[PATCH v3 00/56] Convert DSI code to use drm_mipi_dsi and drm_panel

[Tomi Valkeinen]

On 10/11/2020 23:04, H. Nikolaus Schaller wrote:
> 
>> Am 10.11.2020 um 17:52 schrieb Tomi Valkeinen <tomi.valkeinen at ti.com>:
>>
>> On 10/11/2020 18:49, H. Nikolaus Schaller wrote:
>>
>> I guess you have the same issue. It goes to dsi_bridge_mode_valid, then __dsi_calc_config, and stays
>> there finding good clocks.
> 
> Yes, I could trace it down to exactly this point.
> 
> So the culprit is somehow the panel driver. Because it asks for clocks that the PLL driver does not want to provide...
> Or is it the victim?
> 
> Here is what dmesg reports with even more printk():
> 
> [  276.970635] drm_helper_probe_single_connector_modes 12 count=1
> [  277.003509] drm_mode_validate_pipeline 2 ret=0 status=0
> [  277.038678] drm_bridge_chain_mode_valid: func=dsi_bridge_mode_valid+0x0/0xa0 [omapdrm]
> [  277.047199] dsi_bridge_mode_valid
> [  277.050786] __dsi_calc_config
> [  277.057270] dsi_vm_calc
> [  277.073251] dss_pll_calc_a clkin=19200000 pll_min=1555386656 pll_max=1555410656 func=dsi_vm_calc_pll_cb+0x0/0x48 [omapdrm]
> [  277.084975] dss_pll_calc_a pll_hw_max=1800000000 fint_hw_min=150000 fint_hw_max=52000000
> [  277.093637] dss_pll_calc_a n_start=1 n_inc=1 n_stop=128 pll_max'=1555410656
> [  277.101062] dss_pll_calc_a n=1 clkin=19200000 fint=19200000
> [  277.107152] dss_pll_calc_a m_start=41 m_inc=1 m_stop=40
> 
> Ok, we have to wait quite a while until the for(m;;) loop ends, because m_stop < m_start and m_inc is positive.
> 
> So something in the formulae in dss_pll_calc_a() does not fit or has unintended rounding effects.
> Or the values reported by w677l_get_modes() do not fit anything.
> 
> I think these findings and ideas should help to find a fix.

drm_display_mode.clock is in kHz, but the panel driver writes Hz (w677l_PIXELCLOCK) to it. But
there's more after fixing that. The DSI gets configured in bridge's modeset, which I think is before
w677l_prepare where the panel already sends DSI commands. Also, the dsi driver fails to lock the
PLL, so possibly the clock calcs are still wrong.

 Tomi

-- 
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki