On Sun, Sep 20, 2020 at 10:42 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> Yeah, that looks much easier to explain. Ack.
Btw, one thing that might be a good idea at least initially is to add
a check for p->kmap_ctrl.idx being zero at fork, exit and maybe
syscall return time (but that last one may be too cumbersome to really
worry about).
The kmap_atomic() interface basically has a lot of coverage for leaked
as part of all the "might_sleep()" checks sprinkled around, The new
kmap_temporary/local/whatever wouldn't have that kind of incidental
debug checking, and any leaked kmap indexes would be rather hard to
debug (much) later when they cause index overflows or whatever.
Linus
