sys/kern/sys_pipe.c

   1 /*
   2  * Copyright (c) 1996 John S. Dyson
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice immediately at the beginning of the file, without modification,
  10  *    this list of conditions, and the following disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conditions and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  * 3. Absolutely no warranty of function or purpose is made by the author
  15  *    John S. Dyson.
  16  * 4. Modifications may be freely made to this file if the above conditions
  17  *    are met.
  18  *
  19  * $FreeBSD: src/sys/kern/sys_pipe.c,v 1.60.2.13 2002/08/05 15:05:15 des Exp $
  20  * $DragonFly: src/sys/kern/sys_pipe.c,v 1.50 2008/09/09 04:06:13 dillon Exp $
  21  */
  22
  23 /*
  24  * This file contains a high-performance replacement for the socket-based
  25  * pipes scheme originally used in FreeBSD/4.4Lite.  It does not support
  26  * all features of sockets, but does do everything that pipes normally
  27  * do.
  28  */
  29 #include <sys/param.h>
  30 #include <sys/systm.h>
  31 #include <sys/kernel.h>
  32 #include <sys/proc.h>
  33 #include <sys/fcntl.h>
  34 #include <sys/file.h>
  35 #include <sys/filedesc.h>
  36 #include <sys/filio.h>
  37 #include <sys/ttycom.h>
  38 #include <sys/stat.h>
  39 #include <sys/select.h>
  40 #include <sys/signalvar.h>
  41 #include <sys/sysproto.h>
  42 #include <sys/pipe.h>
  43 #include <sys/vnode.h>
  44 #include <sys/uio.h>
  45 #include <sys/event.h>
  46 #include <sys/globaldata.h>
  47 #include <sys/module.h>
  48 #include <sys/malloc.h>
  49 #include <sys/sysctl.h>
  50 #include <sys/socket.h>
  51
  52 #include <vm/vm.h>
  53 #include <vm/vm_param.h>
  54 #include <sys/lock.h>
  55 #include <vm/vm_object.h>
  56 #include <vm/vm_kern.h>
  57 #include <vm/vm_extern.h>
  58 #include <vm/pmap.h>
  59 #include <vm/vm_map.h>
  60 #include <vm/vm_page.h>
  61 #include <vm/vm_zone.h>
  62
  63 #include <sys/file2.h>
  64 #include <sys/signal2.h>
  65 #include <sys/mplock2.h>
  66
  67 #include <machine/cpufunc.h>
  68
  69 /*
  70  * interfaces to the outside world
  71  */
  72 static int pipe_read (struct file *fp, struct uio *uio,
  73                 struct ucred *cred, int flags);
  74 static int pipe_write (struct file *fp, struct uio *uio,
  75                 struct ucred *cred, int flags);
  76 static int pipe_close (struct file *fp);
  77 static int pipe_shutdown (struct file *fp, int how);
  78 static int pipe_kqfilter (struct file *fp, struct knote *kn);
  79 static int pipe_stat (struct file *fp, struct stat *sb, struct ucred *cred);
  80 static int pipe_ioctl (struct file *fp, u_long cmd, caddr_t data,
  81                 struct ucred *cred, struct sysmsg *msg);
  82
  83 static struct fileops pipeops = {
  84         .fo_read = pipe_read,
  85         .fo_write = pipe_write,
  86         .fo_ioctl = pipe_ioctl,
  87         .fo_kqfilter = pipe_kqfilter,
  88         .fo_stat = pipe_stat,
  89         .fo_close = pipe_close,
  90         .fo_shutdown = pipe_shutdown
  91 };
  92
  93 static void     filt_pipedetach(struct knote *kn);
  94 static int      filt_piperead(struct knote *kn, long hint);
  95 static int      filt_pipewrite(struct knote *kn, long hint);
  96
  97 static struct filterops pipe_rfiltops =
  98         { 1, NULL, filt_pipedetach, filt_piperead };
  99 static struct filterops pipe_wfiltops =
 100         { 1, NULL, filt_pipedetach, filt_pipewrite };
 101
 102 MALLOC_DEFINE(M_PIPE, "pipe", "pipe structures");
 103
 104 /*
 105  * Default pipe buffer size(s), this can be kind-of large now because pipe
 106  * space is pageable.  The pipe code will try to maintain locality of
 107  * reference for performance reasons, so small amounts of outstanding I/O
 108  * will not wipe the cache.
 109  */
 110 #define MINPIPESIZE (PIPE_SIZE/3)
 111 #define MAXPIPESIZE (2*PIPE_SIZE/3)
 112
 113 /*
 114  * Limit the number of "big" pipes
 115  */
 116 #define LIMITBIGPIPES   64
 117 #define PIPEQ_MAX_CACHE 16      /* per-cpu pipe structure cache */
 118
 119 static int pipe_maxbig = LIMITBIGPIPES;
 120 static int pipe_maxcache = PIPEQ_MAX_CACHE;
 121 static int pipe_bigcount;
 122 static int pipe_nbig;
 123 static int pipe_bcache_alloc;
 124 static int pipe_bkmem_alloc;
 125 static int pipe_rblocked_count;
 126 static int pipe_wblocked_count;
 127
 128 SYSCTL_NODE(_kern, OID_AUTO, pipe, CTLFLAG_RW, 0, "Pipe operation");
 129 SYSCTL_INT(_kern_pipe, OID_AUTO, nbig,
 130         CTLFLAG_RD, &pipe_nbig, 0, "numer of big pipes allocated");
 131 SYSCTL_INT(_kern_pipe, OID_AUTO, bigcount,
 132         CTLFLAG_RW, &pipe_bigcount, 0, "number of times pipe expanded");
 133 SYSCTL_INT(_kern_pipe, OID_AUTO, rblocked,
 134         CTLFLAG_RW, &pipe_rblocked_count, 0, "number of times pipe expanded");
 135 SYSCTL_INT(_kern_pipe, OID_AUTO, wblocked,
 136         CTLFLAG_RW, &pipe_wblocked_count, 0, "number of times pipe expanded");
 137 SYSCTL_INT(_kern_pipe, OID_AUTO, maxcache,
 138         CTLFLAG_RW, &pipe_maxcache, 0, "max pipes cached per-cpu");
 139 SYSCTL_INT(_kern_pipe, OID_AUTO, maxbig,
 140         CTLFLAG_RW, &pipe_maxbig, 0, "max number of big pipes");
 141 #ifdef SMP
 142 static int pipe_delay = 5000;   /* 5uS default */
 143 SYSCTL_INT(_kern_pipe, OID_AUTO, delay,
 144         CTLFLAG_RW, &pipe_delay, 0, "SMP delay optimization in ns");
 145 static int pipe_mpsafe = 1;
 146 SYSCTL_INT(_kern_pipe, OID_AUTO, mpsafe,
 147         CTLFLAG_RW, &pipe_mpsafe, 0, "");
 148 #endif
 149 #if !defined(NO_PIPE_SYSCTL_STATS)
 150 SYSCTL_INT(_kern_pipe, OID_AUTO, bcache_alloc,
 151         CTLFLAG_RW, &pipe_bcache_alloc, 0, "pipe buffer from pcpu cache");
 152 SYSCTL_INT(_kern_pipe, OID_AUTO, bkmem_alloc,
 153         CTLFLAG_RW, &pipe_bkmem_alloc, 0, "pipe buffer from kmem");
 154 #endif
 155
 156 static void pipeclose (struct pipe *cpipe);
 157 static void pipe_free_kmem (struct pipe *cpipe);
 158 static int pipe_create (struct pipe **cpipep);
 159 static __inline void pipeselwakeup (struct pipe *cpipe);
 160 static int pipespace (struct pipe *cpipe, int size);
 161
 162 static __inline int
 163 pipeseltest(struct pipe *cpipe)
 164 {
 165         return ((cpipe->pipe_state & PIPE_SEL) ||
 166                 ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) ||
 167                 SLIST_FIRST(&cpipe->pipe_sel.si_note));
 168 }
 169
 170 static __inline void
 171 pipeselwakeup(struct pipe *cpipe)
 172 {
 173         if ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) {
 174                 get_mplock();
 175                 pgsigio(cpipe->pipe_sigio, SIGIO, 0);
 176                 rel_mplock();
 177         }
 178         if (SLIST_FIRST(&cpipe->pipe_sel.si_note)) {
 179                 get_mplock();
 180                 KNOTE(&cpipe->pipe_sel.si_note, 0);
 181                 rel_mplock();
 182         }
 183 }
 184
 185 /*
 186  * These routines are called before and after a UIO.  The UIO
 187  * may block, causing our held tokens to be lost temporarily.
 188  *
 189  * We use these routines to serialize reads against other reads
 190  * and writes against other writes.
 191  *
 192  * The read token is held on entry so *ipp does not race.
 193  */
 194 static __inline int
 195 pipe_start_uio(struct pipe *cpipe, int *ipp)
 196 {
 197         int error;
 198
 199         while (*ipp) {
 200                 *ipp = -1;
 201                 error = tsleep(ipp, PCATCH, "pipexx", 0);
 202                 if (error)
 203                         return (error);
 204         }
 205         *ipp = 1;
 206         return (0);
 207 }
 208
 209 static __inline void
 210 pipe_end_uio(struct pipe *cpipe, int *ipp)
 211 {
 212         if (*ipp < 0) {
 213                 *ipp = 0;
 214                 wakeup(ipp);
 215         } else {
 216                 KKASSERT(*ipp > 0);
 217                 *ipp = 0;
 218         }
 219 }
 220
 221 static __inline void
 222 pipe_get_mplock(int *save)
 223 {
 224 #ifdef SMP
 225         if (pipe_mpsafe == 0) {
 226                 get_mplock();
 227                 *save = 1;
 228         } else
 229 #endif
 230         {
 231                 *save = 0;
 232         }
 233 }
 234
 235 static __inline void
 236 pipe_rel_mplock(int *save)
 237 {
 238 #ifdef SMP
 239         if (*save)
 240                 rel_mplock();
 241 #endif
 242 }
 243
 244
 245 /*
 246  * The pipe system call for the DTYPE_PIPE type of pipes
 247  *
 248  * pipe_args(int dummy)
 249  *
 250  * MPSAFE
 251  */
 252 int
 253 sys_pipe(struct pipe_args *uap)
 254 {
 255         struct thread *td = curthread;
 256         struct filedesc *fdp = td->td_proc->p_fd;
 257         struct file *rf, *wf;
 258         struct pipe *rpipe, *wpipe;
 259         int fd1, fd2, error;
 260
 261         rpipe = wpipe = NULL;
 262         if (pipe_create(&rpipe) || pipe_create(&wpipe)) {
 263                 pipeclose(rpipe);
 264                 pipeclose(wpipe);
 265                 return (ENFILE);
 266         }
 267
 268         error = falloc(td->td_lwp, &rf, &fd1);
 269         if (error) {
 270                 pipeclose(rpipe);
 271                 pipeclose(wpipe);
 272                 return (error);
 273         }
 274         uap->sysmsg_fds[0] = fd1;
 275
 276         /*
 277          * Warning: once we've gotten past allocation of the fd for the
 278          * read-side, we can only drop the read side via fdrop() in order
 279          * to avoid races against processes which manage to dup() the read
 280          * side while we are blocked trying to allocate the write side.
 281          */
 282         rf->f_type = DTYPE_PIPE;
 283         rf->f_flag = FREAD | FWRITE;
 284         rf->f_ops = &pipeops;
 285         rf->f_data = rpipe;
 286         error = falloc(td->td_lwp, &wf, &fd2);
 287         if (error) {
 288                 fsetfd(fdp, NULL, fd1);
 289                 fdrop(rf);
 290                 /* rpipe has been closed by fdrop(). */
 291                 pipeclose(wpipe);
 292                 return (error);
 293         }
 294         wf->f_type = DTYPE_PIPE;
 295         wf->f_flag = FREAD | FWRITE;
 296         wf->f_ops = &pipeops;
 297         wf->f_data = wpipe;
 298         uap->sysmsg_fds[1] = fd2;
 299
 300         rpipe->pipe_slock = kmalloc(sizeof(struct lock),
 301                                     M_PIPE, M_WAITOK|M_ZERO);
 302         wpipe->pipe_slock = rpipe->pipe_slock;
 303         rpipe->pipe_peer = wpipe;
 304         wpipe->pipe_peer = rpipe;
 305         lockinit(rpipe->pipe_slock, "pipecl", 0, 0);
 306
 307         /*
 308          * Once activated the peer relationship remains valid until
 309          * both sides are closed.
 310          */
 311         fsetfd(fdp, rf, fd1);
 312         fsetfd(fdp, wf, fd2);
 313         fdrop(rf);
 314         fdrop(wf);
 315
 316         return (0);
 317 }
 318
 319 /*
 320  * Allocate kva for pipe circular buffer, the space is pageable
 321  * This routine will 'realloc' the size of a pipe safely, if it fails
 322  * it will retain the old buffer.
 323  * If it fails it will return ENOMEM.
 324  */
 325 static int
 326 pipespace(struct pipe *cpipe, int size)
 327 {
 328         struct vm_object *object;
 329         caddr_t buffer;
 330         int npages, error;
 331
 332         npages = round_page(size) / PAGE_SIZE;
 333         object = cpipe->pipe_buffer.object;
 334
 335         /*
 336          * [re]create the object if necessary and reserve space for it
 337          * in the kernel_map.  The object and memory are pageable.  On
 338          * success, free the old resources before assigning the new
 339          * ones.
 340          */
 341         if (object == NULL || object->size != npages) {
 342                 get_mplock();
 343                 object = vm_object_allocate(OBJT_DEFAULT, npages);
 344                 buffer = (caddr_t)vm_map_min(&kernel_map);
 345
 346                 error = vm_map_find(&kernel_map, object, 0,
 347                                     (vm_offset_t *)&buffer,
 348                                     size, PAGE_SIZE,
 349                                     1, VM_MAPTYPE_NORMAL,
 350                                     VM_PROT_ALL, VM_PROT_ALL,
 351                                     0);
 352
 353                 if (error != KERN_SUCCESS) {
 354                         vm_object_deallocate(object);
 355                         rel_mplock();
 356                         return (ENOMEM);
 357                 }
 358                 pipe_free_kmem(cpipe);
 359                 rel_mplock();
 360                 cpipe->pipe_buffer.object = object;
 361                 cpipe->pipe_buffer.buffer = buffer;
 362                 cpipe->pipe_buffer.size = size;
 363                 ++pipe_bkmem_alloc;
 364         } else {
 365                 ++pipe_bcache_alloc;
 366         }
 367         cpipe->pipe_buffer.rindex = 0;
 368         cpipe->pipe_buffer.windex = 0;
 369         return (0);
 370 }
 371
 372 /*
 373  * Initialize and allocate VM and memory for pipe, pulling the pipe from
 374  * our per-cpu cache if possible.  For now make sure it is sized for the
 375  * smaller PIPE_SIZE default.
 376  */
 377 static int
 378 pipe_create(struct pipe **cpipep)
 379 {
 380         globaldata_t gd = mycpu;
 381         struct pipe *cpipe;
 382         int error;
 383
 384         if ((cpipe = gd->gd_pipeq) != NULL) {
 385                 gd->gd_pipeq = cpipe->pipe_peer;
 386                 --gd->gd_pipeqcount;
 387                 cpipe->pipe_peer = NULL;
 388                 cpipe->pipe_wantwcnt = 0;
 389         } else {
 390                 cpipe = kmalloc(sizeof(struct pipe), M_PIPE, M_WAITOK|M_ZERO);
 391         }
 392         *cpipep = cpipe;
 393         if ((error = pipespace(cpipe, PIPE_SIZE)) != 0)
 394                 return (error);
 395         vfs_timestamp(&cpipe->pipe_ctime);
 396         cpipe->pipe_atime = cpipe->pipe_ctime;
 397         cpipe->pipe_mtime = cpipe->pipe_ctime;
 398         lwkt_token_init(&cpipe->pipe_rlock, 1);
 399         lwkt_token_init(&cpipe->pipe_wlock, 1);
 400         return (0);
 401 }
 402
 403 /*
 404  * MPALMOSTSAFE (acquires mplock)
 405  */
 406 static int
 407 pipe_read(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 408 {
 409         struct pipe *rpipe;
 410         int error;
 411         size_t nread = 0;
 412         int nbio;
 413         u_int size;     /* total bytes available */
 414         u_int nsize;    /* total bytes to read */
 415         u_int rindex;   /* contiguous bytes available */
 416         int notify_writer;
 417         int mpsave;
 418         int bigread;
 419         int bigcount;
 420
 421         if (uio->uio_resid == 0)
 422                 return(0);
 423
 424         /*
 425          * Setup locks, calculate nbio
 426          */
 427         pipe_get_mplock(&mpsave);
 428         rpipe = (struct pipe *)fp->f_data;
 429         lwkt_gettoken(&rpipe->pipe_rlock);
 430
 431         if (fflags & O_FBLOCKING)
 432                 nbio = 0;
 433         else if (fflags & O_FNONBLOCKING)
 434                 nbio = 1;
 435         else if (fp->f_flag & O_NONBLOCK)
 436                 nbio = 1;
 437         else
 438                 nbio = 0;
 439
 440         /*
 441          * Reads are serialized.  Note howeverthat pipe_buffer.buffer and
 442          * pipe_buffer.size can change out from under us when the number
 443          * of bytes in the buffer are zero due to the write-side doing a
 444          * pipespace().
 445          */
 446         error = pipe_start_uio(rpipe, &rpipe->pipe_rip);
 447         if (error) {
 448                 pipe_rel_mplock(&mpsave);
 449                 lwkt_reltoken(&rpipe->pipe_rlock);
 450                 return (error);
 451         }
 452         notify_writer = 0;
 453
 454         bigread = (uio->uio_resid > 10 * 1024 * 1024);
 455         bigcount = 10;
 456
 457         while (uio->uio_resid) {
 458                 /*
 459                  * Don't hog the cpu.
 460                  */
 461                 if (bigread && --bigcount == 0) {
 462                         lwkt_user_yield();
 463                         bigcount = 10;
 464                         if (CURSIG(curthread->td_lwp)) {
 465                                 error = EINTR;
 466                                 break;
 467                         }
 468                 }
 469
 470                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 471                 cpu_lfence();
 472                 if (size) {
 473                         rindex = rpipe->pipe_buffer.rindex &
 474                                  (rpipe->pipe_buffer.size - 1);
 475                         nsize = size;
 476                         if (nsize > rpipe->pipe_buffer.size - rindex)
 477                                 nsize = rpipe->pipe_buffer.size - rindex;
 478                         nsize = szmin(nsize, uio->uio_resid);
 479
 480                         error = uiomove(&rpipe->pipe_buffer.buffer[rindex],
 481                                         nsize, uio);
 482                         if (error)
 483                                 break;
 484                         cpu_mfence();
 485                         rpipe->pipe_buffer.rindex += nsize;
 486                         nread += nsize;
 487
 488                         /*
 489                          * If the FIFO is still over half full just continue
 490                          * and do not try to notify the writer yet.
 491                          */
 492                         if (size - nsize >= (rpipe->pipe_buffer.size >> 1)) {
 493                                 notify_writer = 0;
 494                                 continue;
 495                         }
 496
 497                         /*
 498                          * When the FIFO is less then half full notify any
 499                          * waiting writer.  WANTW can be checked while
 500                          * holding just the rlock.
 501                          */
 502                         notify_writer = 1;
 503                         if ((rpipe->pipe_state & PIPE_WANTW) == 0)
 504                                 continue;
 505                 }
 506
 507                 /*
 508                  * If the "write-side" was blocked we wake it up.  This code
 509                  * is reached either when the buffer is completely emptied
 510                  * or if it becomes more then half-empty.
 511                  *
 512                  * Pipe_state can only be modified if both the rlock and
 513                  * wlock are held.
 514                  */
 515                 if (rpipe->pipe_state & PIPE_WANTW) {
 516                         lwkt_gettoken(&rpipe->pipe_wlock);
 517                         if (rpipe->pipe_state & PIPE_WANTW) {
 518                                 notify_writer = 0;
 519                                 rpipe->pipe_state &= ~PIPE_WANTW;
 520                                 lwkt_reltoken(&rpipe->pipe_wlock);
 521                                 wakeup(rpipe);
 522                         } else {
 523                                 lwkt_reltoken(&rpipe->pipe_wlock);
 524                         }
 525                 }
 526
 527                 /*
 528                  * Pick up our copy loop again if the writer sent data to
 529                  * us while we were messing around.
 530                  *
 531                  * On a SMP box poll up to pipe_delay nanoseconds for new
 532                  * data.  Typically a value of 2000 to 4000 is sufficient
 533                  * to eradicate most IPIs/tsleeps/wakeups when a pipe
 534                  * is used for synchronous communications with small packets,
 535                  * and 8000 or so (8uS) will pipeline large buffer xfers
 536                  * between cpus over a pipe.
 537                  *
 538                  * For synchronous communications a hit means doing a
 539                  * full Awrite-Bread-Bwrite-Aread cycle in less then 2uS,
 540                  * where as miss requiring a tsleep/wakeup sequence
 541                  * will take 7uS or more.
 542                  */
 543                 if (rpipe->pipe_buffer.windex != rpipe->pipe_buffer.rindex)
 544                         continue;
 545
 546 #if defined(SMP) && defined(_RDTSC_SUPPORTED_)
 547                 if (pipe_delay) {
 548                         int64_t tsc_target;
 549                         int good = 0;
 550
 551                         tsc_target = tsc_get_target(pipe_delay);
 552                         while (tsc_test_target(tsc_target) == 0) {
 553                                 if (rpipe->pipe_buffer.windex !=
 554                                     rpipe->pipe_buffer.rindex) {
 555                                         good = 1;
 556                                         break;
 557                                 }
 558                         }
 559                         if (good)
 560                                 continue;
 561                 }
 562 #endif
 563
 564                 /*
 565                  * Detect EOF condition, do not set error.
 566                  */
 567                 if (rpipe->pipe_state & PIPE_REOF)
 568                         break;
 569
 570                 /*
 571                  * Break if some data was read, or if this was a non-blocking
 572                  * read.
 573                  */
 574                 if (nread > 0)
 575                         break;
 576
 577                 if (nbio) {
 578                         error = EAGAIN;
 579                         break;
 580                 }
 581
 582                 /*
 583                  * Last chance, interlock with WANTR.
 584                  */
 585                 lwkt_gettoken(&rpipe->pipe_wlock);
 586                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 587                 if (size) {
 588                         lwkt_reltoken(&rpipe->pipe_wlock);
 589                         continue;
 590                 }
 591
 592                 /*
 593                  * Retest EOF - acquiring a new token can temporarily release
 594                  * tokens already held.
 595                  */
 596                 if (rpipe->pipe_state & PIPE_REOF) {
 597                         lwkt_reltoken(&rpipe->pipe_wlock);
 598                         break;
 599                 }
 600
 601                 /*
 602                  * If there is no more to read in the pipe, reset its
 603                  * pointers to the beginning.  This improves cache hit
 604                  * stats.
 605                  *
 606                  * We need both locks to modify both pointers, and there
 607                  * must also not be a write in progress or the uiomove()
 608                  * in the write might block and temporarily release
 609                  * its wlock, then reacquire and update windex.  We are
 610                  * only serialized against reads, not writes.
 611                  *
 612                  * XXX should we even bother resetting the indices?  It
 613                  *     might actually be more cache efficient not to.
 614                  */
 615                 if (rpipe->pipe_buffer.rindex == rpipe->pipe_buffer.windex &&
 616                     rpipe->pipe_wip == 0) {
 617                         rpipe->pipe_buffer.rindex = 0;
 618                         rpipe->pipe_buffer.windex = 0;
 619                 }
 620
 621                 /*
 622                  * Wait for more data.
 623                  *
 624                  * Pipe_state can only be set if both the rlock and wlock
 625                  * are held.
 626                  */
 627                 rpipe->pipe_state |= PIPE_WANTR;
 628                 tsleep_interlock(rpipe, PCATCH);
 629                 lwkt_reltoken(&rpipe->pipe_wlock);
 630                 error = tsleep(rpipe, PCATCH | PINTERLOCKED, "piperd", 0);
 631                 ++pipe_rblocked_count;
 632                 if (error)
 633                         break;
 634         }
 635         pipe_end_uio(rpipe, &rpipe->pipe_rip);
 636
 637         /*
 638          * Uptime last access time
 639          */
 640         if (error == 0 && nread)
 641                 vfs_timestamp(&rpipe->pipe_atime);
 642
 643         /*
 644          * If we drained the FIFO more then half way then handle
 645          * write blocking hysteresis.
 646          *
 647          * Note that PIPE_WANTW cannot be set by the writer without
 648          * it holding both rlock and wlock, so we can test it
 649          * while holding just rlock.
 650          */
 651         if (notify_writer) {
 652                 if (rpipe->pipe_state & PIPE_WANTW) {
 653                         lwkt_gettoken(&rpipe->pipe_wlock);
 654                         if (rpipe->pipe_state & PIPE_WANTW) {
 655                                 rpipe->pipe_state &= ~PIPE_WANTW;
 656                                 lwkt_reltoken(&rpipe->pipe_wlock);
 657                                 wakeup(rpipe);
 658                         } else {
 659                                 lwkt_reltoken(&rpipe->pipe_wlock);
 660                         }
 661                 }
 662                 if (pipeseltest(rpipe)) {
 663                         lwkt_gettoken(&rpipe->pipe_wlock);
 664                         pipeselwakeup(rpipe);
 665                         lwkt_reltoken(&rpipe->pipe_wlock);
 666                 }
 667         }
 668         /*size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;*/
 669         lwkt_reltoken(&rpipe->pipe_rlock);
 670
 671         pipe_rel_mplock(&mpsave);
 672         return (error);
 673 }
 674
 675 /*
 676  * MPALMOSTSAFE - acquires mplock
 677  */
 678 static int
 679 pipe_write(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 680 {
 681         int error;
 682         int orig_resid;
 683         int nbio;
 684         struct pipe *wpipe, *rpipe;
 685         u_int windex;
 686         u_int space;
 687         u_int wcount;
 688         int mpsave;
 689         int bigwrite;
 690         int bigcount;
 691
 692         pipe_get_mplock(&mpsave);
 693
 694         /*
 695          * Writes go to the peer.  The peer will always exist.
 696          */
 697         rpipe = (struct pipe *) fp->f_data;
 698         wpipe = rpipe->pipe_peer;
 699         lwkt_gettoken(&wpipe->pipe_wlock);
 700         if (wpipe->pipe_state & PIPE_WEOF) {
 701                 pipe_rel_mplock(&mpsave);
 702                 lwkt_reltoken(&wpipe->pipe_wlock);
 703                 return (EPIPE);
 704         }
 705
 706         /*
 707          * Degenerate case (EPIPE takes prec)
 708          */
 709         if (uio->uio_resid == 0) {
 710                 pipe_rel_mplock(&mpsave);
 711                 lwkt_reltoken(&wpipe->pipe_wlock);
 712                 return(0);
 713         }
 714
 715         /*
 716          * Writes are serialized (start_uio must be called with wlock)
 717          */
 718         error = pipe_start_uio(wpipe, &wpipe->pipe_wip);
 719         if (error) {
 720                 pipe_rel_mplock(&mpsave);
 721                 lwkt_reltoken(&wpipe->pipe_wlock);
 722                 return (error);
 723         }
 724
 725         if (fflags & O_FBLOCKING)
 726                 nbio = 0;
 727         else if (fflags & O_FNONBLOCKING)
 728                 nbio = 1;
 729         else if (fp->f_flag & O_NONBLOCK)
 730                 nbio = 1;
 731         else
 732                 nbio = 0;
 733
 734         /*
 735          * If it is advantageous to resize the pipe buffer, do
 736          * so.  We are write-serialized so we can block safely.
 737          */
 738         if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 739             (pipe_nbig < pipe_maxbig) &&
 740             wpipe->pipe_wantwcnt > 4 &&
 741             (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 742                 /*
 743                  * Recheck after lock.
 744                  */
 745                 lwkt_gettoken(&wpipe->pipe_rlock);
 746                 if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 747                     (pipe_nbig < pipe_maxbig) &&
 748                     (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 749                         atomic_add_int(&pipe_nbig, 1);
 750                         if (pipespace(wpipe, BIG_PIPE_SIZE) == 0)
 751                                 ++pipe_bigcount;
 752                         else
 753                                 atomic_subtract_int(&pipe_nbig, 1);
 754                 }
 755                 lwkt_reltoken(&wpipe->pipe_rlock);
 756         }
 757
 758         orig_resid = uio->uio_resid;
 759         wcount = 0;
 760
 761         bigwrite = (uio->uio_resid > 10 * 1024 * 1024);
 762         bigcount = 10;
 763
 764         while (uio->uio_resid) {
 765                 if (wpipe->pipe_state & PIPE_WEOF) {
 766                         error = EPIPE;
 767                         break;
 768                 }
 769
 770                 /*
 771                  * Don't hog the cpu.
 772                  */
 773                 if (bigwrite && --bigcount == 0) {
 774                         lwkt_user_yield();
 775                         bigcount = 10;
 776                         if (CURSIG(curthread->td_lwp)) {
 777                                 error = EINTR;
 778                                 break;
 779                         }
 780                 }
 781
 782                 windex = wpipe->pipe_buffer.windex &
 783                          (wpipe->pipe_buffer.size - 1);
 784                 space = wpipe->pipe_buffer.size -
 785                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 786                 cpu_lfence();
 787
 788                 /* Writes of size <= PIPE_BUF must be atomic. */
 789                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 790                         space = 0;
 791
 792                 /*
 793                  * Write to fill, read size handles write hysteresis.  Also
 794                  * additional restrictions can cause select-based non-blocking
 795                  * writes to spin.
 796                  */
 797                 if (space > 0) {
 798                         u_int segsize;
 799
 800                         /*
 801                          * Transfer size is minimum of uio transfer
 802                          * and free space in pipe buffer.
 803                          *
 804                          * Limit each uiocopy to no more then PIPE_SIZE
 805                          * so we can keep the gravy train going on a
 806                          * SMP box.  This doubles the performance for
 807                          * write sizes > 16K.  Otherwise large writes
 808                          * wind up doing an inefficient synchronous
 809                          * ping-pong.
 810                          */
 811                         space = szmin(space, uio->uio_resid);
 812                         if (space > PIPE_SIZE)
 813                                 space = PIPE_SIZE;
 814
 815                         /*
 816                          * First segment to transfer is minimum of
 817                          * transfer size and contiguous space in
 818                          * pipe buffer.  If first segment to transfer
 819                          * is less than the transfer size, we've got
 820                          * a wraparound in the buffer.
 821                          */
 822                         segsize = wpipe->pipe_buffer.size - windex;
 823                         if (segsize > space)
 824                                 segsize = space;
 825
 826 #ifdef SMP
 827                         /*
 828                          * If this is the first loop and the reader is
 829                          * blocked, do a preemptive wakeup of the reader.
 830                          *
 831                          * On SMP the IPI latency plus the wlock interlock
 832                          * on the reader side is the fastest way to get the
 833                          * reader going.  (The scheduler will hard loop on
 834                          * lock tokens).
 835                          *
 836                          * NOTE: We can't clear WANTR here without acquiring
 837                          * the rlock, which we don't want to do here!
 838                          */
 839                         if ((wpipe->pipe_state & PIPE_WANTR) && pipe_mpsafe > 1)
 840                                 wakeup(wpipe);
 841 #endif
 842
 843                         /*
 844                          * Transfer segment, which may include a wrap-around.
 845                          * Update windex to account for both all in one go
 846                          * so the reader can read() the data atomically.
 847                          */
 848                         error = uiomove(&wpipe->pipe_buffer.buffer[windex],
 849                                         segsize, uio);
 850                         if (error == 0 && segsize < space) {
 851                                 segsize = space - segsize;
 852                                 error = uiomove(&wpipe->pipe_buffer.buffer[0],
 853                                                 segsize, uio);
 854                         }
 855                         if (error)
 856                                 break;
 857                         cpu_mfence();
 858                         wpipe->pipe_buffer.windex += space;
 859                         wcount += space;
 860                         continue;
 861                 }
 862
 863                 /*
 864                  * We need both the rlock and the wlock to interlock against
 865                  * the EOF, WANTW, and size checks, and to modify pipe_state.
 866                  *
 867                  * These are token locks so we do not have to worry about
 868                  * deadlocks.
 869                  */
 870                 lwkt_gettoken(&wpipe->pipe_rlock);
 871
 872                 /*
 873                  * If the "read-side" has been blocked, wake it up now
 874                  * and yield to let it drain synchronously rather
 875                  * then block.
 876                  */
 877                 if (wpipe->pipe_state & PIPE_WANTR) {
 878                         wpipe->pipe_state &= ~PIPE_WANTR;
 879                         wakeup(wpipe);
 880                 }
 881
 882                 /*
 883                  * don't block on non-blocking I/O
 884                  */
 885                 if (nbio) {
 886                         lwkt_reltoken(&wpipe->pipe_rlock);
 887                         error = EAGAIN;
 888                         break;
 889                 }
 890
 891                 /*
 892                  * re-test whether we have to block in the writer after
 893                  * acquiring both locks, in case the reader opened up
 894                  * some space.
 895                  */
 896                 space = wpipe->pipe_buffer.size -
 897                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 898                 cpu_lfence();
 899                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 900                         space = 0;
 901
 902                 /*
 903                  * Retest EOF - acquiring a new token can temporarily release
 904                  * tokens already held.
 905                  */
 906                 if (wpipe->pipe_state & PIPE_WEOF) {
 907                         lwkt_reltoken(&wpipe->pipe_rlock);
 908                         error = EPIPE;
 909                         break;
 910                 }
 911
 912                 /*
 913                  * We have no more space and have something to offer,
 914                  * wake up select/poll.
 915                  */
 916                 if (space == 0) {
 917                         wpipe->pipe_state |= PIPE_WANTW;
 918                         ++wpipe->pipe_wantwcnt;
 919                         pipeselwakeup(wpipe);
 920                         if (wpipe->pipe_state & PIPE_WANTW)
 921                                 error = tsleep(wpipe, PCATCH, "pipewr", 0);
 922                         ++pipe_wblocked_count;
 923                 }
 924                 lwkt_reltoken(&wpipe->pipe_rlock);
 925
 926                 /*
 927                  * Break out if we errored or the read side wants us to go
 928                  * away.
 929                  */
 930                 if (error)
 931                         break;
 932                 if (wpipe->pipe_state & PIPE_WEOF) {
 933                         error = EPIPE;
 934                         break;
 935                 }
 936         }
 937         pipe_end_uio(wpipe, &wpipe->pipe_wip);
 938
 939         /*
 940          * If we have put any characters in the buffer, we wake up
 941          * the reader.
 942          *
 943          * Both rlock and wlock are required to be able to modify pipe_state.
 944          */
 945         if (wpipe->pipe_buffer.windex != wpipe->pipe_buffer.rindex) {
 946                 if (wpipe->pipe_state & PIPE_WANTR) {
 947                         lwkt_gettoken(&wpipe->pipe_rlock);
 948                         if (wpipe->pipe_state & PIPE_WANTR) {
 949                                 wpipe->pipe_state &= ~PIPE_WANTR;
 950                                 lwkt_reltoken(&wpipe->pipe_rlock);
 951                                 wakeup(wpipe);
 952                         } else {
 953                                 lwkt_reltoken(&wpipe->pipe_rlock);
 954                         }
 955                 }
 956                 if (pipeseltest(wpipe)) {
 957                         lwkt_gettoken(&wpipe->pipe_rlock);
 958                         pipeselwakeup(wpipe);
 959                         lwkt_reltoken(&wpipe->pipe_rlock);
 960                 }
 961         }
 962
 963         /*
 964          * Don't return EPIPE if I/O was successful
 965          */
 966         if ((wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex) &&
 967             (uio->uio_resid == 0) &&
 968             (error == EPIPE)) {
 969                 error = 0;
 970         }
 971
 972         if (error == 0)
 973                 vfs_timestamp(&wpipe->pipe_mtime);
 974
 975         /*
 976          * We have something to offer,
 977          * wake up select/poll.
 978          */
 979         /*space = wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex;*/
 980         lwkt_reltoken(&wpipe->pipe_wlock);
 981         pipe_rel_mplock(&mpsave);
 982         return (error);
 983 }
 984
 985 /*
 986  * MPALMOSTSAFE - acquires mplock
 987  *
 988  * we implement a very minimal set of ioctls for compatibility with sockets.
 989  */
 990 int
 991 pipe_ioctl(struct file *fp, u_long cmd, caddr_t data,
 992            struct ucred *cred, struct sysmsg *msg)
 993 {
 994         struct pipe *mpipe;
 995         int error;
 996         int mpsave;
 997
 998         pipe_get_mplock(&mpsave);
 999         mpipe = (struct pipe *)fp->f_data;
1000
1001         lwkt_gettoken(&mpipe->pipe_rlock);
1002         lwkt_gettoken(&mpipe->pipe_wlock);
1003
1004         switch (cmd) {
1005         case FIOASYNC:
1006                 if (*(int *)data) {
1007                         mpipe->pipe_state |= PIPE_ASYNC;
1008                 } else {
1009                         mpipe->pipe_state &= ~PIPE_ASYNC;
1010                 }
1011                 error = 0;
1012                 break;
1013         case FIONREAD:
1014                 *(int *)data = mpipe->pipe_buffer.windex -
1015                                 mpipe->pipe_buffer.rindex;
1016                 error = 0;
1017                 break;
1018         case FIOSETOWN:
1019                 get_mplock();
1020                 error = fsetown(*(int *)data, &mpipe->pipe_sigio);
1021                 rel_mplock();
1022                 break;
1023         case FIOGETOWN:
1024                 *(int *)data = fgetown(mpipe->pipe_sigio);
1025                 error = 0;
1026                 break;
1027         case TIOCSPGRP:
1028                 /* This is deprecated, FIOSETOWN should be used instead. */
1029                 get_mplock();
1030                 error = fsetown(-(*(int *)data), &mpipe->pipe_sigio);
1031                 rel_mplock();
1032                 break;
1033
1034         case TIOCGPGRP:
1035                 /* This is deprecated, FIOGETOWN should be used instead. */
1036                 *(int *)data = -fgetown(mpipe->pipe_sigio);
1037                 error = 0;
1038                 break;
1039         default:
1040                 error = ENOTTY;
1041                 break;
1042         }
1043         lwkt_reltoken(&mpipe->pipe_wlock);
1044         lwkt_reltoken(&mpipe->pipe_rlock);
1045         pipe_rel_mplock(&mpsave);
1046
1047         return (error);
1048 }
1049
1050 /*
1051  * MPSAFE
1052  */
1053 static int
1054 pipe_stat(struct file *fp, struct stat *ub, struct ucred *cred)
1055 {
1056         struct pipe *pipe;
1057         int mpsave;
1058
1059         pipe_get_mplock(&mpsave);
1060         pipe = (struct pipe *)fp->f_data;
1061
1062         bzero((caddr_t)ub, sizeof(*ub));
1063         ub->st_mode = S_IFIFO;
1064         ub->st_blksize = pipe->pipe_buffer.size;
1065         ub->st_size = pipe->pipe_buffer.windex - pipe->pipe_buffer.rindex;
1066         ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize;
1067         ub->st_atimespec = pipe->pipe_atime;
1068         ub->st_mtimespec = pipe->pipe_mtime;
1069         ub->st_ctimespec = pipe->pipe_ctime;
1070         /*
1071          * Left as 0: st_dev, st_ino, st_nlink, st_uid, st_gid, st_rdev,
1072          * st_flags, st_gen.
1073          * XXX (st_dev, st_ino) should be unique.
1074          */
1075         pipe_rel_mplock(&mpsave);
1076         return (0);
1077 }
1078
1079 /*
1080  * MPALMOSTSAFE - acquires mplock
1081  */
1082 static int
1083 pipe_close(struct file *fp)
1084 {
1085         struct pipe *cpipe;
1086
1087         get_mplock();
1088         cpipe = (struct pipe *)fp->f_data;
1089         fp->f_ops = &badfileops;
1090         fp->f_data = NULL;
1091         funsetown(cpipe->pipe_sigio);
1092         pipeclose(cpipe);
1093         rel_mplock();
1094         return (0);
1095 }
1096
1097 /*
1098  * Shutdown one or both directions of a full-duplex pipe.
1099  *
1100  * MPALMOSTSAFE - acquires mplock
1101  */
1102 static int
1103 pipe_shutdown(struct file *fp, int how)
1104 {
1105         struct pipe *rpipe;
1106         struct pipe *wpipe;
1107         int error = EPIPE;
1108         int mpsave;
1109
1110         pipe_get_mplock(&mpsave);
1111         rpipe = (struct pipe *)fp->f_data;
1112         wpipe = rpipe->pipe_peer;
1113
1114         /*
1115          * We modify pipe_state on both pipes, which means we need
1116          * all four tokens!
1117          */
1118         lwkt_gettoken(&rpipe->pipe_rlock);
1119         lwkt_gettoken(&rpipe->pipe_wlock);
1120         lwkt_gettoken(&wpipe->pipe_rlock);
1121         lwkt_gettoken(&wpipe->pipe_wlock);
1122
1123         switch(how) {
1124         case SHUT_RDWR:
1125         case SHUT_RD:
1126                 rpipe->pipe_state |= PIPE_REOF;         /* my reads */
1127                 rpipe->pipe_state |= PIPE_WEOF;         /* peer writes */
1128                 if (rpipe->pipe_state & PIPE_WANTR) {
1129                         rpipe->pipe_state &= ~PIPE_WANTR;
1130                         wakeup(rpipe);
1131                 }
1132                 if (rpipe->pipe_state & PIPE_WANTW) {
1133                         rpipe->pipe_state &= ~PIPE_WANTW;
1134                         wakeup(rpipe);
1135                 }
1136                 error = 0;
1137                 if (how == SHUT_RD)
1138                         break;
1139                 /* fall through */
1140         case SHUT_WR:
1141                 wpipe->pipe_state |= PIPE_REOF;         /* peer reads */
1142                 wpipe->pipe_state |= PIPE_WEOF;         /* my writes */
1143                 if (wpipe->pipe_state & PIPE_WANTR) {
1144                         wpipe->pipe_state &= ~PIPE_WANTR;
1145                         wakeup(wpipe);
1146                 }
1147                 if (wpipe->pipe_state & PIPE_WANTW) {
1148                         wpipe->pipe_state &= ~PIPE_WANTW;
1149                         wakeup(wpipe);
1150                 }
1151                 error = 0;
1152                 break;
1153         }
1154         pipeselwakeup(rpipe);
1155         pipeselwakeup(wpipe);
1156
1157         lwkt_reltoken(&wpipe->pipe_wlock);
1158         lwkt_reltoken(&wpipe->pipe_rlock);
1159         lwkt_reltoken(&rpipe->pipe_wlock);
1160         lwkt_reltoken(&rpipe->pipe_rlock);
1161
1162         pipe_rel_mplock(&mpsave);
1163         return (error);
1164 }
1165
1166 static void
1167 pipe_free_kmem(struct pipe *cpipe)
1168 {
1169         if (cpipe->pipe_buffer.buffer != NULL) {
1170                 if (cpipe->pipe_buffer.size > PIPE_SIZE)
1171                         atomic_subtract_int(&pipe_nbig, 1);
1172                 kmem_free(&kernel_map,
1173                         (vm_offset_t)cpipe->pipe_buffer.buffer,
1174                         cpipe->pipe_buffer.size);
1175                 cpipe->pipe_buffer.buffer = NULL;
1176                 cpipe->pipe_buffer.object = NULL;
1177         }
1178 }
1179
1180 /*
1181  * Close the pipe.  The slock must be held to interlock against simultanious
1182  * closes.  The rlock and wlock must be held to adjust the pipe_state.
1183  */
1184 static void
1185 pipeclose(struct pipe *cpipe)
1186 {
1187         globaldata_t gd;
1188         struct pipe *ppipe;
1189
1190         if (cpipe == NULL)
1191                 return;
1192
1193         /*
1194          * The slock may not have been allocated yet (close during
1195          * initialization)
1196          *
1197          * We need both the read and write tokens to modify pipe_state.
1198          */
1199         if (cpipe->pipe_slock)
1200                 lockmgr(cpipe->pipe_slock, LK_EXCLUSIVE);
1201         lwkt_gettoken(&cpipe->pipe_rlock);
1202         lwkt_gettoken(&cpipe->pipe_wlock);
1203
1204         /*
1205          * Set our state, wakeup anyone waiting in select, and
1206          * wakeup anyone blocked on our pipe.
1207          */
1208         cpipe->pipe_state |= PIPE_CLOSED | PIPE_REOF | PIPE_WEOF;
1209         pipeselwakeup(cpipe);
1210         if (cpipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1211                 cpipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1212                 wakeup(cpipe);
1213         }
1214
1215         /*
1216          * Disconnect from peer.
1217          */
1218         if ((ppipe = cpipe->pipe_peer) != NULL) {
1219                 lwkt_gettoken(&ppipe->pipe_rlock);
1220                 lwkt_gettoken(&ppipe->pipe_wlock);
1221                 ppipe->pipe_state |= PIPE_REOF | PIPE_WEOF;
1222                 pipeselwakeup(ppipe);
1223                 if (ppipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1224                         ppipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1225                         wakeup(ppipe);
1226                 }
1227                 if (SLIST_FIRST(&ppipe->pipe_sel.si_note)) {
1228                         get_mplock();
1229                         KNOTE(&ppipe->pipe_sel.si_note, 0);
1230                         rel_mplock();
1231                 }
1232                 lwkt_reltoken(&ppipe->pipe_wlock);
1233                 lwkt_reltoken(&ppipe->pipe_rlock);
1234         }
1235
1236         /*
1237          * If the peer is also closed we can free resources for both
1238          * sides, otherwise we leave our side intact to deal with any
1239          * races (since we only have the slock).
1240          */
1241         if (ppipe && (ppipe->pipe_state & PIPE_CLOSED)) {
1242                 cpipe->pipe_peer = NULL;
1243                 ppipe->pipe_peer = NULL;
1244                 ppipe->pipe_slock = NULL;       /* we will free the slock */
1245                 pipeclose(ppipe);
1246                 ppipe = NULL;
1247         }
1248
1249         lwkt_reltoken(&cpipe->pipe_wlock);
1250         lwkt_reltoken(&cpipe->pipe_rlock);
1251         if (cpipe->pipe_slock)
1252                 lockmgr(cpipe->pipe_slock, LK_RELEASE);
1253
1254         /*
1255          * If we disassociated from our peer we can free resources
1256          */
1257         if (ppipe == NULL) {
1258                 gd = mycpu;
1259                 if (cpipe->pipe_slock) {
1260                         kfree(cpipe->pipe_slock, M_PIPE);
1261                         cpipe->pipe_slock = NULL;
1262                 }
1263                 if (gd->gd_pipeqcount >= pipe_maxcache ||
1264                     cpipe->pipe_buffer.size != PIPE_SIZE
1265                 ) {
1266                         pipe_free_kmem(cpipe);
1267                         kfree(cpipe, M_PIPE);
1268                 } else {
1269                         cpipe->pipe_state = 0;
1270                         cpipe->pipe_peer = gd->gd_pipeq;
1271                         gd->gd_pipeq = cpipe;
1272                         ++gd->gd_pipeqcount;
1273                 }
1274         }
1275 }
1276
1277 /*
1278  * MPALMOSTSAFE - acquires mplock
1279  */
1280 static int
1281 pipe_kqfilter(struct file *fp, struct knote *kn)
1282 {
1283         struct pipe *cpipe;
1284
1285         get_mplock();
1286         cpipe = (struct pipe *)kn->kn_fp->f_data;
1287
1288         switch (kn->kn_filter) {
1289         case EVFILT_READ:
1290                 kn->kn_fop = &pipe_rfiltops;
1291                 break;
1292         case EVFILT_WRITE:
1293                 kn->kn_fop = &pipe_wfiltops;
1294                 cpipe = cpipe->pipe_peer;
1295                 if (cpipe == NULL) {
1296                         /* other end of pipe has been closed */
1297                         rel_mplock();
1298                         return (EPIPE);
1299                 }
1300                 break;
1301         default:
1302                 rel_mplock();
1303                 return (EOPNOTSUPP);
1304         }
1305         kn->kn_hook = (caddr_t)cpipe;
1306
1307         SLIST_INSERT_HEAD(&cpipe->pipe_sel.si_note, kn, kn_selnext);
1308         rel_mplock();
1309         return (0);
1310 }
1311
1312 static void
1313 filt_pipedetach(struct knote *kn)
1314 {
1315         struct pipe *cpipe = (struct pipe *)kn->kn_hook;
1316
1317         SLIST_REMOVE(&cpipe->pipe_sel.si_note, kn, knote, kn_selnext);
1318 }
1319
1320 /*ARGSUSED*/
1321 static int
1322 filt_piperead(struct knote *kn, long hint)
1323 {
1324         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1325
1326         kn->kn_data = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
1327
1328         /* XXX RACE */
1329         if (rpipe->pipe_state & PIPE_REOF) {
1330                 kn->kn_flags |= EV_EOF;
1331                 return (1);
1332         }
1333         return (kn->kn_data > 0);
1334 }
1335
1336 /*ARGSUSED*/
1337 static int
1338 filt_pipewrite(struct knote *kn, long hint)
1339 {
1340         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1341         struct pipe *wpipe = rpipe->pipe_peer;
1342         u_int32_t space;
1343
1344         /* XXX RACE */
1345         if ((wpipe == NULL) || (wpipe->pipe_state & PIPE_WEOF)) {
1346                 kn->kn_data = 0;
1347                 kn->kn_flags |= EV_EOF;
1348                 return (1);
1349         }
1350         space = wpipe->pipe_buffer.windex -
1351                 wpipe->pipe_buffer.rindex;
1352         space = wpipe->pipe_buffer.size - space;
1353         kn->kn_data = space;
1354         return (kn->kn_data >= PIPE_BUF);
1355 }