Re: x86-64 bad pmds in 2.6.11.6 II

[Hugh Dickins <hugh@veritas.com>]

On Tue, 19 Apr 2005, Andi Kleen wrote:
> On Fri, Apr 15, 2005 at 06:58:20PM +0100, Hugh Dickins wrote:
> > 
> > I must confess, with all due respect to Andi, that I don't understand his
> > dismissal of the possibility that load_cr3 in leave_mm might be the fix
> > (to create_elf_tables writing user stack data into the pmd).
> 
> Sorry for the late answer.

Not at all.  I didn't expect you to persist in trying to persuade me,
thank you for doing so, and I apologize for taking your time on this.

> Ok, lets try again. The hole fixed by this patch only covers
> the case of an kernel thread with lazy mm doing some memory access
> (or more likely the CPU doing a prefetch there). But ELF loading
> never happens in lazy mm kernel threads.AFAIK in a "real" process
> the TLB is always fully consistent.
> 
> Does that explanation satisfy you? 

It does.  Well, I needed to restudy exec_mmap and switch_mm in detail,
and having done so, I agree that the only way you can get through
exec_mmap's activate_mm without fully flushing the cpu's TLB, is if
the active_mm matches the newly allocated mm (itself impossible since
there's a reference on the active_mm), and the cpu bit is still set
in cpu_vm_mask - precisely not the case if we went through leave_mm.
Yet I was claiming your leave_mm fix could flush TLB for exec_mmap
where it wasn't already done.

Sorry for letting the neatness of my pmd/stack story blind me
to its impossibility, and for wasting your time.

Hugh
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/