Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take
crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Now I'm at the point of trying to work out if I can make DP MST monitors a possibility before we get atomic,
Myself and Ben discussed this here and he suggested we should make the userspace crtc ids pretty much meaningless and not have them tied to actual hw crtcs, so we can reroute things underneath userspace without changing it.
Any input is welcome!
Dave.
Dave Airlie airlied@gmail.com writes:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
Mostly because X has almost always exposed the hardware at as low a level as possible and left clients to sort things out. Given that we had no experience with this whole structure before RandR got implemented, and that we've been running like this for eight years without terrible trouble, it doesn't seem like an utter failure...
For this particular issue, we've got two choices:
1) Describe the situation through the protocol and let applications sort it out
2) Hide physical CRTCs from applications and create virtual CRTCs for applications.
One reason for exposing physical CRTCs to applications is to let them figure out the full allocation plan before starting the process, so as to minimize screen flicker given an API which doesn't let you specify the whole configuration in one go.
If we hide them, then the kernel may need to shut down monitors while it shuffles things around to match application requests.
From my POV I don't think the modesetting interface needs to take crtcs, just connectors and modes,
I'm fine with making the X server need to be smarter about kernel CRTC allocation, pushing the problem out of the kernel and up into the window system. That seems like the simplest change in the kernel API to me.
Making X hide real CRTCs from clients seems like a fairly simple plan; that would also offer us an opportunity to add 'virtual' CRTCs for use by VNC or other software-defined display surfaces.
so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Having a list of CRTCs means that the application would have a chance of figuring out some of the impossible configurations before asking the kernel; you couldn't light up two single-link monitors and a double-link monitor if you only had three CRTCs.
Now I'm at the point of trying to work out if I can make DP MST monitors a possibility before we get atomic,
I think fixing X to hide the physical CRTCs and only advertise virtual ones should be pretty easy to manage; that would leave the kernel API alone, at least for now.
Ben and I discussed this here and he suggested we should make the userspace crtc ids pretty much meaningless and not have them tied to actual hw crtcs, so we can reroute things underneath userspace without changing it.
It's clear that we need this kind of redirection at some level in the stack; what's unclear to me is whether this should be done in the kernel or up in userspace.
With atomic mode setting in the kernel, I think you're probably right in proposing to eliminate explicit CRTC allocation from that. I do think you'll want to indicate the number of available CRTCs in the display engine, and the number of CRTCs each monitor consumes. Do you know if there are some of these monitors that can display lower resolution modes with only a single CRTC? Or is the hardware so separate that you end up always using multiple CRTCs to drive them?
For the current incremental mode setting API, I think it'd work either way.
Pushing the problem out to user space is always tempting, and I don't think it would be hard to teach the X server to manage this. That would also eliminate the need to construct fake EDID data within the kernel; the X server could do whatever it liked in building suitable video mode lists given complete information about the monitor. Plus, I can see how we'd offer an atomic RandR request that could operate on top of the current API while minimizing flashing. Hiding CRTCs from the X server would make this difficult, as the kernel wouldn't have the full set of configuration information available without the atomic mode kernel API.
Solving this in the kernel would make the X piece simpler, although the kernel would now be constructing fake EDID data to advertise the combined set of modes up to X, and you'd end up with more flashing if the kernel allocated the 'wrong' CRTC to any of the displays and needed to disable/re-enable things to get a new configuration working.
Without RandR additions, the two solutions are effectively identical. Somewhere you have to guess which CRTCs to use during incremental mode setting, and sometimes you're just going to guess wrong and have to correct that later on.
I'd pick whichever was simpler to implement and expect this to all be resolved in the glorious atomic mode setting future we've been promised for so long.
On 09/08/2014 11:37 PM, Keith Packard wrote:
With atomic mode setting in the kernel, I think you're probably right in proposing to eliminate explicit CRTC allocation from that. I do think you'll want to indicate the number of available CRTCs in the display engine, and the number of CRTCs each monitor consumes. Do you know if there are some of these monitors that can display lower resolution modes with only a single CRTC? Or is the hardware so separate that you end up always using multiple CRTCs to drive them?
The one I tried definitely can; indeed it has to because the VBIOS doesn't know how to set up MST and drives the port in DP 1.1 mode. I'm sure someone will build a monitor that only turns on half of the display if you do that, but I'd be kind of surprised if someone made one that just doesn't work if you use a DP 1.1-only GPU.
On 9 September 2014 10:43, Dave Airlie airlied@gmail.com wrote:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Now I'm at the point of trying to work out if I can make DP MST monitors a possibility before we get atomic,
Myself and Ben discussed this here and he suggested we should make the userspace crtc ids pretty much meaningless and not have them tied to actual hw crtcs, so we can reroute things underneath userspace without changing it.
The only caveat we came up with is due to page_flip requiring indices we can't probably move things around as much as I'd like,
I'm not sure if we have same problems further up!
Dave.
On Tue, Sep 09, 2014 at 10:43:35AM +1000, Dave Airlie wrote:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take
crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Not all CRTCs are created equal so the user probably wants know what features to expect from a particular CRTC. Now, often that may have something to do with the planes, but there are other hardware features that we want to expose as CRTC properties. If we make all CRTCs appear uniform to userspace the user may not know beforehand that certain features can only be used on a subset of CRTCs. Also if the driver would initially pick the wrong CRTC, and later the user would enable one of those special features, we'd have to do a full modeset to switch hardware CRTCs which would mean a nasty screen blink for the user.
So no, I don't think this is a good idea given real world hardware constraints.
On Tue, Sep 9, 2014 at 4:16 AM, Ville Syrjälä ville.syrjala@linux.intel.com wrote:
On Tue, Sep 09, 2014 at 10:43:35AM +1000, Dave Airlie wrote:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take
crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Not all CRTCs are created equal so the user probably wants know what features to expect from a particular CRTC. Now, often that may have something to do with the planes, but there are other hardware features that we want to expose as CRTC properties. If we make all CRTCs appear uniform to userspace the user may not know beforehand that certain features can only be used on a subset of CRTCs. Also if the driver would initially pick the wrong CRTC, and later the user would enable one of those special features, we'd have to do a full modeset to switch hardware CRTCs which would mean a nasty screen blink for the user.
first off, I tend to think with the trend towards various different wayland compositors doing kms directly, making it easier for userspace sounds pretty attractive. Ie. would you rather fix a bug w/ picking the right crtc for the job in N compositors, or 1 kernel driver?
But that said, it seems like the real problem w/ kernel picking the right crtc is going to be with non-atomic modeset. And for pre-atomic (future legacy) xrandr, I'm not entirely sure how userspace is supposed to do a better job at this than the kernel could. It would also need up front knowledge of all the modes that would be picked. So you've just pushed non-atomic suck in the kernel to non-atomic suck in x11. Doesn't sound like that fixes anything.
But, I think there is maybe a way to have our cake and eat it too.. to leave extra flexibility for highly customized/specialized userspace, we could just allow for some PICK_ANY_CRTC_FOR_ME type value which would let 99% of userspace push the decision to the kernel, while still allowing for the special cases where userspace knows better.
BR, -R
So no, I don't think this is a good idea given real world hardware constraints.
-- Ville Syrjälä Intel OTC _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel
On Tue, Sep 09, 2014 at 07:22:49PM -0400, Rob Clark wrote:
On Tue, Sep 9, 2014 at 4:16 AM, Ville Syrjälä ville.syrjala@linux.intel.com wrote:
On Tue, Sep 09, 2014 at 10:43:35AM +1000, Dave Airlie wrote:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take
crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Not all CRTCs are created equal so the user probably wants know what features to expect from a particular CRTC. Now, often that may have something to do with the planes, but there are other hardware features that we want to expose as CRTC properties. If we make all CRTCs appear uniform to userspace the user may not know beforehand that certain features can only be used on a subset of CRTCs. Also if the driver would initially pick the wrong CRTC, and later the user would enable one of those special features, we'd have to do a full modeset to switch hardware CRTCs which would mean a nasty screen blink for the user.
first off, I tend to think with the trend towards various different wayland compositors doing kms directly, making it easier for userspace sounds pretty attractive. Ie. would you rather fix a bug w/ picking the right crtc for the job in N compositors, or 1 kernel driver?
But that said, it seems like the real problem w/ kernel picking the right crtc is going to be with non-atomic modeset. And for pre-atomic (future legacy) xrandr, I'm not entirely sure how userspace is supposed to do a better job at this than the kernel could. It would also need up front knowledge of all the modes that would be picked. So you've just pushed non-atomic suck in the kernel to non-atomic suck in x11. Doesn't sound like that fixes anything.
I disagree. User space can be in a much better position to pick the crtc, especially in more product oriented scenarios. Eg. a tablet/phone where user space knows that it should use the more capable crtc for the internal display and less capable one for the external display. The reason for this could be that it will be presenting video in a window on the internal display and fullscreen on the external display.
But, I think there is maybe a way to have our cake and eat it too.. to leave extra flexibility for highly customized/specialized userspace, we could just allow for some PICK_ANY_CRTC_FOR_ME type value which would let 99% of userspace push the decision to the kernel, while still allowing for the special cases where userspace knows better.
Then you get to figure out how to deal with all/some of the properties effectively changing values when the crtc changes. I guess you could just specifiy that this will happen and it's up to userspace to provide the full state if it wants to avoid unexpected property values.
On Wed, Sep 10, 2014 at 4:20 AM, Ville Syrjälä ville.syrjala@linux.intel.com wrote:
On Tue, Sep 09, 2014 at 07:22:49PM -0400, Rob Clark wrote:
On Tue, Sep 9, 2014 at 4:16 AM, Ville Syrjälä ville.syrjala@linux.intel.com wrote:
On Tue, Sep 09, 2014 at 10:43:35AM +1000, Dave Airlie wrote:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take
crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Not all CRTCs are created equal so the user probably wants know what features to expect from a particular CRTC. Now, often that may have something to do with the planes, but there are other hardware features that we want to expose as CRTC properties. If we make all CRTCs appear uniform to userspace the user may not know beforehand that certain features can only be used on a subset of CRTCs. Also if the driver would initially pick the wrong CRTC, and later the user would enable one of those special features, we'd have to do a full modeset to switch hardware CRTCs which would mean a nasty screen blink for the user.
first off, I tend to think with the trend towards various different wayland compositors doing kms directly, making it easier for userspace sounds pretty attractive. Ie. would you rather fix a bug w/ picking the right crtc for the job in N compositors, or 1 kernel driver?
But that said, it seems like the real problem w/ kernel picking the right crtc is going to be with non-atomic modeset. And for pre-atomic (future legacy) xrandr, I'm not entirely sure how userspace is supposed to do a better job at this than the kernel could. It would also need up front knowledge of all the modes that would be picked. So you've just pushed non-atomic suck in the kernel to non-atomic suck in x11. Doesn't sound like that fixes anything.
I disagree. User space can be in a much better position to pick the crtc, especially in more product oriented scenarios. Eg. a tablet/phone where user space knows that it should use the more capable crtc for the internal display and less capable one for the external display. The reason for this could be that it will be presenting video in a window on the internal display and fullscreen on the external display.
Sure, but this is a pretty special case from an upstream perspective. But giving userspace the choice of picking explicitly or letting kernel pick, I think we can accommodate both the weird special cases, as well as the common case.
Not quite sure how it would work from atomic ioctl standpoint, where userspace needs to set (potentially) multiple properties on several unchosen crtc's, but I think we could have userspace pick virtual crtc id's, ie.
#define VIRT_CRTC 0x40000000
then user uses (VIRT_CRTC | 1), (VIRT_CRTC | 2), and so on.. then kernel somehow resolves those into actual crtc id's..
But, I think there is maybe a way to have our cake and eat it too.. to leave extra flexibility for highly customized/specialized userspace, we could just allow for some PICK_ANY_CRTC_FOR_ME type value which would let 99% of userspace push the decision to the kernel, while still allowing for the special cases where userspace knows better.
Then you get to figure out how to deal with all/some of the properties effectively changing values when the crtc changes. I guess you could just specifiy that this will happen and it's up to userspace to provide the full state if it wants to avoid unexpected property values.
Well, we still need to define what the semantics are for unspecified property values, esp. if different ctrc's had different properties. I think "preserve" for pageflip and "reset to default" for modeset would be the sane choice, although it is a bit awkward with both crammed into the same ioctl. Probably we end up adding a flag to control the behaviour.
But yeah, playing switch-a-roo under the hood, we have to sort out those details. But at least it would happen only at modeset where userspace probably wants to use the hypothetical RESET_UNSPECIFIED_TO_DEFAULT flag..
BR, -R
-- Ville Syrjälä Intel OTC
Hi
On Tue, Sep 9, 2014 at 10:16 AM, Ville Syrjälä ville.syrjala@linux.intel.com wrote:
On Tue, Sep 09, 2014 at 10:43:35AM +1000, Dave Airlie wrote:
From my POV I don't think the modesetting interface needs to take crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Not all CRTCs are created equal so the user probably wants know what features to expect from a particular CRTC. Now, often that may have something to do with the planes, but there are other hardware features that we want to expose as CRTC properties. If we make all CRTCs appear uniform to userspace the user may not know beforehand that certain features can only be used on a subset of CRTCs. Also if the driver would initially pick the wrong CRTC, and later the user would enable one of those special features, we'd have to do a full modeset to switch hardware CRTCs which would mean a nasty screen blink for the user.
I agree with Ville here. For generic DRM usage, it's irrelevant which CRTC is used, but for special use-cases you want more capable CRTCs for the display you use for compositing. Sure, this might be solved with simple "hints" you pass to the kernel, but those hints pretty soon become what we now have as CRTCs. And re-configuration requires expensive deep modesets so cannot be done on-the-fly by the kernel once user-space allocates resources for a given output.
However, the way we pick CRTCs and encoders today is already problematic. If you write gpu-specific drivers, you know the hardware perfectly, but generic drivers already need to do trial-and-error. Especially, if you want "clone" mode, you have no control which encoder is picked by the kernel, so you don't know which one is left after the first mode-set, so you cannot know which CRTCs are available for the other connectors. No driver exposed such restricted and backwards interfaces so far, and the CRTC->encoder->connector graph is almost always a tree. But you can't be sure, if you write generic drivers.. so for all stuff I did so far, I already use "connectors" as primary objects and re-evaluate after each single modeset. So it wouldn't change much if CRTCs aren't exposed at all.
I also agree with Keith: Somewhere in the stack, there has to be a layer that adds "screens", which are possibly composed of multiple connectors. If we do this in user-space (like we all do so far), we can support setups where multiple real displays are composed into a single large screen *AND* setups, where one display is internally represented by multiple displays. If we do it in the kernel, we might possible require a second such layer in user-space..
Long story short: I'm totally fine with exposing split hardware as multiple connectors..
Thanks David
On Mon, Sep 8, 2014 at 8:43 PM, Dave Airlie airlied@gmail.com wrote:
Hi,
So I've been attempting to hide the 30" Dell MST monitors in the kernel, and ran into a number of problems, but the major one is how to steal a crtc and get away with it.
The standard scenario I have is
CRTC 0: eDP monitor connected
hotplug 30" monitor, userspace decides to configure things as
CRTC 1: DP-4 - 30" monitor CRTC 2: eDP-1
But since we lack atomic it does this in two steps, so when I get the first modeset to set the 30" monitor up I go and use CRTC-2 as the secondary crtc, as CRTC-0 is in use still, then I have to fail the second modeset, and things end up with me crying.
So this led me to wonder why we expose CRTCs at all, and KMS does it because randr did it, but I've no idea why randr did it (Keith??).
From my POV I don't think the modesetting interface needs to take crtcs, just connectors and modes, so I'm wondering going forward for atomic should we even accept crtcs in the interface, just a list of rectangles, connectors per rectangle, etc.
Now I'm at the point of trying to work out if I can make DP MST monitors a possibility before we get atomic,
Myself and Ben discussed this here and he suggested we should make the userspace crtc ids pretty much meaningless and not have them tied to actual hw crtcs, so we can reroute things underneath userspace without changing it.
Any input is welcome!
What about exposing monitors as a modesetting object? They could have a required_crtc_num attribute or something like that. I guess it's a little late since we already did dynamic connectors for MST and I guess it would be complicated to integrate cleanly with the way we do things today.
E.g., you could have:
crtc -> encoder -> connector -> monitor
or
crtc - -> monitor \ / crtc - --> encoder -> connector -> monitor / \ crtc - -> monitor
or
crtc - \ crtc - --> encoder -> connector -> monitor
etc.
Alex
dri-devel@lists.freedesktop.org
-
Aaron Plattner -
Alex Deucher -
Dave Airlie -
David Herrmann -
Keith Packard -
Rob Clark -
Ville Syrjälä