From e0b7d420f72a66b5299da025be8e8a17e019a557 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2018 10:28:40 -0400 Subject: pNFS: Don't discard layout segments that are marked for return If there are layout segments that are marked for return, then we need to ensure that pnfs_mark_matching_lsegs_return() does not just silently discard them, but it should tell the caller that there is a layoutreturn scheduled. Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 64c214fb9da6..af2322256aa4 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -283,19 +283,22 @@ static u32 initiate_file_draining(struct nfs_client *clp, goto unlock; } - if (pnfs_mark_matching_lsegs_return(lo, &free_me_list, + switch (pnfs_mark_matching_lsegs_return(lo, &free_me_list, &args->cbl_range, be32_to_cpu(args->cbl_stateid.seqid))) { + case 0: + case -EBUSY: + /* There are layout segments that need to be returned */ rv = NFS4_OK; - goto unlock; - } - - /* Embrace your forgetfulness! */ - rv = NFS4ERR_NOMATCHING_LAYOUT; + break; + case -ENOENT: + /* Embrace your forgetfulness! */ + rv = NFS4ERR_NOMATCHING_LAYOUT; - if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) { - NFS_SERVER(ino)->pnfs_curr_ld->return_range(lo, - &args->cbl_range); + if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) { + NFS_SERVER(ino)->pnfs_curr_ld->return_range(lo, + &args->cbl_range); + } } unlock: spin_unlock(&ino->i_lock); -- cgit v1.2.3 From 00bcbe119f915dec256f211f9dbfc93cb64773bc Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Sat, 23 Jun 2018 13:35:28 -0400 Subject: pNFS: Don't update the stateid when replying NFS4ERR_DELAY to a layout recall RFC5661 doesn't state directly that the client should update the layout stateid if it returns NFS4ERR_NOMATCHING_LAYOUT in response to a recall, however it does state that this error will "cleanly indicate completion" on par with returning the layout. For this reason, we assume that the client should update the layout stateid. The Linux pNFS server definitely does expect this behaviour. However, if the client replies NFS4ERR_DELAY, then it is stating that the recall was not processed, so it would be very wrong to update the layout stateid. Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index af2322256aa4..efca3d6c89f2 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -273,7 +273,6 @@ static u32 initiate_file_draining(struct nfs_client *clp, rv = pnfs_check_callback_stateid(lo, &args->cbl_stateid); if (rv != NFS_OK) goto unlock; - pnfs_set_layout_stateid(lo, &args->cbl_stateid, true); /* * Enforce RFC5661 Section 12.5.5.2.1.5 (Bulk Recall and Return) @@ -283,6 +282,7 @@ static u32 initiate_file_draining(struct nfs_client *clp, goto unlock; } + pnfs_set_layout_stateid(lo, &args->cbl_stateid, true); switch (pnfs_mark_matching_lsegs_return(lo, &free_me_list, &args->cbl_range, be32_to_cpu(args->cbl_stateid.seqid))) { -- cgit v1.2.3 From af9b6d7570ca9afbbc6076ab7920d8f00f7e55c1 Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Fri, 29 Jun 2018 12:45:53 -0400 Subject: pNFS: Parse the results of layoutget on open even if permissions checks fail Even if the results of the permissions checks failed, we should parse the results of the layout on open call so that we can return the layout if required. Note that we also want to ignore the sequence counter for whether or not a layout recall occurred. If the recall pertained to our OPEN, then the callback will know, and will attempt to wait for us to finih processing anyway. Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 2 -- fs/nfs/nfs4proc.c | 5 +++-- fs/nfs/pnfs.c | 4 ---- include/linux/nfs_fs_sb.h | 1 - include/linux/nfs_xdr.h | 1 - 5 files changed, 3 insertions(+), 10 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index efca3d6c89f2..43ba390bb653 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -331,8 +331,6 @@ static u32 initiate_bulk_draining(struct nfs_client *clp, static u32 do_callback_layoutrecall(struct nfs_client *clp, struct cb_layoutrecallargs *args) { - write_seqcount_begin(&clp->cl_callback_count); - write_seqcount_end(&clp->cl_callback_count); if (args->cbl_recall_type == RETURN_FILE) return initiate_file_draining(clp, args); return initiate_bulk_draining(clp, args); diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index 6dd146885da9..5a8190ec31a2 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -2843,6 +2843,9 @@ static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata, nfs_save_change_attribute(d_inode(opendata->dir))); } + /* Parse layoutget results before we check for access */ + pnfs_parse_lgopen(state->inode, opendata->lgp, ctx); + ret = nfs4_opendata_access(sp->so_cred, opendata, state, fmode, flags); if (ret != 0) goto out; @@ -2851,8 +2854,6 @@ static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata, nfs_inode_attach_open_context(ctx); if (read_seqcount_retry(&sp->so_reclaim_seqcount, seq)) nfs4_schedule_stateid_recovery(server, state); - else - pnfs_parse_lgopen(state->inode, opendata->lgp, ctx); } out: diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c index 030c39c107c2..7fdac8b504dd 100644 --- a/fs/nfs/pnfs.c +++ b/fs/nfs/pnfs.c @@ -1018,7 +1018,6 @@ pnfs_alloc_init_layoutget_args(struct inode *ino, nfs4_stateid_copy(&lgp->args.stateid, stateid); lgp->gfp_flags = gfp_flags; lgp->cred = get_rpccred(ctx->cred); - lgp->callback_count = raw_seqcount_begin(&server->nfs_client->cl_callback_count); return lgp; } @@ -2181,9 +2180,6 @@ void pnfs_parse_lgopen(struct inode *ino, struct nfs4_layoutget *lgp, } else lo = NFS_I(lgp->args.inode)->layout; - if (read_seqcount_retry(&srv->nfs_client->cl_callback_count, - lgp->callback_count)) - return; lseg = pnfs_layout_process(lgp); if (!IS_ERR(lseg)) { iomode = lgp->args.range.iomode; diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index 74ae3e1d19a0..2c18d618604e 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -28,7 +28,6 @@ struct nfs41_impl_id; struct nfs_client { refcount_t cl_count; atomic_t cl_mds_count; - seqcount_t cl_callback_count; int cl_cons_state; /* current construction state (-ve: init error) */ #define NFS_CS_READY 0 /* ready to be used */ #define NFS_CS_INITING 1 /* busy initialising */ diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 712eed156d09..3b7325cfb291 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -271,7 +271,6 @@ struct nfs4_layoutget { struct nfs4_layoutget_args args; struct nfs4_layoutget_res res; struct rpc_cred *cred; - unsigned callback_count; gfp_t gfp_flags; }; -- cgit v1.2.3 From bd3d16a887b0c19a2a20d35ffed499e3a3637feb Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Thu, 12 Jul 2018 14:19:03 -0400 Subject: NFSv4.1: Fix a potential layoutget/layoutrecall deadlock If the client is sending a layoutget, but the server issues a callback to recall what it thinks may be an outstanding layout, then we may find an uninitialised layout attached to the inode due to the layoutget. In that case, it is appropriate to return NFS4ERR_NOMATCHING_LAYOUT rather than NFS4ERR_DELAY, as the latter can end up deadlocking. Signed-off-by: Trond Myklebust --- fs/nfs/callback_proc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index 43ba390bb653..be8e0d2196a5 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -215,9 +215,9 @@ static u32 pnfs_check_callback_stateid(struct pnfs_layout_hdr *lo, { u32 oldseq, newseq; - /* Is the stateid still not initialised? */ + /* Is the stateid not initialised? */ if (!pnfs_layout_is_valid(lo)) - return NFS4ERR_DELAY; + return NFS4ERR_NOMATCHING_LAYOUT; /* Mismatched stateid? */ if (!nfs4_stateid_match_other(&lo->plh_stateid, new)) -- cgit v1.2.3 From 5178a125f6d5fb0720315ea4f7cca642fb936031 Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Mon, 9 Jul 2018 15:13:28 -0400 Subject: NFS CB_OFFLOAD xdr Signed-off-by: Olga Kornievskaia Signed-off-by: Anna Schumaker --- fs/nfs/callback.h | 12 ++++++++ fs/nfs/callback_proc.c | 7 +++++ fs/nfs/callback_xdr.c | 80 +++++++++++++++++++++++++++++++++++++++++++++++++- 3 files changed, 98 insertions(+), 1 deletion(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback.h b/fs/nfs/callback.h index a20a0bce40a4..8f34daf85f70 100644 --- a/fs/nfs/callback.h +++ b/fs/nfs/callback.h @@ -184,6 +184,18 @@ struct cb_notify_lock_args { extern __be32 nfs4_callback_notify_lock(void *argp, void *resp, struct cb_process_state *cps); #endif /* CONFIG_NFS_V4_1 */ +#ifdef CONFIG_NFS_V4_2 +struct cb_offloadargs { + struct nfs_fh coa_fh; + nfs4_stateid coa_stateid; + uint32_t error; + uint64_t wr_count; + struct nfs_writeverf wr_writeverf; +}; + +extern __be32 nfs4_callback_offload(void *args, void *dummy, + struct cb_process_state *cps); +#endif /* CONFIG_NFS_V4_2 */ extern int check_gss_callback_principal(struct nfs_client *, struct svc_rqst *); extern __be32 nfs4_callback_getattr(void *argp, void *resp, struct cb_process_state *cps); diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index be8e0d2196a5..d6f45bd176a9 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -661,3 +661,10 @@ __be32 nfs4_callback_notify_lock(void *argp, void *resp, return htonl(NFS4_OK); } #endif /* CONFIG_NFS_V4_1 */ +#ifdef CONFIG_NFS_V4_2 +__be32 nfs4_callback_offload(void *args, void *dummy, + struct cb_process_state *cps) +{ + return 0; +} +#endif /* CONFIG_NFS_V4_2 */ diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c index cb905c0e606c..a87a56273407 100644 --- a/fs/nfs/callback_xdr.c +++ b/fs/nfs/callback_xdr.c @@ -38,6 +38,9 @@ #define CB_OP_RECALLSLOT_RES_MAXSZ (CB_OP_HDR_RES_MAXSZ) #define CB_OP_NOTIFY_LOCK_RES_MAXSZ (CB_OP_HDR_RES_MAXSZ) #endif /* CONFIG_NFS_V4_1 */ +#ifdef CONFIG_NFS_V4_2 +#define CB_OP_OFFLOAD_RES_MAXSZ (CB_OP_HDR_RES_MAXSZ) +#endif /* CONFIG_NFS_V4_2 */ #define NFSDBG_FACILITY NFSDBG_CALLBACK @@ -527,7 +530,72 @@ static __be32 decode_notify_lock_args(struct svc_rqst *rqstp, } #endif /* CONFIG_NFS_V4_1 */ +#ifdef CONFIG_NFS_V4_2 +static __be32 decode_write_response(struct xdr_stream *xdr, + struct cb_offloadargs *args) +{ + __be32 *p; + + /* skip the always zero field */ + p = read_buf(xdr, 4); + if (unlikely(!p)) + goto out; + p++; + + /* decode count, stable_how, verifier */ + p = xdr_inline_decode(xdr, 8 + 4); + if (unlikely(!p)) + goto out; + p = xdr_decode_hyper(p, &args->wr_count); + args->wr_writeverf.committed = be32_to_cpup(p); + p = xdr_inline_decode(xdr, NFS4_VERIFIER_SIZE); + if (likely(p)) { + memcpy(&args->wr_writeverf.verifier.data[0], p, + NFS4_VERIFIER_SIZE); + return 0; + } +out: + return htonl(NFS4ERR_RESOURCE); +} + +static __be32 decode_offload_args(struct svc_rqst *rqstp, + struct xdr_stream *xdr, + void *data) +{ + struct cb_offloadargs *args = data; + __be32 *p; + __be32 status; + + /* decode fh */ + status = decode_fh(xdr, &args->coa_fh); + if (unlikely(status != 0)) + return status; + + /* decode stateid */ + status = decode_stateid(xdr, &args->coa_stateid); + if (unlikely(status != 0)) + return status; + /* decode status */ + p = read_buf(xdr, 4); + if (unlikely(!p)) + goto out; + args->error = ntohl(*p++); + if (!args->error) { + status = decode_write_response(xdr, args); + if (unlikely(status != 0)) + return status; + } else { + p = xdr_inline_decode(xdr, 8); + if (unlikely(!p)) + goto out; + p = xdr_decode_hyper(p, &args->wr_count); + } + return 0; +out: + return htonl(NFS4ERR_RESOURCE); +} +#endif /* CONFIG_NFS_V4_2 */ static __be32 encode_string(struct xdr_stream *xdr, unsigned int len, const char *str) { if (unlikely(xdr_stream_encode_opaque(xdr, str, len) < 0)) @@ -773,7 +841,10 @@ preprocess_nfs42_op(int nop, unsigned int op_nr, struct callback_op **op) if (status != htonl(NFS4ERR_OP_ILLEGAL)) return status; - if (op_nr == OP_CB_OFFLOAD) + if (op_nr == OP_CB_OFFLOAD) { + *op = &callback_ops[op_nr]; + return htonl(NFS_OK); + } else return htonl(NFS4ERR_NOTSUPP); return htonl(NFS4ERR_OP_ILLEGAL); } @@ -974,6 +1045,13 @@ static struct callback_op callback_ops[] = { .res_maxsize = CB_OP_NOTIFY_LOCK_RES_MAXSZ, }, #endif /* CONFIG_NFS_V4_1 */ +#ifdef CONFIG_NFS_V4_2 + [OP_CB_OFFLOAD] = { + .process_op = nfs4_callback_offload, + .decode_args = decode_offload_args, + .res_maxsize = CB_OP_OFFLOAD_RES_MAXSZ, + }, +#endif /* CONFIG_NFS_V4_2 */ }; /* -- cgit v1.2.3 From 62164f317972fcd36590578888f33a1994dda519 Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Mon, 9 Jul 2018 15:13:31 -0400 Subject: NFS add support for asynchronous COPY Change xdr to always send COPY asynchronously. Keep the list copies send in a list under a server structure. Once copy is sent, it waits on a completion structure that will be signalled by the callback thread that receives CB_OFFLOAD. If CB_OFFLOAD returned an error and even if it returned partial bytes, ignore them (as we can't commit without a verifier to match) and return an error. Signed-off-by: Olga Kornievskaia Signed-off-by: Anna Schumaker --- fs/nfs/callback_proc.c | 38 +++++++++++++++++++++++++++++++- fs/nfs/client.c | 1 + fs/nfs/nfs42proc.c | 55 ++++++++++++++++++++++++++++++++++++++++++----- fs/nfs/nfs42xdr.c | 8 ++++--- include/linux/nfs_fs.h | 9 ++++++++ include/linux/nfs_fs_sb.h | 1 + include/linux/nfs_xdr.h | 1 + 7 files changed, 104 insertions(+), 9 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index d6f45bd176a9..acdda259912e 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -662,9 +662,45 @@ __be32 nfs4_callback_notify_lock(void *argp, void *resp, } #endif /* CONFIG_NFS_V4_1 */ #ifdef CONFIG_NFS_V4_2 -__be32 nfs4_callback_offload(void *args, void *dummy, +static void nfs4_copy_cb_args(struct nfs4_copy_state *cp_state, + struct cb_offloadargs *args) +{ + cp_state->count = args->wr_count; + cp_state->error = args->error; + if (!args->error) { + cp_state->verf.committed = args->wr_writeverf.committed; + memcpy(&cp_state->verf.verifier.data[0], + &args->wr_writeverf.verifier.data[0], + NFS4_VERIFIER_SIZE); + } +} + +__be32 nfs4_callback_offload(void *data, void *dummy, struct cb_process_state *cps) { + struct cb_offloadargs *args = data; + struct nfs_server *server; + struct nfs4_copy_state *copy; + + rcu_read_lock(); + list_for_each_entry_rcu(server, &cps->clp->cl_superblocks, + client_link) { + spin_lock(&server->nfs_client->cl_lock); + list_for_each_entry(copy, &server->ss_copies, copies) { + if (memcmp(args->coa_stateid.other, + copy->stateid.other, + sizeof(args->coa_stateid.other))) + continue; + nfs4_copy_cb_args(copy, args); + complete(©->completion); + spin_unlock(&server->nfs_client->cl_lock); + goto out; + } + spin_unlock(&server->nfs_client->cl_lock); + } +out: + rcu_read_unlock(); + return 0; } #endif /* CONFIG_NFS_V4_2 */ diff --git a/fs/nfs/client.c b/fs/nfs/client.c index 377a61654a88..96d5f8135eb9 100644 --- a/fs/nfs/client.c +++ b/fs/nfs/client.c @@ -886,6 +886,7 @@ struct nfs_server *nfs_alloc_server(void) INIT_LIST_HEAD(&server->delegations); INIT_LIST_HEAD(&server->layouts); INIT_LIST_HEAD(&server->state_owners_lru); + INIT_LIST_HEAD(&server->ss_copies); atomic_set(&server->active, 0); diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 5f59b6f65a42..023aea8f6cf1 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -130,6 +130,37 @@ out_unlock: return err; } +static int handle_async_copy(struct nfs42_copy_res *res, + struct nfs_server *server, + struct file *src, + struct file *dst, + nfs4_stateid *src_stateid) +{ + struct nfs4_copy_state *copy; + int status = NFS4_OK; + + copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS); + if (!copy) + return -ENOMEM; + memcpy(©->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE); + init_completion(©->completion); + + spin_lock(&server->nfs_client->cl_lock); + list_add_tail(©->copies, &server->ss_copies); + spin_unlock(&server->nfs_client->cl_lock); + + wait_for_completion_interruptible(©->completion); + spin_lock(&server->nfs_client->cl_lock); + list_del_init(©->copies); + spin_unlock(&server->nfs_client->cl_lock); + res->write_res.count = copy->count; + memcpy(&res->write_res.verifier, ©->verf, sizeof(copy->verf)); + status = -copy->error; + + kfree(copy); + return status; +} + static ssize_t _nfs42_proc_copy(struct file *src, struct nfs_lock_context *src_lock, struct file *dst, @@ -168,9 +199,13 @@ static ssize_t _nfs42_proc_copy(struct file *src, if (status) return status; - res->commit_res.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS); - if (!res->commit_res.verf) - return -ENOMEM; + res->commit_res.verf = NULL; + if (args->sync) { + res->commit_res.verf = + kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS); + if (!res->commit_res.verf) + return -ENOMEM; + } status = nfs4_call_sync(server->client, server, &msg, &args->seq_args, &res->seq_res, 0); if (status == -ENOTSUPP) @@ -178,18 +213,27 @@ static ssize_t _nfs42_proc_copy(struct file *src, if (status) goto out; - if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier, + if (args->sync && + nfs_write_verifier_cmp(&res->write_res.verifier.verifier, &res->commit_res.verf->verifier)) { status = -EAGAIN; goto out; } + if (!res->synchronous) { + status = handle_async_copy(res, server, src, dst, + &args->src_stateid); + if (status) + return status; + } + truncate_pagecache_range(dst_inode, pos_dst, pos_dst + res->write_res.count); status = res->write_res.count; out: - kfree(res->commit_res.verf); + if (args->sync) + kfree(res->commit_res.verf); return status; } @@ -206,6 +250,7 @@ ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src, .dst_fh = NFS_FH(file_inode(dst)), .dst_pos = pos_dst, .count = count, + .sync = false, }; struct nfs42_copy_res res; struct nfs4_exception src_exception = { diff --git a/fs/nfs/nfs42xdr.c b/fs/nfs/nfs42xdr.c index 205c3567ea08..69f72ed2bf87 100644 --- a/fs/nfs/nfs42xdr.c +++ b/fs/nfs/nfs42xdr.c @@ -150,7 +150,7 @@ static void encode_copy(struct xdr_stream *xdr, encode_uint64(xdr, args->count); encode_uint32(xdr, 1); /* consecutive = true */ - encode_uint32(xdr, 1); /* synchronous = true */ + encode_uint32(xdr, args->sync); encode_uint32(xdr, 0); /* src server list */ } @@ -273,7 +273,8 @@ static void nfs4_xdr_enc_copy(struct rpc_rqst *req, encode_savefh(xdr, &hdr); encode_putfh(xdr, args->dst_fh, &hdr); encode_copy(xdr, args, &hdr); - encode_copy_commit(xdr, args, &hdr); + if (args->sync) + encode_copy_commit(xdr, args, &hdr); encode_nops(&hdr); } @@ -551,7 +552,8 @@ static int nfs4_xdr_dec_copy(struct rpc_rqst *rqstp, status = decode_copy(xdr, res); if (status) goto out; - status = decode_commit(xdr, &res->commit_res); + if (res->commit_res.verf) + status = decode_commit(xdr, &res->commit_res); out: return status; } diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h index 2f129bbfaae8..645ad8e342f6 100644 --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -185,6 +185,15 @@ struct nfs_inode { struct inode vfs_inode; }; +struct nfs4_copy_state { + struct list_head copies; + nfs4_stateid stateid; + struct completion completion; + uint64_t count; + struct nfs_writeverf verf; + int error; +}; + /* * Access bit flags */ diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index fbc735f08d7e..f88952d7b9fb 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -208,6 +208,7 @@ struct nfs_server { struct list_head state_owners_lru; struct list_head layouts; struct list_head delegations; + struct list_head ss_copies; unsigned long mig_gen; unsigned long mig_status; diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 06ddfa31cbef..bd1c889a9ed9 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -1388,6 +1388,7 @@ struct nfs42_copy_args { u64 dst_pos; u64 count; + bool sync; }; struct nfs42_write_res { -- cgit v1.2.3 From bc0c9079b48ddcf1f8a6e1aaa277288b263c78d8 Mon Sep 17 00:00:00 2001 From: Olga Kornievskaia Date: Mon, 9 Jul 2018 15:13:32 -0400 Subject: NFS handle COPY reply CB_OFFLOAD call race It's possible that server replies back with CB_OFFLOAD call and COPY reply at the same time such that client will process CB_OFFLOAD before reply to COPY. For that keep a list of pending callback stateids received and then before waiting on completion check the pending list. Cleanup any pending copies on the client shutdown. Signed-off-by: Olga Kornievskaia Signed-off-by: Anna Schumaker --- fs/nfs/callback_proc.c | 17 ++++++++++++++--- fs/nfs/nfs42proc.c | 22 ++++++++++++++++++++-- fs/nfs/nfs4client.c | 15 +++++++++++++++ include/linux/nfs_fs_sb.h | 1 + 4 files changed, 50 insertions(+), 5 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index acdda259912e..cd733649646b 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -681,11 +681,12 @@ __be32 nfs4_callback_offload(void *data, void *dummy, struct cb_offloadargs *args = data; struct nfs_server *server; struct nfs4_copy_state *copy; + bool found = false; + spin_lock(&cps->clp->cl_lock); rcu_read_lock(); list_for_each_entry_rcu(server, &cps->clp->cl_superblocks, client_link) { - spin_lock(&server->nfs_client->cl_lock); list_for_each_entry(copy, &server->ss_copies, copies) { if (memcmp(args->coa_stateid.other, copy->stateid.other, @@ -693,13 +694,23 @@ __be32 nfs4_callback_offload(void *data, void *dummy, continue; nfs4_copy_cb_args(copy, args); complete(©->completion); - spin_unlock(&server->nfs_client->cl_lock); + found = true; goto out; } - spin_unlock(&server->nfs_client->cl_lock); } out: rcu_read_unlock(); + if (!found) { + copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS); + if (!copy) { + spin_unlock(&cps->clp->cl_lock); + return htonl(NFS4ERR_SERVERFAULT); + } + memcpy(©->stateid, &args->coa_stateid, NFS4_STATEID_SIZE); + nfs4_copy_cb_args(copy, args); + list_add_tail(©->copies, &cps->clp->pending_cb_stateids); + } + spin_unlock(&cps->clp->cl_lock); return 0; } diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c index 023aea8f6cf1..c7d31f72070e 100644 --- a/fs/nfs/nfs42proc.c +++ b/fs/nfs/nfs42proc.c @@ -138,14 +138,31 @@ static int handle_async_copy(struct nfs42_copy_res *res, { struct nfs4_copy_state *copy; int status = NFS4_OK; + bool found_pending = false; + + spin_lock(&server->nfs_client->cl_lock); + list_for_each_entry(copy, &server->nfs_client->pending_cb_stateids, + copies) { + if (memcmp(&res->write_res.stateid, ©->stateid, + NFS4_STATEID_SIZE)) + continue; + found_pending = true; + list_del(©->copies); + break; + } + if (found_pending) { + spin_unlock(&server->nfs_client->cl_lock); + goto out; + } copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS); - if (!copy) + if (!copy) { + spin_unlock(&server->nfs_client->cl_lock); return -ENOMEM; + } memcpy(©->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE); init_completion(©->completion); - spin_lock(&server->nfs_client->cl_lock); list_add_tail(©->copies, &server->ss_copies); spin_unlock(&server->nfs_client->cl_lock); @@ -153,6 +170,7 @@ static int handle_async_copy(struct nfs42_copy_res *res, spin_lock(&server->nfs_client->cl_lock); list_del_init(©->copies); spin_unlock(&server->nfs_client->cl_lock); +out: res->write_res.count = copy->count; memcpy(&res->write_res.verifier, ©->verf, sizeof(copy->verf)); status = -copy->error; diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c index d7124fb12041..146e30862234 100644 --- a/fs/nfs/nfs4client.c +++ b/fs/nfs/nfs4client.c @@ -156,9 +156,23 @@ nfs4_shutdown_ds_clients(struct nfs_client *clp) } } +static void +nfs4_cleanup_callback(struct nfs_client *clp) +{ + struct nfs4_copy_state *cp_state; + + while (!list_empty(&clp->pending_cb_stateids)) { + cp_state = list_entry(clp->pending_cb_stateids.next, + struct nfs4_copy_state, copies); + list_del(&cp_state->copies); + kfree(cp_state); + } +} + void nfs41_shutdown_client(struct nfs_client *clp) { if (nfs4_has_session(clp)) { + nfs4_cleanup_callback(clp); nfs4_shutdown_ds_clients(clp); nfs4_destroy_session(clp->cl_session); nfs4_destroy_clientid(clp); @@ -202,6 +216,7 @@ struct nfs_client *nfs4_alloc_client(const struct nfs_client_initdata *cl_init) #if IS_ENABLED(CONFIG_NFS_V4_1) init_waitqueue_head(&clp->cl_lock_waitq); #endif + INIT_LIST_HEAD(&clp->pending_cb_stateids); return clp; error: diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h index f88952d7b9fb..bf39d9c92201 100644 --- a/include/linux/nfs_fs_sb.h +++ b/include/linux/nfs_fs_sb.h @@ -121,6 +121,7 @@ struct nfs_client { #endif struct net *cl_net; + struct list_head pending_cb_stateids; }; /* -- cgit v1.2.3 From 8618289c46556fd4dd259a1af02ccc448032f48d Mon Sep 17 00:00:00 2001 From: Trond Myklebust Date: Tue, 14 Aug 2018 17:55:56 -0400 Subject: NFSv4: Fix a sleep in atomic context in nfs4_callback_sequence() We must drop the lock before we can sleep in referring_call_exists(). Reported-by: Jia-Ju Bai Fixes: 045d2a6d076a ("NFSv4.1: Delay callback processing...") Cc: stable@vger.kernel.org # v4.9+ Signed-off-by: Trond Myklebust Signed-off-by: Anna Schumaker --- fs/nfs/callback_proc.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'fs/nfs/callback_proc.c') diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c index cd733649646b..fa515d5ea5ba 100644 --- a/fs/nfs/callback_proc.c +++ b/fs/nfs/callback_proc.c @@ -442,11 +442,14 @@ validate_seqid(const struct nfs4_slot_table *tbl, const struct nfs4_slot *slot, * a match. If the slot is in use and the sequence numbers match, the * client is still waiting for a response to the original request. */ -static bool referring_call_exists(struct nfs_client *clp, +static int referring_call_exists(struct nfs_client *clp, uint32_t nrclists, - struct referring_call_list *rclists) + struct referring_call_list *rclists, + spinlock_t *lock) + __releases(lock) + __acquires(lock) { - bool status = false; + int status = 0; int i, j; struct nfs4_session *session; struct nfs4_slot_table *tbl; @@ -469,8 +472,10 @@ static bool referring_call_exists(struct nfs_client *clp, for (j = 0; j < rclist->rcl_nrefcalls; j++) { ref = &rclist->rcl_refcalls[j]; + spin_unlock(lock); status = nfs4_slot_wait_on_seqid(tbl, ref->rc_slotid, ref->rc_sequenceid, HZ >> 1) < 0; + spin_lock(lock); if (status) goto out; } @@ -547,7 +552,8 @@ __be32 nfs4_callback_sequence(void *argp, void *resp, * related callback was received before the response to the original * call. */ - if (referring_call_exists(clp, args->csa_nrclists, args->csa_rclists)) { + if (referring_call_exists(clp, args->csa_nrclists, args->csa_rclists, + &tbl->slot_tbl_lock) < 0) { status = htonl(NFS4ERR_DELAY); goto out_unlock; } -- cgit v1.2.3