Horton

Re: [PATCH] Identify LWLocks in tracepoints

Craig Ringer <craig.ringer@enterprisedb.com>

View thread View on postgr.es
From: Craig Ringer <craig.ringer@enterprisedb.com>
To: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
Cc: pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2021-04-12T05:46:30Z
Lists: pgsql-hackers 
On Mon, 22 Mar 2021 at 17:00, Peter Eisentraut <
peter.eisentraut@enterprisedb.com> wrote:

> On 20.03.21 01:29, Craig Ringer wrote:
> >     There is already support for that.  See the documentation at the end
> of
> >     this page:
> >
> https://www.postgresql.org/docs/devel/dynamic-trace.html#DEFINING-TRACE-POINTS
> >     <
> https://www.postgresql.org/docs/devel/dynamic-trace.html#DEFINING-TRACE-POINTS
> >
> >
> >
> > Pretty sure it won't work right now.
> >
> > To use systemtap semaphores (the _ENABLED macros) you need to run dtrace
> > -g to generate a probes.o then link that into postgres.
> >
> > I don't think we do that. I'll double check soon.
>
> We do that.  (It's -G.)
>

Huh. I could've sworn we didn't. My mistake, it's there in
src/backend/Makefile .

In that case I'll amend the patch to use semaphore guards.

(On a side note, systemtap's semaphore support is actually a massive pain.
The way it's implemented in <sys/sdt.h> means that a single compilation
unit may not use both probes.d style markers produced by the dtrace script
and use regular DTRACE_PROBE(providername,probename) preprocessor macros.
If it attempts to do so, the DTRACE_PROBE macros will emit inline asm that
tries to reference probename_semaphore symbols that will not exist,
resulting in linker errors or runtime link errors. But that's really a
systemtap problem. Core PostgreSQL doesn't use any explicit
DTRACE_PROBE(...), STAP_PROBE(...) etc.)

Commits

  1. Emit dummy statements for probes.d probes when disabled

  2. Prevent lwlock dtrace probes from unnecessary work

  3. doc: Fix typo in logicaldecoding.sgml.

  4. Move lwlock-release probe back where it belongs

  5. Remove bogus tracepoint

  6. Simplify LWLock tranche machinery by removing array_base/array_stride.

  7. Improve LWLock scalability.