Re: [PATCH] BUG #5566: High levels of savepoint nesting trigger stack overflow in AssignTransactionId

On Monday 19 July 2010 17:58:06 Andres Freund wrote:
> On Monday 19 July 2010 17:26:25 Hans van Kranenburg wrote:
> > When issuing an update statement in a transaction with ~30800 levels of
> > savepoint nesting, (which is insane, but possible), postgresql segfaults
> > due to a stack overflow in the AssignTransactionId function, which
> > recursively assign transaction ids to parent transactions.
>
> It seems easy enough to throw a check_stack_depth() in there - survives
> make check here.
>
> It would be nice to check this earlier on though - or simply impose a
> artificial limit of nested transactions. I severely doubt that there are
> non- bug situations with a nesting of above 1k (so maybe set the limit to
> 3k).
>
> Thats independent from checking stack depth there though - it sounds
> possible to get there after an already relatively deep stack usage (deeply
> nested functions or such).
A patch doing both is attached. No idea whether 3k is a good limit.

A small testprogramm is also attached.

Andres