Re: unable to mmap /dev/kmem

["Franck Bui-Huu" <fbuihuu@gmail.com>]

On 1/19/07, Hugh Dickins <hugh@veritas.com> wrote:
>
> Apology surely accepted, it's a confusing area (inevitably: in a driver
> for mem, the distinction between addresses and offsets become blurred).
>
> But please note, the worst of it was, that your patch comment gave no
> hint that you were knowingly changing its behaviour on the "main"
> architectures: it reads as if you were simply fixing it up on a
> few less popular architectures where an anomaly had been missed.
>

Because I was thinking that the expected behaviour was the one
implemented before 2.6.12. So I really thought to fix a bug, again
sorry for not having checked the history...

That said, it's really confusing to pass a virtual address as an
offset because:

    (a) mmap() has always worked with offset not addresses;

    (b) the kernel will treat this virtual address as an offset
        until kmem driver convert it back to a virtual
        address. And it seems that during this convertion the
        lowest bits of the virtual address will be lost...

Maybe read/write behaviours should be changed to use the offset as an
offset and not as a virtual address.

> > > I guess it's reassuring to know that not many are actually
> > > using mmap of /dev/kmem, so you're the first to notice: thanks.
> >
> > yes it doesn't seems to be used. In my case, I was just playing with
> > it when I submitted the patch but have no real usages.
>
> Have I got it right, that actually the problem you thought you were
> fixing does not even exist?

yes, see above.

>  __pa was already doing the right thing on
> all architectures, wasn't it?  So we can simply ask Linus to revert your
> patch?

yes we can if the desired behaviour is the one introduced by 2.6.12.

> I don't think your PFN_DOWN or virt_to_phys were improvements:
> though mem.c happens to live in drivers/char/, imagine it under mm/.
>

I don't get your point here. Do you mean that virt_to_phys() is only
meant for drivers ? If so, I would have said that virt_to_phys() is
prefered once boot memory init is finished...

		Franck