Do this by tracking being inside gc collection with a
separate flag, GC_COLLECT_FLAG. In gc_free(),
ignore this flag when determining if the heap is locked.
* For finalisers calling gc_free() when heap is otherwise unlocked,
this allows memory to be immediately freed (potentially
avoiding a MemoryError).
* Hard IRQs still can't call gc_free(), as heap will be locked via
gc_lock().
* If finalisers are disabled then all of this code can be compiled
out to save some code size.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Currently a finalizer may run and access memory which has already been
freed. (This happens mostly during gc_sweep_all() but could happen during
any garbage collection pass.)
Includes some speed improvement tweaks to skip empty FTB blocks. These help
compensate for the inherent slowdown of having to walk the heap twice.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
This allows coverage to test MicroPython-specific features such as
the tracked alloc cleanups added in the parent commit.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
On mbedTLS ports with non-baremetal configs (mostly esp32, technically also
unix port), mbedTLS memory is allocated from the libc heap. This means an
old SSL socket may be holding large SSL buffers and preventing a new SSL
socket from being allocated.
As a workaround, trigger a GC pass and retry before failing outright.
This was originally implemented as a global mbedTLS calloc function, but
there is complexity around the possibility of C user modules calling into
mbedTLS without holding the GIL. It would be interesting to try making a
generic version for any malloc which fails, but this would require checking
for a Python thread and probably making the GIL recursive.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Test is for an issue reported on the micropython-lib Discord as
effecting the rp2 port umqtt.simple interface when reconnecting with TLS,
however it's a more generic problem.
Currently this test fails on RPI_PICO_W and ESP32_GENERIC_C3 (and no doubt
others). Fixes are in the subsequent commits.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Follow up to 13b13d1fdd, based on some
testing on godbolt, the manual code optimisation seems unnecessary for code
size, at least on gcc x86_64 and ARM, and it's definitely not good for
clarity.
Signed-off-by: Jeff Epler <jepler@gmail.com>
This lwIP configuration file has options that are common to all ports, and
the ports are updated to use this file. This change is a no-op, the lwIP
configuration remains the same for the four ports using this common file.
This reduces code duplication, keeps the ports in sync, and makes it easier
to update the configuration for all ports at once.
Signed-off-by: Damien George <damien@micropython.org>
The way CMake gathers the submodule list, it can quietly be empty
if the previous step fails. This makes it an explicit error.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Similar to other places, CMake will error out if this file
doesn't exist yet but we don't want this if we're only getting
the list of submodules.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
- ECHO_SUBMODULES=1 exits CMake early. With idf_component_manager 1.x this
seems to leave the managed_components directory in a state that causes
later builds to fail.
- Looks like the component manager isn't needed for this step, so disable
it. This invocation logs a warning (not visible in normal output) but
completes successfully and returns the correct list of submodules.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
For IPv6 connections, the peer address was previously defined as only the
first four bytes of the IP address. For IPv6 addresses, this resulted in an
incorrect IPv4 address. For instance, receiving a packet via `::recvfrom`
from `'fe80::87:e7ff:fe48:629a'` is returned as having a peer address of
`'254.128.0.0'`
Signed-off-by: Jared Hancock <jared.hancock@centeredsolutions.com>
And use it in `enter_raw_repl()`. This prevents waiting forever for a
serial device that does not respond to the Ctrl-C/Ctrl-D/etc commands and
is constantly outputting data.
Signed-off-by: Hans Maerki <buhtig.hans.maerki@ergoinfo.ch>
If the target does not return any data then `read_until()` will block
indefinitely. Fix this by making the initial read part of the general read
look, which always checks `inWaiting() > 0` before reading from the serial
device.
Also added the UART timeout to the constructor. This is not currently used
but may be used as an additional safeguard.
Signed-off-by: Hans Maerki <buhtig.hans.maerki@ergoinfo.ch>
This commit introduces the ability to emit optimised code paths on
Xtensa for load/store operations indexed via an immediate offset.
If an immediate offset for a load/store operation is within a certain
range that allows it to be embedded into an available opcode then said
opcode is emitted instead of the generic code sequence.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This commit improves the RV32 code sequence that is emitted if a
function needs to set up an exception handler as its prologue.
The old code would clear a temporary register and then copy that value
to places that needed to be initialised with zero values. On RV32
there's a dedicated register that's hardwired to be equal to zero, which
allows us to bypass the extra register clear and use the zero register
to initialise values.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This commit improves the emitted code sequences for address generation
in the Viper subsystem when loading/storing 16 and 32 bit values via a
register offset.
The Xtensa opcodes ADDX2 and ADDX4 are used to avoid performing the
extra shifts to align the final operation offset. Those opcodes are
available on both xtensa and xtensawin MicroPython architectures.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This includes making int("01") parse in base 10 like standard Python.
When a base of 0 is specified it means auto-detect based on the prefix, and
literals begining with 0 (except when the literal is all 0's) like "01" are
then invalid and now throw an exception.
The new error message is different from CPython. It says e.g.,
`SyntaxError: invalid syntax for integer with base 0: '09'`
Additional test cases were added to cover the changed & added code.
Co-authored-by: Damien George <damien@micropython.org>
Signed-off-by: Jeff Epler <jepler@gmail.com>
The esp32 IDF toolchain can give a "may be used uninitialized" warning, at
least for ESP32-S3 with gcc 14.2.0. Silence that warning by initializing
the variable with NULL.
Co-authored-by: Daniel van de Giessen <daniel@dvdgiessen.nl>
Signed-off-by: IhorNehrutsa <Ihor.Nehrutsa@gmail.com>
Add WIFI_AUTH_WPA3_ENTERPRISE and WIFI_AUTH_WPA2_WPA3_ENTERPRISE, and
update PPP callback signature for latest lwIP.
Co-authored-by: Daniel van de Giessen <daniel@dvdgiessen.nl>
Signed-off-by: IhorNehrutsa <Ihor.Nehrutsa@gmail.com>
The QSPI baud is derived from the AHB clock, not from the APB (peripheral)
clock.
Datasheet: The QSPI Baud rate clock is generated by dividing the module
clock (CLK_QSPI_AHB) by a value between 1 and 255.
As previously implemented, all baudrates are 2.5 times greater than
expected.
Signed-off-by: I. Tomita <ziceva@me.com>
Improvements to DAC support for SAMD51:
- properly validate DAC id
- correctly use dac_init flag, as a 2-ple for A0, A1 channels
- disable DAC before adjusting settings, see SAMD5x data sheet §47.6.2.3
Co-authored-by: robert-hh <robert@hammelrath.com>
Signed-off-by: Graeme Winter <graeme.winter@gmail.com>
This commit fixes compilation for the ESP8266 port when using a local
toolchain on relatively recent Linux systems.
The documentation asks the user to delete the esptool instance that
comes with the toolchain, in favour of using the one provided by the
system. On Linux systems that are at least two years old (looking at
the CI Ubuntu image as an example), the version of esptool installed
with the package manager isn't called `esptool.py` but just `esptool`.
The Makefile didn't take that into account and used `esptool.py` without
checking if such a command exists, making builds fail.
Now preference is given to the `esptool` command, falling back to
`esptool.py` only if the former command does not exist or it is not
available to the current user, to maintain compatibility with old
setups.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This commit moves `<board>.mk` to `<board>/mpconfigboard.mk` for all qemu
boards, making it the same as other bare-metal ports.
Signed-off-by: Damien George <damien@micropython.org>
The user already has it open, and its customised for their
particular board.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
The previous deploy.md refactors revealed that these boards had a different
"product" entry in boards.json compared to the name given in the board.md
file.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Includes fixing the flashing address for newer SoCs, as reported in
discussion https://github.com/orgs/micropython/discussions/16417
Also removes some redundant or out of date information, and adds links to
the Espressif esptool docs which are quite comprehensive.
Information about ESP32_GENERIC variants is moved to the board page, as it
only applies to that board.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Allows two source files (ports/esp32/boards/deploy.md and
deploy_nativeusb.md for boards with only native USB) for all esp32
installation steps, with templated chip name and flash offset inserted via
string formatting.
The new files add more text to explain the esptool.py port auto-detection,
remove the unnecessary -z feature (already enabled by default), and add
a bit of troubleshooting and port detection info.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
The memory bank addresses used for these are independent, can (and must)
enable both.
Also looks like no need to shrink these if FDCAN2 is added, the Reference
Manual is a bit unclear but looks like the peripheral's RAM multiplies out
for each additional controller.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
This is redundant for bxCAN, but for CAN-FD with BRS it's otherwise unclear
which set of parameters (baudrate & sample_point or brs_baudrate &
brs_sample_point) failed to match. This makes finding a valid combination
extra annoying.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Not every baudrate or sample point combination has an exact match,
but getting within 1% on sample point and .1% on baud rate should
always be good enough.
Because the search goes from shorter bit periods (lowest brp) and
increases, the first match which meets this criteria should still mostly be
the best available.
This work was funded through GitHub Sponsors.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Previously micros with the 'FDCAN' peripheral (as opposed to the older
'CAN' peripheral) needed to rename these pins in the CSVs for the CAN
driver to work.
The following CSVs in MicroPython still had FDCAN in them:
$ rg -t csv -l FDCAN boards
boards/stm32h7b3_af.csv
boards/stm32h743_af.csv
boards/stm32h573_af.csv
boards/stm32h723_af.csv
boards/stm32g0b1_af.csv
Confirmed that this allows CAN to work on NUCLEO_H723ZG board, and that at
least one board based on each of the other chips can still compile. Some of
these boards could possibly have MICROPY_HW_ENABLE_CAN set and work, now.
Signed-off-by: Angus Gratton <angus@redyak.com.au>
Reserved and static SPI buses must remain initialized during a soft reboot
as they may be used for SPI flash storage or XIP.
Signed-off-by: iabdalkader <i.abdalkader@gmail.com>
A board should make this return true if the specified SPI instances should
not be deinitialized on soft-reboot.
Signed-off-by: iabdalkader <i.abdalkader@gmail.com>
This commit re-introduces `tests/extmod/vfs_rom.py` and
`tests/float/math_domain.py` to the test suite, as the issues that made
them fail have now been addressed.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This commit fixes code generation for loading halfwords using an offset
greater than 255.
The old code blindly encoded the offset into a `LDRH Rd, [Rn, #imm]`
opcode, but only the lowest 8 bits would be put into the opcode itself.
This commit instead generates a two-opcodes sequence, a constant load into
R8, and then `LDRH Rd, [Rn, R8]`.
This fixes `tests/extmod/vfs_rom.py` for the qemu/SABRELITE board.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This commit fixes code generation for loading a local's address if its
index is greater than 63.
The old code blindly encoded the offset into an `ADD Rd, Rn, #imm` opcode,
but only the lowest 8 bits would be put into the opcode itself. This
commit instead generates a two-opcodes sequence, a constant load into R8,
and then an `ADD Rd, Rn, R8` opcode.
This fixes `tests/float/math_domain.py` for the qemu/SABRELITE board.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
This commit increases the GC heap size from 120KiB to 140KiB, as it is
needed to make the full test suite pass on SABRELITE when ran through the
armv6 native emitter.
This is needed as the code output by the armv6 native emitter is limited to
4-bytes opcodes and thus takes more space than other ARM emitters.
To keep things aligned, the RV32 port also got its heap size increased even
though it is not needed on that platform right now.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
The Micro:Bit machine definition in Qemu has soft MMU support enabled,
which is currently not compatible with the way MicroPython generates code
that needs to call back into non-emitted code.
As a stop-gap solution, the native code emitter for the MICROBIT board is
turned off.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
When a CPU exception is raised when emulating a Thumb-capable processor,
the default exception handler would simply enter in an endless loop without
providing any further information.
This commit adds a more complete exception handler that dumps to STDOUT the
exception cause and the status of the registers at the moment of the
exception.
Signed-off-by: Alessandro Gatti <a.gatti@frob.it>
