1 files changed, 79 insertions, 57 deletions

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 07f59dc8cb2e..af64b4e6fd1f 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -324,6 +324,21 @@ pnfs_grab_inode_layout_hdr(struct pnfs_layout_hdr *lo)
 	return NULL;
 }
 
+/*
+ * Compare 2 layout stateid sequence ids, to see which is newer,
+ * taking into account wraparound issues.
+ */
+static bool pnfs_seqid_is_newer(u32 s1, u32 s2)
+{
+	return (s32)(s1 - s2) > 0;
+}
+
+static void pnfs_barrier_update(struct pnfs_layout_hdr *lo, u32 newseq)
+{
+	if (pnfs_seqid_is_newer(newseq, lo->plh_barrier))
+		lo->plh_barrier = newseq;
+}
+
 static void
 pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 			 u32 seq)
@@ -335,6 +350,7 @@ pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 	if (seq != 0) {
 		WARN_ON_ONCE(lo->plh_return_seq != 0 && lo->plh_return_seq != seq);
 		lo->plh_return_seq = seq;
+		pnfs_barrier_update(lo, seq);
 	}
 }
 
@@ -639,15 +655,6 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg,
 	return rv;
 }
 
-/*
- * Compare 2 layout stateid sequence ids, to see which is newer,
- * taking into account wraparound issues.
- */
-static bool pnfs_seqid_is_newer(u32 s1, u32 s2)
-{
-	return (s32)(s1 - s2) > 0;
-}
-
 static bool
 pnfs_should_free_range(const struct pnfs_layout_range *lseg_range,
 		 const struct pnfs_layout_range *recall_range)
@@ -984,8 +991,7 @@ pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, const nfs4_stateid *new,
 		new_barrier = be32_to_cpu(new->seqid);
 	else if (new_barrier == 0)
 		return;
-	if (pnfs_seqid_is_newer(new_barrier, lo->plh_barrier))
-		lo->plh_barrier = new_barrier;
+	pnfs_barrier_update(lo, new_barrier);
 }
 
 static bool
@@ -994,7 +1000,7 @@ pnfs_layout_stateid_blocked(const struct pnfs_layout_hdr *lo,
 {
 	u32 seqid = be32_to_cpu(stateid->seqid);
 
-	return !pnfs_seqid_is_newer(seqid, lo->plh_barrier);
+	return !pnfs_seqid_is_newer(seqid, lo->plh_barrier) && lo->plh_barrier;
 }
 
 /* lget is set to 1 if called from inside send_layoutget call chain */
@@ -1152,7 +1158,7 @@ void pnfs_layoutreturn_free_lsegs(struct pnfs_layout_hdr *lo,
 	LIST_HEAD(freeme);
 
 	spin_lock(&inode->i_lock);
-	if (!pnfs_layout_is_valid(lo) || !arg_stateid ||
+	if (!pnfs_layout_is_valid(lo) ||
 	    !nfs4_stateid_match_other(&lo->plh_stateid, arg_stateid))
 		goto out_unlock;
 	if (stateid) {
@@ -1183,20 +1189,17 @@ pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo,
 		return false;
 	set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags);
 	pnfs_get_layout_hdr(lo);
+	nfs4_stateid_copy(stateid, &lo->plh_stateid);
+	*cred = get_cred(lo->plh_lc_cred);
 	if (test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags)) {
-		nfs4_stateid_copy(stateid, &lo->plh_stateid);
-		*cred = get_cred(lo->plh_lc_cred);
 		if (lo->plh_return_seq != 0)
 			stateid->seqid = cpu_to_be32(lo->plh_return_seq);
 		if (iomode != NULL)
 			*iomode = lo->plh_return_iomode;
 		pnfs_clear_layoutreturn_info(lo);
-		return true;
-	}
-	nfs4_stateid_copy(stateid, &lo->plh_stateid);
-	*cred = get_cred(lo->plh_lc_cred);
-	if (iomode != NULL)
+	} else if (iomode != NULL)
 		*iomode = IOMODE_ANY;
+	pnfs_barrier_update(lo, be32_to_cpu(stateid->seqid));
 	return true;
 }
 
@@ -1509,10 +1512,8 @@ out_noroc:
 	return false;
 }
 
-int pnfs_roc_done(struct rpc_task *task, struct inode *inode,
-		struct nfs4_layoutreturn_args **argpp,
-		struct nfs4_layoutreturn_res **respp,
-		int *ret)
+int pnfs_roc_done(struct rpc_task *task, struct nfs4_layoutreturn_args **argpp,
+		  struct nfs4_layoutreturn_res **respp, int *ret)
 {
 	struct nfs4_layoutreturn_args *arg = *argpp;
 	int retval = -EAGAIN;
@@ -1545,7 +1546,7 @@ int pnfs_roc_done(struct rpc_task *task, struct inode *inode,
 		return 0;
 	case -NFS4ERR_OLD_STATEID:
 		if (!nfs4_layout_refresh_old_stateid(&arg->stateid,
-					&arg->range, inode))
+						     &arg->range, arg->inode))
 			break;
 		*ret = -NFS4ERR_NOMATCHING_LAYOUT;
 		return -EAGAIN;
@@ -1560,23 +1561,28 @@ void pnfs_roc_release(struct nfs4_layoutreturn_args *args,
 		int ret)
 {
 	struct pnfs_layout_hdr *lo = args->layout;
-	const nfs4_stateid *arg_stateid = NULL;
+	struct inode *inode = args->inode;
 	const nfs4_stateid *res_stateid = NULL;
 	struct nfs4_xdr_opaque_data *ld_private = args->ld_private;
 
 	switch (ret) {
 	case -NFS4ERR_NOMATCHING_LAYOUT:
+		spin_lock(&inode->i_lock);
+		if (pnfs_layout_is_valid(lo) &&
+		    nfs4_stateid_match_other(&args->stateid, &lo->plh_stateid))
+			pnfs_set_plh_return_info(lo, args->range.iomode, 0);
+		pnfs_clear_layoutreturn_waitbit(lo);
+		spin_unlock(&inode->i_lock);
 		break;
 	case 0:
 		if (res->lrs_present)
 			res_stateid = &res->stateid;
 		fallthrough;
 	default:
-		arg_stateid = &args->stateid;
+		pnfs_layoutreturn_free_lsegs(lo, &args->stateid, &args->range,
+					     res_stateid);
 	}
 	trace_nfs4_layoutreturn_on_close(args->inode, &args->stateid, ret);
-	pnfs_layoutreturn_free_lsegs(lo, arg_stateid, &args->range,
-			res_stateid);
 	if (ld_private && ld_private->ops && ld_private->ops->free)
 		ld_private->ops->free(ld_private);
 	pnfs_put_layout_hdr(lo);
@@ -1906,6 +1912,11 @@ static void nfs_layoutget_end(struct pnfs_layout_hdr *lo)
 		wake_up_var(&lo->plh_outstanding);
 }
 
+static bool pnfs_is_first_layoutget(struct pnfs_layout_hdr *lo)
+{
+	return test_bit(NFS_LAYOUT_FIRST_LAYOUTGET, &lo->plh_flags);
+}
+
 static void pnfs_clear_first_layoutget(struct pnfs_layout_hdr *lo)
 {
 	unsigned long *bitlock = &lo->plh_flags;
@@ -2015,6 +2026,27 @@ lookup_again:
 		goto lookup_again;
 	}
 
+	/*
+	 * Because we free lsegs when sending LAYOUTRETURN, we need to wait
+	 * for LAYOUTRETURN.
+	 */
+	if (test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) {
+		spin_unlock(&ino->i_lock);
+		dprintk("%s wait for layoutreturn\n", __func__);
+		lseg = ERR_PTR(pnfs_prepare_to_retry_layoutget(lo));
+		if (!IS_ERR(lseg)) {
+			pnfs_put_layout_hdr(lo);
+			dprintk("%s retrying\n", __func__);
+			trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+						 lseg,
+						 PNFS_UPDATE_LAYOUT_RETRY);
+			goto lookup_again;
+		}
+		trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg,
+					 PNFS_UPDATE_LAYOUT_RETURN);
+		goto out_put_layout_hdr;
+	}
+
 	lseg = pnfs_find_lseg(lo, &arg, strict_iomode);
 	if (lseg) {
 		trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg,
@@ -2067,28 +2099,6 @@ lookup_again:
 		nfs4_stateid_copy(&stateid, &lo->plh_stateid);
 	}
 
-	/*
-	 * Because we free lsegs before sending LAYOUTRETURN, we need to wait
-	 * for LAYOUTRETURN even if first is true.
-	 */
-	if (test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags)) {
-		spin_unlock(&ino->i_lock);
-		dprintk("%s wait for layoutreturn\n", __func__);
-		lseg = ERR_PTR(pnfs_prepare_to_retry_layoutget(lo));
-		if (!IS_ERR(lseg)) {
-			if (first)
-				pnfs_clear_first_layoutget(lo);
-			pnfs_put_layout_hdr(lo);
-			dprintk("%s retrying\n", __func__);
-			trace_pnfs_update_layout(ino, pos, count, iomode, lo,
-					lseg, PNFS_UPDATE_LAYOUT_RETRY);
-			goto lookup_again;
-		}
-		trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg,
-				PNFS_UPDATE_LAYOUT_RETURN);
-		goto out_put_layout_hdr;
-	}
-
 	if (pnfs_layoutgets_blocked(lo)) {
 		trace_pnfs_update_layout(ino, pos, count, iomode, lo, lseg,
 				PNFS_UPDATE_LAYOUT_BLOCKED);
@@ -2242,6 +2252,7 @@ static void _lgopen_prepare_attached(struct nfs4_opendata *data,
 					     &rng, GFP_KERNEL);
 	if (!lgp) {
 		pnfs_clear_first_layoutget(lo);
+		nfs_layoutget_end(lo);
 		pnfs_put_layout_hdr(lo);
 		return;
 	}
@@ -2380,23 +2391,34 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 		goto out_forget;
 	}
 
-	if (!pnfs_layout_is_valid(lo)) {
-		/* We have a completely new layout */
-		pnfs_set_layout_stateid(lo, &res->stateid, lgp->cred, true);
-	} else if (nfs4_stateid_match_other(&lo->plh_stateid, &res->stateid)) {
+	if (nfs4_stateid_match_other(&lo->plh_stateid, &res->stateid)) {
 		/* existing state ID, make sure the sequence number matches. */
 		if (pnfs_layout_stateid_blocked(lo, &res->stateid)) {
+			if (!pnfs_layout_is_valid(lo) &&
+			    pnfs_is_first_layoutget(lo))
+				lo->plh_barrier = 0;
 			dprintk("%s forget reply due to sequence\n", __func__);
 			goto out_forget;
 		}
 		pnfs_set_layout_stateid(lo, &res->stateid, lgp->cred, false);
-	} else {
+	} else if (pnfs_layout_is_valid(lo)) {
 		/*
 		 * We got an entirely new state ID.  Mark all segments for the
 		 * inode invalid, and retry the layoutget
 		 */
-		pnfs_mark_layout_stateid_invalid(lo, &free_me);
+		struct pnfs_layout_range range = {
+			.iomode = IOMODE_ANY,
+			.length = NFS4_MAX_UINT64,
+		};
+		pnfs_set_plh_return_info(lo, IOMODE_ANY, 0);
+		pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs,
+						&range, 0);
 		goto out_forget;
+	} else {
+		/* We have a completely new layout */
+		if (!pnfs_is_first_layoutget(lo))
+			goto out_forget;
+		pnfs_set_layout_stateid(lo, &res->stateid, lgp->cred, true);
 	}
 
 	pnfs_get_lseg(lseg);