diff mbox series

[v11,01/13] mm/vmalloc: fix HUGE_VMAP regression by enabling huge pages in vmalloc_to_page

Message ID 20210126044510.2491820-2-npiggin@gmail.com (mailing list archive)
State New
Headers show
Series huge vmalloc mappings | expand

Commit Message

Nicholas Piggin Jan. 26, 2021, 4:44 a.m. UTC
vmalloc_to_page returns NULL for addresses mapped by larger pages[*].
Whether or not a vmap is huge depends on the architecture details,
alignments, boot options, etc., which the caller can not be expected
to know. Therefore HUGE_VMAP is a regression for vmalloc_to_page.

This change teaches vmalloc_to_page about larger pages, and returns
the struct page that corresponds to the offset within the large page.
This makes the API agnostic to mapping implementation details.

[*] As explained by commit 029c54b095995 ("mm/vmalloc.c: huge-vmap:
    fail gracefully on unexpected huge vmap mappings")

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 mm/vmalloc.c | 41 ++++++++++++++++++++++++++---------------
 1 file changed, 26 insertions(+), 15 deletions(-)

Comments

Miaohe Lin Jan. 26, 2021, 6:40 a.m. UTC | #1
Hi:
On 2021/1/26 12:44, Nicholas Piggin wrote:
> vmalloc_to_page returns NULL for addresses mapped by larger pages[*].
> Whether or not a vmap is huge depends on the architecture details,
> alignments, boot options, etc., which the caller can not be expected
> to know. Therefore HUGE_VMAP is a regression for vmalloc_to_page.
> 
> This change teaches vmalloc_to_page about larger pages, and returns
> the struct page that corresponds to the offset within the large page.
> This makes the API agnostic to mapping implementation details.
> 
> [*] As explained by commit 029c54b095995 ("mm/vmalloc.c: huge-vmap:
>     fail gracefully on unexpected huge vmap mappings")
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
>  mm/vmalloc.c | 41 ++++++++++++++++++++++++++---------------
>  1 file changed, 26 insertions(+), 15 deletions(-)
> 
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index e6f352bf0498..62372f9e0167 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -34,7 +34,7 @@
>  #include <linux/bitops.h>
>  #include <linux/rbtree_augmented.h>
>  #include <linux/overflow.h>
> -
> +#include <linux/pgtable.h>
>  #include <linux/uaccess.h>
>  #include <asm/tlbflush.h>
>  #include <asm/shmparam.h>
> @@ -343,7 +343,9 @@ int is_vmalloc_or_module_addr(const void *x)
>  }
>  
>  /*
> - * Walk a vmap address to the struct page it maps.
> + * Walk a vmap address to the struct page it maps. Huge vmap mappings will
> + * return the tail page that corresponds to the base page address, which
> + * matches small vmap mappings.
>   */
>  struct page *vmalloc_to_page(const void *vmalloc_addr)
>  {
> @@ -363,25 +365,33 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>  
>  	if (pgd_none(*pgd))
>  		return NULL;
> +	if (WARN_ON_ONCE(pgd_leaf(*pgd)))
> +		return NULL; /* XXX: no allowance for huge pgd */
> +	if (WARN_ON_ONCE(pgd_bad(*pgd)))
> +		return NULL;
> +
>  	p4d = p4d_offset(pgd, addr);
>  	if (p4d_none(*p4d))
>  		return NULL;
> -	pud = pud_offset(p4d, addr);
> +	if (p4d_leaf(*p4d))
> +		return p4d_page(*p4d) + ((addr & ~P4D_MASK) >> PAGE_SHIFT);
> +	if (WARN_ON_ONCE(p4d_bad(*p4d)))
> +		return NULL;
>  
> -	/*
> -	 * Don't dereference bad PUD or PMD (below) entries. This will also
> -	 * identify huge mappings, which we may encounter on architectures
> -	 * that define CONFIG_HAVE_ARCH_HUGE_VMAP=y. Such regions will be
> -	 * identified as vmalloc addresses by is_vmalloc_addr(), but are
> -	 * not [unambiguously] associated with a struct page, so there is
> -	 * no correct value to return for them.
> -	 */
> -	WARN_ON_ONCE(pud_bad(*pud));
> -	if (pud_none(*pud) || pud_bad(*pud))
> +	pud = pud_offset(p4d, addr);
> +	if (pud_none(*pud))
> +		return NULL;
> +	if (pud_leaf(*pud))
> +		return pud_page(*pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
> +	if (WARN_ON_ONCE(pud_bad(*pud)))
>  		return NULL;
> +
>  	pmd = pmd_offset(pud, addr);
> -	WARN_ON_ONCE(pmd_bad(*pmd));
> -	if (pmd_none(*pmd) || pmd_bad(*pmd))
> +	if (pmd_none(*pmd))
> +		return NULL;
> +	if (pmd_leaf(*pmd))
> +		return pmd_page(*pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
> +	if (WARN_ON_ONCE(pmd_bad(*pmd)))
>  		return NULL;
>  
>  	ptep = pte_offset_map(pmd, addr);
> @@ -389,6 +399,7 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>  	if (pte_present(pte))
>  		page = pte_page(pte);
>  	pte_unmap(ptep);
> +
>  	return page;
>  }
>  EXPORT_SYMBOL(vmalloc_to_page);
> 

LGTM. Thanks.

Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Ding Tianhong Jan. 28, 2021, 3:13 a.m. UTC | #2
On 2021/1/26 12:44, Nicholas Piggin wrote:
> vmalloc_to_page returns NULL for addresses mapped by larger pages[*].
> Whether or not a vmap is huge depends on the architecture details,
> alignments, boot options, etc., which the caller can not be expected
> to know. Therefore HUGE_VMAP is a regression for vmalloc_to_page.
> 
> This change teaches vmalloc_to_page about larger pages, and returns
> the struct page that corresponds to the offset within the large page.
> This makes the API agnostic to mapping implementation details.
> 
> [*] As explained by commit 029c54b095995 ("mm/vmalloc.c: huge-vmap:
>     fail gracefully on unexpected huge vmap mappings")
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
>  mm/vmalloc.c | 41 ++++++++++++++++++++++++++---------------
>  1 file changed, 26 insertions(+), 15 deletions(-)
> 
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index e6f352bf0498..62372f9e0167 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -34,7 +34,7 @@
>  #include <linux/bitops.h>
>  #include <linux/rbtree_augmented.h>
>  #include <linux/overflow.h>
> -
> +#include <linux/pgtable.h>
>  #include <linux/uaccess.h>
>  #include <asm/tlbflush.h>
>  #include <asm/shmparam.h>
> @@ -343,7 +343,9 @@ int is_vmalloc_or_module_addr(const void *x)
>  }
>  
>  /*
> - * Walk a vmap address to the struct page it maps.
> + * Walk a vmap address to the struct page it maps. Huge vmap mappings will
> + * return the tail page that corresponds to the base page address, which
> + * matches small vmap mappings.
>   */
>  struct page *vmalloc_to_page(const void *vmalloc_addr)
>  {
> @@ -363,25 +365,33 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>  
>  	if (pgd_none(*pgd))
>  		return NULL;
> +	if (WARN_ON_ONCE(pgd_leaf(*pgd)))
> +		return NULL; /* XXX: no allowance for huge pgd */
> +	if (WARN_ON_ONCE(pgd_bad(*pgd)))
> +		return NULL;
> +
>  	p4d = p4d_offset(pgd, addr);
>  	if (p4d_none(*p4d))
>  		return NULL;
> -	pud = pud_offset(p4d, addr);
> +	if (p4d_leaf(*p4d))
> +		return p4d_page(*p4d) + ((addr & ~P4D_MASK) >> PAGE_SHIFT);
> +	if (WARN_ON_ONCE(p4d_bad(*p4d)))
> +		return NULL;
>  
> -	/*
> -	 * Don't dereference bad PUD or PMD (below) entries. This will also
> -	 * identify huge mappings, which we may encounter on architectures
> -	 * that define CONFIG_HAVE_ARCH_HUGE_VMAP=y. Such regions will be
> -	 * identified as vmalloc addresses by is_vmalloc_addr(), but are
> -	 * not [unambiguously] associated with a struct page, so there is
> -	 * no correct value to return for them.
> -	 */
> -	WARN_ON_ONCE(pud_bad(*pud));
> -	if (pud_none(*pud) || pud_bad(*pud))
> +	pud = pud_offset(p4d, addr);
> +	if (pud_none(*pud))
> +		return NULL;
> +	if (pud_leaf(*pud))
> +		return pud_page(*pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);

Hi Nicho:

/builds/1mzfdQzleCy69KZFb5qHNSEgabZ/mm/vmalloc.c: In function 'vmalloc_to_page':
/builds/1mzfdQzleCy69KZFb5qHNSEgabZ/include/asm-generic/pgtable-nop4d-hack.h:48:27: error: implicit declaration of function 'pud_page'; did you mean 'put_page'? [-Werror=implicit-function-declaration]
   48 | #define pgd_page(pgd)    (pud_page((pud_t){ pgd }))
      |                           ^~~~~~~~

the pug_page is not defined for aarch32 when enabling 2-level page config, it break the system building.


> +	if (WARN_ON_ONCE(pud_bad(*pud)))
>  		return NULL;
> +
>  	pmd = pmd_offset(pud, addr);
> -	WARN_ON_ONCE(pmd_bad(*pmd));
> -	if (pmd_none(*pmd) || pmd_bad(*pmd))
> +	if (pmd_none(*pmd))
> +		return NULL;
> +	if (pmd_leaf(*pmd))
> +		return pmd_page(*pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
> +	if (WARN_ON_ONCE(pmd_bad(*pmd)))
>  		return NULL;
>  
>  	ptep = pte_offset_map(pmd, addr);
> @@ -389,6 +399,7 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>  	if (pte_present(pte))
>  		page = pte_page(pte);
>  	pte_unmap(ptep);
> +
>  	return page;
>  }
>  EXPORT_SYMBOL(vmalloc_to_page);
>
Nicholas Piggin Feb. 2, 2021, 10:22 a.m. UTC | #3
Excerpts from Ding Tianhong's message of January 28, 2021 1:13 pm:
> On 2021/1/26 12:44, Nicholas Piggin wrote:
>> vmalloc_to_page returns NULL for addresses mapped by larger pages[*].
>> Whether or not a vmap is huge depends on the architecture details,
>> alignments, boot options, etc., which the caller can not be expected
>> to know. Therefore HUGE_VMAP is a regression for vmalloc_to_page.
>> 
>> This change teaches vmalloc_to_page about larger pages, and returns
>> the struct page that corresponds to the offset within the large page.
>> This makes the API agnostic to mapping implementation details.
>> 
>> [*] As explained by commit 029c54b095995 ("mm/vmalloc.c: huge-vmap:
>>     fail gracefully on unexpected huge vmap mappings")
>> 
>> Reviewed-by: Christoph Hellwig <hch@lst.de>
>> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
>> ---
>>  mm/vmalloc.c | 41 ++++++++++++++++++++++++++---------------
>>  1 file changed, 26 insertions(+), 15 deletions(-)
>> 
>> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
>> index e6f352bf0498..62372f9e0167 100644
>> --- a/mm/vmalloc.c
>> +++ b/mm/vmalloc.c
>> @@ -34,7 +34,7 @@
>>  #include <linux/bitops.h>
>>  #include <linux/rbtree_augmented.h>
>>  #include <linux/overflow.h>
>> -
>> +#include <linux/pgtable.h>
>>  #include <linux/uaccess.h>
>>  #include <asm/tlbflush.h>
>>  #include <asm/shmparam.h>
>> @@ -343,7 +343,9 @@ int is_vmalloc_or_module_addr(const void *x)
>>  }
>>  
>>  /*
>> - * Walk a vmap address to the struct page it maps.
>> + * Walk a vmap address to the struct page it maps. Huge vmap mappings will
>> + * return the tail page that corresponds to the base page address, which
>> + * matches small vmap mappings.
>>   */
>>  struct page *vmalloc_to_page(const void *vmalloc_addr)
>>  {
>> @@ -363,25 +365,33 @@ struct page *vmalloc_to_page(const void *vmalloc_addr)
>>  
>>  	if (pgd_none(*pgd))
>>  		return NULL;
>> +	if (WARN_ON_ONCE(pgd_leaf(*pgd)))
>> +		return NULL; /* XXX: no allowance for huge pgd */
>> +	if (WARN_ON_ONCE(pgd_bad(*pgd)))
>> +		return NULL;
>> +
>>  	p4d = p4d_offset(pgd, addr);
>>  	if (p4d_none(*p4d))
>>  		return NULL;
>> -	pud = pud_offset(p4d, addr);
>> +	if (p4d_leaf(*p4d))
>> +		return p4d_page(*p4d) + ((addr & ~P4D_MASK) >> PAGE_SHIFT);
>> +	if (WARN_ON_ONCE(p4d_bad(*p4d)))
>> +		return NULL;
>>  
>> -	/*
>> -	 * Don't dereference bad PUD or PMD (below) entries. This will also
>> -	 * identify huge mappings, which we may encounter on architectures
>> -	 * that define CONFIG_HAVE_ARCH_HUGE_VMAP=y. Such regions will be
>> -	 * identified as vmalloc addresses by is_vmalloc_addr(), but are
>> -	 * not [unambiguously] associated with a struct page, so there is
>> -	 * no correct value to return for them.
>> -	 */
>> -	WARN_ON_ONCE(pud_bad(*pud));
>> -	if (pud_none(*pud) || pud_bad(*pud))
>> +	pud = pud_offset(p4d, addr);
>> +	if (pud_none(*pud))
>> +		return NULL;
>> +	if (pud_leaf(*pud))
>> +		return pud_page(*pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
> 
> Hi Nicho:
> 
> /builds/1mzfdQzleCy69KZFb5qHNSEgabZ/mm/vmalloc.c: In function 'vmalloc_to_page':
> /builds/1mzfdQzleCy69KZFb5qHNSEgabZ/include/asm-generic/pgtable-nop4d-hack.h:48:27: error: implicit declaration of function 'pud_page'; did you mean 'put_page'? [-Werror=implicit-function-declaration]
>    48 | #define pgd_page(pgd)    (pud_page((pud_t){ pgd }))
>       |                           ^~~~~~~~
> 
> the pug_page is not defined for aarch32 when enabling 2-level page config, it break the system building.

Hey thanks for finding that, not sure why that didn't trigger any CI.

Anyway newer kernels don't have the ptable-*-hack.h headers, but even so 
it still breaks upstream. arm is using some hand-rolled 2-level folding
of its own (which is fair enough because most 32-bit archs were 2 level
at the time I added pgtable-nopud.h header).

This patch seems to at least make it build.

Thanks,
Nick

---
 arch/arm/include/asm/pgtable-3level.h | 2 --
 arch/arm/include/asm/pgtable.h        | 3 +++
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/arm/include/asm/pgtable-3level.h b/arch/arm/include/asm/pgtable-3level.h
index 2b85d175e999..d4edab51a77c 100644
--- a/arch/arm/include/asm/pgtable-3level.h
+++ b/arch/arm/include/asm/pgtable-3level.h
@@ -186,8 +186,6 @@ static inline pte_t pte_mkspecial(pte_t pte)
 
 #define pmd_write(pmd)		(pmd_isclear((pmd), L_PMD_SECT_RDONLY))
 #define pmd_dirty(pmd)		(pmd_isset((pmd), L_PMD_SECT_DIRTY))
-#define pud_page(pud)		pmd_page(__pmd(pud_val(pud)))
-#define pud_write(pud)		pmd_write(__pmd(pud_val(pud)))
 
 #define pmd_hugewillfault(pmd)	(!pmd_young(pmd) || !pmd_write(pmd))
 #define pmd_thp_or_huge(pmd)	(pmd_huge(pmd) || pmd_trans_huge(pmd))
diff --git a/arch/arm/include/asm/pgtable.h b/arch/arm/include/asm/pgtable.h
index c02f24400369..d63a5bb6bd0c 100644
--- a/arch/arm/include/asm/pgtable.h
+++ b/arch/arm/include/asm/pgtable.h
@@ -166,6 +166,9 @@ extern struct page *empty_zero_page;
 
 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
+#define pud_page(pud)		pmd_page(__pmd(pud_val(pud)))
+#define pud_write(pud)		pmd_write(__pmd(pud_val(pud)))
+
 #define pmd_none(pmd)		(!pmd_val(pmd))
 
 static inline pte_t *pmd_page_vaddr(pmd_t pmd)
diff mbox series

Patch

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index e6f352bf0498..62372f9e0167 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -34,7 +34,7 @@ 
 #include <linux/bitops.h>
 #include <linux/rbtree_augmented.h>
 #include <linux/overflow.h>
-
+#include <linux/pgtable.h>
 #include <linux/uaccess.h>
 #include <asm/tlbflush.h>
 #include <asm/shmparam.h>
@@ -343,7 +343,9 @@  int is_vmalloc_or_module_addr(const void *x)
 }
 
 /*
- * Walk a vmap address to the struct page it maps.
+ * Walk a vmap address to the struct page it maps. Huge vmap mappings will
+ * return the tail page that corresponds to the base page address, which
+ * matches small vmap mappings.
  */
 struct page *vmalloc_to_page(const void *vmalloc_addr)
 {
@@ -363,25 +365,33 @@  struct page *vmalloc_to_page(const void *vmalloc_addr)
 
 	if (pgd_none(*pgd))
 		return NULL;
+	if (WARN_ON_ONCE(pgd_leaf(*pgd)))
+		return NULL; /* XXX: no allowance for huge pgd */
+	if (WARN_ON_ONCE(pgd_bad(*pgd)))
+		return NULL;
+
 	p4d = p4d_offset(pgd, addr);
 	if (p4d_none(*p4d))
 		return NULL;
-	pud = pud_offset(p4d, addr);
+	if (p4d_leaf(*p4d))
+		return p4d_page(*p4d) + ((addr & ~P4D_MASK) >> PAGE_SHIFT);
+	if (WARN_ON_ONCE(p4d_bad(*p4d)))
+		return NULL;
 
-	/*
-	 * Don't dereference bad PUD or PMD (below) entries. This will also
-	 * identify huge mappings, which we may encounter on architectures
-	 * that define CONFIG_HAVE_ARCH_HUGE_VMAP=y. Such regions will be
-	 * identified as vmalloc addresses by is_vmalloc_addr(), but are
-	 * not [unambiguously] associated with a struct page, so there is
-	 * no correct value to return for them.
-	 */
-	WARN_ON_ONCE(pud_bad(*pud));
-	if (pud_none(*pud) || pud_bad(*pud))
+	pud = pud_offset(p4d, addr);
+	if (pud_none(*pud))
+		return NULL;
+	if (pud_leaf(*pud))
+		return pud_page(*pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
+	if (WARN_ON_ONCE(pud_bad(*pud)))
 		return NULL;
+
 	pmd = pmd_offset(pud, addr);
-	WARN_ON_ONCE(pmd_bad(*pmd));
-	if (pmd_none(*pmd) || pmd_bad(*pmd))
+	if (pmd_none(*pmd))
+		return NULL;
+	if (pmd_leaf(*pmd))
+		return pmd_page(*pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
+	if (WARN_ON_ONCE(pmd_bad(*pmd)))
 		return NULL;
 
 	ptep = pte_offset_map(pmd, addr);
@@ -389,6 +399,7 @@  struct page *vmalloc_to_page(const void *vmalloc_addr)
 	if (pte_present(pte))
 		page = pte_page(pte);
 	pte_unmap(ptep);
+
 	return page;
 }
 EXPORT_SYMBOL(vmalloc_to_page);