Rusty Russell wrote:> On Friday 21 March 2008 19:15:47 Christian Borntraeger wrote:>> Am Freitag, 21. März 2008 schrieb Rusty Russell:>>> Hmm, panic on device_register fail, but -ENOMEM on add_shared_memory>>> fail? My theory was that since this is boot time, panic() is the right>>> thing.>> Good spot, but I agree with Carsten. Drivers should not panic. I have>> module load/unload capability on my long term todo list, but I can change>> the panic now.> > Yep, that makes sense. For lguest, we panic: it's always at boot time so if > it fails we should die early to make it easier to diagnose (and that makes > sure it happens before we lose our early console).Diangnostic is easy here at any time during the boot process: we've got our store status ioctl that userspace calls after guest execution has ended. It causes all cpus to store their register content and such into the cpu's lowcore area. Then it writes out our memory to a dump image, which lkcdutils and/or crash can read. _______________________________________________Virtualization mailing listVirtualization@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx://lists.linux-foundation.org/mailman/listinfo/virtualization
