Page MenuHomeFreeBSD

Add iMX8MQ support

Authored by gonzo on Jun 14 2020, 11:40 PM.



This patch adds basic support for for iMX8:

  • CCM driver and clocks implementations for iMX8
  • GPC driver for iMX8
  • clock tree for iMX8MQ
  • clocks support and new compat strings (if required) for existing iMX6 UART, I2C, and GPIO drivers
  • Enables aarch64-compatible drivers form iMX6 to arm64 GENERIC kernel config
  • dtb/imx8 kernel module with DTBs for Nitrogen8M and iMX8MQ EVK
Test Plan

This patch was tested only for basic functionality:

  • Booted on Nitrogen8M board with NFS root
  • Booted on iMX8MQ EVK board with NFS root
  • watchdog tested with watchdogd -t 33 -e /usr/bin/false --debug
  • I2C tested by scanning the bus and checking that there is device
  • GPIO wasn't tested

Also verified that both GENERIC and IMX6 kernel configs still build
and boot with this patch applied.

Diff Detail

rS FreeBSD src repository
Automatic diff as part of commit; lint not applicable.
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

gonzo created this revision.Jun 14 2020, 11:40 PM
gonzo requested review of this revision.Jun 14 2020, 11:40 PM
manu added a comment.Jun 16 2020, 5:09 PM

Added a few nits but otherwise it's looks great, thanks for working on this !

170 ↗(On Diff #73120)

I guess this could be commited separatly.

116 ↗(On Diff #73120)

How different are the IMX8 SoCs ?
Maybe we just need SOC_FREESCALE_IMX8 as I guess all the drivers are more or less the same and only a few clocks/device will be present or not.

151 ↗(On Diff #73120)

This can be move above the divr1 = ... line I suppose.

5 ↗(On Diff #73120)

Do you want the "all rights reserved" mention ?

9 ↗(On Diff #73120)

Where does that comes from ?

56 ↗(On Diff #73120)

It's unfortunately GPLv2 licenced ...
What we did for rockchip is that we redefined those value.
I agree that this file cannot be licenced as it's just name/id representation but ...

61 ↗(On Diff #73120)

Does the dummy clocks represent some unimplemented clocks or is it needed because the parent idx is 2 bits ?

5 ↗(On Diff #73120)

Same comment as above for this mention.

2 ↗(On Diff #73120)


gonzo marked 9 inline comments as done.Jun 16 2020, 7:37 PM
gonzo added inline comments.
116 ↗(On Diff #73120)

Yes, looks like the difference is mostly in set of blocks in the SoC, but overall they're conceptually the same:

9 ↗(On Diff #73120)

I started by using arm/freescale/imx/imx_gpt.c but ended up completely rewriting it, so will remove it.

61 ↗(On Diff #73120)

The latter I think. There is dummy clock id in the header file: IMX8MQ_CLK_DUMMY

gonzo updated this revision to Diff 73181.Jun 16 2020, 7:38 PM
gonzo marked 3 inline comments as done.

Address manu@ comments

manu accepted this revision.Jun 16 2020, 7:43 PM
This revision is now accepted and ready to land.Jun 16 2020, 7:43 PM
This revision was automatically updated to reflect the committed changes.