Re: TLB flushes on fixmap changes

[Masami Hiramatsu <mhiramat@kernel.org>]

On Mon, 27 Aug 2018 10:13:29 +0200
Peter Zijlstra <peterz@infradead.org> wrote:

> On Mon, Aug 27, 2018 at 12:03:05PM +0900, Masami Hiramatsu wrote:
> > On Sun, 26 Aug 2018 11:09:58 +0200
> > Peter Zijlstra <peterz@infradead.org> wrote:
> 
> > > FWIW, before text_poke_bp(), text_poke() would only be used from
> > > stop_machine, so all the other CPUs would be stuck busy-waiting with
> > > IRQs disabled. These days, yeah, that's lots more dodgy, but yes
> > > text_mutex should be serializing all that.
> > 
> > I'm still not sure that speculative page-table walk can be done
> > over the mutex. Also, if the fixmap area is for aliasing
> > pages (which always mapped to memory), what kind of
> > security issue can happen?
> 
> So suppose CPU-A is doing the text_poke (let's say through text_poke_bp,
> such that other CPUs get to continue with whatever they're doing).
> 
> While at that point, CPU-B gets an interrupt, and the CPU's
> branch-trace-buffer for the IRET points to / near our fixmap. Then the
> CPU could do a speculative TLB fill based on the BTB value, either
> directly or indirectly (through speculative driven fault-ahead) of
> whatever is in te fixmap at the time.

Hmm, but how "near" is it enough? Since text_poke just map a non
executable alias page in fixmap, it is hard to suppose that IRET points
there (except for attacker change the IRET address).

I see that Intel CPU sometimes speculatively read-ahead the page tables,
but in that case, I guess we just need to keep fixmap area away from
text area. (Of course, it is hard to estimate how far is enough :( )

Anyway, I agree to introduce new page-table (and kthread?) for fixmap.

> Then CPU-A completes the text_poke and only does a local TLB invalidate
> on CPU-A, leaving CPU-B with an active translation.
> 
> *FAIL*

Ah, I got it. So on CPU-B, it can write-access to fixmap'd pages unless
the CPU-B shoot down the full TLB...

Thank you,

-- 
Masami Hiramatsu <mhiramat@kernel.org>