On Tue, Jun 02, 2020 at 01:43:20PM -0700, Linus Torvalds wrote: > On Tue, Jun 2, 2020 at 1:33 PM Michael S. Tsirkin <m...@redhat.com> wrote: > > > > Hmm are you sure we can drop it? access_ok is done in the context > > of the process. Access itself in the context of a kernel thread > > that borrows the same mm. IIUC if the process can be 32 bit > > while the kernel is 64 bit, access_ok in the context of the > > kernel thread will not DTRT. > > You're historically expected to just "set_fs()" when you do use_mm().
Right and we do that, but that still sets the segment according to the current thread's flags, right? E.g. I see: #define USER_DS MAKE_MM_SEG(TASK_SIZE_MAX) and #define TASK_SIZE (test_thread_flag(TIF_ADDR32) ? \ IA32_PAGE_OFFSET : TASK_SIZE_MAX) so if this is run from a kernel thread on a 64 bit kernel, we get TASK_SIZE_MAX even if we got the pointer from a 32 bit userspace address. > Then we fixed it in commit... > > Oh, when I look for it, I notice that it still hasn't gotten merged. > It's still pending, see > > https://lore.kernel.org/lkml/20200416053158.586887-4-...@lst.de/ > > for the current thing. > > Linus Maybe kthread_use_mm should also get the fs, not just mm. Then we can just use access_ok directly before the access. -- MST