Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1331051 > unrolled thread

Re: [PATCHv4 2/3] arm64: Add support for ARCH_SUPPORTS_DEBUG_PAGEALLOC

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCHv4 2/3] arm64: Add support for  ARCH_SUPPORTS_DEBUG_PAGEALLOC Catalin Marinas <catalin.marinas@arm.com> - 2016-02-10 11:40 +0100
    Re: [PATCHv4 2/3] arm64: Add support for ARCH_SUPPORTS_DEBUG_PAGEALLOC Ard Biesheuvel <ard.biesheuvel@linaro.org> - 2016-02-10 11:50 +0100
    Re: [PATCHv4 2/3] arm64: Add support for  ARCH_SUPPORTS_DEBUG_PAGEALLOC Mark Rutland <mark.rutland@arm.com> - 2016-02-10 12:00 +0100
    Re: [PATCHv4 2/3] arm64: Add support for  ARCH_SUPPORTS_DEBUG_PAGEALLOC Catalin Marinas <catalin.marinas@arm.com> - 2016-02-10 12:10 +0100

#1331051 — Re: [PATCHv4 2/3] arm64: Add support for ARCH_SUPPORTS_DEBUG_PAGEALLOC

On Fri, Feb 05, 2016 at 04:24:47PM -0800, Laura Abbott wrote:
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -149,6 +149,19 @@ static void split_pud(pud_t *old_pud, pmd_t *pmd)
>  	} while (pmd++, i++, i < PTRS_PER_PMD);
>  }
>  
> +bool block_mappings_allowed(phys_addr_t (*pgtable_alloc)(void))
> +{
> +
> +	/*
> +	 * If debug_page_alloc is enabled we must map the linear map
> +	 * using pages. However, other mappings created by
> +	 * create_mapping_noalloc must use sections in some cases. Allow
> +	 * sections to be used in those cases, where no pgtable_alloc
> +	 * function is provided.
> +	 */
> +	return !pgtable_alloc || !debug_pagealloc_enabled();
> +}

This breaks the build when CONFIG_DEBUG_PAGEALLOC is not enabled
(defconfig) since debug_pagealloc_enabled() is not defined. A fix went
in next/master as commit 0987684b855c
("mm-slab-clean-up-debug_pagealloc-processing-code-fix"). I need to
track it down and merge it via the arm64 tree, otherwise I'll add some
#ifdefs in this function.

BTW, shouldn't the block_mappings_allowed() function be static?

-- 
Catalin

[toc] | [next] | [standalone]

#1331062 — Re: [PATCHv4 2/3] arm64: Add support for ARCH_SUPPORTS_DEBUG_PAGEALLOC

On 10 February 2016 at 11:38, Catalin Marinas <catalin.marinas@arm.com> wrote:
> On Fri, Feb 05, 2016 at 04:24:47PM -0800, Laura Abbott wrote:
>> --- a/arch/arm64/mm/mmu.c
>> +++ b/arch/arm64/mm/mmu.c
>> @@ -149,6 +149,19 @@ static void split_pud(pud_t *old_pud, pmd_t *pmd)
>>       } while (pmd++, i++, i < PTRS_PER_PMD);
>>  }
>>
>> +bool block_mappings_allowed(phys_addr_t (*pgtable_alloc)(void))
>> +{
>> +
>> +     /*
>> +      * If debug_page_alloc is enabled we must map the linear map
>> +      * using pages. However, other mappings created by
>> +      * create_mapping_noalloc must use sections in some cases. Allow
>> +      * sections to be used in those cases, where no pgtable_alloc
>> +      * function is provided.
>> +      */
>> +     return !pgtable_alloc || !debug_pagealloc_enabled();
>> +}
>
> This breaks the build when CONFIG_DEBUG_PAGEALLOC is not enabled
> (defconfig) since debug_pagealloc_enabled() is not defined. A fix went
> in next/master as commit 0987684b855c
> ("mm-slab-clean-up-debug_pagealloc-processing-code-fix"). I need to
> track it down and merge it via the arm64 tree, otherwise I'll add some
> #ifdefs in this function.
>
> BTW, shouldn't the block_mappings_allowed() function be static?
>

Yes, it should.

[toc] | [prev] | [next] | [standalone]

#1331064

On Wed, Feb 10, 2016 at 10:38:14AM +0000, Catalin Marinas wrote:
> On Fri, Feb 05, 2016 at 04:24:47PM -0800, Laura Abbott wrote:
> > --- a/arch/arm64/mm/mmu.c
> > +++ b/arch/arm64/mm/mmu.c
> > @@ -149,6 +149,19 @@ static void split_pud(pud_t *old_pud, pmd_t *pmd)
> >  	} while (pmd++, i++, i < PTRS_PER_PMD);
> >  }
> >  
> > +bool block_mappings_allowed(phys_addr_t (*pgtable_alloc)(void))
> > +{
> > +
> > +	/*
> > +	 * If debug_page_alloc is enabled we must map the linear map
> > +	 * using pages. However, other mappings created by
> > +	 * create_mapping_noalloc must use sections in some cases. Allow
> > +	 * sections to be used in those cases, where no pgtable_alloc
> > +	 * function is provided.
> > +	 */
> > +	return !pgtable_alloc || !debug_pagealloc_enabled();
> > +}
> 
> This breaks the build when CONFIG_DEBUG_PAGEALLOC is not enabled
> (defconfig) since debug_pagealloc_enabled() is not defined. A fix went
> in next/master as commit 0987684b855c
> ("mm-slab-clean-up-debug_pagealloc-processing-code-fix"). I need to
> track it down and merge it via the arm64 tree, otherwise I'll add some
> #ifdefs in this function.

In the cover for v2 [1],  it was mentioend that we needed a patch [2]
"mm: provide debug_pagealloc_enabled() without CONFIG_DEBUG_PAGEALLOC".
I assumed that was already queued via the mm tree.

> BTW, shouldn't the block_mappings_allowed() function be static?

That would make sense to me, yes.

Mark.

[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2016-January/403439.html  
[2] http://article.gmane.org/gmane.linux.kernel.mm/1452

[toc] | [prev] | [next] | [standalone]

#1331070

On Wed, Feb 10, 2016 at 10:38:14AM +0000, Catalin Marinas wrote:
> On Fri, Feb 05, 2016 at 04:24:47PM -0800, Laura Abbott wrote:
> > --- a/arch/arm64/mm/mmu.c
> > +++ b/arch/arm64/mm/mmu.c
> > @@ -149,6 +149,19 @@ static void split_pud(pud_t *old_pud, pmd_t *pmd)
> >  	} while (pmd++, i++, i < PTRS_PER_PMD);
> >  }
> >  
> > +bool block_mappings_allowed(phys_addr_t (*pgtable_alloc)(void))
> > +{
> > +
> > +	/*
> > +	 * If debug_page_alloc is enabled we must map the linear map
> > +	 * using pages. However, other mappings created by
> > +	 * create_mapping_noalloc must use sections in some cases. Allow
> > +	 * sections to be used in those cases, where no pgtable_alloc
> > +	 * function is provided.
> > +	 */
> > +	return !pgtable_alloc || !debug_pagealloc_enabled();
> > +}
> 
> This breaks the build when CONFIG_DEBUG_PAGEALLOC is not enabled
> (defconfig) since debug_pagealloc_enabled() is not defined. A fix went
> in next/master as commit 0987684b855c
> ("mm-slab-clean-up-debug_pagealloc-processing-code-fix"). I need to
> track it down and merge it via the arm64 tree, otherwise I'll add some
> #ifdefs in this function.

It seems that there are other patches in the mm tree requiring this
change. I'll fix it locally (#ifdefs) and revert the change after -rc1.

-- 
Catalin

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web