Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
Message-ID: <20231010145130.987300357@linutronix.de>
From:   Thomas Gleixner <tglx@linutronix.de>
To:     LKML <linux-kernel@vger.kernel.org>
Cc:     x86@kernel.org, Borislav Petkov <bp@alien8.de>
Subject: [patch 0/5] x86/microcode/amd: Use cached microcode for early loading
Date:   Tue, 10 Oct 2023 17:08:40 +0200 (CEST)
Precedence: bulk

When Borislav applied the microcode overhaul series:

     https://lore.kernel.org/all/20231002115506.217091296@linutronix.de/

he partially reverted patch 15:

     https://lore.kernel.org/all/20231006132617.GOZSAK+YCEwVsEuUdr@fat_crate.local/

to keep AMD suspend/resume working.

After looking at this in detail it turned out that AMD does not use the
microcode cache for the early loading during AP bringup and tries to find
the microcode in builtin/initrd.

That in turn requires to have a CPU hotplug callback for late hotplug which
ensures that the microcode is updated. That's actually too late as CPUID
bits are evaluated already before that callback is invoked.

The reason why this was done is the horrible 32bit hack which invoked the
early loader before paging is enabled.

That's gone now which in turn allows to remove all this by initializing the
microcode cache before the APs are brought up and use the microcode cache
for early loading on APs during boot, manual hotplug and resume.

This gets finally rid of the hotplug callback and the ugly initrd_gone
hackery which was required to make the post init AP loader work.

The series applies on top of:

 git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86/microcode

Thanks,

	tglx
---
 amd.c      |   34 +++++++++++++++++++++-------------
 core.c     |   44 +++-----------------------------------------
 internal.h |    4 ----
 3 files changed, 24 insertions(+), 58 deletions(-)