[U-Boot] [PATCH] spl: if MMCSD_MODE_RAW fails, try MMCSD_MODE_FS, if available

[Guillaume Gardet]

Hi Tom, Nikita ,

Le 18/02/2016 10:19, Nikita Kiryanov a écrit :
> Hi Tom, Guillaume,
>
> On Wed, Feb 17, 2016 at 03:27:22PM -0500, Tom Rini wrote:
>> On Wed, Feb 17, 2016 at 09:09:27AM +0100, Guillaume GARDET wrote:
>>
>>> Since commit fd61d39970b9901217efc7536d9f3a61b4e1752a:
>>>          spl: mmc: add break statements in spl_mmc_load_image()
>>> RAW and FS boot modes are now exclusive again. So, if MMCSD_MODE_RAW fails, the
>>> board hangs. This patch allows to try MMCSD_MODE_FS then, if available.
>>>
>>> It has been tested on a beaglebone black to boot on an EXT partition.
>>>
>>> Signed-off-by: Guillaume GARDET <guillaume.gardet at free.fr>
>>> Cc: Tom Rini <trini at konsulko.com>
>>> Cc: Nikita Kiryanov <nikita at compulab.co.il>
>>> Cc: Igor Grinberg <grinberg at compulab.co.il>
>>> Cc: Paul Kocialkowski <contact at paulk.fr>
>>> Cc: Pantelis Antoniou <panto at antoniou-consulting.com>
>>> Cc: Simon Glass <sjg at chromium.org>
>>> Cc: Matwey V. Kornilov <matwey.kornilov at gmail.com>
>>>
>>> ---
>>>   common/spl/spl_mmc.c | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/common/spl/spl_mmc.c b/common/spl/spl_mmc.c
>>> index c3931c6..2eef0f2 100644
>>> --- a/common/spl/spl_mmc.c
>>> +++ b/common/spl/spl_mmc.c
>>> @@ -284,7 +284,7 @@ int spl_mmc_load_image(u32 boot_device)
>>>   		if (!err)
>>>   			return err;
>>>   #endif
>>> -		break;
>>> +		/* Fall through */
>>>   	case MMCSD_MODE_FS:
>>>   		debug("spl: mmc boot mode: fs\n");
>> This also essentially reverts fd61d399.  So Nikita, was there a specific
>> use case that was broken before, or was the code just unclear in
>> intentions here?  Thanks!
> There was no broken use case that I'm aware of. The change was made as
> part of a code improvement series and was meant to address what I
> consider to be bad and problematic design. Instead of reverting it
> though, how about implementing something similar to what I did in the
> main common/spl/spl.c:board_init_r()? You would have a weak function
> that will default to the original spl_boot_mode() if not overridden,
> and allow the user to define a sequence of boot modes otherwise.

The thing is you broke a wanted behavior currently in use. So, the priority is to come back to the previous behavior.
Then, if someone (you, me or someone else) wants to improve this code, the way you suggested, it would be very nice.
But it will need a lot more work, tests and reviews.


Guillaume

>
>> -- 
>> Tom
>
>