Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Michael S. Tsirkin]

On Wed, Jan 04, 2017 at 08:25:05AM +0300, Christoph Hellwig wrote:
> Most users of BLOCK_PC requests allocate the sense buffer on the stack,
> so to avoid DMA to the stack copy them to a field in the heap allocated
> virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
> including the SG_IO ioctl from userspace will crash the kernel.  Note that
> this includes running tools like hdparm even when the host does not have
> SCSI passthrough enabled.
> 
> Signed-off-by: Christoph Hellwig 
> ---
>  drivers/block/virtio_blk.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index 5545a67..3c3b8f6 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -56,6 +56,7 @@ struct virtblk_req {
>   struct virtio_blk_outhdr out_hdr;
>   struct virtio_scsi_inhdr in_hdr;
>   u8 status;
> + u8 sense[SCSI_SENSE_BUFFERSIZE];
>   struct scatterlist sg[];
>  };
>  
> @@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
>   }
>  
>   if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
> - sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> + memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> + sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);

I would prefer sizeof(vbr->sense) here to avoid duplication.
Otherwise:

Acked-by: Michael S. Tsirkin 


>   sgs[num_out + num_in++] = 
>   sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
>   sgs[num_out + num_in++] = 
> -- 
> 2.1.4

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Michael S. Tsirkin]

On Wed, Jan 04, 2017 at 08:25:05AM +0300, Christoph Hellwig wrote:
> Most users of BLOCK_PC requests allocate the sense buffer on the stack,
> so to avoid DMA to the stack copy them to a field in the heap allocated
> virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
> including the SG_IO ioctl from userspace will crash the kernel.  Note that
> this includes running tools like hdparm even when the host does not have
> SCSI passthrough enabled.
> 
> Signed-off-by: Christoph Hellwig 
> ---
>  drivers/block/virtio_blk.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index 5545a67..3c3b8f6 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -56,6 +56,7 @@ struct virtblk_req {
>   struct virtio_blk_outhdr out_hdr;
>   struct virtio_scsi_inhdr in_hdr;
>   u8 status;
> + u8 sense[SCSI_SENSE_BUFFERSIZE];
>   struct scatterlist sg[];
>  };
>  
> @@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
>   }
>  
>   if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
> - sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> + memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> + sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);

I would prefer sizeof(vbr->sense) here to avoid duplication.
Otherwise:

Acked-by: Michael S. Tsirkin 


>   sgs[num_out + num_in++] = 
>   sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
>   sgs[num_out + num_in++] = 
> -- 
> 2.1.4

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Jens Axboe]

On 01/09/2017 06:35 AM, Christoph Hellwig wrote:
> Is someone going to pick the patch up and send it to Linus?  I keep
> running into all kinds of boot failures whenever I forget to cherry
> pick it into my development trees..

I'll add it.

-- 
Jens Axboe

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Jens Axboe]

On 01/09/2017 06:35 AM, Christoph Hellwig wrote:
> Is someone going to pick the patch up and send it to Linus?  I keep
> running into all kinds of boot failures whenever I forget to cherry
> pick it into my development trees..

I'll add it.

-- 
Jens Axboe

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

Is someone going to pick the patch up and send it to Linus?  I keep
running into all kinds of boot failures whenever I forget to cherry
pick it into my development trees..

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

Is someone going to pick the patch up and send it to Linus?  I keep
running into all kinds of boot failures whenever I forget to cherry
pick it into my development trees..

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

On Thu, Jan 05, 2017 at 11:37:46AM +0100, 王金浦 wrote:
> Thanks, so it's only relevant to kernel > 4.9, as  CONFIG_VMAP_STACK
> only introduced in 4.9 kernel.

kernel >= 4.9, but otherwise, yes.

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

On Thu, Jan 05, 2017 at 11:37:46AM +0100, 王金浦 wrote:
> Thanks, so it's only relevant to kernel > 4.9, as  CONFIG_VMAP_STACK
> only introduced in 4.9 kernel.

kernel >= 4.9, but otherwise, yes.

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[王金浦]

2017-01-05 10:57 GMT+01:00 Christoph Hellwig :
> On Wed, Jan 04, 2017 at 04:47:03PM +0100, 王金浦 wrote:
>> This sounds scary.
>> Could you share how to reproduce it, this should go into stable if
>> it's the case.
>
> Step 1: Build your kernel with CONFIG_VMAP_STACK=y
> Step 2: issue a SG_IO ioctl, e.g. sg_inq /dev/vda
>
Thanks, so it's only relevant to kernel > 4.9, as  CONFIG_VMAP_STACK
only introduced in 4.9 kernel.

Regards,
Jinpu

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[王金浦]

2017-01-05 10:57 GMT+01:00 Christoph Hellwig :
> On Wed, Jan 04, 2017 at 04:47:03PM +0100, 王金浦 wrote:
>> This sounds scary.
>> Could you share how to reproduce it, this should go into stable if
>> it's the case.
>
> Step 1: Build your kernel with CONFIG_VMAP_STACK=y
> Step 2: issue a SG_IO ioctl, e.g. sg_inq /dev/vda
>
Thanks, so it's only relevant to kernel > 4.9, as  CONFIG_VMAP_STACK
only introduced in 4.9 kernel.

Regards,
Jinpu

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

On Wed, Jan 04, 2017 at 04:47:03PM +0100, 王金浦 wrote:
> This sounds scary.
> Could you share how to reproduce it, this should go into stable if
> it's the case.

Step 1: Build your kernel with CONFIG_VMAP_STACK=y
Step 2: issue a SG_IO ioctl, e.g. sg_inq /dev/vda

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

On Wed, Jan 04, 2017 at 04:47:03PM +0100, 王金浦 wrote:
> This sounds scary.
> Could you share how to reproduce it, this should go into stable if
> it's the case.

Step 1: Build your kernel with CONFIG_VMAP_STACK=y
Step 2: issue a SG_IO ioctl, e.g. sg_inq /dev/vda

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[王金浦]

Hi Christoph,

2017-01-04 6:25 GMT+01:00 Christoph Hellwig :
> Most users of BLOCK_PC requests allocate the sense buffer on the stack,
> so to avoid DMA to the stack copy them to a field in the heap allocated
> virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
> including the SG_IO ioctl from userspace will crash the kernel.  Note that
> this includes running tools like hdparm even when the host does not have
> SCSI passthrough enabled.

This sounds scary.
Could you share how to reproduce it, this should go into stable if
it's the case.

Thanks,
Jinpu

>
> Signed-off-by: Christoph Hellwig 
> ---
>  drivers/block/virtio_blk.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index 5545a67..3c3b8f6 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -56,6 +56,7 @@ struct virtblk_req {
> struct virtio_blk_outhdr out_hdr;
> struct virtio_scsi_inhdr in_hdr;
> u8 status;
> +   u8 sense[SCSI_SENSE_BUFFERSIZE];
> struct scatterlist sg[];
>  };
>
> @@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
> }
>
> if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
> -   sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> +   memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> +   sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);
> sgs[num_out + num_in++] = 
> sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
> sgs[num_out + num_in++] = 
> --
> 2.1.4
>

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[王金浦]

Hi Christoph,

2017-01-04 6:25 GMT+01:00 Christoph Hellwig :
> Most users of BLOCK_PC requests allocate the sense buffer on the stack,
> so to avoid DMA to the stack copy them to a field in the heap allocated
> virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
> including the SG_IO ioctl from userspace will crash the kernel.  Note that
> this includes running tools like hdparm even when the host does not have
> SCSI passthrough enabled.

This sounds scary.
Could you share how to reproduce it, this should go into stable if
it's the case.

Thanks,
Jinpu

>
> Signed-off-by: Christoph Hellwig 
> ---
>  drivers/block/virtio_blk.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index 5545a67..3c3b8f6 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -56,6 +56,7 @@ struct virtblk_req {
> struct virtio_blk_outhdr out_hdr;
> struct virtio_scsi_inhdr in_hdr;
> u8 status;
> +   u8 sense[SCSI_SENSE_BUFFERSIZE];
> struct scatterlist sg[];
>  };
>
> @@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
> }
>
> if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
> -   sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> +   memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
> +   sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);
> sgs[num_out + num_in++] = 
> sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
> sgs[num_out + num_in++] = 
> --
> 2.1.4
>

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Jason Wang]

On 2017年01月04日 13:25, Christoph Hellwig wrote:

Most users of BLOCK_PC requests allocate the sense buffer on the stack,
so to avoid DMA to the stack copy them to a field in the heap allocated
virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
including the SG_IO ioctl from userspace will crash the kernel.  Note that
this includes running tools like hdparm even when the host does not have
SCSI passthrough enabled.

Signed-off-by: Christoph Hellwig 
---
  drivers/block/virtio_blk.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 5545a67..3c3b8f6 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -56,6 +56,7 @@ struct virtblk_req {
struct virtio_blk_outhdr out_hdr;
struct virtio_scsi_inhdr in_hdr;
u8 status;
+   u8 sense[SCSI_SENSE_BUFFERSIZE];
struct scatterlist sg[];
  };
  
@@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,

}
  
  	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {

-   sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);
sgs[num_out + num_in++] = 
sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
sgs[num_out + num_in++] = 


Acked-by: Jason Wang 

Re: [PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Jason Wang]

On 2017年01月04日 13:25, Christoph Hellwig wrote:

Most users of BLOCK_PC requests allocate the sense buffer on the stack,
so to avoid DMA to the stack copy them to a field in the heap allocated
virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
including the SG_IO ioctl from userspace will crash the kernel.  Note that
this includes running tools like hdparm even when the host does not have
SCSI passthrough enabled.

Signed-off-by: Christoph Hellwig 
---
  drivers/block/virtio_blk.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 5545a67..3c3b8f6 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -56,6 +56,7 @@ struct virtblk_req {
struct virtio_blk_outhdr out_hdr;
struct virtio_scsi_inhdr in_hdr;
u8 status;
+   u8 sense[SCSI_SENSE_BUFFERSIZE];
struct scatterlist sg[];
  };
  
@@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,

}
  
  	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {

-   sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);
sgs[num_out + num_in++] = 
sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
sgs[num_out + num_in++] = 


Acked-by: Jason Wang 

[PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

Most users of BLOCK_PC requests allocate the sense buffer on the stack,
so to avoid DMA to the stack copy them to a field in the heap allocated
virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
including the SG_IO ioctl from userspace will crash the kernel.  Note that
this includes running tools like hdparm even when the host does not have
SCSI passthrough enabled.

Signed-off-by: Christoph Hellwig 
---
 drivers/block/virtio_blk.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 5545a67..3c3b8f6 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -56,6 +56,7 @@ struct virtblk_req {
struct virtio_blk_outhdr out_hdr;
struct virtio_scsi_inhdr in_hdr;
u8 status;
+   u8 sense[SCSI_SENSE_BUFFERSIZE];
struct scatterlist sg[];
 };
 
@@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
}
 
if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
-   sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);
sgs[num_out + num_in++] = 
sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
sgs[num_out + num_in++] = 
-- 
2.1.4

[PATCH] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

Most users of BLOCK_PC requests allocate the sense buffer on the stack,
so to avoid DMA to the stack copy them to a field in the heap allocated
virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
including the SG_IO ioctl from userspace will crash the kernel.  Note that
this includes running tools like hdparm even when the host does not have
SCSI passthrough enabled.

Signed-off-by: Christoph Hellwig 
---
 drivers/block/virtio_blk.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 5545a67..3c3b8f6 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -56,6 +56,7 @@ struct virtblk_req {
struct virtio_blk_outhdr out_hdr;
struct virtio_scsi_inhdr in_hdr;
u8 status;
+   u8 sense[SCSI_SENSE_BUFFERSIZE];
struct scatterlist sg[];
 };
 
@@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
}
 
if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
-   sg_init_one(, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+   sg_init_one(, vbr->sense, SCSI_SENSE_BUFFERSIZE);
sgs[num_out + num_in++] = 
sg_init_one(, >in_hdr, sizeof(vbr->in_hdr));
sgs[num_out + num_in++] = 
-- 
2.1.4