Ok, so now we are down to one target left that has not had 6.12 approved as a testing kernel, due to lack of basic testing and debugging: at91. That puts it at risk of being dropped entirely from OpenWrt. @claudiubeznea has said he may test it next weekend, but can’t promise after that.
Is there anyone else with at91 hardware who is willing to test and debug 6.12? Here is a list of at91 devices:
Good news! @claudiubeznea tested at91, and then @Ansuel looked over the output and quickly merged it. Now every target in OpenWrt supports 6.12 at least as a testing kernel! Thanks for working on this, Goetz.
~84% of targets in OpenWrt's main branch now support 6.12 by default. There are only seven targets left to test. Good progress!
The Siflower targets link goes to an empty template page but I’m aware of recent commits by BPI-SINOVOIP for 2 riscv64 based devices. Some, but not all, kernel 6.6 patches have been accepted by the MainStream Linux kernel.
There was an earlier suggestion to drop support on the basis of being “only a few devices”.
I think any new architectures, unless they have a ton of money behind them, would be excluded from the OpenWRT development teams and the community at large.
Specifically, riscv64 based devices have opensource code bases.
So today I got to testing 6.12 on BPi-RV2, and while it works generally, there seem to be a lot of rough edges around. I'm not sure whether these are 6.12 specific (probably not), but here are my findings:
GbE LEDs don't really work (both eth5 and the switch's eth0-eth3); 2.5GbE LEDs are fine.
MT7925 PCIe card is not detected.
I2C driver is not compiled in by default; after adding CONFIG_PACKAGE_kmod-i2c-designware-platform=y to the config file, it works, but (see next point)
PCF8563 driver fails to read RTC:
# hwclock
[ 179.184298] rtc-pcf8563 0-0051: low voltage detected, date/time is not reliable.
hwclock: RTC_RD_TIME: Invalid argument
GPIO 30 seem to be wrong here, since the schematic has it as RGMII_RXD1, whereas GPIO 22 is connected to P1240 reset.
# cat /sys/kernel/debug/gpio
gpiochip0: GPIOs 512-552, parent: platform/c800000.gpio, gpio_siflower:
gpio-513 ( |PHY reset ) out hi ACTIVE LOW
gpio-517 ( |spi0 CS0 ) out hi ACTIVE LOW
gpio-521 ( |reset ) out hi ACTIVE LOW
gpio-531 ( |reset ) out hi ACTIVE LOW
gpio-542 ( |reset ) out lo ACTIVE LOW
gpio-548 ( |spi0 CS1 ) out hi ACTIVE LOW
gpio-549 ( |regulator@0 ) out hi
gpio-550 ( |regulator@1 ) out hi
gpio-551 ( |PHY reset ) out hi ACTIVE LOW
gpio-552 ( |interrupt ) in hi IRQ
FYI, https://openwrt.org/docs/techref/targets/bcm53xx does not contain the R8500, so I did not pay attention to it earlier on. But I saw the R8500 in the makefile, so I just did a quick compile and flash to the R8500. Left a comment in the github ticket.
3v, I obviously checked.
Moreover, RTC gets fed from CPU_3V3 when the device is on (which happen to be the case for the log message I quoted...) so low voltage on 3V3 rail seems implausible.
Respectfully, I feel it is curious that the one guy testing literally the last device for 6.12 is not satisfied with 6 days of uptime and wants 2 weeks.
Meanwhile, all sorts of things get approved with little or possibly no testing, sometimes breaking things for large portions of the community. Which is actually fine; that's just how development works.
Could have been said with more courtesy. Any good will is welcome, even if asking for extra cautiousness. That being said I agree that 6 days of uptime should be enough, assuming nothing serious appears in the log.
