Re: [RFC PATCH 01/17] virtio_ring: Avoid reading unneeded used length

Re: [RFC PATCH 01/17] virtio_ring: Avoid reading unneeded used length

[Johannes Berg]

On Mon, 2021-05-17 at 17:08 +0800, Xie Yongji wrote:
> If device driver doesn't need used length, it can pass a NULL
> len in virtqueue_get_buf()/virtqueue_get_buf_ctx().
> 

Well, actually, it can't right now?

You should probably rephrase this, saying something like

   Allow passing NULL to len in ... if the device driver doesn't need
   the length, and don't read it in that case.

or so?

>  Then
> we can avoid reading and validating the len value in used
> ring entries.

Not sure what that "validating" is about, I only see reading?

johannes

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 17/17] virtio_ring: Add validation for used length

[Michael S. Tsirkin]

On Mon, May 17, 2021 at 05:08:36PM +0800, Xie Yongji wrote:
> This adds validation for used length (might come
> from an untrusted device) when it will be used by
> virtio device driver.
> 
> Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
> ---
>  drivers/virtio/virtio_ring.c | 22 +++++++++++++++++++---
>  1 file changed, 19 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index d999a1d6d271..7d4845d06f21 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -68,11 +68,13 @@
>  struct vring_desc_state_split {
>  	void *data;			/* Data for callback. */
>  	struct vring_desc *indir_desc;	/* Indirect descriptor, if any. */
> +	u32 in_len;			/* Total length of writable buffer */
>  };
>  
>  struct vring_desc_state_packed {
>  	void *data;			/* Data for callback. */
>  	struct vring_packed_desc *indir_desc; /* Indirect descriptor, if any. */
> +	u32 in_len;			/* Total length of writable buffer */
>  	u16 num;			/* Descriptor list length. */
>  	u16 last;			/* The last desc state in a list. */
>  };


Hmm for packed it's aligned to 64 bit anyway, so we are not making it
any worse. But for split this pushes struct size up by 1/3 increasing
cache pressure.


> @@ -486,7 +488,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>  	struct vring_virtqueue *vq = to_vvq(_vq);
>  	struct scatterlist *sg;
>  	struct vring_desc *desc;
> -	unsigned int i, n, avail, descs_used, prev, err_idx;
> +	unsigned int i, n, avail, descs_used, prev, err_idx, in_len = 0;
>  	int head;
>  	bool indirect;
>  
> @@ -570,6 +572,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>  						     VRING_DESC_F_NEXT |
>  						     VRING_DESC_F_WRITE,
>  						     indirect);
> +			in_len += sg->length;
>  		}
>  	}
>  	/* Last one doesn't continue. */
> @@ -604,6 +607,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>  
>  	/* Store token and indirect buffer state. */
>  	vq->split.desc_state[head].data = data;
> +	vq->split.desc_state[head].in_len = in_len;
>  	if (indirect)
>  		vq->split.desc_state[head].indir_desc = desc;
>  	else
> @@ -784,6 +788,10 @@ static void *virtqueue_get_buf_ctx_split(struct virtqueue *_vq,
>  		BAD_RING(vq, "id %u is not a head!\n", i);
>  		return NULL;
>  	}
> +	if (unlikely(len && vq->split.desc_state[i].in_len < *len)) {
> +		BAD_RING(vq, "id %u has invalid length: %u!\n", i, *len);
> +		return NULL;
> +	}
>  
>  	/* detach_buf_split clears data, so grab it now. */
>  	ret = vq->split.desc_state[i].data;
> @@ -1059,7 +1067,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq,
>  {
>  	struct vring_packed_desc *desc;
>  	struct scatterlist *sg;
> -	unsigned int i, n, err_idx;
> +	unsigned int i, n, err_idx, in_len = 0;
>  	u16 head, id;
>  	dma_addr_t addr;
>  
> @@ -1084,6 +1092,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq,
>  			if (vring_mapping_error(vq, addr))
>  				goto unmap_release;
>  
> +			in_len += (n < out_sgs) ? 0 : sg->length;
>  			desc[i].flags = cpu_to_le16(n < out_sgs ?
>  						0 : VRING_DESC_F_WRITE);
>  			desc[i].addr = cpu_to_le64(addr);
> @@ -1141,6 +1150,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq,
>  	vq->packed.desc_state[id].data = data;
>  	vq->packed.desc_state[id].indir_desc = desc;
>  	vq->packed.desc_state[id].last = id;
> +	vq->packed.desc_state[id].in_len = in_len;
>  
>  	vq->num_added += 1;
>  
> @@ -1173,7 +1183,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>  	struct vring_virtqueue *vq = to_vvq(_vq);
>  	struct vring_packed_desc *desc;
>  	struct scatterlist *sg;
> -	unsigned int i, n, c, descs_used, err_idx;
> +	unsigned int i, n, c, descs_used, err_idx, in_len = 0;
>  	__le16 head_flags, flags;
>  	u16 head, id, prev, curr, avail_used_flags;
>  
> @@ -1223,6 +1233,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>  			if (vring_mapping_error(vq, addr))
>  				goto unmap_release;
>  
> +			in_len += (n < out_sgs) ? 0 : sg->length;
>  			flags = cpu_to_le16(vq->packed.avail_used_flags |
>  				    (++c == total_sg ? 0 : VRING_DESC_F_NEXT) |
>  				    (n < out_sgs ? 0 : VRING_DESC_F_WRITE));
> @@ -1268,6 +1279,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>  	vq->packed.desc_state[id].data = data;
>  	vq->packed.desc_state[id].indir_desc = ctx;
>  	vq->packed.desc_state[id].last = prev;
> +	vq->packed.desc_state[id].in_len = in_len;
>  
>  	/*
>  	 * A driver MUST NOT make the first descriptor in the list
> @@ -1456,6 +1468,10 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
>  		BAD_RING(vq, "id %u is not a head!\n", id);
>  		return NULL;
>  	}
> +	if (unlikely(len && vq->packed.desc_state[id].in_len < *len)) {
> +		BAD_RING(vq, "id %u has invalid length: %u!\n", id, *len);
> +		return NULL;
> +	}
>  
>  	/* detach_buf_packed clears data, so grab it now. */
>  	ret = vq->packed.desc_state[id].data;
> -- 
> 2.11.0

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 00/17] Add validation for used length

[Michael S. Tsirkin]

On Mon, May 17, 2021 at 05:08:19PM +0800, Xie Yongji wrote:
> Current virtio device drivers may trust the used length returned
> in virtqueue_get_buf()/virtqueue_get_buf_ctx(). But the used length
> might come from an untrusted device when VDUSE[1] is enabled. To
> protect this case, this series tries to add validation for the
> used length.
> 
> Since many legacy devices will also set the used length incorrectly,
> we did not add the validation unconditionally. Instead, we will do
> the validation only when the device driver needs the used length.
> A NULL len passed to virtqueue_get_buf()/virtqueue_get_buf_ctx()
> will mean the used length is not needed by the device driver.

Can we be more specific? Which drivers have problems when used len
is incorrect? Maybe there's an easier way like validating the length
in the driver ...

> [1] https://lore.kernel.org/kvm/20210331080519.172-1-xieyongji@bytedance.com/
> 
> Xie Yongji (17):
>   virtio_ring: Avoid reading unneeded used length
>   virtio-blk: Remove unused used length
>   virtio_console: Remove unused used length
>   crypto: virtio - Remove unused used length
>   drm/virtio: Remove unused used length
>   caif_virtio: Remove unused used length
>   virtio_net: Remove unused used length
>   mac80211_hwsim: Remove unused used length
>   virtio_pmem: Remove unused used length
>   rpmsg: virtio: Remove unused used length
>   virtio_scsi: Remove unused used length
>   virtio_balloon: Remove unused used length
>   virtio_input: Remove unused used length
>   virtio_mem: Remove unused used length
>   virtiofs: Remove unused used length
>   vsock: Remove unused used length
>   virtio_ring: Add validation for used length
> 
>  drivers/block/virtio_blk.c                 |  3 +--
>  drivers/char/virtio_console.c              | 12 ++++--------
>  drivers/crypto/virtio/virtio_crypto_algs.c |  6 ++----
>  drivers/gpu/drm/virtio/virtgpu_vq.c        |  3 +--
>  drivers/net/caif/caif_virtio.c             |  3 +--
>  drivers/net/virtio_net.c                   | 10 ++++------
>  drivers/net/wireless/mac80211_hwsim.c      |  3 +--
>  drivers/nvdimm/nd_virtio.c                 |  3 +--
>  drivers/rpmsg/virtio_rpmsg_bus.c           |  3 +--
>  drivers/scsi/virtio_scsi.c                 |  3 +--
>  drivers/virtio/virtio_balloon.c            | 21 ++++++++++-----------
>  drivers/virtio/virtio_input.c              |  6 ++----
>  drivers/virtio/virtio_mem.c                |  3 +--
>  drivers/virtio/virtio_ring.c               | 28 +++++++++++++++++++++++-----
>  fs/fuse/virtio_fs.c                        |  6 ++----
>  net/vmw_vsock/virtio_transport.c           |  3 +--
>  16 files changed, 56 insertions(+), 60 deletions(-)
> 
> -- 
> 2.11.0

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: Re: [RFC PATCH 00/17] Add validation for used length

[Michael S. Tsirkin]

On Tue, May 18, 2021 at 04:29:44PM +0800, Yongji Xie wrote:
> On Tue, May 18, 2021 at 7:40 AM Michael S. Tsirkin <mst@redhat.com> wrote:
> >
> > On Mon, May 17, 2021 at 05:08:19PM +0800, Xie Yongji wrote:
> > > Current virtio device drivers may trust the used length returned
> > > in virtqueue_get_buf()/virtqueue_get_buf_ctx(). But the used length
> > > might come from an untrusted device when VDUSE[1] is enabled. To
> > > protect this case, this series tries to add validation for the
> > > used length.
> > >
> > > Since many legacy devices will also set the used length incorrectly,
> > > we did not add the validation unconditionally. Instead, we will do
> > > the validation only when the device driver needs the used length.
> > > A NULL len passed to virtqueue_get_buf()/virtqueue_get_buf_ctx()
> > > will mean the used length is not needed by the device driver.
> >
> > Can we be more specific? Which drivers have problems when used len
> > is incorrect? Maybe there's an easier way like validating the length
> > in the driver ...
> >
> 
> It's ok to me. But this means all future new drivers need to remember
> to do the validation.
> 
> Now only virtio-net and virtio-console drivers have this problem. I
> can send some patches to fix it.
> 
> Thanks,
> Yongji

I'd say let's just document the requirement for now.

-- 
MST

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization

Re: [RFC PATCH 17/17] virtio_ring: Add validation for used length

[Jason Wang]

在 2021/5/18 上午7:39, Michael S. Tsirkin 写道:
> On Mon, May 17, 2021 at 05:08:36PM +0800, Xie Yongji wrote:
>> This adds validation for used length (might come
>> from an untrusted device) when it will be used by
>> virtio device driver.
>>
>> Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
>> ---
>>   drivers/virtio/virtio_ring.c | 22 +++++++++++++++++++---
>>   1 file changed, 19 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
>> index d999a1d6d271..7d4845d06f21 100644
>> --- a/drivers/virtio/virtio_ring.c
>> +++ b/drivers/virtio/virtio_ring.c
>> @@ -68,11 +68,13 @@
>>   struct vring_desc_state_split {
>>   	void *data;			/* Data for callback. */
>>   	struct vring_desc *indir_desc;	/* Indirect descriptor, if any. */
>> +	u32 in_len;			/* Total length of writable buffer */
>>   };
>>   
>>   struct vring_desc_state_packed {
>>   	void *data;			/* Data for callback. */
>>   	struct vring_packed_desc *indir_desc; /* Indirect descriptor, if any. */
>> +	u32 in_len;			/* Total length of writable buffer */
>>   	u16 num;			/* Descriptor list length. */
>>   	u16 last;			/* The last desc state in a list. */
>>   };
>
> Hmm for packed it's aligned to 64 bit anyway, so we are not making it
> any worse. But for split this pushes struct size up by 1/3 increasing
> cache pressure.


We can eliminate this by validating through virtio device driver instead 
of virtio core.

E.g for virtio-net we know the rx buffer size so there's no need to 
store in twice in the core.

Thanks


>
>
>> @@ -486,7 +488,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>>   	struct vring_virtqueue *vq = to_vvq(_vq);
>>   	struct scatterlist *sg;
>>   	struct vring_desc *desc;
>> -	unsigned int i, n, avail, descs_used, prev, err_idx;
>> +	unsigned int i, n, avail, descs_used, prev, err_idx, in_len = 0;
>>   	int head;
>>   	bool indirect;
>>   
>> @@ -570,6 +572,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>>   						     VRING_DESC_F_NEXT |
>>   						     VRING_DESC_F_WRITE,
>>   						     indirect);
>> +			in_len += sg->length;
>>   		}
>>   	}
>>   	/* Last one doesn't continue. */
>> @@ -604,6 +607,7 @@ static inline int virtqueue_add_split(struct virtqueue *_vq,
>>   
>>   	/* Store token and indirect buffer state. */
>>   	vq->split.desc_state[head].data = data;
>> +	vq->split.desc_state[head].in_len = in_len;
>>   	if (indirect)
>>   		vq->split.desc_state[head].indir_desc = desc;
>>   	else
>> @@ -784,6 +788,10 @@ static void *virtqueue_get_buf_ctx_split(struct virtqueue *_vq,
>>   		BAD_RING(vq, "id %u is not a head!\n", i);
>>   		return NULL;
>>   	}
>> +	if (unlikely(len && vq->split.desc_state[i].in_len < *len)) {
>> +		BAD_RING(vq, "id %u has invalid length: %u!\n", i, *len);
>> +		return NULL;
>> +	}
>>   
>>   	/* detach_buf_split clears data, so grab it now. */
>>   	ret = vq->split.desc_state[i].data;
>> @@ -1059,7 +1067,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq,
>>   {
>>   	struct vring_packed_desc *desc;
>>   	struct scatterlist *sg;
>> -	unsigned int i, n, err_idx;
>> +	unsigned int i, n, err_idx, in_len = 0;
>>   	u16 head, id;
>>   	dma_addr_t addr;
>>   
>> @@ -1084,6 +1092,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq,
>>   			if (vring_mapping_error(vq, addr))
>>   				goto unmap_release;
>>   
>> +			in_len += (n < out_sgs) ? 0 : sg->length;
>>   			desc[i].flags = cpu_to_le16(n < out_sgs ?
>>   						0 : VRING_DESC_F_WRITE);
>>   			desc[i].addr = cpu_to_le64(addr);
>> @@ -1141,6 +1150,7 @@ static int virtqueue_add_indirect_packed(struct vring_virtqueue *vq,
>>   	vq->packed.desc_state[id].data = data;
>>   	vq->packed.desc_state[id].indir_desc = desc;
>>   	vq->packed.desc_state[id].last = id;
>> +	vq->packed.desc_state[id].in_len = in_len;
>>   
>>   	vq->num_added += 1;
>>   
>> @@ -1173,7 +1183,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>>   	struct vring_virtqueue *vq = to_vvq(_vq);
>>   	struct vring_packed_desc *desc;
>>   	struct scatterlist *sg;
>> -	unsigned int i, n, c, descs_used, err_idx;
>> +	unsigned int i, n, c, descs_used, err_idx, in_len = 0;
>>   	__le16 head_flags, flags;
>>   	u16 head, id, prev, curr, avail_used_flags;
>>   
>> @@ -1223,6 +1233,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>>   			if (vring_mapping_error(vq, addr))
>>   				goto unmap_release;
>>   
>> +			in_len += (n < out_sgs) ? 0 : sg->length;
>>   			flags = cpu_to_le16(vq->packed.avail_used_flags |
>>   				    (++c == total_sg ? 0 : VRING_DESC_F_NEXT) |
>>   				    (n < out_sgs ? 0 : VRING_DESC_F_WRITE));
>> @@ -1268,6 +1279,7 @@ static inline int virtqueue_add_packed(struct virtqueue *_vq,
>>   	vq->packed.desc_state[id].data = data;
>>   	vq->packed.desc_state[id].indir_desc = ctx;
>>   	vq->packed.desc_state[id].last = prev;
>> +	vq->packed.desc_state[id].in_len = in_len;
>>   
>>   	/*
>>   	 * A driver MUST NOT make the first descriptor in the list
>> @@ -1456,6 +1468,10 @@ static void *virtqueue_get_buf_ctx_packed(struct virtqueue *_vq,
>>   		BAD_RING(vq, "id %u is not a head!\n", id);
>>   		return NULL;
>>   	}
>> +	if (unlikely(len && vq->packed.desc_state[id].in_len < *len)) {
>> +		BAD_RING(vq, "id %u has invalid length: %u!\n", id, *len);
>> +		return NULL;
>> +	}
>>   
>>   	/* detach_buf_packed clears data, so grab it now. */
>>   	ret = vq->packed.desc_state[id].data;
>> -- 
>> 2.11.0

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization