pNFS: Do not free layout segments that are marked for return

We may want to process and transmit layout stat information for the
layout segments that are being returned, so we should defer freeing
them until after the layoutreturn has completed.

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

3 files changed, 73 insertions, 22 deletions

diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 917a6db5c84f..561b21e4a930 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -8572,21 +8572,12 @@ static void nfs4_layoutreturn_release(void *calldata)
 {
 	struct nfs4_layoutreturn *lrp = calldata;
 	struct pnfs_layout_hdr *lo = lrp->args.layout;
-	LIST_HEAD(freeme);
 
 	dprintk("--> %s\n", __func__);
-	spin_lock(&lo->plh_inode->i_lock);
-	if (lrp->res.lrs_present) {
-		pnfs_mark_matching_lsegs_invalid(lo, &freeme,
-				&lrp->args.range,
-				be32_to_cpu(lrp->args.stateid.seqid));
-		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
-	} else
-		pnfs_mark_layout_stateid_invalid(lo, &freeme);
-	pnfs_clear_layoutreturn_waitbit(lo);
-	spin_unlock(&lo->plh_inode->i_lock);
+	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.range,
+			be32_to_cpu(lrp->args.stateid.seqid),
+			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
 	nfs4_sequence_free_slot(&lrp->res.seq_res);
-	pnfs_free_lseg_list(&freeme);
 	pnfs_put_layout_hdr(lrp->args.layout);
 	nfs_iput_and_deactive(lrp->inode);
 	kfree(calldata);
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index d70cc467a87b..555151b6d31b 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -54,6 +54,10 @@ static DEFINE_SPINLOCK(pnfs_spinlock);
 static LIST_HEAD(pnfs_modules_tbl);
 
 static void pnfs_layoutreturn_before_put_layout_hdr(struct pnfs_layout_hdr *lo);
+static void pnfs_free_returned_lsegs(struct pnfs_layout_hdr *lo,
+		struct list_head *free_me,
+		const struct pnfs_layout_range *range,
+		u32 seq);
 
 /* Return the registered pnfs layout driver module matching given id */
 static struct pnfs_layoutdriver_type *
@@ -326,6 +330,7 @@ pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo,
 
 	set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
 	pnfs_clear_layoutreturn_info(lo);
+	pnfs_free_returned_lsegs(lo, lseg_list, &range, 0);
 	return pnfs_mark_matching_lsegs_invalid(lo, lseg_list, &range, 0);
 }
 
@@ -405,9 +410,10 @@ pnfs_init_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg,
 
 static void pnfs_free_lseg(struct pnfs_layout_segment *lseg)
 {
-	struct inode *ino = lseg->pls_layout->plh_inode;
-
-	NFS_SERVER(ino)->pnfs_curr_ld->free_lseg(lseg);
+	if (lseg != NULL) {
+		struct inode *inode = lseg->pls_layout->plh_inode;
+		NFS_SERVER(inode)->pnfs_curr_ld->free_lseg(lseg);
+	}
 }
 
 static void
@@ -430,6 +436,18 @@ pnfs_layout_remove_lseg(struct pnfs_layout_hdr *lo,
 	rpc_wake_up(&NFS_SERVER(inode)->roc_rpcwaitq);
 }
 
+static bool
+pnfs_cache_lseg_for_layoutreturn(struct pnfs_layout_hdr *lo,
+		struct pnfs_layout_segment *lseg)
+{
+	if (test_and_clear_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags) &&
+	    pnfs_layout_is_valid(lo)) {
+		list_move_tail(&lseg->pls_list, &lo->plh_return_segs);
+		return true;
+	}
+	return false;
+}
+
 void
 pnfs_put_lseg(struct pnfs_layout_segment *lseg)
 {
@@ -453,6 +471,8 @@ pnfs_put_lseg(struct pnfs_layout_segment *lseg)
 		}
 		pnfs_get_layout_hdr(lo);
 		pnfs_layout_remove_lseg(lo, lseg);
+		if (pnfs_cache_lseg_for_layoutreturn(lo, lseg))
+			lseg = NULL;
 		spin_unlock(&inode->i_lock);
 		pnfs_free_lseg(lseg);
 		pnfs_put_layout_hdr(lo);
@@ -493,9 +513,11 @@ pnfs_put_lseg_locked(struct pnfs_layout_segment *lseg)
 		struct pnfs_layout_hdr *lo = lseg->pls_layout;
 		if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags))
 			return;
-		pnfs_get_layout_hdr(lo);
 		pnfs_layout_remove_lseg(lo, lseg);
-		pnfs_free_lseg_async(lseg);
+		if (!pnfs_cache_lseg_for_layoutreturn(lo, lseg)) {
+			pnfs_get_layout_hdr(lo);
+			pnfs_free_lseg_async(lseg);
+		}
 	}
 }
 EXPORT_SYMBOL_GPL(pnfs_put_lseg_locked);
@@ -619,6 +641,20 @@ pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 	return remaining;
 }
 
+static void
+pnfs_free_returned_lsegs(struct pnfs_layout_hdr *lo,
+		struct list_head *free_me,
+		const struct pnfs_layout_range *range,
+		u32 seq)
+{
+	struct pnfs_layout_segment *lseg, *next;
+
+	list_for_each_entry_safe(lseg, next, &lo->plh_return_segs, pls_list) {
+		if (pnfs_match_lseg_recall(lseg, range, seq))
+			list_move_tail(&lseg->pls_list, free_me);
+	}
+}
+
 /* note free_me must contain lsegs from a single layout_hdr */
 void
 pnfs_free_lseg_list(struct list_head *free_me)
@@ -915,7 +951,7 @@ static void pnfs_clear_layoutcommit(struct inode *inode,
 	}
 }
 
-void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo)
+static void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo)
 {
 	clear_bit_unlock(NFS_LAYOUT_RETURN, &lo->plh_flags);
 	clear_bit(NFS_LAYOUT_RETURN_LOCK, &lo->plh_flags);
@@ -924,6 +960,27 @@ void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo)
 	rpc_wake_up(&NFS_SERVER(lo->plh_inode)->roc_rpcwaitq);
 }
 
+void pnfs_layoutreturn_free_lsegs(struct pnfs_layout_hdr *lo,
+		const struct pnfs_layout_range *range,
+		u32 seq,
+		const nfs4_stateid *stateid)
+{
+	struct inode *inode = lo->plh_inode;
+	LIST_HEAD(freeme);
+
+	spin_lock(&inode->i_lock);
+	if (stateid) {
+		pnfs_mark_matching_lsegs_invalid(lo, &freeme, range, seq);
+		pnfs_free_returned_lsegs(lo, &freeme, range, seq);
+		pnfs_set_layout_stateid(lo, stateid, true);
+	} else
+		pnfs_mark_layout_stateid_invalid(lo, &freeme);
+	pnfs_clear_layoutreturn_waitbit(lo);
+	spin_unlock(&inode->i_lock);
+	pnfs_free_lseg_list(&freeme);
+
+}
+
 static bool
 pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo,
 		nfs4_stateid *stateid,
@@ -1349,6 +1406,7 @@ alloc_init_layout_hdr(struct inode *ino,
 	atomic_set(&lo->plh_refcount, 1);
 	INIT_LIST_HEAD(&lo->plh_layouts);
 	INIT_LIST_HEAD(&lo->plh_segs);
+	INIT_LIST_HEAD(&lo->plh_return_segs);
 	INIT_LIST_HEAD(&lo->plh_bulk_destroy);
 	lo->plh_inode = ino;
 	lo->plh_lc_cred = get_rpccred(ctx->cred);
@@ -1920,7 +1978,6 @@ void pnfs_error_mark_layout_for_return(struct inode *inode,
 		.offset = 0,
 		.length = NFS4_MAX_UINT64,
 	};
-	LIST_HEAD(free_me);
 	bool return_now = false;
 
 	spin_lock(&inode->i_lock);
@@ -1932,7 +1989,7 @@ void pnfs_error_mark_layout_for_return(struct inode *inode,
 	 * segments at hand when sending layoutreturn. See pnfs_put_lseg()
 	 * for how it works.
 	 */
-	if (!pnfs_mark_matching_lsegs_return(lo, &free_me, &range, 0)) {
+	if (!pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs, &range, 0)) {
 		nfs4_stateid stateid;
 		enum pnfs_iomode iomode;
 
@@ -1944,7 +2001,6 @@ void pnfs_error_mark_layout_for_return(struct inode *inode,
 		spin_unlock(&inode->i_lock);
 		nfs_commit_inode(inode, 0);
 	}
-	pnfs_free_lseg_list(&free_me);
 }
 EXPORT_SYMBOL_GPL(pnfs_error_mark_layout_for_return);
 
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 337dad382b6a..a382710edf40 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -191,6 +191,7 @@ struct pnfs_layout_hdr {
 	struct list_head	plh_layouts;   /* other client layouts */
 	struct list_head	plh_bulk_destroy;
 	struct list_head	plh_segs;      /* layout segments list */
+	struct list_head	plh_return_segs; /* invalid layout segments */
 	unsigned long		plh_block_lgets; /* block LAYOUTGET if >0 */
 	unsigned long		plh_retry_timestamp;
 	unsigned long		plh_flags;
@@ -293,7 +294,10 @@ struct pnfs_layout_segment *pnfs_update_layout(struct inode *ino,
 					       enum pnfs_iomode iomode,
 					       bool strict_iomode,
 					       gfp_t gfp_flags);
-void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo);
+void pnfs_layoutreturn_free_lsegs(struct pnfs_layout_hdr *lo,
+		const struct pnfs_layout_range *range,
+		u32 seq,
+		const nfs4_stateid *stateid);
 
 void pnfs_generic_layout_insert_lseg(struct pnfs_layout_hdr *lo,
 		   struct pnfs_layout_segment *lseg,