virtio_pci: fix capability format, comments
Subsystems (2)
Michael S. Tsirkin
maintainerPATCH 1/1 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).
---
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(-)
include/uapi/linux/virtio_pci.h +2 -1
@@ -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 */
Signature · 1 line
MST
Quoted 1 line · Michael S. Tsirkin
- 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.
Quoted 6 lines
- 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.
drivers/virtio/virtio_pci-common.h +3 -0
@@ -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 */
drivers/virtio/virtio_pci.c +2 -1
@@ -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);
_______________________________________________
Virtualization mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/virtualization