>Jens, I had only a very brief look at your patch, but I don't see you flush
>the TLB anywhere... you'll need to either flush the complete TLB (not
>recommended) or use INVLPG to invalidate the page, in order for your
>changes to the page tables to take effect.
>Can this be the problem ?
Hm, a good point, but not the (only?) solution. I added TLB flushing, it's
not better now. But TLB was flushed due to host->guest context switch and
this virtual address touched the guest first time. Think, there was no wrong
TLB entry, but your hint is important for this scenario with other virtual
addresses! Thanx!
BTW, another bug was
pageTable += ((addr >> 12) & 0x3ff);
which has to be
pageTable += ((addr >> 12) & 0x3ff) << 2;
jens
