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
  65 #include <machine/cpufunc.h>
  66
  67 /*
  68  * interfaces to the outside world
  69  */
  70 static int pipe_read (struct file *fp, struct uio *uio,
  71                 struct ucred *cred, int flags);
  72 static int pipe_write (struct file *fp, struct uio *uio,
  73                 struct ucred *cred, int flags);
  74 static int pipe_close (struct file *fp);
  75 static int pipe_shutdown (struct file *fp, int how);
  76 static int pipe_kqfilter (struct file *fp, struct knote *kn);
  77 static int pipe_stat (struct file *fp, struct stat *sb, struct ucred *cred);
  78 static int pipe_ioctl (struct file *fp, u_long cmd, caddr_t data,
  79                 struct ucred *cred, struct sysmsg *msg);
  80
  81 static struct fileops pipeops = {
  82         .fo_read = pipe_read,
  83         .fo_write = pipe_write,
  84         .fo_ioctl = pipe_ioctl,
  85         .fo_kqfilter = pipe_kqfilter,
  86         .fo_stat = pipe_stat,
  87         .fo_close = pipe_close,
  88         .fo_shutdown = pipe_shutdown
  89 };
  90
  91 static void     filt_pipedetach(struct knote *kn);
  92 static int      filt_piperead(struct knote *kn, long hint);
  93 static int      filt_pipewrite(struct knote *kn, long hint);
  94
  95 static struct filterops pipe_rfiltops =
  96         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, filt_pipedetach, filt_piperead };
  97 static struct filterops pipe_wfiltops =
  98         { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, filt_pipedetach, filt_pipewrite };
  99
 100 MALLOC_DEFINE(M_PIPE, "pipe", "pipe structures");
 101
 102 /*
 103  * Default pipe buffer size(s), this can be kind-of large now because pipe
 104  * space is pageable.  The pipe code will try to maintain locality of
 105  * reference for performance reasons, so small amounts of outstanding I/O
 106  * will not wipe the cache.
 107  */
 108 #define MINPIPESIZE (PIPE_SIZE/3)
 109 #define MAXPIPESIZE (2*PIPE_SIZE/3)
 110
 111 /*
 112  * Limit the number of "big" pipes
 113  */
 114 #define LIMITBIGPIPES   64
 115 #define PIPEQ_MAX_CACHE 16      /* per-cpu pipe structure cache */
 116
 117 static int pipe_maxbig = LIMITBIGPIPES;
 118 static int pipe_maxcache = PIPEQ_MAX_CACHE;
 119 static int pipe_bigcount;
 120 static int pipe_nbig;
 121 static int pipe_bcache_alloc;
 122 static int pipe_bkmem_alloc;
 123 static int pipe_rblocked_count;
 124 static int pipe_wblocked_count;
 125
 126 SYSCTL_NODE(_kern, OID_AUTO, pipe, CTLFLAG_RW, 0, "Pipe operation");
 127 SYSCTL_INT(_kern_pipe, OID_AUTO, nbig,
 128         CTLFLAG_RD, &pipe_nbig, 0, "numer of big pipes allocated");
 129 SYSCTL_INT(_kern_pipe, OID_AUTO, bigcount,
 130         CTLFLAG_RW, &pipe_bigcount, 0, "number of times pipe expanded");
 131 SYSCTL_INT(_kern_pipe, OID_AUTO, rblocked,
 132         CTLFLAG_RW, &pipe_rblocked_count, 0, "number of times pipe expanded");
 133 SYSCTL_INT(_kern_pipe, OID_AUTO, wblocked,
 134         CTLFLAG_RW, &pipe_wblocked_count, 0, "number of times pipe expanded");
 135 SYSCTL_INT(_kern_pipe, OID_AUTO, maxcache,
 136         CTLFLAG_RW, &pipe_maxcache, 0, "max pipes cached per-cpu");
 137 SYSCTL_INT(_kern_pipe, OID_AUTO, maxbig,
 138         CTLFLAG_RW, &pipe_maxbig, 0, "max number of big pipes");
 139 #ifdef SMP
 140 static int pipe_delay = 5000;   /* 5uS default */
 141 SYSCTL_INT(_kern_pipe, OID_AUTO, delay,
 142         CTLFLAG_RW, &pipe_delay, 0, "SMP delay optimization in ns");
 143 #endif
 144 #if !defined(NO_PIPE_SYSCTL_STATS)
 145 SYSCTL_INT(_kern_pipe, OID_AUTO, bcache_alloc,
 146         CTLFLAG_RW, &pipe_bcache_alloc, 0, "pipe buffer from pcpu cache");
 147 SYSCTL_INT(_kern_pipe, OID_AUTO, bkmem_alloc,
 148         CTLFLAG_RW, &pipe_bkmem_alloc, 0, "pipe buffer from kmem");
 149 #endif
 150
 151 /*
 152  * Auto-size pipe cache to reduce kmem allocations and frees.
 153  */
 154 static
 155 void
 156 pipeinit(void *dummy)
 157 {
 158         size_t mbytes = kmem_lim_size();
 159
 160         if (pipe_maxbig == LIMITBIGPIPES) {
 161                 if (mbytes >= 7 * 1024)
 162                         pipe_maxbig *= 2;
 163                 if (mbytes >= 15 * 1024)
 164                         pipe_maxbig *= 2;
 165         }
 166         if (pipe_maxcache == PIPEQ_MAX_CACHE) {
 167                 if (mbytes >= 7 * 1024)
 168                         pipe_maxcache *= 2;
 169                 if (mbytes >= 15 * 1024)
 170                         pipe_maxcache *= 2;
 171         }
 172 }
 173 SYSINIT(kmem, SI_BOOT2_MACHDEP, SI_ORDER_ANY, pipeinit, NULL)
 174
 175 static void pipeclose (struct pipe *cpipe);
 176 static void pipe_free_kmem (struct pipe *cpipe);
 177 static int pipe_create (struct pipe **cpipep);
 178 static int pipespace (struct pipe *cpipe, int size);
 179
 180 static __inline void
 181 pipewakeup(struct pipe *cpipe, int dosigio)
 182 {
 183         if (dosigio && (cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) {
 184                 lwkt_gettoken(&proc_token);
 185                 pgsigio(cpipe->pipe_sigio, SIGIO, 0);
 186                 lwkt_reltoken(&proc_token);
 187         }
 188         KNOTE(&cpipe->pipe_kq.ki_note, 0);
 189 }
 190
 191 /*
 192  * These routines are called before and after a UIO.  The UIO
 193  * may block, causing our held tokens to be lost temporarily.
 194  *
 195  * We use these routines to serialize reads against other reads
 196  * and writes against other writes.
 197  *
 198  * The read token is held on entry so *ipp does not race.
 199  */
 200 static __inline int
 201 pipe_start_uio(struct pipe *cpipe, int *ipp)
 202 {
 203         int error;
 204
 205         while (*ipp) {
 206                 *ipp = -1;
 207                 error = tsleep(ipp, PCATCH, "pipexx", 0);
 208                 if (error)
 209                         return (error);
 210         }
 211         *ipp = 1;
 212         return (0);
 213 }
 214
 215 static __inline void
 216 pipe_end_uio(struct pipe *cpipe, int *ipp)
 217 {
 218         if (*ipp < 0) {
 219                 *ipp = 0;
 220                 wakeup(ipp);
 221         } else {
 222                 KKASSERT(*ipp > 0);
 223                 *ipp = 0;
 224         }
 225 }
 226
 227 /*
 228  * The pipe system call for the DTYPE_PIPE type of pipes
 229  *
 230  * pipe_args(int dummy)
 231  *
 232  * MPSAFE
 233  */
 234 int
 235 sys_pipe(struct pipe_args *uap)
 236 {
 237         struct thread *td = curthread;
 238         struct filedesc *fdp = td->td_proc->p_fd;
 239         struct file *rf, *wf;
 240         struct pipe *rpipe, *wpipe;
 241         int fd1, fd2, error;
 242
 243         rpipe = wpipe = NULL;
 244         if (pipe_create(&rpipe) || pipe_create(&wpipe)) {
 245                 pipeclose(rpipe);
 246                 pipeclose(wpipe);
 247                 return (ENFILE);
 248         }
 249
 250         error = falloc(td->td_lwp, &rf, &fd1);
 251         if (error) {
 252                 pipeclose(rpipe);
 253                 pipeclose(wpipe);
 254                 return (error);
 255         }
 256         uap->sysmsg_fds[0] = fd1;
 257
 258         /*
 259          * Warning: once we've gotten past allocation of the fd for the
 260          * read-side, we can only drop the read side via fdrop() in order
 261          * to avoid races against processes which manage to dup() the read
 262          * side while we are blocked trying to allocate the write side.
 263          */
 264         rf->f_type = DTYPE_PIPE;
 265         rf->f_flag = FREAD | FWRITE;
 266         rf->f_ops = &pipeops;
 267         rf->f_data = rpipe;
 268         error = falloc(td->td_lwp, &wf, &fd2);
 269         if (error) {
 270                 fsetfd(fdp, NULL, fd1);
 271                 fdrop(rf);
 272                 /* rpipe has been closed by fdrop(). */
 273                 pipeclose(wpipe);
 274                 return (error);
 275         }
 276         wf->f_type = DTYPE_PIPE;
 277         wf->f_flag = FREAD | FWRITE;
 278         wf->f_ops = &pipeops;
 279         wf->f_data = wpipe;
 280         uap->sysmsg_fds[1] = fd2;
 281
 282         rpipe->pipe_slock = kmalloc(sizeof(struct lock),
 283                                     M_PIPE, M_WAITOK|M_ZERO);
 284         wpipe->pipe_slock = rpipe->pipe_slock;
 285         rpipe->pipe_peer = wpipe;
 286         wpipe->pipe_peer = rpipe;
 287         lockinit(rpipe->pipe_slock, "pipecl", 0, 0);
 288
 289         /*
 290          * Once activated the peer relationship remains valid until
 291          * both sides are closed.
 292          */
 293         fsetfd(fdp, rf, fd1);
 294         fsetfd(fdp, wf, fd2);
 295         fdrop(rf);
 296         fdrop(wf);
 297
 298         return (0);
 299 }
 300
 301 /*
 302  * Allocate kva for pipe circular buffer, the space is pageable
 303  * This routine will 'realloc' the size of a pipe safely, if it fails
 304  * it will retain the old buffer.
 305  * If it fails it will return ENOMEM.
 306  */
 307 static int
 308 pipespace(struct pipe *cpipe, int size)
 309 {
 310         struct vm_object *object;
 311         caddr_t buffer;
 312         int npages, error;
 313
 314         npages = round_page(size) / PAGE_SIZE;
 315         object = cpipe->pipe_buffer.object;
 316
 317         /*
 318          * [re]create the object if necessary and reserve space for it
 319          * in the kernel_map.  The object and memory are pageable.  On
 320          * success, free the old resources before assigning the new
 321          * ones.
 322          */
 323         if (object == NULL || object->size != npages) {
 324                 object = vm_object_allocate(OBJT_DEFAULT, npages);
 325                 buffer = (caddr_t)vm_map_min(&kernel_map);
 326
 327                 error = vm_map_find(&kernel_map, object, 0,
 328                                     (vm_offset_t *)&buffer,
 329                                     size, PAGE_SIZE,
 330                                     1, VM_MAPTYPE_NORMAL,
 331                                     VM_PROT_ALL, VM_PROT_ALL,
 332                                     0);
 333
 334                 if (error != KERN_SUCCESS) {
 335                         vm_object_deallocate(object);
 336                         return (ENOMEM);
 337                 }
 338                 pipe_free_kmem(cpipe);
 339                 cpipe->pipe_buffer.object = object;
 340                 cpipe->pipe_buffer.buffer = buffer;
 341                 cpipe->pipe_buffer.size = size;
 342                 ++pipe_bkmem_alloc;
 343         } else {
 344                 ++pipe_bcache_alloc;
 345         }
 346         cpipe->pipe_buffer.rindex = 0;
 347         cpipe->pipe_buffer.windex = 0;
 348         return (0);
 349 }
 350
 351 /*
 352  * Initialize and allocate VM and memory for pipe, pulling the pipe from
 353  * our per-cpu cache if possible.  For now make sure it is sized for the
 354  * smaller PIPE_SIZE default.
 355  */
 356 static int
 357 pipe_create(struct pipe **cpipep)
 358 {
 359         globaldata_t gd = mycpu;
 360         struct pipe *cpipe;
 361         int error;
 362
 363         if ((cpipe = gd->gd_pipeq) != NULL) {
 364                 gd->gd_pipeq = cpipe->pipe_peer;
 365                 --gd->gd_pipeqcount;
 366                 cpipe->pipe_peer = NULL;
 367                 cpipe->pipe_wantwcnt = 0;
 368         } else {
 369                 cpipe = kmalloc(sizeof(struct pipe), M_PIPE, M_WAITOK|M_ZERO);
 370         }
 371         *cpipep = cpipe;
 372         if ((error = pipespace(cpipe, PIPE_SIZE)) != 0)
 373                 return (error);
 374         vfs_timestamp(&cpipe->pipe_ctime);
 375         cpipe->pipe_atime = cpipe->pipe_ctime;
 376         cpipe->pipe_mtime = cpipe->pipe_ctime;
 377         lwkt_token_init(&cpipe->pipe_rlock, "piper");
 378         lwkt_token_init(&cpipe->pipe_wlock, "pipew");
 379         return (0);
 380 }
 381
 382 static int
 383 pipe_read(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 384 {
 385         struct pipe *rpipe;
 386         struct pipe *wpipe;
 387         int error;
 388         size_t nread = 0;
 389         int nbio;
 390         u_int size;     /* total bytes available */
 391         u_int nsize;    /* total bytes to read */
 392         u_int rindex;   /* contiguous bytes available */
 393         int notify_writer;
 394         int bigread;
 395         int bigcount;
 396
 397         atomic_set_int(&curthread->td_mpflags, TDF_MP_BATCH_DEMARC);
 398
 399         if (uio->uio_resid == 0)
 400                 return(0);
 401
 402         /*
 403          * Setup locks, calculate nbio
 404          */
 405         rpipe = (struct pipe *)fp->f_data;
 406         wpipe = rpipe->pipe_peer;
 407         lwkt_gettoken(&rpipe->pipe_rlock);
 408
 409         if (fflags & O_FBLOCKING)
 410                 nbio = 0;
 411         else if (fflags & O_FNONBLOCKING)
 412                 nbio = 1;
 413         else if (fp->f_flag & O_NONBLOCK)
 414                 nbio = 1;
 415         else
 416                 nbio = 0;
 417
 418         /*
 419          * Reads are serialized.  Note however that pipe_buffer.buffer and
 420          * pipe_buffer.size can change out from under us when the number
 421          * of bytes in the buffer are zero due to the write-side doing a
 422          * pipespace().
 423          */
 424         error = pipe_start_uio(rpipe, &rpipe->pipe_rip);
 425         if (error) {
 426                 lwkt_reltoken(&rpipe->pipe_rlock);
 427                 return (error);
 428         }
 429         notify_writer = 0;
 430
 431         bigread = (uio->uio_resid > 10 * 1024 * 1024);
 432         bigcount = 10;
 433
 434         while (uio->uio_resid) {
 435                 /*
 436                  * Don't hog the cpu.
 437                  */
 438                 if (bigread && --bigcount == 0) {
 439                         lwkt_user_yield();
 440                         bigcount = 10;
 441                         if (CURSIG(curthread->td_lwp)) {
 442                                 error = EINTR;
 443                                 break;
 444                         }
 445                 }
 446
 447                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 448                 cpu_lfence();
 449                 if (size) {
 450                         rindex = rpipe->pipe_buffer.rindex &
 451                                  (rpipe->pipe_buffer.size - 1);
 452                         nsize = size;
 453                         if (nsize > rpipe->pipe_buffer.size - rindex)
 454                                 nsize = rpipe->pipe_buffer.size - rindex;
 455                         nsize = szmin(nsize, uio->uio_resid);
 456
 457                         error = uiomove(&rpipe->pipe_buffer.buffer[rindex],
 458                                         nsize, uio);
 459                         if (error)
 460                                 break;
 461                         cpu_mfence();
 462                         rpipe->pipe_buffer.rindex += nsize;
 463                         nread += nsize;
 464
 465                         /*
 466                          * If the FIFO is still over half full just continue
 467                          * and do not try to notify the writer yet.
 468                          */
 469                         if (size - nsize >= (rpipe->pipe_buffer.size >> 1)) {
 470                                 notify_writer = 0;
 471                                 continue;
 472                         }
 473
 474                         /*
 475                          * When the FIFO is less then half full notify any
 476                          * waiting writer.  WANTW can be checked while
 477                          * holding just the rlock.
 478                          */
 479                         notify_writer = 1;
 480                         if ((rpipe->pipe_state & PIPE_WANTW) == 0)
 481                                 continue;
 482                 }
 483
 484                 /*
 485                  * If the "write-side" was blocked we wake it up.  This code
 486                  * is reached either when the buffer is completely emptied
 487                  * or if it becomes more then half-empty.
 488                  *
 489                  * Pipe_state can only be modified if both the rlock and
 490                  * wlock are held.
 491                  */
 492                 if (rpipe->pipe_state & PIPE_WANTW) {
 493                         lwkt_gettoken(&rpipe->pipe_wlock);
 494                         if (rpipe->pipe_state & PIPE_WANTW) {
 495                                 rpipe->pipe_state &= ~PIPE_WANTW;
 496                                 lwkt_reltoken(&rpipe->pipe_wlock);
 497                                 wakeup(rpipe);
 498                         } else {
 499                                 lwkt_reltoken(&rpipe->pipe_wlock);
 500                         }
 501                 }
 502
 503                 /*
 504                  * Pick up our copy loop again if the writer sent data to
 505                  * us while we were messing around.
 506                  *
 507                  * On a SMP box poll up to pipe_delay nanoseconds for new
 508                  * data.  Typically a value of 2000 to 4000 is sufficient
 509                  * to eradicate most IPIs/tsleeps/wakeups when a pipe
 510                  * is used for synchronous communications with small packets,
 511                  * and 8000 or so (8uS) will pipeline large buffer xfers
 512                  * between cpus over a pipe.
 513                  *
 514                  * For synchronous communications a hit means doing a
 515                  * full Awrite-Bread-Bwrite-Aread cycle in less then 2uS,
 516                  * where as miss requiring a tsleep/wakeup sequence
 517                  * will take 7uS or more.
 518                  */
 519                 if (rpipe->pipe_buffer.windex != rpipe->pipe_buffer.rindex)
 520                         continue;
 521
 522 #if defined(SMP) && defined(_RDTSC_SUPPORTED_)
 523                 if (pipe_delay) {
 524                         int64_t tsc_target;
 525                         int good = 0;
 526
 527                         tsc_target = tsc_get_target(pipe_delay);
 528                         while (tsc_test_target(tsc_target) == 0) {
 529                                 if (rpipe->pipe_buffer.windex !=
 530                                     rpipe->pipe_buffer.rindex) {
 531                                         good = 1;
 532                                         break;
 533                                 }
 534                         }
 535                         if (good)
 536                                 continue;
 537                 }
 538 #endif
 539
 540                 /*
 541                  * Detect EOF condition, do not set error.
 542                  */
 543                 if (rpipe->pipe_state & PIPE_REOF)
 544                         break;
 545
 546                 /*
 547                  * Break if some data was read, or if this was a non-blocking
 548                  * read.
 549                  */
 550                 if (nread > 0)
 551                         break;
 552
 553                 if (nbio) {
 554                         error = EAGAIN;
 555                         break;
 556                 }
 557
 558                 /*
 559                  * Last chance, interlock with WANTR.
 560                  */
 561                 lwkt_gettoken(&rpipe->pipe_wlock);
 562                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 563                 if (size) {
 564                         lwkt_reltoken(&rpipe->pipe_wlock);
 565                         continue;
 566                 }
 567
 568                 /*
 569                  * Retest EOF - acquiring a new token can temporarily release
 570                  * tokens already held.
 571                  */
 572                 if (rpipe->pipe_state & PIPE_REOF) {
 573                         lwkt_reltoken(&rpipe->pipe_wlock);
 574                         break;
 575                 }
 576
 577                 /*
 578                  * If there is no more to read in the pipe, reset its
 579                  * pointers to the beginning.  This improves cache hit
 580                  * stats.
 581                  *
 582                  * We need both locks to modify both pointers, and there
 583                  * must also not be a write in progress or the uiomove()
 584                  * in the write might block and temporarily release
 585                  * its wlock, then reacquire and update windex.  We are
 586                  * only serialized against reads, not writes.
 587                  *
 588                  * XXX should we even bother resetting the indices?  It
 589                  *     might actually be more cache efficient not to.
 590                  */
 591                 if (rpipe->pipe_buffer.rindex == rpipe->pipe_buffer.windex &&
 592                     rpipe->pipe_wip == 0) {
 593                         rpipe->pipe_buffer.rindex = 0;
 594                         rpipe->pipe_buffer.windex = 0;
 595                 }
 596
 597                 /*
 598                  * Wait for more data.
 599                  *
 600                  * Pipe_state can only be set if both the rlock and wlock
 601                  * are held.
 602                  */
 603                 rpipe->pipe_state |= PIPE_WANTR;
 604                 tsleep_interlock(rpipe, PCATCH);
 605                 lwkt_reltoken(&rpipe->pipe_wlock);
 606                 error = tsleep(rpipe, PCATCH | PINTERLOCKED, "piperd", 0);
 607                 ++pipe_rblocked_count;
 608                 if (error)
 609                         break;
 610         }
 611         pipe_end_uio(rpipe, &rpipe->pipe_rip);
 612
 613         /*
 614          * Uptime last access time
 615          */
 616         if (error == 0 && nread)
 617                 vfs_timestamp(&rpipe->pipe_atime);
 618
 619         /*
 620          * If we drained the FIFO more then half way then handle
 621          * write blocking hysteresis.
 622          *
 623          * Note that PIPE_WANTW cannot be set by the writer without
 624          * it holding both rlock and wlock, so we can test it
 625          * while holding just rlock.
 626          */
 627         if (notify_writer) {
 628                 /*
 629                  * Synchronous blocking is done on the pipe involved
 630                  */
 631                 if (rpipe->pipe_state & PIPE_WANTW) {
 632                         lwkt_gettoken(&rpipe->pipe_wlock);
 633                         if (rpipe->pipe_state & PIPE_WANTW) {
 634                                 rpipe->pipe_state &= ~PIPE_WANTW;
 635                                 lwkt_reltoken(&rpipe->pipe_wlock);
 636                                 wakeup(rpipe);
 637                         } else {
 638                                 lwkt_reltoken(&rpipe->pipe_wlock);
 639                         }
 640                 }
 641
 642                 /*
 643                  * But we may also have to deal with a kqueue which is
 644                  * stored on the same pipe as its descriptor, so a
 645                  * EVFILT_WRITE event waiting for our side to drain will
 646                  * be on the other side.
 647                  */
 648                 lwkt_gettoken(&wpipe->pipe_wlock);
 649                 pipewakeup(wpipe, 0);
 650                 lwkt_reltoken(&wpipe->pipe_wlock);
 651         }
 652         /*size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;*/
 653         lwkt_reltoken(&rpipe->pipe_rlock);
 654
 655         return (error);
 656 }
 657
 658 static int
 659 pipe_write(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 660 {
 661         int error;
 662         int orig_resid;
 663         int nbio;
 664         struct pipe *wpipe;
 665         struct pipe *rpipe;
 666         u_int windex;
 667         u_int space;
 668         u_int wcount;
 669         int bigwrite;
 670         int bigcount;
 671
 672         /*
 673          * Writes go to the peer.  The peer will always exist.
 674          */
 675         rpipe = (struct pipe *) fp->f_data;
 676         wpipe = rpipe->pipe_peer;
 677         lwkt_gettoken(&wpipe->pipe_wlock);
 678         if (wpipe->pipe_state & PIPE_WEOF) {
 679                 lwkt_reltoken(&wpipe->pipe_wlock);
 680                 return (EPIPE);
 681         }
 682
 683         /*
 684          * Degenerate case (EPIPE takes prec)
 685          */
 686         if (uio->uio_resid == 0) {
 687                 lwkt_reltoken(&wpipe->pipe_wlock);
 688                 return(0);
 689         }
 690
 691         /*
 692          * Writes are serialized (start_uio must be called with wlock)
 693          */
 694         error = pipe_start_uio(wpipe, &wpipe->pipe_wip);
 695         if (error) {
 696                 lwkt_reltoken(&wpipe->pipe_wlock);
 697                 return (error);
 698         }
 699
 700         if (fflags & O_FBLOCKING)
 701                 nbio = 0;
 702         else if (fflags & O_FNONBLOCKING)
 703                 nbio = 1;
 704         else if (fp->f_flag & O_NONBLOCK)
 705                 nbio = 1;
 706         else
 707                 nbio = 0;
 708
 709         /*
 710          * If it is advantageous to resize the pipe buffer, do
 711          * so.  We are write-serialized so we can block safely.
 712          */
 713         if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 714             (pipe_nbig < pipe_maxbig) &&
 715             wpipe->pipe_wantwcnt > 4 &&
 716             (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 717                 /*
 718                  * Recheck after lock.
 719                  */
 720                 lwkt_gettoken(&wpipe->pipe_rlock);
 721                 if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 722                     (pipe_nbig < pipe_maxbig) &&
 723                     (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 724                         atomic_add_int(&pipe_nbig, 1);
 725                         if (pipespace(wpipe, BIG_PIPE_SIZE) == 0)
 726                                 ++pipe_bigcount;
 727                         else
 728                                 atomic_subtract_int(&pipe_nbig, 1);
 729                 }
 730                 lwkt_reltoken(&wpipe->pipe_rlock);
 731         }
 732
 733         orig_resid = uio->uio_resid;
 734         wcount = 0;
 735
 736         bigwrite = (uio->uio_resid > 10 * 1024 * 1024);
 737         bigcount = 10;
 738
 739         while (uio->uio_resid) {
 740                 if (wpipe->pipe_state & PIPE_WEOF) {
 741                         error = EPIPE;
 742                         break;
 743                 }
 744
 745                 /*
 746                  * Don't hog the cpu.
 747                  */
 748                 if (bigwrite && --bigcount == 0) {
 749                         lwkt_user_yield();
 750                         bigcount = 10;
 751                         if (CURSIG(curthread->td_lwp)) {
 752                                 error = EINTR;
 753                                 break;
 754                         }
 755                 }
 756
 757                 windex = wpipe->pipe_buffer.windex &
 758                          (wpipe->pipe_buffer.size - 1);
 759                 space = wpipe->pipe_buffer.size -
 760                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 761                 cpu_lfence();
 762
 763                 /* Writes of size <= PIPE_BUF must be atomic. */
 764                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 765                         space = 0;
 766
 767                 /*
 768                  * Write to fill, read size handles write hysteresis.  Also
 769                  * additional restrictions can cause select-based non-blocking
 770                  * writes to spin.
 771                  */
 772                 if (space > 0) {
 773                         u_int segsize;
 774
 775                         /*
 776                          * Transfer size is minimum of uio transfer
 777                          * and free space in pipe buffer.
 778                          *
 779                          * Limit each uiocopy to no more then PIPE_SIZE
 780                          * so we can keep the gravy train going on a
 781                          * SMP box.  This doubles the performance for
 782                          * write sizes > 16K.  Otherwise large writes
 783                          * wind up doing an inefficient synchronous
 784                          * ping-pong.
 785                          */
 786                         space = szmin(space, uio->uio_resid);
 787                         if (space > PIPE_SIZE)
 788                                 space = PIPE_SIZE;
 789
 790                         /*
 791                          * First segment to transfer is minimum of
 792                          * transfer size and contiguous space in
 793                          * pipe buffer.  If first segment to transfer
 794                          * is less than the transfer size, we've got
 795                          * a wraparound in the buffer.
 796                          */
 797                         segsize = wpipe->pipe_buffer.size - windex;
 798                         if (segsize > space)
 799                                 segsize = space;
 800
 801 #ifdef SMP
 802                         /*
 803                          * If this is the first loop and the reader is
 804                          * blocked, do a preemptive wakeup of the reader.
 805                          *
 806                          * On SMP the IPI latency plus the wlock interlock
 807                          * on the reader side is the fastest way to get the
 808                          * reader going.  (The scheduler will hard loop on
 809                          * lock tokens).
 810                          *
 811                          * NOTE: We can't clear WANTR here without acquiring
 812                          * the rlock, which we don't want to do here!
 813                          */
 814                         if ((wpipe->pipe_state & PIPE_WANTR))
 815                                 wakeup(wpipe);
 816 #endif
 817
 818                         /*
 819                          * Transfer segment, which may include a wrap-around.
 820                          * Update windex to account for both all in one go
 821                          * so the reader can read() the data atomically.
 822                          */
 823                         error = uiomove(&wpipe->pipe_buffer.buffer[windex],
 824                                         segsize, uio);
 825                         if (error == 0 && segsize < space) {
 826                                 segsize = space - segsize;
 827                                 error = uiomove(&wpipe->pipe_buffer.buffer[0],
 828                                                 segsize, uio);
 829                         }
 830                         if (error)
 831                                 break;
 832                         cpu_mfence();
 833                         wpipe->pipe_buffer.windex += space;
 834                         wcount += space;
 835                         continue;
 836                 }
 837
 838                 /*
 839                  * We need both the rlock and the wlock to interlock against
 840                  * the EOF, WANTW, and size checks, and to modify pipe_state.
 841                  *
 842                  * These are token locks so we do not have to worry about
 843                  * deadlocks.
 844                  */
 845                 lwkt_gettoken(&wpipe->pipe_rlock);
 846
 847                 /*
 848                  * If the "read-side" has been blocked, wake it up now
 849                  * and yield to let it drain synchronously rather
 850                  * then block.
 851                  */
 852                 if (wpipe->pipe_state & PIPE_WANTR) {
 853                         wpipe->pipe_state &= ~PIPE_WANTR;
 854                         wakeup(wpipe);
 855                 }
 856
 857                 /*
 858                  * don't block on non-blocking I/O
 859                  */
 860                 if (nbio) {
 861                         lwkt_reltoken(&wpipe->pipe_rlock);
 862                         error = EAGAIN;
 863                         break;
 864                 }
 865
 866                 /*
 867                  * re-test whether we have to block in the writer after
 868                  * acquiring both locks, in case the reader opened up
 869                  * some space.
 870                  */
 871                 space = wpipe->pipe_buffer.size -
 872                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 873                 cpu_lfence();
 874                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 875                         space = 0;
 876
 877                 /*
 878                  * Retest EOF - acquiring a new token can temporarily release
 879                  * tokens already held.
 880                  */
 881                 if (wpipe->pipe_state & PIPE_WEOF) {
 882                         lwkt_reltoken(&wpipe->pipe_rlock);
 883                         error = EPIPE;
 884                         break;
 885                 }
 886
 887                 /*
 888                  * We have no more space and have something to offer,
 889                  * wake up select/poll/kq.
 890                  */
 891                 if (space == 0) {
 892                         wpipe->pipe_state |= PIPE_WANTW;
 893                         ++wpipe->pipe_wantwcnt;
 894                         pipewakeup(wpipe, 1);
 895                         if (wpipe->pipe_state & PIPE_WANTW)
 896                                 error = tsleep(wpipe, PCATCH, "pipewr", 0);
 897                         ++pipe_wblocked_count;
 898                 }
 899                 lwkt_reltoken(&wpipe->pipe_rlock);
 900
 901                 /*
 902                  * Break out if we errored or the read side wants us to go
 903                  * away.
 904                  */
 905                 if (error)
 906                         break;
 907                 if (wpipe->pipe_state & PIPE_WEOF) {
 908                         error = EPIPE;
 909                         break;
 910                 }
 911         }
 912         pipe_end_uio(wpipe, &wpipe->pipe_wip);
 913
 914         /*
 915          * If we have put any characters in the buffer, we wake up
 916          * the reader.
 917          *
 918          * Both rlock and wlock are required to be able to modify pipe_state.
 919          */
 920         if (wpipe->pipe_buffer.windex != wpipe->pipe_buffer.rindex) {
 921                 if (wpipe->pipe_state & PIPE_WANTR) {
 922                         lwkt_gettoken(&wpipe->pipe_rlock);
 923                         if (wpipe->pipe_state & PIPE_WANTR) {
 924                                 wpipe->pipe_state &= ~PIPE_WANTR;
 925                                 lwkt_reltoken(&wpipe->pipe_rlock);
 926                                 wakeup(wpipe);
 927                         } else {
 928                                 lwkt_reltoken(&wpipe->pipe_rlock);
 929                         }
 930                 }
 931                 lwkt_gettoken(&wpipe->pipe_rlock);
 932                 pipewakeup(wpipe, 1);
 933                 lwkt_reltoken(&wpipe->pipe_rlock);
 934         }
 935
 936         /*
 937          * Don't return EPIPE if I/O was successful
 938          */
 939         if ((wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex) &&
 940             (uio->uio_resid == 0) &&
 941             (error == EPIPE)) {
 942                 error = 0;
 943         }
 944
 945         if (error == 0)
 946                 vfs_timestamp(&wpipe->pipe_mtime);
 947
 948         /*
 949          * We have something to offer,
 950          * wake up select/poll/kq.
 951          */
 952         /*space = wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex;*/
 953         lwkt_reltoken(&wpipe->pipe_wlock);
 954         return (error);
 955 }
 956
 957 /*
 958  * we implement a very minimal set of ioctls for compatibility with sockets.
 959  */
 960 int
 961 pipe_ioctl(struct file *fp, u_long cmd, caddr_t data,
 962            struct ucred *cred, struct sysmsg *msg)
 963 {
 964         struct pipe *mpipe;
 965         int error;
 966
 967         mpipe = (struct pipe *)fp->f_data;
 968
 969         lwkt_gettoken(&mpipe->pipe_rlock);
 970         lwkt_gettoken(&mpipe->pipe_wlock);
 971
 972         switch (cmd) {
 973         case FIOASYNC:
 974                 if (*(int *)data) {
 975                         mpipe->pipe_state |= PIPE_ASYNC;
 976                 } else {
 977                         mpipe->pipe_state &= ~PIPE_ASYNC;
 978                 }
 979                 error = 0;
 980                 break;
 981         case FIONREAD:
 982                 *(int *)data = mpipe->pipe_buffer.windex -
 983                                 mpipe->pipe_buffer.rindex;
 984                 error = 0;
 985                 break;
 986         case FIOSETOWN:
 987                 error = fsetown(*(int *)data, &mpipe->pipe_sigio);
 988                 break;
 989         case FIOGETOWN:
 990                 *(int *)data = fgetown(&mpipe->pipe_sigio);
 991                 error = 0;
 992                 break;
 993         case TIOCSPGRP:
 994                 /* This is deprecated, FIOSETOWN should be used instead. */
 995                 error = fsetown(-(*(int *)data), &mpipe->pipe_sigio);
 996                 break;
 997
 998         case TIOCGPGRP:
 999                 /* This is deprecated, FIOGETOWN should be used instead. */
1000                 *(int *)data = -fgetown(&mpipe->pipe_sigio);
1001                 error = 0;
1002                 break;
1003         default:
1004                 error = ENOTTY;
1005                 break;
1006         }
1007         lwkt_reltoken(&mpipe->pipe_wlock);
1008         lwkt_reltoken(&mpipe->pipe_rlock);
1009
1010         return (error);
1011 }
1012
1013 /*
1014  * MPSAFE
1015  */
1016 static int
1017 pipe_stat(struct file *fp, struct stat *ub, struct ucred *cred)
1018 {
1019         struct pipe *pipe;
1020
1021         pipe = (struct pipe *)fp->f_data;
1022
1023         bzero((caddr_t)ub, sizeof(*ub));
1024         ub->st_mode = S_IFIFO;
1025         ub->st_blksize = pipe->pipe_buffer.size;
1026         ub->st_size = pipe->pipe_buffer.windex - pipe->pipe_buffer.rindex;
1027         ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize;
1028         ub->st_atimespec = pipe->pipe_atime;
1029         ub->st_mtimespec = pipe->pipe_mtime;
1030         ub->st_ctimespec = pipe->pipe_ctime;
1031         /*
1032          * Left as 0: st_dev, st_ino, st_nlink, st_uid, st_gid, st_rdev,
1033          * st_flags, st_gen.
1034          * XXX (st_dev, st_ino) should be unique.
1035          */
1036         return (0);
1037 }
1038
1039 static int
1040 pipe_close(struct file *fp)
1041 {
1042         struct pipe *cpipe;
1043
1044         cpipe = (struct pipe *)fp->f_data;
1045         fp->f_ops = &badfileops;
1046         fp->f_data = NULL;
1047         funsetown(&cpipe->pipe_sigio);
1048         pipeclose(cpipe);
1049         return (0);
1050 }
1051
1052 /*
1053  * Shutdown one or both directions of a full-duplex pipe.
1054  */
1055 static int
1056 pipe_shutdown(struct file *fp, int how)
1057 {
1058         struct pipe *rpipe;
1059         struct pipe *wpipe;
1060         int error = EPIPE;
1061
1062         rpipe = (struct pipe *)fp->f_data;
1063         wpipe = rpipe->pipe_peer;
1064
1065         /*
1066          * We modify pipe_state on both pipes, which means we need
1067          * all four tokens!
1068          */
1069         lwkt_gettoken(&rpipe->pipe_rlock);
1070         lwkt_gettoken(&rpipe->pipe_wlock);
1071         lwkt_gettoken(&wpipe->pipe_rlock);
1072         lwkt_gettoken(&wpipe->pipe_wlock);
1073
1074         switch(how) {
1075         case SHUT_RDWR:
1076         case SHUT_RD:
1077                 rpipe->pipe_state |= PIPE_REOF;         /* my reads */
1078                 rpipe->pipe_state |= PIPE_WEOF;         /* peer writes */
1079                 if (rpipe->pipe_state & PIPE_WANTR) {
1080                         rpipe->pipe_state &= ~PIPE_WANTR;
1081                         wakeup(rpipe);
1082                 }
1083                 if (rpipe->pipe_state & PIPE_WANTW) {
1084                         rpipe->pipe_state &= ~PIPE_WANTW;
1085                         wakeup(rpipe);
1086                 }
1087                 error = 0;
1088                 if (how == SHUT_RD)
1089                         break;
1090                 /* fall through */
1091         case SHUT_WR:
1092                 wpipe->pipe_state |= PIPE_REOF;         /* peer reads */
1093                 wpipe->pipe_state |= PIPE_WEOF;         /* my writes */
1094                 if (wpipe->pipe_state & PIPE_WANTR) {
1095                         wpipe->pipe_state &= ~PIPE_WANTR;
1096                         wakeup(wpipe);
1097                 }
1098                 if (wpipe->pipe_state & PIPE_WANTW) {
1099                         wpipe->pipe_state &= ~PIPE_WANTW;
1100                         wakeup(wpipe);
1101                 }
1102                 error = 0;
1103                 break;
1104         }
1105         pipewakeup(rpipe, 1);
1106         pipewakeup(wpipe, 1);
1107
1108         lwkt_reltoken(&wpipe->pipe_wlock);
1109         lwkt_reltoken(&wpipe->pipe_rlock);
1110         lwkt_reltoken(&rpipe->pipe_wlock);
1111         lwkt_reltoken(&rpipe->pipe_rlock);
1112
1113         return (error);
1114 }
1115
1116 static void
1117 pipe_free_kmem(struct pipe *cpipe)
1118 {
1119         if (cpipe->pipe_buffer.buffer != NULL) {
1120                 if (cpipe->pipe_buffer.size > PIPE_SIZE)
1121                         atomic_subtract_int(&pipe_nbig, 1);
1122                 kmem_free(&kernel_map,
1123                         (vm_offset_t)cpipe->pipe_buffer.buffer,
1124                         cpipe->pipe_buffer.size);
1125                 cpipe->pipe_buffer.buffer = NULL;
1126                 cpipe->pipe_buffer.object = NULL;
1127         }
1128 }
1129
1130 /*
1131  * Close the pipe.  The slock must be held to interlock against simultanious
1132  * closes.  The rlock and wlock must be held to adjust the pipe_state.
1133  */
1134 static void
1135 pipeclose(struct pipe *cpipe)
1136 {
1137         globaldata_t gd;
1138         struct pipe *ppipe;
1139
1140         if (cpipe == NULL)
1141                 return;
1142
1143         /*
1144          * The slock may not have been allocated yet (close during
1145          * initialization)
1146          *
1147          * We need both the read and write tokens to modify pipe_state.
1148          */
1149         if (cpipe->pipe_slock)
1150                 lockmgr(cpipe->pipe_slock, LK_EXCLUSIVE);
1151         lwkt_gettoken(&cpipe->pipe_rlock);
1152         lwkt_gettoken(&cpipe->pipe_wlock);
1153
1154         /*
1155          * Set our state, wakeup anyone waiting in select/poll/kq, and
1156          * wakeup anyone blocked on our pipe.
1157          */
1158         cpipe->pipe_state |= PIPE_CLOSED | PIPE_REOF | PIPE_WEOF;
1159         pipewakeup(cpipe, 1);
1160         if (cpipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1161                 cpipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1162                 wakeup(cpipe);
1163         }
1164
1165         /*
1166          * Disconnect from peer.
1167          */
1168         if ((ppipe = cpipe->pipe_peer) != NULL) {
1169                 lwkt_gettoken(&ppipe->pipe_rlock);
1170                 lwkt_gettoken(&ppipe->pipe_wlock);
1171                 ppipe->pipe_state |= PIPE_REOF | PIPE_WEOF;
1172                 pipewakeup(ppipe, 1);
1173                 if (ppipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1174                         ppipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1175                         wakeup(ppipe);
1176                 }
1177                 if (SLIST_FIRST(&ppipe->pipe_kq.ki_note))
1178                         KNOTE(&ppipe->pipe_kq.ki_note, 0);
1179                 lwkt_reltoken(&ppipe->pipe_wlock);
1180                 lwkt_reltoken(&ppipe->pipe_rlock);
1181         }
1182
1183         /*
1184          * If the peer is also closed we can free resources for both
1185          * sides, otherwise we leave our side intact to deal with any
1186          * races (since we only have the slock).
1187          */
1188         if (ppipe && (ppipe->pipe_state & PIPE_CLOSED)) {
1189                 cpipe->pipe_peer = NULL;
1190                 ppipe->pipe_peer = NULL;
1191                 ppipe->pipe_slock = NULL;       /* we will free the slock */
1192                 pipeclose(ppipe);
1193                 ppipe = NULL;
1194         }
1195
1196         lwkt_reltoken(&cpipe->pipe_wlock);
1197         lwkt_reltoken(&cpipe->pipe_rlock);
1198         if (cpipe->pipe_slock)
1199                 lockmgr(cpipe->pipe_slock, LK_RELEASE);
1200
1201         /*
1202          * If we disassociated from our peer we can free resources
1203          */
1204         if (ppipe == NULL) {
1205                 gd = mycpu;
1206                 if (cpipe->pipe_slock) {
1207                         kfree(cpipe->pipe_slock, M_PIPE);
1208                         cpipe->pipe_slock = NULL;
1209                 }
1210                 if (gd->gd_pipeqcount >= pipe_maxcache ||
1211                     cpipe->pipe_buffer.size != PIPE_SIZE
1212                 ) {
1213                         pipe_free_kmem(cpipe);
1214                         kfree(cpipe, M_PIPE);
1215                 } else {
1216                         cpipe->pipe_state = 0;
1217                         cpipe->pipe_peer = gd->gd_pipeq;
1218                         gd->gd_pipeq = cpipe;
1219                         ++gd->gd_pipeqcount;
1220                 }
1221         }
1222 }
1223
1224 static int
1225 pipe_kqfilter(struct file *fp, struct knote *kn)
1226 {
1227         struct pipe *cpipe;
1228
1229         cpipe = (struct pipe *)kn->kn_fp->f_data;
1230
1231         switch (kn->kn_filter) {
1232         case EVFILT_READ:
1233                 kn->kn_fop = &pipe_rfiltops;
1234                 break;
1235         case EVFILT_WRITE:
1236                 kn->kn_fop = &pipe_wfiltops;
1237                 if (cpipe->pipe_peer == NULL) {
1238                         /* other end of pipe has been closed */
1239                         return (EPIPE);
1240                 }
1241                 break;
1242         default:
1243                 return (EOPNOTSUPP);
1244         }
1245         kn->kn_hook = (caddr_t)cpipe;
1246
1247         knote_insert(&cpipe->pipe_kq.ki_note, kn);
1248
1249         return (0);
1250 }
1251
1252 static void
1253 filt_pipedetach(struct knote *kn)
1254 {
1255         struct pipe *cpipe = (struct pipe *)kn->kn_hook;
1256
1257         knote_remove(&cpipe->pipe_kq.ki_note, kn);
1258 }
1259
1260 /*ARGSUSED*/
1261 static int
1262 filt_piperead(struct knote *kn, long hint)
1263 {
1264         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1265         int ready = 0;
1266
1267         lwkt_gettoken(&rpipe->pipe_rlock);
1268         lwkt_gettoken(&rpipe->pipe_wlock);
1269
1270         kn->kn_data = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
1271
1272         if (rpipe->pipe_state & PIPE_REOF) {
1273                 /*
1274                  * Only set NODATA if all data has been exhausted
1275                  */
1276                 if (kn->kn_data == 0)
1277                         kn->kn_flags |= EV_NODATA;
1278                 kn->kn_flags |= EV_EOF;
1279                 ready = 1;
1280         }
1281
1282         lwkt_reltoken(&rpipe->pipe_wlock);
1283         lwkt_reltoken(&rpipe->pipe_rlock);
1284
1285         if (!ready)
1286                 ready = kn->kn_data > 0;
1287
1288         return (ready);
1289 }
1290
1291 /*ARGSUSED*/
1292 static int
1293 filt_pipewrite(struct knote *kn, long hint)
1294 {
1295         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1296         struct pipe *wpipe = rpipe->pipe_peer;
1297         int ready = 0;
1298
1299         kn->kn_data = 0;
1300         if (wpipe == NULL) {
1301                 kn->kn_flags |= (EV_EOF | EV_NODATA);
1302                 return (1);
1303         }
1304
1305         lwkt_gettoken(&wpipe->pipe_rlock);
1306         lwkt_gettoken(&wpipe->pipe_wlock);
1307
1308         if (wpipe->pipe_state & PIPE_WEOF) {
1309                 kn->kn_flags |= (EV_EOF | EV_NODATA);
1310                 ready = 1;
1311         }
1312
1313         if (!ready)
1314                 kn->kn_data = wpipe->pipe_buffer.size -
1315                               (wpipe->pipe_buffer.windex -
1316                                wpipe->pipe_buffer.rindex);
1317
1318         lwkt_reltoken(&wpipe->pipe_wlock);
1319         lwkt_reltoken(&wpipe->pipe_rlock);
1320
1321         if (!ready)
1322                 ready = kn->kn_data >= PIPE_BUF;
1323
1324         return (ready);
1325 }