On 02/09/2025 7:42 AM, Michael Paquier wrote:
On Mon, Sep 01, 2025 at 02:35:56PM -0400, Peter Geoghegan wrote:
I remember that when I worked on what became commit 9b42e71376 back in
2019 (which fixed a similar problem caused by the INCLUDE index
patch), Tom suggested that we do things this way defensively (without
being specifically aware of the _bt_getbuf hazard). That does seem
like the best approach.
I'm a little bit worried about the added overhead of allocating an
extra BLCKSZ buffer for this. I wonder if it would make sense to use
BLCKSZ arrays of char for both of the temp pages.
Hmm. Yes, the allocation makes me a bit uneasy. Or it would be
possible to just use a PGAlignedBlock here? That's already used in
some code paths for page manipulations, forcing alignment.
Do you suggest to replace `PageGetTempPage` with using local buffers?
Or may be change `PageGetTempPage*` to accept parameter with provided
local buffer?
But it is used in ~20 places. Change them all?
It seems to be too invasive change for such small fix, but I can do it.
Certainly copying data can be done in more efficient way, but I do not
thing that palloc here can have any noticeable impact on performance.
Please let me know if I should prepare new page avoiding memory
allocation or you prefer to do it yourself.
