Bug 98353

There could be two reasons:
- There is no LSPCON on port B (VBT invalid)
- Not able to detect LSPCON signature. 

How can I see detailed logs (with drm_error=0xe) ?

https://intel-gfx-ci.01.org/CI/CI_DRM_1742/fi-skl-6700hq/dmesg-during-piglit-run-1.log

[   45.793263] [drm:intel_ddi_init [i915]] VBT says port B has lspcon
[   45.793333] [drm:intel_dp_init_connector [i915]] Adding DP connector on port B
[   45.793921] [drm:drm_dp_i2c_do_msg] native defer
[   45.795037] [drm:drm_dp_i2c_do_msg] native defer
[   45.796147] [drm:drm_dp_i2c_do_msg] native defer
[   45.797273] [drm:drm_dp_i2c_do_msg] native defer
[   45.798337] [drm:drm_dp_i2c_do_msg] native defer
[   45.799380] [drm:drm_dp_i2c_do_msg] native defer
[   45.800431] [drm:drm_dp_i2c_do_msg] native defer
[   45.801048] [drm:drm_dp_i2c_do_msg] too many retries, giving up
[   45.801475] [drm:drm_dp_i2c_do_msg] native defer
[   45.802570] [drm:drm_dp_i2c_do_msg] native defer
[   45.803682] [drm:drm_dp_i2c_do_msg] native defer
[   45.804796] [drm:drm_dp_i2c_do_msg] native defer
[   45.805912] [drm:drm_dp_i2c_do_msg] native defer
[   45.807027] [drm:drm_dp_i2c_do_msg] native defer
[   45.808136] [drm:drm_dp_i2c_do_msg] native defer
[   45.808837] [drm:drm_dp_i2c_do_msg] too many retries, giving up
[   45.808870] [drm:lspcon_init [i915]] No LSPCON detected, found unknown
[   45.808901] [drm:lspcon_init [i915]] *ERROR* Failed to probe lspcon

See the drv_module_reload_basic test results across machines and CI runs at [1]. The fi-skl-6700hq row indicates module reload sporadically fails with LSPCON init. The machine obviously has LSPCON and it's indicated in VBT. You can see the successful and failing columns, and check the dmesgs when you click the column titles.

[1] https://intel-gfx-ci.01.org/CI/igt@drv_module_reload_basic.html

Thanks for the logs, Chris. 
From the logs:
[drm:lspcon_init [i915]] No LSPCON detected, found unknown

lspcon layer uses drm_dp_dual_mode implementation to detect if there is a lspcon dongle attached with the port (i2c_over_aux) and once we find a dongle, we try to read lspcon signature to determine if its MCA lspcon. 

Looks like drm_dp_dual_mode API could not detect a dp_dual_mode dongle, and it returned unknown. So LSPCON detection failed. 

This could be due to:
- Faulty HW/dongle 
- Invalid FW on dongle
- Wrong VBT programming (there is no LSPCON on port B)

- Shashank

I believe this is just the same "LSPCON firmware falls over if poked too quickly after suspend" thing that Imre found on his APL. So we need Imre's workaround I believe.

This same is now causing dwarns on 6700hq for 
igt@drv_module_reload_basic
igt@gem_exec_suspend@basic-s3
igt@kms_pipe_crc_basic@suspend-read-crc-pipe-a
igt@kms_pipe_crc_basic@suspend-read-crc-pipe-b
igt@kms_pipe_crc_basic@suspend-read-crc-pipe-c

[  185.090441] [IGT] gem_exec_suspend: starting subtest basic-S3
[  187.261576] [drm:drm_lspcon_get_mode] *ERROR* LSPCON read(0x80, 0x41) failed
[  187.261606] [drm:lspcon_change_mode.constprop.2 [i915]] *ERROR* Error reading LSPCON mode
[  187.261618] [drm:lspcon_resume [i915]] *ERROR* LSPCON resume failed

[  475.664167] [IGT] kms_pipe_crc_basic: starting subtest suspend-read-crc-pipe-A
[  477.560903] [drm:drm_lspcon_get_mode] *ERROR* LSPCON read(0x80, 0x41) failed
[  477.560936] [drm:lspcon_change_mode.constprop.2 [i915]] *ERROR* Error reading LSPCON mode
[  477.560948] [drm:lspcon_resume [i915]] *ERROR* LSPCON resume failed

[  482.292551] [IGT] kms_pipe_crc_basic: starting subtest suspend-read-crc-pipe-B
[  484.285888] [drm:drm_lspcon_get_mode] *ERROR* LSPCON read(0x80, 0x41) failed
[  484.285901] [drm:lspcon_change_mode.constprop.2 [i915]] *ERROR* Error reading LSPCON mode
[  484.285913] [drm:lspcon_resume [i915]] *ERROR* LSPCON resume failed

[  488.994986] [IGT] kms_pipe_crc_basic: starting subtest suspend-read-crc-pipe-C
[  490.675622] [drm:drm_lspcon_get_mode] *ERROR* LSPCON read(0x80, 0x41) failed
[  490.675636] [drm:lspcon_change_mode.constprop.2 [i915]] *ERROR* Error reading LSPCON mode
[  490.675647] [drm:lspcon_resume [i915]] *ERROR* LSPCON resume failed

I've sent an updated version, which is mostly reviewed by Jani. Still waiting for his R-b for 2 out of 8 patches.

Patch series details:
https://lists.freedesktop.org/archives/intel-gfx/2016-October/109619.html

Created attachment 127547 [details]
attachment-30464-0.html

Hello,
I am on Diwali vacation during 25 Oct - 01 Nov, with limited Phone and no mail access.
Please expect delay in response.

If this is something urgent, please contact my manager : Indranil, Mukherjee

Regards
Shashank

The fix is now merged to -nightly.

Thanks Imre.

Re-appeared in; https://intel-gfx-ci.01.org/CI/Patchwork_2888/fi-skl-6770hq/igt@drv_module_reload_basic.html

(In reply to Joonas Lahtinen from comment #14)
> Re-appeared in;
> https://intel-gfx-ci.01.org/CI/Patchwork_2888/fi-skl-6770hq/
> igt@drv_module_reload_basic.html

This is a separate issue from the originally reported (on two different machines). The original one was triggered by S3 and probe failed with the following error:
*ERROR* LSPCON read(0x80, 0x41) failed

The one in patchwork_2888 happens independent of S3 with the following error:
[drm:drm_dp_dual_mode_detect] DP dual mode adaptor ID: 44 (err 0)
[drm:lspcon_init [i915]] No LSPCON detected, found type 1 HDMI
[drm:lspcon_init [i915]] *ERROR* Failed to probe lspcon

In this latter case LSPCON recovers, and looks like the same issue we have with some type 1 dongles where the read offset during I2C-over-AUX transfers is ignored by the dongle.

Let's track the single issue described in comment 15 in this bug.

Reference to Imre's patchset: https://patchwork.freedesktop.org/series/15682/

Fix merged to -nightly.

(In reply to Imre Deak from comment #18)
> Fix merged to -nightly.

Closing as fixed and verified in CI system