[PATCH][v2] uio: Support 36-bit physical addresses on 32-bit systems

From: Kai Jiang <redacted>

To support >32-bit physical addresses for UIO_MEM_PHYS type we need to
extend the width of 'addr' in struct uio_mem.  Numerous platforms like
embedded PPC, ARM, and X86 have support for systems with larger physical
address than logical.

Since 'addr' may contain a physical, logical, or virtual address the
easiest solution is to just change the type to 'unsigned long long'
regardless of which type is utilized.

For physical address we can support up to a 44-bit physical address on a
typical 32-bit system as we utilize remap_pfn_range() for the mapping of
the memory region and pfn's are represnted by shifting the address by
the page size (typically 4k).

Signed-off-by: Kai Jiang <redacted>
Signed-off-by: Minghuan Lian <redacted>
Signed-off-by: Kumar Gala <redacted>
---
v2:
* Use phys_addr_t instead of 'unsigned long long'
* Updated DocBook detail in uio-howto.tmpl

 Documentation/DocBook/uio-howto.tmpl |    2 +-
 drivers/uio/uio.c                    |    8 ++++----
 include/linux/uio_driver.h           |    2 +-
 3 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/Documentation/DocBook/uio-howto.tmpl b/Documentation/DocBook/uio-howto.tmpl
index 7c4b514d..54883de 100644
--- a/Documentation/DocBook/uio-howto.tmpl
+++ b/Documentation/DocBook/uio-howto.tmpl

@@ -529,7 +529,7 @@ memory (e.g. allocated with <function>kmalloc()</function>). There's also
 </para></listitem>
 
 <listitem><para>
-<varname>unsigned long addr</varname>: Required if the mapping is used.
+<varname>phys_addr_t addr</varname>: Required if the mapping is used.
 Fill in the address of your memory block. This address is the one that
 appears in sysfs.
 </para></listitem>

diff --git a/drivers/uio/uio.c b/drivers/uio/uio.c
index 88f4444..43b7096 100644
--- a/drivers/uio/uio.c
+++ b/drivers/uio/uio.c

@@ -69,7 +69,7 @@ static ssize_t map_name_show(struct uio_mem *mem, char *buf)
 
 static ssize_t map_addr_show(struct uio_mem *mem, char *buf)
 {
-	return sprintf(buf, "0x%lx\n", mem->addr);
+	return sprintf(buf, "0x%llx\n", (unsigned long long)mem->addr);
 }
 
 static ssize_t map_size_show(struct uio_mem *mem, char *buf)

@@ -79,7 +79,7 @@ static ssize_t map_size_show(struct uio_mem *mem, char *buf)
 
 static ssize_t map_offset_show(struct uio_mem *mem, char *buf)
 {
-	return sprintf(buf, "0x%lx\n", mem->addr & ~PAGE_MASK);
+	return sprintf(buf, "0x%llx\n", (unsigned long long)mem->addr & ~PAGE_MASK);
 }
 
 struct map_sysfs_entry {

@@ -634,8 +634,8 @@ static int uio_vma_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	if (idev->info->mem[mi].memtype == UIO_MEM_LOGICAL)
 		page = virt_to_page(idev->info->mem[mi].addr + offset);
 	else
-		page = vmalloc_to_page((void *)idev->info->mem[mi].addr
-							+ offset);
+		page = vmalloc_to_page((void *)(unsigned long)
+				idev->info->mem[mi].addr + offset);
 	get_page(page);
 	vmf->page = page;
 	return 0;

diff --git a/include/linux/uio_driver.h b/include/linux/uio_driver.h
index 4c618cd..d9ce796 100644
--- a/include/linux/uio_driver.h
+++ b/include/linux/uio_driver.h

@@ -32,7 +32,7 @@ struct uio_map;
  */
 struct uio_mem {
 	const char		*name;
-	unsigned long		addr;
+	phys_addr_t		addr;
 	unsigned long		size;
 	int			memtype;
 	void __iomem		*internal_addr;

-- 
1.7.3.4

On Wed, Oct 12, 2011 at 3:57 PM, Kumar Gala [off-list ref] wro=
te:

From: Kai Jiang <redacted>

To support >32-bit physical addresses for UIO_MEM_PHYS type we need to
extend the width of 'addr' in struct uio_mem. =A0Numerous platforms like
embedded PPC, ARM, and X86 have support for systems with larger physical
address than logical.

Since 'addr' may contain a physical, logical, or virtual address the
easiest solution is to just change the type to 'unsigned long long'
regardless of which type is utilized.

You forgot to update this description.

=A0struct uio_mem {
=A0 =A0 =A0 =A0const char =A0 =A0 =A0 =A0 =A0 =A0 =A0*name;
- =A0 =A0 =A0 unsigned long =A0 =A0 =A0 =A0 =A0 addr;
+ =A0 =A0 =A0 phys_addr_t =A0 =A0 =A0 =A0 =A0 =A0 addr;

Please add a comment here saying:

1) That 'addr' can be a virtual or physical address
2) That the kernel guarantees that sizeof(phys_addr_t) >=3D sizeof(void
*), so it's safe to use phys_addr_t for a virtual pointer.

--=20
Timur Tabi
Linux kernel developer at Freescale=

On Oct 13, 2011, at 9:37 AM, Tabi Timur-B04825 wrote:

On Wed, Oct 12, 2011 at 3:57 PM, Kumar Gala =

[off-list ref] wrote:

quoted

From: Kai Jiang <redacted>
=20
To support >32-bit physical addresses for UIO_MEM_PHYS type we need =

to

quoted

extend the width of 'addr' in struct uio_mem.  Numerous platforms =

like

quoted

embedded PPC, ARM, and X86 have support for systems with larger =

physical

quoted

address than logical.
=20
Since 'addr' may contain a physical, logical, or virtual address the
easiest solution is to just change the type to 'unsigned long long'
regardless of which type is utilized.

=20
You forgot to update this description.

will fix and update commit message

=20

quoted

 struct uio_mem {
       const char              *name;
-       unsigned long           addr;
+       phys_addr_t             addr;

=20
Please add a comment here saying:
=20
1) That 'addr' can be a virtual or physical address

The code and everything else makes that clear

2) That the kernel guarantees that sizeof(phys_addr_t) >=3D =

sizeof(void

*), so it's safe to use phys_addr_t for a virtual pointer.

The commit message will cover that so I don't plan on add it.

- k=

Kumar Gala wrote:

quoted

quoted

quoted

quoted

+       phys_addr_t             addr;

Please add a comment here saying:

1) That 'addr' can be a virtual or physical address

The code and everything else makes that clear

I'm sorry, but I have to strongly disagree here.  It is *NOT* clear that a
variable of type 'phys_addr_t' can hold something that is not a physical address.

-- 
Timur Tabi
Linux kernel developer at Freescale

David Laight wrote:

Since there is a discriminating field, could a union be used?
At a guess the type of the address is constrained between
produces and consumer??

I don't think we need to complicate the code by changing that variable into a
union.  I just a want a short comment added to the structure definition.  I
don't think that's asking a lot.

-- 
Timur Tabi
Linux kernel developer at Freescale

=20

Kumar Gala wrote:

quoted

quoted

quoted

quoted

quoted

+       phys_addr_t             addr;

=20
Please add a comment here saying:
=20
1) That 'addr' can be a virtual or physical address

The code and everything else makes that clear

=20
I'm sorry, but I have to strongly disagree here.  It is *NOT*=20
clear that a variable of type 'phys_addr_t' can hold something
that is not a physical address.

Since there is a discriminating field, could a union be used?
At a guess the type of the address is constrained between
produces and consumer??

	David

On Oct 13, 2011, at 10:37 AM, David Laight wrote:

quoted

Kumar Gala wrote:

quoted

quoted

quoted

quoted

quoted

+       phys_addr_t             addr;

Please add a comment here saying:

1) That 'addr' can be a virtual or physical address

The code and everything else makes that clear

I'm sorry, but I have to strongly disagree here.  It is *NOT* 
clear that a variable of type 'phys_addr_t' can hold something
that is not a physical address.

Since there is a discriminating field, could a union be used?
At a guess the type of the address is constrained between
produces and consumer??

Uugh.

I'll add a comment to uio_mem.

- k