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

["Michael S. Tsirkin" <mst-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>]

On Tue, Apr 21, 2015 at 05:48:20PM +0200, Greg Kurz wrote:
> On Tue, 21 Apr 2015 16:04:23 +0200
> "Michael S. Tsirkin" <mst-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> 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-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
> > > ---
> > >  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 ?

I just tried and gcc optimizes
s.num != 0 && s.num != 1 to s.num > 1

The former will be more readable once we
replace 0 and 1 with defines.

So ignore my advice, keep code as is but use defines.



> > 
> > > +		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 ?

Fine.

> > > + * other values return EINVAL.

Pls also add a note saying that not all kernel configurations support this ioctl,
but all configurations that support SET also support GET.

> > > + */
> > > +#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. */
> > >  
> > 

I'm inclined to think VHOST_SET_VRING_ENDIAN is a slightly better name.
What do you think?

-- 
MST