pci problem - 2 BARs given same address

[davidshoham]

Hi,

After attaching a pci device and booting the imx6 I receive the following output for the config space of the device:

root@freescale /sys/devices/pci0000:00/0000:00:00.0/0000:01:00.0$ od -X config

0000000         ec301204        00100140        00000000        00000008

0000020         fffc0000        fffc0000        00000000        00000000

0000040         00000000        00000000        00000000        30101204

0000060         00000000        00000050        00000000        0000019b

0000100         00000000        00000000        00000000        00000000

0000120         00037001        00000000        00000000        00000000

0000140         00000000        00000000        00000000        00000000

0000160         00809005        00000000        00000000        00000000

0000200         00000000        00000000        00000000        00000000

0000220         00010010        00008000        00002010        00007411

0000240         00110000        00000000        00000000        00000000

0000260         00000000        00000000        00000000        00000000

BAR0 and BAR1 (marked in red) are given the same address. Is this a kernel bug? If so where should I look in the kernel source to fix it?

Thanks,

David

[YixingKong]

David

If your question has been answered, please click Correct Answer or Helpful Answer.

Regards,

Yixing

[Yuri · Accepted solution]

Looks like this may relate to the next from the i.MX6 Errata :

ERR003757 PCIe: 9000448152—Internal Address Translation Unit (iATU):

Inbound Vendor Defined Message (VDM) ‘ID Match Mode’ is not

functional

ERR003759 PCIe: 9000439510—Internal Address Translation Unit (iATU) can

sometimes overwrite Outbound (Tx) Vendor Messages and MSIs

[richard_zhu]

I didn’t ever encounter such kind of phenomena.

Can you past the log of the “lspci –v ” info at your platform?