Merge branch 'pnfs'

6 files changed, 218 insertions, 136 deletions

diff --git a/fs/nfs/blocklayout/dev.c b/fs/nfs/blocklayout/dev.c
index e5b89675263e..a69ef4e9c24c 100644
--- a/fs/nfs/blocklayout/dev.c
+++ b/fs/nfs/blocklayout/dev.c
@@ -65,8 +65,8 @@ nfs4_block_decode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b)
 		if (!p)
 			return -EIO;
 		b->simple.nr_sigs = be32_to_cpup(p++);
-		if (!b->simple.nr_sigs) {
-			dprintk("no signature\n");
+		if (!b->simple.nr_sigs || b->simple.nr_sigs > PNFS_BLOCK_MAX_UUIDS) {
+			dprintk("Bad signature count: %d\n", b->simple.nr_sigs);
 			return -EIO;
 		}
 
@@ -89,7 +89,8 @@ nfs4_block_decode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b)
 			memcpy(&b->simple.sigs[i].sig, p,
 				b->simple.sigs[i].sig_len);
 
-			b->simple.len += 8 + 4 + b->simple.sigs[i].sig_len;
+			b->simple.len += 8 + 4 + \
+				(XDR_QUADLEN(b->simple.sigs[i].sig_len) << 2);
 		}
 		break;
 	case PNFS_BLOCK_VOLUME_SLICE:
@@ -104,7 +105,12 @@ nfs4_block_decode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b)
 		p = xdr_inline_decode(xdr, 4);
 		if (!p)
 			return -EIO;
+
 		b->concat.volumes_count = be32_to_cpup(p++);
+		if (b->concat.volumes_count > PNFS_BLOCK_MAX_DEVICES) {
+			dprintk("Too many volumes: %d\n", b->concat.volumes_count);
+			return -EIO;
+		}
 
 		p = xdr_inline_decode(xdr, b->concat.volumes_count * 4);
 		if (!p)
@@ -116,8 +122,13 @@ nfs4_block_decode_volume(struct xdr_stream *xdr, struct pnfs_block_volume *b)
 		p = xdr_inline_decode(xdr, 8 + 4);
 		if (!p)
 			return -EIO;
+
 		p = xdr_decode_hyper(p, &b->stripe.chunk_size);
 		b->stripe.volumes_count = be32_to_cpup(p++);
+		if (b->stripe.volumes_count > PNFS_BLOCK_MAX_DEVICES) {
+			dprintk("Too many volumes: %d\n", b->stripe.volumes_count);
+			return -EIO;
+		}
 
 		p = xdr_inline_decode(xdr, b->stripe.volumes_count * 4);
 		if (!p)
@@ -224,18 +235,20 @@ bl_parse_simple(struct nfs_server *server, struct pnfs_block_dev *d,
 		struct pnfs_block_volume *volumes, int idx, gfp_t gfp_mask)
 {
 	struct pnfs_block_volume *v = &volumes[idx];
+	struct block_device *bdev;
 	dev_t dev;
 
 	dev = bl_resolve_deviceid(server, v, gfp_mask);
 	if (!dev)
 		return -EIO;
 
-	d->bdev = blkdev_get_by_dev(dev, FMODE_READ | FMODE_WRITE, NULL);
-	if (IS_ERR(d->bdev)) {
+	bdev = blkdev_get_by_dev(dev, FMODE_READ | FMODE_WRITE, NULL);
+	if (IS_ERR(bdev)) {
 		printk(KERN_WARNING "pNFS: failed to open device %d:%d (%ld)\n",
-			MAJOR(dev), MINOR(dev), PTR_ERR(d->bdev));
-		return PTR_ERR(d->bdev);
+			MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
+		return PTR_ERR(bdev);
 	}
+	d->bdev = bdev;
 
 
 	d->len = i_size_read(d->bdev->bd_inode);
@@ -287,44 +300,71 @@ bl_validate_designator(struct pnfs_block_volume *v)
 	}
 }
 
+/*
+ * Try to open the udev path for the WWN.  At least on Debian the udev
+ * by-id path will always point to the dm-multipath device if one exists.
+ */
+static struct block_device *
+bl_open_udev_path(struct pnfs_block_volume *v)
+{
+	struct block_device *bdev;
+	const char *devname;
+
+	devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%*phN",
+				v->scsi.designator_len, v->scsi.designator);
+	if (!devname)
+		return ERR_PTR(-ENOMEM);
+
+	bdev = blkdev_get_by_path(devname, FMODE_READ | FMODE_WRITE, NULL);
+	if (IS_ERR(bdev)) {
+		pr_warn("pNFS: failed to open device %s (%ld)\n",
+			devname, PTR_ERR(bdev));
+	}
+
+	kfree(devname);
+	return bdev;
+}
+
+/*
+ * Try to open the RH/Fedora specific dm-mpath udev path for this WWN, as the
+ * wwn- links will only point to the first discovered SCSI device there.
+ */
+static struct block_device *
+bl_open_dm_mpath_udev_path(struct pnfs_block_volume *v)
+{
+	struct block_device *bdev;
+	const char *devname;
+
+	devname = kasprintf(GFP_KERNEL,
+			"/dev/disk/by-id/dm-uuid-mpath-%d%*phN",
+			v->scsi.designator_type,
+			v->scsi.designator_len, v->scsi.designator);
+	if (!devname)
+		return ERR_PTR(-ENOMEM);
+
+	bdev = blkdev_get_by_path(devname, FMODE_READ | FMODE_WRITE, NULL);
+	kfree(devname);
+	return bdev;
+}
+
 static int
 bl_parse_scsi(struct nfs_server *server, struct pnfs_block_dev *d,
 		struct pnfs_block_volume *volumes, int idx, gfp_t gfp_mask)
 {
 	struct pnfs_block_volume *v = &volumes[idx];
+	struct block_device *bdev;
 	const struct pr_ops *ops;
-	const char *devname;
 	int error;
 
 	if (!bl_validate_designator(v))
 		return -EINVAL;
 
-	switch (v->scsi.designator_len) {
-	case 8:
-		devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%8phN",
-				v->scsi.designator);
-		break;
-	case 12:
-		devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%12phN",
-				v->scsi.designator);
-		break;
-	case 16:
-		devname = kasprintf(GFP_KERNEL, "/dev/disk/by-id/wwn-0x%16phN",
-				v->scsi.designator);
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	d->bdev = blkdev_get_by_path(devname, FMODE_READ, NULL);
-	if (IS_ERR(d->bdev)) {
-		pr_warn("pNFS: failed to open device %s (%ld)\n",
-			devname, PTR_ERR(d->bdev));
-		kfree(devname);
-		return PTR_ERR(d->bdev);
-	}
-
-	kfree(devname);
+	bdev = bl_open_dm_mpath_udev_path(v);
+	if (IS_ERR(bdev))
+		bdev = bl_open_udev_path(v);
+	if (IS_ERR(bdev))
+		return PTR_ERR(bdev);
+	d->bdev = bdev;
 
 	d->len = i_size_read(d->bdev->bd_inode);
 	d->map = bl_map_simple;
@@ -352,7 +392,7 @@ bl_parse_scsi(struct nfs_server *server, struct pnfs_block_dev *d,
 	return 0;
 
 out_blkdev_put:
-	blkdev_put(d->bdev, FMODE_READ);
+	blkdev_put(d->bdev, FMODE_READ | FMODE_WRITE);
 	return error;
 }
 
diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c
index aaa2e8d3df6f..c92a75e066a6 100644
--- a/fs/nfs/callback_proc.c
+++ b/fs/nfs/callback_proc.c
@@ -119,27 +119,30 @@ out:
  * hashed by filehandle.
  */
 static struct pnfs_layout_hdr * get_layout_by_fh_locked(struct nfs_client *clp,
-		struct nfs_fh *fh, nfs4_stateid *stateid)
+		struct nfs_fh *fh)
 {
 	struct nfs_server *server;
+	struct nfs_inode *nfsi;
 	struct inode *ino;
 	struct pnfs_layout_hdr *lo;
 
+restart:
 	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
 		list_for_each_entry(lo, &server->layouts, plh_layouts) {
-			if (!nfs4_stateid_match_other(&lo->plh_stateid, stateid))
+			nfsi = NFS_I(lo->plh_inode);
+			if (nfs_compare_fh(fh, &nfsi->fh))
 				continue;
-			if (nfs_compare_fh(fh, &NFS_I(lo->plh_inode)->fh))
+			if (nfsi->layout != lo)
 				continue;
 			ino = igrab(lo->plh_inode);
 			if (!ino)
 				break;
 			spin_lock(&ino->i_lock);
 			/* Is this layout in the process of being freed? */
-			if (NFS_I(ino)->layout != lo) {
+			if (nfsi->layout != lo) {
 				spin_unlock(&ino->i_lock);
 				iput(ino);
-				break;
+				goto restart;
 			}
 			pnfs_get_layout_hdr(lo);
 			spin_unlock(&ino->i_lock);
@@ -151,13 +154,13 @@ static struct pnfs_layout_hdr * get_layout_by_fh_locked(struct nfs_client *clp,
 }
 
 static struct pnfs_layout_hdr * get_layout_by_fh(struct nfs_client *clp,
-		struct nfs_fh *fh, nfs4_stateid *stateid)
+		struct nfs_fh *fh)
 {
 	struct pnfs_layout_hdr *lo;
 
 	spin_lock(&clp->cl_lock);
 	rcu_read_lock();
-	lo = get_layout_by_fh_locked(clp, fh, stateid);
+	lo = get_layout_by_fh_locked(clp, fh);
 	rcu_read_unlock();
 	spin_unlock(&clp->cl_lock);
 
@@ -167,17 +170,39 @@ static struct pnfs_layout_hdr * get_layout_by_fh(struct nfs_client *clp,
 /*
  * Enforce RFC5661 section 12.5.5.2.1. (Layout Recall and Return Sequencing)
  */
-static bool pnfs_check_stateid_sequence(struct pnfs_layout_hdr *lo,
+static u32 pnfs_check_callback_stateid(struct pnfs_layout_hdr *lo,
 					const nfs4_stateid *new)
 {
 	u32 oldseq, newseq;
 
-	oldseq = be32_to_cpu(lo->plh_stateid.seqid);
+	/* Is the stateid still not initialised? */
+	if (!pnfs_layout_is_valid(lo))
+		return NFS4ERR_DELAY;
+
+	/* Mismatched stateid? */
+	if (!nfs4_stateid_match_other(&lo->plh_stateid, new))
+		return NFS4ERR_BAD_STATEID;
+
 	newseq = be32_to_cpu(new->seqid);
+	/* Are we already in a layout recall situation? */
+	if (test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags) &&
+	    lo->plh_return_seq != 0) {
+		if (newseq < lo->plh_return_seq)
+			return NFS4ERR_OLD_STATEID;
+		if (newseq > lo->plh_return_seq)
+			return NFS4ERR_DELAY;
+		goto out;
+	}
 
+	/* Check that the stateid matches what we think it should be. */
+	oldseq = be32_to_cpu(lo->plh_stateid.seqid);
 	if (newseq > oldseq + 1)
-		return false;
-	return true;
+		return NFS4ERR_DELAY;
+	/* Crazy server! */
+	if (newseq <= oldseq)
+		return NFS4ERR_OLD_STATEID;
+out:
+	return NFS_OK;
 }
 
 static u32 initiate_file_draining(struct nfs_client *clp,
@@ -188,7 +213,7 @@ static u32 initiate_file_draining(struct nfs_client *clp,
 	u32 rv = NFS4ERR_NOMATCHING_LAYOUT;
 	LIST_HEAD(free_me_list);
 
-	lo = get_layout_by_fh(clp, &args->cbl_fh, &args->cbl_stateid);
+	lo = get_layout_by_fh(clp, &args->cbl_fh);
 	if (!lo) {
 		trace_nfs4_cb_layoutrecall_file(clp, &args->cbl_fh, NULL,
 				&args->cbl_stateid, -rv);
@@ -196,18 +221,15 @@ static u32 initiate_file_draining(struct nfs_client *clp,
 	}
 
 	ino = lo->plh_inode;
+	pnfs_layoutcommit_inode(ino, false);
+
 
 	spin_lock(&ino->i_lock);
-	if (!pnfs_check_stateid_sequence(lo, &args->cbl_stateid)) {
-		rv = NFS4ERR_DELAY;
+	rv = pnfs_check_callback_stateid(lo, &args->cbl_stateid);
+	if (rv != NFS_OK)
 		goto unlock;
-	}
 	pnfs_set_layout_stateid(lo, &args->cbl_stateid, true);
-	spin_unlock(&ino->i_lock);
-
-	pnfs_layoutcommit_inode(ino, false);
 
-	spin_lock(&ino->i_lock);
 	/*
 	 * Enforce RFC5661 Section 12.5.5.2.1.5 (Bulk Recall and Return)
 	 */
@@ -223,11 +245,13 @@ static u32 initiate_file_draining(struct nfs_client *clp,
 		goto unlock;
 	}
 
+	/* Embrace your forgetfulness! */
+	rv = NFS4ERR_NOMATCHING_LAYOUT;
+
 	if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) {
 		NFS_SERVER(ino)->pnfs_curr_ld->return_range(lo,
 			&args->cbl_range);
 	}
-	pnfs_mark_layout_returned_if_empty(lo);
 unlock:
 	spin_unlock(&ino->i_lock);
 	pnfs_free_lseg_list(&free_me_list);
diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c
index 616dc254b38b..33da841a21bb 100644
--- a/fs/nfs/nfs42proc.c
+++ b/fs/nfs/nfs42proc.c
@@ -351,8 +351,7 @@ nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
 			 * Mark the bad layout state as invalid, then retry
 			 * with the current stateid.
 			 */
-			set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
-			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0);
+			pnfs_mark_layout_stateid_invalid(lo, &head);
 			spin_unlock(&inode->i_lock);
 			pnfs_free_lseg_list(&head);
 		} else
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index a1a3b4c9a563..d30f88c667b7 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -7944,8 +7944,7 @@ nfs4_layoutget_handle_exception(struct rpc_task *task,
 		/*
 		 * Mark the bad layout state as invalid, then retry
 		 */
-		set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
-		pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0);
+		pnfs_mark_layout_stateid_invalid(lo, &head);
 		spin_unlock(&inode->i_lock);
 		pnfs_free_lseg_list(&head);
 		status = -EAGAIN;
@@ -8144,8 +8143,7 @@ static void nfs4_layoutreturn_release(void *calldata)
 	spin_lock(&lo->plh_inode->i_lock);
 	pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range,
 			be32_to_cpu(lrp->args.stateid.seqid));
-	pnfs_mark_layout_returned_if_empty(lo);
-	if (lrp->res.lrs_present)
+	if (lrp->res.lrs_present && pnfs_layout_is_valid(lo))
 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
 	pnfs_clear_layoutreturn_waitbit(lo);
 	spin_unlock(&lo->plh_inode->i_lock);
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 4110c1dc8f68..70806cae0d36 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -259,7 +259,7 @@ pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo)
  * is required.
  * Note that caller must hold inode->i_lock.
  */
-static int
+int
 pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo,
 		struct list_head *lseg_list)
 {
@@ -334,14 +334,17 @@ pnfs_layout_io_test_failed(struct pnfs_layout_hdr *lo, u32 iomode)
 }
 
 static void
-init_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg)
+pnfs_init_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg,
+		const struct pnfs_layout_range *range,
+		const nfs4_stateid *stateid)
 {
 	INIT_LIST_HEAD(&lseg->pls_list);
 	INIT_LIST_HEAD(&lseg->pls_lc_list);
 	atomic_set(&lseg->pls_refcount, 1);
-	smp_mb();
 	set_bit(NFS_LSEG_VALID, &lseg->pls_flags);
 	lseg->pls_layout = lo;
+	lseg->pls_range = *range;
+	lseg->pls_seq = be32_to_cpu(stateid->seqid);
 }
 
 static void pnfs_free_lseg(struct pnfs_layout_segment *lseg)
@@ -486,15 +489,6 @@ pnfs_lseg_range_intersecting(const struct pnfs_layout_range *l1,
 	       (end2 == NFS4_MAX_UINT64 || end2 > start1);
 }
 
-static bool
-should_free_lseg(const struct pnfs_layout_range *lseg_range,
-		 const struct pnfs_layout_range *recall_range)
-{
-	return (recall_range->iomode == IOMODE_ANY ||
-		lseg_range->iomode == recall_range->iomode) &&
-	       pnfs_lseg_range_intersecting(lseg_range, recall_range);
-}
-
 static bool pnfs_lseg_dec_and_remove_zero(struct pnfs_layout_segment *lseg,
 		struct list_head *tmp_list)
 {
@@ -533,6 +527,27 @@ 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)
+{
+	return (recall_range->iomode == IOMODE_ANY ||
+		lseg_range->iomode == recall_range->iomode) &&
+	       pnfs_lseg_range_intersecting(lseg_range, recall_range);
+}
+
+static bool
+pnfs_match_lseg_recall(const struct pnfs_layout_segment *lseg,
+		const struct pnfs_layout_range *recall_range,
+		u32 seq)
+{
+	if (seq != 0 && pnfs_seqid_is_newer(lseg->pls_seq, seq))
+		return false;
+	if (recall_range == NULL)
+		return true;
+	return pnfs_should_free_range(&lseg->pls_range, recall_range);
+}
+
 /**
  * pnfs_mark_matching_lsegs_invalid - tear down lsegs or mark them for later
  * @lo: layout header containing the lsegs
@@ -562,10 +577,7 @@ pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 	if (list_empty(&lo->plh_segs))
 		return 0;
 	list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list)
-		if (!recall_range ||
-		    should_free_lseg(&lseg->pls_range, recall_range)) {
-			if (seq && pnfs_seqid_is_newer(lseg->pls_seq, seq))
-				continue;
+		if (pnfs_match_lseg_recall(lseg, recall_range, seq)) {
 			dprintk("%s: freeing lseg %p iomode %d seq %u"
 				"offset %llu length %llu\n", __func__,
 				lseg, lseg->pls_range.iomode, lseg->pls_seq,
@@ -761,24 +773,25 @@ void
 pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, const nfs4_stateid *new,
 			bool update_barrier)
 {
-	u32 oldseq, newseq, new_barrier;
-	int empty = list_empty(&lo->plh_segs);
+	u32 oldseq, newseq, new_barrier = 0;
+	bool invalid = !pnfs_layout_is_valid(lo);
 
 	oldseq = be32_to_cpu(lo->plh_stateid.seqid);
 	newseq = be32_to_cpu(new->seqid);
-	if (empty || pnfs_seqid_is_newer(newseq, oldseq)) {
+	if (invalid || pnfs_seqid_is_newer(newseq, oldseq)) {
 		nfs4_stateid_copy(&lo->plh_stateid, new);
-		if (update_barrier) {
-			new_barrier = be32_to_cpu(new->seqid);
-		} else {
-			/* Because of wraparound, we want to keep the barrier
-			 * "close" to the current seqids.
-			 */
-			new_barrier = newseq - atomic_read(&lo->plh_outstanding);
-		}
-		if (empty || pnfs_seqid_is_newer(new_barrier, lo->plh_barrier))
-			lo->plh_barrier = new_barrier;
+		/*
+		 * Because of wraparound, we want to keep the barrier
+		 * "close" to the current seqids.
+		 */
+		new_barrier = newseq - atomic_read(&lo->plh_outstanding);
 	}
+	if (update_barrier)
+		new_barrier = be32_to_cpu(new->seqid);
+	else if (new_barrier == 0)
+		return;
+	if (invalid || pnfs_seqid_is_newer(new_barrier, lo->plh_barrier))
+		lo->plh_barrier = new_barrier;
 }
 
 static bool
@@ -873,15 +886,37 @@ void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo)
 	rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
 }
 
+static void
+pnfs_clear_layoutreturn_info(struct pnfs_layout_hdr *lo)
+{
+	lo->plh_return_iomode = 0;
+	lo->plh_return_seq = 0;
+	clear_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags);
+}
+
 static bool
-pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo)
+pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo,
+		nfs4_stateid *stateid,
+		enum pnfs_iomode *iomode)
 {
 	if (test_and_set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags))
 		return false;
-	lo->plh_return_iomode = 0;
-	lo->plh_return_seq = 0;
 	pnfs_get_layout_hdr(lo);
-	clear_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags);
+	if (test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags)) {
+		if (stateid != NULL) {
+			nfs4_stateid_copy(stateid, &lo->plh_stateid);
+			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;
+	}
+	if (stateid != NULL)
+		nfs4_stateid_copy(stateid, &lo->plh_stateid);
+	if (iomode != NULL)
+		*iomode = IOMODE_ANY;
 	return true;
 }
 
@@ -949,10 +984,7 @@ static void pnfs_layoutreturn_before_put_layout_hdr(struct pnfs_layout_hdr *lo)
 		enum pnfs_iomode iomode;
 		bool send;
 
-		nfs4_stateid_copy(&stateid, &lo->plh_stateid);
-		stateid.seqid = cpu_to_be32(lo->plh_return_seq);
-		iomode = lo->plh_return_iomode;
-		send = pnfs_prepare_layoutreturn(lo);
+		send = pnfs_prepare_layoutreturn(lo, &stateid, &iomode);
 		spin_unlock(&inode->i_lock);
 		if (send) {
 			/* Send an async layoutreturn so we dont deadlock */
@@ -989,7 +1021,6 @@ _pnfs_return_layout(struct inode *ino)
 		dprintk("NFS: %s no layout to return\n", __func__);
 		goto out;
 	}
-	nfs4_stateid_copy(&stateid, &nfsi->layout->plh_stateid);
 	/* Reference matched in nfs4_layoutreturn_release */
 	pnfs_get_layout_hdr(lo);
 	empty = list_empty(&lo->plh_segs);
@@ -1012,8 +1043,7 @@ _pnfs_return_layout(struct inode *ino)
 		goto out_put_layout_hdr;
 	}
 
-	set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
-	send = pnfs_prepare_layoutreturn(lo);
+	send = pnfs_prepare_layoutreturn(lo, &stateid, NULL);
 	spin_unlock(&ino->i_lock);
 	pnfs_free_lseg_list(&tmp_list);
 	if (send)
@@ -1080,11 +1110,10 @@ bool pnfs_roc(struct inode *ino)
 			goto out_noroc;
 	}
 
-	nfs4_stateid_copy(&stateid, &lo->plh_stateid);
 	/* always send layoutreturn if being marked so */
-	if (test_and_clear_bit(NFS_LAYOUT_RETURN_REQUESTED,
-				   &lo->plh_flags))
-		layoutreturn = pnfs_prepare_layoutreturn(lo);
+	if (test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags))
+		layoutreturn = pnfs_prepare_layoutreturn(lo,
+				&stateid, NULL);
 
 	list_for_each_entry_safe(lseg, tmp, &lo->plh_segs, pls_list)
 		/* If we are sending layoutreturn, invalidate all valid lsegs */
@@ -1132,7 +1161,6 @@ void pnfs_roc_set_barrier(struct inode *ino, u32 barrier)
 
 	spin_lock(&ino->i_lock);
 	lo = NFS_I(ino)->layout;
-	pnfs_mark_layout_returned_if_empty(lo);
 	if (pnfs_seqid_is_newer(barrier, lo->plh_barrier))
 		lo->plh_barrier = barrier;
 	spin_unlock(&ino->i_lock);
@@ -1746,9 +1774,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 		return lseg;
 	}
 
-	init_lseg(lo, lseg);
-	lseg->pls_range = res->range;
-	lseg->pls_seq = be32_to_cpu(res->stateid.seqid);
+	pnfs_init_lseg(lo, lseg, &res->range, &res->stateid);
 
 	spin_lock(&ino->i_lock);
 	if (pnfs_layoutgets_blocked(lo)) {
@@ -1769,16 +1795,19 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 		 * inode invalid, and don't bother validating the stateid
 		 * sequence number.
 		 */
-		pnfs_mark_matching_lsegs_invalid(lo, &free_me, NULL, 0);
+		pnfs_mark_layout_stateid_invalid(lo, &free_me);
 
 		nfs4_stateid_copy(&lo->plh_stateid, &res->stateid);
 		lo->plh_barrier = be32_to_cpu(res->stateid.seqid);
 	}
 
-	clear_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
-
 	pnfs_get_lseg(lseg);
 	pnfs_layout_insert_lseg(lo, lseg, &free_me);
+	if (!pnfs_layout_is_valid(lo)) {
+		pnfs_clear_layoutreturn_info(lo);
+		clear_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
+	}
+
 
 	if (res->return_on_close)
 		set_bit(NFS_LSEG_ROC, &lseg->pls_flags);
@@ -1798,14 +1827,14 @@ static void
 pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 			 u32 seq)
 {
-	if (lo->plh_return_iomode == iomode)
-		return;
-	if (lo->plh_return_iomode != 0)
+	if (lo->plh_return_iomode != 0 && lo->plh_return_iomode != iomode)
 		iomode = IOMODE_ANY;
 	lo->plh_return_iomode = iomode;
 	set_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags);
-	if (!lo->plh_return_seq || pnfs_seqid_is_newer(seq, lo->plh_return_seq))
+	if (seq != 0) {
+		WARN_ON_ONCE(lo->plh_return_seq != 0 && lo->plh_return_seq != seq);
 		lo->plh_return_seq = seq;
+	}
 }
 
 /**
@@ -1835,7 +1864,7 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 	assert_spin_locked(&lo->plh_inode->i_lock);
 
 	list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list)
-		if (should_free_lseg(&lseg->pls_range, return_range)) {
+		if (pnfs_match_lseg_recall(lseg, return_range, seq)) {
 			dprintk("%s: marking lseg %p iomode %d "
 				"offset %llu length %llu\n", __func__,
 				lseg, lseg->pls_range.iomode,
@@ -1866,19 +1895,17 @@ void pnfs_error_mark_layout_for_return(struct inode *inode,
 	bool return_now = false;
 
 	spin_lock(&inode->i_lock);
-	pnfs_set_plh_return_info(lo, range.iomode, lseg->pls_seq);
+	pnfs_set_plh_return_info(lo, range.iomode, 0);
 	/*
 	 * mark all matching lsegs so that we are sure to have no live
 	 * segments at hand when sending layoutreturn. See pnfs_put_lseg()
 	 * for how it works.
 	 */
-	if (!pnfs_mark_matching_lsegs_return(lo, &free_me,
-						&range, lseg->pls_seq)) {
+	if (!pnfs_mark_matching_lsegs_return(lo, &free_me, &range, 0)) {
 		nfs4_stateid stateid;
-		enum pnfs_iomode iomode = lo->plh_return_iomode;
+		enum pnfs_iomode iomode;
 
-		nfs4_stateid_copy(&stateid, &lo->plh_stateid);
-		return_now = pnfs_prepare_layoutreturn(lo);
+		return_now = pnfs_prepare_layoutreturn(lo, &stateid, &iomode);
 		spin_unlock(&inode->i_lock);
 		if (return_now)
 			pnfs_send_layoutreturn(lo, &stateid, iomode, false);
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 181283c4ebc3..31d99b2927b0 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -268,6 +268,8 @@ int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 				struct list_head *tmp_list,
 				const struct pnfs_layout_range *recall_range,
 				u32 seq);
+int pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo,
+		struct list_head *lseg_list);
 bool pnfs_roc(struct inode *ino);
 void pnfs_roc_release(struct inode *ino);
 void pnfs_roc_set_barrier(struct inode *ino, u32 barrier);
@@ -375,6 +377,11 @@ static inline bool nfs_have_layout(struct inode *inode)
 	return NFS_I(inode)->layout != NULL;
 }
 
+static inline bool pnfs_layout_is_valid(const struct pnfs_layout_hdr *lo)
+{
+	return test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) == 0;
+}
+
 static inline struct nfs4_deviceid_node *
 nfs4_get_deviceid(struct nfs4_deviceid_node *d)
 {
@@ -545,19 +552,6 @@ pnfs_calc_offset_length(u64 offset, u64 end)
 	return 1 + end - offset;
 }
 
-/**
- * pnfs_mark_layout_returned_if_empty - marks the layout as returned
- * @lo: layout header
- *
- * Note: Caller must hold inode->i_lock
- */
-static inline void
-pnfs_mark_layout_returned_if_empty(struct pnfs_layout_hdr *lo)
-{
-	if (list_empty(&lo->plh_segs))
-		set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
-}
-
 static inline void
 pnfs_copy_range(struct pnfs_layout_range *dst,
 		const struct pnfs_layout_range *src)