summaryrefslogtreecommitdiffstats
path: root/drivers/vhost/vhost.c
diff options
context:
space:
mode:
authorMichael S. Tsirkin <mst@redhat.com>2011-07-18 03:48:46 +0000
committerDavid S. Miller <davem@davemloft.net>2011-07-18 10:42:32 -0700
commitbab632d69ee48a106e779b60cc01adfe80a72807 (patch)
tree56b8bd3df85cfee8e425abe18963e5aad015e2fa /drivers/vhost/vhost.c
parent5c74501f76360ce6f410730b9b5e5976f38e8504 (diff)
downloadop-kernel-dev-bab632d69ee48a106e779b60cc01adfe80a72807.zip
op-kernel-dev-bab632d69ee48a106e779b60cc01adfe80a72807.tar.gz
vhost: vhost TX zero-copy support
>From: Shirley Ma <mashirle@us.ibm.com> This adds experimental zero copy support in vhost-net, disabled by default. To enable, set experimental_zcopytx module option to 1. This patch maintains the outstanding userspace buffers in the sequence it is delivered to vhost. The outstanding userspace buffers will be marked as done once the lower device buffers DMA has finished. This is monitored through last reference of kfree_skb callback. Two buffer indices are used for this purpose. The vhost-net device passes the userspace buffers info to lower device skb through message control. DMA done status check and guest notification are handled by handle_tx: in the worst case is all buffers in the vq are in pending/done status, so we need to notify guest to release DMA done buffers first before we get any new buffers from the vq. One known problem is that if the guest stops submitting buffers, buffers might never get used until some further action, e.g. device reset. This does not seem to affect linux guests. Signed-off-by: Shirley <xma@us.ibm.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'drivers/vhost/vhost.c')
-rw-r--r--drivers/vhost/vhost.c128
1 files changed, 113 insertions, 15 deletions
diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
index ea966b3..5ef2f62 100644
--- a/drivers/vhost/vhost.c
+++ b/drivers/vhost/vhost.c
@@ -37,6 +37,8 @@ enum {
VHOST_MEMORY_F_LOG = 0x1,
};
+static unsigned vhost_zcopy_mask __read_mostly;
+
#define vhost_used_event(vq) ((u16 __user *)&vq->avail->ring[vq->num])
#define vhost_avail_event(vq) ((u16 __user *)&vq->used->ring[vq->num])
@@ -179,6 +181,9 @@ static void vhost_vq_reset(struct vhost_dev *dev,
vq->call_ctx = NULL;
vq->call = NULL;
vq->log_ctx = NULL;
+ vq->upend_idx = 0;
+ vq->done_idx = 0;
+ vq->ubufs = NULL;
}
static int vhost_worker(void *data)
@@ -225,10 +230,28 @@ static int vhost_worker(void *data)
return 0;
}
+static void vhost_vq_free_iovecs(struct vhost_virtqueue *vq)
+{
+ kfree(vq->indirect);
+ vq->indirect = NULL;
+ kfree(vq->log);
+ vq->log = NULL;
+ kfree(vq->heads);
+ vq->heads = NULL;
+ kfree(vq->ubuf_info);
+ vq->ubuf_info = NULL;
+}
+
+void vhost_enable_zcopy(int vq)
+{
+ vhost_zcopy_mask |= 0x1 << vq;
+}
+
/* Helper to allocate iovec buffers for all vqs. */
static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
{
int i;
+ bool zcopy;
for (i = 0; i < dev->nvqs; ++i) {
dev->vqs[i].indirect = kmalloc(sizeof *dev->vqs[i].indirect *
@@ -237,19 +260,21 @@ static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
GFP_KERNEL);
dev->vqs[i].heads = kmalloc(sizeof *dev->vqs[i].heads *
UIO_MAXIOV, GFP_KERNEL);
-
+ zcopy = vhost_zcopy_mask & (0x1 << i);
+ if (zcopy)
+ dev->vqs[i].ubuf_info =
+ kmalloc(sizeof *dev->vqs[i].ubuf_info *
+ UIO_MAXIOV, GFP_KERNEL);
if (!dev->vqs[i].indirect || !dev->vqs[i].log ||
- !dev->vqs[i].heads)
+ !dev->vqs[i].heads ||
+ (zcopy && !dev->vqs[i].ubuf_info))
goto err_nomem;
}
return 0;
err_nomem:
- for (; i >= 0; --i) {
- kfree(dev->vqs[i].indirect);
- kfree(dev->vqs[i].log);
- kfree(dev->vqs[i].heads);
- }
+ for (; i >= 0; --i)
+ vhost_vq_free_iovecs(&dev->vqs[i]);
return -ENOMEM;
}
@@ -257,14 +282,8 @@ static void vhost_dev_free_iovecs(struct vhost_dev *dev)
{
int i;
- for (i = 0; i < dev->nvqs; ++i) {
- kfree(dev->vqs[i].indirect);
- dev->vqs[i].indirect = NULL;
- kfree(dev->vqs[i].log);
- dev->vqs[i].log = NULL;
- kfree(dev->vqs[i].heads);
- dev->vqs[i].heads = NULL;
- }
+ for (i = 0; i < dev->nvqs; ++i)
+ vhost_vq_free_iovecs(&dev->vqs[i]);
}
long vhost_dev_init(struct vhost_dev *dev,
@@ -287,6 +306,7 @@ long vhost_dev_init(struct vhost_dev *dev,
dev->vqs[i].log = NULL;
dev->vqs[i].indirect = NULL;
dev->vqs[i].heads = NULL;
+ dev->vqs[i].ubuf_info = NULL;
dev->vqs[i].dev = dev;
mutex_init(&dev->vqs[i].mutex);
vhost_vq_reset(dev, dev->vqs + i);
@@ -390,6 +410,30 @@ long vhost_dev_reset_owner(struct vhost_dev *dev)
return 0;
}
+/* In case of DMA done not in order in lower device driver for some reason.
+ * upend_idx is used to track end of used idx, done_idx is used to track head
+ * of used idx. Once lower device DMA done contiguously, we will signal KVM
+ * guest used idx.
+ */
+int vhost_zerocopy_signal_used(struct vhost_virtqueue *vq)
+{
+ int i;
+ int j = 0;
+
+ for (i = vq->done_idx; i != vq->upend_idx; i = (i + 1) % UIO_MAXIOV) {
+ if ((vq->heads[i].len == VHOST_DMA_DONE_LEN)) {
+ vq->heads[i].len = VHOST_DMA_CLEAR_LEN;
+ vhost_add_used_and_signal(vq->dev, vq,
+ vq->heads[i].id, 0);
+ ++j;
+ } else
+ break;
+ }
+ if (j)
+ vq->done_idx = i;
+ return j;
+}
+
/* Caller should have device mutex */
void vhost_dev_cleanup(struct vhost_dev *dev)
{
@@ -400,6 +444,13 @@ void vhost_dev_cleanup(struct vhost_dev *dev)
vhost_poll_stop(&dev->vqs[i].poll);
vhost_poll_flush(&dev->vqs[i].poll);
}
+ /* Wait for all lower device DMAs done. */
+ if (dev->vqs[i].ubufs)
+ vhost_ubuf_put_and_wait(dev->vqs[i].ubufs);
+
+ /* Signal guest as appropriate. */
+ vhost_zerocopy_signal_used(&dev->vqs[i]);
+
if (dev->vqs[i].error_ctx)
eventfd_ctx_put(dev->vqs[i].error_ctx);
if (dev->vqs[i].error)
@@ -1486,3 +1537,50 @@ void vhost_disable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
&vq->used->flags, r);
}
}
+
+static void vhost_zerocopy_done_signal(struct kref *kref)
+{
+ struct vhost_ubuf_ref *ubufs = container_of(kref, struct vhost_ubuf_ref,
+ kref);
+ wake_up(&ubufs->wait);
+}
+
+struct vhost_ubuf_ref *vhost_ubuf_alloc(struct vhost_virtqueue *vq,
+ bool zcopy)
+{
+ struct vhost_ubuf_ref *ubufs;
+ /* No zero copy backend? Nothing to count. */
+ if (!zcopy)
+ return NULL;
+ ubufs = kmalloc(sizeof *ubufs, GFP_KERNEL);
+ if (!ubufs)
+ return ERR_PTR(-ENOMEM);
+ kref_init(&ubufs->kref);
+ kref_get(&ubufs->kref);
+ init_waitqueue_head(&ubufs->wait);
+ ubufs->vq = vq;
+ return ubufs;
+}
+
+void vhost_ubuf_put(struct vhost_ubuf_ref *ubufs)
+{
+ kref_put(&ubufs->kref, vhost_zerocopy_done_signal);
+}
+
+void vhost_ubuf_put_and_wait(struct vhost_ubuf_ref *ubufs)
+{
+ kref_put(&ubufs->kref, vhost_zerocopy_done_signal);
+ wait_event(ubufs->wait, !atomic_read(&ubufs->kref.refcount));
+ kfree(ubufs);
+}
+
+void vhost_zerocopy_callback(void *arg)
+{
+ struct ubuf_info *ubuf = arg;
+ struct vhost_ubuf_ref *ubufs = ubuf->arg;
+ struct vhost_virtqueue *vq = ubufs->vq;
+
+ /* set len = 1 to mark this desc buffers done DMA */
+ vq->heads[ubuf->desc].len = VHOST_DMA_DONE_LEN;
+ kref_put(&ubufs->kref, vhost_zerocopy_done_signal);
+}
OpenPOWER on IntegriCloud