DMAMUX peripheral is a dma request router/trigger, present on g0, wb, h7 and l4+.
Basically it allows to easily map peripheral requests to whatever dma channel we
want to use (similarily to the DMA_CSELR register, but without limitation) but,
it also also adds some clever dma request synchronization and even some dma request
generation logic via internal request generator "channels", allowing some requests
chaining, or triggering reqs from non dma capable peripherals.
nb: g0 only features 1 dmamux bloc, supports 7 irq and 4 generators, l4+ supports 13
dma channels and 3 generators and h7 has two dmamuxes, with support for the 15 dma
channels and 7 generators - so as much CxCR and RGxCR register - but they are bit
to bit compatible - excluding of course the sync/sig and dma requests id mappings.
btw, currently, request generator channels are defined in common header, but maybe
we should define them in device header ? or we dont care (like for dma channels,
only defined in dma_f24 but not for other devices ?).
See ST AN5224 for more information
Add basically what's needed to have some minimal but usefull subset of
function for a timer: irqs, compare, period, out polarity, enable/disable
and start.
F09x and L4 share the same cselr register, as well as some L0s, factorize
definitions in a new shared header and add helpers.
fyi, that register allows to redefine dma channel peripheral mapping - see
device datasheet for mapping tables.
While this appears to be a backward change, this moves the _register_
definitions (their addresses) and the actually specific to f4/f7
numbering back into the explicit headers. Potentially this could be
pulled out again, but it's not much code.
This then allows the stm32l1 to use all the rest of this code, with the
differences really being just the addresses of the registers.
Never seen any reason for these noisy verbose defines. They're not
helpful, and we've never needed them for doing sequence setting code
anyway. Just drop them.
While on all current chips, exticr gpio port mux selection is coded on 4 bits,
stm32g0 EXTI_EXTICR register uses 8 bits. Align all exti header to reference
that value (was previously defined for f0 as SYCFG_EXTICR_SKIP)
flash_unlock_acr allows to unlock RUN_PD bit from FLASH_ACR register. Relock is done automatically
when writing 0 to RUN_PD, so no flash_lock_acr method.
It's always stream on the "new" dma controller (unless it's channel....)
Fix a couple of inconsistent prototypes that had carried over from f1
originally. Reported by vampi on irc.
I2C3 is on many parts, but wasn't properly supported with the register
definitions. Declare them centrally, just depending on the memorymap
defining them. On some parts, the rcc bits were defined, but not the
base registers.
Fixes: https://github.com/libopencm3/libopencm3/issues/820
This shows what is _actually_ different for f7. A couple of option
bits, and a renaming of bit 7 of the status register, from Program
Sequence Error to Erase Sequence Error.
We keep the separate implementation of wait_for_last_operation, to meet
the "suggestions" of the reference manual to insert a DSB instruction.
Keeping the renamed bit/functions also requires us to keep separate
implementations of the flag clearing functions
Move the last few register defines back to their relevant headers, add
doxygen and groups. While these registers _were_ "common" they were the
_only_ common things, so it's simpler for future work (merging f7 with
f2/4) to move them back separately.
F3's flash interface is actually quite different, don't try and force
sharing code that isn't really related. The "PGERR" is a very different
bit than the parallelism error that f2/4/7 have.
