RE: Can anyone boot a system using btrfs root with linux 3.14 or newer? - RESOLVED

RE: Can anyone boot a system using btrfs root with linux 3.14 or newer? - RESOLVED

[Пламен Петров]

The problem reported in this thread has been RESOLVED.

It's not BTRFS's fault.

Debugging on my part led to the actual problem in do_mounts.c - some filesystems mount routines return error codes other than 0, EACCES and EINVAL and such return codes result in the kernel panicking without trying to mount root with all of the available filesystems.

Patch is available as attachment to bug 74901 - https://bugzilla.kernel.org/show_bug.cgi?id=74901 . The bugentry documents how I managed to find the problem.

Also, the patch has been sent to the linux kernel mailing list - see http://news.gmane.org/find-root.php?group=gmane.linux.kernel&article=1691881 Hopefully, it will find its way into the kernel, and later on - in stable releases.

Thanks to you all!
------------------------------
Plamen Petrov

Re: Can anyone boot a system using btrfs root with linux 3.14 or newer? - RESOLVED

[Martin]

On 27/04/14 13:00, Пламен Петров wrote:
> The problem reported in this thread has been RESOLVED.
> 
> It's not BTRFS's fault.
> 
> Debugging on my part led to the actual problem in do_mounts.c - some
> filesystems mount routines return error codes other than 0, EACCES
> and EINVAL and such return codes result in the kernel panicking
> without trying to mount root with all of the available filesystems.
> 
> Patch is available as attachment to bug 74901 -
> https://bugzilla.kernel.org/show_bug.cgi?id=74901 . The bugentry
> documents how I managed to find the problem.

Well deduced and that looks to be a good natural clean fix.

My only question is: What was the original intent to deliberately fail
if something other than EACCES or EINVAL were reported?


> Also, the patch has been sent to the linux kernel mailing list - see
> http://news.gmane.org/find-root.php?group=gmane.linux.kernel&article=1691881
> Hopefully, it will find its way into the kernel, and later on - in
> stable releases.

That all looks very good and very thorough.


> Thanks to you all! ------------------------------ Plamen Petrov

Thanks to you for chasing it through!

AND for posting the "Resolved" to let everyone know. :-)


Regards,
Martin

RE: Can anyone boot a system using btrfs root with linux 3.14 or newer? - RESOLVED

[Пламен Петров]

>On 27/04/14 13:00, Пламен Петров wrote:
>> The problem reported in this thread has been RESOLVED.
>> 
>> It's not BTRFS's fault.
>> 
>> Debugging on my part led to the actual problem in do_mounts.c - some
>> filesystems mount routines return error codes other than 0, EACCES
>> and EINVAL and such return codes result in the kernel panicking
>> without trying to mount root with all of the available filesystems.
>> 
>> Patch is available as attachment to bug 74901 -
>> https://bugzilla.kernel.org/show_bug.cgi?id=74901 . The bugentry
>> documents how I managed to find the problem.
>
>Well deduced and that looks to be a good natural clean fix.
>
>My only question is: What was the original intent to deliberately fail
>if something other than EACCES or EINVAL were reported?
>

This one beats me as well: I looked up current git history for init/do_mounts.c - it turns out that it came in git this way, with 2.6.12-rc2. The actual reasoning behind its behavior - I do not know... and did not bother looking it up further.

>> Also, the patch has been sent to the linux kernel mailing list - see
>> http://news.gmane.org/find-root.php?group=gmane.linux.kernel&article=1691881
>> Hopefully, it will find its way into the kernel, and later on - in
>> stable releases.
>
>>That all looks very good and very thorough.
>>
>> Thanks to you all! ------------------------------ Plamen Petrov
>
>Thanks to you for chasing it through!
>
>AND for posting the "Resolved" to let everyone know. 
>
>Regards,
>Martin

You are welcome!

Regards,
---------------------------------
Plamen Petrov