[PATCH] book3s64/radix : Align section vmemmap start address to PAGE_SIZE

A vmemmap altmap is a device-provided region used to provide
backing storage for struct pages. For each namespace, the altmap
should belong to that same namespace. If the namespaces are
created unaligned, there is a chance that the section vmemmap
start address could also be unaligned. If the section vmemmap
start address is unaligned, the altmap page allocated from the
current namespace might be used by the previous namespace also.
During the free operation, since the altmap is shared between two
namespaces, the previous namespace may detect that the page does
not belong to its altmap and incorrectly assume that the page is a
normal page. It then attempts to free the normal page, which leads
to a kernel crash.

In this patch, we are aligning the section vmemmap start address
to PAGE_SIZE. After alignment, the start address will not be
part of the current namespace, and a normal page will be allocated
for the vmemmap mapping of the current section. For the remaining
sections, altmaps will be allocated. During the free operation,
the normal page will be correctly freed.

Without this patch
==================
NS1 start               NS2 start
 _________________________________________________________
|         NS1               |            NS2              |
 ---------------------------------------------------------
| Altmap| Altmap | .....|Altmap| Altmap | ...........
|  NS1  |  NS1   |      | NS2  |  NS2   |

In the above scenario, NS1 and NS2 are two namespaces. The vmemmap
for NS1 comes from Altmap NS1, which belongs to NS1, and the
vmemmap for NS2 comes from Altmap NS2, which belongs to NS2.

The vmemmap start for NS2 is not aligned, so Altmap NS2 is shared
by both NS1 and NS2. During the free operation in NS1, Altmap NS2
is not part of NS1's altmap, causing it to attempt to free an
invalid page.

With this patch
===============
NS1 start               NS2 start
 _________________________________________________________
|         NS1               |            NS2              |
 ---------------------------------------------------------
| Altmap| Altmap | .....| Normal | Altmap | Altmap |.......
|  NS1  |  NS1   |      |  Page  |  NS2   |  NS2   |

If the vmemmap start for NS2 is not aligned then we are allocating
a normal page. NS1 and NS2 vmemmap will be freed correctly.

Fixes: 368a0590d954("powerpc/book3s64/vmemmap: switch radix to use a different vmemmap handling function")
Co-developed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Donet Tom <redacted>
---
 arch/powerpc/mm/book3s64/radix_pgtable.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/powerpc/mm/book3s64/radix_pgtable.c b/arch/powerpc/mm/book3s64/radix_pgtable.c
index 311e2112d782..b22d5f6147d2 100644
--- a/arch/powerpc/mm/book3s64/radix_pgtable.c
+++ b/arch/powerpc/mm/book3s64/radix_pgtable.c

@@ -1120,6 +1120,8 @@ int __meminit radix__vmemmap_populate(unsigned long start, unsigned long end, in
 	pmd_t *pmd;
 	pte_t *pte;
 
+	start = ALIGN_DOWN(start, PAGE_SIZE);
+
 	for (addr = start; addr < end; addr = next) {
 		next = pmd_addr_end(addr, end);
 

-- 
2.43.5

Donet Tom [off-list ref] writes:

A vmemmap altmap is a device-provided region used to provide
backing storage for struct pages. For each namespace, the altmap
should belong to that same namespace. If the namespaces are
created unaligned, there is a chance that the section vmemmap
start address could also be unaligned. If the section vmemmap
start address is unaligned, the altmap page allocated from the
current namespace might be used by the previous namespace also.
During the free operation, since the altmap is shared between two
namespaces, the previous namespace may detect that the page does
not belong to its altmap and incorrectly assume that the page is a
normal page. It then attempts to free the normal page, which leads
to a kernel crash.

In this patch, we are aligning the section vmemmap start address
to PAGE_SIZE. After alignment, the start address will not be
part of the current namespace, and a normal page will be allocated
for the vmemmap mapping of the current section. For the remaining
sections, altmaps will be allocated. During the free operation,
the normal page will be correctly freed.

Without this patch
==================
NS1 start               NS2 start
 _________________________________________________________
|         NS1               |            NS2              |
 ---------------------------------------------------------
| Altmap| Altmap | .....|Altmap| Altmap | ...........
|  NS1  |  NS1   |      | NS2  |  NS2   |

                            ^^^ this should be allocated in ram?

On 3/3/25 18:32, Aneesh Kumar K.V wrote:

Donet Tom [off-list ref] writes:

quoted

A vmemmap altmap is a device-provided region used to provide
backing storage for struct pages. For each namespace, the altmap
should belong to that same namespace. If the namespaces are
created unaligned, there is a chance that the section vmemmap
start address could also be unaligned. If the section vmemmap
start address is unaligned, the altmap page allocated from the
current namespace might be used by the previous namespace also.
During the free operation, since the altmap is shared between two
namespaces, the previous namespace may detect that the page does
not belong to its altmap and incorrectly assume that the page is a
normal page. It then attempts to free the normal page, which leads
to a kernel crash.

In this patch, we are aligning the section vmemmap start address
to PAGE_SIZE. After alignment, the start address will not be
part of the current namespace, and a normal page will be allocated
for the vmemmap mapping of the current section. For the remaining
sections, altmaps will be allocated. During the free operation,
the normal page will be correctly freed.

Without this patch
==================
NS1 start               NS2 start
  _________________________________________________________
|         NS1               |            NS2              |
  ---------------------------------------------------------
| Altmap| Altmap | .....|Altmap| Altmap | ...........
|  NS1  |  NS1   |      | NS2  |  NS2   |

                             ^^^ this should be allocated in ram?

Yes, it should be allocated from RAM. However, in the current
implementation, an altmap page gets allocated. This is because the
NS2 vmemmap section's start address is unaligned. There is an
altmap_cross_boundary() check. Here, from the vmemmap section
start, we identify the namespace start and check if the namespace start
is within the boundary. Since it is within the boundary, it returns false,
causing an altmap page to be allocated. During the PTE update, the
vmemmap start address is aligned down to PAGE_SIZE, and the PTE is
updated. As a result, the altmap page is shared between the current
and previous namespaces.

If we had aligned the vmemmap start address, the
altmap_cross_boundary() function would return true because the
vmemmap section's start address belongs to the previous
namespace. Therefore normal page gets allocated. During the
PTE set operation, since the address is already aligned, the
PTE will updated.

quoted

In the above scenario, NS1 and NS2 are two namespaces. The vmemmap
for NS1 comes from Altmap NS1, which belongs to NS1, and the
vmemmap for NS2 comes from Altmap NS2, which belongs to NS2.

The vmemmap start for NS2 is not aligned, so Altmap NS2 is shared
by both NS1 and NS2. During the free operation in NS1, Altmap NS2
is not part of NS1's altmap, causing it to attempt to free an
invalid page.

With this patch
===============
NS1 start               NS2 start
  _________________________________________________________
|         NS1               |            NS2              |
  ---------------------------------------------------------
| Altmap| Altmap | .....| Normal | Altmap | Altmap |.......
|  NS1  |  NS1   |      |  Page  |  NS2   |  NS2   |

If the vmemmap start for NS2 is not aligned then we are allocating
a normal page. NS1 and NS2 vmemmap will be freed correctly.

Fixes: 368a0590d954("powerpc/book3s64/vmemmap: switch radix to use a different vmemmap handling function")
Co-developed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Donet Tom <redacted>
---
  arch/powerpc/mm/book3s64/radix_pgtable.c | 2 ++
  1 file changed, 2 insertions(+)

diff --git a/arch/powerpc/mm/book3s64/radix_pgtable.c b/arch/powerpc/mm/book3s64/radix_pgtable.c
index 311e2112d782..b22d5f6147d2 100644
--- a/arch/powerpc/mm/book3s64/radix_pgtable.c
+++ b/arch/powerpc/mm/book3s64/radix_pgtable.c

@@ -1120,6 +1120,8 @@ int __meminit radix__vmemmap_populate(unsigned long start, unsigned long end, in
  	pmd_t *pmd;
  	pte_t *pte;
  
+	start = ALIGN_DOWN(start, PAGE_SIZE);
+
  	for (addr = start; addr < end; addr = next) {
  		next = pmd_addr_end(addr, end);
  

-- 
2.43.5

Donet Tom [off-list ref] writes:

On 3/3/25 18:32, Aneesh Kumar K.V wrote:

quoted

Donet Tom [off-list ref] writes:

quoted

A vmemmap altmap is a device-provided region used to provide
backing storage for struct pages. For each namespace, the altmap
should belong to that same namespace. If the namespaces are
created unaligned, there is a chance that the section vmemmap
start address could also be unaligned. If the section vmemmap
start address is unaligned, the altmap page allocated from the
current namespace might be used by the previous namespace also.
During the free operation, since the altmap is shared between two
namespaces, the previous namespace may detect that the page does
not belong to its altmap and incorrectly assume that the page is a
normal page. It then attempts to free the normal page, which leads
to a kernel crash.

In this patch, we are aligning the section vmemmap start address
to PAGE_SIZE. After alignment, the start address will not be
part of the current namespace, and a normal page will be allocated
for the vmemmap mapping of the current section. For the remaining
sections, altmaps will be allocated. During the free operation,
the normal page will be correctly freed.

Without this patch
==================
NS1 start               NS2 start
  _________________________________________________________
|         NS1               |            NS2              |
  ---------------------------------------------------------
| Altmap| Altmap | .....|Altmap| Altmap | ...........
|  NS1  |  NS1   |      | NS2  |  NS2   |

                             ^^^ this should be allocated in ram?

Yes, it should be allocated from RAM. However, in the current
implementation, an altmap page gets allocated. This is because the
NS2 vmemmap section's start address is unaligned. There is an
altmap_cross_boundary() check. Here, from the vmemmap section
start, we identify the namespace start and check if the namespace start
is within the boundary. Since it is within the boundary, it returns false,
causing an altmap page to be allocated. During the PTE update, the
vmemmap start address is aligned down to PAGE_SIZE, and the PTE is
updated. As a result, the altmap page is shared between the current
and previous namespaces.

If we had aligned the vmemmap start address, the
altmap_cross_boundary() function would return true because the
vmemmap section's start address belongs to the previous
namespace. Therefore normal page gets allocated. During the
PTE set operation, since the address is already aligned, the
PTE will updated.

So the nvdimm driver should ensure that alignment right? I assume other things
will also require that to be properly aligned.?

-aneesh

On 3/6/25 9:41 AM, Aneesh Kumar K.V wrote:

Donet Tom [off-list ref] writes:

quoted

On 3/3/25 18:32, Aneesh Kumar K.V wrote:

quoted

Donet Tom [off-list ref] writes:

quoted

A vmemmap altmap is a device-provided region used to provide
backing storage for struct pages. For each namespace, the altmap
should belong to that same namespace. If the namespaces are
created unaligned, there is a chance that the section vmemmap
start address could also be unaligned. If the section vmemmap
start address is unaligned, the altmap page allocated from the
current namespace might be used by the previous namespace also.
During the free operation, since the altmap is shared between two
namespaces, the previous namespace may detect that the page does
not belong to its altmap and incorrectly assume that the page is a
normal page. It then attempts to free the normal page, which leads
to a kernel crash.

In this patch, we are aligning the section vmemmap start address
to PAGE_SIZE. After alignment, the start address will not be
part of the current namespace, and a normal page will be allocated
for the vmemmap mapping of the current section. For the remaining
sections, altmaps will be allocated. During the free operation,
the normal page will be correctly freed.

Without this patch
==================
NS1 start               NS2 start
   _________________________________________________________
|         NS1               |            NS2              |
   ---------------------------------------------------------
| Altmap| Altmap | .....|Altmap| Altmap | ...........
|  NS1  |  NS1   |      | NS2  |  NS2   |

                              ^^^ this should be allocated in ram?

Yes, it should be allocated from RAM. However, in the current
implementation, an altmap page gets allocated. This is because the
NS2 vmemmap section's start address is unaligned. There is an
altmap_cross_boundary() check. Here, from the vmemmap section
start, we identify the namespace start and check if the namespace start
is within the boundary. Since it is within the boundary, it returns false,
causing an altmap page to be allocated. During the PTE update, the
vmemmap start address is aligned down to PAGE_SIZE, and the PTE is
updated. As a result, the altmap page is shared between the current
and previous namespaces.

If we had aligned the vmemmap start address, the
altmap_cross_boundary() function would return true because the
vmemmap section's start address belongs to the previous
namespace. Therefore normal page gets allocated. During the
PTE set operation, since the address is already aligned, the
PTE will updated.

So the nvdimm driver should ensure that alignment right? I assume other things
will also require that to be properly aligned.?

  #cat /proc/iomem
00000000-63ffffffff : System RAM
40340000000-403401fffff : namespace1.0
40340200000-403a0ffffff : dax1.0
403a1000000-403a11fffff : namespace1.1
403a1200000-40401ffffff : dax1.1
40402000000-404021fffff : namespace1.2
40402200000-40462ffffff : dax1.2
40463000000-404631fffff : namespace1.3
40463200000-404c3ffffff : dax1.3
  #

I have created 4 namespaces with a size of 1552M. As you can see, the 
start of
namespace1.0 is 1G aligned, while namespace1.1, namespace1.2, and 
namespace1.3
are not 1G aligned. If I had created the namespace with a size of 1536M 
(1.5G), then
all the namespaces would have started 1G aligned.

I believe that based on the size we are requesting, the namespaces 
alignments are
being created. They do not always need to be 1G aligned.

Now, if we calculate the vmemmap start for namespace1.1..

Phy start - 0x403a1000000
pfn start - 0x403a1000000 / PAGE_SIZE = 0x403a100

vmemmap start = 0xc00c000000000000 + (0x403a100 * 0x40)
                           =0xC00C000100E84000

This address is not page aligned. This will trigger this issue.

-aneesh

Donet Tom [off-list ref] writes:

       /*
        * Make sure we align the start vmemmap addr so that we calculate
       the correct start_pfn in altmap boundary check to decided whether
       we should use altmap or RAM based backing memory allocation. Also
       the address need to be aligned for set_pte operation.

       If the start addr is already PMD_SIZE aligned we will try to use
       a pmd mapping. We don't want to be too aggressive here beacause
       that will cause more allocations in RAM. So only if the namespace
       vmemmap start addr is PMD_SIZE aligned we will use PMD mapping.

       */

May be with some comments as above?

+	start = ALIGN_DOWN(start, PAGE_SIZE);
+
 	for (addr = start; addr < end; addr = next) {
 		next = pmd_addr_end(addr, end);
 
-- 
2.43.5

On 3/8/25 9:16 AM, Aneesh Kumar K.V wrote:

Donet Tom [off-list ref] writes:

quoted

A vmemmap altmap is a device-provided region used to provide
backing storage for struct pages. For each namespace, the altmap
should belong to that same namespace. If the namespaces are
created unaligned, there is a chance that the section vmemmap
start address could also be unaligned. If the section vmemmap
start address is unaligned, the altmap page allocated from the
current namespace might be used by the previous namespace also.
During the free operation, since the altmap is shared between two
namespaces, the previous namespace may detect that the page does
not belong to its altmap and incorrectly assume that the page is a
normal page. It then attempts to free the normal page, which leads
to a kernel crash.

In this patch, we are aligning the section vmemmap start address
to PAGE_SIZE. After alignment, the start address will not be
part of the current namespace, and a normal page will be allocated
for the vmemmap mapping of the current section. For the remaining
sections, altmaps will be allocated. During the free operation,
the normal page will be correctly freed.

Without this patch
==================
NS1 start               NS2 start
  _________________________________________________________
|         NS1               |            NS2              |
  ---------------------------------------------------------
| Altmap| Altmap | .....|Altmap| Altmap | ...........
|  NS1  |  NS1   |      | NS2  |  NS2   |

In the above scenario, NS1 and NS2 are two namespaces. The vmemmap
for NS1 comes from Altmap NS1, which belongs to NS1, and the
vmemmap for NS2 comes from Altmap NS2, which belongs to NS2.

The vmemmap start for NS2 is not aligned, so Altmap NS2 is shared
by both NS1 and NS2. During the free operation in NS1, Altmap NS2
is not part of NS1's altmap, causing it to attempt to free an
invalid page.

With this patch
===============
NS1 start               NS2 start
  _________________________________________________________
|         NS1               |            NS2              |
  ---------------------------------------------------------
| Altmap| Altmap | .....| Normal | Altmap | Altmap |.......
|  NS1  |  NS1   |      |  Page  |  NS2   |  NS2   |

If the vmemmap start for NS2 is not aligned then we are allocating
a normal page. NS1 and NS2 vmemmap will be freed correctly.

Fixes: 368a0590d954("powerpc/book3s64/vmemmap: switch radix to use a different vmemmap handling function")
Co-developed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Donet Tom <redacted>
---
  arch/powerpc/mm/book3s64/radix_pgtable.c | 2 ++
  1 file changed, 2 insertions(+)

diff --git a/arch/powerpc/mm/book3s64/radix_pgtable.c b/arch/powerpc/mm/book3s64/radix_pgtable.c
index 311e2112d782..b22d5f6147d2 100644
--- a/arch/powerpc/mm/book3s64/radix_pgtable.c
+++ b/arch/powerpc/mm/book3s64/radix_pgtable.c

@@ -1120,6 +1120,8 @@ int __meminit radix__vmemmap_populate(unsigned long start, unsigned long end, in
  	pmd_t *pmd;
  	pte_t *pte;

        /*
         * Make sure we align the start vmemmap addr so that we calculate
        the correct start_pfn in altmap boundary check to decided whether
        we should use altmap or RAM based backing memory allocation. Also
        the address need to be aligned for set_pte operation.

        If the start addr is already PMD_SIZE aligned we will try to use
        a pmd mapping. We don't want to be too aggressive here beacause
        that will cause more allocations in RAM. So only if the namespace
        vmemmap start addr is PMD_SIZE aligned we will use PMD mapping.

        */

May be with some comments as above?

Sure, I will update it and send a V3.

quoted

+	start = ALIGN_DOWN(start, PAGE_SIZE);
+
  	for (addr = start; addr < end; addr = next) {
  		next = pmd_addr_end(addr, end);
  
-- 
2.43.5