Re: [DistroKit] [PATCH 3/5] v8a: images: generate sparse images from rock3a initramfs, kernel and dtb

[Ahmad Fatoum <a.fatoum@pengutronix.de>]

On 06.06.23 17:04, Johannes Zink wrote:
> On 6/6/23 16:51, Ahmad Fatoum wrote:
>> Hello Johannes,
>>
>> On 06.06.23 16:47, Johannes Zink wrote:
>>> On 6/6/23 16:41, Ahmad Fatoum wrote:
>>>> On 06.06.23 16:31, Johannes Zink wrote:
>>>>> This allows to load the images to barebox using android fastboot, since
>>>>> otherwise the sparse header is missing and the image size is not
>>>>> properly aligned as required by fastboot.
>>>>
>>>> IME fastboot just complains and transfers the files anyway. In that
>>>> case building sparse images just adds complication. Did your fastboot
>>>> command fail without this change?
>>>>
>>>
>>> Uploading non-sparse images failed with an error
>>>
>>> Invalid sparse file format at header magic
>>> error: write_sparse_skip_chunk: don't care size 45251584 is not a multiple of the block size 4096
>>>
>>> at the host tool, which is why I decided to use sparse images.
>>
>> Can you describe how to reproduce this with a normal DistroKit build?
> 
> sure, just apply this series (or at least patch 1/5 for initrd generation), run ptxdist images and expose a fastboot partition on the DUT by calling
> 
> usbgadget -A /tmp/initramfs(initramfs)c
> 
> then use fastboot to upload the cpio.gz from your distrokit build dir
> 
> fastboot flash initramfs platform-v8a/images/initramfs-sparse.img

I can reproduce this with fastboot version 28.0.2-debian.
It wasn't always like that. I assume -S 0 wasn't always the
default, but I haven't checked the code. I think it's worth a bug
report.

>>> It saves upload time anyway, at least for the initramfs, so it's a good idea to use sparse images, anyway.
>>
>> Sparse is a compression algorithm, albeit a very primitive one. Your kernel and
>> initrd should already be compressed, so if you have savings from compressing
>> them again with sparse, there's something wrong IMO.
>>
>> The device tree is small enough that it shouldn't matter.

As mentioned, I don't think double compression is useful.
On our side, we can likely just use -S 1 to restore the old behavior.

Cheers,
Ahmad

>>
>> Cheers,
>> Ahmad
>>
>>
>>>
>>> Johannes
>>>
>>>
>>>>>
>>>>> Signed-off-by: Johannes Zink <j.zink@pengutronix.de>
>>>>> ---
>>>>>    configs/platform-v8a/config/images/rock3a.config | 21 +++++++++++++++++++++
>>>>>    1 file changed, 21 insertions(+)
>>>>>
>>>>> diff --git a/configs/platform-v8a/config/images/rock3a.config b/configs/platform-v8a/config/images/rock3a.config
>>>>> index 522d82cdc9c9..cb0c7c4e575d 100644
>>>>> --- a/configs/platform-v8a/config/images/rock3a.config
>>>>> +++ b/configs/platform-v8a/config/images/rock3a.config
>>>>> @@ -44,3 +44,24 @@ image barebox-rock3a.norimg {
>>>>>            image = "barebox-rock3a.img"
>>>>>        }
>>>>>    }
>>>>> +
>>>>> +image initramfs-sparse.img {
>>>>> +    android-sparse {
>>>>> +        image = "root.cpio.gz"
>>>>> +        block-size = 4k
>>>>> +    }
>>>>> +}
>>>>> +
>>>>> +image kernel-sparse.img {
>>>>> +    android-sparse {
>>>>> +        image = "linuximage"
>>>>> +        block-size = 4k
>>>>> +    }
>>>>> +}
>>>>> +
>>>>> +image dtb-rock3a-sparse.img {
>>>>> +    android-sparse {
>>>>> +        image = "rk3568-rock-3a.dtb"
>>>>> +        block-size = 4k
>>>>> +    }
>>>>> +}
>>>>>
>>>>
>>>
>>
> 

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |