[Letux-kernel] compressing kernel

[H. Nikolaus Schaller]

Hi,

> Am 26.10.2019 um 20:11 schrieb Andreas Kemnade <andreas at kemnade.info>:
> 
> Hi Nikolaus,
> 
> On Sat, 26 Oct 2019 19:09:10 +0200
> "H. Nikolaus Schaller" <hns at goldelico.com> wrote:
> 
>> Hi,
>> 
>>> Am 26.10.2019 um 16:11 schrieb Andreas Kemnade <andreas at kemnade.info>:
>>> 
>>> Hi,
>>> 
>>> I have just played around with compression options and found that
>>> we are still  < 4MB with lzma  
>> 
>> Would be nice!
>> 
>> Although we have 6 MB in NAND for GTA04 and plenty of space on µSD
>> for all others.
> 
> well, we are at >5MB, and there might be more stuff to be compiled
> statically (if we want imx6sl* in. 
> # CONFIG_SOC_IMX6SL is not set
> # CONFIG_SOC_IMX6SLL is not set
> 
>> 
>>> 
>>> That can be switched by
>>> # CONFIG_KERNEL_GZIP is not set
>>> CONFIG_KERNEL_LZMA=y  
>> 
>> Yes, it is the default in omap2plus_defconfig AFAIK.
>> 
> And I booted omap2plus_defconfig several times.
> 
> 
>>> Downside is a lot slower compression and a bit slower decompression
>>> (still faster than bzip2).  
>> 
>> And that all our u-boot binaries do not understand it.
>> 
> It is not their business to uncompress it. The kernel extracts itself.
> andi at aktux:~/gta04/gta04-kernel$ file ../5.4-rc1-letux/arch/arm/boot/uImage 
> ../5.4-rc1-letux/arch/arm/boot/uImage: u-boot legacy uImage, Linux-5.4.0-rc4-letux+, Linux/ARM, OS Kernel Image (Not compressed), 3892752 bytes, Sat Oct 26 10:20:31 2019, Load Address: 0x80008000, Entry Point: 0x80008000, Header CRC: 0xE7C5FDF5, Data CRC: 0x7032D32E
> andi at aktux:~/gta04/gta04-kernel$ ls -l ../5.4-rc1-letux/arch/arm/boot/uImage
> -rw-r--r-- 1 andi andi 3892816 Oct 26 12:20 ../5.4-rc1-letux/arch/arm/boot/uImage
> 
> Note: *not compressed*. Means that uboot does not do any decompression
> itself.

Ah, I see... It is too long ago that I looked at these boot phases.
And I am getting older :)

> 
>> Unfortunately, the latest stuff is not always the best :(
>> 
> I do not see any problem in using it.

Ok, I'll give it a try. I remember there was a compile issue (HOST?) but
it may have gone.

BR,
Niklaus