On Fri, Aug 24, 2018 at 06:19:19PM +0200, Matthias B. wrote:
Bisect identified the problem. It's the attached patch. I applied it to 4.4.152 with patch -Rp1 and I'm running the resulting kernel now.
MSB
-- For every idiot-proof system there exists at least one system-proof idiot.
From 02ff2769edbce2261e981effbc3c4b98fae4faf0 Mon Sep 17 00:00:00 2001
From: Andi Kleen ak@linux.intel.com Date: Tue, 7 Aug 2018 15:09:39 -0700 Subject: [PATCH] x86/mm/pat: Make set_memory_np() L1TF safe
commit 958f79b9ee55dfaf00c8106ed1c22a2919e0028b upstream
set_memory_np() is used to mark kernel mappings not present, but it has it's own open coded mechanism which does not have the L1TF protection of inverting the address bits.
Replace the open coded PTE manipulation with the L1TF protecting low level PTE routines.
Passes the CPA self test.
Signed-off-by: Andi Kleen ak@linux.intel.com Signed-off-by: Thomas Gleixner tglx@linutronix.de [ dwmw2: Pull in pud_mkhuge() from commit a00cc7d9dd, and pfn_pud() ] Signed-off-by: David Woodhouse dwmw@amazon.co.uk [groeck: port to 4.4] Signed-off-by: Guenter Roeck linux@roeck-us.net Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
arch/x86/include/asm/pgtable.h | 27 +++++++++++++++++++++++++++ arch/x86/mm/pageattr.c | 8 ++++---- 2 files changed, 31 insertions(+), 4 deletions(-)
<snip>
Guenter, another report of this patch causing an issue. Any ideas? I am away from test systems this weekend, but can push out patches if needed.
thanks,
greg k-h