MIPI DSI kernel driver bug: prepare vs enable

[inojosh]

I've seen some similar posts related to this, but the issue persists still with kernel 6.6.52.
Somebody at NXP should fix this, as it is a pretty significant and annoying bug.

In any panel driver (source/drivers/gpu/drm/panel) for a MIPI DSI display, you will see the 'prepare' function do these things: enable power supplies, bring out of reset, send whatever commands are necessary to initialize the display. 

The problem is, the underlying drivers for the iMX MIPI DSI or some other low level driver (clocking, etc.) have some bug in them where any attempt to do DSI reads/writes within the 'prepare' function fails. You get "imx_sec_dsim_drv 32e10000.mipi_dsi: wait pkthdr tx done time out" and if you look at the D0 pins, they do not transition at all (other than both going high once to enter LP11 mode).
Since 'prepare' is where init code is supposed to happen, it means no MIPI DSI display will work unless you modify the driver code to move any DSI comms from 'prepare' to 'enable'.

Interestingly enough, the one panel that does have init code in the 'enable' function instead of the 'prepare' function is the one used in the iMX EVK... probably because whoever was responsible for getting that working couldn't get it working with the init code in the correct location (in the 'prepare' function).

[Bio_TICFSL]

Hello,

We have not seen anything like you mentioned, please share with board are you using? wich display? and all the step you made to work with it.

Regards

[inojosh]

I am using my own custom board, with iMX8MM.

The display itself doesn't actually matter for this - the point is that no MIPI DSI communications can happen in the driver's 'prepare' function, the D0 pins will simply not toggle.

Other references to this problem:

https://community.toradex.com/t/imx8mm-and-panel-driver-himax8279d-imx-sec-dsim-drv-wait-pkthdr-tx-d...

https://community.nxp.com/t5/i-MX-Processors/iMX8MP-MIPI-DSI-Bring-up-wait-payload-tx-done-time-out/...

https://community.nxp.com/t5/i-MX-Processors/iMX8MM-MIPI-DSI-commands-before-enabling-the-CLK/m-p/10...

NXP's evk device tree:
https://github.com/nxp-imx/linux-imx/blob/lf-6.12.y/arch/arm64/boot/dts/freescale/imx8mm-evk-rm67191...

The display driver for the display used in the evk:
https://github.com/nxp-imx/linux-imx/blob/lf-6.12.y/drivers/gpu/drm/panel/panel-raydium-rm67191.c

notice that no commands are sent in 'prepare', only 'enable'. This means even someone at NXP had to work around this bug.

Pick any other display driver, and notice that commands are sent in 'prepare':
https://github.com/nxp-imx/linux-imx/blob/lf-6.12.y/drivers/gpu/drm/panel/panel-boe-himax8279d.c
http...

If you build the evk+display image with the displays init sequence moved to 'prepare' from 'enable', you will see that the display no longer works. And it should work... that is where the init sequence is supposed to be (as evidenced by every other display driver). Hence, this is a bug.