Re: 2.6.36/2.6.37: broken compatibility with userspace input-utils ?

From: Dmitry Torokhov
Date: Tue Jan 25 2011 - 18:29:28 EST


On Tue, Jan 25, 2011 at 05:22:09PM -0500, Mark Lord wrote:
> On 11-01-25 05:00 PM, Mauro Carvalho Chehab wrote:
> > Em 25-01-2011 18:54, Dmitry Torokhov escreveu:
> >> On Wed, Jan 26, 2011 at 06:09:45AM +1000, Linus Torvalds wrote:
> >>> On Wed, Jan 26, 2011 at 2:48 AM, Dmitry Torokhov
> >>> <dmitry.torokhov@xxxxxxxxx> wrote:
> >>>>
> >>>> We should be able to handle the case where scancode is valid even though
> >>>> it might be unmapped yet. This is regardless of what version of
> >>>> EVIOCGKEYCODE we use, 1 or 2, and whether it is sparse keymap or not.
> >>>>
> >>>> Is it possible to validate the scancode by driver?
> >>>
> >>> More appropriately, why not just revert the thing? The version change
> >
> > Reverting the version increment is a bad thing. I agree with Dmitry that
> > an application that fails just because the API version were incremented
> > is buggy.
> >
> >> Well, then we'll break Ubuntu again as they recompiled their input-utils
> >> package (without fixing the check). And the rest of distros do not seem
> >> to be using that package...
> >
> > Reverting it will also break the ir-keytable userspace program that it is
> > meant to be used by the Remote Controller devices, and uses it to adjust
> > its behaviour to support RC's with more than 16 bits of scancodes.
> >
> > I agree that it is bad that the ABI broke, but reverting it will cause even
> > more damage.
>
> There we disagree. Sure it's a very poorly thought out interface,
> but the way to fix it is to put a new one along side the old,
> and put the old back the way it was before it got broken.
>
> I'm not making a fuss here for myself -- I'm more than capable of working
> around new kernel bugs like these, but for every person like me there are
> likely hundreds of others who simply get frustrated and give up.
>
> If you're worried about Ubuntu's adaptation to the buggy regression,
> then email their developers (kernel and input-utils packagers) explaining
> the revert, and they can coordination their kernel and input-utils updates
> to do the Right Thing.
>
> But for all of the rest of us, our systems are broken by this change.
>
> ...
>
>
> >>> As Mark said, breaking user space simply isn't acceptable. And since
> >>> breaking user space isn't acceptable, then incrementing the version is
> >>> stupid too.
> >>
> >> It might not have been the best idea to increment, however I maintain
> >> that if there exists version is can be changed. Otherwise there is no
> >> point in having version at all.
> >
> > Not arguing in favor of the version numbering, but it is easy to read
> > the version increment at the beginning of the application, and adjust
> > if the code will use EVIOCGKEYCODE or EVIOCGKEYCODE_V2 of the ioctl's,
> > depending on what kernel provides.
> >
> > Ok, we might be just calling the new ioctl and check for -ENOSYS at
> > the beginning, using some fake arguments.
> >
> >> As I said, reverting the version bump will cause yet another wave of
> >> breakages so I propose leaving version as is.
> >>
> >>>
> >>> The way we add new ioctl's is not by incrementing some "ABI version"
> >>> crap. It's by adding new ioctl's or system calls or whatever that
> >>> simply used to return -ENOSYS or other error before, while preserving
> >>> the old ABI. That way old binaries don't break (for _ANY_ reason), and
> >>> new binaries can see "oh, this doesn't support the new thing".
> >>
> >> That has been done as well; we have 2 new ioctls and kept 2 old ioctls.
>
> That's the problem: you did NOT keep the two old ioctls().
> Those got changed too.. so now we have four NEW ioctls(),
> none of which backward compatible with userspace.
>

Please calm down. This, in fact, is not new vs old ioctl problem but
rather particular driver (or rather set of drivers) implementation
issue. Even if we drop the new ioctls and convert the RC code to use the
old ones you'd be observing the same breakage as RC code responds with
-EINVAL to not-yet-established mappings.

I'll see what can be done for these drivers; I guess we could supply a
fake KEY_RESERVED entry for not mapped scancodes if there are mapped
scancodes "above" current one. That should result in the same behavior
for RCs as before.

--
Dmitry
--
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/