Skip to contentSkip to conversation

xen/gntdev: Fix off-by-one error when unmapping with holes

PATCH In review Ross Lagerwall 2 participants 7 messages Jan 9, 2018 – Jan 10, 2018 Xen hypervisor interface

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 xen/gntdev: Fix off-by-one error when unmapping with holes

If the requested range has a hole, the calculation of the number of pages to unmap is off by one. Fix it.

---
 drivers/xen/gntdev.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
drivers/xen/gntdev.c +1 -3
index 57efbd3..d3391a1 100644
@@ -380,10 +380,8 @@ static int unmap_grant_pages(struct grant_map *map, int offset, int pages)
 		}
 		range = 0;
 		while (range < pages) {
-			if (map->unmap_ops[offset+range].handle == -1) {
-				range--;
+			if (map->unmap_ops[offset+range].handle == -1)
 				break;
-			}
 			range++;
 		}
 		err = __unmap_grant_pages(map, offset, range);
Signature · 7 lines
2.9.5


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#1 lore raw

PATCH 2/2 xen/gntdev: Fix partial gntdev_mmap() cleanup

When cleaning up after a partially successful gntdev_mmap(), unmap the successfully mapped grant pages otherwise Xen will kill the domain if in debug mode (Attempt to implicitly unmap a granted PTE) or Linux will kill the process and emit "BUG: Bad page map in process" if Xen is in release mode.

This is only needed when use_ptemod is true because gntdev_put_map() will unmap grant pages itself when use_ptemod is false.

---
 drivers/xen/gntdev.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/xen/gntdev.c +3 -1
index d3391a1..bd56653 100644
@@ -1071,8 +1071,10 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
 out_unlock_put:
 	mutex_unlock(&priv->lock);
 out_put_map:
-	if (use_ptemod)
+	if (use_ptemod) {
 		map->vma = NULL;
+		unmap_grant_pages(map, 0, map->count);
+	}
 	gntdev_put_map(priv, map);
 	return err;
 }
Signature · 7 lines
2.9.5


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Reviewed-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] xen/gntdev: Fix partial gntdev_mmap() cleanup

Quoted 10 lines · Ross Lagerwall

When cleaning up after a partially successful gntdev_mmap(), unmap the successfully mapped grant pages otherwise Xen will kill the domain if in debug mode (Attempt to implicitly unmap a granted PTE) or Linux will kill the process and emit "BUG: Bad page map in process" if Xen is in release mode.

This is only needed when use_ptemod is true because gntdev_put_map() will unmap grant pages itself when use_ptemod is false.

although I wonder whether it may be possible to have gntdev_put_map() figure whether to unmap the pages if use_ptemod is set.

Quoted 11 lines
---
  drivers/xen/gntdev.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)
drivers/xen/gntdev.c +0 -0
index d3391a1..bd56653 100644
@@ -1071,8 +1071,10 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
  out_unlock_put:

drivers/xen/gntdev.c static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma) · line 1080

Quoted 10 lines
drivers/xen/gntdev.c +3 -1
  	mutex_unlock(&priv->lock);
  out_put_map:
-	if (use_ptemod)
+	if (use_ptemod) {
  		map->vma = NULL;
+		unmap_grant_pages(map, 0, map->count);
+	}
  	gntdev_put_map(priv, map);
  	return err;
  }

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#3 lore raw

Re: [PATCH 2/2] xen/gntdev: Fix partial gntdev_mmap() cleanup

Quoted 9 lines · Boris Ostrovsky
Quoted 6 lines · Ross Lagerwall

When cleaning up after a partially successful gntdev_mmap(), unmap the successfully mapped grant pages otherwise Xen will kill the domain if in debug mode (Attempt to implicitly unmap a granted PTE) or Linux will kill the process and emit "BUG: Bad page map in process" if Xen is in release mode.

Quoted 8 lines
Quoted 4 lines

This is only needed when use_ptemod is true because gntdev_put_map() will unmap grant pages itself when use_ptemod is false.

although I wonder whether it may be possible to have gntdev_put_map() figure whether to unmap the pages if use_ptemod is set.

It was a while since I wrote this patch, but IIRC when use_ptemod is set, successfully mmapped pages are unmapped via the mmu_notifier release callback. So doing it in gntdev_put_map() isn't possible without further changes.

Signature · 1 line
Ross Lagerwall
#4 lore raw

Re: [PATCH 2/2] xen/gntdev: Fix partial gntdev_mmap() cleanup

In reply to Boris Ostrovsky
Quoted 9 lines · Boris Ostrovsky
Quoted 6 lines · Ross Lagerwall

When cleaning up after a partially successful gntdev_mmap(), unmap the successfully mapped grant pages otherwise Xen will kill the domain if in debug mode (Attempt to implicitly unmap a granted PTE) or Linux will kill the process and emit "BUG: Bad page map in process" if Xen is in release mode.

Quoted 8 lines
Quoted 4 lines

This is only needed when use_ptemod is true because gntdev_put_map() will unmap grant pages itself when use_ptemod is false.

although I wonder whether it may be possible to have gntdev_put_map() figure whether to unmap the pages if use_ptemod is set.

It was a while since I wrote this patch, but IIRC when use_ptemod is set, successfully mmapped pages are unmapped via the mmu_notifier release callback. So doing it in gntdev_put_map() isn't possible without further changes.

Signature · 6 lines
Ross Lagerwall

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel