[PATCH] arm64:free_initrd_mem should also free the memblock

On Mon, Sep 15, 2014 at 07:40:23PM +0100, Russell King - ARM Linux wrote:

On Mon, Sep 15, 2014 at 07:33:34PM +0100, Will Deacon wrote:

quoted

On Fri, Sep 12, 2014 at 11:17:18AM +0100, Wang, Yalin wrote:

quoted

this patch fix the memblock statics for memblock
in file /sys/kernel/debug/memblock/reserved
if we don't call memblock_free the initrd will still
be marked as reserved, even they are freed.

Signed-off-by: Yalin Wang <redacted>
---
 arch/arm64/mm/init.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c
index 5472c24..34605c8 100644
--- a/arch/arm64/mm/init.c
+++ b/arch/arm64/mm/init.c

@@ -334,8 +334,10 @@ static int keep_initrd;
 
 void free_initrd_mem(unsigned long start, unsigned long end)
 {
-	if (!keep_initrd)
+	if (!keep_initrd) {
 		free_reserved_area((void *)start, (void *)end, 0, "initrd");
+		memblock_free(__pa(start), end - start);
+	}

I don't think it makes any technical difference, but doing the memblock_free
before the free_reserved_area makes more sense to me.

A better question is... should we even be doing this.  The memblock
information is used as a method to bring up the kernel and provide
early allocation.  Once the memory is handed over from memblock to
the normal kernel page allocators, we no longer care what happens to
memblock.

There is no need to free the initrd memory back into memblock.  In
fact, seeing the initrd location in /sys/kernel/debug/memblock/reserved
can be useful debug information in itself.

That's a fair point. Yang: do you have a specific use-case in mind for this?

I wondered if it might interact with our pfn_valid implementation, which
uses memblock_is_memory, however memblock_free only deals with the reserved
regions, so I now I can't see why this change is required either.

Will