From 6b1891052a3f8e3c3217e8512bbed2fd6252977b Mon Sep 17 00:00:00 2001 From: Scott Mayhew Date: Tue, 26 Mar 2019 18:06:26 -0400 Subject: nfsd: make nfs4_client_reclaim use an xdr_netobj instead of a fixed char array This will allow the reclaim_str_hashtbl to store either the recovery directory names used by the legacy client tracking code or the full client strings used by the nfsdcld client tracking code. Signed-off-by: Scott Mayhew Signed-off-by: J. Bruce Fields --- fs/nfsd/nfs4recover.c | 109 +++++++++++++++++++++++++++++++++++++++++--------- 1 file changed, 91 insertions(+), 18 deletions(-) (limited to 'fs/nfsd/nfs4recover.c') diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c index 5188f9f70c78..2243b909b407 100644 --- a/fs/nfsd/nfs4recover.c +++ b/fs/nfsd/nfs4recover.c @@ -169,13 +169,34 @@ legacy_recdir_name_error(struct nfs4_client *clp, int error) } } +static void +__nfsd4_create_reclaim_record_grace(struct nfs4_client *clp, + const char *dname, int len, struct nfsd_net *nn) +{ + struct xdr_netobj name; + struct nfs4_client_reclaim *crp; + + name.data = kmemdup(dname, len, GFP_KERNEL); + if (!name.data) { + dprintk("%s: failed to allocate memory for name.data!\n", + __func__); + return; + } + name.len = len; + crp = nfs4_client_to_reclaim(name, nn); + if (!crp) { + kfree(name.data); + return; + } + crp->cr_clp = clp; +} + static void nfsd4_create_clid_dir(struct nfs4_client *clp) { const struct cred *original_cred; char dname[HEXDIR_LEN]; struct dentry *dir, *dentry; - struct nfs4_client_reclaim *crp; int status; struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); @@ -221,11 +242,9 @@ out_put: out_unlock: inode_unlock(d_inode(dir)); if (status == 0) { - if (nn->in_grace) { - crp = nfs4_client_to_reclaim(dname, nn); - if (crp) - crp->cr_clp = clp; - } + if (nn->in_grace) + __nfsd4_create_reclaim_record_grace(clp, dname, + HEXDIR_LEN, nn); vfs_fsync(nn->rec_file, 0); } else { printk(KERN_ERR "NFSD: failed to write recovery record" @@ -345,11 +364,30 @@ out_unlock: return status; } +static void +__nfsd4_remove_reclaim_record_grace(const char *dname, int len, + struct nfsd_net *nn) +{ + struct xdr_netobj name; + struct nfs4_client_reclaim *crp; + + name.data = kmemdup(dname, len, GFP_KERNEL); + if (!name.data) { + dprintk("%s: failed to allocate memory for name.data!\n", + __func__); + return; + } + name.len = len; + crp = nfsd4_find_reclaim_client(name, nn); + kfree(name.data); + if (crp) + nfs4_remove_reclaim_record(crp, nn); +} + static void nfsd4_remove_clid_dir(struct nfs4_client *clp) { const struct cred *original_cred; - struct nfs4_client_reclaim *crp; char dname[HEXDIR_LEN]; int status; struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); @@ -374,12 +412,9 @@ nfsd4_remove_clid_dir(struct nfs4_client *clp) nfs4_reset_creds(original_cred); if (status == 0) { vfs_fsync(nn->rec_file, 0); - if (nn->in_grace) { - /* remove reclaim record */ - crp = nfsd4_find_reclaim_client(dname, nn); - if (crp) - nfs4_remove_reclaim_record(crp, nn); - } + if (nn->in_grace) + __nfsd4_remove_reclaim_record_grace(dname, + HEXDIR_LEN, nn); } out_drop_write: mnt_drop_write_file(nn->rec_file); @@ -393,14 +428,31 @@ static int purge_old(struct dentry *parent, struct dentry *child, struct nfsd_net *nn) { int status; + struct xdr_netobj name; - if (nfs4_has_reclaimed_state(child->d_name.name, nn)) + if (child->d_name.len != HEXDIR_LEN - 1) { + printk("%s: illegal name %pd in recovery directory\n", + __func__, child); + /* Keep trying; maybe the others are OK: */ return 0; + } + name.data = kmemdup_nul(child->d_name.name, child->d_name.len, GFP_KERNEL); + if (!name.data) { + dprintk("%s: failed to allocate memory for name.data!\n", + __func__); + goto out; + } + name.len = HEXDIR_LEN; + if (nfs4_has_reclaimed_state(name, nn)) + goto out_free; status = vfs_rmdir(d_inode(parent), child); if (status) printk("failed to remove client recovery directory %pd\n", child); +out_free: + kfree(name.data); +out: /* Keep trying, success or failure: */ return 0; } @@ -430,13 +482,24 @@ out: static int load_recdir(struct dentry *parent, struct dentry *child, struct nfsd_net *nn) { + struct xdr_netobj name; + if (child->d_name.len != HEXDIR_LEN - 1) { - printk("nfsd4: illegal name %pd in recovery directory\n", - child); + printk("%s: illegal name %pd in recovery directory\n", + __func__, child); /* Keep trying; maybe the others are OK: */ return 0; } - nfs4_client_to_reclaim(child->d_name.name, nn); + name.data = kmemdup_nul(child->d_name.name, child->d_name.len, GFP_KERNEL); + if (!name.data) { + dprintk("%s: failed to allocate memory for name.data!\n", + __func__); + goto out; + } + name.len = HEXDIR_LEN; + if (!nfs4_client_to_reclaim(name, nn)) + kfree(name.data); +out: return 0; } @@ -616,6 +679,7 @@ nfsd4_check_legacy_client(struct nfs4_client *clp) char dname[HEXDIR_LEN]; struct nfs4_client_reclaim *crp; struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); + struct xdr_netobj name; /* did we already find that this client is stable? */ if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) @@ -628,13 +692,22 @@ nfsd4_check_legacy_client(struct nfs4_client *clp) } /* look for it in the reclaim hashtable otherwise */ - crp = nfsd4_find_reclaim_client(dname, nn); + name.data = kmemdup(dname, HEXDIR_LEN, GFP_KERNEL); + if (!name.data) { + dprintk("%s: failed to allocate memory for name.data!\n", + __func__); + goto out_enoent; + } + name.len = HEXDIR_LEN; + crp = nfsd4_find_reclaim_client(name, nn); + kfree(name.data); if (crp) { set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags); crp->cr_clp = clp; return 0; } +out_enoent: return -ENOENT; } -- cgit v1.2.3 From 74725959c33c14114fdce1e36e3504d106584d53 Mon Sep 17 00:00:00 2001 From: Scott Mayhew Date: Tue, 26 Mar 2019 18:06:27 -0400 Subject: nfsd: un-deprecate nfsdcld When nfsdcld was released, it was quickly deprecated in favor of the nfsdcltrack usermodehelper, so as to not require another running daemon. That prevents NFSv4 clients from reclaiming locks from nfsd's running in containers, since neither nfsdcltrack nor the legacy client tracking code work in containers. This commit un-deprecates the use of nfsdcld, with one twist: we will populate the reclaim_str_hashtbl on startup. During client tracking initialization, do an upcall ("GraceStart") to nfsdcld to get a list of clients from the database. nfsdcld will do one downcall with a status of -EINPROGRESS for each client record in the database, which in turn will cause an nfs4_client_reclaim to be added to the reclaim_str_hashtbl. When complete, nfsdcld will do a final downcall with a status of 0. This will save nfsd from having to do an upcall to the daemon during nfs4_check_open_reclaim() processing. Even though nfsdcld was quickly deprecated, there is a very small chance of old nfsdcld daemons running in the wild. These will respond to the new "GraceStart" upcall with -EOPNOTSUPP, in which case we will log a message and fall back to the original nfsdcld tracking ops (now called nfsd4_cld_tracking_ops_v0). Signed-off-by: Scott Mayhew Signed-off-by: J. Bruce Fields --- fs/nfsd/nfs4recover.c | 225 ++++++++++++++++++++++++++++++++++++++++-- include/uapi/linux/nfsd/cld.h | 1 + 2 files changed, 218 insertions(+), 8 deletions(-) (limited to 'fs/nfsd/nfs4recover.c') diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c index 2243b909b407..89c2a27956d0 100644 --- a/fs/nfsd/nfs4recover.c +++ b/fs/nfsd/nfs4recover.c @@ -779,6 +779,34 @@ cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg) return ret; } +static ssize_t +__cld_pipe_inprogress_downcall(const struct cld_msg __user *cmsg, + struct nfsd_net *nn) +{ + uint8_t cmd; + struct xdr_netobj name; + uint16_t namelen; + + if (get_user(cmd, &cmsg->cm_cmd)) { + dprintk("%s: error when copying cmd from userspace", __func__); + return -EFAULT; + } + if (cmd == Cld_GraceStart) { + if (get_user(namelen, &cmsg->cm_u.cm_name.cn_len)) + return -EFAULT; + name.data = memdup_user(&cmsg->cm_u.cm_name.cn_id, namelen); + if (IS_ERR_OR_NULL(name.data)) + return -EFAULT; + name.len = namelen; + if (!nfs4_client_to_reclaim(name, nn)) { + kfree(name.data); + return -EFAULT; + } + return sizeof(*cmsg); + } + return -EFAULT; +} + static ssize_t cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) { @@ -788,6 +816,7 @@ cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) struct nfsd_net *nn = net_generic(file_inode(filp)->i_sb->s_fs_info, nfsd_net_id); struct cld_net *cn = nn->cld_net; + int16_t status; if (mlen != sizeof(*cmsg)) { dprintk("%s: got %zu bytes, expected %zu\n", __func__, mlen, @@ -801,13 +830,24 @@ cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) return -EFAULT; } + /* + * copy the status so we know whether to remove the upcall from the + * list (for -EINPROGRESS, we just want to make sure the xid is + * valid, not remove the upcall from the list) + */ + if (get_user(status, &cmsg->cm_status)) { + dprintk("%s: error when copying status from userspace", __func__); + return -EFAULT; + } + /* walk the list and find corresponding xid */ cup = NULL; spin_lock(&cn->cn_lock); list_for_each_entry(tmp, &cn->cn_list, cu_list) { if (get_unaligned(&tmp->cu_msg.cm_xid) == xid) { cup = tmp; - list_del_init(&cup->cu_list); + if (status != -EINPROGRESS) + list_del_init(&cup->cu_list); break; } } @@ -819,6 +859,9 @@ cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen) return -EINVAL; } + if (status == -EINPROGRESS) + return __cld_pipe_inprogress_downcall(cmsg, nn); + if (copy_from_user(&cup->cu_msg, src, mlen) != 0) return -EFAULT; @@ -1065,9 +1108,14 @@ out_err: "record from stable storage: %d\n", ret); } -/* Check for presence of a record, and update its timestamp */ +/* + * For older nfsdcld's that do not allow us to "slurp" the clients + * from the tracking database during startup. + * + * Check for presence of a record, and update its timestamp + */ static int -nfsd4_cld_check(struct nfs4_client *clp) +nfsd4_cld_check_v0(struct nfs4_client *clp) { int ret; struct cld_upcall *cup; @@ -1100,8 +1148,61 @@ nfsd4_cld_check(struct nfs4_client *clp) return ret; } +/* + * For newer nfsdcld's that allow us to "slurp" the clients + * from the tracking database during startup. + * + * Check for presence of a record in the reclaim_str_hashtbl + */ +static int +nfsd4_cld_check(struct nfs4_client *clp) +{ + struct nfs4_client_reclaim *crp; + struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); + + /* did we already find that this client is stable? */ + if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) + return 0; + + /* look for it in the reclaim hashtable otherwise */ + crp = nfsd4_find_reclaim_client(clp->cl_name, nn); + if (crp) { + crp->cr_clp = clp; + return 0; + } + + return -ENOENT; +} + +static int +nfsd4_cld_grace_start(struct nfsd_net *nn) +{ + int ret; + struct cld_upcall *cup; + struct cld_net *cn = nn->cld_net; + + cup = alloc_cld_upcall(cn); + if (!cup) { + ret = -ENOMEM; + goto out_err; + } + + cup->cu_msg.cm_cmd = Cld_GraceStart; + ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); + if (!ret) + ret = cup->cu_msg.cm_status; + + free_cld_upcall(cup); +out_err: + if (ret) + dprintk("%s: Unable to get clients from userspace: %d\n", + __func__, ret); + return ret; +} + +/* For older nfsdcld's that need cm_gracetime */ static void -nfsd4_cld_grace_done(struct nfsd_net *nn) +nfsd4_cld_grace_done_v0(struct nfsd_net *nn) { int ret; struct cld_upcall *cup; @@ -1125,11 +1226,118 @@ out_err: printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret); } -static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = { +/* + * For newer nfsdcld's that do not need cm_gracetime. We also need to call + * nfs4_release_reclaim() to clear out the reclaim_str_hashtbl. + */ +static void +nfsd4_cld_grace_done(struct nfsd_net *nn) +{ + int ret; + struct cld_upcall *cup; + struct cld_net *cn = nn->cld_net; + + cup = alloc_cld_upcall(cn); + if (!cup) { + ret = -ENOMEM; + goto out_err; + } + + cup->cu_msg.cm_cmd = Cld_GraceDone; + ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg); + if (!ret) + ret = cup->cu_msg.cm_status; + + free_cld_upcall(cup); +out_err: + nfs4_release_reclaim(nn); + if (ret) + printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret); +} + +static int +nfs4_cld_state_init(struct net *net) +{ + struct nfsd_net *nn = net_generic(net, nfsd_net_id); + int i; + + nn->reclaim_str_hashtbl = kmalloc_array(CLIENT_HASH_SIZE, + sizeof(struct list_head), + GFP_KERNEL); + if (!nn->reclaim_str_hashtbl) + return -ENOMEM; + + for (i = 0; i < CLIENT_HASH_SIZE; i++) + INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]); + nn->reclaim_str_hashtbl_size = 0; + + return 0; +} + +static void +nfs4_cld_state_shutdown(struct net *net) +{ + struct nfsd_net *nn = net_generic(net, nfsd_net_id); + + kfree(nn->reclaim_str_hashtbl); +} + +static int +nfsd4_cld_tracking_init(struct net *net) +{ + int status; + struct nfsd_net *nn = net_generic(net, nfsd_net_id); + + status = nfs4_cld_state_init(net); + if (status) + return status; + + status = nfsd4_init_cld_pipe(net); + if (status) + goto err_shutdown; + + status = nfsd4_cld_grace_start(nn); + if (status) { + if (status == -EOPNOTSUPP) + printk(KERN_WARNING "NFSD: Please upgrade nfsdcld.\n"); + nfs4_release_reclaim(nn); + goto err_remove; + } + return 0; + +err_remove: + nfsd4_remove_cld_pipe(net); +err_shutdown: + nfs4_cld_state_shutdown(net); + return status; +} + +static void +nfsd4_cld_tracking_exit(struct net *net) +{ + struct nfsd_net *nn = net_generic(net, nfsd_net_id); + + nfs4_release_reclaim(nn); + nfsd4_remove_cld_pipe(net); + nfs4_cld_state_shutdown(net); +} + +/* For older nfsdcld's */ +static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops_v0 = { .init = nfsd4_init_cld_pipe, .exit = nfsd4_remove_cld_pipe, .create = nfsd4_cld_create, .remove = nfsd4_cld_remove, + .check = nfsd4_cld_check_v0, + .grace_done = nfsd4_cld_grace_done_v0, +}; + +/* For newer nfsdcld's */ +static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = { + .init = nfsd4_cld_tracking_init, + .exit = nfsd4_cld_tracking_exit, + .create = nfsd4_cld_create, + .remove = nfsd4_cld_remove, .check = nfsd4_cld_check, .grace_done = nfsd4_cld_grace_done, }; @@ -1514,9 +1722,10 @@ nfsd4_client_tracking_init(struct net *net) /* Finally, try to use nfsdcld */ nn->client_tracking_ops = &nfsd4_cld_tracking_ops; - printk(KERN_WARNING "NFSD: the nfsdcld client tracking upcall will be " - "removed in 3.10. Please transition to using " - "nfsdcltrack.\n"); + status = nn->client_tracking_ops->init(net); + if (!status) + return status; + nn->client_tracking_ops = &nfsd4_cld_tracking_ops_v0; do_init: status = nn->client_tracking_ops->init(net); if (status) { diff --git a/include/uapi/linux/nfsd/cld.h b/include/uapi/linux/nfsd/cld.h index f8f5cccad749..b1e9de4f07d5 100644 --- a/include/uapi/linux/nfsd/cld.h +++ b/include/uapi/linux/nfsd/cld.h @@ -36,6 +36,7 @@ enum cld_command { Cld_Remove, /* remove record of this cm_id */ Cld_Check, /* is this cm_id allowed? */ Cld_GraceDone, /* grace period is complete */ + Cld_GraceStart, }; /* representation of long-form NFSv4 client ID */ -- cgit v1.2.3 From 362063a595be959bc08f4163e6405a0266740091 Mon Sep 17 00:00:00 2001 From: Scott Mayhew Date: Tue, 26 Mar 2019 18:06:28 -0400 Subject: nfsd: keep a tally of RECLAIM_COMPLETE operations when using nfsdcld When using nfsdcld for NFSv4 client tracking, track the number of RECLAIM_COMPLETE operations we receive from "known" clients to help in deciding if we can lift the grace period early (or whether we need to start a v4 grace period at all). Signed-off-by: Scott Mayhew Signed-off-by: J. Bruce Fields --- fs/nfsd/netns.h | 3 +++ fs/nfsd/nfs4recover.c | 3 +++ fs/nfsd/nfs4state.c | 33 ++++++++++++++++++++++++++++++++- fs/nfsd/nfsctl.c | 1 + 4 files changed, 39 insertions(+), 1 deletion(-) (limited to 'fs/nfsd/nfs4recover.c') diff --git a/fs/nfsd/netns.h b/fs/nfsd/netns.h index 32cb8c027483..b42aaa22fba2 100644 --- a/fs/nfsd/netns.h +++ b/fs/nfsd/netns.h @@ -104,6 +104,9 @@ struct nfsd_net { time_t nfsd4_grace; bool somebody_reclaimed; + bool track_reclaim_completes; + atomic_t nr_reclaim_complete; + bool nfsd_net_up; bool lockd_up; diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c index 89c2a27956d0..09e3c9352778 100644 --- a/fs/nfsd/nfs4recover.c +++ b/fs/nfsd/nfs4recover.c @@ -1270,6 +1270,8 @@ nfs4_cld_state_init(struct net *net) for (i = 0; i < CLIENT_HASH_SIZE; i++) INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]); nn->reclaim_str_hashtbl_size = 0; + nn->track_reclaim_completes = true; + atomic_set(&nn->nr_reclaim_complete, 0); return 0; } @@ -1279,6 +1281,7 @@ nfs4_cld_state_shutdown(struct net *net) { struct nfsd_net *nn = net_generic(net, nfsd_net_id); + nn->track_reclaim_completes = false; kfree(nn->reclaim_str_hashtbl); } diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c index e601416aaf5f..8078314981f5 100644 --- a/fs/nfsd/nfs4state.c +++ b/fs/nfsd/nfs4state.c @@ -77,6 +77,7 @@ static u64 current_sessionid = 1; /* forward declarations */ static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner); static void nfs4_free_ol_stateid(struct nfs4_stid *stid); +void nfsd4_end_grace(struct nfsd_net *nn); /* Locking: */ @@ -1997,6 +1998,22 @@ destroy_client(struct nfs4_client *clp) __destroy_client(clp); } +static void inc_reclaim_complete(struct nfs4_client *clp) +{ + struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); + + if (!nn->track_reclaim_completes) + return; + if (!nfsd4_find_reclaim_client(clp->cl_name, nn)) + return; + if (atomic_inc_return(&nn->nr_reclaim_complete) == + nn->reclaim_str_hashtbl_size) { + printk(KERN_INFO "NFSD: all clients done reclaiming, ending NFSv4 grace period (net %x)\n", + clp->net->ns.inum); + nfsd4_end_grace(nn); + } +} + static void expire_client(struct nfs4_client *clp) { unhash_client(clp); @@ -3348,6 +3365,7 @@ nfsd4_reclaim_complete(struct svc_rqst *rqstp, status = nfs_ok; nfsd4_client_record_create(cstate->session->se_client); + inc_reclaim_complete(cstate->session->se_client); out: return status; } @@ -4707,7 +4725,6 @@ nfsd4_end_grace(struct nfsd_net *nn) if (nn->grace_ended) return; - dprintk("NFSD: end of grace period\n"); nn->grace_ended = true; /* * If the server goes down again right now, an NFSv4 @@ -4743,6 +4760,10 @@ static bool clients_still_reclaiming(struct nfsd_net *nn) unsigned long double_grace_period_end = nn->boot_time + 2 * nn->nfsd4_lease; + if (nn->track_reclaim_completes && + atomic_read(&nn->nr_reclaim_complete) == + nn->reclaim_str_hashtbl_size) + return false; if (!nn->somebody_reclaimed) return false; nn->somebody_reclaimed = false; @@ -4773,6 +4794,7 @@ nfs4_laundromat(struct nfsd_net *nn) new_timeo = 0; goto out; } + dprintk("NFSD: end of grace period\n"); nfsd4_end_grace(nn); INIT_LIST_HEAD(&reaplist); spin_lock(&nn->client_lock); @@ -7261,10 +7283,19 @@ nfs4_state_start_net(struct net *net) return ret; locks_start_grace(net, &nn->nfsd4_manager); nfsd4_client_tracking_init(net); + if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0) + goto skip_grace; printk(KERN_INFO "NFSD: starting %ld-second grace period (net %x)\n", nn->nfsd4_grace, net->ns.inum); queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ); return 0; + +skip_grace: + printk(KERN_INFO "NFSD: no clients to reclaim, skipping NFSv4 grace period (net %x)\n", + net->ns.inum); + queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_lease * HZ); + nfsd4_end_grace(nn); + return 0; } /* initialization to perform when the nfsd service is started: */ diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c index f2feb2d11bae..6667d3a4f839 100644 --- a/fs/nfsd/nfsctl.c +++ b/fs/nfsd/nfsctl.c @@ -1242,6 +1242,7 @@ static __net_init int nfsd_init_net(struct net *net) nn->nfsd4_lease = 90; /* default lease time */ nn->nfsd4_grace = 90; nn->somebody_reclaimed = false; + nn->track_reclaim_completes = false; nn->clverifier_counter = prandom_u32(); nn->clientid_counter = prandom_u32(); nn->s2s_cp_cl_id = nn->clientid_counter++; -- cgit v1.2.3 From 869216075b63c7a4030ee193c079cb5a69a918cb Mon Sep 17 00:00:00 2001 From: Scott Mayhew Date: Tue, 26 Mar 2019 18:06:29 -0400 Subject: nfsd: re-order client tracking method selection The new order is first nfsdcld, then the UMH upcall, and finally the legacy tracking method. Added some printk's to the tracking initialization functions so it's clear which tracking method was ultimately selected. Signed-off-by: Scott Mayhew Signed-off-by: J. Bruce Fields --- fs/nfsd/nfs4recover.c | 80 +++++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 65 insertions(+), 15 deletions(-) (limited to 'fs/nfsd/nfs4recover.c') diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c index 09e3c9352778..d60001da1dbf 100644 --- a/fs/nfsd/nfs4recover.c +++ b/fs/nfsd/nfs4recover.c @@ -628,6 +628,7 @@ nfsd4_legacy_tracking_init(struct net *net) status = nfsd4_load_reboot_recovery_data(net); if (status) goto err; + printk("NFSD: Using legacy client tracking operations.\n"); return 0; err: @@ -937,7 +938,7 @@ nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe) /* Initialize rpc_pipefs pipe for communication with client tracking daemon */ static int -nfsd4_init_cld_pipe(struct net *net) +__nfsd4_init_cld_pipe(struct net *net) { int ret; struct dentry *dentry; @@ -980,6 +981,17 @@ err: return ret; } +static int +nfsd4_init_cld_pipe(struct net *net) +{ + int status; + + status = __nfsd4_init_cld_pipe(net); + if (!status) + printk("NFSD: Using old nfsdcld client tracking operations.\n"); + return status; +} + static void nfsd4_remove_cld_pipe(struct net *net) { @@ -1285,27 +1297,55 @@ nfs4_cld_state_shutdown(struct net *net) kfree(nn->reclaim_str_hashtbl); } +static bool +cld_running(struct nfsd_net *nn) +{ + struct cld_net *cn = nn->cld_net; + struct rpc_pipe *pipe = cn->cn_pipe; + + return pipe->nreaders || pipe->nwriters; +} + static int nfsd4_cld_tracking_init(struct net *net) { int status; struct nfsd_net *nn = net_generic(net, nfsd_net_id); + bool running; + int retries = 10; status = nfs4_cld_state_init(net); if (status) return status; - status = nfsd4_init_cld_pipe(net); + status = __nfsd4_init_cld_pipe(net); if (status) goto err_shutdown; + /* + * rpc pipe upcalls take 30 seconds to time out, so we don't want to + * queue an upcall unless we know that nfsdcld is running (because we + * want this to fail fast so that nfsd4_client_tracking_init() can try + * the next client tracking method). nfsdcld should already be running + * before nfsd is started, so the wait here is for nfsdcld to open the + * pipefs file we just created. + */ + while (!(running = cld_running(nn)) && retries--) + msleep(100); + + if (!running) { + status = -ETIMEDOUT; + goto err_remove; + } + status = nfsd4_cld_grace_start(nn); if (status) { if (status == -EOPNOTSUPP) printk(KERN_WARNING "NFSD: Please upgrade nfsdcld.\n"); nfs4_release_reclaim(nn); goto err_remove; - } + } else + printk("NFSD: Using nfsdcld client tracking operations.\n"); return 0; err_remove: @@ -1552,6 +1592,8 @@ nfsd4_umh_cltrack_init(struct net *net) ret = nfsd4_umh_cltrack_upcall("init", NULL, grace_start, NULL); kfree(grace_start); + if (!ret) + printk("NFSD: Using UMH upcall client tracking operations.\n"); return ret; } @@ -1701,9 +1743,20 @@ nfsd4_client_tracking_init(struct net *net) if (nn->client_tracking_ops) goto do_init; + /* First, try to use nfsdcld */ + nn->client_tracking_ops = &nfsd4_cld_tracking_ops; + status = nn->client_tracking_ops->init(net); + if (!status) + return status; + if (status != -ETIMEDOUT) { + nn->client_tracking_ops = &nfsd4_cld_tracking_ops_v0; + status = nn->client_tracking_ops->init(net); + if (!status) + return status; + } + /* - * First, try a UMH upcall. It should succeed or fail quickly, so - * there's little harm in trying that first. + * Next, try the UMH upcall. */ nn->client_tracking_ops = &nfsd4_umh_tracking_ops; status = nn->client_tracking_ops->init(net); @@ -1711,26 +1764,23 @@ nfsd4_client_tracking_init(struct net *net) return status; /* - * See if the recoverydir exists and is a directory. If it is, - * then use the legacy ops. + * Finally, See if the recoverydir exists and is a directory. + * If it is, then use the legacy ops. */ nn->client_tracking_ops = &nfsd4_legacy_tracking_ops; status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path); if (!status) { status = d_is_dir(path.dentry); path_put(&path); - if (status) - goto do_init; + if (!status) { + status = -EINVAL; + goto out; + } } - /* Finally, try to use nfsdcld */ - nn->client_tracking_ops = &nfsd4_cld_tracking_ops; - status = nn->client_tracking_ops->init(net); - if (!status) - return status; - nn->client_tracking_ops = &nfsd4_cld_tracking_ops_v0; do_init: status = nn->client_tracking_ops->init(net); +out: if (status) { printk(KERN_WARNING "NFSD: Unable to initialize client " "recovery tracking! (%d)\n", status); -- cgit v1.2.3 From 8a9f4f41248a4dc00a9f0831a50f8b7a2fd86bfc Mon Sep 17 00:00:00 2001 From: Scott Mayhew Date: Tue, 26 Mar 2019 18:06:30 -0400 Subject: nfsd: handle legacy client tracking records sent by nfsdcld The new nfsdcld will do a one-time "upgrade" where it searches for records from nfsdcltrack and the legacy tracking during startup. Legacy records will be prefixed with the string "hash:", which we need to strip off before adding to the reclaim_str_hashtbl. When legacy records are encountered, set the new cn_has_legacy flag in the cld_net. When this flag is set, if the search for a reclaim record based on the client name string fails, then do a second search based on the hash of the name string. Note that if there are legacy records then the grace period will not be lifted early via the tracking of RECLAIM_COMPLETEs. Signed-off-by: Scott Mayhew Signed-off-by: J. Bruce Fields --- fs/nfsd/nfs4recover.c | 39 +++++++++++++++++++++++++++++++++++---- 1 file changed, 35 insertions(+), 4 deletions(-) (limited to 'fs/nfsd/nfs4recover.c') diff --git a/fs/nfsd/nfs4recover.c b/fs/nfsd/nfs4recover.c index d60001da1dbf..3a123c88120a 100644 --- a/fs/nfsd/nfs4recover.c +++ b/fs/nfsd/nfs4recover.c @@ -731,6 +731,7 @@ struct cld_net { spinlock_t cn_lock; struct list_head cn_list; unsigned int cn_xid; + bool cn_has_legacy; }; struct cld_upcall { @@ -787,6 +788,7 @@ __cld_pipe_inprogress_downcall(const struct cld_msg __user *cmsg, uint8_t cmd; struct xdr_netobj name; uint16_t namelen; + struct cld_net *cn = nn->cld_net; if (get_user(cmd, &cmsg->cm_cmd)) { dprintk("%s: error when copying cmd from userspace", __func__); @@ -799,6 +801,11 @@ __cld_pipe_inprogress_downcall(const struct cld_msg __user *cmsg, if (IS_ERR_OR_NULL(name.data)) return -EFAULT; name.len = namelen; + if (name.len > 5 && memcmp(name.data, "hash:", 5) == 0) { + name.len = name.len - 5; + memmove(name.data, name.data + 5, name.len); + cn->cn_has_legacy = true; + } if (!nfs4_client_to_reclaim(name, nn)) { kfree(name.data); return -EFAULT; @@ -969,6 +976,7 @@ __nfsd4_init_cld_pipe(struct net *net) } cn->cn_pipe->dentry = dentry; + cn->cn_has_legacy = false; nn->cld_net = cn; return 0; @@ -1171,6 +1179,10 @@ nfsd4_cld_check(struct nfs4_client *clp) { struct nfs4_client_reclaim *crp; struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); + struct cld_net *cn = nn->cld_net; + int status; + char dname[HEXDIR_LEN]; + struct xdr_netobj name; /* did we already find that this client is stable? */ if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) @@ -1178,12 +1190,31 @@ nfsd4_cld_check(struct nfs4_client *clp) /* look for it in the reclaim hashtable otherwise */ crp = nfsd4_find_reclaim_client(clp->cl_name, nn); - if (crp) { - crp->cr_clp = clp; - return 0; - } + if (crp) + goto found; + + if (cn->cn_has_legacy) { + status = nfs4_make_rec_clidname(dname, &clp->cl_name); + if (status) + return -ENOENT; + + name.data = kmemdup(dname, HEXDIR_LEN, GFP_KERNEL); + if (!name.data) { + dprintk("%s: failed to allocate memory for name.data!\n", + __func__); + return -ENOENT; + } + name.len = HEXDIR_LEN; + crp = nfsd4_find_reclaim_client(name, nn); + kfree(name.data); + if (crp) + goto found; + } return -ENOENT; +found: + crp->cr_clp = clp; + return 0; } static int -- cgit v1.2.3