GIT a49e0450f06325e1bd7fdd5642ae560eded12642 rsync://client.linux-nfs.org/pub/linux/nfs-2.6.git --- diff --git a/fs/lockd/host.c b/fs/lockd/host.c --- a/fs/lockd/host.c +++ b/fs/lockd/host.c @@ -173,11 +173,10 @@ nlm_bind_host(struct nlm_host *host) /* If we've already created an RPC client, check whether * RPC rebind is required - * Note: why keep rebinding if we're on a tcp connection? */ if ((clnt = host->h_rpcclnt) != NULL) { xprt = clnt->cl_xprt; - if (!xprt->stream && time_after_eq(jiffies, host->h_nextrebind)) { + if (time_after_eq(jiffies, host->h_nextrebind)) { clnt->cl_port = 0; host->h_nextrebind = jiffies + NLM_HOST_REBIND; dprintk("lockd: next rebind in %ld jiffies\n", @@ -189,7 +188,6 @@ nlm_bind_host(struct nlm_host *host) goto forgetit; xprt_set_timeout(&xprt->timeout, 5, nlmsvc_timeout); - xprt->nocong = 1; /* No congestion control for NLM */ xprt->resvport = 1; /* NLM requires a reserved port */ /* Existing NLM servers accept AUTH_UNIX only */ diff --git a/fs/locks.c b/fs/locks.c --- a/fs/locks.c +++ b/fs/locks.c @@ -828,12 +828,16 @@ static int __posix_lock_file(struct inod /* Detect adjacent or overlapping regions (if same lock type) */ if (request->fl_type == fl->fl_type) { + /* In all comparisons of start vs end, use + * "start - 1" rather than "end + 1". If end + * is OFFSET_MAX, end + 1 will become negative. + */ if (fl->fl_end < request->fl_start - 1) goto next_lock; /* If the next lock in the list has entirely bigger * addresses than the new one, insert the lock here. */ - if (fl->fl_start > request->fl_end + 1) + if (fl->fl_start - 1 > request->fl_end) break; /* If we come here, the new and old lock are of the diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -565,8 +565,6 @@ static int nfs_readdir(struct file *filp } } unlock_kernel(); - if (desc->error < 0) - return desc->error; if (res < 0) return res; return 0; @@ -1539,7 +1537,8 @@ static int nfs_rename(struct inode *old_ #endif goto out; } - } + } else + new_inode->i_nlink--; go_ahead: /* diff --git a/fs/nfs/idmap.c b/fs/nfs/idmap.c --- a/fs/nfs/idmap.c +++ b/fs/nfs/idmap.c @@ -66,7 +66,6 @@ struct idmap_hashtable { }; struct idmap { - char idmap_path[48]; struct dentry *idmap_dentry; wait_queue_head_t idmap_wq; struct idmap_msg idmap_im; @@ -102,11 +101,8 @@ nfs_idmap_new(struct nfs4_client *clp) memset(idmap, 0, sizeof(*idmap)); - snprintf(idmap->idmap_path, sizeof(idmap->idmap_path), - "%s/idmap", clp->cl_rpcclient->cl_pathname); - - idmap->idmap_dentry = rpc_mkpipe(idmap->idmap_path, - idmap, &idmap_upcall_ops, 0); + idmap->idmap_dentry = rpc_mkpipe(clp->cl_rpcclient->cl_dentry, + "idmap", idmap, &idmap_upcall_ops, 0); if (IS_ERR(idmap->idmap_dentry)) { kfree(idmap); return; @@ -128,7 +124,7 @@ nfs_idmap_delete(struct nfs4_client *clp if (!idmap) return; - rpc_unlink(idmap->idmap_path); + rpc_unlink(idmap->idmap_dentry); clp->cl_idmap = NULL; kfree(idmap); } diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c --- a/fs/nfs/inode.c +++ b/fs/nfs/inode.c @@ -356,6 +356,35 @@ out_no_root: return no_root_error; } +static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans) +{ + to->to_initval = timeo * HZ / 10; + to->to_retries = retrans; + if (!to->to_retries) + to->to_retries = 2; + + switch (proto) { + case IPPROTO_TCP: + if (!to->to_initval) + to->to_initval = 60 * HZ; + if (to->to_initval > NFS_MAX_TCP_TIMEOUT) + to->to_initval = NFS_MAX_TCP_TIMEOUT; + to->to_increment = to->to_initval; + to->to_maxval = to->to_initval + (to->to_increment * to->to_retries); + to->to_exponential = 0; + break; + case IPPROTO_UDP: + default: + if (!to->to_initval) + to->to_initval = 11 * HZ / 10; + if (to->to_initval > NFS_MAX_UDP_TIMEOUT) + to->to_initval = NFS_MAX_UDP_TIMEOUT; + to->to_maxval = NFS_MAX_UDP_TIMEOUT; + to->to_exponential = 1; + break; + } +} + /* * Create an RPC client handle. */ @@ -365,22 +394,12 @@ nfs_create_client(struct nfs_server *ser struct rpc_timeout timeparms; struct rpc_xprt *xprt = NULL; struct rpc_clnt *clnt = NULL; - int tcp = (data->flags & NFS_MOUNT_TCP); + int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP; - /* Initialize timeout values */ - timeparms.to_initval = data->timeo * HZ / 10; - timeparms.to_retries = data->retrans; - timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT; - timeparms.to_exponential = 1; - - if (!timeparms.to_initval) - timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10; - if (!timeparms.to_retries) - timeparms.to_retries = 5; + nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans); /* create transport and client */ - xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP, - &server->addr, &timeparms); + xprt = xprt_create_proto(proto, &server->addr, &timeparms); if (IS_ERR(xprt)) { dprintk("%s: cannot create RPC transport. Error = %ld\n", __FUNCTION__, PTR_ERR(xprt)); @@ -574,7 +593,6 @@ static int nfs_show_options(struct seq_f { NFS_MOUNT_SOFT, ",soft", ",hard" }, { NFS_MOUNT_INTR, ",intr", "" }, { NFS_MOUNT_POSIX, ",posix", "" }, - { NFS_MOUNT_TCP, ",tcp", ",udp" }, { NFS_MOUNT_NOCTO, ",nocto", "" }, { NFS_MOUNT_NOAC, ",noac", "" }, { NFS_MOUNT_NONLM, ",nolock", ",lock" }, @@ -583,6 +601,8 @@ static int nfs_show_options(struct seq_f }; struct proc_nfs_info *nfs_infop; struct nfs_server *nfss = NFS_SB(mnt->mnt_sb); + char buf[12]; + char *proto; seq_printf(m, ",v%d", nfss->rpc_ops->version); seq_printf(m, ",rsize=%d", nfss->rsize); @@ -601,6 +621,18 @@ static int nfs_show_options(struct seq_f else seq_puts(m, nfs_infop->nostr); } + switch (nfss->client->cl_xprt->prot) { + case IPPROTO_TCP: + proto = "tcp"; + break; + case IPPROTO_UDP: + proto = "udp"; + break; + default: + snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot); + proto = buf; + } + seq_printf(m, ",proto=%s", proto); seq_puts(m, ",addr="); seq_escape(m, nfss->hostname, " \t\n\\"); return 0; @@ -1672,7 +1704,7 @@ static int nfs4_fill_super(struct super_ struct rpc_clnt *clnt = NULL; struct rpc_timeout timeparms; rpc_authflavor_t authflavour; - int proto, err = -EIO; + int err = -EIO; sb->s_blocksize_bits = 0; sb->s_blocksize = 0; @@ -1690,30 +1722,8 @@ static int nfs4_fill_super(struct super_ server->acdirmax = data->acdirmax*HZ; server->rpc_ops = &nfs_v4_clientops; - /* Initialize timeout values */ - timeparms.to_initval = data->timeo * HZ / 10; - timeparms.to_retries = data->retrans; - timeparms.to_exponential = 1; - if (!timeparms.to_retries) - timeparms.to_retries = 5; - - proto = data->proto; - /* Which IP protocol do we use? */ - switch (proto) { - case IPPROTO_TCP: - timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT; - if (!timeparms.to_initval) - timeparms.to_initval = 600 * HZ / 10; - break; - case IPPROTO_UDP: - timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT; - if (!timeparms.to_initval) - timeparms.to_initval = 11 * HZ / 10; - break; - default: - return -EINVAL; - } + nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans); clp = nfs4_get_client(&server->addr.sin_addr); if (!clp) { @@ -1738,7 +1748,7 @@ static int nfs4_fill_super(struct super_ down_write(&clp->cl_sem); if (IS_ERR(clp->cl_rpcclient)) { - xprt = xprt_create_proto(proto, &server->addr, &timeparms); + xprt = xprt_create_proto(data->proto, &server->addr, &timeparms); if (IS_ERR(xprt)) { up_write(&clp->cl_sem); err = PTR_ERR(xprt); diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h --- a/include/linux/nfs_fs.h +++ b/include/linux/nfs_fs.h @@ -41,6 +41,10 @@ #define NFS_MAX_FILE_IO_BUFFER_SIZE 32768 #define NFS_DEF_FILE_IO_BUFFER_SIZE 4096 +/* Default timeout values */ +#define NFS_MAX_UDP_TIMEOUT (60*HZ) +#define NFS_MAX_TCP_TIMEOUT (600*HZ) + /* * superblock magic number for NFS */ diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h --- a/include/linux/sunrpc/clnt.h +++ b/include/linux/sunrpc/clnt.h @@ -59,7 +59,7 @@ struct rpc_clnt { int cl_nodelen; /* nodename length */ char cl_nodename[UNX_MAXNODENAME]; - char cl_pathname[30];/* Path in rpc_pipe_fs */ + struct dentry * __cl_parent_dentry; struct dentry * cl_dentry; /* inode */ struct rpc_clnt * cl_parent; /* Points to parent of clones */ struct rpc_rtt cl_rtt_default; diff --git a/include/linux/sunrpc/debug.h b/include/linux/sunrpc/debug.h --- a/include/linux/sunrpc/debug.h +++ b/include/linux/sunrpc/debug.h @@ -32,6 +32,7 @@ #define RPCDBG_AUTH 0x0010 #define RPCDBG_PMAP 0x0020 #define RPCDBG_SCHED 0x0040 +#define RPCDBG_TRANS 0x0080 #define RPCDBG_SVCSOCK 0x0100 #define RPCDBG_SVCDSP 0x0200 #define RPCDBG_MISC 0x0400 @@ -94,6 +95,8 @@ enum { CTL_NLMDEBUG, CTL_SLOTTABLE_UDP, CTL_SLOTTABLE_TCP, + CTL_MIN_RESVPORT, + CTL_MAX_RESVPORT, }; #endif /* _LINUX_SUNRPC_DEBUG_H_ */ diff --git a/include/linux/sunrpc/msg_prot.h b/include/linux/sunrpc/msg_prot.h --- a/include/linux/sunrpc/msg_prot.h +++ b/include/linux/sunrpc/msg_prot.h @@ -76,5 +76,30 @@ enum rpc_auth_stat { #define RPC_MAXNETNAMELEN 256 +/* + * From RFC 1831: + * + * "A record is composed of one or more record fragments. A record + * fragment is a four-byte header followed by 0 to (2**31) - 1 bytes of + * fragment data. The bytes encode an unsigned binary number; as with + * XDR integers, the byte order is from highest to lowest. The number + * encodes two values -- a boolean which indicates whether the fragment + * is the last fragment of the record (bit value 1 implies the fragment + * is the last fragment) and a 31-bit unsigned binary value which is the + * length in bytes of the fragment's data. The boolean value is the + * highest-order bit of the header; the length is the 31 low-order bits. + * (Note that this record specification is NOT in XDR standard form!)" + * + * The Linux RPC client always sends its requests in a single record + * fragment, limiting the maximum payload size for stream transports to + * 2GB. + */ + +typedef u32 rpc_fraghdr; + +#define RPC_LAST_STREAM_FRAGMENT (1U << 31) +#define RPC_FRAGMENT_SIZE_MASK (~RPC_LAST_STREAM_FRAGMENT) +#define RPC_MAX_FRAGMENT_SIZE ((1U << 31) - 1) + #endif /* __KERNEL__ */ #endif /* _LINUX_SUNRPC_MSGPROT_H_ */ diff --git a/include/linux/sunrpc/rpc_pipe_fs.h b/include/linux/sunrpc/rpc_pipe_fs.h --- a/include/linux/sunrpc/rpc_pipe_fs.h +++ b/include/linux/sunrpc/rpc_pipe_fs.h @@ -41,10 +41,11 @@ RPC_I(struct inode *inode) extern int rpc_queue_upcall(struct inode *, struct rpc_pipe_msg *); -extern struct dentry *rpc_mkdir(char *, struct rpc_clnt *); -extern int rpc_rmdir(char *); -extern struct dentry *rpc_mkpipe(char *, void *, struct rpc_pipe_ops *, int flags); -extern int rpc_unlink(char *); +extern struct dentry *rpc_mkdir(struct dentry *, char *, struct rpc_clnt *); +extern void rpc_rmdir(struct dentry *); +extern struct dentry *rpc_mkpipe(struct dentry *, char *, void *, + struct rpc_pipe_ops *, int flags); +extern void rpc_unlink(struct dentry *); #endif #endif diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h --- a/include/linux/sunrpc/xdr.h +++ b/include/linux/sunrpc/xdr.h @@ -161,14 +161,10 @@ typedef struct { typedef size_t (*skb_read_actor_t)(skb_reader_t *desc, void *to, size_t len); +extern int csum_partial_copy_to_xdr(struct xdr_buf *, struct sk_buff *); extern ssize_t xdr_partial_copy_from_skb(struct xdr_buf *, unsigned int, skb_reader_t *, skb_read_actor_t); -struct socket; -struct sockaddr; -extern int xdr_sendpages(struct socket *, struct sockaddr *, int, - struct xdr_buf *, unsigned int, int); - extern int xdr_encode_word(struct xdr_buf *, int, u32); extern int xdr_decode_word(struct xdr_buf *, int, u32 *); diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h --- a/include/linux/sunrpc/xprt.h +++ b/include/linux/sunrpc/xprt.h @@ -1,5 +1,5 @@ /* - * linux/include/linux/sunrpc/clnt_xprt.h + * linux/include/linux/sunrpc/xprt.h * * Declarations for the RPC transport interface. * @@ -15,20 +15,6 @@ #include #include -/* - * The transport code maintains an estimate on the maximum number of out- - * standing RPC requests, using a smoothed version of the congestion - * avoidance implemented in 44BSD. This is basically the Van Jacobson - * congestion algorithm: If a retransmit occurs, the congestion window is - * halved; otherwise, it is incremented by 1/cwnd when - * - * - a reply is received and - * - a full number of requests are outstanding and - * - the congestion window hasn't been updated recently. - * - * Upper procedures may check whether a request would block waiting for - * a free RPC slot by using the RPC_CONGESTED() macro. - */ extern unsigned int xprt_udp_slot_table_entries; extern unsigned int xprt_tcp_slot_table_entries; @@ -36,34 +22,23 @@ extern unsigned int xprt_tcp_slot_table_ #define RPC_DEF_SLOT_TABLE (16U) #define RPC_MAX_SLOT_TABLE (128U) -#define RPC_CWNDSHIFT (8U) -#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT) -#define RPC_INITCWND RPC_CWNDSCALE -#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT) -#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd) - -/* Default timeout values */ -#define RPC_MAX_UDP_TIMEOUT (60*HZ) -#define RPC_MAX_TCP_TIMEOUT (600*HZ) - /* - * Wait duration for an RPC TCP connection to be established. Solaris - * NFS over TCP uses 60 seconds, for example, which is in line with how - * long a server takes to reboot. + * RPC call and reply header size as number of 32bit words (verifier + * size computed separately) */ -#define RPC_CONNECT_TIMEOUT (60*HZ) +#define RPC_CALLHDRSIZE 6 +#define RPC_REPHDRSIZE 4 /* - * Delay an arbitrary number of seconds before attempting to reconnect - * after an error. + * Parameters for choosing a free port */ -#define RPC_REESTABLISH_TIMEOUT (15*HZ) +extern unsigned int xprt_min_resvport; +extern unsigned int xprt_max_resvport; -/* RPC call and reply header size as number of 32bit words (verifier - * size computed separately) - */ -#define RPC_CALLHDRSIZE 6 -#define RPC_REPHDRSIZE 4 +#define RPC_MIN_RESVPORT (1U) +#define RPC_MAX_RESVPORT (65535U) +#define RPC_DEF_MIN_RESVPORT (650U) +#define RPC_DEF_MAX_RESVPORT (1023U) /* * This describes a timeout strategy @@ -76,6 +51,9 @@ struct rpc_timeout { unsigned char to_exponential; }; +struct rpc_task; +struct rpc_xprt; + /* * This describes a complete RPC request */ @@ -121,12 +99,21 @@ struct rpc_rqst { #define rq_svec rq_snd_buf.head #define rq_slen rq_snd_buf.len -#define XPRT_LAST_FRAG (1 << 0) -#define XPRT_COPY_RECM (1 << 1) -#define XPRT_COPY_XID (1 << 2) -#define XPRT_COPY_DATA (1 << 3) +struct rpc_xprt_ops { + void (*set_buffer_size)(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize); + int (*reserve_xprt)(struct rpc_task *task); + void (*release_xprt)(struct rpc_xprt *xprt, struct rpc_task *task); + void (*connect)(struct rpc_task *task); + int (*send_request)(struct rpc_task *task); + void (*set_retrans_timeout)(struct rpc_task *task); + void (*timer)(struct rpc_task *task); + void (*release_request)(struct rpc_task *task); + void (*close)(struct rpc_xprt *xprt); + void (*destroy)(struct rpc_xprt *xprt); +}; struct rpc_xprt { + struct rpc_xprt_ops * ops; /* transport methods */ struct socket * sock; /* BSD socket layer */ struct sock * inet; /* INET layer */ @@ -137,11 +124,13 @@ struct rpc_xprt { unsigned long cong; /* current congestion */ unsigned long cwnd; /* congestion window */ - unsigned int rcvsize, /* socket receive buffer size */ - sndsize; /* socket send buffer size */ + size_t rcvsize, /* transport rcv buffer size */ + sndsize; /* transport send buffer size */ size_t max_payload; /* largest RPC payload size, in bytes */ + unsigned int tsh_size; /* size of transport specific + header */ struct rpc_wait_queue sending; /* requests waiting to send */ struct rpc_wait_queue resend; /* requests waiting to resend */ @@ -150,11 +139,9 @@ struct rpc_xprt { struct list_head free; /* free slots */ struct rpc_rqst * slot; /* slot table storage */ unsigned int max_reqs; /* total slots */ - unsigned long sockstate; /* Socket state */ + unsigned long state; /* transport state */ unsigned char shutdown : 1, /* being shut down */ - nocong : 1, /* no congestion control */ - resvport : 1, /* use a reserved port */ - stream : 1; /* TCP */ + resvport : 1; /* use a reserved port */ /* * XID @@ -171,22 +158,27 @@ struct rpc_xprt { unsigned long tcp_copied, /* copied to request */ tcp_flags; /* - * Connection of sockets + * Connection of transports */ - struct work_struct sock_connect; + unsigned long connect_timeout, + bind_timeout, + reestablish_timeout; + struct work_struct connect_worker; unsigned short port; + /* - * Disconnection of idle sockets + * Disconnection of idle transports */ struct work_struct task_cleanup; struct timer_list timer; - unsigned long last_used; + unsigned long last_used, + idle_timeout; /* * Send stuff */ - spinlock_t sock_lock; /* lock socket info */ - spinlock_t xprt_lock; /* lock xprt info */ + spinlock_t transport_lock; /* lock transport info */ + spinlock_t reserve_lock; /* lock slot table */ struct rpc_task * snd_task; /* Task blocked in send */ struct list_head recv; @@ -195,37 +187,110 @@ struct rpc_xprt { void (*old_data_ready)(struct sock *, int); void (*old_state_change)(struct sock *); void (*old_write_space)(struct sock *); - - wait_queue_head_t cong_wait; }; +#define XPRT_LAST_FRAG (1 << 0) +#define XPRT_COPY_RECM (1 << 1) +#define XPRT_COPY_XID (1 << 2) +#define XPRT_COPY_DATA (1 << 3) + #ifdef __KERNEL__ -struct rpc_xprt * xprt_create_proto(int proto, struct sockaddr_in *addr, - struct rpc_timeout *toparms); -int xprt_destroy(struct rpc_xprt *); -void xprt_set_timeout(struct rpc_timeout *, unsigned int, - unsigned long); - -void xprt_reserve(struct rpc_task *); -int xprt_prepare_transmit(struct rpc_task *); -void xprt_transmit(struct rpc_task *); -void xprt_receive(struct rpc_task *); +/* + * Transport operations used by ULPs + */ +struct rpc_xprt * xprt_create_proto(int proto, struct sockaddr_in *addr, struct rpc_timeout *to); +void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr); + +/* + * Generic internal transport functions + */ +void xprt_connect(struct rpc_task *task); +void xprt_reserve(struct rpc_task *task); +int xprt_reserve_xprt(struct rpc_task *task); +int xprt_reserve_xprt_cong(struct rpc_task *task); +int xprt_prepare_transmit(struct rpc_task *task); +void xprt_transmit(struct rpc_task *task); int xprt_adjust_timeout(struct rpc_rqst *req); -void xprt_release(struct rpc_task *); -void xprt_connect(struct rpc_task *); -void xprt_sock_setbufsize(struct rpc_xprt *); - -#define XPRT_LOCKED 0 -#define XPRT_CONNECT 1 -#define XPRT_CONNECTING 2 - -#define xprt_connected(xp) (test_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_set_connected(xp) (set_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_test_and_set_connected(xp) (test_and_set_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_test_and_clear_connected(xp) \ - (test_and_clear_bit(XPRT_CONNECT, &(xp)->sockstate)) -#define xprt_clear_connected(xp) (clear_bit(XPRT_CONNECT, &(xp)->sockstate)) +void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task); +void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task); +void xprt_release(struct rpc_task *task); +int xprt_destroy(struct rpc_xprt *xprt); + +static inline u32 *xprt_skip_transport_header(struct rpc_xprt *xprt, u32 *p) +{ + return p + xprt->tsh_size; +} + +/* + * Transport switch helper functions + */ +void xprt_set_retrans_timeout_def(struct rpc_task *task); +void xprt_set_retrans_timeout_rtt(struct rpc_task *task); +void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status); +void xprt_wait_for_buffer_space(struct rpc_task *task); +void xprt_write_space(struct rpc_xprt *xprt); +void xprt_update_rtt(struct rpc_task *task); +void xprt_adjust_cwnd(struct rpc_task *task, int result); +struct rpc_rqst * xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid); +void xprt_complete_rqst(struct rpc_task *task, int copied); +void xprt_release_rqst_cong(struct rpc_task *task); +void xprt_disconnect(struct rpc_xprt *xprt); + +/* + * Socket transport setup operations + */ +int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to); +int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to); + +/* + * Reserved bit positions in xprt->state + */ +#define XPRT_LOCKED (0) +#define XPRT_CONNECTED (1) +#define XPRT_CONNECTING (2) + +static inline void xprt_set_connected(struct rpc_xprt *xprt) +{ + set_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline void xprt_clear_connected(struct rpc_xprt *xprt) +{ + clear_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline int xprt_connected(struct rpc_xprt *xprt) +{ + return test_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline int xprt_test_and_set_connected(struct rpc_xprt *xprt) +{ + return test_and_set_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline int xprt_test_and_clear_connected(struct rpc_xprt *xprt) +{ + return test_and_clear_bit(XPRT_CONNECTED, &xprt->state); +} + +static inline void xprt_clear_connecting(struct rpc_xprt *xprt) +{ + smp_mb__before_clear_bit(); + clear_bit(XPRT_CONNECTING, &xprt->state); + smp_mb__after_clear_bit(); +} + +static inline int xprt_connecting(struct rpc_xprt *xprt) +{ + return test_bit(XPRT_CONNECTING, &xprt->state); +} + +static inline int xprt_test_and_set_connecting(struct rpc_xprt *xprt) +{ + return test_and_set_bit(XPRT_CONNECTING, &xprt->state); +} #endif /* __KERNEL__*/ diff --git a/net/sunrpc/Makefile b/net/sunrpc/Makefile --- a/net/sunrpc/Makefile +++ b/net/sunrpc/Makefile @@ -6,7 +6,7 @@ obj-$(CONFIG_SUNRPC) += sunrpc.o obj-$(CONFIG_SUNRPC_GSS) += auth_gss/ -sunrpc-y := clnt.o xprt.o sched.o \ +sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \ auth.o auth_null.o auth_unix.o \ svc.o svcsock.o svcauth.o svcauth_unix.o \ pmap_clnt.o timer.o xdr.o \ diff --git a/net/sunrpc/auth.c b/net/sunrpc/auth.c --- a/net/sunrpc/auth.c +++ b/net/sunrpc/auth.c @@ -11,7 +11,6 @@ #include #include #include -#include #include #include diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c --- a/net/sunrpc/auth_gss/auth_gss.c +++ b/net/sunrpc/auth_gss/auth_gss.c @@ -42,8 +42,6 @@ #include #include #include -#include -#include #include #include #include @@ -89,7 +87,6 @@ struct gss_auth { struct list_head upcalls; struct rpc_clnt *client; struct dentry *dentry; - char path[48]; spinlock_t lock; }; @@ -692,10 +689,8 @@ gss_create(struct rpc_clnt *clnt, rpc_au if (err) goto err_put_mech; - snprintf(gss_auth->path, sizeof(gss_auth->path), "%s/%s", - clnt->cl_pathname, - gss_auth->mech->gm_name); - gss_auth->dentry = rpc_mkpipe(gss_auth->path, clnt, &gss_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN); + gss_auth->dentry = rpc_mkpipe(clnt->cl_dentry, gss_auth->mech->gm_name, + clnt, &gss_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN); if (IS_ERR(gss_auth->dentry)) { err = PTR_ERR(gss_auth->dentry); goto err_put_mech; @@ -720,7 +715,7 @@ gss_destroy(struct rpc_auth *auth) auth, auth->au_flavor); gss_auth = container_of(auth, struct gss_auth, rpc_auth); - rpc_unlink(gss_auth->path); + rpc_unlink(gss_auth->dentry); gss_mech_put(gss_auth->mech); rpcauth_free_credcache(auth); @@ -846,10 +841,8 @@ gss_marshal(struct rpc_task *task, u32 * /* We compute the checksum for the verifier over the xdr-encoded bytes * starting with the xid and ending at the end of the credential: */ - iov.iov_base = req->rq_snd_buf.head[0].iov_base; - if (task->tk_client->cl_xprt->stream) - /* See clnt.c:call_header() */ - iov.iov_base += 4; + iov.iov_base = xprt_skip_transport_header(task->tk_xprt, + req->rq_snd_buf.head[0].iov_base); iov.iov_len = (u8 *)p - (u8 *)iov.iov_base; xdr_buf_from_iov(&iov, &verf_buf); diff --git a/net/sunrpc/auth_gss/gss_krb5_mech.c b/net/sunrpc/auth_gss/gss_krb5_mech.c --- a/net/sunrpc/auth_gss/gss_krb5_mech.c +++ b/net/sunrpc/auth_gss/gss_krb5_mech.c @@ -39,7 +39,6 @@ #include #include #include -#include #include #include #include diff --git a/net/sunrpc/auth_gss/gss_mech_switch.c b/net/sunrpc/auth_gss/gss_mech_switch.c --- a/net/sunrpc/auth_gss/gss_mech_switch.c +++ b/net/sunrpc/auth_gss/gss_mech_switch.c @@ -35,7 +35,6 @@ #include #include -#include #include #include #include diff --git a/net/sunrpc/auth_null.c b/net/sunrpc/auth_null.c --- a/net/sunrpc/auth_null.c +++ b/net/sunrpc/auth_null.c @@ -7,9 +7,7 @@ */ #include -#include #include -#include #include #include #include diff --git a/net/sunrpc/auth_unix.c b/net/sunrpc/auth_unix.c --- a/net/sunrpc/auth_unix.c +++ b/net/sunrpc/auth_unix.c @@ -9,8 +9,6 @@ #include #include #include -#include -#include #include #include diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c --- a/net/sunrpc/clnt.c +++ b/net/sunrpc/clnt.c @@ -1,5 +1,5 @@ /* - * linux/net/sunrpc/rpcclnt.c + * linux/net/sunrpc/clnt.c * * This file contains the high-level RPC interface. * It is modeled as a finite state machine to support both synchronous @@ -27,7 +27,6 @@ #include #include #include -#include #include #include @@ -53,6 +52,7 @@ static void call_allocate(struct rpc_tas static void call_encode(struct rpc_task *task); static void call_decode(struct rpc_task *task); static void call_bind(struct rpc_task *task); +static void call_bind_status(struct rpc_task *task); static void call_transmit(struct rpc_task *task); static void call_status(struct rpc_task *task); static void call_refresh(struct rpc_task *task); @@ -67,26 +67,42 @@ static u32 * call_verify(struct rpc_task static int rpc_setup_pipedir(struct rpc_clnt *clnt, char *dir_name) { - static uint32_t clntid; + static unsigned int clntid; + char name[128]; int error; if (dir_name == NULL) return 0; - for (;;) { - snprintf(clnt->cl_pathname, sizeof(clnt->cl_pathname), - "%s/clnt%x", dir_name, - (unsigned int)clntid++); - clnt->cl_pathname[sizeof(clnt->cl_pathname) - 1] = '\0'; - clnt->cl_dentry = rpc_mkdir(clnt->cl_pathname, clnt); - if (!IS_ERR(clnt->cl_dentry)) - return 0; + + retry_parent: + clnt->__cl_parent_dentry = rpc_mkdir(NULL, dir_name, NULL); + if (IS_ERR(clnt->__cl_parent_dentry)) { + error = PTR_ERR(clnt->__cl_parent_dentry); + if (error == -EEXIST) + goto retry_parent; /* XXX(hch): WTF? */ + + printk(KERN_INFO "RPC: Couldn't create pipefs entry %s, error %d\n", + dir_name, error); + return error; + } + + + retry_child: + snprintf(name, sizeof(name), "clnt%x", clntid++); + name[sizeof(name) - 1] = '\0'; + + clnt->cl_dentry = rpc_mkdir(clnt->__cl_parent_dentry, name, clnt); + if (IS_ERR(clnt->cl_dentry)) { error = PTR_ERR(clnt->cl_dentry); - if (error != -EEXIST) { - printk(KERN_INFO "RPC: Couldn't create pipefs entry %s, error %d\n", - clnt->cl_pathname, error); - return error; - } + if (error == -EEXIST) + goto retry_child; + printk(KERN_INFO "RPC: Couldn't create pipefs entry %s, error %d\n", + name, error); + rpc_rmdir(clnt->__cl_parent_dentry); + return error; } + + return 0; } /* @@ -174,7 +190,8 @@ rpc_new_client(struct rpc_xprt *xprt, ch return clnt; out_no_auth: - rpc_rmdir(clnt->cl_pathname); + rpc_rmdir(clnt->cl_dentry); + rpc_rmdir(clnt->__cl_parent_dentry); out_no_path: if (clnt->cl_server != clnt->cl_inline_name) kfree(clnt->cl_server); @@ -302,8 +319,10 @@ rpc_destroy_client(struct rpc_clnt *clnt rpc_destroy_client(clnt->cl_parent); goto out_free; } - if (clnt->cl_pathname[0]) - rpc_rmdir(clnt->cl_pathname); + if (clnt->cl_dentry) + rpc_rmdir(clnt->cl_dentry); + if (clnt->__cl_parent_dentry) + rpc_rmdir(clnt->__cl_parent_dentry); if (clnt->cl_xprt) { xprt_destroy(clnt->cl_xprt); clnt->cl_xprt = NULL; @@ -517,15 +536,8 @@ void rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize) { struct rpc_xprt *xprt = clnt->cl_xprt; - - xprt->sndsize = 0; - if (sndsize) - xprt->sndsize = sndsize + RPC_SLACK_SPACE; - xprt->rcvsize = 0; - if (rcvsize) - xprt->rcvsize = rcvsize + RPC_SLACK_SPACE; - if (xprt_connected(xprt)) - xprt_sock_setbufsize(xprt); + if (xprt->ops->set_buffer_size) + xprt->ops->set_buffer_size(xprt, sndsize, rcvsize); } /* @@ -734,43 +746,94 @@ static void call_bind(struct rpc_task *task) { struct rpc_clnt *clnt = task->tk_client; - struct rpc_xprt *xprt = clnt->cl_xprt; - dprintk("RPC: %4d call_bind xprt %p %s connected\n", task->tk_pid, - xprt, (xprt_connected(xprt) ? "is" : "is not")); - - task->tk_action = (xprt_connected(xprt)) ? call_transmit : call_connect; + dprintk("RPC: %4d call_bind (status %d)\n", + task->tk_pid, task->tk_status); + task->tk_action = call_connect; if (!clnt->cl_port) { - task->tk_action = call_connect; - task->tk_timeout = RPC_CONNECT_TIMEOUT; + task->tk_action = call_bind_status; + task->tk_timeout = task->tk_xprt->bind_timeout; rpc_getport(task, clnt); } } /* - * 4a. Connect to the RPC server (TCP case) + * 4a. Sort out bind result */ static void -call_connect(struct rpc_task *task) +call_bind_status(struct rpc_task *task) { - struct rpc_clnt *clnt = task->tk_client; - - dprintk("RPC: %4d call_connect status %d\n", - task->tk_pid, task->tk_status); + int status = -EACCES; - if (xprt_connected(clnt->cl_xprt)) { - task->tk_action = call_transmit; + if (task->tk_status >= 0) { + dprintk("RPC: %4d call_bind_status (status %d)\n", + task->tk_pid, task->tk_status); + task->tk_status = 0; + task->tk_action = call_connect; return; } - task->tk_action = call_connect_status; - if (task->tk_status < 0) - return; - xprt_connect(task); + + switch (task->tk_status) { + case -EACCES: + dprintk("RPC: %4d remote rpcbind: RPC program/version unavailable\n", + task->tk_pid); + break; + case -ETIMEDOUT: + dprintk("RPC: %4d rpcbind request timed out\n", + task->tk_pid); + if (RPC_IS_SOFT(task)) { + status = -EIO; + break; + } + goto retry_bind; + case -EPFNOSUPPORT: + dprintk("RPC: %4d remote rpcbind service unavailable\n", + task->tk_pid); + break; + case -EPROTONOSUPPORT: + dprintk("RPC: %4d remote rpcbind version 2 unavailable\n", + task->tk_pid); + break; + default: + dprintk("RPC: %4d unrecognized rpcbind error (%d)\n", + task->tk_pid, -task->tk_status); + status = -EIO; + break; + } + + rpc_exit(task, status); + return; + +retry_bind: + task->tk_status = 0; + task->tk_action = call_bind; + return; +} + +/* + * 4b. Connect to the RPC server + */ +static void +call_connect(struct rpc_task *task) +{ + struct rpc_xprt *xprt = task->tk_xprt; + + dprintk("RPC: %4d call_connect xprt %p %s connected\n", + task->tk_pid, xprt, + (xprt_connected(xprt) ? "is" : "is not")); + + task->tk_action = call_transmit; + if (!xprt_connected(xprt)) { + task->tk_action = call_connect_status; + if (task->tk_status < 0) + return; + xprt_connect(task); + } } /* - * 4b. Sort out connect result + * 4c. Sort out connect result */ static void call_connect_status(struct rpc_task *task) @@ -778,6 +841,9 @@ call_connect_status(struct rpc_task *tas struct rpc_clnt *clnt = task->tk_client; int status = task->tk_status; + dprintk("RPC: %5u call_connect_status (status %d)\n", + task->tk_pid, task->tk_status); + task->tk_status = 0; if (status >= 0) { clnt->cl_stats->netreconn++; @@ -785,17 +851,19 @@ call_connect_status(struct rpc_task *tas return; } - /* Something failed: we may have to rebind */ + /* Something failed: remote service port may have changed */ if (clnt->cl_autobind) clnt->cl_port = 0; + switch (status) { case -ENOTCONN: case -ETIMEDOUT: case -EAGAIN: - task->tk_action = (clnt->cl_port == 0) ? call_bind : call_connect; + task->tk_action = call_bind; break; default: rpc_exit(task, -EIO); + break; } } @@ -1020,13 +1088,12 @@ static u32 * call_header(struct rpc_task *task) { struct rpc_clnt *clnt = task->tk_client; - struct rpc_xprt *xprt = clnt->cl_xprt; struct rpc_rqst *req = task->tk_rqstp; u32 *p = req->rq_svec[0].iov_base; /* FIXME: check buffer size? */ - if (xprt->stream) - *p++ = 0; /* fill in later */ + + p = xprt_skip_transport_header(task->tk_xprt, p); *p++ = req->rq_xid; /* XID */ *p++ = htonl(RPC_CALL); /* CALL */ *p++ = htonl(RPC_VERSION); /* RPC version */ diff --git a/net/sunrpc/pmap_clnt.c b/net/sunrpc/pmap_clnt.c --- a/net/sunrpc/pmap_clnt.c +++ b/net/sunrpc/pmap_clnt.c @@ -208,6 +208,7 @@ pmap_create(char *hostname, struct socka if (IS_ERR(xprt)) return (struct rpc_clnt *)xprt; xprt->addr.sin_port = htons(RPC_PMAP_PORT); + xprt->resvport = 0; /* printk("pmap: create clnt\n"); */ clnt = rpc_new_client(xprt, hostname, diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c --- a/net/sunrpc/rpc_pipe.c +++ b/net/sunrpc/rpc_pipe.c @@ -414,38 +414,6 @@ rpc_put_mount(void) simple_release_fs(&rpc_mount, &rpc_mount_count); } -static int -rpc_lookup_parent(char *path, struct nameidata *nd) -{ - if (path[0] == '\0') - return -ENOENT; - if (rpc_get_mount()) { - printk(KERN_WARNING "%s: %s failed to mount " - "pseudofilesystem \n", __FILE__, __FUNCTION__); - return -ENODEV; - } - nd->mnt = mntget(rpc_mount); - nd->dentry = dget(rpc_mount->mnt_root); - nd->last_type = LAST_ROOT; - nd->flags = LOOKUP_PARENT; - nd->depth = 0; - - if (path_walk(path, nd)) { - printk(KERN_WARNING "%s: %s failed to find path %s\n", - __FILE__, __FUNCTION__, path); - rpc_put_mount(); - return -ENOENT; - } - return 0; -} - -static void -rpc_release_path(struct nameidata *nd) -{ - path_release(nd); - rpc_put_mount(); -} - static struct inode * rpc_get_inode(struct super_block *sb, int mode) { @@ -550,197 +518,149 @@ out_bad: return -ENOMEM; } -static int -__rpc_mkdir(struct inode *dir, struct dentry *dentry) +struct dentry * +rpc_mkdir(struct dentry *parent, char *name, struct rpc_clnt *rpc_client) { + struct inode *dir; + struct dentry *dentry; struct inode *inode; - - inode = rpc_get_inode(dir->i_sb, S_IFDIR | S_IRUSR | S_IXUSR); - if (!inode) - goto out_err; - inode->i_ino = iunique(dir->i_sb, 100); - d_instantiate(dentry, inode); - dir->i_nlink++; - inode_dir_notify(dir, DN_CREATE); - rpc_get_mount(); - return 0; -out_err: - printk(KERN_WARNING "%s: %s failed to allocate inode for dentry %s\n", - __FILE__, __FUNCTION__, dentry->d_name.name); - return -ENOMEM; -} - -static int -__rpc_rmdir(struct inode *dir, struct dentry *dentry) -{ int error; - shrink_dcache_parent(dentry); - if (dentry->d_inode) { - rpc_close_pipes(dentry->d_inode); - rpc_inode_setowner(dentry->d_inode, NULL); - } - if ((error = simple_rmdir(dir, dentry)) != 0) - return error; - if (!error) { - inode_dir_notify(dir, DN_DELETE); - d_drop(dentry); - rpc_put_mount(); - } - return 0; -} - -static struct dentry * -rpc_lookup_negative(char *path, struct nameidata *nd) -{ - struct dentry *dentry; - struct inode *dir; - int error; + if (!parent) + parent = rpc_mount->mnt_root; - if ((error = rpc_lookup_parent(path, nd)) != 0) + dir = parent->d_inode; + + error = rpc_get_mount(); + if (error) return ERR_PTR(error); - dir = nd->dentry->d_inode; + down(&dir->i_sem); - dentry = lookup_hash(&nd->last, nd->dentry); + dentry = lookup_one_len(name, parent, strlen(name)); if (IS_ERR(dentry)) - goto out_err; + goto out_unlock; if (dentry->d_inode) { - dput(dentry); dentry = ERR_PTR(-EEXIST); - goto out_err; + goto out_dput; } - return dentry; -out_err: - up(&dir->i_sem); - rpc_release_path(nd); - return dentry; -} + inode = rpc_get_inode(dir->i_sb, S_IFDIR | S_IRUSR | S_IXUSR); + if (!inode) + goto out_dput; + inode->i_ino = iunique(dir->i_sb, 100); + dir->i_nlink++; + RPC_I(dentry->d_inode)->private = rpc_client; -struct dentry * -rpc_mkdir(char *path, struct rpc_clnt *rpc_client) -{ - struct nameidata nd; - struct dentry *dentry; - struct inode *dir; - int error; + d_instantiate(dentry, inode); + dget(dentry); + up(&dir->i_sem); + + inode_dir_notify(dir, DN_CREATE); - dentry = rpc_lookup_negative(path, &nd); - if (IS_ERR(dentry)) - return dentry; - dir = nd.dentry->d_inode; - if ((error = __rpc_mkdir(dir, dentry)) != 0) - goto err_dput; - RPC_I(dentry->d_inode)->private = rpc_client; error = rpc_populate(dentry, authfiles, RPCAUTH_info, RPCAUTH_EOF); if (error) - goto err_depopulate; -out: - up(&dir->i_sem); - rpc_release_path(&nd); + goto out_depopulate; + return dentry; -err_depopulate: - rpc_depopulate(dentry); - __rpc_rmdir(dir, dentry); -err_dput: + + out_depopulate: + rpc_rmdir(dentry); + out_dput: dput(dentry); - printk(KERN_WARNING "%s: %s() failed to create directory %s (errno = %d)\n", - __FILE__, __FUNCTION__, path, error); - dentry = ERR_PTR(error); - goto out; + out_unlock: + up(&dir->i_sem); + rpc_put_mount(); + return dentry; } -int -rpc_rmdir(char *path) +void +rpc_rmdir(struct dentry *dentry) { - struct nameidata nd; - struct dentry *dentry; - struct inode *dir; - int error; + struct dentry *parent = dentry->d_parent; - if ((error = rpc_lookup_parent(path, &nd)) != 0) - return error; - dir = nd.dentry->d_inode; - down(&dir->i_sem); - dentry = lookup_hash(&nd.last, nd.dentry); - if (IS_ERR(dentry)) { - error = PTR_ERR(dentry); - goto out_release; - } rpc_depopulate(dentry); - error = __rpc_rmdir(dir, dentry); - dput(dentry); -out_release: - up(&dir->i_sem); - rpc_release_path(&nd); - return error; + + down(&parent->d_inode->i_sem); + if (dentry->d_inode) { + rpc_close_pipes(dentry->d_inode); + rpc_inode_setowner(dentry->d_inode, NULL); + simple_rmdir(parent->d_inode, dentry); + } + up(&parent->d_inode->i_sem); + + inode_dir_notify(parent->d_inode, DN_DELETE); + rpc_put_mount(); } struct dentry * -rpc_mkpipe(char *path, void *private, struct rpc_pipe_ops *ops, int flags) +rpc_mkpipe(struct dentry *parent, char *name, void *private, + struct rpc_pipe_ops *ops, int flags) { - struct nameidata nd; + struct inode *dir = parent->d_inode; struct dentry *dentry; - struct inode *dir, *inode; + struct inode *inode; struct rpc_inode *rpci; + int error; - dentry = rpc_lookup_negative(path, &nd); + error = rpc_get_mount(); + if (error) + return ERR_PTR(error); + + down(&parent->d_inode->i_sem); + dentry = lookup_one_len(name, parent, strlen(name)); if (IS_ERR(dentry)) - return dentry; - dir = nd.dentry->d_inode; - inode = rpc_get_inode(dir->i_sb, S_IFSOCK | S_IRUSR | S_IWUSR); - if (!inode) - goto err_dput; + goto out_unlock; + if (dentry->d_inode) { + dentry = ERR_PTR(-EEXIST); + goto out_dput; + } + + inode = rpc_get_inode(parent->d_inode->i_sb, + S_IFSOCK | S_IRUSR | S_IWUSR); + if (!inode) { + dentry = ERR_PTR(-ENOMEM); + goto out_dput; + } + inode->i_ino = iunique(dir->i_sb, 100); inode->i_fop = &rpc_pipe_fops; - d_instantiate(dentry, inode); + rpci = RPC_I(inode); rpci->private = private; rpci->flags = flags; rpci->ops = ops; + + d_instantiate(dentry, inode); + dget(dentry); + up(&parent->d_inode->i_sem); + inode_dir_notify(dir, DN_CREATE); -out: - up(&dir->i_sem); - rpc_release_path(&nd); return dentry; -err_dput: + + out_dput: dput(dentry); - dentry = ERR_PTR(-ENOMEM); - printk(KERN_WARNING "%s: %s() failed to create pipe %s (errno = %d)\n", - __FILE__, __FUNCTION__, path, -ENOMEM); - goto out; + out_unlock: + up(&parent->d_inode->i_sem); + rpc_put_mount(); + return dentry; } -int -rpc_unlink(char *path) +void +rpc_unlink(struct dentry *dentry) { - struct nameidata nd; - struct dentry *dentry; - struct inode *dir; - int error; + struct dentry *parent = dentry->d_parent; - if ((error = rpc_lookup_parent(path, &nd)) != 0) - return error; - dir = nd.dentry->d_inode; - down(&dir->i_sem); - dentry = lookup_hash(&nd.last, nd.dentry); - if (IS_ERR(dentry)) { - error = PTR_ERR(dentry); - goto out_release; - } - d_drop(dentry); + down(&parent->d_inode->i_sem); if (dentry->d_inode) { rpc_close_pipes(dentry->d_inode); rpc_inode_setowner(dentry->d_inode, NULL); - error = simple_unlink(dir, dentry); + simple_unlink(parent->d_inode, dentry); } - dput(dentry); - inode_dir_notify(dir, DN_DELETE); -out_release: - up(&dir->i_sem); - rpc_release_path(&nd); - return error; + up(&parent->d_inode->i_sem); + + inode_dir_notify(parent->d_inode, DN_DELETE); + rpc_put_mount(); } /* diff --git a/net/sunrpc/socklib.c b/net/sunrpc/socklib.c new file mode 100644 --- /dev/null +++ b/net/sunrpc/socklib.c @@ -0,0 +1,175 @@ +/* + * linux/net/sunrpc/socklib.c + * + * Common socket helper routines for RPC client and server + * + * Copyright (C) 1995, 1996 Olaf Kirch + */ + +#include +#include +#include +#include + + +/** + * skb_read_bits - copy some data bits from skb to internal buffer + * @desc: sk_buff copy helper + * @to: copy destination + * @len: number of bytes to copy + * + * Possibly called several times to iterate over an sk_buff and copy + * data out of it. + */ +static size_t skb_read_bits(skb_reader_t *desc, void *to, size_t len) +{ + if (len > desc->count) + len = desc->count; + if (skb_copy_bits(desc->skb, desc->offset, to, len)) + return 0; + desc->count -= len; + desc->offset += len; + return len; +} + +/** + * skb_read_and_csum_bits - copy and checksum from skb to buffer + * @desc: sk_buff copy helper + * @to: copy destination + * @len: number of bytes to copy + * + * Same as skb_read_bits, but calculate a checksum at the same time. + */ +static size_t skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len) +{ + unsigned int csum2, pos; + + if (len > desc->count) + len = desc->count; + pos = desc->offset; + csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0); + desc->csum = csum_block_add(desc->csum, csum2, pos); + desc->count -= len; + desc->offset += len; + return len; +} + +/** + * xdr_partial_copy_from_skb - copy data out of an skb + * @xdr: target XDR buffer + * @base: starting offset + * @desc: sk_buff copy helper + * @copy_actor: virtual method for copying data + * + */ +ssize_t xdr_partial_copy_from_skb(struct xdr_buf *xdr, unsigned int base, skb_reader_t *desc, skb_read_actor_t copy_actor) +{ + struct page **ppage = xdr->pages; + unsigned int len, pglen = xdr->page_len; + ssize_t copied = 0; + int ret; + + len = xdr->head[0].iov_len; + if (base < len) { + len -= base; + ret = copy_actor(desc, (char *)xdr->head[0].iov_base + base, len); + copied += ret; + if (ret != len || !desc->count) + goto out; + base = 0; + } else + base -= len; + + if (unlikely(pglen == 0)) + goto copy_tail; + if (unlikely(base >= pglen)) { + base -= pglen; + goto copy_tail; + } + if (base || xdr->page_base) { + pglen -= base; + base += xdr->page_base; + ppage += base >> PAGE_CACHE_SHIFT; + base &= ~PAGE_CACHE_MASK; + } + do { + char *kaddr; + + /* ACL likes to be lazy in allocating pages - ACLs + * are small by default but can get huge. */ + if (unlikely(*ppage == NULL)) { + *ppage = alloc_page(GFP_ATOMIC); + if (unlikely(*ppage == NULL)) { + if (copied == 0) + copied = -ENOMEM; + goto out; + } + } + + len = PAGE_CACHE_SIZE; + kaddr = kmap_atomic(*ppage, KM_SKB_SUNRPC_DATA); + if (base) { + len -= base; + if (pglen < len) + len = pglen; + ret = copy_actor(desc, kaddr + base, len); + base = 0; + } else { + if (pglen < len) + len = pglen; + ret = copy_actor(desc, kaddr, len); + } + flush_dcache_page(*ppage); + kunmap_atomic(kaddr, KM_SKB_SUNRPC_DATA); + copied += ret; + if (ret != len || !desc->count) + goto out; + ppage++; + } while ((pglen -= len) != 0); +copy_tail: + len = xdr->tail[0].iov_len; + if (base < len) + copied += copy_actor(desc, (char *)xdr->tail[0].iov_base + base, len - base); +out: + return copied; +} + +/** + * csum_partial_copy_to_xdr - checksum and copy data + * @xdr: target XDR buffer + * @skb: source skb + * + * We have set things up such that we perform the checksum of the UDP + * packet in parallel with the copies into the RPC client iovec. -DaveM + */ +int csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb) +{ + skb_reader_t desc; + + desc.skb = skb; + desc.offset = sizeof(struct udphdr); + desc.count = skb->len - desc.offset; + + if (skb->ip_summed == CHECKSUM_UNNECESSARY) + goto no_checksum; + + desc.csum = csum_partial(skb->data, desc.offset, skb->csum); + if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0) + return -1; + if (desc.offset != skb->len) { + unsigned int csum2; + csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0); + desc.csum = csum_block_add(desc.csum, csum2, desc.offset); + } + if (desc.count) + return -1; + if ((unsigned short)csum_fold(desc.csum)) + return -1; + return 0; +no_checksum: + if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0) + return -1; + if (desc.count) + return -1; + return 0; +} diff --git a/net/sunrpc/sunrpc_syms.c b/net/sunrpc/sunrpc_syms.c --- a/net/sunrpc/sunrpc_syms.c +++ b/net/sunrpc/sunrpc_syms.c @@ -10,7 +10,6 @@ #include #include -#include #include #include #include diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c --- a/net/sunrpc/svcsock.c +++ b/net/sunrpc/svcsock.c @@ -549,9 +549,6 @@ svc_write_space(struct sock *sk) /* * Receive a datagram from a UDP socket. */ -extern int -csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb); - static int svc_udp_recvfrom(struct svc_rqst *rqstp) { diff --git a/net/sunrpc/sysctl.c b/net/sunrpc/sysctl.c --- a/net/sunrpc/sysctl.c +++ b/net/sunrpc/sysctl.c @@ -119,8 +119,18 @@ done: return 0; } +unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; +unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE; +unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; +EXPORT_SYMBOL(xprt_min_resvport); +unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; +EXPORT_SYMBOL(xprt_max_resvport); + + static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; +static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; +static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; static ctl_table debug_table[] = { { @@ -177,6 +187,28 @@ static ctl_table debug_table[] = { .extra1 = &min_slot_table_size, .extra2 = &max_slot_table_size }, + { + .ctl_name = CTL_MIN_RESVPORT, + .procname = "min_resvport", + .data = &xprt_min_resvport, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .strategy = &sysctl_intvec, + .extra1 = &xprt_min_resvport_limit, + .extra2 = &xprt_max_resvport_limit + }, + { + .ctl_name = CTL_MAX_RESVPORT, + .procname = "max_resvport", + .data = &xprt_max_resvport, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .strategy = &sysctl_intvec, + .extra1 = &xprt_min_resvport_limit, + .extra2 = &xprt_max_resvport_limit + }, { .ctl_name = 0 } }; diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c --- a/net/sunrpc/xdr.c +++ b/net/sunrpc/xdr.c @@ -6,15 +6,12 @@ * Copyright (C) 1995, 1996 Olaf Kirch */ +#include #include -#include #include #include #include #include -#include -#include -#include #include #include @@ -176,178 +173,6 @@ xdr_inline_pages(struct xdr_buf *xdr, un xdr->buflen += len; } -ssize_t -xdr_partial_copy_from_skb(struct xdr_buf *xdr, unsigned int base, - skb_reader_t *desc, - skb_read_actor_t copy_actor) -{ - struct page **ppage = xdr->pages; - unsigned int len, pglen = xdr->page_len; - ssize_t copied = 0; - int ret; - - len = xdr->head[0].iov_len; - if (base < len) { - len -= base; - ret = copy_actor(desc, (char *)xdr->head[0].iov_base + base, len); - copied += ret; - if (ret != len || !desc->count) - goto out; - base = 0; - } else - base -= len; - - if (pglen == 0) - goto copy_tail; - if (base >= pglen) { - base -= pglen; - goto copy_tail; - } - if (base || xdr->page_base) { - pglen -= base; - base += xdr->page_base; - ppage += base >> PAGE_CACHE_SHIFT; - base &= ~PAGE_CACHE_MASK; - } - do { - char *kaddr; - - /* ACL likes to be lazy in allocating pages - ACLs - * are small by default but can get huge. */ - if (unlikely(*ppage == NULL)) { - *ppage = alloc_page(GFP_ATOMIC); - if (unlikely(*ppage == NULL)) { - if (copied == 0) - copied = -ENOMEM; - goto out; - } - } - - len = PAGE_CACHE_SIZE; - kaddr = kmap_atomic(*ppage, KM_SKB_SUNRPC_DATA); - if (base) { - len -= base; - if (pglen < len) - len = pglen; - ret = copy_actor(desc, kaddr + base, len); - base = 0; - } else { - if (pglen < len) - len = pglen; - ret = copy_actor(desc, kaddr, len); - } - flush_dcache_page(*ppage); - kunmap_atomic(kaddr, KM_SKB_SUNRPC_DATA); - copied += ret; - if (ret != len || !desc->count) - goto out; - ppage++; - } while ((pglen -= len) != 0); -copy_tail: - len = xdr->tail[0].iov_len; - if (base < len) - copied += copy_actor(desc, (char *)xdr->tail[0].iov_base + base, len - base); -out: - return copied; -} - - -int -xdr_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, - struct xdr_buf *xdr, unsigned int base, int msgflags) -{ - struct page **ppage = xdr->pages; - unsigned int len, pglen = xdr->page_len; - int err, ret = 0; - ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int); - - len = xdr->head[0].iov_len; - if (base < len || (addr != NULL && base == 0)) { - struct kvec iov = { - .iov_base = xdr->head[0].iov_base + base, - .iov_len = len - base, - }; - struct msghdr msg = { - .msg_name = addr, - .msg_namelen = addrlen, - .msg_flags = msgflags, - }; - if (xdr->len > len) - msg.msg_flags |= MSG_MORE; - - if (iov.iov_len != 0) - err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); - else - err = kernel_sendmsg(sock, &msg, NULL, 0, 0); - if (ret == 0) - ret = err; - else if (err > 0) - ret += err; - if (err != iov.iov_len) - goto out; - base = 0; - } else - base -= len; - - if (pglen == 0) - goto copy_tail; - if (base >= pglen) { - base -= pglen; - goto copy_tail; - } - if (base || xdr->page_base) { - pglen -= base; - base += xdr->page_base; - ppage += base >> PAGE_CACHE_SHIFT; - base &= ~PAGE_CACHE_MASK; - } - - sendpage = sock->ops->sendpage ? : sock_no_sendpage; - do { - int flags = msgflags; - - len = PAGE_CACHE_SIZE; - if (base) - len -= base; - if (pglen < len) - len = pglen; - - if (pglen != len || xdr->tail[0].iov_len != 0) - flags |= MSG_MORE; - - /* Hmm... We might be dealing with highmem pages */ - if (PageHighMem(*ppage)) - sendpage = sock_no_sendpage; - err = sendpage(sock, *ppage, base, len, flags); - if (ret == 0) - ret = err; - else if (err > 0) - ret += err; - if (err != len) - goto out; - base = 0; - ppage++; - } while ((pglen -= len) != 0); -copy_tail: - len = xdr->tail[0].iov_len; - if (base < len) { - struct kvec iov = { - .iov_base = xdr->tail[0].iov_base + base, - .iov_len = len - base, - }; - struct msghdr msg = { - .msg_flags = msgflags, - }; - err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); - if (ret == 0) - ret = err; - else if (err > 0) - ret += err; - } -out: - return ret; -} - /* * Helper routines for doing 'memmove' like operations on a struct xdr_buf diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c --- a/net/sunrpc/xprt.c +++ b/net/sunrpc/xprt.c @@ -10,12 +10,12 @@ * one is available. Otherwise, it sleeps on the backlog queue * (xprt_reserve). * - Next, the caller puts together the RPC message, stuffs it into - * the request struct, and calls xprt_call(). - * - xprt_call transmits the message and installs the caller on the - * socket's wait list. At the same time, it installs a timer that + * the request struct, and calls xprt_transmit(). + * - xprt_transmit sends the message and installs the caller on the + * transport's wait list. At the same time, it installs a timer that * is run after the packet's timeout has expired. * - When a packet arrives, the data_ready handler walks the list of - * pending requests for that socket. If a matching XID is found, the + * pending requests for that transport. If a matching XID is found, the * caller is woken up, and the timer removed. * - When no reply arrives within the timeout interval, the timer is * fired by the kernel and runs xprt_timer(). It either adjusts the @@ -33,36 +33,17 @@ * * Copyright (C) 1995-1997, Olaf Kirch * - * TCP callback races fixes (C) 1998 Red Hat Software - * TCP send fixes (C) 1998 Red Hat Software - * TCP NFS related read + write fixes - * (C) 1999 Dave Airlie, University of Limerick, Ireland - * - * Rewrite of larges part of the code in order to stabilize TCP stuff. - * Fix behaviour when socket buffer is full. - * (C) 1999 Trond Myklebust + * Transport switch API copyright (C) 2005, Chuck Lever */ +#include + #include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include +#include #include #include -#include -#include -#include -#include +#include /* * Local variables @@ -73,81 +54,90 @@ # define RPCDBG_FACILITY RPCDBG_XPRT #endif -#define XPRT_MAX_BACKOFF (8) -#define XPRT_IDLE_TIMEOUT (5*60*HZ) -#define XPRT_MAX_RESVPORT (800) - /* * Local functions */ static void xprt_request_init(struct rpc_task *, struct rpc_xprt *); static inline void do_xprt_reserve(struct rpc_task *); -static void xprt_disconnect(struct rpc_xprt *); static void xprt_connect_status(struct rpc_task *task); -static struct rpc_xprt * xprt_setup(int proto, struct sockaddr_in *ap, - struct rpc_timeout *to); -static struct socket *xprt_create_socket(struct rpc_xprt *, int, int); -static void xprt_bind_socket(struct rpc_xprt *, struct socket *); static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *); -static int xprt_clear_backlog(struct rpc_xprt *xprt); - -#ifdef RPC_DEBUG_DATA /* - * Print the buffer contents (first 128 bytes only--just enough for - * diropres return). + * The transport code maintains an estimate on the maximum number of out- + * standing RPC requests, using a smoothed version of the congestion + * avoidance implemented in 44BSD. This is basically the Van Jacobson + * congestion algorithm: If a retransmit occurs, the congestion window is + * halved; otherwise, it is incremented by 1/cwnd when + * + * - a reply is received and + * - a full number of requests are outstanding and + * - the congestion window hasn't been updated recently. + */ +#define RPC_CWNDSHIFT (8U) +#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT) +#define RPC_INITCWND RPC_CWNDSCALE +#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT) + +#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd) + +/** + * xprt_reserve_xprt - serialize write access to transports + * @task: task that is requesting access to the transport + * + * This prevents mixing the payload of separate requests, and prevents + * transport connects from colliding with writes. No congestion control + * is provided. */ -static void -xprt_pktdump(char *msg, u32 *packet, unsigned int count) +int xprt_reserve_xprt(struct rpc_task *task) { - u8 *buf = (u8 *) packet; - int j; + struct rpc_xprt *xprt = task->tk_xprt; + struct rpc_rqst *req = task->tk_rqstp; - dprintk("RPC: %s\n", msg); - for (j = 0; j < count && j < 128; j += 4) { - if (!(j & 31)) { - if (j) - dprintk("\n"); - dprintk("0x%04x ", j); - } - dprintk("%02x%02x%02x%02x ", - buf[j], buf[j+1], buf[j+2], buf[j+3]); + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) { + if (task == xprt->snd_task) + return 1; + if (task == NULL) + return 0; + goto out_sleep; } - dprintk("\n"); -} -#else -static inline void -xprt_pktdump(char *msg, u32 *packet, unsigned int count) -{ - /* NOP */ -} -#endif + xprt->snd_task = task; + if (req) { + req->rq_bytes_sent = 0; + req->rq_ntrans++; + } + return 1; -/* - * Look up RPC transport given an INET socket - */ -static inline struct rpc_xprt * -xprt_from_sock(struct sock *sk) -{ - return (struct rpc_xprt *) sk->sk_user_data; +out_sleep: + dprintk("RPC: %4d failed to lock transport %p\n", + task->tk_pid, xprt); + task->tk_timeout = 0; + task->tk_status = -EAGAIN; + if (req && req->rq_ntrans) + rpc_sleep_on(&xprt->resend, task, NULL, NULL); + else + rpc_sleep_on(&xprt->sending, task, NULL, NULL); + return 0; } /* - * Serialize write access to sockets, in order to prevent different - * requests from interfering with each other. - * Also prevents TCP socket connects from colliding with writes. + * xprt_reserve_xprt_cong - serialize write access to transports + * @task: task that is requesting access to the transport + * + * Same as xprt_reserve_xprt, but Van Jacobson congestion control is + * integrated into the decision of whether a request is allowed to be + * woken up and given access to the transport. */ -static int -__xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task) +int xprt_reserve_xprt_cong(struct rpc_task *task) { + struct rpc_xprt *xprt = task->tk_xprt; struct rpc_rqst *req = task->tk_rqstp; - if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) { + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) { if (task == xprt->snd_task) return 1; goto out_sleep; } - if (xprt->nocong || __xprt_get_cong(xprt, task)) { + if (__xprt_get_cong(xprt, task)) { xprt->snd_task = task; if (req) { req->rq_bytes_sent = 0; @@ -156,10 +146,10 @@ __xprt_lock_write(struct rpc_xprt *xprt, return 1; } smp_mb__before_clear_bit(); - clear_bit(XPRT_LOCKED, &xprt->sockstate); + clear_bit(XPRT_LOCKED, &xprt->state); smp_mb__after_clear_bit(); out_sleep: - dprintk("RPC: %4d failed to lock socket %p\n", task->tk_pid, xprt); + dprintk("RPC: %4d failed to lock transport %p\n", task->tk_pid, xprt); task->tk_timeout = 0; task->tk_status = -EAGAIN; if (req && req->rq_ntrans) @@ -169,26 +159,52 @@ out_sleep: return 0; } -static inline int -xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task) +static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task) { int retval; - spin_lock_bh(&xprt->sock_lock); - retval = __xprt_lock_write(xprt, task); - spin_unlock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); + retval = xprt->ops->reserve_xprt(task); + spin_unlock_bh(&xprt->transport_lock); return retval; } +static void __xprt_lock_write_next(struct rpc_xprt *xprt) +{ + struct rpc_task *task; + struct rpc_rqst *req; -static void -__xprt_lock_write_next(struct rpc_xprt *xprt) + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) + return; + + task = rpc_wake_up_next(&xprt->resend); + if (!task) { + task = rpc_wake_up_next(&xprt->sending); + if (!task) + goto out_unlock; + } + + req = task->tk_rqstp; + xprt->snd_task = task; + if (req) { + req->rq_bytes_sent = 0; + req->rq_ntrans++; + } + return; + +out_unlock: + smp_mb__before_clear_bit(); + clear_bit(XPRT_LOCKED, &xprt->state); + smp_mb__after_clear_bit(); +} + +static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt) { struct rpc_task *task; - if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) return; - if (!xprt->nocong && RPCXPRT_CONGESTED(xprt)) + if (RPCXPRT_CONGESTED(xprt)) goto out_unlock; task = rpc_wake_up_next(&xprt->resend); if (!task) { @@ -196,7 +212,7 @@ __xprt_lock_write_next(struct rpc_xprt * if (!task) goto out_unlock; } - if (xprt->nocong || __xprt_get_cong(xprt, task)) { + if (__xprt_get_cong(xprt, task)) { struct rpc_rqst *req = task->tk_rqstp; xprt->snd_task = task; if (req) { @@ -207,87 +223,52 @@ __xprt_lock_write_next(struct rpc_xprt * } out_unlock: smp_mb__before_clear_bit(); - clear_bit(XPRT_LOCKED, &xprt->sockstate); + clear_bit(XPRT_LOCKED, &xprt->state); smp_mb__after_clear_bit(); } -/* - * Releases the socket for use by other requests. +/** + * xprt_release_xprt - allow other requests to use a transport + * @xprt: transport with other tasks potentially waiting + * @task: task that is releasing access to the transport + * + * Note that "task" can be NULL. No congestion control is provided. */ -static void -__xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task) +void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task) { if (xprt->snd_task == task) { xprt->snd_task = NULL; smp_mb__before_clear_bit(); - clear_bit(XPRT_LOCKED, &xprt->sockstate); + clear_bit(XPRT_LOCKED, &xprt->state); smp_mb__after_clear_bit(); __xprt_lock_write_next(xprt); } } -static inline void -xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task) -{ - spin_lock_bh(&xprt->sock_lock); - __xprt_release_write(xprt, task); - spin_unlock_bh(&xprt->sock_lock); -} - -/* - * Write data to socket. +/** + * xprt_release_xprt_cong - allow other requests to use a transport + * @xprt: transport with other tasks potentially waiting + * @task: task that is releasing access to the transport + * + * Note that "task" can be NULL. Another task is awoken to use the + * transport if the transport's congestion window allows it. */ -static inline int -xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req) +void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task) { - struct socket *sock = xprt->sock; - struct xdr_buf *xdr = &req->rq_snd_buf; - struct sockaddr *addr = NULL; - int addrlen = 0; - unsigned int skip; - int result; - - if (!sock) - return -ENOTCONN; - - xprt_pktdump("packet data:", - req->rq_svec->iov_base, - req->rq_svec->iov_len); - - /* For UDP, we need to provide an address */ - if (!xprt->stream) { - addr = (struct sockaddr *) &xprt->addr; - addrlen = sizeof(xprt->addr); + if (xprt->snd_task == task) { + xprt->snd_task = NULL; + smp_mb__before_clear_bit(); + clear_bit(XPRT_LOCKED, &xprt->state); + smp_mb__after_clear_bit(); + __xprt_lock_write_next_cong(xprt); } - /* Dont repeat bytes */ - skip = req->rq_bytes_sent; - - clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); - result = xdr_sendpages(sock, addr, addrlen, xdr, skip, MSG_DONTWAIT); - - dprintk("RPC: xprt_sendmsg(%d) = %d\n", xdr->len - skip, result); - - if (result >= 0) - return result; +} - switch (result) { - case -ECONNREFUSED: - /* When the server has died, an ICMP port unreachable message - * prompts ECONNREFUSED. - */ - case -EAGAIN: - break; - case -ECONNRESET: - case -ENOTCONN: - case -EPIPE: - /* connection broken */ - if (xprt->stream) - result = -ENOTCONN; - break; - default: - printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result); - } - return result; +static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task) +{ + spin_lock_bh(&xprt->transport_lock); + xprt->ops->release_xprt(xprt, task); + spin_unlock_bh(&xprt->transport_lock); } /* @@ -321,26 +302,40 @@ __xprt_put_cong(struct rpc_xprt *xprt, s return; req->rq_cong = 0; xprt->cong -= RPC_CWNDSCALE; - __xprt_lock_write_next(xprt); + __xprt_lock_write_next_cong(xprt); } -/* - * Adjust RPC congestion window +/** + * xprt_release_rqst_cong - housekeeping when request is complete + * @task: RPC request that recently completed + * + * Useful for transports that require congestion control. + */ +void xprt_release_rqst_cong(struct rpc_task *task) +{ + __xprt_put_cong(task->tk_xprt, task->tk_rqstp); +} + +/** + * xprt_adjust_cwnd - adjust transport congestion window + * @task: recently completed RPC request used to adjust window + * @result: result code of completed RPC request + * * We use a time-smoothed congestion estimator to avoid heavy oscillation. */ -static void -xprt_adjust_cwnd(struct rpc_xprt *xprt, int result) +void xprt_adjust_cwnd(struct rpc_task *task, int result) { - unsigned long cwnd; + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = task->tk_xprt; + unsigned long cwnd = xprt->cwnd; - cwnd = xprt->cwnd; if (result >= 0 && cwnd <= xprt->cong) { /* The (cwnd >> 1) term makes sure * the result gets rounded properly. */ cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd; if (cwnd > RPC_MAXCWND(xprt)) cwnd = RPC_MAXCWND(xprt); - __xprt_lock_write_next(xprt); + __xprt_lock_write_next_cong(xprt); } else if (result == -ETIMEDOUT) { cwnd >>= 1; if (cwnd < RPC_CWNDSCALE) @@ -349,11 +344,89 @@ xprt_adjust_cwnd(struct rpc_xprt *xprt, dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n", xprt->cong, xprt->cwnd, cwnd); xprt->cwnd = cwnd; + __xprt_put_cong(xprt, req); +} + +/** + * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue + * @xprt: transport with waiting tasks + * @status: result code to plant in each task before waking it + * + */ +void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status) +{ + if (status < 0) + rpc_wake_up_status(&xprt->pending, status); + else + rpc_wake_up(&xprt->pending); +} + +/** + * xprt_wait_for_buffer_space - wait for transport output buffer to clear + * @task: task to be put to sleep + * + */ +void xprt_wait_for_buffer_space(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + + task->tk_timeout = req->rq_timeout; + rpc_sleep_on(&xprt->pending, task, NULL, NULL); +} + +/** + * xprt_write_space - wake the task waiting for transport output buffer space + * @xprt: transport with waiting tasks + * + * Can be called in a soft IRQ context, so xprt_write_space never sleeps. + */ +void xprt_write_space(struct rpc_xprt *xprt) +{ + if (unlikely(xprt->shutdown)) + return; + + spin_lock_bh(&xprt->transport_lock); + if (xprt->snd_task) { + dprintk("RPC: write space: waking waiting task on xprt %p\n", + xprt); + rpc_wake_up_task(xprt->snd_task); + } + spin_unlock_bh(&xprt->transport_lock); +} + +/** + * xprt_set_retrans_timeout_def - set a request's retransmit timeout + * @task: task whose timeout is to be set + * + * Set a request's retransmit timeout based on the transport's + * default timeout parameters. Used by transports that don't adjust + * the retransmit timeout based on round-trip time estimation. + */ +void xprt_set_retrans_timeout_def(struct rpc_task *task) +{ + task->tk_timeout = task->tk_rqstp->rq_timeout; } /* - * Reset the major timeout value + * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout + * @task: task whose timeout is to be set + * + * Set a request's retransmit timeout using the RTT estimator. */ +void xprt_set_retrans_timeout_rtt(struct rpc_task *task) +{ + int timer = task->tk_msg.rpc_proc->p_timer; + struct rpc_rtt *rtt = task->tk_client->cl_rtt; + struct rpc_rqst *req = task->tk_rqstp; + unsigned long max_timeout = req->rq_xprt->timeout.to_maxval; + + task->tk_timeout = rpc_calc_rto(rtt, timer); + task->tk_timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries; + if (task->tk_timeout > max_timeout || task->tk_timeout == 0) + task->tk_timeout = max_timeout; +} + static void xprt_reset_majortimeo(struct rpc_rqst *req) { struct rpc_timeout *to = &req->rq_xprt->timeout; @@ -368,8 +441,10 @@ static void xprt_reset_majortimeo(struct req->rq_majortimeo += jiffies; } -/* - * Adjust timeout values etc for next retransmit +/** + * xprt_adjust_timeout - adjust timeout values for next retransmit + * @req: RPC request containing parameters to use for the adjustment + * */ int xprt_adjust_timeout(struct rpc_rqst *req) { @@ -391,9 +466,9 @@ int xprt_adjust_timeout(struct rpc_rqst req->rq_retries = 0; xprt_reset_majortimeo(req); /* Reset the RTT counters == "slow start" */ - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval); - spin_unlock_bh(&xprt->sock_lock); + spin_unlock_bh(&xprt->transport_lock); pprintk("RPC: %lu timeout\n", jiffies); status = -ETIMEDOUT; } @@ -405,133 +480,52 @@ int xprt_adjust_timeout(struct rpc_rqst return status; } -/* - * Close down a transport socket - */ -static void -xprt_close(struct rpc_xprt *xprt) -{ - struct socket *sock = xprt->sock; - struct sock *sk = xprt->inet; - - if (!sk) - return; - - write_lock_bh(&sk->sk_callback_lock); - xprt->inet = NULL; - xprt->sock = NULL; - - sk->sk_user_data = NULL; - sk->sk_data_ready = xprt->old_data_ready; - sk->sk_state_change = xprt->old_state_change; - sk->sk_write_space = xprt->old_write_space; - write_unlock_bh(&sk->sk_callback_lock); - - sk->sk_no_check = 0; - - sock_release(sock); -} - -static void -xprt_socket_autoclose(void *args) +static void xprt_autoclose(void *args) { struct rpc_xprt *xprt = (struct rpc_xprt *)args; xprt_disconnect(xprt); - xprt_close(xprt); + xprt->ops->close(xprt); xprt_release_write(xprt, NULL); } -/* - * Mark a transport as disconnected +/** + * xprt_disconnect - mark a transport as disconnected + * @xprt: transport to flag for disconnect + * */ -static void -xprt_disconnect(struct rpc_xprt *xprt) +void xprt_disconnect(struct rpc_xprt *xprt) { dprintk("RPC: disconnected transport %p\n", xprt); - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); xprt_clear_connected(xprt); - rpc_wake_up_status(&xprt->pending, -ENOTCONN); - spin_unlock_bh(&xprt->sock_lock); + xprt_wake_pending_tasks(xprt, -ENOTCONN); + spin_unlock_bh(&xprt->transport_lock); } -/* - * Used to allow disconnection when we've been idle - */ static void xprt_init_autodisconnect(unsigned long data) { struct rpc_xprt *xprt = (struct rpc_xprt *)data; - spin_lock(&xprt->sock_lock); + spin_lock(&xprt->transport_lock); if (!list_empty(&xprt->recv) || xprt->shutdown) goto out_abort; - if (test_and_set_bit(XPRT_LOCKED, &xprt->sockstate)) + if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) goto out_abort; - spin_unlock(&xprt->sock_lock); - /* Let keventd close the socket */ - if (test_bit(XPRT_CONNECTING, &xprt->sockstate) != 0) + spin_unlock(&xprt->transport_lock); + if (xprt_connecting(xprt)) xprt_release_write(xprt, NULL); else schedule_work(&xprt->task_cleanup); return; out_abort: - spin_unlock(&xprt->sock_lock); -} - -static void xprt_socket_connect(void *args) -{ - struct rpc_xprt *xprt = (struct rpc_xprt *)args; - struct socket *sock = xprt->sock; - int status = -EIO; - - if (xprt->shutdown || xprt->addr.sin_port == 0) - goto out; - - /* - * Start by resetting any existing state - */ - xprt_close(xprt); - sock = xprt_create_socket(xprt, xprt->prot, xprt->resvport); - if (sock == NULL) { - /* couldn't create socket or bind to reserved port; - * this is likely a permanent error, so cause an abort */ - goto out; - } - xprt_bind_socket(xprt, sock); - xprt_sock_setbufsize(xprt); - - status = 0; - if (!xprt->stream) - goto out; - - /* - * Tell the socket layer to start connecting... - */ - status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr, - sizeof(xprt->addr), O_NONBLOCK); - dprintk("RPC: %p connect status %d connected %d sock state %d\n", - xprt, -status, xprt_connected(xprt), sock->sk->sk_state); - if (status < 0) { - switch (status) { - case -EINPROGRESS: - case -EALREADY: - goto out_clear; - } - } -out: - if (status < 0) - rpc_wake_up_status(&xprt->pending, status); - else - rpc_wake_up(&xprt->pending); -out_clear: - smp_mb__before_clear_bit(); - clear_bit(XPRT_CONNECTING, &xprt->sockstate); - smp_mb__after_clear_bit(); + spin_unlock(&xprt->transport_lock); } -/* - * Attempt to connect a TCP socket. +/** + * xprt_connect - schedule a transport connect operation + * @task: RPC task that is requesting the connect * */ void xprt_connect(struct rpc_task *task) @@ -552,37 +546,19 @@ void xprt_connect(struct rpc_task *task) if (!xprt_lock_write(xprt, task)) return; if (xprt_connected(xprt)) - goto out_write; - - if (task->tk_rqstp) - task->tk_rqstp->rq_bytes_sent = 0; - - task->tk_timeout = RPC_CONNECT_TIMEOUT; - rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL); - if (!test_and_set_bit(XPRT_CONNECTING, &xprt->sockstate)) { - /* Note: if we are here due to a dropped connection - * we delay reconnecting by RPC_REESTABLISH_TIMEOUT/HZ - * seconds - */ - if (xprt->sock != NULL) - schedule_delayed_work(&xprt->sock_connect, - RPC_REESTABLISH_TIMEOUT); - else { - schedule_work(&xprt->sock_connect); - if (!RPC_IS_ASYNC(task)) - flush_scheduled_work(); - } + xprt_release_write(xprt, task); + else { + if (task->tk_rqstp) + task->tk_rqstp->rq_bytes_sent = 0; + + task->tk_timeout = xprt->connect_timeout; + rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL); + xprt->ops->connect(task); } return; - out_write: - xprt_release_write(xprt, task); } -/* - * We arrive here when awoken from waiting on connection establishment. - */ -static void -xprt_connect_status(struct rpc_task *task) +static void xprt_connect_status(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; @@ -592,31 +568,42 @@ xprt_connect_status(struct rpc_task *tas return; } - /* if soft mounted, just cause this RPC to fail */ - if (RPC_IS_SOFT(task)) - task->tk_status = -EIO; - switch (task->tk_status) { case -ECONNREFUSED: case -ECONNRESET: + dprintk("RPC: %4d xprt_connect_status: server %s refused connection\n", + task->tk_pid, task->tk_client->cl_server); + break; case -ENOTCONN: - return; + dprintk("RPC: %4d xprt_connect_status: connection broken\n", + task->tk_pid); + break; case -ETIMEDOUT: - dprintk("RPC: %4d xprt_connect_status: timed out\n", + dprintk("RPC: %4d xprt_connect_status: connect attempt timed out\n", task->tk_pid); break; default: - printk(KERN_ERR "RPC: error %d connecting to server %s\n", - -task->tk_status, task->tk_client->cl_server); + dprintk("RPC: %4d xprt_connect_status: error %d connecting to server %s\n", + task->tk_pid, -task->tk_status, task->tk_client->cl_server); + xprt_release_write(xprt, task); + task->tk_status = -EIO; + return; + } + + /* if soft mounted, just cause this RPC to fail */ + if (RPC_IS_SOFT(task)) { + xprt_release_write(xprt, task); + task->tk_status = -EIO; } - xprt_release_write(xprt, task); } -/* - * Look up the RPC request corresponding to a reply, and then lock it. +/** + * xprt_lookup_rqst - find an RPC request corresponding to an XID + * @xprt: transport on which the original request was transmitted + * @xid: RPC XID of incoming reply + * */ -static inline struct rpc_rqst * -xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid) +struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid) { struct list_head *pos; struct rpc_rqst *req = NULL; @@ -631,556 +618,68 @@ xprt_lookup_rqst(struct rpc_xprt *xprt, return req; } -/* - * Complete reply received. - * The TCP code relies on us to remove the request from xprt->pending. - */ -static void -xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied) -{ - struct rpc_task *task = req->rq_task; - struct rpc_clnt *clnt = task->tk_client; - - /* Adjust congestion window */ - if (!xprt->nocong) { - unsigned timer = task->tk_msg.rpc_proc->p_timer; - xprt_adjust_cwnd(xprt, copied); - __xprt_put_cong(xprt, req); - if (timer) { - if (req->rq_ntrans == 1) - rpc_update_rtt(clnt->cl_rtt, timer, - (long)jiffies - req->rq_xtime); - rpc_set_timeo(clnt->cl_rtt, timer, req->rq_ntrans - 1); - } - } - -#ifdef RPC_PROFILE - /* Profile only reads for now */ - if (copied > 1024) { - static unsigned long nextstat; - static unsigned long pkt_rtt, pkt_len, pkt_cnt; - - pkt_cnt++; - pkt_len += req->rq_slen + copied; - pkt_rtt += jiffies - req->rq_xtime; - if (time_before(nextstat, jiffies)) { - printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd); - printk("RPC: %ld %ld %ld %ld stat\n", - jiffies, pkt_cnt, pkt_len, pkt_rtt); - pkt_rtt = pkt_len = pkt_cnt = 0; - nextstat = jiffies + 5 * HZ; - } - } -#endif - - dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied); - list_del_init(&req->rq_list); - req->rq_received = req->rq_private_buf.len = copied; - - /* ... and wake up the process. */ - rpc_wake_up_task(task); - return; -} - -static size_t -skb_read_bits(skb_reader_t *desc, void *to, size_t len) -{ - if (len > desc->count) - len = desc->count; - if (skb_copy_bits(desc->skb, desc->offset, to, len)) - return 0; - desc->count -= len; - desc->offset += len; - return len; -} - -static size_t -skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len) -{ - unsigned int csum2, pos; - - if (len > desc->count) - len = desc->count; - pos = desc->offset; - csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0); - desc->csum = csum_block_add(desc->csum, csum2, pos); - desc->count -= len; - desc->offset += len; - return len; -} - -/* - * We have set things up such that we perform the checksum of the UDP - * packet in parallel with the copies into the RPC client iovec. -DaveM - */ -int -csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb) -{ - skb_reader_t desc; - - desc.skb = skb; - desc.offset = sizeof(struct udphdr); - desc.count = skb->len - desc.offset; - - if (skb->ip_summed == CHECKSUM_UNNECESSARY) - goto no_checksum; - - desc.csum = csum_partial(skb->data, desc.offset, skb->csum); - if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits) < 0) - return -1; - if (desc.offset != skb->len) { - unsigned int csum2; - csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0); - desc.csum = csum_block_add(desc.csum, csum2, desc.offset); - } - if (desc.count) - return -1; - if ((unsigned short)csum_fold(desc.csum)) - return -1; - return 0; -no_checksum: - if (xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits) < 0) - return -1; - if (desc.count) - return -1; - return 0; -} - -/* - * Input handler for RPC replies. Called from a bottom half and hence - * atomic. - */ -static void -udp_data_ready(struct sock *sk, int len) -{ - struct rpc_task *task; - struct rpc_xprt *xprt; - struct rpc_rqst *rovr; - struct sk_buff *skb; - int err, repsize, copied; - u32 _xid, *xp; - - read_lock(&sk->sk_callback_lock); - dprintk("RPC: udp_data_ready...\n"); - if (!(xprt = xprt_from_sock(sk))) { - printk("RPC: udp_data_ready request not found!\n"); - goto out; - } - - dprintk("RPC: udp_data_ready client %p\n", xprt); - - if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) - goto out; - - if (xprt->shutdown) - goto dropit; - - repsize = skb->len - sizeof(struct udphdr); - if (repsize < 4) { - printk("RPC: impossible RPC reply size %d!\n", repsize); - goto dropit; - } - - /* Copy the XID from the skb... */ - xp = skb_header_pointer(skb, sizeof(struct udphdr), - sizeof(_xid), &_xid); - if (xp == NULL) - goto dropit; - - /* Look up and lock the request corresponding to the given XID */ - spin_lock(&xprt->sock_lock); - rovr = xprt_lookup_rqst(xprt, *xp); - if (!rovr) - goto out_unlock; - task = rovr->rq_task; - - dprintk("RPC: %4d received reply\n", task->tk_pid); - - if ((copied = rovr->rq_private_buf.buflen) > repsize) - copied = repsize; - - /* Suck it into the iovec, verify checksum if not done by hw. */ - if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) - goto out_unlock; - - /* Something worked... */ - dst_confirm(skb->dst); - - xprt_complete_rqst(xprt, rovr, copied); - - out_unlock: - spin_unlock(&xprt->sock_lock); - dropit: - skb_free_datagram(sk, skb); - out: - read_unlock(&sk->sk_callback_lock); -} - -/* - * Copy from an skb into memory and shrink the skb. - */ -static inline size_t -tcp_copy_data(skb_reader_t *desc, void *p, size_t len) -{ - if (len > desc->count) - len = desc->count; - if (skb_copy_bits(desc->skb, desc->offset, p, len)) { - dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n", - len, desc->count); - return 0; - } - desc->offset += len; - desc->count -= len; - dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n", - len, desc->count); - return len; -} - -/* - * TCP read fragment marker - */ -static inline void -tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - size_t len, used; - char *p; - - p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset; - len = sizeof(xprt->tcp_recm) - xprt->tcp_offset; - used = tcp_copy_data(desc, p, len); - xprt->tcp_offset += used; - if (used != len) - return; - xprt->tcp_reclen = ntohl(xprt->tcp_recm); - if (xprt->tcp_reclen & 0x80000000) - xprt->tcp_flags |= XPRT_LAST_FRAG; - else - xprt->tcp_flags &= ~XPRT_LAST_FRAG; - xprt->tcp_reclen &= 0x7fffffff; - xprt->tcp_flags &= ~XPRT_COPY_RECM; - xprt->tcp_offset = 0; - /* Sanity check of the record length */ - if (xprt->tcp_reclen < 4) { - printk(KERN_ERR "RPC: Invalid TCP record fragment length\n"); - xprt_disconnect(xprt); - } - dprintk("RPC: reading TCP record fragment of length %d\n", - xprt->tcp_reclen); -} - -static void -tcp_check_recm(struct rpc_xprt *xprt) -{ - dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n", - xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags); - if (xprt->tcp_offset == xprt->tcp_reclen) { - xprt->tcp_flags |= XPRT_COPY_RECM; - xprt->tcp_offset = 0; - if (xprt->tcp_flags & XPRT_LAST_FRAG) { - xprt->tcp_flags &= ~XPRT_COPY_DATA; - xprt->tcp_flags |= XPRT_COPY_XID; - xprt->tcp_copied = 0; - } - } -} - -/* - * TCP read xid - */ -static inline void -tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - size_t len, used; - char *p; - - len = sizeof(xprt->tcp_xid) - xprt->tcp_offset; - dprintk("RPC: reading XID (%Zu bytes)\n", len); - p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset; - used = tcp_copy_data(desc, p, len); - xprt->tcp_offset += used; - if (used != len) - return; - xprt->tcp_flags &= ~XPRT_COPY_XID; - xprt->tcp_flags |= XPRT_COPY_DATA; - xprt->tcp_copied = 4; - dprintk("RPC: reading reply for XID %08x\n", - ntohl(xprt->tcp_xid)); - tcp_check_recm(xprt); -} - -/* - * TCP read and complete request - */ -static inline void -tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - struct rpc_rqst *req; - struct xdr_buf *rcvbuf; - size_t len; - ssize_t r; - - /* Find and lock the request corresponding to this xid */ - spin_lock(&xprt->sock_lock); - req = xprt_lookup_rqst(xprt, xprt->tcp_xid); - if (!req) { - xprt->tcp_flags &= ~XPRT_COPY_DATA; - dprintk("RPC: XID %08x request not found!\n", - ntohl(xprt->tcp_xid)); - spin_unlock(&xprt->sock_lock); - return; - } - - rcvbuf = &req->rq_private_buf; - len = desc->count; - if (len > xprt->tcp_reclen - xprt->tcp_offset) { - skb_reader_t my_desc; - - len = xprt->tcp_reclen - xprt->tcp_offset; - memcpy(&my_desc, desc, sizeof(my_desc)); - my_desc.count = len; - r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, - &my_desc, tcp_copy_data); - desc->count -= r; - desc->offset += r; - } else - r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, - desc, tcp_copy_data); - - if (r > 0) { - xprt->tcp_copied += r; - xprt->tcp_offset += r; - } - if (r != len) { - /* Error when copying to the receive buffer, - * usually because we weren't able to allocate - * additional buffer pages. All we can do now - * is turn off XPRT_COPY_DATA, so the request - * will not receive any additional updates, - * and time out. - * Any remaining data from this record will - * be discarded. - */ - xprt->tcp_flags &= ~XPRT_COPY_DATA; - dprintk("RPC: XID %08x truncated request\n", - ntohl(xprt->tcp_xid)); - dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", - xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); - goto out; - } - - dprintk("RPC: XID %08x read %Zd bytes\n", - ntohl(xprt->tcp_xid), r); - dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", - xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); - - if (xprt->tcp_copied == req->rq_private_buf.buflen) - xprt->tcp_flags &= ~XPRT_COPY_DATA; - else if (xprt->tcp_offset == xprt->tcp_reclen) { - if (xprt->tcp_flags & XPRT_LAST_FRAG) - xprt->tcp_flags &= ~XPRT_COPY_DATA; - } - -out: - if (!(xprt->tcp_flags & XPRT_COPY_DATA)) { - dprintk("RPC: %4d received reply complete\n", - req->rq_task->tk_pid); - xprt_complete_rqst(xprt, req, xprt->tcp_copied); - } - spin_unlock(&xprt->sock_lock); - tcp_check_recm(xprt); -} - -/* - * TCP discard extra bytes from a short read - */ -static inline void -tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc) -{ - size_t len; - - len = xprt->tcp_reclen - xprt->tcp_offset; - if (len > desc->count) - len = desc->count; - desc->count -= len; - desc->offset += len; - xprt->tcp_offset += len; - dprintk("RPC: discarded %Zu bytes\n", len); - tcp_check_recm(xprt); -} - -/* - * TCP record receive routine - * We first have to grab the record marker, then the XID, then the data. +/** + * xprt_update_rtt - update an RPC client's RTT state after receiving a reply + * @task: RPC request that recently completed + * */ -static int -tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, - unsigned int offset, size_t len) +void xprt_update_rtt(struct rpc_task *task) { - struct rpc_xprt *xprt = rd_desc->arg.data; - skb_reader_t desc = { - .skb = skb, - .offset = offset, - .count = len, - .csum = 0 - }; - - dprintk("RPC: tcp_data_recv\n"); - do { - /* Read in a new fragment marker if necessary */ - /* Can we ever really expect to get completely empty fragments? */ - if (xprt->tcp_flags & XPRT_COPY_RECM) { - tcp_read_fraghdr(xprt, &desc); - continue; - } - /* Read in the xid if necessary */ - if (xprt->tcp_flags & XPRT_COPY_XID) { - tcp_read_xid(xprt, &desc); - continue; - } - /* Read in the request data */ - if (xprt->tcp_flags & XPRT_COPY_DATA) { - tcp_read_request(xprt, &desc); - continue; - } - /* Skip over any trailing bytes on short reads */ - tcp_read_discard(xprt, &desc); - } while (desc.count); - dprintk("RPC: tcp_data_recv done\n"); - return len - desc.count; -} - -static void tcp_data_ready(struct sock *sk, int bytes) -{ - struct rpc_xprt *xprt; - read_descriptor_t rd_desc; - - read_lock(&sk->sk_callback_lock); - dprintk("RPC: tcp_data_ready...\n"); - if (!(xprt = xprt_from_sock(sk))) { - printk("RPC: tcp_data_ready socket info not found!\n"); - goto out; - } - if (xprt->shutdown) - goto out; - - /* We use rd_desc to pass struct xprt to tcp_data_recv */ - rd_desc.arg.data = xprt; - rd_desc.count = 65536; - tcp_read_sock(sk, &rd_desc, tcp_data_recv); -out: - read_unlock(&sk->sk_callback_lock); -} - -static void -tcp_state_change(struct sock *sk) -{ - struct rpc_xprt *xprt; + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_rtt *rtt = task->tk_client->cl_rtt; + unsigned timer = task->tk_msg.rpc_proc->p_timer; - read_lock(&sk->sk_callback_lock); - if (!(xprt = xprt_from_sock(sk))) - goto out; - dprintk("RPC: tcp_state_change client %p...\n", xprt); - dprintk("RPC: state %x conn %d dead %d zapped %d\n", - sk->sk_state, xprt_connected(xprt), - sock_flag(sk, SOCK_DEAD), - sock_flag(sk, SOCK_ZAPPED)); - - switch (sk->sk_state) { - case TCP_ESTABLISHED: - spin_lock_bh(&xprt->sock_lock); - if (!xprt_test_and_set_connected(xprt)) { - /* Reset TCP record info */ - xprt->tcp_offset = 0; - xprt->tcp_reclen = 0; - xprt->tcp_copied = 0; - xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID; - rpc_wake_up(&xprt->pending); - } - spin_unlock_bh(&xprt->sock_lock); - break; - case TCP_SYN_SENT: - case TCP_SYN_RECV: - break; - default: - xprt_disconnect(xprt); - break; + if (timer) { + if (req->rq_ntrans == 1) + rpc_update_rtt(rtt, timer, + (long)jiffies - req->rq_xtime); + rpc_set_timeo(rtt, timer, req->rq_ntrans - 1); } - out: - read_unlock(&sk->sk_callback_lock); } -/* - * Called when more output buffer space is available for this socket. - * We try not to wake our writers until they can make "significant" - * progress, otherwise we'll waste resources thrashing sock_sendmsg - * with a bunch of small requests. +/** + * xprt_complete_rqst - called when reply processing is complete + * @task: RPC request that recently completed + * @copied: actual number of bytes received from the transport + * + * Caller holds transport lock. */ -static void -xprt_write_space(struct sock *sk) +void xprt_complete_rqst(struct rpc_task *task, int copied) { - struct rpc_xprt *xprt; - struct socket *sock; - - read_lock(&sk->sk_callback_lock); - if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->sk_socket)) - goto out; - if (xprt->shutdown) - goto out; - - /* Wait until we have enough socket memory */ - if (xprt->stream) { - /* from net/core/stream.c:sk_stream_write_space */ - if (sk_stream_wspace(sk) < sk_stream_min_wspace(sk)) - goto out; - } else { - /* from net/core/sock.c:sock_def_write_space */ - if (!sock_writeable(sk)) - goto out; - } + struct rpc_rqst *req = task->tk_rqstp; - if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags)) - goto out; + dprintk("RPC: %5u xid %08x complete (%d bytes received)\n", + task->tk_pid, ntohl(req->rq_xid), copied); - spin_lock_bh(&xprt->sock_lock); - if (xprt->snd_task) - rpc_wake_up_task(xprt->snd_task); - spin_unlock_bh(&xprt->sock_lock); -out: - read_unlock(&sk->sk_callback_lock); + list_del_init(&req->rq_list); + req->rq_received = req->rq_private_buf.len = copied; + rpc_wake_up_task(task); } -/* - * RPC receive timeout handler. - */ -static void -xprt_timer(struct rpc_task *task) +static void xprt_timer(struct rpc_task *task) { - struct rpc_rqst *req = task->tk_rqstp; + struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = req->rq_xprt; - spin_lock(&xprt->sock_lock); - if (req->rq_received) - goto out; + dprintk("RPC: %4d xprt_timer\n", task->tk_pid); - xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT); - __xprt_put_cong(xprt, req); - - dprintk("RPC: %4d xprt_timer (%s request)\n", - task->tk_pid, req ? "pending" : "backlogged"); - - task->tk_status = -ETIMEDOUT; -out: + spin_lock(&xprt->transport_lock); + if (!req->rq_received) { + if (xprt->ops->timer) + xprt->ops->timer(task); + task->tk_status = -ETIMEDOUT; + } task->tk_timeout = 0; rpc_wake_up_task(task); - spin_unlock(&xprt->sock_lock); + spin_unlock(&xprt->transport_lock); } -/* - * Place the actual RPC call. - * We have to copy the iovec because sendmsg fiddles with its contents. +/** + * xprt_prepare_transmit - reserve the transport before sending a request + * @task: RPC task about to send a request + * */ -int -xprt_prepare_transmit(struct rpc_task *task) +int xprt_prepare_transmit(struct rpc_task *task) { struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = req->rq_xprt; @@ -1191,12 +690,12 @@ xprt_prepare_transmit(struct rpc_task *t if (xprt->shutdown) return -EIO; - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); if (req->rq_received && !req->rq_bytes_sent) { err = req->rq_received; goto out_unlock; } - if (!__xprt_lock_write(xprt, task)) { + if (!xprt->ops->reserve_xprt(task)) { err = -EAGAIN; goto out_unlock; } @@ -1206,39 +705,34 @@ xprt_prepare_transmit(struct rpc_task *t goto out_unlock; } out_unlock: - spin_unlock_bh(&xprt->sock_lock); + spin_unlock_bh(&xprt->transport_lock); return err; } -void -xprt_transmit(struct rpc_task *task) +/** + * xprt_transmit - send an RPC request on a transport + * @task: controlling RPC task + * + * We have to copy the iovec because sendmsg fiddles with its contents. + */ +void xprt_transmit(struct rpc_task *task) { - struct rpc_clnt *clnt = task->tk_client; struct rpc_rqst *req = task->tk_rqstp; struct rpc_xprt *xprt = req->rq_xprt; - int status, retry = 0; - + int status; dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen); - /* set up everything as needed. */ - /* Write the record marker */ - if (xprt->stream) { - u32 *marker = req->rq_svec[0].iov_base; - - *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker))); - } - smp_rmb(); if (!req->rq_received) { if (list_empty(&req->rq_list)) { - spin_lock_bh(&xprt->sock_lock); + spin_lock_bh(&xprt->transport_lock); /* Update the softirq receive buffer */ memcpy(&req->rq_private_buf, &req->rq_rcv_buf, sizeof(req->rq_private_buf)); /* Add request to the receive list */ list_add_tail(&req->rq_list, &xprt->recv); - spin_unlock_bh(&xprt->sock_lock); + spin_unlock_bh(&xprt->transport_lock); xprt_reset_majortimeo(req); /* Turn off autodisconnect */ del_singleshot_timer_sync(&xprt->timer); @@ -1246,40 +740,19 @@ xprt_transmit(struct rpc_task *task) } else if (!req->rq_bytes_sent) return; - /* Continue transmitting the packet/record. We must be careful - * to cope with writespace callbacks arriving _after_ we have - * called xprt_sendmsg(). - */ - while (1) { - req->rq_xtime = jiffies; - status = xprt_sendmsg(xprt, req); - - if (status < 0) - break; - - if (xprt->stream) { - req->rq_bytes_sent += status; - - /* If we've sent the entire packet, immediately - * reset the count of bytes sent. */ - if (req->rq_bytes_sent >= req->rq_slen) { - req->rq_bytes_sent = 0; - goto out_receive; - } - } else { - if (status >= req->rq_slen) - goto out_receive; - status = -EAGAIN; - break; - } - - dprintk("RPC: %4d xmit incomplete (%d left of %d)\n", - task->tk_pid, req->rq_slen - req->rq_bytes_sent, - req->rq_slen); - - status = -EAGAIN; - if (retry++ > 50) - break; + status = xprt->ops->send_request(task); + if (status == 0) { + dprintk("RPC: %4d xmit complete\n", task->tk_pid); + spin_lock_bh(&xprt->transport_lock); + xprt->ops->set_retrans_timeout(task); + /* Don't race with disconnect */ + if (!xprt_connected(xprt)) + task->tk_status = -ENOTCONN; + else if (!req->rq_received) + rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer); + xprt->ops->release_xprt(xprt, task); + spin_unlock_bh(&xprt->transport_lock); + return; } /* Note: at this point, task->tk_sleeping has not yet been set, @@ -1289,60 +762,19 @@ xprt_transmit(struct rpc_task *task) task->tk_status = status; switch (status) { - case -EAGAIN: - if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) { - /* Protect against races with xprt_write_space */ - spin_lock_bh(&xprt->sock_lock); - /* Don't race with disconnect */ - if (!xprt_connected(xprt)) - task->tk_status = -ENOTCONN; - else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) { - task->tk_timeout = req->rq_timeout; - rpc_sleep_on(&xprt->pending, task, NULL, NULL); - } - spin_unlock_bh(&xprt->sock_lock); - return; - } - /* Keep holding the socket if it is blocked */ - rpc_delay(task, HZ>>4); - return; case -ECONNREFUSED: - task->tk_timeout = RPC_REESTABLISH_TIMEOUT; rpc_sleep_on(&xprt->sending, task, NULL, NULL); + case -EAGAIN: case -ENOTCONN: return; default: - if (xprt->stream) - xprt_disconnect(xprt); + break; } xprt_release_write(xprt, task); return; - out_receive: - dprintk("RPC: %4d xmit complete\n", task->tk_pid); - /* Set the task's receive timeout value */ - spin_lock_bh(&xprt->sock_lock); - if (!xprt->nocong) { - int timer = task->tk_msg.rpc_proc->p_timer; - task->tk_timeout = rpc_calc_rto(clnt->cl_rtt, timer); - task->tk_timeout <<= rpc_ntimeo(clnt->cl_rtt, timer) + req->rq_retries; - if (task->tk_timeout > xprt->timeout.to_maxval || task->tk_timeout == 0) - task->tk_timeout = xprt->timeout.to_maxval; - } else - task->tk_timeout = req->rq_timeout; - /* Don't race with disconnect */ - if (!xprt_connected(xprt)) - task->tk_status = -ENOTCONN; - else if (!req->rq_received) - rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer); - __xprt_release_write(xprt, task); - spin_unlock_bh(&xprt->sock_lock); } -/* - * Reserve an RPC call slot. - */ -static inline void -do_xprt_reserve(struct rpc_task *task) +static inline void do_xprt_reserve(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; @@ -1362,22 +794,25 @@ do_xprt_reserve(struct rpc_task *task) rpc_sleep_on(&xprt->backlog, task, NULL, NULL); } -void -xprt_reserve(struct rpc_task *task) +/** + * xprt_reserve - allocate an RPC request slot + * @task: RPC task requesting a slot allocation + * + * If no more slots are available, place the task on the transport's + * backlog queue. + */ +void xprt_reserve(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; task->tk_status = -EIO; if (!xprt->shutdown) { - spin_lock(&xprt->xprt_lock); + spin_lock(&xprt->reserve_lock); do_xprt_reserve(task); - spin_unlock(&xprt->xprt_lock); + spin_unlock(&xprt->reserve_lock); } } -/* - * Allocate a 'unique' XID - */ static inline u32 xprt_alloc_xid(struct rpc_xprt *xprt) { return xprt->xid++; @@ -1388,11 +823,7 @@ static inline void xprt_init_xid(struct get_random_bytes(&xprt->xid, sizeof(xprt->xid)); } -/* - * Initialize RPC request - */ -static void -xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt) +static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt) { struct rpc_rqst *req = task->tk_rqstp; @@ -1404,124 +835,97 @@ xprt_request_init(struct rpc_task *task, req, ntohl(req->rq_xid)); } -/* - * Release an RPC call slot +/** + * xprt_release - release an RPC request slot + * @task: task which is finished with the slot + * */ -void -xprt_release(struct rpc_task *task) +void xprt_release(struct rpc_task *task) { struct rpc_xprt *xprt = task->tk_xprt; struct rpc_rqst *req; if (!(req = task->tk_rqstp)) return; - spin_lock_bh(&xprt->sock_lock); - __xprt_release_write(xprt, task); - __xprt_put_cong(xprt, req); + spin_lock_bh(&xprt->transport_lock); + xprt->ops->release_xprt(xprt, task); + if (xprt->ops->release_request) + xprt->ops->release_request(task); if (!list_empty(&req->rq_list)) list_del(&req->rq_list); xprt->last_used = jiffies; if (list_empty(&xprt->recv) && !xprt->shutdown) - mod_timer(&xprt->timer, xprt->last_used + XPRT_IDLE_TIMEOUT); - spin_unlock_bh(&xprt->sock_lock); + mod_timer(&xprt->timer, + xprt->last_used + xprt->idle_timeout); + spin_unlock_bh(&xprt->transport_lock); task->tk_rqstp = NULL; memset(req, 0, sizeof(*req)); /* mark unused */ dprintk("RPC: %4d release request %p\n", task->tk_pid, req); - spin_lock(&xprt->xprt_lock); + spin_lock(&xprt->reserve_lock); list_add(&req->rq_list, &xprt->free); - xprt_clear_backlog(xprt); - spin_unlock(&xprt->xprt_lock); -} - -/* - * Set default timeout parameters - */ -static void -xprt_default_timeout(struct rpc_timeout *to, int proto) -{ - if (proto == IPPROTO_UDP) - xprt_set_timeout(to, 5, 5 * HZ); - else - xprt_set_timeout(to, 5, 60 * HZ); + rpc_wake_up_next(&xprt->backlog); + spin_unlock(&xprt->reserve_lock); } -/* - * Set constant timeout +/** + * xprt_set_timeout - set constant RPC timeout + * @to: RPC timeout parameters to set up + * @retr: number of retries + * @incr: amount of increase after each retry + * */ -void -xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr) +void xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr) { to->to_initval = to->to_increment = incr; - to->to_maxval = incr * retr; + to->to_maxval = to->to_initval + (incr * retr); to->to_retries = retr; to->to_exponential = 0; } -unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; -unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE; - -/* - * Initialize an RPC client - */ -static struct rpc_xprt * -xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to) +static struct rpc_xprt *xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to) { + int result; struct rpc_xprt *xprt; - unsigned int entries; - size_t slot_table_size; struct rpc_rqst *req; - dprintk("RPC: setting up %s transport...\n", - proto == IPPROTO_UDP? "UDP" : "TCP"); - - entries = (proto == IPPROTO_TCP)? - xprt_tcp_slot_table_entries : xprt_udp_slot_table_entries; - if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL) return ERR_PTR(-ENOMEM); memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */ - xprt->max_reqs = entries; - slot_table_size = entries * sizeof(xprt->slot[0]); - xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); - if (xprt->slot == NULL) { - kfree(xprt); - return ERR_PTR(-ENOMEM); - } - memset(xprt->slot, 0, slot_table_size); xprt->addr = *ap; - xprt->prot = proto; - xprt->stream = (proto == IPPROTO_TCP)? 1 : 0; - if (xprt->stream) { - xprt->cwnd = RPC_MAXCWND(xprt); - xprt->nocong = 1; - xprt->max_payload = (1U << 31) - 1; - } else { - xprt->cwnd = RPC_INITCWND; - xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); + + switch (proto) { + case IPPROTO_UDP: + result = xs_setup_udp(xprt, to); + break; + case IPPROTO_TCP: + result = xs_setup_tcp(xprt, to); + break; + default: + printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n", + proto); + result = -EIO; + break; } - spin_lock_init(&xprt->sock_lock); - spin_lock_init(&xprt->xprt_lock); - init_waitqueue_head(&xprt->cong_wait); + if (result) { + kfree(xprt); + return ERR_PTR(result); + } + + spin_lock_init(&xprt->transport_lock); + spin_lock_init(&xprt->reserve_lock); INIT_LIST_HEAD(&xprt->free); INIT_LIST_HEAD(&xprt->recv); - INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt); - INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt); + INIT_WORK(&xprt->task_cleanup, xprt_autoclose, xprt); init_timer(&xprt->timer); xprt->timer.function = xprt_init_autodisconnect; xprt->timer.data = (unsigned long) xprt; xprt->last_used = jiffies; - xprt->port = XPRT_MAX_RESVPORT; - - /* Set timeout parameters */ - if (to) { - xprt->timeout = *to; - } else - xprt_default_timeout(&xprt->timeout, xprt->prot); + xprt->cwnd = RPC_INITCWND; rpc_init_wait_queue(&xprt->pending, "xprt_pending"); rpc_init_wait_queue(&xprt->sending, "xprt_sending"); @@ -1529,139 +933,25 @@ xprt_setup(int proto, struct sockaddr_in rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog"); /* initialize free list */ - for (req = &xprt->slot[entries-1]; req >= &xprt->slot[0]; req--) + for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--) list_add(&req->rq_list, &xprt->free); xprt_init_xid(xprt); - /* Check whether we want to use a reserved port */ - xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; - dprintk("RPC: created transport %p with %u slots\n", xprt, xprt->max_reqs); return xprt; } -/* - * Bind to a reserved port - */ -static inline int xprt_bindresvport(struct rpc_xprt *xprt, struct socket *sock) -{ - struct sockaddr_in myaddr = { - .sin_family = AF_INET, - }; - int err, port; - - /* Were we already bound to a given port? Try to reuse it */ - port = xprt->port; - do { - myaddr.sin_port = htons(port); - err = sock->ops->bind(sock, (struct sockaddr *) &myaddr, - sizeof(myaddr)); - if (err == 0) { - xprt->port = port; - return 0; - } - if (--port == 0) - port = XPRT_MAX_RESVPORT; - } while (err == -EADDRINUSE && port != xprt->port); - - printk("RPC: Can't bind to reserved port (%d).\n", -err); - return err; -} - -static void -xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock) -{ - struct sock *sk = sock->sk; - - if (xprt->inet) - return; - - write_lock_bh(&sk->sk_callback_lock); - sk->sk_user_data = xprt; - xprt->old_data_ready = sk->sk_data_ready; - xprt->old_state_change = sk->sk_state_change; - xprt->old_write_space = sk->sk_write_space; - if (xprt->prot == IPPROTO_UDP) { - sk->sk_data_ready = udp_data_ready; - sk->sk_no_check = UDP_CSUM_NORCV; - xprt_set_connected(xprt); - } else { - tcp_sk(sk)->nonagle = 1; /* disable Nagle's algorithm */ - sk->sk_data_ready = tcp_data_ready; - sk->sk_state_change = tcp_state_change; - xprt_clear_connected(xprt); - } - sk->sk_write_space = xprt_write_space; - - /* Reset to new socket */ - xprt->sock = sock; - xprt->inet = sk; - write_unlock_bh(&sk->sk_callback_lock); - - return; -} - -/* - * Set socket buffer length - */ -void -xprt_sock_setbufsize(struct rpc_xprt *xprt) -{ - struct sock *sk = xprt->inet; - - if (xprt->stream) - return; - if (xprt->rcvsize) { - sk->sk_userlocks |= SOCK_RCVBUF_LOCK; - sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2; - } - if (xprt->sndsize) { - sk->sk_userlocks |= SOCK_SNDBUF_LOCK; - sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2; - sk->sk_write_space(sk); - } -} - -/* - * Datastream sockets are created here, but xprt_connect will create - * and connect stream sockets. - */ -static struct socket * xprt_create_socket(struct rpc_xprt *xprt, int proto, int resvport) -{ - struct socket *sock; - int type, err; - - dprintk("RPC: xprt_create_socket(%s %d)\n", - (proto == IPPROTO_UDP)? "udp" : "tcp", proto); - - type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM; - - if ((err = sock_create_kern(PF_INET, type, proto, &sock)) < 0) { - printk("RPC: can't create socket (%d).\n", -err); - return NULL; - } - - /* If the caller has the capability, bind to a reserved port */ - if (resvport && xprt_bindresvport(xprt, sock) < 0) { - printk("RPC: can't bind to reserved port.\n"); - goto failed; - } - - return sock; - -failed: - sock_release(sock); - return NULL; -} - -/* - * Create an RPC client transport given the protocol and peer address. +/** + * xprt_create_proto - create an RPC client transport + * @proto: requested transport protocol + * @sap: remote peer's address + * @to: timeout parameters for new transport + * */ -struct rpc_xprt * -xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to) +struct rpc_xprt *xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to) { struct rpc_xprt *xprt; @@ -1673,46 +963,26 @@ xprt_create_proto(int proto, struct sock return xprt; } -/* - * Prepare for transport shutdown. - */ -static void -xprt_shutdown(struct rpc_xprt *xprt) +static void xprt_shutdown(struct rpc_xprt *xprt) { xprt->shutdown = 1; rpc_wake_up(&xprt->sending); rpc_wake_up(&xprt->resend); - rpc_wake_up(&xprt->pending); + xprt_wake_pending_tasks(xprt, -EIO); rpc_wake_up(&xprt->backlog); - wake_up(&xprt->cong_wait); del_timer_sync(&xprt->timer); - - /* synchronously wait for connect worker to finish */ - cancel_delayed_work(&xprt->sock_connect); - flush_scheduled_work(); -} - -/* - * Clear the xprt backlog queue - */ -static int -xprt_clear_backlog(struct rpc_xprt *xprt) { - rpc_wake_up_next(&xprt->backlog); - wake_up(&xprt->cong_wait); - return 1; } -/* - * Destroy an RPC transport, killing off all requests. +/** + * xprt_destroy - destroy an RPC transport, killing off all requests. + * @xprt: transport to destroy + * */ -int -xprt_destroy(struct rpc_xprt *xprt) +int xprt_destroy(struct rpc_xprt *xprt) { dprintk("RPC: destroying transport %p\n", xprt); xprt_shutdown(xprt); - xprt_disconnect(xprt); - xprt_close(xprt); - kfree(xprt->slot); + xprt->ops->destroy(xprt); kfree(xprt); return 0; diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c new file mode 100644 --- /dev/null +++ b/net/sunrpc/xprtsock.c @@ -0,0 +1,1252 @@ +/* + * linux/net/sunrpc/xprtsock.c + * + * Client-side transport implementation for sockets. + * + * TCP callback races fixes (C) 1998 Red Hat Software + * TCP send fixes (C) 1998 Red Hat Software + * TCP NFS related read + write fixes + * (C) 1999 Dave Airlie, University of Limerick, Ireland + * + * Rewrite of larges part of the code in order to stabilize TCP stuff. + * Fix behaviour when socket buffer is full. + * (C) 1999 Trond Myklebust + * + * IP socket transport implementation, (C) 2005 Chuck Lever + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include + +/* + * How many times to try sending a request on a socket before waiting + * for the socket buffer to clear. + */ +#define XS_SENDMSG_RETRY (10U) + +/* + * Time out for an RPC UDP socket connect. UDP socket connects are + * synchronous, but we set a timeout anyway in case of resource + * exhaustion on the local host. + */ +#define XS_UDP_CONN_TO (5U * HZ) + +/* + * Wait duration for an RPC TCP connection to be established. Solaris + * NFS over TCP uses 60 seconds, for example, which is in line with how + * long a server takes to reboot. + */ +#define XS_TCP_CONN_TO (60U * HZ) + +/* + * Wait duration for a reply from the RPC portmapper. + */ +#define XS_BIND_TO (60U * HZ) + +/* + * Delay if a UDP socket connect error occurs. This is most likely some + * kind of resource problem on the local host. + */ +#define XS_UDP_REEST_TO (2U * HZ) + +/* + * The reestablish timeout allows clients to delay for a bit before attempting + * to reconnect to a server that just dropped our connection. + * + * We implement an exponential backoff when trying to reestablish a TCP + * transport connection with the server. Some servers like to drop a TCP + * connection when they are overworked, so we start with a short timeout and + * increase over time if the server is down or not responding. + */ +#define XS_TCP_INIT_REEST_TO (3U * HZ) +#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ) + +/* + * TCP idle timeout; client drops the transport socket if it is idle + * for this long. Note that we also timeout UDP sockets to prevent + * holding port numbers when there is no RPC traffic. + */ +#define XS_IDLE_DISC_TO (5U * 60 * HZ) + +#ifdef RPC_DEBUG +# undef RPC_DEBUG_DATA +# define RPCDBG_FACILITY RPCDBG_TRANS +#endif + +#ifdef RPC_DEBUG_DATA +static void xs_pktdump(char *msg, u32 *packet, unsigned int count) +{ + u8 *buf = (u8 *) packet; + int j; + + dprintk("RPC: %s\n", msg); + for (j = 0; j < count && j < 128; j += 4) { + if (!(j & 31)) { + if (j) + dprintk("\n"); + dprintk("0x%04x ", j); + } + dprintk("%02x%02x%02x%02x ", + buf[j], buf[j+1], buf[j+2], buf[j+3]); + } + dprintk("\n"); +} +#else +static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) +{ + /* NOP */ +} +#endif + +#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) + +static inline int xs_send_head(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, unsigned int len) +{ + struct kvec iov = { + .iov_base = xdr->head[0].iov_base + base, + .iov_len = len - base, + }; + struct msghdr msg = { + .msg_name = addr, + .msg_namelen = addrlen, + .msg_flags = XS_SENDMSG_FLAGS, + }; + + if (xdr->len > len) + msg.msg_flags |= MSG_MORE; + + if (likely(iov.iov_len)) + return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); + return kernel_sendmsg(sock, &msg, NULL, 0, 0); +} + +static int xs_send_tail(struct socket *sock, struct xdr_buf *xdr, unsigned int base, unsigned int len) +{ + struct kvec iov = { + .iov_base = xdr->tail[0].iov_base + base, + .iov_len = len - base, + }; + struct msghdr msg = { + .msg_flags = XS_SENDMSG_FLAGS, + }; + + return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); +} + +/** + * xs_sendpages - write pages directly to a socket + * @sock: socket to send on + * @addr: UDP only -- address of destination + * @addrlen: UDP only -- length of destination address + * @xdr: buffer containing this request + * @base: starting position in the buffer + * + */ +static inline int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base) +{ + struct page **ppage = xdr->pages; + unsigned int len, pglen = xdr->page_len; + int err, ret = 0; + ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int); + + if (unlikely(!sock)) + return -ENOTCONN; + + clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); + + len = xdr->head[0].iov_len; + if (base < len || (addr != NULL && base == 0)) { + err = xs_send_head(sock, addr, addrlen, xdr, base, len); + if (ret == 0) + ret = err; + else if (err > 0) + ret += err; + if (err != (len - base)) + goto out; + base = 0; + } else + base -= len; + + if (unlikely(pglen == 0)) + goto copy_tail; + if (unlikely(base >= pglen)) { + base -= pglen; + goto copy_tail; + } + if (base || xdr->page_base) { + pglen -= base; + base += xdr->page_base; + ppage += base >> PAGE_CACHE_SHIFT; + base &= ~PAGE_CACHE_MASK; + } + + sendpage = sock->ops->sendpage ? : sock_no_sendpage; + do { + int flags = XS_SENDMSG_FLAGS; + + len = PAGE_CACHE_SIZE; + if (base) + len -= base; + if (pglen < len) + len = pglen; + + if (pglen != len || xdr->tail[0].iov_len != 0) + flags |= MSG_MORE; + + /* Hmm... We might be dealing with highmem pages */ + if (PageHighMem(*ppage)) + sendpage = sock_no_sendpage; + err = sendpage(sock, *ppage, base, len, flags); + if (ret == 0) + ret = err; + else if (err > 0) + ret += err; + if (err != len) + goto out; + base = 0; + ppage++; + } while ((pglen -= len) != 0); +copy_tail: + len = xdr->tail[0].iov_len; + if (base < len) { + err = xs_send_tail(sock, xdr, base, len); + if (ret == 0) + ret = err; + else if (err > 0) + ret += err; + } +out: + return ret; +} + +/** + * xs_nospace - place task on wait queue if transmit was incomplete + * @task: task to put to sleep + * + */ +static void xs_nospace(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + + dprintk("RPC: %4d xmit incomplete (%u left of %u)\n", + task->tk_pid, req->rq_slen - req->rq_bytes_sent, + req->rq_slen); + + if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) { + /* Protect against races with write_space */ + spin_lock_bh(&xprt->transport_lock); + + /* Don't race with disconnect */ + if (!xprt_connected(xprt)) + task->tk_status = -ENOTCONN; + else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) + xprt_wait_for_buffer_space(task); + + spin_unlock_bh(&xprt->transport_lock); + } else + /* Keep holding the socket if it is blocked */ + rpc_delay(task, HZ>>4); +} + +/** + * xs_udp_send_request - write an RPC request to a UDP socket + * @task: address of RPC task that manages the state of an RPC request + * + * Return values: + * 0: The request has been sent + * EAGAIN: The socket was blocked, please call again later to + * complete the request + * ENOTCONN: Caller needs to invoke connect logic then call again + * other: Some other error occured, the request was not sent + */ +static int xs_udp_send_request(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + struct xdr_buf *xdr = &req->rq_snd_buf; + int status; + + xs_pktdump("packet data:", + req->rq_svec->iov_base, + req->rq_svec->iov_len); + + req->rq_xtime = jiffies; + status = xs_sendpages(xprt->sock, (struct sockaddr *) &xprt->addr, + sizeof(xprt->addr), xdr, req->rq_bytes_sent); + + dprintk("RPC: xs_udp_send_request(%u) = %d\n", + xdr->len - req->rq_bytes_sent, status); + + if (likely(status >= (int) req->rq_slen)) + return 0; + + /* Still some bytes left; set up for a retry later. */ + if (status > 0) + status = -EAGAIN; + + switch (status) { + case -ENETUNREACH: + case -EPIPE: + case -ECONNREFUSED: + /* When the server has died, an ICMP port unreachable message + * prompts ECONNREFUSED. */ + break; + case -EAGAIN: + xs_nospace(task); + break; + default: + dprintk("RPC: sendmsg returned unrecognized error %d\n", + -status); + break; + } + + return status; +} + +static inline void xs_encode_tcp_record_marker(struct xdr_buf *buf) +{ + u32 reclen = buf->len - sizeof(rpc_fraghdr); + rpc_fraghdr *base = buf->head[0].iov_base; + *base = htonl(RPC_LAST_STREAM_FRAGMENT | reclen); +} + +/** + * xs_tcp_send_request - write an RPC request to a TCP socket + * @task: address of RPC task that manages the state of an RPC request + * + * Return values: + * 0: The request has been sent + * EAGAIN: The socket was blocked, please call again later to + * complete the request + * ENOTCONN: Caller needs to invoke connect logic then call again + * other: Some other error occured, the request was not sent + * + * XXX: In the case of soft timeouts, should we eventually give up + * if sendmsg is not able to make progress? + */ +static int xs_tcp_send_request(struct rpc_task *task) +{ + struct rpc_rqst *req = task->tk_rqstp; + struct rpc_xprt *xprt = req->rq_xprt; + struct xdr_buf *xdr = &req->rq_snd_buf; + int status, retry = 0; + + xs_encode_tcp_record_marker(&req->rq_snd_buf); + + xs_pktdump("packet data:", + req->rq_svec->iov_base, + req->rq_svec->iov_len); + + /* Continue transmitting the packet/record. We must be careful + * to cope with writespace callbacks arriving _after_ we have + * called sendmsg(). */ + while (1) { + req->rq_xtime = jiffies; + status = xs_sendpages(xprt->sock, NULL, 0, xdr, + req->rq_bytes_sent); + + dprintk("RPC: xs_tcp_send_request(%u) = %d\n", + xdr->len - req->rq_bytes_sent, status); + + if (unlikely(status < 0)) + break; + + /* If we've sent the entire packet, immediately + * reset the count of bytes sent. */ + req->rq_bytes_sent += status; + if (likely(req->rq_bytes_sent >= req->rq_slen)) { + req->rq_bytes_sent = 0; + return 0; + } + + status = -EAGAIN; + if (retry++ > XS_SENDMSG_RETRY) + break; + } + + switch (status) { + case -EAGAIN: + xs_nospace(task); + break; + case -ECONNREFUSED: + case -ECONNRESET: + case -ENOTCONN: + case -EPIPE: + status = -ENOTCONN; + break; + default: + dprintk("RPC: sendmsg returned unrecognized error %d\n", + -status); + xprt_disconnect(xprt); + break; + } + + return status; +} + +/** + * xs_close - close a socket + * @xprt: transport + * + * This is used when all requests are complete; ie, no DRC state remains + * on the server we want to save. + */ +static void xs_close(struct rpc_xprt *xprt) +{ + struct socket *sock = xprt->sock; + struct sock *sk = xprt->inet; + + if (!sk) + return; + + dprintk("RPC: xs_close xprt %p\n", xprt); + + write_lock_bh(&sk->sk_callback_lock); + xprt->inet = NULL; + xprt->sock = NULL; + + sk->sk_user_data = NULL; + sk->sk_data_ready = xprt->old_data_ready; + sk->sk_state_change = xprt->old_state_change; + sk->sk_write_space = xprt->old_write_space; + write_unlock_bh(&sk->sk_callback_lock); + + sk->sk_no_check = 0; + + sock_release(sock); +} + +/** + * xs_destroy - prepare to shutdown a transport + * @xprt: doomed transport + * + */ +static void xs_destroy(struct rpc_xprt *xprt) +{ + dprintk("RPC: xs_destroy xprt %p\n", xprt); + + cancel_delayed_work(&xprt->connect_worker); + flush_scheduled_work(); + + xprt_disconnect(xprt); + xs_close(xprt); + kfree(xprt->slot); +} + +static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) +{ + return (struct rpc_xprt *) sk->sk_user_data; +} + +/** + * xs_udp_data_ready - "data ready" callback for UDP sockets + * @sk: socket with data to read + * @len: how much data to read + * + */ +static void xs_udp_data_ready(struct sock *sk, int len) +{ + struct rpc_task *task; + struct rpc_xprt *xprt; + struct rpc_rqst *rovr; + struct sk_buff *skb; + int err, repsize, copied; + u32 _xid, *xp; + + read_lock(&sk->sk_callback_lock); + dprintk("RPC: xs_udp_data_ready...\n"); + if (!(xprt = xprt_from_sock(sk))) + goto out; + + if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) + goto out; + + if (xprt->shutdown) + goto dropit; + + repsize = skb->len - sizeof(struct udphdr); + if (repsize < 4) { + dprintk("RPC: impossible RPC reply size %d!\n", repsize); + goto dropit; + } + + /* Copy the XID from the skb... */ + xp = skb_header_pointer(skb, sizeof(struct udphdr), + sizeof(_xid), &_xid); + if (xp == NULL) + goto dropit; + + /* Look up and lock the request corresponding to the given XID */ + spin_lock(&xprt->transport_lock); + rovr = xprt_lookup_rqst(xprt, *xp); + if (!rovr) + goto out_unlock; + task = rovr->rq_task; + + if ((copied = rovr->rq_private_buf.buflen) > repsize) + copied = repsize; + + /* Suck it into the iovec, verify checksum if not done by hw. */ + if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) + goto out_unlock; + + /* Something worked... */ + dst_confirm(skb->dst); + + xprt_adjust_cwnd(task, copied); + xprt_update_rtt(task); + xprt_complete_rqst(task, copied); + + out_unlock: + spin_unlock(&xprt->transport_lock); + dropit: + skb_free_datagram(sk, skb); + out: + read_unlock(&sk->sk_callback_lock); +} + +static inline size_t xs_tcp_copy_data(skb_reader_t *desc, void *p, size_t len) +{ + if (len > desc->count) + len = desc->count; + if (skb_copy_bits(desc->skb, desc->offset, p, len)) { + dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n", + len, desc->count); + return 0; + } + desc->offset += len; + desc->count -= len; + dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n", + len, desc->count); + return len; +} + +static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + size_t len, used; + char *p; + + p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset; + len = sizeof(xprt->tcp_recm) - xprt->tcp_offset; + used = xs_tcp_copy_data(desc, p, len); + xprt->tcp_offset += used; + if (used != len) + return; + + xprt->tcp_reclen = ntohl(xprt->tcp_recm); + if (xprt->tcp_reclen & RPC_LAST_STREAM_FRAGMENT) + xprt->tcp_flags |= XPRT_LAST_FRAG; + else + xprt->tcp_flags &= ~XPRT_LAST_FRAG; + xprt->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK; + + xprt->tcp_flags &= ~XPRT_COPY_RECM; + xprt->tcp_offset = 0; + + /* Sanity check of the record length */ + if (unlikely(xprt->tcp_reclen < 4)) { + dprintk("RPC: invalid TCP record fragment length\n"); + xprt_disconnect(xprt); + return; + } + dprintk("RPC: reading TCP record fragment of length %d\n", + xprt->tcp_reclen); +} + +static void xs_tcp_check_recm(struct rpc_xprt *xprt) +{ + dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n", + xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags); + if (xprt->tcp_offset == xprt->tcp_reclen) { + xprt->tcp_flags |= XPRT_COPY_RECM; + xprt->tcp_offset = 0; + if (xprt->tcp_flags & XPRT_LAST_FRAG) { + xprt->tcp_flags &= ~XPRT_COPY_DATA; + xprt->tcp_flags |= XPRT_COPY_XID; + xprt->tcp_copied = 0; + } + } +} + +static inline void xs_tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + size_t len, used; + char *p; + + len = sizeof(xprt->tcp_xid) - xprt->tcp_offset; + dprintk("RPC: reading XID (%Zu bytes)\n", len); + p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset; + used = xs_tcp_copy_data(desc, p, len); + xprt->tcp_offset += used; + if (used != len) + return; + xprt->tcp_flags &= ~XPRT_COPY_XID; + xprt->tcp_flags |= XPRT_COPY_DATA; + xprt->tcp_copied = 4; + dprintk("RPC: reading reply for XID %08x\n", + ntohl(xprt->tcp_xid)); + xs_tcp_check_recm(xprt); +} + +static inline void xs_tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + struct rpc_rqst *req; + struct xdr_buf *rcvbuf; + size_t len; + ssize_t r; + + /* Find and lock the request corresponding to this xid */ + spin_lock(&xprt->transport_lock); + req = xprt_lookup_rqst(xprt, xprt->tcp_xid); + if (!req) { + xprt->tcp_flags &= ~XPRT_COPY_DATA; + dprintk("RPC: XID %08x request not found!\n", + ntohl(xprt->tcp_xid)); + spin_unlock(&xprt->transport_lock); + return; + } + + rcvbuf = &req->rq_private_buf; + len = desc->count; + if (len > xprt->tcp_reclen - xprt->tcp_offset) { + skb_reader_t my_desc; + + len = xprt->tcp_reclen - xprt->tcp_offset; + memcpy(&my_desc, desc, sizeof(my_desc)); + my_desc.count = len; + r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, + &my_desc, xs_tcp_copy_data); + desc->count -= r; + desc->offset += r; + } else + r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, + desc, xs_tcp_copy_data); + + if (r > 0) { + xprt->tcp_copied += r; + xprt->tcp_offset += r; + } + if (r != len) { + /* Error when copying to the receive buffer, + * usually because we weren't able to allocate + * additional buffer pages. All we can do now + * is turn off XPRT_COPY_DATA, so the request + * will not receive any additional updates, + * and time out. + * Any remaining data from this record will + * be discarded. + */ + xprt->tcp_flags &= ~XPRT_COPY_DATA; + dprintk("RPC: XID %08x truncated request\n", + ntohl(xprt->tcp_xid)); + dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", + xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); + goto out; + } + + dprintk("RPC: XID %08x read %Zd bytes\n", + ntohl(xprt->tcp_xid), r); + dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", + xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); + + if (xprt->tcp_copied == req->rq_private_buf.buflen) + xprt->tcp_flags &= ~XPRT_COPY_DATA; + else if (xprt->tcp_offset == xprt->tcp_reclen) { + if (xprt->tcp_flags & XPRT_LAST_FRAG) + xprt->tcp_flags &= ~XPRT_COPY_DATA; + } + +out: + if (!(xprt->tcp_flags & XPRT_COPY_DATA)) + xprt_complete_rqst(req->rq_task, xprt->tcp_copied); + spin_unlock(&xprt->transport_lock); + xs_tcp_check_recm(xprt); +} + +static inline void xs_tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc) +{ + size_t len; + + len = xprt->tcp_reclen - xprt->tcp_offset; + if (len > desc->count) + len = desc->count; + desc->count -= len; + desc->offset += len; + xprt->tcp_offset += len; + dprintk("RPC: discarded %Zu bytes\n", len); + xs_tcp_check_recm(xprt); +} + +static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) +{ + struct rpc_xprt *xprt = rd_desc->arg.data; + skb_reader_t desc = { + .skb = skb, + .offset = offset, + .count = len, + .csum = 0 + }; + + dprintk("RPC: xs_tcp_data_recv started\n"); + do { + /* Read in a new fragment marker if necessary */ + /* Can we ever really expect to get completely empty fragments? */ + if (xprt->tcp_flags & XPRT_COPY_RECM) { + xs_tcp_read_fraghdr(xprt, &desc); + continue; + } + /* Read in the xid if necessary */ + if (xprt->tcp_flags & XPRT_COPY_XID) { + xs_tcp_read_xid(xprt, &desc); + continue; + } + /* Read in the request data */ + if (xprt->tcp_flags & XPRT_COPY_DATA) { + xs_tcp_read_request(xprt, &desc); + continue; + } + /* Skip over any trailing bytes on short reads */ + xs_tcp_read_discard(xprt, &desc); + } while (desc.count); + dprintk("RPC: xs_tcp_data_recv done\n"); + return len - desc.count; +} + +/** + * xs_tcp_data_ready - "data ready" callback for TCP sockets + * @sk: socket with data to read + * @bytes: how much data to read + * + */ +static void xs_tcp_data_ready(struct sock *sk, int bytes) +{ + struct rpc_xprt *xprt; + read_descriptor_t rd_desc; + + read_lock(&sk->sk_callback_lock); + dprintk("RPC: xs_tcp_data_ready...\n"); + if (!(xprt = xprt_from_sock(sk))) + goto out; + if (xprt->shutdown) + goto out; + + /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ + rd_desc.arg.data = xprt; + rd_desc.count = 65536; + tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); +out: + read_unlock(&sk->sk_callback_lock); +} + +/** + * xs_tcp_state_change - callback to handle TCP socket state changes + * @sk: socket whose state has changed + * + */ +static void xs_tcp_state_change(struct sock *sk) +{ + struct rpc_xprt *xprt; + + read_lock(&sk->sk_callback_lock); + if (!(xprt = xprt_from_sock(sk))) + goto out; + dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); + dprintk("RPC: state %x conn %d dead %d zapped %d\n", + sk->sk_state, xprt_connected(xprt), + sock_flag(sk, SOCK_DEAD), + sock_flag(sk, SOCK_ZAPPED)); + + switch (sk->sk_state) { + case TCP_ESTABLISHED: + spin_lock_bh(&xprt->transport_lock); + if (!xprt_test_and_set_connected(xprt)) { + /* Reset TCP record info */ + xprt->tcp_offset = 0; + xprt->tcp_reclen = 0; + xprt->tcp_copied = 0; + xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID; + xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; + xprt_wake_pending_tasks(xprt, 0); + } + spin_unlock_bh(&xprt->transport_lock); + break; + case TCP_SYN_SENT: + case TCP_SYN_RECV: + break; + default: + xprt_disconnect(xprt); + break; + } + out: + read_unlock(&sk->sk_callback_lock); +} + +/** + * xs_udp_write_space - callback invoked when socket buffer space + * becomes available + * @sk: socket whose state has changed + * + * Called when more output buffer space is available for this socket. + * We try not to wake our writers until they can make "significant" + * progress, otherwise we'll waste resources thrashing kernel_sendmsg + * with a bunch of small requests. + */ +static void xs_udp_write_space(struct sock *sk) +{ + read_lock(&sk->sk_callback_lock); + + /* from net/core/sock.c:sock_def_write_space */ + if (sock_writeable(sk)) { + struct socket *sock; + struct rpc_xprt *xprt; + + if (unlikely(!(sock = sk->sk_socket))) + goto out; + if (unlikely(!(xprt = xprt_from_sock(sk)))) + goto out; + if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) + goto out; + + xprt_write_space(xprt); + } + + out: + read_unlock(&sk->sk_callback_lock); +} + +/** + * xs_tcp_write_space - callback invoked when socket buffer space + * becomes available + * @sk: socket whose state has changed + * + * Called when more output buffer space is available for this socket. + * We try not to wake our writers until they can make "significant" + * progress, otherwise we'll waste resources thrashing kernel_sendmsg + * with a bunch of small requests. + */ +static void xs_tcp_write_space(struct sock *sk) +{ + read_lock(&sk->sk_callback_lock); + + /* from net/core/stream.c:sk_stream_write_space */ + if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) { + struct socket *sock; + struct rpc_xprt *xprt; + + if (unlikely(!(sock = sk->sk_socket))) + goto out; + if (unlikely(!(xprt = xprt_from_sock(sk)))) + goto out; + if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) + goto out; + + xprt_write_space(xprt); + } + + out: + read_unlock(&sk->sk_callback_lock); +} + +static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) +{ + struct sock *sk = xprt->inet; + + if (xprt->rcvsize) { + sk->sk_userlocks |= SOCK_RCVBUF_LOCK; + sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2; + } + if (xprt->sndsize) { + sk->sk_userlocks |= SOCK_SNDBUF_LOCK; + sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2; + sk->sk_write_space(sk); + } +} + +/** + * xs_udp_set_buffer_size - set send and receive limits + * @xprt: generic transport + * @sndsize: requested size of send buffer, in bytes + * @rcvsize: requested size of receive buffer, in bytes + * + * Set socket send and receive buffer size limits. + */ +static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) +{ + xprt->sndsize = 0; + if (sndsize) + xprt->sndsize = sndsize + 1024; + xprt->rcvsize = 0; + if (rcvsize) + xprt->rcvsize = rcvsize + 1024; + + xs_udp_do_set_buffer_size(xprt); +} + +/** + * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport + * @task: task that timed out + * + * Adjust the congestion window after a retransmit timeout has occurred. + */ +static void xs_udp_timer(struct rpc_task *task) +{ + xprt_adjust_cwnd(task, -ETIMEDOUT); +} + +static int xs_bindresvport(struct rpc_xprt *xprt, struct socket *sock) +{ + struct sockaddr_in myaddr = { + .sin_family = AF_INET, + }; + int err; + unsigned short port = xprt->port; + + do { + myaddr.sin_port = htons(port); + err = sock->ops->bind(sock, (struct sockaddr *) &myaddr, + sizeof(myaddr)); + if (err == 0) { + xprt->port = port; + dprintk("RPC: xs_bindresvport bound to port %u\n", + port); + return 0; + } + if (port <= xprt_min_resvport) + port = xprt_max_resvport; + else + port--; + } while (err == -EADDRINUSE && port != xprt->port); + + dprintk("RPC: can't bind to reserved port (%d).\n", -err); + return err; +} + +/** + * xs_udp_connect_worker - set up a UDP socket + * @args: RPC transport to connect + * + * Invoked by a work queue tasklet. + */ +static void xs_udp_connect_worker(void *args) +{ + struct rpc_xprt *xprt = (struct rpc_xprt *) args; + struct socket *sock = xprt->sock; + int err, status = -EIO; + + if (xprt->shutdown || xprt->addr.sin_port == 0) + goto out; + + dprintk("RPC: xs_udp_connect_worker for xprt %p\n", xprt); + + /* Start by resetting any existing state */ + xs_close(xprt); + + if ((err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { + dprintk("RPC: can't create UDP transport socket (%d).\n", -err); + goto out; + } + + if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) { + sock_release(sock); + goto out; + } + + if (!xprt->inet) { + struct sock *sk = sock->sk; + + write_lock_bh(&sk->sk_callback_lock); + + sk->sk_user_data = xprt; + xprt->old_data_ready = sk->sk_data_ready; + xprt->old_state_change = sk->sk_state_change; + xprt->old_write_space = sk->sk_write_space; + sk->sk_data_ready = xs_udp_data_ready; + sk->sk_write_space = xs_udp_write_space; + sk->sk_no_check = UDP_CSUM_NORCV; + + xprt_set_connected(xprt); + + /* Reset to new socket */ + xprt->sock = sock; + xprt->inet = sk; + + write_unlock_bh(&sk->sk_callback_lock); + } + xs_udp_do_set_buffer_size(xprt); + status = 0; +out: + xprt_wake_pending_tasks(xprt, status); + xprt_clear_connecting(xprt); +} + +/* + * We need to preserve the port number so the reply cache on the server can + * find our cached RPC replies when we get around to reconnecting. + */ +static void xs_tcp_reuse_connection(struct rpc_xprt *xprt) +{ + int result; + struct socket *sock = xprt->sock; + struct sockaddr any; + + dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt); + + /* + * Disconnect the transport socket by doing a connect operation + * with AF_UNSPEC. This should return immediately... + */ + memset(&any, 0, sizeof(any)); + any.sa_family = AF_UNSPEC; + result = sock->ops->connect(sock, &any, sizeof(any), 0); + if (result) + dprintk("RPC: AF_UNSPEC connect return code %d\n", + result); +} + +/** + * xs_tcp_connect_worker - connect a TCP socket to a remote endpoint + * @args: RPC transport to connect + * + * Invoked by a work queue tasklet. + */ +static void xs_tcp_connect_worker(void *args) +{ + struct rpc_xprt *xprt = (struct rpc_xprt *)args; + struct socket *sock = xprt->sock; + int err, status = -EIO; + + if (xprt->shutdown || xprt->addr.sin_port == 0) + goto out; + + dprintk("RPC: xs_tcp_connect_worker for xprt %p\n", xprt); + + if (!xprt->sock) { + /* start from scratch */ + if ((err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { + dprintk("RPC: can't create TCP transport socket (%d).\n", -err); + goto out; + } + + if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) { + sock_release(sock); + goto out; + } + } else + /* "close" the socket, preserving the local port */ + xs_tcp_reuse_connection(xprt); + + if (!xprt->inet) { + struct sock *sk = sock->sk; + + write_lock_bh(&sk->sk_callback_lock); + + sk->sk_user_data = xprt; + xprt->old_data_ready = sk->sk_data_ready; + xprt->old_state_change = sk->sk_state_change; + xprt->old_write_space = sk->sk_write_space; + sk->sk_data_ready = xs_tcp_data_ready; + sk->sk_state_change = xs_tcp_state_change; + sk->sk_write_space = xs_tcp_write_space; + + /* socket options */ + sk->sk_userlocks |= SOCK_BINDPORT_LOCK; + sock_reset_flag(sk, SOCK_LINGER); + tcp_sk(sk)->linger2 = 0; + tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; + + xprt_clear_connected(xprt); + + /* Reset to new socket */ + xprt->sock = sock; + xprt->inet = sk; + + write_unlock_bh(&sk->sk_callback_lock); + } + + /* Tell the socket layer to start connecting... */ + status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr, + sizeof(xprt->addr), O_NONBLOCK); + dprintk("RPC: %p connect status %d connected %d sock state %d\n", + xprt, -status, xprt_connected(xprt), sock->sk->sk_state); + if (status < 0) { + switch (status) { + case -EINPROGRESS: + case -EALREADY: + goto out_clear; + case -ECONNREFUSED: + case -ECONNRESET: + /* retry with existing socket, after a delay */ + break; + default: + /* get rid of existing socket, and retry */ + xs_close(xprt); + break; + } + } +out: + xprt_wake_pending_tasks(xprt, status); +out_clear: + xprt_clear_connecting(xprt); +} + +/** + * xs_connect - connect a socket to a remote endpoint + * @task: address of RPC task that manages state of connect request + * + * TCP: If the remote end dropped the connection, delay reconnecting. + * + * UDP socket connects are synchronous, but we use a work queue anyway + * to guarantee that even unprivileged user processes can set up a + * socket on a privileged port. + * + * If a UDP socket connect fails, the delay behavior here prevents + * retry floods (hard mounts). + */ +static void xs_connect(struct rpc_task *task) +{ + struct rpc_xprt *xprt = task->tk_xprt; + + if (xprt_test_and_set_connecting(xprt)) + return; + + if (xprt->sock != NULL) { + dprintk("RPC: xs_connect delayed xprt %p for %lu seconds\n", + xprt, xprt->reestablish_timeout / HZ); + schedule_delayed_work(&xprt->connect_worker, + xprt->reestablish_timeout); + xprt->reestablish_timeout <<= 1; + if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) + xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; + } else { + dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); + schedule_work(&xprt->connect_worker); + + /* flush_scheduled_work can sleep... */ + if (!RPC_IS_ASYNC(task)) + flush_scheduled_work(); + } +} + +static struct rpc_xprt_ops xs_udp_ops = { + .set_buffer_size = xs_udp_set_buffer_size, + .reserve_xprt = xprt_reserve_xprt_cong, + .release_xprt = xprt_release_xprt_cong, + .connect = xs_connect, + .send_request = xs_udp_send_request, + .set_retrans_timeout = xprt_set_retrans_timeout_rtt, + .timer = xs_udp_timer, + .release_request = xprt_release_rqst_cong, + .close = xs_close, + .destroy = xs_destroy, +}; + +static struct rpc_xprt_ops xs_tcp_ops = { + .reserve_xprt = xprt_reserve_xprt, + .release_xprt = xprt_release_xprt, + .connect = xs_connect, + .send_request = xs_tcp_send_request, + .set_retrans_timeout = xprt_set_retrans_timeout_def, + .close = xs_close, + .destroy = xs_destroy, +}; + +/** + * xs_setup_udp - Set up transport to use a UDP socket + * @xprt: transport to set up + * @to: timeout parameters + * + */ +int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to) +{ + size_t slot_table_size; + + dprintk("RPC: setting up udp-ipv4 transport...\n"); + + xprt->max_reqs = xprt_udp_slot_table_entries; + slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]); + xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); + if (xprt->slot == NULL) + return -ENOMEM; + memset(xprt->slot, 0, slot_table_size); + + xprt->prot = IPPROTO_UDP; + xprt->port = xprt_max_resvport; + xprt->tsh_size = 0; + xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; + /* XXX: header size can vary due to auth type, IPv6, etc. */ + xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); + + INIT_WORK(&xprt->connect_worker, xs_udp_connect_worker, xprt); + xprt->bind_timeout = XS_BIND_TO; + xprt->connect_timeout = XS_UDP_CONN_TO; + xprt->reestablish_timeout = XS_UDP_REEST_TO; + xprt->idle_timeout = XS_IDLE_DISC_TO; + + xprt->ops = &xs_udp_ops; + + if (to) + xprt->timeout = *to; + else + xprt_set_timeout(&xprt->timeout, 5, 5 * HZ); + + return 0; +} + +/** + * xs_setup_tcp - Set up transport to use a TCP socket + * @xprt: transport to set up + * @to: timeout parameters + * + */ +int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to) +{ + size_t slot_table_size; + + dprintk("RPC: setting up tcp-ipv4 transport...\n"); + + xprt->max_reqs = xprt_tcp_slot_table_entries; + slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]); + xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); + if (xprt->slot == NULL) + return -ENOMEM; + memset(xprt->slot, 0, slot_table_size); + + xprt->prot = IPPROTO_TCP; + xprt->port = xprt_max_resvport; + xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); + xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; + xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; + + INIT_WORK(&xprt->connect_worker, xs_tcp_connect_worker, xprt); + xprt->bind_timeout = XS_BIND_TO; + xprt->connect_timeout = XS_TCP_CONN_TO; + xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; + xprt->idle_timeout = XS_IDLE_DISC_TO; + + xprt->ops = &xs_tcp_ops; + + if (to) + xprt->timeout = *to; + else + xprt_set_timeout(&xprt->timeout, 2, 60 * HZ); + + return 0; +}