Merge tag 'nfs-for-4.14-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client updates from Trond Myklebust:
 "Hightlights include:

  Stable bugfixes:
   - Fix mirror allocation in the writeback code to avoid a use after
     free
   - Fix the O_DSYNC writes to use the correct byte range
   - Fix 2 use after free issues in the I/O code

  Features:
   - Writeback fixes to split up the inode->i_lock in order to reduce
     contention
   - RPC client receive fixes to reduce the amount of time the
     xprt->transport_lock is held when receiving data from a socket into
     am XDR buffer.
   - Ditto fixes to reduce contention between call side users of the
     rdma rb_lock, and its use in rpcrdma_reply_handler.
   - Re-arrange rdma stats to reduce false cacheline sharing.
   - Various rdma cleanups and optimisations.
   - Refactor the NFSv4.1 exchange id code and clean up the code.
   - Const-ify all instances of struct rpc_xprt_ops

  Bugfixes:
   - Fix the NFSv2 'sec=' mount option.
   - NFSv4.1: don't use machine credentials for CLOSE when using
     'sec=sys'
   - Fix the NFSv3 GRANT callback when the port changes on the server.
   - Fix livelock issues with COMMIT
   - NFSv4: Use correct inode in _nfs4_opendata_to_nfs4_state() when
     doing and NFSv4.1 open by filehandle"

* tag 'nfs-for-4.14-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (69 commits)
  NFS: Count the bytes of skipped subrequests in nfs_lock_and_join_requests()
  NFS: Don't hold the group lock when calling nfs_release_request()
  NFS: Remove pnfs_generic_transfer_commit_list()
  NFS: nfs_lock_and_join_requests and nfs_scan_commit_list can deadlock
  NFS: Fix 2 use after free issues in the I/O code
  NFS: Sync the correct byte range during synchronous writes
  lockd: Delete an error message for a failed memory allocation in reclaimer()
  NFS: remove jiffies field from access cache
  NFS: flush data when locking a file to ensure cache coherence for mmap.
  SUNRPC: remove some dead code.
  NFS: don't expect errors from mempool_alloc().
  xprtrdma: Use xprt_pin_rqst in rpcrdma_reply_handler
  xprtrdma: Re-arrange struct rx_stats
  NFS: Fix NFSv2 security settings
  NFSv4.1: don't use machine credentials for CLOSE when using 'sec=sys'
  SUNRPC: ECONNREFUSED should cause a rebind.
  NFS: Remove unused parameter gfp_flags from nfs_pageio_init()
  NFSv4: Fix up mirror allocation
  SUNRPC: Add a separate spinlock to protect the RPC request receive list
  SUNRPC: Cleanup xs_tcp_read_common()
  ...

fs/lockd/clntlock.c

@@ -235,12 +235,8 @@ reclaimer(void *ptr)
 	struct net *net = host->net;
 
 	req = kmalloc(sizeof(*req), GFP_KERNEL);
-	if (!req) {
-		printk(KERN_ERR "lockd: reclaimer unable to alloc memory."
-				" Locks for %s won't be reclaimed!\n",
-				host->h_name);
+	if (!req)
 		return 0;
-	}
 
 	allow_signal(SIGKILL);
 

fs/nfs/callback_proc.c

@@ -51,7 +51,7 @@ __be32 nfs4_callback_getattr(void *argp, void *resp,
 		goto out_iput;
 	res->size = i_size_read(inode);
 	res->change_attr = delegation->change_attr;
-	if (nfsi->nrequests != 0)
+	if (nfs_have_writebacks(inode))
 		res->change_attr++;
 	res->ctime = inode->i_ctime;
 	res->mtime = inode->i_mtime;

fs/nfs/delegation.c

@@ -1089,7 +1089,7 @@ bool nfs4_delegation_flush_on_close(const struct inode *inode)
 	delegation = rcu_dereference(nfsi->delegation);
 	if (delegation == NULL || !(delegation->type & FMODE_WRITE))
 		goto out;
-	if (nfsi->nrequests < delegation->pagemod_limit)
+	if (atomic_long_read(&nfsi->nrequests) < delegation->pagemod_limit)
 		ret = false;
 out:
 	rcu_read_unlock();

fs/nfs/dir.c

@@ -2260,7 +2260,6 @@ static int nfs_access_get_cached(struct inode *inode, struct rpc_cred *cred, str
 		spin_lock(&inode->i_lock);
 		retry = false;
 	}
-	res->jiffies = cache->jiffies;
 	res->cred = cache->cred;
 	res->mask = cache->mask;
 	list_move_tail(&cache->lru, &nfsi->access_cache_entry_lru);
@@ -2296,7 +2295,6 @@ static int nfs_access_get_cached_rcu(struct inode *inode, struct rpc_cred *cred,
 		goto out;
 	if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_ACCESS))
 		goto out;
-	res->jiffies = cache->jiffies;
 	res->cred = cache->cred;
 	res->mask = cache->mask;
 	err = 0;
@@ -2344,7 +2342,6 @@ void nfs_access_add_cache(struct inode *inode, struct nfs_access_entry *set)
 	if (cache == NULL)
 		return;
 	RB_CLEAR_NODE(&cache->rb_node);
-	cache->jiffies = set->jiffies;
 	cache->cred = get_rpccred(set->cred);
 	cache->mask = set->mask;
 
@@ -2432,7 +2429,6 @@ static int nfs_do_access(struct inode *inode, struct rpc_cred *cred, int mask)
 	cache.mask = NFS_MAY_LOOKUP | NFS_MAY_EXECUTE
 		     | NFS_MAY_WRITE | NFS_MAY_READ;
 	cache.cred = cred;
-	cache.jiffies = jiffies;
 	status = NFS_PROTO(inode)->access(inode, &cache);
 	if (status != 0) {
 		if (status == -ESTALE) {

fs/nfs/direct.c

@@ -616,13 +616,13 @@ nfs_direct_write_scan_commit_list(struct inode *inode,
 				  struct list_head *list,
 				  struct nfs_commit_info *cinfo)
 {
-	spin_lock(&cinfo->inode->i_lock);
+	mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
 #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->inode->i_lock);
+	mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
 }
 
 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)

fs/nfs/file.c

@@ -631,11 +631,11 @@ ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from)
 	if (result <= 0)
 		goto out;
 
-	result = generic_write_sync(iocb, result);
-	if (result < 0)
-		goto out;
 	written = result;
 	iocb->ki_pos += written;
+	result = generic_write_sync(iocb, written);
+	if (result < 0)
+		goto out;
 
 	/* Return error values */
 	if (nfs_need_check_write(file, inode)) {
@@ -744,15 +744,18 @@ do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
 		goto out;
 
 	/*
-	 * Revalidate the cache if the server has time stamps granular
-	 * enough to detect subsecond changes.  Otherwise, clear the
-	 * cache to prevent missing any changes.
+	 * Invalidate cache to prevent missing any changes.  If
+	 * the file is mapped, clear the page cache as well so
+	 * those mappings will be loaded.
 	 *
 	 * This makes locking act as a cache coherency point.
 	 */
 	nfs_sync_mapping(filp->f_mapping);
-	if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
+	if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) {
 		nfs_zap_caches(inode);
+		if (mapping_mapped(filp->f_mapping))
+			nfs_revalidate_mapping(inode, filp->f_mapping);
+	}
 out:
 	return status;
 }

fs/nfs/inode.c

@@ -1285,7 +1285,6 @@ static bool nfs_file_has_buffered_writers(struct nfs_inode *nfsi)
 
 static unsigned long nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 {
-	struct nfs_inode *nfsi = NFS_I(inode);
 	unsigned long ret = 0;
 
 	if ((fattr->valid & NFS_ATTR_FATTR_PRECHANGE)
@@ -1315,7 +1314,7 @@ static unsigned long nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr
 	if ((fattr->valid & NFS_ATTR_FATTR_PRESIZE)
 			&& (fattr->valid & NFS_ATTR_FATTR_SIZE)
 			&& i_size_read(inode) == nfs_size_to_loff_t(fattr->pre_size)
-			&& nfsi->nrequests == 0) {
+			&& !nfs_have_writebacks(inode)) {
 		i_size_write(inode, nfs_size_to_loff_t(fattr->size));
 		ret |= NFS_INO_INVALID_ATTR;
 	}
@@ -1823,7 +1822,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 		if (new_isize != cur_isize) {
 			/* Do we perhaps have any outstanding writes, or has
 			 * the file grown beyond our last write? */
-			if (nfsi->nrequests == 0 || new_isize > cur_isize) {
+			if (!nfs_have_writebacks(inode) || new_isize > cur_isize) {
 				i_size_write(inode, new_isize);
 				if (!have_writers)
 					invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
@@ -2012,10 +2011,11 @@ static void init_once(void *foo)
 	INIT_LIST_HEAD(&nfsi->access_cache_entry_lru);
 	INIT_LIST_HEAD(&nfsi->access_cache_inode_lru);
 	INIT_LIST_HEAD(&nfsi->commit_info.list);
-	nfsi->nrequests = 0;
-	nfsi->commit_info.ncommit = 0;
+	atomic_long_set(&nfsi->nrequests, 0);
+	atomic_long_set(&nfsi->commit_info.ncommit, 0);
 	atomic_set(&nfsi->commit_info.rpcs_out, 0);
 	init_rwsem(&nfsi->rmdir_sem);
+	mutex_init(&nfsi->commit_mutex);
 	nfs4_init_once(nfsi);
 }
 

fs/nfs/internal.h

@@ -251,7 +251,6 @@ int nfs_iocounter_wait(struct nfs_lock_context *l_ctx);
 extern const struct nfs_pageio_ops nfs_pgio_rw_ops;
 struct nfs_pgio_header *nfs_pgio_header_alloc(const struct nfs_rw_ops *);
 void nfs_pgio_header_free(struct nfs_pgio_header *);
-void nfs_pgio_data_destroy(struct nfs_pgio_header *);
 int nfs_generic_pgio(struct nfs_pageio_descriptor *, struct nfs_pgio_header *);
 int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_header *hdr,
 		      struct rpc_cred *cred, const struct nfs_rpc_ops *rpc_ops,

fs/nfs/nfs4_fs.h

@@ -303,6 +303,17 @@ _nfs4_state_protect(struct nfs_client *clp, unsigned long sp4_mode,
 	struct rpc_cred *newcred = NULL;
 	rpc_authflavor_t flavor;
 
+	if (sp4_mode == NFS_SP4_MACH_CRED_CLEANUP ||
+	    sp4_mode == NFS_SP4_MACH_CRED_PNFS_CLEANUP) {
+		/* Using machine creds for cleanup operations
+		 * is only relevent if the client credentials
+		 * might expire. So don't bother for
+		 * RPC_AUTH_UNIX.  If file was only exported to
+		 * sec=sys, the PUTFH would fail anyway.
+		 */
+		if ((*clntp)->cl_auth->au_flavor == RPC_AUTH_UNIX)
+			return false;
+	}
 	if (test_bit(sp4_mode, &clp->cl_sp4_flags)) {
 		spin_lock(&clp->cl_lock);
 		if (clp->cl_machine_cred != NULL)

fs/nfs/pagelist.c

@@ -134,19 +134,14 @@ EXPORT_SYMBOL_GPL(nfs_async_iocounter_wait);
 /*
  * nfs_page_group_lock - lock the head of the page group
  * @req - request in group that is to be locked
- * @nonblock - if true don't block waiting for lock
  *
- * this lock must be held if modifying the page group list
+ * this lock must be held when traversing or modifying the page
+ * group list
  *
- * return 0 on success, < 0 on error: -EDELAY if nonblocking or the
- * result from wait_on_bit_lock
- *
- * NOTE: calling with nonblock=false should always have set the
- *       lock bit (see fs/buffer.c and other uses of wait_on_bit_lock
- *       with TASK_UNINTERRUPTIBLE), so there is no need to check the result.
+ * return 0 on success, < 0 on error
  */
 int
-nfs_page_group_lock(struct nfs_page *req, bool nonblock)
+nfs_page_group_lock(struct nfs_page *req)
 {
 	struct nfs_page *head = req->wb_head;
 
@@ -155,35 +150,10 @@ nfs_page_group_lock(struct nfs_page *req, bool nonblock)
 	if (!test_and_set_bit(PG_HEADLOCK, &head->wb_flags))
 		return 0;
 
-	if (!nonblock) {
-		set_bit(PG_CONTENDED1, &head->wb_flags);
-		smp_mb__after_atomic();
-		return wait_on_bit_lock(&head->wb_flags, PG_HEADLOCK,
-				TASK_UNINTERRUPTIBLE);
-	}
-
-	return -EAGAIN;
-}
-
-/*
- * nfs_page_group_lock_wait - wait for the lock to clear, but don't grab it
- * @req - a request in the group
- *
- * This is a blocking call to wait for the group lock to be cleared.
- */
-void
-nfs_page_group_lock_wait(struct nfs_page *req)
-{
-	struct nfs_page *head = req->wb_head;
-
-	WARN_ON_ONCE(head != head->wb_head);
-
-	if (!test_bit(PG_HEADLOCK, &head->wb_flags))
-		return;
 	set_bit(PG_CONTENDED1, &head->wb_flags);
 	smp_mb__after_atomic();
-	wait_on_bit(&head->wb_flags, PG_HEADLOCK,
-		TASK_UNINTERRUPTIBLE);
+	return wait_on_bit_lock(&head->wb_flags, PG_HEADLOCK,
+				TASK_UNINTERRUPTIBLE);
 }
 
 /*
@@ -246,7 +216,7 @@ bool nfs_page_group_sync_on_bit(struct nfs_page *req, unsigned int bit)
 {
 	bool ret;
 
-	nfs_page_group_lock(req, false);
+	nfs_page_group_lock(req);
 	ret = nfs_page_group_sync_on_bit_locked(req, bit);
 	nfs_page_group_unlock(req);
 
@@ -288,9 +258,7 @@ nfs_page_group_init(struct nfs_page *req, struct nfs_page *prev)
 			inode = page_file_mapping(req->wb_page)->host;
 			set_bit(PG_INODE_REF, &req->wb_flags);
 			kref_get(&req->wb_kref);
-			spin_lock(&inode->i_lock);
-			NFS_I(inode)->nrequests++;
-			spin_unlock(&inode->i_lock);
+			atomic_long_inc(&NFS_I(inode)->nrequests);
 		}
 	}
 }
@@ -306,14 +274,11 @@ static void
 nfs_page_group_destroy(struct kref *kref)
 {
 	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
+	struct nfs_page *head = req->wb_head;
 	struct nfs_page *tmp, *next;
 
-	/* subrequests must release the ref on the head request */
-	if (req->wb_head != req)
-		nfs_release_request(req->wb_head);
-
 	if (!nfs_page_group_sync_on_bit(req, PG_TEARDOWN))
-		return;
+		goto out;
 
 	tmp = req;
 	do {
@@ -324,6 +289,10 @@ nfs_page_group_destroy(struct kref *kref)
 		nfs_free_request(tmp);
 		tmp = next;
 	} while (tmp != req);
+out:
+	/* subrequests must release the ref on the head request */
+	if (head != req)
+		nfs_release_request(head);
 }
 
 /**
@@ -465,6 +434,7 @@ void nfs_release_request(struct nfs_page *req)
 {
 	kref_put(&req->wb_kref, nfs_page_group_destroy);
 }
+EXPORT_SYMBOL_GPL(nfs_release_request);
 
 /**
  * nfs_wait_on_request - Wait for a request to complete.
@@ -483,6 +453,7 @@ nfs_wait_on_request(struct nfs_page *req)
 	return wait_on_bit_io(&req->wb_flags, PG_BUSY,
 			      TASK_UNINTERRUPTIBLE);
 }
+EXPORT_SYMBOL_GPL(nfs_wait_on_request);
 
 /*
  * nfs_generic_pg_test - determine if requests can be coalesced
@@ -530,16 +501,6 @@ struct nfs_pgio_header *nfs_pgio_header_alloc(const struct nfs_rw_ops *ops)
 }
 EXPORT_SYMBOL_GPL(nfs_pgio_header_alloc);
 
-/*
- * nfs_pgio_header_free - Free a read or write header
- * @hdr: The header to free
- */
-void nfs_pgio_header_free(struct nfs_pgio_header *hdr)
-{
-	hdr->rw_ops->rw_free_header(hdr);
-}
-EXPORT_SYMBOL_GPL(nfs_pgio_header_free);
-
 /**
  * nfs_pgio_data_destroy - make @hdr suitable for reuse
  *
@@ -548,14 +509,24 @@ EXPORT_SYMBOL_GPL(nfs_pgio_header_free);
  *
  * @hdr: A header that has had nfs_generic_pgio called
  */
-void nfs_pgio_data_destroy(struct nfs_pgio_header *hdr)
+static void nfs_pgio_data_destroy(struct nfs_pgio_header *hdr)
 {
 	if (hdr->args.context)
 		put_nfs_open_context(hdr->args.context);
 	if (hdr->page_array.pagevec != hdr->page_array.page_array)
 		kfree(hdr->page_array.pagevec);
 }
-EXPORT_SYMBOL_GPL(nfs_pgio_data_destroy);
+
+/*
+ * nfs_pgio_header_free - Free a read or write header
+ * @hdr: The header to free
+ */
+void nfs_pgio_header_free(struct nfs_pgio_header *hdr)
+{
+	nfs_pgio_data_destroy(hdr);
+	hdr->rw_ops->rw_free_header(hdr);
+}
+EXPORT_SYMBOL_GPL(nfs_pgio_header_free);
 
 /**
  * nfs_pgio_rpcsetup - Set up arguments for a pageio call
@@ -669,7 +640,6 @@ EXPORT_SYMBOL_GPL(nfs_initiate_pgio);
 static void nfs_pgio_error(struct nfs_pgio_header *hdr)
 {
 	set_bit(NFS_IOHDR_REDO, &hdr->flags);
-	nfs_pgio_data_destroy(hdr);
 	hdr->completion_ops->completion(hdr);
 }
 
@@ -680,7 +650,6 @@ static void nfs_pgio_error(struct nfs_pgio_header *hdr)
 static void nfs_pgio_release(void *calldata)
 {
 	struct nfs_pgio_header *hdr = calldata;
-	nfs_pgio_data_destroy(hdr);
 	hdr->completion_ops->completion(hdr);
 }
 
@@ -711,12 +680,8 @@ void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
 		     const struct nfs_pgio_completion_ops *compl_ops,
 		     const struct nfs_rw_ops *rw_ops,
 		     size_t bsize,
-		     int io_flags,
-		     gfp_t gfp_flags)
+		     int io_flags)
 {
-	struct nfs_pgio_mirror *new;
-	int i;
-
 	desc->pg_moreio = 0;
 	desc->pg_inode = inode;
 	desc->pg_ops = pg_ops;
@@ -732,23 +697,10 @@ void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
 	desc->pg_mirror_count = 1;
 	desc->pg_mirror_idx = 0;
 
-	if (pg_ops->pg_get_mirror_count) {
-		/* until we have a request, we don't have an lseg and no
-		 * idea how many mirrors there will be */
-		new = kcalloc(NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX,
-			      sizeof(struct nfs_pgio_mirror), gfp_flags);
-		desc->pg_mirrors_dynamic = new;
-		desc->pg_mirrors = new;
-
-		for (i = 0; i < NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX; i++)
-			nfs_pageio_mirror_init(&desc->pg_mirrors[i], bsize);
-	} else {
-		desc->pg_mirrors_dynamic = NULL;
-		desc->pg_mirrors = desc->pg_mirrors_static;
-		nfs_pageio_mirror_init(&desc->pg_mirrors[0], bsize);
-	}
+	desc->pg_mirrors_dynamic = NULL;
+	desc->pg_mirrors = desc->pg_mirrors_static;
+	nfs_pageio_mirror_init(&desc->pg_mirrors[0], bsize);
 }
-EXPORT_SYMBOL_GPL(nfs_pageio_init);
 
 /**
  * nfs_pgio_result - Basic pageio error handling
@@ -865,32 +817,52 @@ static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
 	return ret;
 }
 
+static struct nfs_pgio_mirror *
+nfs_pageio_alloc_mirrors(struct nfs_pageio_descriptor *desc,
+		unsigned int mirror_count)
+{
+	struct nfs_pgio_mirror *ret;
+	unsigned int i;
+
+	kfree(desc->pg_mirrors_dynamic);
+	desc->pg_mirrors_dynamic = NULL;
+	if (mirror_count == 1)
+		return desc->pg_mirrors_static;
+	ret = kmalloc_array(mirror_count, sizeof(*ret), GFP_NOFS);
+	if (ret != NULL) {
+		for (i = 0; i < mirror_count; i++)
+			nfs_pageio_mirror_init(&ret[i], desc->pg_bsize);
+		desc->pg_mirrors_dynamic = ret;
+	}
+	return ret;
+}
+
 /*
  * nfs_pageio_setup_mirroring - determine if mirroring is to be used
  *				by calling the pg_get_mirror_count op
  */
-static int nfs_pageio_setup_mirroring(struct nfs_pageio_descriptor *pgio,
+static void nfs_pageio_setup_mirroring(struct nfs_pageio_descriptor *pgio,
 				       struct nfs_page *req)
 {
-	int mirror_count = 1;
-
-	if (!pgio->pg_ops->pg_get_mirror_count)
-		return 0;
+	unsigned int mirror_count = 1;
 
-	mirror_count = pgio->pg_ops->pg_get_mirror_count(pgio, req);
-
-	if (pgio->pg_error < 0)
-		return pgio->pg_error;
-
-	if (!mirror_count || mirror_count > NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX)
-		return -EINVAL;
+	if (pgio->pg_ops->pg_get_mirror_count)
+		mirror_count = pgio->pg_ops->pg_get_mirror_count(pgio, req);
+	if (mirror_count == pgio->pg_mirror_count || pgio->pg_error < 0)
+		return;
 
-	if (WARN_ON_ONCE(!pgio->pg_mirrors_dynamic))
-		return -EINVAL;
+	if (!mirror_count || mirror_count > NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX) {
+		pgio->pg_error = -EINVAL;
+		return;
+	}
 
+	pgio->pg_mirrors = nfs_pageio_alloc_mirrors(pgio, mirror_count);
+	if (pgio->pg_mirrors == NULL) {
+		pgio->pg_error = -ENOMEM;
+		pgio->pg_mirrors = pgio->pg_mirrors_static;
+		mirror_count = 1;
+	}
 	pgio->pg_mirror_count = mirror_count;
-
-	return 0;
 }
 
 /*
@@ -1036,7 +1008,7 @@ static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 	unsigned int bytes_left = 0;
 	unsigned int offset, pgbase;
 
-	nfs_page_group_lock(req, false);
+	nfs_page_group_lock(req);
 
 	subreq = req;
 	bytes_left = subreq->wb_bytes;
@@ -1058,7 +1030,7 @@ static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 			if (mirror->pg_recoalesce)
 				return 0;
 			/* retry add_request for this subreq */
-			nfs_page_group_lock(req, false);
+			nfs_page_group_lock(req);
 			continue;
 		}
 
@@ -1155,7 +1127,7 @@ int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 
 	for (midx = 0; midx < desc->pg_mirror_count; midx++) {
 		if (midx) {
-			nfs_page_group_lock(req, false);
+			nfs_page_group_lock(req);
 
 			/* find the last request */
 			for (lastreq = req->wb_head;

fs/nfs/pnfs.c

@@ -529,47 +529,6 @@ pnfs_put_lseg(struct pnfs_layout_segment *lseg)
 }
 EXPORT_SYMBOL_GPL(pnfs_put_lseg);
 
-static void pnfs_free_lseg_async_work(struct work_struct *work)
-{
-	struct pnfs_layout_segment *lseg;
-	struct pnfs_layout_hdr *lo;
-
-	lseg = container_of(work, struct pnfs_layout_segment, pls_work);
-	lo = lseg->pls_layout;
-
-	pnfs_free_lseg(lseg);
-	pnfs_put_layout_hdr(lo);
-}
-
-static void pnfs_free_lseg_async(struct pnfs_layout_segment *lseg)
-{
-	INIT_WORK(&lseg->pls_work, pnfs_free_lseg_async_work);
-	schedule_work(&lseg->pls_work);
-}
-
-void
-pnfs_put_lseg_locked(struct pnfs_layout_segment *lseg)
-{
-	if (!lseg)
-		return;
-
-	assert_spin_locked(&lseg->pls_layout->plh_inode->i_lock);
-
-	dprintk("%s: lseg %p ref %d valid %d\n", __func__, lseg,
-		atomic_read(&lseg->pls_refcount),
-		test_bit(NFS_LSEG_VALID, &lseg->pls_flags));
-	if (atomic_dec_and_test(&lseg->pls_refcount)) {
-		struct pnfs_layout_hdr *lo = lseg->pls_layout;
-		if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags))
-			return;
-		pnfs_layout_remove_lseg(lo, lseg);
-		if (!pnfs_cache_lseg_for_layoutreturn(lo, lseg)) {
-			pnfs_get_layout_hdr(lo);
-			pnfs_free_lseg_async(lseg);
-		}
-	}
-}
-
 /*
  * is l2 fully contained in l1?
  *   start1                             end1
@@ -2274,7 +2233,6 @@ pnfs_write_through_mds(struct nfs_pageio_descriptor *desc,
 		nfs_pageio_reset_write_mds(desc);
 		mirror->pg_recoalesce = 1;
 	}
-	nfs_pgio_data_destroy(hdr);
 	hdr->release(hdr);
 }
 
@@ -2398,7 +2356,6 @@ pnfs_read_through_mds(struct nfs_pageio_descriptor *desc,
 		nfs_pageio_reset_read_mds(desc);
 		mirror->pg_recoalesce = 1;
 	}
-	nfs_pgio_data_destroy(hdr);
 	hdr->release(hdr);
 }
 

fs/nfs/pnfs.h

@@ -67,7 +67,6 @@ struct pnfs_layout_segment {
 	u32 pls_seq;
 	unsigned long pls_flags;
 	struct pnfs_layout_hdr *pls_layout;
-	struct work_struct pls_work;
 };
 
 enum pnfs_try_status {
@@ -230,7 +229,6 @@ extern int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync);
 /* pnfs.c */
 void pnfs_get_layout_hdr(struct pnfs_layout_hdr *lo);
 void pnfs_put_lseg(struct pnfs_layout_segment *lseg);
-void pnfs_put_lseg_locked(struct pnfs_layout_segment *lseg);
 
 void set_pnfs_layoutdriver(struct nfs_server *, const struct nfs_fh *, struct nfs_fsinfo *);
 void unset_pnfs_layoutdriver(struct nfs_server *);

fs/nfs/pnfs_nfs.c

@@ -83,33 +83,10 @@ pnfs_generic_clear_request_commit(struct nfs_page *req,
 	}
 out:
 	nfs_request_remove_commit_list(req, cinfo);
-	pnfs_put_lseg_locked(freeme);
+	pnfs_put_lseg(freeme);
 }
 EXPORT_SYMBOL_GPL(pnfs_generic_clear_request_commit);
 
-static int
-pnfs_generic_transfer_commit_list(struct list_head *src, struct list_head *dst,
-				  struct nfs_commit_info *cinfo, int max)
-{
-	struct nfs_page *req, *tmp;
-	int ret = 0;
-
-	list_for_each_entry_safe(req, tmp, src, wb_list) {
-		if (!nfs_lock_request(req))
-			continue;
-		kref_get(&req->wb_kref);
-		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);
-		nfs_list_add_request(req, dst);
-		ret++;
-		if ((ret == max) && !cinfo->dreq)
-			break;
-	}
-	return ret;
-}
-
 static int
 pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket,
 				 struct nfs_commit_info *cinfo,
@@ -119,15 +96,15 @@ pnfs_generic_scan_ds_commit_list(struct pnfs_commit_bucket *bucket,
 	struct list_head *dst = &bucket->committing;
 	int ret;
 
-	lockdep_assert_held(&cinfo->inode->i_lock);
-	ret = pnfs_generic_transfer_commit_list(src, dst, cinfo, max);
+	lockdep_assert_held(&NFS_I(cinfo->inode)->commit_mutex);
+	ret = nfs_scan_commit_list(src, dst, cinfo, max);
 	if (ret) {
 		cinfo->ds->nwritten -= ret;
 		cinfo->ds->ncommitting += ret;
 		if (bucket->clseg == NULL)
 			bucket->clseg = pnfs_get_lseg(bucket->wlseg);
 		if (list_empty(src)) {
-			pnfs_put_lseg_locked(bucket->wlseg);
+			pnfs_put_lseg(bucket->wlseg);
 			bucket->wlseg = NULL;
 		}
 	}
@@ -142,7 +119,7 @@ int pnfs_generic_scan_commit_lists(struct nfs_commit_info *cinfo,
 {
 	int i, rv = 0, cnt;
 
-	lockdep_assert_held(&cinfo->inode->i_lock);
+	lockdep_assert_held(&NFS_I(cinfo->inode)->commit_mutex);
 	for (i = 0; i < cinfo->ds->nbuckets && max != 0; i++) {
 		cnt = pnfs_generic_scan_ds_commit_list(&cinfo->ds->buckets[i],
 						       cinfo, max);
@@ -162,11 +139,10 @@ void pnfs_generic_recover_commit_reqs(struct list_head *dst,
 	int nwritten;
 	int i;
 
-	lockdep_assert_held(&cinfo->inode->i_lock);
+	lockdep_assert_held(&NFS_I(cinfo->inode)->commit_mutex);
 restart:
 	for (i = 0, b = cinfo->ds->buckets; i < cinfo->ds->nbuckets; i++, b++) {
-		nwritten = pnfs_generic_transfer_commit_list(&b->written,
-				dst, cinfo, 0);
+		nwritten = nfs_scan_commit_list(&b->written, dst, cinfo, 0);
 		if (!nwritten)
 			continue;
 		cinfo->ds->nwritten -= nwritten;
@@ -953,12 +929,12 @@ pnfs_layout_mark_request_commit(struct nfs_page *req,
 	struct list_head *list;
 	struct pnfs_commit_bucket *buckets;
 
-	spin_lock(&cinfo->inode->i_lock);
+	mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
 	buckets = cinfo->ds->buckets;
 	list = &buckets[ds_commit_idx].written;
 	if (list_empty(list)) {
 		if (!pnfs_is_valid_lseg(lseg)) {
-			spin_unlock(&cinfo->inode->i_lock);
+			mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
 			cinfo->completion_ops->resched_write(cinfo, req);
 			return;
 		}
@@ -975,7 +951,7 @@ pnfs_layout_mark_request_commit(struct nfs_page *req,
 	cinfo->ds->nwritten++;
 
 	nfs_request_add_commit_list_locked(req, list, cinfo);
-	spin_unlock(&cinfo->inode->i_lock);
+	mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
 	nfs_mark_page_unstable(req->wb_page, cinfo);
 }
 EXPORT_SYMBOL_GPL(pnfs_layout_mark_request_commit);

fs/nfs/read.c

@@ -68,7 +68,7 @@ void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio,
 		pg_ops = server->pnfs_curr_ld->pg_read_ops;
 #endif
 	nfs_pageio_init(pgio, inode, pg_ops, compl_ops, &nfs_rw_read_ops,
-			server->rsize, 0, GFP_KERNEL);
+			server->rsize, 0);
 }
 EXPORT_SYMBOL_GPL(nfs_pageio_init_read);
 

fs/nfs/super.c

@@ -1691,8 +1691,8 @@ static int nfs_verify_authflavors(struct nfs_parsed_mount_data *args,
 			rpc_authflavor_t *server_authlist, unsigned int count)
 {
 	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
+	bool found_auth_null = false;
 	unsigned int i;
-	int use_auth_null = false;
 
 	/*
 	 * If the sec= mount option is used, the specified flavor or AUTH_NULL
@@ -1701,6 +1701,10 @@ static int nfs_verify_authflavors(struct nfs_parsed_mount_data *args,
 	 * AUTH_NULL has a special meaning when it's in the server list - it
 	 * means that the server will ignore the rpc creds, so any flavor
 	 * can be used but still use the sec= that was specified.
+	 *
+	 * Note also that the MNT procedure in MNTv1 does not return a list
+	 * of supported security flavors. In this case, nfs_mount() fabricates
+	 * a security flavor list containing just AUTH_NULL.
 	 */
 	for (i = 0; i < count; i++) {
 		flavor = server_authlist[i];
@@ -1709,11 +1713,11 @@ static int nfs_verify_authflavors(struct nfs_parsed_mount_data *args,
 			goto out;
 
 		if (flavor == RPC_AUTH_NULL)
-			use_auth_null = true;
+			found_auth_null = true;
 	}
 
-	if (use_auth_null) {
-		flavor = RPC_AUTH_NULL;
+	if (found_auth_null) {
+		flavor = args->auth_info.flavors[0];
 		goto out;
 	}
 

include/linux/nfs_fs.h

@@ -49,7 +49,6 @@
 struct nfs_access_entry {
 	struct rb_node		rb_node;
 	struct list_head	lru;
-	unsigned long		jiffies;
 	struct rpc_cred *	cred;
 	__u32			mask;
 	struct rcu_head		rcu_head;
@@ -154,7 +153,7 @@ struct nfs_inode {
 	 */
 	__be32			cookieverf[2];
 
-	unsigned long		nrequests;
+	atomic_long_t		nrequests;
 	struct nfs_mds_commit_info commit_info;
 
 	/* Open contexts for shared mmap writes */
@@ -163,6 +162,7 @@ struct nfs_inode {
 	/* Readers: in-flight sillydelete RPC calls */
 	/* Writers: rmdir */
 	struct rw_semaphore	rmdir_sem;
+	struct mutex		commit_mutex;
 
 #if IS_ENABLED(CONFIG_NFS_V4)
 	struct nfs4_cached_acl	*nfs4_acl;
@@ -510,7 +510,7 @@ extern void nfs_commit_free(struct nfs_commit_data *data);
 static inline int
 nfs_have_writebacks(struct inode *inode)
 {
-	return NFS_I(inode)->nrequests != 0;
+	return atomic_long_read(&NFS_I(inode)->nrequests) != 0;
 }
 
 /*

include/linux/nfs_page.h

@@ -125,8 +125,7 @@ extern	void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
 			     const struct nfs_pgio_completion_ops *compl_ops,
 			     const struct nfs_rw_ops *rw_ops,
 			     size_t bsize,
-			     int how,
-			     gfp_t gfp_flags);
+			     int how);
 extern	int nfs_pageio_add_request(struct nfs_pageio_descriptor *,
 				   struct nfs_page *);
 extern  int nfs_pageio_resend(struct nfs_pageio_descriptor *,
@@ -139,8 +138,7 @@ extern size_t nfs_generic_pg_test(struct nfs_pageio_descriptor *desc,
 extern  int nfs_wait_on_request(struct nfs_page *);
 extern	void nfs_unlock_request(struct nfs_page *req);
 extern	void nfs_unlock_and_release_request(struct nfs_page *);
-extern int nfs_page_group_lock(struct nfs_page *, bool);
-extern void nfs_page_group_lock_wait(struct nfs_page *);
+extern int nfs_page_group_lock(struct nfs_page *);
 extern void nfs_page_group_unlock(struct nfs_page *);
 extern bool nfs_page_group_sync_on_bit(struct nfs_page *, unsigned int);
 extern bool nfs_async_iocounter_wait(struct rpc_task *, struct nfs_lock_context *);

include/linux/nfs_xdr.h

@@ -1476,7 +1476,7 @@ struct nfs_pgio_header {
 
 struct nfs_mds_commit_info {
 	atomic_t rpcs_out;
-	unsigned long		ncommit;
+	atomic_long_t		ncommit;
 	struct list_head	list;
 };
 

include/linux/sunrpc/sched.h

@@ -139,6 +139,8 @@ struct rpc_task_setup {
 #define RPC_TASK_RUNNING	0
 #define RPC_TASK_QUEUED		1
 #define RPC_TASK_ACTIVE		2
+#define RPC_TASK_MSG_RECV	3
+#define RPC_TASK_MSG_RECV_WAIT	4
 
 #define RPC_IS_RUNNING(t)	test_bit(RPC_TASK_RUNNING, &(t)->tk_runstate)
 #define rpc_set_running(t)	set_bit(RPC_TASK_RUNNING, &(t)->tk_runstate)

include/linux/sunrpc/xdr.h

@@ -239,6 +239,19 @@ extern unsigned int xdr_read_pages(struct xdr_stream *xdr, unsigned int len);
 extern void xdr_enter_page(struct xdr_stream *xdr, unsigned int len);
 extern int xdr_process_buf(struct xdr_buf *buf, unsigned int offset, unsigned int len, int (*actor)(struct scatterlist *, void *), void *data);
 
+/**
+ * xdr_stream_remaining - Return the number of bytes remaining in the stream
+ * @xdr: pointer to struct xdr_stream
+ *
+ * Return value:
+ *   Number of bytes remaining in @xdr before xdr->end
+ */
+static inline size_t
+xdr_stream_remaining(const struct xdr_stream *xdr)
+{
+	return xdr->nwords << 2;
+}
+
 ssize_t xdr_stream_decode_string_dup(struct xdr_stream *xdr, char **str,
 		size_t maxlen, gfp_t gfp_flags);
 /**

include/linux/sunrpc/xprt.h

@@ -174,7 +174,7 @@ enum xprt_transports {
 
 struct rpc_xprt {
 	struct kref		kref;		/* Reference count */
-	struct rpc_xprt_ops *	ops;		/* transport methods */
+	const struct rpc_xprt_ops *ops;		/* transport methods */
 
 	const struct rpc_timeout *timeout;	/* timeout parms */
 	struct sockaddr_storage	addr;		/* server address */
@@ -232,6 +232,7 @@ struct rpc_xprt {
 	 */
 	spinlock_t		transport_lock;	/* lock transport info */
 	spinlock_t		reserve_lock;	/* lock slot table */
+	spinlock_t		recv_lock;	/* lock receive list */
 	u32			xid;		/* Next XID value to use */
 	struct rpc_task *	snd_task;	/* Task blocked in send */
 	struct svc_xprt		*bc_xprt;	/* NFSv4.1 backchannel */
@@ -372,6 +373,8 @@ void			xprt_write_space(struct rpc_xprt *xprt);
 void			xprt_adjust_cwnd(struct rpc_xprt *xprt, struct rpc_task *task, int result);
 struct rpc_rqst *	xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid);
 void			xprt_complete_rqst(struct rpc_task *task, int copied);
+void			xprt_pin_rqst(struct rpc_rqst *req);
+void			xprt_unpin_rqst(struct rpc_rqst *req);
 void			xprt_release_rqst_cong(struct rpc_task *task);
 void			xprt_disconnect_done(struct rpc_xprt *xprt);
 void			xprt_force_disconnect(struct rpc_xprt *xprt);

net/sunrpc/backchannel_rqst.c

@@ -171,10 +171,10 @@ int xprt_setup_bc(struct rpc_xprt *xprt, unsigned int min_reqs)
 	/*
 	 * Add the temporary list to the backchannel preallocation list
 	 */
-	spin_lock_bh(&xprt->bc_pa_lock);
+	spin_lock(&xprt->bc_pa_lock);
 	list_splice(&tmp_list, &xprt->bc_pa_list);
 	xprt_inc_alloc_count(xprt, min_reqs);
-	spin_unlock_bh(&xprt->bc_pa_lock);
+	spin_unlock(&xprt->bc_pa_lock);
 
 	dprintk("RPC:       setup backchannel transport done\n");
 	return 0;

net/sunrpc/clnt.c

@@ -1903,6 +1903,14 @@ call_connect_status(struct rpc_task *task)
 	task->tk_status = 0;
 	switch (status) {
 	case -ECONNREFUSED:
+		/* A positive refusal suggests a rebind is needed. */
+		if (RPC_IS_SOFTCONN(task))
+			break;
+		if (clnt->cl_autobind) {
+			rpc_force_rebind(clnt);
+			task->tk_action = call_bind;
+			return;
+		}
 	case -ECONNRESET:
 	case -ECONNABORTED:
 	case -ENETUNREACH:
@@ -2139,10 +2147,6 @@ call_status(struct rpc_task *task)
 		rpc_delay(task, 3*HZ);
 	case -ETIMEDOUT:
 		task->tk_action = call_timeout;
-		if (!(task->tk_flags & RPC_TASK_NO_RETRANS_TIMEOUT)
-		    && task->tk_client->cl_discrtry)
-			xprt_conditional_disconnect(req->rq_xprt,
-					req->rq_connect_cookie);
 		break;
 	case -ECONNREFUSED:
 	case -ECONNRESET:

net/sunrpc/svcsock.c

@@ -1013,7 +1013,7 @@ static int receive_cb_reply(struct svc_sock *svsk, struct svc_rqst *rqstp)
 
 	if (!bc_xprt)
 		return -EAGAIN;
-	spin_lock_bh(&bc_xprt->transport_lock);
+	spin_lock(&bc_xprt->recv_lock);
 	req = xprt_lookup_rqst(bc_xprt, xid);
 	if (!req)
 		goto unlock_notfound;
@@ -1031,7 +1031,7 @@ static int receive_cb_reply(struct svc_sock *svsk, struct svc_rqst *rqstp)
 	memcpy(dst->iov_base, src->iov_base, src->iov_len);
 	xprt_complete_rqst(req->rq_task, rqstp->rq_arg.len);
 	rqstp->rq_arg.len = 0;
-	spin_unlock_bh(&bc_xprt->transport_lock);
+	spin_unlock(&bc_xprt->recv_lock);
 	return 0;
 unlock_notfound:
 	printk(KERN_NOTICE
@@ -1040,7 +1040,7 @@ static int receive_cb_reply(struct svc_sock *svsk, struct svc_rqst *rqstp)
 		__func__, ntohl(calldir),
 		bc_xprt, ntohl(xid));
 unlock_eagain:
-	spin_unlock_bh(&bc_xprt->transport_lock);
+	spin_unlock(&bc_xprt->recv_lock);
 	return -EAGAIN;
 }
 

net/sunrpc/xprt.c

@@ -844,6 +844,50 @@ struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
 }
 EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
 
+/**
+ * xprt_pin_rqst - Pin a request on the transport receive list
+ * @req: Request to pin
+ *
+ * Caller must ensure this is atomic with the call to xprt_lookup_rqst()
+ * so should be holding the xprt transport lock.
+ */
+void xprt_pin_rqst(struct rpc_rqst *req)
+{
+	set_bit(RPC_TASK_MSG_RECV, &req->rq_task->tk_runstate);
+}
+EXPORT_SYMBOL_GPL(xprt_pin_rqst);
+
+/**
+ * xprt_unpin_rqst - Unpin a request on the transport receive list
+ * @req: Request to pin
+ *
+ * Caller should be holding the xprt transport lock.
+ */
+void xprt_unpin_rqst(struct rpc_rqst *req)
+{
+	struct rpc_task *task = req->rq_task;
+
+	clear_bit(RPC_TASK_MSG_RECV, &task->tk_runstate);
+	if (test_bit(RPC_TASK_MSG_RECV_WAIT, &task->tk_runstate))
+		wake_up_bit(&task->tk_runstate, RPC_TASK_MSG_RECV);
+}
+EXPORT_SYMBOL_GPL(xprt_unpin_rqst);
+
+static void xprt_wait_on_pinned_rqst(struct rpc_rqst *req)
+__must_hold(&req->rq_xprt->recv_lock)
+{
+	struct rpc_task *task = req->rq_task;
+	
+	if (task && test_bit(RPC_TASK_MSG_RECV, &task->tk_runstate)) {
+		spin_unlock(&req->rq_xprt->recv_lock);
+		set_bit(RPC_TASK_MSG_RECV_WAIT, &task->tk_runstate);
+		wait_on_bit(&task->tk_runstate, RPC_TASK_MSG_RECV,
+				TASK_UNINTERRUPTIBLE);
+		clear_bit(RPC_TASK_MSG_RECV_WAIT, &task->tk_runstate);
+		spin_lock(&req->rq_xprt->recv_lock);
+	}
+}
+
 static void xprt_update_rtt(struct rpc_task *task)
 {
 	struct rpc_rqst *req = task->tk_rqstp;
@@ -966,13 +1010,13 @@ void xprt_transmit(struct rpc_task *task)
 			/*
 			 * Add to the list only if we're expecting a reply
 			 */
-			spin_lock_bh(&xprt->transport_lock);
 			/* Update the softirq receive buffer */
 			memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
 					sizeof(req->rq_private_buf));
 			/* Add request to the receive list */
+			spin_lock(&xprt->recv_lock);
 			list_add_tail(&req->rq_list, &xprt->recv);
-			spin_unlock_bh(&xprt->transport_lock);
+			spin_unlock(&xprt->recv_lock);
 			xprt_reset_majortimeo(req);
 			/* Turn off autodisconnect */
 			del_singleshot_timer_sync(&xprt->timer);
@@ -1287,12 +1331,16 @@ void xprt_release(struct rpc_task *task)
 		task->tk_ops->rpc_count_stats(task, task->tk_calldata);
 	else if (task->tk_client)
 		rpc_count_iostats(task, task->tk_client->cl_metrics);
+	spin_lock(&xprt->recv_lock);
+	if (!list_empty(&req->rq_list)) {
+		list_del(&req->rq_list);
+		xprt_wait_on_pinned_rqst(req);
+	}
+	spin_unlock(&xprt->recv_lock);
 	spin_lock_bh(&xprt->transport_lock);
 	xprt->ops->release_xprt(xprt, task);
 	if (xprt->ops->release_request)
 		xprt->ops->release_request(task);
-	if (!list_empty(&req->rq_list))
-		list_del(&req->rq_list);
 	xprt->last_used = jiffies;
 	xprt_schedule_autodisconnect(xprt);
 	spin_unlock_bh(&xprt->transport_lock);
@@ -1318,6 +1366,7 @@ static void xprt_init(struct rpc_xprt *xprt, struct net *net)
 
 	spin_lock_init(&xprt->transport_lock);
 	spin_lock_init(&xprt->reserve_lock);
+	spin_lock_init(&xprt->recv_lock);
 
 	INIT_LIST_HEAD(&xprt->free);
 	INIT_LIST_HEAD(&xprt->recv);

net/sunrpc/xprtrdma/backchannel.c

@@ -49,6 +49,7 @@ static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
 	if (IS_ERR(rb))
 		goto out_fail;
 	req->rl_rdmabuf = rb;
+	xdr_buf_init(&req->rl_hdrbuf, rb->rg_base, rdmab_length(rb));
 
 	size = r_xprt->rx_data.inline_rsize;
 	rb = rpcrdma_alloc_regbuf(size, DMA_TO_DEVICE, GFP_KERNEL);
@@ -202,20 +203,24 @@ size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *xprt)
  */
 int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
 {
-	struct rpc_xprt *xprt = rqst->rq_xprt;
-	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
+	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
-	struct rpcrdma_msg *headerp;
-
-	headerp = rdmab_to_msg(req->rl_rdmabuf);
-	headerp->rm_xid = rqst->rq_xid;
-	headerp->rm_vers = rpcrdma_version;
-	headerp->rm_credit =
-			cpu_to_be32(r_xprt->rx_buf.rb_bc_srv_max_requests);
-	headerp->rm_type = rdma_msg;
-	headerp->rm_body.rm_chunks[0] = xdr_zero;
-	headerp->rm_body.rm_chunks[1] = xdr_zero;
-	headerp->rm_body.rm_chunks[2] = xdr_zero;
+	__be32 *p;
+
+	rpcrdma_set_xdrlen(&req->rl_hdrbuf, 0);
+	xdr_init_encode(&req->rl_stream, &req->rl_hdrbuf,
+			req->rl_rdmabuf->rg_base);
+
+	p = xdr_reserve_space(&req->rl_stream, 28);
+	if (unlikely(!p))
+		return -EIO;
+	*p++ = rqst->rq_xid;
+	*p++ = rpcrdma_version;
+	*p++ = cpu_to_be32(r_xprt->rx_buf.rb_bc_srv_max_requests);
+	*p++ = rdma_msg;
+	*p++ = xdr_zero;
+	*p++ = xdr_zero;
+	*p = xdr_zero;
 
 	if (!rpcrdma_prepare_send_sges(&r_xprt->rx_ia, req, RPCRDMA_HDRLEN_MIN,
 				       &rqst->rq_snd_buf, rpcrdma_noch))
@@ -271,9 +276,6 @@ void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst)
  * @xprt: transport receiving the call
  * @rep: receive buffer containing the call
  *
- * Called in the RPC reply handler, which runs in a tasklet.
- * Be quick about it.
- *
  * Operational assumptions:
  *    o Backchannel credits are ignored, just as the NFS server
  *      forechannel currently does
@@ -284,7 +286,6 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 			     struct rpcrdma_rep *rep)
 {
 	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
-	struct rpcrdma_msg *headerp;
 	struct svc_serv *bc_serv;
 	struct rpcrdma_req *req;
 	struct rpc_rqst *rqst;
@@ -292,24 +293,15 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	size_t size;
 	__be32 *p;
 
-	headerp = rdmab_to_msg(rep->rr_rdmabuf);
+	p = xdr_inline_decode(&rep->rr_stream, 0);
+	size = xdr_stream_remaining(&rep->rr_stream);
+
 #ifdef RPCRDMA_BACKCHANNEL_DEBUG
 	pr_info("RPC:       %s: callback XID %08x, length=%u\n",
-		__func__, be32_to_cpu(headerp->rm_xid), rep->rr_len);
-	pr_info("RPC:       %s: %*ph\n", __func__, rep->rr_len, headerp);
+		__func__, be32_to_cpup(p), size);
+	pr_info("RPC:       %s: %*ph\n", __func__, size, p);
 #endif
 
-	/* Sanity check:
-	 * Need at least enough bytes for RPC/RDMA header, as code
-	 * here references the header fields by array offset. Also,
-	 * backward calls are always inline, so ensure there
-	 * are some bytes beyond the RPC/RDMA header.
-	 */
-	if (rep->rr_len < RPCRDMA_HDRLEN_MIN + 24)
-		goto out_short;
-	p = (__be32 *)((unsigned char *)headerp + RPCRDMA_HDRLEN_MIN);
-	size = rep->rr_len - RPCRDMA_HDRLEN_MIN;
-
 	/* Grab a free bc rqst */
 	spin_lock(&xprt->bc_pa_lock);
 	if (list_empty(&xprt->bc_pa_list)) {
@@ -325,7 +317,7 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	/* Prepare rqst */
 	rqst->rq_reply_bytes_recvd = 0;
 	rqst->rq_bytes_sent = 0;
-	rqst->rq_xid = headerp->rm_xid;
+	rqst->rq_xid = *p;
 
 	rqst->rq_private_buf.len = size;
 	set_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
@@ -337,9 +329,9 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	buf->len = size;
 
 	/* The receive buffer has to be hooked to the rpcrdma_req
-	 * so that it can be reposted after the server is done
-	 * parsing it but just before sending the backward
-	 * direction reply.
+	 * so that it is not released while the req is pointing
+	 * to its buffer, and so that it can be reposted after
+	 * the Upper Layer is done decoding it.
 	 */
 	req = rpcr_to_rdmar(rqst);
 	dprintk("RPC:       %s: attaching rep %p to req %p\n",
@@ -367,13 +359,4 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	 * when the connection is re-established.
 	 */
 	return;
-
-out_short:
-	pr_warn("RPC/RDMA short backward direction call\n");
-
-	if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, rep))
-		xprt_disconnect_done(xprt);
-	else
-		pr_warn("RPC:       %s: reposting rep %p\n",
-			__func__, rep);
 }

net/sunrpc/xprtrdma/fmr_ops.c

@@ -177,7 +177,7 @@ fmr_op_maxpages(struct rpcrdma_xprt *r_xprt)
 /* Use the ib_map_phys_fmr() verb to register a memory region
  * for remote access via RDMA READ or RDMA WRITE.
  */
-static int
+static struct rpcrdma_mr_seg *
 fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	   int nsegs, bool writing, struct rpcrdma_mw **out)
 {
@@ -188,7 +188,7 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 
 	mw = rpcrdma_get_mw(r_xprt);
 	if (!mw)
-		return -ENOBUFS;
+		return ERR_PTR(-ENOBUFS);
 
 	pageoff = offset_in_page(seg1->mr_offset);
 	seg1->mr_offset -= pageoff;	/* start of page */
@@ -232,13 +232,13 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	mw->mw_offset = dma_pages[0] + pageoff;
 
 	*out = mw;
-	return mw->mw_nents;
+	return seg;
 
 out_dmamap_err:
 	pr_err("rpcrdma: failed to DMA map sg %p sg_nents %d\n",
 	       mw->mw_sg, i);
 	rpcrdma_put_mw(r_xprt, mw);
-	return -EIO;
+	return ERR_PTR(-EIO);
 
 out_maperr:
 	pr_err("rpcrdma: ib_map_phys_fmr %u@0x%llx+%i (%d) status %i\n",
@@ -247,7 +247,7 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
 			mw->mw_sg, mw->mw_nents, mw->mw_dir);
 	rpcrdma_put_mw(r_xprt, mw);
-	return -EIO;
+	return ERR_PTR(-EIO);
 }
 
 /* Invalidate all memory regions that were registered for "req".

net/sunrpc/xprtrdma/frwr_ops.c

@@ -344,7 +344,7 @@ frwr_wc_localinv_wake(struct ib_cq *cq, struct ib_wc *wc)
 /* Post a REG_MR Work Request to register a memory region
  * for remote access via RDMA READ or RDMA WRITE.
  */
-static int
+static struct rpcrdma_mr_seg *
 frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	    int nsegs, bool writing, struct rpcrdma_mw **out)
 {
@@ -364,7 +364,7 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 			rpcrdma_defer_mr_recovery(mw);
 		mw = rpcrdma_get_mw(r_xprt);
 		if (!mw)
-			return -ENOBUFS;
+			return ERR_PTR(-ENOBUFS);
 	} while (mw->frmr.fr_state != FRMR_IS_INVALID);
 	frmr = &mw->frmr;
 	frmr->fr_state = FRMR_IS_VALID;
@@ -429,25 +429,25 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	mw->mw_offset = mr->iova;
 
 	*out = mw;
-	return mw->mw_nents;
+	return seg;
 
 out_dmamap_err:
 	pr_err("rpcrdma: failed to DMA map sg %p sg_nents %d\n",
 	       mw->mw_sg, i);
 	frmr->fr_state = FRMR_IS_INVALID;
 	rpcrdma_put_mw(r_xprt, mw);
-	return -EIO;
+	return ERR_PTR(-EIO);
 
 out_mapmr_err:
 	pr_err("rpcrdma: failed to map mr %p (%d/%d)\n",
 	       frmr->fr_mr, n, mw->mw_nents);
 	rpcrdma_defer_mr_recovery(mw);
-	return -EIO;
+	return ERR_PTR(-EIO);
 
 out_senderr:
 	pr_err("rpcrdma: FRMR registration ib_post_send returned %i\n", rc);
 	rpcrdma_defer_mr_recovery(mw);
-	return -ENOTCONN;
+	return ERR_PTR(-ENOTCONN);
 }
 
 /* Invalidate all memory regions that were registered for "req".

net/sunrpc/xprtrdma/svc_rdma_backchannel.c

@@ -52,7 +52,7 @@ int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt, __be32 *rdma_resp,
 	if (src->iov_len < 24)
 		goto out_shortreply;
 
-	spin_lock_bh(&xprt->transport_lock);
+	spin_lock(&xprt->recv_lock);
 	req = xprt_lookup_rqst(xprt, xid);
 	if (!req)
 		goto out_notfound;
@@ -69,17 +69,20 @@ int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt, __be32 *rdma_resp,
 	else if (credits > r_xprt->rx_buf.rb_bc_max_requests)
 		credits = r_xprt->rx_buf.rb_bc_max_requests;
 
+	spin_lock_bh(&xprt->transport_lock);
 	cwnd = xprt->cwnd;
 	xprt->cwnd = credits << RPC_CWNDSHIFT;
 	if (xprt->cwnd > cwnd)
 		xprt_release_rqst_cong(req->rq_task);
+	spin_unlock_bh(&xprt->transport_lock);
+
 
 	ret = 0;
 	xprt_complete_rqst(req->rq_task, rcvbuf->len);
 	rcvbuf->len = 0;
 
 out_unlock:
-	spin_unlock_bh(&xprt->transport_lock);
+	spin_unlock(&xprt->recv_lock);
 out:
 	return ret;
 
@@ -266,7 +269,7 @@ xprt_rdma_bc_put(struct rpc_xprt *xprt)
 	module_put(THIS_MODULE);
 }
 
-static struct rpc_xprt_ops xprt_rdma_bc_procs = {
+static const struct rpc_xprt_ops xprt_rdma_bc_procs = {
 	.reserve_xprt		= xprt_reserve_xprt_cong,
 	.release_xprt		= xprt_release_xprt_cong,
 	.alloc_slot		= xprt_alloc_slot,

net/sunrpc/xprtrdma/transport.c

@@ -149,7 +149,7 @@ static struct ctl_table sunrpc_table[] = {
 
 #endif
 
-static struct rpc_xprt_ops xprt_rdma_procs;	/*forward reference */
+static const struct rpc_xprt_ops xprt_rdma_procs;
 
 static void
 xprt_rdma_format_addresses4(struct rpc_xprt *xprt, struct sockaddr *sap)
@@ -559,6 +559,7 @@ rpcrdma_get_rdmabuf(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 
 	r_xprt->rx_stats.hardway_register_count += size;
 	req->rl_rdmabuf = rb;
+	xdr_buf_init(&req->rl_hdrbuf, rb->rg_base, rdmab_length(rb));
 	return true;
 }
 
@@ -684,7 +685,6 @@ xprt_rdma_free(struct rpc_task *task)
 
 	dprintk("RPC:       %s: called on 0x%p\n", __func__, req->rl_reply);
 
-	rpcrdma_remove_req(&r_xprt->rx_buf, req);
 	if (!list_empty(&req->rl_registered))
 		ia->ri_ops->ro_unmap_safe(r_xprt, req, !RPC_IS_ASYNC(task));
 	rpcrdma_unmap_sges(ia, req);
@@ -730,7 +730,7 @@ xprt_rdma_send_request(struct rpc_task *task)
 	if (unlikely(!list_empty(&req->rl_registered)))
 		r_xprt->rx_ia.ri_ops->ro_unmap_safe(r_xprt, req, false);
 
-	rc = rpcrdma_marshal_req(rqst);
+	rc = rpcrdma_marshal_req(r_xprt, rqst);
 	if (rc < 0)
 		goto failed_marshal;
 
@@ -811,7 +811,7 @@ xprt_rdma_disable_swap(struct rpc_xprt *xprt)
  * Plumbing for rpc transport switch and kernel module
  */
 
-static struct rpc_xprt_ops xprt_rdma_procs = {
+static const struct rpc_xprt_ops xprt_rdma_procs = {
 	.reserve_xprt		= xprt_reserve_xprt_cong,
 	.release_xprt		= xprt_release_xprt_cong, /* sunrpc/xprt.c */
 	.alloc_slot		= xprt_alloc_slot,

net/sunrpc/xprtrdma/verbs.c

@@ -139,14 +139,11 @@ rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 static void
 rpcrdma_update_granted_credits(struct rpcrdma_rep *rep)
 {
-	struct rpcrdma_msg *rmsgp = rdmab_to_msg(rep->rr_rdmabuf);
 	struct rpcrdma_buffer *buffer = &rep->rr_rxprt->rx_buf;
+	__be32 *p = rep->rr_rdmabuf->rg_base;
 	u32 credits;
 
-	if (rep->rr_len < RPCRDMA_HDRLEN_ERR)
-		return;
-
-	credits = be32_to_cpu(rmsgp->rm_credit);
+	credits = be32_to_cpup(p + 2);
 	if (credits == 0)
 		credits = 1;	/* don't deadlock */
 	else if (credits > buffer->rb_max_requests)
@@ -173,21 +170,19 @@ rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 		goto out_fail;
 
 	/* status == SUCCESS means all fields in wc are trustworthy */
-	if (wc->opcode != IB_WC_RECV)
-		return;
-
 	dprintk("RPC:       %s: rep %p opcode 'recv', length %u: success\n",
 		__func__, rep, wc->byte_len);
 
-	rep->rr_len = wc->byte_len;
+	rpcrdma_set_xdrlen(&rep->rr_hdrbuf, wc->byte_len);
 	rep->rr_wc_flags = wc->wc_flags;
 	rep->rr_inv_rkey = wc->ex.invalidate_rkey;
 
 	ib_dma_sync_single_for_cpu(rdmab_device(rep->rr_rdmabuf),
 				   rdmab_addr(rep->rr_rdmabuf),
-				   rep->rr_len, DMA_FROM_DEVICE);
+				   wc->byte_len, DMA_FROM_DEVICE);
 
-	rpcrdma_update_granted_credits(rep);
+	if (wc->byte_len >= RPCRDMA_HDRLEN_ERR)
+		rpcrdma_update_granted_credits(rep);
 
 out_schedule:
 	queue_work(rpcrdma_receive_wq, &rep->rr_work);
@@ -198,7 +193,7 @@ rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 		pr_err("rpcrdma: Recv: %s (%u/0x%x)\n",
 		       ib_wc_status_msg(wc->status),
 		       wc->status, wc->vendor_err);
-	rep->rr_len = RPCRDMA_BAD_LEN;
+	rpcrdma_set_xdrlen(&rep->rr_hdrbuf, 0);
 	goto out_schedule;
 }
 
@@ -974,6 +969,8 @@ rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt)
 		rc = PTR_ERR(rep->rr_rdmabuf);
 		goto out_free;
 	}
+	xdr_buf_init(&rep->rr_hdrbuf, rep->rr_rdmabuf->rg_base,
+		     rdmab_length(rep->rr_rdmabuf));
 
 	rep->rr_cqe.done = rpcrdma_wc_receive;
 	rep->rr_rxprt = r_xprt;
@@ -1004,7 +1001,6 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 	spin_lock_init(&buf->rb_recovery_lock);
 	INIT_LIST_HEAD(&buf->rb_mws);
 	INIT_LIST_HEAD(&buf->rb_all);
-	INIT_LIST_HEAD(&buf->rb_pending);
 	INIT_LIST_HEAD(&buf->rb_stale_mrs);
 	INIT_DELAYED_WORK(&buf->rb_refresh_worker,
 			  rpcrdma_mr_refresh_worker);

net/sunrpc/xprtrdma/xprt_rdma.h

@@ -218,18 +218,17 @@ enum {
 
 struct rpcrdma_rep {
 	struct ib_cqe		rr_cqe;
-	unsigned int		rr_len;
 	int			rr_wc_flags;
 	u32			rr_inv_rkey;
+	struct rpcrdma_regbuf	*rr_rdmabuf;
 	struct rpcrdma_xprt	*rr_rxprt;
 	struct work_struct	rr_work;
+	struct xdr_buf		rr_hdrbuf;
+	struct xdr_stream	rr_stream;
 	struct list_head	rr_list;
 	struct ib_recv_wr	rr_recv_wr;
-	struct rpcrdma_regbuf	*rr_rdmabuf;
 };
 
-#define RPCRDMA_BAD_LEN		(~0U)
-
 /*
  * struct rpcrdma_mw - external memory region metadata
  *
@@ -341,11 +340,12 @@ enum {
 struct rpcrdma_buffer;
 struct rpcrdma_req {
 	struct list_head	rl_list;
-	__be32			rl_xid;
 	unsigned int		rl_mapped_sges;
 	unsigned int		rl_connect_cookie;
 	struct rpcrdma_buffer	*rl_buffer;
 	struct rpcrdma_rep	*rl_reply;
+	struct xdr_stream	rl_stream;
+	struct xdr_buf		rl_hdrbuf;
 	struct ib_send_wr	rl_send_wr;
 	struct ib_sge		rl_send_sge[RPCRDMA_MAX_SEND_SGES];
 	struct rpcrdma_regbuf	*rl_rdmabuf;	/* xprt header */
@@ -403,7 +403,6 @@ struct rpcrdma_buffer {
 	int			rb_send_count, rb_recv_count;
 	struct list_head	rb_send_bufs;
 	struct list_head	rb_recv_bufs;
-	struct list_head	rb_pending;
 	u32			rb_max_requests;
 	atomic_t		rb_credits;	/* most recent credit grant */
 
@@ -440,24 +439,27 @@ struct rpcrdma_create_data_internal {
  * Statistics for RPCRDMA
  */
 struct rpcrdma_stats {
+	/* accessed when sending a call */
 	unsigned long		read_chunk_count;
 	unsigned long		write_chunk_count;
 	unsigned long		reply_chunk_count;
-
 	unsigned long long	total_rdma_request;
-	unsigned long long	total_rdma_reply;
 
+	/* rarely accessed error counters */
 	unsigned long long	pullup_copy_count;
-	unsigned long long	fixup_copy_count;
 	unsigned long		hardway_register_count;
 	unsigned long		failed_marshal_count;
 	unsigned long		bad_reply_count;
-	unsigned long		nomsg_call_count;
-	unsigned long		bcall_count;
 	unsigned long		mrs_recovered;
 	unsigned long		mrs_orphaned;
 	unsigned long		mrs_allocated;
+
+	/* accessed when receiving a reply */
+	unsigned long long	total_rdma_reply;
+	unsigned long long	fixup_copy_count;
 	unsigned long		local_inv_needed;
+	unsigned long		nomsg_call_count;
+	unsigned long		bcall_count;
 };
 
 /*
@@ -465,7 +467,8 @@ struct rpcrdma_stats {
  */
 struct rpcrdma_xprt;
 struct rpcrdma_memreg_ops {
-	int		(*ro_map)(struct rpcrdma_xprt *,
+	struct rpcrdma_mr_seg *
+			(*ro_map)(struct rpcrdma_xprt *,
 				  struct rpcrdma_mr_seg *, int, bool,
 				  struct rpcrdma_mw **);
 	void		(*ro_unmap_sync)(struct rpcrdma_xprt *,
@@ -552,34 +555,6 @@ void rpcrdma_destroy_req(struct rpcrdma_req *);
 int rpcrdma_buffer_create(struct rpcrdma_xprt *);
 void rpcrdma_buffer_destroy(struct rpcrdma_buffer *);
 
-static inline void
-rpcrdma_insert_req(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
-{
-	spin_lock(&buffers->rb_lock);
-	if (list_empty(&req->rl_list))
-		list_add_tail(&req->rl_list, &buffers->rb_pending);
-	spin_unlock(&buffers->rb_lock);
-}
-
-static inline struct rpcrdma_req *
-rpcrdma_lookup_req_locked(struct rpcrdma_buffer *buffers, __be32 xid)
-{
-	struct rpcrdma_req *pos;
-
-	list_for_each_entry(pos, &buffers->rb_pending, rl_list)
-		if (pos->rl_xid == xid)
-			return pos;
-	return NULL;
-}
-
-static inline void
-rpcrdma_remove_req(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
-{
-	spin_lock(&buffers->rb_lock);
-	list_del(&req->rl_list);
-	spin_unlock(&buffers->rb_lock);
-}
-
 struct rpcrdma_mw *rpcrdma_get_mw(struct rpcrdma_xprt *);
 void rpcrdma_put_mw(struct rpcrdma_xprt *, struct rpcrdma_mw *);
 struct rpcrdma_req *rpcrdma_buffer_get(struct rpcrdma_buffer *);
@@ -638,10 +613,16 @@ enum rpcrdma_chunktype {
 bool rpcrdma_prepare_send_sges(struct rpcrdma_ia *, struct rpcrdma_req *,
 			       u32, struct xdr_buf *, enum rpcrdma_chunktype);
 void rpcrdma_unmap_sges(struct rpcrdma_ia *, struct rpcrdma_req *);
-int rpcrdma_marshal_req(struct rpc_rqst *);
+int rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst);
 void rpcrdma_set_max_header_sizes(struct rpcrdma_xprt *);
 void rpcrdma_reply_handler(struct work_struct *work);
 
+static inline void rpcrdma_set_xdrlen(struct xdr_buf *xdr, size_t len)
+{
+	xdr->head[0].iov_len = len;
+	xdr->len = len;
+}
+
 /* RPC/RDMA module init - xprtrdma/transport.c
  */
 extern unsigned int xprt_rdma_max_inline_read;

net/sunrpc/xprtsock.c

@@ -969,10 +969,12 @@ static void xs_local_data_read_skb(struct rpc_xprt *xprt,
 		return;
 
 	/* Look up and lock the request corresponding to the given XID */
-	spin_lock_bh(&xprt->transport_lock);
+	spin_lock(&xprt->recv_lock);
 	rovr = xprt_lookup_rqst(xprt, *xp);
 	if (!rovr)
 		goto out_unlock;
+	xprt_pin_rqst(rovr);
+	spin_unlock(&xprt->recv_lock);
 	task = rovr->rq_task;
 
 	copied = rovr->rq_private_buf.buflen;
@@ -981,13 +983,16 @@ static void xs_local_data_read_skb(struct rpc_xprt *xprt,
 
 	if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) {
 		dprintk("RPC:       sk_buff copy failed\n");
-		goto out_unlock;
+		spin_lock(&xprt->recv_lock);
+		goto out_unpin;
 	}
 
+	spin_lock(&xprt->recv_lock);
 	xprt_complete_rqst(task, copied);
-
+out_unpin:
+	xprt_unpin_rqst(rovr);
  out_unlock:
-	spin_unlock_bh(&xprt->transport_lock);
+	spin_unlock(&xprt->recv_lock);
 }
 
 static void xs_local_data_receive(struct sock_xprt *transport)
@@ -1050,10 +1055,12 @@ static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
 		return;
 
 	/* Look up and lock the request corresponding to the given XID */
-	spin_lock_bh(&xprt->transport_lock);
+	spin_lock(&xprt->recv_lock);
 	rovr = xprt_lookup_rqst(xprt, *xp);
 	if (!rovr)
 		goto out_unlock;
+	xprt_pin_rqst(rovr);
+	spin_unlock(&xprt->recv_lock);
 	task = rovr->rq_task;
 
 	if ((copied = rovr->rq_private_buf.buflen) > repsize)
@@ -1062,16 +1069,21 @@ static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
 	/* Suck it into the iovec, verify checksum if not done by hw. */
 	if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
 		__UDPX_INC_STATS(sk, UDP_MIB_INERRORS);
-		goto out_unlock;
+		spin_lock(&xprt->recv_lock);
+		goto out_unpin;
 	}
 
 	__UDPX_INC_STATS(sk, UDP_MIB_INDATAGRAMS);
 
+	spin_lock_bh(&xprt->transport_lock);
 	xprt_adjust_cwnd(xprt, task, copied);
+	spin_unlock_bh(&xprt->transport_lock);
+	spin_lock(&xprt->recv_lock);
 	xprt_complete_rqst(task, copied);
-
+out_unpin:
+	xprt_unpin_rqst(rovr);
  out_unlock:
-	spin_unlock_bh(&xprt->transport_lock);
+	spin_unlock(&xprt->recv_lock);
 }
 
 static void xs_udp_data_receive(struct sock_xprt *transport)
@@ -1277,25 +1289,12 @@ static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
 	}
 
 	len = desc->count;
-	if (len > transport->tcp_reclen - transport->tcp_offset) {
-		struct xdr_skb_reader my_desc;
-
-		len = transport->tcp_reclen - transport->tcp_offset;
-		memcpy(&my_desc, desc, sizeof(my_desc));
-		my_desc.count = len;
-		r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
-					  &my_desc, xdr_skb_read_bits);
-		desc->count -= r;
-		desc->offset += r;
-	} else
-		r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
+	if (len > transport->tcp_reclen - transport->tcp_offset)
+		desc->count = transport->tcp_reclen - transport->tcp_offset;
+	r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
 					  desc, xdr_skb_read_bits);
 
-	if (r > 0) {
-		transport->tcp_copied += r;
-		transport->tcp_offset += r;
-	}
-	if (r != len) {
+	if (desc->count) {
 		/* Error when copying to the receive buffer,
 		 * usually because we weren't able to allocate
 		 * additional buffer pages. All we can do now
@@ -1315,6 +1314,10 @@ static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
 		return;
 	}
 
+	transport->tcp_copied += r;
+	transport->tcp_offset += r;
+	desc->count = len - r;
+
 	dprintk("RPC:       XID %08x read %zd bytes\n",
 			ntohl(transport->tcp_xid), r);
 	dprintk("RPC:       xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
@@ -1343,21 +1346,24 @@ static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
 	dprintk("RPC:       read reply XID %08x\n", ntohl(transport->tcp_xid));
 
 	/* Find and lock the request corresponding to this xid */
-	spin_lock_bh(&xprt->transport_lock);
+	spin_lock(&xprt->recv_lock);
 	req = xprt_lookup_rqst(xprt, transport->tcp_xid);
 	if (!req) {
 		dprintk("RPC:       XID %08x request not found!\n",
 				ntohl(transport->tcp_xid));
-		spin_unlock_bh(&xprt->transport_lock);
+		spin_unlock(&xprt->recv_lock);
 		return -1;
 	}
+	xprt_pin_rqst(req);
+	spin_unlock(&xprt->recv_lock);
 
 	xs_tcp_read_common(xprt, desc, req);
 
+	spin_lock(&xprt->recv_lock);
 	if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
 		xprt_complete_rqst(req->rq_task, transport->tcp_copied);
-
-	spin_unlock_bh(&xprt->transport_lock);
+	xprt_unpin_rqst(req);
+	spin_unlock(&xprt->recv_lock);
 	return 0;
 }
 
@@ -1376,11 +1382,9 @@ static int xs_tcp_read_callback(struct rpc_xprt *xprt,
 				container_of(xprt, struct sock_xprt, xprt);
 	struct rpc_rqst *req;
 
-	/* Look up and lock the request corresponding to the given XID */
-	spin_lock_bh(&xprt->transport_lock);
+	/* Look up the request corresponding to the given XID */
 	req = xprt_lookup_bc_request(xprt, transport->tcp_xid);
 	if (req == NULL) {
-		spin_unlock_bh(&xprt->transport_lock);
 		printk(KERN_WARNING "Callback slot table overflowed\n");
 		xprt_force_disconnect(xprt);
 		return -1;
@@ -1391,7 +1395,6 @@ static int xs_tcp_read_callback(struct rpc_xprt *xprt,
 
 	if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
 		xprt_complete_bc_request(req, transport->tcp_copied);
-	spin_unlock_bh(&xprt->transport_lock);
 
 	return 0;
 }
@@ -1516,6 +1519,7 @@ static void xs_tcp_data_receive(struct sock_xprt *transport)
 		.arg.data = xprt,
 	};
 	unsigned long total = 0;
+	int loop;
 	int read = 0;
 
 	mutex_lock(&transport->recv_mutex);
@@ -1524,20 +1528,20 @@ static void xs_tcp_data_receive(struct sock_xprt *transport)
 		goto out;
 
 	/* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
-	for (;;) {
+	for (loop = 0; loop < 64; loop++) {
 		lock_sock(sk);
 		read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
 		if (read <= 0) {
 			clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
 			release_sock(sk);
-			if (!test_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
-				break;
-		} else {
-			release_sock(sk);
-			total += read;
+			break;
 		}
+		release_sock(sk);
+		total += read;
 		rd_desc.count = 65536;
 	}
+	if (test_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
+		queue_work(xprtiod_workqueue, &transport->recv_worker);
 out:
 	mutex_unlock(&transport->recv_mutex);
 	trace_xs_tcp_data_ready(xprt, read, total);
@@ -2724,7 +2728,7 @@ static void bc_destroy(struct rpc_xprt *xprt)
 	module_put(THIS_MODULE);
 }
 
-static struct rpc_xprt_ops xs_local_ops = {
+static const struct rpc_xprt_ops xs_local_ops = {
 	.reserve_xprt		= xprt_reserve_xprt,
 	.release_xprt		= xs_tcp_release_xprt,
 	.alloc_slot		= xprt_alloc_slot,
@@ -2742,7 +2746,7 @@ static struct rpc_xprt_ops xs_local_ops = {
 	.disable_swap		= xs_disable_swap,
 };
 
-static struct rpc_xprt_ops xs_udp_ops = {
+static const struct rpc_xprt_ops xs_udp_ops = {
 	.set_buffer_size	= xs_udp_set_buffer_size,
 	.reserve_xprt		= xprt_reserve_xprt_cong,
 	.release_xprt		= xprt_release_xprt_cong,
@@ -2764,7 +2768,7 @@ static struct rpc_xprt_ops xs_udp_ops = {
 	.inject_disconnect	= xs_inject_disconnect,
 };
 
-static struct rpc_xprt_ops xs_tcp_ops = {
+static const struct rpc_xprt_ops xs_tcp_ops = {
 	.reserve_xprt		= xprt_reserve_xprt,
 	.release_xprt		= xs_tcp_release_xprt,
 	.alloc_slot		= xprt_lock_and_alloc_slot,
@@ -2795,7 +2799,7 @@ static struct rpc_xprt_ops xs_tcp_ops = {
  * The rpc_xprt_ops for the server backchannel
  */
 
-static struct rpc_xprt_ops bc_tcp_ops = {
+static const struct rpc_xprt_ops bc_tcp_ops = {
 	.reserve_xprt		= xprt_reserve_xprt,
 	.release_xprt		= xprt_release_xprt,
 	.alloc_slot		= xprt_alloc_slot,