[v3,1/9] mm: Introduce new vm_insert_range API

Commit Message

Souptick Joarder Dec. 6, 2018, 6:39 p.m. UTC

Previouly drivers have their own way of mapping range of
kernel pages/memory into user vma and this was done by
invoking vm_insert_page() within a loop.

As this pattern is common across different drivers, it can
be generalized by creating a new function and use it across
the drivers.

vm_insert_range is the new API which will be used to map a
range of kernel memory/pages to user vma.

This API is tested by Heiko for Rockchip drm driver, on rk3188,
rk3288, rk3328 and rk3399 with graphics.

Signed-off-by: Souptick Joarder <[e-mail]>
Reviewed-by: Matthew Wilcox <[e-mail]>
Reviewed-by: Mike Rapoport <[e-mail]>
Tested-by: Heiko Stuebner <[e-mail]>
---
 include/linux/mm.h |  2 ++
 mm/memory.c        | 38 ++++++++++++++++++++++++++++++++++++++
 mm/nommu.c         |  7 +++++++
 3 files changed, 47 insertions(+)

Comments

Mauro Carvalho Chehab Dec. 7, 2018, 2:47 p.m. UTC | #1

Em Fri, 7 Dec 2018 00:09:45 +0530
Souptick Joarder <[e-mail]> escreveu:

> Previouly drivers have their own way of mapping range of
> kernel pages/memory into user vma and this was done by
> invoking vm_insert_page() within a loop.
> 
> As this pattern is common across different drivers, it can
> be generalized by creating a new function and use it across
> the drivers.
> 
> vm_insert_range is the new API which will be used to map a
> range of kernel memory/pages to user vma.
> 
> This API is tested by Heiko for Rockchip drm driver, on rk3188,
> rk3288, rk3328 and rk3399 with graphics.
> 
> Signed-off-by: Souptick Joarder <[e-mail]>
> Reviewed-by: Matthew Wilcox <[e-mail]>
> Reviewed-by: Mike Rapoport <[e-mail]>
> Tested-by: Heiko Stuebner <[e-mail]>

Looks good to me.

Reviewed-by: Mauro Carvalho Chehab <[e-mail]>

> ---
>  include/linux/mm.h |  2 ++
>  mm/memory.c        | 38 ++++++++++++++++++++++++++++++++++++++
>  mm/nommu.c         |  7 +++++++
>  3 files changed, 47 insertions(+)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index fcf9cc9..2bc399f 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2506,6 +2506,8 @@ unsigned long change_prot_numa(struct vm_area_struct *vma,
>  int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
>  			unsigned long pfn, unsigned long size, pgprot_t);
>  int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> +			struct page **pages, unsigned long page_count);
>  vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
>  			unsigned long pfn);
>  vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
> diff --git a/mm/memory.c b/mm/memory.c
> index 15c417e..84ea46c 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -1478,6 +1478,44 @@ static int insert_page(struct vm_area_struct *vma, unsigned long addr,
>  }
>  
>  /**
> + * vm_insert_range - insert range of kernel pages into user vma
> + * @vma: user vma to map to
> + * @addr: target user address of this page
> + * @pages: pointer to array of source kernel pages
> + * @page_count: number of pages need to insert into user vma
> + *
> + * This allows drivers to insert range of kernel pages they've allocated
> + * into a user vma. This is a generic function which drivers can use
> + * rather than using their own way of mapping range of kernel pages into
> + * user vma.
> + *
> + * If we fail to insert any page into the vma, the function will return
> + * immediately leaving any previously-inserted pages present.  Callers
> + * from the mmap handler may immediately return the error as their caller
> + * will destroy the vma, removing any successfully-inserted pages. Other
> + * callers should make their own arrangements for calling unmap_region().
> + *
> + * Context: Process context. Called by mmap handlers.
> + * Return: 0 on success and error code otherwise
> + */
> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> +			struct page **pages, unsigned long page_count)
> +{
> +	unsigned long uaddr = addr;
> +	int ret = 0, i;
> +
> +	for (i = 0; i < page_count; i++) {
> +		ret = vm_insert_page(vma, uaddr, pages[i]);
> +		if (ret < 0)
> +			return ret;
> +		uaddr += PAGE_SIZE;
> +	}
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL(vm_insert_range);
> +
> +/**
>   * vm_insert_page - insert single page into user vma
>   * @vma: user vma to map to
>   * @addr: target user address of this page
> diff --git a/mm/nommu.c b/mm/nommu.c
> index 749276b..d6ef5c7 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -473,6 +473,13 @@ int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
>  }
>  EXPORT_SYMBOL(vm_insert_page);
>  
> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> +			struct page **pages, unsigned long page_count)
> +{
> +	return -EINVAL;
> +}
> +EXPORT_SYMBOL(vm_insert_range);
> +
>  /*
>   *  sys_brk() for the most part doesn't need the global kernel
>   *  lock, except when an application is doing something nasty



Thanks,
Mauro

Robin Murphy Dec. 7, 2018, 3:34 p.m. UTC | #2

On 06/12/2018 18:39, Souptick Joarder wrote:
> Previouly drivers have their own way of mapping range of
> kernel pages/memory into user vma and this was done by
> invoking vm_insert_page() within a loop.
> 
> As this pattern is common across different drivers, it can
> be generalized by creating a new function and use it across
> the drivers.
> 
> vm_insert_range is the new API which will be used to map a
> range of kernel memory/pages to user vma.
> 
> This API is tested by Heiko for Rockchip drm driver, on rk3188,
> rk3288, rk3328 and rk3399 with graphics.
> 
> Signed-off-by: Souptick Joarder <[e-mail]>
> Reviewed-by: Matthew Wilcox <[e-mail]>
> Reviewed-by: Mike Rapoport <[e-mail]>
> Tested-by: Heiko Stuebner <[e-mail]>
> ---
>   include/linux/mm.h |  2 ++
>   mm/memory.c        | 38 ++++++++++++++++++++++++++++++++++++++
>   mm/nommu.c         |  7 +++++++
>   3 files changed, 47 insertions(+)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index fcf9cc9..2bc399f 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2506,6 +2506,8 @@ unsigned long change_prot_numa(struct vm_area_struct *vma,
>   int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
>   			unsigned long pfn, unsigned long size, pgprot_t);
>   int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> +			struct page **pages, unsigned long page_count);
>   vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
>   			unsigned long pfn);
>   vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
> diff --git a/mm/memory.c b/mm/memory.c
> index 15c417e..84ea46c 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -1478,6 +1478,44 @@ static int insert_page(struct vm_area_struct *vma, unsigned long addr,
>   }
>   
>   /**
> + * vm_insert_range - insert range of kernel pages into user vma
> + * @vma: user vma to map to
> + * @addr: target user address of this page
> + * @pages: pointer to array of source kernel pages
> + * @page_count: number of pages need to insert into user vma
> + *
> + * This allows drivers to insert range of kernel pages they've allocated
> + * into a user vma. This is a generic function which drivers can use
> + * rather than using their own way of mapping range of kernel pages into
> + * user vma.
> + *
> + * If we fail to insert any page into the vma, the function will return
> + * immediately leaving any previously-inserted pages present.  Callers
> + * from the mmap handler may immediately return the error as their caller
> + * will destroy the vma, removing any successfully-inserted pages. Other
> + * callers should make their own arrangements for calling unmap_region().
> + *
> + * Context: Process context. Called by mmap handlers.
> + * Return: 0 on success and error code otherwise
> + */
> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> +			struct page **pages, unsigned long page_count)
> +{
> +	unsigned long uaddr = addr;
> +	int ret = 0, i;

Some of the sites being replaced were effectively ensuring that vma and 
pages were mutually compatible as an initial condition - would it be 
worth adding something here for robustness, e.g.:

+	if (page_count != vma_pages(vma))
+		return -ENXIO;

?

(then you could also clean up a couple more places where you're not 
already removing such checks)

Robin.

> +
> +	for (i = 0; i < page_count; i++) {
> +		ret = vm_insert_page(vma, uaddr, pages[i]);
> +		if (ret < 0)
> +			return ret;
> +		uaddr += PAGE_SIZE;
> +	}
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL(vm_insert_range);
> +
> +/**
>    * vm_insert_page - insert single page into user vma
>    * @vma: user vma to map to
>    * @addr: target user address of this page
> diff --git a/mm/nommu.c b/mm/nommu.c
> index 749276b..d6ef5c7 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -473,6 +473,13 @@ int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
>   }
>   EXPORT_SYMBOL(vm_insert_page);
>   
> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> +			struct page **pages, unsigned long page_count)
> +{
> +	return -EINVAL;
> +}
> +EXPORT_SYMBOL(vm_insert_range);
> +
>   /*
>    *  sys_brk() for the most part doesn't need the global kernel
>    *  lock, except when an application is doing something nasty
>

Matthew Wilcox Dec. 7, 2018, 5:11 p.m. UTC | #3

On Fri, Dec 07, 2018 at 03:34:56PM +0000, Robin Murphy wrote:
> > +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> > +			struct page **pages, unsigned long page_count)
> > +{
> > +	unsigned long uaddr = addr;
> > +	int ret = 0, i;
> 
> Some of the sites being replaced were effectively ensuring that vma and
> pages were mutually compatible as an initial condition - would it be worth
> adding something here for robustness, e.g.:
> 
> +	if (page_count != vma_pages(vma))
> +		return -ENXIO;

I think we want to allow this to be used to populate part of a VMA.
So perhaps:

	if (page_count > vma_pages(vma))
		return -ENXIO;

Souptick Joarder Dec. 7, 2018, 7:28 p.m. UTC | #4

On Fri, Dec 7, 2018 at 10:41 PM Matthew Wilcox <[e-mail]> wrote:
>
> On Fri, Dec 07, 2018 at 03:34:56PM +0000, Robin Murphy wrote:
> > > +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> > > +                   struct page **pages, unsigned long page_count)
> > > +{
> > > +   unsigned long uaddr = addr;
> > > +   int ret = 0, i;
> >
> > Some of the sites being replaced were effectively ensuring that vma and
> > pages were mutually compatible as an initial condition - would it be worth
> > adding something here for robustness, e.g.:
> >
> > +     if (page_count != vma_pages(vma))
> > +             return -ENXIO;
>
> I think we want to allow this to be used to populate part of a VMA.
> So perhaps:
>
>         if (page_count > vma_pages(vma))
>                 return -ENXIO;

Ok, This can be added.

I think Patch [2/9] is the only leftover place where this
check could be removed.

Robin Murphy Dec. 7, 2018, 9:10 p.m. UTC | #5

On 2018-12-07 7:28 pm, Souptick Joarder wrote:
> On Fri, Dec 7, 2018 at 10:41 PM Matthew Wilcox <[e-mail]> wrote:
>>
>> On Fri, Dec 07, 2018 at 03:34:56PM +0000, Robin Murphy wrote:
>>>> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
>>>> +                   struct page **pages, unsigned long page_count)
>>>> +{
>>>> +   unsigned long uaddr = addr;
>>>> +   int ret = 0, i;
>>>
>>> Some of the sites being replaced were effectively ensuring that vma and
>>> pages were mutually compatible as an initial condition - would it be worth
>>> adding something here for robustness, e.g.:
>>>
>>> +     if (page_count != vma_pages(vma))
>>> +             return -ENXIO;
>>
>> I think we want to allow this to be used to populate part of a VMA.
>> So perhaps:
>>
>>          if (page_count > vma_pages(vma))
>>                  return -ENXIO;
> 
> Ok, This can be added.
> 
> I think Patch [2/9] is the only leftover place where this
> check could be removed.

Right, 9/9 could also have relied on my stricter check here, but since 
it's really testing whether it actually managed to allocate vma_pages() 
worth of pages earlier, Matthew's more lenient version won't help for 
that one. (Why privcmd_buf_mmap() doesn't clean up and return an error 
as soon as that allocation loop fails, without taking the mutex under 
which it still does a bunch more pointless work to only undo it again, 
is a mind-boggling mystery, but that's not our problem here...)

Robin.

Souptick Joarder Dec. 7, 2018, 9:48 p.m. UTC | #6

On Sat, Dec 8, 2018 at 2:40 AM Robin Murphy <[e-mail]> wrote:
>
> On 2018-12-07 7:28 pm, Souptick Joarder wrote:
> > On Fri, Dec 7, 2018 at 10:41 PM Matthew Wilcox <[e-mail]> wrote:
> >>
> >> On Fri, Dec 07, 2018 at 03:34:56PM +0000, Robin Murphy wrote:
> >>>> +int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
> >>>> +                   struct page **pages, unsigned long page_count)
> >>>> +{
> >>>> +   unsigned long uaddr = addr;
> >>>> +   int ret = 0, i;
> >>>
> >>> Some of the sites being replaced were effectively ensuring that vma and
> >>> pages were mutually compatible as an initial condition - would it be worth
> >>> adding something here for robustness, e.g.:
> >>>
> >>> +     if (page_count != vma_pages(vma))
> >>> +             return -ENXIO;
> >>
> >> I think we want to allow this to be used to populate part of a VMA.
> >> So perhaps:
> >>
> >>          if (page_count > vma_pages(vma))
> >>                  return -ENXIO;
> >
> > Ok, This can be added.
> >
> > I think Patch [2/9] is the only leftover place where this
> > check could be removed.
>
> Right, 9/9 could also have relied on my stricter check here, but since
> it's really testing whether it actually managed to allocate vma_pages()
> worth of pages earlier, Matthew's more lenient version won't help for
> that one.


(Why privcmd_buf_mmap() doesn't clean up and return an error
> as soon as that allocation loop fails, without taking the mutex under
> which it still does a bunch more pointless work to only undo it again,
> is a mind-boggling mystery, but that's not our problem here...)

I think some clean up can be done here in a separate patch.

Patch

diff --git a/include/linux/mm.h b/include/linux/mm.h
index fcf9cc9..2bc399f 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2506,6 +2506,8 @@  unsigned long change_prot_numa(struct vm_area_struct *vma,
 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
 			unsigned long pfn, unsigned long size, pgprot_t);
 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
+int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
+			struct page **pages, unsigned long page_count);
 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
 			unsigned long pfn);
 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
diff --git a/mm/memory.c b/mm/memory.c
index 15c417e..84ea46c 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -1478,6 +1478,44 @@  static int insert_page(struct vm_area_struct *vma, unsigned long addr,
 }
 
 /**
+ * vm_insert_range - insert range of kernel pages into user vma
+ * @vma: user vma to map to
+ * @addr: target user address of this page
+ * @pages: pointer to array of source kernel pages
+ * @page_count: number of pages need to insert into user vma
+ *
+ * This allows drivers to insert range of kernel pages they've allocated
+ * into a user vma. This is a generic function which drivers can use
+ * rather than using their own way of mapping range of kernel pages into
+ * user vma.
+ *
+ * If we fail to insert any page into the vma, the function will return
+ * immediately leaving any previously-inserted pages present.  Callers
+ * from the mmap handler may immediately return the error as their caller
+ * will destroy the vma, removing any successfully-inserted pages. Other
+ * callers should make their own arrangements for calling unmap_region().
+ *
+ * Context: Process context. Called by mmap handlers.
+ * Return: 0 on success and error code otherwise
+ */
+int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
+			struct page **pages, unsigned long page_count)
+{
+	unsigned long uaddr = addr;
+	int ret = 0, i;
+
+	for (i = 0; i < page_count; i++) {
+		ret = vm_insert_page(vma, uaddr, pages[i]);
+		if (ret < 0)
+			return ret;
+		uaddr += PAGE_SIZE;
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(vm_insert_range);
+
+/**
  * vm_insert_page - insert single page into user vma
  * @vma: user vma to map to
  * @addr: target user address of this page
diff --git a/mm/nommu.c b/mm/nommu.c
index 749276b..d6ef5c7 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -473,6 +473,13 @@  int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
 }
 EXPORT_SYMBOL(vm_insert_page);
 
+int vm_insert_range(struct vm_area_struct *vma, unsigned long addr,
+			struct page **pages, unsigned long page_count)
+{
+	return -EINVAL;
+}
+EXPORT_SYMBOL(vm_insert_range);
+
 /*
  *  sys_brk() for the most part doesn't need the global kernel
  *  lock, except when an application is doing something nasty