/******************************************************************************
|
*
|
* Copyright 2009-2012 Broadcom Corporation
|
*
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
* you may not use this file except in compliance with the License.
|
* You may obtain a copy of the License at:
|
*
|
* http://www.apache.org/licenses/LICENSE-2.0
|
*
|
* Unless required by applicable law or agreed to in writing, software
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
* See the License for the specific language governing permissions and
|
* limitations under the License.
|
*
|
******************************************************************************/
|
|
/*******************************************************************************
|
*
|
* Filename: btif_sock_thread.cc
|
*
|
* Description: socket select thread
|
*
|
******************************************************************************/
|
|
#define LOG_TAG "bt_btif_sock"
|
|
#include "btif_sock_thread.h"
|
|
#include <alloca.h>
|
#include <ctype.h>
|
#include <errno.h>
|
#include <fcntl.h>
|
#include <features.h>
|
#include <pthread.h>
|
#include <signal.h>
|
#include <stdio.h>
|
#include <stdlib.h>
|
#include <string.h>
|
#include <sys/poll.h>
|
#include <sys/select.h>
|
#include <sys/socket.h>
|
#include <sys/types.h>
|
#include <sys/un.h>
|
#include <time.h>
|
#include <unistd.h>
|
|
#include <mutex>
|
#include <string>
|
|
#include "bta_api.h"
|
#include "btif_common.h"
|
#include "btif_sock.h"
|
#include "btif_sock_util.h"
|
#include "btif_util.h"
|
#include "osi/include/socket_utils/sockets.h"
|
|
#define asrt(s) \
|
do { \
|
if (!(s)) \
|
APPL_TRACE_ERROR("## %s assert %s failed at line:%d ##", __func__, #s, \
|
__LINE__) \
|
} while (0)
|
|
#define MAX_THREAD 8
|
#define MAX_POLL 64
|
#define POLL_EXCEPTION_EVENTS (POLLHUP | POLLRDHUP | POLLERR | POLLNVAL)
|
#define IS_EXCEPTION(e) ((e)&POLL_EXCEPTION_EVENTS)
|
#define IS_READ(e) ((e)&POLLIN)
|
#define IS_WRITE(e) ((e)&POLLOUT)
|
/*cmd executes in socket poll thread */
|
#define CMD_WAKEUP 1
|
#define CMD_EXIT 2
|
#define CMD_ADD_FD 3
|
#define CMD_REMOVE_FD 4
|
#define CMD_USER_PRIVATE 5
|
|
typedef struct {
|
struct pollfd pfd;
|
uint32_t user_id;
|
int type;
|
int flags;
|
} poll_slot_t;
|
typedef struct {
|
int cmd_fdr, cmd_fdw;
|
int poll_count;
|
poll_slot_t ps[MAX_POLL];
|
int psi[MAX_POLL]; // index of poll slot
|
pthread_t thread_id;
|
btsock_signaled_cb callback;
|
btsock_cmd_cb cmd_callback;
|
int used;
|
} thread_slot_t;
|
static thread_slot_t ts[MAX_THREAD];
|
|
static void* sock_poll_thread(void* arg);
|
static inline void close_cmd_fd(int h);
|
|
static inline void add_poll(int h, int fd, int type, int flags,
|
uint32_t user_id);
|
|
static std::recursive_mutex thread_slot_lock;
|
|
static inline int create_thread(void* (*start_routine)(void*), void* arg,
|
pthread_t* thread_id) {
|
pthread_attr_t thread_attr;
|
pthread_attr_init(&thread_attr);
|
pthread_attr_setdetachstate(&thread_attr, PTHREAD_CREATE_JOINABLE);
|
int policy;
|
int min_pri = 0;
|
int ret = -1;
|
struct sched_param param;
|
|
ret = pthread_create(thread_id, &thread_attr, start_routine, arg);
|
if (ret != 0) {
|
APPL_TRACE_ERROR("pthread_create : %s", strerror(errno));
|
return ret;
|
}
|
/* We need to lower the priority of this thread to ensure the stack gets
|
* priority over transfer to a socket */
|
pthread_getschedparam(*thread_id, &policy, ¶m);
|
min_pri = sched_get_priority_min(policy);
|
if (param.sched_priority > min_pri) {
|
param.sched_priority -= 1;
|
}
|
pthread_setschedparam(*thread_id, policy, ¶m);
|
return ret;
|
}
|
static void init_poll(int cmd_fd);
|
static int alloc_thread_slot() {
|
std::unique_lock<std::recursive_mutex> lock(thread_slot_lock);
|
int i;
|
// reversed order to save guard uninitialized access to 0 index
|
for (i = MAX_THREAD - 1; i >= 0; i--) {
|
APPL_TRACE_DEBUG("ts[%d].used:%d", i, ts[i].used);
|
if (!ts[i].used) {
|
ts[i].used = 1;
|
return i;
|
}
|
}
|
APPL_TRACE_ERROR("execeeded max thread count");
|
return -1;
|
}
|
static void free_thread_slot(int h) {
|
if (0 <= h && h < MAX_THREAD) {
|
close_cmd_fd(h);
|
ts[h].used = 0;
|
} else
|
APPL_TRACE_ERROR("invalid thread handle:%d", h);
|
}
|
int btsock_thread_init() {
|
static int initialized;
|
APPL_TRACE_DEBUG("in initialized:%d", initialized);
|
if (!initialized) {
|
initialized = 1;
|
int h;
|
for (h = 0; h < MAX_THREAD; h++) {
|
ts[h].cmd_fdr = ts[h].cmd_fdw = -1;
|
ts[h].used = 0;
|
ts[h].thread_id = -1;
|
ts[h].poll_count = 0;
|
ts[h].callback = NULL;
|
ts[h].cmd_callback = NULL;
|
}
|
}
|
return true;
|
}
|
int btsock_thread_create(btsock_signaled_cb callback,
|
btsock_cmd_cb cmd_callback) {
|
asrt(callback || cmd_callback);
|
int h = alloc_thread_slot();
|
APPL_TRACE_DEBUG("alloc_thread_slot ret:%d", h);
|
if (h >= 0) {
|
init_poll(h);
|
pthread_t thread;
|
int status = create_thread(sock_poll_thread, (void*)(uintptr_t)h, &thread);
|
if (status) {
|
APPL_TRACE_ERROR("create_thread failed: %s", strerror(status));
|
free_thread_slot(h);
|
return -1;
|
}
|
|
ts[h].thread_id = thread;
|
APPL_TRACE_DEBUG("h:%d, thread id:%d", h, ts[h].thread_id);
|
ts[h].callback = callback;
|
ts[h].cmd_callback = cmd_callback;
|
}
|
return h;
|
}
|
|
/* create dummy socket pair used to wake up select loop */
|
static inline void init_cmd_fd(int h) {
|
asrt(ts[h].cmd_fdr == -1 && ts[h].cmd_fdw == -1);
|
if (socketpair(AF_UNIX, SOCK_STREAM, 0, &ts[h].cmd_fdr) < 0) {
|
APPL_TRACE_ERROR("socketpair failed: %s", strerror(errno));
|
return;
|
}
|
APPL_TRACE_DEBUG("h:%d, cmd_fdr:%d, cmd_fdw:%d", h, ts[h].cmd_fdr,
|
ts[h].cmd_fdw);
|
// add the cmd fd for read & write
|
add_poll(h, ts[h].cmd_fdr, 0, SOCK_THREAD_FD_RD, 0);
|
}
|
static inline void close_cmd_fd(int h) {
|
if (ts[h].cmd_fdr != -1) {
|
close(ts[h].cmd_fdr);
|
ts[h].cmd_fdr = -1;
|
}
|
if (ts[h].cmd_fdw != -1) {
|
close(ts[h].cmd_fdw);
|
ts[h].cmd_fdw = -1;
|
}
|
}
|
typedef struct {
|
int id;
|
int fd;
|
int type;
|
int flags;
|
uint32_t user_id;
|
} sock_cmd_t;
|
int btsock_thread_add_fd(int h, int fd, int type, int flags, uint32_t user_id) {
|
if (h < 0 || h >= MAX_THREAD) {
|
APPL_TRACE_ERROR("invalid bt thread handle:%d", h);
|
return false;
|
}
|
if (ts[h].cmd_fdw == -1) {
|
APPL_TRACE_ERROR(
|
"cmd socket is not created. socket thread may not initialized");
|
return false;
|
}
|
if (flags & SOCK_THREAD_ADD_FD_SYNC) {
|
// must executed in socket poll thread
|
if (ts[h].thread_id == pthread_self()) {
|
// cleanup one-time flags
|
flags &= ~SOCK_THREAD_ADD_FD_SYNC;
|
add_poll(h, fd, type, flags, user_id);
|
return true;
|
}
|
APPL_TRACE_DEBUG(
|
"THREAD_ADD_FD_SYNC is not called in poll thread, fallback to async");
|
}
|
sock_cmd_t cmd = {CMD_ADD_FD, fd, type, flags, user_id};
|
APPL_TRACE_DEBUG("adding fd:%d, flags:0x%x", fd, flags);
|
|
ssize_t ret;
|
OSI_NO_INTR(ret = send(ts[h].cmd_fdw, &cmd, sizeof(cmd), 0));
|
|
return ret == sizeof(cmd);
|
}
|
|
bool btsock_thread_remove_fd_and_close(int thread_handle, int fd) {
|
if (thread_handle < 0 || thread_handle >= MAX_THREAD) {
|
APPL_TRACE_ERROR("%s invalid thread handle: %d", __func__, thread_handle);
|
return false;
|
}
|
if (fd == -1) {
|
APPL_TRACE_ERROR("%s invalid file descriptor.", __func__);
|
return false;
|
}
|
|
sock_cmd_t cmd = {CMD_REMOVE_FD, fd, 0, 0, 0};
|
|
ssize_t ret;
|
OSI_NO_INTR(ret = send(ts[thread_handle].cmd_fdw, &cmd, sizeof(cmd), 0));
|
|
return ret == sizeof(cmd);
|
}
|
|
int btsock_thread_post_cmd(int h, int type, const unsigned char* data, int size,
|
uint32_t user_id) {
|
if (h < 0 || h >= MAX_THREAD) {
|
APPL_TRACE_ERROR("invalid bt thread handle:%d", h);
|
return false;
|
}
|
if (ts[h].cmd_fdw == -1) {
|
APPL_TRACE_ERROR(
|
"cmd socket is not created. socket thread may not initialized");
|
return false;
|
}
|
sock_cmd_t cmd = {CMD_USER_PRIVATE, 0, type, size, user_id};
|
APPL_TRACE_DEBUG("post cmd type:%d, size:%d, h:%d, ", type, size, h);
|
sock_cmd_t* cmd_send = &cmd;
|
int size_send = sizeof(cmd);
|
if (data && size) {
|
size_send = sizeof(cmd) + size;
|
cmd_send = (sock_cmd_t*)alloca(size_send);
|
if (cmd_send) {
|
*cmd_send = cmd;
|
memcpy(cmd_send + 1, data, size);
|
} else {
|
APPL_TRACE_ERROR("alloca failed at h:%d, cmd type:%d, size:%d", h, type,
|
size_send);
|
return false;
|
}
|
}
|
|
ssize_t ret;
|
OSI_NO_INTR(ret = send(ts[h].cmd_fdw, cmd_send, size_send, 0));
|
|
return ret == size_send;
|
}
|
int btsock_thread_wakeup(int h) {
|
if (h < 0 || h >= MAX_THREAD) {
|
APPL_TRACE_ERROR("invalid bt thread handle:%d", h);
|
return false;
|
}
|
if (ts[h].cmd_fdw == -1) {
|
APPL_TRACE_ERROR("thread handle:%d, cmd socket is not created", h);
|
return false;
|
}
|
sock_cmd_t cmd = {CMD_WAKEUP, 0, 0, 0, 0};
|
|
ssize_t ret;
|
OSI_NO_INTR(ret = send(ts[h].cmd_fdw, &cmd, sizeof(cmd), 0));
|
|
return ret == sizeof(cmd);
|
}
|
int btsock_thread_exit(int h) {
|
if (h < 0 || h >= MAX_THREAD) {
|
APPL_TRACE_ERROR("invalid bt thread slot:%d", h);
|
return false;
|
}
|
if (ts[h].cmd_fdw == -1) {
|
APPL_TRACE_ERROR("cmd socket is not created");
|
return false;
|
}
|
sock_cmd_t cmd = {CMD_EXIT, 0, 0, 0, 0};
|
|
ssize_t ret;
|
OSI_NO_INTR(ret = send(ts[h].cmd_fdw, &cmd, sizeof(cmd), 0));
|
|
if (ret == sizeof(cmd)) {
|
if (ts[h].thread_id != -1) {
|
pthread_join(ts[h].thread_id, 0);
|
ts[h].thread_id = -1;
|
}
|
free_thread_slot(h);
|
return true;
|
}
|
return false;
|
}
|
static void init_poll(int h) {
|
int i;
|
ts[h].poll_count = 0;
|
ts[h].thread_id = -1;
|
ts[h].callback = NULL;
|
ts[h].cmd_callback = NULL;
|
for (i = 0; i < MAX_POLL; i++) {
|
ts[h].ps[i].pfd.fd = -1;
|
ts[h].psi[i] = -1;
|
}
|
init_cmd_fd(h);
|
}
|
static inline unsigned int flags2pevents(int flags) {
|
unsigned int pevents = 0;
|
if (flags & SOCK_THREAD_FD_WR) pevents |= POLLOUT;
|
if (flags & SOCK_THREAD_FD_RD) pevents |= POLLIN;
|
pevents |= POLL_EXCEPTION_EVENTS;
|
return pevents;
|
}
|
|
static inline void set_poll(poll_slot_t* ps, int fd, int type, int flags,
|
uint32_t user_id) {
|
ps->pfd.fd = fd;
|
ps->user_id = user_id;
|
if (ps->type != 0 && ps->type != type)
|
APPL_TRACE_ERROR(
|
"poll socket type should not changed! type was:%d, type now:%d",
|
ps->type, type);
|
ps->type = type;
|
ps->flags = flags;
|
ps->pfd.events = flags2pevents(flags);
|
ps->pfd.revents = 0;
|
}
|
static inline void add_poll(int h, int fd, int type, int flags,
|
uint32_t user_id) {
|
asrt(fd != -1);
|
int i;
|
int empty = -1;
|
poll_slot_t* ps = ts[h].ps;
|
|
for (i = 0; i < MAX_POLL; i++) {
|
if (ps[i].pfd.fd == fd) {
|
asrt(ts[h].poll_count < MAX_POLL);
|
|
set_poll(&ps[i], fd, type, flags | ps[i].flags, user_id);
|
return;
|
} else if (empty < 0 && ps[i].pfd.fd == -1)
|
empty = i;
|
}
|
if (empty >= 0) {
|
asrt(ts[h].poll_count < MAX_POLL);
|
set_poll(&ps[empty], fd, type, flags, user_id);
|
++ts[h].poll_count;
|
return;
|
}
|
APPL_TRACE_ERROR("exceeded max poll slot:%d!", MAX_POLL);
|
}
|
static inline void remove_poll(int h, poll_slot_t* ps, int flags) {
|
if (flags == ps->flags) {
|
// all monitored events signaled. To remove it, just clear the slot
|
--ts[h].poll_count;
|
memset(ps, 0, sizeof(*ps));
|
ps->pfd.fd = -1;
|
} else {
|
// one read or one write monitor event signaled, removed the accordding bit
|
ps->flags &= ~flags;
|
// update the poll events mask
|
ps->pfd.events = flags2pevents(ps->flags);
|
}
|
}
|
static int process_cmd_sock(int h) {
|
sock_cmd_t cmd = {-1, 0, 0, 0, 0};
|
int fd = ts[h].cmd_fdr;
|
|
ssize_t ret;
|
OSI_NO_INTR(ret = recv(fd, &cmd, sizeof(cmd), MSG_WAITALL));
|
|
if (ret != sizeof(cmd)) {
|
APPL_TRACE_ERROR("recv cmd errno:%d", errno);
|
return false;
|
}
|
APPL_TRACE_DEBUG("cmd.id:%d", cmd.id);
|
switch (cmd.id) {
|
case CMD_ADD_FD:
|
add_poll(h, cmd.fd, cmd.type, cmd.flags, cmd.user_id);
|
break;
|
case CMD_REMOVE_FD:
|
for (int i = 1; i < MAX_POLL; ++i) {
|
poll_slot_t* poll_slot = &ts[h].ps[i];
|
if (poll_slot->pfd.fd == cmd.fd) {
|
remove_poll(h, poll_slot, poll_slot->flags);
|
break;
|
}
|
}
|
close(cmd.fd);
|
break;
|
case CMD_WAKEUP:
|
break;
|
case CMD_USER_PRIVATE:
|
asrt(ts[h].cmd_callback);
|
if (ts[h].cmd_callback)
|
ts[h].cmd_callback(fd, cmd.type, cmd.flags, cmd.user_id);
|
break;
|
case CMD_EXIT:
|
return false;
|
default:
|
APPL_TRACE_DEBUG("unknown cmd: %d", cmd.id);
|
break;
|
}
|
return true;
|
}
|
|
static void print_events(short events) {
|
std::string flags("");
|
if ((events)&POLLIN) flags += " POLLIN";
|
if ((events)&POLLPRI) flags += " POLLPRI";
|
if ((events)&POLLOUT) flags += " POLLOUT";
|
if ((events)&POLLERR) flags += " POLLERR";
|
if ((events)&POLLHUP) flags += " POLLHUP ";
|
if ((events)&POLLNVAL) flags += " POLLNVAL";
|
if ((events)&POLLRDHUP) flags += " POLLRDHUP";
|
APPL_TRACE_DEBUG("print poll event:%x = %s", (events), flags.c_str());
|
}
|
|
static void process_data_sock(int h, struct pollfd* pfds, int count) {
|
asrt(count <= ts[h].poll_count);
|
int i;
|
for (i = 1; i < ts[h].poll_count; i++) {
|
if (pfds[i].revents) {
|
int ps_i = ts[h].psi[i];
|
asrt(pfds[i].fd == ts[h].ps[ps_i].pfd.fd);
|
uint32_t user_id = ts[h].ps[ps_i].user_id;
|
int type = ts[h].ps[ps_i].type;
|
int flags = 0;
|
print_events(pfds[i].revents);
|
if (IS_READ(pfds[i].revents)) {
|
flags |= SOCK_THREAD_FD_RD;
|
}
|
if (IS_WRITE(pfds[i].revents)) {
|
flags |= SOCK_THREAD_FD_WR;
|
}
|
if (IS_EXCEPTION(pfds[i].revents)) {
|
flags |= SOCK_THREAD_FD_EXCEPTION;
|
// remove the whole slot not flags
|
remove_poll(h, &ts[h].ps[ps_i], ts[h].ps[ps_i].flags);
|
} else if (flags)
|
remove_poll(h, &ts[h].ps[ps_i],
|
flags); // remove the monitor flags that already processed
|
if (flags) ts[h].callback(pfds[i].fd, type, flags, user_id);
|
}
|
}
|
}
|
|
static void prepare_poll_fds(int h, struct pollfd* pfds) {
|
int count = 0;
|
int ps_i = 0;
|
int pfd_i = 0;
|
asrt(ts[h].poll_count <= MAX_POLL);
|
memset(pfds, 0, sizeof(pfds[0]) * ts[h].poll_count);
|
while (count < ts[h].poll_count) {
|
if (ps_i >= MAX_POLL) {
|
APPL_TRACE_ERROR(
|
"exceed max poll range, ps_i:%d, MAX_POLL:%d, count:%d, "
|
"ts[h].poll_count:%d",
|
ps_i, MAX_POLL, count, ts[h].poll_count);
|
return;
|
}
|
if (ts[h].ps[ps_i].pfd.fd >= 0) {
|
pfds[pfd_i] = ts[h].ps[ps_i].pfd;
|
ts[h].psi[pfd_i] = ps_i;
|
count++;
|
pfd_i++;
|
}
|
ps_i++;
|
}
|
}
|
static void* sock_poll_thread(void* arg) {
|
struct pollfd pfds[MAX_POLL];
|
memset(pfds, 0, sizeof(pfds));
|
int h = (intptr_t)arg;
|
for (;;) {
|
prepare_poll_fds(h, pfds);
|
int ret;
|
OSI_NO_INTR(ret = poll(pfds, ts[h].poll_count, -1));
|
if (ret == -1) {
|
APPL_TRACE_ERROR("poll ret -1, exit the thread, errno:%d, err:%s", errno,
|
strerror(errno));
|
break;
|
}
|
if (ret != 0) {
|
int need_process_data_fd = true;
|
if (pfds[0].revents) // cmd fd always is the first one
|
{
|
asrt(pfds[0].fd == ts[h].cmd_fdr);
|
if (!process_cmd_sock(h)) {
|
APPL_TRACE_DEBUG("h:%d, process_cmd_sock return false, exit...", h);
|
break;
|
}
|
if (ret == 1)
|
need_process_data_fd = false;
|
else
|
ret--; // exclude the cmd fd
|
}
|
if (need_process_data_fd) process_data_sock(h, pfds, ret);
|
} else {
|
APPL_TRACE_DEBUG("no data, select ret: %d", ret)
|
};
|
}
|
APPL_TRACE_DEBUG("socket poll thread exiting, h:%d", h);
|
return 0;
|
}
|