Re: [PATCH] pci-dma: Fix x86 dma_alloc_coherent to fully clear all pages returned

From: Alan Stern
Date: Wed Feb 18 2015 - 16:15:43 EST

On Wed, 18 Feb 2015, Andi Kleen wrote:

> > > Have you tried doing the experiments I suggested in
> > >
> > >
> > >
> > > to determine where the problem occurs?
> > >
> >
> > I was bogged down with other things lately and I haven't got a chance to
> > test that. But as you said, there's very few places where xhci
> > call this memory allocation. So I think the problem has been fairly
> > narrowed down for the XHCI folks.
> Also I don't really understand why we're even discussing this. The patch
> only makes an widely used API behave as it was before. Who knows who
> else was broken with this change. There's no sane way to audit all
> users. There is no real advantage of the new behavior.

We are discussing it because fixing problems is better than papering
around them.

> The only good way is to revert to old behavior, like in Tim's
> original patch. And doing it quickly for mainline and stable.

I will agree that applying the patch is a reasonable thing to do.
However, I also believe that it is important to fix the bugs revealed
by the API change.

> FWIW we have a large number of systems here that are broken
> without this change.

For all you know, they will still be broken even after the change is
applied. The breakage may become less obvious, but that doesn't mean
it will disappear entirely.

Alan Stern

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at