Re: [PATCH v4 2/2] ACPI / button: Add document for ACPI control method lid device restrictions

From: Bastien Nocera <hadess@hadess.net>
To: Benjamin Tissoires <benjamin.tissoires@gmail.com>,
	"Zheng, Lv" <lv.zheng@intel.com>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>,
	"Zhang, Rui" <rui.zhang@intel.com>,
	"Wysocki, Rafael J" <rafael.j.wysocki@intel.com>,
	"Rafael J. Wysocki" <rjw@rjwysocki.net>,
	"Brown, Len" <len.brown@intel.com>, Lv Zheng <zetalog@gmail.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-acpi@vger.kernel.org" <linux-acpi@vger.kernel.org>,
	"linux-input@vger.kernel.org" <linux-input@vger.kernel.org>
Subject: Re: [PATCH v4 2/2] ACPI / button: Add document for ACPI control method lid device restrictions
Date: Sun, 24 Jul 2016 13:28:56 +0200	[thread overview]
Message-ID: <1469359736.6761.110.camel@hadess.net> (raw)
In-Reply-To: <CAN+gG=FBE1e96qZsbAvV7gCizMByca3m38-qt=8ftsSXYFyQhA@mail.gmail.com>

On Fri, 2016-07-22 at 11:08 +0200, Benjamin Tissoires wrote:
> 
<snip>
> Then you just need to amend the documentation to say that the
> fallback
> of the KEY events is not the "future" but a way to get events on some
> reduced platforms and it will not be the default.
> Please make sure userspace knows that the default is the good SW_LID,
> and some particular cases will need to be handled through the KEY
> events, not the other way around.
> 
> [few thoughts later]
> 
> How about:
> - you send only one patch with the SW_LID ON/OFF or OFF/ON when we
> receive the notification on buggy platform
> - in the same patch, you add the documentation saying that on most
> platforms, LID is reliable but some don't provide a reliable LID
> state, but you guarantee to send an event when the state changes
> - in userspace, we add the hwdb which says "on this particular
> platform, don't rely on the actual state, but wait for events" ->
> this
> basically removes the polling on these platforms.
> 
> Bastien, Dmitry?
> 
> I still don't like relying on userspace to actually set the SW_LID
> back to open on resume, as we should not rely on some userspace
> program to set the value (but if logind really wants it, it's up to
> them).

>From my point of view, I would only send the events that can actually
be generated by the system, not any synthetic ones, because user-space
would have no way to know that this was synthetic, and how accurate it
would be.

So we'd have a separate API, or a separate event for the "close to
Windows behaviour" devices. We'd then use hwdb in udev to tag the
machines that don't have a reliable LID status, in user-space, so we
can have a quick turn around for those machines.

That should hopefully give us a way to tag test systems, so we can test
the new behaviour, though we'll certainly need to have some changes
made in the stack.

As Benjamin mentioned, it would be nice to have a list of devices that
don't work today, because of this problem.

Cheers