Re: is_absolute_path incorrect on Windows

From: Bruce Momjian <bruce(at)momjian(dot)us>
To: Bruce Momjian <bruce(at)momjian(dot)us>
Cc: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Magnus Hagander <magnus(at)hagander(dot)net>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: is_absolute_path incorrect on Windows
Date: 2011-02-12 14:49:36
Message-ID: 201102121449.p1CEnaZ19618@momjian.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Bruce Momjian wrote:
> Bruce Momjian wrote:
> > Tom Lane wrote:
> > > Bruce Momjian <bruce(at)momjian(dot)us> writes:
> > > > Tom Lane wrote:
> > > >> Bruce Momjian <bruce(at)momjian(dot)us> writes:
> > > >>> I have reviewed is_absolute_path() and have implemented
> > > >>> path_is_relative_and_below_cwd() to cleanly handle cases like 'E:abc' on
> > > >>> Win32; patch attached.
> > > >>
> > > >> This patch appears to remove some security-critical restrictions.
> > > >> Why did you delete the path_contains_parent_reference calls?
> > >
> > > > They are now in path_is_relative_and_below_cwd(),
> > >
> > > ... and thus not invoked in the absolute-path case. This is a security
> > > hole.
> > >
> > > > I don't see a general reason to prevent
> > > > ".." in absolute paths, only relative ones.
> > >
> > > load '/path/to/database/../../../path/to/anywhere'
> >
> > Ah, good point. I was thinking about someone using ".." in the part of
> > the path that is compared to /data or /log, but using it after that
> > would clearly be a security problem.
> >
> > I have attached an updated patch that restructures the code to be
> > clearer and adds the needed checks.
>
> I decided that my convert_and_check_filename() usage was too intertwined
> so I have developed a simplified version that is easier to understand;
> patch attached.

Applied, with a new mention of why we don't use GetFullPathName():

+ /*
+ * On Win32, a drive letter _not_ followed by a slash, e.g. 'E:abc', is
+ * relative to the cwd on that drive, or the drive's root directory
+ * if that drive has no cwd. Because the path itself cannot tell us
+ * which is the case, we have to assume the worst, i.e. that it is not
+ * below the cwd. We could use GetFullPathName() to find the full path
+ * but that could change if the current directory for the drive changes
+ * underneath us, so we just disallow it.
+ */

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Noah Misch 2011-02-12 15:45:09 Re: ALTER TYPE 2: skip already-provable no-work rewrites
Previous Message Robert Haas 2011-02-12 14:32:31 Re: Change pg_last_xlog_receive_location not to move backwards