On Saturday, March 28, 2015 09:25 PM, George Grey wrote:

On Mar 28, 2015, at 5:55 AM, xinliang <xin3liang@qq.com> wrote:

Hi Scott,
Sorry for late replying. I just come back from a travel.


On Friday, March 27, 2015 03:58 PM, Dan zhao wrote:

-----邮件原件-----
发件人: Scott Bambrough [mailto:scott.bambrough@linaro.org]
发送时间: 2015年3月27日 7:07
收件人: Benjamin Gaignard; dev@lists.96boards.org
抄送: Tom Gall; Guodong Xu; Lars-Peter Clausen; George Grey; Dan zhao
主题: Re: [Dev] X fails to load on current alip image....

On Thu, Mar 26, 2015 at 3:03 PM, Benjamin Gaignard
<benjamin.gaignard@linaro.org> wrote:
It could be simple call to an i2c adapter.
For example it is done like this in sti drm/kms driver:
https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/tree/
drivers/gpu/drm/sti/sti_hdmi.c?id=refs/tags/next-20150326#n695

Benjamin

Le jeudi 26 mars 2015, Scott Bambrough <scott.bambrough@linaro.org> a
écrit
:

We have no hdmi ddc client in the kernel.  I suspect this is the
patches that HiSilicon needs to give us Guodong.

Scott

On Thu, Mar 26, 2015 at 1:58 PM, Tom Gall <tgall@gentoo.org> wrote:
FYI

Advise from Benjamin.

Begin forwarded message:

Date: March 26, 2015 at 8:57:34 AM PDT
Subject: Re: [Dev] X fails to load on current alip image....
From: Benjamin Gaignard <benjamin.gaignard@linaro.org>
To: Tom Gall <tgall@gentoo.org>

Hi,

not sound familiar but I think that the hdmi ddc client isn't set
on the correct i2c bus so EDID can't be retrieved.

I suggest to check if the hdmi ddc is set on the correct i2c bus
and if the dispaly is well connected.

Regards,
Benjamin

2015-03-26 16:12 GMT+01:00 Tom Gall <tgall@gentoo.org>:

Hi Benjamin,

Does the drm error below look familiar?

[    2.412661] [drm:hisi_dsi_probe] *ERROR* failed to find slave
encoder i2c client

Is that a real error or is there something else to look at?

Thanks!
Tom

Begin forwarded message:

From: Scott Bambrough <scott.bambrough@linaro.org>
Subject: [Dev] X fails to load on current alip image....
To: <dev@lists.96boards.org>
Date: March 26, 2015 at 7:21:51 AM PDT

I had a look at this failure, debugged this a bit...

Started with latest snapshot (135), X fails to load.

Rebuild kernel with the latest git, hikey-mali branch Rebuild boot
and system images using latest kernel, setup networking via eth0 by
default so I can ssh in.
Flash boot and system partitions using rebuilt images.
Reboot (1)

Looking into Xorg.0.log see the following error:

[    60.310] (==) RandR enabled
[    60.333] (II) SELinux: Disabled on system
[    60.336] (II) AIGLX: Screen 0 is not DRI2 capable
[    60.336] (EE) AIGLX: reverting to software rendering
[    60.336] (EE) AIGLX error: dlopen of
/usr/lib/aarch64-linux-gnu/dri/swrast_dri.so failed
(/usr/lib/aarch64-linux-gnu/dri/swrast_dri.so: cannot open shared
object file: No such file or directory)
[    60.336] (EE) GLX: could not load software renderer
[    60.336] (II) GLX: no usable GL providers found for screen 0
[    60.426] (II) config/udev: Adding drm device (/dev/dri/card0)
[    60.426] (II) xfree86: Adding drm device (/dev/dri/card0)
[    60.429] (EE) /dev/dri/card0: failed to set DRM interface version
1.4: Invalid argument
[    60.590] (EE) Server terminated successfully (0). Closing log file.

Solve that error by installing the following package:
# apt-get install libgl1-mesa-dri

Reboot (2)
Missing shared library is found.

[    43.500] (==) RandR enabled
[    43.523] (II) SELinux: Disabled on system
[    43.526] (II) AIGLX: Screen 0 is not DRI2 capable
[    43.526] (EE) AIGLX: reverting to software rendering
[    44.621] (II) AIGLX: Loaded and initialized swrast
[    44.621] (II) GLX: Initialized DRISWRAST GL provider for screen 0
[    44.970] (II) config/udev: Adding drm device (/dev/dri/card0)
[    44.971] (II) xfree86: Adding drm device (/dev/dri/card0)
[    44.973] (EE) /dev/dri/card0: failed to set DRM interface version
1.4: Invalid argument

Why the error "/dev/dri/card0: failed to set DRM interface version
1.4: Invalid argument"?
dmesg output has the following error

[   44.985050] ------------[ cut here ]------------
[   44.985082] WARNING: CPU: 1 PID: 1453 at
/media/scottb/linaro/hikey/source/linux/drivers/gpu/drm/drm_ioctl.c
:143
drm_setversion+0x168/0x16c()
[   44.985095] No drm_driver.set_busid() implementation provided by
hisi_drm_driver. Use drm_dev_set_unique() to set the unique name
explicitly.
[   44.985100] Modules linked in:
[   44.985106]  btwilink bluetooth st_drv mali wl18xx wlcore mac80211
cfg80211 rfkill wlcore_sdio
[   44.985136] CPU: 1 PID: 1453 Comm: Xorg Tainted: G        W
3.18.0linaro-hikey+ #3
[   44.985141] Call trace:
[   44.987598] [<ffffffc000088538>] dump_backtrace+0x0/0x124
[   44.987606] [<ffffffc00008866c>] show_stack+0x10/0x1c
[   44.987616] [<ffffffc000806bfc>] dump_stack+0x74/0xb8
[   44.987625] [<ffffffc0000afdb4>] warn_slowpath_common+0x90/0xb8
[   44.987631] [<ffffffc0000afe28>] warn_slowpath_fmt+0x4c/0x58
[   44.987639] [<ffffffc0004eaec4>] drm_setversion+0x164/0x16c
[   44.987646] [<ffffffc0004eaaa0>] drm_ioctl+0x254/0x4d0
[   44.987656] [<ffffffc0001d216c>] do_vfs_ioctl+0x368/0x5b0
[   44.987663] [<ffffffc0001d2434>] SyS_ioctl+0x80/0x98
[   44.987669] ---[ end trace bf426acc001551e5 ]---

Solve that error by applying the following patch:

diff --git a/drivers/gpu/drm/hisilicon/hisi_drm_drv.c
b/drivers/gpu/drm/hisilicon/hisi_drm_drv.c
index 987c530..b646bab 100644
--- a/drivers/gpu/drm/hisilicon/hisi_drm_drv.c
+++ b/drivers/gpu/drm/hisilicon/hisi_drm_drv.c
@@ -147,6 +147,7 @@ static struct drm_driver hisi_drm_driver = {
               | DRIVER_PRIME,
   .load            = hisi_drm_load,
   .unload     = hisi_drm_unload,
+    .set_busid    = drm_platform_set_busid,
   .fops            = &hisi_drm_fops,
   .name            = "hisi-drm",
   .desc            = "Hisilicon Terminal SoCs DRM Driver",

Rebuild kernel
Rebuild boot partition image with latest kernel and flash to the
boot partition of the HiKey Reboot (3)

dmesg has the following messages from DRM:

linaro@linaro-alip:~/sdcard/xwork/3$ grep drm dmesg.txt
[    2.410636] calling  drm_core_init+0x0/0x12c @ 1
[    2.410705] [drm] Initialized drm 1.1.0 20060810
[    2.410720] initcall drm_core_init+0x0/0x12c returned 0 after 69
usecs
[    2.410742] calling  hisi_drm_platform_driver_init+0x0/0x20 @ 1
[    2.411790] [drm] Initialized hisi-drm 1.0.0 20141224 on minor 0
[    2.411893] initcall hisi_drm_platform_driver_init+0x0/0x20
returned 0 after 1108 usecs
[    2.412661] [drm:hisi_dsi_probe] *ERROR* failed to find slave
encoder i2c client
[    3.546316] [drm] HDMI: new_status=1,old_status=2,hpd=1,dpms=3
[    3.602363] hisi-drm smb:display-subsystem: DSI-1: EDID block 0
invalid.
[    3.602469] [drm] HDMI: mode=0,format_422=0,format_ycbcr=0
[    3.603762] [drm] May has one empty edid block!
[    3.624268] [drm]
mipi_init,pixcel_clk=75,dphy_freq=641,hsa=42,hbp=117,hline=1760
[    3.624270] [drm] mipi_init , exit success!
[    3.654185] hisi-drm smb:display-subsystem: fb0:  frame buffer
device
[    3.654192] hisi-drm smb:display-subsystem: registered panic notifier
[   43.134240] [drm] phystopstateclklane is not ready.
[   43.134350] [drm]
mipi_init,pixcel_clk=75,dphy_freq=641,hsa=42,hbp=117,hline=1760
[   43.134350] [drm] mipi_init , exit success!

Xorg.0.log error setting DRM version is gone.

[    43.134] (==) RandR enabled
[    43.158] (II) SELinux: Disabled on system
[    43.161] (II) AIGLX: Screen 0 is not DRI2 capable
[    43.161] (EE) AIGLX: reverting to software rendering
[    43.173] (II) AIGLX: Loaded and initialized swrast
[    43.173] (II) GLX: Initialized DRISWRAST GL provider for screen 0

Why do these errors occur?
[    2.412661] [drm:hisi_dsi_probe] *ERROR* failed to find slave
encoder i2c client
This is not a problem, the dsi module is waiting for HDMI module. When dsi module can't find HDMI slave encoder module it will return -EPROBE_DEFER to deffer probe.
After the HDMI module has been initialized. This error msg would not come again.
[    3.602363] hisi-drm smb:display-subsystem: DSI-1: EDID block 0
invalid.
EDID error doesn't matter now. Due to the unfixed signal issue, now we explicitly set the 720P timing parameters. So the timing parameters doesn't come from EDID.

Hello - this is in fact a major problem for us. We want to make the HDMI work “properly” with all monitors. We also need to investigate the possible SoC clock timing issue identified by Lars at Analog Devices. Therefore we want to enable the EDID data so that we can then test different display timings and configurations. 
To do testing, characterization and validation of different monitor timings we also need a command line tool for debugging to enable us to change the DSI timings and clock set ups on the SoC (see Lars emails) for different displays. Here is the use case:

1. Read EDID information on boot 
2. Boot up in known working 720p configuration (this is what happens today) 
3. User reads the EDID data using dmesg
4. Users logs in from remote PC using ssh
5. Using the EDID information and a debug tool which can adjust clock settings and DSI output configuration the user sets the monitor timings for different supported frequency, no of pixels, lines etc. The debug tool makes the appropriate changes in the SoC and the ADV7533. 
6. User records whether the monitor is working correctly
7. If the monitor fails the user can go back to the known working 720p setup using the debug tool. 

This will help us to characterize which timing setups work and which fail so that we can find the root cause of HDMI not working at different frequencies/configurations. 
Can you help on this?
Yes, I can read EDID properly with my local patch, i'll push this patch to 96boards' Github kernel as soon as possible.
Such a debug tool is something like the "modetest" tool in libdrm. I'll try it.
But, the import fact is that:
Recent days i and our SoC guys are analyzing the HDMI signal issue. We have tested many mode timings on TV monitors and PC monitors and we found very a few mode timings can work, including the 720P. SoC guys said that TV monitors are follow the CEA 681 standard and our SoC can't math this standard and during our SoC design phase HDMI is not taken into account.
I'm afraid this signal can't be fixed and our SoC can't support HDMI well.

[   43.134240] [drm] phystopstateclklane is not ready.

Out of my depth here, punting to HiSilicon engineers.   Guodong can
you take up these errors with HiSilicon please.

Attached logs (dmesg, Xorg.0.log) from after every reboot.  In text
above you will see:
Reboot (x) -- logs are in directory x if the tarball

Also attached script (build.sh) I use to rebuild the images for
flashing.  It will need some rework as it is pretty custom to my
system.

Scott

--
Scott Bambrough
Technical Director, Member Services Linaro








--
Benjamin Gaignard

Graphic Working Group

Linaro.org │ Open source software for ARM SoCs

Follow Linaro: Facebook | Twitter | Blog




--
Scott Bambrough
Technical Director, Member Services
Linaro


--

Benjamin Gaignard

Graphic Working Group

Linaro.org │ Open source software for ARM SoCs

Follow Linaro: Facebook | Twitter | Blog


Yeah, I realized that afterwards.  Seems like the Analog Device code
in drivers/gpu/drm/i2c/adv7533.c drivers/media/i2c/adv7533.c registers
one.

See: static int adv7533_probe(struct i2c_client *i2c, const struct
i2c_device_id *id)
       adv7533->i2c_main = i2c;
        adv7533->i2c_edid = i2c_new_dummy(i2c->adapter, edid_i2c_addr >>
1);
        if (!adv7533->i2c_edid)
                return -ENOMEM;

If I grep through the dmesg output:

scottb@eagle:~/tmp/logs/3$ grep "adv7533\|drm" dmesg.txt
[    2.410636] calling  drm_core_init+0x0/0x12c @ 1
[    2.410705] [drm] Initialized drm 1.1.0 20060810
[    2.410720] initcall drm_core_init+0x0/0x12c returned 0 after 69 usecs
[    2.410742] calling  hisi_drm_platform_driver_init+0x0/0x20 @ 1
[    2.411790] [drm] Initialized hisi-drm 1.0.0 20141224 on minor 0
[    2.411893] initcall hisi_drm_platform_driver_init+0x0/0x20
returned 0 after 1108 usecs
[    2.412661] [drm:hisi_dsi_probe] *ERROR* failed to find slave
encoder i2c client
[    2.420194] calling  adv7533_init+0x0/0x24 @ 1
[    2.420254] initcall adv7533_init+0x0/0x24 returned 0 after 44 usecs
[    3.546316] [drm] HDMI: new_status=1,old_status=2,hpd=1,dpms=3
[    3.602363] hisi-drm smb:display-subsystem: DSI-1: EDID block 0 invalid.
[    3.602469] [drm] HDMI: mode=0,format_422=0,format_ycbcr=0
[    3.603762] [drm] May has one empty edid block!
[    3.624268] [drm]
mipi_init,pixcel_clk=75,dphy_freq=641,hsa=42,hbp=117,hline=1760
[    3.624270] [drm] mipi_init , exit success!
[    3.654185] hisi-drm smb:display-subsystem: fb0:  frame buffer device
[    3.654192] hisi-drm smb:display-subsystem: registered panic notifier
[   43.134240] [drm] phystopstateclklane is not ready.
[   43.134350] [drm]
mipi_init,pixcel_clk=75,dphy_freq=641,hsa=42,hbp=117,hline=1760
[   43.134350] [drm] mipi_init , exit success!

It appears the HiSilicon DRM driver is initialized, then the ADV7533.
Could it just be the init order is wrong?

[    2.412661] [drm:hisi_dsi_probe] *ERROR* failed to find slave
encoder i2c client
The above error occurs before the ADV7533 driver is initialized.

[    3.602363] hisi-drm smb:display-subsystem: DSI-1: EDID block 0 invalid.
[    3.602469] [drm] HDMI: mode=0,format_422=0,format_ycbcr=0
[    3.603762] [drm] May has one empty edid block!

The above errors stem from a call to:
static int adv7533_get_modes(struct drm_encoder *encoder, struct
drm_connector *connector)

[    3.654185] hisi-drm smb:display-subsystem: fb0:  frame buffer device
[    3.654192] hisi-drm smb:display-subsystem: registered panic notifier

These indicate the framebuffer is being created.  These stem from a call to:
static int drm_fb_helper_single_fb_probe(struct drm_fb_helper
*fb_helper, int preferred_bpp)

which is called from:
bool drm_fb_helper_initial_config(struct drm_fb_helper *fb_helper, int
bpp_sel)

which is called from:
static struct hisi_drm_fbdev *hisi_drm_fbdev_create(struct drm_device *dev,
        unsigned int preferred_bpp, unsigned int num_crtc,
        unsigned int max_conn_count)

Another thing I found odd is mipi_init seems to be called twice:  Why?

[    3.624268] [drm]
mipi_init,pixcel_clk=75,dphy_freq=641,hsa=42,hbp=117,hline=1760
[    3.624270] [drm] mipi_init , exit success!

[   43.134350] [drm]
mipi_init,pixcel_clk=75,dphy_freq=641,hsa=42,hbp=117,hline=1760
[   43.134350] [drm] mipi_init , exit success!

mipi_init is called from:
static int hisi_dsi_init(struct hisi_dsi *dsi)

which is called from:
static int hisi_dsi_enable(struct hisi_dsi *dsi)

which is called from:
static void hisi_drm_encoder_dpms(struct drm_encoder *encoder, int mode)
when mode = DRM_MODE_DPMS_ON

There is a corresponding
static int hisi_dsi_disable(struct hisi_dsi *dsi)
called when mode is one of DRM_MODE_DPMS_STANDBY,
DRM_MODE_DPMS_SUSPEND, DRM_MODE_DPMS_OFF

It does nothing however:

static int hisi_dsi_disable(struct hisi_dsi *dsi)
{
        DRM_DEBUG_DRIVER("enter.\n");
        DRM_DEBUG_DRIVER("exit success.\n");
        return 0;
}

[   43.134240] [drm] phystopstateclklane is not ready.
This error is from mipi_init.  The first time through mipi_init it
doesn't occur, it does on the second time though.
Seems to indicate a problem with the clocks for the mipi dsi data lanes.

I really think we need to sort this out.

Scott

--
Scott Bambrough
Technical Director, Member Services
Linaro
Anyway,  I'll take a look at the latest debian img to see if the "fail to load " problem has someting about the drm driver changes during recent period.

Thanks very much - the failure happened when the HDMI code was changed to the DRM driver. If this can be fixed we can again bring up the GUI and test Mali GPU acceleration. 

Kind regards
George

Best regards,
-Xinliang Liu