Hello all,
I need some information concerning rootfs and initramfs.
I have read a lot of stuff on this but some questions remain.

I need to understand how rootfs is mounted with the initramfs procedure.

Things I know: __initramfs_start symbol is defined in vmlinux.lds.S (in the .init section)
During boot, "populate_rootfs()" calls "unpack_to_roofs()" and after "decompress_method()" is called to check the magic number of the compression used to pack initramfs_data.cpio.<comp>. This magic number is expected at __initramfs_start adress. If a valid magic number is found, the name of the compressor is returned to "unpack_to_roofs()". Then the un<comp> routine (example: ungzip, unlzma ..) is called.

Are these things correct?

if so, the initramfs_data.cpio archive should already been in memory (at __initramfs_start) during boot.
So how to copy initramfs.cpio archive to __initramfs_start?

Where is the initramfs_data.cpio.<comp> placed in memory after it is decompressed.

Thanks in advance for your help.
Regards

Read the ramfs-rootfs-initramfs.txt file in ../Documentation/filesystems/

There are (nowadays) two distinct initramfs potentially present during the boot process - you are looking at the code for the newer resident one.

Thanks for your answer.

Indeed, I have already read this document.

As I explained, I am confused concerning the initramfs procedure.
I really need to understand how the initramfs cpio archive is provided to the kernel.

__initramfs_start is defined in the vmlinux.lds.S: initramfs_data.cpio.gz should be expanded to this point.
But how and when?
During kernel build, when initramfs_data.cpio.gz is created?
During boot? If so, how and where it is done?

Thanks in advance.