On Wed, Jan 13, 2021 at 5:16 PM Chen-Yu Tsai [off-list ref] wrote:

On Thu, Jan 7, 2021 at 6:27 PM Samuel Holland [off-list ref] wrote:

quoted

On 1/6/21 5:38 AM, Chen-Yu Tsai wrote:

quoted

On Wed, Jan 6, 2021 at 7:06 PM Maxime Ripard [off-list ref] wrote:

quoted

On Mon, Jan 04, 2021 at 10:54:19AM +0000, André Przywara wrote:

quoted

On 03/01/2021 10:00, Samuel Holland wrote:

quoted

On boards where the only peripheral connected to PL0/PL1 is an X-Powers
PMIC, configure the connection to use the RSB bus rather than the I2C
bus. Compared to the I2C controller that shares the pins, the RSB
controller allows a higher bus frequency, and it is more CPU-efficient.

But is it really necessary to change the DTs for those boards in this
way? It means those newer DTs now become incompatible with older
kernels, and I don't know if those reasons above really justify this.

I understand that we officially don't care about "newer DTs on older
kernels", but do we really need to break this deliberately, for no
pressing reasons?

P.S. I am fine with supporting RSB on H6, and even using it on new DTs,
just want to avoid breaking existing ones.

Doing so would also introduce some inconsistencies, one more thing to
consider during reviews, and would require more testing effort.

I'm not sure that stretching our - already fairly sparse - resources
thin would be very wise here, especially for something that we don't
have to do and for a setup that isn't really used that much.

As soon as some software component starts running RSB, (which I assume
is what Samuel is planning to do in Crust?), there's a chance that it
doesn't switch the chip back to I2C. And then Linux won't be able to
access it.

Crust can handle either way via a config option, which currently
defaults to I2C for H6. It must use the same selection as Linux, not
only because of the PMIC mode, but also because of the pinctrl.

Could Crust be made to also handle pinctrl?

quoted

TF-A is already converted to use RSB[1], and it does switch the PMIC
back to I2C before handing off to U-Boot[2]. So new TF-A + old Linux is
fine. However, Linux currently does not switch the PMIC back. So the
most likely problem from this patch is that, with new Linux + old TF-A,
TF-A will be unable to power down the board or access regulators after
an SoC reset.

I expect there will be a TF-A release between now and when 5.12 hits
stable, but people tend not upgrade their U-Boot/TF-A very often.

We could solve this by having the Linux RSB driver switch all child
devices back to I2C in .shutdown, or by dropping this patch and only
using RSB for new boards (which would also address Andre's concern).

This will work for most cases, except in a kernel panic or IIRC direct
reboot using sysrq. So it's not robust as we'd like it to be.

I also wonder what would happen when there are multiple RSB devices, and
we switch them back to I2C one by one. It's not like there's an option
to switch all of them back at the same time, unlike switching from I2C
to RSB. The A80 and A83T are the platforms that would be affected.

So I merged the previous patch, i.e. changes to the .dtsi, but I think
we should delay this one by a release. That would give us more time to
think about it, and let users upgrade U-Boot/TF-A.

ChenYu

ChenYu

quoted

Cheers,
Samuel

[1]: https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/7576
[2]: https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/7575

quoted

So I'm for keeping things consistent and converting all users to RSB.


ChenYu

--
You received this message because you are subscribed to the Google Groups "linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linux-sunxi+unsubscribe@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/linux-sunxi/bc95a8d2-ebec-489c-10af-fd5a80ea1276%40sholland.org.