Re: [PATCH] clk: ingenic-tcu: Properly enable registers before accessing timers

[Aidan MacDonald <aidanmacdonald.0x0@gmail.com>]

Paul Cercueil <paul@crapouillou.net> writes:

> Le ven., juin 10 2022 at 16:43:27 +0100, Aidan MacDonald
> <aidanmacdonald.0x0@gmail.com> a écrit :
>> Stephen Boyd <sboyd@kernel.org> writes:
>> 
>>>  Quoting Aidan MacDonald (2022-06-03 06:47:05)
>>>>  Access to registers is guarded by ingenic_tcu_{enable,disable}_regs()
>>>>  so the stop bit can be cleared before accessing a timer channel, but
>>>>  those functions did not clear the stop bit on SoCs with a global TCU
>>>>  clock gate.
>>>>  Testing on the X1000 has revealed that the stop bits must be cleared
>>>>  _and_ the global TCU clock must be ungated to access timer registers.
>>>>  Programming manuals for the X1000, JZ4740, and JZ4725B specify this
>>>>  behavior. If the stop bit isn't cleared, then writes to registers do
>>>>  not take effect, which can leave clocks with no defined parent when
>>>>  registered and leave clock tree state out of sync with the hardware,
>>>>  triggering bugs in downstream drivers relying on TCU clocks.
>>>>  Fixing this is easy: have ingenic_tcu_{enable,disable}_regs() always
>>>>  clear the stop bit, regardless of the presence of a global TCU gate.
>>>>  Signed-off-by: Aidan MacDonald <aidanmacdonald.0x0@gmail.com>
>>>>  ---
>>>  Any Fixes: tag?
>> Probably 4f89e4b8f121 ("clk: ingenic: Add driver for the TCU clocks")
>> but I don't have docs or hardware to confirm the bug affects the jz4770,
>> which is the only other SoC affected by the change.
>> I think what caused my problem was my bootloader stopping all the timer
>> channels. The stop bits are supposed to be zeroed at reset, so I'd guess
>> the jz4770 relied on that and only worked by accident.
>
> I'll test it on JZ4770 this weekend.
>
>> I'll send a v2 along shortly. Is it worth CC'ing stable as well?
>
> If the bug is in jz-5.18 or earlier, yes.
>
> Cheers,
> -Paul

Thanks. Guess I'll wait for your test results, though I don't expect any
problems.