[fedora-arm] kernel BUG at drivers/mtd/maps/physmap.c:158!

[Jon Masters]

On 06/15/2012 06:43 PM, Peter Robinson wrote:
> On Fri, Jun 15, 2012 at 10:00 PM, Jon Masters <jcm at redhat.com> wrote:
>> On 06/15/2012 04:17 PM, Brendan Conoboy wrote:
>>> On 06/15/2012 10:06 AM, Alex Villací­s Lasso wrote:
>>>> Still present in 3.4.2-3.fc17.armv5tel:
>>>
>>> Jon Masters tracked this down.  For the time being add the following to
>>> the kernel command line (IE, the append=...):
>>>
>>>    physmap.enabled=0
>>
>> For the record, the vexpress model doesn't provide a full emulation of
>> the MTD flash, but the vexpress kernel provides glue that causes udev to
>> try to load the physmap driver. So on real hardware, this is ok, but in
>> the model, it dies during probe when it doesn't see hardware (that's
>> probably a kernel bug - it ought not to oops there). The hack is to
>> specify a kernel module parameter (enabled=0) that doesn't exist (there
>> is no "enabled") since we intentionally engineered the module utilities
>> to handle that case by failing to load the module. So, udev tries to
>> load it, the module gets an unknown parameter, fails, and we're cool.
>>
>> We don't need this fixed, but obviously it does need fixing for general
>> happiness and for us to remove the need for this command line hack.
> 
> Jon do you know the particular mtd module that is failing to load?
> What does it do on a vexpress board?

It's not a flash module failing per se, it's the physmap driver is
failing because the underlying flash isn't actually there on the model.
It shouldn't fail (should just fail out in the probe and continue), but
we can get the kernel fixed post-GA.

Jon.