Home | History | Annotate | Download | only in adb
      1 /* http://frotznet.googlecode.com/svn/trunk/utils/fdevent.c
      2 **
      3 ** Copyright 2006, Brian Swetland <swetland (at) frotz.net>
      4 **
      5 ** Licensed under the Apache License, Version 2.0 (the "License");
      6 ** you may not use this file except in compliance with the License.
      7 ** You may obtain a copy of the License at
      8 **
      9 **     http://www.apache.org/licenses/LICENSE-2.0
     10 **
     11 ** Unless required by applicable law or agreed to in writing, software
     12 ** distributed under the License is distributed on an "AS IS" BASIS,
     13 ** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
     14 ** See the License for the specific language governing permissions and
     15 ** limitations under the License.
     16 */
     17 
     18 #define TRACE_TAG FDEVENT
     19 
     20 #include "sysdeps.h"
     21 #include "fdevent.h"
     22 
     23 #include <fcntl.h>
     24 #include <stdlib.h>
     25 #include <string.h>
     26 #include <unistd.h>
     27 
     28 #include <atomic>
     29 #include <functional>
     30 #include <list>
     31 #include <mutex>
     32 #include <unordered_map>
     33 #include <vector>
     34 
     35 #include <android-base/logging.h>
     36 #include <android-base/stringprintf.h>
     37 #include <android-base/thread_annotations.h>
     38 
     39 #include "adb_io.h"
     40 #include "adb_trace.h"
     41 #include "adb_unique_fd.h"
     42 #include "adb_utils.h"
     43 
     44 #if !ADB_HOST
     45 // This socket is used when a subproc shell service exists.
     46 // It wakes up the fdevent_loop() and cause the correct handling
     47 // of the shell's pseudo-tty master. I.e. force close it.
     48 int SHELL_EXIT_NOTIFY_FD = -1;
     49 #endif // !ADB_HOST
     50 
     51 #define FDE_EVENTMASK  0x00ff
     52 #define FDE_STATEMASK  0xff00
     53 
     54 #define FDE_ACTIVE     0x0100
     55 #define FDE_PENDING    0x0200
     56 #define FDE_CREATED    0x0400
     57 
     58 struct PollNode {
     59   fdevent* fde;
     60   adb_pollfd pollfd;
     61 
     62   explicit PollNode(fdevent* fde) : fde(fde) {
     63       memset(&pollfd, 0, sizeof(pollfd));
     64       pollfd.fd = fde->fd;
     65 
     66 #if defined(__linux__)
     67       // Always enable POLLRDHUP, so the host server can take action when some clients disconnect.
     68       // Then we can avoid leaving many sockets in CLOSE_WAIT state. See http://b/23314034.
     69       pollfd.events = POLLRDHUP;
     70 #endif
     71   }
     72 };
     73 
     74 // All operations to fdevent should happen only in the main thread.
     75 // That's why we don't need a lock for fdevent.
     76 static auto& g_poll_node_map = *new std::unordered_map<int, PollNode>();
     77 static auto& g_pending_list = *new std::list<fdevent*>();
     78 static std::atomic<bool> terminate_loop(false);
     79 static bool main_thread_valid;
     80 static unsigned long main_thread_id;
     81 
     82 static auto& run_queue_notify_fd = *new unique_fd();
     83 static auto& run_queue_mutex = *new std::mutex();
     84 static auto& run_queue GUARDED_BY(run_queue_mutex) = *new std::vector<std::function<void()>>();
     85 
     86 void check_main_thread() {
     87     if (main_thread_valid) {
     88         CHECK_EQ(main_thread_id, adb_thread_id());
     89     }
     90 }
     91 
     92 void set_main_thread() {
     93     main_thread_valid = true;
     94     main_thread_id = adb_thread_id();
     95 }
     96 
     97 static std::string dump_fde(const fdevent* fde) {
     98     std::string state;
     99     if (fde->state & FDE_ACTIVE) {
    100         state += "A";
    101     }
    102     if (fde->state & FDE_PENDING) {
    103         state += "P";
    104     }
    105     if (fde->state & FDE_CREATED) {
    106         state += "C";
    107     }
    108     if (fde->state & FDE_READ) {
    109         state += "R";
    110     }
    111     if (fde->state & FDE_WRITE) {
    112         state += "W";
    113     }
    114     if (fde->state & FDE_ERROR) {
    115         state += "E";
    116     }
    117     if (fde->state & FDE_DONT_CLOSE) {
    118         state += "D";
    119     }
    120     return android::base::StringPrintf("(fdevent %d %s)", fde->fd, state.c_str());
    121 }
    122 
    123 fdevent* fdevent_create(int fd, fd_func func, void* arg) {
    124     check_main_thread();
    125     fdevent *fde = (fdevent*) malloc(sizeof(fdevent));
    126     if(fde == 0) return 0;
    127     fdevent_install(fde, fd, func, arg);
    128     fde->state |= FDE_CREATED;
    129     return fde;
    130 }
    131 
    132 void fdevent_destroy(fdevent* fde) {
    133     check_main_thread();
    134     if(fde == 0) return;
    135     if(!(fde->state & FDE_CREATED)) {
    136         LOG(FATAL) << "destroying fde not created by fdevent_create(): " << dump_fde(fde);
    137     }
    138     fdevent_remove(fde);
    139     free(fde);
    140 }
    141 
    142 void fdevent_install(fdevent* fde, int fd, fd_func func, void* arg) {
    143     check_main_thread();
    144     CHECK_GE(fd, 0);
    145     memset(fde, 0, sizeof(fdevent));
    146     fde->state = FDE_ACTIVE;
    147     fde->fd = fd;
    148     fde->func = func;
    149     fde->arg = arg;
    150     if (!set_file_block_mode(fd, false)) {
    151         // Here is not proper to handle the error. If it fails here, some error is
    152         // likely to be detected by poll(), then we can let the callback function
    153         // to handle it.
    154         LOG(ERROR) << "failed to set non-blocking mode for fd " << fd;
    155     }
    156     auto pair = g_poll_node_map.emplace(fde->fd, PollNode(fde));
    157     CHECK(pair.second) << "install existing fd " << fd;
    158     D("fdevent_install %s", dump_fde(fde).c_str());
    159 }
    160 
    161 void fdevent_remove(fdevent* fde) {
    162     check_main_thread();
    163     D("fdevent_remove %s", dump_fde(fde).c_str());
    164     if (fde->state & FDE_ACTIVE) {
    165         g_poll_node_map.erase(fde->fd);
    166         if (fde->state & FDE_PENDING) {
    167             g_pending_list.remove(fde);
    168         }
    169         if (!(fde->state & FDE_DONT_CLOSE)) {
    170             adb_close(fde->fd);
    171             fde->fd = -1;
    172         }
    173         fde->state = 0;
    174         fde->events = 0;
    175     }
    176 }
    177 
    178 static void fdevent_update(fdevent* fde, unsigned events) {
    179     auto it = g_poll_node_map.find(fde->fd);
    180     CHECK(it != g_poll_node_map.end());
    181     PollNode& node = it->second;
    182     if (events & FDE_READ) {
    183         node.pollfd.events |= POLLIN;
    184     } else {
    185         node.pollfd.events &= ~POLLIN;
    186     }
    187 
    188     if (events & FDE_WRITE) {
    189         node.pollfd.events |= POLLOUT;
    190     } else {
    191         node.pollfd.events &= ~POLLOUT;
    192     }
    193     fde->state = (fde->state & FDE_STATEMASK) | events;
    194 }
    195 
    196 void fdevent_set(fdevent* fde, unsigned events) {
    197     check_main_thread();
    198     events &= FDE_EVENTMASK;
    199     if ((fde->state & FDE_EVENTMASK) == events) {
    200         return;
    201     }
    202     CHECK(fde->state & FDE_ACTIVE);
    203     fdevent_update(fde, events);
    204     D("fdevent_set: %s, events = %u", dump_fde(fde).c_str(), events);
    205 
    206     if (fde->state & FDE_PENDING) {
    207         // If we are pending, make sure we don't signal an event that is no longer wanted.
    208         fde->events &= events;
    209         if (fde->events == 0) {
    210             g_pending_list.remove(fde);
    211             fde->state &= ~FDE_PENDING;
    212         }
    213     }
    214 }
    215 
    216 void fdevent_add(fdevent* fde, unsigned events) {
    217     check_main_thread();
    218     fdevent_set(fde, (fde->state & FDE_EVENTMASK) | events);
    219 }
    220 
    221 void fdevent_del(fdevent* fde, unsigned events) {
    222     check_main_thread();
    223     fdevent_set(fde, (fde->state & FDE_EVENTMASK) & ~events);
    224 }
    225 
    226 static std::string dump_pollfds(const std::vector<adb_pollfd>& pollfds) {
    227     std::string result;
    228     for (const auto& pollfd : pollfds) {
    229         std::string op;
    230         if (pollfd.events & POLLIN) {
    231             op += "R";
    232         }
    233         if (pollfd.events & POLLOUT) {
    234             op += "W";
    235         }
    236         android::base::StringAppendF(&result, " %d(%s)", pollfd.fd, op.c_str());
    237     }
    238     return result;
    239 }
    240 
    241 static void fdevent_process() {
    242     std::vector<adb_pollfd> pollfds;
    243     for (const auto& pair : g_poll_node_map) {
    244         pollfds.push_back(pair.second.pollfd);
    245     }
    246     CHECK_GT(pollfds.size(), 0u);
    247     D("poll(), pollfds = %s", dump_pollfds(pollfds).c_str());
    248     int ret = adb_poll(&pollfds[0], pollfds.size(), -1);
    249     if (ret == -1) {
    250         PLOG(ERROR) << "poll(), ret = " << ret;
    251         return;
    252     }
    253     for (const auto& pollfd : pollfds) {
    254         if (pollfd.revents != 0) {
    255             D("for fd %d, revents = %x", pollfd.fd, pollfd.revents);
    256         }
    257         unsigned events = 0;
    258         if (pollfd.revents & POLLIN) {
    259             events |= FDE_READ;
    260         }
    261         if (pollfd.revents & POLLOUT) {
    262             events |= FDE_WRITE;
    263         }
    264         if (pollfd.revents & (POLLERR | POLLHUP | POLLNVAL)) {
    265             // We fake a read, as the rest of the code assumes that errors will
    266             // be detected at that point.
    267             events |= FDE_READ | FDE_ERROR;
    268         }
    269 #if defined(__linux__)
    270         if (pollfd.revents & POLLRDHUP) {
    271             events |= FDE_READ | FDE_ERROR;
    272         }
    273 #endif
    274         if (events != 0) {
    275             auto it = g_poll_node_map.find(pollfd.fd);
    276             CHECK(it != g_poll_node_map.end());
    277             fdevent* fde = it->second.fde;
    278             CHECK_EQ(fde->fd, pollfd.fd);
    279             fde->events |= events;
    280             D("%s got events %x", dump_fde(fde).c_str(), events);
    281             fde->state |= FDE_PENDING;
    282             g_pending_list.push_back(fde);
    283         }
    284     }
    285 }
    286 
    287 static void fdevent_call_fdfunc(fdevent* fde) {
    288     unsigned events = fde->events;
    289     fde->events = 0;
    290     CHECK(fde->state & FDE_PENDING);
    291     fde->state &= (~FDE_PENDING);
    292     D("fdevent_call_fdfunc %s", dump_fde(fde).c_str());
    293     fde->func(fde->fd, events, fde->arg);
    294 }
    295 
    296 #if !ADB_HOST
    297 
    298 #include <sys/ioctl.h>
    299 
    300 static void fdevent_subproc_event_func(int fd, unsigned ev, void* /* userdata */) {
    301     D("subproc handling on fd = %d, ev = %x", fd, ev);
    302 
    303     CHECK_GE(fd, 0);
    304 
    305     if (ev & FDE_READ) {
    306         int subproc_fd;
    307 
    308         if(!ReadFdExactly(fd, &subproc_fd, sizeof(subproc_fd))) {
    309             LOG(FATAL) << "Failed to read the subproc's fd from " << fd;
    310         }
    311         auto it = g_poll_node_map.find(subproc_fd);
    312         if (it == g_poll_node_map.end()) {
    313             D("subproc_fd %d cleared from fd_table", subproc_fd);
    314             adb_close(subproc_fd);
    315             return;
    316         }
    317         fdevent* subproc_fde = it->second.fde;
    318         if(subproc_fde->fd != subproc_fd) {
    319             // Already reallocated?
    320             LOG(FATAL) << "subproc_fd(" << subproc_fd << ") != subproc_fde->fd(" << subproc_fde->fd
    321                        << ")";
    322             return;
    323         }
    324 
    325         subproc_fde->force_eof = 1;
    326 
    327         int rcount = 0;
    328         ioctl(subproc_fd, FIONREAD, &rcount);
    329         D("subproc with fd %d has rcount=%d, err=%d", subproc_fd, rcount, errno);
    330         if (rcount != 0) {
    331             // If there is data left, it will show up in the select().
    332             // This works because there is no other thread reading that
    333             // data when in this fd_func().
    334             return;
    335         }
    336 
    337         D("subproc_fde %s", dump_fde(subproc_fde).c_str());
    338         subproc_fde->events |= FDE_READ;
    339         if(subproc_fde->state & FDE_PENDING) {
    340             return;
    341         }
    342         subproc_fde->state |= FDE_PENDING;
    343         fdevent_call_fdfunc(subproc_fde);
    344     }
    345 }
    346 
    347 static void fdevent_subproc_setup() {
    348     int s[2];
    349 
    350     if(adb_socketpair(s)) {
    351         PLOG(FATAL) << "cannot create shell-exit socket-pair";
    352     }
    353     D("fdevent_subproc: socket pair (%d, %d)", s[0], s[1]);
    354 
    355     SHELL_EXIT_NOTIFY_FD = s[0];
    356     fdevent *fde = fdevent_create(s[1], fdevent_subproc_event_func, NULL);
    357     CHECK(fde != nullptr) << "cannot create fdevent for shell-exit handler";
    358     fdevent_add(fde, FDE_READ);
    359 }
    360 #endif // !ADB_HOST
    361 
    362 static void fdevent_run_flush() REQUIRES(run_queue_mutex) {
    363     for (auto& f : run_queue) {
    364         f();
    365     }
    366     run_queue.clear();
    367 }
    368 
    369 static void fdevent_run_func(int fd, unsigned ev, void* /* userdata */) {
    370     CHECK_GE(fd, 0);
    371     CHECK(ev & FDE_READ);
    372 
    373     char buf[1024];
    374 
    375     // Empty the fd.
    376     if (adb_read(fd, buf, sizeof(buf)) == -1) {
    377         PLOG(FATAL) << "failed to empty run queue notify fd";
    378     }
    379 
    380     std::lock_guard<std::mutex> lock(run_queue_mutex);
    381     fdevent_run_flush();
    382 }
    383 
    384 static void fdevent_run_setup() {
    385     std::lock_guard<std::mutex> lock(run_queue_mutex);
    386     CHECK(run_queue_notify_fd.get() == -1);
    387     int s[2];
    388     if (adb_socketpair(s) != 0) {
    389         PLOG(FATAL) << "failed to create run queue notify socketpair";
    390     }
    391 
    392     run_queue_notify_fd.reset(s[0]);
    393     fdevent* fde = fdevent_create(s[1], fdevent_run_func, nullptr);
    394     CHECK(fde != nullptr);
    395     fdevent_add(fde, FDE_READ);
    396 
    397     fdevent_run_flush();
    398 }
    399 
    400 void fdevent_run_on_main_thread(std::function<void()> fn) {
    401     std::lock_guard<std::mutex> lock(run_queue_mutex);
    402     run_queue.push_back(std::move(fn));
    403 
    404     // run_queue_notify_fd could still be -1 if we're called before fdevent has finished setting up.
    405     // In that case, rely on the setup code to flush the queue without a notification being needed.
    406     if (run_queue_notify_fd != -1) {
    407         if (adb_write(run_queue_notify_fd.get(), "", 1) != 1) {
    408             PLOG(FATAL) << "failed to write to run queue notify fd";
    409         }
    410     }
    411 }
    412 
    413 void fdevent_loop() {
    414     set_main_thread();
    415 #if !ADB_HOST
    416     fdevent_subproc_setup();
    417 #endif // !ADB_HOST
    418     fdevent_run_setup();
    419 
    420     while (true) {
    421         if (terminate_loop) {
    422             return;
    423         }
    424 
    425         D("--- --- waiting for events");
    426 
    427         fdevent_process();
    428 
    429         while (!g_pending_list.empty()) {
    430             fdevent* fde = g_pending_list.front();
    431             g_pending_list.pop_front();
    432             fdevent_call_fdfunc(fde);
    433         }
    434     }
    435 }
    436 
    437 void fdevent_terminate_loop() {
    438     terminate_loop = true;
    439 }
    440 
    441 size_t fdevent_installed_count() {
    442     return g_poll_node_map.size();
    443 }
    444 
    445 void fdevent_reset() {
    446     g_poll_node_map.clear();
    447     g_pending_list.clear();
    448 
    449     std::lock_guard<std::mutex> lock(run_queue_mutex);
    450     run_queue_notify_fd.reset();
    451     run_queue.clear();
    452 
    453     main_thread_valid = false;
    454     terminate_loop = false;
    455 }
    456