On Mon, 18 Feb 2019 18:01:46 +0100
Martin Schwidefsky <schwidefsky@xxxxxxxxxx> wrote:
> On Mon, 18 Feb 2019 07:46:40 -0800
> Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
>
> > Hi,
> >
> > On Thu, Feb 14, 2019 at 03:40:56PM +0100, Martin Schwidefsky wrote:
> > > The setup_lowcore() function creates a new prefix page for the boot CPU.
> > > The PSW mask for the system_call, external interrupt, i/o interrupt and
> > > the program check handler have the DAT bit set in this new prefix page.
> > >
> > > At the time setup_lowcore is called the system still runs without virtual
> > > address translation, the paging_init() function creates the kernel page
> > > table and loads the CR13 with the kernel ASCE.
> > >
> > > Any code between setup_lowcore() and the end of paging_init() that has
> > > a BUG or WARN statement will create a program check that can not be
> > > handled correctly as there is no kernel page table yet.
> > >
> > > To allow early WARN statements initially setup the lowcore with DAT off
> > > and set the DAT bit only after paging_init() has completed.
> > >
> > > Cc: stable@xxxxxxxxxxxxxxx
> > > Signed-off-by: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
> >
> > This patch causes s390 qemu emulations to crash with a kernel stack overflow.
> > Reverting the patch fixes the problem. Crash log and bisect results below.
>
> Urgs, yes. That is EDAT-1 again that makes it work with 1MB pages but breaks
> with 4K mapping where the prefix page is mapped to absolute zero.
>
> Just using S390_lowcore instead of lowcore_ptr[0] does not work either
> because low-address protection is already active. I'll think of something.
>
> Thanks for bug report!
This patch should fix the problem:
--
>From d4393e82c3ec9b2fe5dba4b0d1b6eef29f8d15c8 Mon Sep 17 00:00:00 2001
From: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
Date: Mon, 18 Feb 2019 18:10:08 +0100
Subject: [PATCH] s390/setup: fix boot crash for machine without EDAT-1
The fix to make WARN work in the early boot code created a problem
on older machines without EDAT-1. The setup_lowcore_dat_on function
uses the pointer from lowcore_ptr[0] to set the DAT bit in the new
PSWs. That does not work if the kernel page table is set up with
4K pages as the prefix address maps to absolute zero.
To make this work the PSWs need to be changed with via address 0 in
form of the S390_lowcore definition.
Cc: stable@xxxxxxxxxxxxxxx
Fixes: 94f85ed3e2 ("s390/setup: fix early warning messages")
Signed-off-by: Martin Schwidefsky <schwidefsky@xxxxxxxxxx>
---
arch/s390/kernel/setup.c | 13 ++++++-------
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c
index 65b22ef5141a..12934e8fbb91 100644
--- a/arch/s390/kernel/setup.c
+++ b/arch/s390/kernel/setup.c
@@ -451,13 +451,12 @@ static void __init setup_lowcore_dat_off(void)
static void __init setup_lowcore_dat_on(void)
{
- struct lowcore *lc;
-
- lc = lowcore_ptr[0];
- lc->external_new_psw.mask |= PSW_MASK_DAT;
- lc->svc_new_psw.mask |= PSW_MASK_DAT;
- lc->program_new_psw.mask |= PSW_MASK_DAT;
- lc->io_new_psw.mask |= PSW_MASK_DAT;
+ __ctl_clear_bit(0, 28);
+ S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
+ S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
+ S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
+ S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
+ __ctl_set_bit(0, 28);
}
static struct resource code_resource = {
--
2.16.4
--
blue skies,
Martin.
"Reality continues to ruin my life." - Calvin.
