Fernando Luis V?zquez Cao <fernando at oss.ntt.co.jp> writes: >> Is just a jump and not a move followed by a jump still 10 bytes? >> I hate to say it but I think this fails miserably for any exception >> after a nmi. > > Thank you for the heads up! Actually, it was working for the > exceptions after the nmi but with a corrupted esi (vector > number). My original intention was to fill the empty space > with nops but forgot to actually implement it... Sorry about > that. Will fix for the next iteration. Sound good, and thank you very much for tackling this. >> I expect the simplest solution is to modify early_idt_handler to test >> for vector == 2. > > That is precisely what I did on a previous version but that would > involve using registers which need to be saved and restored and > I wanted to avoid using the stack in the NMI path. We would also > need to add a "pushq rsi " in early_idt_handlers which implies > modifying "early_idt_handlers" definition in "segment.h". > > If you are OK with it I would like to go with the approach in > the two patches I sent. I am fine with your approach. I suggest a big fat comment mentioning the 10 byte requirement and the register requirement. Neither one is locally obvious which makes it easy to goof when modifying the code. >> Doing something less brittle than: >>> extern const char early_idt_handlers[NUM_EXCEPTION_VECTORS][10]; >> in segment.h might be a good idea as well. > > Yes, I agree. I will give it some thought. Eric
