On Tue, Feb 20, 2007 at 08:10:37PM -0800, Greg Kurtzer wrote:
>
> On Feb 20, 2007, at 6:57 PM, Horms wrote:
>
> >On Tue, Feb 20, 2007 at 04:23:03PM -0800, Greg Kurtzer wrote:
> >>I am seeing an issue where a network device (e1000) won't come up
> >>after a kexec.
> >>
> >>Booted into newly kexec'ed OS I am get the following error when I
> >>`modprobe e1000`:
> >>
> >>e1000: EEPROM initialization failed
> >>e1000: probe of 000:07:07.0 failed with error -5
> >>
> >>My kexec line looks like:
> >>
> >>kexec --initrd=/vnfs.img --force --append="masterip=${MASTERIP}
> >>noapic" /vmlinuz
> >>
> >>note: It is being called from the Perceus (cluster provisioning
> >>solution) initramfs.
> >>
> >>When I use the original kexec-tools and a 2.6.16 kernel I didn't have
> >>this problem. Any ideas would be greatly appreciated!
> >
> >Devices not surviving kexec is a fairly common problem and arguably
> >the most difficult long term problem that kexec faces.
> >
> >I'm a little surprised at an e1000 not coming up. But generally
> >speaking this occurs when the initalisation code assumes
> >(correctly or incorrectly) something about the state of the
> >hardware, and that assumption isn't true after a kexec.
> >This is typically best solved by shutting down the device/bus/whatever
> >correctly before kexec rather than fudging the initialisation code.
> >But its a tricky area.
>
> Good call, you nailed it! OK, I am just gonna run the hardware probe in
> reverse
> (s/modprobe/rmmod/) right before the kexec.
Good plan. Though the code really should handle this without the
need for modprobe/rmmod/...
--
Horms
H: http://www.vergenet.net/~horms/
W: http://www.valinux.co.jp/en/
_______________________________________________
fastboot mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/fastboot
