2010-08-25 05:44:23

by Hugh Dickins

[permalink] [raw]
Subject: [PATCH] x86: fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline

rc2 kernel crashes when booting second cpu on this CONFIG_VMSPLIT_2G_OPT
laptop: whereas cloning from kernel to low mappings pgd range does need
to limit by both KERNEL_PGD_PTRS and KERNEL_PGD_BOUNDARY, cloning kernel
pgd range itself must not be limited by the smaller KERNEL_PGD_BOUNDARY.

Signed-off-by: Hugh Dickins <[email protected]>
---

arch/x86/kernel/trampoline.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

--- 2.6.36-rc2/arch/x86/kernel/trampoline.c.0 2010-08-22 18:56:44.000000000 -0700
+++ linux/arch/x86/kernel/trampoline.c 2010-08-24 22:11:08.000000000 -0700
@@ -45,8 +45,7 @@ void __init setup_trampoline_page_table(
/* Copy kernel address range */
clone_pgd_range(trampoline_pg_dir + KERNEL_PGD_BOUNDARY,
swapper_pg_dir + KERNEL_PGD_BOUNDARY,
- min_t(unsigned long, KERNEL_PGD_PTRS,
- KERNEL_PGD_BOUNDARY));
+ KERNEL_PGD_PTRS);

/* Initialize low mappings */
clone_pgd_range(trampoline_pg_dir,


Subject: Re: [PATCH] x86: fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline

From: Hugh Dickins <[email protected]>
Date: Wed, Aug 25, 2010 at 01:44:12AM -0400

> rc2 kernel crashes when booting second cpu on this CONFIG_VMSPLIT_2G_OPT
> laptop: whereas cloning from kernel to low mappings pgd range does need
> to limit by both KERNEL_PGD_PTRS and KERNEL_PGD_BOUNDARY, cloning kernel
> pgd range itself must not be limited by the smaller KERNEL_PGD_BOUNDARY.

Yep, makes sense.

In looking around for other swapper_pg_dir manipulations, there's this
piece in <arch/x86/kernel/reboot.c:machine_real_restart()>:

/* Remap the kernel at virtual address zero, as well as offset zero
from the kernel segment. This assumes the kernel segment starts at
virtual address PAGE_OFFSET. */
memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);

and we're cloning to low mappings here but still take KERNEL_PGD_PTRS.

In the CONFIG_VMSPLIT_2G_OPT case, KERNEL_PGD_BOUNDARY and
KERNEL_PGD_PTRS are 480 and 544, respectively, and IINM, and we end up
overwriting the virtual mappings beginning at PAGE_OFFSET.

I guess this doesn't matter on reboot or am I missing something else?

>
> Signed-off-by: Hugh Dickins <[email protected]>

Acked-by: Borislav Petkov <[email protected]>

> ---
>
> arch/x86/kernel/trampoline.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> --- 2.6.36-rc2/arch/x86/kernel/trampoline.c.0 2010-08-22 18:56:44.000000000 -0700
> +++ linux/arch/x86/kernel/trampoline.c 2010-08-24 22:11:08.000000000 -0700
> @@ -45,8 +45,7 @@ void __init setup_trampoline_page_table(
> /* Copy kernel address range */
> clone_pgd_range(trampoline_pg_dir + KERNEL_PGD_BOUNDARY,
> swapper_pg_dir + KERNEL_PGD_BOUNDARY,
> - min_t(unsigned long, KERNEL_PGD_PTRS,
> - KERNEL_PGD_BOUNDARY));
> + KERNEL_PGD_PTRS);
>
> /* Initialize low mappings */
> clone_pgd_range(trampoline_pg_dir,
>

--
Regards/Gruss,
Boris.

Advanced Micro Devices GmbH
Einsteinring 24, 85609 Dornach
General Managers: Alberto Bozzo, Andrew Bowd
Registration: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632

2010-08-25 07:06:46

by H. Peter Anvin

[permalink] [raw]
Subject: Re: [PATCH] x86: fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline

On 08/24/2010 11:30 PM, Borislav Petkov wrote:
> From: Hugh Dickins <[email protected]>
> Date: Wed, Aug 25, 2010 at 01:44:12AM -0400
>
>> rc2 kernel crashes when booting second cpu on this CONFIG_VMSPLIT_2G_OPT
>> laptop: whereas cloning from kernel to low mappings pgd range does need
>> to limit by both KERNEL_PGD_PTRS and KERNEL_PGD_BOUNDARY, cloning kernel
>> pgd range itself must not be limited by the smaller KERNEL_PGD_BOUNDARY.
>
> Yep, makes sense.
>
> In looking around for other swapper_pg_dir manipulations, there's this
> piece in <arch/x86/kernel/reboot.c:machine_real_restart()>:
>
> /* Remap the kernel at virtual address zero, as well as offset zero
> from the kernel segment. This assumes the kernel segment starts at
> virtual address PAGE_OFFSET. */
> memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
> sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);
>
> and we're cloning to low mappings here but still take KERNEL_PGD_PTRS.
>
> In the CONFIG_VMSPLIT_2G_OPT case, KERNEL_PGD_BOUNDARY and
> KERNEL_PGD_PTRS are 480 and 544, respectively, and IINM, and we end up
> overwriting the virtual mappings beginning at PAGE_OFFSET.
>
> I guess this doesn't matter on reboot or am I missing something else?
>

It doesn't matter on reboot, but the right thing would be to switch to
the trampoline page table here. Same with suspend.

-hpa

--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.

Subject: Re: [PATCH] x86: fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline

From: "H. Peter Anvin" <[email protected]>
Date: Wed, Aug 25, 2010 at 03:06:34AM -0400

> It doesn't matter on reboot, but the right thing would be to switch to
> the trampoline page table here. Same with suspend.

Yeah, I do that in my other patch (http://lkml.org/lkml/2010/8/24/61)
which adds the initial_page_table and keeps swapper_pg_dir clear of low
mappings. I'll add Hugh's fix there too and will start testing to see
what happens :).

--
Regards/Gruss,
Boris.

Advanced Micro Devices GmbH
Einsteinring 24, 85609 Dornach
General Managers: Alberto Bozzo, Andrew Bowd
Registration: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632

2010-08-25 07:14:45

by Hugh Dickins

[permalink] [raw]
Subject: [tip:x86/urgent] x86, mm: Fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline

Commit-ID: b7d460897739e02f186425b7276e3fdb1595cea7
Gitweb: http://git.kernel.org/tip/b7d460897739e02f186425b7276e3fdb1595cea7
Author: Hugh Dickins <[email protected]>
AuthorDate: Tue, 24 Aug 2010 22:44:12 -0700
Committer: H. Peter Anvin <[email protected]>
CommitDate: Tue, 24 Aug 2010 23:05:17 -0700

x86, mm: Fix CONFIG_VMSPLIT_1G and 2G_OPT trampoline

rc2 kernel crashes when booting second cpu on this CONFIG_VMSPLIT_2G_OPT
laptop: whereas cloning from kernel to low mappings pgd range does need
to limit by both KERNEL_PGD_PTRS and KERNEL_PGD_BOUNDARY, cloning kernel
pgd range itself must not be limited by the smaller KERNEL_PGD_BOUNDARY.

Signed-off-by: Hugh Dickins <[email protected]>
LKML-Reference: <[email protected]>
Signed-off-by: H. Peter Anvin <[email protected]>
---
arch/x86/kernel/trampoline.c | 3 +--
1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/trampoline.c b/arch/x86/kernel/trampoline.c
index a874495..e2a5952 100644
--- a/arch/x86/kernel/trampoline.c
+++ b/arch/x86/kernel/trampoline.c
@@ -45,8 +45,7 @@ void __init setup_trampoline_page_table(void)
/* Copy kernel address range */
clone_pgd_range(trampoline_pg_dir + KERNEL_PGD_BOUNDARY,
swapper_pg_dir + KERNEL_PGD_BOUNDARY,
- min_t(unsigned long, KERNEL_PGD_PTRS,
- KERNEL_PGD_BOUNDARY));
+ KERNEL_PGD_PTRS);

/* Initialize low mappings */
clone_pgd_range(trampoline_pg_dir,