[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20180830180849.nmh646ehnolphq4v@suse.de>
Date: Thu, 30 Aug 2018 20:08:49 +0200
From: Joerg Roedel <jroedel@...e.de>
To: Guenter Roeck <linux@...ck-us.net>
Cc: linux-kernel@...r.kernel.org, Thomas Gleixner <tglx@...utronix.de>,
Michal Hocko <mhocko@...e.com>,
Andi Kleen <ak@...ux.intel.com>,
Linus Torvalds <torvalds@...ux-foundation.org>, x86@...nel.org,
Dave Hansen <dave.hansen@...el.com>,
Pavel Machek <pavel@....cz>
Subject: Re: efi boot failures due to PTI with 32 bit builds and Intel CPUs
On Thu, Aug 30, 2018 at 05:20:22AM -0700, Guenter Roeck wrote:
> Excellent.
>
> In case it matters: I can use the same 32-bit efi image to boot a 64-bit kernel
> without problems.
Okay, I tracked it down. The 32-bit EFI code switches the CPU to the
read-write version of the GDT which is not mapped in the user
page-table. This obviously faults when the kernel entry/exit code
restores the segment registers.
Without a mapped GDT the #PF and #DF handlers also can't be started, so
the machine triple-faults. Below diff fixes it for me, I'll send a
proper patch tomorrow.
diff --git a/arch/x86/platform/efi/efi_32.c b/arch/x86/platform/efi/efi_32.c
index 324b93328b37..892edb4664fb 100644
--- a/arch/x86/platform/efi/efi_32.c
+++ b/arch/x86/platform/efi/efi_32.c
@@ -87,7 +87,7 @@ void __init efi_call_phys_epilog(pgd_t *save_pgd)
{
struct desc_ptr gdt_descr;
- gdt_descr.address = (unsigned long)get_cpu_gdt_rw(0);
+ gdt_descr.address = (unsigned long)get_cpu_gdt_ro(0);
gdt_descr.size = GDT_SIZE - 1;
load_gdt(&gdt_descr);
Powered by blists - more mailing lists