2022-06-14 05:46:54

by 黄杰

Subject: [PATCH] virtio_ring : keep used_wrap_counter in vq->last_used_idx

From: "huangjie.albert" <[email protected]>

the used_wrap_counter and the vq->last_used_idx may get
out of sync if they are separate assignment，and interrupt
might use an incorrect value to check for the used index.

for example:OOB access
ksoftirqd may consume the packet and it will call:
virtnet_poll
-->virtnet_receive
-->virtqueue_get_buf_ctx
-->virtqueue_get_buf_ctx_packed
and in virtqueue_get_buf_ctx_packed:

vq->last_used_idx += vq->packed.desc_state[id].num;
if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
vq->last_used_idx -= vq->packed.vring.num;
vq->packed.used_wrap_counter ^= 1;
}

if at the same time, there comes a vring interrupt，in vring_interrupt:
we will call:
vring_interrupt
-->more_used
-->more_used_packed
-->is_used_desc_packed
in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
so this could case a memory out of bounds bug.

this patch is to keep the used_wrap_counter in vq->last_used_idx
so we can get the correct value to check for used index in interrupt.

Signed-off-by: huangjie.albert <[email protected]>
---
drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++--------------
include/uapi/linux/virtio_ring.h | 6 ++++
2 files changed, 40 insertions(+), 26 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 13a7348cedff..35c3750e89e1 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -111,7 +111,12 @@ struct vring_virtqueue {
/* Number we've added since last sync. */
unsigned int num_added;

- /* Last used index we've seen. */
+ /* Last used index we've seen.
+ * for split ring, it just contains last used index
+ * for packed ring, it not only contains last used index, but also
+ * used_wrap_counter, the VRING_PACKED_USED_INDEX_F_WRAP_CTR is
+ * the bit shift in last_used_idx
+ */
u16 last_used_idx;

/* Hint for event idx: already triggered no need to disable. */
@@ -154,9 +159,6 @@ struct vring_virtqueue {
/* Driver ring wrap counter. */
bool avail_wrap_counter;

- /* Device ring wrap counter. */
- bool used_wrap_counter;
-
/* Avail used flags. */
u16 avail_used_flags;

@@ -1397,6 +1399,9 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
bool avail, used;
u16 flags;

+ if (idx >= vq->packed.vring.num)
+ return false;
+
flags = le16_to_cpu(vq->packed.vring.desc[idx].flags);
avail = !!(flags & (1 << VRING_PACKED_DESC_F_AVAIL));
used = !!(flags & (1 << VRING_PACKED_DESC_F_USED));
@@ -1406,8 +1411,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,

static inline bool more_used_packed(const struct vring_virtqueue *vq)
{
- return is_used_desc_packed(vq, vq->last_used_idx,
- vq->packed.used_wrap_counter);
+ u16 last_used;
+ bool used_wrap_counter;
+
+ last_used = vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
+ used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
+ return is_used_desc_packed(vq, last_used, used_wrap_counter);
}

static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
@@ -1416,6 +1425,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
{
struct vring_virtqueue *vq = to_vvq(_vq);
u16 last_used, id;
+ bool used_wrap_counter;
void *ret;

START_USE(vq);
@@ -1434,7 +1444,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
/* Only get used elements after they have been exposed by host. */
virtio_rmb(vq->weak_barriers);

- last_used = vq->last_used_idx;
+ used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR));
+ last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
*len = le32_to_cpu(vq->packed.vring.desc[last_used].len);

@@ -1451,12 +1462,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
ret = vq->packed.desc_state[id].data;
detach_buf_packed(vq, id, ctx);

- vq->last_used_idx += vq->packed.desc_state[id].num;
- if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
- vq->last_used_idx -= vq->packed.vring.num;
- vq->packed.used_wrap_counter ^= 1;
+ last_used += vq->packed.desc_state[id].num;
+ if (unlikely(last_used >= vq->packed.vring.num)) {
+ last_used -= vq->packed.vring.num;
+ used_wrap_counter ^= 1;
}

+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
+ vq->last_used_idx = last_used;
+
/*
* If we expect an interrupt for the next entry, tell host
* by writing event index and flush out the write before
@@ -1465,9 +1479,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
virtio_store_mb(vq->weak_barriers,
&vq->packed.vring.driver->off_wrap,
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR)));
+ cpu_to_le16(vq->last_used_idx));

LAST_ADD_TIME_INVALID(vq);

@@ -1499,9 +1511,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)

if (vq->event) {
vq->packed.vring.driver->off_wrap =
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR));
+ cpu_to_le16(vq->last_used_idx);
/*
* We need to update event offset and event wrap
* counter first before updating event flags.
@@ -1518,8 +1528,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
}

END_USE(vq);
- return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR);
+ return vq->last_used_idx;
}

static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
@@ -1550,9 +1559,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
if (vq->event) {
/* TODO: tune this threshold */
bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
- wrap_counter = vq->packed.used_wrap_counter;
+ wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);

- used_idx = vq->last_used_idx + bufs;
+ used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR)) + bufs;
if (used_idx >= vq->packed.vring.num) {
used_idx -= vq->packed.vring.num;
wrap_counter ^= 1;
@@ -1582,9 +1591,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
*/
virtio_mb(vq->weak_barriers);

- if (is_used_desc_packed(vq,
- vq->last_used_idx,
- vq->packed.used_wrap_counter)) {
+ wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
+ used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
+ if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
END_USE(vq);
return false;
}
@@ -1689,7 +1698,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
vq->notify = notify;
vq->weak_barriers = weak_barriers;
vq->broken = true;
- vq->last_used_idx = 0;
+ vq->last_used_idx = 0 | (1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
vq->event_triggered = false;
vq->num_added = 0;
vq->packed_ring = true;
@@ -1720,7 +1729,6 @@ static struct virtqueue *vring_create_virtqueue_packed(

vq->packed.next_avail_idx = 0;
vq->packed.avail_wrap_counter = 1;
- vq->packed.used_wrap_counter = 1;
vq->packed.event_flags_shadow = 0;
vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;

diff --git a/include/uapi/linux/virtio_ring.h b/include/uapi/linux/virtio_ring.h
index 476d3e5c0fe7..96bcc4d52fce 100644
--- a/include/uapi/linux/virtio_ring.h
+++ b/include/uapi/linux/virtio_ring.h
@@ -77,6 +77,12 @@
*/
#define VRING_PACKED_EVENT_F_WRAP_CTR 15

+
+/*
+ * used wrap Counter bit shift in vq->last_used_idx for packed ring
+ */
+#define VRING_PACKED_USED_INDEX_F_WRAP_CTR 15
+
/* We support indirect buffer descriptors */
#define VIRTIO_RING_F_INDIRECT_DESC 28

--
2.31.1

2022-06-14 08:33:27

by Jason Wang

Subject: Re: [PATCH] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Tue, Jun 14, 2022 at 4:17 PM 黄杰 <[email protected]> wrote:
>
> Jason Wang <[email protected]> 于2022年6月14日周二 15:45写道：
> >
> > On Tue, Jun 14, 2022 at 1:38 PM Albert Huang
> > <[email protected]> wrote:
> > >
> > > From: "huangjie.albert" <[email protected]>
> > >
> > > the used_wrap_counter and the vq->last_used_idx may get
> > > out of sync if they are separate assignment，and interrupt
> > > might use an incorrect value to check for the used index.
> > >
> > > for example:OOB access
> > > ksoftirqd may consume the packet and it will call:
> > > virtnet_poll
> > > -->virtnet_receive
> > > -->virtqueue_get_buf_ctx
> > > -->virtqueue_get_buf_ctx_packed
> > > and in virtqueue_get_buf_ctx_packed:
> > >
> > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > vq->last_used_idx -= vq->packed.vring.num;
> > > vq->packed.used_wrap_counter ^= 1;
> > > }
> > >
> > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > we will call:
> > > vring_interrupt
> > > -->more_used
> > > -->more_used_packed
> > > -->is_used_desc_packed
> > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > so this could case a memory out of bounds bug.
> > >
> > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > so we can get the correct value to check for used index in interrupt.
> > >
> > > Signed-off-by: huangjie.albert <[email protected]>
> > > ---
> > > drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++--------------
> > > include/uapi/linux/virtio_ring.h | 6 ++++
> > > 2 files changed, 40 insertions(+), 26 deletions(-)
> > >
> > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > index 13a7348cedff..35c3750e89e1 100644
> > > --- a/drivers/virtio/virtio_ring.c
> > > +++ b/drivers/virtio/virtio_ring.c
> > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > /* Number we've added since last sync. */
> > > unsigned int num_added;
> > >
> > > - /* Last used index we've seen. */
> > > + /* Last used index we've seen.
> > > + * for split ring, it just contains last used index
> > > + * for packed ring, it not only contains last used index, but also
> > > + * used_wrap_counter, the VRING_PACKED_USED_INDEX_F_WRAP_CTR is
> > > + * the bit shift in last_used_idx
> > > + */
> > > u16 last_used_idx;
> > >
> > > /* Hint for event idx: already triggered no need to disable. */
> > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > /* Driver ring wrap counter. */
> > > bool avail_wrap_counter;
> > >
> > > - /* Device ring wrap counter. */
> > > - bool used_wrap_counter;
> > > -
> > > /* Avail used flags. */
> > > u16 avail_used_flags;
> > >
> > > @@ -1397,6 +1399,9 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > bool avail, used;
> > > u16 flags;
> > >
> > > + if (idx >= vq->packed.vring.num)
> > > + return false;
> >
> > I wonder if we can avoid this trick, more below.
>
> Yes, the idx pass to the is_used_desc_packed can never >=
> vq->packed.vring.num since we
> tweak the math here :
> >+ last_used += vq->packed.desc_state[id].num;
> >+ if (unlikely(last_used >= vq->packed.vring.num)) {
> >+ last_used -= vq->packed.vring.num;
> >+ used_wrap_counter ^= 1;
> > }
> >+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> >+ vq->last_used_idx = last_used;
>
> >
> > > +
> > > flags = le16_to_cpu(vq->packed.vring.desc[idx].flags);
> > > avail = !!(flags & (1 << VRING_PACKED_DESC_F_AVAIL));
> > > used = !!(flags & (1 << VRING_PACKED_DESC_F_USED));
> > > @@ -1406,8 +1411,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > >
> > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > {
> > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > - vq->packed.used_wrap_counter);
> > > + u16 last_used;
> > > + bool used_wrap_counter;
> > > +
> > > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > }
> > >
> > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > @@ -1416,6 +1425,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > {
> > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > u16 last_used, id;
> > > + bool used_wrap_counter;
> > > void *ret;
> > >
> > > START_USE(vq);
> > > @@ -1434,7 +1444,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > /* Only get used elements after they have been exposed by host. */
> > > virtio_rmb(vq->weak_barriers);
> > >
> > > - last_used = vq->last_used_idx;
> > > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > >
> > > @@ -1451,12 +1462,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > ret = vq->packed.desc_state[id].data;
> > > detach_buf_packed(vq, id, ctx);
> > >
> > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > - vq->last_used_idx -= vq->packed.vring.num;
> > > - vq->packed.used_wrap_counter ^= 1;
> > > + last_used += vq->packed.desc_state[id].num;
> > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> >
> > Can we tweak the math here to avoid the out of bound result?
>
> Yes, this can avoid the out of bound result. because of We just
> assign last_used_idx in the following code：
> vq->last_used_idx = last_used;
> and it is a 16bit/32bit operations,and are guaranteed to be atomic
>
> >
> > > + last_used -= vq->packed.vring.num;
> > > + used_wrap_counter ^= 1;
> > > }
> > >
> > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> > > + vq->last_used_idx = last_used;
> > > +
> > > /*
> > > * If we expect an interrupt for the next entry, tell host
> > > * by writing event index and flush out the write before
> > > @@ -1465,9 +1479,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > virtio_store_mb(vq->weak_barriers,
> > > &vq->packed.vring.driver->off_wrap,
> > > - cpu_to_le16(vq->last_used_idx |
> > > - (vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > + cpu_to_le16(vq->last_used_idx));
> > >
> > > LAST_ADD_TIME_INVALID(vq);
> > >
> > > @@ -1499,9 +1511,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > >
> > > if (vq->event) {
> > > vq->packed.vring.driver->off_wrap =
> > > - cpu_to_le16(vq->last_used_idx |
> > > - (vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + cpu_to_le16(vq->last_used_idx);
> > > /*
> > > * We need to update event offset and event wrap
> > > * counter first before updating event flags.
> > > @@ -1518,8 +1528,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > }
> > >
> > > END_USE(vq);
> > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > + return vq->last_used_idx;
> > > }
> > >
> > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > @@ -1550,9 +1559,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > if (vq->event) {
> > > /* TODO: tune this threshold */
> > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > - wrap_counter = vq->packed.used_wrap_counter;
> > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > >
> > > - used_idx = vq->last_used_idx + bufs;
> > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR)) + bufs;
> > > if (used_idx >= vq->packed.vring.num) {
> > > used_idx -= vq->packed.vring.num;
> > > wrap_counter ^= 1;
> > > @@ -1582,9 +1591,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > */
> > > virtio_mb(vq->weak_barriers);
> > >
> > > - if (is_used_desc_packed(vq,
> > > - vq->last_used_idx,
> > > - vq->packed.used_wrap_counter)) {
> > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > END_USE(vq);
> > > return false;
> > > }
> > > @@ -1689,7 +1698,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > vq->notify = notify;
> > > vq->weak_barriers = weak_barriers;
> > > vq->broken = true;
> > > - vq->last_used_idx = 0;
> > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > vq->event_triggered = false;
> > > vq->num_added = 0;
> > > vq->packed_ring = true;
> > > @@ -1720,7 +1729,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > >
> > > vq->packed.next_avail_idx = 0;
> > > vq->packed.avail_wrap_counter = 1;
> > > - vq->packed.used_wrap_counter = 1;
> > > vq->packed.event_flags_shadow = 0;
> > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > >
> > > diff --git a/include/uapi/linux/virtio_ring.h b/include/uapi/linux/virtio_ring.h
> > > index 476d3e5c0fe7..96bcc4d52fce 100644
> > > --- a/include/uapi/linux/virtio_ring.h
> > > +++ b/include/uapi/linux/virtio_ring.h
> > > @@ -77,6 +77,12 @@
> > > */
> > > #define VRING_PACKED_EVENT_F_WRAP_CTR 15
> > >
> > > +
> > > +/*
> > > + * used wrap Counter bit shift in vq->last_used_idx for packed ring
> > > + */
> > > +#define VRING_PACKED_USED_INDEX_F_WRAP_CTR 15
> >
> > Let's simply reuse VRING_PACKED_EVENT_F_WRAP_CTR ?
> >
> > Thanks
>
> Although the two values are the same, I see the meanings are different,
> so I redefine one. Is it possible to rename
> VRING_PACKED_EVENT_F_WRAP_CTR to VRING_PACKED_F_WRAP_CTR?
>
> Jason, what is your take?

That looks even worse. So I'm fine to define a new macro as 15.

Thanks

>
> >
> > > +
> > > /* We support indirect buffer descriptors */
> > > #define VIRTIO_RING_F_INDIRECT_DESC 28
> > >
> > > --
> > > 2.31.1
> > >
> >
>

2022-06-14 08:37:15

by Jason Wang

Subject: Re: [PATCH] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Tue, Jun 14, 2022 at 1:38 PM Albert Huang
<[email protected]> wrote:
>
> From: "huangjie.albert" <[email protected]>
>
> the used_wrap_counter and the vq->last_used_idx may get
> out of sync if they are separate assignment，and interrupt
> might use an incorrect value to check for the used index.
>
> for example:OOB access
> ksoftirqd may consume the packet and it will call:
> virtnet_poll
> -->virtnet_receive
> -->virtqueue_get_buf_ctx
> -->virtqueue_get_buf_ctx_packed
> and in virtqueue_get_buf_ctx_packed:
>
> vq->last_used_idx += vq->packed.desc_state[id].num;
> if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> vq->last_used_idx -= vq->packed.vring.num;
> vq->packed.used_wrap_counter ^= 1;
> }
>
> if at the same time, there comes a vring interrupt，in vring_interrupt:
> we will call:
> vring_interrupt
> -->more_used
> -->more_used_packed
> -->is_used_desc_packed
> in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> so this could case a memory out of bounds bug.
>
> this patch is to keep the used_wrap_counter in vq->last_used_idx
> so we can get the correct value to check for used index in interrupt.
>
> Signed-off-by: huangjie.albert <[email protected]>
> ---
> drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++--------------
> include/uapi/linux/virtio_ring.h | 6 ++++
> 2 files changed, 40 insertions(+), 26 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 13a7348cedff..35c3750e89e1 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -111,7 +111,12 @@ struct vring_virtqueue {
> /* Number we've added since last sync. */
> unsigned int num_added;
>
> - /* Last used index we've seen. */
> + /* Last used index we've seen.
> + * for split ring, it just contains last used index
> + * for packed ring, it not only contains last used index, but also
> + * used_wrap_counter, the VRING_PACKED_USED_INDEX_F_WRAP_CTR is
> + * the bit shift in last_used_idx
> + */
> u16 last_used_idx;
>
> /* Hint for event idx: already triggered no need to disable. */
> @@ -154,9 +159,6 @@ struct vring_virtqueue {
> /* Driver ring wrap counter. */
> bool avail_wrap_counter;
>
> - /* Device ring wrap counter. */
> - bool used_wrap_counter;
> -
> /* Avail used flags. */
> u16 avail_used_flags;
>
> @@ -1397,6 +1399,9 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> bool avail, used;
> u16 flags;
>
> + if (idx >= vq->packed.vring.num)
> + return false;

I wonder if we can avoid this trick, more below.

> +
> flags = le16_to_cpu(vq->packed.vring.desc[idx].flags);
> avail = !!(flags & (1 << VRING_PACKED_DESC_F_AVAIL));
> used = !!(flags & (1 << VRING_PACKED_DESC_F_USED));
> @@ -1406,8 +1411,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
>
> static inline bool more_used_packed(const struct vring_virtqueue *vq)
> {
> - return is_used_desc_packed(vq, vq->last_used_idx,
> - vq->packed.used_wrap_counter);
> + u16 last_used;
> + bool used_wrap_counter;
> +
> + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> }
>
> static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> @@ -1416,6 +1425,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> u16 last_used, id;
> + bool used_wrap_counter;
> void *ret;
>
> START_USE(vq);
> @@ -1434,7 +1444,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> /* Only get used elements after they have been exposed by host. */
> virtio_rmb(vq->weak_barriers);
>
> - last_used = vq->last_used_idx;
> + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
>
> @@ -1451,12 +1462,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> ret = vq->packed.desc_state[id].data;
> detach_buf_packed(vq, id, ctx);
>
> - vq->last_used_idx += vq->packed.desc_state[id].num;
> - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> - vq->last_used_idx -= vq->packed.vring.num;
> - vq->packed.used_wrap_counter ^= 1;
> + last_used += vq->packed.desc_state[id].num;
> + if (unlikely(last_used >= vq->packed.vring.num)) {

Can we tweak the math here to avoid the out of bound result?

> + last_used -= vq->packed.vring.num;
> + used_wrap_counter ^= 1;
> }
>
> + last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> + vq->last_used_idx = last_used;
> +
> /*
> * If we expect an interrupt for the next entry, tell host
> * by writing event index and flush out the write before
> @@ -1465,9 +1479,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> virtio_store_mb(vq->weak_barriers,
> &vq->packed.vring.driver->off_wrap,
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR)));
> + cpu_to_le16(vq->last_used_idx));
>
> LAST_ADD_TIME_INVALID(vq);
>
> @@ -1499,9 +1511,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
>
> if (vq->event) {
> vq->packed.vring.driver->off_wrap =
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR));
> + cpu_to_le16(vq->last_used_idx);
> /*
> * We need to update event offset and event wrap
> * counter first before updating event flags.
> @@ -1518,8 +1528,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> }
>
> END_USE(vq);
> - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR);
> + return vq->last_used_idx;
> }
>
> static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> @@ -1550,9 +1559,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> if (vq->event) {
> /* TODO: tune this threshold */
> bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> - wrap_counter = vq->packed.used_wrap_counter;
> + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
>
> - used_idx = vq->last_used_idx + bufs;
> + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR)) + bufs;
> if (used_idx >= vq->packed.vring.num) {
> used_idx -= vq->packed.vring.num;
> wrap_counter ^= 1;
> @@ -1582,9 +1591,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> */
> virtio_mb(vq->weak_barriers);
>
> - if (is_used_desc_packed(vq,
> - vq->last_used_idx,
> - vq->packed.used_wrap_counter)) {
> + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> END_USE(vq);
> return false;
> }
> @@ -1689,7 +1698,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> vq->notify = notify;
> vq->weak_barriers = weak_barriers;
> vq->broken = true;
> - vq->last_used_idx = 0;
> + vq->last_used_idx = 0 | (1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> vq->event_triggered = false;
> vq->num_added = 0;
> vq->packed_ring = true;
> @@ -1720,7 +1729,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
>
> vq->packed.next_avail_idx = 0;
> vq->packed.avail_wrap_counter = 1;
> - vq->packed.used_wrap_counter = 1;
> vq->packed.event_flags_shadow = 0;
> vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
>
> diff --git a/include/uapi/linux/virtio_ring.h b/include/uapi/linux/virtio_ring.h
> index 476d3e5c0fe7..96bcc4d52fce 100644
> --- a/include/uapi/linux/virtio_ring.h
> +++ b/include/uapi/linux/virtio_ring.h
> @@ -77,6 +77,12 @@
> */
> #define VRING_PACKED_EVENT_F_WRAP_CTR 15
>
> +
> +/*
> + * used wrap Counter bit shift in vq->last_used_idx for packed ring
> + */
> +#define VRING_PACKED_USED_INDEX_F_WRAP_CTR 15

Let's simply reuse VRING_PACKED_EVENT_F_WRAP_CTR ?

Thanks

> +
> /* We support indirect buffer descriptors */
> #define VIRTIO_RING_F_INDIRECT_DESC 28
>
> --
> 2.31.1
>

2022-06-14 08:39:47

by 黄杰

Subject: Re: [PATCH] virtio_ring : keep used_wrap_counter in vq->last_used_idx

Jason Wang <[email protected]> 于2022年6月14日周二 15:45写道：
>
> On Tue, Jun 14, 2022 at 1:38 PM Albert Huang
> <[email protected]> wrote:
> >
> > From: "huangjie.albert" <[email protected]>
> >
> > the used_wrap_counter and the vq->last_used_idx may get
> > out of sync if they are separate assignment，and interrupt
> > might use an incorrect value to check for the used index.
> >
> > for example:OOB access
> > ksoftirqd may consume the packet and it will call:
> > virtnet_poll
> > -->virtnet_receive
> > -->virtqueue_get_buf_ctx
> > -->virtqueue_get_buf_ctx_packed
> > and in virtqueue_get_buf_ctx_packed:
> >
> > vq->last_used_idx += vq->packed.desc_state[id].num;
> > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > vq->last_used_idx -= vq->packed.vring.num;
> > vq->packed.used_wrap_counter ^= 1;
> > }
> >
> > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > we will call:
> > vring_interrupt
> > -->more_used
> > -->more_used_packed
> > -->is_used_desc_packed
> > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > so this could case a memory out of bounds bug.
> >
> > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > so we can get the correct value to check for used index in interrupt.
> >
> > Signed-off-by: huangjie.albert <[email protected]>
> > ---
> > drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++--------------
> > include/uapi/linux/virtio_ring.h | 6 ++++
> > 2 files changed, 40 insertions(+), 26 deletions(-)
> >
> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > index 13a7348cedff..35c3750e89e1 100644
> > --- a/drivers/virtio/virtio_ring.c
> > +++ b/drivers/virtio/virtio_ring.c
> > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > /* Number we've added since last sync. */
> > unsigned int num_added;
> >
> > - /* Last used index we've seen. */
> > + /* Last used index we've seen.
> > + * for split ring, it just contains last used index
> > + * for packed ring, it not only contains last used index, but also
> > + * used_wrap_counter, the VRING_PACKED_USED_INDEX_F_WRAP_CTR is
> > + * the bit shift in last_used_idx
> > + */
> > u16 last_used_idx;
> >
> > /* Hint for event idx: already triggered no need to disable. */
> > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > /* Driver ring wrap counter. */
> > bool avail_wrap_counter;
> >
> > - /* Device ring wrap counter. */
> > - bool used_wrap_counter;
> > -
> > /* Avail used flags. */
> > u16 avail_used_flags;
> >
> > @@ -1397,6 +1399,9 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > bool avail, used;
> > u16 flags;
> >
> > + if (idx >= vq->packed.vring.num)
> > + return false;
>
> I wonder if we can avoid this trick, more below.

Yes, the idx pass to the is_used_desc_packed can never >=
vq->packed.vring.num since we
tweak the math here :
>+ last_used += vq->packed.desc_state[id].num;
>+ if (unlikely(last_used >= vq->packed.vring.num)) {
>+ last_used -= vq->packed.vring.num;
>+ used_wrap_counter ^= 1;
> }
>+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
>+ vq->last_used_idx = last_used;

>
> > +
> > flags = le16_to_cpu(vq->packed.vring.desc[idx].flags);
> > avail = !!(flags & (1 << VRING_PACKED_DESC_F_AVAIL));
> > used = !!(flags & (1 << VRING_PACKED_DESC_F_USED));
> > @@ -1406,8 +1411,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> >
> > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > {
> > - return is_used_desc_packed(vq, vq->last_used_idx,
> > - vq->packed.used_wrap_counter);
> > + u16 last_used;
> > + bool used_wrap_counter;
> > +
> > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > }
> >
> > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > @@ -1416,6 +1425,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > {
> > struct vring_virtqueue *vq = to_vvq(_vq);
> > u16 last_used, id;
> > + bool used_wrap_counter;
> > void *ret;
> >
> > START_USE(vq);
> > @@ -1434,7 +1444,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > /* Only get used elements after they have been exposed by host. */
> > virtio_rmb(vq->weak_barriers);
> >
> > - last_used = vq->last_used_idx;
> > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> >
> > @@ -1451,12 +1462,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > ret = vq->packed.desc_state[id].data;
> > detach_buf_packed(vq, id, ctx);
> >
> > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > - vq->last_used_idx -= vq->packed.vring.num;
> > - vq->packed.used_wrap_counter ^= 1;
> > + last_used += vq->packed.desc_state[id].num;
> > + if (unlikely(last_used >= vq->packed.vring.num)) {
>
> Can we tweak the math here to avoid the out of bound result?

Yes, this can avoid the out of bound result. because of We just
assign last_used_idx in the following code：
vq->last_used_idx = last_used;
and it is a 16bit/32bit operations,and are guaranteed to be atomic

>
> > + last_used -= vq->packed.vring.num;
> > + used_wrap_counter ^= 1;
> > }
> >
> > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> > + vq->last_used_idx = last_used;
> > +
> > /*
> > * If we expect an interrupt for the next entry, tell host
> > * by writing event index and flush out the write before
> > @@ -1465,9 +1479,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > virtio_store_mb(vq->weak_barriers,
> > &vq->packed.vring.driver->off_wrap,
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > + cpu_to_le16(vq->last_used_idx));
> >
> > LAST_ADD_TIME_INVALID(vq);
> >
> > @@ -1499,9 +1511,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> >
> > if (vq->event) {
> > vq->packed.vring.driver->off_wrap =
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > + cpu_to_le16(vq->last_used_idx);
> > /*
> > * We need to update event offset and event wrap
> > * counter first before updating event flags.
> > @@ -1518,8 +1528,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > }
> >
> > END_USE(vq);
> > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > + return vq->last_used_idx;
> > }
> >
> > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > @@ -1550,9 +1559,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > if (vq->event) {
> > /* TODO: tune this threshold */
> > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > - wrap_counter = vq->packed.used_wrap_counter;
> > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> >
> > - used_idx = vq->last_used_idx + bufs;
> > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR)) + bufs;
> > if (used_idx >= vq->packed.vring.num) {
> > used_idx -= vq->packed.vring.num;
> > wrap_counter ^= 1;
> > @@ -1582,9 +1591,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > */
> > virtio_mb(vq->weak_barriers);
> >
> > - if (is_used_desc_packed(vq,
> > - vq->last_used_idx,
> > - vq->packed.used_wrap_counter)) {
> > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > END_USE(vq);
> > return false;
> > }
> > @@ -1689,7 +1698,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > vq->notify = notify;
> > vq->weak_barriers = weak_barriers;
> > vq->broken = true;
> > - vq->last_used_idx = 0;
> > + vq->last_used_idx = 0 | (1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > vq->event_triggered = false;
> > vq->num_added = 0;
> > vq->packed_ring = true;
> > @@ -1720,7 +1729,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> >
> > vq->packed.next_avail_idx = 0;
> > vq->packed.avail_wrap_counter = 1;
> > - vq->packed.used_wrap_counter = 1;
> > vq->packed.event_flags_shadow = 0;
> > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> >
> > diff --git a/include/uapi/linux/virtio_ring.h b/include/uapi/linux/virtio_ring.h
> > index 476d3e5c0fe7..96bcc4d52fce 100644
> > --- a/include/uapi/linux/virtio_ring.h
> > +++ b/include/uapi/linux/virtio_ring.h
> > @@ -77,6 +77,12 @@
> > */
> > #define VRING_PACKED_EVENT_F_WRAP_CTR 15
> >
> > +
> > +/*
> > + * used wrap Counter bit shift in vq->last_used_idx for packed ring
> > + */
> > +#define VRING_PACKED_USED_INDEX_F_WRAP_CTR 15
>
> Let's simply reuse VRING_PACKED_EVENT_F_WRAP_CTR ?
>
> Thanks

Although the two values are the same, I see the meanings are different,
so I redefine one. Is it possible to rename
VRING_PACKED_EVENT_F_WRAP_CTR to VRING_PACKED_F_WRAP_CTR?

Jason, what is your take?

>
> > +
> > /* We support indirect buffer descriptors */
> > #define VIRTIO_RING_F_INDIRECT_DESC 28
> >
> > --
> > 2.31.1
> >
>

2022-06-15 03:47:37

by 黄杰

Subject: Re: [PATCH] virtio_ring : keep used_wrap_counter in vq->last_used_idx

Jason Wang <[email protected]> 于2022年6月14日周二 16:21写道：
>
> On Tue, Jun 14, 2022 at 4:17 PM 黄杰 <[email protected]> wrote:
> >
> > Jason Wang <[email protected]> 于2022年6月14日周二 15:45写道：
> > >
> > > On Tue, Jun 14, 2022 at 1:38 PM Albert Huang
> > > <[email protected]> wrote:
> > > >
> > > > From: "huangjie.albert" <[email protected]>
> > > >
> > > > the used_wrap_counter and the vq->last_used_idx may get
> > > > out of sync if they are separate assignment，and interrupt
> > > > might use an incorrect value to check for the used index.
> > > >
> > > > for example:OOB access
> > > > ksoftirqd may consume the packet and it will call:
> > > > virtnet_poll
> > > > -->virtnet_receive
> > > > -->virtqueue_get_buf_ctx
> > > > -->virtqueue_get_buf_ctx_packed
> > > > and in virtqueue_get_buf_ctx_packed:
> > > >
> > > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > vq->last_used_idx -= vq->packed.vring.num;
> > > > vq->packed.used_wrap_counter ^= 1;
> > > > }
> > > >
> > > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > > we will call:
> > > > vring_interrupt
> > > > -->more_used
> > > > -->more_used_packed
> > > > -->is_used_desc_packed
> > > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > > so this could case a memory out of bounds bug.
> > > >
> > > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > > so we can get the correct value to check for used index in interrupt.
> > > >
> > > > Signed-off-by: huangjie.albert <[email protected]>
> > > > ---
> > > > drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++--------------
> > > > include/uapi/linux/virtio_ring.h | 6 ++++
> > > > 2 files changed, 40 insertions(+), 26 deletions(-)
> > > >
> > > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > > index 13a7348cedff..35c3750e89e1 100644
> > > > --- a/drivers/virtio/virtio_ring.c
> > > > +++ b/drivers/virtio/virtio_ring.c
> > > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > > /* Number we've added since last sync. */
> > > > unsigned int num_added;
> > > >
> > > > - /* Last used index we've seen. */
> > > > + /* Last used index we've seen.
> > > > + * for split ring, it just contains last used index
> > > > + * for packed ring, it not only contains last used index, but also
> > > > + * used_wrap_counter, the VRING_PACKED_USED_INDEX_F_WRAP_CTR is
> > > > + * the bit shift in last_used_idx
> > > > + */
> > > > u16 last_used_idx;
> > > >
> > > > /* Hint for event idx: already triggered no need to disable. */
> > > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > > /* Driver ring wrap counter. */
> > > > bool avail_wrap_counter;
> > > >
> > > > - /* Device ring wrap counter. */
> > > > - bool used_wrap_counter;
> > > > -
> > > > /* Avail used flags. */
> > > > u16 avail_used_flags;
> > > >
> > > > @@ -1397,6 +1399,9 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > > bool avail, used;
> > > > u16 flags;
> > > >
> > > > + if (idx >= vq->packed.vring.num)
> > > > + return false;
> > >
> > > I wonder if we can avoid this trick, more below.
> >
> > Yes, the idx pass to the is_used_desc_packed can never >=
> > vq->packed.vring.num since we
> > tweak the math here :
> > >+ last_used += vq->packed.desc_state[id].num;
> > >+ if (unlikely(last_used >= vq->packed.vring.num)) {
> > >+ last_used -= vq->packed.vring.num;
> > >+ used_wrap_counter ^= 1;
> > > }
> > >+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> > >+ vq->last_used_idx = last_used;
> >
> > >
> > > > +
> > > > flags = le16_to_cpu(vq->packed.vring.desc[idx].flags);
> > > > avail = !!(flags & (1 << VRING_PACKED_DESC_F_AVAIL));
> > > > used = !!(flags & (1 << VRING_PACKED_DESC_F_USED));
> > > > @@ -1406,8 +1411,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > >
> > > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > > {
> > > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > > - vq->packed.used_wrap_counter);
> > > > + u16 last_used;
> > > > + bool used_wrap_counter;
> > > > +
> > > > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > > }
> > > >
> > > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > @@ -1416,6 +1425,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > {
> > > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > > u16 last_used, id;
> > > > + bool used_wrap_counter;
> > > > void *ret;
> > > >
> > > > START_USE(vq);
> > > > @@ -1434,7 +1444,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > /* Only get used elements after they have been exposed by host. */
> > > > virtio_rmb(vq->weak_barriers);
> > > >
> > > > - last_used = vq->last_used_idx;
> > > > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > > >
> > > > @@ -1451,12 +1462,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > ret = vq->packed.desc_state[id].data;
> > > > detach_buf_packed(vq, id, ctx);
> > > >
> > > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > - vq->last_used_idx -= vq->packed.vring.num;
> > > > - vq->packed.used_wrap_counter ^= 1;
> > > > + last_used += vq->packed.desc_state[id].num;
> > > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > >
> > > Can we tweak the math here to avoid the out of bound result?
> >
> > Yes, this can avoid the out of bound result. because of We just
> > assign last_used_idx in the following code：
> > vq->last_used_idx = last_used;
> > and it is a 16bit/32bit operations,and are guaranteed to be atomic
> >
> > >
> > > > + last_used -= vq->packed.vring.num;
> > > > + used_wrap_counter ^= 1;
> > > > }
> > > >
> > > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> > > > + vq->last_used_idx = last_used;
> > > > +
> > > > /*
> > > > * If we expect an interrupt for the next entry, tell host
> > > > * by writing event index and flush out the write before
> > > > @@ -1465,9 +1479,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > > virtio_store_mb(vq->weak_barriers,
> > > > &vq->packed.vring.driver->off_wrap,
> > > > - cpu_to_le16(vq->last_used_idx |
> > > > - (vq->packed.used_wrap_counter <<
> > > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > > + cpu_to_le16(vq->last_used_idx));
> > > >
> > > > LAST_ADD_TIME_INVALID(vq);
> > > >
> > > > @@ -1499,9 +1511,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > >
> > > > if (vq->event) {
> > > > vq->packed.vring.driver->off_wrap =
> > > > - cpu_to_le16(vq->last_used_idx |
> > > > - (vq->packed.used_wrap_counter <<
> > > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > + cpu_to_le16(vq->last_used_idx);
> > > > /*
> > > > * We need to update event offset and event wrap
> > > > * counter first before updating event flags.
> > > > @@ -1518,8 +1528,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > > }
> > > >
> > > > END_USE(vq);
> > > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > > + return vq->last_used_idx;
> > > > }
> > > >
> > > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > > @@ -1550,9 +1559,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > if (vq->event) {
> > > > /* TODO: tune this threshold */
> > > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > > - wrap_counter = vq->packed.used_wrap_counter;
> > > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > >
> > > > - used_idx = vq->last_used_idx + bufs;
> > > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR)) + bufs;
> > > > if (used_idx >= vq->packed.vring.num) {
> > > > used_idx -= vq->packed.vring.num;
> > > > wrap_counter ^= 1;
> > > > @@ -1582,9 +1591,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > */
> > > > virtio_mb(vq->weak_barriers);
> > > >
> > > > - if (is_used_desc_packed(vq,
> > > > - vq->last_used_idx,
> > > > - vq->packed.used_wrap_counter)) {
> > > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > > END_USE(vq);
> > > > return false;
> > > > }
> > > > @@ -1689,7 +1698,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > > vq->notify = notify;
> > > > vq->weak_barriers = weak_barriers;
> > > > vq->broken = true;
> > > > - vq->last_used_idx = 0;
> > > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > vq->event_triggered = false;
> > > > vq->num_added = 0;
> > > > vq->packed_ring = true;
> > > > @@ -1720,7 +1729,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > >
> > > > vq->packed.next_avail_idx = 0;
> > > > vq->packed.avail_wrap_counter = 1;
> > > > - vq->packed.used_wrap_counter = 1;
> > > > vq->packed.event_flags_shadow = 0;
> > > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > > >
> > > > diff --git a/include/uapi/linux/virtio_ring.h b/include/uapi/linux/virtio_ring.h
> > > > index 476d3e5c0fe7..96bcc4d52fce 100644
> > > > --- a/include/uapi/linux/virtio_ring.h
> > > > +++ b/include/uapi/linux/virtio_ring.h
> > > > @@ -77,6 +77,12 @@
> > > > */
> > > > #define VRING_PACKED_EVENT_F_WRAP_CTR 15
> > > >
> > > > +
> > > > +/*
> > > > + * used wrap Counter bit shift in vq->last_used_idx for packed ring
> > > > + */
> > > > +#define VRING_PACKED_USED_INDEX_F_WRAP_CTR 15
> > >
> > > Let's simply reuse VRING_PACKED_EVENT_F_WRAP_CTR ?
> > >
> > > Thanks
> >
> > Although the two values are the same, I see the meanings are different,
> > so I redefine one. Is it possible to rename
> > VRING_PACKED_EVENT_F_WRAP_CTR to VRING_PACKED_F_WRAP_CTR?
> >
> > Jason, what is your take?
>
> That looks even worse. So I'm fine to define a new macro as 15.
>
> Thanks
>
> >
> > >
> > > > +
> > > > /* We support indirect buffer descriptors */
> > > > #define VIRTIO_RING_F_INDIRECT_DESC 28
> > > >
> > > > --
> > > > 2.31.1
> > > >
> > >
> >
>
hello, Jason

Any other suggestions? I will do some stability and functional
testing before remaking a patch for patch v2.

Thanks

2022-06-15 03:50:09

by Jason Wang

Subject: Re: [PATCH] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Wed, Jun 15, 2022 at 11:25 AM 黄杰 <[email protected]> wrote:
>
> Jason Wang <[email protected]> 于2022年6月14日周二 16:21写道：
> >
> > On Tue, Jun 14, 2022 at 4:17 PM 黄杰 <[email protected]> wrote:
> > >
> > > Jason Wang <[email protected]> 于2022年6月14日周二 15:45写道：
> > > >
> > > > On Tue, Jun 14, 2022 at 1:38 PM Albert Huang
> > > > <[email protected]> wrote:
> > > > >
> > > > > From: "huangjie.albert" <[email protected]>
> > > > >
> > > > > the used_wrap_counter and the vq->last_used_idx may get
> > > > > out of sync if they are separate assignment，and interrupt
> > > > > might use an incorrect value to check for the used index.
> > > > >
> > > > > for example:OOB access
> > > > > ksoftirqd may consume the packet and it will call:
> > > > > virtnet_poll
> > > > > -->virtnet_receive
> > > > > -->virtqueue_get_buf_ctx
> > > > > -->virtqueue_get_buf_ctx_packed
> > > > > and in virtqueue_get_buf_ctx_packed:
> > > > >
> > > > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > > vq->last_used_idx -= vq->packed.vring.num;
> > > > > vq->packed.used_wrap_counter ^= 1;
> > > > > }
> > > > >
> > > > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > > > we will call:
> > > > > vring_interrupt
> > > > > -->more_used
> > > > > -->more_used_packed
> > > > > -->is_used_desc_packed
> > > > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > > > so this could case a memory out of bounds bug.
> > > > >
> > > > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > > > so we can get the correct value to check for used index in interrupt.
> > > > >
> > > > > Signed-off-by: huangjie.albert <[email protected]>
> > > > > ---
> > > > > drivers/virtio/virtio_ring.c | 60 ++++++++++++++++++--------------
> > > > > include/uapi/linux/virtio_ring.h | 6 ++++
> > > > > 2 files changed, 40 insertions(+), 26 deletions(-)
> > > > >
> > > > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > > > index 13a7348cedff..35c3750e89e1 100644
> > > > > --- a/drivers/virtio/virtio_ring.c
> > > > > +++ b/drivers/virtio/virtio_ring.c
> > > > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > > > /* Number we've added since last sync. */
> > > > > unsigned int num_added;
> > > > >
> > > > > - /* Last used index we've seen. */
> > > > > + /* Last used index we've seen.
> > > > > + * for split ring, it just contains last used index
> > > > > + * for packed ring, it not only contains last used index, but also
> > > > > + * used_wrap_counter, the VRING_PACKED_USED_INDEX_F_WRAP_CTR is
> > > > > + * the bit shift in last_used_idx
> > > > > + */
> > > > > u16 last_used_idx;
> > > > >
> > > > > /* Hint for event idx: already triggered no need to disable. */
> > > > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > > > /* Driver ring wrap counter. */
> > > > > bool avail_wrap_counter;
> > > > >
> > > > > - /* Device ring wrap counter. */
> > > > > - bool used_wrap_counter;
> > > > > -
> > > > > /* Avail used flags. */
> > > > > u16 avail_used_flags;
> > > > >
> > > > > @@ -1397,6 +1399,9 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > > > bool avail, used;
> > > > > u16 flags;
> > > > >
> > > > > + if (idx >= vq->packed.vring.num)
> > > > > + return false;
> > > >
> > > > I wonder if we can avoid this trick, more below.
> > >
> > > Yes, the idx pass to the is_used_desc_packed can never >=
> > > vq->packed.vring.num since we
> > > tweak the math here :
> > > >+ last_used += vq->packed.desc_state[id].num;
> > > >+ if (unlikely(last_used >= vq->packed.vring.num)) {
> > > >+ last_used -= vq->packed.vring.num;
> > > >+ used_wrap_counter ^= 1;
> > > > }
> > > >+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> > > >+ vq->last_used_idx = last_used;
> > >
> > > >
> > > > > +
> > > > > flags = le16_to_cpu(vq->packed.vring.desc[idx].flags);
> > > > > avail = !!(flags & (1 << VRING_PACKED_DESC_F_AVAIL));
> > > > > used = !!(flags & (1 << VRING_PACKED_DESC_F_USED));
> > > > > @@ -1406,8 +1411,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > > >
> > > > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > > > {
> > > > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > > > - vq->packed.used_wrap_counter);
> > > > > + u16 last_used;
> > > > > + bool used_wrap_counter;
> > > > > +
> > > > > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > > > }
> > > > >
> > > > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > @@ -1416,6 +1425,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > {
> > > > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > > > u16 last_used, id;
> > > > > + bool used_wrap_counter;
> > > > > void *ret;
> > > > >
> > > > > START_USE(vq);
> > > > > @@ -1434,7 +1444,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > /* Only get used elements after they have been exposed by host. */
> > > > > virtio_rmb(vq->weak_barriers);
> > > > >
> > > > > - last_used = vq->last_used_idx;
> > > > > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > > > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > > > >
> > > > > @@ -1451,12 +1462,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > ret = vq->packed.desc_state[id].data;
> > > > > detach_buf_packed(vq, id, ctx);
> > > > >
> > > > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > > - vq->last_used_idx -= vq->packed.vring.num;
> > > > > - vq->packed.used_wrap_counter ^= 1;
> > > > > + last_used += vq->packed.desc_state[id].num;
> > > > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > > >
> > > > Can we tweak the math here to avoid the out of bound result?
> > >
> > > Yes, this can avoid the out of bound result. because of We just
> > > assign last_used_idx in the following code：
> > > vq->last_used_idx = last_used;
> > > and it is a 16bit/32bit operations,and are guaranteed to be atomic
> > >
> > > >
> > > > > + last_used -= vq->packed.vring.num;
> > > > > + used_wrap_counter ^= 1;
> > > > > }
> > > > >
> > > > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_DESC_F_USED));
> > > > > + vq->last_used_idx = last_used;
> > > > > +
> > > > > /*
> > > > > * If we expect an interrupt for the next entry, tell host
> > > > > * by writing event index and flush out the write before
> > > > > @@ -1465,9 +1479,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > > > virtio_store_mb(vq->weak_barriers,
> > > > > &vq->packed.vring.driver->off_wrap,
> > > > > - cpu_to_le16(vq->last_used_idx |
> > > > > - (vq->packed.used_wrap_counter <<
> > > > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > > > + cpu_to_le16(vq->last_used_idx));
> > > > >
> > > > > LAST_ADD_TIME_INVALID(vq);
> > > > >
> > > > > @@ -1499,9 +1511,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > > >
> > > > > if (vq->event) {
> > > > > vq->packed.vring.driver->off_wrap =
> > > > > - cpu_to_le16(vq->last_used_idx |
> > > > > - (vq->packed.used_wrap_counter <<
> > > > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > > + cpu_to_le16(vq->last_used_idx);
> > > > > /*
> > > > > * We need to update event offset and event wrap
> > > > > * counter first before updating event flags.
> > > > > @@ -1518,8 +1528,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > > > }
> > > > >
> > > > > END_USE(vq);
> > > > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > > > + return vq->last_used_idx;
> > > > > }
> > > > >
> > > > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > > > @@ -1550,9 +1559,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > > if (vq->event) {
> > > > > /* TODO: tune this threshold */
> > > > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > > > - wrap_counter = vq->packed.used_wrap_counter;
> > > > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > >
> > > > > - used_idx = vq->last_used_idx + bufs;
> > > > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR)) + bufs;
> > > > > if (used_idx >= vq->packed.vring.num) {
> > > > > used_idx -= vq->packed.vring.num;
> > > > > wrap_counter ^= 1;
> > > > > @@ -1582,9 +1591,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > > */
> > > > > virtio_mb(vq->weak_barriers);
> > > > >
> > > > > - if (is_used_desc_packed(vq,
> > > > > - vq->last_used_idx,
> > > > > - vq->packed.used_wrap_counter)) {
> > > > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR));
> > > > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > > > END_USE(vq);
> > > > > return false;
> > > > > }
> > > > > @@ -1689,7 +1698,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > > > vq->notify = notify;
> > > > > vq->weak_barriers = weak_barriers;
> > > > > vq->broken = true;
> > > > > - vq->last_used_idx = 0;
> > > > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_USED_INDEX_F_WRAP_CTR);
> > > > > vq->event_triggered = false;
> > > > > vq->num_added = 0;
> > > > > vq->packed_ring = true;
> > > > > @@ -1720,7 +1729,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > > >
> > > > > vq->packed.next_avail_idx = 0;
> > > > > vq->packed.avail_wrap_counter = 1;
> > > > > - vq->packed.used_wrap_counter = 1;
> > > > > vq->packed.event_flags_shadow = 0;
> > > > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > > > >
> > > > > diff --git a/include/uapi/linux/virtio_ring.h b/include/uapi/linux/virtio_ring.h
> > > > > index 476d3e5c0fe7..96bcc4d52fce 100644
> > > > > --- a/include/uapi/linux/virtio_ring.h
> > > > > +++ b/include/uapi/linux/virtio_ring.h
> > > > > @@ -77,6 +77,12 @@
> > > > > */
> > > > > #define VRING_PACKED_EVENT_F_WRAP_CTR 15
> > > > >
> > > > > +
> > > > > +/*
> > > > > + * used wrap Counter bit shift in vq->last_used_idx for packed ring
> > > > > + */
> > > > > +#define VRING_PACKED_USED_INDEX_F_WRAP_CTR 15
> > > >
> > > > Let's simply reuse VRING_PACKED_EVENT_F_WRAP_CTR ?
> > > >
> > > > Thanks
> > >
> > > Although the two values are the same, I see the meanings are different,
> > > so I redefine one. Is it possible to rename
> > > VRING_PACKED_EVENT_F_WRAP_CTR to VRING_PACKED_F_WRAP_CTR?
> > >
> > > Jason, what is your take?
> >
> > That looks even worse. So I'm fine to define a new macro as 15.
> >
> > Thanks
> >
> > >
> > > >
> > > > > +
> > > > > /* We support indirect buffer descriptors */
> > > > > #define VIRTIO_RING_F_INDIRECT_DESC 28
> > > > >
> > > > > --
> > > > > 2.31.1
> > > > >
> > > >
> > >
> >
> hello, Jason
>
> Any other suggestions?

Looking at the current code again, it seems we've already used
VRING_PACKED_EVENT_F_WRAP_CTR for packing information. see
virtqueue_enable_cb_prepare_packed():

return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
VRING_PACKED_EVENT_F_WRAP_CTR);

So I still prefer to simply use VRING_PACKED_EVENT_F_WRAP_CTR.

Others should be fine.

Thanks

> I will do some stability and functional
> testing before remaking a patch for patch v2.
>
> Thanks
>

2022-06-16 05:28:06

by 黄杰

Subject: [PATCH v2] virtio_ring : keep used_wrap_counter in vq->last_used_idx

From: "huangjie.albert" <[email protected]>

the used_wrap_counter and the vq->last_used_idx may get
out of sync if they are separate assignment，and interrupt
might use an incorrect value to check for the used index.

for example:OOB access
ksoftirqd may consume the packet and it will call:
virtnet_poll
-->virtnet_receive
-->virtqueue_get_buf_ctx
-->virtqueue_get_buf_ctx_packed
and in virtqueue_get_buf_ctx_packed:

vq->last_used_idx += vq->packed.desc_state[id].num;
if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
vq->last_used_idx -= vq->packed.vring.num;
vq->packed.used_wrap_counter ^= 1;
}

if at the same time, there comes a vring interrupt，in vring_interrupt:
we will call:
vring_interrupt
-->more_used
-->more_used_packed
-->is_used_desc_packed
in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
so this could case a memory out of bounds bug.

this patch is to keep the used_wrap_counter in vq->last_used_idx
so we can get the correct value to check for used index in interrupt.

v1->v2:
- reuse the VRING_PACKED_EVENT_F_WRAP_CTR
- Remove parameter judgment in is_used_desc_packed,
because it can't be illegal

Signed-off-by: huangjie.albert <[email protected]>
---
drivers/virtio/virtio_ring.c | 57 ++++++++++++++++++++----------------
1 file changed, 31 insertions(+), 26 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 13a7348cedff..b22d97c9a755 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -111,7 +111,12 @@ struct vring_virtqueue {
/* Number we've added since last sync. */
unsigned int num_added;

- /* Last used index we've seen. */
+ /* Last used index we've seen.
+ * for split ring, it just contains last used index
+ * for packed ring, it not only contains last used index, but also
+ * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
+ * the bit shift in last_used_idx
+ */
u16 last_used_idx;

/* Hint for event idx: already triggered no need to disable. */
@@ -154,9 +159,6 @@ struct vring_virtqueue {
/* Driver ring wrap counter. */
bool avail_wrap_counter;

- /* Device ring wrap counter. */
- bool used_wrap_counter;
-
/* Avail used flags. */
u16 avail_used_flags;

@@ -1406,8 +1408,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,

static inline bool more_used_packed(const struct vring_virtqueue *vq)
{
- return is_used_desc_packed(vq, vq->last_used_idx,
- vq->packed.used_wrap_counter);
+ u16 last_used;
+ bool used_wrap_counter;
+
+ last_used = vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);
+ used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);
+ return is_used_desc_packed(vq, last_used, used_wrap_counter);
}

static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
@@ -1416,6 +1422,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
{
struct vring_virtqueue *vq = to_vvq(_vq);
u16 last_used, id;
+ bool used_wrap_counter;
void *ret;

START_USE(vq);
@@ -1434,7 +1441,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
/* Only get used elements after they have been exposed by host. */
virtio_rmb(vq->weak_barriers);

- last_used = vq->last_used_idx;
+ used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR));
+ last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
*len = le32_to_cpu(vq->packed.vring.desc[last_used].len);

@@ -1451,12 +1459,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
ret = vq->packed.desc_state[id].data;
detach_buf_packed(vq, id, ctx);

- vq->last_used_idx += vq->packed.desc_state[id].num;
- if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
- vq->last_used_idx -= vq->packed.vring.num;
- vq->packed.used_wrap_counter ^= 1;
+ last_used += vq->packed.desc_state[id].num;
+ if (unlikely(last_used >= vq->packed.vring.num)) {
+ last_used -= vq->packed.vring.num;
+ used_wrap_counter ^= 1;
}

+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
+ vq->last_used_idx = last_used;
+
/*
* If we expect an interrupt for the next entry, tell host
* by writing event index and flush out the write before
@@ -1465,9 +1476,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
virtio_store_mb(vq->weak_barriers,
&vq->packed.vring.driver->off_wrap,
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR)));
+ cpu_to_le16(vq->last_used_idx));

LAST_ADD_TIME_INVALID(vq);

@@ -1499,9 +1508,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)

if (vq->event) {
vq->packed.vring.driver->off_wrap =
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR));
+ cpu_to_le16(vq->last_used_idx);
/*
* We need to update event offset and event wrap
* counter first before updating event flags.
@@ -1518,8 +1525,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
}

END_USE(vq);
- return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR);
+ return vq->last_used_idx;
}

static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
@@ -1550,9 +1556,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
if (vq->event) {
/* TODO: tune this threshold */
bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
- wrap_counter = vq->packed.used_wrap_counter;
+ wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);

- used_idx = vq->last_used_idx + bufs;
+ used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR)) + bufs;
if (used_idx >= vq->packed.vring.num) {
used_idx -= vq->packed.vring.num;
wrap_counter ^= 1;
@@ -1582,9 +1588,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
*/
virtio_mb(vq->weak_barriers);

- if (is_used_desc_packed(vq,
- vq->last_used_idx,
- vq->packed.used_wrap_counter)) {
+ wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
+ used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+ if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
END_USE(vq);
return false;
}
@@ -1689,7 +1695,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
vq->notify = notify;
vq->weak_barriers = weak_barriers;
vq->broken = true;
- vq->last_used_idx = 0;
+ vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
vq->event_triggered = false;
vq->num_added = 0;
vq->packed_ring = true;
@@ -1720,7 +1726,6 @@ static struct virtqueue *vring_create_virtqueue_packed(

vq->packed.next_avail_idx = 0;
vq->packed.avail_wrap_counter = 1;
- vq->packed.used_wrap_counter = 1;
vq->packed.event_flags_shadow = 0;
vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;

--
2.31.1

2022-06-16 06:29:31

by Michael S. Tsirkin

Subject: Re: [PATCH v2] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Thu, Jun 16, 2022 at 01:12:21PM +0800, Albert Huang wrote:
> From: "huangjie.albert" <[email protected]>
>
> the used_wrap_counter and the vq->last_used_idx may get
> out of sync if they are separate assignment，and interrupt
> might use an incorrect value to check for the used index.
>
> for example:OOB access
> ksoftirqd may consume the packet and it will call:
> virtnet_poll
> -->virtnet_receive
> -->virtqueue_get_buf_ctx
> -->virtqueue_get_buf_ctx_packed
> and in virtqueue_get_buf_ctx_packed:
>
> vq->last_used_idx += vq->packed.desc_state[id].num;
> if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> vq->last_used_idx -= vq->packed.vring.num;
> vq->packed.used_wrap_counter ^= 1;
> }
>
> if at the same time, there comes a vring interrupt，in vring_interrupt:
> we will call:
> vring_interrupt
> -->more_used
> -->more_used_packed
> -->is_used_desc_packed
> in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> so this could case a memory out of bounds bug.
>
> this patch is to keep the used_wrap_counter in vq->last_used_idx
> so we can get the correct value to check for used index in interrupt.
>
> v1->v2:
> - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> - Remove parameter judgment in is_used_desc_packed,
> because it can't be illegal
>
> Signed-off-by: huangjie.albert <[email protected]>


This looks good, just a small suggestion below:

> ---
> drivers/virtio/virtio_ring.c | 57 ++++++++++++++++++++----------------
> 1 file changed, 31 insertions(+), 26 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 13a7348cedff..b22d97c9a755 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -111,7 +111,12 @@ struct vring_virtqueue {
> /* Number we've added since last sync. */
> unsigned int num_added;
>
> - /* Last used index we've seen. */
> + /* Last used index we've seen.
> + * for split ring, it just contains last used index
> + * for packed ring, it not only contains last used index, but also
> + * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
> + * the bit shift in last_used_idx
> + */
> u16 last_used_idx;
>
> /* Hint for event idx: already triggered no need to disable. */
> @@ -154,9 +159,6 @@ struct vring_virtqueue {
> /* Driver ring wrap counter. */
> bool avail_wrap_counter;
>
> - /* Device ring wrap counter. */
> - bool used_wrap_counter;
> -
> /* Avail used flags. */
> u16 avail_used_flags;
>
> @@ -1406,8 +1408,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
>
> static inline bool more_used_packed(const struct vring_virtqueue *vq)
> {
> - return is_used_desc_packed(vq, vq->last_used_idx,
> - vq->packed.used_wrap_counter);
> + u16 last_used;
> + bool used_wrap_counter;
> +
> + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);

This only works if last_used_idx is 16 bit and
VRING_PACKED_EVENT_F_WRAP_CTR is 15.

I think you want
/* all bits below VRING_PACKED_EVENT_F_WRAP_CTR */
vq->last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));


> + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);


A bit more efficient and clear:

!!(q->last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR))



Also this logic is repeated in multiple places. Let's add a couple of inline
functions:

static inline bool packed_used_wrap_counter(vq)

static inline u16 packed_last_used(vq)

then use these everywhere.


> + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> }
>
> static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> @@ -1416,6 +1422,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> u16 last_used, id;
> + bool used_wrap_counter;
> void *ret;
>
> START_USE(vq);
> @@ -1434,7 +1441,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> /* Only get used elements after they have been exposed by host. */
> virtio_rmb(vq->weak_barriers);
>
> - last_used = vq->last_used_idx;
> + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR));
> + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
>
> @@ -1451,12 +1459,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> ret = vq->packed.desc_state[id].data;
> detach_buf_packed(vq, id, ctx);
>
> - vq->last_used_idx += vq->packed.desc_state[id].num;
> - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> - vq->last_used_idx -= vq->packed.vring.num;
> - vq->packed.used_wrap_counter ^= 1;
> + last_used += vq->packed.desc_state[id].num;
> + if (unlikely(last_used >= vq->packed.vring.num)) {
> + last_used -= vq->packed.vring.num;
> + used_wrap_counter ^= 1;
> }
>
> + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> + vq->last_used_idx = last_used;
> +
> /*
> * If we expect an interrupt for the next entry, tell host
> * by writing event index and flush out the write before
> @@ -1465,9 +1476,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> virtio_store_mb(vq->weak_barriers,
> &vq->packed.vring.driver->off_wrap,
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR)));
> + cpu_to_le16(vq->last_used_idx));
>
> LAST_ADD_TIME_INVALID(vq);
>
> @@ -1499,9 +1508,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
>
> if (vq->event) {
> vq->packed.vring.driver->off_wrap =
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR));
> + cpu_to_le16(vq->last_used_idx);
> /*
> * We need to update event offset and event wrap
> * counter first before updating event flags.
> @@ -1518,8 +1525,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> }
>
> END_USE(vq);
> - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR);
> + return vq->last_used_idx;
> }
>
> static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> @@ -1550,9 +1556,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> if (vq->event) {
> /* TODO: tune this threshold */
> bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> - wrap_counter = vq->packed.used_wrap_counter;
> + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
>
> - used_idx = vq->last_used_idx + bufs;
> + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR)) + bufs;
> if (used_idx >= vq->packed.vring.num) {
> used_idx -= vq->packed.vring.num;
> wrap_counter ^= 1;
> @@ -1582,9 +1588,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> */
> virtio_mb(vq->weak_barriers);
>
> - if (is_used_desc_packed(vq,
> - vq->last_used_idx,
> - vq->packed.used_wrap_counter)) {
> + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> END_USE(vq);
> return false;
> }
> @@ -1689,7 +1695,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> vq->notify = notify;
> vq->weak_barriers = weak_barriers;
> vq->broken = true;
> - vq->last_used_idx = 0;
> + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> vq->event_triggered = false;
> vq->num_added = 0;
> vq->packed_ring = true;
> @@ -1720,7 +1726,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
>
> vq->packed.next_avail_idx = 0;
> vq->packed.avail_wrap_counter = 1;
> - vq->packed.used_wrap_counter = 1;
> vq->packed.event_flags_shadow = 0;
> vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
>
> --
> 2.31.1

2022-06-16 06:54:39

by Michael S. Tsirkin

Subject: Re: [PATCH v2] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Thu, Jun 16, 2022 at 01:12:21PM +0800, Albert Huang wrote:
> From: "huangjie.albert" <[email protected]>
>
> the used_wrap_counter and the vq->last_used_idx may get
> out of sync if they are separate assignment，and interrupt
> might use an incorrect value to check for the used index.
>
> for example:OOB access
> ksoftirqd may consume the packet and it will call:
> virtnet_poll
> -->virtnet_receive
> -->virtqueue_get_buf_ctx
> -->virtqueue_get_buf_ctx_packed
> and in virtqueue_get_buf_ctx_packed:
>
> vq->last_used_idx += vq->packed.desc_state[id].num;
> if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> vq->last_used_idx -= vq->packed.vring.num;
> vq->packed.used_wrap_counter ^= 1;
> }
>
> if at the same time, there comes a vring interrupt，in vring_interrupt:
> we will call:
> vring_interrupt
> -->more_used
> -->more_used_packed
> -->is_used_desc_packed
> in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> so this could case a memory out of bounds bug.
>
> this patch is to keep the used_wrap_counter in vq->last_used_idx
> so we can get the correct value to check for used index in interrupt.
>
> v1->v2:
> - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> - Remove parameter judgment in is_used_desc_packed,
> because it can't be illegal
>
> Signed-off-by: huangjie.albert <[email protected]>
> ---
> drivers/virtio/virtio_ring.c | 57 ++++++++++++++++++++----------------
> 1 file changed, 31 insertions(+), 26 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 13a7348cedff..b22d97c9a755 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -111,7 +111,12 @@ struct vring_virtqueue {
> /* Number we've added since last sync. */
> unsigned int num_added;
>
> - /* Last used index we've seen. */
> + /* Last used index we've seen.
> + * for split ring, it just contains last used index
> + * for packed ring, it not only contains last used index, but also
> + * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
> + * the bit shift in last_used_idx
> + */
> u16 last_used_idx;
>
> /* Hint for event idx: already triggered no need to disable. */
> @@ -154,9 +159,6 @@ struct vring_virtqueue {
> /* Driver ring wrap counter. */
> bool avail_wrap_counter;
>
> - /* Device ring wrap counter. */
> - bool used_wrap_counter;
> -
> /* Avail used flags. */
> u16 avail_used_flags;
>
> @@ -1406,8 +1408,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
>
> static inline bool more_used_packed(const struct vring_virtqueue *vq)
> {
> - return is_used_desc_packed(vq, vq->last_used_idx,
> - vq->packed.used_wrap_counter);
> + u16 last_used;
> + bool used_wrap_counter;
> +
> + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);
> + return is_used_desc_packed(vq, last_used, used_wrap_counter);

Hmm.

If vq->last_used_idx is read twice like this the values can be inconsistent,
no idea what the result will be if so.

I think we need to read vq->last_used_idx with READ_ONCE.

And I guess write it with WRITE_ONCE for symmetry.



> }
>
> static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> @@ -1416,6 +1422,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> u16 last_used, id;
> + bool used_wrap_counter;
> void *ret;
>
> START_USE(vq);
> @@ -1434,7 +1441,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> /* Only get used elements after they have been exposed by host. */
> virtio_rmb(vq->weak_barriers);
>
> - last_used = vq->last_used_idx;
> + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR));
> + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
>
> @@ -1451,12 +1459,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> ret = vq->packed.desc_state[id].data;
> detach_buf_packed(vq, id, ctx);
>
> - vq->last_used_idx += vq->packed.desc_state[id].num;
> - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> - vq->last_used_idx -= vq->packed.vring.num;
> - vq->packed.used_wrap_counter ^= 1;
> + last_used += vq->packed.desc_state[id].num;
> + if (unlikely(last_used >= vq->packed.vring.num)) {
> + last_used -= vq->packed.vring.num;
> + used_wrap_counter ^= 1;
> }
>
> + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> + vq->last_used_idx = last_used;
> +
> /*
> * If we expect an interrupt for the next entry, tell host
> * by writing event index and flush out the write before
> @@ -1465,9 +1476,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> virtio_store_mb(vq->weak_barriers,
> &vq->packed.vring.driver->off_wrap,
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR)));
> + cpu_to_le16(vq->last_used_idx));
>
> LAST_ADD_TIME_INVALID(vq);
>
> @@ -1499,9 +1508,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
>
> if (vq->event) {
> vq->packed.vring.driver->off_wrap =
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR));
> + cpu_to_le16(vq->last_used_idx);
> /*
> * We need to update event offset and event wrap
> * counter first before updating event flags.
> @@ -1518,8 +1525,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> }
>
> END_USE(vq);
> - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR);
> + return vq->last_used_idx;
> }
>
> static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> @@ -1550,9 +1556,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> if (vq->event) {
> /* TODO: tune this threshold */
> bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> - wrap_counter = vq->packed.used_wrap_counter;
> + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
>
> - used_idx = vq->last_used_idx + bufs;
> + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR)) + bufs;
> if (used_idx >= vq->packed.vring.num) {
> used_idx -= vq->packed.vring.num;
> wrap_counter ^= 1;
> @@ -1582,9 +1588,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> */
> virtio_mb(vq->weak_barriers);
>
> - if (is_used_desc_packed(vq,
> - vq->last_used_idx,
> - vq->packed.used_wrap_counter)) {
> + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> END_USE(vq);
> return false;
> }
> @@ -1689,7 +1695,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> vq->notify = notify;
> vq->weak_barriers = weak_barriers;
> vq->broken = true;
> - vq->last_used_idx = 0;
> + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> vq->event_triggered = false;
> vq->num_added = 0;
> vq->packed_ring = true;
> @@ -1720,7 +1726,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
>
> vq->packed.next_avail_idx = 0;
> vq->packed.avail_wrap_counter = 1;
> - vq->packed.used_wrap_counter = 1;
> vq->packed.event_flags_shadow = 0;
> vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
>
> --
> 2.31.1

2022-06-16 07:01:18

by Michael S. Tsirkin

Subject: Re: [PATCH v2] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Thu, Jun 16, 2022 at 02:07:19AM -0400, Michael S. Tsirkin wrote:
> On Thu, Jun 16, 2022 at 01:12:21PM +0800, Albert Huang wrote:
> > From: "huangjie.albert" <[email protected]>
> >
> > the used_wrap_counter and the vq->last_used_idx may get
> > out of sync if they are separate assignment，and interrupt
> > might use an incorrect value to check for the used index.
> >
> > for example:OOB access
> > ksoftirqd may consume the packet and it will call:
> > virtnet_poll
> > -->virtnet_receive
> > -->virtqueue_get_buf_ctx
> > -->virtqueue_get_buf_ctx_packed
> > and in virtqueue_get_buf_ctx_packed:
> >
> > vq->last_used_idx += vq->packed.desc_state[id].num;
> > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > vq->last_used_idx -= vq->packed.vring.num;
> > vq->packed.used_wrap_counter ^= 1;
> > }
> >
> > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > we will call:
> > vring_interrupt
> > -->more_used
> > -->more_used_packed
> > -->is_used_desc_packed
> > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > so this could case a memory out of bounds bug.
> >
> > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > so we can get the correct value to check for used index in interrupt.
> >
> > v1->v2:
> > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > - Remove parameter judgment in is_used_desc_packed,
> > because it can't be illegal
> >
> > Signed-off-by: huangjie.albert <[email protected]>
>
>
> This looks good, just a small suggestion below:
>
> > ---
> > drivers/virtio/virtio_ring.c | 57 ++++++++++++++++++++----------------
> > 1 file changed, 31 insertions(+), 26 deletions(-)
> >
> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > index 13a7348cedff..b22d97c9a755 100644
> > --- a/drivers/virtio/virtio_ring.c
> > +++ b/drivers/virtio/virtio_ring.c
> > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > /* Number we've added since last sync. */
> > unsigned int num_added;
> >
> > - /* Last used index we've seen. */
> > + /* Last used index we've seen.
> > + * for split ring, it just contains last used index
> > + * for packed ring, it not only contains last used index, but also
> > + * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
> > + * the bit shift in last_used_idx
> > + */
> > u16 last_used_idx;
> >
> > /* Hint for event idx: already triggered no need to disable. */
> > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > /* Driver ring wrap counter. */
> > bool avail_wrap_counter;
> >
> > - /* Device ring wrap counter. */
> > - bool used_wrap_counter;
> > -
> > /* Avail used flags. */
> > u16 avail_used_flags;
> >
> > @@ -1406,8 +1408,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> >
> > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > {
> > - return is_used_desc_packed(vq, vq->last_used_idx,
> > - vq->packed.used_wrap_counter);
> > + u16 last_used;
> > + bool used_wrap_counter;
> > +
> > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);
>
> This only works if last_used_idx is 16 bit and
> VRING_PACKED_EVENT_F_WRAP_CTR is 15.
>
> I think you want
> /* all bits below VRING_PACKED_EVENT_F_WRAP_CTR */
> vq->last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
>
>
> > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);
>
>
> A bit more efficient and clear:
>
> !!(q->last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR))
>
>
>
> Also this logic is repeated in multiple places. Let's add a couple of inline
> functions:
>
> static inline bool packed_used_wrap_counter(vq)
>
> static inline u16 packed_last_used(vq)

Or better:

packed_used_wrap_counter(u16 last_used_idx)
packed_last_used(u16 last_used_idx)


> then use these everywhere.
>
>
> > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > }
> >
> > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > @@ -1416,6 +1422,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > {
> > struct vring_virtqueue *vq = to_vvq(_vq);
> > u16 last_used, id;
> > + bool used_wrap_counter;
> > void *ret;
> >
> > START_USE(vq);
> > @@ -1434,7 +1441,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > /* Only get used elements after they have been exposed by host. */
> > virtio_rmb(vq->weak_barriers);
> >
> > - last_used = vq->last_used_idx;
> > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR));
> > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> >
> > @@ -1451,12 +1459,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > ret = vq->packed.desc_state[id].data;
> > detach_buf_packed(vq, id, ctx);
> >
> > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > - vq->last_used_idx -= vq->packed.vring.num;
> > - vq->packed.used_wrap_counter ^= 1;
> > + last_used += vq->packed.desc_state[id].num;
> > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > + last_used -= vq->packed.vring.num;
> > + used_wrap_counter ^= 1;
> > }
> >
> > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > + vq->last_used_idx = last_used;
> > +
> > /*
> > * If we expect an interrupt for the next entry, tell host
> > * by writing event index and flush out the write before
> > @@ -1465,9 +1476,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > virtio_store_mb(vq->weak_barriers,
> > &vq->packed.vring.driver->off_wrap,
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > + cpu_to_le16(vq->last_used_idx));
> >
> > LAST_ADD_TIME_INVALID(vq);
> >
> > @@ -1499,9 +1508,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> >
> > if (vq->event) {
> > vq->packed.vring.driver->off_wrap =
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > + cpu_to_le16(vq->last_used_idx);
> > /*
> > * We need to update event offset and event wrap
> > * counter first before updating event flags.
> > @@ -1518,8 +1525,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > }
> >
> > END_USE(vq);
> > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > + return vq->last_used_idx;
> > }
> >
> > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > @@ -1550,9 +1556,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > if (vq->event) {
> > /* TODO: tune this threshold */
> > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > - wrap_counter = vq->packed.used_wrap_counter;
> > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> >
> > - used_idx = vq->last_used_idx + bufs;
> > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR)) + bufs;
> > if (used_idx >= vq->packed.vring.num) {
> > used_idx -= vq->packed.vring.num;
> > wrap_counter ^= 1;
> > @@ -1582,9 +1588,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > */
> > virtio_mb(vq->weak_barriers);
> >
> > - if (is_used_desc_packed(vq,
> > - vq->last_used_idx,
> > - vq->packed.used_wrap_counter)) {
> > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > END_USE(vq);
> > return false;
> > }
> > @@ -1689,7 +1695,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > vq->notify = notify;
> > vq->weak_barriers = weak_barriers;
> > vq->broken = true;
> > - vq->last_used_idx = 0;
> > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > vq->event_triggered = false;
> > vq->num_added = 0;
> > vq->packed_ring = true;
> > @@ -1720,7 +1726,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> >
> > vq->packed.next_avail_idx = 0;
> > vq->packed.avail_wrap_counter = 1;
> > - vq->packed.used_wrap_counter = 1;
> > vq->packed.event_flags_shadow = 0;
> > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> >
> > --
> > 2.31.1

2022-06-16 07:47:59

by 黄杰

Subject: Re: [External] Re: [PATCH v2] virtio_ring : keep used_wrap_counter in vq->last_used_idx

Michael S. Tsirkin <[email protected]> 于2022年6月16日周四 14:42写道：
>
> On Thu, Jun 16, 2022 at 02:07:19AM -0400, Michael S. Tsirkin wrote:
> > On Thu, Jun 16, 2022 at 01:12:21PM +0800, Albert Huang wrote:
> > > From: "huangjie.albert" <[email protected]>
> > >
> > > the used_wrap_counter and the vq->last_used_idx may get
> > > out of sync if they are separate assignment，and interrupt
> > > might use an incorrect value to check for the used index.
> > >
> > > for example:OOB access
> > > ksoftirqd may consume the packet and it will call:
> > > virtnet_poll
> > > -->virtnet_receive
> > > -->virtqueue_get_buf_ctx
> > > -->virtqueue_get_buf_ctx_packed
> > > and in virtqueue_get_buf_ctx_packed:
> > >
> > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > vq->last_used_idx -= vq->packed.vring.num;
> > > vq->packed.used_wrap_counter ^= 1;
> > > }
> > >
> > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > we will call:
> > > vring_interrupt
> > > -->more_used
> > > -->more_used_packed
> > > -->is_used_desc_packed
> > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > so this could case a memory out of bounds bug.
> > >
> > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > so we can get the correct value to check for used index in interrupt.
> > >
> > > v1->v2:
> > > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > > - Remove parameter judgment in is_used_desc_packed,
> > > because it can't be illegal
> > >
> > > Signed-off-by: huangjie.albert <[email protected]>
> >
> >
> > This looks good, just a small suggestion below:
> >
> > > ---
> > > drivers/virtio/virtio_ring.c | 57 ++++++++++++++++++++----------------
> > > 1 file changed, 31 insertions(+), 26 deletions(-)
> > >
> > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > index 13a7348cedff..b22d97c9a755 100644
> > > --- a/drivers/virtio/virtio_ring.c
> > > +++ b/drivers/virtio/virtio_ring.c
> > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > /* Number we've added since last sync. */
> > > unsigned int num_added;
> > >
> > > - /* Last used index we've seen. */
> > > + /* Last used index we've seen.
> > > + * for split ring, it just contains last used index
> > > + * for packed ring, it not only contains last used index, but also
> > > + * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
> > > + * the bit shift in last_used_idx
> > > + */
> > > u16 last_used_idx;
> > >
> > > /* Hint for event idx: already triggered no need to disable. */
> > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > /* Driver ring wrap counter. */
> > > bool avail_wrap_counter;
> > >
> > > - /* Device ring wrap counter. */
> > > - bool used_wrap_counter;
> > > -
> > > /* Avail used flags. */
> > > u16 avail_used_flags;
> > >
> > > @@ -1406,8 +1408,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > >
> > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > {
> > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > - vq->packed.used_wrap_counter);
> > > + u16 last_used;
> > > + bool used_wrap_counter;
> > > +
> > > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> >
> > This only works if last_used_idx is 16 bit and
> > VRING_PACKED_EVENT_F_WRAP_CTR is 15.
> >
> > I think you want
> > /* all bits below VRING_PACKED_EVENT_F_WRAP_CTR */
> > vq->last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> >
> >
> > > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);
> >
> >
> > A bit more efficient and clear:
> >
> > !!(q->last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> >
> >
> >
> > Also this logic is repeated in multiple places. Let's add a couple of inline
> > functions:
> >
> > static inline bool packed_used_wrap_counter(vq)
> >
> > static inline u16 packed_last_used(vq)
>
> Or better:
>
> packed_used_wrap_counter(u16 last_used_idx)
> packed_last_used(u16 last_used_idx)
>

This one does look better, Jason what is your take?

thanks.

> > then use these everywhere.
> >
> >
> > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > }
> > >
> > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > @@ -1416,6 +1422,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > {
> > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > u16 last_used, id;
> > > + bool used_wrap_counter;
> > > void *ret;
> > >
> > > START_USE(vq);
> > > @@ -1434,7 +1441,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > /* Only get used elements after they have been exposed by host. */
> > > virtio_rmb(vq->weak_barriers);
> > >
> > > - last_used = vq->last_used_idx;
> > > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > >
> > > @@ -1451,12 +1459,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > ret = vq->packed.desc_state[id].data;
> > > detach_buf_packed(vq, id, ctx);
> > >
> > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > - vq->last_used_idx -= vq->packed.vring.num;
> > > - vq->packed.used_wrap_counter ^= 1;
> > > + last_used += vq->packed.desc_state[id].num;
> > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > > + last_used -= vq->packed.vring.num;
> > > + used_wrap_counter ^= 1;
> > > }
> > >
> > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + vq->last_used_idx = last_used;
> > > +
> > > /*
> > > * If we expect an interrupt for the next entry, tell host
> > > * by writing event index and flush out the write before
> > > @@ -1465,9 +1476,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > virtio_store_mb(vq->weak_barriers,
> > > &vq->packed.vring.driver->off_wrap,
> > > - cpu_to_le16(vq->last_used_idx |
> > > - (vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > + cpu_to_le16(vq->last_used_idx));
> > >
> > > LAST_ADD_TIME_INVALID(vq);
> > >
> > > @@ -1499,9 +1508,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > >
> > > if (vq->event) {
> > > vq->packed.vring.driver->off_wrap =
> > > - cpu_to_le16(vq->last_used_idx |
> > > - (vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + cpu_to_le16(vq->last_used_idx);
> > > /*
> > > * We need to update event offset and event wrap
> > > * counter first before updating event flags.
> > > @@ -1518,8 +1525,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > }
> > >
> > > END_USE(vq);
> > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > + return vq->last_used_idx;
> > > }
> > >
> > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > @@ -1550,9 +1556,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > if (vq->event) {
> > > /* TODO: tune this threshold */
> > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > - wrap_counter = vq->packed.used_wrap_counter;
> > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> > >
> > > - used_idx = vq->last_used_idx + bufs;
> > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR)) + bufs;
> > > if (used_idx >= vq->packed.vring.num) {
> > > used_idx -= vq->packed.vring.num;
> > > wrap_counter ^= 1;
> > > @@ -1582,9 +1588,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > */
> > > virtio_mb(vq->weak_barriers);
> > >
> > > - if (is_used_desc_packed(vq,
> > > - vq->last_used_idx,
> > > - vq->packed.used_wrap_counter)) {
> > > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> > > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > END_USE(vq);
> > > return false;
> > > }
> > > @@ -1689,7 +1695,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > vq->notify = notify;
> > > vq->weak_barriers = weak_barriers;
> > > vq->broken = true;
> > > - vq->last_used_idx = 0;
> > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > > vq->event_triggered = false;
> > > vq->num_added = 0;
> > > vq->packed_ring = true;
> > > @@ -1720,7 +1726,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > >
> > > vq->packed.next_avail_idx = 0;
> > > vq->packed.avail_wrap_counter = 1;
> > > - vq->packed.used_wrap_counter = 1;
> > > vq->packed.event_flags_shadow = 0;
> > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > >
> > > --
> > > 2.31.1
>

2022-06-16 07:54:35

by 黄杰

Subject: Re: [External] Re: [PATCH v2] virtio_ring : keep used_wrap_counter in vq->last_used_idx

Michael S. Tsirkin <[email protected]> 于2022年6月16日周四 14:41写道：
>
> On Thu, Jun 16, 2022 at 01:12:21PM +0800, Albert Huang wrote:
> > From: "huangjie.albert" <[email protected]>
> >
> > the used_wrap_counter and the vq->last_used_idx may get
> > out of sync if they are separate assignment，and interrupt
> > might use an incorrect value to check for the used index.
> >
> > for example:OOB access
> > ksoftirqd may consume the packet and it will call:
> > virtnet_poll
> > -->virtnet_receive
> > -->virtqueue_get_buf_ctx
> > -->virtqueue_get_buf_ctx_packed
> > and in virtqueue_get_buf_ctx_packed:
> >
> > vq->last_used_idx += vq->packed.desc_state[id].num;
> > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > vq->last_used_idx -= vq->packed.vring.num;
> > vq->packed.used_wrap_counter ^= 1;
> > }
> >
> > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > we will call:
> > vring_interrupt
> > -->more_used
> > -->more_used_packed
> > -->is_used_desc_packed
> > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > so this could case a memory out of bounds bug.
> >
> > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > so we can get the correct value to check for used index in interrupt.
> >
> > v1->v2:
> > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > - Remove parameter judgment in is_used_desc_packed,
> > because it can't be illegal
> >
> > Signed-off-by: huangjie.albert <[email protected]>
> > ---
> > drivers/virtio/virtio_ring.c | 57 ++++++++++++++++++++----------------
> > 1 file changed, 31 insertions(+), 26 deletions(-)
> >
> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > index 13a7348cedff..b22d97c9a755 100644
> > --- a/drivers/virtio/virtio_ring.c
> > +++ b/drivers/virtio/virtio_ring.c
> > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > /* Number we've added since last sync. */
> > unsigned int num_added;
> >
> > - /* Last used index we've seen. */
> > + /* Last used index we've seen.
> > + * for split ring, it just contains last used index
> > + * for packed ring, it not only contains last used index, but also
> > + * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
> > + * the bit shift in last_used_idx
> > + */
> > u16 last_used_idx;
> >
> > /* Hint for event idx: already triggered no need to disable. */
> > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > /* Driver ring wrap counter. */
> > bool avail_wrap_counter;
> >
> > - /* Device ring wrap counter. */
> > - bool used_wrap_counter;
> > -
> > /* Avail used flags. */
> > u16 avail_used_flags;
> >
> > @@ -1406,8 +1408,12 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> >
> > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > {
> > - return is_used_desc_packed(vq, vq->last_used_idx,
> > - vq->packed.used_wrap_counter);
> > + u16 last_used;
> > + bool used_wrap_counter;
> > +
> > + last_used = vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > + used_wrap_counter = !!((vq->last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);
> > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
>
> Hmm.
>
> If vq->last_used_idx is read twice like this the values can be inconsistent,
> no idea what the result will be if so.
>
> I think we need to read vq->last_used_idx with READ_ONCE.
>
> And I guess write it with WRITE_ONCE for symmetry.
>

if we read vq->last_used_idx with READ_ONCE. for example:

1)last_used = READ_ONCE (vq->last_used_idx) & ~(1 <<
VRING_PACKED_EVENT_F_WRAP_CTR);
2)used_wrap_counter = !!((READ_ONCE (vq->last_used_idx)) >>
VRING_PACKED_EVENT_F_WRAP_CTR);

1 and 2 may get different values, so think this will be better:
last_used_idx = vq->last_used_idx;
last_used = last_used_idx& ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR);
used_wrap_counter = !!((last_used_idx) >> VRING_PACKED_EVENT_F_WRAP_CTR);

what do you think？

>
> > }
> >
> > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > @@ -1416,6 +1422,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > {
> > struct vring_virtqueue *vq = to_vvq(_vq);
> > u16 last_used, id;
> > + bool used_wrap_counter;
> > void *ret;
> >
> > START_USE(vq);
> > @@ -1434,7 +1441,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > /* Only get used elements after they have been exposed by host. */
> > virtio_rmb(vq->weak_barriers);
> >
> > - last_used = vq->last_used_idx;
> > + used_wrap_counter = !!((vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR));
> > + last_used = (vq->last_used_idx) & (~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> >
> > @@ -1451,12 +1459,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > ret = vq->packed.desc_state[id].data;
> > detach_buf_packed(vq, id, ctx);
> >
> > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > - vq->last_used_idx -= vq->packed.vring.num;
> > - vq->packed.used_wrap_counter ^= 1;
> > + last_used += vq->packed.desc_state[id].num;
> > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > + last_used -= vq->packed.vring.num;
> > + used_wrap_counter ^= 1;
> > }
> >
> > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > + vq->last_used_idx = last_used;
> > +
> > /*
> > * If we expect an interrupt for the next entry, tell host
> > * by writing event index and flush out the write before
> > @@ -1465,9 +1476,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > virtio_store_mb(vq->weak_barriers,
> > &vq->packed.vring.driver->off_wrap,
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > + cpu_to_le16(vq->last_used_idx));
> >
> > LAST_ADD_TIME_INVALID(vq);
> >
> > @@ -1499,9 +1508,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> >
> > if (vq->event) {
> > vq->packed.vring.driver->off_wrap =
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > + cpu_to_le16(vq->last_used_idx);
> > /*
> > * We need to update event offset and event wrap
> > * counter first before updating event flags.
> > @@ -1518,8 +1525,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > }
> >
> > END_USE(vq);
> > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > + return vq->last_used_idx;
> > }
> >
> > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > @@ -1550,9 +1556,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > if (vq->event) {
> > /* TODO: tune this threshold */
> > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > - wrap_counter = vq->packed.used_wrap_counter;
> > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> >
> > - used_idx = vq->last_used_idx + bufs;
> > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR)) + bufs;
> > if (used_idx >= vq->packed.vring.num) {
> > used_idx -= vq->packed.vring.num;
> > wrap_counter ^= 1;
> > @@ -1582,9 +1588,9 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > */
> > virtio_mb(vq->weak_barriers);
> >
> > - if (is_used_desc_packed(vq,
> > - vq->last_used_idx,
> > - vq->packed.used_wrap_counter)) {
> > + wrap_counter = !!(vq->last_used_idx >> VRING_PACKED_EVENT_F_WRAP_CTR);
> > + used_idx = (vq->last_used_idx & ~(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > END_USE(vq);
> > return false;
> > }
> > @@ -1689,7 +1695,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > vq->notify = notify;
> > vq->weak_barriers = weak_barriers;
> > vq->broken = true;
> > - vq->last_used_idx = 0;
> > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > vq->event_triggered = false;
> > vq->num_added = 0;
> > vq->packed_ring = true;
> > @@ -1720,7 +1726,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> >
> > vq->packed.next_avail_idx = 0;
> > vq->packed.avail_wrap_counter = 1;
> > - vq->packed.used_wrap_counter = 1;
> > vq->packed.event_flags_shadow = 0;
> > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> >
> > --
> > 2.31.1
>

2022-06-16 10:08:52

by 黄杰

Subject: [PATCH v3] virtio_ring : keep used_wrap_counter in vq->last_used_idx

From: "huangjie.albert" <[email protected]>

the used_wrap_counter and the vq->last_used_idx may get
out of sync if they are separate assignment，and interrupt
might use an incorrect value to check for the used index.

for example:OOB access
ksoftirqd may consume the packet and it will call:
virtnet_poll
-->virtnet_receive
-->virtqueue_get_buf_ctx
-->virtqueue_get_buf_ctx_packed
and in virtqueue_get_buf_ctx_packed:

vq->last_used_idx += vq->packed.desc_state[id].num;
if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
vq->last_used_idx -= vq->packed.vring.num;
vq->packed.used_wrap_counter ^= 1;
}

if at the same time, there comes a vring interrupt，in vring_interrupt:
we will call:
vring_interrupt
-->more_used
-->more_used_packed
-->is_used_desc_packed
in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
so this could case a memory out of bounds bug.

this patch is to keep the used_wrap_counter in vq->last_used_idx
so we can get the correct value to check for used index in interrupt.

v2->v3:
- add inline function to get used_wrap_counter and last_used
- when use vq->last_used_idx, only read once
if vq->last_used_idx is read twice, the values can be inconsistent.
- use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR

v1->v2:
- reuse the VRING_PACKED_EVENT_F_WRAP_CTR
- Remove parameter judgment in is_used_desc_packed,
because it can't be illegal

Signed-off-by: huangjie.albert <[email protected]>
---
drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
1 file changed, 47 insertions(+), 28 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 13a7348cedff..0184b5056457 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -111,7 +111,12 @@ struct vring_virtqueue {
/* Number we've added since last sync. */
unsigned int num_added;

- /* Last used index we've seen. */
+ /* Last used index we've seen.
+ * for split ring, it just contains last used index
+ * for packed ring, it not only contains last used index, but also
+ * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
+ * the bit shift in last_used_idx
+ */
u16 last_used_idx;

/* Hint for event idx: already triggered no need to disable. */
@@ -154,9 +159,6 @@ struct vring_virtqueue {
/* Driver ring wrap counter. */
bool avail_wrap_counter;

- /* Device ring wrap counter. */
- bool used_wrap_counter;
-
/* Avail used flags. */
u16 avail_used_flags;

@@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
/*
* Packed ring specific functions - *_packed().
*/
+static inline bool packed_used_wrap_counter(u16 last_used_idx)
+{
+ return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}
+
+static inline u16 packed_last_used(u16 last_used_idx)
+{
+ return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}

static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
struct vring_desc_extra *extra)
@@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,

static inline bool more_used_packed(const struct vring_virtqueue *vq)
{
- return is_used_desc_packed(vq, vq->last_used_idx,
- vq->packed.used_wrap_counter);
+ u16 last_used;
+ u16 last_used_idx;
+ bool used_wrap_counter;
+
+ last_used_idx = vq->last_used_idx;
+ last_used = packed_last_used(last_used_idx);
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ return is_used_desc_packed(vq, last_used, used_wrap_counter);
}

static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
@@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
void **ctx)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 last_used, id;
+ u16 last_used, id, last_used_idx;
+ bool used_wrap_counter;
void *ret;

START_USE(vq);
@@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
/* Only get used elements after they have been exposed by host. */
virtio_rmb(vq->weak_barriers);

- last_used = vq->last_used_idx;
+ last_used_idx = vq->last_used_idx;
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ last_used = packed_last_used(last_used_idx);
id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
*len = le32_to_cpu(vq->packed.vring.desc[last_used].len);

@@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
ret = vq->packed.desc_state[id].data;
detach_buf_packed(vq, id, ctx);

- vq->last_used_idx += vq->packed.desc_state[id].num;
- if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
- vq->last_used_idx -= vq->packed.vring.num;
- vq->packed.used_wrap_counter ^= 1;
+ last_used += vq->packed.desc_state[id].num;
+ if (unlikely(last_used >= vq->packed.vring.num)) {
+ last_used -= vq->packed.vring.num;
+ used_wrap_counter ^= 1;
}

+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
+ vq->last_used_idx = last_used;
+
/*
* If we expect an interrupt for the next entry, tell host
* by writing event index and flush out the write before
@@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
virtio_store_mb(vq->weak_barriers,
&vq->packed.vring.driver->off_wrap,
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR)));
+ cpu_to_le16(vq->last_used_idx));

LAST_ADD_TIME_INVALID(vq);

@@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)

if (vq->event) {
vq->packed.vring.driver->off_wrap =
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR));
+ cpu_to_le16(vq->last_used_idx);
/*
* We need to update event offset and event wrap
* counter first before updating event flags.
@@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
}

END_USE(vq);
- return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR);
+ return vq->last_used_idx;
}

static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
@@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 used_idx, wrap_counter;
+ u16 used_idx, wrap_counter, last_used_idx;
u16 bufs;

START_USE(vq);
@@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
if (vq->event) {
/* TODO: tune this threshold */
bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
- wrap_counter = vq->packed.used_wrap_counter;
+ last_used_idx = vq->last_used_idx;
+ wrap_counter = packed_used_wrap_counter(last_used_idx);

- used_idx = vq->last_used_idx + bufs;
+ used_idx = packed_last_used(last_used_idx) + bufs;
if (used_idx >= vq->packed.vring.num) {
used_idx -= vq->packed.vring.num;
wrap_counter ^= 1;
@@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
*/
virtio_mb(vq->weak_barriers);

- if (is_used_desc_packed(vq,
- vq->last_used_idx,
- vq->packed.used_wrap_counter)) {
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ wrap_counter = packed_used_wrap_counter(last_used_idx);
+ used_idx = packed_last_used(last_used_idx);
+ if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
END_USE(vq);
return false;
}
@@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
vq->notify = notify;
vq->weak_barriers = weak_barriers;
vq->broken = true;
- vq->last_used_idx = 0;
+ vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
vq->event_triggered = false;
vq->num_added = 0;
vq->packed_ring = true;
@@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(

vq->packed.next_avail_idx = 0;
vq->packed.avail_wrap_counter = 1;
- vq->packed.used_wrap_counter = 1;
vq->packed.event_flags_shadow = 0;
vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;

--
2.31.1

2022-06-16 10:25:14

by 黄杰

Subject: [PATCH v3] virtio_ring : keep used_wrap_counter in vq->last_used_idx

From: "huangjie.albert" <[email protected]>

the used_wrap_counter and the vq->last_used_idx may get
out of sync if they are separate assignment，and interrupt
might use an incorrect value to check for the used index.

for example:OOB access
ksoftirqd may consume the packet and it will call:
virtnet_poll
-->virtnet_receive
-->virtqueue_get_buf_ctx
-->virtqueue_get_buf_ctx_packed
and in virtqueue_get_buf_ctx_packed:

vq->last_used_idx += vq->packed.desc_state[id].num;
if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
vq->last_used_idx -= vq->packed.vring.num;
vq->packed.used_wrap_counter ^= 1;
}

if at the same time, there comes a vring interrupt，in vring_interrupt:
we will call:
vring_interrupt
-->more_used
-->more_used_packed
-->is_used_desc_packed
in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
so this could case a memory out of bounds bug.

this patch is to keep the used_wrap_counter in vq->last_used_idx
so we can get the correct value to check for used index in interrupt.

v2->v3:
- add inline function to get used_wrap_counter and last_used
- when use vq->last_used_idx, only read once
if vq->last_used_idx is read twice, the values can be inconsistent.
- use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR

v1->v2:
- reuse the VRING_PACKED_EVENT_F_WRAP_CTR
- Remove parameter judgment in is_used_desc_packed,
because it can't be illegal

Signed-off-by: huangjie.albert <[email protected]>
---
drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
1 file changed, 47 insertions(+), 28 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 13a7348cedff..a253f50b8f86 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -111,7 +111,12 @@ struct vring_virtqueue {
/* Number we've added since last sync. */
unsigned int num_added;

- /* Last used index we've seen. */
+ /* Last used index we've seen.
+ * for split ring, it just contains last used index
+ * for packed ring, it not only contains last used index, but also
+ * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
+ * the bit shift in last_used_idx
+ */
u16 last_used_idx;

/* Hint for event idx: already triggered no need to disable. */
@@ -154,9 +159,6 @@ struct vring_virtqueue {
/* Driver ring wrap counter. */
bool avail_wrap_counter;

- /* Device ring wrap counter. */
- bool used_wrap_counter;
-
/* Avail used flags. */
u16 avail_used_flags;

@@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
/*
* Packed ring specific functions - *_packed().
*/
+static inline bool packed_used_wrap_counter(u16 last_used_idx)
+{
+ return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}
+
+static inline u16 packed_last_used(u16 last_used_idx)
+{
+ return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}

static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
struct vring_desc_extra *extra)
@@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,

static inline bool more_used_packed(const struct vring_virtqueue *vq)
{
- return is_used_desc_packed(vq, vq->last_used_idx,
- vq->packed.used_wrap_counter);
+ u16 last_used;
+ u16 last_used_idx;
+ bool used_wrap_counter;
+
+ last_used_idx = vq->last_used_idx;
+ last_used = packed_last_used(last_used_idx);
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ return is_used_desc_packed(vq, last_used, used_wrap_counter);
}

static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
@@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
void **ctx)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 last_used, id;
+ u16 last_used, id, last_used_idx;
+ bool used_wrap_counter;
void *ret;

START_USE(vq);
@@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
/* Only get used elements after they have been exposed by host. */
virtio_rmb(vq->weak_barriers);

- last_used = vq->last_used_idx;
+ last_used_idx = vq->last_used_idx;
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ last_used = packed_last_used(last_used_idx);
id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
*len = le32_to_cpu(vq->packed.vring.desc[last_used].len);

@@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
ret = vq->packed.desc_state[id].data;
detach_buf_packed(vq, id, ctx);

- vq->last_used_idx += vq->packed.desc_state[id].num;
- if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
- vq->last_used_idx -= vq->packed.vring.num;
- vq->packed.used_wrap_counter ^= 1;
+ last_used += vq->packed.desc_state[id].num;
+ if (unlikely(last_used >= vq->packed.vring.num)) {
+ last_used -= vq->packed.vring.num;
+ used_wrap_counter ^= 1;
}

+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
+ vq->last_used_idx = last_used;
+
/*
* If we expect an interrupt for the next entry, tell host
* by writing event index and flush out the write before
@@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
virtio_store_mb(vq->weak_barriers,
&vq->packed.vring.driver->off_wrap,
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR)));
+ cpu_to_le16(vq->last_used_idx));

LAST_ADD_TIME_INVALID(vq);

@@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)

if (vq->event) {
vq->packed.vring.driver->off_wrap =
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR));
+ cpu_to_le16(vq->last_used_idx);
/*
* We need to update event offset and event wrap
* counter first before updating event flags.
@@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
}

END_USE(vq);
- return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR);
+ return vq->last_used_idx;
}

static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
@@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 used_idx, wrap_counter;
+ u16 used_idx, wrap_counter, last_used_idx;
u16 bufs;

START_USE(vq);
@@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
if (vq->event) {
/* TODO: tune this threshold */
bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
- wrap_counter = vq->packed.used_wrap_counter;
+ last_used_idx = vq->last_used_idx;
+ wrap_counter = packed_used_wrap_counter(last_used_idx);

- used_idx = vq->last_used_idx + bufs;
+ used_idx = packed_last_used(last_used_idx) + bufs;
if (used_idx >= vq->packed.vring.num) {
used_idx -= vq->packed.vring.num;
wrap_counter ^= 1;
@@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
*/
virtio_mb(vq->weak_barriers);

- if (is_used_desc_packed(vq,
- vq->last_used_idx,
- vq->packed.used_wrap_counter)) {
+ last_used_idx = vq->last_used_idx;
+ wrap_counter = packed_used_wrap_counter(last_used_idx);
+ used_idx = packed_last_used(last_used_idx);
+ if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
END_USE(vq);
return false;
}
@@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
vq->notify = notify;
vq->weak_barriers = weak_barriers;
vq->broken = true;
- vq->last_used_idx = 0;
+ vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
vq->event_triggered = false;
vq->num_added = 0;
vq->packed_ring = true;
@@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(

vq->packed.next_avail_idx = 0;
vq->packed.avail_wrap_counter = 1;
- vq->packed.used_wrap_counter = 1;
vq->packed.event_flags_shadow = 0;
vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;

--
2.31.1

2022-06-16 12:50:05

by Michael S. Tsirkin

Subject: Re: [PATCH v3] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Thu, Jun 16, 2022 at 05:54:59PM +0800, Albert Huang wrote:
> From: "huangjie.albert" <[email protected]>
>
> the used_wrap_counter and the vq->last_used_idx may get
> out of sync if they are separate assignment，and interrupt
> might use an incorrect value to check for the used index.
>
> for example:OOB access
> ksoftirqd may consume the packet and it will call:
> virtnet_poll
> -->virtnet_receive
> -->virtqueue_get_buf_ctx
> -->virtqueue_get_buf_ctx_packed
> and in virtqueue_get_buf_ctx_packed:
>
> vq->last_used_idx += vq->packed.desc_state[id].num;
> if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> vq->last_used_idx -= vq->packed.vring.num;
> vq->packed.used_wrap_counter ^= 1;
> }
>
> if at the same time, there comes a vring interrupt，in vring_interrupt:
> we will call:
> vring_interrupt
> -->more_used
> -->more_used_packed
> -->is_used_desc_packed
> in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> so this could case a memory out of bounds bug.
>
> this patch is to keep the used_wrap_counter in vq->last_used_idx
> so we can get the correct value to check for used index in interrupt.
>
> v2->v3:
> - add inline function to get used_wrap_counter and last_used
> - when use vq->last_used_idx, only read once
> if vq->last_used_idx is read twice, the values can be inconsistent.
> - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
>
> v1->v2:
> - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> - Remove parameter judgment in is_used_desc_packed,
> because it can't be illegal
>
> Signed-off-by: huangjie.albert <[email protected]>
> ---
> drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> 1 file changed, 47 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 13a7348cedff..a253f50b8f86 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -111,7 +111,12 @@ struct vring_virtqueue {
> /* Number we've added since last sync. */
> unsigned int num_added;
>
> - /* Last used index we've seen. */
> + /* Last used index we've seen.
> + * for split ring, it just contains last used index
> + * for packed ring, it not only contains last used index, but also
> + * used_wrap_counter, the VRING_PACKED_EVENT_F_WRAP_CTR is
> + * the bit shift in last_used_idx


reword:

for packed ring, bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the
last used index. Bits from VRING_PACKED_EVENT_F_WRAP_CTR include the
used wrap counter.

> + */
> u16 last_used_idx;
>
> /* Hint for event idx: already triggered no need to disable. */
> @@ -154,9 +159,6 @@ struct vring_virtqueue {
> /* Driver ring wrap counter. */
> bool avail_wrap_counter;
>
> - /* Device ring wrap counter. */
> - bool used_wrap_counter;
> -
> /* Avail used flags. */
> u16 avail_used_flags;
>
> @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> /*
> * Packed ring specific functions - *_packed().
> */
> +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> +{
> + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> +}
> +
> +static inline u16 packed_last_used(u16 last_used_idx)
> +{
> + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> +}
>
> static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> struct vring_desc_extra *extra)
> @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
>
> static inline bool more_used_packed(const struct vring_virtqueue *vq)
> {
> - return is_used_desc_packed(vq, vq->last_used_idx,
> - vq->packed.used_wrap_counter);
> + u16 last_used;
> + u16 last_used_idx;
> + bool used_wrap_counter;
> +
> + last_used_idx = vq->last_used_idx;
> + last_used = packed_last_used(last_used_idx);
> + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> }
>
> static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> void **ctx)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> - u16 last_used, id;
> + u16 last_used, id, last_used_idx;
> + bool used_wrap_counter;
> void *ret;
>
> START_USE(vq);
> @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> /* Only get used elements after they have been exposed by host. */
> virtio_rmb(vq->weak_barriers);
>
> - last_used = vq->last_used_idx;
> + last_used_idx = vq->last_used_idx;


I think we need READ_ONCE here. Otherwise compiler is free to still
do two reads.

> + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> + last_used = packed_last_used(last_used_idx);
> id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
>
> @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> ret = vq->packed.desc_state[id].data;
> detach_buf_packed(vq, id, ctx);
>
> - vq->last_used_idx += vq->packed.desc_state[id].num;
> - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> - vq->last_used_idx -= vq->packed.vring.num;
> - vq->packed.used_wrap_counter ^= 1;
> + last_used += vq->packed.desc_state[id].num;
> + if (unlikely(last_used >= vq->packed.vring.num)) {
> + last_used -= vq->packed.vring.num;
> + used_wrap_counter ^= 1;
> }
>
> + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> + vq->last_used_idx = last_used;
> +
> /*
> * If we expect an interrupt for the next entry, tell host
> * by writing event index and flush out the write before
> @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> virtio_store_mb(vq->weak_barriers,
> &vq->packed.vring.driver->off_wrap,
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR)));
> + cpu_to_le16(vq->last_used_idx));
>
> LAST_ADD_TIME_INVALID(vq);
>
> @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
>
> if (vq->event) {
> vq->packed.vring.driver->off_wrap =
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR));
> + cpu_to_le16(vq->last_used_idx);
> /*
> * We need to update event offset and event wrap
> * counter first before updating event flags.
> @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> }
>
> END_USE(vq);
> - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR);
> + return vq->last_used_idx;
> }
>
> static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> - u16 used_idx, wrap_counter;
> + u16 used_idx, wrap_counter, last_used_idx;
> u16 bufs;
>
> START_USE(vq);
> @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> if (vq->event) {
> /* TODO: tune this threshold */
> bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> - wrap_counter = vq->packed.used_wrap_counter;
> + last_used_idx = vq->last_used_idx;
> + wrap_counter = packed_used_wrap_counter(last_used_idx);
>
> - used_idx = vq->last_used_idx + bufs;
> + used_idx = packed_last_used(last_used_idx) + bufs;
> if (used_idx >= vq->packed.vring.num) {
> used_idx -= vq->packed.vring.num;
> wrap_counter ^= 1;
> @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> */
> virtio_mb(vq->weak_barriers);
>
> - if (is_used_desc_packed(vq,
> - vq->last_used_idx,
> - vq->packed.used_wrap_counter)) {
> + last_used_idx = vq->last_used_idx;


same here.

> + wrap_counter = packed_used_wrap_counter(last_used_idx);
> + used_idx = packed_last_used(last_used_idx);
> + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> END_USE(vq);
> return false;
> }
> @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> vq->notify = notify;
> vq->weak_barriers = weak_barriers;
> vq->broken = true;
> - vq->last_used_idx = 0;
> + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> vq->event_triggered = false;
> vq->num_added = 0;
> vq->packed_ring = true;
> @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
>
> vq->packed.next_avail_idx = 0;
> vq->packed.avail_wrap_counter = 1;
> - vq->packed.used_wrap_counter = 1;
> vq->packed.event_flags_shadow = 0;
> vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
>
> --
> 2.31.1

2022-06-16 13:22:37

by 黄杰

Subject: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

From: "huangjie.albert" <[email protected]>

the used_wrap_counter and the vq->last_used_idx may get
out of sync if they are separate assignment，and interrupt
might use an incorrect value to check for the used index.

for example:OOB access
ksoftirqd may consume the packet and it will call:
virtnet_poll
-->virtnet_receive
-->virtqueue_get_buf_ctx
-->virtqueue_get_buf_ctx_packed
and in virtqueue_get_buf_ctx_packed:

vq->last_used_idx += vq->packed.desc_state[id].num;
if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
vq->last_used_idx -= vq->packed.vring.num;
vq->packed.used_wrap_counter ^= 1;
}

if at the same time, there comes a vring interrupt，in vring_interrupt:
we will call:
vring_interrupt
-->more_used
-->more_used_packed
-->is_used_desc_packed
in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
so this could case a memory out of bounds bug.

this patch is to keep the used_wrap_counter in vq->last_used_idx
so we can get the correct value to check for used index in interrupt.

v3->v4:
- use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx

v2->v3:
- add inline function to get used_wrap_counter and last_used
- when use vq->last_used_idx, only read once
if vq->last_used_idx is read twice, the values can be inconsistent.
- use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR

v1->v2:
- reuse the VRING_PACKED_EVENT_F_WRAP_CTR
- Remove parameter judgment in is_used_desc_packed,
because it can't be illegal

Signed-off-by: huangjie.albert <[email protected]>
---
drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
1 file changed, 47 insertions(+), 28 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 13a7348cedff..719fbbe716d6 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -111,7 +111,12 @@ struct vring_virtqueue {
/* Number we've added since last sync. */
unsigned int num_added;

- /* Last used index we've seen. */
+ /* Last used index we've seen.
+ * for split ring, it just contains last used index
+ * for packed ring:
+ * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
+ * bits VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
+ */
u16 last_used_idx;

/* Hint for event idx: already triggered no need to disable. */
@@ -154,9 +159,6 @@ struct vring_virtqueue {
/* Driver ring wrap counter. */
bool avail_wrap_counter;

- /* Device ring wrap counter. */
- bool used_wrap_counter;
-
/* Avail used flags. */
u16 avail_used_flags;

@@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
/*
* Packed ring specific functions - *_packed().
*/
+static inline bool packed_used_wrap_counter(u16 last_used_idx)
+{
+ return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}
+
+static inline u16 packed_last_used(u16 last_used_idx)
+{
+ return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}

static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
struct vring_desc_extra *extra)
@@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,

static inline bool more_used_packed(const struct vring_virtqueue *vq)
{
- return is_used_desc_packed(vq, vq->last_used_idx,
- vq->packed.used_wrap_counter);
+ u16 last_used;
+ u16 last_used_idx;
+ bool used_wrap_counter;
+
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ last_used = packed_last_used(last_used_idx);
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ return is_used_desc_packed(vq, last_used, used_wrap_counter);
}

static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
@@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
void **ctx)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 last_used, id;
+ u16 last_used, id, last_used_idx;
+ bool used_wrap_counter;
void *ret;

START_USE(vq);
@@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
/* Only get used elements after they have been exposed by host. */
virtio_rmb(vq->weak_barriers);

- last_used = vq->last_used_idx;
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ last_used = packed_last_used(last_used_idx);
id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
*len = le32_to_cpu(vq->packed.vring.desc[last_used].len);

@@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
ret = vq->packed.desc_state[id].data;
detach_buf_packed(vq, id, ctx);

- vq->last_used_idx += vq->packed.desc_state[id].num;
- if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
- vq->last_used_idx -= vq->packed.vring.num;
- vq->packed.used_wrap_counter ^= 1;
+ last_used += vq->packed.desc_state[id].num;
+ if (unlikely(last_used >= vq->packed.vring.num)) {
+ last_used -= vq->packed.vring.num;
+ used_wrap_counter ^= 1;
}

+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
+ WRITE_ONCE(vq->last_used_idx, last_used);
+
/*
* If we expect an interrupt for the next entry, tell host
* by writing event index and flush out the write before
@@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
virtio_store_mb(vq->weak_barriers,
&vq->packed.vring.driver->off_wrap,
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR)));
+ cpu_to_le16(vq->last_used_idx));

LAST_ADD_TIME_INVALID(vq);

@@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)

if (vq->event) {
vq->packed.vring.driver->off_wrap =
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR));
+ cpu_to_le16(vq->last_used_idx);
/*
* We need to update event offset and event wrap
* counter first before updating event flags.
@@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
}

END_USE(vq);
- return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR);
+ return vq->last_used_idx;
}

static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
@@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 used_idx, wrap_counter;
+ u16 used_idx, wrap_counter, last_used_idx;
u16 bufs;

START_USE(vq);
@@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
if (vq->event) {
/* TODO: tune this threshold */
bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
- wrap_counter = vq->packed.used_wrap_counter;
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ wrap_counter = packed_used_wrap_counter(last_used_idx);

- used_idx = vq->last_used_idx + bufs;
+ used_idx = packed_last_used(last_used_idx) + bufs;
if (used_idx >= vq->packed.vring.num) {
used_idx -= vq->packed.vring.num;
wrap_counter ^= 1;
@@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
*/
virtio_mb(vq->weak_barriers);

- if (is_used_desc_packed(vq,
- vq->last_used_idx,
- vq->packed.used_wrap_counter)) {
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ wrap_counter = packed_used_wrap_counter(last_used_idx);
+ used_idx = packed_last_used(last_used_idx);
+ if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
END_USE(vq);
return false;
}
@@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
vq->notify = notify;
vq->weak_barriers = weak_barriers;
vq->broken = true;
- vq->last_used_idx = 0;
+ vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
vq->event_triggered = false;
vq->num_added = 0;
vq->packed_ring = true;
@@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(

vq->packed.next_avail_idx = 0;
vq->packed.avail_wrap_counter = 1;
- vq->packed.used_wrap_counter = 1;
vq->packed.event_flags_shadow = 0;
vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;

--
2.31.1

2022-06-16 14:36:33

by Michael S. Tsirkin

Subject: Re: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Thu, Jun 16, 2022 at 08:57:36PM +0800, Albert Huang wrote:
> From: "huangjie.albert" <[email protected]>
>
> the used_wrap_counter and the vq->last_used_idx may get
> out of sync if they are separate assignment，and interrupt
> might use an incorrect value to check for the used index.
>
> for example:OOB access
> ksoftirqd may consume the packet and it will call:
> virtnet_poll
> -->virtnet_receive
> -->virtqueue_get_buf_ctx
> -->virtqueue_get_buf_ctx_packed
> and in virtqueue_get_buf_ctx_packed:
>
> vq->last_used_idx += vq->packed.desc_state[id].num;
> if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> vq->last_used_idx -= vq->packed.vring.num;
> vq->packed.used_wrap_counter ^= 1;
> }
>
> if at the same time, there comes a vring interrupt，in vring_interrupt:
> we will call:
> vring_interrupt
> -->more_used
> -->more_used_packed
> -->is_used_desc_packed
> in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> so this could case a memory out of bounds bug.
>
> this patch is to keep the used_wrap_counter in vq->last_used_idx
> so we can get the correct value to check for used index in interrupt.
>
> v3->v4:
> - use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx
>
> v2->v3:
> - add inline function to get used_wrap_counter and last_used
> - when use vq->last_used_idx, only read once
> if vq->last_used_idx is read twice, the values can be inconsistent.
> - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
>
> v1->v2:
> - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> - Remove parameter judgment in is_used_desc_packed,
> because it can't be illegal
>
> Signed-off-by: huangjie.albert <[email protected]>
> ---
> drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> 1 file changed, 47 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 13a7348cedff..719fbbe716d6 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -111,7 +111,12 @@ struct vring_virtqueue {
> /* Number we've added since last sync. */
> unsigned int num_added;
>
> - /* Last used index we've seen. */
> + /* Last used index we've seen.
> + * for split ring, it just contains last used index
> + * for packed ring:
> + * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
> + * bits VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.

bits from VRING_PACKED_EVENT_F_WRAP_CTR

> + */
> u16 last_used_idx;
>
> /* Hint for event idx: already triggered no need to disable. */
> @@ -154,9 +159,6 @@ struct vring_virtqueue {
> /* Driver ring wrap counter. */
> bool avail_wrap_counter;
>
> - /* Device ring wrap counter. */
> - bool used_wrap_counter;
> -
> /* Avail used flags. */
> u16 avail_used_flags;
>
> @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> /*
> * Packed ring specific functions - *_packed().
> */
> +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> +{
> + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> +}
> +
> +static inline u16 packed_last_used(u16 last_used_idx)
> +{
> + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> +}
>
> static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> struct vring_desc_extra *extra)
> @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
>
> static inline bool more_used_packed(const struct vring_virtqueue *vq)
> {
> - return is_used_desc_packed(vq, vq->last_used_idx,
> - vq->packed.used_wrap_counter);
> + u16 last_used;
> + u16 last_used_idx;
> + bool used_wrap_counter;
> +
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + last_used = packed_last_used(last_used_idx);
> + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> }
>
> static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> void **ctx)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> - u16 last_used, id;
> + u16 last_used, id, last_used_idx;
> + bool used_wrap_counter;
> void *ret;
>
> START_USE(vq);
> @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> /* Only get used elements after they have been exposed by host. */
> virtio_rmb(vq->weak_barriers);
>
> - last_used = vq->last_used_idx;
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> + last_used = packed_last_used(last_used_idx);
> id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
>
> @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> ret = vq->packed.desc_state[id].data;
> detach_buf_packed(vq, id, ctx);
>
> - vq->last_used_idx += vq->packed.desc_state[id].num;
> - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> - vq->last_used_idx -= vq->packed.vring.num;
> - vq->packed.used_wrap_counter ^= 1;
> + last_used += vq->packed.desc_state[id].num;
> + if (unlikely(last_used >= vq->packed.vring.num)) {
> + last_used -= vq->packed.vring.num;
> + used_wrap_counter ^= 1;
> }
>
> + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> + WRITE_ONCE(vq->last_used_idx, last_used);
> +
> /*
> * If we expect an interrupt for the next entry, tell host
> * by writing event index and flush out the write before
> @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> virtio_store_mb(vq->weak_barriers,
> &vq->packed.vring.driver->off_wrap,
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR)));
> + cpu_to_le16(vq->last_used_idx));
>
> LAST_ADD_TIME_INVALID(vq);
>
> @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
>
> if (vq->event) {
> vq->packed.vring.driver->off_wrap =
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR));
> + cpu_to_le16(vq->last_used_idx);
> /*
> * We need to update event offset and event wrap
> * counter first before updating event flags.
> @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> }
>
> END_USE(vq);
> - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR);
> + return vq->last_used_idx;
> }
>
> static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> - u16 used_idx, wrap_counter;
> + u16 used_idx, wrap_counter, last_used_idx;
> u16 bufs;
>
> START_USE(vq);
> @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> if (vq->event) {
> /* TODO: tune this threshold */
> bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> - wrap_counter = vq->packed.used_wrap_counter;
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + wrap_counter = packed_used_wrap_counter(last_used_idx);
>
> - used_idx = vq->last_used_idx + bufs;
> + used_idx = packed_last_used(last_used_idx) + bufs;
> if (used_idx >= vq->packed.vring.num) {
> used_idx -= vq->packed.vring.num;
> wrap_counter ^= 1;
> @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> */
> virtio_mb(vq->weak_barriers);
>
> - if (is_used_desc_packed(vq,
> - vq->last_used_idx,
> - vq->packed.used_wrap_counter)) {
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + wrap_counter = packed_used_wrap_counter(last_used_idx);
> + used_idx = packed_last_used(last_used_idx);
> + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> END_USE(vq);
> return false;
> }
> @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> vq->notify = notify;
> vq->weak_barriers = weak_barriers;
> vq->broken = true;
> - vq->last_used_idx = 0;
> + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> vq->event_triggered = false;
> vq->num_added = 0;
> vq->packed_ring = true;
> @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
>
> vq->packed.next_avail_idx = 0;
> vq->packed.avail_wrap_counter = 1;
> - vq->packed.used_wrap_counter = 1;
> vq->packed.event_flags_shadow = 0;
> vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
>
> --
> 2.31.1

2022-06-17 02:22:22

by 黄杰

Subject: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

From: "huangjie.albert" <[email protected]>

the used_wrap_counter and the vq->last_used_idx may get
out of sync if they are separate assignment，and interrupt
might use an incorrect value to check for the used index.

for example:OOB access
ksoftirqd may consume the packet and it will call:
virtnet_poll
-->virtnet_receive
-->virtqueue_get_buf_ctx
-->virtqueue_get_buf_ctx_packed
and in virtqueue_get_buf_ctx_packed:

vq->last_used_idx += vq->packed.desc_state[id].num;
if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
vq->last_used_idx -= vq->packed.vring.num;
vq->packed.used_wrap_counter ^= 1;
}

if at the same time, there comes a vring interrupt，in vring_interrupt:
we will call:
vring_interrupt
-->more_used
-->more_used_packed
-->is_used_desc_packed
in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
so this could case a memory out of bounds bug.

this patch is to keep the used_wrap_counter in vq->last_used_idx
so we can get the correct value to check for used index in interrupt.

v3->v4:
- use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx

v2->v3:
- add inline function to get used_wrap_counter and last_used
- when use vq->last_used_idx, only read once
if vq->last_used_idx is read twice, the values can be inconsistent.
- use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR

v1->v2:
- reuse the VRING_PACKED_EVENT_F_WRAP_CTR
- Remove parameter judgment in is_used_desc_packed,
because it can't be illegal

Signed-off-by: huangjie.albert <[email protected]>
---
drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
1 file changed, 47 insertions(+), 28 deletions(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 13a7348cedff..719fbbe716d6 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -111,7 +111,12 @@ struct vring_virtqueue {
/* Number we've added since last sync. */
unsigned int num_added;

- /* Last used index we've seen. */
+ /* Last used index we've seen.
+ * for split ring, it just contains last used index
+ * for packed ring:
+ * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
+ * bits from VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
+ */
u16 last_used_idx;

/* Hint for event idx: already triggered no need to disable. */
@@ -154,9 +159,6 @@ struct vring_virtqueue {
/* Driver ring wrap counter. */
bool avail_wrap_counter;

- /* Device ring wrap counter. */
- bool used_wrap_counter;
-
/* Avail used flags. */
u16 avail_used_flags;

@@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
/*
* Packed ring specific functions - *_packed().
*/
+static inline bool packed_used_wrap_counter(u16 last_used_idx)
+{
+ return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}
+
+static inline u16 packed_last_used(u16 last_used_idx)
+{
+ return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
+}

static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
struct vring_desc_extra *extra)
@@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,

static inline bool more_used_packed(const struct vring_virtqueue *vq)
{
- return is_used_desc_packed(vq, vq->last_used_idx,
- vq->packed.used_wrap_counter);
+ u16 last_used;
+ u16 last_used_idx;
+ bool used_wrap_counter;
+
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ last_used = packed_last_used(last_used_idx);
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ return is_used_desc_packed(vq, last_used, used_wrap_counter);
}

static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
@@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
void **ctx)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 last_used, id;
+ u16 last_used, id, last_used_idx;
+ bool used_wrap_counter;
void *ret;

START_USE(vq);
@@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
/* Only get used elements after they have been exposed by host. */
virtio_rmb(vq->weak_barriers);

- last_used = vq->last_used_idx;
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ used_wrap_counter = packed_used_wrap_counter(last_used_idx);
+ last_used = packed_last_used(last_used_idx);
id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
*len = le32_to_cpu(vq->packed.vring.desc[last_used].len);

@@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
ret = vq->packed.desc_state[id].data;
detach_buf_packed(vq, id, ctx);

- vq->last_used_idx += vq->packed.desc_state[id].num;
- if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
- vq->last_used_idx -= vq->packed.vring.num;
- vq->packed.used_wrap_counter ^= 1;
+ last_used += vq->packed.desc_state[id].num;
+ if (unlikely(last_used >= vq->packed.vring.num)) {
+ last_used -= vq->packed.vring.num;
+ used_wrap_counter ^= 1;
}

+ last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
+ WRITE_ONCE(vq->last_used_idx, last_used);
+
/*
* If we expect an interrupt for the next entry, tell host
* by writing event index and flush out the write before
@@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
virtio_store_mb(vq->weak_barriers,
&vq->packed.vring.driver->off_wrap,
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR)));
+ cpu_to_le16(vq->last_used_idx));

LAST_ADD_TIME_INVALID(vq);

@@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)

if (vq->event) {
vq->packed.vring.driver->off_wrap =
- cpu_to_le16(vq->last_used_idx |
- (vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR));
+ cpu_to_le16(vq->last_used_idx);
/*
* We need to update event offset and event wrap
* counter first before updating event flags.
@@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
}

END_USE(vq);
- return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
- VRING_PACKED_EVENT_F_WRAP_CTR);
+ return vq->last_used_idx;
}

static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
@@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
{
struct vring_virtqueue *vq = to_vvq(_vq);
- u16 used_idx, wrap_counter;
+ u16 used_idx, wrap_counter, last_used_idx;
u16 bufs;

START_USE(vq);
@@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
if (vq->event) {
/* TODO: tune this threshold */
bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
- wrap_counter = vq->packed.used_wrap_counter;
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ wrap_counter = packed_used_wrap_counter(last_used_idx);

- used_idx = vq->last_used_idx + bufs;
+ used_idx = packed_last_used(last_used_idx) + bufs;
if (used_idx >= vq->packed.vring.num) {
used_idx -= vq->packed.vring.num;
wrap_counter ^= 1;
@@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
*/
virtio_mb(vq->weak_barriers);

- if (is_used_desc_packed(vq,
- vq->last_used_idx,
- vq->packed.used_wrap_counter)) {
+ last_used_idx = READ_ONCE(vq->last_used_idx);
+ wrap_counter = packed_used_wrap_counter(last_used_idx);
+ used_idx = packed_last_used(last_used_idx);
+ if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
END_USE(vq);
return false;
}
@@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
vq->notify = notify;
vq->weak_barriers = weak_barriers;
vq->broken = true;
- vq->last_used_idx = 0;
+ vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
vq->event_triggered = false;
vq->num_added = 0;
vq->packed_ring = true;
@@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(

vq->packed.next_avail_idx = 0;
vq->packed.avail_wrap_counter = 1;
- vq->packed.used_wrap_counter = 1;
vq->packed.event_flags_shadow = 0;
vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;

--
2.31.1

2022-06-22 09:02:25

by Jason Wang

Subject: Re: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Fri, Jun 17, 2022 at 10:04 AM Albert Huang
<[email protected]> wrote:
>
> From: "huangjie.albert" <[email protected]>
>
> the used_wrap_counter and the vq->last_used_idx may get
> out of sync if they are separate assignment，and interrupt
> might use an incorrect value to check for the used index.
>
> for example:OOB access
> ksoftirqd may consume the packet and it will call:
> virtnet_poll
> -->virtnet_receive
> -->virtqueue_get_buf_ctx
> -->virtqueue_get_buf_ctx_packed
> and in virtqueue_get_buf_ctx_packed:
>
> vq->last_used_idx += vq->packed.desc_state[id].num;
> if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> vq->last_used_idx -= vq->packed.vring.num;
> vq->packed.used_wrap_counter ^= 1;
> }
>
> if at the same time, there comes a vring interrupt，in vring_interrupt:
> we will call:
> vring_interrupt
> -->more_used
> -->more_used_packed
> -->is_used_desc_packed
> in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> so this could case a memory out of bounds bug.
>
> this patch is to keep the used_wrap_counter in vq->last_used_idx
> so we can get the correct value to check for used index in interrupt.
>
> v3->v4:
> - use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx
>
> v2->v3:
> - add inline function to get used_wrap_counter and last_used
> - when use vq->last_used_idx, only read once
> if vq->last_used_idx is read twice, the values can be inconsistent.
> - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
>
> v1->v2:
> - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> - Remove parameter judgment in is_used_desc_packed,
> because it can't be illegal
>
> Signed-off-by: huangjie.albert <[email protected]>
> ---
> drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> 1 file changed, 47 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 13a7348cedff..719fbbe716d6 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -111,7 +111,12 @@ struct vring_virtqueue {
> /* Number we've added since last sync. */
> unsigned int num_added;
>
> - /* Last used index we've seen. */
> + /* Last used index we've seen.
> + * for split ring, it just contains last used index
> + * for packed ring:
> + * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
> + * bits from VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
> + */
> u16 last_used_idx;
>
> /* Hint for event idx: already triggered no need to disable. */
> @@ -154,9 +159,6 @@ struct vring_virtqueue {
> /* Driver ring wrap counter. */
> bool avail_wrap_counter;
>
> - /* Device ring wrap counter. */
> - bool used_wrap_counter;
> -
> /* Avail used flags. */
> u16 avail_used_flags;
>
> @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> /*
> * Packed ring specific functions - *_packed().
> */
> +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> +{
> + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> +}
> +
> +static inline u16 packed_last_used(u16 last_used_idx)
> +{
> + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> +}

Any reason we need a minus after the shift?

Others look good.

Thanks

>
> static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> struct vring_desc_extra *extra)
> @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
>
> static inline bool more_used_packed(const struct vring_virtqueue *vq)
> {
> - return is_used_desc_packed(vq, vq->last_used_idx,
> - vq->packed.used_wrap_counter);
> + u16 last_used;
> + u16 last_used_idx;
> + bool used_wrap_counter;
> +
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + last_used = packed_last_used(last_used_idx);
> + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> }
>
> static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> void **ctx)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> - u16 last_used, id;
> + u16 last_used, id, last_used_idx;
> + bool used_wrap_counter;
> void *ret;
>
> START_USE(vq);
> @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> /* Only get used elements after they have been exposed by host. */
> virtio_rmb(vq->weak_barriers);
>
> - last_used = vq->last_used_idx;
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> + last_used = packed_last_used(last_used_idx);
> id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
>
> @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> ret = vq->packed.desc_state[id].data;
> detach_buf_packed(vq, id, ctx);
>
> - vq->last_used_idx += vq->packed.desc_state[id].num;
> - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> - vq->last_used_idx -= vq->packed.vring.num;
> - vq->packed.used_wrap_counter ^= 1;
> + last_used += vq->packed.desc_state[id].num;
> + if (unlikely(last_used >= vq->packed.vring.num)) {
> + last_used -= vq->packed.vring.num;
> + used_wrap_counter ^= 1;
> }
>
> + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> + WRITE_ONCE(vq->last_used_idx, last_used);
> +
> /*
> * If we expect an interrupt for the next entry, tell host
> * by writing event index and flush out the write before
> @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> virtio_store_mb(vq->weak_barriers,
> &vq->packed.vring.driver->off_wrap,
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR)));
> + cpu_to_le16(vq->last_used_idx));
>
> LAST_ADD_TIME_INVALID(vq);
>
> @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
>
> if (vq->event) {
> vq->packed.vring.driver->off_wrap =
> - cpu_to_le16(vq->last_used_idx |
> - (vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR));
> + cpu_to_le16(vq->last_used_idx);
> /*
> * We need to update event offset and event wrap
> * counter first before updating event flags.
> @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> }
>
> END_USE(vq);
> - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> - VRING_PACKED_EVENT_F_WRAP_CTR);
> + return vq->last_used_idx;
> }
>
> static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
> - u16 used_idx, wrap_counter;
> + u16 used_idx, wrap_counter, last_used_idx;
> u16 bufs;
>
> START_USE(vq);
> @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> if (vq->event) {
> /* TODO: tune this threshold */
> bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> - wrap_counter = vq->packed.used_wrap_counter;
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + wrap_counter = packed_used_wrap_counter(last_used_idx);
>
> - used_idx = vq->last_used_idx + bufs;
> + used_idx = packed_last_used(last_used_idx) + bufs;
> if (used_idx >= vq->packed.vring.num) {
> used_idx -= vq->packed.vring.num;
> wrap_counter ^= 1;
> @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> */
> virtio_mb(vq->weak_barriers);
>
> - if (is_used_desc_packed(vq,
> - vq->last_used_idx,
> - vq->packed.used_wrap_counter)) {
> + last_used_idx = READ_ONCE(vq->last_used_idx);
> + wrap_counter = packed_used_wrap_counter(last_used_idx);
> + used_idx = packed_last_used(last_used_idx);
> + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> END_USE(vq);
> return false;
> }
> @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> vq->notify = notify;
> vq->weak_barriers = weak_barriers;
> vq->broken = true;
> - vq->last_used_idx = 0;
> + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> vq->event_triggered = false;
> vq->num_added = 0;
> vq->packed_ring = true;
> @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
>
> vq->packed.next_avail_idx = 0;
> vq->packed.avail_wrap_counter = 1;
> - vq->packed.used_wrap_counter = 1;
> vq->packed.event_flags_shadow = 0;
> vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
>
> --
> 2.31.1
>

2022-06-22 12:22:08

by Michael S. Tsirkin

Subject: Re: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Wed, Jun 22, 2022 at 04:51:22PM +0800, Jason Wang wrote:
> On Fri, Jun 17, 2022 at 10:04 AM Albert Huang
> <[email protected]> wrote:
> >
> > From: "huangjie.albert" <[email protected]>
> >
> > the used_wrap_counter and the vq->last_used_idx may get
> > out of sync if they are separate assignment，and interrupt
> > might use an incorrect value to check for the used index.
> >
> > for example:OOB access
> > ksoftirqd may consume the packet and it will call:
> > virtnet_poll
> > -->virtnet_receive
> > -->virtqueue_get_buf_ctx
> > -->virtqueue_get_buf_ctx_packed
> > and in virtqueue_get_buf_ctx_packed:
> >
> > vq->last_used_idx += vq->packed.desc_state[id].num;
> > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > vq->last_used_idx -= vq->packed.vring.num;
> > vq->packed.used_wrap_counter ^= 1;
> > }
> >
> > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > we will call:
> > vring_interrupt
> > -->more_used
> > -->more_used_packed
> > -->is_used_desc_packed
> > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > so this could case a memory out of bounds bug.
> >
> > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > so we can get the correct value to check for used index in interrupt.
> >
> > v3->v4:
> > - use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx
> >
> > v2->v3:
> > - add inline function to get used_wrap_counter and last_used
> > - when use vq->last_used_idx, only read once
> > if vq->last_used_idx is read twice, the values can be inconsistent.
> > - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> > to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
> >
> > v1->v2:
> > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > - Remove parameter judgment in is_used_desc_packed,
> > because it can't be illegal
> >
> > Signed-off-by: huangjie.albert <[email protected]>
> > ---
> > drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> > 1 file changed, 47 insertions(+), 28 deletions(-)
> >
> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > index 13a7348cedff..719fbbe716d6 100644
> > --- a/drivers/virtio/virtio_ring.c
> > +++ b/drivers/virtio/virtio_ring.c
> > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > /* Number we've added since last sync. */
> > unsigned int num_added;
> >
> > - /* Last used index we've seen. */
> > + /* Last used index we've seen.
> > + * for split ring, it just contains last used index
> > + * for packed ring:
> > + * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
> > + * bits from VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
> > + */
> > u16 last_used_idx;
> >
> > /* Hint for event idx: already triggered no need to disable. */
> > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > /* Driver ring wrap counter. */
> > bool avail_wrap_counter;
> >
> > - /* Device ring wrap counter. */
> > - bool used_wrap_counter;
> > -
> > /* Avail used flags. */
> > u16 avail_used_flags;
> >
> > @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> > /*
> > * Packed ring specific functions - *_packed().
> > */
> > +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> > +{
> > + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > +}
> > +
> > +static inline u16 packed_last_used(u16 last_used_idx)
> > +{
> > + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > +}
>
> Any reason we need a minus after the shift?

The point is to say "all bits above VRING_PACKED_EVENT_F_WRAP_CTR".
Has no effect currently but will if last_used_idx is extended to 32 bit.


> Others look good.
>
> Thanks
>
> >
> > static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> > struct vring_desc_extra *extra)
> > @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> >
> > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > {
> > - return is_used_desc_packed(vq, vq->last_used_idx,
> > - vq->packed.used_wrap_counter);
> > + u16 last_used;
> > + u16 last_used_idx;
> > + bool used_wrap_counter;
> > +
> > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > + last_used = packed_last_used(last_used_idx);
> > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > }
> >
> > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > void **ctx)
> > {
> > struct vring_virtqueue *vq = to_vvq(_vq);
> > - u16 last_used, id;
> > + u16 last_used, id, last_used_idx;
> > + bool used_wrap_counter;
> > void *ret;
> >
> > START_USE(vq);
> > @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > /* Only get used elements after they have been exposed by host. */
> > virtio_rmb(vq->weak_barriers);
> >
> > - last_used = vq->last_used_idx;
> > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > + last_used = packed_last_used(last_used_idx);
> > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> >
> > @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > ret = vq->packed.desc_state[id].data;
> > detach_buf_packed(vq, id, ctx);
> >
> > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > - vq->last_used_idx -= vq->packed.vring.num;
> > - vq->packed.used_wrap_counter ^= 1;
> > + last_used += vq->packed.desc_state[id].num;
> > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > + last_used -= vq->packed.vring.num;
> > + used_wrap_counter ^= 1;
> > }
> >
> > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > + WRITE_ONCE(vq->last_used_idx, last_used);
> > +
> > /*
> > * If we expect an interrupt for the next entry, tell host
> > * by writing event index and flush out the write before
> > @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > virtio_store_mb(vq->weak_barriers,
> > &vq->packed.vring.driver->off_wrap,
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > + cpu_to_le16(vq->last_used_idx));
> >
> > LAST_ADD_TIME_INVALID(vq);
> >
> > @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> >
> > if (vq->event) {
> > vq->packed.vring.driver->off_wrap =
> > - cpu_to_le16(vq->last_used_idx |
> > - (vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > + cpu_to_le16(vq->last_used_idx);
> > /*
> > * We need to update event offset and event wrap
> > * counter first before updating event flags.
> > @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > }
> >
> > END_USE(vq);
> > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > + return vq->last_used_idx;
> > }
> >
> > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > {
> > struct vring_virtqueue *vq = to_vvq(_vq);
> > - u16 used_idx, wrap_counter;
> > + u16 used_idx, wrap_counter, last_used_idx;
> > u16 bufs;
> >
> > START_USE(vq);
> > @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > if (vq->event) {
> > /* TODO: tune this threshold */
> > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > - wrap_counter = vq->packed.used_wrap_counter;
> > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> >
> > - used_idx = vq->last_used_idx + bufs;
> > + used_idx = packed_last_used(last_used_idx) + bufs;
> > if (used_idx >= vq->packed.vring.num) {
> > used_idx -= vq->packed.vring.num;
> > wrap_counter ^= 1;
> > @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > */
> > virtio_mb(vq->weak_barriers);
> >
> > - if (is_used_desc_packed(vq,
> > - vq->last_used_idx,
> > - vq->packed.used_wrap_counter)) {
> > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > + used_idx = packed_last_used(last_used_idx);
> > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > END_USE(vq);
> > return false;
> > }
> > @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > vq->notify = notify;
> > vq->weak_barriers = weak_barriers;
> > vq->broken = true;
> > - vq->last_used_idx = 0;
> > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > vq->event_triggered = false;
> > vq->num_added = 0;
> > vq->packed_ring = true;
> > @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> >
> > vq->packed.next_avail_idx = 0;
> > vq->packed.avail_wrap_counter = 1;
> > - vq->packed.used_wrap_counter = 1;
> > vq->packed.event_flags_shadow = 0;
> > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> >
> > --
> > 2.31.1
> >

2022-06-23 01:42:47

by Jason Wang

Subject: Re: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Wed, Jun 22, 2022 at 8:16 PM Michael S. Tsirkin <[email protected]> wrote:
>
> On Wed, Jun 22, 2022 at 04:51:22PM +0800, Jason Wang wrote:
> > On Fri, Jun 17, 2022 at 10:04 AM Albert Huang
> > <[email protected]> wrote:
> > >
> > > From: "huangjie.albert" <[email protected]>
> > >
> > > the used_wrap_counter and the vq->last_used_idx may get
> > > out of sync if they are separate assignment，and interrupt
> > > might use an incorrect value to check for the used index.
> > >
> > > for example:OOB access
> > > ksoftirqd may consume the packet and it will call:
> > > virtnet_poll
> > > -->virtnet_receive
> > > -->virtqueue_get_buf_ctx
> > > -->virtqueue_get_buf_ctx_packed
> > > and in virtqueue_get_buf_ctx_packed:
> > >
> > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > vq->last_used_idx -= vq->packed.vring.num;
> > > vq->packed.used_wrap_counter ^= 1;
> > > }
> > >
> > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > we will call:
> > > vring_interrupt
> > > -->more_used
> > > -->more_used_packed
> > > -->is_used_desc_packed
> > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > so this could case a memory out of bounds bug.
> > >
> > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > so we can get the correct value to check for used index in interrupt.
> > >
> > > v3->v4:
> > > - use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx
> > >
> > > v2->v3:
> > > - add inline function to get used_wrap_counter and last_used
> > > - when use vq->last_used_idx, only read once
> > > if vq->last_used_idx is read twice, the values can be inconsistent.
> > > - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> > > to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
> > >
> > > v1->v2:
> > > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > > - Remove parameter judgment in is_used_desc_packed,
> > > because it can't be illegal
> > >
> > > Signed-off-by: huangjie.albert <[email protected]>
> > > ---
> > > drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> > > 1 file changed, 47 insertions(+), 28 deletions(-)
> > >
> > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > index 13a7348cedff..719fbbe716d6 100644
> > > --- a/drivers/virtio/virtio_ring.c
> > > +++ b/drivers/virtio/virtio_ring.c
> > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > /* Number we've added since last sync. */
> > > unsigned int num_added;
> > >
> > > - /* Last used index we've seen. */
> > > + /* Last used index we've seen.
> > > + * for split ring, it just contains last used index
> > > + * for packed ring:
> > > + * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
> > > + * bits from VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
> > > + */
> > > u16 last_used_idx;
> > >
> > > /* Hint for event idx: already triggered no need to disable. */
> > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > /* Driver ring wrap counter. */
> > > bool avail_wrap_counter;
> > >
> > > - /* Device ring wrap counter. */
> > > - bool used_wrap_counter;
> > > -
> > > /* Avail used flags. */
> > > u16 avail_used_flags;
> > >
> > > @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> > > /*
> > > * Packed ring specific functions - *_packed().
> > > */
> > > +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> > > +{
> > > + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > +}
> > > +
> > > +static inline u16 packed_last_used(u16 last_used_idx)
> > > +{
> > > + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > +}
> >
> > Any reason we need a minus after the shift?
>
> The point is to say "all bits above VRING_PACKED_EVENT_F_WRAP_CTR".
> Has no effect currently but will if last_used_idx is extended to 32 bit.

Ok, but we don't do this for other uses for VRING_PACKED_EVENT_F_WRAP_CTR.

I wonder how much value we do it only here.

Thanks

>
>
> > Others look good.
> >
> > Thanks
> >
> > >
> > > static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> > > struct vring_desc_extra *extra)
> > > @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > >
> > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > {
> > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > - vq->packed.used_wrap_counter);
> > > + u16 last_used;
> > > + u16 last_used_idx;
> > > + bool used_wrap_counter;
> > > +
> > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > + last_used = packed_last_used(last_used_idx);
> > > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > }
> > >
> > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > void **ctx)
> > > {
> > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > - u16 last_used, id;
> > > + u16 last_used, id, last_used_idx;
> > > + bool used_wrap_counter;
> > > void *ret;
> > >
> > > START_USE(vq);
> > > @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > /* Only get used elements after they have been exposed by host. */
> > > virtio_rmb(vq->weak_barriers);
> > >
> > > - last_used = vq->last_used_idx;
> > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > + last_used = packed_last_used(last_used_idx);
> > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > >
> > > @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > ret = vq->packed.desc_state[id].data;
> > > detach_buf_packed(vq, id, ctx);
> > >
> > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > - vq->last_used_idx -= vq->packed.vring.num;
> > > - vq->packed.used_wrap_counter ^= 1;
> > > + last_used += vq->packed.desc_state[id].num;
> > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > > + last_used -= vq->packed.vring.num;
> > > + used_wrap_counter ^= 1;
> > > }
> > >
> > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + WRITE_ONCE(vq->last_used_idx, last_used);
> > > +
> > > /*
> > > * If we expect an interrupt for the next entry, tell host
> > > * by writing event index and flush out the write before
> > > @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > virtio_store_mb(vq->weak_barriers,
> > > &vq->packed.vring.driver->off_wrap,
> > > - cpu_to_le16(vq->last_used_idx |
> > > - (vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > + cpu_to_le16(vq->last_used_idx));
> > >
> > > LAST_ADD_TIME_INVALID(vq);
> > >
> > > @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > >
> > > if (vq->event) {
> > > vq->packed.vring.driver->off_wrap =
> > > - cpu_to_le16(vq->last_used_idx |
> > > - (vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > + cpu_to_le16(vq->last_used_idx);
> > > /*
> > > * We need to update event offset and event wrap
> > > * counter first before updating event flags.
> > > @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > }
> > >
> > > END_USE(vq);
> > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > + return vq->last_used_idx;
> > > }
> > >
> > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > {
> > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > - u16 used_idx, wrap_counter;
> > > + u16 used_idx, wrap_counter, last_used_idx;
> > > u16 bufs;
> > >
> > > START_USE(vq);
> > > @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > if (vq->event) {
> > > /* TODO: tune this threshold */
> > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > - wrap_counter = vq->packed.used_wrap_counter;
> > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > >
> > > - used_idx = vq->last_used_idx + bufs;
> > > + used_idx = packed_last_used(last_used_idx) + bufs;
> > > if (used_idx >= vq->packed.vring.num) {
> > > used_idx -= vq->packed.vring.num;
> > > wrap_counter ^= 1;
> > > @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > */
> > > virtio_mb(vq->weak_barriers);
> > >
> > > - if (is_used_desc_packed(vq,
> > > - vq->last_used_idx,
> > > - vq->packed.used_wrap_counter)) {
> > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > + used_idx = packed_last_used(last_used_idx);
> > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > END_USE(vq);
> > > return false;
> > > }
> > > @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > vq->notify = notify;
> > > vq->weak_barriers = weak_barriers;
> > > vq->broken = true;
> > > - vq->last_used_idx = 0;
> > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > > vq->event_triggered = false;
> > > vq->num_added = 0;
> > > vq->packed_ring = true;
> > > @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > >
> > > vq->packed.next_avail_idx = 0;
> > > vq->packed.avail_wrap_counter = 1;
> > > - vq->packed.used_wrap_counter = 1;
> > > vq->packed.event_flags_shadow = 0;
> > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > >
> > > --
> > > 2.31.1
> > >
>

2022-06-24 06:27:35

by Michael S. Tsirkin

Subject: Re: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Thu, Jun 23, 2022 at 09:30:47AM +0800, Jason Wang wrote:
> On Wed, Jun 22, 2022 at 8:16 PM Michael S. Tsirkin <[email protected]> wrote:
> >
> > On Wed, Jun 22, 2022 at 04:51:22PM +0800, Jason Wang wrote:
> > > On Fri, Jun 17, 2022 at 10:04 AM Albert Huang
> > > <[email protected]> wrote:
> > > >
> > > > From: "huangjie.albert" <[email protected]>
> > > >
> > > > the used_wrap_counter and the vq->last_used_idx may get
> > > > out of sync if they are separate assignment，and interrupt
> > > > might use an incorrect value to check for the used index.
> > > >
> > > > for example:OOB access
> > > > ksoftirqd may consume the packet and it will call:
> > > > virtnet_poll
> > > > -->virtnet_receive
> > > > -->virtqueue_get_buf_ctx
> > > > -->virtqueue_get_buf_ctx_packed
> > > > and in virtqueue_get_buf_ctx_packed:
> > > >
> > > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > vq->last_used_idx -= vq->packed.vring.num;
> > > > vq->packed.used_wrap_counter ^= 1;
> > > > }
> > > >
> > > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > > we will call:
> > > > vring_interrupt
> > > > -->more_used
> > > > -->more_used_packed
> > > > -->is_used_desc_packed
> > > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > > so this could case a memory out of bounds bug.
> > > >
> > > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > > so we can get the correct value to check for used index in interrupt.
> > > >
> > > > v3->v4:
> > > > - use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx
> > > >
> > > > v2->v3:
> > > > - add inline function to get used_wrap_counter and last_used
> > > > - when use vq->last_used_idx, only read once
> > > > if vq->last_used_idx is read twice, the values can be inconsistent.
> > > > - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> > > > to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
> > > >
> > > > v1->v2:
> > > > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > > > - Remove parameter judgment in is_used_desc_packed,
> > > > because it can't be illegal
> > > >
> > > > Signed-off-by: huangjie.albert <[email protected]>
> > > > ---
> > > > drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> > > > 1 file changed, 47 insertions(+), 28 deletions(-)
> > > >
> > > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > > index 13a7348cedff..719fbbe716d6 100644
> > > > --- a/drivers/virtio/virtio_ring.c
> > > > +++ b/drivers/virtio/virtio_ring.c
> > > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > > /* Number we've added since last sync. */
> > > > unsigned int num_added;
> > > >
> > > > - /* Last used index we've seen. */
> > > > + /* Last used index we've seen.
> > > > + * for split ring, it just contains last used index
> > > > + * for packed ring:
> > > > + * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
> > > > + * bits from VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
> > > > + */
> > > > u16 last_used_idx;
> > > >
> > > > /* Hint for event idx: already triggered no need to disable. */
> > > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > > /* Driver ring wrap counter. */
> > > > bool avail_wrap_counter;
> > > >
> > > > - /* Device ring wrap counter. */
> > > > - bool used_wrap_counter;
> > > > -
> > > > /* Avail used flags. */
> > > > u16 avail_used_flags;
> > > >
> > > > @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> > > > /*
> > > > * Packed ring specific functions - *_packed().
> > > > */
> > > > +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> > > > +{
> > > > + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > +}
> > > > +
> > > > +static inline u16 packed_last_used(u16 last_used_idx)
> > > > +{
> > > > + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > +}
> > >
> > > Any reason we need a minus after the shift?
> >
> > The point is to say "all bits above VRING_PACKED_EVENT_F_WRAP_CTR".
> > Has no effect currently but will if last_used_idx is extended to 32 bit.
>
> Ok, but we don't do this for other uses for VRING_PACKED_EVENT_F_WRAP_CTR.
>
> I wonder how much value we do it only here.
>
> Thanks

I don't care much either way. Feel free to go ahead and play with
different versions so see which works better.

> >
> >
> > > Others look good.
> > >
> > > Thanks
> > >
> > > >
> > > > static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> > > > struct vring_desc_extra *extra)
> > > > @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > >
> > > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > > {
> > > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > > - vq->packed.used_wrap_counter);
> > > > + u16 last_used;
> > > > + u16 last_used_idx;
> > > > + bool used_wrap_counter;
> > > > +
> > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > + last_used = packed_last_used(last_used_idx);
> > > > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > > }
> > > >
> > > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > void **ctx)
> > > > {
> > > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > > - u16 last_used, id;
> > > > + u16 last_used, id, last_used_idx;
> > > > + bool used_wrap_counter;
> > > > void *ret;
> > > >
> > > > START_USE(vq);
> > > > @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > /* Only get used elements after they have been exposed by host. */
> > > > virtio_rmb(vq->weak_barriers);
> > > >
> > > > - last_used = vq->last_used_idx;
> > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > + last_used = packed_last_used(last_used_idx);
> > > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > > >
> > > > @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > ret = vq->packed.desc_state[id].data;
> > > > detach_buf_packed(vq, id, ctx);
> > > >
> > > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > - vq->last_used_idx -= vq->packed.vring.num;
> > > > - vq->packed.used_wrap_counter ^= 1;
> > > > + last_used += vq->packed.desc_state[id].num;
> > > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > > > + last_used -= vq->packed.vring.num;
> > > > + used_wrap_counter ^= 1;
> > > > }
> > > >
> > > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > + WRITE_ONCE(vq->last_used_idx, last_used);
> > > > +
> > > > /*
> > > > * If we expect an interrupt for the next entry, tell host
> > > > * by writing event index and flush out the write before
> > > > @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > > virtio_store_mb(vq->weak_barriers,
> > > > &vq->packed.vring.driver->off_wrap,
> > > > - cpu_to_le16(vq->last_used_idx |
> > > > - (vq->packed.used_wrap_counter <<
> > > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > > + cpu_to_le16(vq->last_used_idx));
> > > >
> > > > LAST_ADD_TIME_INVALID(vq);
> > > >
> > > > @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > >
> > > > if (vq->event) {
> > > > vq->packed.vring.driver->off_wrap =
> > > > - cpu_to_le16(vq->last_used_idx |
> > > > - (vq->packed.used_wrap_counter <<
> > > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > + cpu_to_le16(vq->last_used_idx);
> > > > /*
> > > > * We need to update event offset and event wrap
> > > > * counter first before updating event flags.
> > > > @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > > }
> > > >
> > > > END_USE(vq);
> > > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > > + return vq->last_used_idx;
> > > > }
> > > >
> > > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > > @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > > static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > {
> > > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > > - u16 used_idx, wrap_counter;
> > > > + u16 used_idx, wrap_counter, last_used_idx;
> > > > u16 bufs;
> > > >
> > > > START_USE(vq);
> > > > @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > if (vq->event) {
> > > > /* TODO: tune this threshold */
> > > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > > - wrap_counter = vq->packed.used_wrap_counter;
> > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > >
> > > > - used_idx = vq->last_used_idx + bufs;
> > > > + used_idx = packed_last_used(last_used_idx) + bufs;
> > > > if (used_idx >= vq->packed.vring.num) {
> > > > used_idx -= vq->packed.vring.num;
> > > > wrap_counter ^= 1;
> > > > @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > */
> > > > virtio_mb(vq->weak_barriers);
> > > >
> > > > - if (is_used_desc_packed(vq,
> > > > - vq->last_used_idx,
> > > > - vq->packed.used_wrap_counter)) {
> > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > + used_idx = packed_last_used(last_used_idx);
> > > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > > END_USE(vq);
> > > > return false;
> > > > }
> > > > @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > > vq->notify = notify;
> > > > vq->weak_barriers = weak_barriers;
> > > > vq->broken = true;
> > > > - vq->last_used_idx = 0;
> > > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > > > vq->event_triggered = false;
> > > > vq->num_added = 0;
> > > > vq->packed_ring = true;
> > > > @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > >
> > > > vq->packed.next_avail_idx = 0;
> > > > vq->packed.avail_wrap_counter = 1;
> > > > - vq->packed.used_wrap_counter = 1;
> > > > vq->packed.event_flags_shadow = 0;
> > > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > > >
> > > > --
> > > > 2.31.1
> > > >
> >

2022-06-27 02:53:54

by Jason Wang

Subject: Re: [PATCH v4] virtio_ring : keep used_wrap_counter in vq->last_used_idx

On Fri, Jun 24, 2022 at 2:23 PM Michael S. Tsirkin <[email protected]> wrote:
>
> On Thu, Jun 23, 2022 at 09:30:47AM +0800, Jason Wang wrote:
> > On Wed, Jun 22, 2022 at 8:16 PM Michael S. Tsirkin <[email protected]> wrote:
> > >
> > > On Wed, Jun 22, 2022 at 04:51:22PM +0800, Jason Wang wrote:
> > > > On Fri, Jun 17, 2022 at 10:04 AM Albert Huang
> > > > <[email protected]> wrote:
> > > > >
> > > > > From: "huangjie.albert" <[email protected]>
> > > > >
> > > > > the used_wrap_counter and the vq->last_used_idx may get
> > > > > out of sync if they are separate assignment，and interrupt
> > > > > might use an incorrect value to check for the used index.
> > > > >
> > > > > for example:OOB access
> > > > > ksoftirqd may consume the packet and it will call:
> > > > > virtnet_poll
> > > > > -->virtnet_receive
> > > > > -->virtqueue_get_buf_ctx
> > > > > -->virtqueue_get_buf_ctx_packed
> > > > > and in virtqueue_get_buf_ctx_packed:
> > > > >
> > > > > vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > > if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > > vq->last_used_idx -= vq->packed.vring.num;
> > > > > vq->packed.used_wrap_counter ^= 1;
> > > > > }
> > > > >
> > > > > if at the same time, there comes a vring interrupt，in vring_interrupt:
> > > > > we will call:
> > > > > vring_interrupt
> > > > > -->more_used
> > > > > -->more_used_packed
> > > > > -->is_used_desc_packed
> > > > > in is_used_desc_packed, the last_used_idx maybe >= vq->packed.vring.num.
> > > > > so this could case a memory out of bounds bug.
> > > > >
> > > > > this patch is to keep the used_wrap_counter in vq->last_used_idx
> > > > > so we can get the correct value to check for used index in interrupt.
> > > > >
> > > > > v3->v4:
> > > > > - use READ_ONCE/WRITE_ONCE to get/set vq->last_used_idx
> > > > >
> > > > > v2->v3:
> > > > > - add inline function to get used_wrap_counter and last_used
> > > > > - when use vq->last_used_idx, only read once
> > > > > if vq->last_used_idx is read twice, the values can be inconsistent.
> > > > > - use last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR))
> > > > > to get the all bits below VRING_PACKED_EVENT_F_WRAP_CTR
> > > > >
> > > > > v1->v2:
> > > > > - reuse the VRING_PACKED_EVENT_F_WRAP_CTR
> > > > > - Remove parameter judgment in is_used_desc_packed,
> > > > > because it can't be illegal
> > > > >
> > > > > Signed-off-by: huangjie.albert <[email protected]>
> > > > > ---
> > > > > drivers/virtio/virtio_ring.c | 75 ++++++++++++++++++++++--------------
> > > > > 1 file changed, 47 insertions(+), 28 deletions(-)
> > > > >
> > > > > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > > > > index 13a7348cedff..719fbbe716d6 100644
> > > > > --- a/drivers/virtio/virtio_ring.c
> > > > > +++ b/drivers/virtio/virtio_ring.c
> > > > > @@ -111,7 +111,12 @@ struct vring_virtqueue {
> > > > > /* Number we've added since last sync. */
> > > > > unsigned int num_added;
> > > > >
> > > > > - /* Last used index we've seen. */
> > > > > + /* Last used index we've seen.
> > > > > + * for split ring, it just contains last used index
> > > > > + * for packed ring:
> > > > > + * bits up to VRING_PACKED_EVENT_F_WRAP_CTR include the last used index.
> > > > > + * bits from VRING_PACKED_EVENT_F_WRAP_CTR include the used wrap counter.
> > > > > + */
> > > > > u16 last_used_idx;
> > > > >
> > > > > /* Hint for event idx: already triggered no need to disable. */
> > > > > @@ -154,9 +159,6 @@ struct vring_virtqueue {
> > > > > /* Driver ring wrap counter. */
> > > > > bool avail_wrap_counter;
> > > > >
> > > > > - /* Device ring wrap counter. */
> > > > > - bool used_wrap_counter;
> > > > > -
> > > > > /* Avail used flags. */
> > > > > u16 avail_used_flags;
> > > > >
> > > > > @@ -973,6 +975,15 @@ static struct virtqueue *vring_create_virtqueue_split(
> > > > > /*
> > > > > * Packed ring specific functions - *_packed().
> > > > > */
> > > > > +static inline bool packed_used_wrap_counter(u16 last_used_idx)
> > > > > +{
> > > > > + return !!(last_used_idx & (1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > > +}
> > > > > +
> > > > > +static inline u16 packed_last_used(u16 last_used_idx)
> > > > > +{
> > > > > + return last_used_idx & ~(-(1 << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > > +}
> > > >
> > > > Any reason we need a minus after the shift?
> > >
> > > The point is to say "all bits above VRING_PACKED_EVENT_F_WRAP_CTR".
> > > Has no effect currently but will if last_used_idx is extended to 32 bit.
> >
> > Ok, but we don't do this for other uses for VRING_PACKED_EVENT_F_WRAP_CTR.
> >
> > I wonder how much value we do it only here.
> >
> > Thanks
>
> I don't care much either way. Feel free to go ahead and play with
> different versions so see which works better.

Ok, I'm fine with either. So

Acked-by: Jason Wang <[email protected]>

>
> > >
> > >
> > > > Others look good.
> > > >
> > > > Thanks
> > > >
> > > > >
> > > > > static void vring_unmap_extra_packed(const struct vring_virtqueue *vq,
> > > > > struct vring_desc_extra *extra)
> > > > > @@ -1406,8 +1417,14 @@ static inline bool is_used_desc_packed(const struct vring_virtqueue *vq,
> > > > >
> > > > > static inline bool more_used_packed(const struct vring_virtqueue *vq)
> > > > > {
> > > > > - return is_used_desc_packed(vq, vq->last_used_idx,
> > > > > - vq->packed.used_wrap_counter);
> > > > > + u16 last_used;
> > > > > + u16 last_used_idx;
> > > > > + bool used_wrap_counter;
> > > > > +
> > > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > > + last_used = packed_last_used(last_used_idx);
> > > > > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > > + return is_used_desc_packed(vq, last_used, used_wrap_counter);
> > > > > }
> > > > >
> > > > > static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > @@ -1415,7 +1432,8 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > void **ctx)
> > > > > {
> > > > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > > > - u16 last_used, id;
> > > > > + u16 last_used, id, last_used_idx;
> > > > > + bool used_wrap_counter;
> > > > > void *ret;
> > > > >
> > > > > START_USE(vq);
> > > > > @@ -1434,7 +1452,9 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > /* Only get used elements after they have been exposed by host. */
> > > > > virtio_rmb(vq->weak_barriers);
> > > > >
> > > > > - last_used = vq->last_used_idx;
> > > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > > + used_wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > > + last_used = packed_last_used(last_used_idx);
> > > > > id = le16_to_cpu(vq->packed.vring.desc[last_used].id);
> > > > > *len = le32_to_cpu(vq->packed.vring.desc[last_used].len);
> > > > >
> > > > > @@ -1451,12 +1471,15 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > ret = vq->packed.desc_state[id].data;
> > > > > detach_buf_packed(vq, id, ctx);
> > > > >
> > > > > - vq->last_used_idx += vq->packed.desc_state[id].num;
> > > > > - if (unlikely(vq->last_used_idx >= vq->packed.vring.num)) {
> > > > > - vq->last_used_idx -= vq->packed.vring.num;
> > > > > - vq->packed.used_wrap_counter ^= 1;
> > > > > + last_used += vq->packed.desc_state[id].num;
> > > > > + if (unlikely(last_used >= vq->packed.vring.num)) {
> > > > > + last_used -= vq->packed.vring.num;
> > > > > + used_wrap_counter ^= 1;
> > > > > }
> > > > >
> > > > > + last_used = (last_used | (used_wrap_counter << VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > > + WRITE_ONCE(vq->last_used_idx, last_used);
> > > > > +
> > > > > /*
> > > > > * If we expect an interrupt for the next entry, tell host
> > > > > * by writing event index and flush out the write before
> > > > > @@ -1465,9 +1488,7 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
> > > > > if (vq->packed.event_flags_shadow == VRING_PACKED_EVENT_FLAG_DESC)
> > > > > virtio_store_mb(vq->weak_barriers,
> > > > > &vq->packed.vring.driver->off_wrap,
> > > > > - cpu_to_le16(vq->last_used_idx |
> > > > > - (vq->packed.used_wrap_counter <<
> > > > > - VRING_PACKED_EVENT_F_WRAP_CTR)));
> > > > > + cpu_to_le16(vq->last_used_idx));
> > > > >
> > > > > LAST_ADD_TIME_INVALID(vq);
> > > > >
> > > > > @@ -1499,9 +1520,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > > >
> > > > > if (vq->event) {
> > > > > vq->packed.vring.driver->off_wrap =
> > > > > - cpu_to_le16(vq->last_used_idx |
> > > > > - (vq->packed.used_wrap_counter <<
> > > > > - VRING_PACKED_EVENT_F_WRAP_CTR));
> > > > > + cpu_to_le16(vq->last_used_idx);
> > > > > /*
> > > > > * We need to update event offset and event wrap
> > > > > * counter first before updating event flags.
> > > > > @@ -1518,8 +1537,7 @@ static unsigned int virtqueue_enable_cb_prepare_packed(struct virtqueue *_vq)
> > > > > }
> > > > >
> > > > > END_USE(vq);
> > > > > - return vq->last_used_idx | ((u16)vq->packed.used_wrap_counter <<
> > > > > - VRING_PACKED_EVENT_F_WRAP_CTR);
> > > > > + return vq->last_used_idx;
> > > > > }
> > > > >
> > > > > static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > > > @@ -1537,7 +1555,7 @@ static bool virtqueue_poll_packed(struct virtqueue *_vq, u16 off_wrap)
> > > > > static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > > {
> > > > > struct vring_virtqueue *vq = to_vvq(_vq);
> > > > > - u16 used_idx, wrap_counter;
> > > > > + u16 used_idx, wrap_counter, last_used_idx;
> > > > > u16 bufs;
> > > > >
> > > > > START_USE(vq);
> > > > > @@ -1550,9 +1568,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > > if (vq->event) {
> > > > > /* TODO: tune this threshold */
> > > > > bufs = (vq->packed.vring.num - vq->vq.num_free) * 3 / 4;
> > > > > - wrap_counter = vq->packed.used_wrap_counter;
> > > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > >
> > > > > - used_idx = vq->last_used_idx + bufs;
> > > > > + used_idx = packed_last_used(last_used_idx) + bufs;
> > > > > if (used_idx >= vq->packed.vring.num) {
> > > > > used_idx -= vq->packed.vring.num;
> > > > > wrap_counter ^= 1;
> > > > > @@ -1582,9 +1601,10 @@ static bool virtqueue_enable_cb_delayed_packed(struct virtqueue *_vq)
> > > > > */
> > > > > virtio_mb(vq->weak_barriers);
> > > > >
> > > > > - if (is_used_desc_packed(vq,
> > > > > - vq->last_used_idx,
> > > > > - vq->packed.used_wrap_counter)) {
> > > > > + last_used_idx = READ_ONCE(vq->last_used_idx);
> > > > > + wrap_counter = packed_used_wrap_counter(last_used_idx);
> > > > > + used_idx = packed_last_used(last_used_idx);
> > > > > + if (is_used_desc_packed(vq, used_idx, wrap_counter)) {
> > > > > END_USE(vq);
> > > > > return false;
> > > > > }
> > > > > @@ -1689,7 +1709,7 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > > > vq->notify = notify;
> > > > > vq->weak_barriers = weak_barriers;
> > > > > vq->broken = true;
> > > > > - vq->last_used_idx = 0;
> > > > > + vq->last_used_idx = 0 | (1 << VRING_PACKED_EVENT_F_WRAP_CTR);
> > > > > vq->event_triggered = false;
> > > > > vq->num_added = 0;
> > > > > vq->packed_ring = true;
> > > > > @@ -1720,7 +1740,6 @@ static struct virtqueue *vring_create_virtqueue_packed(
> > > > >
> > > > > vq->packed.next_avail_idx = 0;
> > > > > vq->packed.avail_wrap_counter = 1;
> > > > > - vq->packed.used_wrap_counter = 1;
> > > > > vq->packed.event_flags_shadow = 0;
> > > > > vq->packed.avail_used_flags = 1 << VRING_PACKED_DESC_F_AVAIL;
> > > > >
> > > > > --
> > > > > 2.31.1
> > > > >
> > >
>