Skip to contentSkip to conversation

um: vector: fix use-after-free in vector_mmsg_rx()

PATCH Applied Michael Bommarito 3 participants 5 messages Jun 22 – Jul 23 User-mode Linux (UML)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 um: vector: fix use-after-free in vector_mmsg_rx()

When vector_mmsg_rx() discards a packet whose overlay header fails verify_header(), it frees the skb and continues the loop:

	if (header_check < 0) {
		dev_kfree_skb_irq(skb);
		vp->estats.rx_encaps_errors++;
		continue;
	}

The normal and short-packet paths fall through to the bottom of the loop body, which clears the consumed slot and advances the cursors:

	(*skbuff_vector) = NULL;
	mmsg_vector++;
	skbuff_vector++;

The verify_header() < 0 path skips that via continue, so the freed skb is left in skbuff_vector[] and the cursors do not advance. The next iteration reads the same slot, gets the freed skb, and frees it again, producing a refcount underflow / use-after-free in the RX path.

Discard the slot the same way the other paths do before continuing.

Only transports whose verify_header() can return negative are affected:
GRE and L2TPv3 do so on a cookie/session-id mismatch (raw/tap do not), so any peer on such a transport can trigger it without authentication.

Fixes 49da7e64f33e ("High Performance UML Vector Network Driver")
Assisted-by Claude:claude-opus-4-8

---
Reproduced on a KASAN + refcount-full UML build: bring up a vec0 l2tpv3 device and send frames from the host with the right cookie but a wrong session id. Stock kernel logs repeated "uml_l2tpv3: session mismatch" then a refcount_t underflow / use-after-free in dev_kfree_skb_irq_reason() <- vector_poll() <- vector_mmsg_rx(); with this patch the mismatch is logged and the splat is gone (0 underflow, 0 use-after-free). Full dmesg and reproducer available on request.

 arch/um/drivers/vector_kern.c | 3 +++
 1 file changed, 3 insertions(+)
arch/um/drivers/vector_kern.c +3 -0
index 2cc90055499a5..8a70b3a625537 100644
@@ -995,6 +995,9 @@ static int vector_mmsg_rx(struct vector_private *vp, int budget)
 				 */
 					dev_kfree_skb_irq(skb);
 					vp->estats.rx_encaps_errors++;
+					(*skbuff_vector) = NULL;
+					mmsg_vector++;
+					skbuff_vector++;
 					continue;
 				}
 				if (header_check > 0) {

base-commit: ef0c9f75a19532d7675384708fc8621e10850104

Signature · 1 line
2.53.0
#1 lore raw
Quoted 49 lines · Michael Bommarito

When vector_mmsg_rx() discards a packet whose overlay header fails verify_header(), it frees the skb and continues the loop:

        if (header_check < 0) {
                dev_kfree_skb_irq(skb);
                vp->estats.rx_encaps_errors++;
                continue;
        }

The normal and short-packet paths fall through to the bottom of the loop body, which clears the consumed slot and advances the cursors:

        (*skbuff_vector) = NULL;
        mmsg_vector++;
        skbuff_vector++;

The verify_header() < 0 path skips that via continue, so the freed skb is left in skbuff_vector[] and the cursors do not advance. The next iteration reads the same slot, gets the freed skb, and frees it again, producing a refcount underflow / use-after-free in the RX path.

Discard the slot the same way the other paths do before continuing.

Only transports whose verify_header() can return negative are affected:
GRE and L2TPv3 do so on a cookie/session-id mismatch (raw/tap do not), so any peer on such a transport can trigger it without authentication.

Fixes 49da7e64f33e ("High Performance UML Vector Network Driver")
Assisted-by Claude:claude-opus-4-8

---
Reproduced on a KASAN + refcount-full UML build: bring up a vec0 l2tpv3 device and send frames from the host with the right cookie but a wrong session id. Stock kernel logs repeated "uml_l2tpv3: session mismatch" then a refcount_t underflow / use-after-free in dev_kfree_skb_irq_reason() <- vector_poll() <- vector_mmsg_rx(); with this patch the mismatch is logged and the splat is gone (0 underflow, 0 use-after-free). Full dmesg and reproducer available on request.

 arch/um/drivers/vector_kern.c | 3 +++
 1 file changed, 3 insertions(+)
arch/um/drivers/vector_kern.c +0 -0
index 2cc90055499a5..8a70b3a625537 100644

arch/um/drivers/vector_kern.c static int vector_mmsg_rx(struct vector_private *vp, int budget) · line 1000

Quoted 6 lines
arch/um/drivers/vector_kern.c +3 -0
@@ -995,6 +995,9 @@ static int vector_mmsg_rx(struct vector_private *vp, int budget)
                                 */
                                        dev_kfree_skb_irq(skb);
                                        vp->estats.rx_encaps_errors++;
+                                       (*skbuff_vector) = NULL;
+                                       mmsg_vector++;
+                                       skbuff_vector++;

Anton, please have a look.

Signature · 2 lines
Thanks,
//richard
maintainer
#2 lore raw
Quoted 46 lines · Richard Weinberger
Quoted 44 lines · Michael Bommarito

When vector_mmsg_rx() discards a packet whose overlay header fails verify_header(), it frees the skb and continues the loop:

         if (header_check < 0) {
                 dev_kfree_skb_irq(skb);
                 vp->estats.rx_encaps_errors++;
                 continue;
         }

The normal and short-packet paths fall through to the bottom of the loop body, which clears the consumed slot and advances the cursors:

         (*skbuff_vector) = NULL;
         mmsg_vector++;
         skbuff_vector++;

The verify_header() < 0 path skips that via continue, so the freed skb is left in skbuff_vector[] and the cursors do not advance. The next iteration reads the same slot, gets the freed skb, and frees it again, producing a refcount underflow / use-after-free in the RX path.

Discard the slot the same way the other paths do before continuing.

Only transports whose verify_header() can return negative are affected:
GRE and L2TPv3 do so on a cookie/session-id mismatch (raw/tap do not), so any peer on such a transport can trigger it without authentication.

Fixes 49da7e64f33e ("High Performance UML Vector Network Driver")
Assisted-by Claude:claude-opus-4-8

---
Reproduced on a KASAN + refcount-full UML build: bring up a vec0 l2tpv3 device and send frames from the host with the right cookie but a wrong session id. Stock kernel logs repeated "uml_l2tpv3: session mismatch" then a refcount_t underflow / use-after-free in dev_kfree_skb_irq_reason() <- vector_poll() <- vector_mmsg_rx(); with this patch the mismatch is logged and the splat is gone (0 underflow, 0 use-after-free). Full dmesg and reproducer available on request.

  arch/um/drivers/vector_kern.c | 3 +++
  1 file changed, 3 insertions(+)
Quoted 14 lines

arch/um/drivers/vector_kern.c static int vector_mmsg_rx(struct vector_private *vp, int budget) · line 1000

Quoted 11 lines
arch/um/drivers/vector_kern.c +3 -0
index 2cc90055499a5..8a70b3a625537 100644
@@ -995,6 +995,9 @@ static int vector_mmsg_rx(struct vector_private *vp, int budget)
                                  */
                                         dev_kfree_skb_irq(skb);
                                         vp->estats.rx_encaps_errors++;
+                                       (*skbuff_vector) = NULL;
+                                       mmsg_vector++;
+                                       skbuff_vector++;

Anton, please have a look.

I already did with Johannes a while back. Valid bug and can be triggered remotely.

I think I acked the patch at the time.

Signature · 3 lines
Anton R. Ivanov
Cambridgegreys Limited. Registered in England. Company Number 10273661
https://www.cambridgegreys.com/
#3 lore raw
Quoted 4 lines · Anton Ivanov

I already did with Johannes a while back. Valid bug and can be triggered remotely.

I think I acked the patch at the time.

Yup, back in April. I think it got lost in the security triage at the time and I forgot to follow up until June.

Sorry for not providing that context.

Thanks,
Mike

#4 lore raw
In reply to Michael Bommarito
Quoted 41 lines · Michael Bommarito

When vector_mmsg_rx() discards a packet whose overlay header fails verify_header(), it frees the skb and continues the loop:

        if (header_check < 0) {
                dev_kfree_skb_irq(skb);
                vp->estats.rx_encaps_errors++;
                continue;
        }

The normal and short-packet paths fall through to the bottom of the loop body, which clears the consumed slot and advances the cursors:

        (*skbuff_vector) = NULL;
        mmsg_vector++;
        skbuff_vector++;

The verify_header() < 0 path skips that via continue, so the freed skb is left in skbuff_vector[] and the cursors do not advance. The next iteration reads the same slot, gets the freed skb, and frees it again, producing a refcount underflow / use-after-free in the RX path.

Discard the slot the same way the other paths do before continuing.

Only transports whose verify_header() can return negative are affected:
GRE and L2TPv3 do so on a cookie/session-id mismatch (raw/tap do not), so any peer on such a transport can trigger it without authentication.

Fixes 49da7e64f33e ("High Performance UML Vector Network Driver")
Assisted-by Claude:claude-opus-4-8

---
Reproduced on a KASAN + refcount-full UML build: bring up a vec0 l2tpv3 device and send frames from the host with the right cookie but a wrong session id. Stock kernel logs repeated "uml_l2tpv3: session mismatch" then a refcount_t underflow / use-after-free in dev_kfree_skb_irq_reason() <- vector_poll() <- vector_mmsg_rx(); with this patch the mismatch is logged and the splat is gone (0 underflow, 0 use-after-free). Full dmesg and reproducer available on request.

Quoted 2 lines
 arch/um/drivers/vector_kern.c | 3 +++
 1 file changed, 3 insertions(+)

Applied to fixes branch.

Signature · 2 lines
Thanks,
//richard