NFS/flexfiles: Speed up read failover when DSes are down

If we notice that a DS may be down, we should attempt to read from the
other mirrors first before we go back to retry the dead DS.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

3 files changed, 73 insertions, 12 deletions

diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c
index 747ab2fafb69..44ccfce3784e 100644
--- a/fs/nfs/flexfilelayout/flexfilelayout.c
+++ b/fs/nfs/flexfilelayout/flexfilelayout.c
@@ -788,10 +788,28 @@ ff_layout_alloc_commit_info(struct pnfs_layout_segment *lseg,
         }
 }
 
+static void
+ff_layout_mark_ds_unreachable(struct pnfs_layout_segment *lseg, int idx)
+{
+        struct nfs4_deviceid_node *devid = FF_LAYOUT_DEVID_NODE(lseg, idx);
+
+        if (devid)
+                nfs4_mark_deviceid_unavailable(devid);
+}
+
+static void
+ff_layout_mark_ds_reachable(struct pnfs_layout_segment *lseg, int idx)
+{
+        struct nfs4_deviceid_node *devid = FF_LAYOUT_DEVID_NODE(lseg, idx);
+
+        if (devid)
+                nfs4_mark_deviceid_available(devid);
+}
+
 static struct nfs4_pnfs_ds *
-ff_layout_choose_best_ds_for_read(struct pnfs_layout_segment *lseg,
-                                  int start_idx,
-                                  int *best_idx)
+ff_layout_choose_ds_for_read(struct pnfs_layout_segment *lseg,
+                             int start_idx, int *best_idx,
+                             bool check_device)
 {
         struct nfs4_ff_layout_segment *fls = FF_LAYOUT_LSEG(lseg);
         struct nfs4_ff_layout_mirror *mirror;
@@ -799,25 +817,53 @@ ff_layout_choose_best_ds_for_read(struct pnfs_layout_segment *lseg,
         bool fail_return = false;
         int idx;
 
-        /* mirrors are sorted by efficiency */
+        /* mirrors are initially sorted by efficiency */
         for (idx = start_idx; idx < fls->mirror_array_cnt; idx++) {
                 if (idx+1 == fls->mirror_array_cnt)
-                        fail_return = true;
+                        fail_return = !check_device;
 
                 mirror = FF_LAYOUT_COMP(lseg, idx);
-                if (ff_layout_test_devid_unavailable(&mirror->mirror_ds->id_node))
+                ds = nfs4_ff_layout_prepare_ds(lseg, mirror, fail_return);
+                if (!ds)
                         continue;
 
-                ds = nfs4_ff_layout_prepare_ds(lseg, mirror, fail_return);
-                if (ds) {
-                        *best_idx = idx;
-                        return ds;
-                }
+                if (check_device &&
+                    nfs4_test_deviceid_unavailable(&mirror->mirror_ds->id_node))
+                        continue;
+
+                *best_idx = idx;
+                return ds;
         }
 
         return NULL;
 }
 
+static struct nfs4_pnfs_ds *
+ff_layout_choose_any_ds_for_read(struct pnfs_layout_segment *lseg,
+                                 int start_idx, int *best_idx)
+{
+        return ff_layout_choose_ds_for_read(lseg, start_idx, best_idx, false);
+}
+
+static struct nfs4_pnfs_ds *
+ff_layout_choose_valid_ds_for_read(struct pnfs_layout_segment *lseg,
+                                   int start_idx, int *best_idx)
+{
+        return ff_layout_choose_ds_for_read(lseg, start_idx, best_idx, true);
+}
+
+static struct nfs4_pnfs_ds *
+ff_layout_choose_best_ds_for_read(struct pnfs_layout_segment *lseg,
+                                  int start_idx, int *best_idx)
+{
+        struct nfs4_pnfs_ds *ds;
+
+        ds = ff_layout_choose_valid_ds_for_read(lseg, start_idx, best_idx);
+        if (ds)
+                return ds;
+        return ff_layout_choose_any_ds_for_read(lseg, start_idx, best_idx);
+}
+
 static void
 ff_layout_pg_get_read(struct nfs_pageio_descriptor *pgio,
                       struct nfs_page *req,
@@ -1167,8 +1213,10 @@ static int ff_layout_async_handle_error(struct rpc_task *task,
 {
         int vers = clp->cl_nfs_mod->rpc_vers->number;
 
-        if (task->tk_status >= 0)
+        if (task->tk_status >= 0) {
+                ff_layout_mark_ds_reachable(lseg, idx);
                 return 0;
+        }
 
         /* Handle the case of an invalid layout segment */
         if (!pnfs_is_valid_lseg(lseg))
@@ -1231,6 +1279,8 @@ static void ff_layout_io_track_ds_error(struct pnfs_layout_segment *lseg,
         err = ff_layout_track_ds_error(FF_LAYOUT_FROM_HDR(lseg->pls_layout),
                                        mirror, offset, length, status, opnum,
                                        GFP_NOIO);
+        if (status == NFS4ERR_NXIO)
+                ff_layout_mark_ds_unreachable(lseg, idx);
         pnfs_error_mark_layout_for_return(lseg->pls_layout->plh_inode, lseg);
         dprintk("%s: err %d op %d status %u\n", __func__, err, opnum, status);
 }
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 56659ccce1d8..c0420b979d88 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -350,6 +350,7 @@ void nfs4_delete_deviceid(const struct pnfs_layoutdriver_type *, const struct nf
 void nfs4_init_deviceid_node(struct nfs4_deviceid_node *, struct nfs_server *,
                              const struct nfs4_deviceid *);
 bool nfs4_put_deviceid_node(struct nfs4_deviceid_node *);
+void nfs4_mark_deviceid_available(struct nfs4_deviceid_node *node);
 void nfs4_mark_deviceid_unavailable(struct nfs4_deviceid_node *node);
 bool nfs4_test_deviceid_unavailable(struct nfs4_deviceid_node *node);
 void nfs4_deviceid_purge_client(const struct nfs_client *);
diff --git a/fs/nfs/pnfs_dev.c b/fs/nfs/pnfs_dev.c
index ec6b607ff1d7..537b80d693f1 100644
--- a/fs/nfs/pnfs_dev.c
+++ b/fs/nfs/pnfs_dev.c
@@ -284,6 +284,16 @@ nfs4_put_deviceid_node(struct nfs4_deviceid_node *d)
 EXPORT_SYMBOL_GPL(nfs4_put_deviceid_node);
 
 void
+nfs4_mark_deviceid_available(struct nfs4_deviceid_node *node)
+{
+        if (test_bit(NFS_DEVICEID_UNAVAILABLE, &node->flags)) {
+                clear_bit(NFS_DEVICEID_UNAVAILABLE, &node->flags);
+                smp_mb__after_atomic();
+        }
+}
+EXPORT_SYMBOL_GPL(nfs4_mark_deviceid_available);
+
+void
 nfs4_mark_deviceid_unavailable(struct nfs4_deviceid_node *node)
 {
         node->timestamp_unavailable = jiffies;