Skip to contentSkip to conversation

binder: fix race between munmap() and direct reclaim

PATCH In review Todd Kjos 4 participants 4 messages Mar 1, 2019 – Mar 5, 2019 Android drivers

lore
Share & tools
lore

Todd Kjos

author maintainer
#0 lore raw

PATCH 1/1 binder: fix race between munmap() and direct reclaim

An munmap() on a binder device causes binder_vma_close() to be called which clears the alloc->vma pointer.

If direct reclaim causes binder_alloc_free_page() to be called, there is a race where alloc->vma is read into a local vma pointer and then used later after the mm->mmap_sem is acquired. This can result in calling zap_page_range() with an invalid vma which manifests as a use-after-free in zap_page_range().

The fix is to check alloc->vma after acquiring the mmap_sem (which we were acquiring anyway) and skip zap_page_range() if it has changed to NULL.

---
 drivers/android/binder_alloc.c | 18 ++++++++----------
 1 file changed, 8 insertions(+), 10 deletions(-)
drivers/android/binder_alloc.c +8 -10
index 6389467670a0b..195f120c4e8c9 100644
@@ -927,14 +927,13 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
 
 	index = page - alloc->pages;
 	page_addr = (uintptr_t)alloc->buffer + index * PAGE_SIZE;
+
+	mm = alloc->vma_vm_mm;
+	if (!mmget_not_zero(mm))
+		goto err_mmget;
+	if (!down_write_trylock(&mm->mmap_sem))
+		goto err_down_write_mmap_sem_failed;
 	vma = binder_alloc_get_vma(alloc);
-	if (vma) {
-		if (!mmget_not_zero(alloc->vma_vm_mm))
-			goto err_mmget;
-		mm = alloc->vma_vm_mm;
-		if (!down_read_trylock(&mm->mmap_sem))
-			goto err_down_write_mmap_sem_failed;
-	}
 
 	list_lru_isolate(lru, item);
 	spin_unlock(lock);
@@ -945,10 +944,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
 		zap_page_range(vma, page_addr, PAGE_SIZE);
 
 		trace_binder_unmap_user_end(alloc, index);
-
-		up_read(&mm->mmap_sem);
-		mmput(mm);
 	}
+	up_write(&mm->mmap_sem);
+	mmput(mm);
 
 	trace_binder_unmap_kernel_start(alloc, index);
 
Signature · 1 line
2.21.0.352.gf09ad66450-goog
maintainer
#1 lore raw
Quoted 9 lines · Todd Kjos

An munmap() on a binder device causes binder_vma_close() to be called which clears the alloc->vma pointer.

If direct reclaim causes binder_alloc_free_page() to be called, there is a race where alloc->vma is read into a local vma pointer and then used later after the mm->mmap_sem is acquired. This can result in calling zap_page_range() with an invalid vma which manifests as a use-after-free in zap_page_range().

Quoted 6 lines

The fix is to check alloc->vma after acquiring the mmap_sem (which we were acquiring anyway) and skip zap_page_range() if it has changed to NULL.

---

Any specific commit that this fixes? And should it be marked for stable releases?

thanks,

greg k-h

#2 lore raw
Quoted 11 lines · Greg KH
Quoted 9 lines · Todd Kjos

An munmap() on a binder device causes binder_vma_close() to be called which clears the alloc->vma pointer.

If direct reclaim causes binder_alloc_free_page() to be called, there is a race where alloc->vma is read into a local vma pointer and then used later after the mm->mmap_sem is acquired. This can result in calling zap_page_range() with an invalid vma which manifests as a use-after-free in zap_page_range().

Quoted 8 lines
Quoted 6 lines

The fix is to check alloc->vma after acquiring the mmap_sem (which we were acquiring anyway) and skip zap_page_range() if it has changed to NULL.

---

Any specific commit that this fixes?

No, it's been there a long time.

Quoted 1 line

And should it be marked for stable releases?

It is needed in stable (back to 4.4), but will need to be backported. Should I post backported versions targeting the specific releases now? I was thinking we'd wait for this one to land. I think we'll need 1 patch for 4.4/4.9 and a second one for 4.14/4.19 (and some of those backported patches will have conflicts when merged down to android-4.X -- I think the 4.14/4.19 version will apply to all the android branches). Let me know how you want to handle this.

Quoted 4 lines

thanks,

greg k-h

Reviewed-by
#3 lore raw
Quoted 12 lines · Todd Kjos
Quoted 11 lines · Greg KH
Quoted 9 lines · Todd Kjos

An munmap() on a binder device causes binder_vma_close() to be called which clears the alloc->vma pointer.

If direct reclaim causes binder_alloc_free_page() to be called, there is a race where alloc->vma is read into a local vma pointer and then used later after the mm->mmap_sem is acquired. This can result in calling zap_page_range() with an invalid vma which manifests as a use-after-free in zap_page_range().

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

The fix is to check alloc->vma after acquiring the mmap_sem (which we were acquiring anyway) and skip zap_page_range() if it has changed to NULL.

Awesome patch,

thanks!

- Joel

Quoted 20 lines
Quoted 3 lines
Quoted 1 line

---

Any specific commit that this fixes?

No, it's been there a long time.

Quoted 1 line

And should it be marked for stable releases?

It is needed in stable (back to 4.4), but will need to be backported. Should I post backported versions targeting the specific releases now? I was thinking we'd wait for this one to land. I think we'll need 1 patch for 4.4/4.9 and a second one for 4.14/4.19 (and some of those backported patches will have conflicts when merged down to android-4.X -- I think the 4.14/4.19 version will apply to all the android branches). Let me know how you want to handle this.

Quoted 4 lines

thanks,

greg k-h