Ethernet driver events properly bounded with ESP NETIF actions to support multiple Ethernet modules used at a time.
Components using Ethernet updated to conform with new API.
Closes https://github.com/espressif/esp-idf/issues/7318
Increase the interrupt reassert level timing so the chances of missing
two consecutive events are minimal.
Enable only SIR_RECV interrupt event, so the SEND events are not used
for GPIO signal.
If the GPIO interrupt is re-asserted too quickly it could be missed. If this happens the driver goes silent and never receives any data. Recover by periodic checks of the IO signal level
Also updated KSZ8851SNL per internal code review:
* Removed Link status change interrupt as it's handled with polling
* Added auto negotiation timeout
* Updated typedefs, moved types to appropriate source, updated
components/esp_eth/src/ksz8851.h to use only inherent device types
* Applied IDF code formatting
* Updated header file order to include first the most generic to more
specific
The issue typically happens for link-down during Tx. Added two retry levels, one before checking the sanity of the w5500 phy register and another for the Tx done itself (if the device is in the sane state)
Closes https://github.com/espressif/esp-idf/issues/6233
Reading SPI data may come in 4-byte units and thus result in unwanted
overwrites if smaller size registers read, especially if multiple placed
one after another. Fixed by using direct reads to `trans` structure for
sizes smaller or equal to 4.
Closes https://github.com/espressif/esp-idf/issues/6579
When `DIS_USB_JTAG` eFuse is NOT burned (`False`), it is not possible
to set pins 18 and 19 as GPIOs. This commit solves this by manually
disabling USB JTAG when using pins 18 or 19.
The functions shall use `gpio_hal_iomux_func_sel` instead of
`PIN_FUNC_SELELECT`.
W5500 features an Ethernet MAC+PHY, plus a HW SOCKET implementation.
But in IDF, we only use software TCP/IP stack, which means we have
to bypass the HW TCP/IP stack in W5500.
So we enables the MAC RAW mode in W5500.
* Target components pull in xtensa component directly
* Use CPU HAL where applicable
* Remove unnecessary xtensa headers
* Compilation changes necessary to support non-xtensa gcc types (ie int32_t/uint32_t is no
longer signed/unsigned int).
Changes come from internal branch commit a6723fc
