Hi Brian, Interesting finding with the stack size limit. Maybe you could summarize your experiences and solutions with the exception handlers in the wiki. With regards to the interrupts our work with the MMU emulation and shadow page tables also indicates that hooking all interrupts might be necessary. In general we might have to re-think some of the assembly hooks to make it work. We had a long talk with Oren on Friday and have started to construct a basic shadow page table (using mostly built in kernel functions) based on his advice. Maybe we can have a talk about asynchronous interrupts and page tables in the meeting tomorrow where we have some extra expertise in the area. I know Christoffer will be away tonight but if you have something you would like to discuss before the meeting you can just give me a call. /Andreas -----Original Message----- From: android-virt-bounces at lists.cs.columbia.edu [mailto:android-virt-bounces at lists.cs.columbia.edu] On Behalf Of Brian Smith Sent: den 19 april 2009 00:34 To: android-virt at lists.cs.columbia.edu Subject: status/comments Hey guys, I wanted to let you know I have made significant changes to arm_interrupts.S to handle more than the software interrupts. The big problem I ran into that caused a little bit of reorganizing was all modes besides SVC and USER have very limited stack space (12 bytes for each mode). Anyway, if you have hits to this part please let me know and I will see how it might fit in. I was also thinking quite a bit on whether we should also intercept asynchronous interrupts (IRQ/FIQ), and I think that we should. The reason being if we didn't we would end up loosing control over the guest. For example if the guest was in some sort of loop with no page faults occurring we (the host user space program) might not get control for a long time. The host kernel would of course get control, but would handle the interrupts and redispatch directly to the guest, and as a result we would never be able to "asynchronously" issue interrupts to the guest. Does that make sense? Or am I overthinking this? Brian _______________________________________________ Android-virt mailing list Android-virt at lists.cs.columbia.edu https://lists.cs.columbia.edu/cucslists/listinfo/android-virt
