[syslog-ng]umask settings?

[Balazs Scheidler]

> > > now..when i start syslog-ng with umask 077, i get files generated as 0600
> > > when i start with umask 022, i get files generated with as 0644.  i though
> > > that perm and dir_perm were set while ignoring umask, this should be
> > > noted in the documentation if it uses umask as well.
> > 
> > umask is enforced by the kernel. syslog-ng doesn't use it, it simply issues
> > chmod(file, perm), kernel masks bits found in umask.
> 
> On which UNIX and where can I get a copy? ;)

To be honest I never checked this directly (ie kernel source), but here's
the relevant part of the Linux kernel source:

struct dentry * open_namei(const char * pathname, int flag, int mode)
{
        int acc_mode, error;
        struct inode *inode;
        struct dentry *dentry;

        mode &= S_IALLUGO & ~current->fs->umask;
        mode |= S_IFREG;

or the man page for open:

OPEN(2)                    System calls                   OPEN(2)

       mode specifies the permissions to use if  a  new  file  is
       created.  It  is  modified  by  the process's umask in the
       usual way: the permissions of the created file are (mode &
       ~umask).


> 
> I have only seen this type of mode enforcement on B1 systems. chmod()
> calls (except symbolic ones without "ugo" specification) will actually
> ignore umask settings. open() in turn, will not, and I guess that's
> why we are seeing what we see :)

you are right, syslog-ng uses simply open with mode set to perm | O_WRONLY.

> I'd have to look at the source, but I suspect something with the chmod
> being wrong. As far as I'm concerned, perm's that were specified
> explicitely should be in place regardless of the umask. I start all my
> daemons with 077, just in case they dump temporary stuff (e.g. cores)
> to disk that I don't want others to read, overwrite, whatever. I still
> might want to create output files with permissions appropriate to
> archive files or make them available for other legitimate users.

I may add a chmod() call to enforce perm bits. What do others think?

-- 
Bazsi
PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1
     url: http://www.balabit.hu/pgpkey.txt