From ca05cbae2a0468e5d78e9b4605936a8bf5da328b Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 10 Jul 2021 18:07:14 -0400 Subject: NFS: Fix up nfs_ctx_key_to_expire() If the cached credential exists but doesn't have any expiration callback then exit early. Fix up atomicity issues when replacing the credential with a new one since the existing code could lead to refcount leaks. Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 41 ++++++++++++++++++++++++++++------------- 1 file changed, 28 insertions(+), 13 deletions(-) (limited to 'fs/nfs/write.c') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index eae9bf114041..773ea2c8504d 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1246,7 +1246,7 @@ nfs_key_timeout_notify(struct file *filp, struct inode *inode) struct nfs_open_context *ctx = nfs_file_open_context(filp); if (nfs_ctx_key_to_expire(ctx, inode) && - !ctx->ll_cred) + !rcu_access_pointer(ctx->ll_cred)) /* Already expired! */ return -EACCES; return 0; @@ -1258,23 +1258,38 @@ nfs_key_timeout_notify(struct file *filp, struct inode *inode) bool nfs_ctx_key_to_expire(struct nfs_open_context *ctx, struct inode *inode) { struct rpc_auth *auth = NFS_SERVER(inode)->client->cl_auth; - struct rpc_cred *cred = ctx->ll_cred; + struct rpc_cred *cred, *new, *old = NULL; struct auth_cred acred = { .cred = ctx->cred, }; + bool ret = false; - if (cred && !cred->cr_ops->crmatch(&acred, cred, 0)) { - put_rpccred(cred); - ctx->ll_cred = NULL; - cred = NULL; - } - if (!cred) - cred = auth->au_ops->lookup_cred(auth, &acred, 0); - if (!cred || IS_ERR(cred)) + rcu_read_lock(); + cred = rcu_dereference(ctx->ll_cred); + if (cred && !(cred->cr_ops->crkey_timeout && + cred->cr_ops->crkey_timeout(cred))) + goto out; + rcu_read_unlock(); + + new = auth->au_ops->lookup_cred(auth, &acred, 0); + if (new == cred) { + put_rpccred(new); return true; - ctx->ll_cred = cred; - return !!(cred->cr_ops->crkey_timeout && - cred->cr_ops->crkey_timeout(cred)); + } + if (IS_ERR_OR_NULL(new)) { + new = NULL; + ret = true; + } else if (new->cr_ops->crkey_timeout && + new->cr_ops->crkey_timeout(new)) + ret = true; + + rcu_read_lock(); + old = rcu_dereference_protected(xchg(&ctx->ll_cred, + RCU_INITIALIZER(new)), 1); +out: + rcu_read_unlock(); + put_rpccred(old); + return ret; } /* -- cgit From 110cb2d2f9326030f13b7ec85d6d482934ea5462 Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Mon, 4 Oct 2021 10:10:16 -0400 Subject: NFS: Instrument i_size_write() Generate a trace event whenever the NFS client modifies the size of a file. These new events aid troubleshooting workloads that trigger races around size updates. There are four new trace points, all named nfs_size_something so they are easy to grep for or enable as a group with a single glob. Size updated on the server: kworker/u24:10-194 [010] 369.939174: nfs_size_update: fileid=00:28:2 fhandle=0x36fbbe51 version=1752899344277980615 cursize=250471 newsize=172083 Server-side size update reported via NFSv3 WCC attributes: fsx-1387 [006] 380.760686: nfs_size_wcc: fileid=00:28:2 fhandle=0x36fbbe51 version=1752899355909932456 cursize=146792 newsize=171216 File has been truncated locally: fsx-1387 [007] 369.437421: nfs_size_truncate: fileid=00:28:2 fhandle=0x36fbbe51 version=1752899231200117272 cursize=215244 newsize=0 File has been extended locally: fsx-1387 [007] 369.439213: nfs_size_grow: fileid=00:28:2 fhandle=0x36fbbe51 version=1752899343704248410 cursize=258048 newsize=262144 Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/nfs/write.c') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 773ea2c8504d..b89d5ef3af0e 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -288,6 +288,7 @@ static void nfs_grow_file(struct page *page, unsigned int offset, unsigned int c end = page_file_offset(page) + ((loff_t)offset+count); if (i_size >= end) goto out; + trace_nfs_size_grow(inode, end); i_size_write(inode, end); NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_SIZE; nfs_inc_stats(inode, NFSIOS_EXTENDWRITE); -- cgit From 64a93dbf25d3a1368bb58ddf0f61d0a92d7479e3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Oct 2021 15:44:16 -0400 Subject: NFS: Fix deadlocks in nfs_scan_commit_list() Partially revert commit 2ce209c42c01 ("NFS: Wait for requests that are locked on the commit list"), since it can lead to deadlocks between commit requests and nfs_join_page_group(). For now we should assume that any locked requests on the commit list are either about to be removed and committed by another task, or the writes they describe are about to be retransmitted. In either case, we should not need to worry. Fixes: 2ce209c42c01 ("NFS: Wait for requests that are locked on the commit list") Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 17 ++--------------- 1 file changed, 2 insertions(+), 15 deletions(-) (limited to 'fs/nfs/write.c') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index b89d5ef3af0e..38f181e1343a 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1039,25 +1039,11 @@ nfs_scan_commit_list(struct list_head *src, struct list_head *dst, struct nfs_page *req, *tmp; int ret = 0; -restart: list_for_each_entry_safe(req, tmp, src, wb_list) { kref_get(&req->wb_kref); if (!nfs_lock_request(req)) { - int status; - - /* Prevent deadlock with nfs_lock_and_join_requests */ - if (!list_empty(dst)) { - nfs_release_request(req); - continue; - } - /* Ensure we make progress to prevent livelock */ - mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); - status = nfs_wait_on_request(req); nfs_release_request(req); - mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); - if (status < 0) - break; - goto restart; + continue; } nfs_request_remove_commit_list(req, cinfo); clear_bit(PG_COMMIT_TO_DS, &req->wb_flags); @@ -1952,6 +1938,7 @@ static int __nfs_commit_inode(struct inode *inode, int how, int may_wait = how & FLUSH_SYNC; int ret, nscan; + how &= ~FLUSH_SYNC; nfs_init_cinfo_from_inode(&cinfo, inode); nfs_commit_begin(cinfo.mds); for (;;) { -- cgit From 133a48abf6ecc535d7eddc6da1c3e4c972445882 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Mon, 4 Oct 2021 15:37:42 -0400 Subject: NFS: Fix up commit deadlocks If O_DIRECT bumps the commit_info rpcs_out field, then that could lead to fsync() hangs. The fix is to ensure that O_DIRECT calls nfs_commit_end(). Fixes: 723c921e7dfc ("sched/wait, fs/nfs: Convert wait_on_atomic_t() usage to the new wait_var_event() API") Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'fs/nfs/write.c') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 38f181e1343a..465220f47142 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1673,10 +1673,13 @@ static void nfs_commit_begin(struct nfs_mds_commit_info *cinfo) atomic_inc(&cinfo->rpcs_out); } -static void nfs_commit_end(struct nfs_mds_commit_info *cinfo) +bool nfs_commit_end(struct nfs_mds_commit_info *cinfo) { - if (atomic_dec_and_test(&cinfo->rpcs_out)) + if (atomic_dec_and_test(&cinfo->rpcs_out)) { wake_up_var(&cinfo->rpcs_out); + return true; + } + return false; } void nfs_commitdata_release(struct nfs_commit_data *data) @@ -1776,6 +1779,7 @@ void nfs_init_commit(struct nfs_commit_data *data, data->res.fattr = &data->fattr; data->res.verf = &data->verf; nfs_fattr_init(&data->fattr); + nfs_commit_begin(cinfo->mds); } EXPORT_SYMBOL_GPL(nfs_init_commit); @@ -1822,7 +1826,6 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how, /* Set up the argument struct */ nfs_init_commit(data, head, NULL, cinfo); - atomic_inc(&cinfo->mds->rpcs_out); if (NFS_SERVER(inode)->nfs_client->cl_minorversion) task_flags = RPC_TASK_MOVEABLE; return nfs_initiate_commit(NFS_CLIENT(inode), data, NFS_PROTO(inode), -- cgit From b40887e10dcacc5e8ae3c1a99dcba20877c4831b Mon Sep 17 00:00:00 2001 From: Chuck Lever Date: Sat, 16 Oct 2021 18:02:57 -0400 Subject: SUNRPC: Trace calls to .rpc_call_done Introduce a single tracepoint that can replace simple dprintk call sites in upper layer "rpc_call_done" callbacks. Example: kworker/u24:2-1254 [001] 771.026677: rpc_stats_latency: task:00000001@00000002 xid=0x16a6f3c0 rpcbindv2 GETPORT backlog=446 rtt=101 execute=555 kworker/u24:2-1254 [001] 771.026677: rpc_task_call_done: task:00000001@00000002 flags=ASYNC|DYNAMIC|SOFT|SOFTCONN|SENT runstate=RUNNING|ACTIVE status=0 action=rpcb_getport_done kworker/u24:2-1254 [001] 771.026678: rpcb_setport: task:00000001@00000002 status=0 port=20048 Signed-off-by: Chuck Lever Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'fs/nfs/write.c') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 465220f47142..82c5b89395f6 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1840,9 +1840,6 @@ static void nfs_commit_done(struct rpc_task *task, void *calldata) { struct nfs_commit_data *data = calldata; - dprintk("NFS: %5u nfs_commit_done (status %d)\n", - task->tk_pid, task->tk_status); - /* Call the NFS version-specific code */ NFS_PROTO(data->inode)->commit_done(task, data); trace_nfs_commit_done(task, data); -- cgit From 4cd27df88af29929cda6e8eb4e0f5bb4e25812bf Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 21 Oct 2021 17:11:37 -0400 Subject: NFS: Remove redundant call to __set_page_dirty_nobuffers Remove a redundant call in nfs_updatepage(). nfs_writepage_setup() will have already called nfs_mark_request_dirty() on success. Signed-off-by: Trond Myklebust --- fs/nfs/write.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'fs/nfs/write.c') diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 82c5b89395f6..9b7619ce17a7 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1384,8 +1384,6 @@ int nfs_updatepage(struct file *file, struct page *page, status = nfs_writepage_setup(ctx, page, offset, count); if (status < 0) nfs_set_pageerror(mapping); - else - __set_page_dirty_nobuffers(page); out: dprintk("NFS: nfs_updatepage returns %d (isize %lld)\n", status, (long long)i_size_read(inode)); -- cgit