Fixes polling thread looping forever hangs preventing new connections.
Don't lose listening sockets on mp resets and re-init.
Keep better separation of "system" and "user" sockets.
Track socket states to prevent re-use of sockets before closed.
Close REST socket when transaction completes. No post-init.
Remove unnecessary state flags.
This reverts commit 7e6e824d56.
Fixes#7770
The change in #7623 needs to be revered; the raise-site added in #7632
is the correct one and the one in socketpool needs to be reverted.
This is not affecting 8.0.x because #7623 was not back-ported to there
before we realized it was not a full fix.
Both #7770 and #7606 should be re-tested. I didn't test.
Based on Feather RP2040 SCORPIO, but new USB PID (from MBAdafruitBoards/Production) and pin name changes for GP16-23 (CKN/CKP etc., from board schematic).
It now handles deinit, never_reset and sharing tracking. PWM
now runs in the WAIT state as well during a time.sleep().
_reset_ok() was removed because it was called in one spot right
before deinit().
Some PWMOut were also switched to a bitmap for use instead of
reference count. That way init and deinit are idempotent.
Fixes#6589. Fixes#4841. Fixes#4541.
Rather than returning the negative error value.
This is intended to close#7606, though I did not test with mqtt.
Instead, I created a simple standalone test program:
```python
import wifi, socketpool, ssl, time
#wifi.radio.connect(<omitted>)
import socketpool
socket = socketpool.SocketPool(wifi.radio)
ctx = ssl.create_default_context()
b = bytearray(8)
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sss = ctx.wrap_socket(s, server_hostname='example.com')
sss.connect(('example.com', 443))
sss.setblocking(False)
r = sss.recv_into(b)
print(r, b) # prints 4294967285 which is -11 as unsigned
sss.close()
```
Before the change, r was the out of range value 4294967285. After the
change, the recv_into call raises OSError instead.
This is comparable to the behavior on standard Python, though an
SSLWantReadError is raised instead.
The original (mis)behavior seems to match what was uncovered deep inside
minimqtt by adding logging:
```
370.578: DEBUG - PKT: _sock_exact_recv: recv_len = 4294967285
```
This 2-in-1 PR started with the goal of support the Bangle.js 2
smartwatch with *no USB*.
* Adds "secure" DFU build support with a committed private key.
* Adds 3-bit color support with one dummy bit for the JDI memory display
* Allows nrf boards to have a board_background_task() run in RUN_BACKGROUND_TASK.
This is needed because the Bangle.js 2 uses the watchdog to reset.
* Renamed port_background_task() to port_background_tick() to indicate it
runs on tick, not RUN_BACKGROUND_TASK.
* Marks serial connected when the display terminal is inited. This means
that safe mode messages show up on the display.
ACep, 7-color epaper displays also pack 3 bits in 4. So, I added that
support as well.
* Adds 3-bit ACeP color support for 7-color e-paper displays. (Not
watch related but similar due to color depth.)
* Allows a refresh sequence instead of a single int command. The 7" ACeP
display requires a data byte for refresh.
* Adds optional delay after resetting the display. The ACeP displays
need this. (Probably to load LUTs from flash.)
* Adds a cleaning phase for ACeP displays before the real refresh.
For both:
* Add dither support to Palette.
* Palette no longer converts colors when set. Instead, it caches
converted colors at each index.
* ColorConverter now caches the last converted color. It should make
conversions faster for repeated colors (not dithering.)
Watchdogs are used to reboot out of the bootloader. There is a
scratch register for user watchdogs. So use sdk functions to better
distinguish these.
Related to #7346
The comment says it is `buffer[start:end]` but it assumed elements
were a single byte long. Now it correctly does multibyte elements
from array.array.
Fixes#4988
Now, open wifi works on Pico W, so this closes#7438.
For the web workflow it's now OK to either
* specify an empty password string (as before)
* not have the CIRCUITPY_WIFI_PASSWORD key at all (new functionality)
reset_port calls reset_all_pins, which disables the built-in pull
down. In theory, this could allow some CYW43 interfacing pin to float
to an inappropriate value. There's no proof of this! but the move can't
really hurt, either
* read() is now readinto() and takes the buffer to write into.
* readinto() returns the number of valid samples.
* readinto() can be interrupted by ctrl-c.
* readinto() API doesn't support signed numbers because it never did.
* sample_rate is now required in the constructor because supported
values will vary per-port.
* 16 bit values are full range. 12 bit samples from RP2040 are stretched
in the same way they are for AnalogIn.
Fixes#7226
This needs thorough testing before it's merged, as we tried
and reverted this once before (#5341 and #5356).
I think that besides checking for tinyusb having "something to do",
the fact that `port_interrupt_after_ticks` and `port_disable_tick`
weren't implemented that was causing a secondary problem.
I've tested this on a pico w over reboot-cycles and ctrl-c-cycles,
with and without drive automounting, with and without serial repl open,
and on a power-only connection.
I didn't notice the problem reported in #5356 after merely implementing
port_idle_until_interrupt; but I did notice that sleeps in general would
take over-long until "something" (like writing to the USB drive) happened;
I think "something" was probably calling port_enable_tick(). When this
problem was happening, sleeps would take a lot longer; for instance,
`sleep(.001)` would take about 1/20s and `sleep(.1)` would take about 1/7s.
Because this must be treated like an in-use pin for all other purposes,
unfortunately a special case must be added in shared-bindings.
Multiple AnalogIn objects for VOLTAGE_MONITOR can be created (because
in use tracking isn't working) but this causes no harm.
Testing performed: Read the monitor, then imported wifi. When the
pin state was insufficiently restored, the second step would fail
with debug messages about do_ioctl timeout.
```
import analogio, board
a = analogio.AnalogIn(board.VOLTAGE_MONITOR)
print(a.value)
import wifi
```
Closes: #7020
This is the lwip no-os version of SO_REUSEADDR, which is set on all
listening sockets in the espressif port; do so here as well,
it makes running servers easier. The "address in use" error does
not occur.
You might wonder how this fixes a problem with PulseIn, when the
changes aren't to any of those files! PulseIn is implemented in terms of
StateMachine, which had some assumptions about the relation between
the index of a pin object in mcu_pin_global_dict_table and its "pin
number". This was true, until some pins were removed from the
microcontroller module on Pico W.
Closes: #7078
Weirdly we have to stop the AP too (which we never started),
or cyw43_tcpip_link_status still reports that STA is connected.
As long as AP mode isn't implemented, this doesn't matter and
we can just do it.
Foamyguy discovered that trying to send >2920 bytes at once consistently
failed. I further discovered that sometimes trying to send >1460 bytes
would fail too. By "fail", I mean that it would take a very long time
(around 200 * 50ms) before erroneously reporting that all bytes were
written.
In my testing, this change causes larger writes to successfully
send either 2920 or 1460 bytes (possibly after doing some 50ms waits
for a previous packet to clear).
The documentation of socket.send always stated that it COULD send fewer
bytes than requested, but adafruit_httpserver assumed that the number
of requested bytes were always sent, so after this change alone,
adafruit_httpserver will still not work properly.
Closes: #7077 (albeit fixes are needed in adafruit_httpserver too)
This reduces power consumption during true deep sleep.
In my measurements with ppk2 and a program that _irrevocably_ entered
deep sleep (no time alarm or pin alarm), power usage as measured on a
ppk2 decreased from ~10mA to ~1mA.
The prefixed versions raise Python exceptions, the un-prefixed return
negative error values. We don't want to raise an exception from here,
it leaves the SSL stack in an undefined state.
## Testing self-signed certificates and `load_verify_locations`
Obtain the badssl "self-signed" certificate in the correct form:
```sh
openssl s_client -servername self-signed.badssl.com -connect untrusted-root.badssl.com:443 < /dev/null | openssl x509 > self-signed.pem
```
Copy it and the script to CIRCUITPY:
```python
import os
import wifi
import socketpool
import ssl
import adafruit_requests
TEXT_URL = "https://self-signed.badssl.com/"
if not wifi.radio.ipv4_address:
wifi.radio.connect(os.getenv('WIFI_SSID'), os.getenv('WIFI_PASSWORD'))
pool = socketpool.SocketPool(wifi.radio)
context = ssl.create_default_context()
requests = adafruit_requests.Session(pool, context)
print(f"Fetching from {TEXT_URL} without certificate (should fail)")
try:
response = requests.get(TEXT_URL)
except Exception as e:
print(f"Failed: {e}")
else:
print(f"{response.status_code=}, should have failed with exception")
print("Loading server certificate")
with open("/self-signed.pem", "rb") as certfile:
context.load_verify_locations(cadata=certfile.read())
requests = adafruit_requests.Session(pool, context)
print(f"Fetching from {TEXT_URL} with certificate (should succeed)")
try:
response = requests.get(TEXT_URL)
except Exception as e:
print(f"Unexpected exception: {e}")
else:
print(f"{response.status_code=}, should be 200 OK")
```
Tested with badssl.com:
1. Get client certificates from https://badssl.com/download/
2. Convert public portion with `openssl x509 -in badssl.com-client.pem -out CIRCUITPY/cert.pem`
3. Convert private portion with `openssl rsa -in badssl.com-client.pem -out CIRCUITPY/privkey.pem` and the password `badssl.com`
4. Put wifi settings in CIRCUITPY/.env
5. Run the below Python script:
```py
import os
import wifi
import socketpool
import ssl
import adafruit_requests
TEXT_URL = "https://client.badssl.com/"
wifi.radio.connect(os.getenv('WIFI_SSID'), os.getenv('WIFI_PASSWORD'))
pool = socketpool.SocketPool(wifi.radio)
context = ssl.create_default_context()
requests = adafruit_requests.Session(pool, context)
print(f"Fetching from {TEXT_URL} without certificate (should fail)")
response = requests.get(TEXT_URL)
print(f"{response.status_code=}, should be 400 Bad Request")
input("hit enter to continue\r")
print("Loading client certificate")
context.load_cert_chain("/cert.pem", "privkey.pem")
requests = adafruit_requests.Session(pool, context)
print(f"Fetching from {TEXT_URL} with certificate (should succeed)")
response = requests.get(TEXT_URL)
print(f"{response.status_code=}, should be 200 OK")
```
Closes: #7017
* Remove the 'GP23' alias for CYW1
* Remove the 'CYW0' alias for CYW0
* Switch VBUS_SENSE to CYW2, remove 'GP24' alias
Code that wants to use SMPS_MODE, VBUS_SENSE and LED while being
portable to the W and non-W variants should use those names, not alias
names.
* Remove A3 / VOLTAGE_MONITOR
Right now this cannot be used. The ability to check the voltage monitor
should be added back in some fashion in the future.
This is intended (but not entirely verified) to match our esp32 builds.
It does fix accessing https://circuitpython.org, which failed before with
"MBEDTLS_ERR_SSL_FATAL_ALERT_MESSAGE".
It still doesn't work on a personal website of mine with valid letsencrypt
certificate but I haven't verified whether it works on esp32s2 with CP.
That site only allows TLS 1.3, while this mbedtls only supports up to
1.2.
The version of mbedtls we adopted based on micropython's use has no
TLS 1.3 support, but the one in espressif esp-idf does.
Note: at this time, the ssl module on pico_w never verifies the server
certificate. This means it does not actually provide a higher security
level than regular socket / http protocols.
Before this, CIRCUITPY would start at 1MB anyway. This appeared to work
only because I hadn't checked the actual size of the CIRCUITPY drive,
and because until now the flash hadn't actually crossed that 1MB
boundary into CIRCUITPY storage.
WARNING: on pico_w, upgrading/downgrading CircuitPython across this commit
boundary will erase the CIRCUITPY filesystem. After this commit,
switching between pico and pico_w firmware will erase the CIRCUITPY
filesystem
.. the value actually needs to be enforced each time the STA or AP
is enabled, because internally there's a call to cyw43_wifi_pm with the
library's defaut power management value, not ours.
Add a getter, though it only returns our idea of what the power
management register is set to, it doesn't read out from the actual
hardware, sadly.
Originally, black_bindings found each contiguous "//|" block and sent
it to black independently. This was slower than it needed to be.
Instead, swap the comment prefix: when running black, take off
"//|" prefixes and put "##|" prefixes on all un-prefixed lines.
Then, after black is run, do the opposite operation
This more than doubles the overall speed of "pre-commit run --all",
from 3m20s to 55s CPU time on my local machine (32.5s to under 10s
"elapsed" time)
It also causes a small amount of churn in the bindings, because
black now sees enough context to know whether one 'def' follows another
or ends the 'def's in a 'class'. In the latter case, it adds an extra
newline, which becomes a "//|" line.
I'm less sure why a trailing comma was omitted before down in
rp2pio/StateMachine.c but let's roll with it.
My pings go out, and then they come back
```py
import os
import wifi
import ipaddress
wifi.radio.connect(os.getenv('WIFI_SSID'), os.getenv('WIFI_PASSWORD'))
ipv4 = ipaddress.ip_address("8.8.4.4")
print("Ping google.com: %f ms" % (wifi.radio.ping(ipv4)*1000))
```
