With the embedded-hal rc3 update I changed them to require `&mut self`, but
in retrospect I think `&self` is better, for extra flexibility.
This PR reverts the changes from the rc3 update to inherent methods.
Update the start and stop flags for all read/write/read_write methods to match
those in the default blocking implementation of these methods (as well as
other RP2040 I2C implementations, and expected I2C behavior).
Also adds a write_read_async method that doesnt require using embedded-hal, as
this is required to use I2C in an idiomatic fashion (see TI Application Report
SLVA704).
The RP2040 datasheet indicates that I2C pins should have a limited
slew rate (Page 440 - 4.3.1.3.). This configures that for both
`I2c` and `I2cSlave`.
In addition, the pin configuration has been centralized to a single
fn.
They're heavily spamming logs for HIL tests, and I don't believe
they're valuable now that the thing they helped debug in their young
age is now solid and mature.
Move i2c to mod, split device and controller
Remove mode generic:
I don't think it's reasonable to use the i2c in device mode while
blocking, so I'm cutting the generic.
this is "generic" in that it doesn't require the user to set up anything
specific to go to dormant sleep, unlike the C sdk which requires clock
sources to be configured explicitly and doesn't much care about PLLs. we
will instead take a snapshot of the current clock configuration, switch
to a known clock source (very slow rosc, in this case), go to sleep, and
on wakeup undo everything we've done (ensuring stability of PLLs and
such).
tested locally, but adding tests to HIL seems infeasible. we'd need at
least another pico or extensive modifications to teleprobe since
dormant-sleep breaks SWD (except to rescue-dp), neither of which is
feasible at this point. if we *did* want to add tests we should check
for both rtc wakeups (with an external rtc clock source) and gpio wakeups.
this temporarily takes ownership of pins because we need to clear edge
interrupts before waiting for them (otherwise we may wait indefinitely),
we want to clean up the dormant-wake bits after a wakeup, and doing
anything *else* with the input while we're waiting for a wakeup isn't
possible at all. doing it like this lets us not impose any cost on those
who don't use dormant wakes without entangling dormant waits too badly
with regular interrupt waits.
with uniform treatment of adc inputs it's easy enough to add a new
sampling method. dma sampling only supports one channel at the moment,
though round-robin sampling would be a simple extension (probably a new
trait that's implemented for Channel and &[Channel]). continuous dma as
proposed in #1608 also isn't done here, we'd expect that to be a
compound dma::Channel that internally splits a buffer in half and
dispatches callbacks or something like that.
this lets us treat pins and the temperature sensor uniformly using the
same interface. uniformity in turn lets us add more adc features without
combinatorial explosion of methods and types needed to handle them all.
so far only bank0 interrupts were processed and configured, even if a
qspi pin was given. this is obviously not the right thing to do, so
let's rectify that. the fact that no problems have shown up so far does
suggest that most, if not all, applications don't use this functionality
at all.
this will be mostly not useful to anyone since flash is attached to
qspi, and using flash chips that don't use the *entire* qspi interface
will severly slow down the chip. the code overhead is minimal right now,
but if we also fix interrupt support on qspi pins this will
change (adding more code to potentially hot paths, using more memory for
wakers that are never used, and preventing the qspi gpio irq from being
used in software interrupts as RTIC applications may want to do).
we'll need access to the pin io bank registers for an upcoming fix, and
having both `io` and `io_bank` or similar can get confusing quickly.
rename `io` to `gpio` to avoid this, and also match the type while there.
this removed the RelocatedProgram construction step from pio uses.
there's not all that much to be said for the extra step because the
origin can be set on the input program itself, and the remaining
information exposed by RelocatedProgram can be exposed from
LoadedProgram instead (even though it's already available on the pio_asm
programs, albeit perhaps less convenient). we do lose access to the
relocated instruction iterator, but we also cannot think of anything
this iterator would actually be useful for outside of program loading.
Make sure that the ptr() function for ROM functions always returns
the actual ROM pointer. This allows the use of flash I/O while the
function cache is enabled.
using these will require some linker script intervention. setting the
core0 stack needs linker intervention anyway (to provide _stack_start),
having it also provide _stack_end for the guard to use is not that much
of a stretch.
the region field of the register is four bits wide followed by the valid
bit that causes the rnr update we rely on for the rasr write. 0x08 is
just a bit short to reach the valid bit, and since rp2040 has only 8
regions it (at best) doesn't do anything at all.
the adc constantly pulls a small but significant amount of current while
the hardware is enabled. this can have quite an effect on sleeping
devices that also use the adc.
Change embassy-rp i2c.rs impl of embedded_hal_async::i2c::I2c::transaction
to only do the call to setup() for address once per call to transactions.
Calling setup multiple times results in I2C transactions being skipped
on the bus, even across calls to transaction() or devices.
- don't require an irq binding for blocking-only adc
- abstract adc pins into an AnyPin like interface, erasing the actual
peripheral type at runtime.
- add pull-up/pull-down functions for adc pins
- add a test (mostly a copy of the example, to be honest)
- configure adc pads according to datasheet
- report conversion errors (although they seem exceedingly rare?)
- drop embedded-hal interfaces. embedded-hal channels can do neither
AnyPin nor pullup/pulldown without encoding both into the type
- Move typelevel interrupts to a special-purpose mod: `embassy_xx::interrupt::typelevel`.
- Reexport the PAC interrupt enum in `embassy_xx::interrupt`.
This has a few advantages:
- The `embassy_xx::interrupt` module is now more "standard".
- It works with `cortex-m` functions for manipulating interrupts, for example.
- It works with RTIC.
- the interrupt enum allows holding value that can be "any interrupt at runtime", this can't be done with typelevel irqs.
- When "const-generics on enums" is stable, we can remove the typelevel interrupts without disruptive changes to `embassy_xx::interrupt`.
1465: rp: continue clock rework r=Dirbaio a=pennae
vastly reduce the code size of initial clock config (over 700 bytes saved!), at the cost of about 48 bytes of ram used to store the frequencies of all clocks in the system. also stop exporting unstable pac items for clock config, fix a few settings that were out of spec, and add missing features (most notably gpin source information).
Co-authored-by: pennae <github@quasiparticle.net>
gpin clock sources aren't going to be very useful during cold boot and
thus require runtime clock reconfig. once we get there we can use this
for reference. or maybe we can't, only time will tell.
we'll take static ownership of an entire pin (not just a limited
reference), otherwise we cannot at all guarantee that the pin will not
be reused for something else while still in use. in theory we could
limit the liftime of this use, but that would require attaching
lifetimes to ClockConfig (and subsequently the main config), passing
those through init(), and returning an object that undoes the gpin
configuration on drop. that's a lot unnecessary support code while we
don't have runtime clock reconfig.
don't recalculate clock frequencies every time they are asked for. while
this is not very often in practice it does consume a bunch of flash
space that cannot be optimized away, and was pulled in unconditionally
previously. while we technically only need the configured rosc, xosc and
gpin frequencies it is easier to store all frequencies (and much cheaper
at runtime too).
if rosc really does run at 140MHz in high at div=1 then these values
were not correct and would've exceeded the chip spec. the HIL test
device seems to run fast (150MHz) so they're still not quite correct,
but rosc has high variance anyway so it's probably fine.
exposing pac items kind of undermines the unstable-pac feature. directly
exposing register structure is also pretty inconvenient since the clock
switching code takes care of the src/aux difference in behavior, so a
user needn't really be forced to write down decomposed register values.
the datasheet says that the xosc may be run by feeding a square wave
into the XIN pin of the chip, but requires that the oscillator be set to
pass through XIN in that case. it does not mention how, the xosc
peripheral does not seem to have any config bits that could be set to
this effect, and pico-sdk seems to have no (or at least no special)
handling for this configuration at all. it can thus be assumed to either
be not supported even by the reference sdk or to not need different
handling.
solvers usually output fbdiv directly, using vco_freq to get back to
fbdiv is not all that necessary or useful. both vco_freq and fbdiv have
hidden constraints, but vco_freq is a lot less accurate because the
fbdiv value resulting from the division may be off by almost a full
ref_freq's worth of frequency.
also fixes the usb pll config, which ran the pll vco way out of (below)
spec.
we might not configure both, so we should put the others into reset
state. leaving them fully as is might leave them running, which might
not be the goal for runtime reconfig (when it comes around). this now
mirrors how we reset all clock-using peripherals and only unreset those
that are properly clocked.
this is only really useful for runtime *re*configuration, which we don't
currently support. even runtime reconfig probably won't need it, unless
we keep taking the sledgehammer approach of reconfiguring everything all
the time.
It is UB to pass `entry` to core1 as `&mut`, because core0 keeps
an aliasing pointer to that memory region, and actually writes to
it (when `spawn_core1` returns, the stack frame gets deallocated and the memory
gets reused). This violates noalias requirements.
Added the fence just in case, een though it works without.
