Rpi 4.4.y dsi stub squash #1556

anholt · 2016-07-07T19:08:22Z

Here's the stripped down series I've come up with for adding DSI support, based on not having ignore_lcd in the config.txt. Since it's requiring that the firmware set up DSI, there's no power management and instead of turning off the DSI and PV during DPMS, we just set the HVS to scan out black.

There's no touch input support yet. I think that's the next step, even before fixing the DSI link management.

clivem · 2016-07-07T19:33:07Z

arch/arm/boot/dts/bcm2708_common.dtsi

@@ -110,9 +111,10 @@
 			reg = <0x7e101000 0x2000>;

 			/* CPRMAN derives everything from the platform's
-			 * oscillator.
+			 * oscillator except for a few clocks that may
+                        * derive from something else derived from CPRMAN..


Cut and paste??????

anholt · 2016-07-07T20:28:38Z

It looks like without disable_touchscreen set in config.txt, the rpi-ft5406 driver keeps working just fine.

anholt · 2016-07-08T18:48:48Z

(force push was to drop the stray "+" in the comment)

chbae · 2016-07-27T15:53:57Z

@anholt When will you merge this change?

anholt · 2016-07-27T18:03:54Z

@chbae I'm the one requesting the merge, it's up to the foundation folks to pull it.

clivem · 2016-07-27T20:37:48Z

@anholt Eric, is there no way of making this fly with CONFIG_DRM=m and CONFIG_DRM_PANEL_RASPBERRYPI_TOUCHSCREEN=m. I haven't checked, but I'm thinking that having DRM compiled-in rather than module, is going to increase the size of the default kernel substantially, when very few people are actually using the functionality...... Just thinking out loud......

popcornmix · 2016-08-08T17:08:07Z

Looks like the =y option on CONFIG_DRM costs us 224K is available memory and 165K in size of kernel7.img.
@anholt - is that something that is essential?
It seems expensive for the users who don't use it.

pi@domnfs:~ $ ls -l /boot/kernel7.img 
-rwxr-xr-x 1 root root 4225384 Aug  8 18:01 /boot/kernel7.img
pi@domnfs:~ $ free -k
             total       used       free     shared    buffers     cached
Mem:        817828      91668     726160       5932         24      50740
-/+ buffers/cache:      40904     776924
Swap:       524284          0     524284

pi@domnfs:~ $ ls -l /boot/kernel7.img 
-rwxr-xr-x 1 root root 4390520 Aug  8 18:05 /boot/kernel7.img
pi@domnfs:~ $ free -k
             total       used       free     shared    buffers     cached
Mem:        817456      91544     725912       5932         24      50764
-/+ buffers/cache:      40756     776700
Swap:            0          0          0

anholt · 2016-08-08T21:58:11Z

I don't think counting on-disk size is appropriate, as you're just moving it from an initrd module to built-in. Memory size in non-vc4 mode is unfortunate, though if we can flip to vc4-by-defaut we should be able to get way more back by reducing gpu_mem.

For getting back to building as a module, hopefully with this series http://www.spinics.net/lists/dri-devel/msg114234.html it will be an option.

anholt · 2016-08-08T22:23:39Z

Note, if you're not up for merging the whole branch, it would still help a lot if you could pull up to before the defconfig change.

popcornmix · 2016-08-09T17:33:07Z

@pelwell any objections?

pelwell · 2016-08-09T18:30:04Z

I'm skeptical about the need for CONFIG_DEBUG_GPIO, which appears to just enable some dmesg KERN_DEBUG output. AFAICT the /sys/kernel/debug/pinctrl/ entries are enabled by CONFIG_DEBUG_FS, which is already defined. I am, of course, open to correction.

While the SDRAM is being driven by its dedicated PLL most of the time, there is a little loop running in the firmware that periodically turns on the CM SDRAM clock (using its pre-initialized parent) and switches SDRAM to using the CM clock to do PVT recalibration. This avoids system hangs if we choose SDRAM's parent for some other clock, then disable that clock. Signed-off-by: Eric Anholt <[email protected]>

Our core PLLs are intended to be configured once and left alone. With the flag set, asking to set the PLLD_DSI1 clock rate would change PLLD just to get closer to the requested DSI clock, thus changing PLLD_PER, the UART and ethernet PHY clock rates downstream of it, and breaking ethernet. Signed-off-by: Eric Anholt <[email protected]>

Signed-off-by: Eric Anholt <[email protected]>

The drm_fbdev_cma_init function always calls the drm_helper_disable_unused_functions. Since it's part of the usual probe process, all the drivers using that helper will end up having their encoder and CRTC disable functions called at probe if their device has not been reported as enabled. This could be fixed by reading out from the registers the current state of the device if it is enabled, but even that will not handle the case where the device is actually disabled. Moreover, the drivers using the atomic modesetting expect that their enable and disable callback to be called when the device is already enabled or disabled (respectively). We can however fix this issue by moving the call to drm_helper_disable_unused_functions out of drm_fbdev_cma_init and make the drivers needing it (all the drivers calling drm_fbdev_cma_init and not using the atomic modesetting) explicitly call it. Signed-off-by: Maxime Ripard <[email protected]> Link: http://patchwork.freedesktop.org/patch/msgid/1452785109-6172-14-git-send-email-maxime.ripard@free-electrons.com Acked-by: Laurent Pinchart <[email protected]> Signed-off-by: Daniel Vetter <[email protected]> (cherry picked from commit 4314e19)

anholt · 2016-08-09T20:55:58Z

Yeah, I must have screwed something up to write that commit. I'll drop it and retest, or you can just drop it.

I'm working on a build against the current rpi-4.4.y right now, and if it works I'll re-push.

pelwell · 2016-08-09T21:04:14Z

drivers/pinctrl/bcm/pinctrl-bcm2835.c

@@ -307,7 +307,7 @@ static inline void bcm2835_pinctrl_fsel_set(
 	u32 val = bcm2835_gpio_rd(pc, FSEL_REG(pin));
 	enum bcm2835_fsel cur = (val >> FSEL_SHIFT(pin)) & BCM2835_FSEL_MASK;

-	dev_dbg(pc->dev, "read %08x (%u => %s)\n", val, pin,
+	dev_info(pc->dev, "read %08x (%u => %s)\n", val, pin,


Hmm - I don't think we want these.

pelwell · 2016-08-09T21:15:26Z

Cool - it's nice to have my sanity confirmed occasionally,

Aside from that and the few inline comments, I have no issues with the patch., but note that I've concentrated on the interactions with other areas rather than reading all of your code.

anholt · 2016-08-09T21:22:43Z

Thanks for the cleanups @pelwell. Pushed with those two fixed and the gpio debug config dropped.

anholt · 2016-08-09T21:26:31Z

Also, anyone have any ideas why the panel built as a module wouldn't automatically load? That's the prereq for dropping CONFIG_DRM_MIPI_DSI to a module, and thus CONFIG_DRM back to module.

pelwell · 2016-08-09T21:48:10Z

Off the top of my head (which is going to bed now), is it because the parent of the touchscreen node doesn't cause subnodes to be treated as potential platform devices, perhaps because it isn't a bus?

This is a preliminary patch for building drm-mipi-dsi as a module. Add the module exit callback to unregister the bus properly. Suggested-by: Thierry Reding <[email protected]> Signed-off-by: Takashi Iwai <[email protected]> Signed-off-by: Eric Anholt <[email protected]>

The drm-mipi-dsi driver has been only built-in although this isn't strictly required to be so. Since it's referred by lots of DRM drivers nowadays, most of distro kernels include the driver as built-in as a result, even though many systems don't need it at all. This patch fixes Kconfig to allow drm-mipi-dsi driver built as a module, so that we can save footprint on systems without such DRM drivers. The probe order is managed by the module dependency, and postcore_initcall() works just fine as a module init call. Signed-off-by: Takashi Iwai <[email protected]> Signed-off-by: Eric Anholt <[email protected]>

Empirically, DSI1 appears to be using HVS channel 0 on my Pi2. Signed-off-by: Eric Anholt <[email protected]>

Signed-off-by: Eric Anholt <[email protected]>

Every file was defining its own little struct and dumping for the regs, when there's a helper in debugfs for doing just this. However, instead of printing: PV_HORZA (0x000c): 0x00000000 we now print: PV_HORZA = 0x00000000 Signed-off-by: Eric Anholt <[email protected]>

anholt · 2016-08-09T21:59:23Z

Yeah, they're not instantiated as platform devices, they're on their own bus. I may have a fix.

The DSI0 and DSI1 blocks on the 2835 are different but very similar hardware blocks. Some registers move around, and the featureset is slightly different, but they're clearly related. This doesn't enable DSI0, but some of the infrastructure is present. Also, this driver doesn't initialize the DSI successfully from poweron, so we currently require that the Raspberry Pi firmware enable it at boot time. From there, we just keep the same settings forever, and when poweroff is requested we just scan out black instead. Signed-off-by: Eric Anholt <[email protected]>

Signed-off-by: Eric Anholt <[email protected]>

In order to make DSI panel modules able to load automatically, we need to emit the OF modalias uevent so that their MODULE_DEVICE_TABLE(of, ...) can match against it. Signed-off-by: Eric Anholt <[email protected]>

This gets the Raspberry Pi panel module to automatically load at boot time. Signed-off-by: Eric Anholt <[email protected]>

anholt · 2016-08-09T22:28:15Z

And fixed a stray debug printf in vc4_dsi.c.

See: raspberrypi/linux#1581 kernel: enable gembird joypad support See: raspberrypi/linux#1589 kernel: Added HiFiBerry Digi+ Pro driver See: raspberrypi/linux#1583 kernel: overlays: Add assert_falling_edge to pps-gpio overlay See: raspberrypi/linux#1590 kernel: Experimental graphics driver DSI support See: raspberrypi/linux#1556 firmware: Add customisation project board type firmware: platform: Enable Bluetooth on custom firmware: gencmd: Add support for hdmi_cvt and hdmi_timings See: #637 firmware: arm_display: Allow source aspect ratio to be configured See: #638 firmware: mmal: Allow /dev/vchiq to be opened initialised prior to mmal_vc_init

See: raspberrypi/linux#1581 kernel: enable gembird joypad support See: raspberrypi/linux#1589 kernel: Added HiFiBerry Digi+ Pro driver See: raspberrypi/linux#1583 kernel: overlays: Add assert_falling_edge to pps-gpio overlay See: raspberrypi/linux#1590 kernel: Experimental graphics driver DSI support See: raspberrypi/linux#1556 firmware: Add customisation project board type firmware: platform: Enable Bluetooth on custom firmware: gencmd: Add support for hdmi_cvt and hdmi_timings See: raspberrypi/firmware#637 firmware: arm_display: Allow source aspect ratio to be configured See: raspberrypi/firmware#638 firmware: mmal: Allow /dev/vchiq to be opened initialised prior to mmal_vc_init

- kernel: fix auto-sense in lirc_rpi driver See: raspberrypi/linux#1581 - kernel: enable gembird joypad support See: raspberrypi/linux#1589 - kernel: Added HiFiBerry Digi+ Pro driver See: raspberrypi/linux#1583 - kernel: overlays: Add assert_falling_edge to pps-gpio overlay See: raspberrypi/linux#1590 - kernel: Experimental graphics driver DSI support See: raspberrypi/linux#1556 - firmware: Add customisation project board type - firmware: platform: Enable Bluetooth on custom - firmware: gencmd: Add support for hdmi_cvt and hdmi_timings See: #637 - firmware: arm_display: Allow source aspect ratio to be configured See: #638 - firmware: mmal: Allow /dev/vchiq to be opened initialised prior to mmal_vc_init

chbae · 2016-08-16T04:35:13Z

@anholt
Thanks for this work.
It works fine on GENIVI platform. (https://github.com/genivi/genivi-dev-platform).
I tested on my raspberrypi2 + Official 7'' touch screen. After adding rpi-ft5406-overlay, touch is also working fine.

tom--pollard · 2016-08-17T10:20:17Z

Can confirm this is working for me also, thanks for the great work @anholt !

See: raspberrypi/linux#1581 kernel: enable gembird joypad support See: raspberrypi/linux#1589 kernel: Added HiFiBerry Digi+ Pro driver See: raspberrypi/linux#1583 kernel: overlays: Add assert_falling_edge to pps-gpio overlay See: raspberrypi/linux#1590 kernel: Experimental graphics driver DSI support See: raspberrypi/linux#1556 firmware: Add customisation project board type firmware: platform: Enable Bluetooth on custom firmware: gencmd: Add support for hdmi_cvt and hdmi_timings See: raspberrypi#637 firmware: arm_display: Allow source aspect ratio to be configured See: raspberrypi#638 firmware: mmal: Allow /dev/vchiq to be opened initialised prior to mmal_vc_init

clivem reviewed Jul 7, 2016
View reviewed changes

anholt force-pushed the rpi-4.4.y-dsi-stub-squash branch from b988b5e to 6e70121 Compare July 8, 2016 18:02

popcornmix mentioned this pull request Aug 8, 2016

7" touch screen with drm (vc4) #1478

Closed

anholt and others added 5 commits August 9, 2016 13:31

clk: bcm2835: Do appropriate name lookups for DSI1's parents as well.

bd7c4fa

Signed-off-by: Eric Anholt <[email protected]>

clk: bcm2835: Add an enum for the DSI1 pixel clock.

12262a8

Signed-off-by: Eric Anholt <[email protected]>

pelwell reviewed Aug 9, 2016
View reviewed changes

anholt force-pushed the rpi-4.4.y-dsi-stub-squash branch from 6e70121 to 8720647 Compare August 9, 2016 21:22

tiwai and others added 4 commits August 9, 2016 14:49

drm/vc4: Fix flipped HVS channels for DSI0/1.

804467c

Empirically, DSI1 appears to be using HVS channel 0 on my Pi2. Signed-off-by: Eric Anholt <[email protected]>

drm/vc4: Add support for feeding DSI encoders from the pixel valve.

b5f1d8f

Signed-off-by: Eric Anholt <[email protected]>

anholt force-pushed the rpi-4.4.y-dsi-stub-squash branch from 8720647 to 1a04975 Compare August 9, 2016 22:20

anholt added 6 commits August 9, 2016 15:23

drm/panel: Add support for the Raspberry Pi 7" Touchscreen.

07ca869

Signed-off-by: Eric Anholt <[email protected]>

BCM270X: Add the DSI panel to the defconfig.

ba09079

Signed-off-by: Eric Anholt <[email protected]>

BCM270X: Add the DSI and DSI panel nodes to the VC4 overlay.

92b7197

Signed-off-by: Eric Anholt <[email protected]>

of: Export of_device_uevent_modalias for DRM_MIPI_DSI.

4c4d16a

In order to make DSI panel modules able to load automatically, we need to emit the OF modalias uevent so that their MODULE_DEVICE_TABLE(of, ...) can match against it. Signed-off-by: Eric Anholt <[email protected]>

drm: Emit modalias uevents for the DSI devices we create.

8018f92

This gets the Raspberry Pi panel module to automatically load at boot time. Signed-off-by: Eric Anholt <[email protected]>

anholt force-pushed the rpi-4.4.y-dsi-stub-squash branch from 1a04975 to 8018f92 Compare August 9, 2016 22:28

popcornmix merged commit 26550dc into raspberrypi:rpi-4.4.y Aug 10, 2016

anholt deleted the rpi-4.4.y-dsi-stub-squash branch September 29, 2016 00:48

Rpi 4.4.y dsi stub squash #1556

Rpi 4.4.y dsi stub squash #1556

Uh oh!

Conversation

anholt commented Jul 7, 2016

Uh oh!

clivem Jul 7, 2016

Choose a reason for hiding this comment

Uh oh!

anholt commented Jul 7, 2016

Uh oh!

anholt commented Jul 8, 2016

Uh oh!

chbae commented Jul 27, 2016

Uh oh!

anholt commented Jul 27, 2016

Uh oh!

clivem commented Jul 27, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

popcornmix commented Aug 8, 2016

Uh oh!

anholt commented Aug 8, 2016

Uh oh!

anholt commented Aug 8, 2016

Uh oh!

popcornmix commented Aug 9, 2016

Uh oh!

pelwell commented Aug 9, 2016

Uh oh!

anholt commented Aug 9, 2016

Uh oh!

pelwell Aug 9, 2016

Choose a reason for hiding this comment

Uh oh!

pelwell commented Aug 9, 2016

Uh oh!

anholt commented Aug 9, 2016

Uh oh!

anholt commented Aug 9, 2016

Uh oh!

pelwell commented Aug 9, 2016

Uh oh!

anholt commented Aug 9, 2016 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

anholt commented Aug 9, 2016

Uh oh!

chbae commented Aug 16, 2016

Uh oh!

tom--pollard commented Aug 17, 2016

Uh oh!

Uh oh!

clivem commented Jul 27, 2016 •

edited

Loading

anholt commented Aug 9, 2016 •

edited

Loading