Re: [AMaViS-user] amavis-milter permissions [patch]

[Adam C. M. <ad...@mi...>]

Mark Martinec wrote:

>Adam,
>
>| FWIW I'd argue that running everything as one user comes with it's own
>| set of security problems.  For example I like that `clamav' is in group
>| `amavis' because it produces a one-way trust relationship from the
>| former to the latter while it doesn't allow amavis the ability to write
>| to the ClamAV virus patterns database.
>
>I agree. Patch accepted, thanks.
>
>|	!(StatBuf.st_mode & (S_IRWXU|S_IRGRP|S_IXGRP)))
>shouldn't this be:
>	(StatBuf.st_mode & ~(S_IRWXU|S_IRGRP|S_IXGRP)) )
>  
>
Yes, it should.  My apologies.  To much %s// without watching where I 
was going...

>| -    if ((priv->mlfi_fp = fopen(priv->mlfi_fname, "w+")) == NULL) {
>| +    if ((priv->mlfi_fp = fopen(priv->mlfi_fname, "w+")) == NULL
>| +       || fchmod(fileno(priv->mlfi_fp), S_IRUSR|S_IWUSR|S_IRGRP) == -1) {
>
>The umask should have take cane of this, or perhaps:
>
>    if ((priv->mlfi_fp = fopen(priv->mlfi_fname, "w+",
>				S_IRUSR|S_IWUSR|S_IRGRP)) == NULL) {
>  
>

Hmmm, not sure what you mean here.  fopen(3) doesn't take a mode_t 
argument so this won't compile as is.  open(2) takes mode_t but would 
leave you with an integer file-descriptor which you could feed to 
fdopen(3) which if nested could look almost like what you have above.  ;-)

With respect to the umask, i set the mode with fchmod(2) because 
amavis-milter generally seems to like to control permissions explicitly 
throughout rather than relying on umask so I stuck with that.

>Mark
>
>  
>


-- 
Adam C. Migus - http://people.migus.org/~adam/