The nfs4_proc_fs_locations() function is invoked during referral processing to perform a GETATTR(fs_locations) on an object's parent directory in order to discover the target of the referral. It performs a LOOKUP, so the client needs to know the parent's file handle a priori. o During migration recovery, we need to probe fs_locations informa- tion on an FSID's root directory. The "parent" of a root directory is not available for a LOOKUP operation. o Recovering from NFS4ERR_LEASE_MOVED is a process of walking over a list of known FSIDs that reside on the server, and probing whether they have migrated. Once the server has detected that the client has probed all migrated file systems, it stops returning NFS4ERR_LEASE_MOVED. A minor version zero server needs to know what client ID is requesting fs_locations information so it can clear the flag that forces it to continue returning NFS4ERR_LEASE_MOVED. This flag is set per client ID and per FSID. However, the client ID is not an argument of either the PUTFH or GETATTR operations. Later minor versions have client ID information embedded in the underlying session. Therefore, by convention, minor version zero clients send a RENEW operation in the same compound as the GETATTR(fs_locations), since RENEW has one argument: the short-form client ID. This allows a minor version zero server to identify correctly the client that is probing for a migration. o For various subtle reasons, servers can't return NFS4ERR_DELAY to state-changing operations while they are actually doing the migration. Instead, they put off clients during the brief window when the data is actually unavailable by returning DELAY to the GETATTR(fs_locations), since GETATTR doesn't mutate NFSv4 state. So our client must be able to deal properly with an NFS4ERR_DELAY reply during a migration status probe. To handle all this random wackiness, we need a variant of nfs4_proc_fs_locations() that can operate directly on a target file handle, rather than taking a name and doing a LOOKUP as part of retrieving fs_locations from the server. It also must properly append a RENEW operation as needed, and it should retry NFS4ERR_DELAY as appropriate. Introduce nfs4_proc_get_mig_status() to fill this role, and add the requisite XDR encoding and decoding paraphenalia. Under the covers, both nfs4_proc_foo functions use the FS_LOCATIONS XDR routines. This is a little awkward, but is necessary because it's currently not straightforward to add new NFSv4 procedures due to a bug in nfsstat. Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx> --- fs/nfs/nfs4_fs.h | 2 + fs/nfs/nfs4proc.c | 89 ++++++++++++++++++++++++++++++++++++++++++++++- fs/nfs/nfs4xdr.c | 48 ++++++++++++++++++------- include/linux/nfs_xdr.h | 7 +++- 4 files changed, 130 insertions(+), 16 deletions(-) diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h index e1c261d..06b7457 100644 --- a/fs/nfs/nfs4_fs.h +++ b/fs/nfs/nfs4_fs.h @@ -241,6 +241,8 @@ extern int nfs4_do_close(struct path *path, struct nfs4_state *state, gfp_t gfp_ 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 int nfs4_proc_get_mig_status(struct nfs_server *server, + struct nfs4_fs_locations *fs_locations, struct page *page); extern void nfs4_release_lockowner(const struct nfs4_lock_state *); extern const struct xattr_handler *nfs4_xattr_handlers[]; diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index aa90145..561ffa1 100644 --- a/fs/nfs/nfs4proc.c +++ b/fs/nfs/nfs4proc.c @@ -4659,6 +4659,16 @@ static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr) fattr->nlink = 2; } +/** + * nfs4_proc_fs_locations - retrieve locations array for a named object + * + * @dir: inode of parent directory + * @name: qstr containing name of object to query + * @locations: result of query + * @page: buffer + * + * Returns zero on success, or a negative errno code + */ int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name, struct nfs4_fs_locations *fs_locations, struct page *page) { @@ -4683,16 +4693,91 @@ int nfs4_proc_fs_locations(struct inode *dir, const struct qstr *name, }; int status; - dprintk("%s: start\n", __func__); + dprintk("--> %s: FSID %llx:%llx on \"%s\"\n", __func__, + (unsigned long long)server->fsid.major, + (unsigned long long)server->fsid.minor, + server->nfs_client->cl_hostname); nfs_fattr_init(&fs_locations->fattr); fs_locations->server = server; fs_locations->nlocations = 0; status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0); nfs_fixup_referral_attributes(&fs_locations->fattr); - dprintk("%s: returned status = %d\n", __func__, status); + dprintk("<-- %s status=%d\n", __func__, status); + return status; +} + +static int _nfs4_proc_get_mig_status(struct nfs_server *server, + struct nfs4_fs_locations *locations, + struct page *page) +{ + u32 bitmask[2] = { + [0] = FATTR4_WORD0_FSID | + FATTR4_WORD0_FS_LOCATIONS, + }; + struct nfs4_fs_locations_arg args = { + .client = server->nfs_client, + .fh = server->rootfh, + .page = page, + .bitmask = bitmask, + }; + struct nfs4_fs_locations_res res = { + .fs_locations = locations, + }; + struct rpc_message msg = { + .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS], + .rpc_argp = &args, + .rpc_resp = &res, + }; + int status; + + dprintk("--> %s: FSID %llx:%llx on \"%s\"\n", __func__, + (unsigned long long)server->fsid.major, + (unsigned long long)server->fsid.minor, + args.client->cl_hostname); + nfs_display_fhandle(args.fh, "Probing file handle"); + + args.mig_status = res.mig_status = 1; + if (args.client->cl_mvops->minor_version == 0) + args.renew = res.renew = 1; + nfs_fattr_init(&locations->fattr); + locations->server = server; + locations->nlocations = 0; + status = nfs4_call_sync(server->client, server, &msg, + &args.seq_args, &res.seq_res, 0); + dprintk("<-- %s status=%d\n", __func__, status); return status; } +/** + * nfs4_proc_get_mig_status - probe migration status of an export + * + * @server: local state for server + * @locations: result of query + * @page: buffer + * + * Returns zero on success, or a negative errno code + * + * Servers often return NFS4ERR_DELAY to our migration probe while a + * migration is in progress. This allows a server to delay client + * activity without returning NFS4ERR_DELAY on a sequence ID mutating + * operation. + */ +int nfs4_proc_get_mig_status(struct nfs_server *server, + struct nfs4_fs_locations *locations, + struct page *page) +{ + struct nfs4_exception exception = { }; + int err; + + do { + err = _nfs4_proc_get_mig_status(server, locations, page); + if (err != -NFS4ERR_DELAY) + break; + nfs4_handle_exception(server, err, &exception); + } while (exception.retry); + return err; +} + static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors) { int status; diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index 76209b2..a3119ed 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -684,13 +684,15 @@ static int nfs4_stat_to_errno(int); encode_sequence_maxsz + \ encode_putfh_maxsz + \ encode_lookup_maxsz + \ - encode_fs_locations_maxsz) + encode_fs_locations_maxsz + \ + encode_renew_maxsz) #define NFS4_dec_fs_locations_sz \ (compound_decode_hdr_maxsz + \ decode_sequence_maxsz + \ decode_putfh_maxsz + \ decode_lookup_maxsz + \ - decode_fs_locations_maxsz) + decode_fs_locations_maxsz + \ + encode_renew_maxsz) #define NFS4_enc_secinfo_sz (compound_encode_hdr_maxsz + \ encode_sequence_maxsz + \ encode_putfh_maxsz + \ @@ -2530,11 +2532,20 @@ static void nfs4_xdr_enc_fs_locations(struct rpc_rqst *req, encode_compound_hdr(xdr, req, &hdr); encode_sequence(xdr, &args->seq_args, &hdr); - encode_putfh(xdr, args->dir_fh, &hdr); - encode_lookup(xdr, args->name, &hdr); - replen = hdr.replen; /* get the attribute into args->page */ - encode_fs_locations(xdr, args->bitmask, &hdr); + if (args->mig_status) { + encode_putfh(xdr, args->fh, &hdr); + replen = hdr.replen; + encode_fs_locations(xdr, args->bitmask, &hdr); + if (args->renew) + encode_renew(xdr, args->client, &hdr); + } else { + encode_putfh(xdr, args->dir_fh, &hdr); + encode_lookup(xdr, args->name, &hdr); + replen = hdr.replen; + encode_fs_locations(xdr, args->bitmask, &hdr); + } + /* Set up reply kvec to capture returned fs_locations array. */ xdr_inline_pages(&req->rq_rcv_buf, replen << 2, &args->page, 0, PAGE_SIZE); encode_nops(&hdr); @@ -6133,13 +6144,24 @@ static int nfs4_xdr_dec_fs_locations(struct rpc_rqst *req, status = decode_putfh(xdr); if (status) goto out; - status = decode_lookup(xdr); - if (status) - goto out; - xdr_enter_page(xdr, PAGE_SIZE); - status = decode_getfattr(xdr, &res->fs_locations->fattr, - res->fs_locations->server, - !RPC_IS_ASYNC(req->rq_task)); + if (res->mig_status) { + xdr_enter_page(xdr, PAGE_SIZE); + status = decode_getfattr(xdr, &res->fs_locations->fattr, + res->fs_locations->server, + !RPC_IS_ASYNC(req->rq_task)); + if (status) + goto out; + if (res->renew) + status = decode_renew(xdr); + } else { + status = decode_lookup(xdr); + if (status) + goto out; + xdr_enter_page(xdr, PAGE_SIZE); + status = decode_getfattr(xdr, &res->fs_locations->fattr, + res->fs_locations->server, + !RPC_IS_ASYNC(req->rq_task)); + } out: return status; } diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h index 26e693e..8168764 100644 --- a/include/linux/nfs_xdr.h +++ b/include/linux/nfs_xdr.h @@ -955,16 +955,21 @@ struct nfs4_fs_locations { }; struct nfs4_fs_locations_arg { + const struct nfs_client *client; const struct nfs_fh *dir_fh; + const struct nfs_fh *fh; const struct qstr *name; struct page *page; const u32 *bitmask; - struct nfs4_sequence_args seq_args; + struct nfs4_sequence_args seq_args; + unsigned char mig_status:1, renew:1; }; struct nfs4_fs_locations_res { struct nfs4_fs_locations *fs_locations; struct nfs4_sequence_res seq_res; + unsigned char mig_status:1, + renew:1; }; struct nfs4_secinfo_oid { -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html