15:20 <nashpa> hello, does anyone have a recipe for fixing firefly-rk3399 PCIe link training timeout (-110) error?

15:21 <nashpa> does this point to my hardware being broken?

15:21 <nashpa> (I'm using mainline U-Boot and linux-next)

15:37 <mmind00> nashpa: good question ... I think there was a problem with pcie gen2 link training having issues in some cases, but I thought there was already a patch limiting that ... let me check

15:40 <mmind00> nashpa: https://patchwork.kernel.org/patch/9477649/ ... and just verified that I did apply it

15:41 <mmind00> nashpa: basic problem was that the pcie can have link-training issues in gen2 with some devices, while other people reported that it works actually nicely ... so we defaulted to gen1 ... so I guess link-training should actually work nicely

15:41 <mmind00> nashpa: can you try some older kernel versions to check if the problem is present on them as well?

15:41 <mmind00> just to check if that is a newly introduced regression

15:42 <nashpa> mmind00: I have that change

15:42 <nashpa> mmind00: let me try a different question then: should the PCIe training work without any device being plugged in?

15:43 <nashpa> mmind00: and if not, does the number of lines in the DTS have to match what the connected device has?

15:44 <nashpa> mmind00: I don't know which older kernel version I should try, the one that Firefly provides in their image doesn't work either (times out as well)

15:44 <mmind00> nashpa: I'm really not that versed in pcie stuff ... I do have one of the m.2 -> sata adapters in my firefly-rk3399 board and it is working

15:44 <mmind00> nashpa: is something actually in your pcie port

15:44 <mmind00> ?

15:44 <nashpa> mmind00: hmm, I wonder if there is any magic in that adapter.

15:44 <nashpa> mmind00: I've put a Toshiba RC100 NVME SSD

15:44 <mmind00> i.e. I'd think link training should fail if nothing is in the slot

15:45 <nashpa> which I think it is an x2 PCIe 3.1 device (hopefully compatible with the PCIe from RK3399 as well)

15:47 <mmind00> nashpa: I'm really not sure, but if the vendor kernel is also not working, and also something like 4.12 or so ... I'd guess there may be some sort of issue with the ssd itself, but as I said, I'm really not sure

15:48 <mmind00> nashpa: I only have the sata adapter in my firefly and a full pcie-usb3 card in my rk3399-puma

15:48 <nashpa> mmind00: OK, thanks for your time, I'll try to see if I can get some alternative to that SSD that matches closer to what Firefly-RK3399 expects

15:49 <nashpa> mmind00: I'm waiting on the Renegade board to show up to have another RK3399, wanted to use the Firefly to prepare for that

15:49 <mmind00> nashpa: glad to help or at least try to :-)

15:50 <nashpa> I've managed to get a fully open-source stack after a lot of pain trying to recover a bad emmc bootloader

15:50 <nashpa> but now I have a method of recovering

15:50 <mmind00> :-)

15:51 <nashpa> unfortunately you still need to use the rk3399_loader_v1.08.106.bin to flash the SPL, U-Boot doesn't seem to be capable of writing to the eMMC

15:51 <nashpa> it can erase it just fine, but nothing else, env in eMMC is a no-go :-(

16:34 <Esmil> Did anyone look at the mwifiex scheduling while atomic bug in 4.19 on kevin? It seems I can avoid them by reverting 5188d5453bc9380ccd4ae1086138dd485d13aef2 and 5631909364e1e74b6188ec860d2a4cf216150a26

20:40 <beeble> nashpa: link training will fail if nothing is attached

20:41 <beeble> devices with a bridge inside will fail often

20:41 <beeble> or mostly

20:42 <beeble> but that would end with a kernel crash in most cases

20:43 <beeble> so at least link training should work if everything in your pcie setup is correct

21:17 <nashpa> beeble: yeah, I'm not sure if the SSD I have (PCIe 3.0) is capable of downgrading to RK3399 PCIe 2.1

21:20 <nashpa> beeble: I've ordered from Aliexpress the PCIe M.2 to SATA adapter from firefly, when it arrives I should be able to validate my HW