Linux Virtualization - May 2013 - [PATCH] virtio_pci: fix capability format, comments

- queue size can actually be 0 which is not a power of 2
- fix capability format. PCI spec says:
	The layout of the information is vendor specific, except that the byte
	immediately following the ?Next? pointer in the capability structure is
	defined to be a length field.
	This length field provides the number of bytes in the capability
	structure (including the ID and Next pointer bytes).

Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
---

This patch is on top of the new layout branch, too

 include/uapi/linux/virtio_pci.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/uapi/linux/virtio_pci.h b/include/uapi/linux/virtio_pci.h
index cda688f..a5ef8cd 100644
--- a/include/uapi/linux/virtio_pci.h
+++ b/include/uapi/linux/virtio_pci.h
@@ -129,6 +129,7 @@
 struct virtio_pci_cap {
 	__u8 cap_vndr;	/* Generic PCI field: PCI_CAP_ID_VNDR */
 	__u8 cap_next;	/* Generic PCI field: next ptr. */
+	__u8 cap_len;	/* Generic PCI field: capability length */
 	__u8 cfg_type;	/* One of the VIRTIO_PCI_CAP_*_CFG. */
 	__u8 bar;		/* Where to find it. */
 	__le32 offset;	/* Offset within bar. */
@@ -154,7 +155,7 @@ struct virtio_pci_common_cfg {
 
 	/* About a specific virtqueue. */
 	__le16 queue_select;		/* read-write */
-	__le16 queue_size;		/* read-write, power of 2. */
+	__le16 queue_size;		/* read-write, power of 2, or 0. */
 	__le16 queue_msix_vector;	/* read-write */
 	__le16 queue_enable;		/* read-write */
 	__le16 queue_notify_off;	/* read-only */
-- 
MST

"Michael S. Tsirkin" <mst at redhat.com> writes:
> - queue size can actually be 0 which is not a power of 2
Actually, that points to a flaw in the code.  When we shut down the
queue, we should ideally reset it to what the device started with,
rather than 0.

See below.
> - fix capability format. PCI spec says:
> 	The layout of the information is vendor specific, except that the byte
> 	immediately following the ?Next? pointer in the capability structure is
> 	defined to be a length field.
> 	This length field provides the number of bytes in the capability
> 	structure (including the ID and Next pointer bytes).
That part's definitely correct: applied.

Thanks,
Rusty.

Subjet: virtio_pci: save the desired ringsize.

MST points out that 0 isn't a power of 2.  This means we can't re-open
a virtio device once we write 0 into the queue length.

We should restore the amount the device originally asked for.

Signed-off-by: Rusty Russell <rusty at rustcorp.com.au>

diff --git a/drivers/virtio/virtio_pci-common.h
b/drivers/virtio/virtio_pci-common.h
index ba1bf81..0e3143b 100644
--- a/drivers/virtio/virtio_pci-common.h
+++ b/drivers/virtio/virtio_pci-common.h
@@ -82,6 +82,9 @@ struct virtio_pci_vq_info {
 
 	/* MSI-X vector (or none) */
 	unsigned msix_vector;
+
+	/* What size did the device *want* this to be? */
+	u16 desired_num;
 };
 
 /* the notify function used when creating a virt queue */
diff --git a/drivers/virtio/virtio_pci.c b/drivers/virtio/virtio_pci.c
index 0f0e3a6..8b35c2e 100644
--- a/drivers/virtio/virtio_pci.c
+++ b/drivers/virtio/virtio_pci.c
@@ -230,6 +230,7 @@ static struct virtqueue *setup_vq(struct virtio_pci_device
*vp_dev,
 		return ERR_PTR(-ENOMEM);
 
 	info->msix_vector = msix_vec;
+	info->desired_num = num;
 
 	/* get offset of notification word for this vq (shouldn't wrap) */
 	off = ioread16(&vp_dev->common->queue_notify_off);
@@ -350,7 +351,7 @@ static void del_vq(struct virtqueue *vq)
 	vring_del_virtqueue(vq);
 
 	/* This is for our own benefit, not the device's! */
-	iowrite16(0, &vp_dev->common->queue_size);
+	iowrite16(info->desired_num, &vp_dev->common->queue_size);
 	iowrite64_twopart(0, &vp_dev->common->queue_desc);
 	iowrite64_twopart(0, &vp_dev->common->queue_avail);
 	iowrite64_twopart(0, &vp_dev->common->queue_used);