Merge tag 'gcc-plugins-v4.9-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git...
[cascardo/linux.git] / drivers / acpi / button.c
index 31abb0b..e19f530 100644 (file)
@@ -19,6 +19,8 @@
  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  */
 
+#define pr_fmt(fmt) "ACPI : button: " fmt
+
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/init.h>
@@ -104,6 +106,8 @@ struct acpi_button {
        struct input_dev *input;
        char phys[32];                  /* for input device */
        unsigned long pushed;
+       int last_state;
+       ktime_t last_time;
        bool suspended;
 };
 
@@ -111,6 +115,10 @@ static BLOCKING_NOTIFIER_HEAD(acpi_lid_notifier);
 static struct acpi_device *lid_device;
 static u8 lid_init_state = ACPI_BUTTON_LID_INIT_METHOD;
 
+static unsigned long lid_report_interval __read_mostly = 500;
+module_param(lid_report_interval, ulong, 0644);
+MODULE_PARM_DESC(lid_report_interval, "Interval (ms) between lid key events");
+
 /* --------------------------------------------------------------------------
                               FS Interface (/proc)
    -------------------------------------------------------------------------- */
@@ -134,10 +142,79 @@ static int acpi_lid_notify_state(struct acpi_device *device, int state)
 {
        struct acpi_button *button = acpi_driver_data(device);
        int ret;
+       ktime_t next_report;
+       bool do_update;
+
+       /*
+        * In lid_init_state=ignore mode, if user opens/closes lid
+        * frequently with "open" missing, and "last_time" is also updated
+        * frequently, "close" cannot be delivered to the userspace.
+        * So "last_time" is only updated after a timeout or an actual
+        * switch.
+        */
+       if (lid_init_state != ACPI_BUTTON_LID_INIT_IGNORE ||
+           button->last_state != !!state)
+               do_update = true;
+       else
+               do_update = false;
+
+       next_report = ktime_add(button->last_time,
+                               ms_to_ktime(lid_report_interval));
+       if (button->last_state == !!state &&
+           ktime_after(ktime_get(), next_report)) {
+               /* Complain the buggy firmware */
+               pr_warn_once("The lid device is not compliant to SW_LID.\n");
 
-       /* input layer checks if event is redundant */
-       input_report_switch(button->input, SW_LID, !state);
-       input_sync(button->input);
+               /*
+                * Send the unreliable complement switch event:
+                *
+                * On most platforms, the lid device is reliable. However
+                * there are exceptions:
+                * 1. Platforms returning initial lid state as "close" by
+                *    default after booting/resuming:
+                *     https://bugzilla.kernel.org/show_bug.cgi?id=89211
+                *     https://bugzilla.kernel.org/show_bug.cgi?id=106151
+                * 2. Platforms never reporting "open" events:
+                *     https://bugzilla.kernel.org/show_bug.cgi?id=106941
+                * On these buggy platforms, the usage model of the ACPI
+                * lid device actually is:
+                * 1. The initial returning value of _LID may not be
+                *    reliable.
+                * 2. The open event may not be reliable.
+                * 3. The close event is reliable.
+                *
+                * But SW_LID is typed as input switch event, the input
+                * layer checks if the event is redundant. Hence if the
+                * state is not switched, the userspace cannot see this
+                * platform triggered reliable event. By inserting a
+                * complement switch event, it then is guaranteed that the
+                * platform triggered reliable one can always be seen by
+                * the userspace.
+                */
+               if (lid_init_state == ACPI_BUTTON_LID_INIT_IGNORE) {
+                       do_update = true;
+                       /*
+                        * Do generate complement switch event for "close"
+                        * as "close" is reliable and wrong "open" won't
+                        * trigger unexpected behaviors.
+                        * Do not generate complement switch event for
+                        * "open" as "open" is not reliable and wrong
+                        * "close" will trigger unexpected behaviors.
+                        */
+                       if (!state) {
+                               input_report_switch(button->input,
+                                                   SW_LID, state);
+                               input_sync(button->input);
+                       }
+               }
+       }
+       /* Send the platform triggered reliable event */
+       if (do_update) {
+               input_report_switch(button->input, SW_LID, !state);
+               input_sync(button->input);
+               button->last_state = !!state;
+               button->last_time = ktime_get();
+       }
 
        if (state)
                pm_wakeup_event(&device->dev, 0);
@@ -411,6 +488,8 @@ static int acpi_button_add(struct acpi_device *device)
                strcpy(name, ACPI_BUTTON_DEVICE_NAME_LID);
                sprintf(class, "%s/%s",
                        ACPI_BUTTON_CLASS, ACPI_BUTTON_SUBCLASS_LID);
+               button->last_state = !!acpi_lid_evaluate_state(device);
+               button->last_time = ktime_get();
        } else {
                printk(KERN_ERR PREFIX "Unsupported hid [%s]\n", hid);
                error = -ENODEV;