]> bbs.cooldavid.org Git - net-next-2.6.git/commitdiff
vhost: max s/g to match qemu
authorJason Wang <jasowang@redhat.com>
Tue, 14 Sep 2010 15:53:05 +0000 (23:53 +0800)
committerMichael S. Tsirkin <mst@redhat.com>
Tue, 5 Oct 2010 11:33:43 +0000 (13:33 +0200)
Qemu supports up to UIO_MAXIOV s/g so we have to match that because guest
drivers may rely on this.

Allocate indirect and log arrays dynamically to avoid using too much contigious
memory and make the length of hdr array to match the header length since each
iovec entry has a least one byte.

Test with copying large files w/ and w/o migration in both linux and windows
guests.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
drivers/vhost/net.c
drivers/vhost/vhost.c
drivers/vhost/vhost.h

index 7c8008225ee3a47ebef2cecaf005c0631fa93da1..72ab71fdf05339e311c14d4004f122ed0ef3828e 100644 (file)
@@ -243,7 +243,7 @@ static int get_rx_bufs(struct vhost_virtqueue *vq,
        int r, nlogs = 0;
 
        while (datalen > 0) {
-               if (unlikely(seg >= VHOST_NET_MAX_SG)) {
+               if (unlikely(seg >= UIO_MAXIOV)) {
                        r = -ENOBUFS;
                        goto err;
                }
index dd3d6f7406f80092cdd8843acdd06c313e9cf164..344019774ddde5dde56a3ccb43868084fc386eb6 100644 (file)
@@ -212,6 +212,45 @@ static int vhost_worker(void *data)
        }
 }
 
+/* Helper to allocate iovec buffers for all vqs. */
+static long vhost_dev_alloc_iovecs(struct vhost_dev *dev)
+{
+       int i;
+       for (i = 0; i < dev->nvqs; ++i) {
+               dev->vqs[i].indirect = kmalloc(sizeof *dev->vqs[i].indirect *
+                                              UIO_MAXIOV, GFP_KERNEL);
+               dev->vqs[i].log = kmalloc(sizeof *dev->vqs[i].log * UIO_MAXIOV,
+                                         GFP_KERNEL);
+               dev->vqs[i].heads = kmalloc(sizeof *dev->vqs[i].heads *
+                                           UIO_MAXIOV, GFP_KERNEL);
+
+               if (!dev->vqs[i].indirect || !dev->vqs[i].log ||
+                       !dev->vqs[i].heads)
+                       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);
+       }
+       return -ENOMEM;
+}
+
+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;
+       }
+}
+
 long vhost_dev_init(struct vhost_dev *dev,
                    struct vhost_virtqueue *vqs, int nvqs)
 {
@@ -229,6 +268,9 @@ long vhost_dev_init(struct vhost_dev *dev,
        dev->worker = NULL;
 
        for (i = 0; i < dev->nvqs; ++i) {
+               dev->vqs[i].log = NULL;
+               dev->vqs[i].indirect = NULL;
+               dev->vqs[i].heads = NULL;
                dev->vqs[i].dev = dev;
                mutex_init(&dev->vqs[i].mutex);
                vhost_vq_reset(dev, dev->vqs + i);
@@ -295,6 +337,10 @@ static long vhost_dev_set_owner(struct vhost_dev *dev)
        if (err)
                goto err_cgroup;
 
+       err = vhost_dev_alloc_iovecs(dev);
+       if (err)
+               goto err_cgroup;
+
        return 0;
 err_cgroup:
        kthread_stop(worker);
@@ -345,6 +391,7 @@ void vhost_dev_cleanup(struct vhost_dev *dev)
                        fput(dev->vqs[i].call);
                vhost_vq_reset(dev, dev->vqs + i);
        }
+       vhost_dev_free_iovecs(dev);
        if (dev->log_ctx)
                eventfd_ctx_put(dev->log_ctx);
        dev->log_ctx = NULL;
@@ -947,7 +994,7 @@ static int get_indirect(struct vhost_dev *dev, struct vhost_virtqueue *vq,
        }
 
        ret = translate_desc(dev, indirect->addr, indirect->len, vq->indirect,
-                            ARRAY_SIZE(vq->indirect));
+                            UIO_MAXIOV);
        if (unlikely(ret < 0)) {
                vq_err(vq, "Translation failure %d in indirect.\n", ret);
                return ret;
index afd77295971ce3044117d0d6e5ea8f4e20f655fe..edc892989992bdfed2a9cd015b2593f2b28616b2 100644 (file)
 
 struct vhost_device;
 
-enum {
-       /* Enough place for all fragments, head, and virtio net header. */
-       VHOST_NET_MAX_SG = MAX_SKB_FRAGS + 2,
-};
-
 struct vhost_work;
 typedef void (*vhost_work_fn_t)(struct vhost_work *work);
 
@@ -93,12 +88,15 @@ struct vhost_virtqueue {
        bool log_used;
        u64 log_addr;
 
-       struct iovec indirect[VHOST_NET_MAX_SG];
-       struct iovec iov[VHOST_NET_MAX_SG];
-       struct iovec hdr[VHOST_NET_MAX_SG];
+       struct iovec iov[UIO_MAXIOV];
+       /* hdr is used to store the virtio header.
+        * Since each iovec has >= 1 byte length, we never need more than
+        * header length entries to store the header. */
+       struct iovec hdr[sizeof(struct virtio_net_hdr_mrg_rxbuf)];
+       struct iovec *indirect;
        size_t vhost_hlen;
        size_t sock_hlen;
-       struct vring_used_elem heads[VHOST_NET_MAX_SG];
+       struct vring_used_elem *heads;
        /* We use a kind of RCU to access private pointer.
         * All readers access it from worker, which makes it possible to
         * flush the vhost_work instead of synchronize_rcu. Therefore readers do
@@ -109,7 +107,7 @@ struct vhost_virtqueue {
        void *private_data;
        /* Log write descriptors */
        void __user *log_base;
-       struct vhost_log log[VHOST_NET_MAX_SG];
+       struct vhost_log *log;
 };
 
 struct vhost_dev {