1 /*
   2  * Copyright (c) 2001, 2008, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.  Oracle designates this
   8  * particular file as subject to the "Classpath" exception as provided
   9  * by Oracle in the LICENSE file that accompanied this code.
  10  *
  11  * This code is distributed in the hope that it will be useful, but WITHOUT
  12  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  13  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  14  * version 2 for more details (a copy is included in the LICENSE file that
  15  * accompanied this code).
  16  *
  17  * You should have received a copy of the GNU General Public License version
  18  * 2 along with this work; if not, write to the Free Software Foundation,
  19  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  20  *
  21  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  22  * or visit www.oracle.com if you need additional information or have any
  23  * questions.
  24  */
  25 
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <signal.h>
  29 #include <pthread.h>
  30 #include <sys/types.h>
  31 #include <sys/socket.h>
  32 #include <sys/time.h>
  33 #include <sys/resource.h>
  34 #include <sys/uio.h>
  35 #include <unistd.h>
  36 #include <errno.h>
  37 
  38 #include <sys/poll.h>
  39 
  40 /*
  41  * Stack allocated by thread when doing blocking operation
  42  */
  43 typedef struct threadEntry {
  44     pthread_t thr;                      /* this thread */
  45     struct threadEntry *next;           /* next thread */
  46     int intr;                           /* interrupted */
  47 } threadEntry_t;
  48 
  49 /*
  50  * Heap allocated during initialized - one entry per fd
  51  */
  52 typedef struct {
  53     pthread_mutex_t lock;               /* fd lock */
  54     threadEntry_t *threads;             /* threads blocked on fd */
  55 } fdEntry_t;
  56 
  57 /*
  58  * Signal to unblock thread
  59  */
  60 static int sigWakeup = (__SIGRTMAX - 2);
  61 
  62 /*
  63  * The fd table and the number of file descriptors
  64  */
  65 static fdEntry_t *fdTable;
  66 static int fdCount;
  67 
  68 /*
  69  * Null signal handler
  70  */
  71 static void sig_wakeup(int sig) {
  72 }
  73 
  74 /*
  75  * Initialization routine (executed when library is loaded)
  76  * Allocate fd tables and sets up signal handler.
  77  */
  78 static void __attribute((constructor)) init() {
  79     struct rlimit nbr_files;
  80     sigset_t sigset;
  81     struct sigaction sa;
  82 
  83     /*
  84      * Allocate table based on the maximum number of
  85      * file descriptors.
  86      */
  87     getrlimit(RLIMIT_NOFILE, &nbr_files);
  88     fdCount = nbr_files.rlim_max;
  89     fdTable = (fdEntry_t *)calloc(fdCount, sizeof(fdEntry_t));
  90     if (fdTable == NULL) {
  91         fprintf(stderr, "library initialization failed - "
  92                 "unable to allocate file descriptor table - out of memory");
  93         abort();
  94     }
  95 
  96     /*
  97      * Setup the signal handler
  98      */
  99     sa.sa_handler = sig_wakeup;
 100     sa.sa_flags   = 0;
 101     sigemptyset(&sa.sa_mask);
 102     sigaction(sigWakeup, &sa, NULL);
 103 
 104     sigemptyset(&sigset);
 105     sigaddset(&sigset, sigWakeup);
 106     sigprocmask(SIG_UNBLOCK, &sigset, NULL);
 107 }
 108 
 109 /*
 110  * Return the fd table for this fd or NULL is fd out
 111  * of range.
 112  */
 113 static inline fdEntry_t *getFdEntry(int fd)
 114 {
 115     if (fd < 0 || fd >= fdCount) {
 116         return NULL;
 117     }
 118     return &fdTable[fd];
 119 }
 120 
 121 /*
 122  * Start a blocking operation :-
 123  *    Insert thread onto thread list for the fd.
 124  */
 125 static inline void startOp(fdEntry_t *fdEntry, threadEntry_t *self)
 126 {
 127     self->thr = pthread_self();
 128     self->intr = 0;
 129 
 130     pthread_mutex_lock(&(fdEntry->lock));
 131     {
 132         self->next = fdEntry->threads;
 133         fdEntry->threads = self;
 134     }
 135     pthread_mutex_unlock(&(fdEntry->lock));
 136 }
 137 
 138 /*
 139  * End a blocking operation :-
 140  *     Remove thread from thread list for the fd
 141  *     If fd has been interrupted then set errno to EBADF
 142  */
 143 static inline void endOp
 144     (fdEntry_t *fdEntry, threadEntry_t *self)
 145 {
 146     int orig_errno = errno;
 147     pthread_mutex_lock(&(fdEntry->lock));
 148     {
 149         threadEntry_t *curr, *prev=NULL;
 150         curr = fdEntry->threads;
 151         while (curr != NULL) {
 152             if (curr == self) {
 153                 if (curr->intr) {
 154                     orig_errno = EBADF;
 155                 }
 156                 if (prev == NULL) {
 157                     fdEntry->threads = curr->next;
 158                 } else {
 159                     prev->next = curr->next;
 160                 }
 161                 break;
 162             }
 163             prev = curr;
 164             curr = curr->next;
 165         }
 166     }
 167     pthread_mutex_unlock(&(fdEntry->lock));
 168     errno = orig_errno;
 169 }
 170 
 171 /*
 172  * Close or dup2 a file descriptor ensuring that all threads blocked on
 173  * the file descriptor are notified via a wakeup signal.
 174  *
 175  *      fd1 < 0    => close(fd2)
 176  *      fd1 >= 0   => dup2(fd1, fd2)
 177  *
 178  * Returns -1 with errno set if operation fails.
 179  */
 180 static int closefd(int fd1, int fd2) {
 181     int rv, orig_errno;
 182     fdEntry_t *fdEntry = getFdEntry(fd2);
 183     if (fdEntry == NULL) {
 184         errno = EBADF;
 185         return -1;
 186     }
 187 
 188     /*
 189      * Lock the fd to hold-off additional I/O on this fd.
 190      */
 191     pthread_mutex_lock(&(fdEntry->lock));
 192 
 193     {
 194         /*
 195          * Send a wakeup signal to all threads blocked on this
 196          * file descriptor.
 197          */
 198         threadEntry_t *curr = fdEntry->threads;
 199         while (curr != NULL) {
 200             curr->intr = 1;
 201             pthread_kill( curr->thr, sigWakeup );
 202             curr = curr->next;
 203         }
 204 
 205         /*
 206          * And close/dup the file descriptor
 207          * (restart if interrupted by signal)
 208          */
 209         do {
 210             if (fd1 < 0) {
 211                 rv = close(fd2);
 212             } else {
 213                 rv = dup2(fd1, fd2);
 214             }
 215         } while (rv == -1 && errno == EINTR);
 216 
 217     }
 218 
 219     /*
 220      * Unlock without destroying errno
 221      */
 222     orig_errno = errno;
 223     pthread_mutex_unlock(&(fdEntry->lock));
 224     errno = orig_errno;
 225 
 226     return rv;
 227 }
 228 
 229 /*
 230  * Wrapper for dup2 - same semantics as dup2 system call except
 231  * that any threads blocked in an I/O system call on fd2 will be
 232  * preempted and return -1/EBADF;
 233  */
 234 int NET_Dup2(int fd, int fd2) {
 235     if (fd < 0) {
 236         errno = EBADF;
 237         return -1;
 238     }
 239     return closefd(fd, fd2);
 240 }
 241 
 242 /*
 243  * Wrapper for close - same semantics as close system call
 244  * except that any threads blocked in an I/O on fd will be
 245  * preempted and the I/O system call will return -1/EBADF.
 246  */
 247 int NET_SocketClose(int fd) {
 248     return closefd(-1, fd);
 249 }
 250 
 251 /************** Basic I/O operations here ***************/
 252 
 253 /*
 254  * Macro to perform a blocking IO operation. Restarts
 255  * automatically if interrupted by signal (other than
 256  * our wakeup signal)
 257  */
 258 #define BLOCKING_IO_RETURN_INT(FD, FUNC) {      \
 259     int ret;                                    \
 260     threadEntry_t self;                         \
 261     fdEntry_t *fdEntry = getFdEntry(FD);        \
 262     if (fdEntry == NULL) {                      \
 263         errno = EBADF;                          \
 264         return -1;                              \
 265     }                                           \
 266     do {                                        \
 267         startOp(fdEntry, &self);                \
 268         ret = FUNC;                             \
 269         endOp(fdEntry, &self);                  \
 270     } while (ret == -1 && errno == EINTR);      \
 271     return ret;                                 \
 272 }
 273 
 274 int NET_Read(int s, void* buf, size_t len) {
 275     BLOCKING_IO_RETURN_INT( s, recv(s, buf, len, 0) );
 276 }
 277 
 278 int NET_ReadV(int s, const struct iovec * vector, int count) {
 279     BLOCKING_IO_RETURN_INT( s, readv(s, vector, count) );
 280 }
 281 
 282 int NET_RecvFrom(int s, void *buf, int len, unsigned int flags,
 283        struct sockaddr *from, int *fromlen) {
 284     socklen_t socklen = *fromlen;
 285     BLOCKING_IO_RETURN_INT( s, recvfrom(s, buf, len, flags, from, &socklen) );
 286     *fromlen = socklen;
 287 }
 288 
 289 int NET_Send(int s, void *msg, int len, unsigned int flags) {
 290     BLOCKING_IO_RETURN_INT( s, send(s, msg, len, flags) );
 291 }
 292 
 293 int NET_WriteV(int s, const struct iovec * vector, int count) {
 294     BLOCKING_IO_RETURN_INT( s, writev(s, vector, count) );
 295 }
 296 
 297 int NET_SendTo(int s, const void *msg, int len,  unsigned  int
 298        flags, const struct sockaddr *to, int tolen) {
 299     BLOCKING_IO_RETURN_INT( s, sendto(s, msg, len, flags, to, tolen) );
 300 }
 301 
 302 int NET_Accept(int s, struct sockaddr *addr, int *addrlen) {
 303     socklen_t socklen = *addrlen;
 304     BLOCKING_IO_RETURN_INT( s, accept(s, addr, &socklen) );
 305     *addrlen = socklen;
 306 }
 307 
 308 int NET_Connect(int s, struct sockaddr *addr, int addrlen) {
 309     BLOCKING_IO_RETURN_INT( s, connect(s, addr, addrlen) );
 310 }
 311 
 312 #ifndef USE_SELECT
 313 int NET_Poll(struct pollfd *ufds, unsigned int nfds, int timeout) {
 314     BLOCKING_IO_RETURN_INT( ufds[0].fd, poll(ufds, nfds, timeout) );
 315 }
 316 #else
 317 int NET_Select(int s, fd_set *readfds, fd_set *writefds,
 318                fd_set *exceptfds, struct timeval *timeout) {
 319     BLOCKING_IO_RETURN_INT( s-1,
 320                             select(s, readfds, writefds, exceptfds, timeout) );
 321 }
 322 #endif
 323 
 324 /*
 325  * Wrapper for poll(s, timeout).
 326  * Auto restarts with adjusted timeout if interrupted by
 327  * signal other than our wakeup signal.
 328  */
 329 int NET_Timeout(int s, long timeout) {
 330     long prevtime = 0, newtime;
 331     struct timeval t;
 332     fdEntry_t *fdEntry = getFdEntry(s);
 333 
 334     /*
 335      * Check that fd hasn't been closed.
 336      */
 337     if (fdEntry == NULL) {
 338         errno = EBADF;
 339         return -1;
 340     }
 341 
 342     /*
 343      * Pick up current time as may need to adjust timeout
 344      */
 345     if (timeout > 0) {
 346         gettimeofday(&t, NULL);
 347         prevtime = t.tv_sec * 1000  +  t.tv_usec / 1000;
 348     }
 349 
 350     for(;;) {
 351         struct pollfd pfd;
 352         int rv;
 353         threadEntry_t self;
 354 
 355         /*
 356          * Poll the fd. If interrupted by our wakeup signal
 357          * errno will be set to EBADF.
 358          */
 359         pfd.fd = s;
 360         pfd.events = POLLIN | POLLERR;
 361 
 362         startOp(fdEntry, &self);
 363         rv = poll(&pfd, 1, timeout);
 364         endOp(fdEntry, &self);
 365 
 366         /*
 367          * If interrupted then adjust timeout. If timeout
 368          * has expired return 0 (indicating timeout expired).
 369          */
 370         if (rv < 0 && errno == EINTR) {
 371             if (timeout > 0) {
 372                 gettimeofday(&t, NULL);
 373                 newtime = t.tv_sec * 1000  +  t.tv_usec / 1000;
 374                 timeout -= newtime - prevtime;
 375                 if (timeout <= 0) {
 376                     return 0;
 377                 }
 378                 prevtime = newtime;
 379             }
 380         } else {
 381             return rv;
 382         }
 383 
 384     }
 385 }