0);
        n4->base.count = 1;
 
        return &n4->children[0];