On Fri, Jul 10, 2020 at 11:13:50AM +0200, Thomas Petazzoni wrote:
On Fri, 10 Jul 2020 10:54:24 +0200 Rasmus Villemoes rasmus.villemoes@prevas.dk wrote:
It's very much like the FAT filesystem case: if you have U-Boot proper and your Linux kernel image in a FAT filesystem,
No, this is very much _not_ like the above. In this paragraph, you combine "U-Boot proper and your Linux kernel", imposing an implicit assumption that they are stored in the same way. Sure, _if_ both these items are stored in squashfs images (possibly the same, possibly distinct), then the thing that loads the respective images obviously needs squashfs (or FAT, or whatnot) support.
My point is that it's possible that, say, U-Boot proper is stored in a FAT file system, and the kernel is stored in a UBI volume. So SPL needs FAT support. Why should I be forced to compile FAT support into U-Boot proper if U-Boot proper never needs to access a FAT filesystem? And the same for squashfs. Or any of the drivers or DM_ frameworks that do that "depends on" or "select".
Ah, I absolutely agree that it should be possible to have Squashfs in both SPL and U-Boot proper, or only in SPL or only in U-Boot proper.
It was not clear in your initial e-mail that this was the issue you were pointing.
Note that on this point the question is, do we have a use case for falcon mode and loading linux from squashfs? I assume the answer is yes, and that's why we would want to have squashfs be enabled in SPL.