[RESEND PATCH] mm/sparsemem: use wrapped macros instead of open-coding

[RESEND PATCH] mm/sparsemem: use wrapped macros instead of open-coding

[qiwuchen55]

From: chenqiwu <chenqiwu@xiaomi.com>

Use wrapped macros instead of open-coding for better code
readability.

Signed-off-by: chenqiwu <chenqiwu@xiaomi.com>
---
 mm/sparse.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/mm/sparse.c b/mm/sparse.c
index 42c18a3..9b14164 100644
--- a/mm/sparse.c
+++ b/mm/sparse.c
@@ -385,8 +385,8 @@ static void __init check_usemap_section_nr(int nid,
 		old_pgdat_snr = NR_MEM_SECTIONS;
 	}
 
-	usemap_snr = pfn_to_section_nr(__pa(usage) >> PAGE_SHIFT);
-	pgdat_snr = pfn_to_section_nr(__pa(pgdat) >> PAGE_SHIFT);
+	usemap_snr = pfn_to_section_nr(virt_to_pfn(usage));
+	pgdat_snr = pfn_to_section_nr(virt_to_pfn(pgdat));
 	if (usemap_snr == pgdat_snr)
 		return;
 
@@ -677,7 +677,7 @@ struct page * __meminit populate_section_memmap(unsigned long pfn,
 
 	return NULL;
 got_map_page:
-	ret = (struct page *)pfn_to_kaddr(page_to_pfn(page));
+	ret = (struct page *)page_to_virt(page);
 got_map_ptr:
 
 	return ret;
-- 
1.9.1

Re: [RESEND PATCH] mm/sparsemem: use wrapped macros instead of open-coding

[John Hubbard]

On 3/6/20 2:19 AM, qiwuchen55@gmail.com wrote:
> From: chenqiwu <chenqiwu@xiaomi.com>
> 
> Use wrapped macros instead of open-coding for better code
> readability.
> 
> Signed-off-by: chenqiwu <chenqiwu@xiaomi.com>


This breaks my x86 (64-bit) build:

mm/sparse.c: In function ‘check_usemap_section_nr’:
mm/sparse.c:389:33: error: implicit declaration of function ‘virt_to_pfn’; did you mean ‘virt_to_phys’? [-Werror=implicit-function-declaration]
  389 |  usemap_snr = pfn_to_section_nr(virt_to_pfn(usage));
      |                                 ^~~~~~~~~~~
      |                                 virt_to_phys


...and I think the reason is that the arch/x86/include/asm/page.h does not
have virt_to_pfn(). For that reason, I'm concerned that the following fix, 
which fixes up x86, may not completely correct for the patch.

Here's a compile-tested (only, and only on x86 64-bit) fix:

diff --git a/arch/x86/include/asm/page.h b/arch/x86/include/asm/page.h
index 7555b48803a8..982bc76bf13c 100644
--- a/arch/x86/include/asm/page.h
+++ b/arch/x86/include/asm/page.h
@@ -62,6 +62,9 @@ static inline void copy_user_page(void *to, void *from, unsigned long vaddr,
 #define __boot_va(x)		__va(x)
 #define __boot_pa(x)		__pa(x)
 
+#define virt_to_pfn(kaddr)	(__pa(kaddr) >> PAGE_SHIFT)
+#define pfn_to_virt(pfn)	__va((pfn) << PAGE_SHIFT)
+
 /*
  * virt_to_page(kaddr) returns a valid pointer if and only if
  * virt_addr_valid(kaddr) returns true.


thanks,
-- 
John Hubbard
NVIDIA

> ---
>  mm/sparse.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/mm/sparse.c b/mm/sparse.c
> index 42c18a3..9b14164 100644
> --- a/mm/sparse.c
> +++ b/mm/sparse.c
> @@ -385,8 +385,8 @@ static void __init check_usemap_section_nr(int nid,
>  		old_pgdat_snr = NR_MEM_SECTIONS;
>  	}
>  
> -	usemap_snr = pfn_to_section_nr(__pa(usage) >> PAGE_SHIFT);
> -	pgdat_snr = pfn_to_section_nr(__pa(pgdat) >> PAGE_SHIFT);
> +	usemap_snr = pfn_to_section_nr(virt_to_pfn(usage));
> +	pgdat_snr = pfn_to_section_nr(virt_to_pfn(pgdat));
>  	if (usemap_snr == pgdat_snr)
>  		return;
>  
> @@ -677,7 +677,7 @@ struct page * __meminit populate_section_memmap(unsigned long pfn,
>  
>  	return NULL;
>  got_map_page:
> -	ret = (struct page *)pfn_to_kaddr(page_to_pfn(page));
> +	ret = (struct page *)page_to_virt(page);
>  got_map_ptr:
>  
>  	return ret;
> 

Re: [RESEND PATCH] mm/sparsemem: use wrapped macros instead of open-coding

[chenqiwu]

On Tue, Mar 10, 2020 at 07:10:10PM -0700, John Hubbard wrote:
> On 3/6/20 2:19 AM, qiwuchen55@gmail.com wrote:
> > From: chenqiwu <chenqiwu@xiaomi.com>
> > 
> > Use wrapped macros instead of open-coding for better code
> > readability.
> > 
> > Signed-off-by: chenqiwu <chenqiwu@xiaomi.com>
> 
> 
> This breaks my x86 (64-bit) build:
> 
> mm/sparse.c: In function ‘check_usemap_section_nr’:
> mm/sparse.c:389:33: error: implicit declaration of function ‘virt_to_pfn’; did you mean ‘virt_to_phys’? [-Werror=implicit-function-declaration]
>   389 |  usemap_snr = pfn_to_section_nr(virt_to_pfn(usage));
>       |                                 ^~~~~~~~~~~
>       |                                 virt_to_phys
> 
> 
> ...and I think the reason is that the arch/x86/include/asm/page.h does not
> have virt_to_pfn(). For that reason, I'm concerned that the following fix, 
> which fixes up x86, may not completely correct for the patch.
> 
> Here's a compile-tested (only, and only on x86 64-bit) fix:
> 
> diff --git a/arch/x86/include/asm/page.h b/arch/x86/include/asm/page.h
> index 7555b48803a8..982bc76bf13c 100644
> --- a/arch/x86/include/asm/page.h
> +++ b/arch/x86/include/asm/page.h
> @@ -62,6 +62,9 @@ static inline void copy_user_page(void *to, void *from, unsigned long vaddr,
>  #define __boot_va(x)		__va(x)
>  #define __boot_pa(x)		__pa(x)
>  
> +#define virt_to_pfn(kaddr)	(__pa(kaddr) >> PAGE_SHIFT)
> +#define pfn_to_virt(pfn)	__va((pfn) << PAGE_SHIFT)
> +
>  /*
>   * virt_to_page(kaddr) returns a valid pointer if and only if
>   * virt_addr_valid(kaddr) returns true.
> 
> 
> thanks,
> -- 
> John Hubbard
> NVIDIA

Hi John,
I tested this patch only on arm64 build, not sure if any other build
breakings. But I think any architecture is worth to support virt_to_pfn()
API. Could you please send your compile-tested patch to x86 upstream?

Thanks
Qiwu

Re: [RESEND PATCH] mm/sparsemem: use wrapped macros instead of open-coding

[John Hubbard]

On 3/11/20 1:01 AM, chenqiwu wrote:
...
> 
> Hi John,
> I tested this patch only on arm64 build, not sure if any other build
> breakings. But I think any architecture is worth to support virt_to_pfn()


If a patch causes a build break (on any architecture), it normally should not
be merged into any maintainer's git tree. That's why the patch submission
checklist, here:

     https://www.kernel.org/doc/html/latest/process/submit-checklist.html

...requests "Builds on multiple CPU architectures".

We don't all always do all of the steps there, but it is still a good baseline
checklist. And conversely, if you routinely skip key steps such as cross-compilation,
then reviewers have a lot more to worry about when looking at your patch submissions,
so that potentially slows down approvals.


> API. Could you please send your compile-tested patch to x86 upstream?
> 


I don't think that will help in this case, unless of course there is something
wrong with my build setup, and I'm seeing a problem that no one else sees.


thanks,
-- 
John Hubbard
NVIDIA