Messages in this thread | | | Date | Mon, 11 Oct 1999 11:28:56 -0700 | From | "H. Peter Anvin" <> | Subject | Re: devd and device events (was Re: My $0.02 on devd and devfs) |
| |
Jeff Garzik wrote: > > "H. Peter Anvin" wrote: > > David Harris wrote: > > > > Depending on the kind of daemon used, it would either be respawned or > > > > the info would get queued up until it is restarted. I don't think it's > > > > particularly anything to worry about, though... > > > > If you re-spawn the daemon and it gets to see all of the events that have been > > > stored up, it still does not have a proper view of the device configuration. > > > The kernel really has to be smarter than just issuing events and buffering > > > them. > > > It doesn't have to -- the additional information it needs is in the file > > system. > > It doesn't have to, yes, but scanning the fs would be ugly and probably > time consuming. > > IMHO /proc/device_events should be an ASCII list of device events, like > described previously in this thread. Each line in that list, though, > would need a timestamp of some sort. > > devd then maintains its own state, [optionally] syncing up after a > restart using the timestamps to handle as-yet-unprocessed events. This > allows for any amount of time to pass between kernel bootup and devd > doing its work. > > To continue along this train of thought: any suggestions for creating > /proc/device_events? Presumeably stealing code from PCMCIA would be a > good place to start. >
I don't believe that devd *needs* to keep this kind of state, however. I'm actually pretty sure it doesn't. That's why I would start by working on the daemon, and feed it simulated data, before implementing the kernel interface -- then I'd know better what the requirements really are, and what the smallest possible implementation would look like.
-hpa
-- <hpa@transmeta.com> at work, <hpa@zytor.com> in private!
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |