mirrors/linux
1
0
Fork 1
mirror of https://github.com/torvalds/linux.git synced 2025-11-04 02:30:34 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

RDS: TCP: Add sysctl tunables for sndbuf/rcvbuf on rds-tcp socket

Browse source 
Add per-net sysctl tunables to set the size of sndbuf and
rcvbuf on the kernel tcp socket.

The tunables are added at /proc/sys/net/rds/tcp/rds_tcp_sndbuf
and /proc/sys/net/rds/tcp/rds_tcp_rcvbuf.

These values must be set before accept() or connect(),
and there may be an arbitrary number of existing rds-tcp
sockets when the tunable is modified. To make sure that all
connections in the netns pick up the same value for the tunable,
we reset existing rds-tcp connections in the netns, so that
they can reconnect with the new parameters.

Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
Sowmini Varadhan 2016-03-16 11:38:12 -07:00 committed by David S. Miller
parent 79d3b59a93
commit c6a58ffed5
1 changed files with 135 additions and 10 deletions
repo.diff.show_diff_stats Download patch file Download diff file Expand all files Collapse all files

145
net/rds/tcp.c
View file

@ -54,6 +54,35 @@ static struct kmem_cache *rds_tcp_conn_slab;
#define RDS_TCP_DEFAULT_BUFSIZE (128 * 1024)
static int rds_tcp_skbuf_handler(struct ctl_table *ctl, int write,
void __user *buffer, size_t *lenp,
loff_t *fpos);
int rds_tcp_min_sndbuf = SOCK_MIN_SNDBUF;
int rds_tcp_min_rcvbuf = SOCK_MIN_RCVBUF;
static struct ctl_table rds_tcp_sysctl_table[] = {
#define RDS_TCP_SNDBUF 0
{
.procname = "rds_tcp_sndbuf",
/* data is per-net pointer */
.maxlen = sizeof(int),
.mode = 0644,
.proc_handler = rds_tcp_skbuf_handler,
.extra1 = &rds_tcp_min_sndbuf,
},
#define RDS_TCP_RCVBUF 1
{
.procname = "rds_tcp_rcvbuf",
/* data is per-net pointer */
.maxlen = sizeof(int),
.mode = 0644,
.proc_handler = rds_tcp_skbuf_handler,
.extra1 = &rds_tcp_min_rcvbuf,
},
{ }
};
/* doing it this way avoids calling tcp_sk() */
void rds_tcp_nonagle(struct socket *sock)
{
@ -66,15 +95,6 @@ void rds_tcp_nonagle(struct socket *sock)
set_fs(oldfs);
}
/* All module specific customizations to the RDS-TCP socket should be done in
* rds_tcp_tune() and applied after socket creation. In general these
* customizations should be tunable via module_param()
*/
void rds_tcp_tune(struct socket *sock)
{
rds_tcp_nonagle(sock);
}
u32 rds_tcp_snd_nxt(struct rds_tcp_connection *tc)
{
return tcp_sk(tc->t_sock->sk)->snd_nxt;
@ -272,8 +292,34 @@ static int rds_tcp_netid;
struct rds_tcp_net {
struct socket *rds_tcp_listen_sock;
struct work_struct rds_tcp_accept_w;
struct ctl_table_header *rds_tcp_sysctl;
struct ctl_table *ctl_table;
int sndbuf_size;
int rcvbuf_size;
};
/* All module specific customizations to the RDS-TCP socket should be done in
* rds_tcp_tune() and applied after socket creation.
*/
void rds_tcp_tune(struct socket *sock)
{
struct sock *sk = sock->sk;
struct net *net = sock_net(sk);
struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
rds_tcp_nonagle(sock);
lock_sock(sk);
if (rtn->sndbuf_size > 0) {
sk->sk_sndbuf = rtn->sndbuf_size;
sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
}
if (rtn->rcvbuf_size > 0) {
sk->sk_sndbuf = rtn->rcvbuf_size;
sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
}
release_sock(sk);
}
static void rds_tcp_accept_worker(struct work_struct *work)
{
struct rds_tcp_net *rtn = container_of(work,
@ -295,20 +341,60 @@ void rds_tcp_accept_work(struct sock *sk)
static __net_init int rds_tcp_init_net(struct net *net)
{
struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
struct ctl_table *tbl;
int err = 0;
memset(rtn, 0, sizeof(*rtn));
/* {snd, rcv}buf_size default to 0, which implies we let the
* stack pick the value, and permit auto-tuning of buffer size.
*/
if (net == &init_net) {
tbl = rds_tcp_sysctl_table;
} else {
tbl = kmemdup(rds_tcp_sysctl_table,
sizeof(rds_tcp_sysctl_table), GFP_KERNEL);
if (!tbl) {
pr_warn("could not set allocate syctl table\n");
return -ENOMEM;
}
rtn->ctl_table = tbl;
}
tbl[RDS_TCP_SNDBUF].data = &rtn->sndbuf_size;
tbl[RDS_TCP_RCVBUF].data = &rtn->rcvbuf_size;
rtn->rds_tcp_sysctl = register_net_sysctl(net, "net/rds/tcp", tbl);
if (!rtn->rds_tcp_sysctl) {
pr_warn("could not register sysctl\n");
err = -ENOMEM;
goto fail;
}
rtn->rds_tcp_listen_sock = rds_tcp_listen_init(net);
if (!rtn->rds_tcp_listen_sock) {
pr_warn("could not set up listen sock\n");
return -EAFNOSUPPORT;
unregister_net_sysctl_table(rtn->rds_tcp_sysctl);
rtn->rds_tcp_sysctl = NULL;
err = -EAFNOSUPPORT;
goto fail;
}
INIT_WORK(&rtn->rds_tcp_accept_w, rds_tcp_accept_worker);
return 0;
fail:
if (net != &init_net)
kfree(tbl);
return err;
}
static void __net_exit rds_tcp_exit_net(struct net *net)
{
struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
if (rtn->rds_tcp_sysctl)
unregister_net_sysctl_table(rtn->rds_tcp_sysctl);
if (net != &init_net && rtn->ctl_table)
kfree(rtn->ctl_table);
/* If rds_tcp_exit_net() is called as a result of netns deletion,
* the rds_tcp_kill_sock() device notifier would already have cleaned
* up the listen socket, thus there is no work to do in this function.
@ -383,6 +469,45 @@ static struct notifier_block rds_tcp_dev_notifier = {
.priority = -10, /* must be called after other network notifiers */
};
/* when sysctl is used to modify some kernel socket parameters,this
* function resets the RDS connections in that netns so that we can
* restart with new parameters. The assumption is that such reset
* events are few and far-between.
*/
static void rds_tcp_sysctl_reset(struct net *net)
{
struct rds_tcp_connection *tc, *_tc;
spin_lock_irq(&rds_tcp_conn_lock);
list_for_each_entry_safe(tc, _tc, &rds_tcp_conn_list, t_tcp_node) {
struct net *c_net = read_pnet(&tc->conn->c_net);
if (net != c_net || !tc->t_sock)
continue;
rds_conn_drop(tc->conn); /* reconnect with new parameters */
}
spin_unlock_irq(&rds_tcp_conn_lock);
}
static int rds_tcp_skbuf_handler(struct ctl_table *ctl, int write,
void __user *buffer, size_t *lenp,
loff_t *fpos)
{
struct net *net = current->nsproxy->net_ns;
int err;
err = proc_dointvec_minmax(ctl, write, buffer, lenp, fpos);
if (err < 0) {
pr_warn("Invalid input. Must be >= %d\n",
*(int *)(ctl->extra1));
return err;
}
if (write)
rds_tcp_sysctl_reset(net);
return 0;
}
static void rds_tcp_exit(void)
{
rds_info_deregister_func(RDS_INFO_TCP_SOCKETS, rds_tcp_tc_info);