Luca, le dim. 28 août 2022 09:54:43 +0200, a ecrit: > Il 27 agosto 2022 20:02:40 UTC, Samuel Thibault <samuel.thiba...@gnu.org> ha > scritto: > > Samuel Thibault, le sam. 27 août 2022 21:41:57 +0200, a ecrit: > > > Mmm, on the set_cr3() in i386at_init call, it reboots... > > > > Ah, reducing the memory size to 2G fixed it. The kernel_pmap->l4base > > address was at 0xFFFD3000, apparently that was going up the roof :) > > It'd be useful to test with larger memory sizes, since that's one of the > > points of going 64bit ;) > > Yes that is still a limitation, I forgot to add it to the commit > description. I think the simplest way to solve it would be to link to the > upper 2G and enhance pmap to handle L4 tables, but I was trying to have > something running until a shell prompt first :)
Sure, but then please put something that caps the memory, so that people don't trap into the pitfall (they don't usually read the source code...) > On 64 bit we could also remove the direct mapping of all the phisical pages, > right? IIRC it's needed only by the linux drivers. We'll probably be able to remove that, yes. Samuel
