All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/2] [v3] mm/mempolicy: fix copy_nodes_to_user() for odd inputs
@ 2026-11-20 18:52 Arnd Bergmann
  2026-11-20 18:52 ` [PATCH 2/2] [v3] mm/mempolicy: replace copy_nodes_to_user() interface with put_nodes() Arnd Bergmann
  0 siblings, 1 reply; 2+ messages in thread
From: Arnd Bergmann @ 2026-11-20 18:52 UTC (permalink / raw)
  To: David Hildenbrand, Gregory Price, Andrew Morton, Arnd Bergmann
  Cc: larips, stable, Zi Yan, Joshua Hahn, Rakie Kim, Ying Huang,
	Lorenzo Stoakes (ARM), Johannes Weiner, linux-mm

From: Arnd Bergmann <arnd@arndb.de>

copy_nodes_to_user() is broken for compat userspace in three
distinct ways:

 - Passing a power-of-two size plus 1 as most normal userspace
   does can end up calling compat_put_bitmap() with the wrong
   size, copying an extra 32-bit word from the stack into
   userspace and overflowing both arrays as compat_put_bitmap()
   expects the actual number of bits, not the number plus 1.

 - Asking for an odd number of 32-bit words rounds up the
   userspace buffer size full 64-bit words, leading to
   an out-of-bounds access in userspace.

Fix the calculation of 'copy' to be based on the userspace
word size and use this to for calculating the final number
of bits to copy as well.

Reported-by: larips <larips950@gmail.com>
Reported-by: Gregory Price (Meta) <gourry@gourry.net>
Fixes: e130242dc351 ("mm: simplify compat numa syscalls")
Link: https://lore.kernel.org/all/20261117101910.1820066-1-gourry@gourry.net/
Cc: stable@vger.kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
v3: don't introduce a bug that skip setting a 32-bit word to zero
    where needed.
v2: as Lorenzo pointed out, the off-by-one mistake is part of the ABI
    and cannot change, we just have to use it consistently.
---
 mm/mempolicy.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 62d4c9e57035..afe614efb77c 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -1772,12 +1772,14 @@ static int get_nodes(nodemask_t *nodes, const unsigned long __user *nmask,
 static int copy_nodes_to_user(unsigned long __user *mask, unsigned long maxnode,
 			      nodemask_t *nodes)
 {
-	unsigned long copy = ALIGN(maxnode-1, 64) / 8;
+	unsigned long copy = ALIGN(maxnode-1, BITS_PER_LONG) / 8;
 	unsigned int nbytes = BITS_TO_LONGS(nr_node_ids) * sizeof(long);
 	bool compat = in_compat_syscall();
 
-	if (compat)
+	if (compat) {
+		copy = ALIGN(maxnode-1, BITS_PER_COMPAT_LONG) / 8;
 		nbytes = BITS_TO_COMPAT_LONGS(nr_node_ids) * sizeof(compat_long_t);
+	}
 
 	if (copy > nbytes) {
 		if (copy > PAGE_SIZE)
@@ -1785,12 +1787,11 @@ static int copy_nodes_to_user(unsigned long __user *mask, unsigned long maxnode,
 		if (clear_user((char __user *)mask + nbytes, copy - nbytes))
 			return -EFAULT;
 		copy = nbytes;
-		maxnode = nr_node_ids;
 	}
 
 	if (compat)
 		return compat_put_bitmap((compat_ulong_t __user *)mask,
-					 nodes_addr(*nodes), maxnode);
+					 nodes_addr(*nodes), copy * 8);
 
 	return copy_to_user(mask, nodes_addr(*nodes), copy) ? -EFAULT : 0;
 }
-- 
2.53.0



^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-11-20 18:53 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-20 18:52 [PATCH 1/2] [v3] mm/mempolicy: fix copy_nodes_to_user() for odd inputs Arnd Bergmann
2026-11-20 18:52 ` [PATCH 2/2] [v3] mm/mempolicy: replace copy_nodes_to_user() interface with put_nodes() Arnd Bergmann

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.