Re: bprm_count and stack_mprotect error when testing BPF LSM on v5.7-rc3

[Ma Xinjian <max.xinjian@intel.com>]

On 5/9/20 9:28 PM, KP Singh wrote:
> On Sat, May 9, 2020 at 11:59 AM Ma Xinjian <max.xinjian@intel.com> wrote:
>>
>> On 5/9/20 5:26 PM, KP Singh wrote:
>>> Do you have bpf in your CONFIG_LSM string?
>> That's the point!
>>
>> I remove bpf from  since I can't boot if bpf in it.
> That does indicate a problem which needs to be fixed.
>
>> seems bpf in CONFIG_LSM conflict with CONFIG_BPF_LSM
>>
>> Here is boot error:
>>
>> "Cannot determine cgroup we are running in: No data available
>> Failed to allocate manager object: No data available
>> [!!!!!!] Failed to allocate manager object, freezing.
> I found some references to these error messages and they seem
> to be coming from systemd but I am not sure.
>
>     https://github.com/lxc/lxc/issues/1669
>     https://github.com/containers/libpod/issues/1226
>
>> Freezing execution.
>> [   35.773797] random: fast init done
>> [  130.560629] random: crng init done"
>>
>>> Also, can you share your Kconfig please?
>> refer to attackment.
>>
>> I doubt sth was wrong with my kconfig, maybe me some suggestion
> I am not saying something is wrong with your Kconfig :)
> I just want to make sure we eliminate as many
> variables as possible.
>
> I was able to boot this successfully using QEMU
> (after I enabled SCSI and VIRTIO). So it's likely
> dependent on some user-space configuration
> (again, I am not saying your config is wrong). But
> I will need more information to reproduce and debug this.
>
> Can you try providing a reliable reproduction with a list
> of steps? e.g.
>
> 1. Download the vanilla image here.
> 2. Compile the kernel with defonconfig and kvmconfig
>     (or your own config)
> 3. Boot the kernel in QEMU with the cmdline (...) and the
>    QEMU args (...)
>
> Thanks!
> - KP

Thank you very much for your kind and quick reply.

I tested on LKP cluster of Intel. Everything works automatically.

https://github.com/intel/lkp-tests

---------------------

And I have found the cause.

It can't boot due to comfliction between cgroup configuration for 
CONFIG_BPF_LSM

and systemd.

similar to https://github.com/elogind/elogind/issues/18

we have decided to skip this test.

Thanks again.

- Ma

>
>> Besides, I tested on both physical machine and vm
> [...]
>
-- 
Best Regards.
Ma Xinjian