On Wed, Feb 13, 2019 at 07:44:48PM +0100, Christoph Hellwig wrote:
> On Wed, Feb 13, 2019 at 08:41:40PM +0200, Mike Rapoport wrote:
> > csky seems to open-code free_reserved_page with the only
> > difference that it's also increments totalram_pages for the freed pages,
> > which doesn't seem correct anyway...
> >
> > That said, I suppose arch/csky can be also added to the party.
>
> Yes, I noticed that. But I'd rather move it over manually in
> another patch post rc1 or for the next merge window.
Fair enough.
> > > +void __weak free_initrd_mem(unsigned long start, unsigned long end)
> > > +{
> > > + free_reserved_area((void *)start, (void *)end, -1, "initrd");
> >
> > Some architectures have pr_info("Freeing initrd memory..."), I'd add it for
> > the generic version as well.
>
> Well, if we think such a printk is useful it should probably be
> moved to the caller in init/initramfs.c instead. I can include a
> patch for that in the next iteration of the series.
I found it useful during board bring ups, this gave some starting point
when everything hangs and you are out to catch the lion in the desert.
> > Another thing that I was thinking of is that x86 has all those memory
> > protection calls in its free_initrd_mem, maybe it'd make sense to have them
> > in the generic version as well?
>
> Maybe. But I'd rather keep it out of the initial series as it looks
> a little more complicated. Having a single implementation
> of free_initrd_mem would be great, though.
Ok.
BTW, the memblock_free() arm64 does, seems to be relevant for architectures
with CONFIG_ARCH_DISCARD_MEMBLOCK=n.
On powerpc the freed initrd region shows up in
/sys/kernel/debug/memblock/reserved.
--
Sincerely yours,
Mike.
