[PATCH v4] lib: xarray: free unused spare node in xas_create_range()

[PATCH v4] lib: xarray: free unused spare node in xas_create_range()

[Shardul Bankar]

xas_create_range() is typically called in a retry loop that uses
xas_nomem() to handle -ENOMEM errors. xas_nomem() may allocate a spare
xa_node and store it in xas->xa_alloc for use in the retry.

If the lock is dropped after xas_nomem(), another thread can expand the
xarray tree in the meantime. On the next retry, xas_create_range() can
then succeed without consuming the spare node stored in xas->xa_alloc.
If the function returns without freeing this spare node, it leaks.

xas_create_range() calls xas_create() multiple times in a loop for
different index ranges. A spare node that isn't needed for one range
iteration might be needed for the next, so we cannot free it after each
xas_create() call. We can only safely free it after xas_create_range()
completes.

Fix this by calling xas_destroy() at the end of xas_create_range() to
free any unused spare node. This makes the API safer by default and
prevents callers from needing to remember cleanup.

This fixes a memory leak in mm/khugepaged.c and potentially other
callers that use xas_nomem() with xas_create_range().

Link: https://syzkaller.appspot.com/bug?id=a274d65fc733448ed518ad15481ed575669dd98c
Link: https://lore.kernel.org/all/20251201074540.3576327-1-shardul.b@mpiricsoftware.com/ ("v3")
Fixes: cae106dd67b9 ("mm/khugepaged: refactor collapse_file control flow")
Signed-off-by: Shardul Bankar <shardul.b@mpiricsoftware.com>
---
 v4:
 - Drop redundant `if (xa_alloc)` around xas_destroy(), as xas_destroy()
   already checks xa_alloc internally.
 v3:
 - Move fix from collapse_file() to xas_create_range() as suggested by Matthew Wilcox
 - Fix in library function makes API safer by default, preventing callers from needing
   to remember cleanup
 - Use shared cleanup label that both restore: and success: paths jump to
 - Clean up unused spare node on both success and error exit paths
 v2:
 - Call xas_destroy() on both success and failure
 - Explained retry semantics and xa_alloc / concurrency risk
 - Dropped cleanup_empty_nodes from previous proposal

 lib/xarray.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/lib/xarray.c b/lib/xarray.c
index 9a8b4916540c..f49ccfa5f57d 100644
--- a/lib/xarray.c
+++ b/lib/xarray.c
@@ -744,11 +744,16 @@ void xas_create_range(struct xa_state *xas)
 	xas->xa_shift = shift;
 	xas->xa_sibs = sibs;
 	xas->xa_index = index;
-	return;
+	goto cleanup;
+
 success:
 	xas->xa_index = index;
 	if (xas->xa_node)
 		xas_set_offset(xas);
+
+cleanup:
+	/* Free any unused spare node from xas_nomem() */
+	xas_destroy(xas);
 }
 EXPORT_SYMBOL_GPL(xas_create_range);
 
-- 
2.34.1

Re: [PATCH v4] lib: xarray: free unused spare node in xas_create_range()

[David Hildenbrand (Red Hat)]

On 12/4/25 15:26, Shardul Bankar wrote:
> xas_create_range() is typically called in a retry loop that uses
> xas_nomem() to handle -ENOMEM errors. xas_nomem() may allocate a spare
> xa_node and store it in xas->xa_alloc for use in the retry.
> 
> If the lock is dropped after xas_nomem(), another thread can expand the
> xarray tree in the meantime. On the next retry, xas_create_range() can
> then succeed without consuming the spare node stored in xas->xa_alloc.
> If the function returns without freeing this spare node, it leaks.
> 
> xas_create_range() calls xas_create() multiple times in a loop for
> different index ranges. A spare node that isn't needed for one range
> iteration might be needed for the next, so we cannot free it after each
> xas_create() call. We can only safely free it after xas_create_range()
> completes.
> 
> Fix this by calling xas_destroy() at the end of xas_create_range() to
> free any unused spare node. This makes the API safer by default and
> prevents callers from needing to remember cleanup.
> 
> This fixes a memory leak in mm/khugepaged.c and potentially other
> callers that use xas_nomem() with xas_create_range().
> 
> Link: https://syzkaller.appspot.com/bug?id=a274d65fc733448ed518ad15481ed575669dd98c
> Link: https://lore.kernel.org/all/20251201074540.3576327-1-shardul.b@mpiricsoftware.com/ ("v3")
> Fixes: cae106dd67b9 ("mm/khugepaged: refactor collapse_file control flow")
> Signed-off-by: Shardul Bankar <shardul.b@mpiricsoftware.com>
> ---
>   v4:
>   - Drop redundant `if (xa_alloc)` around xas_destroy(), as xas_destroy()
>     already checks xa_alloc internally.
>   v3:
>   - Move fix from collapse_file() to xas_create_range() as suggested by Matthew Wilcox
>   - Fix in library function makes API safer by default, preventing callers from needing
>     to remember cleanup
>   - Use shared cleanup label that both restore: and success: paths jump to
>   - Clean up unused spare node on both success and error exit paths
>   v2:
>   - Call xas_destroy() on both success and failure
>   - Explained retry semantics and xa_alloc / concurrency risk
>   - Dropped cleanup_empty_nodes from previous proposal
> 
>   lib/xarray.c | 7 ++++++-
>   1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/lib/xarray.c b/lib/xarray.c
> index 9a8b4916540c..f49ccfa5f57d 100644
> --- a/lib/xarray.c
> +++ b/lib/xarray.c
> @@ -744,11 +744,16 @@ void xas_create_range(struct xa_state *xas)
>   	xas->xa_shift = shift;
>   	xas->xa_sibs = sibs;
>   	xas->xa_index = index;
> -	return;
> +	goto cleanup;
> +
>   success:
>   	xas->xa_index = index;
>   	if (xas->xa_node)
>   		xas_set_offset(xas);
> +
> +cleanup:
> +	/* Free any unused spare node from xas_nomem() */
> +	xas_destroy(xas);
>   }
>   EXPORT_SYMBOL_GPL(xas_create_range);
>   

Nothing jumped at me, except that the label situation is a bit
suboptimal.

Hoping Willy will take another look as well.

Reviewed-by: David Hildenbrand (Red Hat) <david@kernel.org>


BTW, do we have a way to test this in a test case?


A follow-up cleanup that avoids labels could be something like (untested):


diff --git a/lib/xarray.c b/lib/xarray.c
index 9a8b4916540cf..325f264530fb2 100644
--- a/lib/xarray.c
+++ b/lib/xarray.c
@@ -714,6 +714,7 @@ void xas_create_range(struct xa_state *xas)
         unsigned long index = xas->xa_index;
         unsigned char shift = xas->xa_shift;
         unsigned char sibs = xas->xa_sibs;
+       bool success = false;
  
         xas->xa_index |= ((sibs + 1UL) << shift) - 1;
         if (xas_is_node(xas) && xas->xa_node->shift == xas->xa_shift)
@@ -724,9 +725,11 @@ void xas_create_range(struct xa_state *xas)
         for (;;) {
                 xas_create(xas, true);
                 if (xas_error(xas))
-                       goto restore;
-               if (xas->xa_index <= (index | XA_CHUNK_MASK))
-                       goto success;
+                       break
+               if (xas->xa_index <= (index | XA_CHUNK_MASK)) {
+                       succeess = true;
+                       break;
+               }
                 xas->xa_index -= XA_CHUNK_SIZE;
  
                 for (;;) {
@@ -740,15 +743,17 @@ void xas_create_range(struct xa_state *xas)
                 }
         }
  
-restore:
-       xas->xa_shift = shift;
-       xas->xa_sibs = sibs;
-       xas->xa_index = index;
-       return;
-success:
-       xas->xa_index = index;
-       if (xas->xa_node)
-               xas_set_offset(xas);
+       if (success) {
+               xas->xa_index = index;
+               if (xas->xa_node)
+                       xas_set_offset(xas);
+       } else {
+               xas->xa_shift = shift;
+               xas->xa_sibs = sibs;
+               xas->xa_index = index;
+       }
+       /* Free any unused spare node from xas_nomem() */
+       xas_destroy(xas);
  }
  EXPORT_SYMBOL_GPL(xas_create_range);


-- 
Cheers

David

Re: [PATCH v4] lib: xarray: free unused spare node in xas_create_range()

[Shardul Bankar]

On Fri, 2025-12-05 at 08:22 +0100, David Hildenbrand (Red Hat) wrote:
> > Link:
> > https://syzkaller.appspot.com/bug?id=a274d65fc733448ed518ad15481ed575669dd98c
> ...
> Reviewed-by: David Hildenbrand (Red Hat) <david@kernel.org>
> 
> 
> BTW, do we have a way to test this in a test case?
Hi David,

Thanks for the review and the Reviewed-by.

Regarding a test case: I don’t have a focused selftest or fault-
injection setup yet that reliably hits this xas_nomem() +
xas_create_range() corner case. 
I noticed this spare-node leak while analyzing the Syzbot report I
referenced in the Link: tag, but the reproducer I see there doesn’t
isolate this path and reports other kmemleaks.

For now I’d prefer to treat this as a small correctness fix in xarray
itself. If I manage to come up with a robust way to exercise this path
in a selftest (e.g. via targeted fault injection in lib/test_xarray.c),
I can follow up with a separate patch, but I don’t have anything solid
to propose today.

> 
> 
> A follow-up cleanup that avoids labels could be something like
> (untested):
> 
> 
> diff --git a/lib/xarray.c b/lib/xarray.c
> index 9a8b4916540cf..325f264530fb2 100644
> --- a/lib/xarray.c
> +++ b/lib/xarray.c
> @@ -714,6 +714,7 @@ void xas_create_range(struct xa_state *xas)
>          unsigned long index = xas->xa_index;
>          unsigned char shift = xas->xa_shift;
>          unsigned char sibs = xas->xa_sibs;
> +       bool success = false;
>   
>          xas->xa_index |= ((sibs + 1UL) << shift) - 1;
>          if (xas_is_node(xas) && xas->xa_node->shift == xas-
> >xa_shift)
> @@ -724,9 +725,11 @@ void xas_create_range(struct xa_state *xas)
>          for (;;) {
>                  xas_create(xas, true);
>                  if (xas_error(xas))
> -                       goto restore;
> -               if (xas->xa_index <= (index | XA_CHUNK_MASK))
> -                       goto success;
> +                       break
> +               if (xas->xa_index <= (index | XA_CHUNK_MASK)) {
> +                       succeess = true;
> +                       break;
> +               }
>                  xas->xa_index -= XA_CHUNK_SIZE;
>   
>                  for (;;) {
> @@ -740,15 +743,17 @@ void xas_create_range(struct xa_state *xas)
>                  }
>          }
>   
> -restore:
> -       xas->xa_shift = shift;
> -       xas->xa_sibs = sibs;
> -       xas->xa_index = index;
> -       return;
> -success:
> -       xas->xa_index = index;
> -       if (xas->xa_node)
> -               xas_set_offset(xas);
> +       if (success) {
> +               xas->xa_index = index;
> +               if (xas->xa_node)
> +                       xas_set_offset(xas);
> +       } else {
> +               xas->xa_shift = shift;
> +               xas->xa_sibs = sibs;
> +               xas->xa_index = index;
> +       }
> +       /* Free any unused spare node from xas_nomem() */
> +       xas_destroy(xas);
>   }
>   EXPORT_SYMBOL_GPL(xas_create_range);
> 
> 
Your bool-based version reads nicer; I’m happy to follow up with a
small cleanup patch on top that switches xas_create_range() over to
that style (with a Suggested-by tag).

Thanks and Regards,
Shardul