Index | Thread | Search

From:
Mark Kettenis <mark.kettenis@xs4all.nl>
Subject:
Re: rkcomphy: add rk3528 support
To:
Hayk Martirosyan <aoglmf@gmail.com>
Cc:
jonathan@d14n.org, tech@openbsd.org, kettenis@openbsd.org
Date:
Tue, 24 Mar 2026 21:08:49 +0100

Download raw body.

Thread
> From: Hayk Martirosyan <aoglmf@gmail.com>
> Date: Tue, 24 Mar 2026 23:39:53 +0400
> 
> Hi everyone,
> 
> Sorry for chiming in, but shouldn’t DIV(7, 2) be written as DIV(6, 2)
> for RK3528_CLK_PPLL_100M_MATRIX?
> 
> In the Linux definition of CLK_PPLL_100M_MATRIX, the same data is
> given as div_shift = 2, div_width = 5. If I'm not mistaken, that
> should translate to DIV(6, 2) in OpenBSD. The
> RK3528_CLK_MATRIX_50M_SRC clock is already defined that way.

Yes.  The OpenBSD "convention" here is designed to make it easy to
check against the technical reference manuals.  Unfortunately it seems
Rockchip didn't make the TRM for the RK3528 public :(.

> One more question for Jonathan: did you manage to get the Radxa E20C
> to survive warm reboots? Every time I reboot the board, U-Boot fails
> to bring up PCIe and prints the following message:
> 
> pcie_dw_rockchip pcie@fe000000: PCIe-0 Link Fail
> 
> It looks like the current OpenBSD code leaves PCIe in a state that
> U-Boot cannot properly reinitialize.

I'd say that is a bug in U-Boot.