[U-Boot] [PATCH v2] ubi: enable error reporting in initialization

[Hector Palacios]

Hi Andrew,

On 11/05/2014 08:31 PM, Andrew Ruder wrote:
> The UBI layer will disable much of its error reporting when it is
> compiled into the linux kernel to avoid stopping boot.  We want this
> error reporting in U-Boot since we don't initialize the UBI layer until
> it is used and want the error reporting.
> 
> We force this by telling the UBI layer we are building as a module.
> 
> Signed-off-by: Andrew Ruder <andrew.ruder at elecsyscorp.com>
> Cc: Wolfgang Denk <wd at denx.de>
> Cc: Heiko Schocher <hs at denx.de>
> Cc: Kyungmin Park <kmpark at infradead.org>
> ---
>  include/ubi_uboot.h | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/include/ubi_uboot.h b/include/ubi_uboot.h
> index 1fd15f4..324fe72 100644
> --- a/include/ubi_uboot.h
> +++ b/include/ubi_uboot.h
> @@ -51,6 +51,14 @@
>  
>  #undef CONFIG_MTD_UBI_BLOCK
>  
> +/* ubi_init() disables returning error codes when built into the Linux
> + * kernel so that it doesn't hang the Linux kernel boot process.  Since
> + * the U-Boot driver code depends on getting valid error codes from this
> + * function we just tell the UBI layer that we are building as a module
> + * (which only enables the additional error reporting).
> + */
> +#define CONFIG_MTD_UBI_MODULE
> +
>  #if !defined(CONFIG_MTD_UBI_BEB_LIMIT)
>  #define CONFIG_MTD_UBI_BEB_LIMIT	20
>  #endif
> 

I applied this patch but apparently I'm suffering a memory leak in a certain
condition. I wonder if you can reproduce it:

1. Assume you have an empty partition called "MyPart".
2. Write it with some dummy data:
	=> mw.l $loadaddr deadbeed 1000
	=> nand write $loadaddr MyPart 1000
3. Set it as the UBI part:
	=> ubi part MyPart
This (which returned 0 before, despite failing during the attach procedure now should
return an error).
4. Run the command again several times:
	=> ubi part MyPart
	=> ubi part MyPart

In my case, after calling this three times, the target hangs. I think the exit path in
ubi_init() does not properly free every allocated memory. I was not able to find the
root cause, though.

Regards
--
Hector Palacios