[PATCH 1/2] mm: Fix memory size alignment in devm_memremap_pages_release()

[PATCH 1/2] mm: Fix memory size alignment in devm_memremap_pages_release()

[Jan H. Schönherr]

The functions devm_memremap_pages() and devm_memremap_pages_release() use
different ways to calculate the section-aligned amount of memory. The
latter function may use an incorrect size if the memory region is small
but straddles a section border.

Use the same code for both.

Fixes: 5f29a77cd957 ("mm: fix mixed zone detection in devm_memremap_pages")
Signed-off-by: Jan H. Schönherr <jschoenh@amazon.de>
---
 kernel/memremap.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/memremap.c b/kernel/memremap.c
index 403ab9c..4712ce6 100644
--- a/kernel/memremap.c
+++ b/kernel/memremap.c
@@ -301,7 +301,8 @@ static void devm_memremap_pages_release(struct device *dev, void *data)
 
 	/* pages are dead and unused, undo the arch mapping */
 	align_start = res->start & ~(SECTION_SIZE - 1);
-	align_size = ALIGN(resource_size(res), SECTION_SIZE);
+	align_size = ALIGN(res->start + resource_size(res), SECTION_SIZE)
+		- align_start;
 
 	mem_hotplug_begin();
 	arch_remove_memory(align_start, align_size);
-- 
2.9.3.1.gcba166c.dirty

[PATCH 2/2] mm: Fix devm_memremap_pages() collision handling

[Jan H. Schönherr]

If devm_memremap_pages() detects a collision while adding entries
to the radix-tree, we call pgmap_radix_release(). Unfortunately,
the function removes *all* entries for the range -- including the
entries that caused the collision in the first place.

Modify pgmap_radix_release() to take an additional argument to
indicate where to stop, so that only newly added entries are removed
from the tree.

Fixes: 9476df7d80df ("mm: introduce find_dev_pagemap()")
Signed-off-by: Jan H. Schönherr <jschoenh@amazon.de>
---
 kernel/memremap.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/kernel/memremap.c b/kernel/memremap.c
index 4712ce6..2b136d4 100644
--- a/kernel/memremap.c
+++ b/kernel/memremap.c
@@ -248,13 +248,16 @@ int device_private_entry_fault(struct vm_area_struct *vma,
 EXPORT_SYMBOL(device_private_entry_fault);
 #endif /* CONFIG_DEVICE_PRIVATE */
 
-static void pgmap_radix_release(struct resource *res)
+static void pgmap_radix_release(struct resource *res, unsigned long end_pgoff)
 {
 	unsigned long pgoff, order;
 
 	mutex_lock(&pgmap_lock);
-	foreach_order_pgoff(res, order, pgoff)
+	foreach_order_pgoff(res, order, pgoff) {
+		if (pgoff >= end_pgoff)
+			break;
 		radix_tree_delete(&pgmap_radix, PHYS_PFN(res->start) + pgoff);
+	}
 	mutex_unlock(&pgmap_lock);
 
 	synchronize_rcu();
@@ -309,7 +312,7 @@ static void devm_memremap_pages_release(struct device *dev, void *data)
 	mem_hotplug_done();
 
 	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
-	pgmap_radix_release(res);
+	pgmap_radix_release(res, -1);
 	dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc,
 			"%s: failed to free all reserved pages\n", __func__);
 }
@@ -459,7 +462,7 @@ void *devm_memremap_pages(struct device *dev, struct resource *res,
 	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
  err_pfn_remap:
  err_radix:
-	pgmap_radix_release(res);
+	pgmap_radix_release(res, pgoff);
 	devres_free(page_map);
 	return ERR_PTR(error);
 }
-- 
2.9.3.1.gcba166c.dirty

Re: [PATCH 1/2] mm: Fix memory size alignment in devm_memremap_pages_release()

[Dan Williams]

On Wed, Jan 17, 2018 at 4:06 PM, Jan H. Schönherr <jschoenh@amazon.de> wrote:
> The functions devm_memremap_pages() and devm_memremap_pages_release() use
> different ways to calculate the section-aligned amount of memory. The
> latter function may use an incorrect size if the memory region is small
> but straddles a section border.
>
> Use the same code for both.
>
> Fixes: 5f29a77cd957 ("mm: fix mixed zone detection in devm_memremap_pages")
> Signed-off-by: Jan H. Schönherr <jschoenh@amazon.de>

Looks good to me, applied for 4.16.

Re: [PATCH 2/2] mm: Fix devm_memremap_pages() collision handling

[Dan Williams]

On Wed, Jan 17, 2018 at 4:06 PM, Jan H. Schönherr <jschoenh@amazon.de> wrote:
> If devm_memremap_pages() detects a collision while adding entries
> to the radix-tree, we call pgmap_radix_release(). Unfortunately,
> the function removes *all* entries for the range -- including the
> entries that caused the collision in the first place.
>
> Modify pgmap_radix_release() to take an additional argument to
> indicate where to stop, so that only newly added entries are removed
> from the tree.
>
> Fixes: 9476df7d80df ("mm: introduce find_dev_pagemap()")
> Signed-off-by: Jan H. Schönherr <jschoenh@amazon.de>

Looks good to me, applied for 4.16.