[PATCH 0/3] Make kexec work with ibmvscsi/ibmveth

Haren Myneni haren at us.ibm.com
Fri Aug 5 10:44:33 EST 2005


>
> Michael Ellerman writes:
>
> > On our Power5 we have vscsi and veth installed. When trying to kexec 
> a new
> > kernel the vscsi and veth initialisation fails because the previous 
> kernel
> > didn't clean up properly and as far as the Hypervisor is concerned 
> is still
> > connected.
> >
> > The fix is, perhaps, to add a shutdown() function to the vio_driver 
> struct so
> > that vscsi and veth have a chance to disconnect from the Hypervisor 
> at reboot.
>
> But if we are using kexec to do kdump, and the kernel panics, the
> driver's shutdown method won't get called.  We should instead harden
> the driver so that it can cope with the error from the hypervisor and
> recover, I think.
>
> Paul.
>

BTW, When I tested kexec boot without calling device_shutdown(), had the 
device initialization issue for IPR. However, the system was booted with 
some EEH traces after I applied Linas's EEH Error recovery patches 
(reset the PCI slot).  I need to do some more testing.

Thanks
Haren

But, in the normal kexec boot,

> _______________________________________________
> Linuxppc64-dev mailing list
> Linuxppc64-dev at ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc64-dev
>




More information about the Linuxppc64-dev mailing list