Re: [PATCH] x86: uaccess: fix regression in unsafe_get_user

[Andy Lutomirski]

On Tue, Nov 5, 2019 at 8:05 AM Arthur Gautier <baloo@xxxxxxxxx> wrote:
>
> On Tue, Nov 05, 2019 at 03:11:12PM +0100, Borislav Petkov wrote:
> > On Thu, Oct 10, 2019 at 04:49:51PM +0000, Arthur Gautier wrote:
> > > I did not receive neither the patch Andy provided, nor the comments made
> > > on it. But I'd be happy to help and/or take over to fix those if someone could
> > > send me both.
> >
> > Yes, please do, it seems Andy's busy. You can find the whole thread here:
> >
> > https://lore.kernel.org/lkml/20190215235901.23541-1-baloo@xxxxxxxxx/
> >
> > and you can download it in mbox format.
> >
> > Care to take Andy's patch, work in the comments I made to it, test it,
> > write a commit message, i.e., productize it?
> >
> > So that we get this thing moving...
> >
> > Thx.
> >
>
> Hello Boris,
>
> Thank you! But I believe this is the patch I sent, I know Andy sent a
> patchset with two patches, I believe privately (not copied to a public
> ML) to some of the recepients here. I got a copy of the second patch
> but not the first one.
>
> I believe from discussions here, that comments have been made on those
> patchset and because I was not Cc-ed on those patches, I do not have
> neither the full patchset nor the comments.
>
> I cannot take over the work, nor finish the patchset.
>
> Would anyone have a copy of the thread and could send them my way?
>

I forwarded it to you.

Here are the patches in git:

https://git.kernel.org/pub/scm/linux/kernel/git/luto/linux.git/log/?h=task_size

it's "unaligned", "fix test sparse warning", optionally
"strncpy_from_user: Zero any extra output bytes that get written",
"uaccess: Add a selftest for strncpy_from_user()", and
"strncpy_from_user: Don't overrun the input buffer onto the next page"

The basic summary is that Linus didn't like calling it bug fix, but it
might be acceptable as an improvement.  I also thought that the
KERNEL_DS oops was changed so it didn't trigger here.