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/poll.h>
  40 #include <sys/select.h>
  41 #include <sys/signalvar.h>
  42 #include <sys/sysproto.h>
  43 #include <sys/pipe.h>
  44 #include <sys/vnode.h>
  45 #include <sys/uio.h>
  46 #include <sys/event.h>
  47 #include <sys/globaldata.h>
  48 #include <sys/module.h>
  49 #include <sys/malloc.h>
  50 #include <sys/sysctl.h>
  51 #include <sys/socket.h>
  52
  53 #include <vm/vm.h>
  54 #include <vm/vm_param.h>
  55 #include <sys/lock.h>
  56 #include <vm/vm_object.h>
  57 #include <vm/vm_kern.h>
  58 #include <vm/vm_extern.h>
  59 #include <vm/pmap.h>
  60 #include <vm/vm_map.h>
  61 #include <vm/vm_page.h>
  62 #include <vm/vm_zone.h>
  63
  64 #include <sys/file2.h>
  65 #include <sys/signal2.h>
  66
  67 #include <machine/cpufunc.h>
  68
  69 /*
  70  * interfaces to the outside world
  71  */
  72 static int pipe_read (struct file *fp, struct uio *uio,
  73                 struct ucred *cred, int flags);
  74 static int pipe_write (struct file *fp, struct uio *uio,
  75                 struct ucred *cred, int flags);
  76 static int pipe_close (struct file *fp);
  77 static int pipe_shutdown (struct file *fp, int how);
  78 static int pipe_poll (struct file *fp, int events, struct ucred *cred);
  79 static int pipe_kqfilter (struct file *fp, struct knote *kn);
  80 static int pipe_stat (struct file *fp, struct stat *sb, struct ucred *cred);
  81 static int pipe_ioctl (struct file *fp, u_long cmd, caddr_t data,
  82                 struct ucred *cred, struct sysmsg *msg);
  83
  84 static struct fileops pipeops = {
  85         .fo_read = pipe_read,
  86         .fo_write = pipe_write,
  87         .fo_ioctl = pipe_ioctl,
  88         .fo_poll = pipe_poll,
  89         .fo_kqfilter = pipe_kqfilter,
  90         .fo_stat = pipe_stat,
  91         .fo_close = pipe_close,
  92         .fo_shutdown = pipe_shutdown
  93 };
  94
  95 static void     filt_pipedetach(struct knote *kn);
  96 static int      filt_piperead(struct knote *kn, long hint);
  97 static int      filt_pipewrite(struct knote *kn, long hint);
  98
  99 static struct filterops pipe_rfiltops =
 100         { 1, NULL, filt_pipedetach, filt_piperead };
 101 static struct filterops pipe_wfiltops =
 102         { 1, NULL, filt_pipedetach, filt_pipewrite };
 103
 104 MALLOC_DEFINE(M_PIPE, "pipe", "pipe structures");
 105
 106 /*
 107  * Default pipe buffer size(s), this can be kind-of large now because pipe
 108  * space is pageable.  The pipe code will try to maintain locality of
 109  * reference for performance reasons, so small amounts of outstanding I/O
 110  * will not wipe the cache.
 111  */
 112 #define MINPIPESIZE (PIPE_SIZE/3)
 113 #define MAXPIPESIZE (2*PIPE_SIZE/3)
 114
 115 /*
 116  * Limit the number of "big" pipes
 117  */
 118 #define LIMITBIGPIPES   64
 119 #define PIPEQ_MAX_CACHE 16      /* per-cpu pipe structure cache */
 120
 121 static int pipe_maxbig = LIMITBIGPIPES;
 122 static int pipe_maxcache = PIPEQ_MAX_CACHE;
 123 static int pipe_bigcount;
 124 static int pipe_nbig;
 125 static int pipe_bcache_alloc;
 126 static int pipe_bkmem_alloc;
 127 static int pipe_rblocked_count;
 128 static int pipe_wblocked_count;
 129
 130 SYSCTL_NODE(_kern, OID_AUTO, pipe, CTLFLAG_RW, 0, "Pipe operation");
 131 SYSCTL_INT(_kern_pipe, OID_AUTO, nbig,
 132         CTLFLAG_RD, &pipe_nbig, 0, "numer of big pipes allocated");
 133 SYSCTL_INT(_kern_pipe, OID_AUTO, bigcount,
 134         CTLFLAG_RW, &pipe_bigcount, 0, "number of times pipe expanded");
 135 SYSCTL_INT(_kern_pipe, OID_AUTO, rblocked,
 136         CTLFLAG_RW, &pipe_rblocked_count, 0, "number of times pipe expanded");
 137 SYSCTL_INT(_kern_pipe, OID_AUTO, wblocked,
 138         CTLFLAG_RW, &pipe_wblocked_count, 0, "number of times pipe expanded");
 139 SYSCTL_INT(_kern_pipe, OID_AUTO, maxcache,
 140         CTLFLAG_RW, &pipe_maxcache, 0, "max pipes cached per-cpu");
 141 SYSCTL_INT(_kern_pipe, OID_AUTO, maxbig,
 142         CTLFLAG_RW, &pipe_maxbig, 0, "max number of big pipes");
 143 #ifdef SMP
 144 static int pipe_delay = 5000;   /* 5uS default */
 145 SYSCTL_INT(_kern_pipe, OID_AUTO, delay,
 146         CTLFLAG_RW, &pipe_delay, 0, "SMP delay optimization in ns");
 147 static int pipe_mpsafe = 1;
 148 SYSCTL_INT(_kern_pipe, OID_AUTO, mpsafe,
 149         CTLFLAG_RW, &pipe_mpsafe, 0, "");
 150 #endif
 151 #if !defined(NO_PIPE_SYSCTL_STATS)
 152 SYSCTL_INT(_kern_pipe, OID_AUTO, bcache_alloc,
 153         CTLFLAG_RW, &pipe_bcache_alloc, 0, "pipe buffer from pcpu cache");
 154 SYSCTL_INT(_kern_pipe, OID_AUTO, bkmem_alloc,
 155         CTLFLAG_RW, &pipe_bkmem_alloc, 0, "pipe buffer from kmem");
 156 #endif
 157
 158 static void pipeclose (struct pipe *cpipe);
 159 static void pipe_free_kmem (struct pipe *cpipe);
 160 static int pipe_create (struct pipe **cpipep);
 161 static __inline void pipeselwakeup (struct pipe *cpipe);
 162 static int pipespace (struct pipe *cpipe, int size);
 163
 164 static __inline int
 165 pipeseltest(struct pipe *cpipe)
 166 {
 167         return ((cpipe->pipe_state & PIPE_SEL) ||
 168                 ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) ||
 169                 SLIST_FIRST(&cpipe->pipe_sel.si_note));
 170 }
 171
 172 static __inline void
 173 pipeselwakeup(struct pipe *cpipe)
 174 {
 175         if (cpipe->pipe_state & PIPE_SEL) {
 176                 get_mplock();
 177                 cpipe->pipe_state &= ~PIPE_SEL;
 178                 selwakeup(&cpipe->pipe_sel);
 179                 rel_mplock();
 180         }
 181         if ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) {
 182                 get_mplock();
 183                 pgsigio(cpipe->pipe_sigio, SIGIO, 0);
 184                 rel_mplock();
 185         }
 186         if (SLIST_FIRST(&cpipe->pipe_sel.si_note)) {
 187                 get_mplock();
 188                 KNOTE(&cpipe->pipe_sel.si_note, 0);
 189                 rel_mplock();
 190         }
 191 }
 192
 193 /*
 194  * These routines are called before and after a UIO.  The UIO
 195  * may block, causing our held tokens to be lost temporarily.
 196  *
 197  * We use these routines to serialize reads against other reads
 198  * and writes against other writes.
 199  *
 200  * The read token is held on entry so *ipp does not race.
 201  */
 202 static __inline int
 203 pipe_start_uio(struct pipe *cpipe, int *ipp)
 204 {
 205         int error;
 206
 207         while (*ipp) {
 208                 *ipp = -1;
 209                 error = tsleep(ipp, PCATCH, "pipexx", 0);
 210                 if (error)
 211                         return (error);
 212         }
 213         *ipp = 1;
 214         return (0);
 215 }
 216
 217 static __inline void
 218 pipe_end_uio(struct pipe *cpipe, int *ipp)
 219 {
 220         if (*ipp < 0) {
 221                 *ipp = 0;
 222                 wakeup(ipp);
 223         } else {
 224                 KKASSERT(*ipp > 0);
 225                 *ipp = 0;
 226         }
 227 }
 228
 229 static __inline void
 230 pipe_get_mplock(int *save)
 231 {
 232 #ifdef SMP
 233         if (pipe_mpsafe == 0) {
 234                 get_mplock();
 235                 *save = 1;
 236         } else
 237 #endif
 238         {
 239                 *save = 0;
 240         }
 241 }
 242
 243 static __inline void
 244 pipe_rel_mplock(int *save)
 245 {
 246 #ifdef SMP
 247         if (*save)
 248                 rel_mplock();
 249 #endif
 250 }
 251
 252
 253 /*
 254  * The pipe system call for the DTYPE_PIPE type of pipes
 255  *
 256  * pipe_args(int dummy)
 257  *
 258  * MPSAFE
 259  */
 260 int
 261 sys_pipe(struct pipe_args *uap)
 262 {
 263         struct thread *td = curthread;
 264         struct proc *p = td->td_proc;
 265         struct file *rf, *wf;
 266         struct pipe *rpipe, *wpipe;
 267         int fd1, fd2, error;
 268
 269         KKASSERT(p);
 270
 271         rpipe = wpipe = NULL;
 272         if (pipe_create(&rpipe) || pipe_create(&wpipe)) {
 273                 pipeclose(rpipe);
 274                 pipeclose(wpipe);
 275                 return (ENFILE);
 276         }
 277
 278         error = falloc(p, &rf, &fd1);
 279         if (error) {
 280                 pipeclose(rpipe);
 281                 pipeclose(wpipe);
 282                 return (error);
 283         }
 284         uap->sysmsg_fds[0] = fd1;
 285
 286         /*
 287          * Warning: once we've gotten past allocation of the fd for the
 288          * read-side, we can only drop the read side via fdrop() in order
 289          * to avoid races against processes which manage to dup() the read
 290          * side while we are blocked trying to allocate the write side.
 291          */
 292         rf->f_type = DTYPE_PIPE;
 293         rf->f_flag = FREAD | FWRITE;
 294         rf->f_ops = &pipeops;
 295         rf->f_data = rpipe;
 296         error = falloc(p, &wf, &fd2);
 297         if (error) {
 298                 fsetfd(p, NULL, fd1);
 299                 fdrop(rf);
 300                 /* rpipe has been closed by fdrop(). */
 301                 pipeclose(wpipe);
 302                 return (error);
 303         }
 304         wf->f_type = DTYPE_PIPE;
 305         wf->f_flag = FREAD | FWRITE;
 306         wf->f_ops = &pipeops;
 307         wf->f_data = wpipe;
 308         uap->sysmsg_fds[1] = fd2;
 309
 310         rpipe->pipe_slock = kmalloc(sizeof(struct lock),
 311                                     M_PIPE, M_WAITOK|M_ZERO);
 312         wpipe->pipe_slock = rpipe->pipe_slock;
 313         rpipe->pipe_peer = wpipe;
 314         wpipe->pipe_peer = rpipe;
 315         lockinit(rpipe->pipe_slock, "pipecl", 0, 0);
 316
 317         /*
 318          * Once activated the peer relationship remains valid until
 319          * both sides are closed.
 320          */
 321         fsetfd(p, rf, fd1);
 322         fsetfd(p, wf, fd2);
 323         fdrop(rf);
 324         fdrop(wf);
 325
 326         return (0);
 327 }
 328
 329 /*
 330  * Allocate kva for pipe circular buffer, the space is pageable
 331  * This routine will 'realloc' the size of a pipe safely, if it fails
 332  * it will retain the old buffer.
 333  * If it fails it will return ENOMEM.
 334  */
 335 static int
 336 pipespace(struct pipe *cpipe, int size)
 337 {
 338         struct vm_object *object;
 339         caddr_t buffer;
 340         int npages, error;
 341
 342         npages = round_page(size) / PAGE_SIZE;
 343         object = cpipe->pipe_buffer.object;
 344
 345         /*
 346          * [re]create the object if necessary and reserve space for it
 347          * in the kernel_map.  The object and memory are pageable.  On
 348          * success, free the old resources before assigning the new
 349          * ones.
 350          */
 351         if (object == NULL || object->size != npages) {
 352                 get_mplock();
 353                 object = vm_object_allocate(OBJT_DEFAULT, npages);
 354                 buffer = (caddr_t)vm_map_min(&kernel_map);
 355
 356                 error = vm_map_find(&kernel_map, object, 0,
 357                                     (vm_offset_t *)&buffer, size,
 358                                     1,
 359                                     VM_MAPTYPE_NORMAL,
 360                                     VM_PROT_ALL, VM_PROT_ALL,
 361                                     0);
 362
 363                 if (error != KERN_SUCCESS) {
 364                         vm_object_deallocate(object);
 365                         rel_mplock();
 366                         return (ENOMEM);
 367                 }
 368                 pipe_free_kmem(cpipe);
 369                 rel_mplock();
 370                 cpipe->pipe_buffer.object = object;
 371                 cpipe->pipe_buffer.buffer = buffer;
 372                 cpipe->pipe_buffer.size = size;
 373                 ++pipe_bkmem_alloc;
 374         } else {
 375                 ++pipe_bcache_alloc;
 376         }
 377         cpipe->pipe_buffer.rindex = 0;
 378         cpipe->pipe_buffer.windex = 0;
 379         return (0);
 380 }
 381
 382 /*
 383  * Initialize and allocate VM and memory for pipe, pulling the pipe from
 384  * our per-cpu cache if possible.  For now make sure it is sized for the
 385  * smaller PIPE_SIZE default.
 386  */
 387 static int
 388 pipe_create(struct pipe **cpipep)
 389 {
 390         globaldata_t gd = mycpu;
 391         struct pipe *cpipe;
 392         int error;
 393
 394         if ((cpipe = gd->gd_pipeq) != NULL) {
 395                 gd->gd_pipeq = cpipe->pipe_peer;
 396                 --gd->gd_pipeqcount;
 397                 cpipe->pipe_peer = NULL;
 398                 cpipe->pipe_wantwcnt = 0;
 399         } else {
 400                 cpipe = kmalloc(sizeof(struct pipe), M_PIPE, M_WAITOK|M_ZERO);
 401         }
 402         *cpipep = cpipe;
 403         if ((error = pipespace(cpipe, PIPE_SIZE)) != 0)
 404                 return (error);
 405         vfs_timestamp(&cpipe->pipe_ctime);
 406         cpipe->pipe_atime = cpipe->pipe_ctime;
 407         cpipe->pipe_mtime = cpipe->pipe_ctime;
 408         lwkt_token_init(&cpipe->pipe_rlock);
 409         lwkt_token_init(&cpipe->pipe_wlock);
 410         return (0);
 411 }
 412
 413 /*
 414  * MPALMOSTSAFE (acquires mplock)
 415  */
 416 static int
 417 pipe_read(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 418 {
 419         struct pipe *rpipe;
 420         int error;
 421         size_t nread = 0;
 422         int nbio;
 423         u_int size;     /* total bytes available */
 424         u_int nsize;    /* total bytes to read */
 425         u_int rindex;   /* contiguous bytes available */
 426         int notify_writer;
 427         lwkt_tokref rlock;
 428         lwkt_tokref wlock;
 429         int mpsave;
 430         int bigread;
 431         int bigcount;
 432
 433         if (uio->uio_resid == 0)
 434                 return(0);
 435
 436         /*
 437          * Setup locks, calculate nbio
 438          */
 439         pipe_get_mplock(&mpsave);
 440         rpipe = (struct pipe *)fp->f_data;
 441         lwkt_gettoken(&rlock, &rpipe->pipe_rlock);
 442
 443         if (fflags & O_FBLOCKING)
 444                 nbio = 0;
 445         else if (fflags & O_FNONBLOCKING)
 446                 nbio = 1;
 447         else if (fp->f_flag & O_NONBLOCK)
 448                 nbio = 1;
 449         else
 450                 nbio = 0;
 451
 452         /*
 453          * Reads are serialized.  Note howeverthat pipe_buffer.buffer and
 454          * pipe_buffer.size can change out from under us when the number
 455          * of bytes in the buffer are zero due to the write-side doing a
 456          * pipespace().
 457          */
 458         error = pipe_start_uio(rpipe, &rpipe->pipe_rip);
 459         if (error) {
 460                 pipe_rel_mplock(&mpsave);
 461                 lwkt_reltoken(&rlock);
 462                 return (error);
 463         }
 464         notify_writer = 0;
 465
 466         bigread = (uio->uio_resid > 10 * 1024 * 1024);
 467         bigcount = 10;
 468
 469         while (uio->uio_resid) {
 470                 /*
 471                  * Don't hog the cpu.
 472                  */
 473                 if (bigread && --bigcount == 0) {
 474                         lwkt_user_yield();
 475                         bigcount = 10;
 476                         if (CURSIG(curthread->td_lwp)) {
 477                                 error = EINTR;
 478                                 break;
 479                         }
 480                 }
 481
 482                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 483                 cpu_lfence();
 484                 if (size) {
 485                         rindex = rpipe->pipe_buffer.rindex &
 486                                  (rpipe->pipe_buffer.size - 1);
 487                         nsize = size;
 488                         if (nsize > rpipe->pipe_buffer.size - rindex)
 489                                 nsize = rpipe->pipe_buffer.size - rindex;
 490                         nsize = szmin(nsize, uio->uio_resid);
 491
 492                         error = uiomove(&rpipe->pipe_buffer.buffer[rindex],
 493                                         nsize, uio);
 494                         if (error)
 495                                 break;
 496                         cpu_mfence();
 497                         rpipe->pipe_buffer.rindex += nsize;
 498                         nread += nsize;
 499
 500                         /*
 501                          * If the FIFO is still over half full just continue
 502                          * and do not try to notify the writer yet.
 503                          */
 504                         if (size - nsize >= (rpipe->pipe_buffer.size >> 1)) {
 505                                 notify_writer = 0;
 506                                 continue;
 507                         }
 508
 509                         /*
 510                          * When the FIFO is less then half full notify any
 511                          * waiting writer.  WANTW can be checked while
 512                          * holding just the rlock.
 513                          */
 514                         notify_writer = 1;
 515                         if ((rpipe->pipe_state & PIPE_WANTW) == 0)
 516                                 continue;
 517                 }
 518
 519                 /*
 520                  * If the "write-side" was blocked we wake it up.  This code
 521                  * is reached either when the buffer is completely emptied
 522                  * or if it becomes more then half-empty.
 523                  *
 524                  * Pipe_state can only be modified if both the rlock and
 525                  * wlock are held.
 526                  */
 527                 if (rpipe->pipe_state & PIPE_WANTW) {
 528                         lwkt_gettoken(&wlock, &rpipe->pipe_wlock);
 529                         if (rpipe->pipe_state & PIPE_WANTW) {
 530                                 notify_writer = 0;
 531                                 rpipe->pipe_state &= ~PIPE_WANTW;
 532                                 lwkt_reltoken(&wlock);
 533                                 wakeup(rpipe);
 534                         } else {
 535                                 lwkt_reltoken(&wlock);
 536                         }
 537                 }
 538
 539                 /*
 540                  * Pick up our copy loop again if the writer sent data to
 541                  * us while we were messing around.
 542                  *
 543                  * On a SMP box poll up to pipe_delay nanoseconds for new
 544                  * data.  Typically a value of 2000 to 4000 is sufficient
 545                  * to eradicate most IPIs/tsleeps/wakeups when a pipe
 546                  * is used for synchronous communications with small packets,
 547                  * and 8000 or so (8uS) will pipeline large buffer xfers
 548                  * between cpus over a pipe.
 549                  *
 550                  * For synchronous communications a hit means doing a
 551                  * full Awrite-Bread-Bwrite-Aread cycle in less then 2uS,
 552                  * where as miss requiring a tsleep/wakeup sequence
 553                  * will take 7uS or more.
 554                  */
 555                 if (rpipe->pipe_buffer.windex != rpipe->pipe_buffer.rindex)
 556                         continue;
 557
 558 #if defined(SMP) && defined(_RDTSC_SUPPORTED_)
 559                 if (pipe_delay) {
 560                         int64_t tsc_target;
 561                         int good = 0;
 562
 563                         tsc_target = tsc_get_target(pipe_delay);
 564                         while (tsc_test_target(tsc_target) == 0) {
 565                                 if (rpipe->pipe_buffer.windex !=
 566                                     rpipe->pipe_buffer.rindex) {
 567                                         good = 1;
 568                                         break;
 569                                 }
 570                         }
 571                         if (good)
 572                                 continue;
 573                 }
 574 #endif
 575
 576                 /*
 577                  * Detect EOF condition, do not set error.
 578                  */
 579                 if (rpipe->pipe_state & PIPE_REOF)
 580                         break;
 581
 582                 /*
 583                  * Break if some data was read, or if this was a non-blocking
 584                  * read.
 585                  */
 586                 if (nread > 0)
 587                         break;
 588
 589                 if (nbio) {
 590                         error = EAGAIN;
 591                         break;
 592                 }
 593
 594                 /*
 595                  * Last chance, interlock with WANTR.
 596                  */
 597                 lwkt_gettoken(&wlock, &rpipe->pipe_wlock);
 598                 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
 599                 if (size) {
 600                         lwkt_reltoken(&wlock);
 601                         continue;
 602                 }
 603
 604                 /*
 605                  * Retest EOF - acquiring a new token can temporarily release
 606                  * tokens already held.
 607                  */
 608                 if (rpipe->pipe_state & PIPE_REOF) {
 609                         lwkt_reltoken(&wlock);
 610                         break;
 611                 }
 612
 613                 /*
 614                  * If there is no more to read in the pipe, reset its
 615                  * pointers to the beginning.  This improves cache hit
 616                  * stats.
 617                  *
 618                  * We need both locks to modify both pointers, and there
 619                  * must also not be a write in progress or the uiomove()
 620                  * in the write might block and temporarily release
 621                  * its wlock, then reacquire and update windex.  We are
 622                  * only serialized against reads, not writes.
 623                  *
 624                  * XXX should we even bother resetting the indices?  It
 625                  *     might actually be more cache efficient not to.
 626                  */
 627                 if (rpipe->pipe_buffer.rindex == rpipe->pipe_buffer.windex &&
 628                     rpipe->pipe_wip == 0) {
 629                         rpipe->pipe_buffer.rindex = 0;
 630                         rpipe->pipe_buffer.windex = 0;
 631                 }
 632
 633                 /*
 634                  * Wait for more data.
 635                  *
 636                  * Pipe_state can only be set if both the rlock and wlock
 637                  * are held.
 638                  */
 639                 rpipe->pipe_state |= PIPE_WANTR;
 640                 tsleep_interlock(rpipe, PCATCH);
 641                 lwkt_reltoken(&wlock);
 642                 error = tsleep(rpipe, PCATCH | PINTERLOCKED, "piperd", 0);
 643                 ++pipe_rblocked_count;
 644                 if (error)
 645                         break;
 646         }
 647         pipe_end_uio(rpipe, &rpipe->pipe_rip);
 648
 649         /*
 650          * Uptime last access time
 651          */
 652         if (error == 0 && nread)
 653                 vfs_timestamp(&rpipe->pipe_atime);
 654
 655         /*
 656          * If we drained the FIFO more then half way then handle
 657          * write blocking hysteresis.
 658          *
 659          * Note that PIPE_WANTW cannot be set by the writer without
 660          * it holding both rlock and wlock, so we can test it
 661          * while holding just rlock.
 662          */
 663         if (notify_writer) {
 664                 if (rpipe->pipe_state & PIPE_WANTW) {
 665                         lwkt_gettoken(&wlock, &rpipe->pipe_wlock);
 666                         if (rpipe->pipe_state & PIPE_WANTW) {
 667                                 rpipe->pipe_state &= ~PIPE_WANTW;
 668                                 lwkt_reltoken(&wlock);
 669                                 wakeup(rpipe);
 670                         } else {
 671                                 lwkt_reltoken(&wlock);
 672                         }
 673                 }
 674                 if (pipeseltest(rpipe)) {
 675                         lwkt_gettoken(&wlock, &rpipe->pipe_wlock);
 676                         pipeselwakeup(rpipe);
 677                         lwkt_reltoken(&wlock);
 678                 }
 679         }
 680         /*size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;*/
 681         lwkt_reltoken(&rlock);
 682
 683         pipe_rel_mplock(&mpsave);
 684         return (error);
 685 }
 686
 687 /*
 688  * MPALMOSTSAFE - acquires mplock
 689  */
 690 static int
 691 pipe_write(struct file *fp, struct uio *uio, struct ucred *cred, int fflags)
 692 {
 693         int error;
 694         int orig_resid;
 695         int nbio;
 696         struct pipe *wpipe, *rpipe;
 697         lwkt_tokref rlock;
 698         lwkt_tokref wlock;
 699         u_int windex;
 700         u_int space;
 701         u_int wcount;
 702         int mpsave;
 703         int bigwrite;
 704         int bigcount;
 705
 706         pipe_get_mplock(&mpsave);
 707
 708         /*
 709          * Writes go to the peer.  The peer will always exist.
 710          */
 711         rpipe = (struct pipe *) fp->f_data;
 712         wpipe = rpipe->pipe_peer;
 713         lwkt_gettoken(&wlock, &wpipe->pipe_wlock);
 714         if (wpipe->pipe_state & PIPE_WEOF) {
 715                 pipe_rel_mplock(&mpsave);
 716                 lwkt_reltoken(&wlock);
 717                 return (EPIPE);
 718         }
 719
 720         /*
 721          * Degenerate case (EPIPE takes prec)
 722          */
 723         if (uio->uio_resid == 0) {
 724                 pipe_rel_mplock(&mpsave);
 725                 lwkt_reltoken(&wlock);
 726                 return(0);
 727         }
 728
 729         /*
 730          * Writes are serialized (start_uio must be called with wlock)
 731          */
 732         error = pipe_start_uio(wpipe, &wpipe->pipe_wip);
 733         if (error) {
 734                 pipe_rel_mplock(&mpsave);
 735                 lwkt_reltoken(&wlock);
 736                 return (error);
 737         }
 738
 739         if (fflags & O_FBLOCKING)
 740                 nbio = 0;
 741         else if (fflags & O_FNONBLOCKING)
 742                 nbio = 1;
 743         else if (fp->f_flag & O_NONBLOCK)
 744                 nbio = 1;
 745         else
 746                 nbio = 0;
 747
 748         /*
 749          * If it is advantageous to resize the pipe buffer, do
 750          * so.  We are write-serialized so we can block safely.
 751          */
 752         if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 753             (pipe_nbig < pipe_maxbig) &&
 754             wpipe->pipe_wantwcnt > 4 &&
 755             (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 756                 /*
 757                  * Recheck after lock.
 758                  */
 759                 lwkt_gettoken(&rlock, &wpipe->pipe_rlock);
 760                 if ((wpipe->pipe_buffer.size <= PIPE_SIZE) &&
 761                     (pipe_nbig < pipe_maxbig) &&
 762                     (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) {
 763                         atomic_add_int(&pipe_nbig, 1);
 764                         if (pipespace(wpipe, BIG_PIPE_SIZE) == 0)
 765                                 ++pipe_bigcount;
 766                         else
 767                                 atomic_subtract_int(&pipe_nbig, 1);
 768                 }
 769                 lwkt_reltoken(&rlock);
 770         }
 771
 772         orig_resid = uio->uio_resid;
 773         wcount = 0;
 774
 775         bigwrite = (uio->uio_resid > 10 * 1024 * 1024);
 776         bigcount = 10;
 777
 778         while (uio->uio_resid) {
 779                 if (wpipe->pipe_state & PIPE_WEOF) {
 780                         error = EPIPE;
 781                         break;
 782                 }
 783
 784                 /*
 785                  * Don't hog the cpu.
 786                  */
 787                 if (bigwrite && --bigcount == 0) {
 788                         lwkt_user_yield();
 789                         bigcount = 10;
 790                         if (CURSIG(curthread->td_lwp)) {
 791                                 error = EINTR;
 792                                 break;
 793                         }
 794                 }
 795
 796                 windex = wpipe->pipe_buffer.windex &
 797                          (wpipe->pipe_buffer.size - 1);
 798                 space = wpipe->pipe_buffer.size -
 799                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 800                 cpu_lfence();
 801
 802                 /* Writes of size <= PIPE_BUF must be atomic. */
 803                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 804                         space = 0;
 805
 806                 /*
 807                  * Write to fill, read size handles write hysteresis.  Also
 808                  * additional restrictions can cause select-based non-blocking
 809                  * writes to spin.
 810                  */
 811                 if (space > 0) {
 812                         u_int segsize;
 813
 814                         /*
 815                          * Transfer size is minimum of uio transfer
 816                          * and free space in pipe buffer.
 817                          *
 818                          * Limit each uiocopy to no more then PIPE_SIZE
 819                          * so we can keep the gravy train going on a
 820                          * SMP box.  This doubles the performance for
 821                          * write sizes > 16K.  Otherwise large writes
 822                          * wind up doing an inefficient synchronous
 823                          * ping-pong.
 824                          */
 825                         space = szmin(space, uio->uio_resid);
 826                         if (space > PIPE_SIZE)
 827                                 space = PIPE_SIZE;
 828
 829                         /*
 830                          * First segment to transfer is minimum of
 831                          * transfer size and contiguous space in
 832                          * pipe buffer.  If first segment to transfer
 833                          * is less than the transfer size, we've got
 834                          * a wraparound in the buffer.
 835                          */
 836                         segsize = wpipe->pipe_buffer.size - windex;
 837                         if (segsize > space)
 838                                 segsize = space;
 839
 840 #ifdef SMP
 841                         /*
 842                          * If this is the first loop and the reader is
 843                          * blocked, do a preemptive wakeup of the reader.
 844                          *
 845                          * On SMP the IPI latency plus the wlock interlock
 846                          * on the reader side is the fastest way to get the
 847                          * reader going.  (The scheduler will hard loop on
 848                          * lock tokens).
 849                          *
 850                          * NOTE: We can't clear WANTR here without acquiring
 851                          * the rlock, which we don't want to do here!
 852                          */
 853                         if ((wpipe->pipe_state & PIPE_WANTR) && pipe_mpsafe > 1)
 854                                 wakeup(wpipe);
 855 #endif
 856
 857                         /*
 858                          * Transfer segment, which may include a wrap-around.
 859                          * Update windex to account for both all in one go
 860                          * so the reader can read() the data atomically.
 861                          */
 862                         error = uiomove(&wpipe->pipe_buffer.buffer[windex],
 863                                         segsize, uio);
 864                         if (error == 0 && segsize < space) {
 865                                 segsize = space - segsize;
 866                                 error = uiomove(&wpipe->pipe_buffer.buffer[0],
 867                                                 segsize, uio);
 868                         }
 869                         if (error)
 870                                 break;
 871                         cpu_mfence();
 872                         wpipe->pipe_buffer.windex += space;
 873                         wcount += space;
 874                         continue;
 875                 }
 876
 877                 /*
 878                  * We need both the rlock and the wlock to interlock against
 879                  * the EOF, WANTW, and size checks, and to modify pipe_state.
 880                  *
 881                  * These are token locks so we do not have to worry about
 882                  * deadlocks.
 883                  */
 884                 lwkt_gettoken(&rlock, &wpipe->pipe_rlock);
 885
 886                 /*
 887                  * If the "read-side" has been blocked, wake it up now
 888                  * and yield to let it drain synchronously rather
 889                  * then block.
 890                  */
 891                 if (wpipe->pipe_state & PIPE_WANTR) {
 892                         wpipe->pipe_state &= ~PIPE_WANTR;
 893                         wakeup(wpipe);
 894                 }
 895
 896                 /*
 897                  * don't block on non-blocking I/O
 898                  */
 899                 if (nbio) {
 900                         lwkt_reltoken(&rlock);
 901                         error = EAGAIN;
 902                         break;
 903                 }
 904
 905                 /*
 906                  * re-test whether we have to block in the writer after
 907                  * acquiring both locks, in case the reader opened up
 908                  * some space.
 909                  */
 910                 space = wpipe->pipe_buffer.size -
 911                         (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex);
 912                 cpu_lfence();
 913                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
 914                         space = 0;
 915
 916                 /*
 917                  * Retest EOF - acquiring a new token can temporarily release
 918                  * tokens already held.
 919                  */
 920                 if (wpipe->pipe_state & PIPE_WEOF) {
 921                         lwkt_reltoken(&rlock);
 922                         error = EPIPE;
 923                         break;
 924                 }
 925
 926                 /*
 927                  * We have no more space and have something to offer,
 928                  * wake up select/poll.
 929                  */
 930                 if (space == 0) {
 931                         wpipe->pipe_state |= PIPE_WANTW;
 932                         ++wpipe->pipe_wantwcnt;
 933                         pipeselwakeup(wpipe);
 934                         if (wpipe->pipe_state & PIPE_WANTW)
 935                                 error = tsleep(wpipe, PCATCH, "pipewr", 0);
 936                         ++pipe_wblocked_count;
 937                 }
 938                 lwkt_reltoken(&rlock);
 939
 940                 /*
 941                  * Break out if we errored or the read side wants us to go
 942                  * away.
 943                  */
 944                 if (error)
 945                         break;
 946                 if (wpipe->pipe_state & PIPE_WEOF) {
 947                         error = EPIPE;
 948                         break;
 949                 }
 950         }
 951         pipe_end_uio(wpipe, &wpipe->pipe_wip);
 952
 953         /*
 954          * If we have put any characters in the buffer, we wake up
 955          * the reader.
 956          *
 957          * Both rlock and wlock are required to be able to modify pipe_state.
 958          */
 959         if (wpipe->pipe_buffer.windex != wpipe->pipe_buffer.rindex) {
 960                 if (wpipe->pipe_state & PIPE_WANTR) {
 961                         lwkt_gettoken(&rlock, &wpipe->pipe_rlock);
 962                         if (wpipe->pipe_state & PIPE_WANTR) {
 963                                 wpipe->pipe_state &= ~PIPE_WANTR;
 964                                 lwkt_reltoken(&rlock);
 965                                 wakeup(wpipe);
 966                         } else {
 967                                 lwkt_reltoken(&rlock);
 968                         }
 969                 }
 970                 if (pipeseltest(wpipe)) {
 971                         lwkt_gettoken(&rlock, &wpipe->pipe_rlock);
 972                         pipeselwakeup(wpipe);
 973                         lwkt_reltoken(&rlock);
 974                 }
 975         }
 976
 977         /*
 978          * Don't return EPIPE if I/O was successful
 979          */
 980         if ((wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex) &&
 981             (uio->uio_resid == 0) &&
 982             (error == EPIPE)) {
 983                 error = 0;
 984         }
 985
 986         if (error == 0)
 987                 vfs_timestamp(&wpipe->pipe_mtime);
 988
 989         /*
 990          * We have something to offer,
 991          * wake up select/poll.
 992          */
 993         /*space = wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex;*/
 994         lwkt_reltoken(&wlock);
 995         pipe_rel_mplock(&mpsave);
 996         return (error);
 997 }
 998
 999 /*
1000  * MPALMOSTSAFE - acquires mplock
1001  *
1002  * we implement a very minimal set of ioctls for compatibility with sockets.
1003  */
1004 int
1005 pipe_ioctl(struct file *fp, u_long cmd, caddr_t data,
1006            struct ucred *cred, struct sysmsg *msg)
1007 {
1008         struct pipe *mpipe;
1009         lwkt_tokref rlock;
1010         lwkt_tokref wlock;
1011         int error;
1012         int mpsave;
1013
1014         pipe_get_mplock(&mpsave);
1015         mpipe = (struct pipe *)fp->f_data;
1016
1017         lwkt_gettoken(&rlock, &mpipe->pipe_rlock);
1018         lwkt_gettoken(&wlock, &mpipe->pipe_wlock);
1019
1020         switch (cmd) {
1021         case FIOASYNC:
1022                 if (*(int *)data) {
1023                         mpipe->pipe_state |= PIPE_ASYNC;
1024                 } else {
1025                         mpipe->pipe_state &= ~PIPE_ASYNC;
1026                 }
1027                 error = 0;
1028                 break;
1029         case FIONREAD:
1030                 *(int *)data = mpipe->pipe_buffer.windex -
1031                                 mpipe->pipe_buffer.rindex;
1032                 error = 0;
1033                 break;
1034         case FIOSETOWN:
1035                 get_mplock();
1036                 error = fsetown(*(int *)data, &mpipe->pipe_sigio);
1037                 rel_mplock();
1038                 break;
1039         case FIOGETOWN:
1040                 *(int *)data = fgetown(mpipe->pipe_sigio);
1041                 error = 0;
1042                 break;
1043         case TIOCSPGRP:
1044                 /* This is deprecated, FIOSETOWN should be used instead. */
1045                 get_mplock();
1046                 error = fsetown(-(*(int *)data), &mpipe->pipe_sigio);
1047                 rel_mplock();
1048                 break;
1049
1050         case TIOCGPGRP:
1051                 /* This is deprecated, FIOGETOWN should be used instead. */
1052                 *(int *)data = -fgetown(mpipe->pipe_sigio);
1053                 error = 0;
1054                 break;
1055         default:
1056                 error = ENOTTY;
1057                 break;
1058         }
1059         lwkt_reltoken(&rlock);
1060         lwkt_reltoken(&wlock);
1061         pipe_rel_mplock(&mpsave);
1062
1063         return (error);
1064 }
1065
1066 /*
1067  * MPALMOSTSAFE - acquires mplock
1068  *
1069  * poll for events (helper)
1070  */
1071 static int
1072 pipe_poll_events(struct pipe *rpipe, struct pipe *wpipe, int events)
1073 {
1074         int revents = 0;
1075         u_int space;
1076
1077         if (events & (POLLIN | POLLRDNORM)) {
1078                 if ((rpipe->pipe_buffer.windex != rpipe->pipe_buffer.rindex) ||
1079                     (rpipe->pipe_state & PIPE_REOF)) {
1080                         revents |= events & (POLLIN | POLLRDNORM);
1081                 }
1082         }
1083
1084         if (events & (POLLOUT | POLLWRNORM)) {
1085                 if (wpipe == NULL || (wpipe->pipe_state & PIPE_WEOF)) {
1086                         revents |= events & (POLLOUT | POLLWRNORM);
1087                 } else {
1088                         space = wpipe->pipe_buffer.windex -
1089                                 wpipe->pipe_buffer.rindex;
1090                         space = wpipe->pipe_buffer.size - space;
1091                         if (space >= PIPE_BUF)
1092                                 revents |= events & (POLLOUT | POLLWRNORM);
1093                 }
1094         }
1095
1096         if ((rpipe->pipe_state & PIPE_REOF) ||
1097             (wpipe == NULL) ||
1098             (wpipe->pipe_state & PIPE_WEOF)) {
1099                 revents |= POLLHUP;
1100         }
1101         return (revents);
1102 }
1103
1104 /*
1105  * Poll for events from file pointer.
1106  */
1107 int
1108 pipe_poll(struct file *fp, int events, struct ucred *cred)
1109 {
1110         lwkt_tokref rpipe_rlock;
1111         lwkt_tokref rpipe_wlock;
1112         lwkt_tokref wpipe_rlock;
1113         lwkt_tokref wpipe_wlock;
1114         struct pipe *rpipe;
1115         struct pipe *wpipe;
1116         int revents = 0;
1117         int mpsave;
1118
1119         pipe_get_mplock(&mpsave);
1120         rpipe = (struct pipe *)fp->f_data;
1121         wpipe = rpipe->pipe_peer;
1122
1123         revents = pipe_poll_events(rpipe, wpipe, events);
1124         if (revents == 0) {
1125                 if (events & (POLLIN | POLLRDNORM)) {
1126                         lwkt_gettoken(&rpipe_rlock, &rpipe->pipe_rlock);
1127                         lwkt_gettoken(&rpipe_wlock, &rpipe->pipe_wlock);
1128                 }
1129                 if (events & (POLLOUT | POLLWRNORM)) {
1130                         lwkt_gettoken(&wpipe_rlock, &wpipe->pipe_rlock);
1131                         lwkt_gettoken(&wpipe_wlock, &wpipe->pipe_wlock);
1132                 }
1133                 revents = pipe_poll_events(rpipe, wpipe, events);
1134                 if (revents == 0) {
1135                         if (events & (POLLIN | POLLRDNORM)) {
1136                                 selrecord(curthread, &rpipe->pipe_sel);
1137                                 rpipe->pipe_state |= PIPE_SEL;
1138                         }
1139
1140                         if (events & (POLLOUT | POLLWRNORM)) {
1141                                 selrecord(curthread, &wpipe->pipe_sel);
1142                                 wpipe->pipe_state |= PIPE_SEL;
1143                         }
1144                 }
1145                 if (events & (POLLIN | POLLRDNORM)) {
1146                         lwkt_reltoken(&rpipe_rlock);
1147                         lwkt_reltoken(&rpipe_wlock);
1148                 }
1149                 if (events & (POLLOUT | POLLWRNORM)) {
1150                         lwkt_reltoken(&wpipe_rlock);
1151                         lwkt_reltoken(&wpipe_wlock);
1152                 }
1153         }
1154         pipe_rel_mplock(&mpsave);
1155         return (revents);
1156 }
1157
1158 /*
1159  * MPSAFE
1160  */
1161 static int
1162 pipe_stat(struct file *fp, struct stat *ub, struct ucred *cred)
1163 {
1164         struct pipe *pipe;
1165         int mpsave;
1166
1167         pipe_get_mplock(&mpsave);
1168         pipe = (struct pipe *)fp->f_data;
1169
1170         bzero((caddr_t)ub, sizeof(*ub));
1171         ub->st_mode = S_IFIFO;
1172         ub->st_blksize = pipe->pipe_buffer.size;
1173         ub->st_size = pipe->pipe_buffer.windex - pipe->pipe_buffer.rindex;
1174         ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize;
1175         ub->st_atimespec = pipe->pipe_atime;
1176         ub->st_mtimespec = pipe->pipe_mtime;
1177         ub->st_ctimespec = pipe->pipe_ctime;
1178         /*
1179          * Left as 0: st_dev, st_ino, st_nlink, st_uid, st_gid, st_rdev,
1180          * st_flags, st_gen.
1181          * XXX (st_dev, st_ino) should be unique.
1182          */
1183         pipe_rel_mplock(&mpsave);
1184         return (0);
1185 }
1186
1187 /*
1188  * MPALMOSTSAFE - acquires mplock
1189  */
1190 static int
1191 pipe_close(struct file *fp)
1192 {
1193         struct pipe *cpipe;
1194
1195         get_mplock();
1196         cpipe = (struct pipe *)fp->f_data;
1197         fp->f_ops = &badfileops;
1198         fp->f_data = NULL;
1199         funsetown(cpipe->pipe_sigio);
1200         pipeclose(cpipe);
1201         rel_mplock();
1202         return (0);
1203 }
1204
1205 /*
1206  * Shutdown one or both directions of a full-duplex pipe.
1207  *
1208  * MPALMOSTSAFE - acquires mplock
1209  */
1210 static int
1211 pipe_shutdown(struct file *fp, int how)
1212 {
1213         struct pipe *rpipe;
1214         struct pipe *wpipe;
1215         int error = EPIPE;
1216         lwkt_tokref rpipe_rlock;
1217         lwkt_tokref rpipe_wlock;
1218         lwkt_tokref wpipe_rlock;
1219         lwkt_tokref wpipe_wlock;
1220         int mpsave;
1221
1222         pipe_get_mplock(&mpsave);
1223         rpipe = (struct pipe *)fp->f_data;
1224         wpipe = rpipe->pipe_peer;
1225
1226         /*
1227          * We modify pipe_state on both pipes, which means we need
1228          * all four tokens!
1229          */
1230         lwkt_gettoken(&rpipe_rlock, &rpipe->pipe_rlock);
1231         lwkt_gettoken(&rpipe_wlock, &rpipe->pipe_wlock);
1232         lwkt_gettoken(&wpipe_rlock, &wpipe->pipe_rlock);
1233         lwkt_gettoken(&wpipe_wlock, &wpipe->pipe_wlock);
1234
1235         switch(how) {
1236         case SHUT_RDWR:
1237         case SHUT_RD:
1238                 rpipe->pipe_state |= PIPE_REOF;         /* my reads */
1239                 rpipe->pipe_state |= PIPE_WEOF;         /* peer writes */
1240                 if (rpipe->pipe_state & PIPE_WANTR) {
1241                         rpipe->pipe_state &= ~PIPE_WANTR;
1242                         wakeup(rpipe);
1243                 }
1244                 if (rpipe->pipe_state & PIPE_WANTW) {
1245                         rpipe->pipe_state &= ~PIPE_WANTW;
1246                         wakeup(rpipe);
1247                 }
1248                 error = 0;
1249                 if (how == SHUT_RD)
1250                         break;
1251                 /* fall through */
1252         case SHUT_WR:
1253                 wpipe->pipe_state |= PIPE_REOF;         /* peer reads */
1254                 wpipe->pipe_state |= PIPE_WEOF;         /* my writes */
1255                 if (wpipe->pipe_state & PIPE_WANTR) {
1256                         wpipe->pipe_state &= ~PIPE_WANTR;
1257                         wakeup(wpipe);
1258                 }
1259                 if (wpipe->pipe_state & PIPE_WANTW) {
1260                         wpipe->pipe_state &= ~PIPE_WANTW;
1261                         wakeup(wpipe);
1262                 }
1263                 error = 0;
1264                 break;
1265         }
1266         pipeselwakeup(rpipe);
1267         pipeselwakeup(wpipe);
1268
1269         lwkt_reltoken(&rpipe_rlock);
1270         lwkt_reltoken(&rpipe_wlock);
1271         lwkt_reltoken(&wpipe_rlock);
1272         lwkt_reltoken(&wpipe_wlock);
1273
1274         pipe_rel_mplock(&mpsave);
1275         return (error);
1276 }
1277
1278 static void
1279 pipe_free_kmem(struct pipe *cpipe)
1280 {
1281         if (cpipe->pipe_buffer.buffer != NULL) {
1282                 if (cpipe->pipe_buffer.size > PIPE_SIZE)
1283                         atomic_subtract_int(&pipe_nbig, 1);
1284                 kmem_free(&kernel_map,
1285                         (vm_offset_t)cpipe->pipe_buffer.buffer,
1286                         cpipe->pipe_buffer.size);
1287                 cpipe->pipe_buffer.buffer = NULL;
1288                 cpipe->pipe_buffer.object = NULL;
1289         }
1290 }
1291
1292 /*
1293  * Close the pipe.  The slock must be held to interlock against simultanious
1294  * closes.  The rlock and wlock must be held to adjust the pipe_state.
1295  */
1296 static void
1297 pipeclose(struct pipe *cpipe)
1298 {
1299         globaldata_t gd;
1300         struct pipe *ppipe;
1301         lwkt_tokref cpipe_rlock;
1302         lwkt_tokref cpipe_wlock;
1303         lwkt_tokref ppipe_rlock;
1304         lwkt_tokref ppipe_wlock;
1305
1306         if (cpipe == NULL)
1307                 return;
1308
1309         /*
1310          * The slock may not have been allocated yet (close during
1311          * initialization)
1312          *
1313          * We need both the read and write tokens to modify pipe_state.
1314          */
1315         if (cpipe->pipe_slock)
1316                 lockmgr(cpipe->pipe_slock, LK_EXCLUSIVE);
1317         lwkt_gettoken(&cpipe_rlock, &cpipe->pipe_rlock);
1318         lwkt_gettoken(&cpipe_wlock, &cpipe->pipe_wlock);
1319
1320         /*
1321          * Set our state, wakeup anyone waiting in select, and
1322          * wakeup anyone blocked on our pipe.
1323          */
1324         cpipe->pipe_state |= PIPE_CLOSED | PIPE_REOF | PIPE_WEOF;
1325         pipeselwakeup(cpipe);
1326         if (cpipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1327                 cpipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1328                 wakeup(cpipe);
1329         }
1330
1331         /*
1332          * Disconnect from peer.
1333          */
1334         if ((ppipe = cpipe->pipe_peer) != NULL) {
1335                 lwkt_gettoken(&ppipe_rlock, &ppipe->pipe_rlock);
1336                 lwkt_gettoken(&ppipe_wlock, &ppipe->pipe_wlock);
1337                 ppipe->pipe_state |= PIPE_REOF | PIPE_WEOF;
1338                 pipeselwakeup(ppipe);
1339                 if (ppipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) {
1340                         ppipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW);
1341                         wakeup(ppipe);
1342                 }
1343                 if (SLIST_FIRST(&ppipe->pipe_sel.si_note)) {
1344                         get_mplock();
1345                         KNOTE(&ppipe->pipe_sel.si_note, 0);
1346                         rel_mplock();
1347                 }
1348                 lwkt_reltoken(&ppipe_rlock);
1349                 lwkt_reltoken(&ppipe_wlock);
1350         }
1351
1352         /*
1353          * If the peer is also closed we can free resources for both
1354          * sides, otherwise we leave our side intact to deal with any
1355          * races (since we only have the slock).
1356          */
1357         if (ppipe && (ppipe->pipe_state & PIPE_CLOSED)) {
1358                 cpipe->pipe_peer = NULL;
1359                 ppipe->pipe_peer = NULL;
1360                 ppipe->pipe_slock = NULL;       /* we will free the slock */
1361                 pipeclose(ppipe);
1362                 ppipe = NULL;
1363         }
1364
1365         lwkt_reltoken(&cpipe_rlock);
1366         lwkt_reltoken(&cpipe_wlock);
1367         if (cpipe->pipe_slock)
1368                 lockmgr(cpipe->pipe_slock, LK_RELEASE);
1369
1370         /*
1371          * If we disassociated from our peer we can free resources
1372          */
1373         if (ppipe == NULL) {
1374                 gd = mycpu;
1375                 if (cpipe->pipe_slock) {
1376                         kfree(cpipe->pipe_slock, M_PIPE);
1377                         cpipe->pipe_slock = NULL;
1378                 }
1379                 if (gd->gd_pipeqcount >= pipe_maxcache ||
1380                     cpipe->pipe_buffer.size != PIPE_SIZE
1381                 ) {
1382                         pipe_free_kmem(cpipe);
1383                         kfree(cpipe, M_PIPE);
1384                 } else {
1385                         cpipe->pipe_state = 0;
1386                         cpipe->pipe_peer = gd->gd_pipeq;
1387                         gd->gd_pipeq = cpipe;
1388                         ++gd->gd_pipeqcount;
1389                 }
1390         }
1391 }
1392
1393 /*
1394  * MPALMOSTSAFE - acquires mplock
1395  */
1396 static int
1397 pipe_kqfilter(struct file *fp, struct knote *kn)
1398 {
1399         struct pipe *cpipe;
1400
1401         get_mplock();
1402         cpipe = (struct pipe *)kn->kn_fp->f_data;
1403
1404         switch (kn->kn_filter) {
1405         case EVFILT_READ:
1406                 kn->kn_fop = &pipe_rfiltops;
1407                 break;
1408         case EVFILT_WRITE:
1409                 kn->kn_fop = &pipe_wfiltops;
1410                 cpipe = cpipe->pipe_peer;
1411                 if (cpipe == NULL) {
1412                         /* other end of pipe has been closed */
1413                         rel_mplock();
1414                         return (EPIPE);
1415                 }
1416                 break;
1417         default:
1418                 return (1);
1419         }
1420         kn->kn_hook = (caddr_t)cpipe;
1421
1422         SLIST_INSERT_HEAD(&cpipe->pipe_sel.si_note, kn, kn_selnext);
1423         rel_mplock();
1424         return (0);
1425 }
1426
1427 static void
1428 filt_pipedetach(struct knote *kn)
1429 {
1430         struct pipe *cpipe = (struct pipe *)kn->kn_hook;
1431
1432         SLIST_REMOVE(&cpipe->pipe_sel.si_note, kn, knote, kn_selnext);
1433 }
1434
1435 /*ARGSUSED*/
1436 static int
1437 filt_piperead(struct knote *kn, long hint)
1438 {
1439         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1440
1441         kn->kn_data = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;
1442
1443         /* XXX RACE */
1444         if (rpipe->pipe_state & PIPE_REOF) {
1445                 kn->kn_flags |= EV_EOF;
1446                 return (1);
1447         }
1448         return (kn->kn_data > 0);
1449 }
1450
1451 /*ARGSUSED*/
1452 static int
1453 filt_pipewrite(struct knote *kn, long hint)
1454 {
1455         struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data;
1456         struct pipe *wpipe = rpipe->pipe_peer;
1457         u_int32_t space;
1458
1459         /* XXX RACE */
1460         if ((wpipe == NULL) || (wpipe->pipe_state & PIPE_WEOF)) {
1461                 kn->kn_data = 0;
1462                 kn->kn_flags |= EV_EOF;
1463                 return (1);
1464         }
1465         space = wpipe->pipe_buffer.windex -
1466                 wpipe->pipe_buffer.rindex;
1467         space = wpipe->pipe_buffer.size - space;
1468         kn->kn_data = space;
1469         return (kn->kn_data >= PIPE_BUF);
1470 }