On Wed, Aug 05, 2020 at 12:19:58PM +0100, Marc Zyngier wrote:
On 2020-08-05 10:54, Greg Kroah-Hartman wrote:
On Tue, Aug 04, 2020 at 10:23:06PM +0100, Marc Zyngier wrote:
On 2020-08-04 19:33, Linus Torvalds wrote:
On Tue, Aug 4, 2020 at 1:21 AM Greg Kroah-Hartman gregkh@linuxfoundation.org wrote:
So Linus's tree is also broken here.
No, there's 835d1c3a9879 ("arm64: Drop unnecessary include from asm/smp.h") upstream.
My bet is that Greg ended up with this patch backported to 5.7, but doesn't have 62a679cb2825 ("arm64: simplify ptrauth initialization") as the latter isn't a fix.
I don't think any of these two patches are worth backporting, to be honest.
I didn't have either of those patches, so I can try applying them to see if the build errors go away. But if you don't think they should be applied, what should I do?
Here's what I did have queued up:
f227e3ec3b5c ("random32: update the net random state on interrupt and activity") aa54ea903abb ("ARM: percpu.h: fix build error") 1c9df907da83 ("random: fix circular include dependency on arm64 after addition of percpu.h") 83bdc7275e62 ("random32: remove net_rand_state from the latent entropy gcc plugin") c0842fbc1b18 ("random32: move the pseudo-random 32-bit definitions to prandom.h")
Not what I expected, then. I stand corrected.
And that caused the builds to blow up.
So, what should I do here?
OK, this is getting hairy. I solved it by grabbing:
d0055da5266a ("arm64: remove ptrauth_keys_install_kernel sync arg") 62a679cb2825 ("arm64: simplify ptrauth initialization")
and at which point you might as well take 835d1c3a9879 despite everything I said earlier. And backporting that further down the line is fraught with danger.
I came up with yet another "quality" hack, which gets the job done, see below. It is obviously much simpler, but also terribly ugly.
I like it :)
I've taken it for 5.7.y, and modified it a bit for 5.4.y, and don't think it's needed on anything older, but let's see what blows up...
thanks!
greg k-h