Ronald G Minnich wrote: > On 19 Jan 2001, Eric W. Biederman wrote: > > > Hmm. I haven't looked in detail but I believe that wake-on-lan, > > is only that just the wake up not reset or power off. And it is really > > touchy going beyond that because then anyone plugged into your switch > > could reset/reboot any of your nodes. > > This is true. But, reset-on-lan and power-off-on-lan are going to happen. > API does it now on the CS20, and another (to-be-named) company is going to > do it too. > > > And with at least SDRAM timing at issue at stake this is something > > that is that needs solving. > > what do you think of using CMOS as a flag for "last time I reset, linux > got all the way up". > > I'm going to put example code for this out in C form in a bit > I wonder if you >could have > automatic reset after say 3 seconds and > > have linuxbios abort the reset once it is up and running, if it hangs > > it will reset and read the flag in nvram and start with stable > > settings > > this is what I'm thinking but we've found that 3 seconds is too short. > > chipset hardware is there to do this ... but the time should be more like > 30 seconds to make sure linux is up to runlevel 3. > We've been using ethernet controllers with wake-on-lan that also has SMBus for access to system management events. Using this you can check the status of the node over the network and if it's not right reset it with a packet. You could also add a hardware watchdog with an SMBus interface to this to automatically reset the node if it hangs and also be controlled (on/off, reset window, etc.) over the network. Bari
