Re: [PATCH V3 7/7] arm64: dts: imx8mp: add interconnect for hsio blk ctrl

[Greg Ungerer]

Hi Marco,

On 28/3/23 23:51, Marco Felsch wrote:
> On 23-03-28, Marco Felsch wrote:
> > Hi Greg,
> >
> > On 23-03-28, Greg Ungerer wrote:
> > > Hi Marco,
> > >
> > > On 28/3/23 17:33, Marco Felsch wrote:
> > > > Hi Greg,
> > > >
> > > > On 23-03-27, Greg Ungerer wrote:
> > > > > Hi Ahmad,
> > > > >
> > > > > On 27/3/23 17:16, Ahmad Fatoum wrote:
> > > > > > On 27.03.23 08:27, Alexander Stein wrote:
> > > > > > > Am Montag, 27. März 2023, 06:50:37 CEST schrieb Greg Ungerer:
> > > > > > > > Any thoughts on why this breaks USB?
> > > > > > >
> > > > > > > Maybe you are missing CONFIG_INTERCONNECT_IMX8MP?
> > > > > >
> > > > > > And if that's the case, did you check /sys/kernel/debug/devices_deferred
> > > > > > to see if there was any indication that this is the reason?
> > > > >
> > > > > Yeah, it does:
> > > > >
> > > > >       # cat /sys/kernel/debug/devices_deferred
> > > > >       32f10100.usb	platform: supplier 32f10000.blk-ctrl not ready
> > > > >       32f10108.usb	platform: supplier 32f10000.blk-ctrl not ready
> > > > >       32ec0000.blk-ctrl	imx8m-blk-ctrl: failed to get noc entries
> > > > >       381f0040.usb-phy	platform: supplier 32f10000.blk-ctrl not ready
> > > > >       382f0040.usb-phy	platform: supplier 32f10000.blk-ctrl not ready
> > > > >       imx-pgc-domain.11	
> > > > >       imx-pgc-domain.12	
> > > > >       imx-pgc-domain.13	
> > > > >       38330000.blk-ctrl	platform: supplier imx-pgc-domain.11 not ready
> > > > >       32f10000.blk-ctrl	imx8mp-blk-ctrl: failed to get noc entries
> > > > >
> > > > > As far as I can tell blk-ctrl should be good:
> > > > >
> > > > >       #
> > > > >       # i.MX SoC drivers
> > > > >       #
> > > > >       CONFIG_IMX_GPCV2_PM_DOMAINS=y
> > > > >       CONFIG_SOC_IMX8M=y
> > > > >       # CONFIG_SOC_IMX9 is not set
> > > > >       CONFIG_IMX8M_BLK_CTRL=y
> > > > >       # end of i.MX SoC drivers
> > > > >
> > > > >
> > > > > > If you didn't find any hint there, you might want to place a
> > > > > > dev_err_probe with a suitable message at the place where -EPROBE_DEFER
> > > > > > was returned.
> > > > >
> > > > > I will try that.
> > > >
> > > > Can you check that CONFIG_ARM_IMX_BUS_DEVFREQ is enabled? This is the
> > > > noc/interconnect driver. This could also the problem for you vpu issue.
> > >
> > > I do not have that enabled. Enabling that fixes the USB probing.
> > > So that is good, thanks.
> > >
> > > It doesn't fix the other problem I mentioned with the vpu pgc nodes though.
> > > I do get some extra messages now with this enabled and the 6.1 kernel:
> > >
> > >      imx-pgc imx-pgc-domain.8: failed to command PGC
> > >      imx-pgc imx-pgc-domain.8: failed to command PGC
> > >      imx8m-blk-ctrl 38330000.blk-ctrl: deferred probe timeout, ignoring dependency
> > >      imx8m-blk-ctrl 38330000.blk-ctrl: error -110: failed to attach power domain "g1"
> > >      imx8m-blk-ctrl: probe of 38330000.blk-ctrl failed with error -110
> >
> > Okay, this seems more like a "real" issue not related to some missing
> > drivers. I followed the code and found a poll within the
> > imx_pgc_power_up() in gpcv2.c. Power-domain 8 is the vpumix domain which
> > is used as power-domain for the g1 power-domain. My assumption is that
> > this poll does run into the timeout. Maybe Peng can support you here
> > since I didn't had the time for to test the VPUs yet and he did the
> > integration patches.
> >
> > Just ignore the errors if you don't use the VPUs or disable the
> > blk-ctrl@38330000 node via status = "disabled".
>
> I forgot to ask: Does your i.MX8MP have a VPU? There are i.MX8MP devices
> (don't know the name) which don't have support for certain IPs. If this

The hardware platform I have is using the MIMX8ML4CVNKZAB "i.MX 8M Plus QuadLite"
(https://www.nxp.com/part/MIMX8ML4CVNKZAB#/) which does not have the hardware
video encode/decoder module (like the "i.MX 8M Plus Quad" parts).


> is the case the bootloader will fixup your devicetree by disable the
> corresponding nodes, we call this feature-controller:
>
> https://elixir.bootlin.com/barebox/latest/source/arch/arm/dts/imx8mp.dtsi
>
> As you can see the imx8mp.dtsi is missing the feature bits for the VPU
> but you can check the i.mx8mm.dtsi. Here you can see that barebox will
> check the availability of the vpu:
>
> https://elixir.bootlin.com/barebox/latest/source/arch/arm/dts/imx8mm.dtsi

Ok, thanks, I'll take a look.

Regards
Greg