Re: [PATCH] Endianness-aware mkcramfs

[Daniel Quinlan <quinlan@transmeta.com>]

Daniel Marmier <daniel.marmier@lightning.ch> writes:

> Here you are, against kernel 2.4.16. The patch is not as clean as one
> would like it to be, but we use it and it works well for us.
> 
> Basically it adds a "-b" (byteorder option) which can take four parameters:
>    -bb	creates a big-endian cramfs,
>    -bl	creates a little-endian cramfs,
>    -bh	creates a cramfs with the same endianness as the host,
>    -br	creates a cramfs with the reverse endianness as the host,
> where "host" refers to the machine running the mkcramfs program.
> 
> As told above, it could be cleaner, but I don't know of a nice method of
> accessing byteorder dependent data through structures.
> 
> Have a nice day,

Hmm... I've been on vacation, so I'm joining the discussion a little
late here, but as hpa and others have said, cramfs is defined to be
little-endian -- we do not want two different versions of the
filesystem.

Send me a patch so big-endian systems byte-swap metadata (and the
equivalent for the mkcramfs/cramfsck programs) and I'd be happy to
help you clean it up and get it into the kernel.

 Dan