#include <linux/slab.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
+#include <linux/file.h>
+#include <linux/freezer.h>
#include <net/sock.h>
#include <net/checksum.h>
#include <net/ip.h>
/* SMP locking strategy:
*
- * svc_serv->sv_lock protects most stuff for that service.
+ * svc_pool->sp_lock protects most of the fields of that pool.
+ * svc_serv->sv_lock protects sv_tempsocks, sv_permsocks, sv_tmpcnt.
+ * when both need to be taken (rare), svc_serv->sv_lock is first.
+ * BKL protects svc_serv->sv_nrthread.
+ * svc_sock->sk_defer_lock protects the svc_sock->sk_deferred list
+ * svc_sock->sk_flags.SK_BUSY prevents a svc_sock being enqueued multiply.
*
* Some flags can be set to certain values at any time
* providing that certain rules are followed:
*
- * SK_BUSY can be set to 0 at any time.
- * svc_sock_enqueue must be called afterwards
* SK_CONN, SK_DATA, can be set or cleared at any time.
* after a set, svc_sock_enqueue must be called.
* after a clear, the socket must be read/accepted
* if this succeeds, it must be set again.
* SK_CLOSE can set at any time. It is never cleared.
+ * sk_inuse contains a bias of '1' until SK_DEAD is set.
+ * so when sk_inuse hits zero, we know the socket is dead
+ * and no-one is using it.
+ * SK_DEAD can only be set while SK_BUSY is held which ensures
+ * no other thread will be using the socket or will try to
+ * set SK_DEAD.
*
*/
static struct svc_sock *svc_setup_socket(struct svc_serv *, struct socket *,
int *errp, int pmap_reg);
+static void svc_delete_socket(struct svc_sock *svsk);
static void svc_udp_data_ready(struct sock *, int);
static int svc_udp_recvfrom(struct svc_rqst *);
static int svc_udp_sendto(struct svc_rqst *);
+static void svc_close_socket(struct svc_sock *svsk);
static struct svc_deferred_req *svc_deferred_dequeue(struct svc_sock *svsk);
static int svc_deferred_recv(struct svc_rqst *rqstp);
static struct cache_deferred_req *svc_defer(struct cache_req *req);
+/* apparently the "standard" is that clients close
+ * idle connections after 5 minutes, servers after
+ * 6 minutes
+ * http://www.connectathon.org/talks96/nfstcp.pdf
+ */
+static int svc_conn_age_period = 6*60;
+
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+static struct lock_class_key svc_key[2];
+static struct lock_class_key svc_slock_key[2];
+
+static inline void svc_reclassify_socket(struct socket *sock)
+{
+ struct sock *sk = sock->sk;
+ BUG_ON(sk->sk_lock.owner != NULL);
+ switch (sk->sk_family) {
+ case AF_INET:
+ sock_lock_init_class_and_name(sk, "slock-AF_INET-NFSD",
+ &svc_slock_key[0], "sk_lock-AF_INET-NFSD", &svc_key[0]);
+ break;
+
+ case AF_INET6:
+ sock_lock_init_class_and_name(sk, "slock-AF_INET6-NFSD",
+ &svc_slock_key[1], "sk_lock-AF_INET6-NFSD", &svc_key[1]);
+ break;
+
+ default:
+ BUG();
+ }
+}
+#else
+static inline void svc_reclassify_socket(struct socket *sock)
+{
+}
+#endif
+
/*
- * Queue up an idle server thread. Must have serv->sv_lock held.
+ * Queue up an idle server thread. Must have pool->sp_lock held.
* Note: this is really a stack rather than a queue, so that we only
- * use as many different threads as we need, and the rest don't polute
+ * use as many different threads as we need, and the rest don't pollute
* the cache.
*/
static inline void
-svc_serv_enqueue(struct svc_serv *serv, struct svc_rqst *rqstp)
+svc_thread_enqueue(struct svc_pool *pool, struct svc_rqst *rqstp)
{
- list_add(&rqstp->rq_list, &serv->sv_threads);
+ list_add(&rqstp->rq_list, &pool->sp_threads);
}
/*
- * Dequeue an nfsd thread. Must have serv->sv_lock held.
+ * Dequeue an nfsd thread. Must have pool->sp_lock held.
*/
static inline void
-svc_serv_dequeue(struct svc_serv *serv, struct svc_rqst *rqstp)
+svc_thread_dequeue(struct svc_pool *pool, struct svc_rqst *rqstp)
{
list_del(&rqstp->rq_list);
}
svc_sock_enqueue(struct svc_sock *svsk)
{
struct svc_serv *serv = svsk->sk_server;
+ struct svc_pool *pool;
struct svc_rqst *rqstp;
+ int cpu;
if (!(svsk->sk_flags &
( (1<<SK_CONN)|(1<<SK_DATA)|(1<<SK_CLOSE)|(1<<SK_DEFERRED)) ))
if (test_bit(SK_DEAD, &svsk->sk_flags))
return;
- spin_lock_bh(&serv->sv_lock);
+ cpu = get_cpu();
+ pool = svc_pool_for_cpu(svsk->sk_server, cpu);
+ put_cpu();
+
+ spin_lock_bh(&pool->sp_lock);
- if (!list_empty(&serv->sv_threads) &&
- !list_empty(&serv->sv_sockets))
+ if (!list_empty(&pool->sp_threads) &&
+ !list_empty(&pool->sp_sockets))
printk(KERN_ERR
"svc_sock_enqueue: threads and sockets both waiting??\n");
goto out_unlock;
}
- if (test_bit(SK_BUSY, &svsk->sk_flags)) {
- /* Don't enqueue socket while daemon is receiving */
+ /* Mark socket as busy. It will remain in this state until the
+ * server has processed all pending data and put the socket back
+ * on the idle list. We update SK_BUSY atomically because
+ * it also guards against trying to enqueue the svc_sock twice.
+ */
+ if (test_and_set_bit(SK_BUSY, &svsk->sk_flags)) {
+ /* Don't enqueue socket while already enqueued */
dprintk("svc: socket %p busy, not enqueued\n", svsk->sk_sk);
goto out_unlock;
}
+ BUG_ON(svsk->sk_pool != NULL);
+ svsk->sk_pool = pool;
set_bit(SOCK_NOSPACE, &svsk->sk_sock->flags);
- if (((svsk->sk_reserved + serv->sv_bufsz)*2
+ if (((atomic_read(&svsk->sk_reserved) + serv->sv_max_mesg)*2
> svc_sock_wspace(svsk))
&& !test_bit(SK_CLOSE, &svsk->sk_flags)
&& !test_bit(SK_CONN, &svsk->sk_flags)) {
/* Don't enqueue while not enough space for reply */
dprintk("svc: socket %p no space, %d*2 > %ld, not enqueued\n",
- svsk->sk_sk, svsk->sk_reserved+serv->sv_bufsz,
+ svsk->sk_sk, atomic_read(&svsk->sk_reserved)+serv->sv_max_mesg,
svc_sock_wspace(svsk));
+ svsk->sk_pool = NULL;
+ clear_bit(SK_BUSY, &svsk->sk_flags);
goto out_unlock;
}
clear_bit(SOCK_NOSPACE, &svsk->sk_sock->flags);
- /* Mark socket as busy. It will remain in this state until the
- * server has processed all pending data and put the socket back
- * on the idle list.
- */
- set_bit(SK_BUSY, &svsk->sk_flags);
- if (!list_empty(&serv->sv_threads)) {
- rqstp = list_entry(serv->sv_threads.next,
+ if (!list_empty(&pool->sp_threads)) {
+ rqstp = list_entry(pool->sp_threads.next,
struct svc_rqst,
rq_list);
dprintk("svc: socket %p served by daemon %p\n",
svsk->sk_sk, rqstp);
- svc_serv_dequeue(serv, rqstp);
+ svc_thread_dequeue(pool, rqstp);
if (rqstp->rq_sock)
printk(KERN_ERR
"svc_sock_enqueue: server %p, rq_sock=%p!\n",
rqstp, rqstp->rq_sock);
rqstp->rq_sock = svsk;
- svsk->sk_inuse++;
- rqstp->rq_reserved = serv->sv_bufsz;
- svsk->sk_reserved += rqstp->rq_reserved;
+ atomic_inc(&svsk->sk_inuse);
+ rqstp->rq_reserved = serv->sv_max_mesg;
+ atomic_add(rqstp->rq_reserved, &svsk->sk_reserved);
+ BUG_ON(svsk->sk_pool != pool);
wake_up(&rqstp->rq_wait);
} else {
dprintk("svc: socket %p put into queue\n", svsk->sk_sk);
- list_add_tail(&svsk->sk_ready, &serv->sv_sockets);
+ list_add_tail(&svsk->sk_ready, &pool->sp_sockets);
+ BUG_ON(svsk->sk_pool != pool);
}
out_unlock:
- spin_unlock_bh(&serv->sv_lock);
+ spin_unlock_bh(&pool->sp_lock);
}
/*
- * Dequeue the first socket. Must be called with the serv->sv_lock held.
+ * Dequeue the first socket. Must be called with the pool->sp_lock held.
*/
static inline struct svc_sock *
-svc_sock_dequeue(struct svc_serv *serv)
+svc_sock_dequeue(struct svc_pool *pool)
{
struct svc_sock *svsk;
- if (list_empty(&serv->sv_sockets))
+ if (list_empty(&pool->sp_sockets))
return NULL;
- svsk = list_entry(serv->sv_sockets.next,
+ svsk = list_entry(pool->sp_sockets.next,
struct svc_sock, sk_ready);
list_del_init(&svsk->sk_ready);
dprintk("svc: socket %p dequeued, inuse=%d\n",
- svsk->sk_sk, svsk->sk_inuse);
+ svsk->sk_sk, atomic_read(&svsk->sk_inuse));
return svsk;
}
static inline void
svc_sock_received(struct svc_sock *svsk)
{
+ svsk->sk_pool = NULL;
clear_bit(SK_BUSY, &svsk->sk_flags);
svc_sock_enqueue(svsk);
}
if (space < rqstp->rq_reserved) {
struct svc_sock *svsk = rqstp->rq_sock;
- spin_lock_bh(&svsk->sk_server->sv_lock);
- svsk->sk_reserved -= (rqstp->rq_reserved - space);
+ atomic_sub((rqstp->rq_reserved - space), &svsk->sk_reserved);
rqstp->rq_reserved = space;
- spin_unlock_bh(&svsk->sk_server->sv_lock);
svc_sock_enqueue(svsk);
}
static inline void
svc_sock_put(struct svc_sock *svsk)
{
- struct svc_serv *serv = svsk->sk_server;
+ if (atomic_dec_and_test(&svsk->sk_inuse)) {
+ BUG_ON(! test_bit(SK_DEAD, &svsk->sk_flags));
- spin_lock_bh(&serv->sv_lock);
- if (!--(svsk->sk_inuse) && test_bit(SK_DEAD, &svsk->sk_flags)) {
- spin_unlock_bh(&serv->sv_lock);
dprintk("svc: releasing dead socket\n");
- sock_release(svsk->sk_sock);
+ if (svsk->sk_sock->file)
+ sockfd_put(svsk->sk_sock);
+ else
+ sock_release(svsk->sk_sock);
+ if (svsk->sk_info_authunix != NULL)
+ svcauth_unix_info_release(svsk->sk_info_authunix);
kfree(svsk);
}
- else
- spin_unlock_bh(&serv->sv_lock);
}
static void
svc_release_skb(rqstp);
- svc_free_allpages(rqstp);
+ svc_free_res_pages(rqstp);
rqstp->rq_res.page_len = 0;
rqstp->rq_res.page_base = 0;
/*
* External function to wake up a server waiting for data
+ * This really only makes sense for services like lockd
+ * which have exactly one thread anyway.
*/
void
svc_wake_up(struct svc_serv *serv)
{
struct svc_rqst *rqstp;
-
- spin_lock_bh(&serv->sv_lock);
- if (!list_empty(&serv->sv_threads)) {
- rqstp = list_entry(serv->sv_threads.next,
- struct svc_rqst,
- rq_list);
- dprintk("svc: daemon %p woken up.\n", rqstp);
- /*
- svc_serv_dequeue(serv, rqstp);
- rqstp->rq_sock = NULL;
- */
- wake_up(&rqstp->rq_wait);
+ unsigned int i;
+ struct svc_pool *pool;
+
+ for (i = 0; i < serv->sv_nrpools; i++) {
+ pool = &serv->sv_pools[i];
+
+ spin_lock_bh(&pool->sp_lock);
+ if (!list_empty(&pool->sp_threads)) {
+ rqstp = list_entry(pool->sp_threads.next,
+ struct svc_rqst,
+ rq_list);
+ dprintk("svc: daemon %p woken up.\n", rqstp);
+ /*
+ svc_thread_dequeue(pool, rqstp);
+ rqstp->rq_sock = NULL;
+ */
+ wake_up(&rqstp->rq_wait);
+ }
+ spin_unlock_bh(&pool->sp_lock);
}
- spin_unlock_bh(&serv->sv_lock);
}
/*
/* send head */
if (slen == xdr->head[0].iov_len)
flags = 0;
- len = sock->ops->sendpage(sock, rqstp->rq_respages[0], 0, xdr->head[0].iov_len, flags);
+ len = kernel_sendpage(sock, rqstp->rq_respages[0], 0,
+ xdr->head[0].iov_len, flags);
if (len != xdr->head[0].iov_len)
goto out;
slen -= xdr->head[0].iov_len;
while (pglen > 0) {
if (slen == size)
flags = 0;
- result = sock->ops->sendpage(sock, *ppage, base, size, flags);
+ result = kernel_sendpage(sock, *ppage, base, size, flags);
if (result > 0)
len += result;
if (result != size)
}
/* send tail */
if (xdr->tail[0].iov_len) {
- result = sock->ops->sendpage(sock, rqstp->rq_respages[rqstp->rq_restailpage],
- ((unsigned long)xdr->tail[0].iov_base)& (PAGE_SIZE-1),
+ result = kernel_sendpage(sock, rqstp->rq_respages[0],
+ ((unsigned long)xdr->tail[0].iov_base)
+ & (PAGE_SIZE-1),
xdr->tail[0].iov_len, 0);
if (result > 0)
return len;
}
+/*
+ * Report socket names for nfsdfs
+ */
+static int one_sock_name(char *buf, struct svc_sock *svsk)
+{
+ int len;
+
+ switch(svsk->sk_sk->sk_family) {
+ case AF_INET:
+ len = sprintf(buf, "ipv4 %s %u.%u.%u.%u %d\n",
+ svsk->sk_sk->sk_protocol==IPPROTO_UDP?
+ "udp" : "tcp",
+ NIPQUAD(inet_sk(svsk->sk_sk)->rcv_saddr),
+ inet_sk(svsk->sk_sk)->num);
+ break;
+ default:
+ len = sprintf(buf, "*unknown-%d*\n",
+ svsk->sk_sk->sk_family);
+ }
+ return len;
+}
+
+int
+svc_sock_names(char *buf, struct svc_serv *serv, char *toclose)
+{
+ struct svc_sock *svsk, *closesk = NULL;
+ int len = 0;
+
+ if (!serv)
+ return 0;
+ spin_lock_bh(&serv->sv_lock);
+ list_for_each_entry(svsk, &serv->sv_permsocks, sk_list) {
+ int onelen = one_sock_name(buf+len, svsk);
+ if (toclose && strcmp(toclose, buf+len) == 0)
+ closesk = svsk;
+ else
+ len += onelen;
+ }
+ spin_unlock_bh(&serv->sv_lock);
+ if (closesk)
+ /* Should unregister with portmap, but you cannot
+ * unregister just one protocol...
+ */
+ svc_close_socket(closesk);
+ else if (toclose)
+ return -ENOENT;
+ return len;
+}
+EXPORT_SYMBOL(svc_sock_names);
+
/*
* Check input queue length
*/
static int
svc_recv_available(struct svc_sock *svsk)
{
- mm_segment_t oldfs;
struct socket *sock = svsk->sk_sock;
int avail, err;
- oldfs = get_fs(); set_fs(KERNEL_DS);
- err = sock->ops->ioctl(sock, TIOCINQ, (unsigned long) &avail);
- set_fs(oldfs);
+ err = kernel_sock_ioctl(sock, TIOCINQ, (unsigned long) &avail);
return (err >= 0)? avail : err;
}
* at accept time. FIXME
*/
alen = sizeof(rqstp->rq_addr);
- sock->ops->getname(sock, (struct sockaddr *)&rqstp->rq_addr, &alen, 1);
+ kernel_getpeername(sock, (struct sockaddr *)&rqstp->rq_addr, &alen);
dprintk("svc: socket %p recvfrom(%p, %Zu) = %d\n",
rqstp->rq_sock, iov[0].iov_base, iov[0].iov_len, len);
/* udp sockets need large rcvbuf as all pending
* requests are still in that buffer. sndbuf must
* also be large enough that there is enough space
- * for one reply per thread.
+ * for one reply per thread. We count all threads
+ * rather than threads in a particular pool, which
+ * provides an upper bound on the number of threads
+ * which will access the socket.
*/
svc_sock_setbufsize(svsk->sk_sock,
- (serv->sv_nrthreads+3) * serv->sv_bufsz,
- (serv->sv_nrthreads+3) * serv->sv_bufsz);
+ (serv->sv_nrthreads+3) * serv->sv_max_mesg,
+ (serv->sv_nrthreads+3) * serv->sv_max_mesg);
if ((rqstp->rq_deferred = svc_deferred_dequeue(svsk))) {
svc_sock_received(svsk);
return svc_deferred_recv(rqstp);
}
+ if (test_bit(SK_CLOSE, &svsk->sk_flags)) {
+ svc_delete_socket(svsk);
+ return 0;
+ }
+
clear_bit(SK_DATA, &svsk->sk_flags);
while ((skb = skb_recv_datagram(svsk->sk_sk, 0, 1, &err)) == NULL) {
if (err == -EAGAIN) {
if (len <= rqstp->rq_arg.head[0].iov_len) {
rqstp->rq_arg.head[0].iov_len = len;
rqstp->rq_arg.page_len = 0;
+ rqstp->rq_respages = rqstp->rq_pages+1;
} else {
rqstp->rq_arg.page_len = len - rqstp->rq_arg.head[0].iov_len;
- rqstp->rq_argused += (rqstp->rq_arg.page_len + PAGE_SIZE - 1)/ PAGE_SIZE;
+ rqstp->rq_respages = rqstp->rq_pages + 1 +
+ (rqstp->rq_arg.page_len + PAGE_SIZE - 1)/ PAGE_SIZE;
}
if (serv->sv_stats)
* svc_udp_recvfrom will re-adjust if necessary
*/
svc_sock_setbufsize(svsk->sk_sock,
- 3 * svsk->sk_server->sv_bufsz,
- 3 * svsk->sk_server->sv_bufsz);
+ 3 * svsk->sk_server->sv_max_mesg,
+ 3 * svsk->sk_server->sv_max_mesg);
set_bit(SK_DATA, &svsk->sk_flags); /* might have come in before data_ready set up */
set_bit(SK_CHNGBUF, &svsk->sk_flags);
struct svc_serv *serv = svsk->sk_server;
struct socket *sock = svsk->sk_sock;
struct socket *newsock;
- const struct proto_ops *ops;
struct svc_sock *newsvsk;
int err, slen;
if (!sock)
return;
- err = sock_create_lite(PF_INET, SOCK_STREAM, IPPROTO_TCP, &newsock);
- if (err) {
+ clear_bit(SK_CONN, &svsk->sk_flags);
+ err = kernel_accept(sock, &newsock, O_NONBLOCK);
+ if (err < 0) {
if (err == -ENOMEM)
printk(KERN_WARNING "%s: no more sockets!\n",
serv->sv_name);
- return;
- }
-
- dprintk("svc: tcp_accept %p allocated\n", newsock);
- newsock->ops = ops = sock->ops;
-
- clear_bit(SK_CONN, &svsk->sk_flags);
- if ((err = ops->accept(sock, newsock, O_NONBLOCK)) < 0) {
- if (err != -EAGAIN && net_ratelimit())
+ else if (err != -EAGAIN && net_ratelimit())
printk(KERN_WARNING "%s: accept failed (err %d)!\n",
serv->sv_name, -err);
- goto failed; /* aborted connection or whatever */
+ return;
}
+
set_bit(SK_CONN, &svsk->sk_flags);
svc_sock_enqueue(svsk);
slen = sizeof(sin);
- err = ops->getname(newsock, (struct sockaddr *) &sin, &slen, 1);
+ err = kernel_getpeername(newsock, (struct sockaddr *) &sin, &slen);
if (err < 0) {
if (net_ratelimit())
printk(KERN_WARNING "%s: peername failed (err %d)!\n",
struct svc_sock,
sk_list);
set_bit(SK_CLOSE, &svsk->sk_flags);
- svsk->sk_inuse ++;
+ atomic_inc(&svsk->sk_inuse);
}
spin_unlock_bh(&serv->sv_lock);
struct svc_sock *svsk = rqstp->rq_sock;
struct svc_serv *serv = svsk->sk_server;
int len;
- struct kvec vec[RPCSVC_MAXPAGES];
+ struct kvec *vec;
int pnum, vlen;
dprintk("svc: tcp_recv %p data %d conn %d close %d\n",
/* sndbuf needs to have room for one request
* per thread, otherwise we can stall even when the
* network isn't a bottleneck.
+ *
+ * We count all threads rather than threads in a
+ * particular pool, which provides an upper bound
+ * on the number of threads which will access the socket.
+ *
* rcvbuf just needs to be able to hold a few requests.
* Normally they will be removed from the queue
* as soon a a complete request arrives.
*/
svc_sock_setbufsize(svsk->sk_sock,
- (serv->sv_nrthreads+3) * serv->sv_bufsz,
- 3 * serv->sv_bufsz);
+ (serv->sv_nrthreads+3) * serv->sv_max_mesg,
+ 3 * serv->sv_max_mesg);
clear_bit(SK_DATA, &svsk->sk_flags);
* bit set in the fragment length header.
* But apparently no known nfs clients send fragmented
* records. */
- printk(KERN_NOTICE "RPC: bad TCP reclen 0x%08lx (non-terminal)\n",
- (unsigned long) svsk->sk_reclen);
+ if (net_ratelimit())
+ printk(KERN_NOTICE "RPC: bad TCP reclen 0x%08lx"
+ " (non-terminal)\n",
+ (unsigned long) svsk->sk_reclen);
goto err_delete;
}
svsk->sk_reclen &= 0x7fffffff;
dprintk("svc: TCP record, %d bytes\n", svsk->sk_reclen);
- if (svsk->sk_reclen > serv->sv_bufsz) {
- printk(KERN_NOTICE "RPC: bad TCP reclen 0x%08lx (large)\n",
- (unsigned long) svsk->sk_reclen);
+ if (svsk->sk_reclen > serv->sv_max_mesg) {
+ if (net_ratelimit())
+ printk(KERN_NOTICE "RPC: bad TCP reclen 0x%08lx"
+ " (large)\n",
+ (unsigned long) svsk->sk_reclen);
goto err_delete;
}
}
len = svsk->sk_reclen;
set_bit(SK_DATA, &svsk->sk_flags);
+ vec = rqstp->rq_vec;
vec[0] = rqstp->rq_arg.head[0];
vlen = PAGE_SIZE;
pnum = 1;
while (vlen < len) {
- vec[pnum].iov_base = page_address(rqstp->rq_argpages[rqstp->rq_argused++]);
+ vec[pnum].iov_base = page_address(rqstp->rq_pages[pnum]);
vec[pnum].iov_len = PAGE_SIZE;
pnum++;
vlen += PAGE_SIZE;
}
+ rqstp->rq_respages = &rqstp->rq_pages[pnum];
/* Now receive data */
len = svc_recvfrom(rqstp, vec, pnum, len);
{
struct xdr_buf *xbufp = &rqstp->rq_res;
int sent;
- u32 reclen;
+ __be32 reclen;
/* Set up the first element of the reply kvec.
* Any other kvecs that may be in use have been taken
rqstp->rq_sock->sk_server->sv_name,
(sent<0)?"got error":"sent only",
sent, xbufp->len);
- svc_delete_socket(rqstp->rq_sock);
+ set_bit(SK_CLOSE, &rqstp->rq_sock->sk_flags);
+ svc_sock_enqueue(rqstp->rq_sock);
sent = -EAGAIN;
}
return sent;
* svc_tcp_recvfrom will re-adjust if necessary
*/
svc_sock_setbufsize(svsk->sk_sock,
- 3 * svsk->sk_server->sv_bufsz,
- 3 * svsk->sk_server->sv_bufsz);
+ 3 * svsk->sk_server->sv_max_mesg,
+ 3 * svsk->sk_server->sv_max_mesg);
set_bit(SK_CHNGBUF, &svsk->sk_flags);
set_bit(SK_DATA, &svsk->sk_flags);
}
/*
- * Receive the next request on any socket.
+ * Receive the next request on any socket. This code is carefully
+ * organised not to touch any cachelines in the shared svc_serv
+ * structure, only cachelines in the local svc_pool.
*/
int
-svc_recv(struct svc_serv *serv, struct svc_rqst *rqstp, long timeout)
+svc_recv(struct svc_rqst *rqstp, long timeout)
{
struct svc_sock *svsk =NULL;
- int len;
+ struct svc_serv *serv = rqstp->rq_server;
+ struct svc_pool *pool = rqstp->rq_pool;
+ int len, i;
int pages;
struct xdr_buf *arg;
DECLARE_WAITQUEUE(wait, current);
"svc_recv: service %p, wait queue active!\n",
rqstp);
- /* Initialize the buffers */
- /* first reclaim pages that were moved to response list */
- svc_pushback_allpages(rqstp);
/* now allocate needed pages. If we get a failure, sleep briefly */
- pages = 2 + (serv->sv_bufsz + PAGE_SIZE -1) / PAGE_SIZE;
- while (rqstp->rq_arghi < pages) {
- struct page *p = alloc_page(GFP_KERNEL);
- if (!p) {
- schedule_timeout_uninterruptible(msecs_to_jiffies(500));
- continue;
+ pages = (serv->sv_max_mesg + PAGE_SIZE) / PAGE_SIZE;
+ for (i=0; i < pages ; i++)
+ while (rqstp->rq_pages[i] == NULL) {
+ struct page *p = alloc_page(GFP_KERNEL);
+ if (!p)
+ schedule_timeout_uninterruptible(msecs_to_jiffies(500));
+ rqstp->rq_pages[i] = p;
}
- rqstp->rq_argpages[rqstp->rq_arghi++] = p;
- }
+ rqstp->rq_pages[i++] = NULL; /* this might be seen in nfs_read_actor */
+ BUG_ON(pages >= RPCSVC_MAXPAGES);
/* Make arg->head point to first page and arg->pages point to rest */
arg = &rqstp->rq_arg;
- arg->head[0].iov_base = page_address(rqstp->rq_argpages[0]);
+ arg->head[0].iov_base = page_address(rqstp->rq_pages[0]);
arg->head[0].iov_len = PAGE_SIZE;
- rqstp->rq_argused = 1;
- arg->pages = rqstp->rq_argpages + 1;
+ arg->pages = rqstp->rq_pages + 1;
arg->page_base = 0;
/* save at least one page for response */
arg->page_len = (pages-2)*PAGE_SIZE;
if (signalled())
return -EINTR;
- spin_lock_bh(&serv->sv_lock);
- if (!list_empty(&serv->sv_tempsocks)) {
- svsk = list_entry(serv->sv_tempsocks.next,
- struct svc_sock, sk_list);
- /* apparently the "standard" is that clients close
- * idle connections after 5 minutes, servers after
- * 6 minutes
- * http://www.connectathon.org/talks96/nfstcp.pdf
- */
- if (get_seconds() - svsk->sk_lastrecv < 6*60
- || test_bit(SK_BUSY, &svsk->sk_flags))
- svsk = NULL;
- }
- if (svsk) {
- set_bit(SK_BUSY, &svsk->sk_flags);
- set_bit(SK_CLOSE, &svsk->sk_flags);
+ spin_lock_bh(&pool->sp_lock);
+ if ((svsk = svc_sock_dequeue(pool)) != NULL) {
rqstp->rq_sock = svsk;
- svsk->sk_inuse++;
- } else if ((svsk = svc_sock_dequeue(serv)) != NULL) {
- rqstp->rq_sock = svsk;
- svsk->sk_inuse++;
- rqstp->rq_reserved = serv->sv_bufsz;
- svsk->sk_reserved += rqstp->rq_reserved;
+ atomic_inc(&svsk->sk_inuse);
+ rqstp->rq_reserved = serv->sv_max_mesg;
+ atomic_add(rqstp->rq_reserved, &svsk->sk_reserved);
} else {
/* No data pending. Go to sleep */
- svc_serv_enqueue(serv, rqstp);
+ svc_thread_enqueue(pool, rqstp);
/*
* We have to be able to interrupt this wait
*/
set_current_state(TASK_INTERRUPTIBLE);
add_wait_queue(&rqstp->rq_wait, &wait);
- spin_unlock_bh(&serv->sv_lock);
+ spin_unlock_bh(&pool->sp_lock);
schedule_timeout(timeout);
try_to_freeze();
- spin_lock_bh(&serv->sv_lock);
+ spin_lock_bh(&pool->sp_lock);
remove_wait_queue(&rqstp->rq_wait, &wait);
if (!(svsk = rqstp->rq_sock)) {
- svc_serv_dequeue(serv, rqstp);
- spin_unlock_bh(&serv->sv_lock);
+ svc_thread_dequeue(pool, rqstp);
+ spin_unlock_bh(&pool->sp_lock);
dprintk("svc: server %p, no data yet\n", rqstp);
return signalled()? -EINTR : -EAGAIN;
}
}
- spin_unlock_bh(&serv->sv_lock);
+ spin_unlock_bh(&pool->sp_lock);
- dprintk("svc: server %p, socket %p, inuse=%d\n",
- rqstp, svsk, svsk->sk_inuse);
+ dprintk("svc: server %p, pool %u, socket %p, inuse=%d\n",
+ rqstp, pool->sp_id, svsk, atomic_read(&svsk->sk_inuse));
len = svsk->sk_recvfrom(rqstp);
dprintk("svc: got len=%d\n", len);
return -EAGAIN;
}
svsk->sk_lastrecv = get_seconds();
- if (test_bit(SK_TEMP, &svsk->sk_flags)) {
- /* push active sockets to end of list */
- spin_lock_bh(&serv->sv_lock);
- if (!list_empty(&svsk->sk_list))
- list_move_tail(&svsk->sk_list, &serv->sv_tempsocks);
- spin_unlock_bh(&serv->sv_lock);
- }
+ clear_bit(SK_OLD, &svsk->sk_flags);
rqstp->rq_secure = ntohs(rqstp->rq_addr.sin_port) < 1024;
rqstp->rq_chandle.defer = svc_defer;
xb->page_len +
xb->tail[0].iov_len;
- /* Grab svsk->sk_sem to serialize outgoing data. */
- down(&svsk->sk_sem);
+ /* Grab svsk->sk_mutex to serialize outgoing data. */
+ mutex_lock(&svsk->sk_mutex);
if (test_bit(SK_DEAD, &svsk->sk_flags))
len = -ENOTCONN;
else
len = svsk->sk_sendto(rqstp);
- up(&svsk->sk_sem);
+ mutex_unlock(&svsk->sk_mutex);
svc_sock_release(rqstp);
if (len == -ECONNREFUSED || len == -ENOTCONN || len == -EAGAIN)
return len;
}
+/*
+ * Timer function to close old temporary sockets, using
+ * a mark-and-sweep algorithm.
+ */
+static void
+svc_age_temp_sockets(unsigned long closure)
+{
+ struct svc_serv *serv = (struct svc_serv *)closure;
+ struct svc_sock *svsk;
+ struct list_head *le, *next;
+ LIST_HEAD(to_be_aged);
+
+ dprintk("svc_age_temp_sockets\n");
+
+ if (!spin_trylock_bh(&serv->sv_lock)) {
+ /* busy, try again 1 sec later */
+ dprintk("svc_age_temp_sockets: busy\n");
+ mod_timer(&serv->sv_temptimer, jiffies + HZ);
+ return;
+ }
+
+ list_for_each_safe(le, next, &serv->sv_tempsocks) {
+ svsk = list_entry(le, struct svc_sock, sk_list);
+
+ if (!test_and_set_bit(SK_OLD, &svsk->sk_flags))
+ continue;
+ if (atomic_read(&svsk->sk_inuse) || test_bit(SK_BUSY, &svsk->sk_flags))
+ continue;
+ atomic_inc(&svsk->sk_inuse);
+ list_move(le, &to_be_aged);
+ set_bit(SK_CLOSE, &svsk->sk_flags);
+ set_bit(SK_DETACHED, &svsk->sk_flags);
+ }
+ spin_unlock_bh(&serv->sv_lock);
+
+ while (!list_empty(&to_be_aged)) {
+ le = to_be_aged.next;
+ /* fiddling the sk_list node is safe 'cos we're SK_DETACHED */
+ list_del_init(le);
+ svsk = list_entry(le, struct svc_sock, sk_list);
+
+ dprintk("queuing svsk %p for closing, %lu seconds old\n",
+ svsk, get_seconds() - svsk->sk_lastrecv);
+
+ /* a thread will dequeue and close it soon */
+ svc_sock_enqueue(svsk);
+ svc_sock_put(svsk);
+ }
+
+ mod_timer(&serv->sv_temptimer, jiffies + svc_conn_age_period * HZ);
+}
+
/*
* Initialize socket for RPC use and create svc_sock struct
* XXX: May want to setsockopt SO_SNDBUF and SO_RCVBUF.
struct sock *inet;
dprintk("svc: svc_setup_socket %p\n", sock);
- if (!(svsk = kmalloc(sizeof(*svsk), GFP_KERNEL))) {
+ if (!(svsk = kzalloc(sizeof(*svsk), GFP_KERNEL))) {
*errp = -ENOMEM;
return NULL;
}
- memset(svsk, 0, sizeof(*svsk));
inet = sock->sk;
svsk->sk_odata = inet->sk_data_ready;
svsk->sk_owspace = inet->sk_write_space;
svsk->sk_server = serv;
+ atomic_set(&svsk->sk_inuse, 1);
svsk->sk_lastrecv = get_seconds();
+ spin_lock_init(&svsk->sk_defer_lock);
INIT_LIST_HEAD(&svsk->sk_deferred);
INIT_LIST_HEAD(&svsk->sk_ready);
- sema_init(&svsk->sk_sem, 1);
+ mutex_init(&svsk->sk_mutex);
/* Initialize the socket */
if (sock->type == SOCK_DGRAM)
set_bit(SK_TEMP, &svsk->sk_flags);
list_add(&svsk->sk_list, &serv->sv_tempsocks);
serv->sv_tmpcnt++;
+ if (serv->sv_temptimer.function == NULL) {
+ /* setup timer to age temp sockets */
+ setup_timer(&serv->sv_temptimer, svc_age_temp_sockets,
+ (unsigned long)serv);
+ mod_timer(&serv->sv_temptimer,
+ jiffies + svc_conn_age_period * HZ);
+ }
} else {
clear_bit(SK_TEMP, &svsk->sk_flags);
list_add(&svsk->sk_list, &serv->sv_permsocks);
return svsk;
}
+int svc_addsock(struct svc_serv *serv,
+ int fd,
+ char *name_return,
+ int *proto)
+{
+ int err = 0;
+ struct socket *so = sockfd_lookup(fd, &err);
+ struct svc_sock *svsk = NULL;
+
+ if (!so)
+ return err;
+ if (so->sk->sk_family != AF_INET)
+ err = -EAFNOSUPPORT;
+ else if (so->sk->sk_protocol != IPPROTO_TCP &&
+ so->sk->sk_protocol != IPPROTO_UDP)
+ err = -EPROTONOSUPPORT;
+ else if (so->state > SS_UNCONNECTED)
+ err = -EISCONN;
+ else {
+ svsk = svc_setup_socket(serv, so, &err, 1);
+ if (svsk)
+ err = 0;
+ }
+ if (err) {
+ sockfd_put(so);
+ return err;
+ }
+ if (proto) *proto = so->sk->sk_protocol;
+ return one_sock_name(name_return, svsk);
+}
+EXPORT_SYMBOL_GPL(svc_addsock);
+
/*
* Create socket for RPC service.
*/
if ((error = sock_create_kern(PF_INET, type, protocol, &sock)) < 0)
return error;
- if (sin != NULL) {
- if (type == SOCK_STREAM)
- sock->sk->sk_reuse = 1; /* allow address reuse */
- error = sock->ops->bind(sock, (struct sockaddr *) sin,
- sizeof(*sin));
- if (error < 0)
- goto bummer;
- }
+ svc_reclassify_socket(sock);
+
+ if (type == SOCK_STREAM)
+ sock->sk->sk_reuse = 1; /* allow address reuse */
+ error = kernel_bind(sock, (struct sockaddr *) sin,
+ sizeof(*sin));
+ if (error < 0)
+ goto bummer;
if (protocol == IPPROTO_TCP) {
- if ((error = sock->ops->listen(sock, 64)) < 0)
+ if ((error = kernel_listen(sock, 64)) < 0)
goto bummer;
}
/*
* Remove a dead socket
*/
-void
+static void
svc_delete_socket(struct svc_sock *svsk)
{
struct svc_serv *serv;
spin_lock_bh(&serv->sv_lock);
- list_del_init(&svsk->sk_list);
- list_del_init(&svsk->sk_ready);
- if (!test_and_set_bit(SK_DEAD, &svsk->sk_flags))
+ if (!test_and_set_bit(SK_DETACHED, &svsk->sk_flags))
+ list_del_init(&svsk->sk_list);
+ /*
+ * We used to delete the svc_sock from whichever list
+ * it's sk_ready node was on, but we don't actually
+ * need to. This is because the only time we're called
+ * while still attached to a queue, the queue itself
+ * is about to be destroyed (in svc_destroy).
+ */
+ if (!test_and_set_bit(SK_DEAD, &svsk->sk_flags)) {
+ BUG_ON(atomic_read(&svsk->sk_inuse)<2);
+ atomic_dec(&svsk->sk_inuse);
if (test_bit(SK_TEMP, &svsk->sk_flags))
serv->sv_tmpcnt--;
+ }
- if (!svsk->sk_inuse) {
- spin_unlock_bh(&serv->sv_lock);
- sock_release(svsk->sk_sock);
- kfree(svsk);
- } else {
- spin_unlock_bh(&serv->sv_lock);
- dprintk(KERN_NOTICE "svc: server socket destroy delayed\n");
- /* svsk->sk_server = NULL; */
+ spin_unlock_bh(&serv->sv_lock);
+}
+
+static void svc_close_socket(struct svc_sock *svsk)
+{
+ set_bit(SK_CLOSE, &svsk->sk_flags);
+ if (test_and_set_bit(SK_BUSY, &svsk->sk_flags))
+ /* someone else will have to effect the close */
+ return;
+
+ atomic_inc(&svsk->sk_inuse);
+ svc_delete_socket(svsk);
+ clear_bit(SK_BUSY, &svsk->sk_flags);
+ svc_sock_put(svsk);
+}
+
+void svc_force_close_socket(struct svc_sock *svsk)
+{
+ set_bit(SK_CLOSE, &svsk->sk_flags);
+ if (test_bit(SK_BUSY, &svsk->sk_flags)) {
+ /* Waiting to be processed, but no threads left,
+ * So just remove it from the waiting list
+ */
+ list_del_init(&svsk->sk_ready);
+ clear_bit(SK_BUSY, &svsk->sk_flags);
}
+ svc_close_socket(svsk);
}
/*
static void svc_revisit(struct cache_deferred_req *dreq, int too_many)
{
struct svc_deferred_req *dr = container_of(dreq, struct svc_deferred_req, handle);
- struct svc_serv *serv = dreq->owner;
struct svc_sock *svsk;
if (too_many) {
dprintk("revisit queued\n");
svsk = dr->svsk;
dr->svsk = NULL;
- spin_lock_bh(&serv->sv_lock);
+ spin_lock_bh(&svsk->sk_defer_lock);
list_add(&dr->handle.recent, &svsk->sk_deferred);
- spin_unlock_bh(&serv->sv_lock);
+ spin_unlock_bh(&svsk->sk_defer_lock);
set_bit(SK_DEFERRED, &svsk->sk_flags);
svc_sock_enqueue(svsk);
svc_sock_put(svsk);
dr->argslen = rqstp->rq_arg.len >> 2;
memcpy(dr->args, rqstp->rq_arg.head[0].iov_base-skip, dr->argslen<<2);
}
- spin_lock_bh(&rqstp->rq_server->sv_lock);
- rqstp->rq_sock->sk_inuse++;
+ atomic_inc(&rqstp->rq_sock->sk_inuse);
dr->svsk = rqstp->rq_sock;
- spin_unlock_bh(&rqstp->rq_server->sv_lock);
dr->handle.revisit = svc_revisit;
return &dr->handle;
rqstp->rq_prot = dr->prot;
rqstp->rq_addr = dr->addr;
rqstp->rq_daddr = dr->daddr;
+ rqstp->rq_respages = rqstp->rq_pages;
return dr->argslen<<2;
}
static struct svc_deferred_req *svc_deferred_dequeue(struct svc_sock *svsk)
{
struct svc_deferred_req *dr = NULL;
- struct svc_serv *serv = svsk->sk_server;
if (!test_bit(SK_DEFERRED, &svsk->sk_flags))
return NULL;
- spin_lock_bh(&serv->sv_lock);
+ spin_lock_bh(&svsk->sk_defer_lock);
clear_bit(SK_DEFERRED, &svsk->sk_flags);
if (!list_empty(&svsk->sk_deferred)) {
dr = list_entry(svsk->sk_deferred.next,
list_del_init(&dr->handle.recent);
set_bit(SK_DEFERRED, &svsk->sk_flags);
}
- spin_unlock_bh(&serv->sv_lock);
+ spin_unlock_bh(&svsk->sk_defer_lock);
return dr;
}
git://git.onelab.eu / linux-2.6.git / blobdiff