Hi All: This patch tries to validate the used buffer length in the virtio core. This help to eliminate the unexpected result caused by buggy or mailicous devices. For the drivers that can do the validation itself, they can ask the virtio core to suppress the check. Changes since V4: - Fix the out of date description in the commit log Changes since V3: - Initialize the buflen to zero when the validation is done by the driver. Jason Wang (4): virtio_ring: validate used buffer length virtio-net: don't let virtio core to validate used length virtio-blk: don't let virtio core to validate used length virtio-scsi: don't let virtio core to validate used buffer length drivers/block/virtio_blk.c | 1 + drivers/net/virtio_net.c | 1 + drivers/scsi/virtio_scsi.c | 1 + drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++++++++++++++++++++ include/linux/virtio.h | 2 ++ 5 files changed, 65 insertions(+) -- 2.25.1
This patch validate the used buffer length provided by the device
before trying to use it. This is done by record the in buffer length
in a new field in desc_state structure during virtqueue_add(), then we
can fail the virtqueue_get_buf() when we find the device is trying to
give us a used buffer length which is greater than the in buffer
length.
Since some drivers have already done the validation by themselves,
this patch tries to makes the core validation optional. For the driver
that doesn't want the validation, it can set the
suppress_used_validation to be true (which could be overridden by
force_used_validation module parameter). To be more efficient, a
dedicate array is used for storing the validate used length, this
helps to eliminate the cache stress if validation is done by the
driver.
Signed-off-by: Jason Wang <jasowang at redhat.com>
---
drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++++++++++++++++++++
include/linux/virtio.h | 2 ++
2 files changed, 62 insertions(+)
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 4c0ec82cef56..a6e5a3b94337 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -14,6 +14,9 @@
#include <linux/spinlock.h>
#include <xen/xen.h>
+static bool force_used_validation = false;
+module_param(force_used_validation, bool, 0444);
+
#ifdef DEBUG
/* For development, we want to crash whenever the ring is screwed. */
#define BAD_RING(_vq, fmt, args...) \
@@ -182,6 +185,9 @@ struct vring_virtqueue {
} packed;
};
+ /* Per-descriptor in buffer length */
+ u32 *buflen;
+
/* How to notify other side. FIXME: commonalize hcalls! */
bool (*notify)(struct virtqueue *vq);
@@ -490,6 +496,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
unsigned int i, n, avail, descs_used, prev, err_idx;
int head;
bool indirect;
+ u32 buflen = 0;
START_USE(vq);
@@ -571,6 +578,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
VRING_DESC_F_NEXT |
VRING_DESC_F_WRITE,
indirect);
+ buflen += sg->length;
}
}
/* Last one doesn't continue. */
@@ -610,6 +618,10 @@ static inline int virtqueue_add_split(struct virtqueue
*_vq,
else
vq->split.desc_state[head].indir_desc = ctx;
+ /* Store in buffer length if necessary */
+ if (vq->buflen)
+ vq->buflen[head] = buflen;
+
/* Put entry in available array (but don't update avail->idx until they
* do sync). */
avail = vq->split.avail_idx_shadow & (vq->split.vring.num - 1);
@@ -784,6 +796,11 @@ static void *virtqueue_get_buf_ctx_split(struct virtqueue
*_vq,
BAD_RING(vq, "id %u is not a head!\n", i);
return NULL;
}
+ if (vq->buflen && unlikely(*len > vq->buflen[i])) {
+ BAD_RING(vq, "used len %d is larger than in buflen %u\n",
+ *len, vq->buflen[i]);
+ return NULL;
+ }
/* detach_buf_split clears data, so grab it now. */
ret = vq->split.desc_state[i].data;
@@ -1062,6 +1079,7 @@ static int virtqueue_add_indirect_packed(struct
vring_virtqueue *vq,
unsigned int i, n, err_idx;
u16 head, id;
dma_addr_t addr;
+ u32 buflen = 0;
head = vq->packed.next_avail_idx;
desc = alloc_indirect_packed(total_sg, gfp);
@@ -1091,6 +1109,8 @@ static int virtqueue_add_indirect_packed(struct
vring_virtqueue *vq,
desc[i].addr = cpu_to_le64(addr);
desc[i].len = cpu_to_le32(sg->length);
i++;
+ if (n >= out_sgs)
+ buflen += sg->length;
}
}
@@ -1144,6 +1164,10 @@ static int virtqueue_add_indirect_packed(struct
vring_virtqueue *vq,
vq->packed.desc_state[id].indir_desc = desc;
vq->packed.desc_state[id].last = id;
+ /* Store in buffer length if necessary */
+ if (vq->buflen)
+ vq->buflen[id] = buflen;
+
vq->num_added += 1;
pr_debug("Added buffer head %i to %p\n", head, vq);
@@ -1179,6 +1203,7 @@ static inline int virtqueue_add_packed(struct virtqueue
*_vq,
__le16 head_flags, flags;
u16 head, id, prev, curr, avail_used_flags;
int err;
+ u32 buflen = 0;
START_USE(vq);
@@ -1258,6 +1283,8 @@ static inline int virtqueue_add_packed(struct virtqueue
*_vq,
1 << VRING_PACKED_DESC_F_AVAIL |
1 << VRING_PACKED_DESC_F_USED;
}
+ if (n >= out_sgs)
+ buflen += sg->length;
}
}
@@ -1277,6 +1304,10 @@ static inline int virtqueue_add_packed(struct virtqueue
*_vq,
vq->packed.desc_state[id].indir_desc = ctx;
vq->packed.desc_state[id].last = prev;
+ /* Store in buffer length if necessary */
+ if (vq->buflen)
+ vq->buflen[id] = buflen;
+
/*
* A driver MUST NOT make the first descriptor in the list
* available before all subsequent descriptors comprising
@@ -1463,6 +1494,11 @@ static void *virtqueue_get_buf_ctx_packed(struct
virtqueue *_vq,
BAD_RING(vq, "id %u is not a head!\n", id);
return NULL;
}
+ if (vq->buflen && unlikely(*len > vq->buflen[id])) {
+ BAD_RING(vq, "used len %d is larger than in buflen %u\n",
+ *len, vq->buflen[id]);
+ return NULL;
+ }
/* detach_buf_packed clears data, so grab it now. */
ret = vq->packed.desc_state[id].data;
@@ -1668,6 +1704,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
struct vring_virtqueue *vq;
struct vring_packed_desc *ring;
struct vring_packed_desc_event *driver, *device;
+ struct virtio_driver *drv = drv_to_virtio(vdev->dev.driver);
dma_addr_t ring_dma_addr, driver_event_dma_addr, device_event_dma_addr;
size_t ring_size_in_bytes, event_size_in_bytes;
@@ -1757,6 +1794,15 @@ static struct virtqueue *vring_create_virtqueue_packed(
if (!vq->packed.desc_extra)
goto err_desc_extra;
+ if (!drv->suppress_used_validation || force_used_validation) {
+ vq->buflen = kmalloc_array(num, sizeof(*vq->buflen),
+ GFP_KERNEL);
+ if (!vq->buflen)
+ goto err_buflen;
+ } else {
+ vq->buflen = NULL;
+ }
+
/* No callback? Tell other side not to bother us. */
if (!callback) {
vq->packed.event_flags_shadow = VRING_PACKED_EVENT_FLAG_DISABLE;
@@ -1769,6 +1815,8 @@ static struct virtqueue *vring_create_virtqueue_packed(
spin_unlock(&vdev->vqs_list_lock);
return &vq->vq;
+err_buflen:
+ kfree(vq->packed.desc_extra);
err_desc_extra:
kfree(vq->packed.desc_state);
err_desc_state:
@@ -2176,6 +2224,7 @@ struct virtqueue *__vring_new_virtqueue(unsigned int
index,
void (*callback)(struct virtqueue *),
const char *name)
{
+ struct virtio_driver *drv = drv_to_virtio(vdev->dev.driver);
struct vring_virtqueue *vq;
if (virtio_has_feature(vdev, VIRTIO_F_RING_PACKED))
@@ -2235,6 +2284,15 @@ struct virtqueue *__vring_new_virtqueue(unsigned int
index,
if (!vq->split.desc_extra)
goto err_extra;
+ if (!drv->suppress_used_validation || force_used_validation) {
+ vq->buflen = kmalloc_array(vring.num, sizeof(*vq->buflen),
+ GFP_KERNEL);
+ if (!vq->buflen)
+ goto err_buflen;
+ } else {
+ vq->buflen = NULL;
+ }
+
/* Put everything in free lists. */
vq->free_head = 0;
memset(vq->split.desc_state, 0, vring.num *
@@ -2245,6 +2303,8 @@ struct virtqueue *__vring_new_virtqueue(unsigned int
index,
spin_unlock(&vdev->vqs_list_lock);
return &vq->vq;
+err_buflen:
+ kfree(vq->split.desc_extra);
err_extra:
kfree(vq->split.desc_state);
err_state:
diff --git a/include/linux/virtio.h b/include/linux/virtio.h
index 41edbc01ffa4..44d0e09da2d9 100644
--- a/include/linux/virtio.h
+++ b/include/linux/virtio.h
@@ -152,6 +152,7 @@ size_t virtio_max_dma_size(struct virtio_device *vdev);
* @feature_table_size: number of entries in the feature table array.
* @feature_table_legacy: same as feature_table but when working in legacy
mode.
* @feature_table_size_legacy: number of entries in feature table legacy array.
+ * @suppress_used_validation: set to not have core validate used length
* @probe: the function to call when a device is found. Returns 0 or -errno.
* @scan: optional function to call after successful probe; intended
* for virtio-scsi to invoke a scan.
@@ -168,6 +169,7 @@ struct virtio_driver {
unsigned int feature_table_size;
const unsigned int *feature_table_legacy;
unsigned int feature_table_size_legacy;
+ bool suppress_used_validation;
int (*validate)(struct virtio_device *dev);
int (*probe)(struct virtio_device *dev);
void (*scan)(struct virtio_device *dev);
--
2.25.1
Jason Wang
2021-Oct-27 02:21 UTC
[PATCH V5 2/4] virtio-net: don't let virtio core to validate used length
For RX virtuqueue, the used length is validated in all the three paths
(big, small and mergeable). For control vq, we never tries to use used
length. So this patch forbids the core to validate the used length.
Signed-off-by: Jason Wang <jasowang at redhat.com>
---
drivers/net/virtio_net.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 6d8c8745bf24..7c43bfc1ce44 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -3385,6 +3385,7 @@ static struct virtio_driver virtio_net_driver = {
.feature_table_size = ARRAY_SIZE(features),
.feature_table_legacy = features_legacy,
.feature_table_size_legacy = ARRAY_SIZE(features_legacy),
+ .suppress_used_validation = true,
.driver.name = KBUILD_MODNAME,
.driver.owner = THIS_MODULE,
.id_table = id_table,
--
2.25.1
Jason Wang
2021-Oct-27 02:21 UTC
[PATCH V5 3/4] virtio-blk: don't let virtio core to validate used length
We never tries to use used length, so the patch prevents the virtio
core from validating used length.
Signed-off-by: Jason Wang <jasowang at redhat.com>
---
drivers/block/virtio_blk.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index c7d05ff24084..36d645ec5002 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -1041,6 +1041,7 @@ static struct virtio_driver virtio_blk = {
.feature_table_size = ARRAY_SIZE(features),
.feature_table_legacy = features_legacy,
.feature_table_size_legacy = ARRAY_SIZE(features_legacy),
+ .suppress_used_validation = true,
.driver.name = KBUILD_MODNAME,
.driver.owner = THIS_MODULE,
.id_table = id_table,
--
2.25.1
Jason Wang
2021-Oct-27 02:21 UTC
[PATCH V5 4/4] virtio-scsi: don't let virtio core to validate used buffer length
We never tries to use used length, so the patch prevents the virtio
core from validating used length.
Signed-off-by: Jason Wang <jasowang at redhat.com>
---
drivers/scsi/virtio_scsi.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/scsi/virtio_scsi.c b/drivers/scsi/virtio_scsi.c
index 07d0250f17c3..03b09ecea42d 100644
--- a/drivers/scsi/virtio_scsi.c
+++ b/drivers/scsi/virtio_scsi.c
@@ -977,6 +977,7 @@ static unsigned int features[] = {
static struct virtio_driver virtio_scsi_driver = {
.feature_table = features,
.feature_table_size = ARRAY_SIZE(features),
+ .suppress_used_validation = true,
.driver.name = KBUILD_MODNAME,
.driver.owner = THIS_MODULE,
.id_table = id_table,
--
2.25.1
On Wed, 27 Oct 2021 10:21:04 +0800, Jason Wang <jasowang at redhat.com> wrote:> This patch validate the used buffer length provided by the device > before trying to use it. This is done by record the in buffer length > in a new field in desc_state structure during virtqueue_add(), then we > can fail the virtqueue_get_buf() when we find the device is trying to > give us a used buffer length which is greater than the in buffer > length. > > Since some drivers have already done the validation by themselves, > this patch tries to makes the core validation optional. For the driver > that doesn't want the validation, it can set the > suppress_used_validation to be true (which could be overridden by > force_used_validation module parameter). To be more efficient, a > dedicate array is used for storing the validate used length, this > helps to eliminate the cache stress if validation is done by the > driver. > > Signed-off-by: Jason Wang <jasowang at redhat.com> > --- > drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++++++++++++++++++++ > include/linux/virtio.h | 2 ++ > 2 files changed, 62 insertions(+) > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c > index 4c0ec82cef56..a6e5a3b94337 100644 > --- a/drivers/virtio/virtio_ring.c > +++ b/drivers/virtio/virtio_ring.c > @@ -14,6 +14,9 @@ > #include <linux/spinlock.h> > #include <xen/xen.h> > > +static bool force_used_validation = false; > +module_param(force_used_validation, bool, 0444); > + > #ifdef DEBUG > /* For development, we want to crash whenever the ring is screwed. */ > #define BAD_RING(_vq, fmt, args...) \ > @@ -182,6 +185,9 @@ struct vring_virtqueue { > } packed; > }; > > + /* Per-descriptor in buffer length */ > + u32 *buflen; > + > /* How to notify other side. FIXME: commonalize hcalls! */ > bool (*notify)(struct virtqueue *vq); > > @@ -490,6 +496,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, > unsigned int i, n, avail, descs_used, prev, err_idx; > int head; > bool indirect; > + u32 buflen = 0; > > START_USE(vq); > > @@ -571,6 +578,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, > VRING_DESC_F_NEXT | > VRING_DESC_F_WRITE, > indirect); > + buflen += sg->length; > } > } > /* Last one doesn't continue. */ > @@ -610,6 +618,10 @@ static inline int virtqueue_add_split(struct virtqueue *_vq, > else > vq->split.desc_state[head].indir_desc = ctx; > > + /* Store in buffer length if necessary */ > + if (vq->buflen) > + vq->buflen[head] = buflen; > + > /* Put entry in available array (but don't update avail->idx until they > * do sync). */ > avail = vq->split.avail_idx_shadow & (vq->split.vring.num - 1); > @@ -784,6 +796,11 @@ static void *virtqueue_get_buf_ctx_split(struct virtqueue *_vq, > BAD_RING(vq, "id %u is not a head!\n", i); > return NULL; > } > + if (vq->buflen && unlikely(*len > vq->buflen[i])) { > + BAD_RING(vq, "used len %d is larger than in buflen %u\n", > + *len, vq->buflen[i]); > + return NULL; > + } > > /* detach_buf_split clears data, so grab it now. */ > ret = vq->split.desc_state[i].data; > @@ -1062,6 +1079,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq, > unsigned int i, n, err_idx; > u16 head, id; > dma_addr_t addr; > + u32 buflen = 0; > > head = vq->packed.next_avail_idx; > desc = alloc_indirect_packed(total_sg, gfp); > @@ -1091,6 +1109,8 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq, > desc[i].addr = cpu_to_le64(addr); > desc[i].len = cpu_to_le32(sg->length); > i++; > + if (n >= out_sgs) > + buflen += sg->length; > } > } > > @@ -1144,6 +1164,10 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq, > vq->packed.desc_state[id].indir_desc = desc; > vq->packed.desc_state[id].last = id; > > + /* Store in buffer length if necessary */ > + if (vq->buflen) > + vq->buflen[id] = buflen; > + > vq->num_added += 1; > > pr_debug("Added buffer head %i to %p\n", head, vq); > @@ -1179,6 +1203,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq, > __le16 head_flags, flags; > u16 head, id, prev, curr, avail_used_flags; > int err; > + u32 buflen = 0; > > START_USE(vq); > > @@ -1258,6 +1283,8 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq, > 1 << VRING_PACKED_DESC_F_AVAIL | > 1 << VRING_PACKED_DESC_F_USED; > } > + if (n >= out_sgs) > + buflen += sg->length; > } > } > > @@ -1277,6 +1304,10 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq, > vq->packed.desc_state[id].indir_desc = ctx; > vq->packed.desc_state[id].last = prev; > > + /* Store in buffer length if necessary */ > + if (vq->buflen) > + vq->buflen[id] = buflen; > + > /* > * A driver MUST NOT make the first descriptor in the list > * available before all subsequent descriptors comprising > @@ -1463,6 +1494,11 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq, > BAD_RING(vq, "id %u is not a head!\n", id); > return NULL; > } > + if (vq->buflen && unlikely(*len > vq->buflen[id])) { > + BAD_RING(vq, "used len %d is larger than in buflen %u\n", > + *len, vq->buflen[id]); > + return NULL; > + } > > /* detach_buf_packed clears data, so grab it now. */ > ret = vq->packed.desc_state[id].data; > @@ -1668,6 +1704,7 @@ static struct virtqueue *vring_create_virtqueue_packed( > struct vring_virtqueue *vq; > struct vring_packed_desc *ring; > struct vring_packed_desc_event *driver, *device; > + struct virtio_driver *drv = drv_to_virtio(vdev->dev.driver); > dma_addr_t ring_dma_addr, driver_event_dma_addr, device_event_dma_addr; > size_t ring_size_in_bytes, event_size_in_bytes; > > @@ -1757,6 +1794,15 @@ static struct virtqueue *vring_create_virtqueue_packed( > if (!vq->packed.desc_extra) > goto err_desc_extra; > > + if (!drv->suppress_used_validation || force_used_validation) { > + vq->buflen = kmalloc_array(num, sizeof(*vq->buflen), > + GFP_KERNEL); > + if (!vq->buflen) > + goto err_buflen;The reason for not using "struct vring_desc_state_split/ struct vring_desc_state_split" is to occupy less memory when turning off this function? Thanks.> + } else { > + vq->buflen = NULL; > + } > + > /* No callback? Tell other side not to bother us. */ > if (!callback) { > vq->packed.event_flags_shadow = VRING_PACKED_EVENT_FLAG_DISABLE; > @@ -1769,6 +1815,8 @@ static struct virtqueue *vring_create_virtqueue_packed( > spin_unlock(&vdev->vqs_list_lock); > return &vq->vq; > > +err_buflen: > + kfree(vq->packed.desc_extra); > err_desc_extra: > kfree(vq->packed.desc_state); > err_desc_state: > @@ -2176,6 +2224,7 @@ struct virtqueue *__vring_new_virtqueue(unsigned int index, > void (*callback)(struct virtqueue *), > const char *name) > { > + struct virtio_driver *drv = drv_to_virtio(vdev->dev.driver); > struct vring_virtqueue *vq; > > if (virtio_has_feature(vdev, VIRTIO_F_RING_PACKED)) > @@ -2235,6 +2284,15 @@ struct virtqueue *__vring_new_virtqueue(unsigned int index, > if (!vq->split.desc_extra) > goto err_extra; > > + if (!drv->suppress_used_validation || force_used_validation) { > + vq->buflen = kmalloc_array(vring.num, sizeof(*vq->buflen), > + GFP_KERNEL); > + if (!vq->buflen) > + goto err_buflen; > + } else { > + vq->buflen = NULL; > + } > + > /* Put everything in free lists. */ > vq->free_head = 0; > memset(vq->split.desc_state, 0, vring.num * > @@ -2245,6 +2303,8 @@ struct virtqueue *__vring_new_virtqueue(unsigned int index, > spin_unlock(&vdev->vqs_list_lock); > return &vq->vq; > > +err_buflen: > + kfree(vq->split.desc_extra); > err_extra: > kfree(vq->split.desc_state); > err_state: > diff --git a/include/linux/virtio.h b/include/linux/virtio.h > index 41edbc01ffa4..44d0e09da2d9 100644 > --- a/include/linux/virtio.h > +++ b/include/linux/virtio.h > @@ -152,6 +152,7 @@ size_t virtio_max_dma_size(struct virtio_device *vdev); > * @feature_table_size: number of entries in the feature table array. > * @feature_table_legacy: same as feature_table but when working in legacy mode. > * @feature_table_size_legacy: number of entries in feature table legacy array. > + * @suppress_used_validation: set to not have core validate used length > * @probe: the function to call when a device is found. Returns 0 or -errno. > * @scan: optional function to call after successful probe; intended > * for virtio-scsi to invoke a scan. > @@ -168,6 +169,7 @@ struct virtio_driver { > unsigned int feature_table_size; > const unsigned int *feature_table_legacy; > unsigned int feature_table_size_legacy; > + bool suppress_used_validation; > int (*validate)(struct virtio_device *dev); > int (*probe)(struct virtio_device *dev); > void (*scan)(struct virtio_device *dev); > -- > 2.25.1 > > _______________________________________________ > Virtualization mailing list > Virtualization at lists.linux-foundation.org > https://lists.linuxfoundation.org/mailman/listinfo/virtualization
On Wed, 27 Oct 2021 10:21:04 +0800 Jason Wang <jasowang at redhat.com> wrote:> This patch validate the used buffer length provided by the device > before trying to use it. This is done by record the in buffer length > in a new field in desc_state structure during virtqueue_add(), then we > can fail the virtqueue_get_buf() when we find the device is trying to > give us a used buffer length which is greater than the in buffer > length. > > Since some drivers have already done the validation by themselves, > this patch tries to makes the core validation optional. For the driver > that doesn't want the validation, it can set the > suppress_used_validation to be true (which could be overridden by > force_used_validation module parameter). To be more efficient, a > dedicate array is used for storing the validate used length, this > helps to eliminate the cache stress if validation is done by the > driver. > > Signed-off-by: Jason Wang <jasowang at redhat.com>Hi Jason! Our CI has detected, that virtio-vsock became unusable with this patch on s390x. I didn't test on x86 yet. The guest kernel says something like: vmw_vsock_virtio_transport virtio1: tx: used len 44 is larger than in buflen 0 Did you, or anybody else, see something like this on platforms other that s390x? I had a quick look at this code, and I speculate that it probably uncovers a pre-existig bug, rather than introducing a new one. If somebody is already working on this please reach out to me. Regards, Halil