RaspberryPi4 Panic in net_ns_init()

Martin Roukala martin.roukala at mupuf.org
Thu Sep 8 08:14:29 UTC 2022


On 9/5/22 14:18, Maxime Ripard wrote:
> On Wed, Aug 31, 2022 at 04:42:05PM +0200, Maxime Ripard wrote:
>> Sorry for the fairly broad list of recipients, I'm not entirely sure
>> where the issue lies exactly, and it seems like multiple areas are
>> involved.
>>
>> Martin reported me an issue discovered with the VC4 DRM driver that
>> would prevent the RaspberryPi4 from booting entirely. At boot, and
>> apparently before the console initialization, the board would just die.
>>
>> It first appeared when both DYNAMIC_DEBUG and DRM_VC4 were built-in. We
>> started to look into what configuration would trigger it.
>>
>> It looks like a good reproducer is:
>>
>> ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j18 defconfig mod2yesconfig
>> ./scripts/config -e CONFIG_DYNAMIC_DEBUG
>> ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- make -j18 olddefconfig
> 
> Interestingly, this was making the kernel Image cross the 64MB boundary.
> I've compiled the same configuration but with -Os, and then tried to
> boot the failing configuration with U-Boot instead of the RaspberryPi
> firmware, and both of them worked. I guess that leaves us with a
> bootloader bug, and nothing related to Linux after all.
> 
> Sorry for the noise,
> Maxime

Thanks a lot for figuring this out, Maxime! The idea to use -Os is much 
nicer than what I was going for: padding a working kernel with 0s at the 
end!

I unfortunately did not try to reproduce your success with u-boot, but 
I'll get to it in the near future.

Thanks again to everyone involved for all the help and support!
Martin


More information about the dri-devel mailing list