[mtd/sbc_gxx] kernel BUG at include/linux/mtd/map.h:148!

On Mon, Mar 31, 2014 at 07:34:12PM +0800, Fengguang Wu wrote:
> CC Michal and kbuild list.
> 
> On Thu, Mar 27, 2014 at 04:51:53PM -0600, Bjorn Helgaas wrote:
> > [+cc David, Brian]
> > 
> > On Thu, Mar 27, 2014 at 8:01 AM, Fengguang Wu <fengguang.wu@intel.com> wrote:
> > > FYI, here is a very old warning, too old to be bisected.
> > >
> > > [    5.282127] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
> > > [    5.286079] SBC-GXx flash: IO:0x258-0x259 MEM:0xdc000-0xdffff
> > > [    5.288723] ------------[ cut here ]------------
> > > [    5.289649] kernel BUG at include/linux/mtd/map.h:148!
> > 
> > I think the problem is that your randconfig happens to have none of
> > CONFIG_MTD_MAP_BANK_WIDTH_* set (you should have played the lottery
> > today!), and the default implementation of map_bankwidth() in that
> > case is just "BUG()":
> > 
> >   $ grep MTD_MAP config-3.14.0-rc8-wl-03045-gdf16ea4
> >   # CONFIG_MTD_MAP_BANK_WIDTH_1 is not set
> >   # CONFIG_MTD_MAP_BANK_WIDTH_2 is not set
> >   # CONFIG_MTD_MAP_BANK_WIDTH_4 is not set
> >   # CONFIG_MTD_MAP_BANK_WIDTH_8 is not set
> >   # CONFIG_MTD_MAP_BANK_WIDTH_16 is not set
> >   # CONFIG_MTD_MAP_BANK_WIDTH_32 is not set
> > 
> > I don't know enough Kconfig or MTD to fix this.
> 
> Michal, the problem we run into is, how to select one of the above
> CONFIG_MTD_MAP_BANK_WIDTH_xx in "make randconfig".

You can't enforce this in Kconfig. What you can do is to make the logic
more fool-proof and fall back to some sensible default if none of the
CONFIG_MTD_MAP_BANK_WIDTH_* options is enabled. Like this


From 76e66ceea7e2ffbb1d39c01af2eaf6f2b66c2be3 Mon Sep 17 00:00:00 2001
From: Michal Marek <mmarek@suse.cz>
Date: Mon, 31 Mar 2014 16:25:32 +0200
Subject: [PATCH] mtd: Fall back to MTD_MAP_BANK_WIDTH_1 if none is specified

This is mainly to fix make randconfig errors.

Signed-off-by: Michal Marek <mmarek@suse.cz>
---
 drivers/mtd/chips/Kconfig | 8 ++++++++
 include/linux/mtd/map.h   | 4 ++--
 2 files changed, 10 insertions(+), 2 deletions(-)

Hi Michal,

Your patch fixes the problem.

Tested-by: Jet Chen <jet.chen@intel.com>

Thanks,
-Jet

On 03/31/2014 10:35 PM, Michal Marek wrote:
> On Mon, Mar 31, 2014 at 07:34:12PM +0800, Fengguang Wu wrote:
>> CC Michal and kbuild list.
>>
>> On Thu, Mar 27, 2014 at 04:51:53PM -0600, Bjorn Helgaas wrote:
>>> [+cc David, Brian]
>>>
>>> On Thu, Mar 27, 2014 at 8:01 AM, Fengguang Wu <fengguang.wu@intel.com> wrote:
>>>> FYI, here is a very old warning, too old to be bisected.
>>>>
>>>> [    5.282127] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
>>>> [    5.286079] SBC-GXx flash: IO:0x258-0x259 MEM:0xdc000-0xdffff
>>>> [    5.288723] ------------[ cut here ]------------
>>>> [    5.289649] kernel BUG at include/linux/mtd/map.h:148!
>>>
>>> I think the problem is that your randconfig happens to have none of
>>> CONFIG_MTD_MAP_BANK_WIDTH_* set (you should have played the lottery
>>> today!), and the default implementation of map_bankwidth() in that
>>> case is just "BUG()":
>>>
>>>    $ grep MTD_MAP config-3.14.0-rc8-wl-03045-gdf16ea4
>>>    # CONFIG_MTD_MAP_BANK_WIDTH_1 is not set
>>>    # CONFIG_MTD_MAP_BANK_WIDTH_2 is not set
>>>    # CONFIG_MTD_MAP_BANK_WIDTH_4 is not set
>>>    # CONFIG_MTD_MAP_BANK_WIDTH_8 is not set
>>>    # CONFIG_MTD_MAP_BANK_WIDTH_16 is not set
>>>    # CONFIG_MTD_MAP_BANK_WIDTH_32 is not set
>>>
>>> I don't know enough Kconfig or MTD to fix this.
>>
>> Michal, the problem we run into is, how to select one of the above
>> CONFIG_MTD_MAP_BANK_WIDTH_xx in "make randconfig".
>
> You can't enforce this in Kconfig. What you can do is to make the logic
> more fool-proof and fall back to some sensible default if none of the
> CONFIG_MTD_MAP_BANK_WIDTH_* options is enabled. Like this
>
>
>  From 76e66ceea7e2ffbb1d39c01af2eaf6f2b66c2be3 Mon Sep 17 00:00:00 2001
> From: Michal Marek <mmarek@suse.cz>
> Date: Mon, 31 Mar 2014 16:25:32 +0200
> Subject: [PATCH] mtd: Fall back to MTD_MAP_BANK_WIDTH_1 if none is specified
>
> This is mainly to fix make randconfig errors.
>
> Signed-off-by: Michal Marek <mmarek@suse.cz>
> ---
>   drivers/mtd/chips/Kconfig | 8 ++++++++
>   include/linux/mtd/map.h   | 4 ++--
>   2 files changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/mtd/chips/Kconfig b/drivers/mtd/chips/Kconfig
> index e4696b3..219de49 100644
> --- a/drivers/mtd/chips/Kconfig
> +++ b/drivers/mtd/chips/Kconfig
> @@ -115,6 +115,14 @@ config MTD_MAP_BANK_WIDTH_32
>   	  If you wish to support CFI devices on a physical bus which is
>   	  256 bits wide, say 'Y'.
>
> +config HAVE_MTD_MAP_BANK_WIDTH_1
> +	bool
> +	default MTD_MAP_BANK_WIDTH_1 || (!MTD_MAP_BANK_WIDTH_2 && !MTD_MAP_BANK_WIDTH_4 && !MTD_MAP_BANK_WIDTH_8 && !MTD_MAP_BANK_WIDTH_16 && !MTD_MAP_BANK_WIDTH_32)
> +
> +if HAVE_MTD_MAP_BANK_WIDTH_1 && !MTD_MAP_BANK_WIDTH_1
> +comment "no buswidth selected, using 8-bit as a fallback "
> +endif
> +
>   config MTD_CFI_I1
>   	bool "Support 1-chip flash interleave" if MTD_CFI_GEOMETRY
>   	default y
> diff --git a/include/linux/mtd/map.h b/include/linux/mtd/map.h
> index 5f487d7..d4f4f49 100644
> --- a/include/linux/mtd/map.h
> +++ b/include/linux/mtd/map.h
> @@ -32,7 +32,7 @@
>   #include <asm/io.h>
>   #include <asm/barrier.h>
>
> -#ifdef CONFIG_MTD_MAP_BANK_WIDTH_1
> +#ifdef CONFIG_HAVE_MTD_MAP_BANK_WIDTH_1
>   #define map_bankwidth(map) 1
>   #define map_bankwidth_is_1(map) (map_bankwidth(map) == 1)
>   #define map_bankwidth_is_large(map) (0)
> @@ -156,7 +156,7 @@ static inline int map_bankwidth(void *map)
>   static inline int map_bankwidth_supported(int w)
>   {
>   	switch (w) {
> -#ifdef CONFIG_MTD_MAP_BANK_WIDTH_1
> +#ifdef CONFIG_HAVE_MTD_MAP_BANK_WIDTH_1
>   	case 1:
>   #endif
>   #ifdef CONFIG_MTD_MAP_BANK_WIDTH_2
>