freebsd-dev/sys/cam/ctl/ctl_ha.c

1009 lines
24 KiB
C
Raw Normal View History

/*-
2018-01-16 23:19:57 +00:00
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2015 Alexander Motin <mav@FreeBSD.org>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer,
* without modification, immediately at the beginning of the file.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/types.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/module.h>
#include <sys/mutex.h>
#include <sys/condvar.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/proc.h>
#include <sys/conf.h>
#include <sys/queue.h>
#include <sys/sysctl.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/uio.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#include <vm/uma.h>
#include <cam/cam.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_da.h>
#include <cam/ctl/ctl_io.h>
#include <cam/ctl/ctl.h>
#include <cam/ctl/ctl_frontend.h>
#include <cam/ctl/ctl_util.h>
#include <cam/ctl/ctl_backend.h>
#include <cam/ctl/ctl_ioctl.h>
#include <cam/ctl/ctl_ha.h>
#include <cam/ctl/ctl_private.h>
#include <cam/ctl/ctl_debug.h>
#include <cam/ctl/ctl_error.h>
#if (__FreeBSD_version < 1100000)
struct mbufq {
struct mbuf *head;
struct mbuf *tail;
};
static void
mbufq_init(struct mbufq *q, int limit)
{
q->head = q->tail = NULL;
}
static void
mbufq_drain(struct mbufq *q)
{
struct mbuf *m;
while ((m = q->head) != NULL) {
q->head = m->m_nextpkt;
m_freem(m);
}
q->tail = NULL;
}
static struct mbuf *
mbufq_dequeue(struct mbufq *q)
{
struct mbuf *m;
m = q->head;
if (m) {
if (q->tail == m)
q->tail = NULL;
q->head = m->m_nextpkt;
m->m_nextpkt = NULL;
}
return (m);
}
static void
mbufq_enqueue(struct mbufq *q, struct mbuf *m)
{
m->m_nextpkt = NULL;
if (q->tail)
q->tail->m_nextpkt = m;
else
q->head = m;
q->tail = m;
}
static u_int
sbavail(struct sockbuf *sb)
{
return (sb->sb_cc);
}
#if (__FreeBSD_version < 1000000)
#define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
#endif
#endif
struct ha_msg_wire {
uint32_t channel;
uint32_t length;
};
struct ha_dt_msg_wire {
ctl_ha_dt_cmd command;
uint32_t size;
uint8_t *local;
uint8_t *remote;
};
struct ha_softc {
struct ctl_softc *ha_ctl_softc;
ctl_evt_handler ha_handler[CTL_HA_CHAN_MAX];
char ha_peer[128];
struct sockaddr_in ha_peer_in;
struct socket *ha_lso;
struct socket *ha_so;
struct mbufq ha_sendq;
struct mbuf *ha_sending;
struct mtx ha_lock;
int ha_connect;
int ha_listen;
int ha_connected;
int ha_receiving;
int ha_wakeup;
int ha_disconnect;
int ha_shutdown;
eventhandler_tag ha_shutdown_eh;
TAILQ_HEAD(, ctl_ha_dt_req) ha_dts;
} ha_softc;
static void
ctl_ha_conn_wake(struct ha_softc *softc)
{
mtx_lock(&softc->ha_lock);
softc->ha_wakeup = 1;
mtx_unlock(&softc->ha_lock);
wakeup(&softc->ha_wakeup);
}
static int
ctl_ha_lupcall(struct socket *so, void *arg, int waitflag)
{
struct ha_softc *softc = arg;
ctl_ha_conn_wake(softc);
return (SU_OK);
}
static int
ctl_ha_rupcall(struct socket *so, void *arg, int waitflag)
{
struct ha_softc *softc = arg;
wakeup(&softc->ha_receiving);
return (SU_OK);
}
static int
ctl_ha_supcall(struct socket *so, void *arg, int waitflag)
{
struct ha_softc *softc = arg;
ctl_ha_conn_wake(softc);
return (SU_OK);
}
static void
ctl_ha_evt(struct ha_softc *softc, ctl_ha_channel ch, ctl_ha_event evt,
int param)
{
int i;
if (ch < CTL_HA_CHAN_MAX) {
if (softc->ha_handler[ch])
softc->ha_handler[ch](ch, evt, param);
return;
}
for (i = 0; i < CTL_HA_CHAN_MAX; i++) {
if (softc->ha_handler[i])
softc->ha_handler[i](i, evt, param);
}
}
static void
ctl_ha_close(struct ha_softc *softc)
{
struct socket *so = softc->ha_so;
int report = 0;
if (softc->ha_connected || softc->ha_disconnect) {
softc->ha_connected = 0;
mbufq_drain(&softc->ha_sendq);
m_freem(softc->ha_sending);
softc->ha_sending = NULL;
report = 1;
}
if (so) {
SOCKBUF_LOCK(&so->so_rcv);
soupcall_clear(so, SO_RCV);
while (softc->ha_receiving) {
wakeup(&softc->ha_receiving);
msleep(&softc->ha_receiving, SOCKBUF_MTX(&so->so_rcv),
0, "ha_rx exit", 0);
}
SOCKBUF_UNLOCK(&so->so_rcv);
SOCKBUF_LOCK(&so->so_snd);
soupcall_clear(so, SO_SND);
SOCKBUF_UNLOCK(&so->so_snd);
softc->ha_so = NULL;
if (softc->ha_connect)
pause("reconnect", hz / 2);
soclose(so);
}
if (report) {
ctl_ha_evt(softc, CTL_HA_CHAN_MAX, CTL_HA_EVT_LINK_CHANGE,
(softc->ha_connect || softc->ha_listen) ?
CTL_HA_LINK_UNKNOWN : CTL_HA_LINK_OFFLINE);
}
}
static void
ctl_ha_lclose(struct ha_softc *softc)
{
if (softc->ha_lso) {
SOCKBUF_LOCK(&softc->ha_lso->so_rcv);
soupcall_clear(softc->ha_lso, SO_RCV);
SOCKBUF_UNLOCK(&softc->ha_lso->so_rcv);
soclose(softc->ha_lso);
softc->ha_lso = NULL;
}
}
static void
ctl_ha_rx_thread(void *arg)
{
struct ha_softc *softc = arg;
struct socket *so = softc->ha_so;
struct ha_msg_wire wire_hdr;
struct uio uio;
struct iovec iov;
int error, flags, next;
bzero(&wire_hdr, sizeof(wire_hdr));
while (1) {
if (wire_hdr.length > 0)
next = wire_hdr.length;
else
next = sizeof(wire_hdr);
SOCKBUF_LOCK(&so->so_rcv);
while (sbavail(&so->so_rcv) < next || softc->ha_disconnect) {
if (softc->ha_connected == 0 || softc->ha_disconnect ||
so->so_error ||
(so->so_rcv.sb_state & SBS_CANTRCVMORE)) {
goto errout;
}
so->so_rcv.sb_lowat = next;
msleep(&softc->ha_receiving, SOCKBUF_MTX(&so->so_rcv),
0, "-", 0);
}
SOCKBUF_UNLOCK(&so->so_rcv);
if (wire_hdr.length == 0) {
iov.iov_base = &wire_hdr;
iov.iov_len = sizeof(wire_hdr);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_rw = UIO_READ;
uio.uio_segflg = UIO_SYSSPACE;
uio.uio_td = curthread;
uio.uio_resid = sizeof(wire_hdr);
flags = MSG_DONTWAIT;
error = soreceive(softc->ha_so, NULL, &uio, NULL,
NULL, &flags);
if (error != 0) {
printf("%s: header receive error %d\n",
__func__, error);
SOCKBUF_LOCK(&so->so_rcv);
goto errout;
}
} else {
ctl_ha_evt(softc, wire_hdr.channel,
CTL_HA_EVT_MSG_RECV, wire_hdr.length);
wire_hdr.length = 0;
}
}
errout:
softc->ha_receiving = 0;
wakeup(&softc->ha_receiving);
SOCKBUF_UNLOCK(&so->so_rcv);
ctl_ha_conn_wake(softc);
kthread_exit();
}
static void
ctl_ha_send(struct ha_softc *softc)
{
struct socket *so = softc->ha_so;
int error;
while (1) {
if (softc->ha_sending == NULL) {
mtx_lock(&softc->ha_lock);
softc->ha_sending = mbufq_dequeue(&softc->ha_sendq);
mtx_unlock(&softc->ha_lock);
if (softc->ha_sending == NULL) {
so->so_snd.sb_lowat = so->so_snd.sb_hiwat + 1;
break;
}
}
SOCKBUF_LOCK(&so->so_snd);
if (sbspace(&so->so_snd) < softc->ha_sending->m_pkthdr.len) {
so->so_snd.sb_lowat = softc->ha_sending->m_pkthdr.len;
SOCKBUF_UNLOCK(&so->so_snd);
break;
}
SOCKBUF_UNLOCK(&so->so_snd);
error = sosend(softc->ha_so, NULL, NULL, softc->ha_sending,
NULL, MSG_DONTWAIT, curthread);
softc->ha_sending = NULL;
if (error != 0) {
printf("%s: sosend() error %d\n", __func__, error);
return;
}
}
}
static void
ctl_ha_sock_setup(struct ha_softc *softc)
{
struct sockopt opt;
struct socket *so = softc->ha_so;
int error, val;
val = 1024 * 1024;
error = soreserve(so, val, val);
if (error)
printf("%s: soreserve failed %d\n", __func__, error);
SOCKBUF_LOCK(&so->so_rcv);
so->so_rcv.sb_lowat = sizeof(struct ha_msg_wire);
soupcall_set(so, SO_RCV, ctl_ha_rupcall, softc);
SOCKBUF_UNLOCK(&so->so_rcv);
SOCKBUF_LOCK(&so->so_snd);
so->so_snd.sb_lowat = sizeof(struct ha_msg_wire);
soupcall_set(so, SO_SND, ctl_ha_supcall, softc);
SOCKBUF_UNLOCK(&so->so_snd);
bzero(&opt, sizeof(struct sockopt));
opt.sopt_dir = SOPT_SET;
opt.sopt_level = SOL_SOCKET;
opt.sopt_name = SO_KEEPALIVE;
opt.sopt_val = &val;
opt.sopt_valsize = sizeof(val);
val = 1;
error = sosetopt(so, &opt);
if (error)
printf("%s: KEEPALIVE setting failed %d\n", __func__, error);
opt.sopt_level = IPPROTO_TCP;
opt.sopt_name = TCP_NODELAY;
val = 1;
error = sosetopt(so, &opt);
if (error)
printf("%s: NODELAY setting failed %d\n", __func__, error);
opt.sopt_name = TCP_KEEPINIT;
val = 3;
error = sosetopt(so, &opt);
if (error)
printf("%s: KEEPINIT setting failed %d\n", __func__, error);
opt.sopt_name = TCP_KEEPIDLE;
val = 1;
error = sosetopt(so, &opt);
if (error)
printf("%s: KEEPIDLE setting failed %d\n", __func__, error);
opt.sopt_name = TCP_KEEPINTVL;
val = 1;
error = sosetopt(so, &opt);
if (error)
printf("%s: KEEPINTVL setting failed %d\n", __func__, error);
opt.sopt_name = TCP_KEEPCNT;
val = 5;
error = sosetopt(so, &opt);
if (error)
printf("%s: KEEPCNT setting failed %d\n", __func__, error);
}
static int
ctl_ha_connect(struct ha_softc *softc)
{
struct thread *td = curthread;
struct sockaddr_in sa;
struct socket *so;
int error;
/* Create the socket */
error = socreate(PF_INET, &so, SOCK_STREAM,
IPPROTO_TCP, td->td_ucred, td);
if (error != 0) {
printf("%s: socreate() error %d\n", __func__, error);
return (error);
}
softc->ha_so = so;
ctl_ha_sock_setup(softc);
memcpy(&sa, &softc->ha_peer_in, sizeof(sa));
error = soconnect(so, (struct sockaddr *)&sa, td);
if (error != 0) {
if (bootverbose)
printf("%s: soconnect() error %d\n", __func__, error);
goto out;
}
return (0);
out:
ctl_ha_close(softc);
return (error);
}
static int
ctl_ha_accept(struct ha_softc *softc)
{
Listening sockets improvements. o Separate fields of struct socket that belong to listening from fields that belong to normal dataflow, and unionize them. This shrinks the structure a bit. - Take out selinfo's from the socket buffers into the socket. The first reason is to support braindamaged scenario when a socket is added to kevent(2) and then listen(2) is cast on it. The second reason is that there is future plan to make socket buffers pluggable, so that for a dataflow socket a socket buffer can be changed, and in this case we also want to keep same selinfos through the lifetime of a socket. - Remove struct struct so_accf. Since now listening stuff no longer affects struct socket size, just move its fields into listening part of the union. - Provide sol_upcall field and enforce that so_upcall_set() may be called only on a dataflow socket, which has buffers, and for listening sockets provide solisten_upcall_set(). o Remove ACCEPT_LOCK() global. - Add a mutex to socket, to be used instead of socket buffer lock to lock fields of struct socket that don't belong to a socket buffer. - Allow to acquire two socket locks, but the first one must belong to a listening socket. - Make soref()/sorele() to use atomic(9). This allows in some situations to do soref() without owning socket lock. There is place for improvement here, it is possible to make sorele() also to lock optionally. - Most protocols aren't touched by this change, except UNIX local sockets. See below for more information. o Reduce copy-and-paste in kernel modules that accept connections from listening sockets: provide function solisten_dequeue(), and use it in the following modules: ctl(4), iscsi(4), ng_btsocket(4), ng_ksocket(4), infiniband, rpc. o UNIX local sockets. - Removal of ACCEPT_LOCK() global uncovered several races in the UNIX local sockets. Most races exist around spawning a new socket, when we are connecting to a local listening socket. To cover them, we need to hold locks on both PCBs when spawning a third one. This means holding them across sonewconn(). This creates a LOR between pcb locks and unp_list_lock. - To fix the new LOR, abandon the global unp_list_lock in favor of global unp_link_lock. Indeed, separating these two locks didn't provide us any extra parralelism in the UNIX sockets. - Now call into uipc_attach() may happen with unp_link_lock hold if, we are accepting, or without unp_link_lock in case if we are just creating a socket. - Another problem in UNIX sockets is that uipc_close() basicly did nothing for a listening socket. The vnode remained opened for connections. This is fixed by removing vnode in uipc_close(). Maybe the right way would be to do it for all sockets (not only listening), simply move the vnode teardown from uipc_detach() to uipc_close()? Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D9770
2017-06-08 21:30:34 +00:00
struct socket *lso, *so;
struct sockaddr *sap;
int error;
Listening sockets improvements. o Separate fields of struct socket that belong to listening from fields that belong to normal dataflow, and unionize them. This shrinks the structure a bit. - Take out selinfo's from the socket buffers into the socket. The first reason is to support braindamaged scenario when a socket is added to kevent(2) and then listen(2) is cast on it. The second reason is that there is future plan to make socket buffers pluggable, so that for a dataflow socket a socket buffer can be changed, and in this case we also want to keep same selinfos through the lifetime of a socket. - Remove struct struct so_accf. Since now listening stuff no longer affects struct socket size, just move its fields into listening part of the union. - Provide sol_upcall field and enforce that so_upcall_set() may be called only on a dataflow socket, which has buffers, and for listening sockets provide solisten_upcall_set(). o Remove ACCEPT_LOCK() global. - Add a mutex to socket, to be used instead of socket buffer lock to lock fields of struct socket that don't belong to a socket buffer. - Allow to acquire two socket locks, but the first one must belong to a listening socket. - Make soref()/sorele() to use atomic(9). This allows in some situations to do soref() without owning socket lock. There is place for improvement here, it is possible to make sorele() also to lock optionally. - Most protocols aren't touched by this change, except UNIX local sockets. See below for more information. o Reduce copy-and-paste in kernel modules that accept connections from listening sockets: provide function solisten_dequeue(), and use it in the following modules: ctl(4), iscsi(4), ng_btsocket(4), ng_ksocket(4), infiniband, rpc. o UNIX local sockets. - Removal of ACCEPT_LOCK() global uncovered several races in the UNIX local sockets. Most races exist around spawning a new socket, when we are connecting to a local listening socket. To cover them, we need to hold locks on both PCBs when spawning a third one. This means holding them across sonewconn(). This creates a LOR between pcb locks and unp_list_lock. - To fix the new LOR, abandon the global unp_list_lock in favor of global unp_link_lock. Indeed, separating these two locks didn't provide us any extra parralelism in the UNIX sockets. - Now call into uipc_attach() may happen with unp_link_lock hold if, we are accepting, or without unp_link_lock in case if we are just creating a socket. - Another problem in UNIX sockets is that uipc_close() basicly did nothing for a listening socket. The vnode remained opened for connections. This is fixed by removing vnode in uipc_close(). Maybe the right way would be to do it for all sockets (not only listening), simply move the vnode teardown from uipc_detach() to uipc_close()? Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D9770
2017-06-08 21:30:34 +00:00
lso = softc->ha_lso;
SOLISTEN_LOCK(lso);
error = solisten_dequeue(lso, &so, 0);
if (error == EWOULDBLOCK)
return (error);
if (error) {
printf("%s: socket error %d\n", __func__, error);
goto out;
}
sap = NULL;
error = soaccept(so, &sap);
if (error != 0) {
printf("%s: soaccept() error %d\n", __func__, error);
if (sap != NULL)
free(sap, M_SONAME);
goto out;
}
if (sap != NULL)
free(sap, M_SONAME);
softc->ha_so = so;
ctl_ha_sock_setup(softc);
return (0);
out:
ctl_ha_lclose(softc);
return (error);
}
static int
ctl_ha_listen(struct ha_softc *softc)
{
struct thread *td = curthread;
struct sockaddr_in sa;
struct sockopt opt;
int error, val;
/* Create the socket */
if (softc->ha_lso == NULL) {
error = socreate(PF_INET, &softc->ha_lso, SOCK_STREAM,
IPPROTO_TCP, td->td_ucred, td);
if (error != 0) {
printf("%s: socreate() error %d\n", __func__, error);
return (error);
}
bzero(&opt, sizeof(struct sockopt));
opt.sopt_dir = SOPT_SET;
opt.sopt_level = SOL_SOCKET;
opt.sopt_name = SO_REUSEADDR;
opt.sopt_val = &val;
opt.sopt_valsize = sizeof(val);
val = 1;
error = sosetopt(softc->ha_lso, &opt);
if (error) {
printf("%s: REUSEADDR setting failed %d\n",
__func__, error);
}
bzero(&opt, sizeof(struct sockopt));
opt.sopt_dir = SOPT_SET;
opt.sopt_level = SOL_SOCKET;
opt.sopt_name = SO_REUSEPORT;
opt.sopt_val = &val;
opt.sopt_valsize = sizeof(val);
val = 1;
error = sosetopt(softc->ha_lso, &opt);
if (error) {
printf("%s: REUSEPORT setting failed %d\n",
__func__, error);
}
}
memcpy(&sa, &softc->ha_peer_in, sizeof(sa));
error = sobind(softc->ha_lso, (struct sockaddr *)&sa, td);
if (error != 0) {
printf("%s: sobind() error %d\n", __func__, error);
goto out;
}
error = solisten(softc->ha_lso, 1, td);
if (error != 0) {
printf("%s: solisten() error %d\n", __func__, error);
goto out;
}
Listening sockets improvements. o Separate fields of struct socket that belong to listening from fields that belong to normal dataflow, and unionize them. This shrinks the structure a bit. - Take out selinfo's from the socket buffers into the socket. The first reason is to support braindamaged scenario when a socket is added to kevent(2) and then listen(2) is cast on it. The second reason is that there is future plan to make socket buffers pluggable, so that for a dataflow socket a socket buffer can be changed, and in this case we also want to keep same selinfos through the lifetime of a socket. - Remove struct struct so_accf. Since now listening stuff no longer affects struct socket size, just move its fields into listening part of the union. - Provide sol_upcall field and enforce that so_upcall_set() may be called only on a dataflow socket, which has buffers, and for listening sockets provide solisten_upcall_set(). o Remove ACCEPT_LOCK() global. - Add a mutex to socket, to be used instead of socket buffer lock to lock fields of struct socket that don't belong to a socket buffer. - Allow to acquire two socket locks, but the first one must belong to a listening socket. - Make soref()/sorele() to use atomic(9). This allows in some situations to do soref() without owning socket lock. There is place for improvement here, it is possible to make sorele() also to lock optionally. - Most protocols aren't touched by this change, except UNIX local sockets. See below for more information. o Reduce copy-and-paste in kernel modules that accept connections from listening sockets: provide function solisten_dequeue(), and use it in the following modules: ctl(4), iscsi(4), ng_btsocket(4), ng_ksocket(4), infiniband, rpc. o UNIX local sockets. - Removal of ACCEPT_LOCK() global uncovered several races in the UNIX local sockets. Most races exist around spawning a new socket, when we are connecting to a local listening socket. To cover them, we need to hold locks on both PCBs when spawning a third one. This means holding them across sonewconn(). This creates a LOR between pcb locks and unp_list_lock. - To fix the new LOR, abandon the global unp_list_lock in favor of global unp_link_lock. Indeed, separating these two locks didn't provide us any extra parralelism in the UNIX sockets. - Now call into uipc_attach() may happen with unp_link_lock hold if, we are accepting, or without unp_link_lock in case if we are just creating a socket. - Another problem in UNIX sockets is that uipc_close() basicly did nothing for a listening socket. The vnode remained opened for connections. This is fixed by removing vnode in uipc_close(). Maybe the right way would be to do it for all sockets (not only listening), simply move the vnode teardown from uipc_detach() to uipc_close()? Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D9770
2017-06-08 21:30:34 +00:00
SOLISTEN_LOCK(softc->ha_lso);
softc->ha_lso->so_state |= SS_NBIO;
solisten_upcall_set(softc->ha_lso, ctl_ha_lupcall, softc);
SOLISTEN_UNLOCK(softc->ha_lso);
return (0);
out:
ctl_ha_lclose(softc);
return (error);
}
static void
ctl_ha_conn_thread(void *arg)
{
struct ha_softc *softc = arg;
int error;
while (1) {
if (softc->ha_disconnect || softc->ha_shutdown) {
ctl_ha_close(softc);
if (softc->ha_disconnect == 2 || softc->ha_shutdown)
ctl_ha_lclose(softc);
softc->ha_disconnect = 0;
if (softc->ha_shutdown)
break;
} else if (softc->ha_so != NULL &&
(softc->ha_so->so_error ||
softc->ha_so->so_rcv.sb_state & SBS_CANTRCVMORE))
ctl_ha_close(softc);
if (softc->ha_so == NULL) {
if (softc->ha_lso != NULL)
ctl_ha_accept(softc);
else if (softc->ha_listen)
ctl_ha_listen(softc);
else if (softc->ha_connect)
ctl_ha_connect(softc);
}
if (softc->ha_so != NULL) {
if (softc->ha_connected == 0 &&
softc->ha_so->so_error == 0 &&
(softc->ha_so->so_state & SS_ISCONNECTING) == 0) {
softc->ha_connected = 1;
ctl_ha_evt(softc, CTL_HA_CHAN_MAX,
CTL_HA_EVT_LINK_CHANGE,
CTL_HA_LINK_ONLINE);
softc->ha_receiving = 1;
error = kproc_kthread_add(ctl_ha_rx_thread,
softc, &softc->ha_ctl_softc->ctl_proc,
NULL, 0, 0, "ctl", "ha_rx");
if (error != 0) {
printf("Error creating CTL HA rx thread!\n");
softc->ha_receiving = 0;
softc->ha_disconnect = 1;
}
}
ctl_ha_send(softc);
}
mtx_lock(&softc->ha_lock);
if (softc->ha_so != NULL &&
(softc->ha_so->so_error ||
softc->ha_so->so_rcv.sb_state & SBS_CANTRCVMORE))
;
else if (!softc->ha_wakeup)
msleep(&softc->ha_wakeup, &softc->ha_lock, 0, "-", hz);
softc->ha_wakeup = 0;
mtx_unlock(&softc->ha_lock);
}
mtx_lock(&softc->ha_lock);
softc->ha_shutdown = 2;
wakeup(&softc->ha_wakeup);
mtx_unlock(&softc->ha_lock);
kthread_exit();
}
static int
ctl_ha_peer_sysctl(SYSCTL_HANDLER_ARGS)
{
struct ha_softc *softc = (struct ha_softc *)arg1;
struct sockaddr_in *sa;
int error, b1, b2, b3, b4, p, num;
char buf[128];
strlcpy(buf, softc->ha_peer, sizeof(buf));
error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
if ((error != 0) || (req->newptr == NULL) ||
strncmp(buf, softc->ha_peer, sizeof(buf)) == 0)
return (error);
sa = &softc->ha_peer_in;
mtx_lock(&softc->ha_lock);
if ((num = sscanf(buf, "connect %d.%d.%d.%d:%d",
&b1, &b2, &b3, &b4, &p)) >= 4) {
softc->ha_connect = 1;
softc->ha_listen = 0;
} else if ((num = sscanf(buf, "listen %d.%d.%d.%d:%d",
&b1, &b2, &b3, &b4, &p)) >= 4) {
softc->ha_connect = 0;
softc->ha_listen = 1;
} else {
softc->ha_connect = 0;
softc->ha_listen = 0;
if (buf[0] != 0) {
buf[0] = 0;
error = EINVAL;
}
}
strlcpy(softc->ha_peer, buf, sizeof(softc->ha_peer));
if (softc->ha_connect || softc->ha_listen) {
memset(sa, 0, sizeof(*sa));
sa->sin_len = sizeof(struct sockaddr_in);
sa->sin_family = AF_INET;
sa->sin_port = htons((num >= 5) ? p : 999);
sa->sin_addr.s_addr =
htonl((b1 << 24) + (b2 << 16) + (b3 << 8) + b4);
}
softc->ha_disconnect = 2;
softc->ha_wakeup = 1;
mtx_unlock(&softc->ha_lock);
wakeup(&softc->ha_wakeup);
return (error);
}
ctl_ha_status
ctl_ha_msg_register(ctl_ha_channel channel, ctl_evt_handler handler)
{
struct ha_softc *softc = &ha_softc;
KASSERT(channel < CTL_HA_CHAN_MAX,
("Wrong CTL HA channel %d", channel));
softc->ha_handler[channel] = handler;
return (CTL_HA_STATUS_SUCCESS);
}
ctl_ha_status
ctl_ha_msg_deregister(ctl_ha_channel channel)
{
struct ha_softc *softc = &ha_softc;
KASSERT(channel < CTL_HA_CHAN_MAX,
("Wrong CTL HA channel %d", channel));
softc->ha_handler[channel] = NULL;
return (CTL_HA_STATUS_SUCCESS);
}
/*
* Receive a message of the specified size.
*/
ctl_ha_status
ctl_ha_msg_recv(ctl_ha_channel channel, void *addr, size_t len,
int wait)
{
struct ha_softc *softc = &ha_softc;
struct uio uio;
struct iovec iov;
int error, flags;
if (!softc->ha_connected)
return (CTL_HA_STATUS_DISCONNECT);
iov.iov_base = addr;
iov.iov_len = len;
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_rw = UIO_READ;
uio.uio_segflg = UIO_SYSSPACE;
uio.uio_td = curthread;
uio.uio_resid = len;
flags = wait ? 0 : MSG_DONTWAIT;
error = soreceive(softc->ha_so, NULL, &uio, NULL, NULL, &flags);
if (error == 0)
return (CTL_HA_STATUS_SUCCESS);
/* Consider all errors fatal for HA sanity. */
mtx_lock(&softc->ha_lock);
if (softc->ha_connected) {
softc->ha_disconnect = 1;
softc->ha_wakeup = 1;
wakeup(&softc->ha_wakeup);
}
mtx_unlock(&softc->ha_lock);
return (CTL_HA_STATUS_ERROR);
}
/*
* Send a message of the specified size.
*/
ctl_ha_status
ctl_ha_msg_send2(ctl_ha_channel channel, const void *addr, size_t len,
const void *addr2, size_t len2, int wait)
{
struct ha_softc *softc = &ha_softc;
struct mbuf *mb, *newmb;
struct ha_msg_wire hdr;
size_t copylen, off;
if (!softc->ha_connected)
return (CTL_HA_STATUS_DISCONNECT);
newmb = m_getm2(NULL, sizeof(hdr) + len + len2, wait, MT_DATA,
M_PKTHDR);
if (newmb == NULL) {
/* Consider all errors fatal for HA sanity. */
mtx_lock(&softc->ha_lock);
if (softc->ha_connected) {
softc->ha_disconnect = 1;
softc->ha_wakeup = 1;
wakeup(&softc->ha_wakeup);
}
mtx_unlock(&softc->ha_lock);
printf("%s: Can't allocate mbuf chain\n", __func__);
return (CTL_HA_STATUS_ERROR);
}
hdr.channel = channel;
hdr.length = len + len2;
mb = newmb;
memcpy(mtodo(mb, 0), &hdr, sizeof(hdr));
mb->m_len += sizeof(hdr);
off = 0;
for (; mb != NULL && off < len; mb = mb->m_next) {
copylen = min(M_TRAILINGSPACE(mb), len - off);
memcpy(mtodo(mb, mb->m_len), (const char *)addr + off, copylen);
mb->m_len += copylen;
off += copylen;
if (off == len)
break;
}
KASSERT(off == len, ("%s: off (%zu) != len (%zu)", __func__,
off, len));
off = 0;
for (; mb != NULL && off < len2; mb = mb->m_next) {
copylen = min(M_TRAILINGSPACE(mb), len2 - off);
memcpy(mtodo(mb, mb->m_len), (const char *)addr2 + off, copylen);
mb->m_len += copylen;
off += copylen;
}
KASSERT(off == len2, ("%s: off (%zu) != len2 (%zu)", __func__,
off, len2));
newmb->m_pkthdr.len = sizeof(hdr) + len + len2;
mtx_lock(&softc->ha_lock);
if (!softc->ha_connected) {
mtx_unlock(&softc->ha_lock);
m_freem(newmb);
return (CTL_HA_STATUS_DISCONNECT);
}
mbufq_enqueue(&softc->ha_sendq, newmb);
softc->ha_wakeup = 1;
mtx_unlock(&softc->ha_lock);
wakeup(&softc->ha_wakeup);
return (CTL_HA_STATUS_SUCCESS);
}
ctl_ha_status
ctl_ha_msg_send(ctl_ha_channel channel, const void *addr, size_t len,
int wait)
{
return (ctl_ha_msg_send2(channel, addr, len, NULL, 0, wait));
}
ctl_ha_status
ctl_ha_msg_abort(ctl_ha_channel channel)
{
struct ha_softc *softc = &ha_softc;
mtx_lock(&softc->ha_lock);
softc->ha_disconnect = 1;
softc->ha_wakeup = 1;
mtx_unlock(&softc->ha_lock);
wakeup(&softc->ha_wakeup);
return (CTL_HA_STATUS_SUCCESS);
}
/*
* Allocate a data transfer request structure.
*/
struct ctl_ha_dt_req *
ctl_dt_req_alloc(void)
{
return (malloc(sizeof(struct ctl_ha_dt_req), M_CTL, M_WAITOK | M_ZERO));
}
/*
* Free a data transfer request structure.
*/
void
ctl_dt_req_free(struct ctl_ha_dt_req *req)
{
free(req, M_CTL);
}
/*
* Issue a DMA request for a single buffer.
*/
ctl_ha_status
ctl_dt_single(struct ctl_ha_dt_req *req)
{
struct ha_softc *softc = &ha_softc;
struct ha_dt_msg_wire wire_dt;
ctl_ha_status status;
wire_dt.command = req->command;
wire_dt.size = req->size;
wire_dt.local = req->local;
wire_dt.remote = req->remote;
if (req->command == CTL_HA_DT_CMD_READ && req->callback != NULL) {
mtx_lock(&softc->ha_lock);
TAILQ_INSERT_TAIL(&softc->ha_dts, req, links);
mtx_unlock(&softc->ha_lock);
ctl_ha_msg_send(CTL_HA_CHAN_DATA, &wire_dt, sizeof(wire_dt),
M_WAITOK);
return (CTL_HA_STATUS_WAIT);
}
if (req->command == CTL_HA_DT_CMD_READ) {
status = ctl_ha_msg_send(CTL_HA_CHAN_DATA, &wire_dt,
sizeof(wire_dt), M_WAITOK);
} else {
status = ctl_ha_msg_send2(CTL_HA_CHAN_DATA, &wire_dt,
sizeof(wire_dt), req->local, req->size, M_WAITOK);
}
return (status);
}
static void
ctl_dt_event_handler(ctl_ha_channel channel, ctl_ha_event event, int param)
{
struct ha_softc *softc = &ha_softc;
struct ctl_ha_dt_req *req;
ctl_ha_status isc_status;
if (event == CTL_HA_EVT_MSG_RECV) {
struct ha_dt_msg_wire wire_dt;
uint8_t *tmp;
int size;
size = min(sizeof(wire_dt), param);
isc_status = ctl_ha_msg_recv(CTL_HA_CHAN_DATA, &wire_dt,
size, M_WAITOK);
if (isc_status != CTL_HA_STATUS_SUCCESS) {
printf("%s: Error receiving message: %d\n",
__func__, isc_status);
return;
}
if (wire_dt.command == CTL_HA_DT_CMD_READ) {
wire_dt.command = CTL_HA_DT_CMD_WRITE;
tmp = wire_dt.local;
wire_dt.local = wire_dt.remote;
wire_dt.remote = tmp;
ctl_ha_msg_send2(CTL_HA_CHAN_DATA, &wire_dt,
sizeof(wire_dt), wire_dt.local, wire_dt.size,
M_WAITOK);
} else if (wire_dt.command == CTL_HA_DT_CMD_WRITE) {
isc_status = ctl_ha_msg_recv(CTL_HA_CHAN_DATA,
wire_dt.remote, wire_dt.size, M_WAITOK);
mtx_lock(&softc->ha_lock);
TAILQ_FOREACH(req, &softc->ha_dts, links) {
if (req->local == wire_dt.remote) {
TAILQ_REMOVE(&softc->ha_dts, req, links);
break;
}
}
mtx_unlock(&softc->ha_lock);
if (req) {
req->ret = isc_status;
req->callback(req);
}
}
} else if (event == CTL_HA_EVT_LINK_CHANGE) {
CTL_DEBUG_PRINT(("%s: Link state change to %d\n", __func__,
param));
if (param != CTL_HA_LINK_ONLINE) {
mtx_lock(&softc->ha_lock);
while ((req = TAILQ_FIRST(&softc->ha_dts)) != NULL) {
TAILQ_REMOVE(&softc->ha_dts, req, links);
mtx_unlock(&softc->ha_lock);
req->ret = CTL_HA_STATUS_DISCONNECT;
req->callback(req);
mtx_lock(&softc->ha_lock);
}
mtx_unlock(&softc->ha_lock);
}
} else {
printf("%s: Unknown event %d\n", __func__, event);
}
}
ctl_ha_status
ctl_ha_msg_init(struct ctl_softc *ctl_softc)
{
struct ha_softc *softc = &ha_softc;
int error;
softc->ha_ctl_softc = ctl_softc;
mtx_init(&softc->ha_lock, "CTL HA mutex", NULL, MTX_DEF);
mbufq_init(&softc->ha_sendq, INT_MAX);
TAILQ_INIT(&softc->ha_dts);
error = kproc_kthread_add(ctl_ha_conn_thread, softc,
&ctl_softc->ctl_proc, NULL, 0, 0, "ctl", "ha_tx");
if (error != 0) {
printf("error creating CTL HA connection thread!\n");
mtx_destroy(&softc->ha_lock);
return (CTL_HA_STATUS_ERROR);
}
softc->ha_shutdown_eh = EVENTHANDLER_REGISTER(shutdown_pre_sync,
ctl_ha_msg_shutdown, ctl_softc, SHUTDOWN_PRI_FIRST);
SYSCTL_ADD_PROC(&ctl_softc->sysctl_ctx,
SYSCTL_CHILDREN(ctl_softc->sysctl_tree),
OID_AUTO, "ha_peer", CTLTYPE_STRING | CTLFLAG_RWTUN,
softc, 0, ctl_ha_peer_sysctl, "A", "HA peer connection method");
if (ctl_ha_msg_register(CTL_HA_CHAN_DATA, ctl_dt_event_handler)
!= CTL_HA_STATUS_SUCCESS) {
printf("%s: ctl_ha_msg_register failed.\n", __func__);
}
return (CTL_HA_STATUS_SUCCESS);
};
void
ctl_ha_msg_shutdown(struct ctl_softc *ctl_softc)
{
struct ha_softc *softc = &ha_softc;
/* Disconnect and shutdown threads. */
mtx_lock(&softc->ha_lock);
if (softc->ha_shutdown < 2) {
softc->ha_shutdown = 1;
softc->ha_wakeup = 1;
wakeup(&softc->ha_wakeup);
while (softc->ha_shutdown < 2 && !SCHEDULER_STOPPED()) {
msleep(&softc->ha_wakeup, &softc->ha_lock, 0,
"shutdown", hz);
}
}
mtx_unlock(&softc->ha_lock);
};
ctl_ha_status
ctl_ha_msg_destroy(struct ctl_softc *ctl_softc)
{
struct ha_softc *softc = &ha_softc;
if (softc->ha_shutdown_eh != NULL) {
EVENTHANDLER_DEREGISTER(shutdown_pre_sync,
softc->ha_shutdown_eh);
softc->ha_shutdown_eh = NULL;
}
ctl_ha_msg_shutdown(ctl_softc); /* Just in case. */
if (ctl_ha_msg_deregister(CTL_HA_CHAN_DATA) != CTL_HA_STATUS_SUCCESS)
printf("%s: ctl_ha_msg_deregister failed.\n", __func__);
mtx_destroy(&softc->ha_lock);
return (CTL_HA_STATUS_SUCCESS);
};