Horton

Thread

Commits

  1. pg_upgrade: Fix inconsistency in memory freeing

  2. pg_amcheck: Fix inconsistency in memory freeing

  3. Fix some inconsistencies with memory freeing in pg_createsubscriber

  4. Drop pre-existing subscriptions from the converted subscriber.

  1. Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-10T16:31:20Z

    Hi.

Per Coverity.

Coverity has some reports about pg_createsubcriber.

CID 1591322: (#1 of 1): Resource leak (RESOURCE_LEAK)
10. leaked_storage: Variable dbname going out of scope leaks the storage it
points to.

Additionally there are several calls that are out of the ordinary,
according to the Postgres API.

According to the documentation:
libpq-exec <https://www.postgresql.org/docs/current/libpq-exec.html>

The correct function to free memory when using PQescapeLiteral and
PQescapeIdentifier would be PQfreemem.

Trivial fixes attached.

best regards,
Ranier Vilela

  2. Re: Small memory fixes for pg_createsubcriber

    Euler Taveira <euler@eulerto.com> — 2025-02-11T16:32:32Z

    On Mon, Feb 10, 2025, at 1:31 PM, Ranier Vilela wrote:
> Coverity has some reports about pg_createsubcriber.
> 
> CID 1591322: (#1 of 1): Resource leak (RESOURCE_LEAK)
> 10. leaked_storage: Variable dbname going out of scope leaks the storage it points to.
> Additionally there are several calls that are out of the ordinary, according to the Postgres API.
> 
> According to the documentation:
> libpq-exec <https://www.postgresql.org/docs/current/libpq-exec.html>
> 
> 
> The correct function to free memory when using PQescapeLiteral and PQescapeIdentifier would be PQfreemem.
> 

Hi,

>From src/common/fe_memutils.c:

void
pg_free(void *ptr)
{
    free(ptr);
}

>From src/interfaces/libpq/fe-exec.c:

void
PQfreemem(void *ptr)
{
    free(ptr);
}

There is no bug. They are the same behind the scenes. I'm fine changing it. It
is a new code and it wouldn't cause a lot of pain to backpatch patches in the
future.


@@ -1130,6 +1130,7 @@ check_and_drop_existing_subscriptions(PGconn *conn,
 
PQclear(res);
destroyPQExpBuffer(query);
+ PQfreemem(dbname);
}

Even if the pg_createsubscriber aims to run in a small amount of time, hence,
it is fine to leak memory, the initial commit cleaned up all variables but a
subsequent commit 4867f8a555c apparently didn't. Although it is just a low
impact improvement, it is better to be strict and shut up SASTs.


--
Euler Taveira
EDB   https://www.enterprisedb.com/

  3. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-12T03:53:51Z

    On Tue, Feb 11, 2025 at 01:32:32PM -0300, Euler Taveira wrote:
> There is no bug. They are the same behind the scenes. I'm fine changing it. It
> is a new code and it wouldn't cause a lot of pain to backpatch patches in the
> future.

On consistency grounds, and as this is documented in fe-exec.c at the
top of PQfreemem(), I can get behind the switch.

> Even if the pg_createsubscriber aims to run in a small amount of time, hence,
> it is fine to leak memory, the initial commit cleaned up all variables but a
> subsequent commit 4867f8a555c apparently didn't. Although it is just a low
> impact improvement, it is better to be strict and shut up SASTs.

check_and_drop_existing_subscriptions() is called once per database in
setup_subscriber(), and we are not going to have millions of them in
this list.  We don't usually care for such short-lived things, but as
the original commit did the effort in d44032d01463, I don't see why we
cannot do it here, either.
--
Michael

  4. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-12T14:06:03Z

    Em qua., 12 de fev. de 2025 às 00:54, Michael Paquier <michael@paquier.xyz>
escreveu:

> On Tue, Feb 11, 2025 at 01:32:32PM -0300, Euler Taveira wrote:
> > There is no bug. They are the same behind the scenes. I'm fine changing
> it. It
> > is a new code and it wouldn't cause a lot of pain to backpatch patches
> in the
> > future.
>
> On consistency grounds, and as this is documented in fe-exec.c at the
> top of PQfreemem(), I can get behind the switch.
>
> > Even if the pg_createsubscriber aims to run in a small amount of time,
> hence,
> > it is fine to leak memory, the initial commit cleaned up all variables
> but a
> > subsequent commit 4867f8a555c apparently didn't. Although it is just a
> low
> > impact improvement, it is better to be strict and shut up SASTs.
>
> check_and_drop_existing_subscriptions() is called once per database in
> setup_subscriber(), and we are not going to have millions of them in
> this list.  We don't usually care for such short-lived things, but as
> the original commit did the effort in d44032d01463, I don't see why we
> cannot do it here, either.
>
Thanks Michael.

best regards,
Ranier Vilela

  5. Re: Small memory fixes for pg_createsubcriber

    Andres Freund <andres@anarazel.de> — 2025-02-12T14:34:24Z

    Hi,

On 2025-02-11 13:32:32 -0300, Euler Taveira wrote:
> On Mon, Feb 10, 2025, at 1:31 PM, Ranier Vilela wrote:
> > Coverity has some reports about pg_createsubcriber.
> > 
> > CID 1591322: (#1 of 1): Resource leak (RESOURCE_LEAK)
> > 10. leaked_storage: Variable dbname going out of scope leaks the storage it points to.
> > Additionally there are several calls that are out of the ordinary, according to the Postgres API.
> > 
> > According to the documentation:
> > libpq-exec <https://www.postgresql.org/docs/current/libpq-exec.html>
> > 
> > 
> > The correct function to free memory when using PQescapeLiteral and PQescapeIdentifier would be PQfreemem.
> > 
> 
> Hi,
> 
> From src/common/fe_memutils.c:
> 
> void
> pg_free(void *ptr)
> {
>     free(ptr);
> }
> 
> From src/interfaces/libpq/fe-exec.c:
> 
> void
> PQfreemem(void *ptr)
> {
>     free(ptr);
> }
> 
> There is no bug. They are the same behind the scenes. I'm fine changing it. It
> is a new code and it wouldn't cause a lot of pain to backpatch patches in the
> future.

That *is* a bug. On windows the allocator that a shared library (i.e. libpq)
uses, may *not* be the same as the one that an executable
(i.e. pg_createsubscriber).  It's not correct to free memory allocated in a
shared library just with free, it has to go through the library's free.

Greetings,

Andres Freund

  6. Re: Small memory fixes for pg_createsubcriber

    Tom Lane <tgl@sss.pgh.pa.us> — 2025-02-12T16:02:04Z

    Andres Freund <andres@anarazel.de> writes:
> On 2025-02-11 13:32:32 -0300, Euler Taveira wrote:
>> There is no bug. They are the same behind the scenes.

> That *is* a bug. On windows the allocator that a shared library (i.e. libpq)
> uses, may *not* be the same as the one that an executable
> (i.e. pg_createsubscriber).  It's not correct to free memory allocated in a
> shared library just with free, it has to go through the library's free.

Indeed.  This is particularly pernicious because it will work even on
Windows under common scenarios (which no doubt explains the lack of
field reports).  From the last time we discussed this [1]:

    It seems to work fine as long as a debug-readline is paired with a
    debug-psql or a release-readline is paired with a release-psql.

I wish we had some way to detect misuses automatically ...

This seems like the sort of bug that Coverity could detect if only it
knew to look, but I have no idea if it could be configured that way.
Maybe some weird lashup with a debugging malloc library would be
another way.

			regards, tom lane

[1] https://www.postgresql.org/message-id/20240709225934.746y5fg3kgxkyant@awork3.anarazel.de

  7. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-12T16:46:38Z

    Hi.

Em qua., 12 de fev. de 2025 às 13:02, Tom Lane <tgl@sss.pgh.pa.us> escreveu:

> Andres Freund <andres@anarazel.de> writes:
> > On 2025-02-11 13:32:32 -0300, Euler Taveira wrote:
> >> There is no bug. They are the same behind the scenes.
>
> > That *is* a bug. On windows the allocator that a shared library (i.e.
> libpq)
> > uses, may *not* be the same as the one that an executable
> > (i.e. pg_createsubscriber).  It's not correct to free memory allocated
> in a
> > shared library just with free, it has to go through the library's free.
>
> Indeed.  This is particularly pernicious because it will work even on
> Windows under common scenarios (which no doubt explains the lack of
> field reports).  From the last time we discussed this [1]:
>
>     It seems to work fine as long as a debug-readline is paired with a
>     debug-psql or a release-readline is paired with a release-psql.
>
> I wish we had some way to detect misuses automatically ...
>
Unfortunately, for now, there is only manual search.

In fact, after your post, I did a new search and found two other
occurrences.
(src/bin/pg_amcheck/pg_amcheck.c)

One fixed in the patch attached.
Another *dat->amcheck_schema* is it not worth the effort,
since the memory is not released at any point?

Trivial patch attached.

best regards,
Ranier Vilela

  8. Re: Small memory fixes for pg_createsubcriber

    Andres Freund <andres@anarazel.de> — 2025-02-12T17:00:03Z

    Hi,

On 2025-02-12 11:02:04 -0500, Tom Lane wrote:
> I wish we had some way to detect misuses automatically ...
>
> This seems like the sort of bug that Coverity could detect if only it
> knew to look, but I have no idea if it could be configured that way.
> Maybe some weird lashup with a debugging malloc library would be
> another way.

Recent gcc actually has a fairly good way to detect this kind of issue:
https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#index-malloc-function-attribute
in particular, the variant of the attribute with "deallocator".

I'd started on a patch to add those, but ran out of cycles for 18.


I quickly checked out my branch and added the relevant attributes to
PQescapeLiteral(), PQescapeIdentifier() and that indeed finds the issue
pointed out in this thread:

../../../../../home/andres/src/postgresql/src/bin/pg_basebackup/pg_createsubscriber.c: In function 'create_logical_replication_slot':
../../../../../home/andres/src/postgresql/src/bin/pg_basebackup/pg_createsubscriber.c:1332:9: warning: 'pg_free' called on pointer returned from a mismatched allocation function [-Wmismatched-dealloc]
 1332 |         pg_free(slot_name_esc);
      |         ^~~~~~~~~~~~~~~~~~~~~~
../../../../../home/andres/src/postgresql/src/bin/pg_basebackup/pg_createsubscriber.c:1326:25: note: returned from 'PQescapeLiteral'
 1326 |         slot_name_esc = PQescapeLiteral(conn, slot_name, strlen(slot_name));
      |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...


but also finds one more:
[62/214 42  28%] Compiling C object src/bin/pg_amcheck/pg_amcheck.p/pg_amcheck.c.o
../../../../../home/andres/src/postgresql/src/bin/pg_amcheck/pg_amcheck.c: In function 'main':
../../../../../home/andres/src/postgresql/src/bin/pg_amcheck/pg_amcheck.c:563:25: warning: 'pfree' called on pointer returned from a mismatched allocation function [-Wmismatched-dealloc]
  563 |                         pfree(schema);
      |                         ^~~~~~~~~~~~~
../../../../../home/andres/src/postgresql/src/bin/pg_amcheck/pg_amcheck.c:556:34: note: returned from 'PQescapeIdentifier'
  556 |                         schema = PQescapeIdentifier(conn, opts.install_schema,
      |                                  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  557 |                                                                                 strlen(opts.install_schema));
      |                                                                                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~



Note that that doesn't just require adding the attributes to
PQescapeIdentifier() etc, but also to pg_malloc(), as the latter is how gcc
knows that pg_pfree() is a deallocating function.


Particularly for something like libpq it's not quitetrivial to add
attributes like this, of course. We can't even depend on pg_config.h.

One way would be to define them in libpq-fe.h, guarded by an #ifdef, that's
"armed" by a commandline -D flag, if the compiler is supported?

Greetings,

Andres Freund

  9. Re: Small memory fixes for pg_createsubcriber

    Dagfinn Ilmari Mannsåker <ilmari@ilmari.org> — 2025-02-12T17:17:49Z

    Andres Freund <andres@anarazel.de> writes:

> Hi,
>
> On 2025-02-12 11:02:04 -0500, Tom Lane wrote:
>> I wish we had some way to detect misuses automatically ...
>>
>> This seems like the sort of bug that Coverity could detect if only it
>> knew to look, but I have no idea if it could be configured that way.
>> Maybe some weird lashup with a debugging malloc library would be
>> another way.
>
> Recent gcc actually has a fairly good way to detect this kind of issue:
> https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#index-malloc-function-attribute
> in particular, the variant of the attribute with "deallocator".

[...]

> Note that that doesn't just require adding the attributes to
> PQescapeIdentifier() etc, but also to pg_malloc(), as the latter is how gcc
> knows that pg_pfree() is a deallocating function.
>
>
> Particularly for something like libpq it's not quitetrivial to add
> attributes like this, of course. We can't even depend on pg_config.h.
>
> One way would be to define them in libpq-fe.h, guarded by an #ifdef, that's
> "armed" by a commandline -D flag, if the compiler is supported?

Does it need a -D flag, wouldn't __has_attribute(malloc) (with the
fallback definition in c.h) be enough?

- ilmari

  10. Re: Small memory fixes for pg_createsubcriber

    Tom Lane <tgl@sss.pgh.pa.us> — 2025-02-12T17:23:17Z

    =?utf-8?Q?Dagfinn_Ilmari_Manns=C3=A5ker?= <ilmari@ilmari.org> writes:
> Andres Freund <andres@anarazel.de> writes:
>> Particularly for something like libpq it's not quitetrivial to add
>> attributes like this, of course. We can't even depend on pg_config.h.
>> One way would be to define them in libpq-fe.h, guarded by an #ifdef, that's
>> "armed" by a commandline -D flag, if the compiler is supported?

> Does it need a -D flag, wouldn't __has_attribute(malloc) (with the
> fallback definition in c.h) be enough?

libpq-fe.h has to be compilable by application code that has never
heard of pg_config.h let alone c.h, so we'd have to tread carefully
about not breaking that property.  But it seems like this would be
worth looking into.

			regards, tom lane

  11. Re: Small memory fixes for pg_createsubcriber

    Dagfinn Ilmari Mannsåker <ilmari@ilmari.org> — 2025-02-12T17:29:57Z

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

> =?utf-8?Q?Dagfinn_Ilmari_Manns=C3=A5ker?= <ilmari@ilmari.org> writes:
>> Andres Freund <andres@anarazel.de> writes:
>>> Particularly for something like libpq it's not quitetrivial to add
>>> attributes like this, of course. We can't even depend on pg_config.h.
>>> One way would be to define them in libpq-fe.h, guarded by an #ifdef, that's
>>> "armed" by a commandline -D flag, if the compiler is supported?
>
>> Does it need a -D flag, wouldn't __has_attribute(malloc) (with the
>> fallback definition in c.h) be enough?
>
> libpq-fe.h has to be compilable by application code that has never
> heard of pg_config.h let alone c.h, so we'd have to tread carefully
> about not breaking that property.  But it seems like this would be
> worth looking into.

The fallback code isn't exactly complicated, so we could just duplicate
it in libpq-fe.h:

#ifndef __has_attribute
#define __has_attribute(attribute) 0
#endif

And then do something like this:

#if __has_attribute (malloc)
#define pg_attribute_malloc __attribute__((malloc))
#define pg_attribute_deallocator(...) __attribute__((malloc(__VA_ARGS__)))
#else
#define pg_attribute_malloc
#define pg_attribute_deallocator(...)
#endif

> 			regards, tom lane

- ilmari

  12. Re: Small memory fixes for pg_createsubcriber

    Tom Lane <tgl@sss.pgh.pa.us> — 2025-02-12T17:38:21Z

    =?utf-8?Q?Dagfinn_Ilmari_Manns=C3=A5ker?= <ilmari@ilmari.org> writes:
> Tom Lane <tgl@sss.pgh.pa.us> writes:
>> libpq-fe.h has to be compilable by application code that has never
>> heard of pg_config.h let alone c.h, so we'd have to tread carefully
>> about not breaking that property.  But it seems like this would be
>> worth looking into.

> The fallback code isn't exactly complicated, so we could just duplicate
> it in libpq-fe.h:

> #ifndef __has_attribute
> #define __has_attribute(attribute) 0
> #endif

The problem with that approach is the likelihood of stomping on
symbols that a calling application will use later.  I think we
really need a controlling #ifdef check on some PG_FOO symbol
that we can be sure no outside application will have defined.
Roughly speaking,

#ifdef PG_USE_ALLOCATOR_CHECKS
#define pg_attribute_malloc __attribute__((malloc))
...
#else
#define pg_attribute_malloc
...
#endif

and then we could make definition of PG_USE_ALLOCATOR_CHECKS
be conditional on having the right compiler behavior, rather
than trusting that a nest of #ifdef checks is sufficient to
detect that.

			regards, tom lane

  13. Re: Small memory fixes for pg_createsubcriber

    Tom Lane <tgl@sss.pgh.pa.us> — 2025-02-12T18:35:20Z

    I experimented with the other approach: hack libpq.so to depend on
dmalloc, leaving the rest of the system alone, so that libpq's
allocations could not be freed elsewhere nor vice versa.

It could not even get through initdb, crashing here:

replace_guc_value(char **lines, const char *guc_name, const char *guc_value,
                  bool mark_as_comment)
{
    PQExpBuffer newline = createPQExpBuffer();

    ...
    free(newline);            /* but don't free newline->data */

which upon investigation is expected, because createPQExpBuffer is
exported by libpq and therefore is returning space allocated within
the shlib.  Diking out that particular free() call just allows it
to crash a bit later on, because initdb is totally full of direct
manipulations of PQExpBuffer contents.

This indicates to me that we have a *far* larger problem than
we thought, if we'd like to be totally clean on this point.

Realistically, it's not going to happen that way; it's just
not worth the trouble and notational mess.  I think if we're
honest we should just document that such-and-such combinations
of libpq and our client programs will work on Windows.

For amusement's sake, totally dirty hack-and-slash patch attached.
(I tested this on macOS, with dmalloc from MacPorts; adjust SHLIB_LINK
to suit on other platforms.)

			regards, tom lane

  14. Re: Small memory fixes for pg_createsubcriber

    Dagfinn Ilmari Mannsåker <ilmari@ilmari.org> — 2025-02-12T18:49:52Z

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

> =?utf-8?Q?Dagfinn_Ilmari_Manns=C3=A5ker?= <ilmari@ilmari.org> writes:
>> Tom Lane <tgl@sss.pgh.pa.us> writes:
>>> libpq-fe.h has to be compilable by application code that has never
>>> heard of pg_config.h let alone c.h, so we'd have to tread carefully
>>> about not breaking that property.  But it seems like this would be
>>> worth looking into.
>
>> The fallback code isn't exactly complicated, so we could just duplicate
>> it in libpq-fe.h:
>
>> #ifndef __has_attribute
>> #define __has_attribute(attribute) 0
>> #endif
>
> The problem with that approach is the likelihood of stomping on
> symbols that a calling application will use later.  I think we
> really need a controlling #ifdef check on some PG_FOO symbol
> that we can be sure no outside application will have defined.
> Roughly speaking,
>
> #ifdef PG_USE_ALLOCATOR_CHECKS

As long as we're not checking for too many attributes, we could avoid
introducing the __has_attribute symbol by doing:

#if defined(__has_attribute) && __has_attribute(malloc)

> #define pg_attribute_malloc __attribute__((malloc))
> ...
> #else
> #define pg_attribute_malloc
> ...
> #endif
>
> and then we could make definition of PG_USE_ALLOCATOR_CHECKS
> be conditional on having the right compiler behavior, rather
> than trusting that a nest of #ifdef checks is sufficient to
> detect that.

But I just looked at the clang attribute docs
(https://clang.llvm.org/docs/AttributeReference.html#malloc), and it
only has the argumentless version, not the one that that declares the
matching deallocator, so you're right we need a more comprehensive
check.

> 			regards, tom lane

- ilmari

  15. Re: Small memory fixes for pg_createsubcriber

    Tom Lane <tgl@sss.pgh.pa.us> — 2025-02-12T21:17:53Z

    Ranier Vilela <ranier.vf@gmail.com> writes:
> Coverity has some reports about pg_createsubcriber.

> CID 1591322: (#1 of 1): Resource leak (RESOURCE_LEAK)
> 10. leaked_storage: Variable dbname going out of scope leaks the storage it
> points to.

FTR, the security team's Coverity instance also complained about that.
I was planning to fix it after the release freeze lifted, but you
beat me to it, which is fine.  Our report turned up a couple other
things that I just pushed fixes for.

(It seems like Coverity must've updated their rules recently,
because we also got a bunch of false-positive reports that were
not there before, mostly in pre-existing code.)

			regards, tom lane

  16. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-12T23:56:53Z

    On Wed, Feb 12, 2025 at 12:00:03PM -0500, Andres Freund wrote:
> Particularly for something like libpq it's not quitetrivial to add
> attributes like this, of course. We can't even depend on pg_config.h.
> 
> One way would be to define them in libpq-fe.h, guarded by an #ifdef, that's
> "armed" by a commandline -D flag, if the compiler is supported?

Interesting set of tricks.

I have looked at bit at the uses of PQescapeLiteral() and
PQescapeIdentifier() in the tree.  On top of the one in pg_amcheck you
are just pointing to, there is an inconsistency in pg_upgrade.c for
set_locale_and_encoding() where datlocale_literal may be allocated
with a pg_strdup() or a PQescapeLiteral() depending on the path.  The
code has been using PQfreemem() for the pg_strdup() allocation, which
is logically incorrect.

The pg_upgrade path is fancy as designed this way, for sure, but
that's less invasive than hardcoding three times "NULL".  Any thoughts
about backpatching something like that for both of them?
--
Michael

  17. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-12T23:59:58Z

    On Wed, Feb 12, 2025 at 01:35:20PM -0500, Tom Lane wrote:
> For amusement's sake, totally dirty hack-and-slash patch attached.
> (I tested this on macOS, with dmalloc from MacPorts; adjust SHLIB_LINK
> to suit on other platforms.)

Ugh.  Fun one.
--
Michael

  18. Re: Small memory fixes for pg_createsubcriber

    Tom Lane <tgl@sss.pgh.pa.us> — 2025-02-13T00:08:31Z

    Michael Paquier <michael@paquier.xyz> writes:
> I have looked at bit at the uses of PQescapeLiteral() and
> PQescapeIdentifier() in the tree.  On top of the one in pg_amcheck you
> are just pointing to, there is an inconsistency in pg_upgrade.c for
> set_locale_and_encoding() where datlocale_literal may be allocated
> with a pg_strdup() or a PQescapeLiteral() depending on the path.  The
> code has been using PQfreemem() for the pg_strdup() allocation, which
> is logically incorrect.

Yeah, I suspected there would be places like that.  It just hasn't
mattered in practice up to now.  (I have a vague recollection that
Windows used to be pickier about this, but evidently not in recent
years.)

I spent a little time earlier today seeing what I could do with the
use-dmalloc patch I posted earlier.  It turns out you can get through
initdb after s/free/PQfreemem/ in just two places, and then the
backend works fine.  But psql is a frickin' disaster --- there's
free's of strings made with PQExpBuffer all over its backslash-command
handling, and no easy way to clean it up.  Maybe other clients will
be less of a mess, but I'm not betting on that.

			regards, tom lane

  19. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-13T04:50:29Z

    On Wed, Feb 12, 2025 at 07:08:31PM -0500, Tom Lane wrote:
> I spent a little time earlier today seeing what I could do with the
> use-dmalloc patch I posted earlier.  It turns out you can get through
> initdb after s/free/PQfreemem/ in just two places, and then the
> backend works fine.  But psql is a frickin' disaster --- there's
> free's of strings made with PQExpBuffer all over its backslash-command
> handling, and no easy way to clean it up.  Maybe other clients will
> be less of a mess, but I'm not betting on that.

Hmm.  Okay.  It sounds like it would be better to group everything
that has been pointed together towards what should be a more generic
solution than what I have posted.  So I am holding on touching
anything.
--
Michael

  20. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-13T11:16:47Z

    Em qua., 12 de fev. de 2025 às 18:17, Tom Lane <tgl@sss.pgh.pa.us> escreveu:

> Ranier Vilela <ranier.vf@gmail.com> writes:
> > Coverity has some reports about pg_createsubcriber.
>
> > CID 1591322: (#1 of 1): Resource leak (RESOURCE_LEAK)
> > 10. leaked_storage: Variable dbname going out of scope leaks the storage
> it
> > points to.
>
> FTR, the security team's Coverity instance also complained about that.
> I was planning to fix it after the release freeze lifted, but you
> beat me to it, which is fine.  Our report turned up a couple other
> things that I just pushed fixes for.
>
Yeah, I see the commits, thanks for that.
I still have some reports that I could post that Coverity thinks are bugs.
They are not, but I think it is worth the effort to fix them because the
code is confusing.
I think it would improve readability and future maintainability.


>
> (It seems like Coverity must've updated their rules recently,
> because we also got a bunch of false-positive reports that were
> not there before, mostly in pre-existing code.)
>
 I believe they are trying to innovate at some point.
Many of these false positives come from a risky coding style,
I am much more cautious in my analyses.

best regards,
Ranier Vilela

  21. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-25T06:02:24Z

    On Thu, Feb 13, 2025 at 01:50:29PM +0900, Michael Paquier wrote:
> On Wed, Feb 12, 2025 at 07:08:31PM -0500, Tom Lane wrote:
>> I spent a little time earlier today seeing what I could do with the
>> use-dmalloc patch I posted earlier.  It turns out you can get through
>> initdb after s/free/PQfreemem/ in just two places, and then the
>> backend works fine.  But psql is a frickin' disaster --- there's
>> free's of strings made with PQExpBuffer all over its backslash-command
>> handling, and no easy way to clean it up.  Maybe other clients will
>> be less of a mess, but I'm not betting on that.
> 
> Hmm.  Okay.  It sounds like it would be better to group everything
> that has been pointed together towards what should be a more generic
> solution than what I have posted.  So I am holding on touching
> anything.

Two weeks later.  Would there be any objections for doing something in
the lines of [1] for pg_upgrade and pg_amcheck?  The pg_upgrade bit is
a bit strange, for sure, still it's better than the current state of
things.

[1]: https://www.postgresql.org/message-id/Z601RQxTmIUohdkV@paquier.xyz
--
Michael

  22. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-25T11:54:31Z

    Em ter., 25 de fev. de 2025 às 03:02, Michael Paquier <michael@paquier.xyz>
escreveu:

> On Thu, Feb 13, 2025 at 01:50:29PM +0900, Michael Paquier wrote:
> > On Wed, Feb 12, 2025 at 07:08:31PM -0500, Tom Lane wrote:
> >> I spent a little time earlier today seeing what I could do with the
> >> use-dmalloc patch I posted earlier.  It turns out you can get through
> >> initdb after s/free/PQfreemem/ in just two places, and then the
> >> backend works fine.  But psql is a frickin' disaster --- there's
> >> free's of strings made with PQExpBuffer all over its backslash-command
> >> handling, and no easy way to clean it up.  Maybe other clients will
> >> be less of a mess, but I'm not betting on that.
> >
> > Hmm.  Okay.  It sounds like it would be better to group everything
> > that has been pointed together towards what should be a more generic
> > solution than what I have posted.  So I am holding on touching
> > anything.
>
> Two weeks later.  Would there be any objections for doing something in
> the lines of [1] for pg_upgrade and pg_amcheck?

I prefer not to mix api.

We already have a branch for decision.
Let's use it then.

diff --git a/src/bin/pg_upgrade/pg_upgrade.c
b/src/bin/pg_upgrade/pg_upgrade.c
index d95c491fb5..b2c8e8e420 100644
--- a/src/bin/pg_upgrade/pg_upgrade.c
+++ b/src/bin/pg_upgrade/pg_upgrade.c
@@ -455,7 +455,9 @@ set_locale_and_encoding(void)
  locale->db_locale,
  strlen(locale->db_locale));
  else
- datlocale_literal = pg_strdup("NULL");
+ datlocale_literal = PQescapeLiteral(conn_new_template1,
+ "NULL",
+ strlen("NULL"));

best regards,
Ranier Vilela

  23. Re: Small memory fixes for pg_createsubcriber

    Andres Freund <andres@anarazel.de> — 2025-02-25T12:39:28Z

    Hi,

On 2025-02-12 19:08:31 -0500, Tom Lane wrote:
> Michael Paquier <michael@paquier.xyz> writes:
> > I have looked at bit at the uses of PQescapeLiteral() and
> > PQescapeIdentifier() in the tree.  On top of the one in pg_amcheck you
> > are just pointing to, there is an inconsistency in pg_upgrade.c for
> > set_locale_and_encoding() where datlocale_literal may be allocated
> > with a pg_strdup() or a PQescapeLiteral() depending on the path.  The
> > code has been using PQfreemem() for the pg_strdup() allocation, which
> > is logically incorrect.
> 
> Yeah, I suspected there would be places like that.  It just hasn't
> mattered in practice up to now.  (I have a vague recollection that
> Windows used to be pickier about this, but evidently not in recent
> years.)

It's a question of compiler / link options. If you e.g. have a debug psql
runtime linked against a non-debug libpq, the allocators for both will
differ. Or if you link an application statically while a library is built for
a shared C runtime. Or a few other such things.

But when building in the BF / CI we'll not typically encounter this issue
between libraries and binaries in our source tree, because they'll all be
built the same.

However, we have more recently hit this with external libraries we link
to. While it's not recommended, we commonly build (in BF/CI) a debug postgres
against production openssl, zstd etc.  Which means they don't share
allocators, file descriptors etc.

Greetings,

Andres Freund

  24. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-27T05:50:39Z

    On Tue, Feb 25, 2025 at 08:54:31AM -0300, Ranier Vilela wrote:
> @@ -455,7 +455,9 @@ set_locale_and_encoding(void)
>   locale->db_locale,
>   strlen(locale->db_locale));
>   else
> - datlocale_literal = pg_strdup("NULL");
> + datlocale_literal = PQescapeLiteral(conn_new_template1,
> + "NULL",
> + strlen("NULL"));

Yeah, I've considered that but hardcoding NULL twice felt a bit weird,
as well.  Perhaps it's slightly cleaner to use an intermediate
variable given then as an argument of PQescapeLiteral()?
--
Michael

  25. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-27T13:23:31Z

    Em qui., 27 de fev. de 2025 às 02:51, Michael Paquier <michael@paquier.xyz>
escreveu:

> On Tue, Feb 25, 2025 at 08:54:31AM -0300, Ranier Vilela wrote:
> > @@ -455,7 +455,9 @@ set_locale_and_encoding(void)
> >   locale->db_locale,
> >   strlen(locale->db_locale));
> >   else
> > - datlocale_literal = pg_strdup("NULL");
> > + datlocale_literal = PQescapeLiteral(conn_new_template1,
> > + "NULL",
> > + strlen("NULL"));
>
> Yeah, I've considered that but hardcoding NULL twice felt a bit weird,
> as well.  Perhaps it's slightly cleaner to use an intermediate
> variable given then as an argument of PQescapeLiteral()?
>
Yeah, I also think it would look good like this.

v1 attached.

best regards,
Ranier Vilela

  26. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-02-28T01:18:53Z

    On Thu, Feb 27, 2025 at 10:23:31AM -0300, Ranier Vilela wrote:
> Yeah, I also think it would look good like this.

It's the least confusing option, indeed.  I've reduced a bit the diffs
and done that down to v16 for the pg_upgrade part where this exists.

Double-checking the tree, it does not seem that we have simolar holes
now..  I hope that I'm not wrong.
--
Michael

  27. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-02-28T11:03:22Z

    Em qui., 27 de fev. de 2025 às 22:19, Michael Paquier <michael@paquier.xyz>
escreveu:

> On Thu, Feb 27, 2025 at 10:23:31AM -0300, Ranier Vilela wrote:
> > Yeah, I also think it would look good like this.
>
> It's the least confusing option, indeed.  I've reduced a bit the diffs
> and done that down to v16 for the pg_upgrade part where this exists.
>
Thanks Michael.


> Double-checking the tree, it does not seem that we have simolar holes
> now..  I hope that I'm not wrong.
>
I think so.

best regards,
Ranier Vilela

  28. Re: Small memory fixes for pg_createsubcriber

    Noah Misch <noah@leadboat.com> — 2025-04-01T18:39:51Z

    On Thu, Feb 27, 2025 at 10:23:31AM -0300, Ranier Vilela wrote:
> Em qui., 27 de fev. de 2025 às 02:51, Michael Paquier <michael@paquier.xyz>
> escreveu:
> 
> > On Tue, Feb 25, 2025 at 08:54:31AM -0300, Ranier Vilela wrote:
> > > @@ -455,7 +455,9 @@ set_locale_and_encoding(void)
> > >   locale->db_locale,
> > >   strlen(locale->db_locale));
> > >   else
> > > - datlocale_literal = pg_strdup("NULL");
> > > + datlocale_literal = PQescapeLiteral(conn_new_template1,
> > > + "NULL",
> > > + strlen("NULL"));
> >
> > Yeah, I've considered that but hardcoding NULL twice felt a bit weird,
> > as well.  Perhaps it's slightly cleaner to use an intermediate
> > variable given then as an argument of PQescapeLiteral()?
> >
> Yeah, I also think it would look good like this.

This became commit 2a083ab "pg_upgrade: Fix inconsistency in memory freeing".
PQescapeLiteral("NULL") is "'NULL'", so this causes pg_database.datlocale to
contain datlocale='NULL'::text instead of datlocale IS NULL.

  29. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-04-01T18:53:08Z

    Em ter., 1 de abr. de 2025 às 15:39, Noah Misch <noah@leadboat.com>
escreveu:

> On Thu, Feb 27, 2025 at 10:23:31AM -0300, Ranier Vilela wrote:
> > Em qui., 27 de fev. de 2025 às 02:51, Michael Paquier <
> michael@paquier.xyz>
> > escreveu:
> >
> > > On Tue, Feb 25, 2025 at 08:54:31AM -0300, Ranier Vilela wrote:
> > > > @@ -455,7 +455,9 @@ set_locale_and_encoding(void)
> > > >   locale->db_locale,
> > > >   strlen(locale->db_locale));
> > > >   else
> > > > - datlocale_literal = pg_strdup("NULL");
> > > > + datlocale_literal = PQescapeLiteral(conn_new_template1,
> > > > + "NULL",
> > > > + strlen("NULL"));
> > >
> > > Yeah, I've considered that but hardcoding NULL twice felt a bit weird,
> > > as well.  Perhaps it's slightly cleaner to use an intermediate
> > > variable given then as an argument of PQescapeLiteral()?
> > >
> > Yeah, I also think it would look good like this.
>
> This became commit 2a083ab "pg_upgrade: Fix inconsistency in memory
> freeing".
> PQescapeLiteral("NULL") is "'NULL'", so this causes pg_database.datlocale
> to
> contain datlocale='NULL'::text instead of datlocale IS NULL.
>
Yeah, thanks for pointing this out.
The patch maintained the previous behavior.
I'll take a look.

best regards,
Ranier Vilela

  30. Re: Small memory fixes for pg_createsubcriber

    Ranier Vilela <ranier.vf@gmail.com> — 2025-04-01T19:28:34Z

    Em ter., 1 de abr. de 2025 às 15:39, Noah Misch <noah@leadboat.com>
escreveu:

> On Thu, Feb 27, 2025 at 10:23:31AM -0300, Ranier Vilela wrote:
> > Em qui., 27 de fev. de 2025 às 02:51, Michael Paquier <
> michael@paquier.xyz>
> > escreveu:
> >
> > > On Tue, Feb 25, 2025 at 08:54:31AM -0300, Ranier Vilela wrote:
> > > > @@ -455,7 +455,9 @@ set_locale_and_encoding(void)
> > > >   locale->db_locale,
> > > >   strlen(locale->db_locale));
> > > >   else
> > > > - datlocale_literal = pg_strdup("NULL");
> > > > + datlocale_literal = PQescapeLiteral(conn_new_template1,
> > > > + "NULL",
> > > > + strlen("NULL"));
> > >
> > > Yeah, I've considered that but hardcoding NULL twice felt a bit weird,
> > > as well.  Perhaps it's slightly cleaner to use an intermediate
> > > variable given then as an argument of PQescapeLiteral()?
> > >
> > Yeah, I also think it would look good like this.
>
> This became commit 2a083ab "pg_upgrade: Fix inconsistency in memory
> freeing".
> PQescapeLiteral("NULL") is "'NULL'", so this causes pg_database.datlocale
> to
> contain datlocale='NULL'::text instead of datlocale IS NULL.
>
I believe the intention of the query is:
For example:
UPDATE pg_catalog.pg_database
SET encoding = 6,
datlocprovider = 'c',
datlocale = NULL
WHERE datname = 'template0';

Where datlocale with NULL is correct  no?

Because:
UPDATE pg_catalog.pg_database
SET encoding = 6,
datlocprovider = 'c',
datlocale IS NULL
WHERE datname = 'template0';

ERROR:  syntax error at or near "IS"
LINE 4: datlocale IS NULL

best regards,
Ranier Vilela

  31. Re: Small memory fixes for pg_createsubcriber

    Noah Misch <noah@leadboat.com> — 2025-04-01T20:14:03Z

    On Tue, Apr 01, 2025 at 04:28:34PM -0300, Ranier Vilela wrote:
> Em ter., 1 de abr. de 2025 às 15:39, Noah Misch <noah@leadboat.com>
> escreveu:
> 
> > On Thu, Feb 27, 2025 at 10:23:31AM -0300, Ranier Vilela wrote:
> > > Em qui., 27 de fev. de 2025 às 02:51, Michael Paquier <
> > michael@paquier.xyz>
> > > escreveu:
> > >
> > > > On Tue, Feb 25, 2025 at 08:54:31AM -0300, Ranier Vilela wrote:
> > > > > @@ -455,7 +455,9 @@ set_locale_and_encoding(void)
> > > > >   locale->db_locale,
> > > > >   strlen(locale->db_locale));
> > > > >   else
> > > > > - datlocale_literal = pg_strdup("NULL");
> > > > > + datlocale_literal = PQescapeLiteral(conn_new_template1,
> > > > > + "NULL",
> > > > > + strlen("NULL"));

> > This became  2a083ab "pg_upgrade: Fix inconsistency in memory
> > freeing".
> > PQescapeLiteral("NULL") is "'NULL'", so this causes pg_database.datlocale
> > to
> > contain datlocale='NULL'::text instead of datlocale IS NULL.
> >
> I believe the intention of the query is:
> For example:
> UPDATE pg_catalog.pg_database
> SET encoding = 6,
> datlocprovider = 'c',
> datlocale = NULL
> WHERE datname = 'template0';
> 
> Where datlocale with NULL is correct  no?

Right.  Commit 2a083ab changed it to:

  UPDATE pg_catalog.pg_database
  SET encoding = 6,
  datlocprovider = 'c',
  datlocale = 'NULL'
  WHERE datname = 'template0';

> Because:
> UPDATE pg_catalog.pg_database
> SET encoding = 6,
> datlocprovider = 'c',
> datlocale IS NULL
> WHERE datname = 'template0';
> 
> ERROR:  syntax error at or near "IS"
> LINE 4: datlocale IS NULL

Yes, pg_upgrade should not do that.  I was trying to explain the difference
between NULL and 'NULL'.  I didn't mean pg_upgrade should emit "IS NULL".

  32. Re: Small memory fixes for pg_createsubcriber

    Michael Paquier <michael@paquier.xyz> — 2025-04-01T23:22:47Z

      On Apr 2, 2025, at 5:14, Noah Misch <noah@leadboat.com> wrote:


  
    Yes, pg_upgrade should not do that.  I was trying to explain the difference

    between NULL and 'NULL'.  I didn't mean pg_upgrade should emit "IS NULL".

  


(Apologies for the probably-weirdly-formatted message)

  



  This issue has been already mentioned around here, with patches exchanged:


  
    
    
      
        
          
            
          
        
        
          
            
              
                
                  
                    
                      
                        Re: pgsql: pg_upgrade: Fix inconsistency in memory freeing
                      
                      
                        postgresql.org
                      
                    
                  
                
              
            
          
        
      
    
  



  --
  
    Michael