[PATCH 1/8] NFS: Fix initialisation of I/O result struct in nfs_pgio_rpcsetup

[PATCH 1/8] NFS: Fix initialisation of I/O result struct in nfs_pgio_rpcsetup

[Trond Myklebust]

Initialise the result count to 0 rather than initialising it to the
argument count. The reason is that we want to ensure we record the
I/O stats correctly in the case where an error is returned (for
instance in the layoutstats).

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pagelist.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index 56cefa0ab804..20b3717cd7ca 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -590,7 +590,7 @@ static void nfs_pgio_rpcsetup(struct nfs_pgio_header *hdr,
 	}
 
 	hdr->res.fattr   = &hdr->fattr;
-	hdr->res.count   = count;
+	hdr->res.count   = 0;
 	hdr->res.eof     = 0;
 	hdr->res.verf    = &hdr->verf;
 	nfs_fattr_init(&hdr->fattr);
-- 
2.21.0

[PATCH 2/8] NFS: On fatal writeback errors, we need to call nfs_inode_remove_request()

[Trond Myklebust]

If the writeback error is fatal, we need to remove the tracking structures
(i.e. the nfs_page) from the inode.

Fixes: 6fbda89b257f ("NFS: Replace custom error reporting mechanism...")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/write.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 92d9cadc6102..3399149435ce 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -57,6 +57,7 @@ static const struct rpc_call_ops nfs_commit_ops;
 static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops;
 static const struct nfs_commit_completion_ops nfs_commit_completion_ops;
 static const struct nfs_rw_ops nfs_rw_write_ops;
+static void nfs_inode_remove_request(struct nfs_page *req);
 static void nfs_clear_request_commit(struct nfs_page *req);
 static void nfs_init_cinfo_from_inode(struct nfs_commit_info *cinfo,
 				      struct inode *inode);
@@ -591,7 +592,9 @@ nfs_lock_and_join_requests(struct page *page)
 
 static void nfs_write_error(struct nfs_page *req, int error)
 {
+	nfs_set_pageerror(page_file_mapping(req->wb_page));
 	nfs_mapping_set_error(req->wb_page, error);
+	nfs_inode_remove_request(req);
 	nfs_end_page_writeback(req);
 	nfs_release_request(req);
 }
-- 
2.21.0

[PATCH 3/8] NFS: Fix spurious EIO read errors

[Trond Myklebust]

If the client attempts to read a page, but the read fails due to some
spurious error (e.g. an ACCESS error or a timeout, ...) then we need
to allow other processes to retry.
Also try to report errors correctly when doing a synchronous readpage.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/read.c | 35 ++++++++++++++++++++++++++---------
 1 file changed, 26 insertions(+), 9 deletions(-)

diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index c19841c82b6a..252f31bfd497 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -91,19 +91,25 @@ void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio)
 }
 EXPORT_SYMBOL_GPL(nfs_pageio_reset_read_mds);
 
-static void nfs_readpage_release(struct nfs_page *req)
+static void nfs_readpage_release(struct nfs_page *req, int error)
 {
 	struct inode *inode = d_inode(nfs_req_openctx(req)->dentry);
+	struct page *page = req->wb_page;
 
 	dprintk("NFS: read done (%s/%llu %d@%lld)\n", inode->i_sb->s_id,
 		(unsigned long long)NFS_FILEID(inode), req->wb_bytes,
 		(long long)req_offset(req));
 
+	if (error < 0)
+		SetPageError(page);
 	if (nfs_page_group_sync_on_bit(req, PG_UNLOCKPAGE)) {
-		if (PageUptodate(req->wb_page))
-			nfs_readpage_to_fscache(inode, req->wb_page, 0);
+		struct address_space *mapping = page_file_mapping(page);
 
-		unlock_page(req->wb_page);
+		if (PageUptodate(page))
+			nfs_readpage_to_fscache(inode, page, 0);
+		else if (PageError(page) && !PagePrivate(page))
+			generic_error_remove_page(mapping, page);
+		unlock_page(page);
 	}
 	nfs_release_request(req);
 }
@@ -131,7 +137,7 @@ int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
 			     &nfs_async_read_completion_ops);
 	if (!nfs_pageio_add_request(&pgio, new)) {
 		nfs_list_remove_request(new);
-		nfs_readpage_release(new);
+		nfs_readpage_release(new, pgio.pg_error);
 	}
 	nfs_pageio_complete(&pgio);
 
@@ -153,6 +159,7 @@ static void nfs_page_group_set_uptodate(struct nfs_page *req)
 static void nfs_read_completion(struct nfs_pgio_header *hdr)
 {
 	unsigned long bytes = 0;
+	int error;
 
 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
 		goto out;
@@ -179,14 +186,19 @@ static void nfs_read_completion(struct nfs_pgio_header *hdr)
 				zero_user_segment(page, start, end);
 			}
 		}
+		error = 0;
 		bytes += req->wb_bytes;
 		if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
 			if (bytes <= hdr->good_bytes)
 				nfs_page_group_set_uptodate(req);
+			else {
+				error = hdr->error;
+				xchg(&nfs_req_openctx(req)->error, error);
+			}
 		} else
 			nfs_page_group_set_uptodate(req);
 		nfs_list_remove_request(req);
-		nfs_readpage_release(req);
+		nfs_readpage_release(req, error);
 	}
 out:
 	hdr->release(hdr);
@@ -213,7 +225,7 @@ nfs_async_read_error(struct list_head *head, int error)
 	while (!list_empty(head)) {
 		req = nfs_list_entry(head->next);
 		nfs_list_remove_request(req);
-		nfs_readpage_release(req);
+		nfs_readpage_release(req, error);
 	}
 }
 
@@ -337,8 +349,13 @@ int nfs_readpage(struct file *file, struct page *page)
 			goto out;
 	}
 
+	xchg(&ctx->error, 0);
 	error = nfs_readpage_async(ctx, inode, page);
-
+	if (!error) {
+		error = wait_on_page_locked_killable(page);
+		if (!PageUptodate(page) && !error)
+			error = xchg(&ctx->error, 0);
+	}
 out:
 	put_nfs_open_context(ctx);
 	return error;
@@ -372,8 +389,8 @@ readpage_async_filler(void *data, struct page *page)
 		zero_user_segment(page, len, PAGE_SIZE);
 	if (!nfs_pageio_add_request(desc->pgio, new)) {
 		nfs_list_remove_request(new);
-		nfs_readpage_release(new);
 		error = desc->pgio->pg_error;
+		nfs_readpage_release(new, error);
 		goto out;
 	}
 	return 0;
-- 
2.21.0

[PATCH 4/8] SUNRPC: Don't handle errors if the bind/connect succeeded

[Trond Myklebust]

Don't handle errors in call_bind_status()/call_connect_status()
if it turns out that a previous call caused it to succeed.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: stable@vger.kernel.org # v5.1+
---
 net/sunrpc/clnt.c | 26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index d8679b6027e9..92c1c4e02855 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1970,6 +1970,7 @@ call_bind(struct rpc_task *task)
 static void
 call_bind_status(struct rpc_task *task)
 {
+	struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 	int status = -EIO;
 
 	if (rpc_task_transmitted(task)) {
@@ -1977,14 +1978,13 @@ call_bind_status(struct rpc_task *task)
 		return;
 	}
 
-	if (task->tk_status >= 0) {
-		dprint_status(task);
-		task->tk_status = 0;
+	dprint_status(task);
+	trace_rpc_bind_status(task);
+	if (task->tk_status >= 0 || xprt_bound(xprt)) {
 		task->tk_action = call_connect;
 		return;
 	}
 
-	trace_rpc_bind_status(task);
 	switch (task->tk_status) {
 	case -ENOMEM:
 		dprintk("RPC: %5u rpcbind out of memory\n", task->tk_pid);
@@ -2043,7 +2043,6 @@ call_bind_status(struct rpc_task *task)
 
 	rpc_call_rpcerror(task, status);
 	return;
-
 retry_timeout:
 	task->tk_status = 0;
 	task->tk_action = call_bind;
@@ -2090,6 +2089,7 @@ call_connect(struct rpc_task *task)
 static void
 call_connect_status(struct rpc_task *task)
 {
+	struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 	struct rpc_clnt *clnt = task->tk_client;
 	int status = task->tk_status;
 
@@ -2099,8 +2099,15 @@ call_connect_status(struct rpc_task *task)
 	}
 
 	dprint_status(task);
-
 	trace_rpc_connect_status(task);
+
+	if (task->tk_status == 0) {
+		clnt->cl_stats->netreconn++;
+		goto out_next;
+	}
+	if (xprt_connected(xprt))
+		goto out_next;
+
 	task->tk_status = 0;
 	switch (status) {
 	case -ECONNREFUSED:
@@ -2131,13 +2138,12 @@ call_connect_status(struct rpc_task *task)
 	case -EAGAIN:
 	case -ETIMEDOUT:
 		goto out_retry;
-	case 0:
-		clnt->cl_stats->netreconn++;
-		task->tk_action = call_transmit;
-		return;
 	}
 	rpc_call_rpcerror(task, status);
 	return;
+out_next:
+	task->tk_action = call_transmit;
+	return;
 out_retry:
 	/* Check for timeouts before looping back to call_bind */
 	task->tk_action = call_bind;
-- 
2.21.0

[PATCH 5/8] pNFS/flexfiles: Turn off soft RPC calls

[Trond Myklebust]

The pNFS/flexfiles I/O requests are sent with the SOFTCONN flag set, so
they automatically time out if the connection breaks. It should
therefore not be necessary to have the soft flag set in addition.

Fixes: 5f01d9539496 ("nfs41: create NFSv3 DS connection if specified")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs_nfs.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c
index c0046c348910..82af4809b869 100644
--- a/fs/nfs/pnfs_nfs.c
+++ b/fs/nfs/pnfs_nfs.c
@@ -627,11 +627,16 @@ static int _nfs4_pnfs_v3_ds_connect(struct nfs_server *mds_srv,
 			/* Add this address as an alias */
 			rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
 					rpc_clnt_test_and_add_xprt, NULL);
-		} else
-			clp = get_v3_ds_connect(mds_srv,
-					(struct sockaddr *)&da->da_addr,
-					da->da_addrlen, IPPROTO_TCP,
-					timeo, retrans);
+			continue;
+		}
+		clp = get_v3_ds_connect(mds_srv,
+				(struct sockaddr *)&da->da_addr,
+				da->da_addrlen, IPPROTO_TCP,
+				timeo, retrans);
+		if (IS_ERR(clp))
+			continue;
+		clp->cl_rpcclient->cl_softerr = 0;
+		clp->cl_rpcclient->cl_softrtry = 0;
 	}
 
 	if (IS_ERR(clp)) {
-- 
2.21.0

[PATCH 6/8] SUNRPC: Handle EADDRINUSE and ENOBUFS correctly

[Trond Myklebust]

If a connect or bind attempt returns EADDRINUSE, that means we want to
retry with a different port. It is not a fatal connection error.
Similarly, ENOBUFS is not fatal, but just indicates a memory allocation
issue. Retry after a short delay.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 net/sunrpc/clnt.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 92c1c4e02855..9546a4c72838 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -2003,6 +2003,9 @@ call_bind_status(struct rpc_task *task)
 		task->tk_rebind_retry--;
 		rpc_delay(task, 3*HZ);
 		goto retry_timeout;
+	case -ENOBUFS:
+		rpc_delay(task, HZ >> 2);
+		goto retry_timeout;
 	case -EAGAIN:
 		goto retry_timeout;
 	case -ETIMEDOUT:
@@ -2026,7 +2029,6 @@ call_bind_status(struct rpc_task *task)
 	case -ENETDOWN:
 	case -EHOSTUNREACH:
 	case -ENETUNREACH:
-	case -ENOBUFS:
 	case -EPIPE:
 		dprintk("RPC: %5u remote rpcbind unreachable: %d\n",
 				task->tk_pid, task->tk_status);
@@ -2124,8 +2126,6 @@ call_connect_status(struct rpc_task *task)
 	case -ENETDOWN:
 	case -ENETUNREACH:
 	case -EHOSTUNREACH:
-	case -EADDRINUSE:
-	case -ENOBUFS:
 	case -EPIPE:
 		xprt_conditional_disconnect(task->tk_rqstp->rq_xprt,
 					    task->tk_rqstp->rq_connect_cookie);
@@ -2134,10 +2134,14 @@ call_connect_status(struct rpc_task *task)
 		/* retry with existing socket, after a delay */
 		rpc_delay(task, 3*HZ);
 		/* fall through */
+	case -EADDRINUSE:
 	case -ENOTCONN:
 	case -EAGAIN:
 	case -ETIMEDOUT:
 		goto out_retry;
+	case -ENOBUFS:
+		rpc_delay(task, HZ >> 2);
+		goto out_retry;
 	}
 	rpc_call_rpcerror(task, status);
 	return;
-- 
2.21.0

[PATCH 7/8] Revert "NFSv4/flexfiles: Abort I/O early if the layout segment was invalidated"

[Trond Myklebust]

This reverts commit a79f194aa4879e9baad118c3f8bb2ca24dbef765.
The mechanism for aborting I/O is racy, since we are not guaranteed that
the request is asleep while we're changing both task->tk_status and
task->tk_action.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: stable@vger.kernel.org # v5.1
---
 fs/nfs/flexfilelayout/flexfilelayout.c | 17 -----------------
 include/linux/sunrpc/sched.h           |  1 -
 net/sunrpc/xprt.c                      |  7 -------
 3 files changed, 25 deletions(-)

diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c
index b04e20d28162..2c7e1eca1ed7 100644
--- a/fs/nfs/flexfilelayout/flexfilelayout.c
+++ b/fs/nfs/flexfilelayout/flexfilelayout.c
@@ -1148,8 +1148,6 @@ static int ff_layout_async_handle_error_v4(struct rpc_task *task,
 		break;
 	case -NFS4ERR_RETRY_UNCACHED_REP:
 		break;
-	case -EAGAIN:
-		return -NFS4ERR_RESET_TO_PNFS;
 	/* Invalidate Layout errors */
 	case -NFS4ERR_PNFS_NO_LAYOUT:
 	case -ESTALE:           /* mapped NFS4ERR_STALE */
@@ -1210,7 +1208,6 @@ static int ff_layout_async_handle_error_v3(struct rpc_task *task,
 	case -EBADHANDLE:
 	case -ELOOP:
 	case -ENOSPC:
-	case -EAGAIN:
 		break;
 	case -EJUKEBOX:
 		nfs_inc_stats(lseg->pls_layout->plh_inode, NFSIOS_DELAY);
@@ -1445,16 +1442,6 @@ static void ff_layout_read_prepare_v4(struct rpc_task *task, void *data)
 	ff_layout_read_prepare_common(task, hdr);
 }
 
-static void
-ff_layout_io_prepare_transmit(struct rpc_task *task,
-		void *data)
-{
-	struct nfs_pgio_header *hdr = data;
-
-	if (!pnfs_is_valid_lseg(hdr->lseg))
-		rpc_exit(task, -EAGAIN);
-}
-
 static void ff_layout_read_call_done(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
@@ -1740,7 +1727,6 @@ static void ff_layout_commit_release(void *data)
 
 static const struct rpc_call_ops ff_layout_read_call_ops_v3 = {
 	.rpc_call_prepare = ff_layout_read_prepare_v3,
-	.rpc_call_prepare_transmit = ff_layout_io_prepare_transmit,
 	.rpc_call_done = ff_layout_read_call_done,
 	.rpc_count_stats = ff_layout_read_count_stats,
 	.rpc_release = ff_layout_read_release,
@@ -1748,7 +1734,6 @@ static const struct rpc_call_ops ff_layout_read_call_ops_v3 = {
 
 static const struct rpc_call_ops ff_layout_read_call_ops_v4 = {
 	.rpc_call_prepare = ff_layout_read_prepare_v4,
-	.rpc_call_prepare_transmit = ff_layout_io_prepare_transmit,
 	.rpc_call_done = ff_layout_read_call_done,
 	.rpc_count_stats = ff_layout_read_count_stats,
 	.rpc_release = ff_layout_read_release,
@@ -1756,7 +1741,6 @@ static const struct rpc_call_ops ff_layout_read_call_ops_v4 = {
 
 static const struct rpc_call_ops ff_layout_write_call_ops_v3 = {
 	.rpc_call_prepare = ff_layout_write_prepare_v3,
-	.rpc_call_prepare_transmit = ff_layout_io_prepare_transmit,
 	.rpc_call_done = ff_layout_write_call_done,
 	.rpc_count_stats = ff_layout_write_count_stats,
 	.rpc_release = ff_layout_write_release,
@@ -1764,7 +1748,6 @@ static const struct rpc_call_ops ff_layout_write_call_ops_v3 = {
 
 static const struct rpc_call_ops ff_layout_write_call_ops_v4 = {
 	.rpc_call_prepare = ff_layout_write_prepare_v4,
-	.rpc_call_prepare_transmit = ff_layout_io_prepare_transmit,
 	.rpc_call_done = ff_layout_write_call_done,
 	.rpc_count_stats = ff_layout_write_count_stats,
 	.rpc_release = ff_layout_write_release,
diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h
index baa3ecdb882f..27536b961552 100644
--- a/include/linux/sunrpc/sched.h
+++ b/include/linux/sunrpc/sched.h
@@ -98,7 +98,6 @@ typedef void			(*rpc_action)(struct rpc_task *);
 
 struct rpc_call_ops {
 	void (*rpc_call_prepare)(struct rpc_task *, void *);
-	void (*rpc_call_prepare_transmit)(struct rpc_task *, void *);
 	void (*rpc_call_done)(struct rpc_task *, void *);
 	void (*rpc_count_stats)(struct rpc_task *, void *);
 	void (*rpc_release)(void *);
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 783748dc5e6f..2e71f5455c6c 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -1408,13 +1408,6 @@ xprt_request_transmit(struct rpc_rqst *req, struct rpc_task *snd_task)
 			status = -EBADMSG;
 			goto out_dequeue;
 		}
-		if (task->tk_ops->rpc_call_prepare_transmit) {
-			task->tk_ops->rpc_call_prepare_transmit(task,
-					task->tk_calldata);
-			status = task->tk_status;
-			if (status < 0)
-				goto out_dequeue;
-		}
 		if (RPC_SIGNALLED(task)) {
 			status = -ERESTARTSYS;
 			goto out_dequeue;
-- 
2.21.0

[PATCH 8/8] SUNRPC: Handle connection breakages correctly in call_status()

[Trond Myklebust]

If the connection breaks while we're waiting for a reply from the
server, then we want to immediately try to reconnect.

Fixes: ec6017d90359 ("SUNRPC fix regression in umount of a secure mount")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 net/sunrpc/clnt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 9546a4c72838..f13ec73c8299 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -2375,7 +2375,7 @@ call_status(struct rpc_task *task)
 	case -ECONNABORTED:
 	case -ENOTCONN:
 		rpc_force_rebind(clnt);
-		/* fall through */
+		break;
 	case -EADDRINUSE:
 		rpc_delay(task, 3*HZ);
 		/* fall through */
-- 
2.21.0