Re: Note describing poor dcache utilization under high memory pressure

Daniel Phillips (phillips@bonn-fries.net)
Tue, 29 Jan 2002 23:53:04 +0100


On January 29, 2002 10:50 pm, Linus Torvalds wrote:
> On Tue, 29 Jan 2002, Oliver Xymoron wrote:
> >
> > I don't think read-only for the tables is sufficient if the pages
> > themselves are writable.
>
> At least on x86, the WRITE bit in the page directory entries will override
> any bits int he PTE. In other words, it doesn't make the page directory
> entries thmselves unwritable - it makes the final pages unwritable.

Ah, didn't know that.

> Which are exactly the semantics we want.

Yes. This feature might be useful to grab back a few cycles at the expense
of some extra complexity. It is not however, a fundamental change to my
algorithm, just a decoration of it. It's also possible that the cost of the
resulting extra fault will wipe out the (small) saving of setting pte entries
RO in the average case.

It's likely there are architectures where this won't work, and just not doing
this optimization is the correct approach.

-- 
Daniel
-
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/