Problem with meson-pcie patch

Hi @numbqq and @Frank

Seems there is a problem (at least for myself ) with a patch introduced by commit in Fenix build system

If I build a kernel with this patch I’m getting the following messages in my kernel log

    0.481935] dw-pcie fc000000.pcie: IRQ index 1 not found
[    0.484338] meson-pcie fc000000.pcie: host bridge /soc/pcie@fc000000 ranges:
[    0.484369] meson-pcie fc000000.pcie:       IO 0x00fc600000..0x00fc6fffff -> 0x0000000000
[    0.484383] meson-pcie fc000000.pcie:      MEM 0x00fc700000..0x00fdffffff -> 0x00fc700000
[    0.484454] meson-pcie fc000000.pcie: invalid resource
[    0.486713] meson-pcie fc000000.pcie: iATU unroll: enabled
[    0.486716] meson-pcie fc000000.pcie: Detected iATU regions: 4 outbound, 4 inbound
[    0.527447] meson-pcie fc000000.pcie: error: wait linkup timeout
[    0.568993] meson-pcie fc000000.pcie: error: wait linkup timeout


    1.268690] meson-pcie fc000000.pcie: error: wait linkup timeout
[    1.409751] meson-pcie fc000000.pcie: error: wait linkup timeout
[    1.550798] meson-pcie fc000000.pcie: error: wait linkup timeout
[    1.691855] meson-pcie fc000000.pcie: error: wait linkup timeout
[    1.832912] meson-pcie fc000000.pcie: error: wait linkup timeout
[    1.933278] meson-pcie fc000000.pcie: Phy link never came up
[    1.933418] meson-pcie fc000000.pcie: PCI host bridge to bus 0000:00
[    1.933426] pci_bus 0000:00: root bus resource [bus 00-ff]
[    1.933431] pci_bus 0000:00: root bus resource [io  0x0000-0xfffff]
[    1.933435] pci_bus 0000:00: root bus resource [mem 0xfc700000-0xfdffffff]
[    1.933460] pci 0000:00:00.0: [16c3:abcd] type 01 class 0x060400
[    1.933476] pci 0000:00:00.0: reg 0x38: [mem 0x00000000-0x0000ffff pref]
[    1.933509] pci 0000:00:00.0: supports D1
[    1.933513] pci 0000:00:00.0: PME# supported from D0 D1 D3hot D3cold
[    1.976998] meson-pcie fc000000.pcie: error: wait linkup timeout
[    1.979821] pci 0000:00:00.0: BAR 6: assigned [mem 0xfc700000-0xfc70ffff pref]
[    1.979842] pci 0000:00:00.0: PCI bridge to [bus 01-ff]
[    1.979858] pci 0000:00:00.0: Max Payload Size set to  256/ 256 (was  256), Max Read Rq  256
[    1.980240] pcieport 0000:00:00.0: PME: Signaling with IRQ 50
[    1.980493] pcieport 0000:00:00.0: AER: enabled with IRQ 50

And as result my NVME SSD stops working and isn’t visible to system

If I undo this patch - everything works absolutely fine and my system gets NVME SSD back

Tested on several kernels 5.15 and 5.16-RC8

@OldNavi I will check it

Just to mention - HW used: VIM3L → New M2X → NVME SSD.


OK, I will check it today.

@OldNavi I tested it . This patch will make the SSD unusable. I will check this patch here and find a way to solve this bug. If you don’t need to use coral, you can remove this patch first.

Hi @Frank - thanks for confirming this defect, of course I don’t use coral and already disabled this patch.

1 Like

Is there a timeline for getting this patch reverted in one or more of the existing Krescue images? Barring that, is there a known-good image I should try instead that isn’t using an ancient kernel?

I’m trying to evaluate my VIM3(L) boards alongside some other ARM SBCs for an embedded hardware application, but if I can’t use a mainline kernel + SSD it’s going to be a short and unsuccessful test for the Khadas option. (I’ll still use them for my hobby projects, but only those that don’t require more storage than the built-in MMC, which rules out anything server or media-related.)

So I went back down the list of images in Krescue and tried basically everything running a 5.x kernel…and the latest Ubuntu 20.04 GNOME image worked!

Now, I don’t need or want the GNOME environment for this project, but I can work around it and build a new golden image from here, at least to get through my evaluation.

try to use Generic ubuntu images via Krescue or fenix

and dont forget install mainline uboot into emmc or spi

Hi OldNavi, could you give me some idea how to disable this patch?

I’m not sure if that patch now in the latest mainline kernel? Better check with @Frank regarding that. And perhaps it is worth to know what kernel version you’re trying to build with Fenix.

Thanks! But now I can load the HP P410 raid card in newest mainline kernel no matter of this patch. I found that I need to boot khadas vim3 and then power on the PCIe 12v for raid card, finally restart khadas vim3 again. The raid card will be recoginzed in the system or the system be hung on pcie initialization and fail to start.

Sorry for the noise!