[PATCH v2 0/6] MIPS: mm: Fix some memory-related issues

[PATCH v2 0/6] MIPS: mm: Fix some memory-related issues

[Serge Semin]

Just recently I've rebased my MIPS32-related work from kernel 6.5-rc4 onto
the latest kernel 6.7-rc1 and immediately got into a bootup-time
mm-related bug (see patches 3-4 in this series). After fixing it I decided
it was time to submit for review the generic MIPS code fixes which I have
been collecting in my local repo for the last year. I was going to submit
them a bit later after I finished working on a patchset with my SoC
arch-specific changes, but since it was getting bigger and bigger, it
turned to be reasonable to spill out the generic part of series right away
especially seeing it might get to be useful in the most recent kernel.

So this series starts with the MIPS-specific dmi_early_remap()
implementation fix. It is utilized by the DMI driver in the framework of
the dmi_setup() method, which is called at the very early boot stage - in
setup_arch((). No VM and slab available at that stage which is required
for the ioremap_cache() to properly work. Thus it was a mistake to have
the dmi_early_remap() macro-function defined as ioremap_cache(). It should
have been ioremap() in first place.

After that goes a fix for the high-memory zone PFNs calculation procedure
on MIPS. It turned out that after some not that recent commit the
IO-memory or just non-memory PFNs got to the high-memory even though they
were directly reachable, thus should have been left in the normal zone.

Then a series of fixes for the recently discovered mm-bug is presented.
Any attempt to re-map the IO-memory with the cached attribute caused the
bootup procedure to crash with the "Unhandled kernel unaligned access"
message. After some digging I found out that the problem was in the
uninitialized IO-memory pages. Please see the patch "mips: Fix max_mapnr
being uninitialized on early stages" description for the detailed
explanation of the problem and suggested fix.

After that goes a patch which adds the slab availability check into the
ioremap_prot() method. Indeed VM mapping performs the slab allocation in
the framework of the get_vm_area() method. Thus any other than uncached
IO-remappings must be proceeded only at the stages when slab is available.
A similar fix was just recently added to the generic version of
ioremap_prot() in the framework of commit a5f616483110 ("mm/ioremap: add
slab availability checking in ioremap_prot").

The patchset is closed with a small improvement which sets the MIPS
board/machine name to the dump-stack module in order to print
arch-personalized oopses in the same way as it's done on ARM, ARM64,
RISC-V, etc.

That's it for today.) Thanks for review in advance. Any tests are very
welcome.

Link: https://lore.kernel.org/linux-mips/20231122182419.30633-1-fancer.lancer@gmail.com/
Changelog v2:
- Drop the patches:
  [PATCH 5/7] mm/mm_init.c: Extend init unavailable range doc info
  [PATCH 6/7] mm/mm_init.c: Append '\n' to the unavailable ranges log-message
  since they have been picked up by Andrew. (@Andrew)
- Replace ioremap_uc() with using ioremap() due to having the former one
  deprecated. (@Arnd)
- Add a new patch:
  [PATCH v2 5/6] mips: mm: add slab availability checking in ioremap_prot
  picked up from the generic ioremap_prot() implementation.
- Extend early DMI mem remapping patch log with a note regarding the unsynched
  caches concern. (@Jiaxun)

Signed-off-by: Serge Semin <fancer.lancer@gmail.com>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Aleksandar Rikalo <aleksandar.rikalo@syrmia.com>
Cc: Aleksandar Rikalo <arikalo@gmail.com>
Cc: Dragan Mladjenovic <dragan.mladjenovic@syrmia.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Baoquan He <bhe@redhat.com>
Cc: Chao-ying Fu <cfu@wavecomp.com>
Cc: Yinglu Yang <yangyinglu@loongson.cn>
Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Marc Zyngier <maz@kernel.org>
Cc: linux-mips@vger.kernel.org
Cc: linux-mm@kvack.org
Cc: linux-kernel@vger.kernel.org

Serge Semin (6):
  mips: dmi: Fix early remap on MIPS32
  mips: Fix incorrect max_low_pfn adjustment
  mips: Fix max_mapnr being uninitialized on early stages
  mips: Optimize max_mapnr init procedure
  mips: mm: add slab availability checking in ioremap_prot
  mips: Set dump-stack arch description

 arch/mips/include/asm/dmi.h |  2 +-
 arch/mips/kernel/prom.c     |  2 ++
 arch/mips/kernel/setup.c    |  4 ++--
 arch/mips/mm/init.c         | 16 +++++++++-------
 arch/mips/mm/ioremap.c      |  4 ++++
 5 files changed, 18 insertions(+), 10 deletions(-)

-- 
2.42.1

[PATCH v2 1/6] mips: dmi: Fix early remap on MIPS32

[Serge Semin]

dmi_early_remap() has been defined as ioremap_cache() which on MIPS32 gets
to be converted to the VM-based mapping. DMI early remapping is performed
at the setup_arch() stage with no VM available. So calling the
dmi_early_remap() for MIPS32 causes the system to crash at the early boot
time. Fix that by converting dmi_early_remap() to the uncached remapping
which is always available on both 32 and 64-bits MIPS systems.

Note this change shall not cause any regressions on the current DMI
support implementation because on the early boot-up stage neither MIPS32
nor MIPS64 has the cacheable ioremapping support anyway.

Fixes: be8fa1cb444c ("MIPS: Add support for Desktop Management Interface (DMI)")
Signed-off-by: Serge Semin <fancer.lancer@gmail.com>

---

Note even though this patch is fully correct from the current
ioremap()-based semantics point of view and shall fix the denoted problem,
Jiaxun thinks that it's better to provide a different fix since
dmi_early_remap() doesn't work correctly on even Loongson64 - the only
currently DMI-equipped platform. In v1 discussion he promised to provide a
better fix for the problem. Until then let's consider this patch as the
only currently available solution.

Changelog v2:
- Replace ioremap_uc() with using ioremap() due to having the former one
  deprecated. (@Arnd)
- Extend patch log with a note regarding the unsynched caches concern.
  (@Jiaxun)
---
 arch/mips/include/asm/dmi.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/mips/include/asm/dmi.h b/arch/mips/include/asm/dmi.h
index 27415a288adf..dc397f630c66 100644
--- a/arch/mips/include/asm/dmi.h
+++ b/arch/mips/include/asm/dmi.h
@@ -5,7 +5,7 @@
 #include <linux/io.h>
 #include <linux/memblock.h>
 
-#define dmi_early_remap(x, l)		ioremap_cache(x, l)
+#define dmi_early_remap(x, l)		ioremap(x, l)
 #define dmi_early_unmap(x, l)		iounmap(x)
 #define dmi_remap(x, l)			ioremap_cache(x, l)
 #define dmi_unmap(x)			iounmap(x)
-- 
2.42.1

[PATCH v2 6/6] mips: Set dump-stack arch description

[Serge Semin]

In the framework of the MIPS architecture the mips_set_machine_name()
method is defined to set the machine name. The name currently is only used
in the /proc/cpuinfo file content generation. Let's have it utilized to
mach-personalize the dump-stack data too in a way it's done on ARM, ARM64,
RISC-V, etc.

Signed-off-by: Serge Semin <fancer.lancer@gmail.com>
---
 arch/mips/kernel/prom.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/mips/kernel/prom.c b/arch/mips/kernel/prom.c
index f88ce78e13e3..6062e6fa589a 100644
--- a/arch/mips/kernel/prom.c
+++ b/arch/mips/kernel/prom.c
@@ -28,6 +28,8 @@ __init void mips_set_machine_name(const char *name)
 
 	strscpy(mips_machine_name, name, sizeof(mips_machine_name));
 	pr_info("MIPS: machine is %s\n", mips_get_machine_name());
+
+	dump_stack_set_arch_desc(name);
 }
 
 char *mips_get_machine_name(void)
-- 
2.42.1

Re: [PATCH v2 0/6] MIPS: mm: Fix some memory-related issues

[Thomas Bogendoerfer]

On Sat, Dec 02, 2023 at 02:14:17PM +0300, Serge Semin wrote:
> Just recently I've rebased my MIPS32-related work from kernel 6.5-rc4 onto
> the latest kernel 6.7-rc1 and immediately got into a bootup-time
> mm-related bug (see patches 3-4 in this series). After fixing it I decided
> it was time to submit for review the generic MIPS code fixes which I have
> been collecting in my local repo for the last year. I was going to submit
> them a bit later after I finished working on a patchset with my SoC
> arch-specific changes, but since it was getting bigger and bigger, it
> turned to be reasonable to spill out the generic part of series right away
> especially seeing it might get to be useful in the most recent kernel.
> 
> So this series starts with the MIPS-specific dmi_early_remap()
> implementation fix. It is utilized by the DMI driver in the framework of
> the dmi_setup() method, which is called at the very early boot stage - in
> setup_arch((). No VM and slab available at that stage which is required
> for the ioremap_cache() to properly work. Thus it was a mistake to have
> the dmi_early_remap() macro-function defined as ioremap_cache(). It should
> have been ioremap() in first place.
> 
> After that goes a fix for the high-memory zone PFNs calculation procedure
> on MIPS. It turned out that after some not that recent commit the
> IO-memory or just non-memory PFNs got to the high-memory even though they
> were directly reachable, thus should have been left in the normal zone.
> 
> Then a series of fixes for the recently discovered mm-bug is presented.
> Any attempt to re-map the IO-memory with the cached attribute caused the
> bootup procedure to crash with the "Unhandled kernel unaligned access"
> message. After some digging I found out that the problem was in the
> uninitialized IO-memory pages. Please see the patch "mips: Fix max_mapnr
> being uninitialized on early stages" description for the detailed
> explanation of the problem and suggested fix.
> 
> After that goes a patch which adds the slab availability check into the
> ioremap_prot() method. Indeed VM mapping performs the slab allocation in
> the framework of the get_vm_area() method. Thus any other than uncached
> IO-remappings must be proceeded only at the stages when slab is available.
> A similar fix was just recently added to the generic version of
> ioremap_prot() in the framework of commit a5f616483110 ("mm/ioremap: add
> slab availability checking in ioremap_prot").
> 
> The patchset is closed with a small improvement which sets the MIPS
> board/machine name to the dump-stack module in order to print
> arch-personalized oopses in the same way as it's done on ARM, ARM64,
> RISC-V, etc.
> 
> That's it for today.) Thanks for review in advance. Any tests are very
> welcome.
> 
> Link: https://lore.kernel.org/linux-mips/20231122182419.30633-1-fancer.lancer@gmail.com/
> Changelog v2:
> - Drop the patches:
>   [PATCH 5/7] mm/mm_init.c: Extend init unavailable range doc info
>   [PATCH 6/7] mm/mm_init.c: Append '\n' to the unavailable ranges log-message
>   since they have been picked up by Andrew. (@Andrew)
> - Replace ioremap_uc() with using ioremap() due to having the former one
>   deprecated. (@Arnd)
> - Add a new patch:
>   [PATCH v2 5/6] mips: mm: add slab availability checking in ioremap_prot
>   picked up from the generic ioremap_prot() implementation.
> - Extend early DMI mem remapping patch log with a note regarding the unsynched
>   caches concern. (@Jiaxun)
> 
> Signed-off-by: Serge Semin <fancer.lancer@gmail.com>
> Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
> Cc: Arnd Bergmann <arnd@arndb.de>
> Cc: Aleksandar Rikalo <aleksandar.rikalo@syrmia.com>
> Cc: Aleksandar Rikalo <arikalo@gmail.com>
> Cc: Dragan Mladjenovic <dragan.mladjenovic@syrmia.com>
> Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
> Cc: Baoquan He <bhe@redhat.com>
> Cc: Chao-ying Fu <cfu@wavecomp.com>
> Cc: Yinglu Yang <yangyinglu@loongson.cn>
> Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
> Cc: Mike Rapoport <rppt@kernel.org>
> Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
> Cc: Marc Zyngier <maz@kernel.org>
> Cc: linux-mips@vger.kernel.org
> Cc: linux-mm@kvack.org
> Cc: linux-kernel@vger.kernel.org
> 
> Serge Semin (6):
>   mips: dmi: Fix early remap on MIPS32
>   mips: Fix incorrect max_low_pfn adjustment
>   mips: Fix max_mapnr being uninitialized on early stages
>   mips: Optimize max_mapnr init procedure
>   mips: mm: add slab availability checking in ioremap_prot
>   mips: Set dump-stack arch description
> 
>  arch/mips/include/asm/dmi.h |  2 +-
>  arch/mips/kernel/prom.c     |  2 ++
>  arch/mips/kernel/setup.c    |  4 ++--
>  arch/mips/mm/init.c         | 16 +++++++++-------
>  arch/mips/mm/ioremap.c      |  4 ++++
>  5 files changed, 18 insertions(+), 10 deletions(-)

series applied to mips-next.

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]