Re: [PATCH 4.19 200/321] mm/page_alloc.c: deduplicate __memblock_free_early() and memblock_free()

[Greg Kroah-Hartman]

On Thu, Dec 05, 2019 at 03:11:28PM +0200, Mike Rapoport wrote:
> On Thu, Dec 05, 2019 at 12:50:43PM +0100, Pavel Machek wrote:
> > Hi!
> > On Tue 2019-12-03 23:34:26, Greg Kroah-Hartman wrote:
> > > From: Wentao Wang <witallwang@xxxxxxxxx>
> > > 
> > > [ Upstream commit d31cfe7bff9109476da92c245b56083e9b48d60a ]
> > 
> > 
> > > @@ -1537,12 +1537,7 @@ void * __init memblock_virt_alloc_try_nid(
> > >   */
> > >  void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
> > >  {
> > > -	phys_addr_t end = base + size - 1;
> > > -
> > > -	memblock_dbg("%s: [%pa-%pa] %pF\n",
> > > -		     __func__, &base, &end, (void *)_RET_IP_);
> > > -	kmemleak_free_part_phys(base, size);
> > > -	memblock_remove_range(&memblock.reserved, base, size);
> > > +	memblock_free(base, size);
> > >  }
> > 
> > This makes the memblock_dbg() less useful: _RET_IP_ will now be one of
> > __memblock_free_early(), not of the original caller.
> > 
> > That may be okay, but I guess it should be mentioned in changelog, and
> > I don't really see why it is queued for -stable.
> 
> Not sure why this one was picked for -stable, but in upstream there is a
> followup commit 4d72868c8f7c ("memblock: replace usage of
> __memblock_free_early() with memblock_free()") that completely eliminates
> __memblock_free_early(). IMHO it would make sense to either to take both or
> to drop both.

Ok, I'll try, but that commit does not apply cleanly to 5.0, so it might
take a bit of time...

thanks,

greg k-h