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 int pipespace (struct pipe *cpipe, int size);
 159
 160 static __inline void
 161 pipewakeup(struct pipe *cpipe, int dosigio)
 162 {
 163         if (dosigio && (cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) {
 164                 get_mplock();
 165                 pgsigio(cpipe->pipe_sigio, SIGIO, 0);
 166                 rel_mplock();
 167         }
 168         KNOTE(&cpipe->pipe_kq.ki_note, 0);
 169 }
 170
 171 /*
 172  * These routines are called before and after a UIO.  The UIO
 173  * may block, causing our held tokens to be lost temporarily.
 174  *
 175  * We use these routines to serialize reads against other reads
 176  * and writes against other writes.
 177  *
 178  * The read token is held on entry so *ipp does not race.
 179  */
 180 static __inline int
 181 pipe_start_uio(struct pipe *cpipe, int *ipp)
 182 {
 183         int error;
 184
 185         while (*ipp) {
 186                 *ipp = -1;
 187                 error = tsleep(ipp, PCATCH, "pipexx", 0);
 188                 if (error)
 189                         return (error);
 190         }
 191         *ipp = 1;
 192         return (0);
 193 }
 194
 195 static __inline void
 196 pipe_end_uio(struct pipe *cpipe, int *ipp)
 197 {
 198         if (*ipp < 0) {
 199                 *ipp = 0;
 200                 wakeup(ipp);
 201         } else {
 202                 KKASSERT(*ipp > 0);
 203                 *ipp = 0;
 204         }
 205 }
 206
 207 static __inline void
 208 pipe_get_mplock(int *save)
 209 {
 210 #ifdef SMP
 211         if (pipe_mpsafe == 0) {
 212                 get_mplock();
 213                 *save = 1;
 214         } else
 215 #endif
 216         {
 217                 *save = 0;
 218         }
 219 }
 220
 221 static __inline void
 222 pipe_rel_mplock(int *save)
 223 {
 224 #ifdef SMP
 225         if (*save)
 226                 rel_mplock();
 227 #endif
 228 }
 229
 230
 231 /*
 232  * The pipe system call for the DTYPE_PIPE type of pipes
 233  *
 234  * pipe_args(int dummy)
 235  *
 236  * MPSAFE
 237  */
 238 int
 239 sys_pipe(struct pipe_args *uap)
 240 {
 241         struct thread *td = curthread;
 242         struct filedesc *fdp = td->td_proc->p_fd;
 243         struct file *rf, *wf;
 244         struct pipe *rpipe, *wpipe;
 245         int fd1, fd2, error;
 246
 247         rpipe = wpipe = NULL;
 248         if (pipe_create(&rpipe) || pipe_create(&wpipe)) {
 249                 pipeclose(rpipe);
 250                 pipeclose(wpipe);
 251                 return (ENFILE);
 252         }
 253
 254         error = falloc(td->td_lwp, &rf, &fd1);
 255         if (error) {
 256                 pipeclose(rpipe);
 257                 pipeclose(wpipe);
 258                 return (error);
 259         }
 260         uap->sysmsg_fds[0] = fd1;
 261
 262         /*
 263          * Warning: once we've gotten past allocation of the fd for the
 264          * read-side, we can only drop the read side via fdrop() in order
 265          * to avoid races against processes which manage to dup() the read
 266          * side while we are blocked trying to allocate the write side.
 267          */
 268         rf->f_type = DTYPE_PIPE;
 269         rf->f_flag = FREAD | FWRITE;
 270         rf->f_ops = &pipeops;
 271         rf->f_data = rpipe;
 272         error = falloc(td->td_lwp, &wf, &fd2);
 273         if (error) {
 274                 fsetfd(fdp, NULL, fd1);
 275                 fdrop(rf);
 276                 /* rpipe has been closed by fdrop(). */
 277                 pipeclose(wpipe);
 278                 return (error);
 279         }
 280         wf->f_type = DTYPE_PIPE;
 281         wf->f_flag = FREAD | FWRITE;
 282         wf->f_ops = &pipeops;
 283         wf->f_data = wpipe;
 284         uap->sysmsg_fds[1] = fd2;
 285
 286         rpipe->pipe_slock = kmalloc(sizeof(struct lock),
 287                                     M_PIPE, M_WAITOK|M_ZERO);
 288         wpipe->pipe_slock = rpipe->pipe_slock;
 289         rpipe->pipe_peer = wpipe;
 290         wpipe->pipe_peer = rpipe;
 291         lockinit(rpipe->pipe_slock, "pipecl", 0, 0);
 292
 293         /*
 294          * Once activated the peer relationship remains valid until
 295          * both sides are closed.
 296          */
 297         fsetfd(fdp, rf, fd1);
 298         fsetfd(fdp, wf, fd2);
 299         fdrop(rf);
 300         fdrop(wf);
 301
 302         return (0);
 303 }
 304
 305 /*
 306  * Allocate kva for pipe circular buffer, the space is pageable
 307  * This routine will 'realloc' the size of a pipe safely, if it fails
 308  * it will retain the old buffer.
 309  * If it fails it will return ENOMEM.
 310  */
 311 static int
 312 pipespace(struct pipe *cpipe, int size)
 313 {
 314         struct vm_object *object;
 315         caddr_t buffer;
 316         int npages, error;
 317
 318         npages = round_page(size) / PAGE_SIZE;
 319         object = cpipe->pipe_buffer.object;
 320
 321         /*
 322          * [re]create the object if necessary and reserve space for it
 323          * in the kernel_map.  The object and memory are pageable.  On
 324          * success, free the old resources before assigning the new
 325          * ones.
 326          */
 327         if (object == NULL || object->size != npages) {
 328                 get_mplock();
 329                 object = vm_object_allocate(OBJT_DEFAULT, npages);
 330                 buffer = (caddr_t)vm_map_min(&kernel_map);
 331
 332                 error = vm_map_find(&kernel_map, object, 0,
 333                                     (vm_offset_t *)&buffer,
 334                                     size, PAGE_SIZE,
 335                                     1, VM_MAPTYPE_NORMAL,
 336                                     VM_PROT_ALL, VM_PROT_ALL,
 337                                     0);
 338
 339                 if (error != KERN_SUCCESS) {
 340                         vm_object_deallocate(object);
 341                         rel_mplock();
 342                         return (ENOMEM);
 343                 }
 344                 pipe_free_kmem(cpipe);
 345                 rel_mplock();
 346                 cpipe->pipe_buffer.object = object;
 347                 cpipe->pipe_buffer.buffer = buffer;
 348                 cpipe->pipe_buffer.size = size;
 349                 ++pipe_bkmem_alloc;
 350         } else {
 351                 ++pipe_bcache_alloc;
 352         }
 353         cpipe->pipe_buffer.rindex = 0;
 354         cpipe->pipe_buffer.windex = 0;
 355         return (0);
 356 }
 357
 358 /*
 359  * Initialize and allocate VM and memory for pipe, pulling the pipe from
 360  * our per-cpu cache if possible.  For now make sure it is sized for the
 361  * smaller PIPE_SIZE default.
 362  */
 363 static int
 364 pipe_create(struct pipe **cpipep)
 365 {
 366         globaldata_t gd = mycpu;
 367         struct pipe *cpipe;
 368         int error;
 369
 370         if ((cpipe = gd->gd_pipeq) != NULL) {
 371                 gd->gd_pipeq = cpipe->pipe_peer;
 372                 --gd->gd_pipeqcount;
 373                 cpipe->pipe_peer = NULL;
 374                 cpipe->pipe_wantwcnt = 0;
 375         } else {
 376                 cpipe = kmalloc(sizeof(struct pipe), M_PIPE, M_WAITOK|M_ZERO);
 377         }
 378         *cpipep = cpipe;
 379         if ((error = pipespace(cpipe, PIPE_SIZE)) != 0)
 380                 return (error);
 381         vfs_timestamp(&cpipe->pipe_ctime);
 382         cpipe->pipe_atime = cpipe->pipe_ctime;
 383         cpipe->pipe_mtime = cpipe->pipe_ctime;
 384         lwkt_token_init(&cpipe->pipe_rlock, 1);
 385         lwkt_token_init(&cpipe->pipe_wlock, 1);
 386         return (0);
 387 }
 388
 389 /*
 390  * MPALMOSTSAFE (acquires mplock)
 391  */
 392 static int
 393 pipe_read(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 394 {
 395         struct pipe *rpipe;
 396         struct pipe *wpipe;
 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         wpipe = rpipe->pipe_peer;
 417         lwkt_gettoken(&rpipe->pipe_rlock);
 418
 419         if (fflags & O_FBLOCKING)
 420                 nbio = 0;
 421         else if (fflags & O_FNONBLOCKING)
 422                 nbio = 1;
 423         else if (fp->f_flag & O_NONBLOCK)
 424                 nbio = 1;
 425         else
 426                 nbio = 0;
 427
 428         /*
 429          * Reads are serialized.  Note however that pipe_buffer.buffer and
 430          * pipe_buffer.size can change out from under us when the number
 431          * of bytes in the buffer are zero due to the write-side doing a
 432          * pipespace().
 433          */
 434         error = pipe_start_uio(rpipe, &rpipe->pipe_rip);
 435         if (error) {
 436                 pipe_rel_mplock(&mpsave);
 437                 lwkt_reltoken(&rpipe->pipe_rlock);
 438                 return (error);
 439         }
 440         notify_writer = 0;
 441
 442         bigread = (uio->uio_resid > 10 * 1024 * 1024);
 443         bigcount = 10;
 444
 445         while (uio->uio_resid) {
 446                 /*
 447                  * Don't hog the cpu.
 448                  */
 449                 if (bigread && --bigcount == 0) {
 450                         lwkt_user_yield();
 451                         bigcount = 10;
 452                         if (CURSIG(curthread->td_lwp)) {
 453                                 error = EINTR;
 454                                 break;
 455                         }
 456                 }
 457
 458                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 459                 cpu_lfence();
 460                 if (size) {
 461                         rindex = rpipe->pipe_buffer.rindex &
 462                                  (rpipe->pipe_buffer.size - 1);
 463                         nsize = size;
 464                         if (nsize > rpipe->pipe_buffer.size - rindex)
 465                                 nsize = rpipe->pipe_buffer.size - rindex;
 466                         nsize = szmin(nsize, uio->uio_resid);
 467
 468                         error = uiomove(&rpipe->pipe_buffer.buffer[rindex],
 469                                         nsize, uio);
 470                         if (error)
 471                                 break;
 472                         cpu_mfence();
 473                         rpipe->pipe_buffer.rindex += nsize;
 474                         nread += nsize;
 475
 476                         /*
 477                          * If the FIFO is still over half full just continue
 478                          * and do not try to notify the writer yet.
 479                          */
 480                         if (size - nsize >= (rpipe->pipe_buffer.size >> 1)) {
 481                                 notify_writer = 0;
 482                                 continue;
 483                         }
 484
 485                         /*
 486                          * When the FIFO is less then half full notify any
 487                          * waiting writer.  WANTW can be checked while
 488                          * holding just the rlock.
 489                          */
 490                         notify_writer = 1;
 491                         if ((rpipe->pipe_state & PIPE_WANTW) == 0)
 492                                 continue;
 493                 }
 494
 495                 /*
 496                  * If the "write-side" was blocked we wake it up.  This code
 497                  * is reached either when the buffer is completely emptied
 498                  * or if it becomes more then half-empty.
 499                  *
 500                  * Pipe_state can only be modified if both the rlock and
 501                  * wlock are held.
 502                  */
 503                 if (rpipe->pipe_state & PIPE_WANTW) {
 504                         lwkt_gettoken(&rpipe->pipe_wlock);
 505                         if (rpipe->pipe_state & PIPE_WANTW) {
 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                 /*
 640                  * Synchronous blocking is done on the pipe involved
 641                  */
 642                 if (rpipe->pipe_state & PIPE_WANTW) {
 643                         lwkt_gettoken(&rpipe->pipe_wlock);
 644                         if (rpipe->pipe_state & PIPE_WANTW) {
 645                                 rpipe->pipe_state &= ~PIPE_WANTW;
 646                                 lwkt_reltoken(&rpipe->pipe_wlock);
 647                                 wakeup(rpipe);
 648                         } else {
 649                                 lwkt_reltoken(&rpipe->pipe_wlock);
 650                         }
 651                 }
 652
 653                 /*
 654                  * But we may also have to deal with a kqueue which is
 655                  * stored on the same pipe as its descriptor, so a
 656                  * EVFILT_WRITE event waiting for our side to drain will
 657                  * be on the other side.
 658                  */
 659                 lwkt_gettoken(&wpipe->pipe_wlock);
 660                 pipewakeup(wpipe, 0);
 661                 lwkt_reltoken(&wpipe->pipe_wlock);
 662         }
 663         /*size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;*/
 664         lwkt_reltoken(&rpipe->pipe_rlock);
 665
 666         pipe_rel_mplock(&mpsave);
 667         return (error);
 668 }
 669
 670 /*
 671  * MPALMOSTSAFE - acquires mplock
 672  */
 673 static int
 674 pipe_write(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 675 {
 676         int error;
 677         int orig_resid;
 678         int nbio;
 679         struct pipe *wpipe;
 680         struct pipe *rpipe;
 681         u_int windex;
 682         u_int space;
 683         u_int wcount;
 684         int mpsave;
 685         int bigwrite;
 686         int bigcount;
 687
 688         pipe_get_mplock(&mpsave);
 689
 690         /*
 691          * Writes go to the peer.  The peer will always exist.
 692          */
 693         rpipe = (struct pipe *) fp->f_data;
 694         wpipe = rpipe->pipe_peer;
 695         lwkt_gettoken(&wpipe->pipe_wlock);
 696         if (wpipe->pipe_state & PIPE_WEOF) {
 697                 pipe_rel_mplock(&mpsave);
 698                 lwkt_reltoken(&wpipe->pipe_wlock);
 699                 return (EPIPE);
 700         }
 701
 702         /*
 703          * Degenerate case (EPIPE takes prec)
 704          */
 705         if (uio->uio_resid == 0) {
 706                 pipe_rel_mplock(&mpsave);
 707                 lwkt_reltoken(&wpipe->pipe_wlock);
 708                 return(0);
 709         }
 710
 711         /*
 712          * Writes are serialized (start_uio must be called with wlock)
 713          */
 714         error = pipe_start_uio(wpipe, &wpipe->pipe_wip);
 715         if (error) {
 716                 pipe_rel_mplock(&mpsave);
 717                 lwkt_reltoken(&wpipe->pipe_wlock);
 718                 return (error);
 719         }
 720
 721         if (fflags & O_FBLOCKING)
 722                 nbio = 0;
 723         else if (fflags & O_FNONBLOCKING)
 724                 nbio = 1;
 725         else if (fp->f_flag & O_NONBLOCK)
 726                 nbio = 1;
 727         else
 728                 nbio = 0;
 729
 730         /*
 731          * If it is advantageous to resize the pipe buffer, do
 732          * so.  We are write-serialized so we can block safely.
 733          */
 734         if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 735             (pipe_nbig < pipe_maxbig) &&
 736             wpipe->pipe_wantwcnt > 4 &&
 737             (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 738                 /*
 739                  * Recheck after lock.
 740                  */
 741                 lwkt_gettoken(&wpipe->pipe_rlock);
 742                 if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 743                     (pipe_nbig < pipe_maxbig) &&
 744                     (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 745                         atomic_add_int(&pipe_nbig, 1);
 746                         if (pipespace(wpipe, BIG_PIPE_SIZE) == 0)
 747                                 ++pipe_bigcount;
 748                         else
 749                                 atomic_subtract_int(&pipe_nbig, 1);
 750                 }
 751                 lwkt_reltoken(&wpipe->pipe_rlock);
 752         }
 753
 754         orig_resid = uio->uio_resid;
 755         wcount = 0;
 756
 757         bigwrite = (uio->uio_resid > 10 * 1024 * 1024);
 758         bigcount = 10;
 759
 760         while (uio->uio_resid) {
 761                 if (wpipe->pipe_state & PIPE_WEOF) {
 762                         error = EPIPE;
 763                         break;
 764                 }
 765
 766                 /*
 767                  * Don't hog the cpu.
 768                  */
 769                 if (bigwrite && --bigcount == 0) {
 770                         lwkt_user_yield();
 771                         bigcount = 10;
 772                         if (CURSIG(curthread->td_lwp)) {
 773                                 error = EINTR;
 774                                 break;
 775                         }
 776                 }
 777
 778                 windex = wpipe->pipe_buffer.windex &
 779                          (wpipe->pipe_buffer.size - 1);
 780                 space = wpipe->pipe_buffer.size -
 781                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 782                 cpu_lfence();
 783
 784                 /* Writes of size <= PIPE_BUF must be atomic. */
 785                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 786                         space = 0;
 787
 788                 /*
 789                  * Write to fill, read size handles write hysteresis.  Also
 790                  * additional restrictions can cause select-based non-blocking
 791                  * writes to spin.
 792                  */
 793                 if (space > 0) {
 794                         u_int segsize;
 795
 796                         /*
 797                          * Transfer size is minimum of uio transfer
 798                          * and free space in pipe buffer.
 799                          *
 800                          * Limit each uiocopy to no more then PIPE_SIZE
 801                          * so we can keep the gravy train going on a
 802                          * SMP box.  This doubles the performance for
 803                          * write sizes > 16K.  Otherwise large writes
 804                          * wind up doing an inefficient synchronous
 805                          * ping-pong.
 806                          */
 807                         space = szmin(space, uio->uio_resid);
 808                         if (space > PIPE_SIZE)
 809                                 space = PIPE_SIZE;
 810
 811                         /*
 812                          * First segment to transfer is minimum of
 813                          * transfer size and contiguous space in
 814                          * pipe buffer.  If first segment to transfer
 815                          * is less than the transfer size, we've got
 816                          * a wraparound in the buffer.
 817                          */
 818                         segsize = wpipe->pipe_buffer.size - windex;
 819                         if (segsize > space)
 820                                 segsize = space;
 821
 822 #ifdef SMP
 823                         /*
 824                          * If this is the first loop and the reader is
 825                          * blocked, do a preemptive wakeup of the reader.
 826                          *
 827                          * On SMP the IPI latency plus the wlock interlock
 828                          * on the reader side is the fastest way to get the
 829                          * reader going.  (The scheduler will hard loop on
 830                          * lock tokens).
 831                          *
 832                          * NOTE: We can't clear WANTR here without acquiring
 833                          * the rlock, which we don't want to do here!
 834                          */
 835                         if ((wpipe->pipe_state & PIPE_WANTR) && pipe_mpsafe > 1)
 836                                 wakeup(wpipe);
 837 #endif
 838
 839                         /*
 840                          * Transfer segment, which may include a wrap-around.
 841                          * Update windex to account for both all in one go
 842                          * so the reader can read() the data atomically.
 843                          */
 844                         error = uiomove(&wpipe->pipe_buffer.buffer[windex],
 845                                         segsize, uio);
 846                         if (error == 0 && segsize < space) {
 847                                 segsize = space - segsize;
 848                                 error = uiomove(&wpipe->pipe_buffer.buffer[0],
 849                                                 segsize, uio);
 850                         }
 851                         if (error)
 852                                 break;
 853                         cpu_mfence();
 854                         wpipe->pipe_buffer.windex += space;
 855                         wcount += space;
 856                         continue;
 857                 }
 858
 859                 /*
 860                  * We need both the rlock and the wlock to interlock against
 861                  * the EOF, WANTW, and size checks, and to modify pipe_state.
 862                  *
 863                  * These are token locks so we do not have to worry about
 864                  * deadlocks.
 865                  */
 866                 lwkt_gettoken(&wpipe->pipe_rlock);
 867
 868                 /*
 869                  * If the "read-side" has been blocked, wake it up now
 870                  * and yield to let it drain synchronously rather
 871                  * then block.
 872                  */
 873                 if (wpipe->pipe_state & PIPE_WANTR) {
 874                         wpipe->pipe_state &= ~PIPE_WANTR;
 875                         wakeup(wpipe);
 876                 }
 877
 878                 /*
 879                  * don't block on non-blocking I/O
 880                  */
 881                 if (nbio) {
 882                         lwkt_reltoken(&wpipe->pipe_rlock);
 883                         error = EAGAIN;
 884                         break;
 885                 }
 886
 887                 /*
 888                  * re-test whether we have to block in the writer after
 889                  * acquiring both locks, in case the reader opened up
 890                  * some space.
 891                  */
 892                 space = wpipe->pipe_buffer.size -
 893                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 894                 cpu_lfence();
 895                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 896                         space = 0;
 897
 898                 /*
 899                  * Retest EOF - acquiring a new token can temporarily release
 900                  * tokens already held.
 901                  */
 902                 if (wpipe->pipe_state & PIPE_WEOF) {
 903                         lwkt_reltoken(&wpipe->pipe_rlock);
 904                         error = EPIPE;
 905                         break;
 906                 }
 907
 908                 /*
 909                  * We have no more space and have something to offer,
 910                  * wake up select/poll/kq.
 911                  */
 912                 if (space == 0) {
 913                         wpipe->pipe_state |= PIPE_WANTW;
 914                         ++wpipe->pipe_wantwcnt;
 915                         pipewakeup(wpipe, 1);
 916                         if (wpipe->pipe_state & PIPE_WANTW)
 917                                 error = tsleep(wpipe, PCATCH, "pipewr", 0);
 918                         ++pipe_wblocked_count;
 919                 }
 920                 lwkt_reltoken(&wpipe->pipe_rlock);
 921
 922                 /*
 923                  * Break out if we errored or the read side wants us to go
 924                  * away.
 925                  */
 926                 if (error)
 927                         break;
 928                 if (wpipe->pipe_state & PIPE_WEOF) {
 929                         error = EPIPE;
 930                         break;
 931                 }
 932         }
 933         pipe_end_uio(wpipe, &wpipe->pipe_wip);
 934
 935         /*
 936          * If we have put any characters in the buffer, we wake up
 937          * the reader.
 938          *
 939          * Both rlock and wlock are required to be able to modify pipe_state.
 940          */
 941         if (wpipe->pipe_buffer.windex != wpipe->pipe_buffer.rindex) {
 942                 if (wpipe->pipe_state & PIPE_WANTR) {
 943                         lwkt_gettoken(&wpipe->pipe_rlock);
 944                         if (wpipe->pipe_state & PIPE_WANTR) {
 945                                 wpipe->pipe_state &= ~PIPE_WANTR;
 946                                 lwkt_reltoken(&wpipe->pipe_rlock);
 947                                 wakeup(wpipe);
 948                         } else {
 949                                 lwkt_reltoken(&wpipe->pipe_rlock);
 950                         }
 951                 }
 952                 lwkt_gettoken(&wpipe->pipe_rlock);
 953                 pipewakeup(wpipe, 1);
 954                 lwkt_reltoken(&wpipe->pipe_rlock);
 955         }
 956
 957         /*
 958          * Don't return EPIPE if I/O was successful
 959          */
 960         if ((wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex) &&
 961             (uio->uio_resid == 0) &&
 962             (error == EPIPE)) {
 963                 error = 0;
 964         }
 965
 966         if (error == 0)
 967                 vfs_timestamp(&wpipe->pipe_mtime);
 968
 969         /*
 970          * We have something to offer,
 971          * wake up select/poll/kq.
 972          */
 973         /*space = wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex;*/
 974         lwkt_reltoken(&wpipe->pipe_wlock);
 975         pipe_rel_mplock(&mpsave);
 976         return (error);
 977 }
 978
 979 /*
 980  * MPALMOSTSAFE - acquires mplock
 981  *
 982  * we implement a very minimal set of ioctls for compatibility with sockets.
 983  */
 984 int
 985 pipe_ioctl(struct file *fp, u_long cmd, caddr_t data,
 986            struct ucred *cred, struct sysmsg *msg)
 987 {
 988         struct pipe *mpipe;
 989         int error;
 990         int mpsave;
 991
 992         pipe_get_mplock(&mpsave);
 993         mpipe = (struct pipe *)fp->f_data;
 994
 995         lwkt_gettoken(&mpipe->pipe_rlock);
 996         lwkt_gettoken(&mpipe->pipe_wlock);
 997
 998         switch (cmd) {
 999         case FIOASYNC:
1000                 if (*(int *)data) {
1001                         mpipe->pipe_state |= PIPE_ASYNC;
1002                 } else {
1003                         mpipe->pipe_state &= ~PIPE_ASYNC;
1004                 }
1005                 error = 0;
1006                 break;
1007         case FIONREAD:
1008                 *(int *)data = mpipe->pipe_buffer.windex -
1009                                 mpipe->pipe_buffer.rindex;
1010                 error = 0;
1011                 break;
1012         case FIOSETOWN:
1013                 get_mplock();
1014                 error = fsetown(*(int *)data, &mpipe->pipe_sigio);
1015                 rel_mplock();
1016                 break;
1017         case FIOGETOWN:
1018                 *(int *)data = fgetown(mpipe->pipe_sigio);
1019                 error = 0;
1020                 break;
1021         case TIOCSPGRP:
1022                 /* This is deprecated, FIOSETOWN should be used instead. */
1023                 get_mplock();
1024                 error = fsetown(-(*(int *)data), &mpipe->pipe_sigio);
1025                 rel_mplock();
1026                 break;
1027
1028         case TIOCGPGRP:
1029                 /* This is deprecated, FIOGETOWN should be used instead. */
1030                 *(int *)data = -fgetown(mpipe->pipe_sigio);
1031                 error = 0;
1032                 break;
1033         default:
1034                 error = ENOTTY;
1035                 break;
1036         }
1037         lwkt_reltoken(&mpipe->pipe_wlock);
1038         lwkt_reltoken(&mpipe->pipe_rlock);
1039         pipe_rel_mplock(&mpsave);
1040
1041         return (error);
1042 }
1043
1044 /*
1045  * MPSAFE
1046  */
1047 static int
1048 pipe_stat(struct file *fp, struct stat *ub, struct ucred *cred)
1049 {
1050         struct pipe *pipe;
1051         int mpsave;
1052
1053         pipe_get_mplock(&mpsave);
1054         pipe = (struct pipe *)fp->f_data;
1055
1056         bzero((caddr_t)ub, sizeof(*ub));
1057         ub->st_mode = S_IFIFO;
1058         ub->st_blksize = pipe->pipe_buffer.size;
1059         ub->st_size = pipe->pipe_buffer.windex - pipe->pipe_buffer.rindex;
1060         ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize;
1061         ub->st_atimespec = pipe->pipe_atime;
1062         ub->st_mtimespec = pipe->pipe_mtime;
1063         ub->st_ctimespec = pipe->pipe_ctime;
1064         /*
1065          * Left as 0: st_dev, st_ino, st_nlink, st_uid, st_gid, st_rdev,
1066          * st_flags, st_gen.
1067          * XXX (st_dev, st_ino) should be unique.
1068          */
1069         pipe_rel_mplock(&mpsave);
1070         return (0);
1071 }
1072
1073 /*
1074  * MPALMOSTSAFE - acquires mplock
1075  */
1076 static int
1077 pipe_close(struct file *fp)
1078 {
1079         struct pipe *cpipe;
1080
1081         get_mplock();
1082         cpipe = (struct pipe *)fp->f_data;
1083         fp->f_ops = &badfileops;
1084         fp->f_data = NULL;
1085         funsetown(cpipe->pipe_sigio);
1086         pipeclose(cpipe);
1087         rel_mplock();
1088         return (0);
1089 }
1090
1091 /*
1092  * Shutdown one or both directions of a full-duplex pipe.
1093  *
1094  * MPALMOSTSAFE - acquires mplock
1095  */
1096 static int
1097 pipe_shutdown(struct file *fp, int how)
1098 {
1099         struct pipe *rpipe;
1100         struct pipe *wpipe;
1101         int error = EPIPE;
1102         int mpsave;
1103
1104         pipe_get_mplock(&mpsave);
1105         rpipe = (struct pipe *)fp->f_data;
1106         wpipe = rpipe->pipe_peer;
1107
1108         /*
1109          * We modify pipe_state on both pipes, which means we need
1110          * all four tokens!
1111          */
1112         lwkt_gettoken(&rpipe->pipe_rlock);
1113         lwkt_gettoken(&rpipe->pipe_wlock);
1114         lwkt_gettoken(&wpipe->pipe_rlock);
1115         lwkt_gettoken(&wpipe->pipe_wlock);
1116
1117         switch(how) {
1118         case SHUT_RDWR:
1119         case SHUT_RD:
1120                 rpipe->pipe_state |= PIPE_REOF;         /* my reads */
1121                 rpipe->pipe_state |= PIPE_WEOF;         /* peer writes */
1122                 if (rpipe->pipe_state & PIPE_WANTR) {
1123                         rpipe->pipe_state &= ~PIPE_WANTR;
1124                         wakeup(rpipe);
1125                 }
1126                 if (rpipe->pipe_state & PIPE_WANTW) {
1127                         rpipe->pipe_state &= ~PIPE_WANTW;
1128                         wakeup(rpipe);
1129                 }
1130                 error = 0;
1131                 if (how == SHUT_RD)
1132                         break;
1133                 /* fall through */
1134         case SHUT_WR:
1135                 wpipe->pipe_state |= PIPE_REOF;         /* peer reads */
1136                 wpipe->pipe_state |= PIPE_WEOF;         /* my writes */
1137                 if (wpipe->pipe_state & PIPE_WANTR) {
1138                         wpipe->pipe_state &= ~PIPE_WANTR;
1139                         wakeup(wpipe);
1140                 }
1141                 if (wpipe->pipe_state & PIPE_WANTW) {
1142                         wpipe->pipe_state &= ~PIPE_WANTW;
1143                         wakeup(wpipe);
1144                 }
1145                 error = 0;
1146                 break;
1147         }
1148         pipewakeup(rpipe, 1);
1149         pipewakeup(wpipe, 1);
1150
1151         lwkt_reltoken(&wpipe->pipe_wlock);
1152         lwkt_reltoken(&wpipe->pipe_rlock);
1153         lwkt_reltoken(&rpipe->pipe_wlock);
1154         lwkt_reltoken(&rpipe->pipe_rlock);
1155
1156         pipe_rel_mplock(&mpsave);
1157         return (error);
1158 }
1159
1160 static void
1161 pipe_free_kmem(struct pipe *cpipe)
1162 {
1163         if (cpipe->pipe_buffer.buffer != NULL) {
1164                 if (cpipe->pipe_buffer.size > PIPE_SIZE)
1165                         atomic_subtract_int(&pipe_nbig, 1);
1166                 kmem_free(&kernel_map,
1167                         (vm_offset_t)cpipe->pipe_buffer.buffer,
1168                         cpipe->pipe_buffer.size);
1169                 cpipe->pipe_buffer.buffer = NULL;
1170                 cpipe->pipe_buffer.object = NULL;
1171         }
1172 }
1173
1174 /*
1175  * Close the pipe.  The slock must be held to interlock against simultanious
1176  * closes.  The rlock and wlock must be held to adjust the pipe_state.
1177  */
1178 static void
1179 pipeclose(struct pipe *cpipe)
1180 {
1181         globaldata_t gd;
1182         struct pipe *ppipe;
1183
1184         if (cpipe == NULL)
1185                 return;
1186
1187         /*
1188          * The slock may not have been allocated yet (close during
1189          * initialization)
1190          *
1191          * We need both the read and write tokens to modify pipe_state.
1192          */
1193         if (cpipe->pipe_slock)
1194                 lockmgr(cpipe->pipe_slock, LK_EXCLUSIVE);
1195         lwkt_gettoken(&cpipe->pipe_rlock);
1196         lwkt_gettoken(&cpipe->pipe_wlock);
1197
1198         /*
1199          * Set our state, wakeup anyone waiting in select/poll/kq, and
1200          * wakeup anyone blocked on our pipe.
1201          */
1202         cpipe->pipe_state |= PIPE_CLOSED | PIPE_REOF | PIPE_WEOF;
1203         pipewakeup(cpipe, 1);
1204         if (cpipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1205                 cpipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1206                 wakeup(cpipe);
1207         }
1208
1209         /*
1210          * Disconnect from peer.
1211          */
1212         if ((ppipe = cpipe->pipe_peer) != NULL) {
1213                 lwkt_gettoken(&ppipe->pipe_rlock);
1214                 lwkt_gettoken(&ppipe->pipe_wlock);
1215                 ppipe->pipe_state |= PIPE_REOF | PIPE_WEOF;
1216                 pipewakeup(ppipe, 1);
1217                 if (ppipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1218                         ppipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1219                         wakeup(ppipe);
1220                 }
1221                 if (SLIST_FIRST(&ppipe->pipe_kq.ki_note))
1222                         KNOTE(&ppipe->pipe_kq.ki_note, 0);
1223                 lwkt_reltoken(&ppipe->pipe_wlock);
1224                 lwkt_reltoken(&ppipe->pipe_rlock);
1225         }
1226
1227         /*
1228          * If the peer is also closed we can free resources for both
1229          * sides, otherwise we leave our side intact to deal with any
1230          * races (since we only have the slock).
1231          */
1232         if (ppipe && (ppipe->pipe_state & PIPE_CLOSED)) {
1233                 cpipe->pipe_peer = NULL;
1234                 ppipe->pipe_peer = NULL;
1235                 ppipe->pipe_slock = NULL;       /* we will free the slock */
1236                 pipeclose(ppipe);
1237                 ppipe = NULL;
1238         }
1239
1240         lwkt_reltoken(&cpipe->pipe_wlock);
1241         lwkt_reltoken(&cpipe->pipe_rlock);
1242         if (cpipe->pipe_slock)
1243                 lockmgr(cpipe->pipe_slock, LK_RELEASE);
1244
1245         /*
1246          * If we disassociated from our peer we can free resources
1247          */
1248         if (ppipe == NULL) {
1249                 gd = mycpu;
1250                 if (cpipe->pipe_slock) {
1251                         kfree(cpipe->pipe_slock, M_PIPE);
1252                         cpipe->pipe_slock = NULL;
1253                 }
1254                 if (gd->gd_pipeqcount >= pipe_maxcache ||
1255                     cpipe->pipe_buffer.size != PIPE_SIZE
1256                 ) {
1257                         pipe_free_kmem(cpipe);
1258                         kfree(cpipe, M_PIPE);
1259                 } else {
1260                         cpipe->pipe_state = 0;
1261                         cpipe->pipe_peer = gd->gd_pipeq;
1262                         gd->gd_pipeq = cpipe;
1263                         ++gd->gd_pipeqcount;
1264                 }
1265         }
1266 }
1267
1268 /*
1269  * MPALMOSTSAFE - acquires mplock
1270  */
1271 static int
1272 pipe_kqfilter(struct file *fp, struct knote *kn)
1273 {
1274         struct pipe *cpipe;
1275
1276         cpipe = (struct pipe *)kn->kn_fp->f_data;
1277
1278         switch (kn->kn_filter) {
1279         case EVFILT_READ:
1280                 kn->kn_fop = &pipe_rfiltops;
1281                 break;
1282         case EVFILT_WRITE:
1283                 kn->kn_fop = &pipe_wfiltops;
1284                 if (cpipe->pipe_peer == NULL) {
1285                         /* other end of pipe has been closed */
1286                         return (EPIPE);
1287                 }
1288                 break;
1289         default:
1290                 return (EOPNOTSUPP);
1291         }
1292         kn->kn_hook = (caddr_t)cpipe;
1293
1294         knote_insert(&cpipe->pipe_kq.ki_note, kn);
1295
1296         return (0);
1297 }
1298
1299 static void
1300 filt_pipedetach(struct knote *kn)
1301 {
1302         struct pipe *cpipe = (struct pipe *)kn->kn_hook;
1303
1304         knote_remove(&cpipe->pipe_kq.ki_note, kn);
1305 }
1306
1307 /*ARGSUSED*/
1308 static int
1309 filt_piperead(struct knote *kn, long hint)
1310 {
1311         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1312         int ready = 0;
1313
1314         lwkt_gettoken(&rpipe->pipe_rlock);
1315         lwkt_gettoken(&rpipe->pipe_wlock);
1316
1317         kn->kn_data = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
1318
1319         /*
1320          * Only set EOF if all data has been exhausted
1321          */
1322         if ((rpipe->pipe_state & PIPE_REOF) && kn->kn_data == 0) {
1323                 kn->kn_flags |= EV_EOF;
1324                 ready = 1;
1325         }
1326
1327         lwkt_reltoken(&rpipe->pipe_wlock);
1328         lwkt_reltoken(&rpipe->pipe_rlock);
1329
1330         if (!ready)
1331                 ready = kn->kn_data > 0;
1332
1333         return (ready);
1334 }
1335
1336 /*ARGSUSED*/
1337 static int
1338 filt_pipewrite(struct knote *kn, long hint)
1339 {
1340         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1341         struct pipe *wpipe = rpipe->pipe_peer;
1342         int ready = 0;
1343
1344         kn->kn_data = 0;
1345         if (wpipe == NULL) {
1346                 kn->kn_flags |= EV_EOF;
1347                 return (1);
1348         }
1349
1350         lwkt_gettoken(&wpipe->pipe_rlock);
1351         lwkt_gettoken(&wpipe->pipe_wlock);
1352
1353         if (wpipe->pipe_state & PIPE_WEOF) {
1354                 kn->kn_flags |= EV_EOF;
1355                 ready = 1;
1356         }
1357
1358         if (!ready)
1359                 kn->kn_data = wpipe->pipe_buffer.size -
1360                               (wpipe->pipe_buffer.windex -
1361                                wpipe->pipe_buffer.rindex);
1362
1363         lwkt_reltoken(&wpipe->pipe_wlock);
1364         lwkt_reltoken(&wpipe->pipe_rlock);
1365
1366         if (!ready)
1367                 ready = kn->kn_data >= PIPE_BUF;
1368
1369         return (ready);
1370 }