Thread: BUG #5810: non-accessible hardlink creation was attempted by: postgres (fsuid 115)

The following bug has been logged online:

Bug reference:      5810
Logged by:          Vincent
Email address:      vmaury79@yahoo.fr
PostgreSQL version: 8.4
Operating system:   Ubuntu 10.10
Description:        non-accessible hardlink creation was attempted by:
postgres (fsuid 115)
Details:

Hi,

I'm using postgresql 8.4 on a Ubuntu 10.10 live USB. When I call 'dmesg', I
frequently see this line:

non-accessible hardlink creation was attempted by: postgres (fsuid 115)

Do you know what it means? How can I fix this?
Thank you very much for your help,

Vincent
On Sunday 02 January 2011 15:16:28 Vincent wrote:
> The following bug has been logged online:
>
> Bug reference:      5810
> Logged by:          Vincent
> Email address:      vmaury79@yahoo.fr
> PostgreSQL version: 8.4
> Operating system:   Ubuntu 10.10
> Description:        non-accessible hardlink creation was attempted by:
> postgres (fsuid 115)
> Details:
>
> Hi,
>
> I'm using postgresql 8.4 on a Ubuntu 10.10 live USB. When I call 'dmesg', I
> frequently see this line:
>
> non-accessible hardlink creation was attempted by: postgres (fsuid 115)
>
> Do you know what it means? How can I fix this?
> Thank you very much for your help,
Afaik this is caused by a strange security module of ubuntu. You can disable
the check by running something like
sysctl -w kernel.yama.protected_nonaccess_hardlinks=0
.

I would be interested to see where pg does this - did you notice any problems
on the postgres side? In /var/log/postgresql/8.4/main/postgresql.log or such?

Andres
"Vincent" <vmaury79@yahoo.fr> writes:
> I'm using postgresql 8.4 on a Ubuntu 10.10 live USB. When I call 'dmesg', I
> frequently see this line:
> non-accessible hardlink creation was attempted by: postgres (fsuid 115)

There aren't very many link() calls in Postgres ... am I right to guess
that you're getting matching occurrences of this message in the
postmaster's log?

                 errmsg("could not link file \"%s\" to \"%s\" (initialization of log file %u, segment %u): %m",

A bit of googling suggests it's probably the same Ubuntu bug described
here:

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/663069

According to that, you should be able to work around it by disabling
this rather-dubious-in-the-first-place security check.

            regards, tom lane