Horton

Re: elog(DEBUG2 in SpinLocked section.

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>
Cc: Kyotaro Horiguchi <horikyota.ntt@gmail.com>, masao.fujii@oss.nttdata.com, amit.kapila16@gmail.com, pasim@vmware.com, pgsql-hackers@lists.postgresql.org
Date: 2020-06-04T01:57:33Z
Lists: pgsql-hackers 
Michael Paquier <michael@paquier.xyz> writes:
> On Wed, Jun 03, 2020 at 12:36:34AM -0400, Tom Lane wrote:
>> Should we think about adding automated detection of this type of
>> mistake?  I don't like the attached as-is because of the #include
>> footprint expansion, but maybe we can find a better way.

> I think that this one first boils down to the FRONTEND dependency in
> those headers.  Or in short, spin.h may get loaded by the frontend but
> we have a backend-only API, no?

I think the #include bloat comes from wanting to declare the global
state variable as "slock_t *".  We could give up on that and write
something like this in a central place like c.h:

#if defined(USE_ASSERT_CHECKING) && !defined(FRONTEND)
extern void *held_spinlock;
#define NotHoldingSpinLock() Assert(held_spinlock == NULL)
#else
#define NotHoldingSpinLock() ((void) 0)
#endif

Then throwing NotHoldingSpinLock() into relevant places costs
nothing new include-wise.

			regards, tom lane

Commits

  1. Fix instance of elog() called while holding a spinlock

  2. Don't call palloc() while holding a spinlock, either.

  3. Don't call elog() while holding spinlock.