On Thu, Jan 12, 2017 at 12:01:11PM +0000, Will Deacon wrote:

On Thu, Jan 12, 2017 at 01:21:44PM +0900, AKASHI Takahiro wrote:

quoted

On Wed, Jan 11, 2017 at 10:54:05AM +0000, Will Deacon wrote:

quoted

On Wed, Jan 11, 2017 at 03:36:28PM +0900, AKASHI Takahiro wrote:

quoted

On Tue, Jan 10, 2017 at 11:32:48AM +0000, Will Deacon wrote:

quoted

On Wed, Dec 28, 2016 at 01:36:01PM +0900, AKASHI Takahiro wrote:

quoted

@@ -22,6 +25,7 @@
 extern const unsigned char arm64_relocate_new_kernel[];
 extern const unsigned long arm64_relocate_new_kernel_size;
 
+static bool in_crash_kexec;

Do you actually need this bool? Why not call kexec_crash_loaded() instead?

The two have different meanings:
"in_crash_kexec" indicates that kdump is taking place, while
kexec_crash_loaded() tells us only whether crash dump kernel has been
loaded or not.

It is crucial to distinguish them especially for machine_kexec()
which can be called on normal kexec even if kdump has been set up.

Ah, I see. So how about just doing:

  if (kimage == kexec_crash_image)

in machine_kexec?

Yeah, it should work.
Do you want to merge the following hunk,
or expect that I will re-send the whole patch series
(with other changes if any)?

Thanks, I'll fold it in and shout if I run into any problems. My plan is
to queue this for 4.11.

Given the DT discussion with Mark, I assume you'll post a new version with
this rolled in.

Will