Re: [RFC PATCH v1 01/12] Revert "drm/sysfs: Link DRM connectors to corresponding Type-C connectors"

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Dmitry,

On Tue, Sep 12, 2023 at 08:39:45PM +0300, Dmitry Baryshkov wrote:
> On 12/09/2023 14:05, Heikki Krogerus wrote:
> > On Tue, Sep 12, 2023 at 12:15:10AM +0300, Dmitry Baryshkov wrote:
> > > On 06/09/2023 16:38, Heikki Krogerus wrote:
> > > > On Wed, Sep 06, 2023 at 03:48:35PM +0300, Dmitry Baryshkov wrote:
> > > > > On Wed, 6 Sept 2023 at 15:44, Heikki Krogerus
> > > > > <heikki.krogerus@xxxxxxxxxxxxxxx> wrote:
> > > > > > 
> > > > > > On Tue, Sep 05, 2023 at 01:56:59PM +0300, Dmitry Baryshkov wrote:
> > > > > > > Hi Heikki,
> > > > > > > 
> > > > > > > On Tue, 5 Sept 2023 at 11:50, Heikki Krogerus
> > > > > > > <heikki.krogerus@xxxxxxxxxxxxxxx> wrote:
> > > > > > > > 
> > > > > > > > Hi Dmitry,
> > > > > > > > 
> > > > > > > > On Mon, Sep 04, 2023 at 12:41:39AM +0300, Dmitry Baryshkov wrote:
> > > > > > > > > The kdev->fwnode pointer is never set in drm_sysfs_connector_add(), so
> > > > > > > > > dev_fwnode() checks never succeed, making the respective commit NOP.
> > > > > > > > 
> > > > > > > > That's not true. The dev->fwnode is assigned when the device is
> > > > > > > > created on ACPI platforms automatically. If the drm_connector fwnode
> > > > > > > > member is assigned before the device is registered, then that fwnode
> > > > > > > > is assigned also to the device - see drm_connector_acpi_find_companion().
> > > > > > > > 
> > > > > > > > But please note that even if drm_connector does not have anything in
> > > > > > > > its fwnode member, the device may still be assigned fwnode, just based
> > > > > > > > on some other logic (maybe in drivers/acpi/acpi_video.c?).
> > > > > > > > 
> > > > > > > > > And if drm_sysfs_connector_add() is modified to set kdev->fwnode, it
> > > > > > > > > breaks drivers already using components (as it was pointed at [1]),
> > > > > > > > > resulting in a deadlock. Lockdep trace is provided below.
> > > > > > > > > 
> > > > > > > > > Granted these two issues, it seems impractical to fix this commit in any
> > > > > > > > > sane way. Revert it instead.
> > > > > > > > 
> > > > > > > > I think there is already user space stuff that relies on these links,
> > > > > > > > so I'm not sure you can just remove them like that. If the component
> > > > > > > > framework is not the correct tool here, then I think you need to
> > > > > > > > suggest some other way of creating them.
> > > > > > > 
> > > > > > > The issue (that was pointed out during review) is that having a
> > > > > > > component code in the framework code can lead to lockups. With the
> > > > > > > patch #2 in place (which is the only logical way to set kdev->fwnode
> > > > > > > for non-ACPI systems) probing of drivers which use components and set
> > > > > > > drm_connector::fwnode breaks immediately.
> > > > > > > 
> > > > > > > Can we move the component part to the respective drivers? With the
> > > > > > > patch 2 in place, connector->fwnode will be copied to the created
> > > > > > > kdev's fwnode pointer.
> > > > > > > 
> > > > > > > Another option might be to make this drm_sysfs component registration optional.
> > > > > > 
> > > > > > You don't need to use the component framework at all if there is
> > > > > > a better way of determining the connection between the DP and its
> > > > > > Type-C connector (I'm assuming that that's what this series is about).
> > > > > > You just need the symlinks, not the component.
> > > > > 
> > > > > The problem is that right now this component registration has become
> > > > > mandatory. And if I set the kdev->fwnode manually (like in the patch
> > > > > 2), the kernel hangs inside the component code.
> > > > > That's why I proposed to move the components to the place where they
> > > > > are really necessary, e.g. i915 and amd drivers.
> > > > 
> > > > So why can't we replace the component with the method you are
> > > > proposing in this series of finding out the Type-C port also with
> > > > i915, AMD, or whatever driver and platform (that's the only thing that
> > > > component is used for)?
> > > 
> > > The drm/msm driver uses drm_bridge for the pipeline (including the last DP
> > > entry) and the drm_bridge_connector to create the connector. I think that
> > > enabling i915 and AMD drivers to use drm_bridge fells out of scope for this
> > > series.
> > > 
> > > 
> > > > Determining the connection between a DP and its Type-C connector is
> > > > starting to get really important, so ideally we have a common solution
> > > > for that.
> > > 
> > > Yes. This is what we have been discussing with Simon for quite some time on
> > > #dri-devel.
> > > 
> > > Unfortunately I think the solution that got merged was pretty much hastened
> > > in instead of being well-thought. For example, it is also not always
> > > possible to provide the drm_connector / typec_connector links (as you can
> > > see from the patch7. Sometimes we can only express that this is a Type-C DP
> > > connector, but we can not easily point it to the particular USB-C port.
> > > 
> > > So, I'm not sure, how can we proceed here. Currently merged patch breaks
> > > drm/msm if we even try to use it by setting kdef->fwnode to
> > > drm_connector->fwnode. The pointed out `drivers/usb/typec/port-mapper.c` is
> > > an ACPI-only thing, which is not expected to work in a non-ACPI cases.
> > 
> > You really have to always supply not only the Type-C ports and partners,
> > but also the alt modes. You need them, firstly to keep things sane
> > inside kernel, but more importantly, so they are always exposed to the
> > user space, AND, always the same way. We have ABIs for all this stuff,
> > including the DP alt mode. Use them. No shortcuts.
> > 
> > So here's what you need to do. UCSI does not seem to bring you
> > anything useful, so just disable it for now. You don't need it. Your
> > port driver is clearly drivers/soc/qcom/pmic_glink_altmode.c, so
> > that's where you need to register all these components - the ports,
> > partners and alt modes. You have all the needed information there.
> 
> To make things even more complicate, UCSI is necessary for the USB part of
> the story. It handles vbus and direction.
> 
> > Only after you've done that we can start to look at how should the
> > connection between the DPs and their USB Type-C connectors be handled.
> 
> But sure enough, I can add typec port registration to the altmode driver.
> This will solve the 'port not existing' part of the story.
> 
> I'd like to hear your opinion on:
> 
> - components. Using them breaks drm/msm. How can we proceed?

I don't think replacing the components is going to be a problem once
you have described everything properly in you DT. I'm fairly certain now
that that is the main problem here. You don't have this connection
described in your DT as it should.

> - PATH property usage. This way we make USB-C DisplayPort behave like the
> MST ports.

That looks to me like an attempt to exploit a feature that is not
designed for this purposes at all. Just drop all that.

The connection has to be first described in your DT, and the way you
usually describe connections in DT is by using the device graph (OF
graph). It seems that you have everything needed for that - the USB
Type-C connectors have their own OF nodes (what you register as
drm_bridges are in fact USB Type-C connectors), and presumable you
also have OF nodes for all your video ports (DisplayPorts) - so
applying the graph between the two really should not be a problem. The
DP is endpoint for the USB Type-C connector, and vice versa.

After you have everything needed in your DT, the problem here isn't
actually much of a problem at all. We will have options how to move
forward after that.

Br,

-- 
heikki



[Index of Archives]     [Linux Media]     [Linux Input]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [Old Linux USB Devel Archive]

  Powered by Linux