Merge tag 'nfs-for-4.7-1' of git://git.linux-nfs.org/projects/anna/linux-nfs

Pull NFS client updates from Anna Schumaker:
 "Highlights include:

  Features:
   - Add support for the NFS v4.2 COPY operation
   - Add support for NFS/RDMA over IPv6

  Bugfixes and cleanups:
   - Avoid race that crashes nfs_init_commit()
   - Fix oops in callback path
   - Fix LOCK/OPEN race when unlinking an open file
   - Choose correct stateids when using delegations in setattr, read and
     write
   - Don't send empty SETATTR after OPEN_CREATE
   - xprtrdma: Prevent server from writing a reply into memory client
     has released
   - xprtrdma: Support using Read list and Reply chunk in one RPC call"

* tag 'nfs-for-4.7-1' of git://git.linux-nfs.org/projects/anna/linux-nfs: (61 commits)
  pnfs: pnfs_update_layout needs to consider if strict iomode checking is on
  nfs/flexfiles: Use the layout segment for reading unless it a IOMODE_RW and reading is disabled
  nfs/flexfiles: Helper function to detect FF_FLAGS_NO_READ_IO
  nfs: avoid race that crashes nfs_init_commit
  NFS: checking for NULL instead of IS_ERR() in nfs_commit_file()
  pnfs: make pnfs_layout_process more robust
  pnfs: rework LAYOUTGET retry handling
  pnfs: lift retry logic from send_layoutget to pnfs_update_layout
  pnfs: fix bad error handling in send_layoutget
  flexfiles: add kerneldoc header to nfs4_ff_layout_prepare_ds
  flexfiles: remove pointless setting of NFS_LAYOUT_RETURN_REQUESTED
  pnfs: only tear down lsegs that precede seqid in LAYOUTRETURN args
  pnfs: keep track of the return sequence number in pnfs_layout_hdr
  pnfs: record sequence in pnfs_layout_segment when it's created
  pnfs: don't merge new ff lsegs with ones that have LAYOUTRETURN bit set
  pNFS/flexfiles: When initing reads or writes, we might have to retry connecting to DSes
  pNFS/flexfiles: When checking for available DSes, conditionally check for MDS io
  pNFS/flexfile: Fix erroneous fall back to read/write through the MDS
  NFS: Reclaim writes via writepage are opportunistic
  NFSv4: Use the right stateid for delegations in setattr, read and write
  ...

fs/nfs/callback_proc.c

@@ -217,7 +217,8 @@ static u32 initiate_file_draining(struct nfs_client *clp,
 	}
 
 	if (pnfs_mark_matching_lsegs_return(lo, &free_me_list,
-					&args->cbl_range)) {
+				&args->cbl_range,
+				be32_to_cpu(args->cbl_stateid.seqid))) {
 		rv = NFS4_OK;
 		goto unlock;
 	}
@@ -500,8 +501,10 @@ __be32 nfs4_callback_sequence(struct cb_sequenceargs *args,
 	cps->slot = slot;
 
 	/* The ca_maxresponsesize_cached is 0 with no DRC */
-	if (args->csa_cachethis != 0)
-		return htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE);
+	if (args->csa_cachethis != 0) {
+		status = htonl(NFS4ERR_REP_TOO_BIG_TO_CACHE);
+		goto out_unlock;
+	}
 
 	/*
 	 * Check for pending referring calls.  If a match is found, a

fs/nfs/callback_xdr.c

@@ -146,10 +146,16 @@ static __be32 decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
 	p = read_buf(xdr, NFS4_STATEID_SIZE);
 	if (unlikely(p == NULL))
 		return htonl(NFS4ERR_RESOURCE);
-	memcpy(stateid, p, NFS4_STATEID_SIZE);
+	memcpy(stateid->data, p, NFS4_STATEID_SIZE);
 	return 0;
 }
 
+static __be32 decode_delegation_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	stateid->type = NFS4_DELEGATION_STATEID_TYPE;
+	return decode_stateid(xdr, stateid);
+}
+
 static __be32 decode_compound_hdr_arg(struct xdr_stream *xdr, struct cb_compound_hdr_arg *hdr)
 {
 	__be32 *p;
@@ -211,7 +217,7 @@ static __be32 decode_recall_args(struct svc_rqst *rqstp, struct xdr_stream *xdr,
 	__be32 *p;
 	__be32 status;
 
-	status = decode_stateid(xdr, &args->stateid);
+	status = decode_delegation_stateid(xdr, &args->stateid);
 	if (unlikely(status != 0))
 		goto out;
 	p = read_buf(xdr, 4);
@@ -227,6 +233,11 @@ static __be32 decode_recall_args(struct svc_rqst *rqstp, struct xdr_stream *xdr,
 }
 
 #if defined(CONFIG_NFS_V4_1)
+static __be32 decode_layout_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	stateid->type = NFS4_LAYOUT_STATEID_TYPE;
+	return decode_stateid(xdr, stateid);
+}
 
 static __be32 decode_layoutrecall_args(struct svc_rqst *rqstp,
 				       struct xdr_stream *xdr,
@@ -263,7 +274,7 @@ static __be32 decode_layoutrecall_args(struct svc_rqst *rqstp,
 		}
 		p = xdr_decode_hyper(p, &args->cbl_range.offset);
 		p = xdr_decode_hyper(p, &args->cbl_range.length);
-		status = decode_stateid(xdr, &args->cbl_stateid);
+		status = decode_layout_stateid(xdr, &args->cbl_stateid);
 		if (unlikely(status != 0))
 			goto out;
 	} else if (args->cbl_recall_type == RETURN_FSID) {

fs/nfs/delegation.c

@@ -875,15 +875,16 @@ int nfs_delegations_present(struct nfs_client *clp)
 
 /**
  * nfs4_copy_delegation_stateid - Copy inode's state ID information
- * @dst: stateid data structure to fill in
  * @inode: inode to check
  * @flags: delegation type requirement
+ * @dst: stateid data structure to fill in
+ * @cred: optional argument to retrieve credential
  *
  * Returns "true" and fills in "dst->data" * if inode had a delegation,
  * otherwise "false" is returned.
  */
-bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode,
-		fmode_t flags)
+bool nfs4_copy_delegation_stateid(struct inode *inode, fmode_t flags,
+		nfs4_stateid *dst, struct rpc_cred **cred)
 {
 	struct nfs_inode *nfsi = NFS_I(inode);
 	struct nfs_delegation *delegation;
@@ -896,6 +897,8 @@ bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode,
 	if (ret) {
 		nfs4_stateid_copy(dst, &delegation->stateid);
 		nfs_mark_delegation_referenced(delegation);
+		if (cred)
+			*cred = get_rpccred(delegation->cred);
 	}
 	rcu_read_unlock();
 	return ret;

fs/nfs/delegation.h

@@ -56,7 +56,7 @@ void nfs_delegation_reap_unclaimed(struct nfs_client *clp);
 int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4_stateid *stateid, int issync);
 int nfs4_open_delegation_recall(struct nfs_open_context *ctx, struct nfs4_state *state, const nfs4_stateid *stateid, fmode_t type);
 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid);
-bool nfs4_copy_delegation_stateid(nfs4_stateid *dst, struct inode *inode, fmode_t flags);
+bool nfs4_copy_delegation_stateid(struct inode *inode, fmode_t flags, nfs4_stateid *dst, struct rpc_cred **cred);
 
 void nfs_mark_delegation_referenced(struct nfs_delegation *delegation);
 int nfs4_have_delegation(struct inode *inode, fmode_t flags);

fs/nfs/direct.c

@@ -87,6 +87,7 @@ struct nfs_direct_req {
 	int			mirror_count;
 
 	ssize_t			count,		/* bytes actually processed */
+				max_count,	/* max expected count */
 				bytes_left,	/* bytes left to be sent */
 				io_start,	/* start of IO */
 				error;		/* any reported error */
@@ -123,6 +124,8 @@ nfs_direct_good_bytes(struct nfs_direct_req *dreq, struct nfs_pgio_header *hdr)
 	int i;
 	ssize_t count;
 
+	WARN_ON_ONCE(dreq->count >= dreq->max_count);
+
 	if (dreq->mirror_count == 1) {
 		dreq->mirrors[hdr->pgio_mirror_idx].count += hdr->good_bytes;
 		dreq->count += hdr->good_bytes;
@@ -275,7 +278,7 @@ static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
 void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
 			      struct nfs_direct_req *dreq)
 {
-	cinfo->lock = &dreq->inode->i_lock;
+	cinfo->inode = dreq->inode;
 	cinfo->mds = &dreq->mds_cinfo;
 	cinfo->ds = &dreq->ds_cinfo;
 	cinfo->dreq = dreq;
@@ -591,7 +594,7 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter)
 		goto out_unlock;
 
 	dreq->inode = inode;
-	dreq->bytes_left = count;
+	dreq->bytes_left = dreq->max_count = count;
 	dreq->io_start = iocb->ki_pos;
 	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 	l_ctx = nfs_get_lock_context(dreq->ctx);
@@ -630,13 +633,13 @@ nfs_direct_write_scan_commit_list(struct inode *inode,
 				  struct list_head *list,
 				  struct nfs_commit_info *cinfo)
 {
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 #ifdef CONFIG_NFS_V4_1
 	if (cinfo->ds != NULL && cinfo->ds->nwritten != 0)
 		NFS_SERVER(inode)->pnfs_curr_ld->recover_commit_reqs(list, cinfo);
 #endif
 	nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0);
-	spin_unlock(cinfo->lock);
+	spin_unlock(&cinfo->inode->i_lock);
 }
 
 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
@@ -671,13 +674,13 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 		if (!nfs_pageio_add_request(&desc, req)) {
 			nfs_list_remove_request(req);
 			nfs_list_add_request(req, &failed);
-			spin_lock(cinfo.lock);
+			spin_lock(&cinfo.inode->i_lock);
 			dreq->flags = 0;
 			if (desc.pg_error < 0)
 				dreq->error = desc.pg_error;
 			else
 				dreq->error = -EIO;
-			spin_unlock(cinfo.lock);
+			spin_unlock(&cinfo.inode->i_lock);
 		}
 		nfs_release_request(req);
 	}
@@ -1023,7 +1026,7 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter)
 		goto out_unlock;
 
 	dreq->inode = inode;
-	dreq->bytes_left = iov_iter_count(iter);
+	dreq->bytes_left = dreq->max_count = iov_iter_count(iter);
 	dreq->io_start = pos;
 	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 	l_ctx = nfs_get_lock_context(dreq->ctx);

fs/nfs/filelayout/filelayout.c

@@ -795,7 +795,7 @@ filelayout_alloc_commit_info(struct pnfs_layout_segment *lseg,
 		buckets[i].direct_verf.committed = NFS_INVALID_STABLE_HOW;
 	}
 
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 	if (cinfo->ds->nbuckets >= size)
 		goto out;
 	for (i = 0; i < cinfo->ds->nbuckets; i++) {
@@ -811,7 +811,7 @@ filelayout_alloc_commit_info(struct pnfs_layout_segment *lseg,
 	swap(cinfo->ds->buckets, buckets);
 	cinfo->ds->nbuckets = size;
 out:
-	spin_unlock(cinfo->lock);
+	spin_unlock(&cinfo->inode->i_lock);
 	kfree(buckets);
 	return 0;
 }
@@ -890,6 +890,7 @@ filelayout_pg_init_read(struct nfs_pageio_descriptor *pgio,
 					   0,
 					   NFS4_MAX_UINT64,
 					   IOMODE_READ,
+					   false,
 					   GFP_KERNEL);
 		if (IS_ERR(pgio->pg_lseg)) {
 			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
@@ -915,6 +916,7 @@ filelayout_pg_init_write(struct nfs_pageio_descriptor *pgio,
 					   0,
 					   NFS4_MAX_UINT64,
 					   IOMODE_RW,
+					   false,
 					   GFP_NOFS);
 		if (IS_ERR(pgio->pg_lseg)) {
 			pgio->pg_error = PTR_ERR(pgio->pg_lseg);

fs/nfs/flexfilelayout/flexfilelayout.h

@@ -10,7 +10,8 @@
 #define FS_NFS_NFS4FLEXFILELAYOUT_H
 
 #define FF_FLAGS_NO_LAYOUTCOMMIT 1
-#define FF_FLAGS_NO_IO_THRU_MDS 2
+#define FF_FLAGS_NO_IO_THRU_MDS  2
+#define FF_FLAGS_NO_READ_IO      4
 
 #include "../pnfs.h"
 
@@ -76,9 +77,8 @@ struct nfs4_ff_layout_mirror {
 	u32				fh_versions_cnt;
 	struct nfs_fh			*fh_versions;
 	nfs4_stateid			stateid;
-	u32				uid;
-	u32				gid;
-	struct rpc_cred			*cred;
+	struct rpc_cred	__rcu		*ro_cred;
+	struct rpc_cred	__rcu		*rw_cred;
 	atomic_t			ref;
 	spinlock_t			lock;
 	struct nfs4_ff_layoutstat	read_stat;
@@ -153,6 +153,12 @@ ff_layout_no_fallback_to_mds(struct pnfs_layout_segment *lseg)
 	return FF_LAYOUT_LSEG(lseg)->flags & FF_FLAGS_NO_IO_THRU_MDS;
 }
 
+static inline bool
+ff_layout_no_read_on_rw(struct pnfs_layout_segment *lseg)
+{
+	return FF_LAYOUT_LSEG(lseg)->flags & FF_FLAGS_NO_READ_IO;
+}
+
 static inline bool
 ff_layout_test_devid_unavailable(struct nfs4_deviceid_node *node)
 {
@@ -192,4 +198,7 @@ nfs4_ff_find_or_create_ds_client(struct pnfs_layout_segment *lseg,
 struct rpc_cred *ff_layout_get_ds_cred(struct pnfs_layout_segment *lseg,
 				       u32 ds_idx, struct rpc_cred *mdscred);
 bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg);
+bool ff_layout_avoid_mds_available_ds(struct pnfs_layout_segment *lseg);
+bool ff_layout_avoid_read_on_rw(struct pnfs_layout_segment *lseg);
+
 #endif /* FS_NFS_NFS4FLEXFILELAYOUT_H */

fs/nfs/flexfilelayout/flexfilelayoutdev.c

@@ -228,7 +228,8 @@ ff_ds_error_match(const struct nfs4_ff_layout_ds_err *e1,
 		return e1->opnum < e2->opnum ? -1 : 1;
 	if (e1->status != e2->status)
 		return e1->status < e2->status ? -1 : 1;
-	ret = memcmp(&e1->stateid, &e2->stateid, sizeof(e1->stateid));
+	ret = memcmp(e1->stateid.data, e2->stateid.data,
+			sizeof(e1->stateid.data));
 	if (ret != 0)
 		return ret;
 	ret = memcmp(&e1->deviceid, &e2->deviceid, sizeof(e1->deviceid));
@@ -302,40 +303,26 @@ int ff_layout_track_ds_error(struct nfs4_flexfile_layout *flo,
 	return 0;
 }
 
-/* currently we only support AUTH_NONE and AUTH_SYS */
-static rpc_authflavor_t
-nfs4_ff_layout_choose_authflavor(struct nfs4_ff_layout_mirror *mirror)
+static struct rpc_cred *
+ff_layout_get_mirror_cred(struct nfs4_ff_layout_mirror *mirror, u32 iomode)
 {
-	if (mirror->uid == (u32)-1)
-		return RPC_AUTH_NULL;
-	return RPC_AUTH_UNIX;
-}
+	struct rpc_cred *cred, __rcu **pcred;
 
-/* fetch cred for NFSv3 DS */
-static int ff_layout_update_mirror_cred(struct nfs4_ff_layout_mirror *mirror,
-				      struct nfs4_pnfs_ds *ds)
-{
-	if (ds->ds_clp && !mirror->cred &&
-	    mirror->mirror_ds->ds_versions[0].version == 3) {
-		struct rpc_auth *auth = ds->ds_clp->cl_rpcclient->cl_auth;
-		struct rpc_cred *cred;
-		struct auth_cred acred = {
-			.uid = make_kuid(&init_user_ns, mirror->uid),
-			.gid = make_kgid(&init_user_ns, mirror->gid),
-		};
-
-		/* AUTH_NULL ignores acred */
-		cred = auth->au_ops->lookup_cred(auth, &acred, 0);
-		if (IS_ERR(cred)) {
-			dprintk("%s: lookup_cred failed with %ld\n",
-				__func__, PTR_ERR(cred));
-			return PTR_ERR(cred);
-		} else {
-			if (cmpxchg(&mirror->cred, NULL, cred))
-				put_rpccred(cred);
-		}
-	}
-	return 0;
+	if (iomode == IOMODE_READ)
+		pcred = &mirror->ro_cred;
+	else
+		pcred = &mirror->rw_cred;
+
+	rcu_read_lock();
+	do {
+		cred = rcu_dereference(*pcred);
+		if (!cred)
+			break;
+
+		cred = get_rpccred_rcu(cred);
+	} while(!cred);
+	rcu_read_unlock();
+	return cred;
 }
 
 struct nfs_fh *
@@ -356,7 +343,23 @@ nfs4_ff_layout_select_ds_fh(struct pnfs_layout_segment *lseg, u32 mirror_idx)
 	return fh;
 }
 
-/* Upon return, either ds is connected, or ds is NULL */
+/**
+ * nfs4_ff_layout_prepare_ds - prepare a DS connection for an RPC call
+ * @lseg: the layout segment we're operating on
+ * @ds_idx: index of the DS to use
+ * @fail_return: return layout on connect failure?
+ *
+ * Try to prepare a DS connection to accept an RPC call. This involves
+ * selecting a mirror to use and connecting the client to it if it's not
+ * already connected.
+ *
+ * Since we only need a single functioning mirror to satisfy a read, we don't
+ * want to return the layout if there is one. For writes though, any down
+ * mirror should result in a LAYOUTRETURN. @fail_return is how we distinguish
+ * between the two cases.
+ *
+ * Returns a pointer to a connected DS object on success or NULL on failure.
+ */
 struct nfs4_pnfs_ds *
 nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx,
 			  bool fail_return)
@@ -367,7 +370,6 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx,
 	struct inode *ino = lseg->pls_layout->plh_inode;
 	struct nfs_server *s = NFS_SERVER(ino);
 	unsigned int max_payload;
-	rpc_authflavor_t flavor;
 
 	if (!ff_layout_mirror_valid(lseg, mirror)) {
 		pr_err_ratelimited("NFS: %s: No data server for offset index %d\n",
@@ -383,9 +385,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx,
 	/* matching smp_wmb() in _nfs4_pnfs_v3/4_ds_connect */
 	smp_rmb();
 	if (ds->ds_clp)
-		goto out_update_creds;
-
-	flavor = nfs4_ff_layout_choose_authflavor(mirror);
+		goto out;
 
 	/* FIXME: For now we assume the server sent only one version of NFS
 	 * to use for the DS.
@@ -394,7 +394,7 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx,
 			     dataserver_retrans,
 			     mirror->mirror_ds->ds_versions[0].version,
 			     mirror->mirror_ds->ds_versions[0].minor_version,
-			     flavor);
+			     RPC_AUTH_UNIX);
 
 	/* connect success, check rsize/wsize limit */
 	if (ds->ds_clp) {
@@ -410,20 +410,10 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx,
 					 mirror, lseg->pls_range.offset,
 					 lseg->pls_range.length, NFS4ERR_NXIO,
 					 OP_ILLEGAL, GFP_NOIO);
-		if (!fail_return) {
-			if (ff_layout_has_available_ds(lseg))
-				set_bit(NFS_LAYOUT_RETURN_REQUESTED,
-					&lseg->pls_layout->plh_flags);
-			else
-				pnfs_error_mark_layout_for_return(ino, lseg);
-		} else
+		if (fail_return || !ff_layout_has_available_ds(lseg))
 			pnfs_error_mark_layout_for_return(ino, lseg);
 		ds = NULL;
-		goto out;
 	}
-out_update_creds:
-	if (ff_layout_update_mirror_cred(mirror, ds))
-		ds = NULL;
 out:
 	return ds;
 }
@@ -433,16 +423,15 @@ ff_layout_get_ds_cred(struct pnfs_layout_segment *lseg, u32 ds_idx,
 		      struct rpc_cred *mdscred)
 {
 	struct nfs4_ff_layout_mirror *mirror = FF_LAYOUT_COMP(lseg, ds_idx);
-	struct rpc_cred *cred = ERR_PTR(-EINVAL);
-
-	if (!nfs4_ff_layout_prepare_ds(lseg, ds_idx, true))
-		goto out;
+	struct rpc_cred *cred;
 
-	if (mirror && mirror->cred)
-		cred = mirror->cred;
-	else
-		cred = mdscred;
-out:
+	if (mirror) {
+		cred = ff_layout_get_mirror_cred(mirror, lseg->pls_range.iomode);
+		if (!cred)
+			cred = get_rpccred(mdscred);
+	} else {
+		cred = get_rpccred(mdscred);
+	}
 	return cred;
 }
 
@@ -562,6 +551,18 @@ bool ff_layout_has_available_ds(struct pnfs_layout_segment *lseg)
 	return ff_rw_layout_has_available_ds(lseg);
 }
 
+bool ff_layout_avoid_mds_available_ds(struct pnfs_layout_segment *lseg)
+{
+	return ff_layout_no_fallback_to_mds(lseg) ||
+	       ff_layout_has_available_ds(lseg);
+}
+
+bool ff_layout_avoid_read_on_rw(struct pnfs_layout_segment *lseg)
+{
+	return lseg->pls_range.iomode == IOMODE_RW &&
+	       ff_layout_no_read_on_rw(lseg);
+}
+
 module_param(dataserver_retrans, uint, 0644);
 MODULE_PARM_DESC(dataserver_retrans, "The  number of times the NFSv4.1 client "
 			"retries a request before it attempts further "

fs/nfs/internal.h

@@ -477,6 +477,7 @@ void nfs_mark_request_commit(struct nfs_page *req,
 			     u32 ds_commit_idx);
 int nfs_write_need_commit(struct nfs_pgio_header *);
 void nfs_writeback_update_inode(struct nfs_pgio_header *hdr);
+int nfs_commit_file(struct file *file, struct nfs_write_verifier *verf);
 int nfs_generic_commit_list(struct inode *inode, struct list_head *head,
 			    int how, struct nfs_commit_info *cinfo);
 void nfs_retry_commit(struct list_head *page_list,

fs/nfs/nfs42.h

@@ -13,6 +13,7 @@
 
 /* nfs4.2proc.c */
 int nfs42_proc_allocate(struct file *, loff_t, loff_t);
+ssize_t nfs42_proc_copy(struct file *, loff_t, struct file *, loff_t, size_t);
 int nfs42_proc_deallocate(struct file *, loff_t, loff_t);
 loff_t nfs42_proc_llseek(struct file *, loff_t, int);
 int nfs42_proc_layoutstats_generic(struct nfs_server *,

fs/nfs/nfs42proc.c

@@ -126,6 +126,111 @@ int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
 	return err;
 }
 
+static ssize_t _nfs42_proc_copy(struct file *src, loff_t pos_src,
+				struct nfs_lock_context *src_lock,
+				struct file *dst, loff_t pos_dst,
+				struct nfs_lock_context *dst_lock,
+				size_t count)
+{
+	struct nfs42_copy_args args = {
+		.src_fh		= NFS_FH(file_inode(src)),
+		.src_pos	= pos_src,
+		.dst_fh		= NFS_FH(file_inode(dst)),
+		.dst_pos	= pos_dst,
+		.count		= count,
+	};
+	struct nfs42_copy_res res;
+	struct rpc_message msg = {
+		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
+		.rpc_argp = &args,
+		.rpc_resp = &res,
+	};
+	struct inode *dst_inode = file_inode(dst);
+	struct nfs_server *server = NFS_SERVER(dst_inode);
+	int status;
+
+	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
+				     src_lock, FMODE_READ);
+	if (status)
+		return status;
+
+	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
+				     dst_lock, FMODE_WRITE);
+	if (status)
+		return status;
+
+	status = nfs4_call_sync(server->client, server, &msg,
+				&args.seq_args, &res.seq_res, 0);
+	if (status == -ENOTSUPP)
+		server->caps &= ~NFS_CAP_COPY;
+	if (status)
+		return status;
+
+	if (res.write_res.verifier.committed != NFS_FILE_SYNC) {
+		status = nfs_commit_file(dst, &res.write_res.verifier.verifier);
+		if (status)
+			return status;
+	}
+
+	truncate_pagecache_range(dst_inode, pos_dst,
+				 pos_dst + res.write_res.count);
+
+	return res.write_res.count;
+}
+
+ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
+			struct file *dst, loff_t pos_dst,
+			size_t count)
+{
+	struct nfs_server *server = NFS_SERVER(file_inode(dst));
+	struct nfs_lock_context *src_lock;
+	struct nfs_lock_context *dst_lock;
+	struct nfs4_exception src_exception = { };
+	struct nfs4_exception dst_exception = { };
+	ssize_t err, err2;
+
+	if (!nfs_server_capable(file_inode(dst), NFS_CAP_COPY))
+		return -EOPNOTSUPP;
+
+	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
+	if (IS_ERR(src_lock))
+		return PTR_ERR(src_lock);
+
+	src_exception.inode = file_inode(src);
+	src_exception.state = src_lock->open_context->state;
+
+	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
+	if (IS_ERR(dst_lock)) {
+		err = PTR_ERR(dst_lock);
+		goto out_put_src_lock;
+	}
+
+	dst_exception.inode = file_inode(dst);
+	dst_exception.state = dst_lock->open_context->state;
+
+	do {
+		inode_lock(file_inode(dst));
+		err = _nfs42_proc_copy(src, pos_src, src_lock,
+				       dst, pos_dst, dst_lock, count);
+		inode_unlock(file_inode(dst));
+
+		if (err == -ENOTSUPP) {
+			err = -EOPNOTSUPP;
+			break;
+		}
+
+		err2 = nfs4_handle_exception(server, err, &src_exception);
+		err  = nfs4_handle_exception(server, err, &dst_exception);
+		if (!err)
+			err = err2;
+	} while (src_exception.retry || dst_exception.retry);
+
+	nfs_put_lock_context(dst_lock);
+out_put_src_lock:
+	nfs_put_lock_context(src_lock);
+	return err;
+}
+
 static loff_t _nfs42_proc_llseek(struct file *filep,
 		struct nfs_lock_context *lock, loff_t offset, int whence)
 {
@@ -232,7 +337,7 @@ nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
 			 * with the current stateid.
 			 */
 			set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
-			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
+			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL, 0);
 			spin_unlock(&inode->i_lock);
 			pnfs_free_lseg_list(&head);
 		} else

fs/nfs/nfs42xdr.c

@@ -9,9 +9,22 @@
 #define encode_fallocate_maxsz		(encode_stateid_maxsz + \
 					 2 /* offset */ + \
 					 2 /* length */)
+#define NFS42_WRITE_RES_SIZE		(1 /* wr_callback_id size */ +\
+					 XDR_QUADLEN(NFS4_STATEID_SIZE) + \
+					 2 /* wr_count */ + \
+					 1 /* wr_committed */ + \
+					 XDR_QUADLEN(NFS4_VERIFIER_SIZE))
 #define encode_allocate_maxsz		(op_encode_hdr_maxsz + \
 					 encode_fallocate_maxsz)
 #define decode_allocate_maxsz		(op_decode_hdr_maxsz)
+#define encode_copy_maxsz		(op_encode_hdr_maxsz +          \
+					 XDR_QUADLEN(NFS4_STATEID_SIZE) + \
+					 XDR_QUADLEN(NFS4_STATEID_SIZE) + \
+					 2 + 2 + 2 + 1 + 1 + 1)
+#define decode_copy_maxsz		(op_decode_hdr_maxsz + \
+					 NFS42_WRITE_RES_SIZE + \
+					 1 /* cr_consecutive */ + \
+					 1 /* cr_synchronous */)
 #define encode_deallocate_maxsz		(op_encode_hdr_maxsz + \
 					 encode_fallocate_maxsz)
 #define decode_deallocate_maxsz		(op_decode_hdr_maxsz)
@@ -49,6 +62,16 @@
 					 decode_putfh_maxsz + \
 					 decode_allocate_maxsz + \
 					 decode_getattr_maxsz)
+#define NFS4_enc_copy_sz		(compound_encode_hdr_maxsz + \
+					 encode_putfh_maxsz + \
+					 encode_savefh_maxsz + \
+					 encode_putfh_maxsz + \
+					 encode_copy_maxsz)
+#define NFS4_dec_copy_sz		(compound_decode_hdr_maxsz + \
+					 decode_putfh_maxsz + \
+					 decode_savefh_maxsz + \
+					 decode_putfh_maxsz + \
+					 decode_copy_maxsz)
 #define NFS4_enc_deallocate_sz		(compound_encode_hdr_maxsz + \
 					 encode_putfh_maxsz + \
 					 encode_deallocate_maxsz + \
@@ -102,6 +125,23 @@ static void encode_allocate(struct xdr_stream *xdr,
 	encode_fallocate(xdr, args);
 }
 
+static void encode_copy(struct xdr_stream *xdr,
+			struct nfs42_copy_args *args,
+			struct compound_hdr *hdr)
+{
+	encode_op_hdr(xdr, OP_COPY, decode_copy_maxsz, hdr);
+	encode_nfs4_stateid(xdr, &args->src_stateid);
+	encode_nfs4_stateid(xdr, &args->dst_stateid);
+
+	encode_uint64(xdr, args->src_pos);
+	encode_uint64(xdr, args->dst_pos);
+	encode_uint64(xdr, args->count);
+
+	encode_uint32(xdr, 1); /* consecutive = true */
+	encode_uint32(xdr, 1); /* synchronous = true */
+	encode_uint32(xdr, 0); /* src server list */
+}
+
 static void encode_deallocate(struct xdr_stream *xdr,
 			      struct nfs42_falloc_args *args,
 			      struct compound_hdr *hdr)
@@ -181,6 +221,26 @@ static void nfs4_xdr_enc_allocate(struct rpc_rqst *req,
 	encode_nops(&hdr);
 }
 
+/*
+ * Encode COPY request
+ */
+static void nfs4_xdr_enc_copy(struct rpc_rqst *req,
+			      struct xdr_stream *xdr,
+			      struct nfs42_copy_args *args)
+{
+	struct compound_hdr hdr = {
+		.minorversion = nfs4_xdr_minorversion(&args->seq_args),
+	};
+
+	encode_compound_hdr(xdr, req, &hdr);
+	encode_sequence(xdr, &args->seq_args, &hdr);
+	encode_putfh(xdr, args->src_fh, &hdr);
+	encode_savefh(xdr, &hdr);
+	encode_putfh(xdr, args->dst_fh, &hdr);
+	encode_copy(xdr, args, &hdr);
+	encode_nops(&hdr);
+}
+
 /*
  * Encode DEALLOCATE request
  */
@@ -266,6 +326,62 @@ static int decode_allocate(struct xdr_stream *xdr, struct nfs42_falloc_res *res)
 	return decode_op_hdr(xdr, OP_ALLOCATE);
 }
 
+static int decode_write_response(struct xdr_stream *xdr,
+				 struct nfs42_write_res *res)
+{
+	__be32 *p;
+	int stateids;
+
+	p = xdr_inline_decode(xdr, 4 + 8 + 4);
+	if (unlikely(!p))
+		goto out_overflow;
+
+	stateids = be32_to_cpup(p++);
+	p = xdr_decode_hyper(p, &res->count);
+	res->verifier.committed = be32_to_cpup(p);
+	return decode_verifier(xdr, &res->verifier.verifier);
+
+out_overflow:
+	print_overflow_msg(__func__, xdr);
+	return -EIO;
+}
+
+static int decode_copy_requirements(struct xdr_stream *xdr,
+				    struct nfs42_copy_res *res) {
+	__be32 *p;
+
+	p = xdr_inline_decode(xdr, 4 + 4);
+	if (unlikely(!p))
+		goto out_overflow;
+
+	res->consecutive = be32_to_cpup(p++);
+	res->synchronous = be32_to_cpup(p++);
+	return 0;
+out_overflow:
+	print_overflow_msg(__func__, xdr);
+	return -EIO;
+}
+
+static int decode_copy(struct xdr_stream *xdr, struct nfs42_copy_res *res)
+{
+	int status;
+
+	status = decode_op_hdr(xdr, OP_COPY);
+	if (status == NFS4ERR_OFFLOAD_NO_REQS) {
+		status = decode_copy_requirements(xdr, res);
+		if (status)
+			return status;
+		return NFS4ERR_OFFLOAD_NO_REQS;
+	} else if (status)
+		return status;
+
+	status = decode_write_response(xdr, &res->write_res);
+	if (status)
+		return status;
+
+	return decode_copy_requirements(xdr, res);
+}
+
 static int decode_deallocate(struct xdr_stream *xdr, struct nfs42_falloc_res *res)
 {
 	return decode_op_hdr(xdr, OP_DEALLOCATE);
@@ -330,6 +446,36 @@ static int nfs4_xdr_dec_allocate(struct rpc_rqst *rqstp,
 	return status;
 }
 
+/*
+ * Decode COPY response
+ */
+static int nfs4_xdr_dec_copy(struct rpc_rqst *rqstp,
+			     struct xdr_stream *xdr,
+			     struct nfs42_copy_res *res)
+{
+	struct compound_hdr hdr;
+	int status;
+
+	status = decode_compound_hdr(xdr, &hdr);
+	if (status)
+		goto out;
+	status = decode_sequence(xdr, &res->seq_res, rqstp);
+	if (status)
+		goto out;
+	status = decode_putfh(xdr);
+	if (status)
+		goto out;
+	status = decode_savefh(xdr);
+	if (status)
+		goto out;
+	status = decode_putfh(xdr);
+	if (status)
+		goto out;
+	status = decode_copy(xdr, res);
+out:
+	return status;
+}
+
 /*
  * Decode DEALLOCATE request
  */

fs/nfs/nfs4_fs.h

@@ -438,8 +438,9 @@ extern void nfs41_handle_server_scope(struct nfs_client *,
 				      struct nfs41_server_scope **);
 extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp);
 extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl);
-extern int nfs4_select_rw_stateid(nfs4_stateid *, struct nfs4_state *,
-		fmode_t, const struct nfs_lockowner *);
+extern int nfs4_select_rw_stateid(struct nfs4_state *, fmode_t,
+		const struct nfs_lockowner *, nfs4_stateid *,
+		struct rpc_cred **);
 
 extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask);
 extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task);
@@ -496,12 +497,15 @@ extern struct svc_version nfs4_callback_version4;
 
 static inline void nfs4_stateid_copy(nfs4_stateid *dst, const nfs4_stateid *src)
 {
-	memcpy(dst, src, sizeof(*dst));
+	memcpy(dst->data, src->data, sizeof(dst->data));
+	dst->type = src->type;
 }
 
 static inline bool nfs4_stateid_match(const nfs4_stateid *dst, const nfs4_stateid *src)
 {
-	return memcmp(dst, src, sizeof(*dst)) == 0;
+	if (dst->type != src->type)
+		return false;
+	return memcmp(dst->data, src->data, sizeof(dst->data)) == 0;
 }
 
 static inline bool nfs4_stateid_match_other(const nfs4_stateid *dst, const nfs4_stateid *src)

fs/nfs/nfs4file.c

@@ -129,6 +129,28 @@ nfs4_file_flush(struct file *file, fl_owner_t id)
 }
 
 #ifdef CONFIG_NFS_V4_2
+static ssize_t nfs4_copy_file_range(struct file *file_in, loff_t pos_in,
+				    struct file *file_out, loff_t pos_out,
+				    size_t count, unsigned int flags)
+{
+	struct inode *in_inode = file_inode(file_in);
+	struct inode *out_inode = file_inode(file_out);
+	int ret;
+
+	if (in_inode == out_inode)
+		return -EINVAL;
+
+	/* flush any pending writes */
+	ret = nfs_sync_inode(in_inode);
+	if (ret)
+		return ret;
+	ret = nfs_sync_inode(out_inode);
+	if (ret)
+		return ret;
+
+	return nfs42_proc_copy(file_in, pos_in, file_out, pos_out, count);
+}
+
 static loff_t nfs4_file_llseek(struct file *filep, loff_t offset, int whence)
 {
 	loff_t ret;
@@ -243,6 +265,7 @@ const struct file_operations nfs4_file_operations = {
 	.check_flags	= nfs_check_flags,
 	.setlease	= simple_nosetlease,
 #ifdef CONFIG_NFS_V4_2
+	.copy_file_range = nfs4_copy_file_range,
 	.llseek		= nfs4_file_llseek,
 	.fallocate	= nfs42_fallocate,
 	.clone_file_range = nfs42_clone_file_range,

fs/nfs/nfs4state.c

@@ -65,7 +65,10 @@
 
 #define OPENOWNER_POOL_SIZE	8
 
-const nfs4_stateid zero_stateid;
+const nfs4_stateid zero_stateid = {
+	.data = { 0 },
+	.type = NFS4_SPECIAL_STATEID_TYPE,
+};
 static DEFINE_MUTEX(nfs_clid_init_mutex);
 
 int nfs4_init_clientid(struct nfs_client *clp, struct rpc_cred *cred)
@@ -985,15 +988,20 @@ static void nfs4_copy_open_stateid(nfs4_stateid *dst, struct nfs4_state *state)
  * Byte-range lock aware utility to initialize the stateid of read/write
  * requests.
  */
-int nfs4_select_rw_stateid(nfs4_stateid *dst, struct nfs4_state *state,
-		fmode_t fmode, const struct nfs_lockowner *lockowner)
+int nfs4_select_rw_stateid(struct nfs4_state *state,
+		fmode_t fmode, const struct nfs_lockowner *lockowner,
+		nfs4_stateid *dst, struct rpc_cred **cred)
 {
-	int ret = nfs4_copy_lock_stateid(dst, state, lockowner);
+	int ret;
+
+	if (cred != NULL)
+		*cred = NULL;
+	ret = nfs4_copy_lock_stateid(dst, state, lockowner);
 	if (ret == -EIO)
 		/* A lost lock - don't even consider delegations */
 		goto out;
 	/* returns true if delegation stateid found and copied */
-	if (nfs4_copy_delegation_stateid(dst, state->inode, fmode)) {
+	if (nfs4_copy_delegation_stateid(state->inode, fmode, dst, cred)) {
 		ret = 0;
 		goto out;
 	}

fs/nfs/nfs4trace.h

@@ -1520,6 +1520,8 @@ DEFINE_NFS4_INODE_EVENT(nfs4_layoutreturn_on_close);
 		{ PNFS_UPDATE_LAYOUT_FOUND_CACHED, "found cached" },	\
 		{ PNFS_UPDATE_LAYOUT_RETURN, "layoutreturn" },		\
 		{ PNFS_UPDATE_LAYOUT_BLOCKED, "layouts blocked" },	\
+		{ PNFS_UPDATE_LAYOUT_INVALID_OPEN, "invalid open" },	\
+		{ PNFS_UPDATE_LAYOUT_RETRY, "retrying" },	\
 		{ PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET, "sent layoutget" })
 
 TRACE_EVENT(pnfs_update_layout,
@@ -1528,9 +1530,10 @@ TRACE_EVENT(pnfs_update_layout,
 			u64 count,
 			enum pnfs_iomode iomode,
 			struct pnfs_layout_hdr *lo,
+			struct pnfs_layout_segment *lseg,
 			enum pnfs_update_layout_reason reason
 		),
-		TP_ARGS(inode, pos, count, iomode, lo, reason),
+		TP_ARGS(inode, pos, count, iomode, lo, lseg, reason),
 		TP_STRUCT__entry(
 			__field(dev_t, dev)
 			__field(u64, fileid)
@@ -1540,6 +1543,7 @@ TRACE_EVENT(pnfs_update_layout,
 			__field(enum pnfs_iomode, iomode)
 			__field(int, layoutstateid_seq)
 			__field(u32, layoutstateid_hash)
+			__field(long, lseg)
 			__field(enum pnfs_update_layout_reason, reason)
 		),
 		TP_fast_assign(
@@ -1559,11 +1563,12 @@ TRACE_EVENT(pnfs_update_layout,
 				__entry->layoutstateid_seq = 0;
 				__entry->layoutstateid_hash = 0;
 			}
+			__entry->lseg = (long)lseg;
 		),
 		TP_printk(
 			"fileid=%02x:%02x:%llu fhandle=0x%08x "
 			"iomode=%s pos=%llu count=%llu "
-			"layoutstateid=%d:0x%08x (%s)",
+			"layoutstateid=%d:0x%08x lseg=0x%lx (%s)",
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
 			__entry->fhandle,
@@ -1571,6 +1576,7 @@ TRACE_EVENT(pnfs_update_layout,
 			(unsigned long long)__entry->pos,
 			(unsigned long long)__entry->count,
 			__entry->layoutstateid_seq, __entry->layoutstateid_hash,
+			__entry->lseg,
 			show_pnfs_update_layout_reason(__entry->reason)
 		)
 );

fs/nfs/nfs4xdr.c

@@ -4270,6 +4270,24 @@ static int decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
 	return decode_opaque_fixed(xdr, stateid, NFS4_STATEID_SIZE);
 }
 
+static int decode_open_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	stateid->type = NFS4_OPEN_STATEID_TYPE;
+	return decode_stateid(xdr, stateid);
+}
+
+static int decode_lock_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	stateid->type = NFS4_LOCK_STATEID_TYPE;
+	return decode_stateid(xdr, stateid);
+}
+
+static int decode_delegation_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	stateid->type = NFS4_DELEGATION_STATEID_TYPE;
+	return decode_stateid(xdr, stateid);
+}
+
 static int decode_close(struct xdr_stream *xdr, struct nfs_closeres *res)
 {
 	int status;
@@ -4278,7 +4296,7 @@ static int decode_close(struct xdr_stream *xdr, struct nfs_closeres *res)
 	if (status != -EIO)
 		nfs_increment_open_seqid(status, res->seqid);
 	if (!status)
-		status = decode_stateid(xdr, &res->stateid);
+		status = decode_open_stateid(xdr, &res->stateid);
 	return status;
 }
 
@@ -4937,7 +4955,7 @@ static int decode_lock(struct xdr_stream *xdr, struct nfs_lock_res *res)
 	if (status == -EIO)
 		goto out;
 	if (status == 0) {
-		status = decode_stateid(xdr, &res->stateid);
+		status = decode_lock_stateid(xdr, &res->stateid);
 		if (unlikely(status))
 			goto out;
 	} else if (status == -NFS4ERR_DENIED)
@@ -4966,7 +4984,7 @@ static int decode_locku(struct xdr_stream *xdr, struct nfs_locku_res *res)
 	if (status != -EIO)
 		nfs_increment_lock_seqid(status, res->seqid);
 	if (status == 0)
-		status = decode_stateid(xdr, &res->stateid);
+		status = decode_lock_stateid(xdr, &res->stateid);
 	return status;
 }
 
@@ -5016,7 +5034,7 @@ static int decode_rw_delegation(struct xdr_stream *xdr,
 	__be32 *p;
 	int status;
 
-	status = decode_stateid(xdr, &res->delegation);
+	status = decode_delegation_stateid(xdr, &res->delegation);
 	if (unlikely(status))
 		return status;
 	p = xdr_inline_decode(xdr, 4);
@@ -5096,7 +5114,7 @@ static int decode_open(struct xdr_stream *xdr, struct nfs_openres *res)
 	nfs_increment_open_seqid(status, res->seqid);
 	if (status)
 		return status;
-	status = decode_stateid(xdr, &res->stateid);
+	status = decode_open_stateid(xdr, &res->stateid);
 	if (unlikely(status))
 		return status;
 
@@ -5136,7 +5154,7 @@ static int decode_open_confirm(struct xdr_stream *xdr, struct nfs_open_confirmre
 	if (status != -EIO)
 		nfs_increment_open_seqid(status, res->seqid);
 	if (!status)
-		status = decode_stateid(xdr, &res->stateid);
+		status = decode_open_stateid(xdr, &res->stateid);
 	return status;
 }
 
@@ -5148,7 +5166,7 @@ static int decode_open_downgrade(struct xdr_stream *xdr, struct nfs_closeres *re
 	if (status != -EIO)
 		nfs_increment_open_seqid(status, res->seqid);
 	if (!status)
-		status = decode_stateid(xdr, &res->stateid);
+		status = decode_open_stateid(xdr, &res->stateid);
 	return status;
 }
 
@@ -5838,6 +5856,12 @@ static int decode_sequence(struct xdr_stream *xdr,
 }
 
 #if defined(CONFIG_NFS_V4_1)
+static int decode_layout_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	stateid->type = NFS4_LAYOUT_STATEID_TYPE;
+	return decode_stateid(xdr, stateid);
+}
+
 static int decode_getdeviceinfo(struct xdr_stream *xdr,
 				struct nfs4_getdeviceinfo_res *res)
 {
@@ -5919,7 +5943,7 @@ static int decode_layoutget(struct xdr_stream *xdr, struct rpc_rqst *req,
 	if (unlikely(!p))
 		goto out_overflow;
 	res->return_on_close = be32_to_cpup(p);
-	decode_stateid(xdr, &res->stateid);
+	decode_layout_stateid(xdr, &res->stateid);
 	p = xdr_inline_decode(xdr, 4);
 	if (unlikely(!p))
 		goto out_overflow;
@@ -5985,7 +6009,7 @@ static int decode_layoutreturn(struct xdr_stream *xdr,
 		goto out_overflow;
 	res->lrs_present = be32_to_cpup(p);
 	if (res->lrs_present)
-		status = decode_stateid(xdr, &res->stateid);
+		status = decode_layout_stateid(xdr, &res->stateid);
 	return status;
 out_overflow:
 	print_overflow_msg(__func__, xdr);
@@ -7515,6 +7539,7 @@ struct rpc_procinfo	nfs4_procedures[] = {
 	PROC(DEALLOCATE,	enc_deallocate,		dec_deallocate),
 	PROC(LAYOUTSTATS,	enc_layoutstats,	dec_layoutstats),
 	PROC(CLONE,		enc_clone,		dec_clone),
+	PROC(COPY,		enc_copy,		dec_copy),
 #endif /* CONFIG_NFS_V4_2 */
 };
 

fs/nfs/pagelist.c

@@ -341,8 +341,10 @@ nfs_create_request(struct nfs_open_context *ctx, struct page *page,
 	 * long write-back delay. This will be adjusted in
 	 * update_nfs_request below if the region is not locked. */
 	req->wb_page    = page;
-	req->wb_index	= page_file_index(page);
-	get_page(page);
+	if (page) {
+		req->wb_index = page_file_index(page);
+		get_page(page);
+	}
 	req->wb_offset  = offset;
 	req->wb_pgbase	= offset;
 	req->wb_bytes   = count;

fs/nfs/pnfs.h

@@ -64,6 +64,7 @@ struct pnfs_layout_segment {
 	struct list_head pls_lc_list;
 	struct pnfs_layout_range pls_range;
 	atomic_t pls_refcount;
+	u32 pls_seq;
 	unsigned long pls_flags;
 	struct pnfs_layout_hdr *pls_layout;
 	struct work_struct pls_work;
@@ -194,6 +195,7 @@ struct pnfs_layout_hdr {
 	unsigned long		plh_flags;
 	nfs4_stateid		plh_stateid;
 	u32			plh_barrier; /* ignore lower seqids */
+	u32			plh_return_seq;
 	enum pnfs_iomode	plh_return_iomode;
 	loff_t			plh_lwb; /* last write byte for layoutcommit */
 	struct rpc_cred		*plh_lc_cred; /* layoutcommit cred */
@@ -226,7 +228,7 @@ extern void pnfs_unregister_layoutdriver(struct pnfs_layoutdriver_type *);
 extern int nfs4_proc_getdeviceinfo(struct nfs_server *server,
 				   struct pnfs_device *dev,
 				   struct rpc_cred *cred);
-extern struct pnfs_layout_segment* nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags);
+extern struct pnfs_layout_segment* nfs4_proc_layoutget(struct nfs4_layoutget *lgp, long *timeout, gfp_t gfp_flags);
 extern int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync);
 
 /* pnfs.c */
@@ -258,16 +260,14 @@ void pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo);
 void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo,
 			     const nfs4_stateid *new,
 			     bool update_barrier);
-int pnfs_choose_layoutget_stateid(nfs4_stateid *dst,
-				  struct pnfs_layout_hdr *lo,
-				  const struct pnfs_layout_range *range,
-				  struct nfs4_state *open_state);
 int pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 				struct list_head *tmp_list,
-				const struct pnfs_layout_range *recall_range);
+				const struct pnfs_layout_range *recall_range,
+				u32 seq);
 int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 				struct list_head *tmp_list,
-				const struct pnfs_layout_range *recall_range);
+				const struct pnfs_layout_range *recall_range,
+				u32 seq);
 bool pnfs_roc(struct inode *ino);
 void pnfs_roc_release(struct inode *ino);
 void pnfs_roc_set_barrier(struct inode *ino, u32 barrier);
@@ -282,12 +282,13 @@ int _pnfs_return_layout(struct inode *);
 int pnfs_commit_and_return_layout(struct inode *);
 void pnfs_ld_write_done(struct nfs_pgio_header *);
 void pnfs_ld_read_done(struct nfs_pgio_header *);
-int pnfs_read_resend_pnfs(struct nfs_pgio_header *);
+void pnfs_read_resend_pnfs(struct nfs_pgio_header *);
 struct pnfs_layout_segment *pnfs_update_layout(struct inode *ino,
 					       struct nfs_open_context *ctx,
 					       loff_t pos,
 					       u64 count,
 					       enum pnfs_iomode iomode,
+					       bool strict_iomode,
 					       gfp_t gfp_flags);
 void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo);
 

fs/nfs/pnfs_nfs.c

@@ -61,7 +61,7 @@ EXPORT_SYMBOL_GPL(pnfs_generic_commit_release);
 
 /* The generic layer is about to remove the req from the commit list.
  * If this will make the bucket empty, it will need to put the lseg reference.
- * Note this must be called holding the inode (/cinfo) lock
+ * Note this must be called holding i_lock
  */
 void
 pnfs_generic_clear_request_commit(struct nfs_page *req,
@@ -98,7 +98,7 @@ pnfs_generic_transfer_commit_list(struct list_head *src, struct list_head *dst,
 		if (!nfs_lock_request(req))
 			continue;
 		kref_get(&req->wb_kref);
-		if (cond_resched_lock(cinfo->lock))
+		if (cond_resched_lock(&cinfo->inode->i_lock))
 			list_safe_reset_next(req, tmp, wb_list);
 		nfs_request_remove_commit_list(req, cinfo);
 		clear_bit(PG_COMMIT_TO_DS, &req->wb_flags);
@@ -119,7 +119,7 @@ pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket,
 	struct list_head *dst = &bucket->committing;
 	int ret;
 
-	lockdep_assert_held(cinfo->lock);
+	lockdep_assert_held(&cinfo->inode->i_lock);
 	ret = pnfs_generic_transfer_commit_list(src, dst, cinfo, max);
 	if (ret) {
 		cinfo->ds->nwritten -= ret;
@@ -142,7 +142,7 @@ int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo,
 {
 	int i, rv = 0, cnt;
 
-	lockdep_assert_held(cinfo->lock);
+	lockdep_assert_held(&cinfo->inode->i_lock);
 	for (i = 0; i < cinfo->ds->nbuckets && max != 0; i++) {
 		cnt = pnfs_generic_scan_ds_commit_list(&cinfo->ds->buckets[i],
 						       cinfo, max);
@@ -161,16 +161,16 @@ void pnfs_generic_recover_commit_reqs(struct list_head *dst,
 	struct pnfs_layout_segment *freeme;
 	int i;
 
-	lockdep_assert_held(cinfo->lock);
+	lockdep_assert_held(&cinfo->inode->i_lock);
 restart:
 	for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) {
 		if (pnfs_generic_transfer_commit_list(&b->written, dst,
 						      cinfo, 0)) {
 			freeme = b->wlseg;
 			b->wlseg = NULL;
-			spin_unlock(cinfo->lock);
+			spin_unlock(&cinfo->inode->i_lock);
 			pnfs_put_lseg(freeme);
-			spin_lock(cinfo->lock);
+			spin_lock(&cinfo->inode->i_lock);
 			goto restart;
 		}
 	}
@@ -186,7 +186,7 @@ static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx)
 	LIST_HEAD(pages);
 	int i;
 
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 	for (i = idx; i < fl_cinfo->nbuckets; i++) {
 		bucket = &fl_cinfo->buckets[i];
 		if (list_empty(&bucket->committing))
@@ -194,12 +194,12 @@ static void pnfs_generic_retry_commit(struct nfs_commit_info *cinfo, int idx)
 		freeme = bucket->clseg;
 		bucket->clseg = NULL;
 		list_splice_init(&bucket->committing, &pages);
-		spin_unlock(cinfo->lock);
+		spin_unlock(&cinfo->inode->i_lock);
 		nfs_retry_commit(&pages, freeme, cinfo, i);
 		pnfs_put_lseg(freeme);
-		spin_lock(cinfo->lock);
+		spin_lock(&cinfo->inode->i_lock);
 	}
-	spin_unlock(cinfo->lock);
+	spin_unlock(&cinfo->inode->i_lock);
 }
 
 static unsigned int
@@ -238,14 +238,31 @@ void pnfs_fetch_commit_bucket_list(struct list_head *pages,
 	struct pnfs_commit_bucket *bucket;
 
 	bucket = &cinfo->ds->buckets[data->ds_commit_index];
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 	list_splice_init(&bucket->committing, pages);
 	data->lseg = bucket->clseg;
 	bucket->clseg = NULL;
-	spin_unlock(cinfo->lock);
+	spin_unlock(&cinfo->inode->i_lock);
 
 }
 
+/* Helper function for pnfs_generic_commit_pagelist to catch an empty
+ * page list. This can happen when two commits race. */
+static bool
+pnfs_generic_commit_cancel_empty_pagelist(struct list_head *pages,
+					  struct nfs_commit_data *data,
+					  struct nfs_commit_info *cinfo)
+{
+	if (list_empty(pages)) {
+		if (atomic_dec_and_test(&cinfo->mds->rpcs_out))
+			wake_up_atomic_t(&cinfo->mds->rpcs_out);
+		nfs_commitdata_release(data);
+		return true;
+	}
+
+	return false;
+}
+
 /* This follows nfs_commit_list pretty closely */
 int
 pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
@@ -280,6 +297,11 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
 	list_for_each_entry_safe(data, tmp, &list, pages) {
 		list_del_init(&data->pages);
 		if (data->ds_commit_index < 0) {
+			/* another commit raced with us */
+			if (pnfs_generic_commit_cancel_empty_pagelist(mds_pages,
+				data, cinfo))
+				continue;
+
 			nfs_init_commit(data, mds_pages, NULL, cinfo);
 			nfs_initiate_commit(NFS_CLIENT(inode), data,
 					    NFS_PROTO(data->inode),
@@ -288,6 +310,12 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
 			LIST_HEAD(pages);
 
 			pnfs_fetch_commit_bucket_list(&pages, data, cinfo);
+
+			/* another commit raced with us */
+			if (pnfs_generic_commit_cancel_empty_pagelist(&pages,
+				data, cinfo))
+				continue;
+
 			nfs_init_commit(data, &pages, data->lseg, cinfo);
 			initiate_commit(data, how);
 		}
@@ -874,12 +902,12 @@ pnfs_layout_mark_request_commit(struct nfs_page *req,
 	struct list_head *list;
 	struct pnfs_commit_bucket *buckets;
 
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 	buckets = cinfo->ds->buckets;
 	list = &buckets[ds_commit_idx].written;
 	if (list_empty(list)) {
 		if (!pnfs_is_valid_lseg(lseg)) {
-			spin_unlock(cinfo->lock);
+			spin_unlock(&cinfo->inode->i_lock);
 			cinfo->completion_ops->resched_write(cinfo, req);
 			return;
 		}
@@ -896,7 +924,7 @@ pnfs_layout_mark_request_commit(struct nfs_page *req,
 	cinfo->ds->nwritten++;
 
 	nfs_request_add_commit_list_locked(req, list, cinfo);
-	spin_unlock(cinfo->lock);
+	spin_unlock(&cinfo->inode->i_lock);
 	nfs_mark_page_unstable(req->wb_page, cinfo);
 }
 EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit);

fs/nfs/super.c

@@ -191,6 +191,7 @@ static const match_table_t nfs_mount_option_tokens = {
 
 enum {
 	Opt_xprt_udp, Opt_xprt_udp6, Opt_xprt_tcp, Opt_xprt_tcp6, Opt_xprt_rdma,
+	Opt_xprt_rdma6,
 
 	Opt_xprt_err
 };
@@ -201,6 +202,7 @@ static const match_table_t nfs_xprt_protocol_tokens = {
 	{ Opt_xprt_tcp, "tcp" },
 	{ Opt_xprt_tcp6, "tcp6" },
 	{ Opt_xprt_rdma, "rdma" },
+	{ Opt_xprt_rdma6, "rdma6" },
 
 	{ Opt_xprt_err, NULL }
 };
@@ -1456,6 +1458,8 @@ static int nfs_parse_mount_options(char *raw,
 				mnt->flags |= NFS_MOUNT_TCP;
 				mnt->nfs_server.protocol = XPRT_TRANSPORT_TCP;
 				break;
+			case Opt_xprt_rdma6:
+				protofamily = AF_INET6;
 			case Opt_xprt_rdma:
 				/* vector side protocols to TCP */
 				mnt->flags |= NFS_MOUNT_TCP;
@@ -2408,6 +2412,11 @@ static int nfs_compare_super_address(struct nfs_server *server1,
 				     struct nfs_server *server2)
 {
 	struct sockaddr *sap1, *sap2;
+	struct rpc_xprt *xprt1 = server1->client->cl_xprt;
+	struct rpc_xprt *xprt2 = server2->client->cl_xprt;
+
+	if (!net_eq(xprt1->xprt_net, xprt2->xprt_net))
+		return 0;
 
 	sap1 = (struct sockaddr *)&server1->nfs_client->cl_addr;
 	sap2 = (struct sockaddr *)&server2->nfs_client->cl_addr;

fs/nfs/write.c

@@ -245,8 +245,7 @@ static void nfs_mark_uptodate(struct nfs_page *req)
 static int wb_priority(struct writeback_control *wbc)
 {
 	int ret = 0;
-	if (wbc->for_reclaim)
-		return FLUSH_HIGHPRI | FLUSH_COND_STABLE;
+
 	if (wbc->sync_mode == WB_SYNC_ALL)
 		ret = FLUSH_COND_STABLE;
 	return ret;
@@ -737,7 +736,7 @@ static void nfs_inode_remove_request(struct nfs_page *req)
 		head = req->wb_head;
 
 		spin_lock(&inode->i_lock);
-		if (likely(!PageSwapCache(head->wb_page))) {
+		if (likely(head->wb_page && !PageSwapCache(head->wb_page))) {
 			set_page_private(head->wb_page, 0);
 			ClearPagePrivate(head->wb_page);
 			smp_mb__after_atomic();
@@ -759,7 +758,8 @@ static void nfs_inode_remove_request(struct nfs_page *req)
 static void
 nfs_mark_request_dirty(struct nfs_page *req)
 {
-	__set_page_dirty_nobuffers(req->wb_page);
+	if (req->wb_page)
+		__set_page_dirty_nobuffers(req->wb_page);
 }
 
 /*
@@ -804,7 +804,7 @@ nfs_page_search_commits_for_head_request_locked(struct nfs_inode *nfsi,
  * number of outstanding requests requiring a commit as well as
  * the MM page stats.
  *
- * The caller must hold the cinfo->lock, and the nfs_page lock.
+ * The caller must hold cinfo->inode->i_lock, and the nfs_page lock.
  */
 void
 nfs_request_add_commit_list_locked(struct nfs_page *req, struct list_head *dst,
@@ -832,10 +832,11 @@ EXPORT_SYMBOL_GPL(nfs_request_add_commit_list_locked);
 void
 nfs_request_add_commit_list(struct nfs_page *req, struct nfs_commit_info *cinfo)
 {
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 	nfs_request_add_commit_list_locked(req, &cinfo->mds->list, cinfo);
-	spin_unlock(cinfo->lock);
-	nfs_mark_page_unstable(req->wb_page, cinfo);
+	spin_unlock(&cinfo->inode->i_lock);
+	if (req->wb_page)
+		nfs_mark_page_unstable(req->wb_page, cinfo);
 }
 EXPORT_SYMBOL_GPL(nfs_request_add_commit_list);
 
@@ -864,7 +865,7 @@ EXPORT_SYMBOL_GPL(nfs_request_remove_commit_list);
 static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
 				      struct inode *inode)
 {
-	cinfo->lock = &inode->i_lock;
+	cinfo->inode = inode;
 	cinfo->mds = &NFS_I(inode)->commit_info;
 	cinfo->ds = pnfs_get_ds_info(inode);
 	cinfo->dreq = NULL;
@@ -967,7 +968,7 @@ nfs_reqs_to_commit(struct nfs_commit_info *cinfo)
 	return cinfo->mds->ncommit;
 }
 
-/* cinfo->lock held by caller */
+/* cinfo->inode->i_lock held by caller */
 int
 nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
 		     struct nfs_commit_info *cinfo, int max)
@@ -979,7 +980,7 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst,
 		if (!nfs_lock_request(req))
 			continue;
 		kref_get(&req->wb_kref);
-		if (cond_resched_lock(cinfo->lock))
+		if (cond_resched_lock(&cinfo->inode->i_lock))
 			list_safe_reset_next(req, tmp, wb_list);
 		nfs_request_remove_commit_list(req, cinfo);
 		nfs_list_add_request(req, dst);
@@ -1005,7 +1006,7 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst,
 {
 	int ret = 0;
 
-	spin_lock(cinfo->lock);
+	spin_lock(&cinfo->inode->i_lock);
 	if (cinfo->mds->ncommit > 0) {
 		const int max = INT_MAX;
 
@@ -1013,7 +1014,7 @@ nfs_scan_commit(struct inode *inode, struct list_head *dst,
 					   cinfo, max);
 		ret += pnfs_scan_commit_lists(inode, cinfo, max - ret);
 	}
-	spin_unlock(cinfo->lock);
+	spin_unlock(&cinfo->inode->i_lock);
 	return ret;
 }
 
@@ -1709,6 +1710,10 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how,
 {
 	struct nfs_commit_data	*data;
 
+	/* another commit raced with us */
+	if (list_empty(head))
+		return 0;
+
 	data = nfs_commitdata_alloc();
 
 	if (!data)
@@ -1724,6 +1729,36 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how,
 	return -ENOMEM;
 }
 
+int nfs_commit_file(struct file *file, struct nfs_write_verifier *verf)
+{
+	struct inode *inode = file_inode(file);
+	struct nfs_open_context *open;
+	struct nfs_commit_info cinfo;
+	struct nfs_page *req;
+	int ret;
+
+	open = get_nfs_open_context(nfs_file_open_context(file));
+	req  = nfs_create_request(open, NULL, NULL, 0, i_size_read(inode));
+	if (IS_ERR(req)) {
+		ret = PTR_ERR(req);
+		goto out_put;
+	}
+
+	nfs_init_cinfo_from_inode(&cinfo, inode);
+
+	memcpy(&req->wb_verf, verf, sizeof(struct nfs_write_verifier));
+	nfs_request_add_commit_list(req, &cinfo);
+	ret = nfs_commit_inode(inode, FLUSH_SYNC);
+	if (ret > 0)
+		ret = 0;
+
+	nfs_free_request(req);
+out_put:
+	put_nfs_open_context(open);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(nfs_commit_file);
+
 /*
  * COMMIT call returned
  */
@@ -1748,7 +1783,8 @@ static void nfs_commit_release_pages(struct nfs_commit_data *data)
 	while (!list_empty(&data->pages)) {
 		req = nfs_list_entry(data->pages.next);
 		nfs_list_remove_request(req);
-		nfs_clear_page_commit(req->wb_page);
+		if (req->wb_page)
+			nfs_clear_page_commit(req->wb_page);
 
 		dprintk("NFS:       commit (%s/%llu %d@%lld)",
 			req->wb_context->dentry->d_sb->s_id,

include/linux/errno.h

@@ -28,5 +28,6 @@
 #define EBADTYPE	527	/* Type not supported by server */
 #define EJUKEBOX	528	/* Request initiated, but will not complete before timeout */
 #define EIOCBQUEUED	529	/* iocb queued, will get completion event */
+#define ERECALLCONFLICT	530	/* conflict with recalled state */
 
 #endif

include/linux/nfs4.h

@@ -50,12 +50,27 @@ struct nfs4_label {
 
 typedef struct { char data[NFS4_VERIFIER_SIZE]; } nfs4_verifier;
 
-struct nfs_stateid4 {
-	__be32 seqid;
-	char other[NFS4_STATEID_OTHER_SIZE];
-} __attribute__ ((packed));
+struct nfs4_stateid_struct {
+	union {
+		char data[NFS4_STATEID_SIZE];
+		struct {
+			__be32 seqid;
+			char other[NFS4_STATEID_OTHER_SIZE];
+		} __attribute__ ((packed));
+	};
+
+	enum {
+		NFS4_INVALID_STATEID_TYPE = 0,
+		NFS4_SPECIAL_STATEID_TYPE,
+		NFS4_OPEN_STATEID_TYPE,
+		NFS4_LOCK_STATEID_TYPE,
+		NFS4_DELEGATION_STATEID_TYPE,
+		NFS4_LAYOUT_STATEID_TYPE,
+		NFS4_PNFS_DS_STATEID_TYPE,
+	} type;
+};
 
-typedef struct nfs_stateid4 nfs4_stateid;
+typedef struct nfs4_stateid_struct nfs4_stateid;
 
 enum nfs_opnum4 {
 	OP_ACCESS = 3,
@@ -504,6 +519,7 @@ enum {
 	NFSPROC4_CLNT_DEALLOCATE,
 	NFSPROC4_CLNT_LAYOUTSTATS,
 	NFSPROC4_CLNT_CLONE,
+	NFSPROC4_CLNT_COPY,
 };
 
 /* nfs41 types */
@@ -621,7 +637,9 @@ enum pnfs_update_layout_reason {
 	PNFS_UPDATE_LAYOUT_IO_TEST_FAIL,
 	PNFS_UPDATE_LAYOUT_FOUND_CACHED,
 	PNFS_UPDATE_LAYOUT_RETURN,
+	PNFS_UPDATE_LAYOUT_RETRY,
 	PNFS_UPDATE_LAYOUT_BLOCKED,
+	PNFS_UPDATE_LAYOUT_INVALID_OPEN,
 	PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET,
 };
 

include/linux/nfs_fs_sb.h

@@ -246,5 +246,6 @@ struct nfs_server {
 #define NFS_CAP_DEALLOCATE	(1U << 21)
 #define NFS_CAP_LAYOUTSTATS	(1U << 22)
 #define NFS_CAP_CLONE		(1U << 23)
+#define NFS_CAP_COPY		(1U << 24)
 
 #endif

include/linux/nfs_xdr.h

@@ -233,7 +233,6 @@ struct nfs4_layoutget_args {
 	struct inode *inode;
 	struct nfs_open_context *ctx;
 	nfs4_stateid stateid;
-	unsigned long timestamp;
 	struct nfs4_layoutdriver_data layout;
 };
 
@@ -251,7 +250,6 @@ struct nfs4_layoutget {
 	struct nfs4_layoutget_res res;
 	struct rpc_cred *cred;
 	gfp_t gfp_flags;
-	long timeout;
 };
 
 struct nfs4_getdeviceinfo_args {
@@ -1343,6 +1341,32 @@ struct nfs42_falloc_res {
 	const struct nfs_server		*falloc_server;
 };
 
+struct nfs42_copy_args {
+	struct nfs4_sequence_args	seq_args;
+
+	struct nfs_fh			*src_fh;
+	nfs4_stateid			src_stateid;
+	u64				src_pos;
+
+	struct nfs_fh			*dst_fh;
+	nfs4_stateid			dst_stateid;
+	u64				dst_pos;
+
+	u64				count;
+};
+
+struct nfs42_write_res {
+	u64			count;
+	struct nfs_writeverf	verifier;
+};
+
+struct nfs42_copy_res {
+	struct nfs4_sequence_res	seq_res;
+	struct nfs42_write_res		write_res;
+	bool				consecutive;
+	bool				synchronous;
+};
+
 struct nfs42_seek_args {
 	struct nfs4_sequence_args	seq_args;
 
@@ -1431,7 +1455,7 @@ struct nfs_commit_completion_ops {
 };
 
 struct nfs_commit_info {
-	spinlock_t			*lock;	/* inode->i_lock */
+	struct inode 			*inode;	/* Needed for inode->i_lock */
 	struct nfs_mds_commit_info	*mds;
 	struct pnfs_ds_commit_info	*ds;
 	struct nfs_direct_req		*dreq;	/* O_DIRECT request */

include/linux/sunrpc/auth.h

@@ -127,7 +127,7 @@ struct rpc_authops {
 	void			(*destroy)(struct rpc_auth *);
 
 	struct rpc_cred *	(*lookup_cred)(struct rpc_auth *, struct auth_cred *, int);
-	struct rpc_cred *	(*crcreate)(struct rpc_auth*, struct auth_cred *, int);
+	struct rpc_cred *	(*crcreate)(struct rpc_auth*, struct auth_cred *, int, gfp_t);
 	int			(*list_pseudoflavors)(rpc_authflavor_t *, int);
 	rpc_authflavor_t	(*info2flavor)(struct rpcsec_gss_info *);
 	int			(*flavor2info)(rpc_authflavor_t,
@@ -167,6 +167,7 @@ void 			rpc_destroy_authunix(void);
 
 struct rpc_cred *	rpc_lookup_cred(void);
 struct rpc_cred *	rpc_lookup_cred_nonblock(void);
+struct rpc_cred *	rpc_lookup_generic_cred(struct auth_cred *, int, gfp_t);
 struct rpc_cred *	rpc_lookup_machine_cred(const char *service_name);
 int			rpcauth_register(const struct rpc_authops *);
 int			rpcauth_unregister(const struct rpc_authops *);
@@ -178,7 +179,7 @@ rpc_authflavor_t	rpcauth_get_pseudoflavor(rpc_authflavor_t,
 int			rpcauth_get_gssinfo(rpc_authflavor_t,
 				struct rpcsec_gss_info *);
 int			rpcauth_list_flavors(rpc_authflavor_t *, int);
-struct rpc_cred *	rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int);
+struct rpc_cred *	rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int, gfp_t);
 void			rpcauth_init_cred(struct rpc_cred *, const struct auth_cred *, struct rpc_auth *, const struct rpc_credops *);
 struct rpc_cred *	rpcauth_lookupcred(struct rpc_auth *, int);
 struct rpc_cred *	rpcauth_generic_bind_cred(struct rpc_task *, struct rpc_cred *, int);
@@ -201,9 +202,28 @@ char *			rpcauth_stringify_acceptor(struct rpc_cred *);
 static inline
 struct rpc_cred *	get_rpccred(struct rpc_cred *cred)
 {
-	atomic_inc(&cred->cr_count);
+	if (cred != NULL)
+		atomic_inc(&cred->cr_count);
 	return cred;
 }
 
+/**
+ * get_rpccred_rcu - get a reference to a cred using rcu-protected pointer
+ * @cred: cred of which to take a reference
+ *
+ * In some cases, we may have a pointer to a credential to which we
+ * want to take a reference, but don't already have one. Because these
+ * objects are freed using RCU, we can access the cr_count while its
+ * on its way to destruction and only take a reference if it's not already
+ * zero.
+ */
+static inline struct rpc_cred *
+get_rpccred_rcu(struct rpc_cred *cred)
+{
+	if (atomic_inc_not_zero(&cred->cr_count))
+		return cred;
+	return NULL;
+}
+
 #endif /* __KERNEL__ */
 #endif /* _LINUX_SUNRPC_AUTH_H */

include/linux/sunrpc/clnt.h

@@ -176,6 +176,7 @@ void		rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int);
 int		rpc_protocol(struct rpc_clnt *);
 struct net *	rpc_net_ns(struct rpc_clnt *);
 size_t		rpc_max_payload(struct rpc_clnt *);
+size_t		rpc_max_bc_payload(struct rpc_clnt *);
 unsigned long	rpc_get_timeout(struct rpc_clnt *clnt);
 void		rpc_force_rebind(struct rpc_clnt *);
 size_t		rpc_peeraddr(struct rpc_clnt *, struct sockaddr *, size_t);

include/linux/sunrpc/msg_prot.h

@@ -158,9 +158,9 @@ typedef __be32	rpc_fraghdr;
 
 /*
  * Note that RFC 1833 does not put any size restrictions on the
- * netid string, but all currently defined netid's fit in 4 bytes.
+ * netid string, but all currently defined netid's fit in 5 bytes.
  */
-#define RPCBIND_MAXNETIDLEN	(4u)
+#define RPCBIND_MAXNETIDLEN	(5u)
 
 /*
  * Universal addresses are introduced in RFC 1833 and further spelled

include/linux/sunrpc/xprt.h

@@ -142,6 +142,7 @@ struct rpc_xprt_ops {
 	int		(*bc_setup)(struct rpc_xprt *xprt,
 				    unsigned int min_reqs);
 	int		(*bc_up)(struct svc_serv *serv, struct net *net);
+	size_t		(*bc_maxpayload)(struct rpc_xprt *xprt);
 	void		(*bc_free_rqst)(struct rpc_rqst *rqst);
 	void		(*bc_destroy)(struct rpc_xprt *xprt,
 				      unsigned int max_reqs);

include/linux/sunrpc/xprtrdma.h

@@ -52,7 +52,9 @@
 #define RPCRDMA_DEF_SLOT_TABLE	(128U)
 #define RPCRDMA_MAX_SLOT_TABLE	(256U)
 
-#define RPCRDMA_DEF_INLINE  (1024)	/* default inline max */
+#define RPCRDMA_MIN_INLINE  (1024)	/* min inline thresh */
+#define RPCRDMA_DEF_INLINE  (1024)	/* default inline thresh */
+#define RPCRDMA_MAX_INLINE  (3068)	/* max inline thresh */
 
 /* Memory registration strategies, by number.
  * This is part of a kernel / user space API. Do not remove. */

net/sunrpc/auth.c

@@ -543,7 +543,7 @@ rpcauth_cache_enforce_limit(void)
  */
 struct rpc_cred *
 rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred,
-		int flags)
+		int flags, gfp_t gfp)
 {
 	LIST_HEAD(free);
 	struct rpc_cred_cache *cache = auth->au_credcache;
@@ -580,7 +580,7 @@ rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred,
 	if (flags & RPCAUTH_LOOKUP_RCU)
 		return ERR_PTR(-ECHILD);
 
-	new = auth->au_ops->crcreate(auth, acred, flags);
+	new = auth->au_ops->crcreate(auth, acred, flags, gfp);
 	if (IS_ERR(new)) {
 		cred = new;
 		goto out;
@@ -703,8 +703,7 @@ rpcauth_bindcred(struct rpc_task *task, struct rpc_cred *cred, int flags)
 		new = rpcauth_bind_new_cred(task, lookupflags);
 	if (IS_ERR(new))
 		return PTR_ERR(new);
-	if (req->rq_cred != NULL)
-		put_rpccred(req->rq_cred);
+	put_rpccred(req->rq_cred);
 	req->rq_cred = new;
 	return 0;
 }
@@ -712,6 +711,8 @@ rpcauth_bindcred(struct rpc_task *task, struct rpc_cred *cred, int flags)
 void
 put_rpccred(struct rpc_cred *cred)
 {
+	if (cred == NULL)
+		return;
 	/* Fast path for unhashed credentials */
 	if (test_bit(RPCAUTH_CRED_HASHED, &cred->cr_flags) == 0) {
 		if (atomic_dec_and_test(&cred->cr_count))

net/sunrpc/auth_generic.c

@@ -38,6 +38,13 @@ struct rpc_cred *rpc_lookup_cred(void)
 }
 EXPORT_SYMBOL_GPL(rpc_lookup_cred);
 
+struct rpc_cred *
+rpc_lookup_generic_cred(struct auth_cred *acred, int flags, gfp_t gfp)
+{
+	return rpcauth_lookup_credcache(&generic_auth, acred, flags, gfp);
+}
+EXPORT_SYMBOL_GPL(rpc_lookup_generic_cred);
+
 struct rpc_cred *rpc_lookup_cred_nonblock(void)
 {
 	return rpcauth_lookupcred(&generic_auth, RPCAUTH_LOOKUP_RCU);
@@ -77,15 +84,15 @@ static struct rpc_cred *generic_bind_cred(struct rpc_task *task,
 static struct rpc_cred *
 generic_lookup_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
 {
-	return rpcauth_lookup_credcache(&generic_auth, acred, flags);
+	return rpcauth_lookup_credcache(&generic_auth, acred, flags, GFP_KERNEL);
 }
 
 static struct rpc_cred *
-generic_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
+generic_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags, gfp_t gfp)
 {
 	struct generic_cred *gcred;
 
-	gcred = kmalloc(sizeof(*gcred), GFP_KERNEL);
+	gcred = kmalloc(sizeof(*gcred), gfp);
 	if (gcred == NULL)
 		return ERR_PTR(-ENOMEM);
 

net/sunrpc/auth_gss/auth_gss.c

@@ -1299,11 +1299,11 @@ gss_destroy_cred(struct rpc_cred *cred)
 static struct rpc_cred *
 gss_lookup_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
 {
-	return rpcauth_lookup_credcache(auth, acred, flags);
+	return rpcauth_lookup_credcache(auth, acred, flags, GFP_NOFS);
 }
 
 static struct rpc_cred *
-gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
+gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags, gfp_t gfp)
 {
 	struct gss_auth *gss_auth = container_of(auth, struct gss_auth, rpc_auth);
 	struct gss_cred	*cred = NULL;
@@ -1313,7 +1313,7 @@ gss_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
 		__func__, from_kuid(&init_user_ns, acred->uid),
 		auth->au_flavor);
 
-	if (!(cred = kzalloc(sizeof(*cred), GFP_NOFS)))
+	if (!(cred = kzalloc(sizeof(*cred), gfp)))
 		goto out_err;
 
 	rpcauth_init_cred(&cred->gc_base, acred, auth, &gss_credops);

net/sunrpc/auth_unix.c

@@ -52,11 +52,11 @@ unx_destroy(struct rpc_auth *auth)
 static struct rpc_cred *
 unx_lookup_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
 {
-	return rpcauth_lookup_credcache(auth, acred, flags);
+	return rpcauth_lookup_credcache(auth, acred, flags, GFP_NOFS);
 }
 
 static struct rpc_cred *
-unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
+unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags, gfp_t gfp)
 {
 	struct unx_cred	*cred;
 	unsigned int groups = 0;
@@ -66,7 +66,7 @@ unx_create_cred(struct rpc_auth *auth, struct auth_cred *acred, int flags)
 			from_kuid(&init_user_ns, acred->uid),
 			from_kgid(&init_user_ns, acred->gid));
 
-	if (!(cred = kmalloc(sizeof(*cred), GFP_NOFS)))
+	if (!(cred = kmalloc(sizeof(*cred), gfp)))
 		return ERR_PTR(-ENOMEM);
 
 	rpcauth_init_cred(&cred->uc_base, acred, auth, &unix_credops);

net/sunrpc/clnt.c

@@ -1413,6 +1413,23 @@ size_t rpc_max_payload(struct rpc_clnt *clnt)
 }
 EXPORT_SYMBOL_GPL(rpc_max_payload);
 
+/**
+ * rpc_max_bc_payload - Get maximum backchannel payload size, in bytes
+ * @clnt: RPC client to query
+ */
+size_t rpc_max_bc_payload(struct rpc_clnt *clnt)
+{
+	struct rpc_xprt *xprt;
+	size_t ret;
+
+	rcu_read_lock();
+	xprt = rcu_dereference(clnt->cl_xprt);
+	ret = xprt->ops->bc_maxpayload(xprt);
+	rcu_read_unlock();
+	return ret;
+}
+EXPORT_SYMBOL_GPL(rpc_max_bc_payload);
+
 /**
  * rpc_get_timeout - Get timeout for transport in units of HZ
  * @clnt: RPC client to query

net/sunrpc/xdr.c

@@ -797,6 +797,8 @@ void xdr_init_decode(struct xdr_stream *xdr, struct xdr_buf *buf, __be32 *p)
 		xdr_set_iov(xdr, buf->head, buf->len);
 	else if (buf->page_len != 0)
 		xdr_set_page_base(xdr, 0, buf->len);
+	else
+		xdr_set_iov(xdr, buf->head, buf->len);
 	if (p != NULL && p > xdr->p && xdr->end >= p) {
 		xdr->nwords -= p - xdr->p;
 		xdr->p = p;

net/sunrpc/xprtrdma/backchannel.c

@@ -191,6 +191,22 @@ int xprt_rdma_bc_up(struct svc_serv *serv, struct net *net)
 	return 0;
 }
 
+/**
+ * xprt_rdma_bc_maxpayload - Return maximum backchannel message size
+ * @xprt: transport
+ *
+ * Returns maximum size, in bytes, of a backchannel message
+ */
+size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *xprt)
+{
+	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
+	struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
+	size_t maxmsg;
+
+	maxmsg = min_t(unsigned int, cdata->inline_rsize, cdata->inline_wsize);
+	return maxmsg - RPCRDMA_HDRLEN_MIN;
+}
+
 /**
  * rpcrdma_bc_marshal_reply - Send backwards direction reply
  * @rqst: buffer containing RPC reply data

net/sunrpc/xprtrdma/fmr_ops.c

@@ -35,10 +35,71 @@
 /* Maximum scatter/gather per FMR */
 #define RPCRDMA_MAX_FMR_SGES	(64)
 
+static struct workqueue_struct *fmr_recovery_wq;
+
+#define FMR_RECOVERY_WQ_FLAGS		(WQ_UNBOUND)
+
+int
+fmr_alloc_recovery_wq(void)
+{
+	fmr_recovery_wq = alloc_workqueue("fmr_recovery", WQ_UNBOUND, 0);
+	return !fmr_recovery_wq ? -ENOMEM : 0;
+}
+
+void
+fmr_destroy_recovery_wq(void)
+{
+	struct workqueue_struct *wq;
+
+	if (!fmr_recovery_wq)
+		return;
+
+	wq = fmr_recovery_wq;
+	fmr_recovery_wq = NULL;
+	destroy_workqueue(wq);
+}
+
+static int
+__fmr_unmap(struct rpcrdma_mw *mw)
+{
+	LIST_HEAD(l);
+
+	list_add(&mw->fmr.fmr->list, &l);
+	return ib_unmap_fmr(&l);
+}
+
+/* Deferred reset of a single FMR. Generate a fresh rkey by
+ * replacing the MR. There's no recovery if this fails.
+ */
+static void
+__fmr_recovery_worker(struct work_struct *work)
+{
+	struct rpcrdma_mw *mw = container_of(work, struct rpcrdma_mw,
+					    mw_work);
+	struct rpcrdma_xprt *r_xprt = mw->mw_xprt;
+
+	__fmr_unmap(mw);
+	rpcrdma_put_mw(r_xprt, mw);
+	return;
+}
+
+/* A broken MR was discovered in a context that can't sleep.
+ * Defer recovery to the recovery worker.
+ */
+static void
+__fmr_queue_recovery(struct rpcrdma_mw *mw)
+{
+	INIT_WORK(&mw->mw_work, __fmr_recovery_worker);
+	queue_work(fmr_recovery_wq, &mw->mw_work);
+}
+
 static int
 fmr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 	    struct rpcrdma_create_data_internal *cdata)
 {
+	rpcrdma_set_max_header_sizes(ia, cdata, max_t(unsigned int, 1,
+						      RPCRDMA_MAX_DATA_SEGS /
+						      RPCRDMA_MAX_FMR_SGES));
 	return 0;
 }
 
@@ -48,7 +109,7 @@ static size_t
 fmr_op_maxpages(struct rpcrdma_xprt *r_xprt)
 {
 	return min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS,
-		     rpcrdma_max_segments(r_xprt) * RPCRDMA_MAX_FMR_SGES);
+		     RPCRDMA_MAX_HDR_SEGS * RPCRDMA_MAX_FMR_SGES);
 }
 
 static int
@@ -89,6 +150,7 @@ fmr_op_init(struct rpcrdma_xprt *r_xprt)
 		if (IS_ERR(r->fmr.fmr))
 			goto out_fmr_err;
 
+		r->mw_xprt = r_xprt;
 		list_add(&r->mw_list, &buf->rb_mws);
 		list_add(&r->mw_all, &buf->rb_all);
 	}
@@ -104,15 +166,6 @@ fmr_op_init(struct rpcrdma_xprt *r_xprt)
 	return rc;
 }
 
-static int
-__fmr_unmap(struct rpcrdma_mw *r)
-{
-	LIST_HEAD(l);
-
-	list_add(&r->fmr.fmr->list, &l);
-	return ib_unmap_fmr(&l);
-}
-
 /* Use the ib_map_phys_fmr() verb to register a memory region
  * for remote access via RDMA READ or RDMA WRITE.
  */
@@ -183,15 +236,10 @@ static void
 __fmr_dma_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
 {
 	struct ib_device *device = r_xprt->rx_ia.ri_device;
-	struct rpcrdma_mw *mw = seg->rl_mw;
 	int nsegs = seg->mr_nsegs;
 
-	seg->rl_mw = NULL;
-
 	while (nsegs--)
 		rpcrdma_unmap_one(device, seg++);
-
-	rpcrdma_put_mw(r_xprt, mw);
 }
 
 /* Invalidate all memory regions that were registered for "req".
@@ -234,42 +282,50 @@ fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 		seg = &req->rl_segments[i];
 
 		__fmr_dma_unmap(r_xprt, seg);
+		rpcrdma_put_mw(r_xprt, seg->rl_mw);
 
 		i += seg->mr_nsegs;
 		seg->mr_nsegs = 0;
+		seg->rl_mw = NULL;
 	}
 
 	req->rl_nchunks = 0;
 }
 
-/* Use the ib_unmap_fmr() verb to prevent further remote
- * access via RDMA READ or RDMA WRITE.
+/* Use a slow, safe mechanism to invalidate all memory regions
+ * that were registered for "req".
+ *
+ * In the asynchronous case, DMA unmapping occurs first here
+ * because the rpcrdma_mr_seg is released immediately after this
+ * call. It's contents won't be available in __fmr_dma_unmap later.
+ * FIXME.
  */
-static int
-fmr_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
+static void
+fmr_op_unmap_safe(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
+		  bool sync)
 {
-	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
-	struct rpcrdma_mr_seg *seg1 = seg;
-	struct rpcrdma_mw *mw = seg1->rl_mw;
-	int rc, nsegs = seg->mr_nsegs;
+	struct rpcrdma_mr_seg *seg;
+	struct rpcrdma_mw *mw;
+	unsigned int i;
 
-	dprintk("RPC:       %s: FMR %p\n", __func__, mw);
+	for (i = 0; req->rl_nchunks; req->rl_nchunks--) {
+		seg = &req->rl_segments[i];
+		mw = seg->rl_mw;
 
-	seg1->rl_mw = NULL;
-	while (seg1->mr_nsegs--)
-		rpcrdma_unmap_one(ia->ri_device, seg++);
-	rc = __fmr_unmap(mw);
-	if (rc)
-		goto out_err;
-	rpcrdma_put_mw(r_xprt, mw);
-	return nsegs;
+		if (sync) {
+			/* ORDER */
+			__fmr_unmap(mw);
+			__fmr_dma_unmap(r_xprt, seg);
+			rpcrdma_put_mw(r_xprt, mw);
+		} else {
+			__fmr_dma_unmap(r_xprt, seg);
+			__fmr_queue_recovery(mw);
+		}
 
-out_err:
-	/* The FMR is abandoned, but remains in rb_all. fmr_op_destroy
-	 * will attempt to release it when the transport is destroyed.
-	 */
-	dprintk("RPC:       %s: ib_unmap_fmr status %i\n", __func__, rc);
-	return nsegs;
+		i += seg->mr_nsegs;
+		seg->mr_nsegs = 0;
+		seg->rl_mw = NULL;
+	}
 }
 
 static void
@@ -295,7 +351,7 @@ fmr_op_destroy(struct rpcrdma_buffer *buf)
 const struct rpcrdma_memreg_ops rpcrdma_fmr_memreg_ops = {
 	.ro_map				= fmr_op_map,
 	.ro_unmap_sync			= fmr_op_unmap_sync,
-	.ro_unmap			= fmr_op_unmap,
+	.ro_unmap_safe			= fmr_op_unmap_safe,
 	.ro_open			= fmr_op_open,
 	.ro_maxpages			= fmr_op_maxpages,
 	.ro_init			= fmr_op_init,

net/sunrpc/xprtrdma/physical_ops.c

@@ -36,8 +36,11 @@ physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 		       __func__, PTR_ERR(mr));
 		return -ENOMEM;
 	}
-
 	ia->ri_dma_mr = mr;
+
+	rpcrdma_set_max_header_sizes(ia, cdata, min_t(unsigned int,
+						      RPCRDMA_MAX_DATA_SEGS,
+						      RPCRDMA_MAX_HDR_SEGS));
 	return 0;
 }
 
@@ -47,7 +50,7 @@ static size_t
 physical_op_maxpages(struct rpcrdma_xprt *r_xprt)
 {
 	return min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS,
-		     rpcrdma_max_segments(r_xprt));
+		     RPCRDMA_MAX_HDR_SEGS);
 }
 
 static int
@@ -71,17 +74,6 @@ physical_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	return 1;
 }
 
-/* Unmap a memory region, but leave it registered.
- */
-static int
-physical_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
-{
-	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
-
-	rpcrdma_unmap_one(ia->ri_device, seg);
-	return 1;
-}
-
 /* DMA unmap all memory regions that were mapped for "req".
  */
 static void
@@ -94,6 +86,25 @@ physical_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 		rpcrdma_unmap_one(device, &req->rl_segments[i++]);
 }
 
+/* Use a slow, safe mechanism to invalidate all memory regions
+ * that were registered for "req".
+ *
+ * For physical memory registration, there is no good way to
+ * fence a single MR that has been advertised to the server. The
+ * client has already handed the server an R_key that cannot be
+ * invalidated and is shared by all MRs on this connection.
+ * Tearing down the PD might be the only safe choice, but it's
+ * not clear that a freshly acquired DMA R_key would be different
+ * than the one used by the PD that was just destroyed.
+ * FIXME.
+ */
+static void
+physical_op_unmap_safe(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
+		       bool sync)
+{
+	physical_op_unmap_sync(r_xprt, req);
+}
+
 static void
 physical_op_destroy(struct rpcrdma_buffer *buf)
 {
@@ -102,7 +113,7 @@ physical_op_destroy(struct rpcrdma_buffer *buf)
 const struct rpcrdma_memreg_ops rpcrdma_physical_memreg_ops = {
 	.ro_map				= physical_op_map,
 	.ro_unmap_sync			= physical_op_unmap_sync,
-	.ro_unmap			= physical_op_unmap,
+	.ro_unmap_safe			= physical_op_unmap_safe,
 	.ro_open			= physical_op_open,
 	.ro_maxpages			= physical_op_maxpages,
 	.ro_init			= physical_op_init,

net/sunrpc/xprtrdma/transport.c

@@ -73,6 +73,8 @@ static unsigned int xprt_rdma_memreg_strategy = RPCRDMA_FRMR;
 
 static unsigned int min_slot_table_size = RPCRDMA_MIN_SLOT_TABLE;
 static unsigned int max_slot_table_size = RPCRDMA_MAX_SLOT_TABLE;
+static unsigned int min_inline_size = RPCRDMA_MIN_INLINE;
+static unsigned int max_inline_size = RPCRDMA_MAX_INLINE;
 static unsigned int zero;
 static unsigned int max_padding = PAGE_SIZE;
 static unsigned int min_memreg = RPCRDMA_BOUNCEBUFFERS;
@@ -96,6 +98,8 @@ static struct ctl_table xr_tunables_table[] = {
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
+		.extra1		= &min_inline_size,
+		.extra2		= &max_inline_size,
 	},
 	{
 		.procname	= "rdma_max_inline_write",
@@ -103,6 +107,8 @@ static struct ctl_table xr_tunables_table[] = {
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
+		.extra1		= &min_inline_size,
+		.extra2		= &max_inline_size,
 	},
 	{
 		.procname	= "rdma_inline_write_padding",
@@ -508,6 +514,7 @@ xprt_rdma_allocate(struct rpc_task *task, size_t size)
 out:
 	dprintk("RPC:       %s: size %zd, request 0x%p\n", __func__, size, req);
 	req->rl_connect_cookie = 0;	/* our reserved value */
+	req->rl_task = task;
 	return req->rl_sendbuf->rg_base;
 
 out_rdmabuf:
@@ -564,7 +571,6 @@ xprt_rdma_free(void *buffer)
 	struct rpcrdma_req *req;
 	struct rpcrdma_xprt *r_xprt;
 	struct rpcrdma_regbuf *rb;
-	int i;
 
 	if (buffer == NULL)
 		return;
@@ -578,11 +584,8 @@ xprt_rdma_free(void *buffer)
 
 	dprintk("RPC:       %s: called on 0x%p\n", __func__, req->rl_reply);
 
-	for (i = 0; req->rl_nchunks;) {
-		--req->rl_nchunks;
-		i += r_xprt->rx_ia.ri_ops->ro_unmap(r_xprt,
-						    &req->rl_segments[i]);
-	}
+	r_xprt->rx_ia.ri_ops->ro_unmap_safe(r_xprt, req,
+					    !RPC_IS_ASYNC(req->rl_task));
 
 	rpcrdma_buffer_put(req);
 }
@@ -707,6 +710,7 @@ static struct rpc_xprt_ops xprt_rdma_procs = {
 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
 	.bc_setup		= xprt_rdma_bc_setup,
 	.bc_up			= xprt_rdma_bc_up,
+	.bc_maxpayload		= xprt_rdma_bc_maxpayload,
 	.bc_free_rqst		= xprt_rdma_bc_free_rqst,
 	.bc_destroy		= xprt_rdma_bc_destroy,
 #endif

net/sunrpc/xprtrdma/verbs.c

@@ -203,15 +203,6 @@ rpcrdma_receive_wc(struct ib_cq *cq, struct ib_wc *wc)
 	goto out_schedule;
 }
 
-static void
-rpcrdma_flush_cqs(struct rpcrdma_ep *ep)
-{
-	struct ib_wc wc;
-
-	while (ib_poll_cq(ep->rep_attr.recv_cq, 1, &wc) > 0)
-		rpcrdma_receive_wc(NULL, &wc);
-}
-
 static int
 rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 {
@@ -373,23 +364,6 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt,
 	return ERR_PTR(rc);
 }
 
-/*
- * Drain any cq, prior to teardown.
- */
-static void
-rpcrdma_clean_cq(struct ib_cq *cq)
-{
-	struct ib_wc wc;
-	int count = 0;
-
-	while (1 == ib_poll_cq(cq, 1, &wc))
-		++count;
-
-	if (count)
-		dprintk("RPC:       %s: flushed %d events (last 0x%x)\n",
-			__func__, count, wc.opcode);
-}
-
 /*
  * Exported functions.
  */
@@ -459,7 +433,6 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg)
 	dprintk("RPC:       %s: memory registration strategy is '%s'\n",
 		__func__, ia->ri_ops->ro_displayname);
 
-	rwlock_init(&ia->ri_qplock);
 	return 0;
 
 out3:
@@ -515,7 +488,7 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 			__func__);
 		return -ENOMEM;
 	}
-	max_qp_wr = ia->ri_device->attrs.max_qp_wr - RPCRDMA_BACKWARD_WRS;
+	max_qp_wr = ia->ri_device->attrs.max_qp_wr - RPCRDMA_BACKWARD_WRS - 1;
 
 	/* check provider's send/recv wr limits */
 	if (cdata->max_requests > max_qp_wr)
@@ -526,11 +499,13 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 	ep->rep_attr.srq = NULL;
 	ep->rep_attr.cap.max_send_wr = cdata->max_requests;
 	ep->rep_attr.cap.max_send_wr += RPCRDMA_BACKWARD_WRS;
+	ep->rep_attr.cap.max_send_wr += 1;	/* drain cqe */
 	rc = ia->ri_ops->ro_open(ia, ep, cdata);
 	if (rc)
 		return rc;
 	ep->rep_attr.cap.max_recv_wr = cdata->max_requests;
 	ep->rep_attr.cap.max_recv_wr += RPCRDMA_BACKWARD_WRS;
+	ep->rep_attr.cap.max_recv_wr += 1;	/* drain cqe */
 	ep->rep_attr.cap.max_send_sge = RPCRDMA_MAX_IOVS;
 	ep->rep_attr.cap.max_recv_sge = 1;
 	ep->rep_attr.cap.max_inline_data = 0;
@@ -578,6 +553,7 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 	ep->rep_attr.recv_cq = recvcq;
 
 	/* Initialize cma parameters */
+	memset(&ep->rep_remote_cma, 0, sizeof(ep->rep_remote_cma));
 
 	/* RPC/RDMA does not use private data */
 	ep->rep_remote_cma.private_data = NULL;
@@ -591,7 +567,16 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 		ep->rep_remote_cma.responder_resources =
 						ia->ri_device->attrs.max_qp_rd_atom;
 
-	ep->rep_remote_cma.retry_count = 7;
+	/* Limit transport retries so client can detect server
+	 * GID changes quickly. RPC layer handles re-establishing
+	 * transport connection and retransmission.
+	 */
+	ep->rep_remote_cma.retry_count = 6;
+
+	/* RPC-over-RDMA handles its own flow control. In addition,
+	 * make all RNR NAKs visible so we know that RPC-over-RDMA
+	 * flow control is working correctly (no NAKs should be seen).
+	 */
 	ep->rep_remote_cma.flow_control = 0;
 	ep->rep_remote_cma.rnr_retry_count = 0;
 
@@ -622,13 +607,8 @@ rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 
 	cancel_delayed_work_sync(&ep->rep_connect_worker);
 
-	if (ia->ri_id->qp)
-		rpcrdma_ep_disconnect(ep, ia);
-
-	rpcrdma_clean_cq(ep->rep_attr.recv_cq);
-	rpcrdma_clean_cq(ep->rep_attr.send_cq);
-
 	if (ia->ri_id->qp) {
+		rpcrdma_ep_disconnect(ep, ia);
 		rdma_destroy_qp(ia->ri_id);
 		ia->ri_id->qp = NULL;
 	}
@@ -659,7 +639,6 @@ rpcrdma_ep_connect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 		dprintk("RPC:       %s: reconnecting...\n", __func__);
 
 		rpcrdma_ep_disconnect(ep, ia);
-		rpcrdma_flush_cqs(ep);
 
 		xprt = container_of(ia, struct rpcrdma_xprt, rx_ia);
 		id = rpcrdma_create_id(xprt, ia,
@@ -692,10 +671,8 @@ rpcrdma_ep_connect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 			goto out;
 		}
 
-		write_lock(&ia->ri_qplock);
 		old = ia->ri_id;
 		ia->ri_id = id;
-		write_unlock(&ia->ri_qplock);
 
 		rdma_destroy_qp(old);
 		rpcrdma_destroy_id(old);
@@ -785,7 +762,6 @@ rpcrdma_ep_disconnect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 {
 	int rc;
 
-	rpcrdma_flush_cqs(ep);
 	rc = rdma_disconnect(ia->ri_id);
 	if (!rc) {
 		/* returns without wait if not connected */
@@ -797,6 +773,8 @@ rpcrdma_ep_disconnect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 		dprintk("RPC:       %s: rdma_disconnect %i\n", __func__, rc);
 		ep->rep_connected = rc;
 	}
+
+	ib_drain_qp(ia->ri_id->qp);
 }
 
 struct rpcrdma_req *
@@ -1271,25 +1249,3 @@ rpcrdma_ep_post_extra_recv(struct rpcrdma_xprt *r_xprt, unsigned int count)
 	rpcrdma_recv_buffer_put(rep);
 	return rc;
 }
-
-/* How many chunk list items fit within our inline buffers?
- */
-unsigned int
-rpcrdma_max_segments(struct rpcrdma_xprt *r_xprt)
-{
-	struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
-	int bytes, segments;
-
-	bytes = min_t(unsigned int, cdata->inline_wsize, cdata->inline_rsize);
-	bytes -= RPCRDMA_HDRLEN_MIN;
-	if (bytes < sizeof(struct rpcrdma_segment) * 2) {
-		pr_warn("RPC:       %s: inline threshold too small\n",
-			__func__);
-		return 0;
-	}
-
-	segments = 1 << (fls(bytes / sizeof(struct rpcrdma_segment)) - 1);
-	dprintk("RPC:       %s: max chunk list size = %d segments\n",
-		__func__, segments);
-	return segments;
-}

net/sunrpc/xprtrdma/xprt_rdma.h

@@ -65,7 +65,6 @@
  */
 struct rpcrdma_ia {
 	const struct rpcrdma_memreg_ops	*ri_ops;
-	rwlock_t		ri_qplock;
 	struct ib_device	*ri_device;
 	struct rdma_cm_id 	*ri_id;
 	struct ib_pd		*ri_pd;
@@ -73,6 +72,8 @@ struct rpcrdma_ia {
 	struct completion	ri_done;
 	int			ri_async_rc;
 	unsigned int		ri_max_frmr_depth;
+	unsigned int		ri_max_inline_write;
+	unsigned int		ri_max_inline_read;
 	struct ib_qp_attr	ri_qp_attr;
 	struct ib_qp_init_attr	ri_qp_init_attr;
 };
@@ -144,6 +145,26 @@ rdmab_to_msg(struct rpcrdma_regbuf *rb)
 
 #define RPCRDMA_DEF_GFP		(GFP_NOIO | __GFP_NOWARN)
 
+/* To ensure a transport can always make forward progress,
+ * the number of RDMA segments allowed in header chunk lists
+ * is capped at 8. This prevents less-capable devices and
+ * memory registrations from overrunning the Send buffer
+ * while building chunk lists.
+ *
+ * Elements of the Read list take up more room than the
+ * Write list or Reply chunk. 8 read segments means the Read
+ * list (or Write list or Reply chunk) cannot consume more
+ * than
+ *
+ * ((8 + 2) * read segment size) + 1 XDR words, or 244 bytes.
+ *
+ * And the fixed part of the header is another 24 bytes.
+ *
+ * The smallest inline threshold is 1024 bytes, ensuring that
+ * at least 750 bytes are available for RPC messages.
+ */
+#define RPCRDMA_MAX_HDR_SEGS	(8)
+
 /*
  * struct rpcrdma_rep -- this structure encapsulates state required to recv
  * and complete a reply, asychronously. It needs several pieces of
@@ -162,7 +183,9 @@ rdmab_to_msg(struct rpcrdma_regbuf *rb)
  */
 
 #define RPCRDMA_MAX_DATA_SEGS	((1 * 1024 * 1024) / PAGE_SIZE)
-#define RPCRDMA_MAX_SEGS 	(RPCRDMA_MAX_DATA_SEGS + 2) /* head+tail = 2 */
+
+/* data segments + head/tail for Call + head/tail for Reply */
+#define RPCRDMA_MAX_SEGS 	(RPCRDMA_MAX_DATA_SEGS + 4)
 
 struct rpcrdma_buffer;
 
@@ -198,14 +221,13 @@ enum rpcrdma_frmr_state {
 };
 
 struct rpcrdma_frmr {
-	struct scatterlist		*sg;
-	int				sg_nents;
+	struct scatterlist		*fr_sg;
+	int				fr_nents;
+	enum dma_data_direction		fr_dir;
 	struct ib_mr			*fr_mr;
 	struct ib_cqe			fr_cqe;
 	enum rpcrdma_frmr_state		fr_state;
 	struct completion		fr_linv_done;
-	struct work_struct		fr_work;
-	struct rpcrdma_xprt		*fr_xprt;
 	union {
 		struct ib_reg_wr	fr_regwr;
 		struct ib_send_wr	fr_invwr;
@@ -222,6 +244,8 @@ struct rpcrdma_mw {
 		struct rpcrdma_fmr	fmr;
 		struct rpcrdma_frmr	frmr;
 	};
+	struct work_struct	mw_work;
+	struct rpcrdma_xprt	*mw_xprt;
 	struct list_head	mw_list;
 	struct list_head	mw_all;
 };
@@ -270,12 +294,14 @@ struct rpcrdma_req {
 	unsigned int		rl_niovs;
 	unsigned int		rl_nchunks;
 	unsigned int		rl_connect_cookie;
+	struct rpc_task		*rl_task;
 	struct rpcrdma_buffer	*rl_buffer;
 	struct rpcrdma_rep	*rl_reply;/* holder for reply buffer */
 	struct ib_sge		rl_send_iov[RPCRDMA_MAX_IOVS];
 	struct rpcrdma_regbuf	*rl_rdmabuf;
 	struct rpcrdma_regbuf	*rl_sendbuf;
 	struct rpcrdma_mr_seg	rl_segments[RPCRDMA_MAX_SEGS];
+	struct rpcrdma_mr_seg	*rl_nextseg;
 
 	struct ib_cqe		rl_cqe;
 	struct list_head	rl_all;
@@ -372,8 +398,8 @@ struct rpcrdma_memreg_ops {
 				  struct rpcrdma_mr_seg *, int, bool);
 	void		(*ro_unmap_sync)(struct rpcrdma_xprt *,
 					 struct rpcrdma_req *);
-	int		(*ro_unmap)(struct rpcrdma_xprt *,
-				    struct rpcrdma_mr_seg *);
+	void		(*ro_unmap_safe)(struct rpcrdma_xprt *,
+					 struct rpcrdma_req *, bool);
 	int		(*ro_open)(struct rpcrdma_ia *,
 				   struct rpcrdma_ep *,
 				   struct rpcrdma_create_data_internal *);
@@ -456,7 +482,6 @@ struct rpcrdma_regbuf *rpcrdma_alloc_regbuf(struct rpcrdma_ia *,
 void rpcrdma_free_regbuf(struct rpcrdma_ia *,
 			 struct rpcrdma_regbuf *);
 
-unsigned int rpcrdma_max_segments(struct rpcrdma_xprt *);
 int rpcrdma_ep_post_extra_recv(struct rpcrdma_xprt *, unsigned int);
 
 int frwr_alloc_recovery_wq(void);
@@ -519,6 +544,9 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *);
  * RPC/RDMA protocol calls - xprtrdma/rpc_rdma.c
  */
 int rpcrdma_marshal_req(struct rpc_rqst *);
+void rpcrdma_set_max_header_sizes(struct rpcrdma_ia *,
+				  struct rpcrdma_create_data_internal *,
+				  unsigned int);
 
 /* RPC/RDMA module init - xprtrdma/transport.c
  */
@@ -534,6 +562,7 @@ void xprt_rdma_cleanup(void);
 #if defined(CONFIG_SUNRPC_BACKCHANNEL)
 int xprt_rdma_bc_setup(struct rpc_xprt *, unsigned int);
 int xprt_rdma_bc_up(struct svc_serv *, struct net *);
+size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *);
 int rpcrdma_bc_post_recv(struct rpcrdma_xprt *, unsigned int);
 void rpcrdma_bc_receive_call(struct rpcrdma_xprt *, struct rpcrdma_rep *);
 int rpcrdma_bc_marshal_reply(struct rpc_rqst *);

net/sunrpc/xprtsock.c

@@ -1364,6 +1364,11 @@ static int xs_tcp_bc_up(struct svc_serv *serv, struct net *net)
 		return ret;
 	return 0;
 }
+
+static size_t xs_tcp_bc_maxpayload(struct rpc_xprt *xprt)
+{
+	return PAGE_SIZE;
+}
 #else
 static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
 					struct xdr_skb_reader *desc)
@@ -2661,6 +2666,7 @@ static struct rpc_xprt_ops xs_tcp_ops = {
 #ifdef CONFIG_SUNRPC_BACKCHANNEL
 	.bc_setup		= xprt_setup_bc,
 	.bc_up			= xs_tcp_bc_up,
+	.bc_maxpayload		= xs_tcp_bc_maxpayload,
 	.bc_free_rqst		= xprt_free_bc_rqst,
 	.bc_destroy		= xprt_destroy_bc,
 #endif