Re: b276527539 ("staging: most: move core files out of the staging .."): [ 12.247349] BUG: kernel NULL pointer dereference, address: 00000000

[Christian.Gromm]

On Fri, 2020-04-24 at 12:16 +0200, Greg KH wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you
> know the content is safe
> 
> On Fri, Apr 24, 2020 at 09:41:36AM +0000, 
> Christian.Gromm@xxxxxxxxxxxxx wrote:
> > On Sun, 2020-03-29 at 21:39 +0800, kernel test robot wrote:
> > > EXTERNAL EMAIL: Do not click links or open attachments unless you
> > > know the content is safe
> > > 
> > > Greetings,
> > > 
> > > 0day kernel testing robot got the below dmesg and the first bad
> > > commit is
> > > 
> > > https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging.git
> > > staging-next
> > > 
> > > commit b276527539188f1f61c082ebef27803db93e536d
> > > Author:     Christian Gromm <christian.gromm@xxxxxxxxxxxxx>
> > > AuthorDate: Tue Mar 10 14:02:40 2020 +0100
> > > Commit:     Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > CommitDate: Tue Mar 24 13:42:44 2020 +0100
> > > 
> > >     staging: most: move core files out of the staging area
> > > 
> > >     This patch moves the core module to the /drivers/most
> > > directory
> > >     and makes all necessary changes in order to not break the
> > > build.
> > > 
> > >     Signed-off-by: Christian Gromm <christian.gromm@xxxxxxxxxxxxx
> > > >
> > >     Link:
> > > https://lore.kernel.org/r/1583845362-26707-2-git-send-email-christian.gromm@xxxxxxxxxxxxx
> > >     Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx
> > > >
> > > 
> > > 22dd4acc80  Staging: speakup: Add identifier name to function
> > > declaration arguments.
> > > b276527539  staging: most: move core files out of the staging
> > > area
> > > e681bb287f  staging: vt6656: Use DIV_ROUND_UP macro instead of
> > > specific code
> > > +-------------------------------------------------------+------
> > > ----
> > > --+------------+------------+
> > > >                                                       |
> > > > 22dd4acc80
> > > > > b276527539 | e681bb287f |
> > > +-------------------------------------------------------+------
> > > ----
> > > --+------------+------------+
> > > > boot_successes                                        |
> > > > 26         | 0          | 0          |
> > > > boot_failures                                         |
> > > > 8          | 11         | 11         |
> > > > WARNING:possible_circular_locking_dependency_detected |
> > > > 8          |            |            |
> > > > BUG:kernel_NULL_pointer_dereference,address           |
> > > > 0          | 11         | 11         |
> > > > Oops:#[##]                                            |
> > > > 0          | 11         | 11         |
> > > > EIP:__list_add_valid                                  |
> > > > 0          | 11         | 11         |
> > > > Kernel_panic-not_syncing:Fatal_exception              |
> > > > 0          | 11         | 11         |
> > > +-------------------------------------------------------+------
> > > ----
> > > --+------------+------------+
> > > 
> > > If you fix the issue, kindly add following tag
> > > Reported-by: kernel test robot <lkp@xxxxxxxxx>
> > > 
> > > [   12.242090] no options.
> > > [   12.245364] FPGA DOWNLOAD --->
> > > [   12.245723] FPGA image file name: xlinx_fpga_firmware.bit
> > > [   12.246548] GPIO INIT FAIL!!
> > > [   12.246995] most_sound: init()
> > > [   12.247349] BUG: kernel NULL pointer dereference, address:
> > 
> > The init order of the modules is wrong in case the driver is
> > being built in-tree.
> > 
> > The init function of module most_sound is called before the
> > core itself is being initialized.
> > 
> > [    5.179189] most_sound: init()
> > [    5.180205] mostcore: __init()
> > 
> > Hence the list used in the core to store and track the
> > registered components has not been initialized with
> > INIT_LIST_HEAD(&comp_list) by the time the sound module
> > tries to register itself with the core.
> > 
> > The Kconfig of most_sound, however, has a dependency to
> > MOST. How can the build system be forced to initialize the
> > core module first?
> 
> Linker order is the thing here.
> 
> You can mess with the init levels here, and use subsys_initcall() for
> mostcore, will that fix it?

I already gave it a try and it works. But is it ok to use the
subsys_initcall() function when the driver is being built as a
module?

thanks,
Chris