On Thu, Dec 14, 2017 at 11:43 AM, Linus Torvalds <torva...@linux-foundation.org> wrote: > On Thu, Dec 14, 2017 at 8:20 AM, Andy Lutomirski <l...@kernel.org> wrote: >> >> If this turns out to need reverting because it breaks Wine or >> something, we're really going to regret it. > > I really don't see that as very likely. We already play other (much > more fundamental) games with segments. >
I dunno. Maybe Wine or DOSEMU apps expect to be able to create a non-accessed segment and then read out the accessed bit using LAR or modify_ldt() later. > But I do agree that it would be good to consider this "turn LDT > read-only" a separate series just in case. Which kind of kills the whole thing. There's no way the idea of putting the LDT in a VMA is okay if it's RW. You just get the kernel to put_user() a call gate into it and it's game over. I have a competing patch that just aliases the LDT high up in kernel land and shares it in the user tables. I like a lot of the cleanups in this series, but I don't like the actual LDT-in-a-VMA part.
