On Mon, 2005-03-21 at 09:47 +0000, Simon Kelley wrote: > I suspect the problem is that _nothing_ happens at suspend/resume. > Dnsmasq just checks the mtime of the file, and if that's more recent > than the last time it checked the file, it re-reads it. Are you > expecting something to happen over a suspend/resume cycle that would > update the file? Does it occur? What happens to the mtime of the file?
The mtime changes; we are talking about a case where the file _is_ updated because I have moved to a different location with a different wireless network. I also tried "touch"ing the file again. > One thing to bear in mind: dnsmasq doesn't spin checking the file all > the time: it only polls it when it has work to do, so touching the file > ins't enough to get it to re-read, you have to touch the file and then > do a DNS query. The query will wake up dnsmasq, and it will then check > resolv.conf and re-read it just before dealing with the query. Understood. It was dnsmasq's failure to resolve hostnames that alerted me to the fact that there was a problem. After suspend-and-resume I have problems with other applications such as squid and pan which access the network. In their case they use 100% CPU until they are restarted. dnsmasq, in contrast, does not use 100% CPU and it responds to signals; but it doesn't notice that resolv.conf has changed. I am beginning to suspect that there is an obscure problem here involving the particular kernel version and the particular firmware I have. -- Thomas Hood <[EMAIL PROTECTED]> -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]