Fix bug in bulk extending temp relation after failure

A ResourceOwnerEnlarge() call was missing. That led to an error:

ERROR:  ResourceOwnerRemember called but array was full

and an assertion failure, if you tried to extend a temp relation again
after a failure. Alexander's test case used running out of disk space
to trigger the original failure.

This bug was introduced in the large ResourceOwner rewrite commit
b8bff07daa. Before that, the UnpinLocalBuffer() call guaranteed that
the subsequent PinLocalBuffer() will succeed, but after the rewrite,
releasing an old resource doesn't guarantee that there is space for a
new one.

Add a comment explaining why the UnpinBuffer + PinBuffer calls in
BufferAlloc(), with no ResourceOwnerEnlarge() in between, are safe.

Reported-by: Alexander Lakhin
Discussion: https://www.postgresql.org/message-id/dc574fea-c83e-a600-08cd-10881762e4fa@gmail.com
This commit is contained in:
Heikki Linnakangas 2024-02-02 21:12:30 +02:00
parent 7e0ade0ffe
commit d212957254
2 changed files with 11 additions and 1 deletions

View File

@ -374,6 +374,9 @@ ExtendBufferedRelLocal(BufferManagerRelation bmr,
victim_buf_id = -buffers[i] - 1;
victim_buf_hdr = GetLocalBufferDescriptor(victim_buf_id);
/* in case we need to pin an existing buffer below */
ResourceOwnerEnlarge(CurrentResourceOwner);
InitBufferTag(&tag, &bmr.smgr->smgr_rlocator.locator, fork, first_block + i);
hresult = (LocalBufferLookupEnt *)
@ -646,6 +649,8 @@ InitLocalBuffers(void)
* XXX: We could have a slightly more efficient version of PinLocalBuffer()
* that does not support adjusting the usagecount - but so far it does not
* seem worth the trouble.
*
* Note that ResourceOwnerEnlarge() must have been done already.
*/
bool
PinLocalBuffer(BufferDesc *buf_hdr, bool adjust_usagecount)

View File

@ -548,8 +548,13 @@ ResourceOwnerRemember(ResourceOwner owner, Datum value, const ResourceOwnerDesc
/*
* Forget that an object is owned by a ResourceOwner
*
* Note: if same resource ID is associated with the ResourceOwner more than
* Note: If same resource ID is associated with the ResourceOwner more than
* once, one instance is removed.
*
* Note: Forgetting a resource does not guarantee that there is room to
* remember a new resource. One exception is when you forget the most
* recently remembered resource; that does make room for a new remember call.
* Some code callers rely on that exception.
*/
void
ResourceOwnerForget(ResourceOwner owner, Datum value, const ResourceOwnerDesc *kind)