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