Re: [PATCH v4 7/8] vhost: feature to set the vring endianness

[Greg Kurz]

On Tue, 21 Apr 2015 16:04:23 +0200
"Michael S. Tsirkin" <mst@xxxxxxxxxx> wrote:

> On Fri, Apr 10, 2015 at 12:19:16PM +0200, Greg Kurz wrote:
> > This patch brings cross-endian support to vhost when used to implement
> > legacy virtio devices. Since it is a relatively rare situation, the
> > feature availability is controlled by a kernel config option (not set
> > by default).
> > 
> > The vq->is_le boolean field is added to cache the endianness to be
> > used for ring accesses. It defaults to native endian, as expected
> > by legacy virtio devices. When the ring gets active, we force little
> > endian if the device is modern. When the ring is deactivated, we
> > revert to the native endian default.
> > 
> > If cross-endian was compiled in, a vq->user_be boolean field is added
> > so that userspace may request a specific endianness. This field is
> > used to override the default when activating the ring of a legacy
> > device. It has no effect on modern devices.
> > 
> > Signed-off-by: Greg Kurz <gkurz@xxxxxxxxxxxxxxxxxx>
> > ---
> >  drivers/vhost/Kconfig      |   10 ++++++the
> >  drivers/vhost/vhost.c      |   76 +++++++++++++++++++++++++++++++++++++++++++-
> >  drivers/vhost/vhost.h      |   12 +++++--
> >  include/uapi/linux/vhost.h |    9 +++++
> >  4 files changed, 103 insertions(+), 4 deletions(-)
> > 
> > Changes since v3:
> > - VHOST_SET_VRING_ENDIAN_LEGACY ioctl renamed to VHOST_SET_VRING_BIG_ENDIAN
> > - ioctl API is now: 0 for le, 1 for be, other values are EINVAL
> > - ioctl doesn't filter out modern devices
> > - ioctl stubs return ENOIOCTLCMD
> > - forbid endianness changes when vring is active
> > - logic now handled with vq->is_le and vq->user_be according to device
> >   start/stop as suggested by Michael
> > 
> > diff --git a/drivers/vhost/Kconfig b/drivers/vhost/Kconfig
> > index 017a1e8..0aec88c 100644
> > --- a/drivers/vhost/Kconfig
> > +++ b/drivers/vhost/Kconfig
> > @@ -32,3 +32,13 @@ config VHOST
> >  	---help---
> >  	  This option is selected by any driver which needs to access
> >  	  the core of vhost.
> > +
> > +config VHOST_SET_ENDIAN_LEGACY
> 
> I'd prefer namin this VHOST_CROSS_ENDIAN_LEGACY
> 

Yes, makes sense. I'll make sure most of the cross-endian changes are
easy to grep.

> > +	bool "Cross-endian support for host kernel accelerator"
> > +	default n
> > +	---help---
> > +	  This option allows vhost to support guests with a different byte
> > +	  ordering from host. It is disabled by default since it adds overhead
> > +	  and it is only needed by a few platforms (powerpc and arm).
> 
> and is only useful on a few platforms (powerpc and arm).
> 
> "it" seems to refer to "overhead", which is rarely needed.
> needed is a bit too strong, you can always e.g. run virtio
> in userspace.
> 

My poor English again... it seems you understood what I wanted to
write though :)

> > +
> > +	  If unsure, say "N".
> > diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> > index 2ee2826..3eb756b 100644the
> > --- a/drivers/vhost/vhost.c
> > +++ b/drivers/vhost/vhost.c
> > @@ -199,6 +199,10 @@ static void vhost_vq_reset(struct vhost_dev *dev,
> >  	vq->call = NULL;
> >  	vq->log_ctx = NULL;
> >  	vq->memory = NULL;
> > +	vq->is_le = virtio_legacy_is_little_endian();
> > +#ifdef CONFIG_VHOST_SET_ENDIAN_LEGACY
> > +	vq->user_be = !vq->is_le;
> > +#endif
> 
> add a wrapper for this too?
> 

Will do.

> >  }
> >  
> >  static int vhost_worker(void *data)
> > @@ -630,6 +634,53 @@ static long vhost_set_memory(struct vhost_dev *d, struct vhost_memory __user *m)
> >  	return 0;
> >  }
> >  
> > +#ifdef CONFIG_VHOST_SET_ENDIAN_LEGACY
> > +static long vhost_set_vring_big_endian(struct vhost_virtqueue *vq,
> > +				       int __user *argp)
> > +{
> > +	struct vhost_vring_state s;
> > +
> > +	if (vq->private_data)
> > +		return -EBUSY;
> > +
> > +	if (copy_from_user(&s, argp, sizeof(s)))
> > +		return -EFAULT;
> > +
> > +	if (s.num && s.num != 1)
> 
> s.num & ~0x1
> 

Since s.num is unsigned and I assume this won't change, what about
s.num > 1 as suggested by Cornelia ?

> 
> > +		return -EINVAL;
> > +
> > +	vq->user_be = s.num;
> > +
> > +	return 0;
> > +}
> > +
> > +static long vhost_get_vring_big_endian(struct vhost_virtqueue *vq, u32 idx,
> > +				       int __user *argp)
> > +{
> > +	struct vhost_vring_state s = {
> > +		.index = idx,
> > +		.num = vq->user_be
> > +	};
> > +
> > +	if (copy_to_user(argp, &s, sizeof(s)))
> > +		return -EFAULT;
> > +
> > +	return 0;
> > +}
> > +#else
> > +static long vhost_set_vring_big_endian(struct vhost_virtqueue *vq,
> > +				       int __user *argp)
> > +{
> > +	return -ENOIOCTLCMD;
> > +}
> > +
> > +static long vhost_get_vring_big_endian(struct vhost_virtqueue *vq, u32 idx,
> > +				       int __user *argp)
> > +{
> > +	return -ENOIOCTLCMD;
> > +}
> > +#endif /* CONFIG_VHOST_SET_ENDIAN_LEGACY */
> > +
> >  long vhost_vring_ioctl(struct vhost_dev *d, int ioctl, void __user *argp)
> >  {
> >  	struct file *eventfp, *filep = NULL;
> > @@ -806,6 +857,12 @@ long vhost_vring_ioctl(struct vhost_dev *d, int ioctl, void __user *argp)
> >  		} else
> >  			filep = eventfp;the
> >  		break;
> > +	case VHOST_SET_VRING_BIG_ENDIAN:
> > +		r = vhost_set_vring_big_endian(vq, argp);
> > +		break;
> > +	case VHOST_GET_VRING_BIG_ENDIAN:
> > +		r = vhost_get_vring_big_endian(vq, idx, argp);
> > +		break;
> >  	default:
> >  		r = -ENOIOCTLCMD;
> >  	}
> > @@ -1040,12 +1097,29 @@ static int vhost_update_avail_event(struct vhost_virtqueue *vq, u16 avail_event)
> >  	return 0;
> >  }
> >  
> > +#ifdef CONFIG_VHOST_SET_ENDIAN_LEGACY
> > +static void vhost_init_is_le(struct vhost_virtqueue *vq)
> > +{
> > +	vq->is_le = vhost_has_feature(vq, VIRTIO_F_VERSION_1) || !vq->user_be;
> > +}
> > +#else
> > +static void vhost_init_is_le(struct vhost_virtqueue *vq)
> > +{
> > +	if (vhost_has_feature(vq, VIRTIO_F_VERSION_1))
> > +		vq->is_le = true;
> > +}
> > +#endif
> > +
> 
> I'd prefer localizing ifdefery somewhere near top of file.
> 

Will do.

> >  int vhost_init_used(struct vhost_virtqueue *vq)
> >  {
> >  	__virtio16 last_used_idx;
> >  	int r;
> > -	if (!vq->private_data)
> > +	if (!vq->private_data) {
> > +		vq->is_le = virtio_legacy_is_little_endian();
> >  		return 0;
> > +	}
> > +
> > +	vhost_init_is_le(vq);
> >  
> >  	r = vhost_update_used_flags(vq);
> >  	if (r)
> > diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
> > index 4e9a186..04b2add 100644
> > --- a/drivers/vhost/vhost.h
> > +++ b/drivers/vhost/vhost.h
> > @@ -106,6 +106,14 @@ struct vhost_virtqueue {
> >  	/* Log write descriptors */
> >  	void __user *log_base;
> >  	struct vhost_log *log;
> > +
> > +	/* Ring endianness. Defaults to legacy native endianness.
> > +	 * Set to true when starting a modern virtio device. */
> > +	bool is_le;
> > +#ifdef CONFIG_VHOST_SET_ENDIAN_LEGACY
> > +	/* Ring endianness requested by userspace for cross-endian support. */
> > +	bool user_be;
> > +#endif
> >  };
> >  
> >  struct vhost_dev {
> > @@ -175,9 +183,7 @@ static inline bool vhost_has_feature(struct vhost_virtqueue *vq, int bit)
> >  
> >  static inline bool vhost_is_little_endian(struct vhost_virtqueue *vq)
> >  {
> > -	if (vhost_has_feature(vq, VIRTIO_F_VERSION_1))
> > -		return true;
> > -	return virtio_legacy_is_little_endian();
> > +	return vq->is_le;
> >  }
> >  
> >  /* Memory accessors */
> > diff --git a/include/uapi/linux/vhost.h b/include/uapi/linux/vhost.h
> > index bb6a5b4..5cdebbc 100644
> > --- a/include/uapi/linux/vhost.h
> > +++ b/include/uapi/linux/vhost.h
> > @@ -103,6 +103,15 @@ struct vhost_memory {
> >  /* Get accessor: reads index, writes value in num */
> >  #define VHOST_GET_VRING_BASE _IOWR(VHOST_VIRTIO, 0x12, struct vhost_vring_state)
> >  
> > +/* Set the vring byte order in num. This is a legacy only API that is simply
> > + * ignored when VIRTIO_F_VERSION_1 is set.
> > + * 0 to set to little-endian
> > + * 1 to set to big-endian
> 
> How about defines for these?
> 

Ok. I'll put the defines here so that all the cross-endian stuff
lies in the same hunk. Is it ok for you ?

> > + * other values return EINVAL.
> > + */
> > +#define VHOST_SET_VRING_BIG_ENDIAN _IOW(VHOST_VIRTIO, 0x13, struct vhost_vring_state)
> > +#define VHOST_GET_VRING_BIG_ENDIAN _IOW(VHOST_VIRTIO, 0x14, struct vhost_vring_state)
> > +
> >  /* The following ioctls use eventfd file descriptors to signal and poll
> >   * for events. */
> >  
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/