From e0639dc5805a9d4faaa2c07ad98fa853b9529dd3 Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Fri, 20 Jul 2018 18:19:20 -0400 Subject: [PATCH] NFSD introduce async copy feature Upon receiving a request for async copy, create a new kthread. If we get asynchronous request, make sure to copy the needed arguments/state from the stack before starting the copy. Then start the thread and reply back to the client indicating copy is asynchronous. nfsd_copy_file_range() will copy in a loop over the total number of bytes is needed to copy. In case a failure happens in the middle, we ignore the error and return how much we copied so far. Once done creating a workitem for the callback workqueue and send CB_OFFLOAD with the results. The lifetime of the copy stateid is bound to the vfs copy. This way we don't need to keep the nfsd_net structure for the callback. We could keep it around longer so that an OFFLOAD_STATUS that came late would still get results, but clients should be able to deal without that. We handle OFFLOAD_CANCEL by sending a signal to the copy thread and calling kthread_stop. A client should cancel any ongoing copies before calling DESTROY_CLIENT; if not, we return a CLIENT_BUSY error. If the client is destroyed for some other reason (lease expiration, or server shutdown), we must clean up any ongoing copies ourselves. Signed-off-by: Olga Kornievskaia [colin.king@canonical.com: fix leak in error case] [bfields@fieldses.org: remove signalling, merge patches] Signed-off-by: J. Bruce Fields --- fs/nfsd/netns.h | 8 ++ fs/nfsd/nfs4proc.c | 261 ++++++++++++++++++++++++++++++++++++++++---- fs/nfsd/nfs4state.c | 38 ++++++- fs/nfsd/nfs4xdr.c | 23 +++- fs/nfsd/nfsctl.c | 1 + fs/nfsd/state.h | 9 ++ fs/nfsd/xdr4.h | 12 ++ 7 files changed, 327 insertions(+), 25 deletions(-) diff --git a/fs/nfsd/netns.h b/fs/nfsd/netns.h index 426f55005697..32cb8c027483 100644 --- a/fs/nfsd/netns.h +++ b/fs/nfsd/netns.h @@ -123,6 +123,14 @@ struct nfsd_net { wait_queue_head_t ntf_wq; atomic_t ntf_refcnt; + + /* + * clientid and stateid data for construction of net unique COPY + * stateids. + */ + u32 s2s_cp_cl_id; + struct idr s2s_cp_stateids; + spinlock_t s2s_cp_lock; }; /* Simple check to find out if a given net was properly initialized */ diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c index 0c7832321010..edff074d38c7 100644 --- a/fs/nfsd/nfs4proc.c +++ b/fs/nfsd/nfs4proc.c @@ -36,6 +36,7 @@ #include #include #include +#include #include "idmap.h" #include "cache.h" @@ -1089,37 +1090,236 @@ nfsd4_clone(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, return status; } +void nfs4_put_copy(struct nfsd4_copy *copy) +{ + if (!refcount_dec_and_test(©->refcount)) + return; + kfree(copy); +} + +static bool +check_and_set_stop_copy(struct nfsd4_copy *copy) +{ + bool value; + + spin_lock(©->cp_clp->async_lock); + value = copy->stopped; + if (!copy->stopped) + copy->stopped = true; + spin_unlock(©->cp_clp->async_lock); + return value; +} + +static void nfsd4_stop_copy(struct nfsd4_copy *copy) +{ + /* only 1 thread should stop the copy */ + if (!check_and_set_stop_copy(copy)) + kthread_stop(copy->copy_task); + nfs4_put_copy(copy); +} + +static struct nfsd4_copy *nfsd4_get_copy(struct nfs4_client *clp) +{ + struct nfsd4_copy *copy = NULL; + + spin_lock(&clp->async_lock); + if (!list_empty(&clp->async_copies)) { + copy = list_first_entry(&clp->async_copies, struct nfsd4_copy, + copies); + refcount_inc(©->refcount); + } + spin_unlock(&clp->async_lock); + return copy; +} + +void nfsd4_shutdown_copy(struct nfs4_client *clp) +{ + struct nfsd4_copy *copy; + + while ((copy = nfsd4_get_copy(clp)) != NULL) + nfsd4_stop_copy(copy); +} + +static void nfsd4_cb_offload_release(struct nfsd4_callback *cb) +{ + struct nfsd4_copy *copy = container_of(cb, struct nfsd4_copy, cp_cb); + + nfs4_put_copy(copy); +} + +static int nfsd4_cb_offload_done(struct nfsd4_callback *cb, + struct rpc_task *task) +{ + return 1; +} + +static const struct nfsd4_callback_ops nfsd4_cb_offload_ops = { + .release = nfsd4_cb_offload_release, + .done = nfsd4_cb_offload_done +}; + +static void nfsd4_init_copy_res(struct nfsd4_copy *copy, bool sync) +{ + copy->cp_res.wr_stable_how = NFS_UNSTABLE; + copy->cp_synchronous = sync; + gen_boot_verifier(©->cp_res.wr_verifier, copy->cp_clp->net); +} + +static ssize_t _nfsd_copy_file_range(struct nfsd4_copy *copy) +{ + ssize_t bytes_copied = 0; + size_t bytes_total = copy->cp_count; + u64 src_pos = copy->cp_src_pos; + u64 dst_pos = copy->cp_dst_pos; + + do { + if (kthread_should_stop()) + break; + bytes_copied = nfsd_copy_file_range(copy->file_src, src_pos, + copy->file_dst, dst_pos, bytes_total); + if (bytes_copied <= 0) + break; + bytes_total -= bytes_copied; + copy->cp_res.wr_bytes_written += bytes_copied; + src_pos += bytes_copied; + dst_pos += bytes_copied; + } while (bytes_total > 0 && !copy->cp_synchronous); + return bytes_copied; +} + +static __be32 nfsd4_do_copy(struct nfsd4_copy *copy, bool sync) +{ + __be32 status; + ssize_t bytes; + + bytes = _nfsd_copy_file_range(copy); + /* for async copy, we ignore the error, client can always retry + * to get the error + */ + if (bytes < 0 && !copy->cp_res.wr_bytes_written) + status = nfserrno(bytes); + else { + nfsd4_init_copy_res(copy, sync); + status = nfs_ok; + } + + fput(copy->file_src); + fput(copy->file_dst); + return status; +} + +static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst) +{ + dst->cp_src_pos = src->cp_src_pos; + dst->cp_dst_pos = src->cp_dst_pos; + dst->cp_count = src->cp_count; + dst->cp_synchronous = src->cp_synchronous; + memcpy(&dst->cp_res, &src->cp_res, sizeof(src->cp_res)); + memcpy(&dst->fh, &src->fh, sizeof(src->fh)); + dst->cp_clp = src->cp_clp; + dst->file_dst = get_file(src->file_dst); + dst->file_src = get_file(src->file_src); + memcpy(&dst->cp_stateid, &src->cp_stateid, sizeof(src->cp_stateid)); +} + +static void cleanup_async_copy(struct nfsd4_copy *copy) +{ + nfs4_free_cp_state(copy); + fput(copy->file_dst); + fput(copy->file_src); + spin_lock(©->cp_clp->async_lock); + list_del(©->copies); + spin_unlock(©->cp_clp->async_lock); + nfs4_put_copy(copy); +} + +static int nfsd4_do_async_copy(void *data) +{ + struct nfsd4_copy *copy = (struct nfsd4_copy *)data; + struct nfsd4_copy *cb_copy; + + copy->nfserr = nfsd4_do_copy(copy, 0); + cb_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL); + if (!cb_copy) + goto out; + memcpy(&cb_copy->cp_res, ©->cp_res, sizeof(copy->cp_res)); + cb_copy->cp_clp = copy->cp_clp; + cb_copy->nfserr = copy->nfserr; + memcpy(&cb_copy->fh, ©->fh, sizeof(copy->fh)); + nfsd4_init_cb(&cb_copy->cp_cb, cb_copy->cp_clp, + &nfsd4_cb_offload_ops, NFSPROC4_CLNT_CB_OFFLOAD); + nfsd4_run_cb(&cb_copy->cp_cb); +out: + cleanup_async_copy(copy); + return 0; +} + static __be32 nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, union nfsd4_op_u *u) { struct nfsd4_copy *copy = &u->copy; - struct file *src, *dst; __be32 status; - ssize_t bytes; + struct nfsd4_copy *async_copy = NULL; - status = nfsd4_verify_copy(rqstp, cstate, ©->cp_src_stateid, &src, - ©->cp_dst_stateid, &dst); + status = nfsd4_verify_copy(rqstp, cstate, ©->cp_src_stateid, + ©->file_src, ©->cp_dst_stateid, + ©->file_dst); if (status) goto out; - bytes = nfsd_copy_file_range(src, copy->cp_src_pos, - dst, copy->cp_dst_pos, copy->cp_count); + copy->cp_clp = cstate->clp; + memcpy(©->fh, &cstate->current_fh.fh_handle, + sizeof(struct knfsd_fh)); + if (!copy->cp_synchronous) { + struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id); - if (bytes < 0) - status = nfserrno(bytes); - else { - copy->cp_res.wr_bytes_written = bytes; - copy->cp_res.wr_stable_how = NFS_UNSTABLE; - copy->cp_synchronous = 1; - gen_boot_verifier(©->cp_res.wr_verifier, SVC_NET(rqstp)); + status = nfserrno(-ENOMEM); + async_copy = kzalloc(sizeof(struct nfsd4_copy), GFP_KERNEL); + if (!async_copy) + goto out; + if (!nfs4_init_cp_state(nn, copy)) { + kfree(async_copy); + goto out; + } + refcount_set(&async_copy->refcount, 1); + memcpy(©->cp_res.cb_stateid, ©->cp_stateid, + sizeof(copy->cp_stateid)); + dup_copy_fields(copy, async_copy); + async_copy->copy_task = kthread_create(nfsd4_do_async_copy, + async_copy, "%s", "copy thread"); + if (IS_ERR(async_copy->copy_task)) + goto out_err; + spin_lock(&async_copy->cp_clp->async_lock); + list_add(&async_copy->copies, + &async_copy->cp_clp->async_copies); + spin_unlock(&async_copy->cp_clp->async_lock); + wake_up_process(async_copy->copy_task); status = nfs_ok; - } - - fput(src); - fput(dst); + } else + status = nfsd4_do_copy(copy, 1); out: return status; +out_err: + cleanup_async_copy(async_copy); + goto out; +} + +struct nfsd4_copy * +find_async_copy(struct nfs4_client *clp, stateid_t *stateid) +{ + struct nfsd4_copy *copy; + + spin_lock(&clp->async_lock); + list_for_each_entry(copy, &clp->async_copies, copies) { + if (memcmp(©->cp_stateid, stateid, NFS4_STATEID_SIZE)) + continue; + refcount_inc(©->refcount); + spin_unlock(&clp->async_lock); + return copy; + } + spin_unlock(&clp->async_lock); + return NULL; } static __be32 @@ -1127,7 +1327,18 @@ nfsd4_offload_cancel(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, union nfsd4_op_u *u) { - return 0; + struct nfsd4_offload_status *os = &u->offload_status; + __be32 status = 0; + struct nfsd4_copy *copy; + struct nfs4_client *clp = cstate->clp; + + copy = find_async_copy(clp, &os->stateid); + if (copy) + nfsd4_stop_copy(copy); + else + status = nfserr_bad_stateid; + + return status; } static __be32 @@ -1157,7 +1368,19 @@ nfsd4_offload_status(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, union nfsd4_op_u *u) { - return nfserr_notsupp; + struct nfsd4_offload_status *os = &u->offload_status; + __be32 status = 0; + struct nfsd4_copy *copy; + struct nfs4_client *clp = cstate->clp; + + copy = find_async_copy(clp, &os->stateid); + if (copy) { + os->count = copy->cp_res.wr_bytes_written; + nfs4_put_copy(copy); + } else + status = nfserr_bad_stateid; + + return status; } static __be32 diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c index b0ca0efd2875..07a57d024f95 100644 --- a/fs/nfsd/nfs4state.c +++ b/fs/nfsd/nfs4state.c @@ -713,6 +713,36 @@ struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *sla return NULL; } +/* + * Create a unique stateid_t to represent each COPY. + */ +int nfs4_init_cp_state(struct nfsd_net *nn, struct nfsd4_copy *copy) +{ + int new_id; + + idr_preload(GFP_KERNEL); + spin_lock(&nn->s2s_cp_lock); + new_id = idr_alloc_cyclic(&nn->s2s_cp_stateids, copy, 0, 0, GFP_NOWAIT); + spin_unlock(&nn->s2s_cp_lock); + idr_preload_end(); + if (new_id < 0) + return 0; + copy->cp_stateid.si_opaque.so_id = new_id; + copy->cp_stateid.si_opaque.so_clid.cl_boot = nn->boot_time; + copy->cp_stateid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id; + return 1; +} + +void nfs4_free_cp_state(struct nfsd4_copy *copy) +{ + struct nfsd_net *nn; + + nn = net_generic(copy->cp_clp->net, nfsd_net_id); + spin_lock(&nn->s2s_cp_lock); + idr_remove(&nn->s2s_cp_stateids, copy->cp_stateid.si_opaque.so_id); + spin_unlock(&nn->s2s_cp_lock); +} + static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp) { struct nfs4_stid *stid; @@ -1827,6 +1857,8 @@ static struct nfs4_client *alloc_client(struct xdr_netobj name) #ifdef CONFIG_NFSD_PNFS INIT_LIST_HEAD(&clp->cl_lo_states); #endif + INIT_LIST_HEAD(&clp->async_copies); + spin_lock_init(&clp->async_lock); spin_lock_init(&clp->cl_lock); rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table"); return clp; @@ -1942,6 +1974,7 @@ __destroy_client(struct nfs4_client *clp) } } nfsd4_return_all_client_layouts(clp); + nfsd4_shutdown_copy(clp); nfsd4_shutdown_callback(clp); if (clp->cl_cb_conn.cb_xprt) svc_xprt_put(clp->cl_cb_conn.cb_xprt); @@ -2475,7 +2508,8 @@ static bool client_has_state(struct nfs4_client *clp) || !list_empty(&clp->cl_lo_states) #endif || !list_empty(&clp->cl_delegations) - || !list_empty(&clp->cl_sessions); + || !list_empty(&clp->cl_sessions) + || !list_empty(&clp->async_copies); } __be32 @@ -7161,6 +7195,8 @@ static int nfs4_state_create_net(struct net *net) INIT_LIST_HEAD(&nn->close_lru); INIT_LIST_HEAD(&nn->del_recall_lru); spin_lock_init(&nn->client_lock); + spin_lock_init(&nn->s2s_cp_lock); + idr_init(&nn->s2s_cp_stateids); spin_lock_init(&nn->blocked_locks_lock); INIT_LIST_HEAD(&nn->blocked_locks_lru); diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c index b78280a8af73..3de42a729093 100644 --- a/fs/nfsd/nfs4xdr.c +++ b/fs/nfsd/nfs4xdr.c @@ -4231,15 +4231,27 @@ nfsd4_encode_layoutreturn(struct nfsd4_compoundres *resp, __be32 nfserr, #endif /* CONFIG_NFSD_PNFS */ static __be32 -nfsd42_encode_write_res(struct nfsd4_compoundres *resp, struct nfsd42_write_res *write) +nfsd42_encode_write_res(struct nfsd4_compoundres *resp, + struct nfsd42_write_res *write, bool sync) { __be32 *p; - - p = xdr_reserve_space(&resp->xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE); + p = xdr_reserve_space(&resp->xdr, 4); + if (!p) + return nfserr_resource; + + if (sync) + *p++ = cpu_to_be32(0); + else { + __be32 nfserr; + *p++ = cpu_to_be32(1); + nfserr = nfsd4_encode_stateid(&resp->xdr, &write->cb_stateid); + if (nfserr) + return nfserr; + } + p = xdr_reserve_space(&resp->xdr, 8 + 4 + NFS4_VERIFIER_SIZE); if (!p) return nfserr_resource; - *p++ = cpu_to_be32(0); p = xdr_encode_hyper(p, write->wr_bytes_written); *p++ = cpu_to_be32(write->wr_stable_how); p = xdr_encode_opaque_fixed(p, write->wr_verifier.data, @@ -4253,7 +4265,8 @@ nfsd4_encode_copy(struct nfsd4_compoundres *resp, __be32 nfserr, { __be32 *p; - nfserr = nfsd42_encode_write_res(resp, ©->cp_res); + nfserr = nfsd42_encode_write_res(resp, ©->cp_res, + copy->cp_synchronous); if (nfserr) return nfserr; diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c index 7fb9f7c667b1..6384c9b94898 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->somebody_reclaimed = false; nn->clverifier_counter = prandom_u32(); nn->clientid_counter = prandom_u32(); + nn->s2s_cp_cl_id = nn->clientid_counter++; atomic_set(&nn->ntf_refcnt, 0); init_waitqueue_head(&nn->ntf_wq); diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h index 6e38d9927448..6aacb325b6a0 100644 --- a/fs/nfsd/state.h +++ b/fs/nfsd/state.h @@ -355,6 +355,8 @@ struct nfs4_client { struct rpc_wait_queue cl_cb_waitq; /* backchannel callers may */ /* wait here for slots */ struct net *net; + struct list_head async_copies; /* list of async copies */ + spinlock_t async_lock; /* lock for async copies */ }; /* struct nfs4_client_reset @@ -600,6 +602,7 @@ struct nfsd4_blocked_lock { struct nfsd4_compound_state; struct nfsd_net; +struct nfsd4_copy; extern __be32 nfs4_preprocess_stateid_op(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct svc_fh *fhp, @@ -609,6 +612,8 @@ __be32 nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate, struct nfs4_stid **s, struct nfsd_net *nn); struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab, void (*sc_free)(struct nfs4_stid *)); +int nfs4_init_cp_state(struct nfsd_net *nn, struct nfsd4_copy *copy); +void nfs4_free_cp_state(struct nfsd4_copy *copy); void nfs4_unhash_stid(struct nfs4_stid *s); void nfs4_put_stid(struct nfs4_stid *s); void nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid); @@ -627,6 +632,7 @@ extern void nfsd4_run_cb(struct nfsd4_callback *cb); extern int nfsd4_create_callback_queue(void); extern void nfsd4_destroy_callback_queue(void); extern void nfsd4_shutdown_callback(struct nfs4_client *); +extern void nfsd4_shutdown_copy(struct nfs4_client *clp); extern void nfsd4_prepare_cb_recall(struct nfs4_delegation *dp); extern struct nfs4_client_reclaim *nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn); @@ -634,6 +640,9 @@ extern bool nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn); struct nfs4_file *find_file(struct knfsd_fh *fh); void put_nfs4_file(struct nfs4_file *fi); +extern void nfs4_put_copy(struct nfsd4_copy *copy); +extern struct nfsd4_copy * +find_async_copy(struct nfs4_client *clp, stateid_t *staetid); static inline void get_nfs4_file(struct nfs4_file *fi) { refcount_inc(&fi->fi_ref); diff --git a/fs/nfsd/xdr4.h b/fs/nfsd/xdr4.h index 06cf218944c5..feeb6d4bdffd 100644 --- a/fs/nfsd/xdr4.h +++ b/fs/nfsd/xdr4.h @@ -532,6 +532,18 @@ struct nfsd4_copy { struct nfsd4_callback cp_cb; __be32 nfserr; struct knfsd_fh fh; + + struct nfs4_client *cp_clp; + + struct file *file_src; + struct file *file_dst; + + stateid_t cp_stateid; + + struct list_head copies; + struct task_struct *copy_task; + refcount_t refcount; + bool stopped; }; struct nfsd4_seek {