Date:	Wed, 18 May 2016 01:25:50 +0000
From:	"Zheng, Lv" <lv.zheng@...el.com>
To:	"Rafael J. Wysocki" <rafael@...nel.org>
CC:	"Wysocki, Rafael J" <rafael.j.wysocki@...el.com>,
	"Rafael J. Wysocki" <rjw@...ysocki.net>,
	"Brown, Len" <len.brown@...el.com>, Lv Zheng <zetalog@...il.com>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	ACPI Devel Maling List <linux-acpi@...r.kernel.org>,
	"Bastien Nocera:" <hadess@...ess.net>
Subject: RE: [RFC PATCH 1/2] ACPI / button: Send "open" state after
 boot/resume

Hi, Rafael

Thanks for the review.

> From: rjwysocki@...il.com [mailto:rjwysocki@...il.com] On Behalf Of
> Rafael J. Wysocki
> Sent: Wednesday, May 18, 2016 7:37 AM
> To: Zheng, Lv <lv.zheng@...el.com>
> Cc: Wysocki, Rafael J <rafael.j.wysocki@...el.com>; Rafael J. Wysocki
> <rjw@...ysocki.net>; Brown, Len <len.brown@...el.com>; Lv Zheng
> <zetalog@...il.com>; Linux Kernel Mailing List <linux-
> kernel@...r.kernel.org>; ACPI Devel Maling List <linux-acpi@...r.kernel.org>;
> Bastien Nocera: <hadess@...ess.net>
> Subject: Re: [RFC PATCH 1/2] ACPI / button: Send "open" state after
> boot/resume
> 
> On Tue, May 17, 2016 at 10:27 AM, Lv Zheng <lv.zheng@...el.com> wrote:
> > (This patch hasn't been tested, it's sent for comments)
> 
> I have a couple of concerns (see below).
> 
> > Linux userspace (systemd-logind) keeps on rechecking lid state when the
> > lid state is closed. If it failed to update the lid state to open after
> > boot/resume, it could suspend the system. But as:
> > 1. Traditional ACPI platform may not generate the lid open event after
> >    resuming as the open event is actually handled by the BIOS and the system
> >    is then resumed from a FACS vector.
> > 2. The _LID control method's initial returning value is not reliable. The
> >    _LID control method is described to return the "current" lid state,
> >    however the word of "current" has ambiguity, many BIOSen return lid
> >    state upon last lid notification while the developers may think the
> >    BIOSen should always return the lid state upon last _LID evaluation.
> >    There won't be difference when we evaluate _LID during the runtime, the
> >    problem is the initial returning value of this function. When the BIOSen
> >    implement this control method with cached value, the initial returning
> >    value is likely not reliable.
> > Thus there is no mean for the ACPI lid driver to provide such an event
> > conveying correct current lid state. When there is no such an event or the
> > event conveys wrong result, false suspending can be examined.
> >
> > The root cause of the issue is systemd itself, it could handle the ACPI
> > control method lid device by implementing a special option like
> > LidSwitchLevelTriggered=False when it detected the ACPI lid device. However
> > there is no explicit documentation clarified the ambiguity, we need to
> > work it around in the kernel before systemd changing its mind.
> 
> The above doesn't explain how the issue is addressed here.
[Lv Zheng] 
The story is a bit long.
We can see several issues that some platform suspends right after boot/resume.
We noticed that on that platforms, _LID is always implemented with cached lid state returned.
And it's initial returning value may be "closed" after boot/resume.

It appears the acpi_lid_send_state() sent after boot/resume is the culprit to report the wrong lid state to the userspace.
But to our surprise, after delete the 2 lines, reporters still can see suspends after boot/resume.
That's because of systemd implementation.
It contains code logic that:
When the lid state is closed, a re-checking mechanism is installed.
So if we do not send any notification after boot/resume and the old lid state is "closed".
systemd determines to suspend in the re-checking mechanism.


> 
> > Link 1: https://lkml.org/2016/3/7/460
> > Link 2: https://github.com/systemd/systemd/issues/2087
> > Link 3: https://bugzilla.kernel.org/show_bug.cgi?id=89211
> >         https://bugzilla.kernel.org/show_bug.cgi?id=106151
> >         https://bugzilla.kernel.org/show_bug.cgi?id=106941
> > Signed-off-by: Lv Zheng <lv.zheng@...el.com>
> > Cc: Bastien Nocera: <hadess@...ess.net>
> > ---
> >  drivers/acpi/button.c |   63
> +++++++++++++++++++++++++++++++++++++++++++++----
> >  1 file changed, 59 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c
> > index 5c3b091..bb14ca5 100644
> > --- a/drivers/acpi/button.c
> > +++ b/drivers/acpi/button.c
> > @@ -53,6 +53,10 @@
> >  #define ACPI_BUTTON_DEVICE_NAME_LID    "Lid Switch"
> >  #define ACPI_BUTTON_TYPE_LID           0x05
> >
> > +#define ACPI_BUTTON_LID_INIT_IGNORE    0x00
> > +#define ACPI_BUTTON_LID_INIT_OPEN      0x01
> > +#define ACPI_BUTTON_LID_INIT_METHOD    0x02
> > +
> >  #define _COMPONENT             ACPI_BUTTON_COMPONENT
> >  ACPI_MODULE_NAME("button");
> >
> > @@ -105,6 +109,7 @@ struct acpi_button {
> >
> >  static BLOCKING_NOTIFIER_HEAD(acpi_lid_notifier);
> >  static struct acpi_device *lid_device;
> > +static u8 lid_init_state = ACPI_BUTTON_LID_INIT_OPEN;
> >
> >  /* --------------------------------------------------------------------------
> >                                FS Interface (/proc)
> > @@ -246,7 +251,8 @@ int acpi_lid_open(void)
> >  }
> >  EXPORT_SYMBOL(acpi_lid_open);
> >
> > -static int acpi_lid_send_state(struct acpi_device *device)
> > +static int acpi_lid_send_state(struct acpi_device *device,
> > +                              bool notify_init_state)
> >  {
> >         struct acpi_button *button = acpi_driver_data(device);
> >         unsigned long long state;
> > @@ -257,6 +263,10 @@ static int acpi_lid_send_state(struct acpi_device
> *device)
> >         if (ACPI_FAILURE(status))
> >                 return -ENODEV;
> >
> > +       if (notify_init_state &&
> > +           lid_init_state == ACPI_BUTTON_LID_INIT_OPEN)
> > +               state = 1;
> > +
> 
> Why do we need to complicate this function?
> 
> Can't we have a separate function for sending the fake "lid open" event?
[Lv Zheng] 

Yes, we can.
But I put the code here for reasons.

I intentionally kept the _LID evaluation right after boot/resume.
Because I validated Windows behavior.
It seems Windows evaluates _LID right after boot.
So I kept _LID evaluated right after boot to prevent compliance issues.

> 
> >         /* input layer checks if event is redundant */
> >         input_report_switch(button->input, SW_LID, !state);
> >         input_sync(button->input);
> > @@ -278,6 +288,13 @@ static int acpi_lid_send_state(struct acpi_device
> *device)
> >         return ret;
> >  }
> >
> > +static int acpi_lid_send_init_state(struct acpi_device *device)
> > +{
> > +       if (lid_init_state != ACPI_BUTTON_LID_INIT_IGNORE)
> > +               return acpi_lid_send_state(device, true);
> > +       return 0;
> > +}
> > +
> >  static void acpi_button_notify(struct acpi_device *device, u32 event)
> >  {
> >         struct acpi_button *button = acpi_driver_data(device);
> > @@ -290,7 +307,7 @@ static void acpi_button_notify(struct acpi_device
> *device, u32 event)
> >         case ACPI_BUTTON_NOTIFY_STATUS:
> >                 input = button->input;
> >                 if (button->type == ACPI_BUTTON_TYPE_LID) {
> > -                       acpi_lid_send_state(device);
> > +                       acpi_lid_send_state(device, false);
> 
> I wouldn't change this code at all.
> 
> >                 } else {
> >                         int keycode;
> >
> > @@ -335,7 +352,7 @@ static int acpi_button_resume(struct device *dev)
> >
> >         button->suspended = false;
> >         if (button->type == ACPI_BUTTON_TYPE_LID)
> > -               return acpi_lid_send_state(device);
> > +               return acpi_lid_send_init_state(device);
> >         return 0;
> >  }
> >  #endif
> > @@ -416,7 +433,7 @@ static int acpi_button_add(struct acpi_device *device)
> >         if (error)
> >                 goto err_remove_fs;
> >         if (button->type == ACPI_BUTTON_TYPE_LID) {
> > -               acpi_lid_send_state(device);
> > +               acpi_lid_send_init_state(device);
> >                 /*
> >                  * This assumes there's only one lid device, or if there are
> >                  * more we only care about the last one...
> > @@ -446,4 +463,42 @@ static int acpi_button_remove(struct acpi_device
> *device)
> >         return 0;
> >  }
> >
> > +static int param_set_lid_init_state(const char *val, struct kernel_param *kp)
> > +{
> > +       int result = 0;
> > +
> > +       if (!strncmp(val, "open", sizeof("open") - 1)) {
> > +               lid_init_state = ACPI_BUTTON_LID_INIT_OPEN;
> > +               pr_info("Notify initial lid state as open\n");
> > +       } else if (!strncmp(val, "method", sizeof("method") - 1)) {
> > +               lid_init_state = ACPI_BUTTON_LID_INIT_METHOD;
> > +               pr_info("Notify initial lid state with _LID return value\n");
> > +       } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) {
> > +               lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE;
> > +               pr_info("Do not notify initial lid state\n");
> > +       } else
> > +               result = -EINVAL;
> > +       return result;
> > +}
> > +
> > +static int param_get_lid_init_state(char *buffer, struct kernel_param *kp)
> > +{
> > +       switch (lid_init_state) {
> > +       case ACPI_BUTTON_LID_INIT_OPEN:
> > +               return sprintf(buffer, "open");
> > +       case ACPI_BUTTON_LID_INIT_METHOD:
> > +               return sprintf(buffer, "method");
> > +       case ACPI_BUTTON_LID_INIT_IGNORE:
> > +               return sprintf(buffer, "ignore");
> > +       default:
> > +               return sprintf(buffer, "invalid");
> > +       }
> > +       return 0;
> > +}
> > +
> > +module_param_call(lid_init_state,
> > +                 param_set_lid_init_state, param_get_lid_init_state,
> > +                 NULL, 0644);
> > +MODULE_PARM_DESC(lid_init_state, "Behavior for reporting LID initial
> state");
> > +
> 
> I'm not seeing a particular value in having this command line switch
> to be honest.  Apparently, the issue can be worked around from user
> space in any case, so why do we need one more way to work around it?
> 
> >  module_acpi_driver(acpi_button_driver);
> > --
> 
> The main concern is general, though.  Evidently, we send fake lid
> input events to user space on init and resume.  I don't think this is
> a good idea, because it may confuse systems like Chrome that want to
> implement "dark resume" scenarios and may rely on input events to
> decide whether to start a UI or suspend again.
> 
> Thus it might be better to simply drop the sending of those fake input
> events from the code.
[Lv Zheng] 
If we did this right now, many other userspace could be broken.
So we prepared the options to allow users to choose.

Thanks and best regards
-Lv

Powered by blists - more mailing lists