[Dnsmasq-discuss] Extension to hosts-dir and dhcohosts-dir

[Andy Hawkins]

Hi,

In article <slrnp80aau.nkj.andy at xcp-mailnews.gently.org.uk>,
           Andy Hawkins<andy at gently.org.uk> wrote:
> In inotify.c, around line 236 is the following code block:
>
> 	  /* ignore emacs backups and dotfiles */
> 	  if (in->len == 0 || 
> 	      in->name[in->len - 1] == '~' ||
> 	      (in->name[0] == '#' && in->name[in->len - 1] == '#') ||
> 	      in->name[0] == '.')
> 	    continue;
> 	  
> However, if I create a file called 'fred~' in the directory I've specified
> using dhcp-hostsdir I still get an event in syslog that shows this file is
> being processed:
>
> Feb 11 11:14:34 xcp-gateway dnsmasq[1039]: inotify, new or changed file 
> 	/etc/dnsmasq/dhcp-hosts.d/fred~

Ok, I've done some debugging. I added the following lines:

          my_syslog(LOG_INFO, "ADH: len: %d", in->len);
          my_syslog(LOG_INFO, "ADH: name: %s", in->name);
          my_syslog(LOG_INFO, "ADH: last char: %c", in->name[in->len - 1]);

And I get the following output:

dnsmasq: ADH: len: 16
dnsmasq: ADH: name: fred4~
dnsmasq: ADH: last char:

So, it appears that the length in in->len is being interpreted correctly.

According to the inotify man page:

       The len field counts all of the bytes in name, including the null
       bytes; the length of each inotify_event structure is thus
       sizeof(struct inotify_event)+len.

So in fact, 'len' seems to be a fixed length, irrespective of the length of
the file name in the 'name' field.

It looks like the check should actually be something like:

          /* ignore emacs backups and dotfiles */
          if (in->len == 0 ||
              in->name[strlen(in->name) - 1] == '~' ||
              (in->name[0] == '#' && in->name[strlen(in->name) - 1] == '#') ||
              in->name[0] == '.')

I guess you may need to check that there's a null in the name somewhere
before using strlen, otherwise you might end up running off the end of the
string. I don't know inotify well enough to know if there's guaranteed to be
a null in there somewhere. The manpage does say that the name field is null
terminated, but I don't know if that's guaranteed or not.

I could have a look at submitting a patch, but my editor is showing some
very strange indentation of the source, so I suspect I have my tab settings
incorrect. What is the standard setting for tabs on the dnasmasq source
files?

Thanks

Andy