cog on imx8mini - failed to schedule a page flip: Permission denied

SebDE commented 2 years ago

I try to get cog on an imx8-mini platform up and running on a wayland / weston only system to act as a fullscreen html/js gui

I used the meta-webkit yocto layer: meta-webkit = "hardknott:56b0544f7d9610866ca67ee574b7bf01b7fded34" with the following build configuration:

Build Configuration:
BB_VERSION           = "1.50.0"
BUILD_SYS            = "x86_64-linux"
NATIVELSBSTRING      = "universal"
TARGET_SYS           = "aarch64-fslc-linux"
MACHINE              = "imx8mm-var-dart"
DISTRO               = "fslc-wayland"
DISTRO_VERSION       = "3.3"
TUNE_FEATURES        = "aarch64 armv8a crc cortexa53 crypto"
TARGET_FPU           = ""

and here is my local.conf

MACHINE ??= 'imx8mm-var-dart'
DISTRO ?= 'fslc-wayland'
DISTRO_FEATURES_append = " opengl wayland"
DISTRO_FEATURES_append = " egl"
PACKAGE_CLASSES ?= 'package_rpm'
IMAGE_INSTALL_append = " gstreamer1.0-plugins-imx gstreamer1.0-plugins-imx-meta"
IMAGE_INSTALL_append = " wpewebkit cog wpebackend-fdo"
PREFERRED_PROVIDER_virtual/wpebackend = "wpebackend-fdo"
EXTRA_IMAGE_FEATURES ?= "debug-tweaks ssh-server-openssh"
USER_CLASSES ?= "buildstats image-mklibs image-prelink"
PATCHRESOLVE = "noop"
BB_DISKMON_DIRS ??= "\
    STOPTASKS,${TMPDIR},1G,100K \
    STOPTASKS,${DL_DIR},1G,100K \
    STOPTASKS,${SSTATE_DIR},1G,100K \
    STOPTASKS,/tmp,100M,100K \
    ABORT,${TMPDIR},100M,1K \
    ABORT,${DL_DIR},100M,1K \
    ABORT,${SSTATE_DIR},100M,1K \
    ABORT,/tmp,10M,1K"
PACKAGECONFIG_append_pn-qemu-system-native = " sdl"
CONF_VERSION = "1"

DL_DIR ?= "/opt/yocto_downloads/"
ACCEPT_FSL_EULA = "1"
LICENSE_FLAGS_WHITELIST_append = " commercial"

But if I try to start cog I just got those messages and no website appears on the attached lcd

root@imx8mm-var-dart:~# cog http://neverssl.com

(cog:574): GLib-GIO-WARNING **: 20:14:20.180: Your application does not implement g_application_activate() and has no handlers connected to the 'activate' signal.  It should do one of these.

(cog:574): Cog-DRM-WARNING **: 20:14:20.415: failed to schedule a page flip: Permission denied
Cog-Core-Message: 20:14:20.415: <http://neverssl.com/> Load started.
Cog-Core-Message: 20:14:20.422: <http://neverssl.com/> Loading...
Cog-Core-Message: 20:14:20.435: <http://neverssl.com/> Loaded successfully.
Cog-Core-Message: 20:14:20.437: <http://neverssl.com/> Load started.
Cog-Core-Message: 20:14:20.646: <http://quietsilverwholelight.neverssl.com/online> Loading...
Cog-Core-Message: 20:14:20.732: <http://quietsilverwholelight.neverssl.com/online> Loaded successfully.

Is the following warning the problem? (cog:574): Cog-DRM-WARNING **: 20:14:20.415: failed to schedule a page flip: Permission denied

Here is the version info

root@imx8mm-var-dart:~# cog --version
0.12.0 (WPE WebKit 2.34.4)

Hope someone can point me in the right direction - many thanks !

aperezdc commented 2 years ago

@SebDE A couple of things to check:

Is there something else running that could be using the device (Weston, X11, etc.)? When running with --platform=drm, Cog will take care of handling the video output directly, a graphics server/compositor is not needed.
Could you set G_MESSAGES_DEBUG=Cog-DRM in the environment and run Cog again? The log could be useful to determine what the issue is.
Which GPU model and, more importantly, which driver is your system using?

Thanks in advance!

SebDE commented 2 years ago

@aperezdc Many thanks for your reply and the tipps:

There was already a weston process running - I was thinking that COG needs / uses a weston session, but now you made it clear for me. Now cog starts without the permission denied warning and renders something to the screen - but it looks very strange:
The imx8m-mini should have the following GPUs: Graphic Processing Unit: Vivante GCNanoUltra for 3D acceleration, Vivante GC320 for 2D acceleration

In case of the driver I'm not really sure: If I look at the loaded kernel modules I see the galcore module, which should be the driver for the vivante gpus


root@imx8mm-var-dart:~# lsmod
Module                  Size  Used by
rfcomm                 53248  12
bnep                   28672  2
hci_uart               40960  1
btbcm                  20480  1 hci_uart
brcmfmac              253952  0
cfg80211              393216  1 brcmfmac
brcmutil               16384  1 brcmfmac
hid_generic            16384  0
bluetooth             450560  41 hci_uart,btbcm,bnep,rfcomm
usbhid                 32768  0
ecdh_generic           16384  1 bluetooth
ecc                    32768  1 ecdh_generic
libaes                 16384  1 bluetooth
mcp251xfd              45056  0
ads7846                20480  0
can_dev                36864  1 mcp251xfd
spidev                 24576  0
ci_hdrc_imx            24576  0
ci_hdrc                61440  1 ci_hdrc_imx
ehci_hcd               65536  1 ci_hdrc
ulpi                   16384  1 ci_hdrc
udc_core               32768  1 ci_hdrc
roles                  16384  1 ci_hdrc
usbcore               266240  3 usbhid,ehci_hcd,ci_hdrc
phy_generic            16384  2
usbmisc_imx            24576  1 ci_hdrc_imx
snd_soc_simple_card    20480  0
snd_soc_simple_card_utils    16384  1 snd_soc_simple_card
snd_soc_fsl_sai        32768  2
snd_soc_wm8904         65536  1
imx_pcm_dma_v2         16384  1 snd_soc_fsl_sai
snd_soc_core          180224  5 imx_pcm_dma_v2,snd_soc_simple_card_utils,snd_soc_simple_card,snd_soc_fsl_sai,snd_soc_wm8904
snd_pcm_dmaengine      20480  2 imx_pcm_dma_v2,snd_soc_core
snd_pcm               126976  5 imx_pcm_dma_v2,snd_soc_core,snd_soc_fsl_sai,snd_soc_wm8904,snd_pcm_dmaengine
snd_timer              45056  1 snd_pcm
snd                    81920  4 snd_timer,snd_soc_core,snd_pcm,snd_soc_wm8904
imx_sdma               36864  2
spi_imx                28672  0
hx280enc               24576  0
extcon_usb_gpio        16384  0
hantrodec_845s         36864  0
nfsd                  106496  1
galcore               421888  0
fuse                  131072  1
ipv6                  454656  50

root@imx8mm-var-dart:~# find /dev -group video /dev/galcore /dev/fb0 /dev/dri/card1 /dev/dri/card0 /dev/video0


Here is the output of:

root@imx8mm-var-dart:~# G_MESSAGES_DEBUG=Cog-DRM cog http://neverssl.com Cog-INFO: 08:23:42.021: com.igalia.Cog.Platform: Cog-INFO: 08:23:42.026: drm - 200/CogDrmPlatform (cog:863): Cog-DEBUG: 08:23:42.047: platform_setup: Platform name: (null) (cog:863): Cog-DRM-DEBUG: 08:23:42.055: init_config: overriding device_scale value, using 1.00 from shell (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: enumerated device 0xaaaaecefeab0, available_nodes 5 (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: DRM_NODE_PRIMARY: /dev/dri/card1 (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: DRM_NODE_RENDER: /dev/dri/renderD128 (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: enumerated device 0xaaaaecefefd0, available_nodes 1 (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: DRM_NODE_PRIMARY: /dev/dri/card0 (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: using device 0xaaaaecefefd0, DRM_NODE_PRIMARY /dev/dri/card0 (cog:863): Cog-DRM-DEBUG: 08:23:42.056: init_drm: 1 connectors available (cog:863): Cog-DRM-DEBUG: 08:23:42.057: init_drm: connector id 35, type 16, connected, 1 usable modes (cog:863): Cog-DRM-DEBUG: 08:23:42.057: init_drm: [0]: '800x480', 800x480@68, flags 10, type 72 (preferred) (cog:863): Cog-DRM-DEBUG: 08:23:42.057: init_drm: using connector id 35, type 16 (cog:863): Cog-DRM-DEBUG: 08:23:42.057: init_drm: using mode [0] '800x480' libinput error: event4 - ADS7846 Touchscreen: kernel bug: device has min == max on ABS_PRESSURE (cog:863): Cog-DEBUG: 08:23:42.160: platform_setup: Selected CogDrmPlatform @ 0xaaaaecea06a0

(cog:863): GLib-GIO-WARNING **: 08:23:42.174: Your application does not implement g_application_activate() and has no handlers connected to the 'activate' signal. It should do one of these. Cog-Core-Message: 08:23:42.407: http://neverssl.com/ Load started. Cog-Core-Message: 08:23:42.414: http://neverssl.com/ Loading... Cog-Core-Message: 08:23:42.427: http://neverssl.com/ Loaded successfully. Cog-Core-Message: 08:23:42.428: http://neverssl.com/ Load started. Cog-Core-Message: 08:23:42.609: http://funclearsublimesong.neverssl.com/online Loading... Cog-Core-Message: 08:23:42.696: http://funclearsublimesong.neverssl.com/online Loaded successfully.



Many thanks !

Ps: But if I remeber right @philn wrote somewhere else that he has already an imx8 module with a running configuration? Maybe he can share the configuration or have a quick look?

philn commented 2 years ago

Sorry I can't help much about this, I haven't worked on imx8 for a couple years and it was on a downstream, heavily patched BSP anyway.

aperezdc commented 2 years ago

@SebDE: Thanks for commenting and trying out things 😺 — a couple of observations:

That something is shown on the screen is good; the garbled output looks like some issue related to the chosen pixel format and/or the parameters being used for graphics buffers.
The driver you are using is indeed the Vivante proprietary driver — the galcore kernel module is part of it. The other option would be using the etnaviv driver (which is the open source one). While in theory things should work, the proprietary driver with the DRM plug-in is not as tested as other setups, which is why we have this issue still open: https://github.com/Igalia/cog/issues/211
Looking at the output with G_MESSAGES_DEBUG=Cog-DRM there aren't further hints about the reason for the garbled screen.
The refresh rate of 68Hz (as indicated by 800x480@68 being the mode) is a little bit suspicious, I would expect a more standard refresh rate like 60, 72, or 75Hz. If you run Weston, does it also report 68Hz being used? Your output seems to be a touch screen panel (as indicated by the libinput warning), I am curious to know how it is connected to the SoC, maybe it is using the LVDS interface? If that's the case, is there any chance that the touch screen panel video timings are a bit off? (for example the way in which they are declared in the device-tree are slightly out of spec, or the timer being assigned in the device-tree for the LVDS bus cannot be used to produce a standard frequency?) — Note that I am not saying things are necessarily wrong in your case: if Weston can produce output, we should be able to fix Cog to be able to drive your panel; I am just curious about the strange 68Hz refresh rate 😺

aperezdc commented 2 years ago

@SebDE One more comment: after I fix #401 another option will be to try building Cog from the repository and running with cog -P drm -O renderer=gles which should be somewhat more compatible with certain drivers and has some additional checks that allow that mode be more careful choosing buffer formats that the driver can handle.

jclsn commented 2 years ago

Same issue here. How to change the pixel format? @aperezdc

cog http://localhost:8081                                                                                                                                                                                                                                                                               
[  205.508219] imx-drm 32c00000.bus:display-subsystem: sn65dsi83_connector_mode_valid: mode: 800*480@25000 is valid
[  205.518614] imx-drm 32c00000.bus:display-subsystem: sn65dsi83_connector_mode_valid: mode: 800*480@25000 is valid

(cog:1498): GLib-GIO-WARNING **: 07:31:02.304: Your application does not implement g_application_activate() and has no handlers connected to the 'activate' signal.  It should do one of these.
Cog-Core-Message: 07:31:02.624: <http://localhost:8081/> Load started.
[  205.956077] sn65dsi83 1-002d: DSI clock [ 75000000 ] Hz
[  205.961383] sn65dsi83 1-002d: GeoMetry [ 800 x 480 ] Hz
Cog-Core-Message: 07:31:02.864: <http://localhost:8081/> Loading...
Cog-Core-Message: 07:31:04.254: <http://localhost:8081/> Loaded successfully.

My display seems to be at 25kHz even, which can't be right.

aperezdc commented 2 years ago

@jclsn Could you set G_MESSAGES_DEBUG=Cog-DRM in the environment and re-run? I am curious to see if there are other modes enumerated in your case.

I am not familiar with the sn65dsi83 driver, but doesn't this indicate that 75Hz gets picked in the end?

[  205.956077] sn65dsi83 1-002d: DSI clock [ 75000000 ] Hz

jclsn commented 2 years ago

@aperezdc Will try on Monday.

I btw hear often that you are focusing more on Chromium than on this project. Is that true? Asking because I need to evaluate cog

jclsn commented 2 years ago

@aperezdc I set the environment variable you asked for. Here is the output

Cog-INFO: 08:01:13.836: com.igalia.Cog.Platform:
Cog-INFO: 08:01:13.845:   drm - 200/CogDrmPlatform
(cog:813): Cog-DEBUG: 08:01:13.980: platform_setup: Platform name: (null)
(cog:813): Cog-DRM-DEBUG: 08:01:13.990: init_config: overriding device_scale value, using 1.00 from shell
(cog:813): Cog-DRM-DEBUG: 08:01:13.992: init_drm: enumerated device 0xaaaae670d6b0, available_nodes 5
(cog:813): Cog-DRM-DEBUG: 08:01:13.992: init_drm:   DRM_NODE_PRIMARY: /dev/dri/card1
(cog:813): Cog-DRM-DEBUG: 08:01:13.993: init_drm:   DRM_NODE_RENDER: /dev/dri/renderD128
(cog:813): Cog-DRM-DEBUG: 08:01:13.993: init_drm: enumerated device 0xaaaae670dbd0, available_nodes 1
(cog:813): Cog-DRM-DEBUG: 08:01:13.993: init_drm:   DRM_NODE_PRIMARY: /dev/dri/card0
(cog:813): Cog-DRM-DEBUG: 08:01:13.993: init_drm: using device 0xaaaae670dbd0, DRM_NODE_PRIMARY /dev/dri/card0
(cog:813): Cog-DRM-DEBUG: 08:01:13.993: init_drm: 1 connectors available
(cog:813): Cog-DRM-DEBUG: 08:01:14.004: init_drm:  connector id 35, type 16, connected, 1 usable modes
(cog:813): Cog-DRM-DEBUG: 08:01:14.004: init_drm:    [0]: '800x480', 800x480@59, flags 10, type 72 (preferred) 
(cog:813): Cog-DRM-DEBUG: 08:01:14.014: init_drm: using connector id 35, type 16
(cog:813): Cog-DRM-DEBUG: 08:01:14.015: init_drm: using mode [0] '800x480'
(cog:813): Cog-DEBUG: 08:01:14.040: platform_setup: Selected CogDrmPlatform @ 0xaaaae66af6a0

(cog:813): GLib-GIO-WARNING **: 08:01:14.086: Your application does not implement g_application_activate() and has no handlers connected to the 'activate' signal.  It should do one of these.
Cog-Core-Message: 08:01:14.662: <http://localhost:8081/> Load started.
Cog-Core-Message: 08:01:14.939: <http://localhost:8081/> Loading...
Cog-Core-Message: 08:01:16.381: <http://localhost:8081/> Loaded successfully.

I am not familiar with the sn65dsi83 driver, but doesn't this indicate that 75Hz gets picked in the end?

Not sure. It says Hz not µHz. I would assume that everything is correct here though, because Chromium works without issues.

SebDE commented 2 years ago

@aperezdc - Thanks for the hints and explanation. I will try to make a build with the etnaviv driver (at the moment I don't know how, but hopefully I will find the needed informations). I also collected some informations about the lcd, but first of all: I'm able to run a session with a wayland terminal without problems. But your totally right, the 68Hz are very weird. I'm using an evaluation kit from Variscite and there is also a wiki with some informations.

... equipped with SN65DSI84 MIPI-DSI to LVDS bridge ... evaluation kit comes with 800x480 LCD panel. DART-MX8M-MINI DTS file was created with this panel in mind

Here is the configuration from the device-tree

dsi_lvds_bridge: sn65dsi84@2c {
        compatible = "ti,sn65dsi83";
        reg = <0x2c>;
        ti,dsi-lanes = <4>;
        ti,lvds-format = <1>;
        ti,lvds-bpp = <24>;
        ti,lvds-channels = <1>;
        ti,width-mm = <154>;
        ti,height-mm = <87>;
        enable-gpios = <&gpio2 11 GPIO_ACTIVE_HIGH>;
        pinctrl-names = "default";
        pinctrl-0 = <&pinctrl_lvds>;
        status = "okay";

        display-timings {
            lvds {
                clock-frequency = <33000000>;
                hactive = <800>;
                vactive = <480>;
                hback-porch = <40>;
                hfront-porch = <40>;
                vback-porch = <29>;
                vfront-porch = <13>;
                hsync-len = <48>;
                vsync-len = <3>;
                hsync-active = <0>;
                vsync-active = <0>;
                de-active = <1>;
                pixelclk-active = <0>;
            };
        };

        port {
            lvds_in: endpoint {
                remote-endpoint = <&dsi_out>;
            };
        };
    };

I also searched for some informations in /var/log/messages

Mar 01 17:14:25 imx8mm-var-dart kernel: sn65dsi83 0-002c: DSI clock [ 99000000 ] Hz
Mar 01 17:14:25 imx8mm-var-dart kernel: sn65dsi83 0-002c: Resolution [ 800 x 480 ]
Mar 01 17:14:25 imx8mm-var-dart kernel: sn65dsi83 0-002c: lvds_clk=33000000 lvds_clk_range=0x0
Mar 01 17:14:25 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_range=0x13
Mar 01 17:14:25 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_div=0x2
...
Mar 01 17:15:47 imx8mm-var-dart kernel: sn65dsi83 0-002c: DSI clock [ 99000000 ] Hz
Mar 01 17:15:47 imx8mm-var-dart kernel: sn65dsi83 0-002c: Resolution [ 800 x 480 ]
Mar 01 17:15:47 imx8mm-var-dart kernel: sn65dsi83 0-002c: lvds_clk=33000000 lvds_clk_range=0x0
Mar 01 17:15:47 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_range=0x13
Mar 01 17:15:47 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_div=0x2
Mar 01 17:16:05 imx8mm-var-dart kernel: sn65dsi83 0-002c: DSI clock [ 99000000 ] Hz
Mar 01 17:16:05 imx8mm-var-dart kernel: sn65dsi83 0-002c: Resolution [ 800 x 480 ]
Mar 01 17:16:05 imx8mm-var-dart kernel: sn65dsi83 0-002c: lvds_clk=33000000 lvds_clk_range=0x0
Mar 01 17:16:05 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_range=0x13
Mar 01 17:16:05 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_div=0x2
Mar 01 17:20:44 imx8mm-var-dart kernel: sn65dsi83 0-002c: DSI clock [ 99000000 ] Hz
Mar 01 17:20:44 imx8mm-var-dart kernel: sn65dsi83 0-002c: Resolution [ 800 x 480 ]
Mar 01 17:20:44 imx8mm-var-dart kernel: sn65dsi83 0-002c: lvds_clk=33000000 lvds_clk_range=0x0
Mar 01 17:20:44 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_range=0x13
Mar 01 17:20:44 imx8mm-var-dart kernel: sn65dsi83 0-002c: dsi_clk_div=0x2

@jclsn : What module / panel do you use? Did you make a custom board?

Igalia / cog

cog on imx8mini - failed to schedule a page flip: Permission denied #400