DRM modesetting & sysfs

DRM modesetting & sysfs

[Jesse Barnes]

[Sorry for the DUP, forgot to cc lkml]

I just pushed a few changes updating the DRM modesetting sysfs support, both 
for debugging and eventual HAL friendliness.

So far, the support is limited to describing outputs and generating hotplug 
events.  A typical "card0" directory now looks like this:

.
|-- card0-DAC-1
|   |-- device -> ../../../../../../devices/pci0000:00/0000:00:02.0/drm/card0
|   |-- dpms
|   |-- edid
|   |-- modes
|   |-- power
|   |   `-- wakeup
|   |-- status
|   |-- subsystem -> ../../../../../../class/drm
|   `-- uevent
|-- card0-LVDS-1
|   |-- device -> ../../../../../../devices/pci0000:00/0000:00:02.0/drm/card0
|   |-- dpms
|   |-- edid
|   |-- modes
|   |-- power
|   |   `-- wakeup
|   |-- status
|   |-- subsystem -> ../../../../../../class/drm
|   `-- uevent
|-- dev
|-- device -> ../../../../../devices/pci0000:00/0000:00:02.0
|-- dri_library_name
|-- power
|   `-- wakeup
|-- subsystem -> ../../../../../class/drm
`-- uevent

Each output is listed, prefixed with its associated card number to 
disambiguate it at the top level /sys/class/drm directory (where it also has 
some symlinks).  Currently, only DPMS, EDID, mode list and connection status 
properties are exported, but we could list more as needed.  You can 
use 'udevmonitor --env' to look for hotplug events; you should see several at 
load time, and presumably some at connection hotplug time (though I haven't 
gotten that working yet, any ideas Jakob?).

Any thoughts on the interface?  Anything in particular people would like to 
see?

Thanks,
Jesse

Re: DRM modesetting & sysfs

[Jakob Bornecrantz]

On Tue, Apr 8, 2008 at 10:17 PM, Jesse Barnes <jbarnes@virtuousgeek.org> wrote:
>
>  I just pushed a few changes updating the DRM modesetting sysfs support, both
>  for debugging and eventual HAL friendliness.
>
>  So far, the support is limited to describing outputs and generating hotplug
>  events.  A typical "card0" directory now looks like this:
>
>  .
>  |-- card0-DAC-1
>  |   |-- device -> ../../../../../../devices/pci0000:00/0000:00:02.0/drm/card0
>  |   |-- dpms
>  |   |-- edid
>  |   |-- modes
>  |   |-- power
>  |   |   `-- wakeup
>  |   |-- status
>  |   |-- subsystem -> ../../../../../../class/drm
>  |   `-- uevent
>  |-- card0-LVDS-1
>  |   |-- device -> ../../../../../../devices/pci0000:00/0000:00:02.0/drm/card0
>  |   |-- dpms
>  |   |-- edid
>  |   |-- modes
>  |   |-- power
>  |   |   `-- wakeup
>  |   |-- status
>  |   |-- subsystem -> ../../../../../../class/drm
>  |   `-- uevent
>  |-- dev
>  |-- device -> ../../../../../devices/pci0000:00/0000:00:02.0
>  |-- dri_library_name
>  |-- power
>  |   `-- wakeup
>  |-- subsystem -> ../../../../../class/drm
>  `-- uevent
>
>  Each output is listed, prefixed with its associated card number to
>  disambiguate it at the top level /sys/class/drm directory (where it also has
>  some symlinks).  Currently, only DPMS, EDID, mode list and connection status
>  properties are exported, but we could list more as needed.  You can
>  use 'udevmonitor --env' to look for hotplug events; you should see several at
>  load time, and presumably some at connection hotplug time (though I haven't
>  gotten that working yet, any ideas Jakob?).
>

I was going to suggest that you plug it into the hotplug_stage_two
function but it looks like you have already done that. Things might be
routed differently now then since the last time I looked at the code,
are you sure that stage_two is being run?

>  Any thoughts on the interface?  Anything in particular people would like to
>  see?
>

Looks good, can you write to dpms, that is turn a monitor on and of,
that would be slightly cool but not overly useful. On the whole I
think exporting properties are fine, but I don't think setting modes
and crtc's are to be done in sysfs. However it is entirely possible.

>  Thanks,
>  Jesse

Cheers Jakob.

Re: DRM modesetting & sysfs

[Jesse Barnes]

On Wednesday, April 09, 2008 1:57 am Jakob Bornecrantz wrote:
> I was going to suggest that you plug it into the hotplug_stage_two
> function but it looks like you have already done that. Things might be
> routed differently now then since the last time I looked at the code,
> are you sure that stage_two is being run?

I'll have to check, I'm not sure I'm even getting interrupts...

> >  Any thoughts on the interface?  Anything in particular people would like
> > to see?
>
> Looks good, can you write to dpms, that is turn a monitor on and of,
> that would be slightly cool but not overly useful. On the whole I
> think exporting properties are fine, but I don't think setting modes
> and crtc's are to be done in sysfs. However it is entirely possible.

Agreed.  I view this more as a way of exporting information about the DRM 
configuration to userspace rather than a way to control the whole thing.  I 
think we should stick with ioctls for the latter...  Anyway thanks for 
checking it out.

Jesse

Re: DRM modesetting & sysfs

[Alan Hourihane]

On Wed, 2008-04-09 at 08:17 -0700, Jesse Barnes wrote:
> On Wednesday, April 09, 2008 1:57 am Jakob Bornecrantz wrote:
> > I was going to suggest that you plug it into the hotplug_stage_two
> > function but it looks like you have already done that. Things might be
> > routed differently now then since the last time I looked at the code,
> > are you sure that stage_two is being run?
> 
> I'll have to check, I'm not sure I'm even getting interrupts...

Jesse,

Is this i915 or i945/i965 you are testing on ??

Alan.

Re: DRM modesetting & sysfs

[Jesse Barnes]

On Wednesday, April 09, 2008 9:15 am Alan Hourihane wrote:
> On Wed, 2008-04-09 at 08:17 -0700, Jesse Barnes wrote:
> > On Wednesday, April 09, 2008 1:57 am Jakob Bornecrantz wrote:
> > > I was going to suggest that you plug it into the hotplug_stage_two
> > > function but it looks like you have already done that. Things might be
> > > routed differently now then since the last time I looked at the code,
> > > are you sure that stage_two is being run?
> >
> > I'll have to check, I'm not sure I'm even getting interrupts...
>
> Jesse,
>
> Is this i915 or i945/i965 you are testing on ??

I've been testing on my i915 based laptop, haven't tried my other systems 
yet...

Jesse

Re: DRM modesetting & sysfs

[Alan Hourihane]

On Wed, 2008-04-09 at 09:34 -0700, Jesse Barnes wrote:
> On Wednesday, April 09, 2008 9:15 am Alan Hourihane wrote:
> > On Wed, 2008-04-09 at 08:17 -0700, Jesse Barnes wrote:
> > > On Wednesday, April 09, 2008 1:57 am Jakob Bornecrantz wrote:
> > > > I was going to suggest that you plug it into the hotplug_stage_two
> > > > function but it looks like you have already done that. Things might be
> > > > routed differently now then since the last time I looked at the code,
> > > > are you sure that stage_two is being run?
> > >
> > > I'll have to check, I'm not sure I'm even getting interrupts...
> >
> > Jesse,
> >
> > Is this i915 or i945/i965 you are testing on ??
> 
> I've been testing on my i915 based laptop, haven't tried my other systems 
> yet...

With VGA (i.e. ADPA) output ?

If so, that's the problem. From what I can tell of i915 docs, they only
support hotplug for SDVO devices only, so that's how it's coded up at
the moment. For i945/i965 it can go the whole hog.

Alan.

Re: DRM modesetting & sysfs

[Jesse Barnes]

On Wednesday, April 09, 2008 10:23 am Alan Hourihane wrote:
> On Wed, 2008-04-09 at 09:34 -0700, Jesse Barnes wrote:
> > On Wednesday, April 09, 2008 9:15 am Alan Hourihane wrote:
> > > On Wed, 2008-04-09 at 08:17 -0700, Jesse Barnes wrote:
> > > > On Wednesday, April 09, 2008 1:57 am Jakob Bornecrantz wrote:
> > > > > I was going to suggest that you plug it into the hotplug_stage_two
> > > > > function but it looks like you have already done that. Things might
> > > > > be routed differently now then since the last time I looked at the
> > > > > code, are you sure that stage_two is being run?
> > > >
> > > > I'll have to check, I'm not sure I'm even getting interrupts...
> > >
> > > Jesse,
> > >
> > > Is this i915 or i945/i965 you are testing on ??
> >
> > I've been testing on my i915 based laptop, haven't tried my other systems
> > yet...
>
> With VGA (i.e. ADPA) output ?
>
> If so, that's the problem. From what I can tell of i915 docs, they only
> support hotplug for SDVO devices only, so that's how it's coded up at
> the moment. For i945/i965 it can go the whole hog.

Ah, maybe that's it, yeah I'm using VGA via ADPA.  I don't see any IIR or 
PIPEASTAT bits set when I plug in the display, so I guess we're stuck with 
manually probing on these devices.

Thanks,
Jesse

Re: DRM modesetting & sysfs

[Jesse Barnes]

On Wednesday, April 09, 2008 10:39 am Jesse Barnes wrote:
> > With VGA (i.e. ADPA) output ?
> >
> > If so, that's the problem. From what I can tell of i915 docs, they only
> > support hotplug for SDVO devices only, so that's how it's coded up at
> > the moment. For i945/i965 it can go the whole hog.
>
> Ah, maybe that's it, yeah I'm using VGA via ADPA.  I don't see any IIR or
> PIPEASTAT bits set when I plug in the display, so I guess we're stuck with
> manually probing on these devices.

Looks like i915 may support TV hotplug too, though I haven't finished the TV 
bits yet; I'll see about that today.

However, on my 965 I get hotplug interrupts, the output is configured, and I 
even get uevents with the new sysfs code, yay! :)

Jesse