Hi Russ, I'm trying to fix http://bugs.debian.org/338348 but I'm sort of clueless.
The problem we have (and still it has to be decided if udhcp is responsible for this bug) is that when a user ifup an interface, udhcpc is called which in turn calls resolvconf to keep track of the assigned ip address (resolvconf is a native debian package that manages dynamically assigned DNS servers) which in turn creates a /etc/resolvconf/run/interface/eth0.udhcpc following this scheme: ifup -> udhcpc -> resolvconf -> eth0.udhcpc Now when you ifdown the interface (in this case eth0), ifdown simply sends a TERM signal to udhcpc which dies silently. The problem is that while dying it doesn't call resolvconf to clean the DNS dynamic table (ie, deleting /etc/resolvconf/run/interface/eth0.udhcpc) Now I thought about having ifdown sending a SIGUSR2 to udhcpc instead of TERM. This effectively calls /etc/udhcpc/default.deconfig which callls resolvconf which in turn delete /etc/resolvconf/run/interface/eth0.udhcpc The problem in this case is that we do dismiss udhcpc, which is still in the process table. Now if we had to ifup the interface again, we'd have 2 udhcpc running on the system, then 3, 4, ... So we have 2 options: using TERM we get rid of udhcpc.pid but not of /etc/resolvconf/run/interface/eth0.udhcpc using SIGUSR2 we get rid of /etc/resolvconf/run/interface/eth0.udhcpc but not of udhcpc.pid Is there a third way ? Do you have any suggestion ? Best regards, -- "Un disco para gobernarlos a todos. Un disco para encontrarlos y atarlos en las tinieblas. En la tierra de Redmond donde se extienden las sombras" Eric VAN BUGGENHAUT [EMAIL PROTECTED] -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
