[PATCH 0/3] vhost: VDUSE-related fixes

[PATCH 0/3] vhost: VDUSE-related fixes

[Maxime Coquelin]

This series contains 3 fixes for issues spotted by Claude Code.

The first one is to avoid out-of-bound accesses in virtqueues array
in the case we have the maximum supported queue pairs and control queue.

Second one is a security issue that could result in theory in a denial
of service, but a CVE was not created because the control queue support
cannot currently be negotiated with the Kernel VDUSE driver.

Last patch is fixing mmap error handling in the VDUSE IOTLB miss handler.

Maxime Coquelin (3):
  vhost: fix virtqueue array size for control queue
  vhost: fix descriptor chain bounds check in control queue
  vhost: fix mmap error check in VDUSE IOTLB miss handler

 lib/vhost/vduse.c           |  5 +++--
 lib/vhost/vhost.h           |  5 +++--
 lib/vhost/virtio_net_ctrl.c | 22 ++++++++++++++++++++--
 3 files changed, 26 insertions(+), 6 deletions(-)

-- 
2.52.0

[PATCH 1/3] vhost: fix virtqueue array size for control queue

[Maxime Coquelin]

When max_queue_pairs is set to VHOST_MAX_QUEUE_PAIRS (128), VDUSE
calculates total_queues as max_queue_pairs * 2 + 1 = 257 to account
for the control queue. However, the virtqueue array was sized as
VHOST_MAX_QUEUE_PAIRS * 2, causing an out-of-bounds array access.

Fix by defining VHOST_MAX_VRING to explicitly account for the control
queue (VHOST_MAX_QUEUE_PAIRS * 2 + 1) and using it for the virtqueue
array size.

Fixes: f0a37cc6a1e2 ("vhost: add multiqueue support to VDUSE")
Cc: stable@dpdk.org

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
 lib/vhost/vhost.h | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/lib/vhost/vhost.h b/lib/vhost/vhost.h
index e9e71c1707..ee61f7415e 100644
--- a/lib/vhost/vhost.h
+++ b/lib/vhost/vhost.h
@@ -261,8 +261,9 @@ struct vhost_async {
 };
 
 #define VHOST_RECONNECT_VERSION		0x0
-#define VHOST_MAX_VRING			0x100
 #define VHOST_MAX_QUEUE_PAIRS		0x80
+/* Max vring count: 2 per queue pair plus 1 control queue */
+#define VHOST_MAX_VRING			(VHOST_MAX_QUEUE_PAIRS * 2 + 1)
 
 struct __rte_cache_aligned vhost_reconnect_vring {
 	uint16_t last_avail_idx;
@@ -501,7 +502,7 @@ struct __rte_cache_aligned virtio_net {
 
 	int			extbuf;
 	int			linearbuf;
-	struct vhost_virtqueue	*virtqueue[VHOST_MAX_QUEUE_PAIRS * 2];
+	struct vhost_virtqueue	*virtqueue[VHOST_MAX_VRING];
 
 	rte_rwlock_t	iotlb_pending_lock;
 	struct vhost_iotlb_entry *iotlb_pool;
-- 
2.52.0

Re: [PATCH 1/3] vhost: fix virtqueue array size for control queue

[David Marchand]

On Thu, 8 Jan 2026 at 14:50, Maxime Coquelin <maxime.coquelin@redhat.com> wrote:
>
> When max_queue_pairs is set to VHOST_MAX_QUEUE_PAIRS (128), VDUSE
> calculates total_queues as max_queue_pairs * 2 + 1 = 257 to account
> for the control queue. However, the virtqueue array was sized as
> VHOST_MAX_QUEUE_PAIRS * 2, causing an out-of-bounds array access.
>
> Fix by defining VHOST_MAX_VRING to explicitly account for the control
> queue (VHOST_MAX_QUEUE_PAIRS * 2 + 1) and using it for the virtqueue
> array size.
>
> Fixes: f0a37cc6a1e2 ("vhost: add multiqueue support to VDUSE")
> Cc: stable@dpdk.org
>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Reviewed-by: David Marchand <david.marchand@redhat.com>


-- 
David Marchand

[PATCH 2/3] vhost: fix descriptor chain bounds check in control queue

[Maxime Coquelin]

The virtio_net_ctrl_pop() function traverses descriptor chains from
guest-controlled memory without validating that the descriptor index
stays within bounds and without a counter to prevent infinite loops
from circular chains.

A malicious guest could craft descriptors with a next field pointing
out of bounds causing memory corruption, or create circular descriptor
chains causing an infinite loop and denial of service.

Add bounds checking and a loop counter to both descriptor chain
traversal loops, similar to the existing protection in virtio_net.c
fill_vec_buf_split().

Fixes: 474f4d7840ad ("vhost: add control virtqueue")
Cc: stable@dpdk.org

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
 lib/vhost/virtio_net_ctrl.c | 22 ++++++++++++++++++++--
 1 file changed, 20 insertions(+), 2 deletions(-)

diff --git a/lib/vhost/virtio_net_ctrl.c b/lib/vhost/virtio_net_ctrl.c
index 603a8db728..8149885384 100644
--- a/lib/vhost/virtio_net_ctrl.c
+++ b/lib/vhost/virtio_net_ctrl.c
@@ -28,7 +28,7 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
 		struct virtio_net_ctrl_elem *ctrl_elem)
 	__rte_requires_shared_capability(&cvq->iotlb_lock)
 {
-	uint16_t avail_idx, desc_idx, n_descs = 0;
+	uint16_t avail_idx, desc_idx, n_descs = 0, nr_descs, cnt = 0;
 	uint64_t desc_len, desc_addr, desc_iova, data_len = 0;
 	uint8_t *ctrl_req;
 	struct vring_desc *descs;
@@ -59,12 +59,19 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
 			goto err;
 		}
 
+		nr_descs = desc_len / sizeof(struct vring_desc);
 		desc_idx = 0;
 	} else {
 		descs = cvq->desc;
+		nr_descs = cvq->size;
 	}
 
 	while (1) {
+		if (unlikely(desc_idx >= nr_descs || ++cnt > nr_descs)) {
+			VHOST_CONFIG_LOG(dev->ifname, ERR, "Invalid ctrl descriptor chain");
+			goto err;
+		}
+
 		desc_len = descs[desc_idx].len;
 		desc_iova = descs[desc_idx].addr;
 
@@ -142,12 +149,23 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
 			goto free_err;
 		}
 
+		nr_descs = desc_len / sizeof(struct vring_desc);
 		desc_idx = 0;
 	} else {
 		descs = cvq->desc;
+		nr_descs = cvq->size;
 	}
 
-	while (!(descs[desc_idx].flags & VRING_DESC_F_WRITE)) {
+	cnt = 0;
+	while (1) {
+		if (unlikely(desc_idx >= nr_descs || ++cnt > nr_descs)) {
+			VHOST_CONFIG_LOG(dev->ifname, ERR, "Invalid ctrl descriptor chain");
+			goto free_err;
+		}
+
+		if (descs[desc_idx].flags & VRING_DESC_F_WRITE)
+			break;
+
 		desc_len = descs[desc_idx].len;
 		desc_iova = descs[desc_idx].addr;
 
-- 
2.52.0

Re: [PATCH 2/3] vhost: fix descriptor chain bounds check in control queue

[David Marchand]

On Thu, 8 Jan 2026 at 14:50, Maxime Coquelin <maxime.coquelin@redhat.com> wrote:
>
> The virtio_net_ctrl_pop() function traverses descriptor chains from
> guest-controlled memory without validating that the descriptor index
> stays within bounds and without a counter to prevent infinite loops
> from circular chains.
>
> A malicious guest could craft descriptors with a next field pointing
> out of bounds causing memory corruption, or create circular descriptor
> chains causing an infinite loop and denial of service.
>
> Add bounds checking and a loop counter to both descriptor chain
> traversal loops, similar to the existing protection in virtio_net.c
> fill_vec_buf_split().

Too bad we have two separate implementations..

>
> Fixes: 474f4d7840ad ("vhost: add control virtqueue")
> Cc: stable@dpdk.org
>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> ---
>  lib/vhost/virtio_net_ctrl.c | 22 ++++++++++++++++++++--
>  1 file changed, 20 insertions(+), 2 deletions(-)
>
> diff --git a/lib/vhost/virtio_net_ctrl.c b/lib/vhost/virtio_net_ctrl.c
> index 603a8db728..8149885384 100644
> --- a/lib/vhost/virtio_net_ctrl.c
> +++ b/lib/vhost/virtio_net_ctrl.c
> @@ -28,7 +28,7 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
>                 struct virtio_net_ctrl_elem *ctrl_elem)
>         __rte_requires_shared_capability(&cvq->iotlb_lock)
>  {
> -       uint16_t avail_idx, desc_idx, n_descs = 0;
> +       uint16_t avail_idx, desc_idx, n_descs = 0, nr_descs, cnt = 0;
>         uint64_t desc_len, desc_addr, desc_iova, data_len = 0;
>         uint8_t *ctrl_req;
>         struct vring_desc *descs;
> @@ -59,12 +59,19 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
>                         goto err;
>                 }
>
> +               nr_descs = desc_len / sizeof(struct vring_desc);
>                 desc_idx = 0;
>         } else {
>                 descs = cvq->desc;
> +               nr_descs = cvq->size;
>         }
>
>         while (1) {
> +               if (unlikely(desc_idx >= nr_descs || ++cnt > nr_descs)) {

I don't like pre increment.
Can we use post increment and stay aligned with known to work
fill_vec_buf_split?


> +                       VHOST_CONFIG_LOG(dev->ifname, ERR, "Invalid ctrl descriptor chain");
> +                       goto err;
> +               }
> +
>                 desc_len = descs[desc_idx].len;
>                 desc_iova = descs[desc_idx].addr;
>
> @@ -142,12 +149,23 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
>                         goto free_err;
>                 }
>
> +               nr_descs = desc_len / sizeof(struct vring_desc);
>                 desc_idx = 0;
>         } else {
>                 descs = cvq->desc;
> +               nr_descs = cvq->size;
>         }
>
> -       while (!(descs[desc_idx].flags & VRING_DESC_F_WRITE)) {
> +       cnt = 0;
> +       while (1) {
> +               if (unlikely(desc_idx >= nr_descs || ++cnt > nr_descs)) {

Idem.


> +                       VHOST_CONFIG_LOG(dev->ifname, ERR, "Invalid ctrl descriptor chain");
> +                       goto free_err;
> +               }
> +
> +               if (descs[desc_idx].flags & VRING_DESC_F_WRITE)
> +                       break;
> +
>                 desc_len = descs[desc_idx].len;
>                 desc_iova = descs[desc_idx].addr;
>

I wonder if we are missing some call to vhost_alloc_copy_ind_table() as well.


-- 
David Marchand

[PATCH 3/3] vhost: fix mmap error check in VDUSE IOTLB miss handler

[Maxime Coquelin]

The mmap() function returns MAP_FAILED on failure, not NULL.
The current check for !mmap_addr will never detect mmap failures.

When mmap fails but the error is not detected, an invalid address (-1)
is inserted into the IOTLB cache via vhost_user_iotlb_cache_insert().
Subsequent attempts to access this address will cause memory
corruption or crash.

Fix by checking for MAP_FAILED instead of NULL. Also add strerror to
the error message for easier debugging.

Fixes: f27d5206c598 ("vhost: add VDUSE callback for IOTLB miss")
Cc: stable@dpdk.org

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
 lib/vhost/vduse.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/lib/vhost/vduse.c b/lib/vhost/vduse.c
index 897dee9f1b..0b5d158fee 100644
--- a/lib/vhost/vduse.c
+++ b/lib/vhost/vduse.c
@@ -86,9 +86,10 @@ vduse_iotlb_miss(struct virtio_net *dev, uint64_t iova, uint8_t perm __rte_unuse
 
 	size = entry.last - entry.start + 1;
 	mmap_addr = mmap(0, size + entry.offset, entry.perm, MAP_SHARED, fd, 0);
-	if (!mmap_addr) {
+	if (mmap_addr == MAP_FAILED) {
 		VHOST_CONFIG_LOG(dev->ifname, ERR,
-				"Failed to mmap IOTLB entry for 0x%" PRIx64, iova);
+				"Failed to mmap IOTLB entry for 0x%" PRIx64 ": %s",
+				iova, strerror(errno));
 		ret = -1;
 		goto close_fd;
 	}
-- 
2.52.0

Re: [PATCH 3/3] vhost: fix mmap error check in VDUSE IOTLB miss handler

[David Marchand]

On Thu, 8 Jan 2026 at 14:50, Maxime Coquelin <maxime.coquelin@redhat.com> wrote:
>
> The mmap() function returns MAP_FAILED on failure, not NULL.
> The current check for !mmap_addr will never detect mmap failures.
>
> When mmap fails but the error is not detected, an invalid address (-1)
> is inserted into the IOTLB cache via vhost_user_iotlb_cache_insert().
> Subsequent attempts to access this address will cause memory
> corruption or crash.
>
> Fix by checking for MAP_FAILED instead of NULL. Also add strerror to
> the error message for easier debugging.
>
> Fixes: f27d5206c598 ("vhost: add VDUSE callback for IOTLB miss")
> Cc: stable@dpdk.org
>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

Ugh.. I should have caught it when reviewing initial commit..
Reviewed-by: David Marchand <david.marchand@redhat.com>


-- 
David Marchand