Re: BUG #18146: Rows reappearing in Tables after Auto-Vacuum Failure in PostgreSQL on Windows

On Mon, Sep 9, 2024 at 7:21 PM Michael Paquier <michael@paquier.xyz> wrote:
> The trick to use _mdfd_segpath() through smgrpreparetruncate() smells
> too much of magic to me.  I don't have an idea that does not involve
> tweaking the interface of smgr.c or some of its structures on top of
> my mind, though, which is to save the path before the critical section
> and pass it through (like a memory area that can be used by
> md.c)..  Or do a MemoryContextAllowInCriticalSection, even if it is
> not its original purpose.

It's funny that it's so hard to get or store the pathnames when every
single File knows its pathname, and segment 0 could give it to you:
FilePathName().  Thinking about that made me notice that the vfd layer
does raw strdup() and ERRORs on failure, so the thing I proposed might
be adhering to the letter of the law ("no palloc in critical
sections"), but not the spirit (the fd.c file opening stuff will throw
ERRORs that are promoted to PANICs, exactly what that law intended to
prevent).  Mleugh.

Another option would be to introduce mdtruncate2(old_nblocks,
new_nblocks), so that it doesn't have to call smgrnblocks() itself,
and you can obtain old_nblocks outside the CS...