Laptop screen doesn't come on when lid is reopened

Hi,

Like many Dell's, my Inspiron 640m turns off the LCD display when the lid is closed, but doesn't power it back up when the lid is opened.

This has been discussed before, with the conclusion that it's a BIOS bug which Linux can't fix (i.e. including a link in the kernel from ACPI to video isn't practical). However I'd like to revisit this as I believe I've found some information which wasn't discussed/observed before:

The lid notification method can be seen here: http://bugzilla.kernel.org/show_bug.cgi?id=5155#c6

Of specific interest, the lid notification comes with video notifications - the DSDT seems to be expecting the OS to do something with the video hardware.

I dug further and found that the first Store in that method is what turns the display off, and the following If always triggers (for both open and close events) in my testing. Sidenote: this means that the DSDT turns the display off both on close and on open.

I also found that no link to the video layer of the kernel would be needed to restore the display here: the _DSS method of the LCD device will happily turn the display back on.

I fixed my DSDT with this modification:

Method (LIDE, 0, NotSerialized)
{
+ // If lid open, enable video
+ If (LNotEqual(\_SB.LID._LID(), 0)) {

+            \_SB.PCI0.VID.LCD._DSS(0x80000001)
+            Notify (\_SB.LID, 0x80)
+            Return

+ }
+
+ // If lid closed, continue
+
+ // The following Store turns the LCD off

         Store (SMI (0x43, 0x00), Local0)
         If (LAnd (LNotEqual (Local0, 0x00), LNotEqual (Local0, 0x0F)))

Works nicely, woohoo!

I'd now like to investigate moving this modification into the kernel so that no DSDT changes are required. I'm running into some problems here:

If I create an acpi_driver alongside the button driver, the button driver claims the lid and my driver doesn't. I presume this is by design -- I was hoping that he subsystem would allow 2 drivers to work on the same device.

I then tried creating a new module which would use acpi_get_handle() to find the lid handle, and then acpi_install_notify_handler() to set up an event handler. I was hoping that the subsystem would allow multiple handlers on one event, but it appears I'm out of luck again: While the button driver is loaded and subscribed to this event, my driver gets the AE_ALREADY_EXISTS error code.

Are there other approaches I can take here? I'm hoping to create some kind of "Dell extras" driver, which contains a list of buggy systems, and listens for the lid events. When it detects a lid open event, it will call _DSS on the LCD device.

Thanks,
Daniel

First the generic acpi video driver needs to be enhanced to support the ACPI display switching features. Luming has posed patches along these lines already.

Then we should use that to discover if invoking _DSS fixes more machines than this pair of Dell laptops, and if it causes any others to break.

Assuming that it works on the Dells, hopefully more than the Dells, and doesn't break any others...

Then the button driver should take the lid event and cause the code in the video driver to be invoked. I'm not sure what the best method for inter-driver event like this, since both of them can be modules.

Finally, the video driver should have suspend/resume methods that poke _DSS and that should fix the Dell video on S3.

-Len

...

It will break all laptops which don't do anything special on lid events other than report it to the OS, so that we can do whatever we want (typically tell HAL or some script to suspend if the lid is closed). Like, say, thinkpads.

Yes, that seems like a good idea.

  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

Why would it do that? The obvious risk is that calling _DSS would break things, but I haven't seen that on Thinkpads. I don't think Len's proposing not passing the lid event through to userspace as well.

Matthew Garrett | [email protected]

I was not talking about calling _DSS per se, but calling it when you get lid events. Nevertheless, it was actually a big thinko. It probably won't break anything, as you said. Sorry about that.

And, if it is actually needed to turn the backlight back on when you press the lid *without triggering a suspend*, then it is probably going to be necessary.

Correct. As I said, it was a thinko on my part.

  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

I have taken a stab at this approach and it is working nicely. I'm sending patches to the linux-acpi list shortly.

Daniel