On Mon, May 25, 2020 at 08:21:58AM +0800, kernel test robot wrote:
> Greetings,
>
> 0day kernel testing robot got the below dmesg and the first bad commit is
>
> https://github.com/0day-ci/linux/commits/Mike-Rapoport/mm-consolidate-definitions-of-page-table-accessors/20200513-025551
>
> commit 379706875d28bf7fc90b067355981de242b7bff1
> Author: Mike Rapoport <[email protected]>
> AuthorDate: Tue May 12 21:44:17 2020 +0300
> Commit: 0day robot <[email protected]>
> CommitDate: Wed May 13 02:55:59 2020 +0800
>
> x86/mm: simplify init_trampoline() and surrounding logic
>
> There are three cases for the trampoline initialization:
> * 32-bit does nothing
> * 64-bit with kaslr disabled simply copies a PGD entry from the direct map
> to the trampoline PGD
> * 64-bit with kaslr enabled maps the real mode trampoline at PUD level
>
> These cases are currently differentiated by a bunch of ifdefs inside
> asm/include/pgtable.h and the case of 64-bits with kaslr on uses
> pgd_index() helper.
>
> Replacing the ifdefs with a static function in arch/x86/mm/init.c gives
> clearer code and allows moving pgd_index() to the generic implementation in
> include/linux/pgtable.h
>
> Signed-off-by: Mike Rapoport <[email protected]>
>
> 7cc33e59db m68k/mm: move {cache,nocahe}_page() definitions close to their user
> 379706875d x86/mm: simplify init_trampoline() and surrounding logic
> 6498f3f0af mm: consolidate pgd_index() and pgd_offset{_k}() definitions
> +----------------------------------------------------------------------------+------------+------------+------------+
> | | 7cc33e59db | 379706875d | 6498f3f0af |
> +----------------------------------------------------------------------------+------------+------------+------------+
> | boot_successes | 0 | 0 | 0 |
> | boot_failures | 50 | 17 | 21 |
The patch below fixes this. Its generated vs v5.7-rc6-mmots-2020-05-22-22-39.
From e8877945b2a0b56861e9a263df9d1bc84f792210 Mon Sep 17 00:00:00 2001
From: Mike Rapoport <[email protected]>
Date: Mon, 25 May 2020 12:18:21 +0300
Subject: [PATCH] x86/kaslr: take CONFIG_RANDOMIZE_MEMORY into account in kaslr_enabled()
Until refactoring of the init_trampoline(), kaslr_enabled() was used only
inside '#ifdef CONFIG_RANDOMIZE_MEMORY' and reported only the state of the
command line override.
The simplification of init_trampoline() made kaslr_enabled() visible for
builds with CONFIG_RANDOMIZE_MEMORY=n and in this case it will return true
if the randomization was not disabled on the command line.
This in turn will call a dummy init_trampoline_kaslr() instead of setting
trampline_pgd_entry directly.
As the result, the kernel reboots while starting the secondary CPUs, e.g.
kernel test bot reported:
...
[ 0.244628] Performance Events: unsupported p6 CPU model 60 no PMU driver, software events only.
[ 0.244993] rcu: Hierarchical SRCU implementation.
[ 0.245617] smp: Bringing up secondary CPUs ...
[ 0.247388] x86: Booting SMP configuration:
BUG: kernel reboot-without-warning in boot stage
Adding a check for IS_ENABLED(CONFIG_RANDOMIZE_MEMORY) to kaslr_enabled()
resolves the issue.
Signed-off-by: Mike Rapoport <[email protected]>
Reported-by: kbuild test robot <[email protected]>
---
arch/x86/include/asm/setup.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/arch/x86/include/asm/setup.h b/arch/x86/include/asm/setup.h
index d95cacf210bb..84b645cc8bc9 100644
--- a/arch/x86/include/asm/setup.h
+++ b/arch/x86/include/asm/setup.h
@@ -75,7 +75,8 @@ extern char _text[];
static inline bool kaslr_enabled(void)
{
- return !!(boot_params.hdr.loadflags & KASLR_FLAG);
+ return IS_ENABLED(CONFIG_RANDOMIZE_MEMORY) &&
+ !!(boot_params.hdr.loadflags & KASLR_FLAG);
}
/*
--
2.26.2