[PATCH 6.12 008/139] pNFS: Fix a deadlock when returning a delegation during open()

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

6.12-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Trond Myklebust <trond.myklebust@hammerspace.com>

[ Upstream commit 857bf9056291a16785ae3be1d291026b2437fc48 ]

Ben Coddington reports seeing a hang in the following stack trace:
  0 [ffffd0b50e1774e0] __schedule at ffffffff9ca05415
  1 [ffffd0b50e177548] schedule at ffffffff9ca05717
  2 [ffffd0b50e177558] bit_wait at ffffffff9ca061e1
  3 [ffffd0b50e177568] __wait_on_bit at ffffffff9ca05cfb
  4 [ffffd0b50e1775c8] out_of_line_wait_on_bit at ffffffff9ca05ea5
  5 [ffffd0b50e177618] pnfs_roc at ffffffffc154207b [nfsv4]
  6 [ffffd0b50e1776b8] _nfs4_proc_delegreturn at ffffffffc1506586 [nfsv4]
  7 [ffffd0b50e177788] nfs4_proc_delegreturn at ffffffffc1507480 [nfsv4]
  8 [ffffd0b50e1777f8] nfs_do_return_delegation at ffffffffc1523e41 [nfsv4]
  9 [ffffd0b50e177838] nfs_inode_set_delegation at ffffffffc1524a75 [nfsv4]
 10 [ffffd0b50e177888] nfs4_process_delegation at ffffffffc14f41dd [nfsv4]
 11 [ffffd0b50e1778a0] _nfs4_opendata_to_nfs4_state at ffffffffc1503edf [nfsv4]
 12 [ffffd0b50e1778c0] _nfs4_open_and_get_state at ffffffffc1504e56 [nfsv4]
 13 [ffffd0b50e177978] _nfs4_do_open at ffffffffc15051b8 [nfsv4]
 14 [ffffd0b50e1779f8] nfs4_do_open at ffffffffc150559c [nfsv4]
 15 [ffffd0b50e177a80] nfs4_atomic_open at ffffffffc15057fb [nfsv4]
 16 [ffffd0b50e177ad0] nfs4_file_open at ffffffffc15219be [nfsv4]
 17 [ffffd0b50e177b78] do_dentry_open at ffffffff9c09e6ea
 18 [ffffd0b50e177ba8] vfs_open at ffffffff9c0a082e
 19 [ffffd0b50e177bd0] dentry_open at ffffffff9c0a0935

The issue is that the delegreturn is being asked to wait for a layout
return that cannot complete because a state recovery was initiated. The
state recovery cannot complete until the open() finishes processing the
delegations it was given.

The solution is to propagate the existing flags that indicate a
non-blocking call to the function pnfs_roc(), so that it knows not to
wait in this situation.

Reported-by: Benjamin Coddington <bcodding@hammerspace.com>
Fixes: 29ade5db1293 ("pNFS: Wait on outstanding layoutreturns to complete in pnfs_roc()")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 fs/nfs/nfs4proc.c |  6 ++---
 fs/nfs/pnfs.c     | 58 +++++++++++++++++++++++++++++++++--------------
 fs/nfs/pnfs.h     | 17 ++++++--------
 3 files changed, 51 insertions(+), 30 deletions(-)

diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 89f779f16f0dc..c76acd537be05 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -3868,8 +3868,8 @@ int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
 	calldata->res.seqid = calldata->arg.seqid;
 	calldata->res.server = server;
 	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
-	calldata->lr.roc = pnfs_roc(state->inode,
-			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
+	calldata->lr.roc = pnfs_roc(state->inode, &calldata->lr.arg,
+				    &calldata->lr.res, msg.rpc_cred, wait);
 	if (calldata->lr.roc) {
 		calldata->arg.lr_args = &calldata->lr.arg;
 		calldata->res.lr_res = &calldata->lr.res;
@@ -6895,7 +6895,7 @@ static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
 	data->inode = nfs_igrab_and_active(inode);
 	if (data->inode || issync) {
 		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
-					cred);
+					cred, issync);
 		if (data->lr.roc) {
 			data->args.lr_args = &data->lr.arg;
 			data->res.lr_res = &data->lr.res;
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 7a742bcff687b..16981d0389c4c 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -1546,10 +1546,9 @@ static int pnfs_layout_return_on_reboot(struct pnfs_layout_hdr *lo)
 				      PNFS_FL_LAYOUTRETURN_PRIVILEGED);
 }
 
-bool pnfs_roc(struct inode *ino,
-		struct nfs4_layoutreturn_args *args,
-		struct nfs4_layoutreturn_res *res,
-		const struct cred *cred)
+bool pnfs_roc(struct inode *ino, struct nfs4_layoutreturn_args *args,
+	      struct nfs4_layoutreturn_res *res, const struct cred *cred,
+	      bool sync)
 {
 	struct nfs_inode *nfsi = NFS_I(ino);
 	struct nfs_open_context *ctx;
@@ -1560,7 +1559,7 @@ bool pnfs_roc(struct inode *ino,
 	nfs4_stateid stateid;
 	enum pnfs_iomode iomode = 0;
 	bool layoutreturn = false, roc = false;
-	bool skip_read = false;
+	bool skip_read;
 
 	if (!nfs_have_layout(ino))
 		return false;
@@ -1573,20 +1572,14 @@ bool pnfs_roc(struct inode *ino,
 		lo = NULL;
 		goto out_noroc;
 	}
-	pnfs_get_layout_hdr(lo);
-	if (test_bit(NFS_LAYOUT_RETURN_LOCK, &lo->plh_flags)) {
-		spin_unlock(&ino->i_lock);
-		rcu_read_unlock();
-		wait_on_bit(&lo->plh_flags, NFS_LAYOUT_RETURN,
-				TASK_UNINTERRUPTIBLE);
-		pnfs_put_layout_hdr(lo);
-		goto retry;
-	}
 
 	/* no roc if we hold a delegation */
+	skip_read = false;
 	if (nfs4_check_delegation(ino, FMODE_READ)) {
-		if (nfs4_check_delegation(ino, FMODE_WRITE))
+		if (nfs4_check_delegation(ino, FMODE_WRITE)) {
+			lo = NULL;
 			goto out_noroc;
+		}
 		skip_read = true;
 	}
 
@@ -1595,12 +1588,43 @@ bool pnfs_roc(struct inode *ino,
 		if (state == NULL)
 			continue;
 		/* Don't return layout if there is open file state */
-		if (state->state & FMODE_WRITE)
+		if (state->state & FMODE_WRITE) {
+			lo = NULL;
 			goto out_noroc;
+		}
 		if (state->state & FMODE_READ)
 			skip_read = true;
 	}
 
+	if (skip_read) {
+		bool writes = false;
+
+		list_for_each_entry(lseg, &lo->plh_segs, pls_list) {
+			if (lseg->pls_range.iomode != IOMODE_READ) {
+				writes = true;
+				break;
+			}
+		}
+		if (!writes) {
+			lo = NULL;
+			goto out_noroc;
+		}
+	}
+
+	pnfs_get_layout_hdr(lo);
+	if (test_bit(NFS_LAYOUT_RETURN_LOCK, &lo->plh_flags)) {
+		if (!sync) {
+			pnfs_set_plh_return_info(
+				lo, skip_read ? IOMODE_RW : IOMODE_ANY, 0);
+			goto out_noroc;
+		}
+		spin_unlock(&ino->i_lock);
+		rcu_read_unlock();
+		wait_on_bit(&lo->plh_flags, NFS_LAYOUT_RETURN,
+			    TASK_UNINTERRUPTIBLE);
+		pnfs_put_layout_hdr(lo);
+		goto retry;
+	}
 
 	list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list) {
 		if (skip_read && lseg->pls_range.iomode == IOMODE_READ)
@@ -1640,7 +1664,7 @@ bool pnfs_roc(struct inode *ino,
 out_noroc:
 	spin_unlock(&ino->i_lock);
 	rcu_read_unlock();
-	pnfs_layoutcommit_inode(ino, true);
+	pnfs_layoutcommit_inode(ino, sync);
 	if (roc) {
 		struct pnfs_layoutdriver_type *ld = NFS_SERVER(ino)->pnfs_curr_ld;
 		if (ld->prepare_layoutreturn)
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 91ff877185c8a..3db8f13d8fe4e 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -303,10 +303,9 @@ int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 				u32 seq);
 int pnfs_mark_layout_stateid_invalid(struct pnfs_layout_hdr *lo,
 		struct list_head *lseg_list);
-bool pnfs_roc(struct inode *ino,
-		struct nfs4_layoutreturn_args *args,
-		struct nfs4_layoutreturn_res *res,
-		const struct cred *cred);
+bool pnfs_roc(struct inode *ino, struct nfs4_layoutreturn_args *args,
+	      struct nfs4_layoutreturn_res *res, const struct cred *cred,
+	      bool sync);
 int pnfs_roc_done(struct rpc_task *task, struct nfs4_layoutreturn_args **argpp,
 		  struct nfs4_layoutreturn_res **respp, int *ret);
 void pnfs_roc_release(struct nfs4_layoutreturn_args *args,
@@ -773,12 +772,10 @@ pnfs_layoutcommit_outstanding(struct inode *inode)
 	return false;
 }
 
-
-static inline bool
-pnfs_roc(struct inode *ino,
-		struct nfs4_layoutreturn_args *args,
-		struct nfs4_layoutreturn_res *res,
-		const struct cred *cred)
+static inline bool pnfs_roc(struct inode *ino,
+			    struct nfs4_layoutreturn_args *args,
+			    struct nfs4_layoutreturn_res *res,
+			    const struct cred *cred, bool sync)
 {
 	return false;
 }
-- 
2.51.0