On Mon, Jun 26, 2006 at 04:10:44PM +0800, Zou, Nanhai wrote: > The list is a bit wrong.., I notice that we don't need to dump > unwind segment to core file for stack unwind to work... I am working > on full register dumping and fixing the stack unwind issue. Awsome, if you have anything that needs testing I am more than happy to help. I am a bit confused about which registers are saved and which aren't in the current code, but I figure that mainly relates to the code being incomplete. > The PCI device shutdown code was to un-master all the PCI devices so that no DMA transaction will be issued by Device. However I think maybe we can remove this code because the new kernel memory space is invisible to first kernel. Ok, that makes a lot of sense. I was really puzzling over why ia64 does this but x86_32 doesn't. I also suspect it can be safely removed. Would you like me to test that out and send a patch if it works? > There is another problem that I call irq->end for each devices, it is > not safe to touch any pointer belong to previous kernel at the crash > time. But without this code, MPT fusion driver is very likely unable > to restart. It sometimes failed to restart even with the irq->end > code. This is an open issue need to be fixed. Right, that does sound quite nicely, especially as that device is quite common, right? -- Horms H: http://www.vergenet.net/~horms/ W: http://www.valinux.co.jp/en/ - : send the line "unsubscribe linux-ia64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
|