drivers: xen-blkback: delay pending_req allocation to connect_ring
authorBob Liu <bob.liu@oracle.com>
Wed, 3 Jun 2015 05:40:01 +0000 (13:40 +0800)
committerKonrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Sat, 6 Jun 2015 01:14:05 +0000 (21:14 -0400)
This is a pre-patch for multi-page ring feature.
In connect_ring, we can know exactly how many pages are used for the shared
ring, delay pending_req allocation here so that we won't waste too much memory.

Signed-off-by: Bob Liu <bob.liu@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
drivers/block/xen-blkback/common.h
drivers/block/xen-blkback/xenbus.c

index f620b5d3f77c0270585f5cc4d86ba71bb4320092..043f13b7b7b046a358bc9dd27ae466f40a79914c 100644 (file)
@@ -248,7 +248,7 @@ struct backend_info;
 #define PERSISTENT_GNT_WAS_ACTIVE      1
 
 /* Number of requests that we can fit in a ring */
-#define XEN_BLKIF_REQS                 32
+#define XEN_BLKIF_REQS_PER_PAGE                32
 
 struct persistent_gnt {
        struct page *page;
index 6ab69ad61ee126c6f62a63f77674b13489de7f73..c212d41fd5bd4c025daf09cd21b08067065d9153 100644 (file)
@@ -124,8 +124,6 @@ static void xen_update_blkif_status(struct xen_blkif *blkif)
 static struct xen_blkif *xen_blkif_alloc(domid_t domid)
 {
        struct xen_blkif *blkif;
-       struct pending_req *req, *n;
-       int i, j;
 
        BUILD_BUG_ON(MAX_INDIRECT_PAGES > BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST);
 
@@ -151,51 +149,11 @@ static struct xen_blkif *xen_blkif_alloc(domid_t domid)
 
        INIT_LIST_HEAD(&blkif->pending_free);
        INIT_WORK(&blkif->free_work, xen_blkif_deferred_free);
-
-       for (i = 0; i < XEN_BLKIF_REQS; i++) {
-               req = kzalloc(sizeof(*req), GFP_KERNEL);
-               if (!req)
-                       goto fail;
-               list_add_tail(&req->free_list,
-                             &blkif->pending_free);
-               for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) {
-                       req->segments[j] = kzalloc(sizeof(*req->segments[0]),
-                                                  GFP_KERNEL);
-                       if (!req->segments[j])
-                               goto fail;
-               }
-               for (j = 0; j < MAX_INDIRECT_PAGES; j++) {
-                       req->indirect_pages[j] = kzalloc(sizeof(*req->indirect_pages[0]),
-                                                        GFP_KERNEL);
-                       if (!req->indirect_pages[j])
-                               goto fail;
-               }
-       }
        spin_lock_init(&blkif->pending_free_lock);
        init_waitqueue_head(&blkif->pending_free_wq);
        init_waitqueue_head(&blkif->shutdown_wq);
 
        return blkif;
-
-fail:
-       list_for_each_entry_safe(req, n, &blkif->pending_free, free_list) {
-               list_del(&req->free_list);
-               for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) {
-                       if (!req->segments[j])
-                               break;
-                       kfree(req->segments[j]);
-               }
-               for (j = 0; j < MAX_INDIRECT_PAGES; j++) {
-                       if (!req->indirect_pages[j])
-                               break;
-                       kfree(req->indirect_pages[j]);
-               }
-               kfree(req);
-       }
-
-       kmem_cache_free(xen_blkif_cachep, blkif);
-
-       return ERR_PTR(-ENOMEM);
 }
 
 static int xen_blkif_map(struct xen_blkif *blkif, grant_ref_t gref,
@@ -312,7 +270,7 @@ static void xen_blkif_free(struct xen_blkif *blkif)
                i++;
        }
 
-       WARN_ON(i != XEN_BLKIF_REQS);
+       WARN_ON(i != XEN_BLKIF_REQS_PER_PAGE);
 
        kmem_cache_free(xen_blkif_cachep, blkif);
 }
@@ -864,7 +822,8 @@ static int connect_ring(struct backend_info *be)
        unsigned int evtchn;
        unsigned int pers_grants;
        char protocol[64] = "";
-       int err;
+       struct pending_req *req, *n;
+       int err, i, j;
 
        pr_debug("%s %s\n", __func__, dev->otherend);
 
@@ -905,6 +864,24 @@ static int connect_ring(struct backend_info *be)
                ring_ref, evtchn, be->blkif->blk_protocol, protocol,
                pers_grants ? "persistent grants" : "");
 
+       for (i = 0; i < XEN_BLKIF_REQS_PER_PAGE; i++) {
+               req = kzalloc(sizeof(*req), GFP_KERNEL);
+               if (!req)
+                       goto fail;
+               list_add_tail(&req->free_list, &be->blkif->pending_free);
+               for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) {
+                       req->segments[j] = kzalloc(sizeof(*req->segments[0]), GFP_KERNEL);
+                       if (!req->segments[j])
+                               goto fail;
+               }
+               for (j = 0; j < MAX_INDIRECT_PAGES; j++) {
+                       req->indirect_pages[j] = kzalloc(sizeof(*req->indirect_pages[0]),
+                                                        GFP_KERNEL);
+                       if (!req->indirect_pages[j])
+                               goto fail;
+               }
+       }
+
        /* Map the shared frame, irq etc. */
        err = xen_blkif_map(be->blkif, ring_ref, evtchn);
        if (err) {
@@ -914,6 +891,23 @@ static int connect_ring(struct backend_info *be)
        }
 
        return 0;
+
+fail:
+       list_for_each_entry_safe(req, n, &be->blkif->pending_free, free_list) {
+               list_del(&req->free_list);
+               for (j = 0; j < MAX_INDIRECT_SEGMENTS; j++) {
+                       if (!req->segments[j])
+                               break;
+                       kfree(req->segments[j]);
+               }
+               for (j = 0; j < MAX_INDIRECT_PAGES; j++) {
+                       if (!req->indirect_pages[j])
+                               break;
+                       kfree(req->indirect_pages[j]);
+               }
+               kfree(req);
+       }
+       return -ENOMEM;
 }
 
 static const struct xenbus_device_id xen_blkbk_ids[] = {