tcm_vhost: Wait for pending requests in vhost_scsi_flush()
authorAsias He <asias@redhat.com>
Sat, 27 Apr 2013 03:16:49 +0000 (11:16 +0800)
committerMichael S. Tsirkin <mst@redhat.com>
Wed, 1 May 2013 07:02:51 +0000 (10:02 +0300)
Unlike tcm_vhost_evt requests, tcm_vhost_cmd requests are passed to the
target core system, we can not make sure all the pending requests will
be finished by flushing the virt queue.

In this patch, we do refcount for every tcm_vhost_cmd requests to make
vhost_scsi_flush() wait for all the pending requests issued before the
flush operation to be finished.

This is useful when we call vhost_scsi_clear_endpoint() to stop
tcm_vhost. No new requests will be passed to target core system because
we clear the endpoint by setting vs_tpg to NULL. And we wait for all the
old requests. These guarantee no requests will be leaked and existing
requests will be completed.

Signed-off-by: Asias He <asias@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
drivers/vhost/tcm_vhost.c
drivers/vhost/tcm_vhost.h

index 99d3480..afb5308 100644 (file)
@@ -74,8 +74,19 @@ enum {
 #define VHOST_SCSI_MAX_VQ      128
 #define VHOST_SCSI_MAX_EVENT   128
 
+struct vhost_scsi_inflight {
+       /* Wait for the flush operation to finish */
+       struct completion comp;
+       /* Refcount for the inflight reqs */
+       struct kref kref;
+};
+
 struct vhost_scsi_virtqueue {
        struct vhost_virtqueue vq;
+       /* Track inflight reqs, protected by vq->mutex */
+       struct vhost_scsi_inflight inflights[2];
+       /* Indicate current inflight in use, protected by vq->mutex */
+       int inflight_idx;
 };
 
 struct vhost_scsi {
@@ -111,6 +122,59 @@ static int iov_num_pages(struct iovec *iov)
               ((unsigned long)iov->iov_base & PAGE_MASK)) >> PAGE_SHIFT;
 }
 
+void tcm_vhost_done_inflight(struct kref *kref)
+{
+       struct vhost_scsi_inflight *inflight;
+
+       inflight = container_of(kref, struct vhost_scsi_inflight, kref);
+       complete(&inflight->comp);
+}
+
+static void tcm_vhost_init_inflight(struct vhost_scsi *vs,
+                                   struct vhost_scsi_inflight *old_inflight[])
+{
+       struct vhost_scsi_inflight *new_inflight;
+       struct vhost_virtqueue *vq;
+       int idx, i;
+
+       for (i = 0; i < VHOST_SCSI_MAX_VQ; i++) {
+               vq = &vs->vqs[i].vq;
+
+               mutex_lock(&vq->mutex);
+
+               /* store old infight */
+               idx = vs->vqs[i].inflight_idx;
+               if (old_inflight)
+                       old_inflight[i] = &vs->vqs[i].inflights[idx];
+
+               /* setup new infight */
+               vs->vqs[i].inflight_idx = idx ^ 1;
+               new_inflight = &vs->vqs[i].inflights[idx ^ 1];
+               kref_init(&new_inflight->kref);
+               init_completion(&new_inflight->comp);
+
+               mutex_unlock(&vq->mutex);
+       }
+}
+
+static struct vhost_scsi_inflight *
+tcm_vhost_get_inflight(struct vhost_virtqueue *vq)
+{
+       struct vhost_scsi_inflight *inflight;
+       struct vhost_scsi_virtqueue *svq;
+
+       svq = container_of(vq, struct vhost_scsi_virtqueue, vq);
+       inflight = &svq->inflights[svq->inflight_idx];
+       kref_get(&inflight->kref);
+
+       return inflight;
+}
+
+static void tcm_vhost_put_inflight(struct vhost_scsi_inflight *inflight)
+{
+       kref_put(&inflight->kref, tcm_vhost_done_inflight);
+}
+
 static int tcm_vhost_check_true(struct se_portal_group *se_tpg)
 {
        return 1;
@@ -407,6 +471,8 @@ static void vhost_scsi_free_cmd(struct tcm_vhost_cmd *tv_cmd)
                kfree(tv_cmd->tvc_sgl);
        }
 
+       tcm_vhost_put_inflight(tv_cmd->inflight);
+
        kfree(tv_cmd);
 }
 
@@ -533,6 +599,7 @@ static void vhost_scsi_complete_cmd_work(struct vhost_work *work)
 }
 
 static struct tcm_vhost_cmd *vhost_scsi_allocate_cmd(
+       struct vhost_virtqueue *vq,
        struct tcm_vhost_tpg *tv_tpg,
        struct virtio_scsi_cmd_req *v_req,
        u32 exp_data_len,
@@ -557,6 +624,7 @@ static struct tcm_vhost_cmd *vhost_scsi_allocate_cmd(
        tv_cmd->tvc_exp_data_len = exp_data_len;
        tv_cmd->tvc_data_direction = data_direction;
        tv_cmd->tvc_nexus = tv_nexus;
+       tv_cmd->inflight = tcm_vhost_get_inflight(vq);
 
        return tv_cmd;
 }
@@ -812,7 +880,7 @@ static void vhost_scsi_handle_vq(struct vhost_scsi *vs,
                for (i = 0; i < data_num; i++)
                        exp_data_len += vq->iov[data_first + i].iov_len;
 
-               tv_cmd = vhost_scsi_allocate_cmd(tv_tpg, &v_req,
+               tv_cmd = vhost_scsi_allocate_cmd(vq, tv_tpg, &v_req,
                                        exp_data_len, data_direction);
                if (IS_ERR(tv_cmd)) {
                        vq_err(vq, "vhost_scsi_allocate_cmd failed %ld\n",
@@ -949,12 +1017,29 @@ static void vhost_scsi_flush_vq(struct vhost_scsi *vs, int index)
 
 static void vhost_scsi_flush(struct vhost_scsi *vs)
 {
+       struct vhost_scsi_inflight *old_inflight[VHOST_SCSI_MAX_VQ];
        int i;
 
+       /* Init new inflight and remember the old inflight */
+       tcm_vhost_init_inflight(vs, old_inflight);
+
+       /*
+        * The inflight->kref was initialized to 1. We decrement it here to
+        * indicate the start of the flush operation so that it will reach 0
+        * when all the reqs are finished.
+        */
+       for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
+               kref_put(&old_inflight[i]->kref, tcm_vhost_done_inflight);
+
+       /* Flush both the vhost poll and vhost work */
        for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
                vhost_scsi_flush_vq(vs, i);
        vhost_work_flush(&vs->dev, &vs->vs_completion_work);
        vhost_work_flush(&vs->dev, &vs->vs_event_work);
+
+       /* Wait for all reqs issued before the flush to be finished */
+       for (i = 0; i < VHOST_SCSI_MAX_VQ; i++)
+               wait_for_completion(&old_inflight[i]->comp);
 }
 
 /*
@@ -1185,6 +1270,9 @@ static int vhost_scsi_open(struct inode *inode, struct file *f)
                s->vqs[i].vq.handle_kick = vhost_scsi_handle_kick;
        }
        r = vhost_dev_init(&s->dev, vqs, VHOST_SCSI_MAX_VQ);
+
+       tcm_vhost_init_inflight(s, NULL);
+
        if (r < 0) {
                kfree(vqs);
                kfree(s);
index 514b9fd..26a57c2 100644 (file)
@@ -2,6 +2,7 @@
 #define TCM_VHOST_NAMELEN 256
 #define TCM_VHOST_MAX_CDB_SIZE 32
 
+struct vhost_scsi_inflight;
 struct tcm_vhost_cmd {
        /* Descriptor from vhost_get_vq_desc() for virt_queue segment */
        int tvc_vq_desc;
@@ -37,6 +38,8 @@ struct tcm_vhost_cmd {
        unsigned char tvc_sense_buf[TRANSPORT_SENSE_BUFFER];
        /* Completed commands list, serviced from vhost worker thread */
        struct llist_node tvc_completion_list;
+       /* Used to track inflight cmd */
+       struct vhost_scsi_inflight *inflight;
 };
 
 struct tcm_vhost_nexus {