[PATCH] arm64:free_initrd_mem should also free the memblock

Russell King - ARM Linux linux at arm.linux.org.uk
Wed Sep 17 11:12:54 PDT 2014


On Wed, Sep 17, 2014 at 05:28:23PM +0100, Catalin Marinas wrote:
> On Tue, Sep 16, 2014 at 02:53:55AM +0100, Wang, Yalin wrote:
> > The reason that a want merge this patch is that
> > It confuse me when I debug memory issue by 
> > /sys/kernel/debug/memblock/reserved  debug file,
> > It show lots of un-correct reserved memory.
> > In fact, I also send a patch to cma driver part
> > For this issue too:
> > http://ozlabs.org/~akpm/mmots/broken-out/free-the-reserved-memblock-when-free-cma-pages.patch
> > 
> > I want to remove these un-correct memblock parts as much as possible,
> > so that I can see more correct info from /sys/kernel/debug/memblock/reserved
> > debug file .
> 
> Could we not always call memblock_free() from free_reserved_area() (with
> a dummy definition when !CONFIG_HAVE_MEMBLOCK)?

Why bother?

The next thing is that people will want to have memblock's reserved areas
track whether the kernel allocates a page so that the memblock debugging
follows the kernel's allocation state.

This is utterly rediculous.  Memblock is purely a method to get the system
up and running.  Once it hands memory over to the normal kernel allocators,
the reservation information in memblock is no longer valid.

The /useful/ information that it provides is the state of memory passed
over to the kernel allocators, which in itself is valuable information.
Destroying it by freeing stuff after that point is not useful.

-- 
FTTC broadband for 0.8mile line: currently at 9.5Mbps down 400kbps up
according to speedtest.net.



More information about the linux-arm-kernel mailing list