Re: MRF kernel module, failed to map BARS

[Michael Davidsaver via Tech-talk <tech-talk at aps.anl.gov>]

On 3/23/21 8:34 AM, GAGET Alexis via Tech-talk wrote:
> Dear All,
> 
> Here at saclay we are using MRF timing system with MTCA.4 in R2 NAT crate.
> 
> I’m using the COMex-E3 and a MCH-PHYS80. The IOC using the mrfioc2 driver is running on the Comex and the mrf kernel module is loaded on it. I have forked from ESS version of mrfioc2, and the kernel module I built seems the same.
> 
> My kernel version is : 3.10.0-1160.15.2.el7.x86_64 on a CenTOS Linux 7.9.2009
> 
> It works globally very well, but I have a reproducible issue:
> 
> When I reboot the whole R2 crate and when I load the MRF kernel module I have the following message :
> 
>  [  347.344081] mrf: module verification failed: signature and/or required key missing - tainting kernel
> [  347.345504] mrf-pci 0000:04:00.0: Attaching MRF device w/o PLX bridge (00007011)
> [  347.345581] mrf-pci 0000:04:00.0: can't ioremap BAR 0: [??? 0x00000000 flags 0x0]
> [  347.345646] mrf-pci 0000:04:00.0: Failed to map BARS! 0 0
> [  347.345918] mrf-pci 0000:06:00.0: Attaching MRF device w/o PLX bridge (00007011)
> [  347.345989] mrf-pci 0000:06:00.0: can't ioremap BAR 0: [??? 0x00000000 flags 0x0]
> [  347.346053] mrf-pci 0000:06:00.0: Failed to map BARS! 0 0
> 
> Consequence, my Comex can’t find some of my EVR or EVM (some are found..).
> 
> To make it work I just have to reboot the Comex and after that it’s working. Meaning after a shutdown I have to reboot it twice…
> 
> Does anyone have this kind of trouble ? Knows how to solve it ? or even understand this message ?

If you look back earlier in the kernel log, you will probably find other
errors mentioning this devices at the point where the kernel is probing
these cards.  Search for the CPU IDs.

As others have mentioned, this is probably an initialization order issue,
which can be addressed in MCH configuration to change card startup
delays so that the MRF cards power up earlier than the CPU.  From my
experience this process isn't as scientific as I would like.