From 7e10cc25bfa0dd3602bbcf5cc9c759a90eb675dc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 9 Aug 2019 12:06:43 -0400 Subject: [PATCH 01/18] NFS: Don't refresh attributes with mounted-on-file information If we've been given the attributes of the mounted-on-file, then do not use those to check or update the attributes on the application-visible inode. Signed-off-by: Trond Myklebust --- fs/nfs/inode.c | 33 +++++++++++++++++++-------------- 1 file changed, 19 insertions(+), 14 deletions(-) diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 8a1758200b57..c764cfe456e5 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -1403,12 +1403,21 @@ static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fat if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) return 0; + /* No fileid? Just exit */ + if (!(fattr->valid & NFS_ATTR_FATTR_FILEID)) + return 0; /* Has the inode gone and changed behind our back? */ - if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid) + if (nfsi->fileid != fattr->fileid) { + /* Is this perhaps the mounted-on fileid? */ + if ((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) && + nfsi->fileid == fattr->mounted_on_fileid) + return 0; return -ESTALE; + } if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) return -ESTALE; + if (!nfs_file_has_buffered_writers(nfsi)) { /* Verify a few of the more important attributes */ if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 && !inode_eq_iversion_raw(inode, fattr->change_attr)) @@ -1768,18 +1777,6 @@ int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fa EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc); -static inline bool nfs_fileid_valid(struct nfs_inode *nfsi, - struct nfs_fattr *fattr) -{ - bool ret1 = true, ret2 = true; - - if (fattr->valid & NFS_ATTR_FATTR_FILEID) - ret1 = (nfsi->fileid == fattr->fileid); - if (fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) - ret2 = (nfsi->fileid == fattr->mounted_on_fileid); - return ret1 || ret2; -} - /* * Many nfs protocol calls return the new file attributes after * an operation. Here we update the inode to reflect the state @@ -1810,7 +1807,15 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr) nfs_display_fhandle_hash(NFS_FH(inode)), atomic_read(&inode->i_count), fattr->valid); - if (!nfs_fileid_valid(nfsi, fattr)) { + /* No fileid? Just exit */ + if (!(fattr->valid & NFS_ATTR_FATTR_FILEID)) + return 0; + /* Has the inode gone and changed behind our back? */ + if (nfsi->fileid != fattr->fileid) { + /* Is this perhaps the mounted-on fileid? */ + if ((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) && + nfsi->fileid == fattr->mounted_on_fileid) + return 0; printk(KERN_ERR "NFS: server %s error: fileid changed\n" "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n", NFS_SERVER(inode)->nfs_client->cl_hostname, From 90cf500e338ab3f3c0f126ba37e36fb6a9058441 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 9 Aug 2019 15:03:11 -0400 Subject: [PATCH 02/18] NFSv4: Fix return values for nfs4_file_open() Currently, we are translating RPC level errors such as timeouts, as well as interrupts etc into EOPENSTALE, which forces a single replay of the open attempt. What we actually want to do is force the replay only in the cases where the returned error indicates that the file may have changed on the server. So the fix is to spell out the exact set of errors where we want to return EOPENSTALE. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4file.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/fs/nfs/nfs4file.c b/fs/nfs/nfs4file.c index 96db471ca2e5..339663d04bf8 100644 --- a/fs/nfs/nfs4file.c +++ b/fs/nfs/nfs4file.c @@ -73,13 +73,13 @@ nfs4_file_open(struct inode *inode, struct file *filp) if (IS_ERR(inode)) { err = PTR_ERR(inode); switch (err) { - case -EPERM: - case -EACCES: - case -EDQUOT: - case -ENOSPC: - case -EROFS: - goto out_put_ctx; default: + goto out_put_ctx; + case -ENOENT: + case -ESTALE: + case -EISDIR: + case -ENOTDIR: + case -ELOOP: goto out_drop; } } From 9821421a291b548ef4369c6998745baa36ddecd5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 9 Aug 2019 12:15:07 -0400 Subject: [PATCH 03/18] NFSv4: Fix return value in nfs_finish_open() If the file turns out to be of the wrong type after opening, we want to revalidate the path and retry, so return EOPENSTALE rather than ESTALE. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 8d501093660f..0adfd8840110 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1487,7 +1487,7 @@ static int nfs_finish_open(struct nfs_open_context *ctx, if (S_ISREG(file->f_path.dentry->d_inode->i_mode)) nfs_file_set_open_context(file, ctx); else - err = -ESTALE; + err = -EOPENSTALE; out: return err; } From f4340e9314dbfadc48758945f85fc3b16612d06f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 12 Aug 2019 15:19:54 -0400 Subject: [PATCH 04/18] NFSv4/pnfs: Fix a page lock leak in nfs_pageio_resend() If the attempt to resend the pages fails, we need to ensure that we clean up those pages that were not transmitted. Fixes: d600ad1f2bdb ("NFS41: pop some layoutget errors to application") Signed-off-by: Trond Myklebust Cc: stable@vger.kernel.org # v4.5+ --- fs/nfs/pagelist.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index ed4e1b07447b..15c254753f88 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -1251,20 +1251,22 @@ static void nfs_pageio_complete_mirror(struct nfs_pageio_descriptor *desc, int nfs_pageio_resend(struct nfs_pageio_descriptor *desc, struct nfs_pgio_header *hdr) { - LIST_HEAD(failed); + LIST_HEAD(pages); desc->pg_io_completion = hdr->io_completion; desc->pg_dreq = hdr->dreq; - while (!list_empty(&hdr->pages)) { - struct nfs_page *req = nfs_list_entry(hdr->pages.next); + list_splice_init(&hdr->pages, &pages); + while (!list_empty(&pages)) { + struct nfs_page *req = nfs_list_entry(pages.next); if (!nfs_pageio_add_request(desc, req)) - nfs_list_move_request(req, &failed); + break; } nfs_pageio_complete(desc); - if (!list_empty(&failed)) { - list_move(&failed, &hdr->pages); - return desc->pg_error < 0 ? desc->pg_error : -EIO; + if (!list_empty(&pages)) { + int err = desc->pg_error < 0 ? desc->pg_error : -EIO; + hdr->completion_ops->error_cleanup(&pages, err); + return err; } return 0; } From eb2c50da9e256dbbb3ff27694440e4c1900cfef8 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 12 Aug 2019 18:04:36 -0400 Subject: [PATCH 05/18] NFS: Ensure O_DIRECT reports an error if the bytes read/written is 0 If the attempt to resend the I/O results in no bytes being read/written, we must ensure that we report the error. Signed-off-by: Trond Myklebust Fixes: 0a00b77b331a ("nfs: mirroring support for direct io") Cc: stable@vger.kernel.org # v3.20+ --- fs/nfs/direct.c | 27 ++++++++++++++++++--------- fs/nfs/pagelist.c | 1 + 2 files changed, 19 insertions(+), 9 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index 0cb442406168..222d7115db71 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -401,15 +401,21 @@ static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) unsigned long bytes = 0; struct nfs_direct_req *dreq = hdr->dreq; - if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) - goto out_put; - spin_lock(&dreq->lock); - if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0)) + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) dreq->error = hdr->error; - else + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) { + spin_unlock(&dreq->lock); + goto out_put; + } + + if (hdr->good_bytes != 0) nfs_direct_good_bytes(dreq, hdr); + if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) + dreq->error = 0; + spin_unlock(&dreq->lock); while (!list_empty(&hdr->pages)) { @@ -782,16 +788,19 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) bool request_commit = false; struct nfs_page *req = nfs_list_entry(hdr->pages.next); - if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) - goto out_put; - nfs_init_cinfo_from_dreq(&cinfo, dreq); spin_lock(&dreq->lock); if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) dreq->error = hdr->error; - if (dreq->error == 0) { + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) { + spin_unlock(&dreq->lock); + goto out_put; + } + + if (hdr->good_bytes != 0) { nfs_direct_good_bytes(dreq, hdr); if (nfs_write_need_commit(hdr)) { if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index 15c254753f88..56cefa0ab804 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -1266,6 +1266,7 @@ int nfs_pageio_resend(struct nfs_pageio_descriptor *desc, if (!list_empty(&pages)) { int err = desc->pg_error < 0 ? desc->pg_error : -EIO; hdr->completion_ops->error_cleanup(&pages, err); + nfs_set_pgio_error(hdr, err, hdr->io_start); return err; } return 0; From 17d8c5d145000070c581f2a8aa01edc7998582ab Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 14 Aug 2019 14:19:09 -0400 Subject: [PATCH 06/18] NFS: Fix initialisation of I/O result struct in nfs_pgio_rpcsetup 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 --- 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); From 06c9fdf3b9f1acc6e53753c99c54c39764cc979f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 14 Aug 2019 15:42:43 -0400 Subject: [PATCH 07/18] NFS: On fatal writeback errors, we need to call nfs_inode_remove_request() 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 --- 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); } From bd736ed3e2d1088d9b4050f727342e1e619c3841 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Aug 2019 17:26:17 -0400 Subject: [PATCH 08/18] SUNRPC: Don't handle errors if the bind/connect succeeded 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 Cc: stable@vger.kernel.org # v5.1+ --- net/sunrpc/clnt.c | 35 ++++++++++++++++++++++++----------- 1 file changed, 24 insertions(+), 11 deletions(-) diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index d8679b6027e9..3b731411d8e8 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 +1978,15 @@ call_bind_status(struct rpc_task *task) return; } - if (task->tk_status >= 0) { - dprint_status(task); + dprint_status(task); + trace_rpc_bind_status(task); + if (task->tk_status >= 0) + goto out_next; + if (xprt_bound(xprt)) { task->tk_status = 0; - task->tk_action = call_connect; - return; + goto out_next; } - trace_rpc_bind_status(task); switch (task->tk_status) { case -ENOMEM: dprintk("RPC: %5u rpcbind out of memory\n", task->tk_pid); @@ -2043,7 +2045,9 @@ call_bind_status(struct rpc_task *task) rpc_call_rpcerror(task, status); return; - +out_next: + task->tk_action = call_connect; + return; retry_timeout: task->tk_status = 0; task->tk_action = call_bind; @@ -2090,6 +2094,7 @@ 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 +2104,17 @@ 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)) { + task->tk_status = 0; + goto out_next; + } + task->tk_status = 0; switch (status) { case -ECONNREFUSED: @@ -2131,13 +2145,12 @@ 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; From bf2bf9b80e0cd3568ddc85a241abe0dd8b46ebdc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Aug 2019 20:18:48 -0400 Subject: [PATCH 09/18] pNFS/flexfiles: Turn off soft RPC calls 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 --- 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)) { From 80f455da6cd0998a5be30a8af24ea2a22815c212 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Aug 2019 22:55:19 -0400 Subject: [PATCH 10/18] SUNRPC: Handle EADDRINUSE and ENOBUFS correctly 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 --- 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 3b731411d8e8..8a25440b771c 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -2005,6 +2005,9 @@ 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: @@ -2028,7 +2031,6 @@ 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); @@ -2131,8 +2133,6 @@ 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); @@ -2141,10 +2141,14 @@ 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; From d5711920ec6e578f51db95caa6f185f5090b865e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 16 Aug 2019 08:37:26 -0400 Subject: [PATCH 11/18] Revert "NFSv4/flexfiles: Abort I/O early if the layout segment was invalidated" 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 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; From c82e5472c9980e0e483f4b689044150eefaca408 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 16 Aug 2019 08:58:48 -0400 Subject: [PATCH 12/18] SUNRPC: Handle connection breakages correctly in call_status() 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 --- 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 8a25440b771c..a07b516e503a 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -2382,7 +2382,7 @@ 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 */ From 7af46292dadcf8870946916f79fdddf79bd7267f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 23 Aug 2019 08:37:50 -0400 Subject: [PATCH 13/18] pNFS/flexfiles: Don't time out requests on hard mounts If the mount is hard, we should ignore the 'io_maxretrans' module parameter so that we always keep retrying. Signed-off-by: Trond Myklebust --- fs/nfs/flexfilelayout/flexfilelayout.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c index 2c7e1eca1ed7..5657b7f2611f 100644 --- a/fs/nfs/flexfilelayout/flexfilelayout.c +++ b/fs/nfs/flexfilelayout/flexfilelayout.c @@ -8,6 +8,7 @@ */ #include +#include #include #include #include @@ -928,7 +929,9 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, pgm = &pgio->pg_mirrors[0]; pgm->pg_bsize = mirror->mirror_ds->ds_versions[0].rsize; - pgio->pg_maxretrans = io_maxretrans; + if (NFS_SERVER(pgio->pg_inode)->flags & + (NFS_MOUNT_SOFT|NFS_MOUNT_SOFTERR)) + pgio->pg_maxretrans = io_maxretrans; return; out_nolseg: if (pgio->pg_error < 0) @@ -940,6 +943,7 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, pgio->pg_lseg); pnfs_put_lseg(pgio->pg_lseg); pgio->pg_lseg = NULL; + pgio->pg_maxretrans = 0; nfs_pageio_reset_read_mds(pgio); } @@ -1000,7 +1004,9 @@ ff_layout_pg_init_write(struct nfs_pageio_descriptor *pgio, pgm->pg_bsize = mirror->mirror_ds->ds_versions[0].wsize; } - pgio->pg_maxretrans = io_maxretrans; + if (NFS_SERVER(pgio->pg_inode)->flags & + (NFS_MOUNT_SOFT|NFS_MOUNT_SOFTERR)) + pgio->pg_maxretrans = io_maxretrans; return; out_mds: @@ -1010,6 +1016,7 @@ ff_layout_pg_init_write(struct nfs_pageio_descriptor *pgio, pgio->pg_lseg); pnfs_put_lseg(pgio->pg_lseg); pgio->pg_lseg = NULL; + pgio->pg_maxretrans = 0; nfs_pageio_reset_write_mds(pgio); } From 8f54c7a4babf58bbaf849e126f7ae9664bdc9e04 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 15 Aug 2019 12:26:05 -0400 Subject: [PATCH 14/18] NFS: Fix spurious EIO read errors 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 --- fs/nfs/internal.h | 10 ++++++++++ fs/nfs/read.c | 35 ++++++++++++++++++++++++++--------- fs/nfs/write.c | 12 ------------ 3 files changed, 36 insertions(+), 21 deletions(-) diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index a2346a2f8361..e64f810223be 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -775,3 +775,13 @@ static inline bool nfs_error_is_fatal(int err) } } +static inline bool nfs_error_is_fatal_on_server(int err) +{ + switch (err) { + case 0: + case -ERESTARTSYS: + case -EINTR: + return false; + } + return nfs_error_is_fatal(err); +} diff --git a/fs/nfs/read.c b/fs/nfs/read.c index c19841c82b6a..cfe0b586eadd 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 (nfs_error_is_fatal_on_server(error) && error != -ETIMEDOUT) + 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; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 3399149435ce..cee9905e419c 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -599,18 +599,6 @@ static void nfs_write_error(struct nfs_page *req, int error) nfs_release_request(req); } -static bool -nfs_error_is_fatal_on_server(int err) -{ - switch (err) { - case 0: - case -ERESTARTSYS: - case -EINTR: - return false; - } - return nfs_error_is_fatal(err); -} - /* * Find an associated nfs write request, and prepare to flush it out * May return an error if the user signalled nfs_wait_on_request(). From 96c4145599b30c0eb6cbeaa24207802452dd1872 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 24 Aug 2019 10:39:00 -0400 Subject: [PATCH 15/18] NFS: Fix writepage(s) error handling to not report errors twice If writepage()/writepages() saw an error, but handled it without reporting it, we should not be re-reporting that error on exit. Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index cee9905e419c..d193042fa228 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -621,12 +621,12 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags)); /* If there is a fatal error that covers this write, just exit */ - ret = 0; mapping = page_file_mapping(page); - if (test_bit(AS_ENOSPC, &mapping->flags) || - test_bit(AS_EIO, &mapping->flags)) + ret = pgio->pg_error; + if (nfs_error_is_fatal_on_server(ret)) goto out_launder; + ret = 0; if (!nfs_pageio_add_request(pgio, req)) { ret = pgio->pg_error; /* @@ -638,6 +638,7 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, } else ret = -EAGAIN; nfs_redirty_request(req); + pgio->pg_error = 0; } else nfs_add_stats(page_file_mapping(page)->host, NFSIOS_WRITEPAGES, 1); @@ -657,7 +658,7 @@ static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, ret = nfs_page_async_flush(pgio, page); if (ret == -EAGAIN) { redirty_page_for_writepage(wbc, page); - ret = 0; + ret = AOP_WRITEPAGE_ACTIVATE; } return ret; } @@ -676,10 +677,11 @@ static int nfs_writepage_locked(struct page *page, nfs_pageio_init_write(&pgio, inode, 0, false, &nfs_async_write_completion_ops); err = nfs_do_writepage(page, wbc, &pgio); + pgio.pg_error = 0; nfs_pageio_complete(&pgio); if (err < 0) return err; - if (pgio.pg_error < 0) + if (nfs_error_is_fatal(pgio.pg_error)) return pgio.pg_error; return 0; } @@ -689,7 +691,8 @@ int nfs_writepage(struct page *page, struct writeback_control *wbc) int ret; ret = nfs_writepage_locked(page, wbc); - unlock_page(page); + if (ret != AOP_WRITEPAGE_ACTIVATE) + unlock_page(page); return ret; } @@ -698,7 +701,8 @@ static int nfs_writepages_callback(struct page *page, struct writeback_control * int ret; ret = nfs_do_writepage(page, wbc, data); - unlock_page(page); + if (ret != AOP_WRITEPAGE_ACTIVATE) + unlock_page(page); return ret; } @@ -724,13 +728,14 @@ int nfs_writepages(struct address_space *mapping, struct writeback_control *wbc) &nfs_async_write_completion_ops); pgio.pg_io_completion = ioc; err = write_cache_pages(mapping, wbc, nfs_writepages_callback, &pgio); + pgio.pg_error = 0; nfs_pageio_complete(&pgio); nfs_io_completion_put(ioc); if (err < 0) goto out_err; err = pgio.pg_error; - if (err < 0) + if (nfs_error_is_fatal(err)) goto out_err; return 0; out_err: From 71affe9be45a5c60b9772e1b2701710712637274 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 26 Aug 2019 20:41:16 -0400 Subject: [PATCH 16/18] NFSv2: Fix eof handling If we received a reply from the server with a zero length read and no error, then that implies we are at eof. Signed-off-by: Trond Myklebust --- fs/nfs/proc.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index 5552fa8b6e12..ec79d2214a78 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -594,7 +594,8 @@ static int nfs_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr) /* Emulate the eof flag, which isn't normally needed in NFSv2 * as it is guaranteed to always return the file attributes */ - if (hdr->args.offset + hdr->res.count >= hdr->res.fattr->size) + if ((hdr->res.count == 0 && hdr->args.count > 0) || + hdr->args.offset + hdr->res.count >= hdr->res.fattr->size) hdr->res.eof = 1; } return 0; From d33d4beb522987d1c305c12500796f9be3687dee Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 27 Aug 2019 07:03:28 -0400 Subject: [PATCH 17/18] NFSv2: Fix write regression Ensure we update the write result count on success, since the RPC call itself does not do so. Reported-by: Jan Stancek Reported-by: Naresh Kamboju Signed-off-by: Trond Myklebust Tested-by: Jan Stancek --- fs/nfs/proc.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c index ec79d2214a78..0f7288b94633 100644 --- a/fs/nfs/proc.c +++ b/fs/nfs/proc.c @@ -616,8 +616,10 @@ static int nfs_proc_pgio_rpc_prepare(struct rpc_task *task, static int nfs_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr) { - if (task->tk_status >= 0) + if (task->tk_status >= 0) { + hdr->res.count = hdr->args.count; nfs_writeback_update_inode(hdr); + } return 0; } From 99300a85260c2b7febd57082a617d1062532067e Mon Sep 17 00:00:00 2001 From: YueHaibing Date: Tue, 27 Aug 2019 15:16:36 +0800 Subject: [PATCH 18/18] NFS: remove set but not used variable 'mapping' Fixes gcc '-Wunused-but-set-variable' warning: fs/nfs/write.c: In function nfs_page_async_flush: fs/nfs/write.c:609:24: warning: variable mapping set but not used [-Wunused-but-set-variable] It is not use since commit aefb623c422e ("NFS: Fix writepage(s) error handling to not report errors twice") Reported-by: Hulk Robot Signed-off-by: YueHaibing Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index d193042fa228..85ca49549b39 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -606,7 +606,6 @@ static void nfs_write_error(struct nfs_page *req, int error) static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, struct page *page) { - struct address_space *mapping; struct nfs_page *req; int ret = 0; @@ -621,7 +620,6 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio, WARN_ON_ONCE(test_bit(PG_CLEAN, &req->wb_flags)); /* If there is a fatal error that covers this write, just exit */ - mapping = page_file_mapping(page); ret = pgio->pg_error; if (nfs_error_is_fatal_on_server(ret)) goto out_launder;