BUG #5267: initdb fails on AIX: could not identify current directory

Michael Felt wrote:

The following bug has been logged online:

Bug reference: 5267
Logged by: Michael Felt
Email address: mamfelt@gmail.com
PostgreSQL version: 8.4.2
Operating system: AIX
Description: initdb fails on AIX: could not identify current
directory
Details:

I have compiled postgresql on AIX 6.1 without a great deal of difficulity
(./configure -without-readline -without-zlib).

However, the fist step - initdb - fails. I have not been able to determine
which directory it is complaining about.

Suggestions welcome.
========
michael@x054:[/data/home/michael]su - postgres
postgres@x054:[/home/postgres]/usr/local/pgsql/bin/initdb -D
/usr/local/pgsql/data
could not identify current directory: The file access permissions do not
allow the specified action.
could not identify current directory: The file access permissions do not
allow the specified action.
could not identify current directory: The file access permissions do not
allow the specified action.
The program "postgres" is needed by initdb but was not found in the
same directory as "initdb".
Check your installation.
postgres@x054:[/home/postgres]ls -l /usr/local/pgsql/bin
total 18744
-rwxr-xr-x 1 root system 79898 Jan 05 13:33 clusterdb
-rwxr-xr-x 1 root system 79554 Jan 05 13:33 createdb
-rwxr-xr-x 1 root system 79732 Jan 05 13:33 createlang
-rwxr-xr-x 1 root system 82515 Jan 05 13:33 createuser
-rwxr-xr-x 1 root system 76661 Jan 05 13:33 dropdb
-rwxr-xr-x 1 root system 82087 Jan 05 13:33 droplang
-rwxr-xr-x 1 root system 76613 Jan 05 13:33 dropuser
-rwxr-xr-x 1 root system 686273 Jan 05 13:33 ecpg
-rwxr-xr-x 1 root system 101041 Jan 05 13:33 initdb
-rwxr-xr-x 1 root system 37341 Jan 05 13:33 pg_config
-rwxr-xr-x 1 root system 31437 Jan 05 13:33 pg_controldata
-rwxr-xr-x 1 root system 52157 Jan 05 13:33 pg_ctl
-rwxr-xr-x 1 root system 392643 Jan 05 13:33 pg_dump
-rwxr-xr-x 1 root system 103892 Jan 05 13:33 pg_dumpall
-rwxr-xr-x 1 root system 46554 Jan 05 13:33 pg_resetxlog
-rwxr-xr-x 1 root system 186965 Jan 05 13:33 pg_restore
-rwxr-xr-x 1 root system 6842028 Jan 05 13:33 postgres
lrwxrwxrwx 1 root system 8 Jan 05 13:44 postmaster ->
postgres
-rwxr-xr-x 1 root system 388374 Jan 05 13:33 psql
-rwxr-xr-x 1 root system 82335 Jan 05 13:33 reindexdb
-rwxr-xr-x 1 root system 47359 Jan 05 13:33 vacuumdb
postgres@x054:[/home/postgres]ls -l .
total 0
postgres@x054:[/home/postgres]ls -la .
total 24
drwxr-xr-x 2 postgres staff 256 Jan 05 13:45 .
drwxr-xr-x 31 bin bin 4096 Jan 05 13:44 ..
-rwx------ 1 postgres staff 254 Jan 05 13:44 .profile
-rw------- 1 postgres staff 884 Jan 05 18:38 .sh_history
postgres@x054:[/home/postgres]mkgroup postgres
ksh: mkgroup: 0403-006 Execute permission denied.
postgres@x054:[/home/postgres]
michael@x054:[/data/home/michael]mkgroup postgres
michael@x054:[/data/home/michael]chuser pgrp=postgres postgres
michael@x054:[/data/home/michael]chgrp postgres /usr/local/pgsql/data
michael@x054:[/data/home/michael]su - postgres
postgres@x054:[/home/postgres]id
uid=2010(postgres) gid=2013(postgres) groups=1(staff)
postgres@x054:[/home/postgres]/usr/local/pgsql/bin/initdb -D
/usr/local/pgsql/data
could not identify current directory: The file access permissions do not
allow the specified action.
could not identify current directory: The file access permissions do not
allow the specified action.
could not identify current directory: The file access permissions do not
allow the specified action.
The program "postgres" is needed by initdb but was not found in the
same directory as "initdb".
Check your installation.
postgres@x054:[/home/postgres]ls -ld /usr
drwxr-xr-x 43 root system 4096 Jan 05 13:40 /usr
postgres@x054:[/home/postgres]ls -ld /usr/local
drwxr-xr-x 19 root system 4096 Jan 05 13:32 /usr/local
postgres@x054:[/home/postgres]ls -ld /usr/local/pgsql
drwxr-xr-x 7 root system 256 Jan 05 13:45 /usr/local/pgsql
postgres@x054:[/home/postgres]ls -ld /usr/local/pgsql/data
drwxr-xr-x 2 postgres postgres 256 Jan 05 13:45
/usr/local/pgsql/data
postgres@x054:[/home/postgres]ls -ld /usr/local/pgsql/bin
drwxr-xr-x 2 root system 4096 Jan 05 13:33
/usr/local/pgsql/bin
postgres@x054:[/home/postgres]/usr/local/pgsql/bin/initdb -D
/usr/local/pgsql/data
postgres@x054:[/home/postgres]

Just thought I would add - I sent this in 24 hours ago using this link
in the FAQ - and obviously, for me, it never arrived. The text may need
updating:
===

Where to report bugs

In general, send bug reports to the bug report mailing list at
|<pgsql-bugs@postgresql.org <mailto:pgsql-bugs@postgresql.org>>|. You
are requested to use a descriptive subject for your email message,
perhaps parts of the error message.
===

Michael Felt escribi�:

The following bug has been logged online:

Bug reference: 5267
Logged by: Michael Felt
Email address: mamfelt@gmail.com
PostgreSQL version: 8.4.2
Operating system: AIX
Description: initdb fails on AIX: could not identify current
directory
Details:

I have compiled postgresql on AIX 6.1 without a great deal of difficulity
(./configure -without-readline -without-zlib).

However, the fist step - initdb - fails. I have not been able to determine
which directory it is complaining about.

Suggestions welcome.

What user was /usr/local/pgsql/data owned by originally?

--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

"Michael Felt" <mamfelt@gmail.com> writes:

I have compiled postgresql on AIX 6.1 without a great deal of difficulity
(./configure -without-readline -without-zlib).

However, the fist step - initdb - fails. I have not been able to determine
which directory it is complaining about.

could not identify current directory: The file access permissions do not
allow the specified action.

This message means that getcwd() failed, apparently with an EACCESS
code. Typically what that means is that some parent directory of
the current directory is not readable or searchable by the user you
ran initdb as.

I believe that initdb also invokes getcwd for the directory that
its own executable is in, and possibly for the target data directory,
so you should check the paths leading to those as well.

regards, tom lane

michael@x054:[/data/home/
michael]ls -ld /
drwxr-xr-x 27 root system 4096 Jan 04 17:20 /
michael@x054:[/data/home/michael]ls -ld /usr

drwxr-xr-x 43 root system 4096 Jan 05 13:40 /usr
michael@x054:[/data/home/michael]ls -ld /usr/local

drwxr-xr-x 19 root system 4096 Jan 05 13:32 /usr/local
michael@x054:[/data/home/michael]ls -ld /usr/local/pgsql

drwxr-xr-x 7 root system 256 Jan 05 13:45 /usr/local/pgsql
michael@x054:[/data/home/michael]ls -ld /usr/local/pgsql/data

drwxr-xr-x 2 postgres postgres 256 Jan 05 13:45
/usr/local/pgsql/data
michael@x054:[/data/home/michael]ls -ld /usr/local/pgsql/bin
drwxr-xr-x 2 root system 4096 Jan 05 13:33
/usr/local/pgsql/bin

To answer the other question: root:system was the original owner. Changed to
pgsql before calling application. Because of the error I made the group,
changed user postgres to be postgres:postgres by default and made the
directory postgres:postgres.

I suppose I could turn on audit and see if it is trying to access a hard
coded directory. But, in any case, I would update the error message to at
least mention the directory name it is having issues with.

On Wed, Jan 6, 2010 at 5:03 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Michael Felt <mamfelt@gmail.com> writes:

I suppose I could turn on audit and see if it is trying to access a hard
coded directory. But, in any case, I would update the error message to at
least mention the directory name it is having issues with.

Well, the problem is what to print? The failure we are trying to report
is exactly that we *can't get* the name of the directory.

regards, tom lane

Well, there is an argument that a system call is using to get somewhere?
Even if it is a number, it is something. I could do an ncheck or whatever to
at least find what it is calling.

As I am not at all familiar with the code - just give me source to debug,
and I'll work from that.

On Thu, Jan 7, 2010 at 3:35 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Michael Felt <mamfelt@gmail.com> writes:

On Thu, Jan 7, 2010 at 3:35 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Well, the problem is what to print? The failure we are trying to report
is exactly that we *can't get* the name of the directory.

Well, there is an argument that a system call is using to get somewhere?

getcwd() has no input parameters.

regards, tom lane

michael@x054:[/data/prj/postgresql-8.4.2/src]grep cwd */*.c
Well, unless you redefine it...
port/exec.c:#define getcwd(cwd,len) GetCurrentDirectory(len, cwd)
port/exec.c: char cwd[MAXPGPATH],
port/exec.c: if (!getcwd(cwd, MAXPGPATH))
port/exec.c: join_path_components(retpath, cwd, argv0);
port/exec.c: join_path_components(retpath, cwd, argv0);
port/exec.c: join_path_components(retpath, cwd,
test_path);
port/exec.c: * getcwd() to figure out where the heck we're at.
port/exec.c: * getcwd() to give us an accurate, symlink-free path.
port/exec.c: if (!getcwd(orig_wd, MAXPGPATH))
port/exec.c: if (!getcwd(path, MAXPGPATH))

Now I have no idea what is being called. I hope you do.

On Thu, Jan 7, 2010 at 3:57 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

On Thu, Jan 7, 2010 at 10:00 AM, Michael Felt <mamfelt@gmail.com> wrote:

michael@x054:[/data/prj/postgresql-8.4.2/src]grep cwd */*.c
Well, unless you redefine it...
port/exec.c:#define getcwd(cwd,len) GetCurrentDirectory(len, cwd)

If you look at the context of this #define you'll see that it only
applies to Windows.

port/exec.c:    char            cwd[MAXPGPATH],
port/exec.c:    if (!getcwd(cwd, MAXPGPATH))
port/exec.c:                    join_path_components(retpath, cwd, argv0);
port/exec.c:    join_path_components(retpath, cwd, argv0);
port/exec.c:                            join_path_components(retpath, cwd,
test_path);
port/exec.c:     * getcwd() to figure out where the heck we're at.
port/exec.c:     * getcwd() to give us an accurate, symlink-free path.
port/exec.c:    if (!getcwd(orig_wd, MAXPGPATH))
port/exec.c:    if (!getcwd(path, MAXPGPATH))

Now I have no idea what is being called. I hope you do.

It looks to me like bin/initdb/initdb.c:main() is calling
port/exec.c:find_other_exec() which is calling
port/exec.c:find_my_exec() which is calling getcwd(). So it's
probably the directory you were in when you ran initdb that is the
problem. For example:

cd $HOME
mkdir delete-me
cd delete-me
rmdir $HOME/delete-me
initdb

Produces:

could not identify current directory: No such file or directory
could not identify current directory: No such file or directory
could not identify current directory: No such file or directory
The program "postgres" is needed by initdb but was not found in the
same directory as "initdb".
Check your installation.

This is very similar to what you got except that for you it's
complaining about permissions rather than existence. I would try
running initdb from someplace like / or /tmp and see if that works.

I have to say that the error message that is produced by the above
test case could easily send one looking in the wrong direction, and
could perhaps stand to be improved. Could we just do getcwd() once,
bail out if it fails, and then stash the results, rather than
continuing on and eventually producing a misleading error message?

...Robert

I wrote a simple program - just calling getcwd() and I added some extra text
to exec.c - to know where it is at. You should recognize it.

michael@x054:[/data/home/michael]ls -l /usr/local/pgsql/bin/mytest
-rwx------ 1 root system 4793 Jan 07 15:39
/usr/local/pgsql/bin/mytest
michael@x054:[/data/home/michael]chmod a+rx /usr/local/pgsql/bin/mytest
michael@x054:[/data/home/michael]r su
su - postgres
postgres@x054:[/home/postgres]/usr/local/pgsql/bin/mytest
/usr/local/pgsql/bin/mytest: 1
PATH_MAX:1023
/home/postgres
postgres@x054:[/home/postgres]
michael@x054:[/data/home/michael]
michael@x054:[/data/home/michael]cat test.c
#include <limits.h>
#include <unistd.h>
main(int argc, char **argv)
{
char buf[PATH_MAX];

printf("%s: %d\n", argv[0], argc);
printf("PATH_MAX:%d\n", PATH_MAX);
printf("%s\n", getcwd(buf, PATH_MAX));
}

===========
And I am only running the command from /home/postgres

postgres@x054:[/home/postgres]ls -ld .
drwxr-xr-x 2 postgres staff 256 Jan 05 13:45 .

postgres@x054:[/home/postgres]ls -l
/usr/local/pgsql/bin
total 18760
-rwxr-xr-x 1 root system 79898 Jan 05 13:33 clusterdb
-rwxr-xr-x 1 root system 79554 Jan 05 13:33 createdb
-rwxr-xr-x 1 root system 79732 Jan 05 13:33 createlang
-rwxr-xr-x 1 root system 82515 Jan 05 13:33 createuser
-rwxr-xr-x 1 root system 76661 Jan 05 13:33 dropdb
-rwxr-xr-x 1 root system 82087 Jan 05 13:33 droplang
-rwxr-xr-x 1 root system 76613 Jan 05 13:33 dropuser
-rwxr-xr-x 1 root system 686273 Jan 05 13:33 ecpg
-rwxr-xr-x 1 root system 101639 Jan 07 15:34 initdb
-rwxr-xr-x 1 root system 4793 Jan 07 15:39 mytest
-rwxr-xr-x 1 root system 37341 Jan 05 13:33 pg_config
-rwxr-xr-x 1 root system 31437 Jan 05 13:33 pg_controldata
-rwxr-xr-x 1 root system 52157 Jan 05 13:33 pg_ctl
-rwxr-xr-x 1 root system 392643 Jan 05 13:33 pg_dump
-rwxr-xr-x 1 root system 103892 Jan 05 13:33 pg_dumpall
-rwxr-xr-x 1 root system 46554 Jan 05 13:33 pg_resetxlog
-rwxr-xr-x 1 root system 186965 Jan 05 13:33 pg_restore
-rwxr-xr-x 1 root system 6842028 Jan 05 13:33 postgres
lrwxrwxrwx 1 root system 8 Jan 05 13:44 postmaster ->
postgres
-rwxr-xr-x 1 root system 388374 Jan 05 13:33 psql
-rwxr-xr-x 1 root system 82335 Jan 05 13:33 reindexdb
-rwxr-xr-x 1 root system 47359 Jan 05 13:33 vacuumdb

postgres@x054:[/home/postgres]/usr/local/pgsql/bin/initdb -D
/usr/local/pgsql/data
resolve_symlinks() orig_wd: MAXPGPATH == 1024
could not identify path directory: The file access permissions do not allow
the specified action.
find_my_exec(): MAXPGPATH == 1024
could not identify current directory: The file access permissions do not
allow the specified action.
find_my_exec(): MAXPGPATH == 1024
could not identify current directory: The file access permissions do not
allow the specified action.
The program "postgres" is needed by initdb but was not found in the
same directory as "initdb".
Check your installation.
postgres@x054:[/home/postgres]

=========

From /tmp gives the same result.

postgres@x054:[/home/postgres]cd /tmp
postgres@x054:[/tmp]/usr/local/pgsql/bin/initdb -D /usr/local/pgsql/data
resolve_symlinks() orig_wd: MAXPGPATH == 1024
could not identify path directory: The file access permissions do not allow
the specified action.
find_my_exec(): MAXPGPATH == 1024
could not identify current directory: The file access permissions do not
allow the specified action.
find_my_exec(): MAXPGPATH == 1024
could not identify current directory: The file access permissions do not
allow the specified action.
The program "postgres" is needed by initdb but was not found in the
same directory as "initdb".
Check your installation.
postgres@x054:[/tmp]

=========

I ran my test program with larger and smaller MAXPGPATH constants. 2046
(1023 * 2) was the largest I tested and it worked fine. When it was shorted
the call failed. I did not test the error message.

On Thu, Jan 7, 2010 at 4:25 PM, Robert Haas <robertmhaas@gmail.com> wrote:

I turned on audit - it continues to say michael as user for accountability.
Notice: su changes to /home/postgres and initdb changes to
/usr/local/pgsql/bin

FS_Chdir michael OK Thu Jan 07 16:06:35 2010
su Global
change current directory to: /home/postgres
MLS Data:
Not supported

FS_Chdir michael OK Thu Jan 07 16:06:38 2010
initdb Global
change current directory to: /usr/local/pgsql/bin
MLS Data:
Not supported

On Thu, Jan 7, 2010 at 3:41 PM, Michael Felt <mamfelt@gmail.com> wrote:

Robert Haas <robertmhaas@gmail.com> writes:

I have to say that the error message that is produced by the above
test case could easily send one looking in the wrong direction, and
could perhaps stand to be improved. Could we just do getcwd() once,
bail out if it fails, and then stash the results, rather than
continuing on and eventually producing a misleading error message?

How does that help? We still can't print the directory name.

regards, tom lane

Michael Felt <mamfelt@gmail.com> writes:

I ran my test program with larger and smaller MAXPGPATH constants. 2046
(1023 * 2) was the largest I tested and it worked fine. When it was shorted
the call failed. I did not test the error message.

[ scratches head... ] This seems to be misbehavior of getcwd() itself,
since the paths involved are certainly not 2K long.
Perhaps you need to file a bug with IBM.

regards, tom lane

On Thu, Jan 7, 2010 at 11:39 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Robert Haas <robertmhaas@gmail.com> writes:

I have to say that the error message that is produced by the above
test case could easily send one looking in the wrong direction, and
could perhaps stand to be improved.  Could we just do getcwd() once,
bail out if it fails, and then stash the results, rather than
continuing on and eventually producing a misleading error message?

How does that help?  We still can't print the directory name.

Well, as it is, it looks like the failure of getcwd() might be an
incidental problem, and the inability to find postgres was what sunk
the ship. In fact, the inability to find postgres is an entirely
illusory problem created by the failure of getcwd(). If you just got
one error message saying "getcwd failed", I think it would be more
clear what the problem was. I had to go read the code to figure out
that the failure of getcwd() would result in a guaranteed failure to
find the postgres executable.

...Robert

Robert Haas <robertmhaas@gmail.com> writes:

On Thu, Jan 7, 2010 at 11:39 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

How does that help? �We still can't print the directory name.

Well, as it is, it looks like the failure of getcwd() might be an
incidental problem, and the inability to find postgres was what sunk
the ship. In fact, the inability to find postgres is an entirely
illusory problem created by the failure of getcwd(). If you just got
one error message saying "getcwd failed", I think it would be more
clear what the problem was. I had to go read the code to figure out
that the failure of getcwd() would result in a guaranteed failure to
find the postgres executable.

Should we just turn find_my_exec() into a routine that elogs/exits on
failure, instead of returning an error code? There are a couple of call
sites that have the idea that they can survive a failure, but I think
they're pretty bogus.

There are actually two distinct cases that we need to worry about (and
I'm not entirely certain that I know which one Michael is hitting).
Case 1 is where getcwd() fails on the program's starting current
directory. Case 2 is where it fails after we do a series of chdir's
following symlinks. In case 1 there really is no additional information
available, whereas in case 2 we could perhaps print the name of the
first or last symlink we tried to follow. Also, while I think it might
be fair to treat case 1 as a hard error, it's a bit more plausible that
a caller might have a recovery strategy for case 2. So maybe treating
these two cases differently would be a good thing.

regards, tom lane

Update: I have reinstalled my server and now it works fine. (I did not
recomply the initdb program).
As I did this to enable a security feature, not related to this (called
Trusted Executition) I can only guess what might have been to problem. As I
dont have the old system - and I am not going to reinstall it to "test" - we
are at a dead end here.

I am glad that my build procedure worked properly though!

I am looking for other AIX users to test the build as I am building up a
repository of pre-built AIX open source packages - especially those not
found elsewhere.

Where can I best make this known within postgres?

Thanks for your assistance,
Michael

On Mon, Jan 11, 2010 at 8:37 AM, Michael Felt <mamfelt@gmail.com> wrote: