Hello Matthew, Michael, Fine work on the bug hunting, Matthew! Posts like yours are always fun to read, particularly because I doubt I ever would have tracked that down myself :)
Michael, I've been doing some research recently on x86 processor modes, and while I couldn't begin to offer any solutions due to my noobliness, some quick Google-Fu brought up these: http://forums.devshed.com/embedded-programming-163/unreal-mode-issue-672039.html *and* http://forum.sysinternals.com/topic21792_post115156.html While the threads are approaching a year old and appear unresolved, it's interesting that they're saying (I think) that access to expanded memory seems to work fine /without/ first setting up the CPU registers while PE=1. Neat. Obnoxious, but neat. I guess the real question is if this is a VirtualBox bug, feature, or both. Praise the almighty Oracle ;) Might it be possible to probe SMBIOS before attempting to set up the CPU, and if VirtualBox is detected, skip that setup? ...Not that I've any clue if it would work, of course. I've only been exploring x86 modes and peeking at assembly for a few weeks now :) Best Regards, Andrew Bobulsky On Mon, Jan 24, 2011 at 9:18 AM, Matthew Walster <[email protected]> wrote: > On 24 January 2011 12:45, Michael Brown <[email protected]> wrote: >> Thanks for tracking this down, Matthew. It's annoying that flat real mode >> seems to cause problems in some hypervisors, even though it works on real >> hardware. > > No problem at all - I had an hour spare and it was a good time to try > "git bisect" out ;) > >> I can't just revert this change, since some later commits (e.g. "[romprefix] >> Add .mrom format") depend upon it. I'll rework it to use some form of >> protected mode instead. > > I'm sorry to hear that. Please do let me know if there's any way I can > be of assistance, even if it's just testing on a range of platforms. > > Matthew Walster > _______________________________________________ > ipxe-devel mailing list > [email protected] > https://lists.ipxe.org/mailman/listinfo/ipxe-devel > _______________________________________________ ipxe-devel mailing list [email protected] https://lists.ipxe.org/mailman/listinfo/ipxe-devel
