drm/doc: Document uapi requirements in DRM

Commit Message

Daniel Vetter Aug. 19, 2016, 8:50 p.m. UTC

Everyone knows them, except all the new folks joining from the ARM
side haven't lived through all the pain of the past years and are
entirely surprised when I raise this. Definitely time to document
this.

Last time this was a big discussion was about 6 years ago, when qcom
tried to land a kernel driver without userspace. Dave Airlie made the
rules really clear:

http://airlied.livejournal.com/73115.html

This write-up here is essentially what I've put into a presentation a
while ago, which was also reviewed by Dave:

http://blog.ffwll.ch/2015/05/gfx-kernel-upstreaming-requirements.html

Cc: Dave Airlie <[e-mail]>
Cc: Oded Gabbay <[e-mail]>
Cc: Russell King <[e-mail]>
Cc: Tomi Valkeinen <[e-mail]>
Cc: Eric Anholt <[e-mail]>
Cc: Thomas Hellstrom <[e-mail]>
Cc: Sinclair Yeh <[e-mail]>
Cc: Lucas Stach <[e-mail]>
Cc: Benjamin Gaignard <[e-mail]>
Cc: Mark Yao <[e-mail]>
Cc: Laurent Pinchart <[e-mail]>
Cc: Ben Skeggs <[e-mail]>
Cc: Rob Clark <[e-mail]>
Cc: CK Hu <[e-mail]>
Cc: Xinliang Liu <[e-mail]>
Cc: Philipp Zabel <[e-mail]>
Cc: Stefan Agner <[e-mail]>
Cc: Inki Dae <[e-mail]>
Cc: Maxime Ripard  <[e-mail]>
Cc: Boris Brezillon <[e-mail]>
Cc: Jani Nikula <[e-mail]>
Cc: Daniel Vetter <[e-mail]>
Cc: Thierry Reding <[e-mail]>
Cc: Christian König <[e-mail]>
Cc: Alex Deucher <[e-mail]>
Cc: Gerd Hoffmann <[e-mail]>
Cc: Brian Starkey <[e-mail]>
Cc: Liviu Dudau <[e-mail]>
Cc: Alexey Brodkin <[e-mail]>
Acked-by: Dave Airlie <[e-mail]>
Signed-off-by: Daniel Vetter <[e-mail]>
---
 Documentation/gpu/drm-uapi.rst | 67 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 67 insertions(+)

Comments

Sinclair Yeh Aug. 26, 2016, 2:09 p.m. UTC | #1

Reviewed-by: Sinclair Yeh <[e-mail]>

On Fri, Aug 19, 2016 at 10:50:38PM +0200, Daniel Vetter wrote:
> Everyone knows them, except all the new folks joining from the ARM
> side haven't lived through all the pain of the past years and are
> entirely surprised when I raise this. Definitely time to document
> this.
> 
> Last time this was a big discussion was about 6 years ago, when qcom
> tried to land a kernel driver without userspace. Dave Airlie made the
> rules really clear:
> 
> https://urldefense.proofpoint.com/v2/url?u=http-3A__airlied.livejournal.com_73115.html&d=CwIDaQ&c=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEs&r=w9Iu3o4zAy-3-s8MFvrNSQ&m=YlKw0M9a4htB0LJuAWmS2fxNN9GUjDewsVWvDM3bkI8&s=sbslwGx1LSh92YTxx8DnoKFau_MPXjwM_uK6S4l4y8o&e= 
> 
> This write-up here is essentially what I've put into a presentation a
> while ago, which was also reviewed by Dave:
> 
> https://urldefense.proofpoint.com/v2/url?u=http-3A__blog.ffwll.ch_2015_05_gfx-2Dkernel-2Dupstreaming-2Drequirements.html&d=CwIDaQ&c=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEs&r=w9Iu3o4zAy-3-s8MFvrNSQ&m=YlKw0M9a4htB0LJuAWmS2fxNN9GUjDewsVWvDM3bkI8&s=EZbXCRJDTIcS1Rp5LG_A2GxBXRvTwl5jFYy6tLkfg-g&e= 
> 
> Cc: Dave Airlie <[e-mail]>
> Cc: Oded Gabbay <[e-mail]>
> Cc: Russell King <[e-mail]>
> Cc: Tomi Valkeinen <[e-mail]>
> Cc: Eric Anholt <[e-mail]>
> Cc: Thomas Hellstrom <[e-mail]>
> Cc: Sinclair Yeh <[e-mail]>
> Cc: Lucas Stach <[e-mail]>
> Cc: Benjamin Gaignard <[e-mail]>
> Cc: Mark Yao <[e-mail]>
> Cc: Laurent Pinchart <[e-mail]>
> Cc: Ben Skeggs <[e-mail]>
> Cc: Rob Clark <[e-mail]>
> Cc: CK Hu <[e-mail]>
> Cc: Xinliang Liu <[e-mail]>
> Cc: Philipp Zabel <[e-mail]>
> Cc: Stefan Agner <[e-mail]>
> Cc: Inki Dae <[e-mail]>
> Cc: Maxime Ripard  <[e-mail]>
> Cc: Boris Brezillon <[e-mail]>
> Cc: Jani Nikula <[e-mail]>
> Cc: Daniel Vetter <[e-mail]>
> Cc: Thierry Reding <[e-mail]>
> Cc: Christian König <[e-mail]>
> Cc: Alex Deucher <[e-mail]>
> Cc: Gerd Hoffmann <[e-mail]>
> Cc: Brian Starkey <[e-mail]>
> Cc: Liviu Dudau <[e-mail]>
> Cc: Alexey Brodkin <[e-mail]>
> Acked-by: Dave Airlie <[e-mail]>
> Signed-off-by: Daniel Vetter <[e-mail]>
> ---
>  Documentation/gpu/drm-uapi.rst | 67 ++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 67 insertions(+)
> 
> diff --git a/Documentation/gpu/drm-uapi.rst b/Documentation/gpu/drm-uapi.rst
> index 94876938aef3..a7e3aa27167d 100644
> --- a/Documentation/gpu/drm-uapi.rst
> +++ b/Documentation/gpu/drm-uapi.rst
> @@ -36,6 +36,73 @@ Primary Nodes, DRM Master and Authentication
>  Open-Source Userspace Requirements
>  ==================================
>  
> +The DRM subsystem has stricter requirements on what the userspace side for new
> +uAPI needs to look like. This section here explains what exactly those
> +requirements are, and why they exist.
> +
> +The short summary is that any addition of DRM uAPI requires corresponding
> +open-sourced userspace patches, and those patches must be reviewed and ready for
> +merging into a suitable and canonical upstream project.
> +
> +GFX devices (both display and render/GPU side) are really complex bits of hardware,
> +with userspace and kernel by necessity having to work together really closely.
> +The interfaces, for rendering and modesetting must be extremely wide and
> +flexible, and therefore it is almost always impossible to precisely define them
> +for every possible corner case. This in turns makes it really practically
> +infeasible to differentiate between behaviour that's required by userspace, and
> +which must not be changed to avoid regressions, and behaviour which is only an
> +accidental artifact of the current implementation.
> +
> +Without access to the full source code of all userspace users that means it
> +becomes impossible to change the implementation details, since userspace could
> +depend upon the accidental behaviour of the current implementation in minute
> +details. And debugging such regressions without access to source code is pretty
> +much impossible. As a consequence this means:
> +
> +- The Linux kernel's "no regression" policy holds in practice only for
> +  open-source userspace of the DRM subsystem. DRM developers are perfectly fine
> +  if closed-source blob drivers in userspace use the same uAPI as the open
> +  drivers, but they must do so in the exact same way as the open drivers.
> +  Creative (ab)use of the interfaces will, and in the past routinely has, lead
> +  to breakage.
> +
> +- Any new userspace interface must have an open-source implementation as
> +  demonstration vehicle.
> +
> +The other reason for requiring open-source userspace is uAPI review. Since the
> +kernel and userspace parts of a GFX stack must work together so closely, code
> +review can only assess whether a new interface achieves its goals by looking at
> +both sides. Making sure that the interface indeed covers the use-case fully
> +leads to a few additional requirements:
> +
> +- The open-source userspace must not be a toy/test application, but the real
> +  thing. Specifically it needs to handle all the usual error and corner cases.
> +  These are often the places where new uAPI falls apart and hence essential to
> +  assess the fitness of a proposed interface.
> +
> +- The userspace side must be fully reviewed and tested to the standards of that
> +  userspace project. For e.g. mesa this means piglit testcases and review on the
> +  mailing list. This is again to ensure that the new interface actually gets the
> +  job done.
> +
> +- The userspace patches must be against the canonical upstream, not some vendor
> +  fork. This is to make sure that no one cheats on the review and testing
> +  requirements by doing a quick fork.
> +
> +- The kernel patch can only be merged after all the above requirements are met,
> +  but it **must** be merged **before** the userspace patches land. uAPI always flows
> +  from the kernel, doing things the other way round risks divergence of the uAPI
> +  definitions and header files.
> +
> +These are fairly steep requirements, but have grown out from years of shared
> +pain and experience with uAPI added hastily, and almost always regretted about
> +as fast. GFX devices change really fast, requiring a paradigm shift and entire
> +new set of uAPI interfaces every few years at least. Together with the Linux
> +kernel's guarantee to keep existing userspace running for 10+ years this is
> +already rather painful for the DRM subsystem, with multiple different uAPIs for
> +the same thing co-existing. If we'd add a few more complete mistakes into the
> +mix every year it would be entirely unmanagable.
> +
>  Render nodes
>  ============
>  
> -- 
> 2.8.1
>

Laurent Pinchart Jan. 1, 2017, 10:40 p.m. UTC | #2

Hi Daniel,

Thank you for the patch.

On Friday 19 Aug 2016 22:50:38 Daniel Vetter wrote:
> Everyone knows them, except all the new folks joining from the ARM
> side haven't lived through all the pain of the past years and are
> entirely surprised when I raise this. Definitely time to document
> this.
> 
> Last time this was a big discussion was about 6 years ago, when qcom
> tried to land a kernel driver without userspace. Dave Airlie made the
> rules really clear:
> 
> http://airlied.livejournal.com/73115.html
> 
> This write-up here is essentially what I've put into a presentation a
> while ago, which was also reviewed by Dave:
> 
> http://blog.ffwll.ch/2015/05/gfx-kernel-upstreaming-requirements.html
> 
> Cc: Dave Airlie <[e-mail]>
> Cc: Oded Gabbay <[e-mail]>
> Cc: Russell King <[e-mail]>
> Cc: Tomi Valkeinen <[e-mail]>
> Cc: Eric Anholt <[e-mail]>
> Cc: Thomas Hellstrom <[e-mail]>
> Cc: Sinclair Yeh <[e-mail]>
> Cc: Lucas Stach <[e-mail]>
> Cc: Benjamin Gaignard <[e-mail]>
> Cc: Mark Yao <[e-mail]>
> Cc: Laurent Pinchart <[e-mail]>
> Cc: Ben Skeggs <[e-mail]>
> Cc: Rob Clark <[e-mail]>
> Cc: CK Hu <[e-mail]>
> Cc: Xinliang Liu <[e-mail]>
> Cc: Philipp Zabel <[e-mail]>
> Cc: Stefan Agner <[e-mail]>
> Cc: Inki Dae <[e-mail]>
> Cc: Maxime Ripard  <[e-mail]>
> Cc: Boris Brezillon <[e-mail]>
> Cc: Jani Nikula <[e-mail]>
> Cc: Daniel Vetter <[e-mail]>
> Cc: Thierry Reding <[e-mail]>
> Cc: Christian König <[e-mail]>
> Cc: Alex Deucher <[e-mail]>
> Cc: Gerd Hoffmann <[e-mail]>
> Cc: Brian Starkey <[e-mail]>
> Cc: Liviu Dudau <[e-mail]>
> Cc: Alexey Brodkin <[e-mail]>
> Acked-by: Dave Airlie <[e-mail]>
> Signed-off-by: Daniel Vetter <[e-mail]>
> ---
>  Documentation/gpu/drm-uapi.rst | 67 +++++++++++++++++++++++++++++++++++++++
>  1 file changed, 67 insertions(+)
> 
> diff --git a/Documentation/gpu/drm-uapi.rst b/Documentation/gpu/drm-uapi.rst
> index 94876938aef3..a7e3aa27167d 100644
> --- a/Documentation/gpu/drm-uapi.rst
> +++ b/Documentation/gpu/drm-uapi.rst
> @@ -36,6 +36,73 @@ Primary Nodes, DRM Master and Authentication
>  Open-Source Userspace Requirements
>  ==================================
> 
> +The DRM subsystem has stricter requirements on what the userspace side for
> new +uAPI needs to look like. This section here explains what exactly those
> +requirements are, and why they exist.
> +
> +The short summary is that any addition of DRM uAPI requires corresponding
> +open-sourced userspace patches, and those patches must be reviewed and
> ready for +merging into a suitable and canonical upstream project.
> +
> +GFX devices (both display and render/GPU side) are really complex bits of
> hardware, +with userspace and kernel by necessity having to work together
> really closely. +The interfaces, for rendering and modesetting must be
> extremely wide and +flexible, and therefore it is almost always impossible
> to precisely define them +for every possible corner case. This in turns
> makes it really practically +infeasible to differentiate between behaviour
> that's required by userspace, and +which must not be changed to avoid
> regressions, and behaviour which is only an +accidental artifact of the
> current implementation.

While I agree that this is a proper description of the current state of the 
DRM/KMS subsystem, I don't like how it implies that shipping code is an 
acceptable replacement for documentation. We need to aim at documenting the 
DRM/KMS userspace API to a level of detail that will cover the vast majority 
of cases (be they corner or round), if not all of them.

> +Without access to the full source code of all userspace users that means it
> +becomes impossible to change the implementation details, since userspace
> could +depend upon the accidental behaviour of the current implementation
> in minute +details. And debugging such regressions without access to source
> code is pretty +much impossible. As a consequence this means:
> +
> +- The Linux kernel's "no regression" policy holds in practice only for
> +  open-source userspace of the DRM subsystem. DRM developers are perfectly
> fine +  if closed-source blob drivers in userspace use the same uAPI as the
> open +  drivers,

In which category do you put the open-source userspace code that is not part 
of the mainline version of a major userspace framework (X11, Weston, Android 
hwcomposer, ...) ? I'm thinking about open-source vendor forks of those 
projects, or home-brew implementation of a tailor-made display stack for a 
product (I'm intentionally using the word display instead of GFX here, if a 
full tailor-made GFX stack including 3D rendering is quite unlikely, a display 
stack or even just an application driving the display through the DRM/KMS API 
is much easier to write from scratch).

> but they must do so in the exact same way as the open drivers.
> +  Creative (ab)use of the interfaces will, and in the past routinely has,
> lead +  to breakage.

This in my opinion calls for documentation, otherwise how can userspace 
developers tell what is an abuse without copying open-source code verbatim ?

> +- Any new userspace interface must have an open-source implementation as
> +  demonstration vehicle.
> +
> +The other reason for requiring open-source userspace is uAPI review. Since
> the +kernel and userspace parts of a GFX stack must work together so
> closely, code +review can only assess whether a new interface achieves its
> goals by looking at +both sides. Making sure that the interface indeed
> covers the use-case fully +leads to a few additional requirements:
> +
> +- The open-source userspace must not be a toy/test application, but the
> real +  thing. Specifically it needs to handle all the usual error and
> corner cases. +  These are often the places where new uAPI falls apart and
> hence essential to +  assess the fitness of a proposed interface.
> +
> +- The userspace side must be fully reviewed and tested to the standards of
> that +  userspace project. For e.g. mesa this means piglit testcases and
> review on the +  mailing list. This is again to ensure that the new
> interface actually gets the +  job done.
> +
> +- The userspace patches must be against the canonical upstream, not some
> vendor +  fork. This is to make sure that no one cheats on the review and
> testing +  requirements by doing a quick fork.
> +
> +- The kernel patch can only be merged after all the above requirements are
> met, +  but it **must** be merged **before** the userspace patches land.
> uAPI always flows +  from the kernel, doing things the other way round
> risks divergence of the uAPI +  definitions and header files.
> +
> +These are fairly steep requirements, but have grown out from years of
> shared +pain and experience with uAPI added hastily, and almost always
> regretted about +as fast. GFX devices change really fast, requiring a
> paradigm shift and entire +new set of uAPI interfaces every few years at
> least. Together with the Linux +kernel's guarantee to keep existing
> userspace running for 10+ years this is +already rather painful for the DRM
> subsystem, with multiple different uAPIs for +the same thing co-existing.
> If we'd add a few more complete mistakes into the +mix every year it would
> be entirely unmanagable.
> +
>  Render nodes
>  ============

Daniel Vetter Jan. 2, 2017, 8:09 a.m. UTC | #3

On Sun, Jan 1, 2017 at 11:40 PM, Laurent Pinchart
<[e-mail]> wrote:
> Hi Daniel,
>
> Thank you for the patch.
>
> On Friday 19 Aug 2016 22:50:38 Daniel Vetter wrote:
>> Everyone knows them, except all the new folks joining from the ARM
>> side haven't lived through all the pain of the past years and are
>> entirely surprised when I raise this. Definitely time to document
>> this.
>>
>> Last time this was a big discussion was about 6 years ago, when qcom
>> tried to land a kernel driver without userspace. Dave Airlie made the
>> rules really clear:
>>
>> http://airlied.livejournal.com/73115.html
>>
>> This write-up here is essentially what I've put into a presentation a
>> while ago, which was also reviewed by Dave:
>>
>> http://blog.ffwll.ch/2015/05/gfx-kernel-upstreaming-requirements.html
>>
>> Cc: Dave Airlie <[e-mail]>
>> Cc: Oded Gabbay <[e-mail]>
>> Cc: Russell King <[e-mail]>
>> Cc: Tomi Valkeinen <[e-mail]>
>> Cc: Eric Anholt <[e-mail]>
>> Cc: Thomas Hellstrom <[e-mail]>
>> Cc: Sinclair Yeh <[e-mail]>
>> Cc: Lucas Stach <[e-mail]>
>> Cc: Benjamin Gaignard <[e-mail]>
>> Cc: Mark Yao <[e-mail]>
>> Cc: Laurent Pinchart <[e-mail]>
>> Cc: Ben Skeggs <[e-mail]>
>> Cc: Rob Clark <[e-mail]>
>> Cc: CK Hu <[e-mail]>
>> Cc: Xinliang Liu <[e-mail]>
>> Cc: Philipp Zabel <[e-mail]>
>> Cc: Stefan Agner <[e-mail]>
>> Cc: Inki Dae <[e-mail]>
>> Cc: Maxime Ripard  <[e-mail]>
>> Cc: Boris Brezillon <[e-mail]>
>> Cc: Jani Nikula <[e-mail]>
>> Cc: Daniel Vetter <[e-mail]>
>> Cc: Thierry Reding <[e-mail]>
>> Cc: Christian König <[e-mail]>
>> Cc: Alex Deucher <[e-mail]>
>> Cc: Gerd Hoffmann <[e-mail]>
>> Cc: Brian Starkey <[e-mail]>
>> Cc: Liviu Dudau <[e-mail]>
>> Cc: Alexey Brodkin <[e-mail]>
>> Acked-by: Dave Airlie <[e-mail]>
>> Signed-off-by: Daniel Vetter <[e-mail]>
>> ---
>>  Documentation/gpu/drm-uapi.rst | 67 +++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 67 insertions(+)
>>
>> diff --git a/Documentation/gpu/drm-uapi.rst b/Documentation/gpu/drm-uapi.rst
>> index 94876938aef3..a7e3aa27167d 100644
>> --- a/Documentation/gpu/drm-uapi.rst
>> +++ b/Documentation/gpu/drm-uapi.rst
>> @@ -36,6 +36,73 @@ Primary Nodes, DRM Master and Authentication
>>  Open-Source Userspace Requirements
>>  ==================================
>>
>> +The DRM subsystem has stricter requirements on what the userspace side for
>> new +uAPI needs to look like. This section here explains what exactly those
>> +requirements are, and why they exist.
>> +
>> +The short summary is that any addition of DRM uAPI requires corresponding
>> +open-sourced userspace patches, and those patches must be reviewed and
>> ready for +merging into a suitable and canonical upstream project.
>> +
>> +GFX devices (both display and render/GPU side) are really complex bits of
>> hardware, +with userspace and kernel by necessity having to work together
>> really closely. +The interfaces, for rendering and modesetting must be
>> extremely wide and +flexible, and therefore it is almost always impossible
>> to precisely define them +for every possible corner case. This in turns
>> makes it really practically +infeasible to differentiate between behaviour
>> that's required by userspace, and +which must not be changed to avoid
>> regressions, and behaviour which is only an +accidental artifact of the
>> current implementation.
>
> While I agree that this is a proper description of the current state of the
> DRM/KMS subsystem, I don't like how it implies that shipping code is an
> acceptable replacement for documentation. We need to aim at documenting the
> DRM/KMS userspace API to a level of detail that will cover the vast majority
> of cases (be they corner or round), if not all of them.

I agree that this is a good idea, the trouble is in convincing
everyone else, and more important, that this is important enough that
they should spent time typing these docs. I think the only way to do
that is to 1) type those docs for existing ioctls yourself, and then
start and RFC to make it a standard. For properties I'm already
starting with at least some not-too-formal docs within the existing
kernel-doc stuff. Not there yet either. I somewhat plan to get around
to ioctl in the next few years, but would be real awesome if someone
beats me to it ;-)

>> +Without access to the full source code of all userspace users that means it
>> +becomes impossible to change the implementation details, since userspace
>> could +depend upon the accidental behaviour of the current implementation
>> in minute +details. And debugging such regressions without access to source
>> code is pretty +much impossible. As a consequence this means:
>> +
>> +- The Linux kernel's "no regression" policy holds in practice only for
>> +  open-source userspace of the DRM subsystem. DRM developers are perfectly
>> fine +  if closed-source blob drivers in userspace use the same uAPI as the
>> open +  drivers,
>
> In which category do you put the open-source userspace code that is not part
> of the mainline version of a major userspace framework (X11, Weston, Android
> hwcomposer, ...) ? I'm thinking about open-source vendor forks of those
> projects, or home-brew implementation of a tailor-made display stack for a
> product (I'm intentionally using the word display instead of GFX here, if a
> full tailor-made GFX stack including 3D rendering is quite unlikely, a display
> stack or even just an application driving the display through the DRM/KMS API
> is much easier to write from scratch).

I think if it's just a vendor fork of an existing open-source project,
then that existing open-source project should be the target. The idea
behind requiring review&ready for acceptance by the userspace side is
to make sure it'll be useful for more than just 1 vendor, just
applying it to a vendor fork somewhat defeats that idea. Best option
would be to either reintegrate the vendor for, or if that's not
possible, have some shared library or other means of sharing code.
Afaiui that's the entire point of libweston, so assuming your vendor
fork is tracking upstream libweston (it really should) this shouldn't
be a problem. If there's no direct code sharing, but at least common
history, porting the patches to upstream as demonstration vehicle also
shouldn't be that bad a burden.

An entirely different case is 2 independent codebases that solve the
exact same problem. We have no precedence for that, but I expect we'll
have that real soon with the community radv vulkan driver vs. the one
from amd (that should be released as open source any minute now).
Personally I'm leaning towards "why don't you just collaborate" and
trying to look cute&cuddly, but no sure where that one will fall tbh.
We will see soon.

>> but they must do so in the exact same way as the open drivers.
>> +  Creative (ab)use of the interfaces will, and in the past routinely has,
>> lead +  to breakage.
>
> This in my opinion calls for documentation, otherwise how can userspace
> developers tell what is an abuse without copying open-source code verbatim ?

If there's a regression, and code disagrees with the documentation,
the code wins. I agree that good docs would help, but the reality is
that very often you get to review 10+ years of git history in a bunch
of projects. I & others have done this plenty of times, and we
generally get things mostly right. But again not having tons of forks
and different implementations helps a lot with this, so "pls
collaborate" is a good idea (for the community overall, and that's the
point here) too. We have a bunch of forks and independent
reimplemenations (e.g. the gallium intel driver), but most often those
forks tend to lack in features so much that they're not hitting any of
the uapi corner cases that might break. I'll be interesting when this
happens for the first time for real, but personally I really hope this
doesn't happen, and that everyone understands the long-term benefits
of collaborating.

Cheers, Daniel

Patch

diff --git a/Documentation/gpu/drm-uapi.rst b/Documentation/gpu/drm-uapi.rst
index 94876938aef3..a7e3aa27167d 100644
--- a/Documentation/gpu/drm-uapi.rst
+++ b/Documentation/gpu/drm-uapi.rst
@@ -36,6 +36,73 @@  Primary Nodes, DRM Master and Authentication
 Open-Source Userspace Requirements
 ==================================
 
+The DRM subsystem has stricter requirements on what the userspace side for new
+uAPI needs to look like. This section here explains what exactly those
+requirements are, and why they exist.
+
+The short summary is that any addition of DRM uAPI requires corresponding
+open-sourced userspace patches, and those patches must be reviewed and ready for
+merging into a suitable and canonical upstream project.
+
+GFX devices (both display and render/GPU side) are really complex bits of hardware,
+with userspace and kernel by necessity having to work together really closely.
+The interfaces, for rendering and modesetting must be extremely wide and
+flexible, and therefore it is almost always impossible to precisely define them
+for every possible corner case. This in turns makes it really practically
+infeasible to differentiate between behaviour that's required by userspace, and
+which must not be changed to avoid regressions, and behaviour which is only an
+accidental artifact of the current implementation.
+
+Without access to the full source code of all userspace users that means it
+becomes impossible to change the implementation details, since userspace could
+depend upon the accidental behaviour of the current implementation in minute
+details. And debugging such regressions without access to source code is pretty
+much impossible. As a consequence this means:
+
+- The Linux kernel's "no regression" policy holds in practice only for
+  open-source userspace of the DRM subsystem. DRM developers are perfectly fine
+  if closed-source blob drivers in userspace use the same uAPI as the open
+  drivers, but they must do so in the exact same way as the open drivers.
+  Creative (ab)use of the interfaces will, and in the past routinely has, lead
+  to breakage.
+
+- Any new userspace interface must have an open-source implementation as
+  demonstration vehicle.
+
+The other reason for requiring open-source userspace is uAPI review. Since the
+kernel and userspace parts of a GFX stack must work together so closely, code
+review can only assess whether a new interface achieves its goals by looking at
+both sides. Making sure that the interface indeed covers the use-case fully
+leads to a few additional requirements:
+
+- The open-source userspace must not be a toy/test application, but the real
+  thing. Specifically it needs to handle all the usual error and corner cases.
+  These are often the places where new uAPI falls apart and hence essential to
+  assess the fitness of a proposed interface.
+
+- The userspace side must be fully reviewed and tested to the standards of that
+  userspace project. For e.g. mesa this means piglit testcases and review on the
+  mailing list. This is again to ensure that the new interface actually gets the
+  job done.
+
+- The userspace patches must be against the canonical upstream, not some vendor
+  fork. This is to make sure that no one cheats on the review and testing
+  requirements by doing a quick fork.
+
+- The kernel patch can only be merged after all the above requirements are met,
+  but it **must** be merged **before** the userspace patches land. uAPI always flows
+  from the kernel, doing things the other way round risks divergence of the uAPI
+  definitions and header files.
+
+These are fairly steep requirements, but have grown out from years of shared
+pain and experience with uAPI added hastily, and almost always regretted about
+as fast. GFX devices change really fast, requiring a paradigm shift and entire
+new set of uAPI interfaces every few years at least. Together with the Linux
+kernel's guarantee to keep existing userspace running for 10+ years this is
+already rather painful for the DRM subsystem, with multiple different uAPIs for
+the same thing co-existing. If we'd add a few more complete mistakes into the
+mix every year it would be entirely unmanagable.
+
 Render nodes
 ============