[PATCH] Make configuration file "include" directive handling more robust

Hello.

At Wed, 17 Jul 2019 12:29:43 +0900, Ian Barwick <ian.barwick@2ndquadrant.com> wrote in <8c8bcbca-3bd9-dc6e-8986-04a5abdef142@2ndquadrant.com>

Hi

While poking about with [1], I noticed a few potential issues with the
inclusion handling for configuration files; another issue is
demonstrated in [2].

[1]
/messages/by-id/aed6cc9f-98f3-2693-ac81-52bb0052307e@2ndquadrant.com
("Stop ALTER SYSTEM from making bad assumptions")
[2]
/messages/by-id/CY4PR1301MB22001D3FAAB3499C5D41DE23A9E50@CY4PR1301MB2200.namprd13.prod.outlook.com
("First Time Starting Up PostgreSQL and Having Problems")

Yeah! That's annoying..

Specifically these are:

1) Provision of empty include directives
=========================================

The default postgresql.conf file includes these thusly:

#include_dir = '' # include files ending in '.conf' from
# a directory, e.g., 'conf.d'
#include_if_exists = '' # include file only if it exists
#include = '' # include file

Currently, uncommenting them but leaving the value empty (as happened
in [2] above) can
result in unexpected behaviour.

For "include" and "include_if_exists", it's a not critical issue as
non-existent
files are, well, non-existent, however this will leave behind the
cryptic
message "input in flex scanner failed" in pg_file_settings's "error"
column, e.g.:

postgres=# SELECT sourceline, seqno, name, setting, applied, error
FROM pg_file_settings
WHERE error IS NOT NULL;
sourceline | seqno | name | setting | applied | error
------------+-------+------+---------+---------+------------------------------
1 | 45 | | | f | input in flex scanner failed
1 | 46 | | | f | input in flex scanner failed
(2 rows)

However, an empty value for "include_dir" will result in the current
configuration
file's directory being read, which can result in circular inclusion
and triggering
of the nesting depth check.

Patch {1} makes provision of an empty value for any of these
directives cause
configuration file processing to report an approprate error, e.g.:

postgres=# SELECT sourceline, seqno, name, setting, applied, error
FROM pg_file_settings
WHERE error IS NOT NULL;
sourceline | seqno | name | setting | applied | error
------------+-------+------+---------+---------+---------------------------------------
757 | 45 | | | f | "include" must not be empty
758 | 46 | | | f | "include_if_exists" must not be empty
759 | 47 | | | f | "include_dir" must not be empty

The patch 1 looks somewhat superficial. All the problems are
reduced to creating unexpected filepath for
inclusion. AbsoluteConfigLocation does the core work, and it can
issue generic error message covering all the cases like:

invalid parameter "<param>" at <calling_file>:<calling_lineno>

which seems sufficient. (The function needs some additional
parameters.)

2) Circular inclusion of configuration files
============================================

Currently there is a simple maximum nesting threshold (currently 10)
which
will stop runaway circular inclusions. However, if triggered, it's not
always obvious what triggered it, and sometimes resource exhaustion
might kick in beforehand (as appeared to be the case in [2] above).

Patch {2} attempts to handle this situation by keeping track of which
files have already been included (based on their absolute, canonical
path) and reporting an error if they were encountered again.

This seems to me to be overkill. The issue [2] is prevented by
the patch 1's amendment. (I don't think it's not worth donig to
add protection from explicit inclusion of pg_hba.conf from
postgresql.conf or itself or such like.)

On server startup:

2019-07-11 09:13:25.610 GMT [71140] LOG: configuration file
"/var/lib/pgsql/data/postgresql.conf" was previously parsed
2019-07-11 09:13:25.610 GMT [71140] FATAL: configuration file
"/var/lib/pgsql/data/postgresql.conf" contains errors

After sending SIGHUP:

postgres=# SELECT sourceline, seqno, name, setting, applied, error
FROM pg_file_settings WHERE error IS NOT NULL;
sourceline | seqno | name | setting | applied | error
------------+-------+------+---------+---------+--------------------------------------------------------------------------------
757 | 45 | | | f | configuration file
"/var/lib/pgsql/data/postgresql.conf" was previously parsed
(1 row)

3) "include" directives in postgresql.auto.conf and extension control
files
===========================================================================

Currently these are parsed and acted on, even though it makes no sense
for further
config files to be included in either case.

Anyway manual edit is explicitly prohibited for auto.conf. And,
even if it is added, the 10-depth limitation would protect from
infinite loop.

With "postgresql.auto.conf", if a file is successfully included, its
contents
will then be written to "postgresql.auto.conf" and the include
directive will be
removed, which seems like a recipe for confusion.

These are admittedly unlikely corner cases, but it's easy enough to
stop this
happening on the offchance someone tries to use this to solve some
problem in
completely the wrong way.

Patch {3} implements this (note that this patch depends on patch {2}).

Extension example:

postgres=# CREATE EXTENSION repmgr;
ERROR: "include" not permitted in file
"/home/barwick/devel/postgres/builds/HEAD/share/extension/repmgr.control"
line 8
postgres=# CREATE EXTENSION repmgr;
ERROR: "include_dir" not permitted in file
"/home/barwick/devel/postgres/builds/HEAD/share/extension/repmgr.control"
line 9
postgres=# CREATE EXTENSION repmgr;
ERROR: "include_if_exists" not permitted in file
"/home/barwick/devel/postgres/builds/HEAD/share/extension/repmgr.control"
line 10

pg.auto.conf example:

postgres=# ALTER SYSTEM SET default_tablespace ='pg_default';
ERROR: could not parse contents of file "postgresql.auto.conf"
postgres=# SELECT regexp_replace(sourcefile, '^/.+/','') AS
sourcefile,
seqno, name, setting, applied, error
FROM pg_file_settings WHERE error IS NOT NULL;
sourcefile | seqno | name | setting | applied | error
----------------------+-------+------+---------+---------+-------------------------
postgresql.auto.conf | 45 | | | f | "include" not permitted
(1 row)

The patch also has the side-effect that "include" directives are no
longer
(silently) removed from "postgresql.auto.conf"; as the only way they
can get
into the file in the first place is by manually editing it, I feel
it's
reasonable for the user to be made aware that they're not valid and
have to
manually remove them.

Patches
=======

Code:

{1} disallow-empty-include-directives.v1.patch
{2} track-included-files.v1.patch
{3} prevent-disallowed-includes.v1.patch

TAP tests:
{1} tap-test-configuration.v1.patch
{2} tap-test-disallow-empty-include-directives.v1.patch
{3} tap-test-track-included-files.v1.patch
{4} tap-test-prevent-disallowed-includes.v1.patch

Patches apply cleanly to REL_12_STABLE/HEAD, they could be modfied for
all supported versions if required. I can consolidate the patches
if preferred.

I don't think this is new to 12.

regards.

--
Kyotaro Horiguchi
NTT Open Source Software Center

On 7/17/19 5:34 PM, Kyotaro Horiguchi wrote:> Hello.

At Wed, 17 Jul 2019 12:29:43 +0900, Ian Barwick <ian.barwick@2ndquadrant.com> wrote in <8c8bcbca-3bd9-dc6e-8986-04a5abdef142@2ndquadrant.com>

Hi

While poking about with [1], I noticed a few potential issues with the
inclusion handling for configuration files; another issue is
demonstrated in [2].

[1]
/messages/by-id/aed6cc9f-98f3-2693-ac81-52bb0052307e@2ndquadrant.com
("Stop ALTER SYSTEM from making bad assumptions")
[2]
/messages/by-id/CY4PR1301MB22001D3FAAB3499C5D41DE23A9E50@CY4PR1301MB2200.namprd13.prod.outlook.com
("First Time Starting Up PostgreSQL and Having Problems")

Yeah! That's annoying..

Specifically these are:

1) Provision of empty include directives
=========================================

The default postgresql.conf file includes these thusly:

#include_dir = '' # include files ending in '.conf' from
# a directory, e.g., 'conf.d'
#include_if_exists = '' # include file only if it exists
#include = '' # include file

Currently, uncommenting them but leaving the value empty (as happened
in [2] above) can
result in unexpected behaviour.

For "include" and "include_if_exists", it's a not critical issue as
non-existent
files are, well, non-existent, however this will leave behind the
cryptic
message "input in flex scanner failed" in pg_file_settings's "error"
column, e.g.:

postgres=# SELECT sourceline, seqno, name, setting, applied, error
FROM pg_file_settings
WHERE error IS NOT NULL;
sourceline | seqno | name | setting | applied | error
------------+-------+------+---------+---------+------------------------------
1 | 45 | | | f | input in flex scanner failed
1 | 46 | | | f | input in flex scanner failed
(2 rows)

However, an empty value for "include_dir" will result in the current
configuration
file's directory being read, which can result in circular inclusion
and triggering
of the nesting depth check.

Patch {1} makes provision of an empty value for any of these
directives cause
configuration file processing to report an approprate error, e.g.:

postgres=# SELECT sourceline, seqno, name, setting, applied, error
FROM pg_file_settings
WHERE error IS NOT NULL;
sourceline | seqno | name | setting | applied | error
------------+-------+------+---------+---------+---------------------------------------
757 | 45 | | | f | "include" must not be empty
758 | 46 | | | f | "include_if_exists" must not be empty
759 | 47 | | | f | "include_dir" must not be empty

The patch 1 looks somewhat superficial. All the problems are
reduced to creating unexpected filepath for
inclusion. AbsoluteConfigLocation does the core work, and it can
issue generic error message covering all the cases like:

invalid parameter "<param>" at <calling_file>:<calling_lineno>

which seems sufficient. (The function needs some additional
parameters.)

That seems unnecessarily complex to me, as we'd be overloading a
function with a single purpose (to manipulate a path) with some
of the parsing logic/control.

2) Circular inclusion of configuration files
============================================

Currently there is a simple maximum nesting threshold (currently 10)
which
will stop runaway circular inclusions. However, if triggered, it's not
always obvious what triggered it, and sometimes resource exhaustion
might kick in beforehand (as appeared to be the case in [2] above).

Patch {2} attempts to handle this situation by keeping track of which
files have already been included (based on their absolute, canonical
path) and reporting an error if they were encountered again.

This seems to me to be overkill. The issue [2] is prevented by
the patch 1's amendment.

Yes, that particular issue is prevented, but this patch is intended to
provide better protection against explicit circular inclusions, e.g.
if someone adds "include 'postgresql.conf'" at the end of "postgresql.conf"
(or more realistically has a complex setup with multiple included
configuration files where something gets mixed up).

Currently the nesting threshold stops it becoming a runaway
problem, but is not very user-friendly. E.g. with "include 'postgresql.conf'"
added to the end of "postgresql.conf", without patch on startup you get:

LOG: could not open configuration file "postgresql.conf": maximum nesting depth exceeded
FATAL: configuration file "/path/to/postgresql.conf" contains errors

(cue panicking user with production server down: "OMG the file can't be opened,
is my filesystem corrupted?" etc.)

With the patch:

LOG: configuration file "/path/to/postgresql.conf" was previously parsed
FATAL: configuration file "/path/to/postgresql.conf" contains errors

(actually maybe we could add a bit more detail such as line number there).

(I don't think it's not worth donig to
add protection from explicit inclusion of pg_hba.conf from
postgresql.conf or itself or such like.)

I thought about that, but came to the same conclusion.

On server startup:

2019-07-11 09:13:25.610 GMT [71140] LOG: configuration file
"/var/lib/pgsql/data/postgresql.conf" was previously parsed
2019-07-11 09:13:25.610 GMT [71140] FATAL: configuration file
"/var/lib/pgsql/data/postgresql.conf" contains errors

After sending SIGHUP:

postgres=# SELECT sourceline, seqno, name, setting, applied, error
FROM pg_file_settings WHERE error IS NOT NULL;
sourceline | seqno | name | setting | applied | error
------------+-------+------+---------+---------+--------------------------------------------------------------------------------
757 | 45 | | | f | configuration file
"/var/lib/pgsql/data/postgresql.conf" was previously parsed
(1 row)

3) "include" directives in postgresql.auto.conf and extension control
files
===========================================================================

Currently these are parsed and acted on, even though it makes no sense
for further
config files to be included in either case.

Anyway manual edit is explicitly prohibited for auto.conf.

Indeed, but there are many things we tell people not to do, such as
removing tablespace directories, but they still do them...

And even if it is added, the 10-depth limitation would protect from
infinite loop.

It's not just about protecting againt infinite loops - if you do something
like "include 'postgresql.conf'", as-is the code will happily slurp in
all the items from "postgresql.conf" into "postgresql.auto.conf", which
is going to cause some headscratching if it ever happens.

Like I said in the original mail these are extremely unlikely corner cases;
but if patch {2} is in place, it's trivial to prevent them ever becoming a
problem.

With "postgresql.auto.conf", if a file is successfully included, its
contents
will then be written to "postgresql.auto.conf" and the include
directive will be
removed, which seems like a recipe for confusion.

These are admittedly unlikely corner cases, but it's easy enough to
stop this
happening on the offchance someone tries to use this to solve some
problem in
completely the wrong way.

Patch {3} implements this (note that this patch depends on patch {2}).

Extension example:

postgres=# CREATE EXTENSION repmgr;
ERROR: "include" not permitted in file
"/home/barwick/devel/postgres/builds/HEAD/share/extension/repmgr.control"
line 8
postgres=# CREATE EXTENSION repmgr;
ERROR: "include_dir" not permitted in file
"/home/barwick/devel/postgres/builds/HEAD/share/extension/repmgr.control"
line 9
postgres=# CREATE EXTENSION repmgr;
ERROR: "include_if_exists" not permitted in file
"/home/barwick/devel/postgres/builds/HEAD/share/extension/repmgr.control"
line 10

pg.auto.conf example:

postgres=# ALTER SYSTEM SET default_tablespace ='pg_default';
ERROR: could not parse contents of file "postgresql.auto.conf"
postgres=# SELECT regexp_replace(sourcefile, '^/.+/','') AS
sourcefile,
seqno, name, setting, applied, error
FROM pg_file_settings WHERE error IS NOT NULL;
sourcefile | seqno | name | setting | applied | error
----------------------+-------+------+---------+---------+-------------------------
postgresql.auto.conf | 45 | | | f | "include" not permitted
(1 row)

The patch also has the side-effect that "include" directives are no
longer
(silently) removed from "postgresql.auto.conf"; as the only way they
can get
into the file in the first place is by manually editing it, I feel
it's
reasonable for the user to be made aware that they're not valid and
have to
manually remove them.

Patches
=======

Code:

{1} disallow-empty-include-directives.v1.patch
{2} track-included-files.v1.patch
{3} prevent-disallowed-includes.v1.patch

TAP tests:
{1} tap-test-configuration.v1.patch
{2} tap-test-disallow-empty-include-directives.v1.patch
{3} tap-test-track-included-files.v1.patch
{4} tap-test-prevent-disallowed-includes.v1.patch

Patches apply cleanly to REL_12_STABLE/HEAD, they could be modfied for
all supported versions if required. I can consolidate the patches
if preferred.

I don't think this is new to 12.

No, though I'm not sure how much this would be seen as a bugfix
and how far back it would be sensible to patch.

Regards

Ian Barwick

--
Ian Barwick https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services

Ian Barwick <ian.barwick@2ndquadrant.com> writes:

On 7/17/19 5:34 PM, Kyotaro Horiguchi wrote:> Hello.

I don't think this is new to 12.

No, though I'm not sure how much this would be seen as a bugfix
and how far back it would be sensible to patch.

I think this is worth considering as a bugfix; although I'm afraid
we can't change the signature of ParseConfigFile/ParseConfigFp in
released branches, since extensions could possibly be using those.
That limits what we can do --- but it's still possible to detect
direct recursion, which seems like enough to produce a nice error
message in typical cases.

I concur with Kyotaro-san that disallow-empty-include-directives.v1.patch
seems a bit brute-force, but where I would put the checks is in
ParseConfigFile and ParseConfigDirectory.

Also, I don't agree with the goals of prevent-disallowed-includes.patch.
I'm utterly not on board with breaking use of "include" in extension
files, for instance; while that may not be documented, it works fine,
and maybe somebody out there is relying on it. Likewise, while "include"
in pg.auto.conf is not really considered supported, I don't see the
point of going out of our way to break the historical behavior.

That leads me to propose the attached simplified patch. While I haven't
actually tried, I'm pretty sure this should back-patch without trouble.

regards, tom lane

On 8/25/19 4:39 AM, Tom Lane wrote:

Ian Barwick <ian.barwick@2ndquadrant.com> writes:

On 7/17/19 5:34 PM, Kyotaro Horiguchi wrote:> Hello.

I don't think this is new to 12.

No, though I'm not sure how much this would be seen as a bugfix
and how far back it would be sensible to patch.

I think this is worth considering as a bugfix; although I'm afraid
we can't change the signature of ParseConfigFile/ParseConfigFp in
released branches, since extensions could possibly be using those.
That limits what we can do --- but it's still possible to detect
direct recursion, which seems like enough to produce a nice error
message in typical cases.

Makes sense.

I concur with Kyotaro-san that disallow-empty-include-directives.v1.patch
seems a bit brute-force, but where I would put the checks is in
ParseConfigFile and ParseConfigDirectory.

Also, I don't agree with the goals of prevent-disallowed-includes.patch.
I'm utterly not on board with breaking use of "include" in extension
files, for instance; while that may not be documented, it works fine,
and maybe somebody out there is relying on it.

I couldn't for the life of me think of any reason for using it.
But if there's undocumented functionality we think someone might
be using, shouldn't that be documented somewhere, if only as a note
in the code to prevent its accidental removal at a later date?

Likewise, while "include"
in pg.auto.conf is not really considered supported, I don't see the
point of going out of our way to break the historical behavior.

The amusing thing about that of course is that the include directive
will disappear the next time ALTER SYSTEM is run and the values from
the included file will appear in pg.auto.conf, which may cause some
headscratching. But I guess hasn't been an actual real-world
issue so far.

That leads me to propose the attached simplified patch. While I haven't
actually tried, I'm pretty sure this should back-patch without trouble.

Ah, I see it's been applied already, thanks!

Regards

Ian Barwick

--
Ian Barwick https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services