Re: [PATCH] Input: add support for polling to input devices

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Tue, Aug 20, 2019 at 01:56:53PM +0200, Michal Vokáč wrote:
> On 13. 08. 19 16:04, Benjamin Tissoires wrote:
> > On Mon, Aug 12, 2019 at 7:11 PM Dmitry Torokhov
> > <dmitry.torokhov@gmail.com> wrote:
> > > 
> > > Hi Benjamin,
> > > 
> > > On Mon, Aug 12, 2019 at 06:50:38PM +0200, Benjamin Tissoires wrote:
> > > > Hi Dmitry,
> > > > 
> > > > On Fri, Aug 9, 2019 at 7:35 PM Dmitry Torokhov
> > > > <dmitry.torokhov@gmail.com> wrote:
> > > > > 
> > > > > Separating "normal" and "polled" input devices was a mistake, as often we
> > > > > want to allow the very same device work on both interrupt-driven and
> > > > > polled mode, depending on the board on which the device is used.
> > > > > 
> > > > > This introduces new APIs:
> > > > > 
> > > > > - input_setup_polling
> > > > > - input_set_poll_interval
> > > > > - input_set_min_poll_interval
> > > > > - input_set_max_poll_interval
> > > > > 
> > > > > These new APIs allow switching an input device into polled mode with sysfs
> > > > > attributes matching drivers using input_polled_dev APIs that will be
> > > > > eventually removed.
> > > > 
> > > > Are you sure that using sysfs is the correct way here?
> > > > I would think using generic properties that can be overwritten by the
> > > > DSDT or by the device tree would make things easier from a driver
> > > > point of view.
> > > 
> > > Couple of points: I wanted it to be compatible with input-polldev.c so
> > > the sysfs attributes must match (so that we can convert existing drivers
> > > and zap input-polldev).
> > 
> > Oh, I missed that. Good point.
> > 
> > > I also am not sure if polling parameters are
> > > property of board, or it is either fundamental hardware limitation or
> > > simply desired system behavior.
> > 
> > I think it's a combination of everything: sometimes the board misses
> > the capability to not do IRQs for that device, and using properties
> > would be better here: you can define them where you need (board,
> > platform or device level), and have a working platfrom from the kernel
> > description entirely.
> > However, it doesn't solve the issue of input-polldev, so maybe
> > properties should be added on top of this sysfs.
> > 
> > > If Rob is OK with adding device
> > > properties I'd be fine adding them as a followup, otherwise we can have
> > > udev adjust the behavior as needed for given box shortly after boot.
> > 
> > Fair enough.
> > 
> > > 
> > > > 
> > > > Also, checkpatch complains about a few octal permissions that are
> > > > preferred over symbolic ones, and there is a possible 'out of memory'
> > > > nessage at drivers/input/input-poller.c:75.
> > > 
> > > Yes, there is. It is there so we would know what device we were trying
> > > to set up when OOM happened. You can probable decipher the driver from
> > > the stack trace, but figuring particular device instance is harder.
> > 
> > Could you add a comment there explaining this choice? I have a feeling
> > you'll have to refuse a few patches of people running coccinelle
> > scripts and be too happy to send a kernel patch.

Done.

> > 
> > Other than that:
> > Acked-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
> > 
> 
> Hi Dmitry,
> 
> what is the status of this patch? Are we still waiting for Rob to
> comment on the device properties or is this ready to land?

I applied it just now.

> 
> Little bit OT question: what tree/branch do you use to apply patches?
> According to the mailing list you recently applied some patches but
> I can not find them here [1].

Patches that go into current cycle will be in 'for-linus' branch.
Patches that will be merged in the next merge window (like this one)
will end up in 'next' branch.

Thanks.

-- 
Dmitry