Re: [PATCH v4 5/6] rockchip: kylin: Enable boot with android boot image

[Jeffy Chen <jeffy.chen@rock-chips.com>]

Hi Tom,

On 2016-1-15 0:22, Tom Rini wrote:
> On Thu, Jan 14, 2016 at 10:31:34AM +0800, Jeffy Chen wrote:
>> Hi Tom,
>>
>> On 2016-1-13 23:28, Tom Rini wrote:
>>> On Wed, Jan 13, 2016 at 04:53:19PM +0800, Jeffy Chen wrote:
>>>
>>>> The android kernel is using appended dtb by default, and store
>>>> ramdisk right after kernel & dtb.
>>>> So we needs to relocate ramdisk, and use atags to pass params.
>>>>
>>>> Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
>>>> Acked-by: Simon Glass <sjg@chromium.org>
>>>> ---
>>>>
>>>> Changes in v4: None
>>>> Changes in v3: None
>>>> Changes in v2: None
>>>>
>>>>   include/configs/kylin_rk3036.h | 23 +++++++++++++++++++++++
>>>>   1 file changed, 23 insertions(+)
>>>>
>>>> diff --git a/include/configs/kylin_rk3036.h b/include/configs/kylin_rk3036.h
>>>> index b750b26..49997ec 100644
>>>> --- a/include/configs/kylin_rk3036.h
>>>> +++ b/include/configs/kylin_rk3036.h
>>>> @@ -35,6 +35,29 @@
>>>>   #undef CONFIG_EXTRA_ENV_SETTINGS
>>>>   #define CONFIG_EXTRA_ENV_SETTINGS \
>>>>   	"partitions=" PARTS_DEFAULT \
>>>> +	"mmcdev=0\0" \
>>>> +	"mmcpart=5\0" \
>>>> +	"loadaddr=" __stringify(CONFIG_SYS_LOAD_ADDR) "\0" \
>>>> +
>>>> +#define CONFIG_ANDROID_BOOT_IMAGE
>>>> +#define CONFIG_SYS_BOOT_RAMDISK_HIGH
>>> This should already be set.
>> Right, i'll remove it...
>>>> +#define CONFIG_SYS_HUSH_PARSER
>>>> +
>>>> +#undef CONFIG_BOOTCOMMAND
>>>> +#define CONFIG_BOOTCOMMAND \
>>>> +	"mmc dev ${mmcdev}; if mmc rescan; then " \
>>>> +		"part start mmc ${mmcdev} ${mmcpart} boot_start;" \
>>>> +		"part size mmc ${mmcdev} ${mmcpart} boot_size;" \
>>>> +		"mmc read ${loadaddr} ${boot_start} ${boot_size};" \
>>>> +		"bootm start ${loadaddr}; bootm ramdisk;" \
>>>> +		"bootm prep; bootm go;" \
>>>> +	"fi;" \
>>>> +
>>>> +/* Enable atags */
>>>> +#define CONFIG_SYS_BOOTPARAMS_LEN	(64*1024)
>>>> +#define CONFIG_INITRD_TAG
>>>> +#define CONFIG_SETUP_MEMORY_TAGS
>>>> +#define CONFIG_CMDLINE_TAG
>>> But I'm confused as to what exactly is going on here.  Appended dtb is
>>> not the same as ATAGS.  And you shouldn't need to split up bootm like
>>> that.  Can you please explain a bit more?  Thanks!
>> The u-boot will pass atags to kernel, and kernel will merge those
>> atags into the appended dtb(fdt).
>>
>> The default bootm flow would not pass ramdisk state, but we need it,
>> so we should add this state into default flow, or just use split
>> bootm cmds :)
> That seems very strange.  Is the ramdisk concatenated with the kernel
> and dtb as well (and that's why bootm ramdisk somehow finds it but
> normal bootm doesn't as you aren't passing in a ramdisk address) ?
Yes, the ramdisk concatenated with the kernel and dtb as 
well(u-boot/include/android_image.h: struct andr_img_hdr).

And the normal bootm cmd would find it by parsing andr_img_hdr struct.
But we still need bootm ramdisk state, because it will call 
boot_ramdisk_high to relocate ramdisk area :)

I found if not relocate it to somewhere else, it would be corrupted 
after kernel's decompressing(during update fdt area).