17:25 <Kwiboo> CounterPillow: thanks for sending the rk356x pcie range fix, will runtime test and reply to list later this weekend :-)

17:26 <CounterPillow> \o/

17:32 <CounterPillow> I don't have a PCIe device that was known to be broken previously but I did hand a patched devicetree to someone and for now it seems to have improved things on the PCIe2 controller of the RK3566 for them

17:33 <Kwiboo> Not sure I have any broken or complex device, my main concern is related to the following issue in u-boot: https://patchwork.ozlabs.org/project/uboot/patch/20230517225337.2089287-6-jonas@kwiboo.se/

17:35 <Kwiboo> not sure if that will happen or is an issue in linux

17:36 <robmur01> I have a vague feeling that might be a relatively standard quirk

17:41 <robmur01> (or at least I'm convinced that I've seen at least one other controller where endpoint mode BARs were erroneously visible in root complex mode)

17:41 <Kwiboo> good to know, have seen similar workarounds for some other pcie drivers in uboot, was hoping linux handles this better in core or have a quirk flag ready

17:46 <robmur01> ah, found a pattern between pci-xgene and pcie-altera at least, trouble is there's probably about 17 other ways of doing it too :)

17:48 <CounterPillow> ... maybe I should re-image my ODROID M1 and check there that this doesn't break NVMe support at least :D

18:25 <CounterPillow> Kwiboo: for your odroid m1 tests, what did you do to make sure the vendor SPI u-boot doesn't run? It seems for me their SPL runs and then your rk35xx branch gets loaded and then fails to voltage select the SD card

18:33 <Kwiboo> CounterPillow: I usually press down on recovery button when connecting power or rebooting, then it loaded from sd-card or emmc

18:33 <CounterPillow> nice, thanks

18:34 <Kwiboo> it was some time since I tested on all devices, planning on re-testing all devices during weekend :-)

18:35 <CounterPillow> ok recovery button works, I just still get Card did not respond to voltage select! : -110 :(

18:37 <Kwiboo> strange, may be some important patch missing, I have tried to isolate rk356x patches in my rk3568-2023.07 branch

18:37 <CounterPillow> Okay, I'll try building that

18:38 <Kwiboo> ahh, yes may have been my latest rk35xx have some issue with unbalanced regulators in mmc core

18:40 <CounterPillow> rk3568-2023.07 fails in the same way

18:40 <CounterPillow> (this is with just odroid-m1-rk3568_defconfig)

18:41 <Kwiboo> strange, will give it a try on my odroid-m1

18:49 <CounterPillow> since I think you got USB to work I can just use a USB root device, so I'll try that :)

18:59 <Kwiboo> I did not get any sd voltage select error with my emmc or sd-card, using a sandisk ultra 32gb a1 card, SDSQUA4-032G-GN6MA, will try to locate some other sd-cards to test with

19:02 <Kwiboo> I am using latest ddr v1.16 and bl31 v1.42 blobs, if that could make any difference

19:05 <Kwiboo> I do see some, USB-error: DEVICENOTRESPONDING: Device did not respond to token (IN) or did not provide a handshake (OUT), that I have not seen before, possible related to fixed/gpio regulator ref count feature missing

19:25 <CounterPillow> I'm using a SanDisk Max Endurance 64GB card

19:27 <CounterPillow> kernel loads from USB though (but my kernel config is missing something, running into a [ 1.912236] /dev/root: Can't open blockdev panic)

19:42 <Kwiboo> tested with different sd-cards: sandisk ultra class 10, samsung pro endurance and samsung evo, not seeing any -110 error, did however see a -110 error when booting into linux from emmc

19:43 <Kwiboo> using extlinux boot method so see: Booting bootflow 'mmc@fe2b0000.bootdev.part_1' with extlinux, when loading kernel

19:44 <Kwiboo> CounterPillow: are you using extlinux, boot script or efi to load kernel?

19:54 <CounterPillow> extlinux

20:17 <CounterPillow> Okay, managed to use a USB root to test the PCIe ranges patch on an ODROID M1, works fine with an NVMe drive inserted.

20:22 <diederik> Later tonight or this weekend I'll send my Tested-By as it's the exact same patch I've been using for a while now