linux/net/sunrpc/xprtsock.c

3035 lines
79 KiB
C
Raw Normal View History

/*
* linux/net/sunrpc/xprtsock.c
*
* Client-side transport implementation for sockets.
*
* TCP callback races fixes (C) 1998 Red Hat
* TCP send fixes (C) 1998 Red Hat
* TCP NFS related read + write fixes
* (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
*
* Rewrite of larges part of the code in order to stabilize TCP stuff.
* Fix behaviour when socket buffer is full.
* (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
*
* IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com>
*
* IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005.
* <gilles.quillard@bull.net>
*/
#include <linux/types.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/capability.h>
#include <linux/pagemap.h>
#include <linux/errno.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/net.h>
#include <linux/mm.h>
#include <linux/un.h>
#include <linux/udp.h>
#include <linux/tcp.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/addr.h>
#include <linux/sunrpc/sched.h>
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
#include <linux/sunrpc/svcsock.h>
#include <linux/sunrpc/xprtsock.h>
#include <linux/file.h>
#ifdef CONFIG_SUNRPC_BACKCHANNEL
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
#include <linux/sunrpc/bc_xprt.h>
#endif
#include <net/sock.h>
#include <net/checksum.h>
#include <net/udp.h>
#include <net/tcp.h>
#include <trace/events/sunrpc.h>
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
#include "sunrpc.h"
static void xs_close(struct rpc_xprt *xprt);
/*
* xprtsock tunables
*/
static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
#define XS_TCP_LINGER_TO (15U * HZ)
static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
/*
* We can register our own files under /proc/sys/sunrpc by
* calling register_sysctl_table() again. The files in that
* directory become the union of all files registered there.
*
* We simply need to make sure that we don't collide with
* someone else's file names!
*/
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 max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
static struct ctl_table_header *sunrpc_table_header;
/*
* FIXME: changing the UDP slot table size should also resize the UDP
* socket buffers for existing UDP transports
*/
static struct ctl_table xs_tunables_table[] = {
{
.procname = "udp_slot_table_entries",
.data = &xprt_udp_slot_table_entries,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_dointvec_minmax,
.extra1 = &min_slot_table_size,
.extra2 = &max_slot_table_size
},
{
.procname = "tcp_slot_table_entries",
.data = &xprt_tcp_slot_table_entries,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_dointvec_minmax,
.extra1 = &min_slot_table_size,
.extra2 = &max_slot_table_size
},
{
.procname = "tcp_max_slot_table_entries",
.data = &xprt_max_tcp_slot_table_entries,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_dointvec_minmax,
.extra1 = &min_slot_table_size,
.extra2 = &max_tcp_slot_table_limit
},
{
.procname = "min_resvport",
.data = &xprt_min_resvport,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_dointvec_minmax,
.extra1 = &xprt_min_resvport_limit,
.extra2 = &xprt_max_resvport_limit
},
{
.procname = "max_resvport",
.data = &xprt_max_resvport,
.maxlen = sizeof(unsigned int),
.mode = 0644,
.proc_handler = proc_dointvec_minmax,
.extra1 = &xprt_min_resvport_limit,
.extra2 = &xprt_max_resvport_limit
},
{
.procname = "tcp_fin_timeout",
.data = &xs_tcp_fin_timeout,
.maxlen = sizeof(xs_tcp_fin_timeout),
.mode = 0644,
.proc_handler = proc_dointvec_jiffies,
},
{ },
};
static struct ctl_table sunrpc_table[] = {
{
.procname = "sunrpc",
.mode = 0555,
.child = xs_tunables_table
},
{ },
};
#endif
/*
* 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)
#if IS_ENABLED(CONFIG_SUNRPC_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
static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
{
return (struct rpc_xprt *) sk->sk_user_data;
}
static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
{
return (struct sockaddr *) &xprt->addr;
}
static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
{
return (struct sockaddr_un *) &xprt->addr;
}
static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
{
return (struct sockaddr_in *) &xprt->addr;
}
static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
{
return (struct sockaddr_in6 *) &xprt->addr;
}
static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
{
struct sockaddr *sap = xs_addr(xprt);
struct sockaddr_in6 *sin6;
struct sockaddr_in *sin;
struct sockaddr_un *sun;
char buf[128];
switch (sap->sa_family) {
case AF_LOCAL:
sun = xs_addr_un(xprt);
strlcpy(buf, sun->sun_path, sizeof(buf));
xprt->address_strings[RPC_DISPLAY_ADDR] =
kstrdup(buf, GFP_KERNEL);
break;
case AF_INET:
(void)rpc_ntop(sap, buf, sizeof(buf));
xprt->address_strings[RPC_DISPLAY_ADDR] =
kstrdup(buf, GFP_KERNEL);
sin = xs_addr_in(xprt);
snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
break;
case AF_INET6:
(void)rpc_ntop(sap, buf, sizeof(buf));
xprt->address_strings[RPC_DISPLAY_ADDR] =
kstrdup(buf, GFP_KERNEL);
sin6 = xs_addr_in6(xprt);
snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
break;
default:
BUG();
}
xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
}
static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
{
struct sockaddr *sap = xs_addr(xprt);
char buf[128];
snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
}
static void xs_format_peer_addresses(struct rpc_xprt *xprt,
const char *protocol,
const char *netid)
{
xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
xprt->address_strings[RPC_DISPLAY_NETID] = netid;
xs_format_common_peer_addresses(xprt);
xs_format_common_peer_ports(xprt);
}
static void xs_update_peer_port(struct rpc_xprt *xprt)
{
kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
xs_format_common_peer_ports(xprt);
}
static void xs_free_peer_addresses(struct rpc_xprt *xprt)
{
unsigned int i;
for (i = 0; i < RPC_DISPLAY_MAX; i++)
switch (i) {
case RPC_DISPLAY_PROTO:
case RPC_DISPLAY_NETID:
continue;
default:
kfree(xprt->address_strings[i]);
}
}
#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more)
{
struct msghdr msg = {
.msg_name = addr,
.msg_namelen = addrlen,
.msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0),
};
struct kvec iov = {
.iov_base = vec->iov_base + base,
.iov_len = vec->iov_len - base,
};
if (iov.iov_len != 0)
return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
return kernel_sendmsg(sock, &msg, NULL, 0, 0);
}
static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more, bool zerocopy, int *sent_p)
{
SUNRPC: Fix a data corruption issue when retransmitting RPC calls The following scenario can cause silent data corruption when doing NFS writes. It has mainly been observed when doing database writes using O_DIRECT. 1) The RPC client uses sendpage() to do zero-copy of the page data. 2) Due to networking issues, the reply from the server is delayed, and so the RPC client times out. 3) The client issues a second sendpage of the page data as part of an RPC call retransmission. 4) The reply to the first transmission arrives from the server _before_ the client hardware has emptied the TCP socket send buffer. 5) After processing the reply, the RPC state machine rules that the call to be done, and triggers the completion callbacks. 6) The application notices the RPC call is done, and reuses the pages to store something else (e.g. a new write). 7) The client NIC drains the TCP socket send buffer. Since the page data has now changed, it reads a corrupted version of the initial RPC call, and puts it on the wire. This patch fixes the problem in the following manner: The ordering guarantees of TCP ensure that when the server sends a reply, then we know that the _first_ transmission has completed. Using zero-copy in that situation is therefore safe. If a time out occurs, we then send the retransmission using sendmsg() (i.e. no zero-copy), We then know that the socket contains a full copy of the data, and so it will retransmit a faithful reproduction even if the RPC call completes, and the application reuses the O_DIRECT buffer in the meantime. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: stable@vger.kernel.org
2013-11-08 22:03:50 +01:00
ssize_t (*do_sendpage)(struct socket *sock, struct page *page,
int offset, size_t size, int flags);
struct page **ppage;
unsigned int remainder;
int err;
remainder = xdr->page_len - base;
base += xdr->page_base;
ppage = xdr->pages + (base >> PAGE_SHIFT);
base &= ~PAGE_MASK;
SUNRPC: Fix a data corruption issue when retransmitting RPC calls The following scenario can cause silent data corruption when doing NFS writes. It has mainly been observed when doing database writes using O_DIRECT. 1) The RPC client uses sendpage() to do zero-copy of the page data. 2) Due to networking issues, the reply from the server is delayed, and so the RPC client times out. 3) The client issues a second sendpage of the page data as part of an RPC call retransmission. 4) The reply to the first transmission arrives from the server _before_ the client hardware has emptied the TCP socket send buffer. 5) After processing the reply, the RPC state machine rules that the call to be done, and triggers the completion callbacks. 6) The application notices the RPC call is done, and reuses the pages to store something else (e.g. a new write). 7) The client NIC drains the TCP socket send buffer. Since the page data has now changed, it reads a corrupted version of the initial RPC call, and puts it on the wire. This patch fixes the problem in the following manner: The ordering guarantees of TCP ensure that when the server sends a reply, then we know that the _first_ transmission has completed. Using zero-copy in that situation is therefore safe. If a time out occurs, we then send the retransmission using sendmsg() (i.e. no zero-copy), We then know that the socket contains a full copy of the data, and so it will retransmit a faithful reproduction even if the RPC call completes, and the application reuses the O_DIRECT buffer in the meantime. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: stable@vger.kernel.org
2013-11-08 22:03:50 +01:00
do_sendpage = sock->ops->sendpage;
if (!zerocopy)
do_sendpage = sock_no_sendpage;
for(;;) {
unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder);
int flags = XS_SENDMSG_FLAGS;
remainder -= len;
if (remainder != 0 || more)
flags |= MSG_MORE;
SUNRPC: Fix a data corruption issue when retransmitting RPC calls The following scenario can cause silent data corruption when doing NFS writes. It has mainly been observed when doing database writes using O_DIRECT. 1) The RPC client uses sendpage() to do zero-copy of the page data. 2) Due to networking issues, the reply from the server is delayed, and so the RPC client times out. 3) The client issues a second sendpage of the page data as part of an RPC call retransmission. 4) The reply to the first transmission arrives from the server _before_ the client hardware has emptied the TCP socket send buffer. 5) After processing the reply, the RPC state machine rules that the call to be done, and triggers the completion callbacks. 6) The application notices the RPC call is done, and reuses the pages to store something else (e.g. a new write). 7) The client NIC drains the TCP socket send buffer. Since the page data has now changed, it reads a corrupted version of the initial RPC call, and puts it on the wire. This patch fixes the problem in the following manner: The ordering guarantees of TCP ensure that when the server sends a reply, then we know that the _first_ transmission has completed. Using zero-copy in that situation is therefore safe. If a time out occurs, we then send the retransmission using sendmsg() (i.e. no zero-copy), We then know that the socket contains a full copy of the data, and so it will retransmit a faithful reproduction even if the RPC call completes, and the application reuses the O_DIRECT buffer in the meantime. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: stable@vger.kernel.org
2013-11-08 22:03:50 +01:00
err = do_sendpage(sock, *ppage, base, len, flags);
if (remainder == 0 || err != len)
break;
*sent_p += err;
ppage++;
base = 0;
}
if (err > 0) {
*sent_p += err;
err = 0;
}
return err;
}
/**
* 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
SUNRPC: Fix a data corruption issue when retransmitting RPC calls The following scenario can cause silent data corruption when doing NFS writes. It has mainly been observed when doing database writes using O_DIRECT. 1) The RPC client uses sendpage() to do zero-copy of the page data. 2) Due to networking issues, the reply from the server is delayed, and so the RPC client times out. 3) The client issues a second sendpage of the page data as part of an RPC call retransmission. 4) The reply to the first transmission arrives from the server _before_ the client hardware has emptied the TCP socket send buffer. 5) After processing the reply, the RPC state machine rules that the call to be done, and triggers the completion callbacks. 6) The application notices the RPC call is done, and reuses the pages to store something else (e.g. a new write). 7) The client NIC drains the TCP socket send buffer. Since the page data has now changed, it reads a corrupted version of the initial RPC call, and puts it on the wire. This patch fixes the problem in the following manner: The ordering guarantees of TCP ensure that when the server sends a reply, then we know that the _first_ transmission has completed. Using zero-copy in that situation is therefore safe. If a time out occurs, we then send the retransmission using sendmsg() (i.e. no zero-copy), We then know that the socket contains a full copy of the data, and so it will retransmit a faithful reproduction even if the RPC call completes, and the application reuses the O_DIRECT buffer in the meantime. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: stable@vger.kernel.org
2013-11-08 22:03:50 +01:00
* @zerocopy: true if it is safe to use sendpage()
* @sent_p: return the total number of bytes successfully queued for sending
*
*/
static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, bool zerocopy, int *sent_p)
{
unsigned int remainder = xdr->len - base;
int err = 0;
int sent = 0;
if (unlikely(!sock))
return -ENOTSOCK;
clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
if (base != 0) {
addr = NULL;
addrlen = 0;
}
if (base < xdr->head[0].iov_len || addr != NULL) {
unsigned int len = xdr->head[0].iov_len - base;
remainder -= len;
err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0);
if (remainder == 0 || err != len)
goto out;
*sent_p += err;
base = 0;
} else
base -= xdr->head[0].iov_len;
if (base < xdr->page_len) {
unsigned int len = xdr->page_len - base;
remainder -= len;
err = xs_send_pagedata(sock, xdr, base, remainder != 0, zerocopy, &sent);
*sent_p += sent;
if (remainder == 0 || sent != len)
goto out;
base = 0;
} else
base -= xdr->page_len;
if (base >= xdr->tail[0].iov_len)
return 0;
err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0);
out:
if (err > 0) {
*sent_p += err;
err = 0;
}
return err;
}
static void xs_nospace_callback(struct rpc_task *task)
{
struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt);
transport->inet->sk_write_pending--;
clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
}
/**
* xs_nospace - place task on wait queue if transmit was incomplete
* @task: task to put to sleep
*
*/
static int xs_nospace(struct rpc_task *task)
{
struct rpc_rqst *req = task->tk_rqstp;
struct rpc_xprt *xprt = req->rq_xprt;
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
struct sock *sk = transport->inet;
int ret = -EAGAIN;
dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
task->tk_pid, req->rq_slen - req->rq_bytes_sent,
req->rq_slen);
/* Protect against races with write_space */
spin_lock_bh(&xprt->transport_lock);
/* Don't race with disconnect */
if (xprt_connected(xprt)) {
if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) {
/*
* Notify TCP that we're limited by the application
* window size
*/
set_bit(SOCK_NOSPACE, &transport->sock->flags);
sk->sk_write_pending++;
/* ...and wait for more buffer space */
xprt_wait_for_buffer_space(task, xs_nospace_callback);
}
} else {
clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
ret = -ENOTCONN;
}
spin_unlock_bh(&xprt->transport_lock);
/* Race breaker in case memory is freed before above code is called */
sk->sk_write_space(sk);
return ret;
}
/*
* Construct a stream transport record marker in @buf.
*/
static inline void xs_encode_stream_record_marker(struct xdr_buf *buf)
{
u32 reclen = buf->len - sizeof(rpc_fraghdr);
rpc_fraghdr *base = buf->head[0].iov_base;
*base = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | reclen);
}
/**
* xs_local_send_request - write an RPC request to an AF_LOCAL socket
* @task: 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_local_send_request(struct rpc_task *task)
{
struct rpc_rqst *req = task->tk_rqstp;
struct rpc_xprt *xprt = req->rq_xprt;
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
struct xdr_buf *xdr = &req->rq_snd_buf;
int status;
int sent = 0;
xs_encode_stream_record_marker(&req->rq_snd_buf);
xs_pktdump("packet data:",
req->rq_svec->iov_base, req->rq_svec->iov_len);
status = xs_sendpages(transport->sock, NULL, 0, xdr, req->rq_bytes_sent,
true, &sent);
dprintk("RPC: %s(%u) = %d\n",
__func__, xdr->len - req->rq_bytes_sent, status);
if (likely(sent > 0) || status == 0) {
req->rq_bytes_sent += sent;
req->rq_xmit_bytes_sent += sent;
if (likely(req->rq_bytes_sent >= req->rq_slen)) {
req->rq_bytes_sent = 0;
return 0;
}
status = -EAGAIN;
}
switch (status) {
case -ENOBUFS:
case -EAGAIN:
status = xs_nospace(task);
break;
default:
dprintk("RPC: sendmsg returned unrecognized error %d\n",
-status);
case -EPIPE:
xs_close(xprt);
status = -ENOTCONN;
}
return status;
}
/**
* 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 occurred, 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 sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
struct xdr_buf *xdr = &req->rq_snd_buf;
int sent = 0;
int status;
xs_pktdump("packet data:",
req->rq_svec->iov_base,
req->rq_svec->iov_len);
if (!xprt_bound(xprt))
return -ENOTCONN;
status = xs_sendpages(transport->sock, xs_addr(xprt), xprt->addrlen,
xdr, req->rq_bytes_sent, true, &sent);
dprintk("RPC: xs_udp_send_request(%u) = %d\n",
xdr->len - req->rq_bytes_sent, status);
rpc: Add -EPERM processing for xs_udp_send_request() If an iptables drop rule is added for an nfs server, the client can end up in a softlockup. Because of the way that xs_sendpages() is structured, the -EPERM is ignored since the prior bits of the packet may have been successfully queued and thus xs_sendpages() returns a non-zero value. Then, xs_udp_send_request() thinks that because some bits were queued it should return -EAGAIN. We then try the request again and again, resulting in cpu spinning. Reproducer: 1) open a file on the nfs server '/nfs/foo' (mounted using udp) 2) iptables -A OUTPUT -d <nfs server ip> -j DROP 3) write to /nfs/foo 4) close /nfs/foo 5) iptables -D OUTPUT -d <nfs server ip> -j DROP The softlockup occurs in step 4 above. The previous patch, allows xs_sendpages() to return both a sent count and any error values that may have occurred. Thus, if we get an -EPERM, return that to the higher level code. With this patch in place we can successfully abort the above sequence and avoid the softlockup. I also tried the above test case on an nfs mount on tcp and although the system does not softlockup, I still ended up with the 'hung_task' firing after 120 seconds, due to the i/o being stuck. The tcp case appears a bit harder to fix, since -EPERM appears to get ignored much lower down in the stack and does not propogate up to xs_sendpages(). This case is not quite as insidious as the softlockup and it is not addressed here. Reported-by: Yigong Lou <ylou@akamai.com> Signed-off-by: Jason Baron <jbaron@akamai.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-09-24 20:08:04 +02:00
/* firewall is blocking us, don't return -EAGAIN or we end up looping */
if (status == -EPERM)
goto process_status;
if (sent > 0 || status == 0) {
req->rq_xmit_bytes_sent += sent;
if (sent >= req->rq_slen)
return 0;
/* Still some bytes left; set up for a retry later. */
status = -EAGAIN;
}
rpc: Add -EPERM processing for xs_udp_send_request() If an iptables drop rule is added for an nfs server, the client can end up in a softlockup. Because of the way that xs_sendpages() is structured, the -EPERM is ignored since the prior bits of the packet may have been successfully queued and thus xs_sendpages() returns a non-zero value. Then, xs_udp_send_request() thinks that because some bits were queued it should return -EAGAIN. We then try the request again and again, resulting in cpu spinning. Reproducer: 1) open a file on the nfs server '/nfs/foo' (mounted using udp) 2) iptables -A OUTPUT -d <nfs server ip> -j DROP 3) write to /nfs/foo 4) close /nfs/foo 5) iptables -D OUTPUT -d <nfs server ip> -j DROP The softlockup occurs in step 4 above. The previous patch, allows xs_sendpages() to return both a sent count and any error values that may have occurred. Thus, if we get an -EPERM, return that to the higher level code. With this patch in place we can successfully abort the above sequence and avoid the softlockup. I also tried the above test case on an nfs mount on tcp and although the system does not softlockup, I still ended up with the 'hung_task' firing after 120 seconds, due to the i/o being stuck. The tcp case appears a bit harder to fix, since -EPERM appears to get ignored much lower down in the stack and does not propogate up to xs_sendpages(). This case is not quite as insidious as the softlockup and it is not addressed here. Reported-by: Yigong Lou <ylou@akamai.com> Signed-off-by: Jason Baron <jbaron@akamai.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-09-24 20:08:04 +02:00
process_status:
switch (status) {
case -ENOTSOCK:
status = -ENOTCONN;
/* Should we call xs_close() here? */
break;
case -EAGAIN:
status = xs_nospace(task);
break;
default:
dprintk("RPC: sendmsg returned unrecognized error %d\n",
-status);
case -ENETUNREACH:
case -ENOBUFS:
case -EPIPE:
case -ECONNREFUSED:
rpc: Add -EPERM processing for xs_udp_send_request() If an iptables drop rule is added for an nfs server, the client can end up in a softlockup. Because of the way that xs_sendpages() is structured, the -EPERM is ignored since the prior bits of the packet may have been successfully queued and thus xs_sendpages() returns a non-zero value. Then, xs_udp_send_request() thinks that because some bits were queued it should return -EAGAIN. We then try the request again and again, resulting in cpu spinning. Reproducer: 1) open a file on the nfs server '/nfs/foo' (mounted using udp) 2) iptables -A OUTPUT -d <nfs server ip> -j DROP 3) write to /nfs/foo 4) close /nfs/foo 5) iptables -D OUTPUT -d <nfs server ip> -j DROP The softlockup occurs in step 4 above. The previous patch, allows xs_sendpages() to return both a sent count and any error values that may have occurred. Thus, if we get an -EPERM, return that to the higher level code. With this patch in place we can successfully abort the above sequence and avoid the softlockup. I also tried the above test case on an nfs mount on tcp and although the system does not softlockup, I still ended up with the 'hung_task' firing after 120 seconds, due to the i/o being stuck. The tcp case appears a bit harder to fix, since -EPERM appears to get ignored much lower down in the stack and does not propogate up to xs_sendpages(). This case is not quite as insidious as the softlockup and it is not addressed here. Reported-by: Yigong Lou <ylou@akamai.com> Signed-off-by: Jason Baron <jbaron@akamai.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-09-24 20:08:04 +02:00
case -EPERM:
/* When the server has died, an ICMP port unreachable message
* prompts ECONNREFUSED. */
clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
}
return status;
}
/**
* xs_tcp_shutdown - gracefully shut down a TCP socket
* @xprt: transport
*
* Initiates a graceful shutdown of the TCP socket by calling the
* equivalent of shutdown(SHUT_RDWR);
*/
static void xs_tcp_shutdown(struct rpc_xprt *xprt)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
struct socket *sock = transport->sock;
if (sock != NULL) {
kernel_sock_shutdown(sock, SHUT_RDWR);
trace_rpc_socket_shutdown(xprt, sock);
}
}
/**
* 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 occurred, 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 sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
struct xdr_buf *xdr = &req->rq_snd_buf;
SUNRPC: Fix a data corruption issue when retransmitting RPC calls The following scenario can cause silent data corruption when doing NFS writes. It has mainly been observed when doing database writes using O_DIRECT. 1) The RPC client uses sendpage() to do zero-copy of the page data. 2) Due to networking issues, the reply from the server is delayed, and so the RPC client times out. 3) The client issues a second sendpage of the page data as part of an RPC call retransmission. 4) The reply to the first transmission arrives from the server _before_ the client hardware has emptied the TCP socket send buffer. 5) After processing the reply, the RPC state machine rules that the call to be done, and triggers the completion callbacks. 6) The application notices the RPC call is done, and reuses the pages to store something else (e.g. a new write). 7) The client NIC drains the TCP socket send buffer. Since the page data has now changed, it reads a corrupted version of the initial RPC call, and puts it on the wire. This patch fixes the problem in the following manner: The ordering guarantees of TCP ensure that when the server sends a reply, then we know that the _first_ transmission has completed. Using zero-copy in that situation is therefore safe. If a time out occurs, we then send the retransmission using sendmsg() (i.e. no zero-copy), We then know that the socket contains a full copy of the data, and so it will retransmit a faithful reproduction even if the RPC call completes, and the application reuses the O_DIRECT buffer in the meantime. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: stable@vger.kernel.org
2013-11-08 22:03:50 +01:00
bool zerocopy = true;
int status;
int sent;
xs_encode_stream_record_marker(&req->rq_snd_buf);
xs_pktdump("packet data:",
req->rq_svec->iov_base,
req->rq_svec->iov_len);
SUNRPC: Fix a data corruption issue when retransmitting RPC calls The following scenario can cause silent data corruption when doing NFS writes. It has mainly been observed when doing database writes using O_DIRECT. 1) The RPC client uses sendpage() to do zero-copy of the page data. 2) Due to networking issues, the reply from the server is delayed, and so the RPC client times out. 3) The client issues a second sendpage of the page data as part of an RPC call retransmission. 4) The reply to the first transmission arrives from the server _before_ the client hardware has emptied the TCP socket send buffer. 5) After processing the reply, the RPC state machine rules that the call to be done, and triggers the completion callbacks. 6) The application notices the RPC call is done, and reuses the pages to store something else (e.g. a new write). 7) The client NIC drains the TCP socket send buffer. Since the page data has now changed, it reads a corrupted version of the initial RPC call, and puts it on the wire. This patch fixes the problem in the following manner: The ordering guarantees of TCP ensure that when the server sends a reply, then we know that the _first_ transmission has completed. Using zero-copy in that situation is therefore safe. If a time out occurs, we then send the retransmission using sendmsg() (i.e. no zero-copy), We then know that the socket contains a full copy of the data, and so it will retransmit a faithful reproduction even if the RPC call completes, and the application reuses the O_DIRECT buffer in the meantime. Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: stable@vger.kernel.org
2013-11-08 22:03:50 +01:00
/* Don't use zero copy if this is a resend. If the RPC call
* completes while the socket holds a reference to the pages,
* then we may end up resending corrupted data.
*/
if (task->tk_flags & RPC_TASK_SENT)
zerocopy = false;
/* Continue transmitting the packet/record. We must be careful
* to cope with writespace callbacks arriving _after_ we have
* called sendmsg(). */
while (1) {
sent = 0;
status = xs_sendpages(transport->sock, NULL, 0, xdr,
req->rq_bytes_sent, zerocopy, &sent);
dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
xdr->len - req->rq_bytes_sent, status);
if (unlikely(sent == 0 && status < 0))
break;
/* If we've sent the entire packet, immediately
* reset the count of bytes sent. */
req->rq_bytes_sent += sent;
req->rq_xmit_bytes_sent += sent;
if (likely(req->rq_bytes_sent >= req->rq_slen)) {
req->rq_bytes_sent = 0;
return 0;
}
if (sent != 0)
continue;
status = -EAGAIN;
break;
}
switch (status) {
case -ENOTSOCK:
status = -ENOTCONN;
/* Should we call xs_close() here? */
break;
case -ENOBUFS:
case -EAGAIN:
status = xs_nospace(task);
break;
default:
dprintk("RPC: sendmsg returned unrecognized error %d\n",
-status);
case -ECONNRESET:
case -ECONNREFUSED:
case -ENOTCONN:
case -EADDRINUSE:
case -EPIPE:
clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
}
return status;
}
/**
* xs_tcp_release_xprt - clean up after a tcp transmission
* @xprt: transport
* @task: rpc task
*
* This cleans up if an error causes us to abort the transmission of a request.
* In this case, the socket may need to be reset in order to avoid confusing
* the server.
*/
static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
{
struct rpc_rqst *req;
if (task != xprt->snd_task)
return;
if (task == NULL)
goto out_release;
req = task->tk_rqstp;
if (req == NULL)
goto out_release;
if (req->rq_bytes_sent == 0)
goto out_release;
if (req->rq_bytes_sent == req->rq_snd_buf.len)
goto out_release;
set_bit(XPRT_CLOSE_WAIT, &xprt->state);
out_release:
xprt_release_xprt(xprt, task);
}
static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
{
transport->old_data_ready = sk->sk_data_ready;
transport->old_state_change = sk->sk_state_change;
transport->old_write_space = sk->sk_write_space;
transport->old_error_report = sk->sk_error_report;
}
static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
{
sk->sk_data_ready = transport->old_data_ready;
sk->sk_state_change = transport->old_state_change;
sk->sk_write_space = transport->old_write_space;
sk->sk_error_report = transport->old_error_report;
}
static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
{
smp_mb__before_atomic();
clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
clear_bit(XPRT_CLOSING, &xprt->state);
smp_mb__after_atomic();
}
static void xs_sock_mark_closed(struct rpc_xprt *xprt)
{
xs_sock_reset_connection_flags(xprt);
/* Mark transport as closed and wake up all pending tasks */
xprt_disconnect_done(xprt);
}
/**
* xs_error_report - callback to handle TCP socket state errors
* @sk: socket
*
* Note: we don't call sock_error() since there may be a rpc_task
* using the socket, and so we don't want to clear sk->sk_err.
*/
static void xs_error_report(struct sock *sk)
{
struct rpc_xprt *xprt;
int err;
read_lock_bh(&sk->sk_callback_lock);
if (!(xprt = xprt_from_sock(sk)))
goto out;
err = -sk->sk_err;
if (err == 0)
goto out;
/* Is this a reset event? */
if (sk->sk_state == TCP_CLOSE)
xs_sock_mark_closed(xprt);
dprintk("RPC: xs_error_report client %p, error=%d...\n",
xprt, -err);
trace_rpc_socket_error(xprt, sk->sk_socket, err);
xprt_wake_pending_tasks(xprt, err);
out:
read_unlock_bh(&sk->sk_callback_lock);
}
static void xs_reset_transport(struct sock_xprt *transport)
{
struct socket *sock = transport->sock;
struct sock *sk = transport->inet;
struct rpc_xprt *xprt = &transport->xprt;
if (sk == NULL)
return;
write_lock_bh(&sk->sk_callback_lock);
transport->inet = NULL;
transport->sock = NULL;
sk->sk_user_data = NULL;
xs_restore_old_callbacks(transport, sk);
write_unlock_bh(&sk->sk_callback_lock);
xs_sock_reset_connection_flags(xprt);
trace_rpc_socket_close(xprt, sock);
sock_release(sock);
}
/**
* 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.
*
* The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
* xs_reset_transport() zeroing the socket from underneath a writer.
*/
static void xs_close(struct rpc_xprt *xprt)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
dprintk("RPC: xs_close xprt %p\n", xprt);
xs_reset_transport(transport);
NFS/RPC: fix problems with reestablish_timeout and related code. [[resending with correct cc: - "vfs.kernel.org" just isn't right!]] xprt->reestablish_timeout is used to cause TCP connection attempts to back off if the connection fails so as not to hammer the network, but to still allow immediate connections when there is no reason to believe there is a problem. It is not used for the first connection (when transport->sock is NULL) but only on reconnects. It is currently set: a/ to 0 when xs_tcp_state_change finds a state of TCP_FIN_WAIT1 on the assumption that the client has closed the connection so the reconnect should be immediate when needed. b/ to at least XS_TCP_INIT_REEST_TO when xs_tcp_state_change detects TCP_CLOSING or TCP_CLOSE_WAIT on the assumption that the server closed the connection so a small delay at least is required. c/ as above when xs_tcp_state_change detects TCP_SYN_SENT, so that it is never 0 while a connection has been attempted, else the doubling will produce 0 and there will be no backoff. d/ to double is value (up to a limit) when delaying a connection, thus providing exponential backoff and e/ to XS_TCP_INIT_REEST_TO in xs_setup_tcp as simple initialisation. So you can see it is highly dependant on xs_tcp_state_change being called as expected. However experimental evidence shows that xs_tcp_state_change does not see all state changes. ("rpcdebug -m rpc trans" can help show what actually happens). Results show: TCP_ESTABLISHED is reported when a connection is made. TCP_SYN_SENT is never reported, so rule 'c' above is never effective. When the server closes the connection, TCP_CLOSE_WAIT and TCP_LAST_ACK *might* be reported, and TCP_CLOSE is always reported. This rule 'b' above will sometimes be effective, but not reliably. When the client closes the connection, it used to result in TCP_FIN_WAIT1, TCP_FIN_WAIT2, TCP_CLOSE. However since commit f75e674 (SUNRPC: Fix the problem of EADDRNOTAVAIL syslog floods on reconnect) we don't see *any* events on client-close. I think this is because xs_restore_old_callbacks is called to disconnect xs_tcp_state_change before the socket is closed. In any case, rule 'a' no longer applies. So all that is left are rule d, which successfully doubles the timeout which is never rest, and rule e which initialises the timeout. Even if the rules worked as expected, there would be a problem because a successful connection does not reset the timeout, so a sequence of events where the server closes the connection (e.g. during failover testing) will cause longer and longer timeouts with no good reason. This patch: - sets reestablish_timeout to 0 in xs_close thus effecting rule 'a' - sets it to 0 in xs_tcp_data_ready to ensure that a successful connection resets the timeout - sets it to at least XS_TCP_INIT_REEST_TO after it is doubled, thus effecting rule c I have not reimplemented rule b and the new version of rule c seems sufficient. I suspect other code in xs_tcp_data_ready needs to be revised as well. For example I don't think connect_cookie is being incremented as often as it should be. Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-09-23 20:36:37 +02:00
xprt->reestablish_timeout = 0;
xprt_disconnect_done(xprt);
}
static void xs_xprt_free(struct rpc_xprt *xprt)
{
xs_free_peer_addresses(xprt);
xprt_free(xprt);
}
/**
* 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);
xs_close(xprt);
xs_xprt_free(xprt);
module_put(THIS_MODULE);
}
static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
{
struct xdr_skb_reader desc = {
.skb = skb,
.offset = sizeof(rpc_fraghdr),
.count = skb->len - sizeof(rpc_fraghdr),
};
if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0)
return -1;
if (desc.count)
return -1;
return 0;
}
/**
* xs_local_data_ready - "data ready" callback for AF_LOCAL sockets
* @sk: socket with data to read
* @len: how much data to read
*
* Currently this assumes we can read the whole reply in a single gulp.
*/
static void xs_local_data_ready(struct sock *sk)
{
struct rpc_task *task;
struct rpc_xprt *xprt;
struct rpc_rqst *rovr;
struct sk_buff *skb;
int err, repsize, copied;
u32 _xid;
__be32 *xp;
read_lock_bh(&sk->sk_callback_lock);
dprintk("RPC: %s...\n", __func__);
xprt = xprt_from_sock(sk);
if (xprt == NULL)
goto out;
skb = skb_recv_datagram(sk, 0, 1, &err);
if (skb == NULL)
goto out;
repsize = skb->len - sizeof(rpc_fraghdr);
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(rpc_fraghdr), 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;
copied = rovr->rq_private_buf.buflen;
if (copied > repsize)
copied = repsize;
if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) {
dprintk("RPC: sk_buff copy failed\n");
goto out_unlock;
}
xprt_complete_rqst(task, copied);
out_unlock:
spin_unlock(&xprt->transport_lock);
dropit:
skb_free_datagram(sk, skb);
out:
read_unlock_bh(&sk->sk_callback_lock);
}
/**
* 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)
{
struct rpc_task *task;
struct rpc_xprt *xprt;
struct rpc_rqst *rovr;
struct sk_buff *skb;
int err, repsize, copied;
u32 _xid;
__be32 *xp;
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_lock_bh(&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;
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)) {
UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS);
goto out_unlock;
}
UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
xprt_adjust_cwnd(xprt, task, copied);
xprt_complete_rqst(task, copied);
out_unlock:
spin_unlock(&xprt->transport_lock);
dropit:
skb_free_datagram(sk, skb);
out:
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_unlock_bh(&sk->sk_callback_lock);
}
/*
* Helper function to force a TCP close if the server is sending
* junk and/or it has put us in CLOSE_WAIT
*/
static void xs_tcp_force_close(struct rpc_xprt *xprt)
{
xprt_force_disconnect(xprt);
}
static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
size_t len, used;
char *p;
p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
used = xdr_skb_read_bits(desc, p, len);
transport->tcp_offset += used;
if (used != len)
return;
transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
transport->tcp_flags |= TCP_RCV_LAST_FRAG;
else
transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
transport->tcp_offset = 0;
/* Sanity check of the record length */
if (unlikely(transport->tcp_reclen < 8)) {
dprintk("RPC: invalid TCP record fragment length\n");
xs_tcp_force_close(xprt);
return;
}
dprintk("RPC: reading TCP record fragment of length %d\n",
transport->tcp_reclen);
}
static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
{
if (transport->tcp_offset == transport->tcp_reclen) {
transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
transport->tcp_offset = 0;
if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
transport->tcp_flags |= TCP_RCV_COPY_XID;
transport->tcp_copied = 0;
}
}
}
static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
{
size_t len, used;
char *p;
len = sizeof(transport->tcp_xid) - transport->tcp_offset;
dprintk("RPC: reading XID (%Zu bytes)\n", len);
p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
used = xdr_skb_read_bits(desc, p, len);
transport->tcp_offset += used;
if (used != len)
return;
transport->tcp_flags &= ~TCP_RCV_COPY_XID;
nfs41: Process the RPC call direction Reading and storing the RPC direction is a three step process. 1. xs_tcp_read_calldir() reads the RPC direction, but it will not store it in the XDR buffer since the 'struct rpc_rqst' is not yet available. 2. The 'struct rpc_rqst' is obtained during the TCP_RCV_COPY_DATA state. This state need not necessarily be preceeded by the TCP_RCV_READ_CALLDIR. For example, we may be reading a continuation packet to a large reply. Therefore, we can't simply obtain the 'struct rpc_rqst' during the TCP_RCV_READ_CALLDIR state and assume it's available during TCP_RCV_COPY_DATA. This patch adds a new TCP_RCV_READ_CALLDIR flag to indicate the need to read the RPC direction. It then uses TCP_RCV_COPY_CALLDIR to indicate the RPC direction needs to be saved after the 'struct rpc_rqst' has been allocated. 3. The 'struct rpc_rqst' is obtained by the xs_tcp_read_data() helper functions. xs_tcp_read_common() then saves the RPC direction in the XDR buffer if TCP_RCV_COPY_CALLDIR is set. This will happen when we're reading the data immediately after the direction was read. xs_tcp_read_common() then clears this flag. [was nfs41: Skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Add RPC direction back into the XDR buffer] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Don't skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:22:54 +02:00
transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
transport->tcp_copied = 4;
dprintk("RPC: reading %s XID %08x\n",
(transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
: "request with",
ntohl(transport->tcp_xid));
xs_tcp_check_fraghdr(transport);
}
static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
struct xdr_skb_reader *desc)
{
size_t len, used;
u32 offset;
char *p;
/*
* We want transport->tcp_offset to be 8 at the end of this routine
* (4 bytes for the xid and 4 bytes for the call/reply flag).
* When this function is called for the first time,
* transport->tcp_offset is 4 (after having already read the xid).
*/
offset = transport->tcp_offset - sizeof(transport->tcp_xid);
len = sizeof(transport->tcp_calldir) - offset;
dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len);
p = ((char *) &transport->tcp_calldir) + offset;
used = xdr_skb_read_bits(desc, p, len);
transport->tcp_offset += used;
if (used != len)
return;
nfs41: Process the RPC call direction Reading and storing the RPC direction is a three step process. 1. xs_tcp_read_calldir() reads the RPC direction, but it will not store it in the XDR buffer since the 'struct rpc_rqst' is not yet available. 2. The 'struct rpc_rqst' is obtained during the TCP_RCV_COPY_DATA state. This state need not necessarily be preceeded by the TCP_RCV_READ_CALLDIR. For example, we may be reading a continuation packet to a large reply. Therefore, we can't simply obtain the 'struct rpc_rqst' during the TCP_RCV_READ_CALLDIR state and assume it's available during TCP_RCV_COPY_DATA. This patch adds a new TCP_RCV_READ_CALLDIR flag to indicate the need to read the RPC direction. It then uses TCP_RCV_COPY_CALLDIR to indicate the RPC direction needs to be saved after the 'struct rpc_rqst' has been allocated. 3. The 'struct rpc_rqst' is obtained by the xs_tcp_read_data() helper functions. xs_tcp_read_common() then saves the RPC direction in the XDR buffer if TCP_RCV_COPY_CALLDIR is set. This will happen when we're reading the data immediately after the direction was read. xs_tcp_read_common() then clears this flag. [was nfs41: Skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Add RPC direction back into the XDR buffer] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Don't skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:22:54 +02:00
transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
/*
* We don't yet have the XDR buffer, so we will write the calldir
* out after we get the buffer from the 'struct rpc_rqst'
*/
switch (ntohl(transport->tcp_calldir)) {
case RPC_REPLY:
transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
transport->tcp_flags |= TCP_RCV_COPY_DATA;
transport->tcp_flags |= TCP_RPC_REPLY;
break;
case RPC_CALL:
transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
transport->tcp_flags |= TCP_RCV_COPY_DATA;
transport->tcp_flags &= ~TCP_RPC_REPLY;
break;
default:
dprintk("RPC: invalid request message type\n");
xs_tcp_force_close(&transport->xprt);
}
xs_tcp_check_fraghdr(transport);
}
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
struct xdr_skb_reader *desc,
struct rpc_rqst *req)
{
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
struct xdr_buf *rcvbuf;
size_t len;
ssize_t r;
rcvbuf = &req->rq_private_buf;
nfs41: Process the RPC call direction Reading and storing the RPC direction is a three step process. 1. xs_tcp_read_calldir() reads the RPC direction, but it will not store it in the XDR buffer since the 'struct rpc_rqst' is not yet available. 2. The 'struct rpc_rqst' is obtained during the TCP_RCV_COPY_DATA state. This state need not necessarily be preceeded by the TCP_RCV_READ_CALLDIR. For example, we may be reading a continuation packet to a large reply. Therefore, we can't simply obtain the 'struct rpc_rqst' during the TCP_RCV_READ_CALLDIR state and assume it's available during TCP_RCV_COPY_DATA. This patch adds a new TCP_RCV_READ_CALLDIR flag to indicate the need to read the RPC direction. It then uses TCP_RCV_COPY_CALLDIR to indicate the RPC direction needs to be saved after the 'struct rpc_rqst' has been allocated. 3. The 'struct rpc_rqst' is obtained by the xs_tcp_read_data() helper functions. xs_tcp_read_common() then saves the RPC direction in the XDR buffer if TCP_RCV_COPY_CALLDIR is set. This will happen when we're reading the data immediately after the direction was read. xs_tcp_read_common() then clears this flag. [was nfs41: Skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Add RPC direction back into the XDR buffer] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Don't skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:22:54 +02:00
if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
/*
* Save the RPC direction in the XDR buffer
*/
memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
&transport->tcp_calldir,
sizeof(transport->tcp_calldir));
transport->tcp_copied += sizeof(transport->tcp_calldir);
nfs41: Process the RPC call direction Reading and storing the RPC direction is a three step process. 1. xs_tcp_read_calldir() reads the RPC direction, but it will not store it in the XDR buffer since the 'struct rpc_rqst' is not yet available. 2. The 'struct rpc_rqst' is obtained during the TCP_RCV_COPY_DATA state. This state need not necessarily be preceeded by the TCP_RCV_READ_CALLDIR. For example, we may be reading a continuation packet to a large reply. Therefore, we can't simply obtain the 'struct rpc_rqst' during the TCP_RCV_READ_CALLDIR state and assume it's available during TCP_RCV_COPY_DATA. This patch adds a new TCP_RCV_READ_CALLDIR flag to indicate the need to read the RPC direction. It then uses TCP_RCV_COPY_CALLDIR to indicate the RPC direction needs to be saved after the 'struct rpc_rqst' has been allocated. 3. The 'struct rpc_rqst' is obtained by the xs_tcp_read_data() helper functions. xs_tcp_read_common() then saves the RPC direction in the XDR buffer if TCP_RCV_COPY_CALLDIR is set. This will happen when we're reading the data immediately after the direction was read. xs_tcp_read_common() then clears this flag. [was nfs41: Skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Add RPC direction back into the XDR buffer] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Don't skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:22:54 +02:00
transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
}
len = desc->count;
if (len > transport->tcp_reclen - transport->tcp_offset) {
struct xdr_skb_reader my_desc;
len = transport->tcp_reclen - transport->tcp_offset;
memcpy(&my_desc, desc, sizeof(my_desc));
my_desc.count = len;
r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
&my_desc, xdr_skb_read_bits);
desc->count -= r;
desc->offset += r;
} else
r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
desc, xdr_skb_read_bits);
if (r > 0) {
transport->tcp_copied += r;
transport->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 TCP_RCV_COPY_DATA, so the request
* will not receive any additional updates,
* and time out.
* Any remaining data from this record will
* be discarded.
*/
transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
dprintk("RPC: XID %08x truncated request\n",
ntohl(transport->tcp_xid));
dprintk("RPC: xprt = %p, tcp_copied = %lu, "
"tcp_offset = %u, tcp_reclen = %u\n",
xprt, transport->tcp_copied,
transport->tcp_offset, transport->tcp_reclen);
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
return;
}
dprintk("RPC: XID %08x read %Zd bytes\n",
ntohl(transport->tcp_xid), r);
dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
"tcp_reclen = %u\n", xprt, transport->tcp_copied,
transport->tcp_offset, transport->tcp_reclen);
if (transport->tcp_copied == req->rq_private_buf.buflen)
transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
else if (transport->tcp_offset == transport->tcp_reclen) {
if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
}
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
}
/*
* Finds the request corresponding to the RPC xid and invokes the common
* tcp read code to read the data.
*/
static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
struct xdr_skb_reader *desc)
{
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
struct rpc_rqst *req;
dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid));
/* Find and lock the request corresponding to this xid */
spin_lock(&xprt->transport_lock);
req = xprt_lookup_rqst(xprt, transport->tcp_xid);
if (!req) {
dprintk("RPC: XID %08x request not found!\n",
ntohl(transport->tcp_xid));
spin_unlock(&xprt->transport_lock);
return -1;
}
xs_tcp_read_common(xprt, desc, req);
if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
xprt_complete_rqst(req->rq_task, transport->tcp_copied);
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
spin_unlock(&xprt->transport_lock);
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
return 0;
}
#if defined(CONFIG_SUNRPC_BACKCHANNEL)
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
/*
* Obtains an rpc_rqst previously allocated and invokes the common
* tcp read code to read the data. The result is placed in the callback
* queue.
* If we're unable to obtain the rpc_rqst we schedule the closing of the
* connection and return -1.
*/
static int xs_tcp_read_callback(struct rpc_xprt *xprt,
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
struct xdr_skb_reader *desc)
{
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
struct rpc_rqst *req;
/* Look up and lock the request corresponding to the given XID */
spin_lock(&xprt->transport_lock);
req = xprt_lookup_bc_request(xprt, transport->tcp_xid);
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
if (req == NULL) {
spin_unlock(&xprt->transport_lock);
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
printk(KERN_WARNING "Callback slot table overflowed\n");
xprt_force_disconnect(xprt);
return -1;
}
dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid));
xs_tcp_read_common(xprt, desc, req);
if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
xprt_complete_bc_request(req, transport->tcp_copied);
spin_unlock(&xprt->transport_lock);
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
return 0;
}
static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
struct xdr_skb_reader *desc)
{
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
return (transport->tcp_flags & TCP_RPC_REPLY) ?
xs_tcp_read_reply(xprt, desc) :
xs_tcp_read_callback(xprt, desc);
}
#else
static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
struct xdr_skb_reader *desc)
{
return xs_tcp_read_reply(xprt, desc);
}
#endif /* CONFIG_SUNRPC_BACKCHANNEL */
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
/*
* Read data off the transport. This can be either an RPC_CALL or an
* RPC_REPLY. Relay the processing to helper functions.
*/
static void xs_tcp_read_data(struct rpc_xprt *xprt,
struct xdr_skb_reader *desc)
{
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
if (_xs_tcp_read_data(xprt, desc) == 0)
xs_tcp_check_fraghdr(transport);
else {
/*
* The transport_lock protects the request handling.
* There's no need to hold it to update the tcp_flags.
*/
transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
}
}
static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
{
size_t len;
len = transport->tcp_reclen - transport->tcp_offset;
if (len > desc->count)
len = desc->count;
desc->count -= len;
desc->offset += len;
transport->tcp_offset += len;
dprintk("RPC: discarded %Zu bytes\n", len);
xs_tcp_check_fraghdr(transport);
}
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;
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
struct xdr_skb_reader desc = {
.skb = skb,
.offset = offset,
.count = len,
};
dprintk("RPC: xs_tcp_data_recv started\n");
do {
trace_xs_tcp_data_recv(transport);
/* Read in a new fragment marker if necessary */
/* Can we ever really expect to get completely empty fragments? */
if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
xs_tcp_read_fraghdr(xprt, &desc);
continue;
}
/* Read in the xid if necessary */
if (transport->tcp_flags & TCP_RCV_COPY_XID) {
xs_tcp_read_xid(transport, &desc);
continue;
}
/* Read in the call/reply flag */
nfs41: Process the RPC call direction Reading and storing the RPC direction is a three step process. 1. xs_tcp_read_calldir() reads the RPC direction, but it will not store it in the XDR buffer since the 'struct rpc_rqst' is not yet available. 2. The 'struct rpc_rqst' is obtained during the TCP_RCV_COPY_DATA state. This state need not necessarily be preceeded by the TCP_RCV_READ_CALLDIR. For example, we may be reading a continuation packet to a large reply. Therefore, we can't simply obtain the 'struct rpc_rqst' during the TCP_RCV_READ_CALLDIR state and assume it's available during TCP_RCV_COPY_DATA. This patch adds a new TCP_RCV_READ_CALLDIR flag to indicate the need to read the RPC direction. It then uses TCP_RCV_COPY_CALLDIR to indicate the RPC direction needs to be saved after the 'struct rpc_rqst' has been allocated. 3. The 'struct rpc_rqst' is obtained by the xs_tcp_read_data() helper functions. xs_tcp_read_common() then saves the RPC direction in the XDR buffer if TCP_RCV_COPY_CALLDIR is set. This will happen when we're reading the data immediately after the direction was read. xs_tcp_read_common() then clears this flag. [was nfs41: Skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Add RPC direction back into the XDR buffer] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: Don't skip past the RPC call direction] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:22:54 +02:00
if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
xs_tcp_read_calldir(transport, &desc);
continue;
}
/* Read in the request data */
if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
nfs41: New xs_tcp_read_data() Handles RPC replies and backchannel callbacks. Traditionally the NFS client has expected only RPC replies on its open connections. With NFSv4.1, callbacks can arrive over an existing open connection. This patch refactors the old xs_tcp_read_request() into an RPC reply handler: xs_tcp_read_reply(), a new backchannel callback handler: xs_tcp_read_callback(), and a common routine to read the data off the transport: xs_tcp_read_common(). The new xs_tcp_read_callback() queues callback requests onto a queue where the callback service (a separate thread) is listening for the processing. This patch incorporates work and suggestions from Rahul Iyer (iyer@netapp.com) and Benny Halevy (bhalevy@panasas.com). xs_tcp_read_callback() drops the connection when the number of expected callbacks is exceeded. Use xprt_force_disconnect(), ensuring tasks on the pending queue are awaken on disconnect. [nfs41: Keep track of RPC call/reply direction with a flag] [nfs41: Preallocate rpc_rqst receive buffer for handling callbacks] Signed-off-by: Ricardo Labiaga <ricardo.labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfs41: sunrpc: xs_tcp_read_callback() should use xprt_force_disconnect()] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [Moves embedded #ifdefs into #ifdef function blocks] Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
2009-04-01 15:23:02 +02:00
xs_tcp_read_data(xprt, &desc);
continue;
}
/* Skip over any trailing bytes on short reads */
xs_tcp_read_discard(transport, &desc);
} while (desc.count);
trace_xs_tcp_data_recv(transport);
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)
{
struct rpc_xprt *xprt;
read_descriptor_t rd_desc;
int read;
unsigned long total = 0;
dprintk("RPC: xs_tcp_data_ready...\n");
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_lock_bh(&sk->sk_callback_lock);
if (!(xprt = xprt_from_sock(sk))) {
read = 0;
goto out;
}
NFS/RPC: fix problems with reestablish_timeout and related code. [[resending with correct cc: - "vfs.kernel.org" just isn't right!]] xprt->reestablish_timeout is used to cause TCP connection attempts to back off if the connection fails so as not to hammer the network, but to still allow immediate connections when there is no reason to believe there is a problem. It is not used for the first connection (when transport->sock is NULL) but only on reconnects. It is currently set: a/ to 0 when xs_tcp_state_change finds a state of TCP_FIN_WAIT1 on the assumption that the client has closed the connection so the reconnect should be immediate when needed. b/ to at least XS_TCP_INIT_REEST_TO when xs_tcp_state_change detects TCP_CLOSING or TCP_CLOSE_WAIT on the assumption that the server closed the connection so a small delay at least is required. c/ as above when xs_tcp_state_change detects TCP_SYN_SENT, so that it is never 0 while a connection has been attempted, else the doubling will produce 0 and there will be no backoff. d/ to double is value (up to a limit) when delaying a connection, thus providing exponential backoff and e/ to XS_TCP_INIT_REEST_TO in xs_setup_tcp as simple initialisation. So you can see it is highly dependant on xs_tcp_state_change being called as expected. However experimental evidence shows that xs_tcp_state_change does not see all state changes. ("rpcdebug -m rpc trans" can help show what actually happens). Results show: TCP_ESTABLISHED is reported when a connection is made. TCP_SYN_SENT is never reported, so rule 'c' above is never effective. When the server closes the connection, TCP_CLOSE_WAIT and TCP_LAST_ACK *might* be reported, and TCP_CLOSE is always reported. This rule 'b' above will sometimes be effective, but not reliably. When the client closes the connection, it used to result in TCP_FIN_WAIT1, TCP_FIN_WAIT2, TCP_CLOSE. However since commit f75e674 (SUNRPC: Fix the problem of EADDRNOTAVAIL syslog floods on reconnect) we don't see *any* events on client-close. I think this is because xs_restore_old_callbacks is called to disconnect xs_tcp_state_change before the socket is closed. In any case, rule 'a' no longer applies. So all that is left are rule d, which successfully doubles the timeout which is never rest, and rule e which initialises the timeout. Even if the rules worked as expected, there would be a problem because a successful connection does not reset the timeout, so a sequence of events where the server closes the connection (e.g. during failover testing) will cause longer and longer timeouts with no good reason. This patch: - sets reestablish_timeout to 0 in xs_close thus effecting rule 'a' - sets it to 0 in xs_tcp_data_ready to ensure that a successful connection resets the timeout - sets it to at least XS_TCP_INIT_REEST_TO after it is doubled, thus effecting rule c I have not reimplemented rule b and the new version of rule c seems sufficient. I suspect other code in xs_tcp_data_ready needs to be revised as well. For example I don't think connect_cookie is being incremented as often as it should be. Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-09-23 20:36:37 +02:00
/* Any data means we had a useful conversation, so
* the we don't need to delay the next reconnect
*/
if (xprt->reestablish_timeout)
xprt->reestablish_timeout = 0;
/* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
rd_desc.arg.data = xprt;
do {
rd_desc.count = 65536;
read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
if (read > 0)
total += read;
} while (read > 0);
out:
trace_xs_tcp_data_ready(xprt, read, total);
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_unlock_bh(&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;
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_lock_bh(&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 sk_shutdown %d\n",
sk->sk_state, xprt_connected(xprt),
sock_flag(sk, SOCK_DEAD),
sock_flag(sk, SOCK_ZAPPED),
sk->sk_shutdown);
trace_rpc_socket_state_change(xprt, sk->sk_socket);
switch (sk->sk_state) {
case TCP_ESTABLISHED:
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
spin_lock(&xprt->transport_lock);
if (!xprt_test_and_set_connected(xprt)) {
struct sock_xprt *transport = container_of(xprt,
struct sock_xprt, xprt);
/* Reset TCP record info */
transport->tcp_offset = 0;
transport->tcp_reclen = 0;
transport->tcp_copied = 0;
transport->tcp_flags =
TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
xprt->connect_cookie++;
xprt_wake_pending_tasks(xprt, -EAGAIN);
}
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
spin_unlock(&xprt->transport_lock);
break;
case TCP_FIN_WAIT1:
/* The client initiated a shutdown of the socket */
xprt->connect_cookie++;
xprt->reestablish_timeout = 0;
set_bit(XPRT_CLOSING, &xprt->state);
smp_mb__before_atomic();
clear_bit(XPRT_CONNECTED, &xprt->state);
clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
smp_mb__after_atomic();
break;
case TCP_CLOSE_WAIT:
/* The server initiated a shutdown of the socket */
xprt->connect_cookie++;
clear_bit(XPRT_CONNECTED, &xprt->state);
xs_tcp_force_close(xprt);
case TCP_CLOSING:
/*
* If the server closed down the connection, make sure that
* we back off before reconnecting
*/
if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
break;
case TCP_LAST_ACK:
set_bit(XPRT_CLOSING, &xprt->state);
smp_mb__before_atomic();
clear_bit(XPRT_CONNECTED, &xprt->state);
smp_mb__after_atomic();
break;
case TCP_CLOSE:
xs_sock_mark_closed(xprt);
}
out:
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_unlock_bh(&sk->sk_callback_lock);
}
static void xs_write_space(struct sock *sk)
{
struct socket *sock;
struct rpc_xprt *xprt;
if (unlikely(!(sock = sk->sk_socket)))
return;
clear_bit(SOCK_NOSPACE, &sock->flags);
if (unlikely(!(xprt = xprt_from_sock(sk))))
return;
if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0)
return;
xprt_write_space(xprt);
}
/**
* 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)
{
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_lock_bh(&sk->sk_callback_lock);
/* from net/core/sock.c:sock_def_write_space */
if (sock_writeable(sk))
xs_write_space(sk);
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_unlock_bh(&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)
{
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_lock_bh(&sk->sk_callback_lock);
/* from net/core/stream.c:sk_stream_write_space */
if (sk_stream_is_writeable(sk))
xs_write_space(sk);
net: fix a lockdep splat We have for each socket : One spinlock (sk_slock.slock) One rwlock (sk_callback_lock) Possible scenarios are : (A) (this is used in net/sunrpc/xprtsock.c) read_lock(&sk->sk_callback_lock) (without blocking BH) <BH> spin_lock(&sk->sk_slock.slock); ... read_lock(&sk->sk_callback_lock); ... (B) write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) (C) spin_lock_bh(&sk->sk_slock) ... write_lock_bh(&sk->sk_callback_lock) stuff write_unlock_bh(&sk->sk_callback_lock) spin_unlock_bh(&sk->sk_slock) This (C) case conflicts with (A) : CPU1 [A] CPU2 [C] read_lock(callback_lock) <BH> spin_lock_bh(slock) <wait to spin_lock(slock)> <wait to write_lock_bh(callback_lock)> We have one problematic (C) use case in inet_csk_listen_stop() : local_bh_disable(); bh_lock_sock(child); // spin_lock_bh(&sk->sk_slock) WARN_ON(sock_owned_by_user(child)); ... sock_orphan(child); // write_lock_bh(&sk->sk_callback_lock) lockdep is not happy with this, as reported by Tetsuo Handa It seems only way to deal with this is to use read_lock_bh(callbacklock) everywhere. Thanks to Jarek for pointing a bug in my first attempt and suggesting this solution. Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> CC: Jarek Poplawski <jarkao2@gmail.com> Tested-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-09-22 14:43:39 +02:00
read_unlock_bh(&sk->sk_callback_lock);
}
static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
struct sock *sk = transport->inet;
if (transport->rcvsize) {
sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
}
if (transport->sndsize) {
sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
sk->sk_sndbuf = transport->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)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
transport->sndsize = 0;
if (sndsize)
transport->sndsize = sndsize + 1024;
transport->rcvsize = 0;
if (rcvsize)
transport->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_xprt *xprt, struct rpc_task *task)
{
xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
}
static unsigned short xs_get_random_port(void)
{
unsigned short range = xprt_max_resvport - xprt_min_resvport;
unsigned short rand = (unsigned short) prandom_u32() % range;
return rand + xprt_min_resvport;
}
/**
* xs_set_reuseaddr_port - set the socket's port and address reuse options
* @sock: socket
*
* Note that this function has to be called on all sockets that share the
* same port, and it must be called before binding.
*/
static void xs_sock_set_reuseport(struct socket *sock)
{
int opt = 1;
kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
(char *)&opt, sizeof(opt));
}
static unsigned short xs_sock_getport(struct socket *sock)
{
struct sockaddr_storage buf;
int buflen;
unsigned short port = 0;
if (kernel_getsockname(sock, (struct sockaddr *)&buf, &buflen) < 0)
goto out;
switch (buf.ss_family) {
case AF_INET6:
port = ntohs(((struct sockaddr_in6 *)&buf)->sin6_port);
break;
case AF_INET:
port = ntohs(((struct sockaddr_in *)&buf)->sin_port);
}
out:
return port;
}
/**
* xs_set_port - reset the port number in the remote endpoint address
* @xprt: generic transport
* @port: new port number
*
*/
static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
{
dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
rpc_set_port(xs_addr(xprt), port);
xs_update_peer_port(xprt);
}
static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock)
{
if (transport->srcport == 0)
transport->srcport = xs_sock_getport(sock);
}
static unsigned short xs_get_srcport(struct sock_xprt *transport)
{
unsigned short port = transport->srcport;
if (port == 0 && transport->xprt.resvport)
port = xs_get_random_port();
return port;
}
static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
{
if (transport->srcport != 0)
transport->srcport = 0;
if (!transport->xprt.resvport)
return 0;
if (port <= xprt_min_resvport || port > xprt_max_resvport)
return xprt_max_resvport;
return --port;
}
static int xs_bind(struct sock_xprt *transport, struct socket *sock)
{
struct sockaddr_storage myaddr;
int err, nloop = 0;
unsigned short port = xs_get_srcport(transport);
unsigned short last;
/*
* If we are asking for any ephemeral port (i.e. port == 0 &&
* transport->xprt.resvport == 0), don't bind. Let the local
* port selection happen implicitly when the socket is used
* (for example at connect time).
*
* This ensures that we can continue to establish TCP
* connections even when all local ephemeral ports are already
* a part of some TCP connection. This makes no difference
* for UDP sockets, but also doens't harm them.
*
* If we're asking for any reserved port (i.e. port == 0 &&
* transport->xprt.resvport == 1) xs_get_srcport above will
* ensure that port is non-zero and we will bind as needed.
*/
if (port == 0)
return 0;
memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
do {
rpc_set_port((struct sockaddr *)&myaddr, port);
err = kernel_bind(sock, (struct sockaddr *)&myaddr,
transport->xprt.addrlen);
if (err == 0) {
transport->srcport = port;
break;
}
last = port;
port = xs_next_srcport(transport, port);
if (port > last)
nloop++;
} while (err == -EADDRINUSE && nloop != 2);
if (myaddr.ss_family == AF_INET)
dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
&((struct sockaddr_in *)&myaddr)->sin_addr,
port, err ? "failed" : "ok", err);
else
dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
&((struct sockaddr_in6 *)&myaddr)->sin6_addr,
port, err ? "failed" : "ok", err);
return err;
}
/*
* We don't support autobind on AF_LOCAL sockets
*/
static void xs_local_rpcbind(struct rpc_task *task)
{
rcu_read_lock();
xprt_set_bound(rcu_dereference(task->tk_client->cl_xprt));
rcu_read_unlock();
}
static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
{
}
#ifdef CONFIG_DEBUG_LOCK_ALLOC
static struct lock_class_key xs_key[2];
static struct lock_class_key xs_slock_key[2];
static inline void xs_reclassify_socketu(struct socket *sock)
{
struct sock *sk = sock->sk;
sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
&xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
}
static inline void xs_reclassify_socket4(struct socket *sock)
{
struct sock *sk = sock->sk;
sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
&xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
}
static inline void xs_reclassify_socket6(struct socket *sock)
{
struct sock *sk = sock->sk;
sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
&xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
}
static inline void xs_reclassify_socket(int family, struct socket *sock)
{
WARN_ON_ONCE(sock_owned_by_user(sock->sk));
if (sock_owned_by_user(sock->sk))
return;
switch (family) {
case AF_LOCAL:
xs_reclassify_socketu(sock);
break;
case AF_INET:
xs_reclassify_socket4(sock);
break;
case AF_INET6:
xs_reclassify_socket6(sock);
break;
}
}
#else
static inline void xs_reclassify_socketu(struct socket *sock)
{
}
static inline void xs_reclassify_socket4(struct socket *sock)
{
}
static inline void xs_reclassify_socket6(struct socket *sock)
{
}
static inline void xs_reclassify_socket(int family, struct socket *sock)
{
}
#endif
static void xs_dummy_setup_socket(struct work_struct *work)
{
}
static struct socket *xs_create_sock(struct rpc_xprt *xprt,
struct sock_xprt *transport, int family, int type,
int protocol, bool reuseport)
{
struct socket *sock;
int err;
err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
if (err < 0) {
dprintk("RPC: can't create %d transport socket (%d).\n",
protocol, -err);
goto out;
}
xs_reclassify_socket(family, sock);
if (reuseport)
xs_sock_set_reuseport(sock);
err = xs_bind(transport, sock);
if (err) {
sock_release(sock);
goto out;
}
return sock;
out:
return ERR_PTR(err);
}
static int xs_local_finish_connecting(struct rpc_xprt *xprt,
struct socket *sock)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
xprt);
if (!transport->inet) {
struct sock *sk = sock->sk;
write_lock_bh(&sk->sk_callback_lock);
xs_save_old_callbacks(transport, sk);
sk->sk_user_data = xprt;
sk->sk_data_ready = xs_local_data_ready;
sk->sk_write_space = xs_udp_write_space;
sk->sk_error_report = xs_error_report;
sk->sk_allocation = GFP_ATOMIC;
xprt_clear_connected(xprt);
/* Reset to new socket */
transport->sock = sock;
transport->inet = sk;
write_unlock_bh(&sk->sk_callback_lock);
}
/* Tell the socket layer to start connecting... */
xprt->stat.connect_count++;
xprt->stat.connect_start = jiffies;
return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
}
/**
* xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
* @xprt: RPC transport to connect
* @transport: socket transport to connect
* @create_sock: function to create a socket of the correct type
*/
static int xs_local_setup_socket(struct sock_xprt *transport)
{
struct rpc_xprt *xprt = &transport->xprt;
struct socket *sock;
int status = -EIO;
status = __sock_create(xprt->xprt_net, AF_LOCAL,
SOCK_STREAM, 0, &sock, 1);
if (status < 0) {
dprintk("RPC: can't create AF_LOCAL "
"transport socket (%d).\n", -status);
goto out;
}
xs_reclassify_socketu(sock);
dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
status = xs_local_finish_connecting(xprt, sock);
trace_rpc_socket_connect(xprt, sock, status);
switch (status) {
case 0:
dprintk("RPC: xprt %p connected to %s\n",
xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
xprt_set_connected(xprt);
case -ENOBUFS:
break;
case -ENOENT:
dprintk("RPC: xprt %p: socket %s does not exist\n",
xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
break;
case -ECONNREFUSED:
dprintk("RPC: xprt %p: connection refused for %s\n",
xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
break;
default:
printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
__func__, -status,
xprt->address_strings[RPC_DISPLAY_ADDR]);
}
out:
xprt_clear_connecting(xprt);
xprt_wake_pending_tasks(xprt, status);
return status;
}
Merge branch 'for-3.9' of git://linux-nfs.org/~bfields/linux Pull nfsd changes from J Bruce Fields: "Miscellaneous bugfixes, plus: - An overhaul of the DRC cache by Jeff Layton. The main effect is just to make it larger. This decreases the chances of intermittent errors especially in the UDP case. But we'll need to watch for any reports of performance regressions. - Containerized nfsd: with some limitations, we now support per-container nfs-service, thanks to extensive work from Stanislav Kinsbursky over the last year." Some notes about conflicts, since there were *two* non-data semantic conflicts here: - idr_remove_all() had been added by a memory leak fix, but has since become deprecated since idr_destroy() does it for us now. - xs_local_connect() had been added by this branch to make AF_LOCAL connections be synchronous, but in the meantime Trond had changed the calling convention in order to avoid a RCU dereference. There were a couple of more obvious actual source-level conflicts due to the hlist traversal changes and one just due to code changes next to each other, but those were trivial. * 'for-3.9' of git://linux-nfs.org/~bfields/linux: (49 commits) SUNRPC: make AF_LOCAL connect synchronous nfsd: fix compiler warning about ambiguous types in nfsd_cache_csum svcrpc: fix rpc server shutdown races svcrpc: make svc_age_temp_xprts enqueue under sv_lock lockd: nlmclnt_reclaim(): avoid stack overflow nfsd: enable NFSv4 state in containers nfsd: disable usermode helper client tracker in container nfsd: use proper net while reading "exports" file nfsd: containerize NFSd filesystem nfsd: fix comments on nfsd_cache_lookup SUNRPC: move cache_detail->cache_request callback call to cache_read() SUNRPC: remove "cache_request" argument in sunrpc_cache_pipe_upcall() function SUNRPC: rework cache upcall logic SUNRPC: introduce cache_detail->cache_request callback NFS: simplify and clean cache library NFS: use SUNRPC cache creation and destruction helper for DNS cache nfsd4: free_stid can be static nfsd: keep a checksum of the first 256 bytes of request sunrpc: trim off trailing checksum before returning decrypted or integrity authenticated buffer sunrpc: fix comment in struct xdr_buf definition ...
2013-03-01 03:02:55 +01:00
static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
int ret;
if (RPC_IS_ASYNC(task)) {
/*
* We want the AF_LOCAL connect to be resolved in the
* filesystem namespace of the process making the rpc
* call. Thus we connect synchronously.
*
* If we want to support asynchronous AF_LOCAL calls,
* we'll need to figure out how to pass a namespace to
* connect.
*/
rpc_exit(task, -ENOTCONN);
return;
}
ret = xs_local_setup_socket(transport);
if (ret && !RPC_IS_SOFTCONN(task))
msleep_interruptible(15000);
}
#ifdef CONFIG_SUNRPC_SWAP
static void xs_set_memalloc(struct rpc_xprt *xprt)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
xprt);
if (xprt->swapper)
sk_set_memalloc(transport->inet);
}
/**
* xs_swapper - Tag this transport as being used for swap.
* @xprt: transport to tag
* @enable: enable/disable
*
*/
int xs_swapper(struct rpc_xprt *xprt, int enable)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
xprt);
int err = 0;
if (enable) {
xprt->swapper++;
xs_set_memalloc(xprt);
} else if (xprt->swapper) {
xprt->swapper--;
sk_clear_memalloc(transport->inet);
}
return err;
}
EXPORT_SYMBOL_GPL(xs_swapper);
#else
static void xs_set_memalloc(struct rpc_xprt *xprt)
{
}
#endif
static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
if (!transport->inet) {
struct sock *sk = sock->sk;
write_lock_bh(&sk->sk_callback_lock);
xs_save_old_callbacks(transport, sk);
sk->sk_user_data = xprt;
sk->sk_data_ready = xs_udp_data_ready;
sk->sk_write_space = xs_udp_write_space;
sk->sk_allocation = GFP_ATOMIC;
xprt_set_connected(xprt);
/* Reset to new socket */
transport->sock = sock;
transport->inet = sk;
xs_set_memalloc(xprt);
write_unlock_bh(&sk->sk_callback_lock);
}
xs_udp_do_set_buffer_size(xprt);
}
static void xs_udp_setup_socket(struct work_struct *work)
{
struct sock_xprt *transport =
container_of(work, struct sock_xprt, connect_worker.work);
struct rpc_xprt *xprt = &transport->xprt;
struct socket *sock = transport->sock;
int status = -EIO;
sock = xs_create_sock(xprt, transport,
xs_addr(xprt)->sa_family, SOCK_DGRAM,
IPPROTO_UDP, false);
if (IS_ERR(sock))
goto out;
dprintk("RPC: worker connecting xprt %p via %s to "
"%s (port %s)\n", xprt,
xprt->address_strings[RPC_DISPLAY_PROTO],
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PORT]);
xs_udp_finish_connecting(xprt, sock);
trace_rpc_socket_connect(xprt, sock, 0);
status = 0;
out:
xprt_unlock_connect(xprt, transport);
xprt_clear_connecting(xprt);
xprt_wake_pending_tasks(xprt, status);
}
static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
int ret = -ENOTCONN;
if (!transport->inet) {
struct sock *sk = sock->sk;
unsigned int keepidle = xprt->timeout->to_initval / HZ;
unsigned int keepcnt = xprt->timeout->to_retries + 1;
unsigned int opt_on = 1;
/* TCP Keepalive options */
kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
(char *)&opt_on, sizeof(opt_on));
kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
(char *)&keepidle, sizeof(keepidle));
kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
(char *)&keepidle, sizeof(keepidle));
kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
(char *)&keepcnt, sizeof(keepcnt));
write_lock_bh(&sk->sk_callback_lock);
xs_save_old_callbacks(transport, sk);
sk->sk_user_data = xprt;
sk->sk_data_ready = xs_tcp_data_ready;
sk->sk_state_change = xs_tcp_state_change;
sk->sk_write_space = xs_tcp_write_space;
sk->sk_error_report = xs_error_report;
sk->sk_allocation = GFP_ATOMIC;
/* socket options */
sock_reset_flag(sk, SOCK_LINGER);
tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
xprt_clear_connected(xprt);
/* Reset to new socket */
transport->sock = sock;
transport->inet = sk;
write_unlock_bh(&sk->sk_callback_lock);
}
if (!xprt_bound(xprt))
goto out;
xs_set_memalloc(xprt);
/* Tell the socket layer to start connecting... */
xprt->stat.connect_count++;
xprt->stat.connect_start = jiffies;
ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
switch (ret) {
case 0:
xs_set_srcport(transport, sock);
case -EINPROGRESS:
/* SYN_SENT! */
if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
}
out:
return ret;
}
/**
* xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
* @xprt: RPC transport to connect
* @transport: socket transport to connect
* @create_sock: function to create a socket of the correct type
*
* Invoked by a work queue tasklet.
*/
static void xs_tcp_setup_socket(struct work_struct *work)
{
struct sock_xprt *transport =
container_of(work, struct sock_xprt, connect_worker.work);
struct socket *sock = transport->sock;
struct rpc_xprt *xprt = &transport->xprt;
int status = -EIO;
if (!sock) {
sock = xs_create_sock(xprt, transport,
xs_addr(xprt)->sa_family, SOCK_STREAM,
IPPROTO_TCP, true);
if (IS_ERR(sock)) {
status = PTR_ERR(sock);
goto out;
}
}
dprintk("RPC: worker connecting xprt %p via %s to "
"%s (port %s)\n", xprt,
xprt->address_strings[RPC_DISPLAY_PROTO],
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PORT]);
status = xs_tcp_finish_connecting(xprt, sock);
trace_rpc_socket_connect(xprt, sock, status);
dprintk("RPC: %p connect status %d connected %d sock state %d\n",
xprt, -status, xprt_connected(xprt),
sock->sk->sk_state);
switch (status) {
default:
printk("%s: connect returned unhandled error %d\n",
__func__, status);
case -EADDRNOTAVAIL:
/* We're probably in TIME_WAIT. Get rid of existing socket,
* and retry
*/
xs_tcp_force_close(xprt);
break;
case 0:
case -EINPROGRESS:
case -EALREADY:
xprt_unlock_connect(xprt, transport);
xprt_clear_connecting(xprt);
return;
case -EINVAL:
/* Happens, for instance, if the user specified a link
* local IPv6 address without a scope-id.
*/
case -ECONNREFUSED:
case -ECONNRESET:
case -ENETUNREACH:
case -EADDRINUSE:
case -ENOBUFS:
/* retry with existing socket, after a delay */
xs_tcp_force_close(xprt);
goto out;
}
status = -EAGAIN;
out:
xprt_unlock_connect(xprt, transport);
xprt_clear_connecting(xprt);
xprt_wake_pending_tasks(xprt, status);
}
/**
* xs_connect - connect a socket to a remote endpoint
* @xprt: pointer to transport structure
* @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_xprt *xprt, struct rpc_task *task)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport));
/* Start by resetting any existing state */
xs_reset_transport(transport);
if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {
dprintk("RPC: xs_connect delayed xprt %p for %lu "
"seconds\n",
xprt, xprt->reestablish_timeout / HZ);
queue_delayed_work(rpciod_workqueue,
&transport->connect_worker,
xprt->reestablish_timeout);
xprt->reestablish_timeout <<= 1;
NFS/RPC: fix problems with reestablish_timeout and related code. [[resending with correct cc: - "vfs.kernel.org" just isn't right!]] xprt->reestablish_timeout is used to cause TCP connection attempts to back off if the connection fails so as not to hammer the network, but to still allow immediate connections when there is no reason to believe there is a problem. It is not used for the first connection (when transport->sock is NULL) but only on reconnects. It is currently set: a/ to 0 when xs_tcp_state_change finds a state of TCP_FIN_WAIT1 on the assumption that the client has closed the connection so the reconnect should be immediate when needed. b/ to at least XS_TCP_INIT_REEST_TO when xs_tcp_state_change detects TCP_CLOSING or TCP_CLOSE_WAIT on the assumption that the server closed the connection so a small delay at least is required. c/ as above when xs_tcp_state_change detects TCP_SYN_SENT, so that it is never 0 while a connection has been attempted, else the doubling will produce 0 and there will be no backoff. d/ to double is value (up to a limit) when delaying a connection, thus providing exponential backoff and e/ to XS_TCP_INIT_REEST_TO in xs_setup_tcp as simple initialisation. So you can see it is highly dependant on xs_tcp_state_change being called as expected. However experimental evidence shows that xs_tcp_state_change does not see all state changes. ("rpcdebug -m rpc trans" can help show what actually happens). Results show: TCP_ESTABLISHED is reported when a connection is made. TCP_SYN_SENT is never reported, so rule 'c' above is never effective. When the server closes the connection, TCP_CLOSE_WAIT and TCP_LAST_ACK *might* be reported, and TCP_CLOSE is always reported. This rule 'b' above will sometimes be effective, but not reliably. When the client closes the connection, it used to result in TCP_FIN_WAIT1, TCP_FIN_WAIT2, TCP_CLOSE. However since commit f75e674 (SUNRPC: Fix the problem of EADDRNOTAVAIL syslog floods on reconnect) we don't see *any* events on client-close. I think this is because xs_restore_old_callbacks is called to disconnect xs_tcp_state_change before the socket is closed. In any case, rule 'a' no longer applies. So all that is left are rule d, which successfully doubles the timeout which is never rest, and rule e which initialises the timeout. Even if the rules worked as expected, there would be a problem because a successful connection does not reset the timeout, so a sequence of events where the server closes the connection (e.g. during failover testing) will cause longer and longer timeouts with no good reason. This patch: - sets reestablish_timeout to 0 in xs_close thus effecting rule 'a' - sets it to 0 in xs_tcp_data_ready to ensure that a successful connection resets the timeout - sets it to at least XS_TCP_INIT_REEST_TO after it is doubled, thus effecting rule c I have not reimplemented rule b and the new version of rule c seems sufficient. I suspect other code in xs_tcp_data_ready needs to be revised as well. For example I don't think connect_cookie is being incremented as often as it should be. Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2009-09-23 20:36:37 +02:00
if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
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);
queue_delayed_work(rpciod_workqueue,
&transport->connect_worker, 0);
}
}
/**
* xs_local_print_stats - display AF_LOCAL socket-specifc stats
* @xprt: rpc_xprt struct containing statistics
* @seq: output file
*
*/
static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
{
long idle_time = 0;
if (xprt_connected(xprt))
idle_time = (long)(jiffies - xprt->last_used) / HZ;
seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
"%llu %llu %lu %llu %llu\n",
xprt->stat.bind_count,
xprt->stat.connect_count,
xprt->stat.connect_time,
idle_time,
xprt->stat.sends,
xprt->stat.recvs,
xprt->stat.bad_xids,
xprt->stat.req_u,
xprt->stat.bklog_u,
xprt->stat.max_slots,
xprt->stat.sending_u,
xprt->stat.pending_u);
}
/**
* xs_udp_print_stats - display UDP socket-specifc stats
* @xprt: rpc_xprt struct containing statistics
* @seq: output file
*
*/
static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
"%lu %llu %llu\n",
transport->srcport,
xprt->stat.bind_count,
xprt->stat.sends,
xprt->stat.recvs,
xprt->stat.bad_xids,
xprt->stat.req_u,
xprt->stat.bklog_u,
xprt->stat.max_slots,
xprt->stat.sending_u,
xprt->stat.pending_u);
}
/**
* xs_tcp_print_stats - display TCP socket-specifc stats
* @xprt: rpc_xprt struct containing statistics
* @seq: output file
*
*/
static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
{
struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
long idle_time = 0;
if (xprt_connected(xprt))
idle_time = (long)(jiffies - xprt->last_used) / HZ;
seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
"%llu %llu %lu %llu %llu\n",
transport->srcport,
xprt->stat.bind_count,
xprt->stat.connect_count,
xprt->stat.connect_time,
idle_time,
xprt->stat.sends,
xprt->stat.recvs,
xprt->stat.bad_xids,
xprt->stat.req_u,
xprt->stat.bklog_u,
xprt->stat.max_slots,
xprt->stat.sending_u,
xprt->stat.pending_u);
}
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
/*
* Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
* we allocate pages instead doing a kmalloc like rpc_malloc is because we want
* to use the server side send routines.
*/
static void *bc_malloc(struct rpc_task *task, size_t size)
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
{
struct page *page;
struct rpc_buffer *buf;
WARN_ON_ONCE(size > PAGE_SIZE - sizeof(struct rpc_buffer));
if (size > PAGE_SIZE - sizeof(struct rpc_buffer))
return NULL;
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
page = alloc_page(GFP_KERNEL);
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
if (!page)
return NULL;
buf = page_address(page);
buf->len = PAGE_SIZE;
return buf->data;
}
/*
* Free the space allocated in the bc_alloc routine
*/
static void bc_free(void *buffer)
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
{
struct rpc_buffer *buf;
if (!buffer)
return;
buf = container_of(buffer, struct rpc_buffer, data);
free_page((unsigned long)buf);
}
/*
* Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
* held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
*/
static int bc_sendto(struct rpc_rqst *req)
{
int len;
struct xdr_buf *xbufp = &req->rq_snd_buf;
struct rpc_xprt *xprt = req->rq_xprt;
struct sock_xprt *transport =
container_of(xprt, struct sock_xprt, xprt);
struct socket *sock = transport->sock;
unsigned long headoff;
unsigned long tailoff;
xs_encode_stream_record_marker(xbufp);
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
len = svc_send_common(sock, xbufp,
virt_to_page(xbufp->head[0].iov_base), headoff,
xbufp->tail[0].iov_base, tailoff);
if (len != xbufp->len) {
printk(KERN_NOTICE "Error sending entire callback!\n");
len = -EAGAIN;
}
return len;
}
/*
* The send routine. Borrows from svc_send
*/
static int bc_send_request(struct rpc_task *task)
{
struct rpc_rqst *req = task->tk_rqstp;
struct svc_xprt *xprt;
u32 len;
dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
/*
* Get the server socket associated with this callback xprt
*/
xprt = req->rq_xprt->bc_xprt;
/*
* Grab the mutex to serialize data as the connection is shared
* with the fore channel
*/
if (!mutex_trylock(&xprt->xpt_mutex)) {
rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
if (!mutex_trylock(&xprt->xpt_mutex))
return -EAGAIN;
rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
}
if (test_bit(XPT_DEAD, &xprt->xpt_flags))
len = -ENOTCONN;
else
len = bc_sendto(req);
mutex_unlock(&xprt->xpt_mutex);
if (len > 0)
len = 0;
return len;
}
/*
* The close routine. Since this is client initiated, we do nothing
*/
static void bc_close(struct rpc_xprt *xprt)
{
}
/*
* The xprt destroy routine. Again, because this connection is client
* initiated, we do nothing
*/
static void bc_destroy(struct rpc_xprt *xprt)
{
dprintk("RPC: bc_destroy xprt %p\n", xprt);
xs_xprt_free(xprt);
module_put(THIS_MODULE);
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
}
static struct rpc_xprt_ops xs_local_ops = {
.reserve_xprt = xprt_reserve_xprt,
.release_xprt = xs_tcp_release_xprt,
.alloc_slot = xprt_alloc_slot,
.rpcbind = xs_local_rpcbind,
.set_port = xs_local_set_port,
.connect = xs_local_connect,
.buf_alloc = rpc_malloc,
.buf_free = rpc_free,
.send_request = xs_local_send_request,
.set_retrans_timeout = xprt_set_retrans_timeout_def,
.close = xs_close,
.destroy = xs_destroy,
.print_stats = xs_local_print_stats,
};
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,
.alloc_slot = xprt_alloc_slot,
.rpcbind = rpcb_getport_async,
.set_port = xs_set_port,
.connect = xs_connect,
.buf_alloc = rpc_malloc,
.buf_free = rpc_free,
.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,
.print_stats = xs_udp_print_stats,
};
static struct rpc_xprt_ops xs_tcp_ops = {
.reserve_xprt = xprt_reserve_xprt,
.release_xprt = xs_tcp_release_xprt,
.alloc_slot = xprt_lock_and_alloc_slot,
.rpcbind = rpcb_getport_async,
.set_port = xs_set_port,
.connect = xs_connect,
.buf_alloc = rpc_malloc,
.buf_free = rpc_free,
.send_request = xs_tcp_send_request,
.set_retrans_timeout = xprt_set_retrans_timeout_def,
.close = xs_tcp_shutdown,
.destroy = xs_destroy,
.print_stats = xs_tcp_print_stats,
};
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
/*
* The rpc_xprt_ops for the server backchannel
*/
static struct rpc_xprt_ops bc_tcp_ops = {
.reserve_xprt = xprt_reserve_xprt,
.release_xprt = xprt_release_xprt,
.alloc_slot = xprt_alloc_slot,
nfsd41: sunrpc: Added rpc server-side backchannel handling When the call direction is a reply, copy the xid and call direction into the req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns rpc_garbage. Signed-off-by: Rahul Iyer <iyer@netapp.com> Signed-off-by: Mike Sager <sager@netapp.com> Signed-off-by: Marc Eshel <eshel@almaden.ibm.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Andy Adamson <andros@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [get rid of CONFIG_NFSD_V4_1] [sunrpc: refactoring of svc_tcp_recvfrom] [nfsd41: sunrpc: create common send routine for the fore and the back channels] [nfsd41: sunrpc: Use free_page() to free server backchannel pages] [nfsd41: sunrpc: Document server backchannel locking] [nfsd41: sunrpc: remove bc_connect_worker()] [nfsd41: sunrpc: Define xprt_server_backchannel()[ [nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions] [nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()] [nfsd41: sunrpc: Don't auto close the server backchannel connection] [nfsd41: sunrpc: Remove unused functions] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> [nfsd41: change bc_sock to bc_xprt] [nfsd41: sunrpc: move struct rpc_buffer def into a common header file] [nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex] [removed cosmetic changes] Signed-off-by: Benny Halevy <bhalevy@panasas.com> [sunrpc: add new xprt class for nfsv4.1 backchannel] [sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> [reverted more cosmetic leftovers] [got rid of xprt_server_backchannel] [separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"] Signed-off-by: Benny Halevy <bhalevy@panasas.com> Cc: Trond Myklebust <trond.myklebust@netapp.com> [sunrpc: change idle timeout value for the backchannel] Signed-off-by: Alexandros Batsakis <batsakis@netapp.com> Signed-off-by: Benny Halevy <bhalevy@panasas.com> Acked-by: Trond Myklebust <trond.myklebust@netapp.com> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
2009-09-10 16:32:28 +02:00
.buf_alloc = bc_malloc,
.buf_free = bc_free,
.send_request = bc_send_request,
.set_retrans_timeout = xprt_set_retrans_timeout_def,
.close = bc_close,
.destroy = bc_destroy,
.print_stats = xs_tcp_print_stats,
};
SUNRPC: Properly initialize sock_xprt.srcaddr in all cases The source address field in the transport's sock_xprt is initialized ONLY IF the RPC application passed a pointer to a source address during the call to rpc_create(). However, xs_bind() subsequently uses the value of this field without regard to whether the source address was initialized during transport creation or not. So far we've been lucky: the uninitialized value of this field is zeroes. xs_bind(), until recently, used only the sin[6]_addr field in this sockaddr, and all zeroes is a valid value for this: it means ANYADDR. This is a happy coincidence. However, xs_bind() now wants to use the sa_family field as well, and expects it to be initialized to something other than zero. Therefore, the source address sockaddr field should be fully initialized at transport create time in _every_ case, not just when the RPC application wants to use a specific bind address. Bruce added a workaround for this missing initialization by adjusting commit 6bc9638a, but the "right" way to do this is to ensure that the source address sockaddr is always correctly initialized from the get-go. This patch doesn't introduce a behavior change. It's simply a clean-up of Bruce's fix, to prevent future problems of this kind. It may look like overkill, but a) it clearly documents the default initial value of this field, b) it doesn't assume that the sockaddr_storage memory is first initialized to any particular value, and c) it will fail verbosely if some unknown address family is passed in Originally introduced by commit d3bc9a1d. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
2010-10-20 17:53:01 +02:00
static int xs_init_anyaddr(const int family, struct sockaddr *sap)
{
static const struct sockaddr_in sin = {
.sin_family = AF_INET,
.sin_addr.s_addr = htonl(INADDR_ANY),
};
static const struct sockaddr_in6 sin6 = {
.sin6_family = AF_INET6,
.sin6_addr = IN6ADDR_ANY_INIT,
};
switch (family) {
case AF_LOCAL:
break;
SUNRPC: Properly initialize sock_xprt.srcaddr in all cases The source address field in the transport's sock_xprt is initialized ONLY IF the RPC application passed a pointer to a source address during the call to rpc_create(). However, xs_bind() subsequently uses the value of this field without regard to whether the source address was initialized during transport creation or not. So far we've been lucky: the uninitialized value of this field is zeroes. xs_bind(), until recently, used only the sin[6]_addr field in this sockaddr, and all zeroes is a valid value for this: it means ANYADDR. This is a happy coincidence. However, xs_bind() now wants to use the sa_family field as well, and expects it to be initialized to something other than zero. Therefore, the source address sockaddr field should be fully initialized at transport create time in _every_ case, not just when the RPC application wants to use a specific bind address. Bruce added a workaround for this missing initialization by adjusting commit 6bc9638a, but the "right" way to do this is to ensure that the source address sockaddr is always correctly initialized from the get-go. This patch doesn't introduce a behavior change. It's simply a clean-up of Bruce's fix, to prevent future problems of this kind. It may look like overkill, but a) it clearly documents the default initial value of this field, b) it doesn't assume that the sockaddr_storage memory is first initialized to any particular value, and c) it will fail verbosely if some unknown address family is passed in Originally introduced by commit d3bc9a1d. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
2010-10-20 17:53:01 +02:00
case AF_INET:
memcpy(sap, &sin, sizeof(sin));
break;
case AF_INET6:
memcpy(sap, &sin6, sizeof(sin6));
break;
default:
dprintk("RPC: %s: Bad address family\n", __func__);
return -EAFNOSUPPORT;
}
return 0;
}
static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
unsigned int slot_table_size,
unsigned int max_slot_table_size)
{
struct rpc_xprt *xprt;
struct sock_xprt *new;
if (args->addrlen > sizeof(xprt->addr)) {
dprintk("RPC: xs_setup_xprt: address too large\n");
return ERR_PTR(-EBADF);
}
xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
max_slot_table_size);
if (xprt == NULL) {
dprintk("RPC: xs_setup_xprt: couldn't allocate "
"rpc_xprt\n");
return ERR_PTR(-ENOMEM);
}
new = container_of(xprt, struct sock_xprt, xprt);
memcpy(&xprt->addr, args->dstaddr, args->addrlen);
xprt->addrlen = args->addrlen;
if (args->srcaddr)
memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
SUNRPC: Properly initialize sock_xprt.srcaddr in all cases The source address field in the transport's sock_xprt is initialized ONLY IF the RPC application passed a pointer to a source address during the call to rpc_create(). However, xs_bind() subsequently uses the value of this field without regard to whether the source address was initialized during transport creation or not. So far we've been lucky: the uninitialized value of this field is zeroes. xs_bind(), until recently, used only the sin[6]_addr field in this sockaddr, and all zeroes is a valid value for this: it means ANYADDR. This is a happy coincidence. However, xs_bind() now wants to use the sa_family field as well, and expects it to be initialized to something other than zero. Therefore, the source address sockaddr field should be fully initialized at transport create time in _every_ case, not just when the RPC application wants to use a specific bind address. Bruce added a workaround for this missing initialization by adjusting commit 6bc9638a, but the "right" way to do this is to ensure that the source address sockaddr is always correctly initialized from the get-go. This patch doesn't introduce a behavior change. It's simply a clean-up of Bruce's fix, to prevent future problems of this kind. It may look like overkill, but a) it clearly documents the default initial value of this field, b) it doesn't assume that the sockaddr_storage memory is first initialized to any particular value, and c) it will fail verbosely if some unknown address family is passed in Originally introduced by commit d3bc9a1d. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
2010-10-20 17:53:01 +02:00
else {
int err;
err = xs_init_anyaddr(args->dstaddr->sa_family,
(struct sockaddr *)&new->srcaddr);
if (err != 0) {
xprt_free(xprt);
SUNRPC: Properly initialize sock_xprt.srcaddr in all cases The source address field in the transport's sock_xprt is initialized ONLY IF the RPC application passed a pointer to a source address during the call to rpc_create(). However, xs_bind() subsequently uses the value of this field without regard to whether the source address was initialized during transport creation or not. So far we've been lucky: the uninitialized value of this field is zeroes. xs_bind(), until recently, used only the sin[6]_addr field in this sockaddr, and all zeroes is a valid value for this: it means ANYADDR. This is a happy coincidence. However, xs_bind() now wants to use the sa_family field as well, and expects it to be initialized to something other than zero. Therefore, the source address sockaddr field should be fully initialized at transport create time in _every_ case, not just when the RPC application wants to use a specific bind address. Bruce added a workaround for this missing initialization by adjusting commit 6bc9638a, but the "right" way to do this is to ensure that the source address sockaddr is always correctly initialized from the get-go. This patch doesn't introduce a behavior change. It's simply a clean-up of Bruce's fix, to prevent future problems of this kind. It may look like overkill, but a) it clearly documents the default initial value of this field, b) it doesn't assume that the sockaddr_storage memory is first initialized to any particular value, and c) it will fail verbosely if some unknown address family is passed in Originally introduced by commit d3bc9a1d. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
2010-10-20 17:53:01 +02:00
return ERR_PTR(err);
}
SUNRPC: Properly initialize sock_xprt.srcaddr in all cases The source address field in the transport's sock_xprt is initialized ONLY IF the RPC application passed a pointer to a source address during the call to rpc_create(). However, xs_bind() subsequently uses the value of this field without regard to whether the source address was initialized during transport creation or not. So far we've been lucky: the uninitialized value of this field is zeroes. xs_bind(), until recently, used only the sin[6]_addr field in this sockaddr, and all zeroes is a valid value for this: it means ANYADDR. This is a happy coincidence. However, xs_bind() now wants to use the sa_family field as well, and expects it to be initialized to something other than zero. Therefore, the source address sockaddr field should be fully initialized at transport create time in _every_ case, not just when the RPC application wants to use a specific bind address. Bruce added a workaround for this missing initialization by adjusting commit 6bc9638a, but the "right" way to do this is to ensure that the source address sockaddr is always correctly initialized from the get-go. This patch doesn't introduce a behavior change. It's simply a clean-up of Bruce's fix, to prevent future problems of this kind. It may look like overkill, but a) it clearly documents the default initial value of this field, b) it doesn't assume that the sockaddr_storage memory is first initialized to any particular value, and c) it will fail verbosely if some unknown address family is passed in Originally introduced by commit d3bc9a1d. Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
2010-10-20 17:53:01 +02:00
}
return xprt;
}
static const struct rpc_timeout xs_local_default_timeout = {
.to_initval = 10 * HZ,
.to_maxval = 10 * HZ,
.to_retries = 2,
};
/**
* xs_setup_local - Set up transport to use an AF_LOCAL socket
* @args: rpc transport creation arguments
*
* AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
*/
static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
{
struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
struct sock_xprt *transport;
struct rpc_xprt *xprt;
struct rpc_xprt *ret;
xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
xprt_max_tcp_slot_table_entries);
if (IS_ERR(xprt))
return xprt;
transport = container_of(xprt, struct sock_xprt, xprt);
xprt->prot = 0;
xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
xprt->bind_timeout = XS_BIND_TO;
xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
xprt->idle_timeout = XS_IDLE_DISC_TO;
xprt->ops = &xs_local_ops;
xprt->timeout = &xs_local_default_timeout;
INIT_DELAYED_WORK(&transport->connect_worker,
xs_dummy_setup_socket);
switch (sun->sun_family) {
case AF_LOCAL:
if (sun->sun_path[0] != '/') {
dprintk("RPC: bad AF_LOCAL address: %s\n",
sun->sun_path);
ret = ERR_PTR(-EINVAL);
goto out_err;
}
xprt_set_bound(xprt);
xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
ret = ERR_PTR(xs_local_setup_socket(transport));
if (ret)
goto out_err;
break;
default:
ret = ERR_PTR(-EAFNOSUPPORT);
goto out_err;
}
dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
xprt->address_strings[RPC_DISPLAY_ADDR]);
if (try_module_get(THIS_MODULE))
return xprt;
ret = ERR_PTR(-EINVAL);
out_err:
xs_xprt_free(xprt);
return ret;
}
static const struct rpc_timeout xs_udp_default_timeout = {
.to_initval = 5 * HZ,
.to_maxval = 30 * HZ,
.to_increment = 5 * HZ,
.to_retries = 5,
};
/**
* xs_setup_udp - Set up transport to use a UDP socket
* @args: rpc transport creation arguments
*
*/
static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
{
struct sockaddr *addr = args->dstaddr;
struct rpc_xprt *xprt;
struct sock_xprt *transport;
struct rpc_xprt *ret;
xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
xprt_udp_slot_table_entries);
if (IS_ERR(xprt))
return xprt;
transport = container_of(xprt, struct sock_xprt, xprt);
xprt->prot = IPPROTO_UDP;
xprt->tsh_size = 0;
/* XXX: header size can vary due to auth type, IPv6, etc. */
xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
xprt->bind_timeout = XS_BIND_TO;
xprt->reestablish_timeout = XS_UDP_REEST_TO;
xprt->idle_timeout = XS_IDLE_DISC_TO;
xprt->ops = &xs_udp_ops;
xprt->timeout = &xs_udp_default_timeout;
switch (addr->sa_family) {
case AF_INET:
if (((struct sockaddr_in *)addr)->sin_port != htons(0))
xprt_set_bound(xprt);
INIT_DELAYED_WORK(&transport->connect_worker,
xs_udp_setup_socket);
xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
break;
case AF_INET6:
if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
xprt_set_bound(xprt);
INIT_DELAYED_WORK(&transport->connect_worker,
xs_udp_setup_socket);
xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
break;
default:
ret = ERR_PTR(-EAFNOSUPPORT);
goto out_err;
}
if (xprt_bound(xprt))
dprintk("RPC: set up xprt to %s (port %s) via %s\n",
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PORT],
xprt->address_strings[RPC_DISPLAY_PROTO]);
else
dprintk("RPC: set up xprt to %s (autobind) via %s\n",
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PROTO]);
if (try_module_get(THIS_MODULE))
return xprt;
ret = ERR_PTR(-EINVAL);
out_err:
xs_xprt_free(xprt);
return ret;
}
static const struct rpc_timeout xs_tcp_default_timeout = {
.to_initval = 60 * HZ,
.to_maxval = 60 * HZ,
.to_retries = 2,
};
/**
* xs_setup_tcp - Set up transport to use a TCP socket
* @args: rpc transport creation arguments
*
*/
static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
{
struct sockaddr *addr = args->dstaddr;
struct rpc_xprt *xprt;
struct sock_xprt *transport;
struct rpc_xprt *ret;
unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
max_slot_table_size);
if (IS_ERR(xprt))
return xprt;
transport = container_of(xprt, struct sock_xprt, xprt);
xprt->prot = IPPROTO_TCP;
xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
xprt->bind_timeout = XS_BIND_TO;
xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
xprt->idle_timeout = XS_IDLE_DISC_TO;
xprt->ops = &xs_tcp_ops;
xprt->timeout = &xs_tcp_default_timeout;
switch (addr->sa_family) {
case AF_INET:
if (((struct sockaddr_in *)addr)->sin_port != htons(0))
xprt_set_bound(xprt);
INIT_DELAYED_WORK(&transport->connect_worker,
xs_tcp_setup_socket);
xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
break;
case AF_INET6:
if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
xprt_set_bound(xprt);
INIT_DELAYED_WORK(&transport->connect_worker,
xs_tcp_setup_socket);
xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
break;
default:
ret = ERR_PTR(-EAFNOSUPPORT);
goto out_err;
}
if (xprt_bound(xprt))
dprintk("RPC: set up xprt to %s (port %s) via %s\n",
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PORT],
xprt->address_strings[RPC_DISPLAY_PROTO]);
else
dprintk("RPC: set up xprt to %s (autobind) via %s\n",
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PROTO]);
if (try_module_get(THIS_MODULE))
return xprt;
ret = ERR_PTR(-EINVAL);
out_err:
xs_xprt_free(xprt);
return ret;
}
/**
* xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
* @args: rpc transport creation arguments
*
*/
static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
{
struct sockaddr *addr = args->dstaddr;
struct rpc_xprt *xprt;
struct sock_xprt *transport;
struct svc_sock *bc_sock;
struct rpc_xprt *ret;
xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
xprt_tcp_slot_table_entries);
if (IS_ERR(xprt))
return xprt;
transport = container_of(xprt, struct sock_xprt, xprt);
xprt->prot = IPPROTO_TCP;
xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
xprt->timeout = &xs_tcp_default_timeout;
/* backchannel */
xprt_set_bound(xprt);
xprt->bind_timeout = 0;
xprt->reestablish_timeout = 0;
xprt->idle_timeout = 0;
xprt->ops = &bc_tcp_ops;
switch (addr->sa_family) {
case AF_INET:
xs_format_peer_addresses(xprt, "tcp",
RPCBIND_NETID_TCP);
break;
case AF_INET6:
xs_format_peer_addresses(xprt, "tcp",
RPCBIND_NETID_TCP6);
break;
default:
ret = ERR_PTR(-EAFNOSUPPORT);
goto out_err;
}
dprintk("RPC: set up xprt to %s (port %s) via %s\n",
xprt->address_strings[RPC_DISPLAY_ADDR],
xprt->address_strings[RPC_DISPLAY_PORT],
xprt->address_strings[RPC_DISPLAY_PROTO]);
/*
* Once we've associated a backchannel xprt with a connection,
* we want to keep it around as long as the connection lasts,
* in case we need to start using it for a backchannel again;
* this reference won't be dropped until bc_xprt is destroyed.
*/
xprt_get(xprt);
args->bc_xprt->xpt_bc_xprt = xprt;
xprt->bc_xprt = args->bc_xprt;
bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
transport->sock = bc_sock->sk_sock;
transport->inet = bc_sock->sk_sk;
/*
* Since we don't want connections for the backchannel, we set
* the xprt status to connected
*/
xprt_set_connected(xprt);
if (try_module_get(THIS_MODULE))
return xprt;
args->bc_xprt->xpt_bc_xprt = NULL;
xprt_put(xprt);
ret = ERR_PTR(-EINVAL);
out_err:
xs_xprt_free(xprt);
return ret;
}
static struct xprt_class xs_local_transport = {
.list = LIST_HEAD_INIT(xs_local_transport.list),
.name = "named UNIX socket",
.owner = THIS_MODULE,
.ident = XPRT_TRANSPORT_LOCAL,
.setup = xs_setup_local,
};
static struct xprt_class xs_udp_transport = {
.list = LIST_HEAD_INIT(xs_udp_transport.list),
.name = "udp",
.owner = THIS_MODULE,
.ident = XPRT_TRANSPORT_UDP,
.setup = xs_setup_udp,
};
static struct xprt_class xs_tcp_transport = {
.list = LIST_HEAD_INIT(xs_tcp_transport.list),
.name = "tcp",
.owner = THIS_MODULE,
.ident = XPRT_TRANSPORT_TCP,
.setup = xs_setup_tcp,
};
static struct xprt_class xs_bc_tcp_transport = {
.list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
.name = "tcp NFSv4.1 backchannel",
.owner = THIS_MODULE,
.ident = XPRT_TRANSPORT_BC_TCP,
.setup = xs_setup_bc_tcp,
};
/**
* init_socket_xprt - set up xprtsock's sysctls, register with RPC client
*
*/
int init_socket_xprt(void)
{
#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
if (!sunrpc_table_header)
[PATCH] sysctl: remove insert_at_head from register_sysctl The semantic effect of insert_at_head is that it would allow new registered sysctl entries to override existing sysctl entries of the same name. Which is pain for caching and the proc interface never implemented. I have done an audit and discovered that none of the current users of register_sysctl care as (excpet for directories) they do not register duplicate sysctl entries. So this patch simply removes the support for overriding existing entries in the sys_sysctl interface since no one uses it or cares and it makes future enhancments harder. Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Acked-by: Ralf Baechle <ralf@linux-mips.org> Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Russell King <rmk@arm.linux.org.uk> Cc: David Howells <dhowells@redhat.com> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andi Kleen <ak@muc.de> Cc: Jens Axboe <axboe@kernel.dk> Cc: Corey Minyard <minyard@acm.org> Cc: Neil Brown <neilb@suse.de> Cc: "John W. Linville" <linville@tuxdriver.com> Cc: James Bottomley <James.Bottomley@steeleye.com> Cc: Jan Kara <jack@ucw.cz> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Mark Fasheh <mark.fasheh@oracle.com> Cc: David Chinner <dgc@sgi.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Patrick McHardy <kaber@trash.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-14 09:34:09 +01:00
sunrpc_table_header = register_sysctl_table(sunrpc_table);
#endif
xprt_register_transport(&xs_local_transport);
xprt_register_transport(&xs_udp_transport);
xprt_register_transport(&xs_tcp_transport);
xprt_register_transport(&xs_bc_tcp_transport);
return 0;
}
/**
* cleanup_socket_xprt - remove xprtsock's sysctls, unregister
*
*/
void cleanup_socket_xprt(void)
{
#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
if (sunrpc_table_header) {
unregister_sysctl_table(sunrpc_table_header);
sunrpc_table_header = NULL;
}
#endif
xprt_unregister_transport(&xs_local_transport);
xprt_unregister_transport(&xs_udp_transport);
xprt_unregister_transport(&xs_tcp_transport);
xprt_unregister_transport(&xs_bc_tcp_transport);
}
static int param_set_uint_minmax(const char *val,
const struct kernel_param *kp,
unsigned int min, unsigned int max)
{
unsigned int num;
int ret;
if (!val)
return -EINVAL;
ret = kstrtouint(val, 0, &num);
if (ret == -EINVAL || num < min || num > max)
return -EINVAL;
*((unsigned int *)kp->arg) = num;
return 0;
}
static int param_set_portnr(const char *val, const struct kernel_param *kp)
{
return param_set_uint_minmax(val, kp,
RPC_MIN_RESVPORT,
RPC_MAX_RESVPORT);
}
static struct kernel_param_ops param_ops_portnr = {
.set = param_set_portnr,
.get = param_get_uint,
};
#define param_check_portnr(name, p) \
__param_check(name, p, unsigned int);
module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
static int param_set_slot_table_size(const char *val,
const struct kernel_param *kp)
{
return param_set_uint_minmax(val, kp,
RPC_MIN_SLOT_TABLE,
RPC_MAX_SLOT_TABLE);
}
static struct kernel_param_ops param_ops_slot_table_size = {
.set = param_set_slot_table_size,
.get = param_get_uint,
};
#define param_check_slot_table_size(name, p) \
__param_check(name, p, unsigned int);
static int param_set_max_slot_table_size(const char *val,
const struct kernel_param *kp)
{
return param_set_uint_minmax(val, kp,
RPC_MIN_SLOT_TABLE,
RPC_MAX_SLOT_TABLE_LIMIT);
}
static struct kernel_param_ops param_ops_max_slot_table_size = {
.set = param_set_max_slot_table_size,
.get = param_get_uint,
};
#define param_check_max_slot_table_size(name, p) \
__param_check(name, p, unsigned int);
module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
slot_table_size, 0644);
module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
max_slot_table_size, 0644);
module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
slot_table_size, 0644);