On Sat, Mar 05, 2016 at 12:21:54AM +0000, Paul Burton wrote:
> On Fri, Mar 04, 2016 at 07:02:24PM +0000, Lars Persson wrote:
> > Hi
> >
> > Some further thoughts on the matter. You have so far not showed a
> > valid example of a race condition. The two examples you give in the
> > commit message are for a _single_ thread existing in the address space
> > (fork and execve).
>
> Hi Lars,
>
> Neither fork nor exec are limited to a single thread existing in the
> address space - I'm not sure what you're saying? fork by its very
> definition results in 2.
Ok, exec kinda is (it's late...). Still, fork clearly isn't.
Again - fork doesn't copy any user page. Copy_page_range just
manipulates PTE tree.
So, no user page cache flushes are needed on MIPS during fork, at least now.
- Leonid.
