2021-03-01 11:51:35

by David Hildenbrand

[permalink] [raw]
Subject: [PATCH v1] microblaze: tag highmem_setup() with __meminit

With commit a0cd7a7c4bc0 ("mm: simplify free_highmem_page() and
free_reserved_page()") the kernel test robot complains about a warning:

WARNING: modpost: vmlinux.o(.text.unlikely+0x23ac): Section mismatch in
reference from the function highmem_setup() to the function
.meminit.text:memblock_is_reserved()

This has been broken ever since microblaze added highmem support,
because memblock_is_reserved() was already tagged with "__init" back then -
most probably the function always got inlined, so we never stumbled over
it.

Reported-by: kernel test robot <[email protected]>
Fixes: 2f2f371f8907 ("microblaze: Highmem support")
Cc: Andrew Morton <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Arvind Sankar <[email protected]>
Cc: Ira Weiny <[email protected]>
Cc: Randy Dunlap <[email protected]>
Cc: Oscar Salvador <[email protected]>
Cc: Anshuman Khandual <[email protected]>
Signed-off-by: David Hildenbrand <[email protected]>
---
arch/microblaze/mm/init.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/microblaze/mm/init.c b/arch/microblaze/mm/init.c
index 181e48782e6c..05cf1fb3f5ff 100644
--- a/arch/microblaze/mm/init.c
+++ b/arch/microblaze/mm/init.c
@@ -52,7 +52,7 @@ static void __init highmem_init(void)
pkmap_page_table = virt_to_kpte(PKMAP_BASE);
}

-static void highmem_setup(void)
+static void __meminit highmem_setup(void)
{
unsigned long pfn;

--
2.29.2


2021-03-02 17:49:49

by Oscar Salvador

[permalink] [raw]
Subject: Re: [PATCH v1] microblaze: tag highmem_setup() with __meminit

On Mon, Mar 01, 2021 at 12:47:49PM +0100, David Hildenbrand wrote:
> With commit a0cd7a7c4bc0 ("mm: simplify free_highmem_page() and
> free_reserved_page()") the kernel test robot complains about a warning:
>
> WARNING: modpost: vmlinux.o(.text.unlikely+0x23ac): Section mismatch in
> reference from the function highmem_setup() to the function
> .meminit.text:memblock_is_reserved()
>
> This has been broken ever since microblaze added highmem support,
> because memblock_is_reserved() was already tagged with "__init" back then -
> most probably the function always got inlined, so we never stumbled over
> it.

It might be good to point out that we need __meminit instead of __init
because microblaze platform does not define CONFIG_ARCH_KEEP_MEMBLOCK,
and __init_memblock fallsback to that.

(I had to go and look as I was puzzled :-) )

Reviewed-by: Oscar Salvador <[email protected]>

>
> Reported-by: kernel test robot <[email protected]>
> Fixes: 2f2f371f8907 ("microblaze: Highmem support")
> Cc: Andrew Morton <[email protected]>
> Cc: Michal Simek <[email protected]>
> Cc: Mike Rapoport <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: Thomas Gleixner <[email protected]>
> Cc: Arvind Sankar <[email protected]>
> Cc: Ira Weiny <[email protected]>
> Cc: Randy Dunlap <[email protected]>
> Cc: Oscar Salvador <[email protected]>
> Cc: Anshuman Khandual <[email protected]>
> Signed-off-by: David Hildenbrand <[email protected]>
> ---
> arch/microblaze/mm/init.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/microblaze/mm/init.c b/arch/microblaze/mm/init.c
> index 181e48782e6c..05cf1fb3f5ff 100644
> --- a/arch/microblaze/mm/init.c
> +++ b/arch/microblaze/mm/init.c
> @@ -52,7 +52,7 @@ static void __init highmem_init(void)
> pkmap_page_table = virt_to_kpte(PKMAP_BASE);
> }
>
> -static void highmem_setup(void)
> +static void __meminit highmem_setup(void)
> {
> unsigned long pfn;
>
> --
> 2.29.2
>
>

--
Oscar Salvador
SUSE L3

2021-03-02 21:26:28

by David Hildenbrand

[permalink] [raw]
Subject: Re: [PATCH v1] microblaze: tag highmem_setup() with __meminit

On 01.03.21 23:18, Oscar Salvador wrote:
> On Mon, Mar 01, 2021 at 12:47:49PM +0100, David Hildenbrand wrote:
>> With commit a0cd7a7c4bc0 ("mm: simplify free_highmem_page() and
>> free_reserved_page()") the kernel test robot complains about a warning:
>>
>> WARNING: modpost: vmlinux.o(.text.unlikely+0x23ac): Section mismatch in
>> reference from the function highmem_setup() to the function
>> .meminit.text:memblock_is_reserved()
>>
>> This has been broken ever since microblaze added highmem support,
>> because memblock_is_reserved() was already tagged with "__init" back then -
>> most probably the function always got inlined, so we never stumbled over
>> it.
>
> It might be good to point out that we need __meminit instead of __init
> because microblaze platform does not define CONFIG_ARCH_KEEP_MEMBLOCK,
> and __init_memblock fallsback to that.
>
> (I had to go and look as I was puzzled :-) )
>
> Reviewed-by: Oscar Salvador <[email protected]>

Thanks!

Whoever feels like picking this up (@Andrew?) can you add

"We need __meminit because __init_memblock defaults to that without
CONFIG_ARCH_KEEP_MEMBLOCK" and __init_memblock is not used outside
memblock code.

--
Thanks,

David / dhildenb

2021-03-04 06:25:47

by Michal Simek

[permalink] [raw]
Subject: Re: [PATCH v1] microblaze: tag highmem_setup() with __meminit



On 3/2/21 10:04 AM, David Hildenbrand wrote:
> On 01.03.21 23:18, Oscar Salvador wrote:
>> On Mon, Mar 01, 2021 at 12:47:49PM +0100, David Hildenbrand wrote:
>>> With commit a0cd7a7c4bc0 ("mm: simplify free_highmem_page() and
>>> free_reserved_page()") the kernel test robot complains about a warning:
>>>
>>> WARNING: modpost: vmlinux.o(.text.unlikely+0x23ac): Section mismatch in
>>>    reference from the function highmem_setup() to the function
>>>    .meminit.text:memblock_is_reserved()
>>>
>>> This has been broken ever since microblaze added highmem support,
>>> because memblock_is_reserved() was already tagged with "__init" back
>>> then -
>>> most probably the function always got inlined, so we never stumbled over
>>> it.
>>
>> It might be good to point out that we need __meminit instead of __init
>> because microblaze platform does not define CONFIG_ARCH_KEEP_MEMBLOCK,
>> and __init_memblock fallsback to that.
>>
>> (I had to go and look as I was puzzled :-) )
>>
>> Reviewed-by: Oscar Salvador <[email protected]>
>
> Thanks!
>
> Whoever feels like picking this up (@Andrew?) can you add
>
> "We need __meminit because __init_memblock defaults to that without
> CONFIG_ARCH_KEEP_MEMBLOCK" and __init_memblock is not used outside
> memblock code.
>

Applied with this line added.

Thanks,
Michal


--
Michal Simek, Ing. (M.Eng), OpenPGP -> KeyID: FE3D1F91
w: http://www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel - Xilinx Microblaze
Maintainer of Linux kernel - Xilinx Zynq ARM and ZynqMP ARM64 SoCs
U-Boot custodian - Xilinx Microblaze/Zynq/ZynqMP/Versal SoCs