19.01.2021 20:40, Andy Lutomirski пишет:
The implementation was rather buggy. It unconditionally marked PTEs read-only, even for VM_SHARED mappings. I'm not sure whether this is actually a problem, but it certainly seems unwise. More importantly, it released the mmap lock before flushing the TLB, which could allow a racing CoW operation to falsely believe that the underlying memory was not writable. I can't find any users at all of this mechanism, so just remove it. Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx> Cc: Linux-MM <linux-mm@xxxxxxxxx> Cc: Jason Gunthorpe <jgg@xxxxxxxx> Cc: x86@xxxxxxxxxx Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx> Cc: Jann Horn <jannh@xxxxxxxxxx> Cc: Jan Kara <jack@xxxxxxx> Cc: Yu Zhao <yuzhao@xxxxxxxxxx> Cc: Peter Xu <peterx@xxxxxxxxxx> Cc: Stas Sergeev <stsp@xxxxxxx> Cc: Brian Gerst <brgerst@xxxxxxxxx> Signed-off-by: Andy Lutomirski <luto@xxxxxxxxxx> ---
Acked-by: stsp2@xxxxxxxxx dosemu2 just uses mprotect(). The BIOSSEG stuff in vm86_32.c is also unused btw.
