inline uint32_t
EventRingBuffer<Time>::write(Time time, EventType type, uint32_t size, const uint8_t* buf)
{
if (write_space() < (sizeof(Time) + sizeof(EventType) + sizeof(uint32_t) + size)) {
return 0;
} else {
PBD::RingBufferNPT<uint8_t>::write ((uint8_t*)&time, sizeof(Time));
PBD::RingBufferNPT<uint8_t>::write ((uint8_t*)&type, sizeof(EventType));
PBD::RingBufferNPT<uint8_t>::write ((uint8_t*)&size, sizeof(uint32_t));
PBD::RingBufferNPT<uint8_t>::write (buf, size);
return size;
}
}
/* elf ----------------------------------------------------------------------*/
void
elf(int fd_s, int fd_d, Elf32_Word p_type, Elf32_Word p_flags,
Elf32_Addr* addr, int* size, int flag)
{
int i;
Elf32_Ehdr e_eh;
static Elf32_Phdr e_ph[32];
lseek(fd_s, 0, SEEK_SET); /* rewind */
readn(fd_s, &e_eh, sizeof(Elf32_Ehdr));
lseek(fd_s, e_eh.e_phoff, SEEK_SET);
readn(fd_s, (char*)e_ph, e_eh.e_phentsize * e_eh.e_phnum);
printf("#-----------------------------------------------------------------#\n");
printf("# [t]ype, [f]lag, [v]addr, [m]emsz, [s]filesz, [o]ffset #\n");
printf("# e_eh.e_phnum = %d\n", e_eh.e_phnum);
for (i = 0 ; i < e_eh.e_phnum ; i ++) {
printf("# ph[%d]: t = 0x%x, f = 0x%x, v = 0x%x, m = 0x%x, s = 0x%x o = %d\n",
i, e_ph[i].p_type, e_ph[i].p_flags, e_ph[i].p_vaddr, e_ph[i].p_memsz,
e_ph[i].p_filesz, e_ph[i].p_offset);
if (e_ph[i].p_type == p_type &&
e_ph[i].p_flags == p_flags) {
/* do everything ------------------------------------*/
if (flag) {
write_space(fd_d,
(int)(e_ph[i].p_vaddr - *addr) - *size);
}
lseek(fd_s, e_ph[i].p_offset, SEEK_SET);
readn(fd_s, buf, e_ph[i].p_filesz);
writen(fd_d, buf, e_ph[i].p_filesz);
*addr = e_ph[i].p_vaddr;
*size = e_ph[i].p_memsz;
printf("# virtual address = 0x%x, size = 0x%x\n", *addr, *size);
break;
}
}
}
/* the core send_sem serializes this with other xmit and shutdown */
static int rds_tcp_sendmsg(struct socket *sock, void *data, unsigned int len)
{
struct kvec vec = {
.iov_base = data,
.iov_len = len,
};
struct msghdr msg = {
.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
};
return kernel_sendmsg(sock, &msg, &vec, 1, vec.iov_len);
}
/* the core send_sem serializes this with other xmit and shutdown */
int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
unsigned int hdr_off, unsigned int sg, unsigned int off)
{
struct rds_conn_path *cp = rm->m_inc.i_conn_path;
struct rds_tcp_connection *tc = cp->cp_transport_data;
int done = 0;
int ret = 0;
int more;
if (hdr_off == 0) {
/*
* m_ack_seq is set to the sequence number of the last byte of
* header and data. see rds_tcp_is_acked().
*/
tc->t_last_sent_nxt = rds_tcp_snd_nxt(tc);
rm->m_ack_seq = tc->t_last_sent_nxt +
sizeof(struct rds_header) +
be32_to_cpu(rm->m_inc.i_hdr.h_len) - 1;
smp_mb__before_atomic();
set_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags);
tc->t_last_expected_una = rm->m_ack_seq + 1;
if (test_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags))
rm->m_inc.i_hdr.h_flags |= RDS_FLAG_RETRANSMITTED;
rdsdebug("rm %p tcp nxt %u ack_seq %llu\n",
rm, rds_tcp_snd_nxt(tc),
(unsigned long long)rm->m_ack_seq);
}
if (hdr_off < sizeof(struct rds_header)) {
/* see rds_tcp_write_space() */
set_bit(SOCK_NOSPACE, &tc->t_sock->sk->sk_socket->flags);
ret = rds_tcp_sendmsg(tc->t_sock,
(void *)&rm->m_inc.i_hdr + hdr_off,
sizeof(rm->m_inc.i_hdr) - hdr_off);
if (ret < 0)
goto out;
done += ret;
if (hdr_off + done != sizeof(struct rds_header))
goto out;
}
more = rm->data.op_nents > 1 ? (MSG_MORE | MSG_SENDPAGE_NOTLAST) : 0;
while (sg < rm->data.op_nents) {
int flags = MSG_DONTWAIT | MSG_NOSIGNAL | more;
ret = tc->t_sock->ops->sendpage(tc->t_sock,
sg_page(&rm->data.op_sg[sg]),
rm->data.op_sg[sg].offset + off,
rm->data.op_sg[sg].length - off,
flags);
rdsdebug("tcp sendpage %p:%u:%u ret %d\n", (void *)sg_page(&rm->data.op_sg[sg]),
rm->data.op_sg[sg].offset + off, rm->data.op_sg[sg].length - off,
ret);
if (ret <= 0)
break;
off += ret;
done += ret;
if (off == rm->data.op_sg[sg].length) {
off = 0;
sg++;
}
if (sg == rm->data.op_nents - 1)
more = 0;
}
out:
if (ret <= 0) {
/* write_space will hit after EAGAIN, all else fatal */
if (ret == -EAGAIN) {
rds_tcp_stats_inc(s_tcp_sndbuf_full);
ret = 0;
} else {
/* No need to disconnect/reconnect if path_drop
* has already been triggered, because, e.g., of
* an incoming RST.
*/
if (rds_conn_path_up(cp)) {
pr_warn("RDS/tcp: send to %pI4 on cp [%d]"
"returned %d, "
"disconnecting and reconnecting\n",
&conn->c_faddr, cp->cp_index, ret);
rds_conn_path_drop(cp);
}
}
}
if (done == 0)
done = ret;
return done;
}
/*
* rm->m_ack_seq is set to the tcp sequence number that corresponds to the
* last byte of the message, including the header. This means that the
* entire message has been received if rm->m_ack_seq is "before" the next
* unacked byte of the TCP sequence space. We have to do very careful
* wrapping 32bit comparisons here.
*/
static int rds_tcp_is_acked(struct rds_message *rm, uint64_t ack)
{
if (!test_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags))
return 0;
return (__s32)((u32)rm->m_ack_seq - (u32)ack) < 0;
}
void rds_tcp_write_space(struct sock *sk)
{
void (*write_space)(struct sock *sk);
struct rds_conn_path *cp;
struct rds_tcp_connection *tc;
read_lock_bh(&sk->sk_callback_lock);
cp = sk->sk_user_data;
if (!cp) {
write_space = sk->sk_write_space;
goto out;
}
tc = cp->cp_transport_data;
rdsdebug("write_space for tc %p\n", tc);
write_space = tc->t_orig_write_space;
rds_tcp_stats_inc(s_tcp_write_space_calls);
rdsdebug("tcp una %u\n", rds_tcp_snd_una(tc));
tc->t_last_seen_una = rds_tcp_snd_una(tc);
rds_send_path_drop_acked(cp, rds_tcp_snd_una(tc), rds_tcp_is_acked);
if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf)
queue_delayed_work(rds_wq, &cp->cp_send_w, 0);
out:
read_unlock_bh(&sk->sk_callback_lock);
/*
* write_space is only called when data leaves tcp's send queue if
* SOCK_NOSPACE is set. We set SOCK_NOSPACE every time we put
* data in tcp's send queue because we use write_space to parse the
* sequence numbers and notice that rds messages have been fully
* received.
*
* tcp's write_space clears SOCK_NOSPACE if the send queue has more
* than a certain amount of space. So we need to set it again *after*
* we call tcp's write_space or else we might only get called on the
* first of a series of incoming tcp acks.
*/
write_space(sk);
if (sk->sk_socket)
set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
}
/* the core send_sem serializes this with other xmit and shutdown */
int rds_tcp_sendmsg(struct socket *sock, void *data, unsigned int len)
{
struct kvec vec = {
.iov_base = data,
.iov_len = len,
};
struct msghdr msg = {
.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
};
return kernel_sendmsg(sock, &msg, &vec, 1, vec.iov_len);
}
/* the core send_sem serializes this with other xmit and shutdown */
int rds_tcp_xmit_cong_map(struct rds_connection *conn,
struct rds_cong_map *map, unsigned long offset)
{
static struct rds_header rds_tcp_map_header = {
.h_flags = RDS_FLAG_CONG_BITMAP,
};
struct rds_tcp_connection *tc = conn->c_transport_data;
unsigned long i;
int ret;
int copied = 0;
/* Some problem claims cpu_to_be32(constant) isn't a constant. */
rds_tcp_map_header.h_len = cpu_to_be32(RDS_CONG_MAP_BYTES);
if (offset < sizeof(struct rds_header)) {
ret = rds_tcp_sendmsg(tc->t_sock,
(void *)&rds_tcp_map_header + offset,
sizeof(struct rds_header) - offset);
if (ret <= 0)
return ret;
offset += ret;
copied = ret;
if (offset < sizeof(struct rds_header))
return ret;
}
offset -= sizeof(struct rds_header);
i = offset / PAGE_SIZE;
offset = offset % PAGE_SIZE;
BUG_ON(i >= RDS_CONG_MAP_PAGES);
do {
ret = tc->t_sock->ops->sendpage(tc->t_sock,
virt_to_page(map->m_page_addrs[i]),
offset, PAGE_SIZE - offset,
MSG_DONTWAIT);
if (ret <= 0)
break;
copied += ret;
offset += ret;
if (offset == PAGE_SIZE) {
offset = 0;
i++;
}
} while (i < RDS_CONG_MAP_PAGES);
return copied ? copied : ret;
}
/* the core send_sem serializes this with other xmit and shutdown */
int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
unsigned int hdr_off, unsigned int sg, unsigned int off)
{
struct rds_tcp_connection *tc = conn->c_transport_data;
int done = 0;
int ret = 0;
if (hdr_off == 0) {
/*
* m_ack_seq is set to the sequence number of the last byte of
* header and data. see rds_tcp_is_acked().
*/
tc->t_last_sent_nxt = rds_tcp_snd_nxt(tc);
rm->m_ack_seq = tc->t_last_sent_nxt +
sizeof(struct rds_header) +
be32_to_cpu(rm->m_inc.i_hdr.h_len) - 1;
smp_mb__before_clear_bit();
set_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags);
tc->t_last_expected_una = rm->m_ack_seq + 1;
rdsdebug("rm %p tcp nxt %u ack_seq %llu\n",
rm, rds_tcp_snd_nxt(tc),
(unsigned long long)rm->m_ack_seq);
}
if (hdr_off < sizeof(struct rds_header)) {
/* see rds_tcp_write_space() */
set_bit(SOCK_NOSPACE, &tc->t_sock->sk->sk_socket->flags);
ret = rds_tcp_sendmsg(tc->t_sock,
(void *)&rm->m_inc.i_hdr + hdr_off,
sizeof(rm->m_inc.i_hdr) - hdr_off);
if (ret < 0)
goto out;
done += ret;
if (hdr_off + done != sizeof(struct rds_header))
goto out;
}
while (sg < rm->m_nents) {
ret = tc->t_sock->ops->sendpage(tc->t_sock,
sg_page(&rm->m_sg[sg]),
rm->m_sg[sg].offset + off,
rm->m_sg[sg].length - off,
MSG_DONTWAIT|MSG_NOSIGNAL);
rdsdebug("tcp sendpage %p:%u:%u ret %d\n", (void *)sg_page(&rm->m_sg[sg]),
rm->m_sg[sg].offset + off, rm->m_sg[sg].length - off,
ret);
if (ret <= 0)
break;
off += ret;
done += ret;
if (off == rm->m_sg[sg].length) {
off = 0;
sg++;
}
}
out:
if (ret <= 0) {
/* write_space will hit after EAGAIN, all else fatal */
if (ret == -EAGAIN) {
rds_tcp_stats_inc(s_tcp_sndbuf_full);
ret = 0;
} else {
printk(KERN_WARNING "RDS/tcp: send to %pI4 "
"returned %d, disconnecting and reconnecting\n",
&conn->c_faddr, ret);
rds_conn_drop(conn);
}
}
if (done == 0)
done = ret;
return done;
}
/*
* rm->m_ack_seq is set to the tcp sequence number that corresponds to the
* last byte of the message, including the header. This means that the
* entire message has been received if rm->m_ack_seq is "before" the next
* unacked byte of the TCP sequence space. We have to do very careful
* wrapping 32bit comparisons here.
*/
static int rds_tcp_is_acked(struct rds_message *rm, uint64_t ack)
{
if (!test_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags))
return 0;
return (__s32)((u32)rm->m_ack_seq - (u32)ack) < 0;
}
void rds_tcp_write_space(struct sock *sk)
{
void (*write_space)(struct sock *sk);
struct rds_connection *conn;
struct rds_tcp_connection *tc;
read_lock_bh(&sk->sk_callback_lock);
conn = sk->sk_user_data;
if (conn == NULL) {
write_space = sk->sk_write_space;
goto out;
}
tc = conn->c_transport_data;
rdsdebug("write_space for tc %p\n", tc);
write_space = tc->t_orig_write_space;
rds_tcp_stats_inc(s_tcp_write_space_calls);
rdsdebug("tcp una %u\n", rds_tcp_snd_una(tc));
tc->t_last_seen_una = rds_tcp_snd_una(tc);
rds_send_drop_acked(conn, rds_tcp_snd_una(tc), rds_tcp_is_acked);
if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf)
queue_delayed_work(rds_wq, &conn->c_send_w, 0);
out:
read_unlock_bh(&sk->sk_callback_lock);
/*
* write_space is only called when data leaves tcp's send queue if
* SOCK_NOSPACE is set. We set SOCK_NOSPACE every time we put
* data in tcp's send queue because we use write_space to parse the
* sequence numbers and notice that rds messages have been fully
* received.
*
* tcp's write_space clears SOCK_NOSPACE if the send queue has more
* than a certain amount of space. So we need to set it again *after*
* we call tcp's write_space or else we might only get called on the
* first of a series of incoming tcp acks.
*/
write_space(sk);
if (sk->sk_socket)
set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
}
static int rds_tcp_sendmsg(struct socket *sock, void *data, unsigned int len)
{
struct kvec vec = {
.iov_base = data,
.iov_len = len,
};
struct msghdr msg = {
.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
};
return kernel_sendmsg(sock, &msg, &vec, 1, vec.iov_len);
}
int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
unsigned int hdr_off, unsigned int sg, unsigned int off)
{
struct rds_tcp_connection *tc = conn->c_transport_data;
int done = 0;
int ret = 0;
if (hdr_off == 0) {
tc->t_last_sent_nxt = rds_tcp_snd_nxt(tc);
rm->m_ack_seq = tc->t_last_sent_nxt +
sizeof(struct rds_header) +
be32_to_cpu(rm->m_inc.i_hdr.h_len) - 1;
smp_mb__before_clear_bit();
set_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags);
tc->t_last_expected_una = rm->m_ack_seq + 1;
rdsdebug("rm %p tcp nxt %u ack_seq %llu\n",
rm, rds_tcp_snd_nxt(tc),
(unsigned long long)rm->m_ack_seq);
}
if (hdr_off < sizeof(struct rds_header)) {
set_bit(SOCK_NOSPACE, &tc->t_sock->sk->sk_socket->flags);
ret = rds_tcp_sendmsg(tc->t_sock,
(void *)&rm->m_inc.i_hdr + hdr_off,
sizeof(rm->m_inc.i_hdr) - hdr_off);
if (ret < 0)
goto out;
done += ret;
if (hdr_off + done != sizeof(struct rds_header))
goto out;
}
while (sg < rm->data.op_nents) {
ret = tc->t_sock->ops->sendpage(tc->t_sock,
sg_page(&rm->data.op_sg[sg]),
rm->data.op_sg[sg].offset + off,
rm->data.op_sg[sg].length - off,
MSG_DONTWAIT|MSG_NOSIGNAL);
rdsdebug("tcp sendpage %p:%u:%u ret %d\n", (void *)sg_page(&rm->data.op_sg[sg]),
rm->data.op_sg[sg].offset + off, rm->data.op_sg[sg].length - off,
ret);
if (ret <= 0)
break;
off += ret;
done += ret;
if (off == rm->data.op_sg[sg].length) {
off = 0;
sg++;
}
}
out:
if (ret <= 0) {
if (ret == -EAGAIN) {
rds_tcp_stats_inc(s_tcp_sndbuf_full);
ret = 0;
} else {
printk(KERN_WARNING "RDS/tcp: send to %pI4 "
"returned %d, disconnecting and reconnecting\n",
&conn->c_faddr, ret);
rds_conn_drop(conn);
}
}
if (done == 0)
done = ret;
return done;
}
static int rds_tcp_is_acked(struct rds_message *rm, uint64_t ack)
{
if (!test_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags))
return 0;
return (__s32)((u32)rm->m_ack_seq - (u32)ack) < 0;
}
void rds_tcp_write_space(struct sock *sk)
{
void (*write_space)(struct sock *sk);
struct rds_connection *conn;
struct rds_tcp_connection *tc;
read_lock_bh(&sk->sk_callback_lock);
conn = sk->sk_user_data;
if (!conn) {
write_space = sk->sk_write_space;
goto out;
}
tc = conn->c_transport_data;
rdsdebug("write_space for tc %p\n", tc);
write_space = tc->t_orig_write_space;
rds_tcp_stats_inc(s_tcp_write_space_calls);
rdsdebug("tcp una %u\n", rds_tcp_snd_una(tc));
tc->t_last_seen_una = rds_tcp_snd_una(tc);
rds_send_drop_acked(conn, rds_tcp_snd_una(tc), rds_tcp_is_acked);
if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf)
queue_delayed_work(rds_wq, &conn->c_send_w, 0);
out:
read_unlock_bh(&sk->sk_callback_lock);
write_space(sk);
if (sk->sk_socket)
set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
}
