From d5f8d3fe72594f2e896b407f78daf24f37ef4d53 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:25 -0400 Subject: [PATCH 01/42] NFSv41: Fix a memory leak in nfs41_proc_async_sequence() If the call to rpc_call_async() fails, then the arguments will not be freed, since there will be no call to nfs41_sequence_call_done Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 39 +++++++++++++++++++++------------------ 1 file changed, 21 insertions(+), 18 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 70015dd60a98..89be778a6543 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5078,18 +5078,27 @@ static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred) &res, args.sa_cache_this, 1); } +struct nfs4_sequence_data { + struct nfs_client *clp; + struct nfs4_sequence_args args; + struct nfs4_sequence_res res; +}; + static void nfs41_sequence_release(void *data) { - struct nfs_client *clp = (struct nfs_client *)data; + struct nfs4_sequence_data *calldata = data; + struct nfs_client *clp = calldata->clp; if (atomic_read(&clp->cl_count) > 1) nfs4_schedule_state_renewal(clp); nfs_put_client(clp); + kfree(calldata); } static void nfs41_sequence_call_done(struct rpc_task *task, void *data) { - struct nfs_client *clp = (struct nfs_client *)data; + struct nfs4_sequence_data *calldata = data; + struct nfs_client *clp = calldata->clp; nfs41_sequence_done(clp, task->tk_msg.rpc_resp, task->tk_status); @@ -5106,19 +5115,16 @@ static void nfs41_sequence_call_done(struct rpc_task *task, void *data) } dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred); out: - kfree(task->tk_msg.rpc_argp); - kfree(task->tk_msg.rpc_resp); - dprintk("<-- %s\n", __func__); } static void nfs41_sequence_prepare(struct rpc_task *task, void *data) { - struct nfs_client *clp; + struct nfs4_sequence_data *calldata = data; + struct nfs_client *clp = calldata->clp; struct nfs4_sequence_args *args; struct nfs4_sequence_res *res; - clp = (struct nfs_client *)data; args = task->tk_msg.rpc_argp; res = task->tk_msg.rpc_resp; @@ -5136,8 +5142,7 @@ static const struct rpc_call_ops nfs41_sequence_ops = { static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred) { - struct nfs4_sequence_args *args; - struct nfs4_sequence_res *res; + struct nfs4_sequence_data *calldata; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE], .rpc_cred = cred, @@ -5145,20 +5150,18 @@ static int nfs41_proc_async_sequence(struct nfs_client *clp, if (!atomic_inc_not_zero(&clp->cl_count)) return -EIO; - args = kzalloc(sizeof(*args), GFP_NOFS); - res = kzalloc(sizeof(*res), GFP_NOFS); - if (!args || !res) { - kfree(args); - kfree(res); + calldata = kmalloc(sizeof(*calldata), GFP_NOFS); + if (calldata == NULL) { nfs_put_client(clp); return -ENOMEM; } - res->sr_slotid = NFS4_MAX_SLOT_TABLE; - msg.rpc_argp = args; - msg.rpc_resp = res; + calldata->res.sr_slotid = NFS4_MAX_SLOT_TABLE; + msg.rpc_argp = &calldata->args; + msg.rpc_resp = &calldata->res; + calldata->clp = clp; return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT, - &nfs41_sequence_ops, (void *)clp); + &nfs41_sequence_ops, calldata); } struct nfs4_reclaim_complete_data { From 2a6e26cdb8f17b1075c2dfd8f2f3c341bac4f441 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:25 -0400 Subject: [PATCH 02/42] NFSv4.1: Clean up nfs4_setup_sequence Firstly, there is little point in first zeroing out the entire struct nfs4_sequence_res, and then initialising all fields save one. Just initialise the last field to zero... Secondly, nfs41_setup_sequence() has only 2 possible return values: 0, or -EAGAIN, so there is no 'terminate rpc task' case. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 89be778a6543..4bfc0b7c428f 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -480,7 +480,6 @@ static int nfs41_setup_sequence(struct nfs4_session *session, if (res->sr_slotid != NFS4_MAX_SLOT_TABLE) return 0; - memset(res, 0, sizeof(*res)); res->sr_slotid = NFS4_MAX_SLOT_TABLE; tbl = &session->fc_slot_table; @@ -525,6 +524,7 @@ static int nfs41_setup_sequence(struct nfs4_session *session, res->sr_session = session; res->sr_slotid = slotid; res->sr_renewal_time = jiffies; + res->sr_status_flags = 0; /* * sr_status is only set in decode_sequence, and so will remain * set to 1 if an rpc level failure occurs. @@ -548,11 +548,6 @@ int nfs4_setup_sequence(struct nfs_client *clp, goto out; ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply, task); - if (ret && ret != -EAGAIN) { - /* terminate rpc task */ - task->tk_status = ret; - task->tk_action = NULL; - } out: dprintk("<-- %s status=%d\n", __func__, ret); return ret; From d185a334c748b3ca9de1f3a293fd8a9cf68378ab Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:25 -0400 Subject: [PATCH 03/42] NFSv4.1: Simplify nfs41_sequence_done() Nobody uses the rpc_status parameter. It is not obvious why we need the struct nfs_client argument either, when we already have that information in the session. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 4bfc0b7c428f..fc653c3bc557 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -370,12 +370,11 @@ static void nfs41_check_drain_session_complete(struct nfs4_session *ses) complete(&ses->complete); } -static void nfs41_sequence_free_slot(const struct nfs_client *clp, - struct nfs4_sequence_res *res) +static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res) { struct nfs4_slot_table *tbl; - tbl = &clp->cl_session->fc_slot_table; + tbl = &res->sr_session->fc_slot_table; if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) { /* just wake up the next guy waiting since * we may have not consumed a slot after all */ @@ -385,14 +384,12 @@ static void nfs41_sequence_free_slot(const struct nfs_client *clp, spin_lock(&tbl->slot_tbl_lock); nfs4_free_slot(tbl, res->sr_slotid); - nfs41_check_drain_session_complete(clp->cl_session); + nfs41_check_drain_session_complete(res->sr_session); spin_unlock(&tbl->slot_tbl_lock); res->sr_slotid = NFS4_MAX_SLOT_TABLE; } -static void nfs41_sequence_done(struct nfs_client *clp, - struct nfs4_sequence_res *res, - int rpc_status) +static void nfs41_sequence_done(struct nfs4_sequence_res *res) { unsigned long timestamp; struct nfs4_slot_table *tbl; @@ -413,7 +410,8 @@ static void nfs41_sequence_done(struct nfs_client *clp, /* Check the SEQUENCE operation status */ if (res->sr_status == 0) { - tbl = &clp->cl_session->fc_slot_table; + struct nfs_client *clp = res->sr_session->clp; + tbl = &res->sr_session->fc_slot_table; slot = tbl->slots + res->sr_slotid; /* Update the slot's sequence and clientid lease timer */ ++slot->seq_nr; @@ -429,7 +427,7 @@ static void nfs41_sequence_done(struct nfs_client *clp, out: /* The session may be reset by one of the error handlers. */ dprintk("%s: Error %d free the slot \n", __func__, res->sr_status); - nfs41_sequence_free_slot(clp, res); + nfs41_sequence_free_slot(res); } /* @@ -582,7 +580,7 @@ static void nfs41_call_sync_done(struct rpc_task *task, void *calldata) { struct nfs41_call_sync_data *data = calldata; - nfs41_sequence_done(data->clp, data->seq_res, task->tk_status); + nfs41_sequence_done(data->seq_res); } struct rpc_call_ops nfs41_call_sync_ops = { @@ -662,7 +660,7 @@ static void nfs4_sequence_done(const struct nfs_server *server, { #ifdef CONFIG_NFS_V4_1 if (nfs4_has_session(server->nfs_client)) - nfs41_sequence_done(server->nfs_client, res, rpc_status); + nfs41_sequence_done(res); #endif /* CONFIG_NFS_V4_1 */ } @@ -4606,7 +4604,7 @@ static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata) (struct nfs4_get_lease_time_data *)calldata; dprintk("--> %s\n", __func__); - nfs41_sequence_done(data->clp, &data->res->lr_seq_res, task->tk_status); + nfs41_sequence_done(&data->res->lr_seq_res); switch (task->tk_status) { case -NFS4ERR_DELAY: case -NFS4ERR_GRACE: @@ -5095,7 +5093,7 @@ static void nfs41_sequence_call_done(struct rpc_task *task, void *data) struct nfs4_sequence_data *calldata = data; struct nfs_client *clp = calldata->clp; - nfs41_sequence_done(clp, task->tk_msg.rpc_resp, task->tk_status); + nfs41_sequence_done(task->tk_msg.rpc_resp); if (task->tk_status < 0) { dprintk("%s ERROR %d\n", __func__, task->tk_status); @@ -5184,7 +5182,7 @@ static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data) struct nfs4_sequence_res *res = &calldata->res.seq_res; dprintk("--> %s\n", __func__); - nfs41_sequence_done(clp, res, task->tk_status); + nfs41_sequence_done(res); switch (task->tk_status) { case 0: case -NFS4ERR_COMPLETE_ALREADY: From aa5190d0ed7d042c6d7d89fe8101558a912eee73 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:25 -0400 Subject: [PATCH 04/42] NFSv4: Kill nfs4_async_handle_error() abuses by NFSv4.1 Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 78 +++++++++++++++++++++++++---------------------- 1 file changed, 41 insertions(+), 37 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index fc653c3bc557..a00932c1215e 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -3457,9 +3457,11 @@ static int nfs4_proc_set_acl(struct inode *inode, const void *buf, size_t buflen } static int -_nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs_client *clp, struct nfs4_state *state) +nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state) { - if (!clp || task->tk_status >= 0) + struct nfs_client *clp = server->nfs_client; + + if (task->tk_status >= 0) return 0; switch(task->tk_status) { case -NFS4ERR_ADMIN_REVOKED: @@ -3491,8 +3493,7 @@ _nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, return -EAGAIN; #endif /* CONFIG_NFS_V4_1 */ case -NFS4ERR_DELAY: - if (server) - nfs_inc_server_stats(server, NFSIOS_DELAY); + nfs_inc_server_stats(server, NFSIOS_DELAY); case -NFS4ERR_GRACE: case -EKEYEXPIRED: rpc_delay(task, NFS4_POLL_RETRY_MAX); @@ -3513,12 +3514,6 @@ do_state_recovery: return -EAGAIN; } -static int -nfs4_async_handle_error(struct rpc_task *task, const struct nfs_server *server, struct nfs4_state *state) -{ - return _nfs4_async_handle_error(task, server, server->nfs_client, state); -} - int nfs4_proc_setclientid(struct nfs_client *clp, u32 program, unsigned short port, struct rpc_cred *cred, struct nfs4_setclientid_res *res) @@ -5088,6 +5083,19 @@ static void nfs41_sequence_release(void *data) kfree(calldata); } +static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp) +{ + switch(task->tk_status) { + case -NFS4ERR_DELAY: + case -EKEYEXPIRED: + rpc_delay(task, NFS4_POLL_RETRY_MAX); + return -EAGAIN; + default: + nfs4_schedule_state_recovery(clp); + } + return 0; +} + static void nfs41_sequence_call_done(struct rpc_task *task, void *data) { struct nfs4_sequence_data *calldata = data; @@ -5100,9 +5108,8 @@ static void nfs41_sequence_call_done(struct rpc_task *task, void *data) if (atomic_read(&clp->cl_count) == 1) goto out; - if (_nfs4_async_handle_error(task, NULL, clp, NULL) - == -EAGAIN) { - nfs_restart_rpc(task, clp); + if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) { + rpc_restart_call_prepare(task); return; } } @@ -5175,6 +5182,23 @@ static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data) rpc_call_start(task); } +static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp) +{ + switch(task->tk_status) { + case 0: + case -NFS4ERR_COMPLETE_ALREADY: + case -NFS4ERR_WRONG_CRED: /* What to do here? */ + break; + case -NFS4ERR_DELAY: + case -EKEYEXPIRED: + rpc_delay(task, NFS4_POLL_RETRY_MAX); + return -EAGAIN; + default: + nfs4_schedule_state_recovery(clp); + } + return 0; +} + static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data) { struct nfs4_reclaim_complete_data *calldata = data; @@ -5183,31 +5207,11 @@ static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data) dprintk("--> %s\n", __func__); nfs41_sequence_done(res); - switch (task->tk_status) { - case 0: - case -NFS4ERR_COMPLETE_ALREADY: - break; - case -NFS4ERR_BADSESSION: - case -NFS4ERR_DEADSESSION: - /* - * Handle the session error, but do not retry the operation, as - * we have no way of telling whether the clientid had to be - * reset before we got our reply. If reset, a new wave of - * reclaim operations will follow, containing their own reclaim - * complete. We don't want our retry to get on the way of - * recovery by incorrectly indicating to the server that we're - * done reclaiming state since the process had to be restarted. - */ - _nfs4_async_handle_error(task, NULL, clp, NULL); - break; - default: - if (_nfs4_async_handle_error( - task, NULL, clp, NULL) == -EAGAIN) { - rpc_restart_call_prepare(task); - return; - } - } + if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) { + rpc_restart_call_prepare(task); + return; + } dprintk("<-- %s\n", __func__); } From 71ac6da9944e2c9ec73ca64ca7bca74428858585 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:26 -0400 Subject: [PATCH 05/42] NFSv4.1: Merge the nfs41_proc_async_sequence() and nfs4_proc_sequence() Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 68 ++++++++++++++++++++++++++++++----------------- 1 file changed, 44 insertions(+), 24 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a00932c1215e..75847e4898a0 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5048,24 +5048,6 @@ int nfs4_init_session(struct nfs_server *server) /* * Renew the cl_session lease. */ -static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred) -{ - struct nfs4_sequence_args args; - struct nfs4_sequence_res res; - - struct rpc_message msg = { - .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE], - .rpc_argp = &args, - .rpc_resp = &res, - .rpc_cred = cred, - }; - - args.sa_cache_this = 0; - - return nfs4_call_sync_sequence(clp, clp->cl_rpcclient, &msg, &args, - &res, args.sa_cache_this, 1); -} - struct nfs4_sequence_data { struct nfs_client *clp; struct nfs4_sequence_args args; @@ -5139,29 +5121,67 @@ static const struct rpc_call_ops nfs41_sequence_ops = { .rpc_release = nfs41_sequence_release, }; -static int nfs41_proc_async_sequence(struct nfs_client *clp, - struct rpc_cred *cred) +static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred) { struct nfs4_sequence_data *calldata; struct rpc_message msg = { .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE], .rpc_cred = cred, }; + struct rpc_task_setup task_setup_data = { + .rpc_client = clp->cl_rpcclient, + .rpc_message = &msg, + .callback_ops = &nfs41_sequence_ops, + .flags = RPC_TASK_ASYNC | RPC_TASK_SOFT, + }; if (!atomic_inc_not_zero(&clp->cl_count)) - return -EIO; + return ERR_PTR(-EIO); calldata = kmalloc(sizeof(*calldata), GFP_NOFS); if (calldata == NULL) { nfs_put_client(clp); - return -ENOMEM; + return ERR_PTR(-ENOMEM); } calldata->res.sr_slotid = NFS4_MAX_SLOT_TABLE; msg.rpc_argp = &calldata->args; msg.rpc_resp = &calldata->res; calldata->clp = clp; + task_setup_data.callback_data = calldata; - return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_SOFT, - &nfs41_sequence_ops, calldata); + return rpc_run_task(&task_setup_data); +} + +static int nfs41_proc_async_sequence(struct nfs_client *clp, struct rpc_cred *cred) +{ + struct rpc_task *task; + int ret = 0; + + task = _nfs41_proc_sequence(clp, cred); + if (IS_ERR(task)) + ret = PTR_ERR(task); + else + rpc_put_task(task); + dprintk("<-- %s status=%d\n", __func__, ret); + return ret; +} + +static int nfs4_proc_sequence(struct nfs_client *clp, struct rpc_cred *cred) +{ + struct rpc_task *task; + int ret; + + task = _nfs41_proc_sequence(clp, cred); + if (IS_ERR(task)) { + ret = PTR_ERR(task); + goto out; + } + ret = rpc_wait_for_completion_task(task); + if (!ret) + ret = task->tk_status; + rpc_put_task(task); +out: + dprintk("<-- %s status=%d\n", __func__, ret); + return ret; } struct nfs4_reclaim_complete_data { From 035168ab39f66e4946d493f9ee20d11e154f332a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:26 -0400 Subject: [PATCH 06/42] NFSv4.1: Make nfs4_setup_sequence take a nfs_server argument In anticipation of the day when we have per-filesystem sessions, and also in order to allow the session to change in the event of a filesystem migration event. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 14 +++++++++++-- fs/nfs/nfs4proc.c | 51 ++++++++++++++++++++++++++--------------------- fs/nfs/read.c | 2 +- fs/nfs/unlink.c | 2 +- fs/nfs/write.c | 4 ++-- 5 files changed, 44 insertions(+), 29 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index c538c6106e16..9c0aa818b41e 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -223,7 +223,12 @@ extern int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name, extern struct nfs4_state_recovery_ops *nfs4_reboot_recovery_ops[]; extern struct nfs4_state_recovery_ops *nfs4_nograce_recovery_ops[]; #if defined(CONFIG_NFS_V4_1) -extern int nfs4_setup_sequence(struct nfs_client *clp, +static inline struct nfs4_session *nfs4_get_session(const struct nfs_server *server) +{ + return server->nfs_client->cl_session; +} + +extern int nfs4_setup_sequence(const struct nfs_server *server, struct nfs4_sequence_args *args, struct nfs4_sequence_res *res, int cache_reply, struct rpc_task *task); extern void nfs4_destroy_session(struct nfs4_session *session); @@ -234,7 +239,12 @@ extern int nfs4_init_session(struct nfs_server *server); extern int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo); #else /* CONFIG_NFS_v4_1 */ -static inline int nfs4_setup_sequence(struct nfs_client *clp, +static inline struct nfs4_session *nfs4_get_session(const struct nfs_server *server) +{ + return NULL; +} + +static inline int nfs4_setup_sequence(const struct nfs_server *server, struct nfs4_sequence_args *args, struct nfs4_sequence_res *res, int cache_reply, struct rpc_task *task) { diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 75847e4898a0..0168007b78d2 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -531,20 +531,25 @@ static int nfs41_setup_sequence(struct nfs4_session *session, return 0; } -int nfs4_setup_sequence(struct nfs_client *clp, +int nfs4_setup_sequence(const struct nfs_server *server, struct nfs4_sequence_args *args, struct nfs4_sequence_res *res, int cache_reply, struct rpc_task *task) { + struct nfs4_session *session = nfs4_get_session(server); int ret = 0; - dprintk("--> %s clp %p session %p sr_slotid %d\n", - __func__, clp, clp->cl_session, res->sr_slotid); - - if (!nfs4_has_session(clp)) + if (session == NULL) { + args->sa_session = NULL; + res->sr_session = NULL; goto out; - ret = nfs41_setup_sequence(clp->cl_session, args, res, cache_reply, + } + + dprintk("--> %s clp %p session %p sr_slotid %d\n", + __func__, session->clp, session, res->sr_slotid); + + ret = nfs41_setup_sequence(session, args, res, cache_reply, task); out: dprintk("<-- %s status=%d\n", __func__, ret); @@ -552,7 +557,7 @@ out: } struct nfs41_call_sync_data { - struct nfs_client *clp; + const struct nfs_server *seq_server; struct nfs4_sequence_args *seq_args; struct nfs4_sequence_res *seq_res; int cache_reply; @@ -562,9 +567,9 @@ static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata) { struct nfs41_call_sync_data *data = calldata; - dprintk("--> %s data->clp->cl_session %p\n", __func__, - data->clp->cl_session); - if (nfs4_setup_sequence(data->clp, data->seq_args, + dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server); + + if (nfs4_setup_sequence(data->seq_server, data->seq_args, data->seq_res, data->cache_reply, task)) return; rpc_call_start(task); @@ -593,8 +598,7 @@ struct rpc_call_ops nfs41_call_priv_sync_ops = { .rpc_call_done = nfs41_call_sync_done, }; -static int nfs4_call_sync_sequence(struct nfs_client *clp, - struct rpc_clnt *clnt, +static int nfs4_call_sync_sequence(struct nfs_server *server, struct rpc_message *msg, struct nfs4_sequence_args *args, struct nfs4_sequence_res *res, @@ -604,13 +608,13 @@ static int nfs4_call_sync_sequence(struct nfs_client *clp, int ret; struct rpc_task *task; struct nfs41_call_sync_data data = { - .clp = clp, + .seq_server = server, .seq_args = args, .seq_res = res, .cache_reply = cache_reply, }; struct rpc_task_setup task_setup = { - .rpc_client = clnt, + .rpc_client = server->client, .rpc_message = msg, .callback_ops = &nfs41_call_sync_ops, .callback_data = &data @@ -635,8 +639,7 @@ int _nfs4_call_sync_session(struct nfs_server *server, struct nfs4_sequence_res *res, int cache_reply) { - return nfs4_call_sync_sequence(server->nfs_client, server->client, - msg, args, res, cache_reply, 0); + return nfs4_call_sync_sequence(server, msg, args, res, cache_reply, 0); } #endif /* CONFIG_NFS_V4_1 */ @@ -1355,7 +1358,7 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) nfs_copy_fh(&data->o_res.fh, data->o_arg.fh); } data->timestamp = jiffies; - if (nfs4_setup_sequence(data->o_arg.server->nfs_client, + if (nfs4_setup_sequence(data->o_arg.server, &data->o_arg.seq_args, &data->o_res.seq_res, 1, task)) return; @@ -1896,7 +1899,7 @@ static void nfs4_close_prepare(struct rpc_task *task, void *data) nfs_fattr_init(calldata->res.fattr); calldata->timestamp = jiffies; - if (nfs4_setup_sequence((NFS_SERVER(calldata->inode))->nfs_client, + if (nfs4_setup_sequence(NFS_SERVER(calldata->inode), &calldata->arg.seq_args, &calldata->res.seq_res, 1, task)) return; @@ -3660,7 +3663,7 @@ static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data) d_data = (struct nfs4_delegreturndata *)data; - if (nfs4_setup_sequence(d_data->res.server->nfs_client, + if (nfs4_setup_sequence(d_data->res.server, &d_data->args.seq_args, &d_data->res.seq_res, 1, task)) return; @@ -3915,7 +3918,7 @@ static void nfs4_locku_prepare(struct rpc_task *task, void *data) return; } calldata->timestamp = jiffies; - if (nfs4_setup_sequence(calldata->server->nfs_client, + if (nfs4_setup_sequence(calldata->server, &calldata->arg.seq_args, &calldata->res.seq_res, 1, task)) return; @@ -4070,7 +4073,8 @@ static void nfs4_lock_prepare(struct rpc_task *task, void *calldata) } else data->arg.new_lock_owner = 0; data->timestamp = jiffies; - if (nfs4_setup_sequence(data->server->nfs_client, &data->arg.seq_args, + if (nfs4_setup_sequence(data->server, + &data->arg.seq_args, &data->res.seq_res, 1, task)) return; rpc_call_start(task); @@ -5110,7 +5114,7 @@ static void nfs41_sequence_prepare(struct rpc_task *task, void *data) args = task->tk_msg.rpc_argp; res = task->tk_msg.rpc_resp; - if (nfs4_setup_sequence(clp, args, res, 0, task)) + if (nfs41_setup_sequence(clp->cl_session, args, res, 0, task)) return; rpc_call_start(task); } @@ -5195,7 +5199,8 @@ static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data) struct nfs4_reclaim_complete_data *calldata = data; rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED); - if (nfs4_setup_sequence(calldata->clp, &calldata->arg.seq_args, + if (nfs41_setup_sequence(calldata->clp->cl_session, + &calldata->arg.seq_args, &calldata->res.seq_res, 0, task)) return; diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 6e2b06e6ca79..5a33a92e8168 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -410,7 +410,7 @@ void nfs_read_prepare(struct rpc_task *task, void *calldata) { struct nfs_read_data *data = calldata; - if (nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client, + if (nfs4_setup_sequence(NFS_SERVER(data->inode), &data->args.seq_args, &data->res.seq_res, 0, task)) return; diff --git a/fs/nfs/unlink.c b/fs/nfs/unlink.c index a2242af6a17d..2f84adaad427 100644 --- a/fs/nfs/unlink.c +++ b/fs/nfs/unlink.c @@ -110,7 +110,7 @@ void nfs_unlink_prepare(struct rpc_task *task, void *calldata) struct nfs_unlinkdata *data = calldata; struct nfs_server *server = NFS_SERVER(data->dir); - if (nfs4_setup_sequence(server->nfs_client, &data->args.seq_args, + if (nfs4_setup_sequence(server, &data->args.seq_args, &data->res.seq_res, 1, task)) return; rpc_call_start(task); diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 91679e2631ee..03df22822c4c 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1036,9 +1036,9 @@ out: void nfs_write_prepare(struct rpc_task *task, void *calldata) { struct nfs_write_data *data = calldata; - struct nfs_client *clp = (NFS_SERVER(data->inode))->nfs_client; - if (nfs4_setup_sequence(clp, &data->args.seq_args, + if (nfs4_setup_sequence(NFS_SERVER(data->inode), + &data->args.seq_args, &data->res.seq_res, 1, task)) return; rpc_call_start(task); From df8964554a4e19c8ddcc4d9c642c4d267662d770 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:26 -0400 Subject: [PATCH 07/42] NFSv41: Further cleanup for nfs4_sequence_done Instead of testing if the nfs_client has a session, we should be testing if the struct nfs4_sequence_res was set up with one. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 0168007b78d2..a896bfbd520a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -430,6 +430,13 @@ out: nfs41_sequence_free_slot(res); } +static void nfs4_sequence_done(const struct nfs_server *server, + struct nfs4_sequence_res *res, int rpc_status) +{ + if (res->sr_session != NULL) + nfs41_sequence_done(res); +} + /* * nfs4_find_slot - efficiently look for a free slot * @@ -642,6 +649,11 @@ int _nfs4_call_sync_session(struct nfs_server *server, return nfs4_call_sync_sequence(server, msg, args, res, cache_reply, 0); } +#else +static void nfs4_sequence_done(const struct nfs_server *server, + struct nfs4_sequence_res *res, int rpc_status) +{ +} #endif /* CONFIG_NFS_V4_1 */ int _nfs4_call_sync(struct nfs_server *server, @@ -658,15 +670,6 @@ int _nfs4_call_sync(struct nfs_server *server, (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \ &(res)->seq_res, (cache_reply)) -static void nfs4_sequence_done(const struct nfs_server *server, - struct nfs4_sequence_res *res, int rpc_status) -{ -#ifdef CONFIG_NFS_V4_1 - if (nfs4_has_session(server->nfs_client)) - nfs41_sequence_done(res); -#endif /* CONFIG_NFS_V4_1 */ -} - static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo) { struct nfs_inode *nfsi = NFS_I(dir); From a2118c33aad6c447ad5e0a60cfaea3939b52ce0a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:26 -0400 Subject: [PATCH 08/42] NFSv41: Don't store session state in the nfs_client->cl_state Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 5 ++++- fs/nfs/nfs4proc.c | 4 ++-- fs/nfs/nfs4state.c | 6 ++++-- 3 files changed, 10 insertions(+), 5 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 9c0aa818b41e..bb1e95530699 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -45,10 +45,13 @@ enum nfs4_client_state { NFS4CLNT_RECLAIM_NOGRACE, NFS4CLNT_DELEGRETURN, NFS4CLNT_SESSION_RESET, - NFS4CLNT_SESSION_DRAINING, NFS4CLNT_RECALL_SLOT, }; +enum nfs4_session_state { + NFS4_SESSION_DRAINING, +}; + /* * struct rpc_sequence ensures that RPC calls are sent in the exact * order that they appear on the list. diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a896bfbd520a..fc972c6f1ce9 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -356,7 +356,7 @@ static void nfs41_check_drain_session_complete(struct nfs4_session *ses) { struct rpc_task *task; - if (!test_bit(NFS4CLNT_SESSION_DRAINING, &ses->clp->cl_state)) { + if (!test_bit(NFS4_SESSION_DRAINING, &ses->session_state)) { task = rpc_wake_up_next(&ses->fc_slot_table.slot_tbl_waitq); if (task) rpc_task_set_priority(task, RPC_PRIORITY_PRIVILEGED); @@ -489,7 +489,7 @@ static int nfs41_setup_sequence(struct nfs4_session *session, tbl = &session->fc_slot_table; spin_lock(&tbl->slot_tbl_lock); - if (test_bit(NFS4CLNT_SESSION_DRAINING, &session->clp->cl_state) && + if (test_bit(NFS4_SESSION_DRAINING, &session->session_state) && !rpc_task_has_priority(task, RPC_PRIORITY_PRIVILEGED)) { /* * The state manager will wait until the slot table is empty. diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 34acf5926fdc..4538c56a7f05 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -145,7 +145,9 @@ static void nfs4_end_drain_session(struct nfs_client *clp) struct nfs4_session *ses = clp->cl_session; int max_slots; - if (test_and_clear_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state)) { + if (ses == NULL) + return; + if (test_and_clear_bit(NFS4_SESSION_DRAINING, &ses->session_state)) { spin_lock(&ses->fc_slot_table.slot_tbl_lock); max_slots = ses->fc_slot_table.max_slots; while (max_slots--) { @@ -167,7 +169,7 @@ static int nfs4_begin_drain_session(struct nfs_client *clp) struct nfs4_slot_table *tbl = &ses->fc_slot_table; spin_lock(&tbl->slot_tbl_lock); - set_bit(NFS4CLNT_SESSION_DRAINING, &clp->cl_state); + set_bit(NFS4_SESSION_DRAINING, &ses->session_state); if (tbl->highest_used_slotid != -1) { INIT_COMPLETION(ses->complete); spin_unlock(&tbl->slot_tbl_lock); From 97dc135947181a6670949a480da56c3ebf8d3715 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:26 -0400 Subject: [PATCH 09/42] NFSv41: Clean up the NFSv4.1 minor version specific operations Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 14 ++++++-------- fs/nfs/nfs4_fs.h | 11 +++++++++++ fs/nfs/nfs4proc.c | 21 ++++++++++++++++++++- include/linux/nfs_fs_sb.h | 7 ++----- 4 files changed, 39 insertions(+), 14 deletions(-) diff --git a/fs/nfs/client.c b/fs/nfs/client.c index d25b5257b7a1..1df708fd4205 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -150,6 +150,7 @@ static struct nfs_client *nfs_alloc_client(const struct nfs_client_initdata *cl_ clp->cl_boot_time = CURRENT_TIME; clp->cl_state = 1 << NFS4CLNT_LEASE_EXPIRED; clp->cl_minorversion = cl_init->minorversion; + clp->cl_mvops = nfs_v4_minor_ops[cl_init->minorversion]; #endif cred = rpc_lookup_machine_cred(); if (!IS_ERR(cred)) @@ -178,7 +179,7 @@ static void nfs4_clear_client_minor_version(struct nfs_client *clp) clp->cl_session = NULL; } - clp->cl_call_sync = _nfs4_call_sync; + clp->cl_mvops = nfs_v4_minor_ops[0]; #endif /* CONFIG_NFS_V4_1 */ } @@ -188,7 +189,7 @@ static void nfs4_clear_client_minor_version(struct nfs_client *clp) static void nfs4_destroy_callback(struct nfs_client *clp) { if (__test_and_clear_bit(NFS_CS_CALLBACK, &clp->cl_res_state)) - nfs_callback_down(clp->cl_minorversion); + nfs_callback_down(clp->cl_mvops->minor_version); } static void nfs4_shutdown_client(struct nfs_client *clp) @@ -1126,7 +1127,7 @@ static int nfs4_init_callback(struct nfs_client *clp) return error; } - error = nfs_callback_up(clp->cl_minorversion, + error = nfs_callback_up(clp->cl_mvops->minor_version, clp->cl_rpcclient->cl_xprt); if (error < 0) { dprintk("%s: failed to start callback. Error = %d\n", @@ -1143,10 +1144,8 @@ static int nfs4_init_callback(struct nfs_client *clp) */ static int nfs4_init_client_minor_version(struct nfs_client *clp) { - clp->cl_call_sync = _nfs4_call_sync; - #if defined(CONFIG_NFS_V4_1) - if (clp->cl_minorversion) { + if (clp->cl_mvops->minor_version) { struct nfs4_session *session = NULL; /* * Create the session and mark it expired. @@ -1158,7 +1157,6 @@ static int nfs4_init_client_minor_version(struct nfs_client *clp) return -ENOMEM; clp->cl_session = session; - clp->cl_call_sync = _nfs4_call_sync_session; } #endif /* CONFIG_NFS_V4_1 */ @@ -1454,7 +1452,7 @@ struct nfs_server *nfs4_create_referral_server(struct nfs_clone_mount *data, data->authflavor, parent_server->client->cl_xprt->prot, parent_server->client->cl_timeout, - parent_client->cl_minorversion); + parent_client->cl_mvops->minor_version); if (error < 0) goto error; diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index bb1e95530699..5b01705e30f9 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -52,6 +52,16 @@ enum nfs4_session_state { NFS4_SESSION_DRAINING, }; +struct nfs4_minor_version_ops { + u32 minor_version; + + int (*call_sync)(struct nfs_server *server, + struct rpc_message *msg, + struct nfs4_sequence_args *args, + struct nfs4_sequence_res *res, + int cache_reply); +}; + /* * struct rpc_sequence ensures that RPC calls are sent in the exact * order that they appear on the list. @@ -260,6 +270,7 @@ static inline int nfs4_init_session(struct nfs_server *server) } #endif /* CONFIG_NFS_V4_1 */ +extern const struct nfs4_minor_version_ops *nfs_v4_minor_ops[]; extern struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[]; extern const u32 nfs4_fattr_bitmap[2]; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index fc972c6f1ce9..a938daf333da 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -667,7 +667,7 @@ int _nfs4_call_sync(struct nfs_server *server, } #define nfs4_call_sync(server, msg, args, res, cache_reply) \ - (server)->nfs_client->cl_call_sync((server), (msg), &(args)->seq_args, \ + (server)->nfs_client->cl_mvops->call_sync((server), (msg), &(args)->seq_args, \ &(res)->seq_res, (cache_reply)) static void update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo) @@ -5353,6 +5353,18 @@ struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = { }; #endif +static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = { + .minor_version = 0, + .call_sync = _nfs4_call_sync, +}; + +#if defined(CONFIG_NFS_V4_1) +static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = { + .minor_version = 1, + .call_sync = _nfs4_call_sync_session, +}; +#endif + /* * Per minor version reboot and network partition recovery ops */ @@ -5378,6 +5390,13 @@ struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[] = { #endif }; +const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = { + [0] = &nfs_v4_0_minor_ops, +#if defined(CONFIG_NFS_V4_1) + [1] = &nfs_v4_1_minor_ops, +#endif +}; + static const struct inode_operations nfs4_file_inode_operations = { .permission = nfs_permission, .getattr = nfs_getattr, diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index d6e10a4c06e5..c82ee7cd6288 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -15,6 +15,7 @@ struct nlm_host; struct nfs4_sequence_args; struct nfs4_sequence_res; struct nfs_server; +struct nfs4_minor_version_ops; /* * The nfs_client identifies our client state to the server. @@ -70,11 +71,7 @@ struct nfs_client { */ char cl_ipaddr[48]; unsigned char cl_id_uniquifier; - int (* cl_call_sync)(struct nfs_server *server, - struct rpc_message *msg, - struct nfs4_sequence_args *args, - struct nfs4_sequence_res *res, - int cache_reply); + const struct nfs4_minor_version_ops *cl_mvops; #endif /* CONFIG_NFS_V4 */ #ifdef CONFIG_NFS_V4_1 From c48f4f3541e67881c9eb7c46e052f5ece48ef530 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:27 -0400 Subject: [PATCH 10/42] NFSv41: Convert the various reboot recovery ops etc to minor version ops Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 6 +++--- fs/nfs/nfs4proc.c | 31 ++++++------------------------- fs/nfs/nfs4renewd.c | 4 ++-- fs/nfs/nfs4state.c | 15 +++++++-------- 4 files changed, 18 insertions(+), 38 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 5b01705e30f9..1ff0ea2a8461 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -60,6 +60,9 @@ struct nfs4_minor_version_ops { struct nfs4_sequence_args *args, struct nfs4_sequence_res *res, int cache_reply); + const struct nfs4_state_recovery_ops *reboot_recovery_ops; + const struct nfs4_state_recovery_ops *nograce_recovery_ops; + const struct nfs4_state_maintenance_ops *state_renewal_ops; }; /* @@ -233,8 +236,6 @@ extern int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fh extern int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name, struct nfs4_fs_locations *fs_locations, struct page *page); -extern struct nfs4_state_recovery_ops *nfs4_reboot_recovery_ops[]; -extern struct nfs4_state_recovery_ops *nfs4_nograce_recovery_ops[]; #if defined(CONFIG_NFS_V4_1) static inline struct nfs4_session *nfs4_get_session(const struct nfs_server *server) { @@ -271,7 +272,6 @@ static inline int nfs4_init_session(struct nfs_server *server) #endif /* CONFIG_NFS_V4_1 */ extern const struct nfs4_minor_version_ops *nfs_v4_minor_ops[]; -extern struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[]; extern const u32 nfs4_fattr_bitmap[2]; extern const u32 nfs4_statfs_bitmap[2]; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index a938daf333da..9f222f57e75b 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5356,40 +5356,21 @@ struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = { static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = { .minor_version = 0, .call_sync = _nfs4_call_sync, + .reboot_recovery_ops = &nfs40_reboot_recovery_ops, + .nograce_recovery_ops = &nfs40_nograce_recovery_ops, + .state_renewal_ops = &nfs40_state_renewal_ops, }; #if defined(CONFIG_NFS_V4_1) static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = { .minor_version = 1, .call_sync = _nfs4_call_sync_session, + .reboot_recovery_ops = &nfs41_reboot_recovery_ops, + .nograce_recovery_ops = &nfs41_nograce_recovery_ops, + .state_renewal_ops = &nfs41_state_renewal_ops, }; #endif -/* - * Per minor version reboot and network partition recovery ops - */ - -struct nfs4_state_recovery_ops *nfs4_reboot_recovery_ops[] = { - &nfs40_reboot_recovery_ops, -#if defined(CONFIG_NFS_V4_1) - &nfs41_reboot_recovery_ops, -#endif -}; - -struct nfs4_state_recovery_ops *nfs4_nograce_recovery_ops[] = { - &nfs40_nograce_recovery_ops, -#if defined(CONFIG_NFS_V4_1) - &nfs41_nograce_recovery_ops, -#endif -}; - -struct nfs4_state_maintenance_ops *nfs4_state_renewal_ops[] = { - &nfs40_state_renewal_ops, -#if defined(CONFIG_NFS_V4_1) - &nfs41_state_renewal_ops, -#endif -}; - const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = { [0] = &nfs_v4_0_minor_ops, #if defined(CONFIG_NFS_V4_1) diff --git a/fs/nfs/nfs4renewd.c b/fs/nfs/nfs4renewd.c index d87f10327b72..72b6c580af13 100644 --- a/fs/nfs/nfs4renewd.c +++ b/fs/nfs/nfs4renewd.c @@ -54,14 +54,14 @@ void nfs4_renew_state(struct work_struct *work) { - struct nfs4_state_maintenance_ops *ops; + const struct nfs4_state_maintenance_ops *ops; struct nfs_client *clp = container_of(work, struct nfs_client, cl_renewd.work); struct rpc_cred *cred; long lease; unsigned long last, now; - ops = nfs4_state_renewal_ops[clp->cl_minorversion]; + ops = clp->cl_mvops->state_renewal_ops; dprintk("%s: start\n", __func__); /* Are there any active superblocks? */ if (list_empty(&clp->cl_superblocks)) diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 4538c56a7f05..c3c6b8aeb209 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -1122,8 +1122,7 @@ static void nfs4_state_end_reclaim_reboot(struct nfs_client *clp) if (!test_and_clear_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) return; - nfs4_reclaim_complete(clp, - nfs4_reboot_recovery_ops[clp->cl_minorversion]); + nfs4_reclaim_complete(clp, clp->cl_mvops->reboot_recovery_ops); for (pos = rb_first(&clp->cl_state_owners); pos != NULL; pos = rb_next(pos)) { sp = rb_entry(pos, struct nfs4_state_owner, so_client_node); @@ -1213,8 +1212,8 @@ restart: static int nfs4_check_lease(struct nfs_client *clp) { struct rpc_cred *cred; - struct nfs4_state_maintenance_ops *ops = - nfs4_state_renewal_ops[clp->cl_minorversion]; + const struct nfs4_state_maintenance_ops *ops = + clp->cl_mvops->state_renewal_ops; int status = -NFS4ERR_EXPIRED; /* Is the client already known to have an expired lease? */ @@ -1237,8 +1236,8 @@ out: static int nfs4_reclaim_lease(struct nfs_client *clp) { struct rpc_cred *cred; - struct nfs4_state_recovery_ops *ops = - nfs4_reboot_recovery_ops[clp->cl_minorversion]; + const struct nfs4_state_recovery_ops *ops = + clp->cl_mvops->reboot_recovery_ops; int status = -ENOENT; cred = ops->get_clid_cred(clp); @@ -1446,7 +1445,7 @@ static void nfs4_state_manager(struct nfs_client *clp) /* First recover reboot state... */ if (test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) { status = nfs4_do_reclaim(clp, - nfs4_reboot_recovery_ops[clp->cl_minorversion]); + clp->cl_mvops->reboot_recovery_ops); if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) || test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state)) continue; @@ -1460,7 +1459,7 @@ static void nfs4_state_manager(struct nfs_client *clp) /* Now recover expired state... */ if (test_and_clear_bit(NFS4CLNT_RECLAIM_NOGRACE, &clp->cl_state)) { status = nfs4_do_reclaim(clp, - nfs4_nograce_recovery_ops[clp->cl_minorversion]); + clp->cl_mvops->nograce_recovery_ops); if (test_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state) || test_bit(NFS4CLNT_SESSION_RESET, &clp->cl_state) || test_bit(NFS4CLNT_RECLAIM_REBOOT, &clp->cl_state)) From e047a10c1293ee0ab20258154e7f3ebf8ad502d6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:27 -0400 Subject: [PATCH 11/42] NFSv41: Fix nfs_async_inode_return_delegation() ugliness Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 13 +------------ fs/nfs/delegation.c | 6 ++---- fs/nfs/delegation.h | 4 +--- fs/nfs/nfs4_fs.h | 2 ++ fs/nfs/nfs4proc.c | 2 ++ 5 files changed, 8 insertions(+), 19 deletions(-) diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index a08770a7e857..7445dd0ae3f3 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -62,16 +62,6 @@ out: return res->status; } -static int (*nfs_validate_delegation_stateid(struct nfs_client *clp))(struct nfs_delegation *, const nfs4_stateid *) -{ -#if defined(CONFIG_NFS_V4_1) - if (clp->cl_minorversion > 0) - return nfs41_validate_delegation_stateid; -#endif - return nfs4_validate_delegation_stateid; -} - - __be32 nfs4_callback_recall(struct cb_recallargs *args, void *dummy) { struct nfs_client *clp; @@ -92,8 +82,7 @@ __be32 nfs4_callback_recall(struct cb_recallargs *args, void *dummy) inode = nfs_delegation_find_inode(clp, &args->fh); if (inode != NULL) { /* Set up a helper thread to actually return the delegation */ - switch (nfs_async_inode_return_delegation(inode, &args->stateid, - nfs_validate_delegation_stateid(clp))) { + switch (nfs_async_inode_return_delegation(inode, &args->stateid)) { case 0: res = 0; break; diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index 301634543974..f34f4ac52b81 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -471,9 +471,7 @@ void nfs_expire_unreferenced_delegations(struct nfs_client *clp) /* * Asynchronous delegation recall! */ -int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid, - int (*validate_stateid)(struct nfs_delegation *delegation, - const nfs4_stateid *stateid)) +int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid) { struct nfs_client *clp = NFS_SERVER(inode)->nfs_client; struct nfs_delegation *delegation; @@ -481,7 +479,7 @@ int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *s rcu_read_lock(); delegation = rcu_dereference(NFS_I(inode)->delegation); - if (!validate_stateid(delegation, stateid)) { + if (!clp->cl_mvops->validate_stateid(delegation, stateid)) { rcu_read_unlock(); return -ENOENT; } diff --git a/fs/nfs/delegation.h b/fs/nfs/delegation.h index 69e7b8140122..2026304bda19 100644 --- a/fs/nfs/delegation.h +++ b/fs/nfs/delegation.h @@ -34,9 +34,7 @@ enum { int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res); int nfs_inode_return_delegation(struct inode *inode); -int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid, - int (*validate_stateid)(struct nfs_delegation *delegation, - const nfs4_stateid *stateid)); +int nfs_async_inode_return_delegation(struct inode *inode, const nfs4_stateid *stateid); void nfs_inode_return_delegation_noreclaim(struct inode *inode); struct inode *nfs_delegation_find_inode(struct nfs_client *clp, const struct nfs_fh *fhandle); diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 1ff0ea2a8461..04eff16b34d3 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -60,6 +60,8 @@ struct nfs4_minor_version_ops { struct nfs4_sequence_args *args, struct nfs4_sequence_res *res, int cache_reply); + int (*validate_stateid)(struct nfs_delegation *, + const nfs4_stateid *); const struct nfs4_state_recovery_ops *reboot_recovery_ops; const struct nfs4_state_recovery_ops *nograce_recovery_ops; const struct nfs4_state_maintenance_ops *state_renewal_ops; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 9f222f57e75b..d1ab0c36e939 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -5356,6 +5356,7 @@ struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = { static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = { .minor_version = 0, .call_sync = _nfs4_call_sync, + .validate_stateid = nfs4_validate_delegation_stateid, .reboot_recovery_ops = &nfs40_reboot_recovery_ops, .nograce_recovery_ops = &nfs40_nograce_recovery_ops, .state_renewal_ops = &nfs40_state_renewal_ops, @@ -5365,6 +5366,7 @@ static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = { static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = { .minor_version = 1, .call_sync = _nfs4_call_sync_session, + .validate_stateid = nfs41_validate_delegation_stateid, .reboot_recovery_ops = &nfs41_reboot_recovery_ops, .nograce_recovery_ops = &nfs41_nograce_recovery_ops, .state_renewal_ops = &nfs41_state_renewal_ops, From a4432345352c2be157ed844603147ac2c82f209c Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:27 -0400 Subject: [PATCH 12/42] NFSv41: Deprecate nfs_client->cl_minorversion Signed-off-by: Trond Myklebust --- fs/nfs/nfs4xdr.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 65c8dae4b267..1f7781d636ae 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1172,7 +1172,7 @@ static inline void encode_createmode(struct xdr_stream *xdr, const struct nfs_op break; default: clp = arg->server->nfs_client; - if (clp->cl_minorversion > 0) { + if (clp->cl_mvops->minor_version > 0) { if (nfs4_has_persistent_session(clp)) { *p = cpu_to_be32(NFS4_CREATE_GUARDED); encode_attrs(xdr, arg->u.attrs, arg->server); @@ -1704,7 +1704,7 @@ static u32 nfs4_xdr_minorversion(const struct nfs4_sequence_args *args) { #if defined(CONFIG_NFS_V4_1) if (args->sa_session) - return args->sa_session->clp->cl_minorversion; + return args->sa_session->clp->cl_mvops->minor_version; #endif /* CONFIG_NFS_V4_1 */ return 0; } @@ -2395,7 +2395,7 @@ static int nfs4_xdr_enc_exchange_id(struct rpc_rqst *req, uint32_t *p, { struct xdr_stream xdr; struct compound_hdr hdr = { - .minorversion = args->client->cl_minorversion, + .minorversion = args->client->cl_mvops->minor_version, }; xdr_init_encode(&xdr, &req->rq_snd_buf, p); @@ -2413,7 +2413,7 @@ static int nfs4_xdr_enc_create_session(struct rpc_rqst *req, uint32_t *p, { struct xdr_stream xdr; struct compound_hdr hdr = { - .minorversion = args->client->cl_minorversion, + .minorversion = args->client->cl_mvops->minor_version, }; xdr_init_encode(&xdr, &req->rq_snd_buf, p); @@ -2431,7 +2431,7 @@ static int nfs4_xdr_enc_destroy_session(struct rpc_rqst *req, uint32_t *p, { struct xdr_stream xdr; struct compound_hdr hdr = { - .minorversion = session->clp->cl_minorversion, + .minorversion = session->clp->cl_mvops->minor_version, }; xdr_init_encode(&xdr, &req->rq_snd_buf, p); From d77d76ffb638bd013782138cca6d8f4918c5afd6 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:27 -0400 Subject: [PATCH 13/42] NFSv41: Clean up exclusive create Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 17 ++++++----------- include/linux/nfs_xdr.h | 6 ++++-- 2 files changed, 10 insertions(+), 13 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index d1ab0c36e939..5d87563d0c1a 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -744,19 +744,14 @@ static struct nfs4_opendata *nfs4_opendata_alloc(struct path *path, p->o_arg.server = server; p->o_arg.bitmask = server->attr_bitmask; p->o_arg.claim = NFS4_OPEN_CLAIM_NULL; - if (flags & O_EXCL) { - if (nfs4_has_persistent_session(server->nfs_client)) { - /* GUARDED */ - p->o_arg.u.attrs = &p->attrs; - memcpy(&p->attrs, attrs, sizeof(p->attrs)); - } else { /* EXCLUSIVE4_1 */ - u32 *s = (u32 *) p->o_arg.u.verifier.data; - s[0] = jiffies; - s[1] = current->pid; - } - } else if (flags & O_CREAT) { + if (flags & O_CREAT) { + u32 *s; + p->o_arg.u.attrs = &p->attrs; memcpy(&p->attrs, attrs, sizeof(p->attrs)); + s = (u32 *) p->o_arg.u.verifier.data; + s[0] = jiffies; + s[1] = current->pid; } p->c_arg.fh = &p->o_res.fh; p->c_arg.stateid = &p->o_res.stateid; diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 51914d7d6cc4..a319cb926abf 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -196,8 +196,10 @@ struct nfs_openargs { __u64 clientid; __u64 id; union { - struct iattr * attrs; /* UNCHECKED, GUARDED */ - nfs4_verifier verifier; /* EXCLUSIVE */ + struct { + struct iattr * attrs; /* UNCHECKED, GUARDED */ + nfs4_verifier verifier; /* EXCLUSIVE */ + }; nfs4_stateid delegation; /* CLAIM_DELEGATE_CUR */ fmode_t delegation_type; /* CLAIM_PREVIOUS */ } u; From fe74ba3a8db003410e48863a5cafa6ac90674540 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:27 -0400 Subject: [PATCH 14/42] NFSv41: Cleanup for nfs4_alloc_session. There is no reason to change the nfs_client state every time we allocate a new session. Move that line into nfs4_init_client_minor_version. Signed-off-by: Trond Myklebust --- fs/nfs/client.c | 7 +++++++ fs/nfs/nfs4proc.c | 7 ------- 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 1df708fd4205..4e7df2adb212 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -1157,6 +1157,13 @@ static int nfs4_init_client_minor_version(struct nfs_client *clp) return -ENOMEM; clp->cl_session = session; + /* + * The create session reply races with the server back + * channel probe. Mark the client NFS_CS_SESSION_INITING + * so that the client back channel can find the + * nfs_client struct + */ + clp->cl_cons_state = NFS_CS_SESSION_INITING; } #endif /* CONFIG_NFS_V4_1 */ diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 5d87563d0c1a..b4132fe14282 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -4795,13 +4795,6 @@ struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp) if (!session) return NULL; - /* - * The create session reply races with the server back - * channel probe. Mark the client NFS_CS_SESSION_INITING - * so that the client back channel can find the - * nfs_client struct - */ - clp->cl_cons_state = NFS_CS_SESSION_INITING; init_completion(&session->complete); tbl = &session->fc_slot_table; From 1055d76d91e69c8ea9fb748db4d4a53b6384df31 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 16 Jun 2010 09:52:27 -0400 Subject: [PATCH 15/42] NFSv4.1: There is no need to init the session more than once... Set up a flag to ensure that is indeed the case. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 1 + fs/nfs/nfs4proc.c | 7 ++++++- 2 files changed, 7 insertions(+), 1 deletion(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index 04eff16b34d3..a986f13e8039 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -49,6 +49,7 @@ enum nfs4_client_state { }; enum nfs4_session_state { + NFS4_SESSION_INITING, NFS4_SESSION_DRAINING, }; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index b4132fe14282..243d6c72cfa1 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -4807,6 +4807,8 @@ struct nfs4_session *nfs4_alloc_session(struct nfs_client *clp) spin_lock_init(&tbl->slot_tbl_lock); rpc_init_wait_queue(&tbl->slot_tbl_waitq, "BackChannel Slot table"); + session->session_state = 1<clp = clp; return session; } @@ -5023,6 +5025,10 @@ int nfs4_init_session(struct nfs_server *server) if (!nfs4_has_session(clp)) return 0; + session = clp->cl_session; + if (!test_and_clear_bit(NFS4_SESSION_INITING, &session->session_state)) + return 0; + rsize = server->rsize; if (rsize == 0) rsize = NFS_MAX_FILE_IO_SIZE; @@ -5030,7 +5036,6 @@ int nfs4_init_session(struct nfs_server *server) if (wsize == 0) wsize = NFS_MAX_FILE_IO_SIZE; - session = clp->cl_session; session->fc_attrs.max_rqst_sz = wsize + nfs41_maxwrite_overhead; session->fc_attrs.max_resp_sz = rsize + nfs41_maxread_overhead; From 1f0e890dba5b0f543fea47732116b1c65d55614e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 24 Jun 2010 15:11:43 -0400 Subject: [PATCH 16/42] NFSv4: Clean up struct nfs4_state_owner The 'so_delegations' list appears to be unused. Also eliminate so_client. If we already have so_server, we can get to the nfs_client structure. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 2 -- fs/nfs/nfs4proc.c | 2 +- fs/nfs/nfs4state.c | 14 ++++++-------- 3 files changed, 7 insertions(+), 11 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index a986f13e8039..cee871471e8d 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -108,7 +108,6 @@ struct nfs_unique_id { */ struct nfs4_state_owner { struct nfs_unique_id so_owner_id; - struct nfs_client *so_client; struct nfs_server *so_server; struct rb_node so_client_node; @@ -118,7 +117,6 @@ struct nfs4_state_owner { atomic_t so_count; unsigned long so_flags; struct list_head so_states; - struct list_head so_delegations; struct nfs_seqid_counter so_seqid; struct rpc_sequence so_sequence; }; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 243d6c72cfa1..de9ff1505a24 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1350,7 +1350,7 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata) } /* Update sequence id. */ data->o_arg.id = sp->so_owner_id.id; - data->o_arg.clientid = sp->so_client->cl_clientid; + data->o_arg.clientid = sp->so_server->nfs_client->cl_clientid; if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS) { task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR]; nfs_copy_fh(&data->o_res.fh, data->o_arg.fh); diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index c3c6b8aeb209..13e17e32e3e4 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -373,7 +373,6 @@ nfs4_alloc_state_owner(void) return NULL; spin_lock_init(&sp->so_lock); INIT_LIST_HEAD(&sp->so_states); - INIT_LIST_HEAD(&sp->so_delegations); rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue"); sp->so_seqid.sequence = &sp->so_sequence; spin_lock_init(&sp->so_sequence.lock); @@ -386,7 +385,7 @@ static void nfs4_drop_state_owner(struct nfs4_state_owner *sp) { if (!RB_EMPTY_NODE(&sp->so_client_node)) { - struct nfs_client *clp = sp->so_client; + struct nfs_client *clp = sp->so_server->nfs_client; spin_lock(&clp->cl_lock); rb_erase(&sp->so_client_node, &clp->cl_state_owners); @@ -408,7 +407,6 @@ struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct new = nfs4_alloc_state_owner(); if (new == NULL) return NULL; - new->so_client = clp; new->so_server = server; new->so_cred = cred; spin_lock(&clp->cl_lock); @@ -425,7 +423,7 @@ struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct void nfs4_put_state_owner(struct nfs4_state_owner *sp) { - struct nfs_client *clp = sp->so_client; + struct nfs_client *clp = sp->so_server->nfs_client; struct rpc_cred *cred = sp->so_cred; if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock)) @@ -624,7 +622,7 @@ __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) { struct nfs4_lock_state *lsp; - struct nfs_client *clp = state->owner->so_client; + struct nfs_client *clp = state->owner->so_server->nfs_client; lsp = kzalloc(sizeof(*lsp), GFP_NOFS); if (lsp == NULL) @@ -645,7 +643,7 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f static void nfs4_free_lock_state(struct nfs4_lock_state *lsp) { - struct nfs_client *clp = lsp->ls_state->owner->so_client; + struct nfs_client *clp = lsp->ls_state->owner->so_server->nfs_client; spin_lock(&clp->cl_lock); nfs_free_unique_id(&clp->cl_lockowner_id, &lsp->ls_id); @@ -1043,11 +1041,11 @@ restart: case -NFS4ERR_BAD_STATEID: case -NFS4ERR_RECLAIM_BAD: case -NFS4ERR_RECLAIM_CONFLICT: - nfs4_state_mark_reclaim_nograce(sp->so_client, state); + nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state); break; case -NFS4ERR_EXPIRED: case -NFS4ERR_NO_GRACE: - nfs4_state_mark_reclaim_nograce(sp->so_client, state); + nfs4_state_mark_reclaim_nograce(sp->so_server->nfs_client, state); case -NFS4ERR_STALE_CLIENTID: case -NFS4ERR_BADSESSION: case -NFS4ERR_BADSLOT: From f11ac8db5d07b6e99d41ff4aa39d878ee5cef1c5 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 25 Jun 2010 16:35:53 -0400 Subject: [PATCH 17/42] NFSv4: Ensure that we track the NFSv4 lock state in read/write requests. This patch fixes bugzilla entry 14501: https://bugzilla.kernel.org/show_bug.cgi?id=14501 Signed-off-by: Trond Myklebust --- fs/nfs/direct.c | 29 +++++++++++++---- fs/nfs/inode.c | 70 +++++++++++++++++++++++++++++++++++++--- fs/nfs/nfs4xdr.c | 8 ++--- fs/nfs/pagelist.c | 8 ++++- fs/nfs/read.c | 1 + fs/nfs/write.c | 5 ++- include/linux/nfs_fs.h | 13 ++++++-- include/linux/nfs_page.h | 1 + include/linux/nfs_xdr.h | 2 ++ 9 files changed, 118 insertions(+), 19 deletions(-) diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c index ad4cd31d6050..064a80961677 100644 --- a/fs/nfs/direct.c +++ b/fs/nfs/direct.c @@ -69,6 +69,7 @@ struct nfs_direct_req { /* I/O parameters */ struct nfs_open_context *ctx; /* file open context info */ + struct nfs_lock_context *l_ctx; /* Lock context info */ struct kiocb * iocb; /* controlling i/o request */ struct inode * inode; /* target file of i/o */ @@ -160,6 +161,7 @@ static inline struct nfs_direct_req *nfs_direct_req_alloc(void) INIT_LIST_HEAD(&dreq->rewrite_list); dreq->iocb = NULL; dreq->ctx = NULL; + dreq->l_ctx = NULL; spin_lock_init(&dreq->lock); atomic_set(&dreq->io_count, 0); dreq->count = 0; @@ -173,6 +175,8 @@ static void nfs_direct_req_free(struct kref *kref) { struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref); + if (dreq->l_ctx != NULL) + nfs_put_lock_context(dreq->l_ctx); if (dreq->ctx != NULL) put_nfs_open_context(dreq->ctx); kmem_cache_free(nfs_direct_cachep, dreq); @@ -336,6 +340,7 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq, data->cred = msg.rpc_cred; data->args.fh = NFS_FH(inode); data->args.context = ctx; + data->args.lock_context = dreq->l_ctx; data->args.offset = pos; data->args.pgbase = pgbase; data->args.pages = data->pagevec; @@ -416,24 +421,28 @@ static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov, unsigned long nr_segs, loff_t pos) { - ssize_t result = 0; + ssize_t result = -ENOMEM; struct inode *inode = iocb->ki_filp->f_mapping->host; struct nfs_direct_req *dreq; dreq = nfs_direct_req_alloc(); - if (!dreq) - return -ENOMEM; + if (dreq == NULL) + goto out; dreq->inode = inode; dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); + dreq->l_ctx = nfs_get_lock_context(dreq->ctx); + if (dreq->l_ctx == NULL) + goto out_release; if (!is_sync_kiocb(iocb)) dreq->iocb = iocb; result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos); if (!result) result = nfs_direct_wait(dreq); +out_release: nfs_direct_req_release(dreq); - +out: return result; } @@ -574,6 +583,7 @@ static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq) data->args.offset = 0; data->args.count = 0; data->args.context = dreq->ctx; + data->args.lock_context = dreq->l_ctx; data->res.count = 0; data->res.fattr = &data->fattr; data->res.verf = &data->verf; @@ -761,6 +771,7 @@ static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq, data->cred = msg.rpc_cred; data->args.fh = NFS_FH(inode); data->args.context = ctx; + data->args.lock_context = dreq->l_ctx; data->args.offset = pos; data->args.pgbase = pgbase; data->args.pages = data->pagevec; @@ -845,7 +856,7 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov, unsigned long nr_segs, loff_t pos, size_t count) { - ssize_t result = 0; + ssize_t result = -ENOMEM; struct inode *inode = iocb->ki_filp->f_mapping->host; struct nfs_direct_req *dreq; size_t wsize = NFS_SERVER(inode)->wsize; @@ -853,7 +864,7 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov, dreq = nfs_direct_req_alloc(); if (!dreq) - return -ENOMEM; + goto out; nfs_alloc_commit_data(dreq); if (dreq->commit_data == NULL || count < wsize) @@ -861,14 +872,18 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov, dreq->inode = inode; dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); + dreq->l_ctx = nfs_get_lock_context(dreq->ctx); + if (dreq->l_ctx != NULL) + goto out_release; if (!is_sync_kiocb(iocb)) dreq->iocb = iocb; result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, sync); if (!result) result = nfs_direct_wait(dreq); +out_release: nfs_direct_req_release(dreq); - +out: return result; } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index 099b3518feea..ec7a8f96a2c2 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -530,6 +530,68 @@ out: return err; } +static void nfs_init_lock_context(struct nfs_lock_context *l_ctx) +{ + atomic_set(&l_ctx->count, 1); + l_ctx->lockowner = current->files; + l_ctx->pid = current->tgid; + INIT_LIST_HEAD(&l_ctx->list); +} + +static struct nfs_lock_context *__nfs_find_lock_context(struct nfs_open_context *ctx) +{ + struct nfs_lock_context *pos; + + list_for_each_entry(pos, &ctx->lock_context.list, list) { + if (pos->lockowner != current->files) + continue; + if (pos->pid != current->tgid) + continue; + atomic_inc(&pos->count); + return pos; + } + return NULL; +} + +struct nfs_lock_context *nfs_get_lock_context(struct nfs_open_context *ctx) +{ + struct nfs_lock_context *res, *new = NULL; + struct inode *inode = ctx->path.dentry->d_inode; + + spin_lock(&inode->i_lock); + res = __nfs_find_lock_context(ctx); + if (res == NULL) { + spin_unlock(&inode->i_lock); + new = kmalloc(sizeof(*new), GFP_KERNEL); + if (new == NULL) + return NULL; + nfs_init_lock_context(new); + spin_lock(&inode->i_lock); + res = __nfs_find_lock_context(ctx); + if (res == NULL) { + list_add_tail(&new->list, &ctx->lock_context.list); + new->open_context = ctx; + res = new; + new = NULL; + } + } + spin_unlock(&inode->i_lock); + kfree(new); + return res; +} + +void nfs_put_lock_context(struct nfs_lock_context *l_ctx) +{ + struct nfs_open_context *ctx = l_ctx->open_context; + struct inode *inode = ctx->path.dentry->d_inode; + + if (!atomic_dec_and_lock(&l_ctx->count, &inode->i_lock)) + return; + list_del(&l_ctx->list); + spin_unlock(&inode->i_lock); + kfree(l_ctx); +} + /** * nfs_close_context - Common close_context() routine NFSv2/v3 * @ctx: pointer to context @@ -566,11 +628,11 @@ static struct nfs_open_context *alloc_nfs_open_context(struct path *path, struct path_get(&ctx->path); ctx->cred = get_rpccred(cred); ctx->state = NULL; - ctx->lockowner = current->files; ctx->flags = 0; ctx->error = 0; ctx->dir_cookie = 0; - atomic_set(&ctx->count, 1); + nfs_init_lock_context(&ctx->lock_context); + ctx->lock_context.open_context = ctx; } return ctx; } @@ -578,7 +640,7 @@ static struct nfs_open_context *alloc_nfs_open_context(struct path *path, struct struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx) { if (ctx != NULL) - atomic_inc(&ctx->count); + atomic_inc(&ctx->lock_context.count); return ctx; } @@ -586,7 +648,7 @@ static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync) { struct inode *inode = ctx->path.dentry->d_inode; - if (!atomic_dec_and_lock(&ctx->count, &inode->i_lock)) + if (!atomic_dec_and_lock(&ctx->lock_context.count, &inode->i_lock)) return; list_del(&ctx->list); spin_unlock(&inode->i_lock); diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 1f7781d636ae..873b62f209ea 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1324,14 +1324,14 @@ static void encode_putrootfh(struct xdr_stream *xdr, struct compound_hdr *hdr) hdr->replen += decode_putrootfh_maxsz; } -static void encode_stateid(struct xdr_stream *xdr, const struct nfs_open_context *ctx) +static void encode_stateid(struct xdr_stream *xdr, const struct nfs_open_context *ctx, const struct nfs_lock_context *l_ctx) { nfs4_stateid stateid; __be32 *p; p = reserve_space(xdr, NFS4_STATEID_SIZE); if (ctx->state != NULL) { - nfs4_copy_stateid(&stateid, ctx->state, ctx->lockowner); + nfs4_copy_stateid(&stateid, ctx->state, l_ctx->lockowner); xdr_encode_opaque_fixed(p, stateid.data, NFS4_STATEID_SIZE); } else xdr_encode_opaque_fixed(p, zero_stateid.data, NFS4_STATEID_SIZE); @@ -1344,7 +1344,7 @@ static void encode_read(struct xdr_stream *xdr, const struct nfs_readargs *args, p = reserve_space(xdr, 4); *p = cpu_to_be32(OP_READ); - encode_stateid(xdr, args->context); + encode_stateid(xdr, args->context, args->lock_context); p = reserve_space(xdr, 12); p = xdr_encode_hyper(p, args->offset); @@ -1523,7 +1523,7 @@ static void encode_write(struct xdr_stream *xdr, const struct nfs_writeargs *arg p = reserve_space(xdr, 4); *p = cpu_to_be32(OP_WRITE); - encode_stateid(xdr, args->context); + encode_stateid(xdr, args->context, args->lock_context); p = reserve_space(xdr, 16); p = xdr_encode_hyper(p, args->offset); diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index a3654e57b589..919490232e17 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -79,6 +79,7 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode, req->wb_pgbase = offset; req->wb_bytes = count; req->wb_context = get_nfs_open_context(ctx); + req->wb_lock_context = nfs_get_lock_context(ctx); kref_init(&req->wb_kref); return req; } @@ -141,11 +142,16 @@ void nfs_clear_request(struct nfs_page *req) { struct page *page = req->wb_page; struct nfs_open_context *ctx = req->wb_context; + struct nfs_lock_context *l_ctx = req->wb_lock_context; if (page != NULL) { page_cache_release(page); req->wb_page = NULL; } + if (l_ctx != NULL) { + nfs_put_lock_context(l_ctx); + req->wb_lock_context = NULL; + } if (ctx != NULL) { put_nfs_open_context(ctx); req->wb_context = NULL; @@ -235,7 +241,7 @@ static int nfs_can_coalesce_requests(struct nfs_page *prev, { if (req->wb_context->cred != prev->wb_context->cred) return 0; - if (req->wb_context->lockowner != prev->wb_context->lockowner) + if (req->wb_lock_context->lockowner != prev->wb_lock_context->lockowner) return 0; if (req->wb_context->state != prev->wb_context->state) return 0; diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 5a33a92e8168..87adc2744246 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -190,6 +190,7 @@ static int nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, data->args.pages = data->pagevec; data->args.count = count; data->args.context = get_nfs_open_context(req->wb_context); + data->args.lock_context = req->wb_lock_context; data->res.fattr = &data->fattr; data->res.count = count; diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 03df22822c4c..5eccea127cac 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -689,7 +689,9 @@ int nfs_flush_incompatible(struct file *file, struct page *page) req = nfs_page_find_request(page); if (req == NULL) return 0; - do_flush = req->wb_page != page || req->wb_context != ctx; + do_flush = req->wb_page != page || req->wb_context != ctx || + req->wb_lock_context->lockowner != current->files || + req->wb_lock_context->pid != current->tgid; nfs_release_request(req); if (!do_flush) return 0; @@ -813,6 +815,7 @@ static int nfs_write_rpcsetup(struct nfs_page *req, data->args.pages = data->pagevec; data->args.count = count; data->args.context = get_nfs_open_context(req->wb_context); + data->args.lock_context = req->wb_lock_context; data->args.stable = NFS_UNSTABLE; if (how & FLUSH_STABLE) { data->args.stable = NFS_DATA_SYNC; diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 77c2ae53431c..a9d802615082 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -72,13 +72,20 @@ struct nfs_access_entry { int mask; }; +struct nfs_lock_context { + atomic_t count; + struct list_head list; + struct nfs_open_context *open_context; + fl_owner_t lockowner; + pid_t pid; +}; + struct nfs4_state; struct nfs_open_context { - atomic_t count; + struct nfs_lock_context lock_context; struct path path; struct rpc_cred *cred; struct nfs4_state *state; - fl_owner_t lockowner; fmode_t mode; unsigned long flags; @@ -353,6 +360,8 @@ extern void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr); extern struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx); extern void put_nfs_open_context(struct nfs_open_context *ctx); extern struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, fmode_t mode); +extern struct nfs_lock_context *nfs_get_lock_context(struct nfs_open_context *ctx); +extern void nfs_put_lock_context(struct nfs_lock_context *l_ctx); extern u64 nfs_compat_user_ino64(u64 fileid); extern void nfs_fattr_init(struct nfs_fattr *fattr); diff --git a/include/linux/nfs_page.h b/include/linux/nfs_page.h index 3c60685d972b..f8b60e7f4c44 100644 --- a/include/linux/nfs_page.h +++ b/include/linux/nfs_page.h @@ -39,6 +39,7 @@ struct nfs_page { struct list_head wb_list; /* Defines state of page: */ struct page *wb_page; /* page to read in/write out */ struct nfs_open_context *wb_context; /* File state context info */ + struct nfs_lock_context *wb_lock_context; /* lock context info */ atomic_t wb_complete; /* i/os we're waiting for */ pgoff_t wb_index; /* Offset >> PAGE_CACHE_SHIFT */ unsigned int wb_offset, /* Offset & ~PAGE_CACHE_MASK */ diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index a319cb926abf..87202c7026e3 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -334,6 +334,7 @@ struct nfs4_delegreturnres { struct nfs_readargs { struct nfs_fh * fh; struct nfs_open_context *context; + struct nfs_lock_context *lock_context; __u64 offset; __u32 count; unsigned int pgbase; @@ -354,6 +355,7 @@ struct nfs_readres { struct nfs_writeargs { struct nfs_fh * fh; struct nfs_open_context *context; + struct nfs_lock_context *lock_context; __u64 offset; __u32 count; enum nfs3_stable_how stable; From daccbded7f153ec84a3baf3136052e41d0eab555 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 25 Jun 2010 18:11:43 -0400 Subject: [PATCH 18/42] NFSv4: Clean up for lockowner XDR encoding Signed-off-by: Trond Myklebust --- fs/nfs/nfs4xdr.c | 32 ++++++++++++++++++++------------ 1 file changed, 20 insertions(+), 12 deletions(-) diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 873b62f209ea..49df05afdc64 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -202,14 +202,17 @@ static int nfs4_stat_to_errno(int); #define encode_link_maxsz (op_encode_hdr_maxsz + \ nfs4_name_maxsz) #define decode_link_maxsz (op_decode_hdr_maxsz + decode_change_info_maxsz) +#define encode_lockowner_maxsz (7) #define encode_lock_maxsz (op_encode_hdr_maxsz + \ 7 + \ - 1 + encode_stateid_maxsz + 8) + 1 + encode_stateid_maxsz + 1 + \ + encode_lockowner_maxsz) #define decode_lock_denied_maxsz \ (8 + decode_lockowner_maxsz) #define decode_lock_maxsz (op_decode_hdr_maxsz + \ decode_lock_denied_maxsz) -#define encode_lockt_maxsz (op_encode_hdr_maxsz + 12) +#define encode_lockt_maxsz (op_encode_hdr_maxsz + 5 + \ + encode_lockowner_maxsz) #define decode_lockt_maxsz (op_decode_hdr_maxsz + \ decode_lock_denied_maxsz) #define encode_locku_maxsz (op_encode_hdr_maxsz + 3 + \ @@ -1042,6 +1045,17 @@ static inline uint64_t nfs4_lock_length(struct file_lock *fl) return fl->fl_end - fl->fl_start + 1; } +static void encode_lockowner(struct xdr_stream *xdr, const struct nfs_lowner *lowner) +{ + __be32 *p; + + p = reserve_space(xdr, 28); + p = xdr_encode_hyper(p, lowner->clientid); + *p++ = cpu_to_be32(16); + p = xdr_encode_opaque_fixed(p, "lock id:", 8); + xdr_encode_hyper(p, lowner->id); +} + /* * opcode,type,reclaim,offset,length,new_lock_owner = 32 * open_seqid,open_stateid,lock_seqid,lock_owner.clientid, lock_owner.id = 40 @@ -1058,14 +1072,11 @@ static void encode_lock(struct xdr_stream *xdr, const struct nfs_lock_args *args p = xdr_encode_hyper(p, nfs4_lock_length(args->fl)); *p = cpu_to_be32(args->new_lock_owner); if (args->new_lock_owner){ - p = reserve_space(xdr, 4+NFS4_STATEID_SIZE+32); + p = reserve_space(xdr, 4+NFS4_STATEID_SIZE+4); *p++ = cpu_to_be32(args->open_seqid->sequence->counter); p = xdr_encode_opaque_fixed(p, args->open_stateid->data, NFS4_STATEID_SIZE); *p++ = cpu_to_be32(args->lock_seqid->sequence->counter); - p = xdr_encode_hyper(p, args->lock_owner.clientid); - *p++ = cpu_to_be32(16); - p = xdr_encode_opaque_fixed(p, "lock id:", 8); - xdr_encode_hyper(p, args->lock_owner.id); + encode_lockowner(xdr, &args->lock_owner); } else { p = reserve_space(xdr, NFS4_STATEID_SIZE+4); @@ -1080,15 +1091,12 @@ static void encode_lockt(struct xdr_stream *xdr, const struct nfs_lockt_args *ar { __be32 *p; - p = reserve_space(xdr, 52); + p = reserve_space(xdr, 24); *p++ = cpu_to_be32(OP_LOCKT); *p++ = cpu_to_be32(nfs4_lock_type(args->fl, 0)); p = xdr_encode_hyper(p, args->fl->fl_start); p = xdr_encode_hyper(p, nfs4_lock_length(args->fl)); - p = xdr_encode_hyper(p, args->lock_owner.clientid); - *p++ = cpu_to_be32(16); - p = xdr_encode_opaque_fixed(p, "lock id:", 8); - xdr_encode_hyper(p, args->lock_owner.id); + encode_lockowner(xdr, &args->lock_owner); hdr->nops++; hdr->replen += decode_lockt_maxsz; } From d3c7b7ccc199ee564177ee914c04771d6bc00295 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Jul 2010 12:49:01 -0400 Subject: [PATCH 19/42] NFSv4: Add support for the RELEASE_LOCKOWNER operation This is needed by NFSv4.0 servers in order to keep the number of locking stateids at a manageable level. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 1 + fs/nfs/nfs4proc.c | 28 +++++++++++++++++++++ fs/nfs/nfs4state.c | 2 ++ fs/nfs/nfs4xdr.c | 55 +++++++++++++++++++++++++++++++++++++++++ include/linux/nfs4.h | 1 + include/linux/nfs_xdr.h | 4 +++ 6 files changed, 91 insertions(+) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index cee871471e8d..deaf37f5a7a9 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -236,6 +236,7 @@ extern int nfs4_open_revalidate(struct inode *, struct dentry *, int, struct nam extern int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle); extern int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name, struct nfs4_fs_locations *fs_locations, struct page *page); +extern void nfs4_release_lockowner(const struct nfs4_lock_state *); #if defined(CONFIG_NFS_V4_1) static inline struct nfs4_session *nfs4_get_session(const struct nfs_server *server) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index de9ff1505a24..5d3e8a2db99f 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -4414,6 +4414,34 @@ out: return err; } +static void nfs4_release_lockowner_release(void *calldata) +{ + kfree(calldata); +} + +const struct rpc_call_ops nfs4_release_lockowner_ops = { + .rpc_release = nfs4_release_lockowner_release, +}; + +void nfs4_release_lockowner(const struct nfs4_lock_state *lsp) +{ + struct nfs_server *server = lsp->ls_state->owner->so_server; + struct nfs_release_lockowner_args *args; + struct rpc_message msg = { + .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER], + }; + + if (server->nfs_client->cl_mvops->minor_version != 0) + return; + args = kmalloc(sizeof(*args), GFP_NOFS); + if (!args) + return; + args->lock_owner.clientid = server->nfs_client->cl_clientid; + args->lock_owner.id = lsp->ls_id.id; + msg.rpc_argp = args; + rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, args); +} + #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl" int nfs4_setxattr(struct dentry *dentry, const char *key, const void *buf, diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 13e17e32e3e4..13a4f27e7271 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -701,6 +701,8 @@ void nfs4_put_lock_state(struct nfs4_lock_state *lsp) if (list_empty(&state->lock_states)) clear_bit(LK_STATE_IN_USE, &state->flags); spin_unlock(&state->state_lock); + if (lsp->ls_flags & NFS_LOCK_INITIALIZED) + nfs4_release_lockowner(lsp); nfs4_free_lock_state(lsp); } diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 49df05afdc64..15185c2abd11 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -220,6 +220,11 @@ static int nfs4_stat_to_errno(int); 4) #define decode_locku_maxsz (op_decode_hdr_maxsz + \ decode_stateid_maxsz) +#define encode_release_lockowner_maxsz \ + (op_encode_hdr_maxsz + \ + encode_lockowner_maxsz) +#define decode_release_lockowner_maxsz \ + (op_decode_hdr_maxsz) #define encode_access_maxsz (op_encode_hdr_maxsz + 1) #define decode_access_maxsz (op_decode_hdr_maxsz + 2) #define encode_symlink_maxsz (op_encode_hdr_maxsz + \ @@ -474,6 +479,12 @@ static int nfs4_stat_to_errno(int); decode_sequence_maxsz + \ decode_putfh_maxsz + \ decode_locku_maxsz) +#define NFS4_enc_release_lockowner_sz \ + (compound_encode_hdr_maxsz + \ + encode_lockowner_maxsz) +#define NFS4_dec_release_lockowner_sz \ + (compound_decode_hdr_maxsz + \ + decode_lockowner_maxsz) #define NFS4_enc_access_sz (compound_encode_hdr_maxsz + \ encode_sequence_maxsz + \ encode_putfh_maxsz + \ @@ -1116,6 +1127,17 @@ static void encode_locku(struct xdr_stream *xdr, const struct nfs_locku_args *ar hdr->replen += decode_locku_maxsz; } +static void encode_release_lockowner(struct xdr_stream *xdr, const struct nfs_lowner *lowner, struct compound_hdr *hdr) +{ + __be32 *p; + + p = reserve_space(xdr, 4); + *p = cpu_to_be32(OP_RELEASE_LOCKOWNER); + encode_lockowner(xdr, lowner); + hdr->nops++; + hdr->replen += decode_release_lockowner_maxsz; +} + static void encode_lookup(struct xdr_stream *xdr, const struct qstr *name, struct compound_hdr *hdr) { int len = name->len; @@ -2056,6 +2078,20 @@ static int nfs4_xdr_enc_locku(struct rpc_rqst *req, __be32 *p, struct nfs_locku_ return 0; } +static int nfs4_xdr_enc_release_lockowner(struct rpc_rqst *req, __be32 *p, struct nfs_release_lockowner_args *args) +{ + struct xdr_stream xdr; + struct compound_hdr hdr = { + .minorversion = 0, + }; + + xdr_init_encode(&xdr, &req->rq_snd_buf, p); + encode_compound_hdr(&xdr, req, &hdr); + encode_release_lockowner(&xdr, &args->lock_owner, &hdr); + encode_nops(&hdr); + return 0; +} + /* * Encode a READLINK request */ @@ -3981,6 +4017,11 @@ static int decode_locku(struct xdr_stream *xdr, struct nfs_locku_res *res) return status; } +static int decode_release_lockowner(struct xdr_stream *xdr) +{ + return decode_op_hdr(xdr, OP_RELEASE_LOCKOWNER); +} + static int decode_lookup(struct xdr_stream *xdr) { return decode_op_hdr(xdr, OP_LOOKUP); @@ -5267,6 +5308,19 @@ out: return status; } +static int nfs4_xdr_dec_release_lockowner(struct rpc_rqst *rqstp, __be32 *p, void *dummy) +{ + struct xdr_stream xdr; + struct compound_hdr hdr; + int status; + + xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p); + status = decode_compound_hdr(&xdr, &hdr); + if (!status) + status = decode_release_lockowner(&xdr); + return status; +} + /* * Decode READLINK response */ @@ -5874,6 +5928,7 @@ struct rpc_procinfo nfs4_procedures[] = { PROC(GETACL, enc_getacl, dec_getacl), PROC(SETACL, enc_setacl, dec_setacl), PROC(FS_LOCATIONS, enc_fs_locations, dec_fs_locations), + PROC(RELEASE_LOCKOWNER, enc_release_lockowner, dec_release_lockowner), #if defined(CONFIG_NFS_V4_1) PROC(EXCHANGE_ID, enc_exchange_id, dec_exchange_id), PROC(CREATE_SESSION, enc_create_session, dec_create_session), diff --git a/include/linux/nfs4.h b/include/linux/nfs4.h index 9b8299af3741..07e40c625972 100644 --- a/include/linux/nfs4.h +++ b/include/linux/nfs4.h @@ -523,6 +523,7 @@ enum { NFSPROC4_CLNT_GETACL, NFSPROC4_CLNT_SETACL, NFSPROC4_CLNT_FS_LOCATIONS, + NFSPROC4_CLNT_RELEASE_LOCKOWNER, /* nfs41 */ NFSPROC4_CLNT_EXCHANGE_ID, diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 87202c7026e3..fc461926c412 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -315,6 +315,10 @@ struct nfs_lockt_res { struct nfs4_sequence_res seq_res; }; +struct nfs_release_lockowner_args { + struct nfs_lowner lock_owner; +}; + struct nfs4_delegreturnargs { const struct nfs_fh *fhandle; const nfs4_stateid *stateid; From 77041ed9b49a9e10f374bfa6e482d30ee7a3d46e Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 1 Jul 2010 12:49:11 -0400 Subject: [PATCH 20/42] NFSv4: Ensure the lockowners are labelled using the fl_owner and/or fl_pid flock locks want to be labelled using the process pid, while posix locks want to be labelled using the fl_owner. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4_fs.h | 15 +++++++++++++-- fs/nfs/nfs4proc.c | 2 +- fs/nfs/nfs4state.c | 45 +++++++++++++++++++++++++++++++++++---------- fs/nfs/nfs4xdr.c | 2 +- 4 files changed, 50 insertions(+), 14 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index deaf37f5a7a9..311e15cc8af0 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -142,10 +142,20 @@ enum { * LOCK: one nfs4_state (LOCK) to hold the lock stateid nfs4_state(OPEN) */ +struct nfs4_lock_owner { + unsigned int lo_type; +#define NFS4_ANY_LOCK_TYPE (0U) +#define NFS4_FLOCK_LOCK_TYPE (1U << 0) +#define NFS4_POSIX_LOCK_TYPE (1U << 1) + union { + fl_owner_t posix_owner; + pid_t flock_owner; + } lo_u; +}; + struct nfs4_lock_state { struct list_head ls_locks; /* Other lock stateids */ struct nfs4_state * ls_state; /* Pointer to open state */ - fl_owner_t ls_owner; /* POSIX lock owner */ #define NFS_LOCK_INITIALIZED 1 int ls_flags; struct nfs_seqid_counter ls_seqid; @@ -153,6 +163,7 @@ struct nfs4_lock_state { struct nfs_unique_id ls_id; nfs4_stateid ls_stateid; atomic_t ls_count; + struct nfs4_lock_owner ls_owner; }; /* bits for nfs4_state->flags */ @@ -310,7 +321,7 @@ extern void nfs41_handle_sequence_flag_errors(struct nfs_client *clp, u32 flags) extern void nfs41_handle_recall_slot(struct nfs_client *clp); extern void nfs4_put_lock_state(struct nfs4_lock_state *lsp); extern int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl); -extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t); +extern void nfs4_copy_stateid(nfs4_stateid *, struct nfs4_state *, fl_owner_t, pid_t); extern struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter, gfp_t gfp_mask); extern int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 5d3e8a2db99f..d6413b48b057 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1767,7 +1767,7 @@ static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred, if (nfs4_copy_delegation_stateid(&arg.stateid, inode)) { /* Use that stateid */ } else if (state != NULL) { - nfs4_copy_stateid(&arg.stateid, state, current->files); + nfs4_copy_stateid(&arg.stateid, state, current->files, current->tgid); } else memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid)); diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c index 13a4f27e7271..3e2f19b04c06 100644 --- a/fs/nfs/nfs4state.c +++ b/fs/nfs/nfs4state.c @@ -602,12 +602,21 @@ void nfs4_close_sync(struct path *path, struct nfs4_state *state, fmode_t fmode) * that is compatible with current->files */ static struct nfs4_lock_state * -__nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) +__nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type) { struct nfs4_lock_state *pos; list_for_each_entry(pos, &state->lock_states, ls_locks) { - if (pos->ls_owner != fl_owner) + if (type != NFS4_ANY_LOCK_TYPE && pos->ls_owner.lo_type != type) continue; + switch (pos->ls_owner.lo_type) { + case NFS4_POSIX_LOCK_TYPE: + if (pos->ls_owner.lo_u.posix_owner != fl_owner) + continue; + break; + case NFS4_FLOCK_LOCK_TYPE: + if (pos->ls_owner.lo_u.flock_owner != fl_pid) + continue; + } atomic_inc(&pos->ls_count); return pos; } @@ -619,7 +628,7 @@ __nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) * exists, return an uninitialized one. * */ -static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner) +static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid, unsigned int type) { struct nfs4_lock_state *lsp; struct nfs_client *clp = state->owner->so_server->nfs_client; @@ -633,7 +642,18 @@ static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, f lsp->ls_seqid.sequence = &lsp->ls_sequence; atomic_set(&lsp->ls_count, 1); lsp->ls_state = state; - lsp->ls_owner = fl_owner; + lsp->ls_owner.lo_type = type; + switch (lsp->ls_owner.lo_type) { + case NFS4_FLOCK_LOCK_TYPE: + lsp->ls_owner.lo_u.flock_owner = fl_pid; + break; + case NFS4_POSIX_LOCK_TYPE: + lsp->ls_owner.lo_u.posix_owner = fl_owner; + break; + default: + kfree(lsp); + return NULL; + } spin_lock(&clp->cl_lock); nfs_alloc_unique_id(&clp->cl_lockowner_id, &lsp->ls_id, 1, 64); spin_unlock(&clp->cl_lock); @@ -657,13 +677,13 @@ static void nfs4_free_lock_state(struct nfs4_lock_state *lsp) * exists, return an uninitialized one. * */ -static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner) +static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner, pid_t pid, unsigned int type) { struct nfs4_lock_state *lsp, *new = NULL; for(;;) { spin_lock(&state->state_lock); - lsp = __nfs4_find_lock_state(state, owner); + lsp = __nfs4_find_lock_state(state, owner, pid, type); if (lsp != NULL) break; if (new != NULL) { @@ -674,7 +694,7 @@ static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_ break; } spin_unlock(&state->state_lock); - new = nfs4_alloc_lock_state(state, owner); + new = nfs4_alloc_lock_state(state, owner, pid, type); if (new == NULL) return NULL; } @@ -730,7 +750,12 @@ int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl) if (fl->fl_ops != NULL) return 0; - lsp = nfs4_get_lock_state(state, fl->fl_owner); + if (fl->fl_flags & FL_POSIX) + lsp = nfs4_get_lock_state(state, fl->fl_owner, 0, NFS4_POSIX_LOCK_TYPE); + else if (fl->fl_flags & FL_FLOCK) + lsp = nfs4_get_lock_state(state, 0, fl->fl_pid, NFS4_FLOCK_LOCK_TYPE); + else + return -EINVAL; if (lsp == NULL) return -ENOMEM; fl->fl_u.nfs4_fl.owner = lsp; @@ -742,7 +767,7 @@ int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl) * Byte-range lock aware utility to initialize the stateid of read/write * requests. */ -void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner) +void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner, pid_t fl_pid) { struct nfs4_lock_state *lsp; int seq; @@ -755,7 +780,7 @@ void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t f return; spin_lock(&state->state_lock); - lsp = __nfs4_find_lock_state(state, fl_owner); + lsp = __nfs4_find_lock_state(state, fl_owner, fl_pid, NFS4_ANY_LOCK_TYPE); if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0) memcpy(dst, &lsp->ls_stateid, sizeof(*dst)); spin_unlock(&state->state_lock); diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 15185c2abd11..257c1811feb4 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -1361,7 +1361,7 @@ static void encode_stateid(struct xdr_stream *xdr, const struct nfs_open_context p = reserve_space(xdr, NFS4_STATEID_SIZE); if (ctx->state != NULL) { - nfs4_copy_stateid(&stateid, ctx->state, l_ctx->lockowner); + nfs4_copy_stateid(&stateid, ctx->state, l_ctx->lockowner, l_ctx->pid); xdr_encode_opaque_fixed(p, stateid.data, NFS4_STATEID_SIZE); } else xdr_encode_opaque_fixed(p, zero_stateid.data, NFS4_STATEID_SIZE); From af7fa16506bf9b6323e862a61e14c20555152bb3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:06 -0400 Subject: [PATCH 21/42] NFS: Fix up the fsync code Christoph points out that the VFS will always flush out data before calling nfs_fsync(), so we can dispense with a full call to nfs_wb_all(), and replace that with a simpler call to nfs_commit_inode(). Signed-off-by: Trond Myklebust --- fs/nfs/file.c | 51 +++++++++++++++++++++------------------------------ 1 file changed, 21 insertions(+), 30 deletions(-) diff --git a/fs/nfs/file.c b/fs/nfs/file.c index 36a5e74f51b4..f36581cd4767 100644 --- a/fs/nfs/file.c +++ b/fs/nfs/file.c @@ -201,38 +201,12 @@ static loff_t nfs_file_llseek(struct file *filp, loff_t offset, int origin) return loff; } -/* - * Helper for nfs_file_flush() and nfs_file_fsync() - * - * Notice that it clears the NFS_CONTEXT_ERROR_WRITE before synching to - * disk, but it retrieves and clears ctx->error after synching, despite - * the two being set at the same time in nfs_context_set_write_error(). - * This is because the former is used to notify the _next_ call to - * nfs_file_write() that a write error occured, and hence cause it to - * fall back to doing a synchronous write. - */ -static int nfs_do_fsync(struct nfs_open_context *ctx, struct inode *inode) -{ - int have_error, status; - int ret = 0; - - have_error = test_and_clear_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags); - status = nfs_wb_all(inode); - have_error |= test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags); - if (have_error) - ret = xchg(&ctx->error, 0); - if (!ret) - ret = status; - return ret; -} - /* * Flush all dirty pages, and check for write errors. */ static int nfs_file_flush(struct file *file, fl_owner_t id) { - struct nfs_open_context *ctx = nfs_file_open_context(file); struct dentry *dentry = file->f_path.dentry; struct inode *inode = dentry->d_inode; @@ -245,7 +219,7 @@ nfs_file_flush(struct file *file, fl_owner_t id) return 0; /* Flush writes to the server and return any errors */ - return nfs_do_fsync(ctx, inode); + return vfs_fsync(file, 0); } static ssize_t @@ -320,6 +294,13 @@ nfs_file_mmap(struct file * file, struct vm_area_struct * vma) * Flush any dirty pages for this process, and check for write errors. * The return status from this call provides a reliable indication of * whether any write errors occurred for this process. + * + * Notice that it clears the NFS_CONTEXT_ERROR_WRITE before synching to + * disk, but it retrieves and clears ctx->error after synching, despite + * the two being set at the same time in nfs_context_set_write_error(). + * This is because the former is used to notify the _next_ call to + * nfs_file_write() that a write error occured, and hence cause it to + * fall back to doing a synchronous write. */ static int nfs_file_fsync(struct file *file, int datasync) @@ -327,13 +308,23 @@ nfs_file_fsync(struct file *file, int datasync) struct dentry *dentry = file->f_path.dentry; struct nfs_open_context *ctx = nfs_file_open_context(file); struct inode *inode = dentry->d_inode; + int have_error, status; + int ret = 0; + dprintk("NFS: fsync file(%s/%s) datasync %d\n", dentry->d_parent->d_name.name, dentry->d_name.name, datasync); nfs_inc_stats(inode, NFSIOS_VFSFSYNC); - return nfs_do_fsync(ctx, inode); + have_error = test_and_clear_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags); + status = nfs_commit_inode(inode, FLUSH_SYNC); + have_error |= test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags); + if (have_error) + ret = xchg(&ctx->error, 0); + if (!ret) + ret = status; + return ret; } /* @@ -639,7 +630,7 @@ static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov, /* Return error values for O_DSYNC and IS_SYNC() */ if (result >= 0 && nfs_need_sync_write(iocb->ki_filp, inode)) { - int err = nfs_do_fsync(nfs_file_open_context(iocb->ki_filp), inode); + int err = vfs_fsync(iocb->ki_filp, 0); if (err < 0) result = err; } @@ -675,7 +666,7 @@ static ssize_t nfs_file_splice_write(struct pipe_inode_info *pipe, written = ret; if (ret >= 0 && nfs_need_sync_write(filp, inode)) { - int err = nfs_do_fsync(nfs_file_open_context(filp), inode); + int err = vfs_fsync(filp, 0); if (err < 0) ret = err; } From 1b924e5f878e3cd62a20bd8dbf3b911a40185a99 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:06 -0400 Subject: [PATCH 22/42] NFS: Clean up the callers of nfs_wb_all() There is no need to flush out writes before calling nfs_wb_all(). Signed-off-by: Trond Myklebust --- fs/nfs/delegation.c | 10 +--------- fs/nfs/inode.c | 4 +--- 2 files changed, 2 insertions(+), 12 deletions(-) diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c index f34f4ac52b81..b9c3c43cea1d 100644 --- a/fs/nfs/delegation.c +++ b/fs/nfs/delegation.c @@ -268,14 +268,6 @@ out: return status; } -/* Sync all data to disk upon delegation return */ -static void nfs_msync_inode(struct inode *inode) -{ - filemap_fdatawrite(inode->i_mapping); - nfs_wb_all(inode); - filemap_fdatawait(inode->i_mapping); -} - /* * Basic procedure for returning a delegation to the server */ @@ -367,7 +359,7 @@ int nfs_inode_return_delegation(struct inode *inode) delegation = nfs_detach_delegation_locked(nfsi, NULL, clp); spin_unlock(&clp->cl_lock); if (delegation != NULL) { - nfs_msync_inode(inode); + nfs_wb_all(inode); err = __nfs_inode_return_delegation(inode, delegation, 1); } } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c index ec7a8f96a2c2..581d8f081e68 100644 --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -413,10 +413,8 @@ nfs_setattr(struct dentry *dentry, struct iattr *attr) return 0; /* Write all dirty data */ - if (S_ISREG(inode->i_mode)) { - filemap_write_and_wait(inode->i_mapping); + if (S_ISREG(inode->i_mode)) nfs_wb_all(inode); - } fattr = nfs_alloc_fattr(); if (fattr == NULL) From 0a8ebba943dd89bdd57c5dab5a66932f690847d9 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:06 -0400 Subject: [PATCH 23/42] NFS: nfs_rename() should not have to flush out writebacks We don't really support nfs servers that invalidate the file handle after a rename, so precautions such as flushing out dirty data before renaming the file are superfluous. Signed-off-by: Trond Myklebust --- fs/nfs/dir.c | 9 --------- 1 file changed, 9 deletions(-) diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 782b431ef91c..067a051397ba 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -1652,16 +1652,7 @@ static int nfs_rename(struct inode *old_dir, struct dentry *old_dentry, } } - /* - * ... prune child dentries and writebacks if needed. - */ - if (atomic_read(&old_dentry->d_count) > 1) { - if (S_ISREG(old_inode->i_mode)) - nfs_wb_all(old_inode); - shrink_dcache_parent(old_dentry); - } nfs_inode_return_delegation(old_inode); - if (new_inode != NULL) nfs_inode_return_delegation(new_inode); From 14516c3a30e256e8d4e7a9af271c8df644ac3222 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:06 -0400 Subject: [PATCH 24/42] NFSv4.1: Handle NFS4ERR_DELAY on SEQUENCE correctly In RFC5661, an NFS4ERR_DELAY error on a SEQUENCE operation has the special meaning that the server is not finished processing the request. In this case we want to just retry the request without touching the slot. Also fix a bug whereby we would fail to update the sequence id if the server returned any error other than NFS_OK/NFS4ERR_DELAY. Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 92 +++++++++++++++++++++++++++++++---------------- 1 file changed, 62 insertions(+), 30 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index d6413b48b057..cab0eb915145 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -389,11 +389,12 @@ static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res) res->sr_slotid = NFS4_MAX_SLOT_TABLE; } -static void nfs41_sequence_done(struct nfs4_sequence_res *res) +static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res) { unsigned long timestamp; struct nfs4_slot_table *tbl; struct nfs4_slot *slot; + struct nfs_client *clp; /* * sr_status remains 1 if an RPC level error occurred. The server @@ -408,14 +409,16 @@ static void nfs41_sequence_done(struct nfs4_sequence_res *res) if (res->sr_slotid == NFS4_MAX_SLOT_TABLE) goto out; + tbl = &res->sr_session->fc_slot_table; + slot = tbl->slots + res->sr_slotid; + /* Check the SEQUENCE operation status */ - if (res->sr_status == 0) { - struct nfs_client *clp = res->sr_session->clp; - tbl = &res->sr_session->fc_slot_table; - slot = tbl->slots + res->sr_slotid; + switch (res->sr_status) { + case 0: /* Update the slot's sequence and clientid lease timer */ ++slot->seq_nr; timestamp = res->sr_renewal_time; + clp = res->sr_session->clp; spin_lock(&clp->cl_lock); if (time_before(clp->cl_last_renewal, timestamp)) clp->cl_last_renewal = timestamp; @@ -423,18 +426,39 @@ static void nfs41_sequence_done(struct nfs4_sequence_res *res) /* Check sequence flags */ if (atomic_read(&clp->cl_count) > 1) nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags); + break; + case -NFS4ERR_DELAY: + /* The server detected a resend of the RPC call and + * returned NFS4ERR_DELAY as per Section 2.10.6.2 + * of RFC5661. + */ + dprintk("%s: slot=%d seq=%d: Operation in progress\n", + __func__, res->sr_slotid, slot->seq_nr); + goto out_retry; + default: + /* Just update the slot sequence no. */ + ++slot->seq_nr; } out: /* The session may be reset by one of the error handlers. */ dprintk("%s: Error %d free the slot \n", __func__, res->sr_status); nfs41_sequence_free_slot(res); + return 1; +out_retry: + rpc_restart_call(task); + /* FIXME: rpc_restart_call() should be made to return success/fail */ + if (task->tk_action == NULL) + goto out; + rpc_delay(task, NFS4_POLL_RETRY_MAX); + return 0; } -static void nfs4_sequence_done(const struct nfs_server *server, - struct nfs4_sequence_res *res, int rpc_status) +static int nfs4_sequence_done(struct rpc_task *task, + struct nfs4_sequence_res *res) { - if (res->sr_session != NULL) - nfs41_sequence_done(res); + if (res->sr_session == NULL) + return 1; + return nfs41_sequence_done(task, res); } /* @@ -592,7 +616,7 @@ static void nfs41_call_sync_done(struct rpc_task *task, void *calldata) { struct nfs41_call_sync_data *data = calldata; - nfs41_sequence_done(data->seq_res); + nfs41_sequence_done(task, data->seq_res); } struct rpc_call_ops nfs41_call_sync_ops = { @@ -650,9 +674,10 @@ int _nfs4_call_sync_session(struct nfs_server *server, } #else -static void nfs4_sequence_done(const struct nfs_server *server, - struct nfs4_sequence_res *res, int rpc_status) +static int nfs4_sequence_done(struct rpc_task *task, + struct nfs4_sequence_res *res) { + return 1; } #endif /* CONFIG_NFS_V4_1 */ @@ -1379,8 +1404,8 @@ static void nfs4_open_done(struct rpc_task *task, void *calldata) data->rpc_status = task->tk_status; - nfs4_sequence_done(data->o_arg.server, &data->o_res.seq_res, - task->tk_status); + if (!nfs4_sequence_done(task, &data->o_res.seq_res)) + return; if (RPC_ASSASSINATED(task)) return; @@ -1832,7 +1857,8 @@ static void nfs4_close_done(struct rpc_task *task, void *data) struct nfs4_state *state = calldata->state; struct nfs_server *server = NFS_SERVER(calldata->inode); - nfs4_sequence_done(server, &calldata->res.seq_res, task->tk_status); + if (!nfs4_sequence_done(task, &calldata->res.seq_res)) + return; if (RPC_ASSASSINATED(task)) return; /* hmm. we are done with the inode, and in the process of freeing @@ -2642,7 +2668,8 @@ static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir) { struct nfs_removeres *res = task->tk_msg.rpc_resp; - nfs4_sequence_done(res->server, &res->seq_res, task->tk_status); + if (!nfs4_sequence_done(task, &res->seq_res)) + return 0; if (nfs4_async_handle_error(task, res->server, NULL) == -EAGAIN) return 0; update_changeattr(dir, &res->cinfo); @@ -3087,7 +3114,8 @@ static int nfs4_read_done(struct rpc_task *task, struct nfs_read_data *data) dprintk("--> %s\n", __func__); - nfs4_sequence_done(server, &data->res.seq_res, task->tk_status); + if (!nfs4_sequence_done(task, &data->res.seq_res)) + return -EAGAIN; if (nfs4_async_handle_error(task, server, data->args.context->state) == -EAGAIN) { nfs_restart_rpc(task, server->nfs_client); @@ -3110,8 +3138,8 @@ static int nfs4_write_done(struct rpc_task *task, struct nfs_write_data *data) { struct inode *inode = data->inode; - nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res, - task->tk_status); + if (!nfs4_sequence_done(task, &data->res.seq_res)) + return -EAGAIN; if (nfs4_async_handle_error(task, NFS_SERVER(inode), data->args.context->state) == -EAGAIN) { nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client); @@ -3139,8 +3167,9 @@ static int nfs4_commit_done(struct rpc_task *task, struct nfs_write_data *data) { struct inode *inode = data->inode; - nfs4_sequence_done(NFS_SERVER(inode), &data->res.seq_res, - task->tk_status); + if (!nfs4_sequence_done(task, &data->res.seq_res)) + return -EAGAIN; + if (nfs4_async_handle_error(task, NFS_SERVER(inode), NULL) == -EAGAIN) { nfs_restart_rpc(task, NFS_SERVER(inode)->nfs_client); return -EAGAIN; @@ -3630,8 +3659,8 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata) { struct nfs4_delegreturndata *data = calldata; - nfs4_sequence_done(data->res.server, &data->res.seq_res, - task->tk_status); + if (!nfs4_sequence_done(task, &data->res.seq_res)) + return; switch (task->tk_status) { case -NFS4ERR_STALE_STATEID: @@ -3881,8 +3910,8 @@ static void nfs4_locku_done(struct rpc_task *task, void *data) { struct nfs4_unlockdata *calldata = data; - nfs4_sequence_done(calldata->server, &calldata->res.seq_res, - task->tk_status); + if (!nfs4_sequence_done(task, &calldata->res.seq_res)) + return; if (RPC_ASSASSINATED(task)) return; switch (task->tk_status) { @@ -4091,8 +4120,8 @@ static void nfs4_lock_done(struct rpc_task *task, void *calldata) dprintk("%s: begin!\n", __func__); - nfs4_sequence_done(data->server, &data->res.seq_res, - task->tk_status); + if (!nfs4_sequence_done(task, &data->res.seq_res)) + return; data->rpc_status = task->tk_status; if (RPC_ASSASSINATED(task)) @@ -4629,7 +4658,8 @@ static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata) (struct nfs4_get_lease_time_data *)calldata; dprintk("--> %s\n", __func__); - nfs41_sequence_done(&data->res->lr_seq_res); + if (!nfs41_sequence_done(task, &data->res->lr_seq_res)) + return; switch (task->tk_status) { case -NFS4ERR_DELAY: case -NFS4ERR_GRACE: @@ -5111,7 +5141,8 @@ static void nfs41_sequence_call_done(struct rpc_task *task, void *data) struct nfs4_sequence_data *calldata = data; struct nfs_client *clp = calldata->clp; - nfs41_sequence_done(task->tk_msg.rpc_resp); + if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp)) + return; if (task->tk_status < 0) { dprintk("%s ERROR %d\n", __func__, task->tk_status); @@ -5255,7 +5286,8 @@ static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data) struct nfs4_sequence_res *res = &calldata->res.seq_res; dprintk("--> %s\n", __func__); - nfs41_sequence_done(res); + if (!nfs41_sequence_done(task, res)) + return; if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) { rpc_restart_call_prepare(task); From 452e93523d9433f83670e7b42cbe75319c208762 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:06 -0400 Subject: [PATCH 25/42] NFSv4: Clean up the process of renewing the NFSv4 lease Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 23 +++++++++-------------- 1 file changed, 9 insertions(+), 14 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index cab0eb915145..d2d20fbeb086 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -303,15 +303,19 @@ do_state_recovery: } -static void renew_lease(const struct nfs_server *server, unsigned long timestamp) +static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp) { - struct nfs_client *clp = server->nfs_client; spin_lock(&clp->cl_lock); if (time_before(clp->cl_last_renewal,timestamp)) clp->cl_last_renewal = timestamp; spin_unlock(&clp->cl_lock); } +static void renew_lease(const struct nfs_server *server, unsigned long timestamp) +{ + do_renew_lease(server->nfs_client, timestamp); +} + #if defined(CONFIG_NFS_V4_1) /* @@ -419,10 +423,7 @@ static int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res * ++slot->seq_nr; timestamp = res->sr_renewal_time; clp = res->sr_session->clp; - spin_lock(&clp->cl_lock); - if (time_before(clp->cl_last_renewal, timestamp)) - clp->cl_last_renewal = timestamp; - spin_unlock(&clp->cl_lock); + do_renew_lease(clp, timestamp); /* Check sequence flags */ if (atomic_read(&clp->cl_count) > 1) nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags); @@ -3219,10 +3220,7 @@ static void nfs4_renew_done(struct rpc_task *task, void *calldata) nfs4_schedule_state_recovery(clp); return; } - spin_lock(&clp->cl_lock); - if (time_before(clp->cl_last_renewal,timestamp)) - clp->cl_last_renewal = timestamp; - spin_unlock(&clp->cl_lock); + do_renew_lease(clp, timestamp); } static const struct rpc_call_ops nfs4_renew_ops = { @@ -3263,10 +3261,7 @@ int nfs4_proc_renew(struct nfs_client *clp, struct rpc_cred *cred) status = rpc_call_sync(clp->cl_rpcclient, &msg, 0); if (status < 0) return status; - spin_lock(&clp->cl_lock); - if (time_before(clp->cl_last_renewal,now)) - clp->cl_last_renewal = now; - spin_unlock(&clp->cl_lock); + do_renew_lease(clp, now); return 0; } From a6f03393ec86fd25523c79497a9a773bda170d1d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:07 -0400 Subject: [PATCH 26/42] NFSv4: Get rid of the bogus RPC_ASSASSINATED(task) checks There is no real reason to have RPC_ASSASSINATED() checks in the NFS code. As far as it is concerned, this is just an RPC error... Signed-off-by: Trond Myklebust --- fs/nfs/nfs4proc.c | 10 ---------- 1 file changed, 10 deletions(-) diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index d2d20fbeb086..101bf403f507 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -1275,8 +1275,6 @@ static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata) struct nfs4_opendata *data = calldata; data->rpc_status = task->tk_status; - if (RPC_ASSASSINATED(task)) - return; if (data->rpc_status == 0) { memcpy(data->o_res.stateid.data, data->c_res.stateid.data, sizeof(data->o_res.stateid.data)); @@ -1408,8 +1406,6 @@ static void nfs4_open_done(struct rpc_task *task, void *calldata) if (!nfs4_sequence_done(task, &data->o_res.seq_res)) return; - if (RPC_ASSASSINATED(task)) - return; if (task->tk_status == 0) { switch (data->o_res.f_attr->mode & S_IFMT) { case S_IFREG: @@ -1860,8 +1856,6 @@ static void nfs4_close_done(struct rpc_task *task, void *data) if (!nfs4_sequence_done(task, &calldata->res.seq_res)) return; - if (RPC_ASSASSINATED(task)) - return; /* hmm. we are done with the inode, and in the process of freeing * the state_owner. we keep this around to process errors */ @@ -3907,8 +3901,6 @@ static void nfs4_locku_done(struct rpc_task *task, void *data) if (!nfs4_sequence_done(task, &calldata->res.seq_res)) return; - if (RPC_ASSASSINATED(task)) - return; switch (task->tk_status) { case 0: memcpy(calldata->lsp->ls_stateid.data, @@ -4119,8 +4111,6 @@ static void nfs4_lock_done(struct rpc_task *task, void *calldata) return; data->rpc_status = task->tk_status; - if (RPC_ASSASSINATED(task)) - goto out; if (data->arg.new_lock_owner != 0) { if (data->rpc_status == 0) nfs_confirm_seqid(&data->lsp->ls_seqid, 0); From f1f88fc7e818c6678c6799a2edb8f1aeccc124aa Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:07 -0400 Subject: [PATCH 27/42] SUNRPC: The function rpc_restart_call() should return success/failure Both rpc_restart_call_prepare() and rpc_restart_call() test for the RPC_TASK_KILLED flag, and fail to restart the RPC call if that flag is set. This patch allows callers to know whether or not the restart was successful, so that they can perform cleanups etc in case of failure. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 4 ++-- net/sunrpc/clnt.c | 11 ++++++----- 2 files changed, 8 insertions(+), 7 deletions(-) diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index 8ed9642a5a76..debe75532199 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -148,8 +148,8 @@ int rpc_call_sync(struct rpc_clnt *clnt, const struct rpc_message *msg, int flags); struct rpc_task *rpc_call_null(struct rpc_clnt *clnt, struct rpc_cred *cred, int flags); -void rpc_restart_call_prepare(struct rpc_task *); -void rpc_restart_call(struct rpc_task *); +int rpc_restart_call_prepare(struct rpc_task *); +int rpc_restart_call(struct rpc_task *); void rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int); size_t rpc_max_payload(struct rpc_clnt *); void rpc_force_rebind(struct rpc_clnt *); diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 756fc324db9e..234c40c15f69 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -756,12 +756,13 @@ EXPORT_SYMBOL_GPL(rpc_force_rebind); * Restart an (async) RPC call from the call_prepare state. * Usually called from within the exit handler. */ -void +int rpc_restart_call_prepare(struct rpc_task *task) { if (RPC_ASSASSINATED(task)) - return; + return 0; task->tk_action = rpc_prepare_task; + return 1; } EXPORT_SYMBOL_GPL(rpc_restart_call_prepare); @@ -769,13 +770,13 @@ EXPORT_SYMBOL_GPL(rpc_restart_call_prepare); * Restart an (async) RPC call. Usually called from within the * exit handler. */ -void +int rpc_restart_call(struct rpc_task *task) { if (RPC_ASSASSINATED(task)) - return; - + return 0; task->tk_action = call_start; + return 1; } EXPORT_SYMBOL_GPL(rpc_restart_call); From d05dd4e98f0dd30ee933e05ac9363614c47df83a Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:07 -0400 Subject: [PATCH 28/42] NFS: Fix the NFS users of rpc_restart_call() Fix up those functions that depend on knowing whether or not rpc_restart_call is successful or not. Signed-off-by: Trond Myklebust --- fs/nfs/internal.h | 7 +++---- fs/nfs/nfs4proc.c | 4 +--- 2 files changed, 4 insertions(+), 7 deletions(-) diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h index d8bd619e386c..699725fee34f 100644 --- a/fs/nfs/internal.h +++ b/fs/nfs/internal.h @@ -369,10 +369,9 @@ unsigned int nfs_page_array_len(unsigned int base, size_t len) * Helper for restarting RPC calls in the possible presence of NFSv4.1 * sessions. */ -static inline void nfs_restart_rpc(struct rpc_task *task, const struct nfs_client *clp) +static inline int nfs_restart_rpc(struct rpc_task *task, const struct nfs_client *clp) { if (nfs4_has_session(clp)) - rpc_restart_call_prepare(task); - else - rpc_restart_call(task); + return rpc_restart_call_prepare(task); + return rpc_restart_call(task); } diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 101bf403f507..7ffbb98ddec3 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -446,9 +446,7 @@ out: nfs41_sequence_free_slot(res); return 1; out_retry: - rpc_restart_call(task); - /* FIXME: rpc_restart_call() should be made to return success/fail */ - if (task->tk_action == NULL) + if (!rpc_restart_call(task)) goto out; rpc_delay(task, NFS4_POLL_RETRY_MAX); return 0; From 5d8d9a4d9ff74c55901642b4e2ac5124830ddafe Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:07 -0400 Subject: [PATCH 29/42] NFS: Ensure the AUTH_UNIX credcache is allocated dynamically Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 7 ++++--- net/sunrpc/auth.c | 19 ++++++++++++++++--- net/sunrpc/auth_generic.c | 12 +++--------- net/sunrpc/auth_unix.c | 15 +++++++-------- net/sunrpc/sunrpc_syms.c | 15 ++++++++++----- 5 files changed, 40 insertions(+), 28 deletions(-) diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 87d7ec0bf779..784e78c73ec5 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -125,11 +125,12 @@ struct rpc_credops { extern const struct rpc_authops authunix_ops; extern const struct rpc_authops authnull_ops; -void __init rpc_init_authunix(void); -void __init rpc_init_generic_auth(void); -void __init rpcauth_init_module(void); +int __init rpc_init_authunix(void); +int __init rpc_init_generic_auth(void); +int __init rpcauth_init_module(void); void __exit rpcauth_remove_module(void); void __exit rpc_destroy_generic_auth(void); +void rpc_destroy_authunix(void); struct rpc_cred * rpc_lookup_cred(void); struct rpc_cred * rpc_lookup_machine_cred(void); diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index 73affb8624fa..db135543d21e 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -587,14 +587,27 @@ static struct shrinker rpc_cred_shrinker = { .seeks = DEFAULT_SEEKS, }; -void __init rpcauth_init_module(void) +int __init rpcauth_init_module(void) { - rpc_init_authunix(); - rpc_init_generic_auth(); + int err; + + err = rpc_init_authunix(); + if (err < 0) + goto out1; + err = rpc_init_generic_auth(); + if (err < 0) + goto out2; register_shrinker(&rpc_cred_shrinker); + return 0; +out2: + rpc_destroy_authunix(); +out1: + return err; } void __exit rpcauth_remove_module(void) { + rpc_destroy_authunix(); + rpc_destroy_generic_auth(); unregister_shrinker(&rpc_cred_shrinker); } diff --git a/net/sunrpc/auth_generic.c b/net/sunrpc/auth_generic.c index 8f623b0f03dd..8bae33b36cc6 100644 --- a/net/sunrpc/auth_generic.c +++ b/net/sunrpc/auth_generic.c @@ -27,7 +27,6 @@ struct generic_cred { }; static struct rpc_auth generic_auth; -static struct rpc_cred_cache generic_cred_cache; static const struct rpc_credops generic_credops; /* @@ -159,20 +158,16 @@ out_nomatch: return 0; } -void __init rpc_init_generic_auth(void) +int __init rpc_init_generic_auth(void) { - spin_lock_init(&generic_cred_cache.lock); + return rpcauth_init_credcache(&generic_auth); } void __exit rpc_destroy_generic_auth(void) { - rpcauth_clear_credcache(&generic_cred_cache); + rpcauth_destroy_credcache(&generic_auth); } -static struct rpc_cred_cache generic_cred_cache = { - {{ NULL, },}, -}; - static const struct rpc_authops generic_auth_ops = { .owner = THIS_MODULE, .au_name = "Generic", @@ -183,7 +178,6 @@ static const struct rpc_authops generic_auth_ops = { static struct rpc_auth generic_auth = { .au_ops = &generic_auth_ops, .au_count = ATOMIC_INIT(0), - .au_credcache = &generic_cred_cache, }; static const struct rpc_credops generic_credops = { diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index aac2f8b4ee21..d5e37dbf207b 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -29,7 +29,6 @@ struct unx_cred { #endif static struct rpc_auth unix_auth; -static struct rpc_cred_cache unix_cred_cache; static const struct rpc_credops unix_credops; static struct rpc_auth * @@ -203,9 +202,14 @@ unx_validate(struct rpc_task *task, __be32 *p) return p; } -void __init rpc_init_authunix(void) +int __init rpc_init_authunix(void) { - spin_lock_init(&unix_cred_cache.lock); + return rpcauth_init_credcache(&unix_auth); +} + +void rpc_destroy_authunix(void) +{ + rpcauth_destroy_credcache(&unix_auth); } const struct rpc_authops authunix_ops = { @@ -218,10 +222,6 @@ const struct rpc_authops authunix_ops = { .crcreate = unx_create_cred, }; -static -struct rpc_cred_cache unix_cred_cache = { -}; - static struct rpc_auth unix_auth = { .au_cslack = UNX_WRITESLACK, @@ -229,7 +229,6 @@ struct rpc_auth unix_auth = { .au_ops = &authunix_ops, .au_flavor = RPC_AUTH_UNIX, .au_count = ATOMIC_INIT(0), - .au_credcache = &unix_cred_cache, }; static diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c index f438347d817b..34b58f9e704a 100644 --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -33,10 +33,11 @@ init_sunrpc(void) if (err) goto out; err = rpc_init_mempool(); - if (err) { - unregister_rpc_pipefs(); - goto out; - } + if (err) + goto out2; + err = rpcauth_init_module(); + if (err) + goto out3; #ifdef RPC_DEBUG rpc_register_sysctl(); #endif @@ -47,7 +48,11 @@ init_sunrpc(void) cache_register(&unix_gid_cache); svc_init_xprt_sock(); /* svc sock transport */ init_socket_xprt(); /* clnt sock transport */ - rpcauth_init_module(); + return 0; +out3: + rpc_destroy_mempool(); +out2: + unregister_rpc_pipefs(); out: return err; } From 988664a0f6bbfc356e6ce55f7a87b8594050012f Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:07 -0400 Subject: [PATCH 30/42] SUNRPC: Store the hashtable size in struct rpc_cred_cache Cleanup in preparation for allowing the user to determine the maximum hash table size. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 1 + net/sunrpc/auth.c | 10 +++++++--- 2 files changed, 8 insertions(+), 3 deletions(-) diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 784e78c73ec5..84d64b6926a9 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -65,6 +65,7 @@ struct rpc_cred { #define RPC_CREDCACHE_NR (1 << RPC_CREDCACHE_HASHBITS) struct rpc_cred_cache { struct hlist_head hashtable[RPC_CREDCACHE_NR]; + unsigned int hashbits; spinlock_t lock; }; diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index db135543d21e..eef76a1f1dd6 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -145,12 +145,15 @@ int rpcauth_init_credcache(struct rpc_auth *auth) { struct rpc_cred_cache *new; + unsigned int hashsize; int i; new = kmalloc(sizeof(*new), GFP_KERNEL); if (!new) return -ENOMEM; - for (i = 0; i < RPC_CREDCACHE_NR; i++) + new->hashbits = RPC_CREDCACHE_HASHBITS; + hashsize = 1U << new->hashbits; + for (i = 0; i < hashsize; i++) INIT_HLIST_HEAD(&new->hashtable[i]); spin_lock_init(&new->lock); auth->au_credcache = new; @@ -183,11 +186,12 @@ rpcauth_clear_credcache(struct rpc_cred_cache *cache) LIST_HEAD(free); struct hlist_head *head; struct rpc_cred *cred; + unsigned int hashsize = 1U << cache->hashbits; int i; spin_lock(&rpc_credcache_lock); spin_lock(&cache->lock); - for (i = 0; i < RPC_CREDCACHE_NR; i++) { + for (i = 0; i < hashsize; i++) { head = &cache->hashtable[i]; while (!hlist_empty(head)) { cred = hlist_entry(head->first, struct rpc_cred, cr_hash); @@ -297,7 +301,7 @@ rpcauth_lookup_credcache(struct rpc_auth *auth, struct auth_cred * acred, *entry, *new; unsigned int nr; - nr = hash_long(acred->uid, RPC_CREDCACHE_HASHBITS); + nr = hash_long(acred->uid, cache->hashbits); rcu_read_lock(); hlist_for_each_entry_rcu(entry, pos, &cache->hashtable[nr], cr_hash) { From 241269bd0b580faae71575443d9ab38df7469126 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:08 -0400 Subject: [PATCH 31/42] SUNRPC: Make the credential cache hashtable size configurable This patch allows the user to configure the credential cache hashtable size using a new module parameter: auth_hashtable_size When set, this parameter will be rounded up to the nearest power of two, with a maximum allowed value of 1024 elements. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 9 +----- net/sunrpc/auth.c | 60 +++++++++++++++++++++++++++++++++---- 2 files changed, 56 insertions(+), 13 deletions(-) diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 84d64b6926a9..d2737625a247 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -61,14 +61,7 @@ struct rpc_cred { /* * Client authentication handle */ -#define RPC_CREDCACHE_HASHBITS 4 -#define RPC_CREDCACHE_NR (1 << RPC_CREDCACHE_HASHBITS) -struct rpc_cred_cache { - struct hlist_head hashtable[RPC_CREDCACHE_NR]; - unsigned int hashbits; - spinlock_t lock; -}; - +struct rpc_cred_cache; struct rpc_authops; struct rpc_auth { unsigned int au_cslack; /* call cred size estimate */ diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index eef76a1f1dd6..d80f01725fc2 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -19,6 +19,15 @@ # define RPCDBG_FACILITY RPCDBG_AUTH #endif +#define RPC_CREDCACHE_DEFAULT_HASHBITS (4) +struct rpc_cred_cache { + struct hlist_head *hashtable; + unsigned int hashbits; + spinlock_t lock; +}; + +static unsigned int auth_hashbits = RPC_CREDCACHE_DEFAULT_HASHBITS; + static DEFINE_SPINLOCK(rpc_authflavor_lock); static const struct rpc_authops *auth_flavors[RPC_AUTH_MAXFLAVOR] = { &authnull_ops, /* AUTH_NULL */ @@ -29,6 +38,42 @@ static const struct rpc_authops *auth_flavors[RPC_AUTH_MAXFLAVOR] = { static LIST_HEAD(cred_unused); static unsigned long number_cred_unused; +#define MAX_HASHTABLE_BITS (10) +static int param_set_hashtbl_sz(const char *val, struct kernel_param *kp) +{ + unsigned long num; + unsigned int nbits; + int ret; + + if (!val) + goto out_inval; + ret = strict_strtoul(val, 0, &num); + if (ret == -EINVAL) + goto out_inval; + nbits = fls(num); + if (num > (1U << nbits)) + nbits++; + if (nbits > MAX_HASHTABLE_BITS || nbits < 2) + goto out_inval; + *(unsigned int *)kp->arg = nbits; + return 0; +out_inval: + return -EINVAL; +} + +static int param_get_hashtbl_sz(char *buffer, struct kernel_param *kp) +{ + unsigned int nbits; + + nbits = *(unsigned int *)kp->arg; + return sprintf(buffer, "%u", 1U << nbits); +} + +#define param_check_hashtbl_sz(name, p) __param_check(name, p, unsigned int); + +module_param_named(auth_hashtable_size, auth_hashbits, hashtbl_sz, 0644); +MODULE_PARM_DESC(auth_hashtable_size, "RPC credential cache hashtable size"); + static u32 pseudoflavor_to_flavor(u32 flavor) { if (flavor >= RPC_AUTH_MAXFLAVOR) @@ -146,18 +191,22 @@ rpcauth_init_credcache(struct rpc_auth *auth) { struct rpc_cred_cache *new; unsigned int hashsize; - int i; new = kmalloc(sizeof(*new), GFP_KERNEL); if (!new) - return -ENOMEM; - new->hashbits = RPC_CREDCACHE_HASHBITS; + goto out_nocache; + new->hashbits = auth_hashbits; hashsize = 1U << new->hashbits; - for (i = 0; i < hashsize; i++) - INIT_HLIST_HEAD(&new->hashtable[i]); + new->hashtable = kcalloc(hashsize, sizeof(new->hashtable[0]), GFP_KERNEL); + if (!new->hashtable) + goto out_nohashtbl; spin_lock_init(&new->lock); auth->au_credcache = new; return 0; +out_nohashtbl: + kfree(new); +out_nocache: + return -ENOMEM; } EXPORT_SYMBOL_GPL(rpcauth_init_credcache); @@ -220,6 +269,7 @@ rpcauth_destroy_credcache(struct rpc_auth *auth) if (cache) { auth->au_credcache = NULL; rpcauth_clear_credcache(cache); + kfree(cache->hashtable); kfree(cache); } } From d9b6cd94601e1d17273f93a326a135fbf487a918 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:08 -0400 Subject: [PATCH 32/42] SUNRPC: Ensure that rpc_exit() always wakes up a sleeping task Make rpc_exit() non-inline, and ensure that it always wakes up a task that has been queued. Kill off the now unused rpc_wake_up_task(). Signed-off-by: Trond Myklebust --- include/linux/sunrpc/sched.h | 7 +------ net/sunrpc/sched.c | 20 +++++++++----------- 2 files changed, 10 insertions(+), 17 deletions(-) diff --git a/include/linux/sunrpc/sched.h b/include/linux/sunrpc/sched.h index 7be4f3a6d246..88513fd8e208 100644 --- a/include/linux/sunrpc/sched.h +++ b/include/linux/sunrpc/sched.h @@ -213,6 +213,7 @@ struct rpc_task *rpc_run_bc_task(struct rpc_rqst *req, const struct rpc_call_ops *ops); void rpc_put_task(struct rpc_task *); void rpc_exit_task(struct rpc_task *); +void rpc_exit(struct rpc_task *, int); void rpc_release_calldata(const struct rpc_call_ops *, void *); void rpc_killall_tasks(struct rpc_clnt *); void rpc_execute(struct rpc_task *); @@ -241,12 +242,6 @@ void rpc_destroy_mempool(void); extern struct workqueue_struct *rpciod_workqueue; void rpc_prepare_task(struct rpc_task *task); -static inline void rpc_exit(struct rpc_task *task, int status) -{ - task->tk_status = status; - task->tk_action = rpc_exit_task; -} - static inline int rpc_wait_for_completion_task(struct rpc_task *task) { return __rpc_wait_for_completion_task(task, NULL); diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 4a843b883b89..37452762af70 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -405,14 +405,6 @@ void rpc_wake_up_queued_task(struct rpc_wait_queue *queue, struct rpc_task *task } EXPORT_SYMBOL_GPL(rpc_wake_up_queued_task); -/* - * Wake up the specified task - */ -static void rpc_wake_up_task(struct rpc_task *task) -{ - rpc_wake_up_queued_task(task->tk_waitqueue, task); -} - /* * Wake up the next task on a priority queue. */ @@ -600,7 +592,15 @@ void rpc_exit_task(struct rpc_task *task) } } } -EXPORT_SYMBOL_GPL(rpc_exit_task); + +void rpc_exit(struct rpc_task *task, int status) +{ + task->tk_status = status; + task->tk_action = rpc_exit_task; + if (RPC_IS_QUEUED(task)) + rpc_wake_up_queued_task(task->tk_waitqueue, task); +} +EXPORT_SYMBOL_GPL(rpc_exit); void rpc_release_calldata(const struct rpc_call_ops *ops, void *calldata) { @@ -690,7 +690,6 @@ static void __rpc_execute(struct rpc_task *task) dprintk("RPC: %5u got signal\n", task->tk_pid); task->tk_flags |= RPC_TASK_KILLED; rpc_exit(task, -ERESTARTSYS); - rpc_wake_up_task(task); } rpc_set_running(task); dprintk("RPC: %5u sync task resuming\n", task->tk_pid); @@ -950,7 +949,6 @@ void rpc_killall_tasks(struct rpc_clnt *clnt) if (!(rovr->tk_flags & RPC_TASK_KILLED)) { rovr->tk_flags |= RPC_TASK_KILLED; rpc_exit(rovr, -EIO); - rpc_wake_up_task(rovr); } } spin_unlock(&clnt->cl_lock); From 58f9612c6ea858f532021a0ce42ec53cb0a493b3 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:08 -0400 Subject: [PATCH 33/42] SUNRPC: Move remaining RPC client related task initialisation into clnt.c Now that rpc_run_task() is the sole entry point for RPC calls, we can move the remaining rpc_client-related initialisation of struct rpc_task from sched.c into clnt.c. Also move rpc_killall_tasks() into the same file, since that too is relative to the rpc_clnt. Signed-off-by: Trond Myklebust --- include/linux/sunrpc/clnt.h | 1 + net/sunrpc/clnt.c | 84 +++++++++++++++++++++++++++++++++++++ net/sunrpc/sched.c | 77 ++-------------------------------- 3 files changed, 89 insertions(+), 73 deletions(-) diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h index debe75532199..569dc722a600 100644 --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -131,6 +131,7 @@ struct rpc_clnt *rpc_bind_new_program(struct rpc_clnt *, struct rpc_clnt *rpc_clone_client(struct rpc_clnt *); void rpc_shutdown_client(struct rpc_clnt *); void rpc_release_client(struct rpc_clnt *); +void rpc_task_release_client(struct rpc_task *); int rpcb_register(u32, u32, int, unsigned short); int rpcb_v4_register(const u32 program, const u32 version, diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 234c40c15f69..3647c81fd689 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -413,6 +413,35 @@ out_no_clnt: } EXPORT_SYMBOL_GPL(rpc_clone_client); +/* + * Kill all tasks for the given client. + * XXX: kill their descendants as well? + */ +void rpc_killall_tasks(struct rpc_clnt *clnt) +{ + struct rpc_task *rovr; + + + if (list_empty(&clnt->cl_tasks)) + return; + dprintk("RPC: killing all tasks for client %p\n", clnt); + /* + * Spin lock all_tasks to prevent changes... + */ + spin_lock(&clnt->cl_lock); + list_for_each_entry(rovr, &clnt->cl_tasks, tk_task) { + if (!RPC_IS_ACTIVATED(rovr)) + continue; + if (!(rovr->tk_flags & RPC_TASK_KILLED)) { + rovr->tk_flags |= RPC_TASK_KILLED; + rpc_exit(rovr, -EIO); + rpc_wake_up_queued_task(rovr->tk_waitqueue, rovr); + } + } + spin_unlock(&clnt->cl_lock); +} +EXPORT_SYMBOL_GPL(rpc_killall_tasks); + /* * Properly shut down an RPC client, terminating all outstanding * requests. @@ -538,6 +567,49 @@ out: } EXPORT_SYMBOL_GPL(rpc_bind_new_program); +void rpc_task_release_client(struct rpc_task *task) +{ + struct rpc_clnt *clnt = task->tk_client; + + if (clnt != NULL) { + /* Remove from client task list */ + spin_lock(&clnt->cl_lock); + list_del(&task->tk_task); + spin_unlock(&clnt->cl_lock); + task->tk_client = NULL; + + rpc_release_client(clnt); + } +} + +static +void rpc_task_set_client(struct rpc_task *task, struct rpc_clnt *clnt) +{ + if (clnt != NULL) { + rpc_task_release_client(task); + task->tk_client = clnt; + kref_get(&clnt->cl_kref); + if (clnt->cl_softrtry) + task->tk_flags |= RPC_TASK_SOFT; + /* Add to the client's list of all tasks */ + spin_lock(&clnt->cl_lock); + list_add_tail(&task->tk_task, &clnt->cl_tasks); + spin_unlock(&clnt->cl_lock); + } +} + +static void +rpc_task_set_rpc_message(struct rpc_task *task, const struct rpc_message *msg) +{ + if (msg != NULL) { + task->tk_msg.rpc_proc = msg->rpc_proc; + task->tk_msg.rpc_argp = msg->rpc_argp; + task->tk_msg.rpc_resp = msg->rpc_resp; + /* Bind the user cred */ + rpcauth_bindcred(task, msg->rpc_cred, task->tk_flags); + } +} + /* * Default callback for async RPC calls */ @@ -562,6 +634,18 @@ struct rpc_task *rpc_run_task(const struct rpc_task_setup *task_setup_data) if (IS_ERR(task)) goto out; + rpc_task_set_client(task, task_setup_data->rpc_client); + rpc_task_set_rpc_message(task, task_setup_data->rpc_message); + + if (task->tk_status != 0) { + int ret = task->tk_status; + rpc_put_task(task); + return ERR_PTR(ret); + } + + if (task->tk_action == NULL) + rpc_call_start(task); + atomic_inc(&task->tk_count); rpc_execute(task); out: diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index 37452762af70..a42296db2ecd 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -246,17 +246,8 @@ static inline void rpc_task_set_debuginfo(struct rpc_task *task) static void rpc_set_active(struct rpc_task *task) { - struct rpc_clnt *clnt; - if (test_and_set_bit(RPC_TASK_ACTIVE, &task->tk_runstate) != 0) - return; rpc_task_set_debuginfo(task); - /* Add to global list of all tasks */ - clnt = task->tk_client; - if (clnt != NULL) { - spin_lock(&clnt->cl_lock); - list_add_tail(&task->tk_task, &clnt->cl_tasks); - spin_unlock(&clnt->cl_lock); - } + set_bit(RPC_TASK_ACTIVE, &task->tk_runstate); } /* @@ -319,11 +310,6 @@ static void __rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task, dprintk("RPC: %5u sleep_on(queue \"%s\" time %lu)\n", task->tk_pid, rpc_qname(q), jiffies); - if (!RPC_IS_ASYNC(task) && !RPC_IS_ACTIVATED(task)) { - printk(KERN_ERR "RPC: Inactive synchronous task put to sleep!\n"); - return; - } - __rpc_add_wait_queue(q, task); BUG_ON(task->tk_callback != NULL); @@ -334,8 +320,8 @@ static void __rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task, void rpc_sleep_on(struct rpc_wait_queue *q, struct rpc_task *task, rpc_action action) { - /* Mark the task as being activated if so needed */ - rpc_set_active(task); + /* We shouldn't ever put an inactive task to sleep */ + BUG_ON(!RPC_IS_ACTIVATED(task)); /* * Protect the queue operations. @@ -807,26 +793,9 @@ static void rpc_init_task(struct rpc_task *task, const struct rpc_task_setup *ta /* Initialize workqueue for async tasks */ task->tk_workqueue = task_setup_data->workqueue; - task->tk_client = task_setup_data->rpc_client; - if (task->tk_client != NULL) { - kref_get(&task->tk_client->cl_kref); - if (task->tk_client->cl_softrtry) - task->tk_flags |= RPC_TASK_SOFT; - } - if (task->tk_ops->rpc_call_prepare != NULL) task->tk_action = rpc_prepare_task; - if (task_setup_data->rpc_message != NULL) { - task->tk_msg.rpc_proc = task_setup_data->rpc_message->rpc_proc; - task->tk_msg.rpc_argp = task_setup_data->rpc_message->rpc_argp; - task->tk_msg.rpc_resp = task_setup_data->rpc_message->rpc_resp; - /* Bind the user cred */ - rpcauth_bindcred(task, task_setup_data->rpc_message->rpc_cred, task_setup_data->flags); - if (task->tk_action == NULL) - rpc_call_start(task); - } - /* starting timestamp */ task->tk_start = ktime_get(); @@ -896,10 +865,7 @@ void rpc_put_task(struct rpc_task *task) xprt_release(task); if (task->tk_msg.rpc_cred) rpcauth_unbindcred(task); - if (task->tk_client) { - rpc_release_client(task->tk_client); - task->tk_client = NULL; - } + rpc_task_release_client(task); if (task->tk_workqueue != NULL) { INIT_WORK(&task->u.tk_work, rpc_async_release); queue_work(task->tk_workqueue, &task->u.tk_work); @@ -912,13 +878,6 @@ static void rpc_release_task(struct rpc_task *task) { dprintk("RPC: %5u release task\n", task->tk_pid); - if (!list_empty(&task->tk_task)) { - struct rpc_clnt *clnt = task->tk_client; - /* Remove from client task list */ - spin_lock(&clnt->cl_lock); - list_del(&task->tk_task); - spin_unlock(&clnt->cl_lock); - } BUG_ON (RPC_IS_QUEUED(task)); /* Wake up anyone who is waiting for task completion */ @@ -927,34 +886,6 @@ static void rpc_release_task(struct rpc_task *task) rpc_put_task(task); } -/* - * Kill all tasks for the given client. - * XXX: kill their descendants as well? - */ -void rpc_killall_tasks(struct rpc_clnt *clnt) -{ - struct rpc_task *rovr; - - - if (list_empty(&clnt->cl_tasks)) - return; - dprintk("RPC: killing all tasks for client %p\n", clnt); - /* - * Spin lock all_tasks to prevent changes... - */ - spin_lock(&clnt->cl_lock); - list_for_each_entry(rovr, &clnt->cl_tasks, tk_task) { - if (! RPC_IS_ACTIVATED(rovr)) - continue; - if (!(rovr->tk_flags & RPC_TASK_KILLED)) { - rovr->tk_flags |= RPC_TASK_KILLED; - rpc_exit(rovr, -EIO); - } - } - spin_unlock(&clnt->cl_lock); -} -EXPORT_SYMBOL_GPL(rpc_killall_tasks); - int rpciod_up(void) { return try_module_get(THIS_MODULE) ? 0 : -EINVAL; From 8572b8e2e3c5f3d990122348c4d2c64dad338611 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:08 -0400 Subject: [PATCH 34/42] SUNRPC: Clean up of rpc_bindcred() Signed-off-by: Trond Myklebust --- include/linux/sunrpc/auth.h | 6 +++--- net/sunrpc/auth.c | 37 +++++++++++++++++-------------------- net/sunrpc/auth_generic.c | 11 +++-------- net/sunrpc/clnt.c | 2 +- 4 files changed, 24 insertions(+), 32 deletions(-) diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index d2737625a247..90e4c3827ac0 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -106,7 +106,7 @@ struct rpc_credops { void (*crdestroy)(struct rpc_cred *); int (*crmatch)(struct auth_cred *, struct rpc_cred *, int); - void (*crbind)(struct rpc_task *, struct rpc_cred *, int); + struct rpc_cred * (*crbind)(struct rpc_task *, struct rpc_cred *, int); __be32 * (*crmarshal)(struct rpc_task *, __be32 *); int (*crrefresh)(struct rpc_task *); __be32 * (*crvalidate)(struct rpc_task *, __be32 *); @@ -135,8 +135,8 @@ void rpcauth_release(struct rpc_auth *); struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int); void rpcauth_init_cred(struct rpc_cred *, const struct auth_cred *, struct rpc_auth *, const struct rpc_credops *); struct rpc_cred * rpcauth_lookupcred(struct rpc_auth *, int); -void rpcauth_bindcred(struct rpc_task *, struct rpc_cred *, int); -void rpcauth_generic_bind_cred(struct rpc_task *, struct rpc_cred *, int); +int rpcauth_bindcred(struct rpc_task *, struct rpc_cred *, int); +struct rpc_cred * rpcauth_generic_bind_cred(struct rpc_task *, struct rpc_cred *, int); void put_rpccred(struct rpc_cred *); void rpcauth_unbindcred(struct rpc_task *); __be32 * rpcauth_marshcred(struct rpc_task *, __be32 *); diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index d80f01725fc2..d8968faf5ccf 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -444,16 +444,16 @@ rpcauth_init_cred(struct rpc_cred *cred, const struct auth_cred *acred, } EXPORT_SYMBOL_GPL(rpcauth_init_cred); -void +struct rpc_cred * rpcauth_generic_bind_cred(struct rpc_task *task, struct rpc_cred *cred, int lookupflags) { - task->tk_msg.rpc_cred = get_rpccred(cred); dprintk("RPC: %5u holding %s cred %p\n", task->tk_pid, cred->cr_auth->au_ops->au_name, cred); + return get_rpccred(cred); } EXPORT_SYMBOL_GPL(rpcauth_generic_bind_cred); -static void +static struct rpc_cred * rpcauth_bind_root_cred(struct rpc_task *task, int lookupflags) { struct rpc_auth *auth = task->tk_client->cl_auth; @@ -461,45 +461,42 @@ rpcauth_bind_root_cred(struct rpc_task *task, int lookupflags) .uid = 0, .gid = 0, }; - struct rpc_cred *ret; dprintk("RPC: %5u looking up %s cred\n", task->tk_pid, task->tk_client->cl_auth->au_ops->au_name); - ret = auth->au_ops->lookup_cred(auth, &acred, lookupflags); - if (!IS_ERR(ret)) - task->tk_msg.rpc_cred = ret; - else - task->tk_status = PTR_ERR(ret); + return auth->au_ops->lookup_cred(auth, &acred, lookupflags); } -static void +static struct rpc_cred * rpcauth_bind_new_cred(struct rpc_task *task, int lookupflags) { struct rpc_auth *auth = task->tk_client->cl_auth; - struct rpc_cred *ret; dprintk("RPC: %5u looking up %s cred\n", task->tk_pid, auth->au_ops->au_name); - ret = rpcauth_lookupcred(auth, lookupflags); - if (!IS_ERR(ret)) - task->tk_msg.rpc_cred = ret; - else - task->tk_status = PTR_ERR(ret); + return rpcauth_lookupcred(auth, lookupflags); } -void +int rpcauth_bindcred(struct rpc_task *task, struct rpc_cred *cred, int flags) { + struct rpc_cred *new; int lookupflags = 0; if (flags & RPC_TASK_ASYNC) lookupflags |= RPCAUTH_LOOKUP_NEW; if (cred != NULL) - cred->cr_ops->crbind(task, cred, lookupflags); + new = cred->cr_ops->crbind(task, cred, lookupflags); else if (flags & RPC_TASK_ROOTCREDS) - rpcauth_bind_root_cred(task, lookupflags); + new = rpcauth_bind_root_cred(task, lookupflags); else - rpcauth_bind_new_cred(task, lookupflags); + new = rpcauth_bind_new_cred(task, lookupflags); + if (IS_ERR(new)) + return PTR_ERR(new); + if (task->tk_msg.rpc_cred != NULL) + put_rpccred(task->tk_msg.rpc_cred); + task->tk_msg.rpc_cred = new; + return 0; } void diff --git a/net/sunrpc/auth_generic.c b/net/sunrpc/auth_generic.c index 8bae33b36cc6..43162bb3b78f 100644 --- a/net/sunrpc/auth_generic.c +++ b/net/sunrpc/auth_generic.c @@ -54,18 +54,13 @@ struct rpc_cred *rpc_lookup_machine_cred(void) } EXPORT_SYMBOL_GPL(rpc_lookup_machine_cred); -static void -generic_bind_cred(struct rpc_task *task, struct rpc_cred *cred, int lookupflags) +static struct rpc_cred *generic_bind_cred(struct rpc_task *task, + struct rpc_cred *cred, int lookupflags) { struct rpc_auth *auth = task->tk_client->cl_auth; struct auth_cred *acred = &container_of(cred, struct generic_cred, gc_base)->acred; - struct rpc_cred *ret; - ret = auth->au_ops->lookup_cred(auth, acred, lookupflags); - if (!IS_ERR(ret)) - task->tk_msg.rpc_cred = ret; - else - task->tk_status = PTR_ERR(ret); + return auth->au_ops->lookup_cred(auth, acred, lookupflags); } /* diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index 3647c81fd689..f34b5e3823c0 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -606,7 +606,7 @@ rpc_task_set_rpc_message(struct rpc_task *task, const struct rpc_message *msg) task->tk_msg.rpc_argp = msg->rpc_argp; task->tk_msg.rpc_resp = msg->rpc_resp; /* Bind the user cred */ - rpcauth_bindcred(task, msg->rpc_cred, task->tk_flags); + task->tk_status = rpcauth_bindcred(task, msg->rpc_cred, task->tk_flags); } } From a17c2153d2e271b0cbacae9bed83b0eaa41db7e1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:08 -0400 Subject: [PATCH 35/42] SUNRPC: Move the bound cred to struct rpc_rqst This will allow us to save the original generic cred in rpc_message, so that if we migrate from one server to another, we can generate a new bound cred without having to punt back to the NFS layer. Signed-off-by: Trond Myklebust --- fs/nfs/nfs2xdr.c | 7 ++- fs/nfs/nfs3xdr.c | 8 +-- fs/nfs/nfs4xdr.c | 2 +- include/linux/sunrpc/auth.h | 2 - include/linux/sunrpc/xprt.h | 1 + net/sunrpc/auth.c | 43 ++++++++-------- net/sunrpc/auth_gss/auth_gss.c | 22 ++++---- net/sunrpc/auth_null.c | 2 +- net/sunrpc/auth_unix.c | 6 +-- net/sunrpc/clnt.c | 91 +++++++++++++++++----------------- net/sunrpc/sched.c | 2 +- net/sunrpc/xprt.c | 2 + 12 files changed, 92 insertions(+), 96 deletions(-) diff --git a/fs/nfs/nfs2xdr.c b/fs/nfs/nfs2xdr.c index 81cf14257916..db8846a0e82e 100644 --- a/fs/nfs/nfs2xdr.c +++ b/fs/nfs/nfs2xdr.c @@ -233,7 +233,7 @@ nfs_xdr_removeargs(struct rpc_rqst *req, __be32 *p, const struct nfs_removeargs static int nfs_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) { - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; u32 offset = (u32)args->offset; u32 count = args->count; @@ -393,8 +393,7 @@ nfs_xdr_symlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs_symlinkargs *arg static int nfs_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs_readdirargs *args) { - struct rpc_task *task = req->rq_task; - struct rpc_auth *auth = task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; u32 count = args->count; @@ -575,7 +574,7 @@ nfs_xdr_diropres(struct rpc_rqst *req, __be32 *p, struct nfs_diropok *res) static int nfs_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs_readlinkargs *args) { - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; p = xdr_encode_fhandle(p, args->fh); diff --git a/fs/nfs/nfs3xdr.c b/fs/nfs/nfs3xdr.c index 75dcfc7da365..9769704f8ce6 100644 --- a/fs/nfs/nfs3xdr.c +++ b/fs/nfs/nfs3xdr.c @@ -330,7 +330,7 @@ nfs3_xdr_accessargs(struct rpc_rqst *req, __be32 *p, struct nfs3_accessargs *arg static int nfs3_xdr_readargs(struct rpc_rqst *req, __be32 *p, struct nfs_readargs *args) { - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; u32 count = args->count; @@ -471,7 +471,7 @@ nfs3_xdr_linkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_linkargs *args) static int nfs3_xdr_readdirargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readdirargs *args) { - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; u32 count = args->count; @@ -675,7 +675,7 @@ static int nfs3_xdr_getaclargs(struct rpc_rqst *req, __be32 *p, struct nfs3_getaclargs *args) { - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; p = xdr_encode_fhandle(p, args->fh); @@ -802,7 +802,7 @@ nfs3_xdr_accessres(struct rpc_rqst *req, __be32 *p, struct nfs3_accessres *res) static int nfs3_xdr_readlinkargs(struct rpc_rqst *req, __be32 *p, struct nfs3_readlinkargs *args) { - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; unsigned int replen; p = xdr_encode_fhandle(p, args->fh); diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 257c1811feb4..08ef91291132 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -758,7 +758,7 @@ static void encode_compound_hdr(struct xdr_stream *xdr, struct compound_hdr *hdr) { __be32 *p; - struct rpc_auth *auth = req->rq_task->tk_msg.rpc_cred->cr_auth; + struct rpc_auth *auth = req->rq_cred->cr_auth; /* initialize running count of expected bytes in reply. * NOTE: the replied tag SHOULD be the same is the one sent, diff --git a/include/linux/sunrpc/auth.h b/include/linux/sunrpc/auth.h index 90e4c3827ac0..5bbc447175dc 100644 --- a/include/linux/sunrpc/auth.h +++ b/include/linux/sunrpc/auth.h @@ -135,10 +135,8 @@ void rpcauth_release(struct rpc_auth *); struct rpc_cred * rpcauth_lookup_credcache(struct rpc_auth *, struct auth_cred *, int); void rpcauth_init_cred(struct rpc_cred *, const struct auth_cred *, struct rpc_auth *, const struct rpc_credops *); struct rpc_cred * rpcauth_lookupcred(struct rpc_auth *, int); -int rpcauth_bindcred(struct rpc_task *, struct rpc_cred *, int); struct rpc_cred * rpcauth_generic_bind_cred(struct rpc_task *, struct rpc_cred *, int); void put_rpccred(struct rpc_cred *); -void rpcauth_unbindcred(struct rpc_task *); __be32 * rpcauth_marshcred(struct rpc_task *, __be32 *); __be32 * rpcauth_checkverf(struct rpc_task *, __be32 *); int rpcauth_wrap_req(struct rpc_task *task, kxdrproc_t encode, void *rqstp, __be32 *data, void *obj); diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h index b51470302399..ff5a77b28c50 100644 --- a/include/linux/sunrpc/xprt.h +++ b/include/linux/sunrpc/xprt.h @@ -64,6 +64,7 @@ struct rpc_rqst { * This is the private part */ struct rpc_task * rq_task; /* RPC task data */ + struct rpc_cred * rq_cred; /* Bound cred */ __be32 rq_xid; /* request XID */ int rq_cong; /* has incremented xprt->cong */ u32 rq_seqno; /* gss seq no. used on req. */ diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c index d8968faf5ccf..95721426296d 100644 --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -477,9 +477,10 @@ rpcauth_bind_new_cred(struct rpc_task *task, int lookupflags) return rpcauth_lookupcred(auth, lookupflags); } -int +static int rpcauth_bindcred(struct rpc_task *task, struct rpc_cred *cred, int flags) { + struct rpc_rqst *req = task->tk_rqstp; struct rpc_cred *new; int lookupflags = 0; @@ -493,9 +494,9 @@ rpcauth_bindcred(struct rpc_task *task, struct rpc_cred *cred, int flags) new = rpcauth_bind_new_cred(task, lookupflags); if (IS_ERR(new)) return PTR_ERR(new); - if (task->tk_msg.rpc_cred != NULL) - put_rpccred(task->tk_msg.rpc_cred); - task->tk_msg.rpc_cred = new; + if (req->rq_cred != NULL) + put_rpccred(req->rq_cred); + req->rq_cred = new; return 0; } @@ -535,22 +536,10 @@ out_nodestroy: } EXPORT_SYMBOL_GPL(put_rpccred); -void -rpcauth_unbindcred(struct rpc_task *task) -{ - struct rpc_cred *cred = task->tk_msg.rpc_cred; - - dprintk("RPC: %5u releasing %s cred %p\n", - task->tk_pid, cred->cr_auth->au_ops->au_name, cred); - - put_rpccred(cred); - task->tk_msg.rpc_cred = NULL; -} - __be32 * rpcauth_marshcred(struct rpc_task *task, __be32 *p) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; dprintk("RPC: %5u marshaling %s cred %p\n", task->tk_pid, cred->cr_auth->au_ops->au_name, cred); @@ -561,7 +550,7 @@ rpcauth_marshcred(struct rpc_task *task, __be32 *p) __be32 * rpcauth_checkverf(struct rpc_task *task, __be32 *p) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; dprintk("RPC: %5u validating %s cred %p\n", task->tk_pid, cred->cr_auth->au_ops->au_name, cred); @@ -573,7 +562,7 @@ int rpcauth_wrap_req(struct rpc_task *task, kxdrproc_t encode, void *rqstp, __be32 *data, void *obj) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; dprintk("RPC: %5u using %s cred %p to wrap rpc data\n", task->tk_pid, cred->cr_ops->cr_name, cred); @@ -587,7 +576,7 @@ int rpcauth_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp, __be32 *data, void *obj) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; dprintk("RPC: %5u using %s cred %p to unwrap rpc data\n", task->tk_pid, cred->cr_ops->cr_name, cred); @@ -601,13 +590,21 @@ rpcauth_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp, int rpcauth_refreshcred(struct rpc_task *task) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; int err; + cred = task->tk_rqstp->rq_cred; + if (cred == NULL) { + err = rpcauth_bindcred(task, task->tk_msg.rpc_cred, task->tk_flags); + if (err < 0) + goto out; + cred = task->tk_rqstp->rq_cred; + }; dprintk("RPC: %5u refreshing %s cred %p\n", task->tk_pid, cred->cr_auth->au_ops->au_name, cred); err = cred->cr_ops->crrefresh(task); +out: if (err < 0) task->tk_status = err; return err; @@ -616,7 +613,7 @@ rpcauth_refreshcred(struct rpc_task *task) void rpcauth_invalcred(struct rpc_task *task) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; dprintk("RPC: %5u invalidating %s cred %p\n", task->tk_pid, cred->cr_auth->au_ops->au_name, cred); @@ -627,7 +624,7 @@ rpcauth_invalcred(struct rpc_task *task) int rpcauth_uptodatecred(struct rpc_task *task) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; return cred == NULL || test_bit(RPCAUTH_CRED_UPTODATE, &cred->cr_flags) != 0; diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 8da2a0e68574..096e1260bc67 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -373,7 +373,7 @@ gss_handle_downcall_result(struct gss_cred *gss_cred, struct gss_upcall_msg *gss static void gss_upcall_callback(struct rpc_task *task) { - struct gss_cred *gss_cred = container_of(task->tk_msg.rpc_cred, + struct gss_cred *gss_cred = container_of(task->tk_rqstp->rq_cred, struct gss_cred, gc_base); struct gss_upcall_msg *gss_msg = gss_cred->gc_upcall; struct inode *inode = &gss_msg->inode->vfs_inode; @@ -502,7 +502,7 @@ static void warn_gssd(void) static inline int gss_refresh_upcall(struct rpc_task *task) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; struct gss_auth *gss_auth = container_of(cred->cr_auth, struct gss_auth, rpc_auth); struct gss_cred *gss_cred = container_of(cred, @@ -1064,12 +1064,12 @@ out: static __be32 * gss_marshal(struct rpc_task *task, __be32 *p) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_cred *cred = req->rq_cred; struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); struct gss_cl_ctx *ctx = gss_cred_get_ctx(cred); __be32 *cred_len; - struct rpc_rqst *req = task->tk_rqstp; u32 maj_stat = 0; struct xdr_netobj mic; struct kvec iov; @@ -1119,7 +1119,7 @@ out_put_ctx: static int gss_renew_cred(struct rpc_task *task) { - struct rpc_cred *oldcred = task->tk_msg.rpc_cred; + struct rpc_cred *oldcred = task->tk_rqstp->rq_cred; struct gss_cred *gss_cred = container_of(oldcred, struct gss_cred, gc_base); @@ -1133,7 +1133,7 @@ static int gss_renew_cred(struct rpc_task *task) new = gss_lookup_cred(auth, &acred, RPCAUTH_LOOKUP_NEW); if (IS_ERR(new)) return PTR_ERR(new); - task->tk_msg.rpc_cred = new; + task->tk_rqstp->rq_cred = new; put_rpccred(oldcred); return 0; } @@ -1161,7 +1161,7 @@ static int gss_cred_is_negative_entry(struct rpc_cred *cred) static int gss_refresh(struct rpc_task *task) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; int ret = 0; if (gss_cred_is_negative_entry(cred)) @@ -1172,7 +1172,7 @@ gss_refresh(struct rpc_task *task) ret = gss_renew_cred(task); if (ret < 0) goto out; - cred = task->tk_msg.rpc_cred; + cred = task->tk_rqstp->rq_cred; } if (test_bit(RPCAUTH_CRED_NEW, &cred->cr_flags)) @@ -1191,7 +1191,7 @@ gss_refresh_null(struct rpc_task *task) static __be32 * gss_validate(struct rpc_task *task, __be32 *p) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; struct gss_cl_ctx *ctx = gss_cred_get_ctx(cred); __be32 seq; struct kvec iov; @@ -1400,7 +1400,7 @@ static int gss_wrap_req(struct rpc_task *task, kxdrproc_t encode, void *rqstp, __be32 *p, void *obj) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); struct gss_cl_ctx *ctx = gss_cred_get_ctx(cred); @@ -1503,7 +1503,7 @@ static int gss_unwrap_resp(struct rpc_task *task, kxdrproc_t decode, void *rqstp, __be32 *p, void *obj) { - struct rpc_cred *cred = task->tk_msg.rpc_cred; + struct rpc_cred *cred = task->tk_rqstp->rq_cred; struct gss_cred *gss_cred = container_of(cred, struct gss_cred, gc_base); struct gss_cl_ctx *ctx = gss_cred_get_ctx(cred); diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c index 1db618f56ecb..a5c36c01707b 100644 --- a/net/sunrpc/auth_null.c +++ b/net/sunrpc/auth_null.c @@ -75,7 +75,7 @@ nul_marshal(struct rpc_task *task, __be32 *p) static int nul_refresh(struct rpc_task *task) { - set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_msg.rpc_cred->cr_flags); + set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_rqstp->rq_cred->cr_flags); return 0; } diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c index d5e37dbf207b..4cb70dc6e7ad 100644 --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -140,7 +140,7 @@ static __be32 * unx_marshal(struct rpc_task *task, __be32 *p) { struct rpc_clnt *clnt = task->tk_client; - struct unx_cred *cred = container_of(task->tk_msg.rpc_cred, struct unx_cred, uc_base); + struct unx_cred *cred = container_of(task->tk_rqstp->rq_cred, struct unx_cred, uc_base); __be32 *base, *hold; int i; @@ -173,7 +173,7 @@ unx_marshal(struct rpc_task *task, __be32 *p) static int unx_refresh(struct rpc_task *task) { - set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_msg.rpc_cred->cr_flags); + set_bit(RPCAUTH_CRED_UPTODATE, &task->tk_rqstp->rq_cred->cr_flags); return 0; } @@ -196,7 +196,7 @@ unx_validate(struct rpc_task *task, __be32 *p) printk("RPC: giant verf size: %u\n", size); return NULL; } - task->tk_msg.rpc_cred->cr_auth->au_rslack = (size >> 2) + 2; + task->tk_rqstp->rq_cred->cr_auth->au_rslack = (size >> 2) + 2; p += (size >> 2); return p; diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c index f34b5e3823c0..2388d83b68ff 100644 --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -605,8 +605,8 @@ rpc_task_set_rpc_message(struct rpc_task *task, const struct rpc_message *msg) task->tk_msg.rpc_proc = msg->rpc_proc; task->tk_msg.rpc_argp = msg->rpc_argp; task->tk_msg.rpc_resp = msg->rpc_resp; - /* Bind the user cred */ - task->tk_status = rpcauth_bindcred(task, msg->rpc_cred, task->tk_flags); + if (msg->rpc_cred != NULL) + task->tk_msg.rpc_cred = get_rpccred(msg->rpc_cred); } } @@ -909,11 +909,6 @@ call_reserve(struct rpc_task *task) { dprint_status(task); - if (!rpcauth_uptodatecred(task)) { - task->tk_action = call_refresh; - return; - } - task->tk_status = 0; task->tk_action = call_reserveresult; xprt_reserve(task); @@ -977,7 +972,7 @@ call_reserveresult(struct rpc_task *task) static void call_allocate(struct rpc_task *task) { - unsigned int slack = task->tk_msg.rpc_cred->cr_auth->au_cslack; + unsigned int slack = task->tk_client->cl_auth->au_cslack; struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = task->tk_xprt; struct rpc_procinfo *proc = task->tk_msg.rpc_proc; @@ -985,7 +980,7 @@ call_allocate(struct rpc_task *task) dprint_status(task); task->tk_status = 0; - task->tk_action = call_bind; + task->tk_action = call_refresh; if (req->rq_buffer) return; @@ -1022,6 +1017,47 @@ call_allocate(struct rpc_task *task) rpc_exit(task, -ERESTARTSYS); } +/* + * 2a. Bind and/or refresh the credentials + */ +static void +call_refresh(struct rpc_task *task) +{ + dprint_status(task); + + task->tk_action = call_refreshresult; + task->tk_status = 0; + task->tk_client->cl_stats->rpcauthrefresh++; + rpcauth_refreshcred(task); +} + +/* + * 2b. Process the results of a credential refresh + */ +static void +call_refreshresult(struct rpc_task *task) +{ + int status = task->tk_status; + + dprint_status(task); + + task->tk_status = 0; + task->tk_action = call_bind; + if (status >= 0 && rpcauth_uptodatecred(task)) + return; + switch (status) { + case -EACCES: + rpc_exit(task, -EACCES); + return; + case -ENOMEM: + rpc_exit(task, -ENOMEM); + return; + case -ETIMEDOUT: + rpc_delay(task, 3*HZ); + } + task->tk_action = call_refresh; +} + static inline int rpc_task_need_encode(struct rpc_task *task) { @@ -1557,43 +1593,6 @@ out_retry: } } -/* - * 8. Refresh the credentials if rejected by the server - */ -static void -call_refresh(struct rpc_task *task) -{ - dprint_status(task); - - task->tk_action = call_refreshresult; - task->tk_status = 0; - task->tk_client->cl_stats->rpcauthrefresh++; - rpcauth_refreshcred(task); -} - -/* - * 8a. Process the results of a credential refresh - */ -static void -call_refreshresult(struct rpc_task *task) -{ - int status = task->tk_status; - - dprint_status(task); - - task->tk_status = 0; - task->tk_action = call_reserve; - if (status >= 0 && rpcauth_uptodatecred(task)) - return; - if (status == -EACCES) { - rpc_exit(task, -EACCES); - return; - } - task->tk_action = call_refresh; - if (status != -ETIMEDOUT) - rpc_delay(task, 3*HZ); -} - static __be32 * rpc_encode_header(struct rpc_task *task) { diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index a42296db2ecd..f6db6131fb2e 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -864,7 +864,7 @@ void rpc_put_task(struct rpc_task *task) if (task->tk_rqstp) xprt_release(task); if (task->tk_msg.rpc_cred) - rpcauth_unbindcred(task); + put_rpccred(task->tk_msg.rpc_cred); rpc_task_release_client(task); if (task->tk_workqueue != NULL) { INIT_WORK(&task->u.tk_work, rpc_async_release); diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c index dcd0132396ba..70297836a191 100644 --- a/net/sunrpc/xprt.c +++ b/net/sunrpc/xprt.c @@ -1032,6 +1032,8 @@ void xprt_release(struct rpc_task *task) spin_unlock_bh(&xprt->transport_lock); if (req->rq_buffer) xprt->ops->buf_free(req->rq_buffer); + if (req->rq_cred != NULL) + put_rpccred(req->rq_cred); task->tk_rqstp = NULL; if (req->rq_release_snd_buf) req->rq_release_snd_buf(req); From d6a1ed08c6acea647b3f2aaebe0261f079e1be49 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 31 Jul 2010 14:29:08 -0400 Subject: [PATCH 36/42] SUNRPC: Reduce asynchronous RPC task stack usage We should just farm out asynchronous RPC tasks immediately to rpciod... Signed-off-by: Trond Myklebust --- net/sunrpc/sched.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c index f6db6131fb2e..cace6049e4a5 100644 --- a/net/sunrpc/sched.c +++ b/net/sunrpc/sched.c @@ -699,8 +699,9 @@ static void __rpc_execute(struct rpc_task *task) void rpc_execute(struct rpc_task *task) { rpc_set_active(task); - rpc_set_running(task); - __rpc_execute(task); + rpc_make_runnable(task); + if (!RPC_IS_ASYNC(task)) + __rpc_execute(task); } static void rpc_async_schedule(struct work_struct *work) From c3ae62ae08bb0db3639d8c579e4ff0967d908199 Mon Sep 17 00:00:00 2001 From: "J. Bruce Fields" Date: Tue, 3 Aug 2010 17:22:20 -0400 Subject: [PATCH 37/42] SUNRPC: prevent task_cleanup running on freed xprt We saw a report of a NULL dereference in xprt_autoclose: https://bugzilla.redhat.com/show_bug.cgi?id=611938 This appears to be the result of an xprt's task_cleanup running after the xprt is destroyed. Nothing in the current code appears to prevent that. Signed-off-by: J. Bruce Fields Signed-off-by: Trond Myklebust --- net/sunrpc/xprt.c | 1 + 1 file changed, 1 insertion(+) diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c index 70297836a191..970fb00f388c 100644 --- a/net/sunrpc/xprt.c +++ b/net/sunrpc/xprt.c @@ -1131,6 +1131,7 @@ static void xprt_destroy(struct kref *kref) rpc_destroy_wait_queue(&xprt->sending); rpc_destroy_wait_queue(&xprt->resend); rpc_destroy_wait_queue(&xprt->backlog); + cancel_work_sync(&xprt->task_cleanup); /* * Tear down transport state and free the rpc_xprt */ From 0d8a374673c9bc62bc78a2d8fe64553a51542cb4 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sun, 1 Aug 2010 20:14:55 -0400 Subject: [PATCH 38/42] SUNRPC: Defer deleting the security context until gss_do_free_ctx() There is no need to delete the gss context separately from the rest of the security context information, and doing so gives rise to a an rcu_dereference_check() warning. Signed-off-by: Trond Myklebust --- net/sunrpc/auth_gss/auth_gss.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c index 096e1260bc67..dcfc66bab2bb 100644 --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -928,6 +928,7 @@ gss_do_free_ctx(struct gss_cl_ctx *ctx) { dprintk("RPC: gss_free_ctx\n"); + gss_delete_sec_context(&ctx->gc_gss_ctx); kfree(ctx->gc_wire_ctx.data); kfree(ctx); } @@ -942,13 +943,7 @@ gss_free_ctx_callback(struct rcu_head *head) static void gss_free_ctx(struct gss_cl_ctx *ctx) { - struct gss_ctx *gc_gss_ctx; - - gc_gss_ctx = rcu_dereference(ctx->gc_gss_ctx); - rcu_assign_pointer(ctx->gc_gss_ctx, NULL); call_rcu(&ctx->gc_rcu, gss_free_ctx_callback); - if (gc_gss_ctx) - gss_delete_sec_context(&gc_gss_ctx); } static void From 761fe93cdfa29071879d882c92e966ae692c0048 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 4 Aug 2010 09:29:52 -0400 Subject: [PATCH 39/42] NFS: Fix the locking in nfs4_callback_getattr The delegation is protected by RCU now, so we need to replace the nfsi->rwsem protection with an rcu protected section. Reported-by: Fred Isaman Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 7445dd0ae3f3..930d10fecdaf 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -37,8 +37,8 @@ __be32 nfs4_callback_getattr(struct cb_getattrargs *args, struct cb_getattrres * if (inode == NULL) goto out_putclient; nfsi = NFS_I(inode); - down_read(&nfsi->rwsem); - delegation = nfsi->delegation; + rcu_read_lock(); + delegation = rcu_dereference(nfsi->delegation); if (delegation == NULL || (delegation->type & FMODE_WRITE) == 0) goto out_iput; res->size = i_size_read(inode); @@ -53,7 +53,7 @@ __be32 nfs4_callback_getattr(struct cb_getattrargs *args, struct cb_getattrres * args->bitmap[1]; res->status = 0; out_iput: - up_read(&nfsi->rwsem); + rcu_read_unlock(); iput(inode); out_putclient: nfs_put_client(clp); From d5eff1a3412f6d75bf28f423c5015ece8055407a Mon Sep 17 00:00:00 2001 From: Bryan Schumaker Date: Tue, 3 Aug 2010 13:04:00 -0400 Subject: [PATCH 40/42] NFS: Fix /proc/mount for legacy binary interface Add a flag so we know if we mounted the NFS server using the legacy binary interface. If we used the legacy interface, then we should not show the mountd options. Signed-off-by: Bryan Schumaker Signed-off-by: Trond Myklebust --- fs/nfs/super.c | 4 ++++ include/linux/nfs_mount.h | 1 + 2 files changed, 5 insertions(+) diff --git a/fs/nfs/super.c b/fs/nfs/super.c index f9df16de4a56..f1ae39f6cb02 100644 --- a/fs/nfs/super.c +++ b/fs/nfs/super.c @@ -546,6 +546,9 @@ static void nfs_show_mountd_options(struct seq_file *m, struct nfs_server *nfss, { struct sockaddr *sap = (struct sockaddr *)&nfss->mountd_address; + if (nfss->flags & NFS_MOUNT_LEGACY_INTERFACE) + return; + switch (sap->sa_family) { case AF_INET: { struct sockaddr_in *sin = (struct sockaddr_in *)sap; @@ -1780,6 +1783,7 @@ static int nfs_validate_mount_data(void *options, * can deal with. */ args->flags = data->flags & NFS_MOUNT_FLAGMASK; + args->flags |= NFS_MOUNT_LEGACY_INTERFACE; args->rsize = data->rsize; args->wsize = data->wsize; args->timeo = data->timeo; diff --git a/include/linux/nfs_mount.h b/include/linux/nfs_mount.h index 4499016e6d0d..5d59ae861aa6 100644 --- a/include/linux/nfs_mount.h +++ b/include/linux/nfs_mount.h @@ -69,5 +69,6 @@ struct nfs_mount_data { #define NFS_MOUNT_LOOKUP_CACHE_NONEG 0x10000 #define NFS_MOUNT_LOOKUP_CACHE_NONE 0x20000 #define NFS_MOUNT_NORESVPORT 0x40000 +#define NFS_MOUNT_LEGACY_INTERFACE 0x80000 #endif From b3edc2bc19b4856c705f4aea3f5472970b99f386 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 4 Aug 2010 14:38:01 -0400 Subject: [PATCH 41/42] NFS: NFS_V4 is no longer an EXPERIMENTAL feature Signed-off-by: Trond Myklebust --- fs/nfs/Kconfig | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/nfs/Kconfig b/fs/nfs/Kconfig index a43d07e7b924..d55f3f5371b2 100644 --- a/fs/nfs/Kconfig +++ b/fs/nfs/Kconfig @@ -61,8 +61,8 @@ config NFS_V3_ACL If unsure, say N. config NFS_V4 - bool "NFS client support for NFS version 4 (EXPERIMENTAL)" - depends on NFS_FS && EXPERIMENTAL + bool "NFS client support for NFS version 4" + depends on NFS_FS select RPCSEC_GSS_KRB5 help This option enables support for version 4 of the NFS protocol @@ -72,7 +72,7 @@ config NFS_V4 space programs which can be found in the Linux nfs-utils package, available from http://linux-nfs.org/. - If unsure, say N. + If unsure, say Y. config NFS_V4_1 bool "NFS client support for NFSv4.1 (DEVELOPER ONLY)" From 3dce9a5c3a39a664e372886ea86c42ae7ae33dfc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Wed, 4 Aug 2010 14:39:16 -0400 Subject: [PATCH 42/42] NFS: NFSv4.1 is no longer a "developer only" feature Mark it as 'experimental' instead, since in practice, NFSv4.1 should now be relatively stable. Signed-off-by: Trond Myklebust --- fs/nfs/Kconfig | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/nfs/Kconfig b/fs/nfs/Kconfig index d55f3f5371b2..cc1bb33b59b8 100644 --- a/fs/nfs/Kconfig +++ b/fs/nfs/Kconfig @@ -75,13 +75,13 @@ config NFS_V4 If unsure, say Y. config NFS_V4_1 - bool "NFS client support for NFSv4.1 (DEVELOPER ONLY)" + bool "NFS client support for NFSv4.1 (EXPERIMENTAL)" depends on NFS_V4 && EXPERIMENTAL help This option enables support for minor version 1 of the NFSv4 protocol (draft-ietf-nfsv4-minorversion1) in the kernel's NFS client. - Unless you're an NFS developer, say N. + If unsure, say N. config ROOT_NFS bool "Root file system on NFS"