On Sat, 24 Nov 2018, Michael Schmitz wrote:
Am 20.11.2018 um 23:02 schrieb Andreas Schwab:
On Nov 20 2018, Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:
Yes you already see the same as I see: this chip MK68901 has no less
than four timers. I bet the kernel is just using one of them, out of
habit.
Note that not all timers can be used freely. Some of them are
hardwired to generate the clock for the serial interfaces.
Timer A is used by the DMA sound driver - no workaround possible there.
Timer B is used by the framebuffer driver, but it's used only once to
reprogram the screen base address at driver init. This one could
potentially be used after framebuffer init to improve the clocksource
accuracy.
I don't think it would make a good clocksource as MFP timers are all 8
bits wide. It could be used as a clock event device in the course of a
GENERIC_CLOCKEVENTS conversion (discussed elsewhere in this thread),
though the longest timed interrupt inverval would be only 21 ms.
--
Timer D is already used to generate timer interrupts used to poll the
ROM port network card / USB adapters. This timer is initialized early in
the boot process, which prevents using the MFP UART as serial console
(something that I hadn't properly considered before). I'll send a patch
for that. I'll also consider using timer B or timer C interrupts instead
to poll ROM port hardware.
There are no serial drivers anymore that could use the MFP UART
hardware, so that point is somewhat moot at present.
Cheers,
Michael
Andreas.
