Re: [PATCH 04/15] binfmt_flat: remove flat_old_ram_flag

[Greg Ungerer]

On 11/6/19 5:36 pm, Christoph Hellwig wrote:
> On Tue, Jun 11, 2019 at 04:04:39PM +1000, Greg Ungerer wrote:
> > > index c0e4535dc1ec..18d82fd5f57c 100644
> > > --- a/fs/binfmt_flat.c
> > > +++ b/fs/binfmt_flat.c
> > > @@ -488,7 +488,8 @@ static int load_flat_file(struct linux_binprm *bprm,
> > >    	 * fix up the flags for the older format,  there were all kinds
> > >    	 * of endian hacks,  this only works for the simple cases
> > >    	 */
> > > -	if (rev == OLD_FLAT_VERSION && flat_old_ram_flag(flags))
> > > +	if (IS_ENABLED(CONFIG_BINFMT_FLAT_OLD_ALWAYS_RAM) &&
> > > +	    rev == OLD_FLAT_VERSION)
> >
> > The flags are from the binary file header here, so this is going to lose
> > that check for most platforms (except h8300 where it would always have
> > been true).
>
> Indeed.  The old code is:
>
> 	if (rev == OLD_FLAT_VERSION && flat_old_ram_flag(flags))
> 		flags = FLAT_FLAG_RAM;
>
> which for !h8300 evaluates to:
>
> 	if (rev == OLD_FLAT_VERSION && flags)
> 		flags = FLAT_FLAG_RAM;
>
> so basically if any flag was set it was turned into FLAT_FLAG_RAM.
> Was that really intentional?

Probably not, looking at the flags. For the compressed flag it
makes some sense. But I don't think many of the others need load
to RAM behavior.


>  I guess even if it wasn't the is no
> point in changing this historic behavior now.
>
> So I guess what we could do it something like:
>
> 	if (rev == OLD_FLAT_VERSION &&
> 	    (flags || IS_ENABLED(CONFIG_BINFMT_FLAT_OLD_ALWAYS_RAM)))
> 		flags = FLAT_FLAG_RAM;

Yeah, that to looks to preserve the old behavior.

Regards
Greg