On Sat, Sep 15, 2001 at 12:19:38PM +0400, Borsenkow Andrej wrote:
>
> You do understand the code, unfortunately. Before making such claims you
^
not
I am pressuming you meant. You are/were right.
> should look into mount_root() first.
I spent some more time last night and came to this same conclusion.
The problem is that mount_root() is overloaded to mount both real root
and initrd root. Before "rootflags" this was OK. It is also possible
to use mount_root() overloaded if the value of "root_mount_data" is
massaged according to whether the call to mount_root() is for the
initrd or the real_root.
> The value of rootflags is passed as
> opaque mount_data parameter
root_mount_data actually.
> that is never touched in the code and is
> meant for fs-specific flags. What is preserved is standard flags like
> ro.
> You have not proved anything because you quote wrong code.
You are right. I was not going to continue that conversation with
"gc" after his crude remarks however. It was apparent that he was not
interested in any further comments from me.
> rootflags is
> meant for direct mounting of root; any kernel with initrd will have
> problems at this point.
It is also something that has been introduced quite recently (post
2.4.4 at least). It's implementation was likely not completely
thought through. This is something that likely needs to be brought up
on the kernel list.
> It is not standard kernel feature. I asked on lkml and it appears to be
> part of ext3 patch.
That is what I was suspecting.
b.
--
Brian J. Murrell
