Bug 68643

Summary: [ILK/SNB/IVB/HSW DP] 1856x1392 and 1792x1344 modes can't light up
Product: DRI Reporter: cancan,feng <cancan.feng>
Component: DRM/IntelAssignee: Intel GFX Bugs mailing list <intel-gfx-bugs>
Status: CLOSED WORKSFORME QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: low CC: christophe.prigent, jbarnes, jinxianx.guo, lei.a.liu, przanoni, qingshuai.tian, yi.sun
Version: unspecified   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: HSW, ILK, IVB, SNB i915 features: display/DP
Attachments:
Description Flags
system boots and sets display modes
none
1792x1344@60
none
1856x1392@60
none
xrandr --verbose
none
dmesg with log_buf_len=4M
none
new-dmesg-hsw none

Description cancan,feng 2013-08-28 08:17:00 UTC
Created attachment 84775 [details]
system boots and sets display modes

System Environment:
--------------------------
Kernel: (drm-intel-nightly)94814d0f36b38c38700bee6ef64de44a32601cba
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Aug 26 19:51:00 2013 -0300

    drm/i915: Always prefer CPU relocations with LLC

Bug Details:
--------------------------
These two modes can't light up on DP. 

[2]  1856x1392 60 1856 1952 2176 2528 1392 1393 1396 1439 0x6 0x40 218250
[3]  1792x1344 60 1792 1920 2120 2448 1344 1345 1348 1394 0x6 0x40 204750

This is the first time I use the monitor which has these two modes, we bought this monitor some time ago.

I attach the dmesg of system booting and setting this two modes.
Comment 1 Daniel Vetter 2013-08-28 09:01:46 UTC
Does this only happen on SNB or also on other platforms like IVB/HSW?
Comment 2 cancan,feng 2013-08-29 02:36:08 UTC
(In reply to comment #1)
> Does this only happen on SNB or also on other platforms like IVB/HSW?

I specifically borrowed machine with DP and found that all platforms including ILK/SNB/IVB/HSW have this issue.
Comment 3 Daniel Vetter 2013-08-29 06:48:26 UTC
Can you please check with a Windows machine whether it works there with these resolutions? Could be that we need to quirk the monitor ...
Comment 4 cancan,feng 2013-08-29 09:10:32 UTC
(In reply to comment #3)
> Can you please check with a Windows machine whether it works there with
> these resolutions? Could be that we need to quirk the monitor ...

I use a windows 8 machine(Intel Graphics Driver PV9.10) but it doesn't seem to support custom resolution. This custom resolution feature is supported in production versions 14.34 and 15.9 and later versions of the Intel graphics driver. I can't get any of these versions..  So, I can't say if the two modes can work right on Windows machine.
Comment 5 Paulo Zanoni 2013-09-02 14:53:12 UTC
Hi

Your dmesg is not full, it doesn't contain the very beginning. Please increase the value of CONFIG_LOG_BUF_SHIFT.

Also, can you please send us the output of "xrandr --verbose"?

I think the first step is to discover why we're creating these modes: they don't look like standard modes, so maybe we shouldn't add these modes to the list at all.

Thanks,
Paulo
Comment 6 cancan,feng 2013-09-04 03:22:21 UTC
Created attachment 85163 [details]
1792x1344@60
Comment 7 cancan,feng 2013-09-04 03:23:48 UTC
Created attachment 85164 [details]
1856x1392@60
Comment 8 cancan,feng 2013-09-04 03:25:50 UTC
(In reply to comment #3)
> Can you please check with a Windows machine whether it works there with
> these resolutions? Could be that we need to quirk the monitor ...

I retest it with a Windows 8 professional machine and find mode 1856x1392@60 can display normal, but mode 1792x1344@60 display center. So I think the monitor is ok. I attached two pictures of these two modes.
Comment 9 cancan,feng 2013-09-04 05:32:42 UTC
(In reply to comment #5)
> Hi
> 
> Your dmesg is not full, it doesn't contain the very beginning. Please
> increase the value of CONFIG_LOG_BUF_SHIFT.
> 
> Also, can you please send us the output of "xrandr --verbose"?
> 
> I think the first step is to discover why we're creating these modes: they
> don't look like standard modes, so maybe we shouldn't add these modes to the
> list at all.
> 
> Thanks,
> Paulo

Hi, I added log_buf_len=4M to kernel command and grabbed the dmesg again, and I hope it will be useful.

I also attached the "xrandr --verbose", this two modes can display at a windows machine like the two pictures I attached. :)
Comment 10 cancan,feng 2013-09-04 05:33:07 UTC
Created attachment 85165 [details]
xrandr --verbose
Comment 11 cancan,feng 2013-09-04 05:33:46 UTC
Created attachment 85166 [details]
dmesg with log_buf_len=4M
Comment 12 Jani Nikula 2014-09-05 09:18:43 UTC
Timeout. Please re-test with current drm-intel-nightly.
Comment 13 liulei 2014-09-09 05:56:55 UTC
(In reply to comment #12)
> Timeout. Please re-test with current drm-intel-nightly.
I find that only one monitor supports 1792x1344 and 1856x1392 resolution, which failed to be lighted up by latest -nightly kernel on one HSW machine. I-G-T testdisplay can detect the DP monitor.
Comment 14 liulei 2014-09-09 05:59:25 UTC
Created attachment 105951 [details]
new-dmesg-hsw
Comment 15 Jesse Barnes 2014-12-05 19:15:14 UTC
Woo more link training failures and pipe off wait failures.  Something is still wrong with our mode set sequence for DP...
Comment 16 Jesse Barnes 2015-03-30 20:50:14 UTC
Any update on this with current bits?  Maybe some of Jani's AUX fixes might help here, dunno.
Comment 17 xubin 2015-04-01 06:28:33 UTC
The bug still exists on latest drm-intel-testing.
Comment 18 liulei 2015-04-02 03:12:34 UTC
Monitor model No: ViewSonic VS14703.

 EDID:
                00ffffffffffff005a632b8301010101
                33160104b53c22783f2595a9544fa126
                0a5054bfef80e1c0d100d1c0b300a940
                a9c095008180565e00a0a0a029503020
                350055502100001a000000ff00544450
                3132353130303031390a000000fc0056
                503237373020534552494553000000fd
                00324b0f5e1b000a20202020202001e8
                020320f153900504030207060f0e1f14
                13121116151e1d0123097f0783010000
                023a801871382d40582c450055502100
                001e011d007251d01e206e2855005550
                2100001e8c0ad08a20e02d10103e9600
                555021000018011d8018711c1620582c
                250055502100009e023a80d072382d40
                102c458055502100001e000000000077
Comment 19 liulei 2015-04-02 15:35:21 UTC
Notice that DMT v1.0 write:
+------------+----------------+----------+--------------+---------------+
| Pixel      |  Refresh Rate  |  DMT ID  |  STD         |    CVT        |
| Format     |                |  Codes   |2 Byte Codes  |3 Byte Codes   |
+------------+----------------+----------+--------------+---------------+
|1856 x 1392 |   60 Hz        |  41h     |   (C9, 40)h  |  n/a          |
+------------+----------------+----------+--------------+---------------+
|1792 x 1344 |   60 Hz        |   3Eh    |   (C1, 40)h  |   n/a         |
+------------+----------------+----------+--------------+---------------+

This two modes is n/a in CVT table. Does this have some influence on display?

I ever did a test as below steps:
1. cvt 1856 1392 60(I get 8 display timing)
2. replace 1856x1392 display timing in drm display mode table which is
   defined in drm_edid.c .
3. build kernel 
4. 1856x1392 mode can be lighted up with the kernel.
1792x1344 mode is the same.
Comment 20 Jani Nikula 2015-10-23 09:34:50 UTC
Timeout, closing. Please reopen if the problem persists with latest kernels.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.