Re: PTE access rules & abstraction

[Hugh Dickins]

On Tue, 23 Sep 2008, Benjamin Herrenschmidt wrote:
> 
> The bug may have been there, as I said, lots of unwritten rules...
> sometimes broken. I'm not necessarily blaming you, but there have been
> lots of changes to the PTE accessors over the last 2 years and not
> always under any control :-)
> 
> In our case, the consequence is that the entry can be re-hashed because
> the fact that it was already hashed and where it was hashed, which is
> encoded in the PTE, gets lost by the clear. That means a potential
> duplicate entry in the hash. A hard to hit race, but possible. Such a
> condition is architecturally illegal and can cause things ranging from
> incorrect translation to machine checks or checkstops (generally, on
> LPAR machines, what will happen is your partition will get killed).

The powerpc bug whereof you write appears to have been there since ...
linux-2.4.0 or earlier:
			entry = ptep_get_and_clear(pte);
			set_pte(pte, pte_modify(entry, newprot));

But perhaps powerpc was slightly different back in those days.
It sounds to me like a bug in your current ptep_get_and_clear(),
not checking if already hashed?

> I know s390 has different issues & constraints. Martin told me during
> Plumbers that mprotect was probably also broken for him.

Then I hope he will probably send Linus the fix.

Though what we already have falls somewhat short of perfection,
I've much more enthusiasm for fixing its bugs, than for any fancy
redesign introducing its own bugs.  Others have more stamina!

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