sys_pipe.c revision 91968
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: head/sys/kern/sys_pipe.c 91968 2002-03-09 22:06:31Z alfred $ 20 */ 21 22/* 23 * This file contains a high-performance replacement for the socket-based 24 * pipes scheme originally used in FreeBSD/4.4Lite. It does not support 25 * all features of sockets, but does do everything that pipes normally 26 * do. 27 */ 28 29/* 30 * This code has two modes of operation, a small write mode and a large 31 * write mode. The small write mode acts like conventional pipes with 32 * a kernel buffer. If the buffer is less than PIPE_MINDIRECT, then the 33 * "normal" pipe buffering is done. If the buffer is between PIPE_MINDIRECT 34 * and PIPE_SIZE in size, it is fully mapped and wired into the kernel, and 35 * the receiving process can copy it directly from the pages in the sending 36 * process. 37 * 38 * If the sending process receives a signal, it is possible that it will 39 * go away, and certainly its address space can change, because control 40 * is returned back to the user-mode side. In that case, the pipe code 41 * arranges to copy the buffer supplied by the user process, to a pageable 42 * kernel buffer, and the receiving process will grab the data from the 43 * pageable kernel buffer. Since signals don't happen all that often, 44 * the copy operation is normally eliminated. 45 * 46 * The constant PIPE_MINDIRECT is chosen to make sure that buffering will 47 * happen for small transfers so that the system will not spend all of 48 * its time context switching. PIPE_SIZE is constrained by the 49 * amount of kernel virtual memory. 50 */ 51 52#include <sys/param.h> 53#include <sys/systm.h> 54#include <sys/fcntl.h> 55#include <sys/file.h> 56#include <sys/filedesc.h> 57#include <sys/filio.h> 58#include <sys/kernel.h> 59#include <sys/lock.h> 60#include <sys/mutex.h> 61#include <sys/ttycom.h> 62#include <sys/stat.h> 63#include <sys/malloc.h> 64#include <sys/poll.h> 65#include <sys/selinfo.h> 66#include <sys/signalvar.h> 67#include <sys/sysproto.h> 68#include <sys/pipe.h> 69#include <sys/proc.h> 70#include <sys/vnode.h> 71#include <sys/uio.h> 72#include <sys/event.h> 73 74#include <vm/vm.h> 75#include <vm/vm_param.h> 76#include <vm/vm_object.h> 77#include <vm/vm_kern.h> 78#include <vm/vm_extern.h> 79#include <vm/pmap.h> 80#include <vm/vm_map.h> 81#include <vm/vm_page.h> 82#include <vm/vm_zone.h> 83 84/* 85 * Use this define if you want to disable *fancy* VM things. Expect an 86 * approx 30% decrease in transfer rate. This could be useful for 87 * NetBSD or OpenBSD. 88 */ 89/* #define PIPE_NODIRECT */ 90 91/* 92 * interfaces to the outside world 93 */ 94static int pipe_read(struct file *fp, struct uio *uio, 95 struct ucred *cred, int flags, struct thread *td); 96static int pipe_write(struct file *fp, struct uio *uio, 97 struct ucred *cred, int flags, struct thread *td); 98static int pipe_close(struct file *fp, struct thread *td); 99static int pipe_poll(struct file *fp, int events, struct ucred *cred, 100 struct thread *td); 101static int pipe_kqfilter(struct file *fp, struct knote *kn); 102static int pipe_stat(struct file *fp, struct stat *sb, struct thread *td); 103static int pipe_ioctl(struct file *fp, u_long cmd, caddr_t data, struct thread *td); 104 105static struct fileops pipeops = { 106 pipe_read, pipe_write, pipe_ioctl, pipe_poll, pipe_kqfilter, 107 pipe_stat, pipe_close 108}; 109 110static void filt_pipedetach(struct knote *kn); 111static int filt_piperead(struct knote *kn, long hint); 112static int filt_pipewrite(struct knote *kn, long hint); 113 114static struct filterops pipe_rfiltops = 115 { 1, NULL, filt_pipedetach, filt_piperead }; 116static struct filterops pipe_wfiltops = 117 { 1, NULL, filt_pipedetach, filt_pipewrite }; 118 119#define PIPE_GET_GIANT(pipe) \ 120 do { \ 121 PIPE_UNLOCK(wpipe); \ 122 mtx_lock(&Giant); \ 123 } while (0) 124 125#define PIPE_DROP_GIANT(pipe) \ 126 do { \ 127 mtx_unlock(&Giant); \ 128 PIPE_LOCK(wpipe); \ 129 } while (0) 130 131/* 132 * Default pipe buffer size(s), this can be kind-of large now because pipe 133 * space is pageable. The pipe code will try to maintain locality of 134 * reference for performance reasons, so small amounts of outstanding I/O 135 * will not wipe the cache. 136 */ 137#define MINPIPESIZE (PIPE_SIZE/3) 138#define MAXPIPESIZE (2*PIPE_SIZE/3) 139 140/* 141 * Maximum amount of kva for pipes -- this is kind-of a soft limit, but 142 * is there so that on large systems, we don't exhaust it. 143 */ 144#define MAXPIPEKVA (8*1024*1024) 145 146/* 147 * Limit for direct transfers, we cannot, of course limit 148 * the amount of kva for pipes in general though. 149 */ 150#define LIMITPIPEKVA (16*1024*1024) 151 152/* 153 * Limit the number of "big" pipes 154 */ 155#define LIMITBIGPIPES 32 156static int nbigpipe; 157 158static int amountpipekva; 159 160static void pipeinit(void *dummy __unused); 161static void pipeclose(struct pipe *cpipe); 162static void pipe_free_kmem(struct pipe *cpipe); 163static int pipe_create(struct pipe **cpipep); 164static __inline int pipelock(struct pipe *cpipe, int catch); 165static __inline void pipeunlock(struct pipe *cpipe); 166static __inline void pipeselwakeup(struct pipe *cpipe); 167#ifndef PIPE_NODIRECT 168static int pipe_build_write_buffer(struct pipe *wpipe, struct uio *uio); 169static void pipe_destroy_write_buffer(struct pipe *wpipe); 170static int pipe_direct_write(struct pipe *wpipe, struct uio *uio); 171static void pipe_clone_write_buffer(struct pipe *wpipe); 172#endif 173static int pipespace(struct pipe *cpipe, int size); 174 175static vm_zone_t pipe_zone; 176 177SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_ANY, pipeinit, NULL); 178 179static void 180pipeinit(void *dummy __unused) 181{ 182 183 pipe_zone = zinit("PIPE", sizeof(struct pipe), 0, 0, 4); 184} 185 186/* 187 * The pipe system call for the DTYPE_PIPE type of pipes 188 */ 189 190/* ARGSUSED */ 191int 192pipe(td, uap) 193 struct thread *td; 194 struct pipe_args /* { 195 int dummy; 196 } */ *uap; 197{ 198 struct filedesc *fdp = td->td_proc->p_fd; 199 struct file *rf, *wf; 200 struct pipe *rpipe, *wpipe; 201 struct mtx *pmtx; 202 int fd, error; 203 204 KASSERT(pipe_zone != NULL, ("pipe_zone not initialized")); 205 206 pmtx = malloc(sizeof(*pmtx), M_TEMP, M_WAITOK | M_ZERO); 207 208 rpipe = wpipe = NULL; 209 if (pipe_create(&rpipe) || pipe_create(&wpipe)) { 210 pipeclose(rpipe); 211 pipeclose(wpipe); 212 free(pmtx, M_TEMP); 213 return (ENFILE); 214 } 215 216 rpipe->pipe_state |= PIPE_DIRECTOK; 217 wpipe->pipe_state |= PIPE_DIRECTOK; 218 219 error = falloc(td, &rf, &fd); 220 if (error) { 221 pipeclose(rpipe); 222 pipeclose(wpipe); 223 free(pmtx, M_TEMP); 224 return (error); 225 } 226 fhold(rf); 227 td->td_retval[0] = fd; 228 229 /* 230 * Warning: once we've gotten past allocation of the fd for the 231 * read-side, we can only drop the read side via fdrop() in order 232 * to avoid races against processes which manage to dup() the read 233 * side while we are blocked trying to allocate the write side. 234 */ 235 FILE_LOCK(rf); 236 rf->f_flag = FREAD | FWRITE; 237 rf->f_type = DTYPE_PIPE; 238 rf->f_data = (caddr_t)rpipe; 239 rf->f_ops = &pipeops; 240 FILE_UNLOCK(rf); 241 error = falloc(td, &wf, &fd); 242 if (error) { 243 FILEDESC_LOCK(fdp); 244 if (fdp->fd_ofiles[td->td_retval[0]] == rf) { 245 fdp->fd_ofiles[td->td_retval[0]] = NULL; 246 FILEDESC_UNLOCK(fdp); 247 fdrop(rf, td); 248 } else 249 FILEDESC_UNLOCK(fdp); 250 fdrop(rf, td); 251 /* rpipe has been closed by fdrop(). */ 252 pipeclose(wpipe); 253 free(pmtx, M_TEMP); 254 return (error); 255 } 256 FILE_LOCK(wf); 257 wf->f_flag = FREAD | FWRITE; 258 wf->f_type = DTYPE_PIPE; 259 wf->f_data = (caddr_t)wpipe; 260 wf->f_ops = &pipeops; 261 FILE_UNLOCK(wf); 262 td->td_retval[1] = fd; 263 rpipe->pipe_peer = wpipe; 264 wpipe->pipe_peer = rpipe; 265 mtx_init(pmtx, "pipe mutex", MTX_DEF); 266 rpipe->pipe_mtxp = wpipe->pipe_mtxp = pmtx; 267 fdrop(rf, td); 268 269 return (0); 270} 271 272/* 273 * Allocate kva for pipe circular buffer, the space is pageable 274 * This routine will 'realloc' the size of a pipe safely, if it fails 275 * it will retain the old buffer. 276 * If it fails it will return ENOMEM. 277 */ 278static int 279pipespace(cpipe, size) 280 struct pipe *cpipe; 281 int size; 282{ 283 struct vm_object *object; 284 caddr_t buffer; 285 int npages, error; 286 287 GIANT_REQUIRED; 288 KASSERT(cpipe->pipe_mtxp == NULL || !mtx_owned(PIPE_MTX(cpipe)), 289 ("pipespace: pipe mutex locked")); 290 291 npages = round_page(size)/PAGE_SIZE; 292 /* 293 * Create an object, I don't like the idea of paging to/from 294 * kernel_object. 295 * XXX -- minor change needed here for NetBSD/OpenBSD VM systems. 296 */ 297 object = vm_object_allocate(OBJT_DEFAULT, npages); 298 buffer = (caddr_t) vm_map_min(kernel_map); 299 300 /* 301 * Insert the object into the kernel map, and allocate kva for it. 302 * The map entry is, by default, pageable. 303 * XXX -- minor change needed here for NetBSD/OpenBSD VM systems. 304 */ 305 error = vm_map_find(kernel_map, object, 0, 306 (vm_offset_t *) &buffer, size, 1, 307 VM_PROT_ALL, VM_PROT_ALL, 0); 308 309 if (error != KERN_SUCCESS) { 310 vm_object_deallocate(object); 311 return (ENOMEM); 312 } 313 314 /* free old resources if we're resizing */ 315 pipe_free_kmem(cpipe); 316 cpipe->pipe_buffer.object = object; 317 cpipe->pipe_buffer.buffer = buffer; 318 cpipe->pipe_buffer.size = size; 319 cpipe->pipe_buffer.in = 0; 320 cpipe->pipe_buffer.out = 0; 321 cpipe->pipe_buffer.cnt = 0; 322 amountpipekva += cpipe->pipe_buffer.size; 323 return (0); 324} 325 326/* 327 * initialize and allocate VM and memory for pipe 328 */ 329static int 330pipe_create(cpipep) 331 struct pipe **cpipep; 332{ 333 struct pipe *cpipe; 334 int error; 335 336 *cpipep = zalloc(pipe_zone); 337 if (*cpipep == NULL) 338 return (ENOMEM); 339 340 cpipe = *cpipep; 341 342 /* so pipespace()->pipe_free_kmem() doesn't follow junk pointer */ 343 cpipe->pipe_buffer.object = NULL; 344#ifndef PIPE_NODIRECT 345 cpipe->pipe_map.kva = NULL; 346#endif 347 /* 348 * protect so pipeclose() doesn't follow a junk pointer 349 * if pipespace() fails. 350 */ 351 bzero(&cpipe->pipe_sel, sizeof(cpipe->pipe_sel)); 352 cpipe->pipe_state = 0; 353 cpipe->pipe_peer = NULL; 354 cpipe->pipe_busy = 0; 355 356#ifndef PIPE_NODIRECT 357 /* 358 * pipe data structure initializations to support direct pipe I/O 359 */ 360 cpipe->pipe_map.cnt = 0; 361 cpipe->pipe_map.kva = 0; 362 cpipe->pipe_map.pos = 0; 363 cpipe->pipe_map.npages = 0; 364 /* cpipe->pipe_map.ms[] = invalid */ 365#endif 366 367 cpipe->pipe_mtxp = NULL; /* avoid pipespace assertion */ 368 error = pipespace(cpipe, PIPE_SIZE); 369 if (error) 370 return (error); 371 372 vfs_timestamp(&cpipe->pipe_ctime); 373 cpipe->pipe_atime = cpipe->pipe_ctime; 374 cpipe->pipe_mtime = cpipe->pipe_ctime; 375 376 return (0); 377} 378 379 380/* 381 * lock a pipe for I/O, blocking other access 382 */ 383static __inline int 384pipelock(cpipe, catch) 385 struct pipe *cpipe; 386 int catch; 387{ 388 int error; 389 390 PIPE_LOCK_ASSERT(cpipe, MA_OWNED); 391 while (cpipe->pipe_state & PIPE_LOCKFL) { 392 cpipe->pipe_state |= PIPE_LWANT; 393 error = msleep(cpipe, PIPE_MTX(cpipe), 394 catch ? (PRIBIO | PCATCH) : PRIBIO, 395 "pipelk", 0); 396 if (error != 0) 397 return (error); 398 } 399 cpipe->pipe_state |= PIPE_LOCKFL; 400 return (0); 401} 402 403/* 404 * unlock a pipe I/O lock 405 */ 406static __inline void 407pipeunlock(cpipe) 408 struct pipe *cpipe; 409{ 410 411 PIPE_LOCK_ASSERT(cpipe, MA_OWNED); 412 cpipe->pipe_state &= ~PIPE_LOCKFL; 413 if (cpipe->pipe_state & PIPE_LWANT) { 414 cpipe->pipe_state &= ~PIPE_LWANT; 415 wakeup(cpipe); 416 } 417} 418 419static __inline void 420pipeselwakeup(cpipe) 421 struct pipe *cpipe; 422{ 423 424 if (cpipe->pipe_state & PIPE_SEL) { 425 cpipe->pipe_state &= ~PIPE_SEL; 426 selwakeup(&cpipe->pipe_sel); 427 } 428 if ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) 429 pgsigio(cpipe->pipe_sigio, SIGIO, 0); 430 KNOTE(&cpipe->pipe_sel.si_note, 0); 431} 432 433/* ARGSUSED */ 434static int 435pipe_read(fp, uio, cred, flags, td) 436 struct file *fp; 437 struct uio *uio; 438 struct ucred *cred; 439 struct thread *td; 440 int flags; 441{ 442 struct pipe *rpipe = (struct pipe *) fp->f_data; 443 int error; 444 int nread = 0; 445 u_int size; 446 447 PIPE_LOCK(rpipe); 448 ++rpipe->pipe_busy; 449 error = pipelock(rpipe, 1); 450 if (error) 451 goto unlocked_error; 452 453 while (uio->uio_resid) { 454 /* 455 * normal pipe buffer receive 456 */ 457 if (rpipe->pipe_buffer.cnt > 0) { 458 size = rpipe->pipe_buffer.size - rpipe->pipe_buffer.out; 459 if (size > rpipe->pipe_buffer.cnt) 460 size = rpipe->pipe_buffer.cnt; 461 if (size > (u_int) uio->uio_resid) 462 size = (u_int) uio->uio_resid; 463 464 PIPE_UNLOCK(rpipe); 465 error = uiomove(&rpipe->pipe_buffer.buffer[rpipe->pipe_buffer.out], 466 size, uio); 467 PIPE_LOCK(rpipe); 468 if (error) 469 break; 470 471 rpipe->pipe_buffer.out += size; 472 if (rpipe->pipe_buffer.out >= rpipe->pipe_buffer.size) 473 rpipe->pipe_buffer.out = 0; 474 475 rpipe->pipe_buffer.cnt -= size; 476 477 /* 478 * If there is no more to read in the pipe, reset 479 * its pointers to the beginning. This improves 480 * cache hit stats. 481 */ 482 if (rpipe->pipe_buffer.cnt == 0) { 483 rpipe->pipe_buffer.in = 0; 484 rpipe->pipe_buffer.out = 0; 485 } 486 nread += size; 487#ifndef PIPE_NODIRECT 488 /* 489 * Direct copy, bypassing a kernel buffer. 490 */ 491 } else if ((size = rpipe->pipe_map.cnt) && 492 (rpipe->pipe_state & PIPE_DIRECTW)) { 493 caddr_t va; 494 if (size > (u_int) uio->uio_resid) 495 size = (u_int) uio->uio_resid; 496 497 va = (caddr_t) rpipe->pipe_map.kva + 498 rpipe->pipe_map.pos; 499 PIPE_UNLOCK(rpipe); 500 error = uiomove(va, size, uio); 501 PIPE_LOCK(rpipe); 502 if (error) 503 break; 504 nread += size; 505 rpipe->pipe_map.pos += size; 506 rpipe->pipe_map.cnt -= size; 507 if (rpipe->pipe_map.cnt == 0) { 508 rpipe->pipe_state &= ~PIPE_DIRECTW; 509 wakeup(rpipe); 510 } 511#endif 512 } else { 513 /* 514 * detect EOF condition 515 * read returns 0 on EOF, no need to set error 516 */ 517 if (rpipe->pipe_state & PIPE_EOF) 518 break; 519 520 /* 521 * If the "write-side" has been blocked, wake it up now. 522 */ 523 if (rpipe->pipe_state & PIPE_WANTW) { 524 rpipe->pipe_state &= ~PIPE_WANTW; 525 wakeup(rpipe); 526 } 527 528 /* 529 * Break if some data was read. 530 */ 531 if (nread > 0) 532 break; 533 534 /* 535 * Unlock the pipe buffer for our remaining processing. We 536 * will either break out with an error or we will sleep and 537 * relock to loop. 538 */ 539 pipeunlock(rpipe); 540 541 /* 542 * Handle non-blocking mode operation or 543 * wait for more data. 544 */ 545 if (fp->f_flag & FNONBLOCK) { 546 error = EAGAIN; 547 } else { 548 rpipe->pipe_state |= PIPE_WANTR; 549 if ((error = msleep(rpipe, PIPE_MTX(rpipe), 550 PRIBIO | PCATCH, 551 "piperd", 0)) == 0) 552 error = pipelock(rpipe, 1); 553 } 554 if (error) 555 goto unlocked_error; 556 } 557 } 558 pipeunlock(rpipe); 559 560 /* XXX: should probably do this before getting any locks. */ 561 if (error == 0) 562 vfs_timestamp(&rpipe->pipe_atime); 563unlocked_error: 564 --rpipe->pipe_busy; 565 566 /* 567 * PIPE_WANT processing only makes sense if pipe_busy is 0. 568 */ 569 if ((rpipe->pipe_busy == 0) && (rpipe->pipe_state & PIPE_WANT)) { 570 rpipe->pipe_state &= ~(PIPE_WANT|PIPE_WANTW); 571 wakeup(rpipe); 572 } else if (rpipe->pipe_buffer.cnt < MINPIPESIZE) { 573 /* 574 * Handle write blocking hysteresis. 575 */ 576 if (rpipe->pipe_state & PIPE_WANTW) { 577 rpipe->pipe_state &= ~PIPE_WANTW; 578 wakeup(rpipe); 579 } 580 } 581 582 if ((rpipe->pipe_buffer.size - rpipe->pipe_buffer.cnt) >= PIPE_BUF) 583 pipeselwakeup(rpipe); 584 585 PIPE_UNLOCK(rpipe); 586 return (error); 587} 588 589#ifndef PIPE_NODIRECT 590/* 591 * Map the sending processes' buffer into kernel space and wire it. 592 * This is similar to a physical write operation. 593 */ 594static int 595pipe_build_write_buffer(wpipe, uio) 596 struct pipe *wpipe; 597 struct uio *uio; 598{ 599 u_int size; 600 int i; 601 vm_offset_t addr, endaddr, paddr; 602 603 GIANT_REQUIRED; 604 PIPE_LOCK_ASSERT(wpipe, MA_NOTOWNED); 605 606 size = (u_int) uio->uio_iov->iov_len; 607 if (size > wpipe->pipe_buffer.size) 608 size = wpipe->pipe_buffer.size; 609 610 endaddr = round_page((vm_offset_t)uio->uio_iov->iov_base + size); 611 addr = trunc_page((vm_offset_t)uio->uio_iov->iov_base); 612 for (i = 0; addr < endaddr; addr += PAGE_SIZE, i++) { 613 vm_page_t m; 614 615 if (vm_fault_quick((caddr_t)addr, VM_PROT_READ) < 0 || 616 (paddr = pmap_kextract(addr)) == 0) { 617 int j; 618 619 for (j = 0; j < i; j++) 620 vm_page_unwire(wpipe->pipe_map.ms[j], 1); 621 return (EFAULT); 622 } 623 624 m = PHYS_TO_VM_PAGE(paddr); 625 vm_page_wire(m); 626 wpipe->pipe_map.ms[i] = m; 627 } 628 629/* 630 * set up the control block 631 */ 632 wpipe->pipe_map.npages = i; 633 wpipe->pipe_map.pos = 634 ((vm_offset_t) uio->uio_iov->iov_base) & PAGE_MASK; 635 wpipe->pipe_map.cnt = size; 636 637/* 638 * and map the buffer 639 */ 640 if (wpipe->pipe_map.kva == 0) { 641 /* 642 * We need to allocate space for an extra page because the 643 * address range might (will) span pages at times. 644 */ 645 wpipe->pipe_map.kva = kmem_alloc_pageable(kernel_map, 646 wpipe->pipe_buffer.size + PAGE_SIZE); 647 amountpipekva += wpipe->pipe_buffer.size + PAGE_SIZE; 648 } 649 pmap_qenter(wpipe->pipe_map.kva, wpipe->pipe_map.ms, 650 wpipe->pipe_map.npages); 651 652/* 653 * and update the uio data 654 */ 655 656 uio->uio_iov->iov_len -= size; 657 uio->uio_iov->iov_base += size; 658 if (uio->uio_iov->iov_len == 0) 659 uio->uio_iov++; 660 uio->uio_resid -= size; 661 uio->uio_offset += size; 662 return (0); 663} 664 665/* 666 * unmap and unwire the process buffer 667 */ 668static void 669pipe_destroy_write_buffer(wpipe) 670 struct pipe *wpipe; 671{ 672 int i; 673 674 GIANT_REQUIRED; 675 PIPE_LOCK_ASSERT(wpipe, MA_NOTOWNED); 676 677 if (wpipe->pipe_map.kva) { 678 pmap_qremove(wpipe->pipe_map.kva, wpipe->pipe_map.npages); 679 680 if (amountpipekva > MAXPIPEKVA) { 681 vm_offset_t kva = wpipe->pipe_map.kva; 682 wpipe->pipe_map.kva = 0; 683 kmem_free(kernel_map, kva, 684 wpipe->pipe_buffer.size + PAGE_SIZE); 685 amountpipekva -= wpipe->pipe_buffer.size + PAGE_SIZE; 686 } 687 } 688 for (i = 0; i < wpipe->pipe_map.npages; i++) 689 vm_page_unwire(wpipe->pipe_map.ms[i], 1); 690 wpipe->pipe_map.npages = 0; 691} 692 693/* 694 * In the case of a signal, the writing process might go away. This 695 * code copies the data into the circular buffer so that the source 696 * pages can be freed without loss of data. 697 */ 698static void 699pipe_clone_write_buffer(wpipe) 700 struct pipe *wpipe; 701{ 702 int size; 703 int pos; 704 705 PIPE_LOCK_ASSERT(wpipe, MA_OWNED); 706 size = wpipe->pipe_map.cnt; 707 pos = wpipe->pipe_map.pos; 708 bcopy((caddr_t) wpipe->pipe_map.kva + pos, 709 (caddr_t) wpipe->pipe_buffer.buffer, size); 710 711 wpipe->pipe_buffer.in = size; 712 wpipe->pipe_buffer.out = 0; 713 wpipe->pipe_buffer.cnt = size; 714 wpipe->pipe_state &= ~PIPE_DIRECTW; 715 716 PIPE_GET_GIANT(wpipe); 717 pipe_destroy_write_buffer(wpipe); 718 PIPE_DROP_GIANT(wpipe); 719} 720 721/* 722 * This implements the pipe buffer write mechanism. Note that only 723 * a direct write OR a normal pipe write can be pending at any given time. 724 * If there are any characters in the pipe buffer, the direct write will 725 * be deferred until the receiving process grabs all of the bytes from 726 * the pipe buffer. Then the direct mapping write is set-up. 727 */ 728static int 729pipe_direct_write(wpipe, uio) 730 struct pipe *wpipe; 731 struct uio *uio; 732{ 733 int error; 734 735retry: 736 PIPE_LOCK_ASSERT(wpipe, MA_OWNED); 737 while (wpipe->pipe_state & PIPE_DIRECTW) { 738 if (wpipe->pipe_state & PIPE_WANTR) { 739 wpipe->pipe_state &= ~PIPE_WANTR; 740 wakeup(wpipe); 741 } 742 wpipe->pipe_state |= PIPE_WANTW; 743 error = msleep(wpipe, PIPE_MTX(wpipe), 744 PRIBIO | PCATCH, "pipdww", 0); 745 if (error) 746 goto error1; 747 if (wpipe->pipe_state & PIPE_EOF) { 748 error = EPIPE; 749 goto error1; 750 } 751 } 752 wpipe->pipe_map.cnt = 0; /* transfer not ready yet */ 753 if (wpipe->pipe_buffer.cnt > 0) { 754 if (wpipe->pipe_state & PIPE_WANTR) { 755 wpipe->pipe_state &= ~PIPE_WANTR; 756 wakeup(wpipe); 757 } 758 759 wpipe->pipe_state |= PIPE_WANTW; 760 error = msleep(wpipe, PIPE_MTX(wpipe), 761 PRIBIO | PCATCH, "pipdwc", 0); 762 if (error) 763 goto error1; 764 if (wpipe->pipe_state & PIPE_EOF) { 765 error = EPIPE; 766 goto error1; 767 } 768 goto retry; 769 } 770 771 wpipe->pipe_state |= PIPE_DIRECTW; 772 773 PIPE_GET_GIANT(wpipe); 774 error = pipe_build_write_buffer(wpipe, uio); 775 PIPE_DROP_GIANT(wpipe); 776 if (error) { 777 wpipe->pipe_state &= ~PIPE_DIRECTW; 778 goto error1; 779 } 780 781 error = 0; 782 while (!error && (wpipe->pipe_state & PIPE_DIRECTW)) { 783 if (wpipe->pipe_state & PIPE_EOF) { 784 pipelock(wpipe, 0); 785 PIPE_GET_GIANT(wpipe); 786 pipe_destroy_write_buffer(wpipe); 787 PIPE_DROP_GIANT(wpipe); 788 pipeunlock(wpipe); 789 pipeselwakeup(wpipe); 790 error = EPIPE; 791 goto error1; 792 } 793 if (wpipe->pipe_state & PIPE_WANTR) { 794 wpipe->pipe_state &= ~PIPE_WANTR; 795 wakeup(wpipe); 796 } 797 pipeselwakeup(wpipe); 798 error = msleep(wpipe, PIPE_MTX(wpipe), PRIBIO | PCATCH, 799 "pipdwt", 0); 800 } 801 802 pipelock(wpipe,0); 803 if (wpipe->pipe_state & PIPE_DIRECTW) { 804 /* 805 * this bit of trickery substitutes a kernel buffer for 806 * the process that might be going away. 807 */ 808 pipe_clone_write_buffer(wpipe); 809 } else { 810 PIPE_GET_GIANT(wpipe); 811 pipe_destroy_write_buffer(wpipe); 812 PIPE_DROP_GIANT(wpipe); 813 } 814 pipeunlock(wpipe); 815 return (error); 816 817error1: 818 wakeup(wpipe); 819 return (error); 820} 821#endif 822 823static int 824pipe_write(fp, uio, cred, flags, td) 825 struct file *fp; 826 struct uio *uio; 827 struct ucred *cred; 828 struct thread *td; 829 int flags; 830{ 831 int error = 0; 832 int orig_resid; 833 struct pipe *wpipe, *rpipe; 834 835 rpipe = (struct pipe *) fp->f_data; 836 wpipe = rpipe->pipe_peer; 837 838 PIPE_LOCK(rpipe); 839 /* 840 * detect loss of pipe read side, issue SIGPIPE if lost. 841 */ 842 if ((wpipe == NULL) || (wpipe->pipe_state & PIPE_EOF)) { 843 PIPE_UNLOCK(rpipe); 844 return (EPIPE); 845 } 846 ++wpipe->pipe_busy; 847 848 /* 849 * If it is advantageous to resize the pipe buffer, do 850 * so. 851 */ 852 if ((uio->uio_resid > PIPE_SIZE) && 853 (nbigpipe < LIMITBIGPIPES) && 854 (wpipe->pipe_state & PIPE_DIRECTW) == 0 && 855 (wpipe->pipe_buffer.size <= PIPE_SIZE) && 856 (wpipe->pipe_buffer.cnt == 0)) { 857 858 if ((error = pipelock(wpipe,1)) == 0) { 859 PIPE_GET_GIANT(rpipe); 860 if (pipespace(wpipe, BIG_PIPE_SIZE) == 0) 861 nbigpipe++; 862 PIPE_DROP_GIANT(rpipe); 863 pipeunlock(wpipe); 864 } 865 } 866 867 /* 868 * If an early error occured unbusy and return, waking up any pending 869 * readers. 870 */ 871 if (error) { 872 --wpipe->pipe_busy; 873 if ((wpipe->pipe_busy == 0) && 874 (wpipe->pipe_state & PIPE_WANT)) { 875 wpipe->pipe_state &= ~(PIPE_WANT | PIPE_WANTR); 876 wakeup(wpipe); 877 } 878 PIPE_UNLOCK(rpipe); 879 return(error); 880 } 881 882 KASSERT(wpipe->pipe_buffer.buffer != NULL, ("pipe buffer gone")); 883 884 orig_resid = uio->uio_resid; 885 886 while (uio->uio_resid) { 887 int space; 888 889#ifndef PIPE_NODIRECT 890 /* 891 * If the transfer is large, we can gain performance if 892 * we do process-to-process copies directly. 893 * If the write is non-blocking, we don't use the 894 * direct write mechanism. 895 * 896 * The direct write mechanism will detect the reader going 897 * away on us. 898 */ 899 if ((uio->uio_iov->iov_len >= PIPE_MINDIRECT) && 900 (fp->f_flag & FNONBLOCK) == 0 && 901 (wpipe->pipe_map.kva || (amountpipekva < LIMITPIPEKVA)) && 902 (uio->uio_iov->iov_len >= PIPE_MINDIRECT)) { 903 error = pipe_direct_write( wpipe, uio); 904 if (error) 905 break; 906 continue; 907 } 908#endif 909 910 /* 911 * Pipe buffered writes cannot be coincidental with 912 * direct writes. We wait until the currently executing 913 * direct write is completed before we start filling the 914 * pipe buffer. We break out if a signal occurs or the 915 * reader goes away. 916 */ 917 retrywrite: 918 while (wpipe->pipe_state & PIPE_DIRECTW) { 919 if (wpipe->pipe_state & PIPE_WANTR) { 920 wpipe->pipe_state &= ~PIPE_WANTR; 921 wakeup(wpipe); 922 } 923 error = msleep(wpipe, PIPE_MTX(rpipe), PRIBIO | PCATCH, 924 "pipbww", 0); 925 if (wpipe->pipe_state & PIPE_EOF) 926 break; 927 if (error) 928 break; 929 } 930 if (wpipe->pipe_state & PIPE_EOF) { 931 error = EPIPE; 932 break; 933 } 934 935 space = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt; 936 937 /* Writes of size <= PIPE_BUF must be atomic. */ 938 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF)) 939 space = 0; 940 941 if (space > 0 && (wpipe->pipe_buffer.cnt < PIPE_SIZE)) { 942 if ((error = pipelock(wpipe,1)) == 0) { 943 int size; /* Transfer size */ 944 int segsize; /* first segment to transfer */ 945 946 /* 947 * It is possible for a direct write to 948 * slip in on us... handle it here... 949 */ 950 if (wpipe->pipe_state & PIPE_DIRECTW) { 951 pipeunlock(wpipe); 952 goto retrywrite; 953 } 954 /* 955 * If a process blocked in uiomove, our 956 * value for space might be bad. 957 * 958 * XXX will we be ok if the reader has gone 959 * away here? 960 */ 961 if (space > wpipe->pipe_buffer.size - 962 wpipe->pipe_buffer.cnt) { 963 pipeunlock(wpipe); 964 goto retrywrite; 965 } 966 967 /* 968 * Transfer size is minimum of uio transfer 969 * and free space in pipe buffer. 970 */ 971 if (space > uio->uio_resid) 972 size = uio->uio_resid; 973 else 974 size = space; 975 /* 976 * First segment to transfer is minimum of 977 * transfer size and contiguous space in 978 * pipe buffer. If first segment to transfer 979 * is less than the transfer size, we've got 980 * a wraparound in the buffer. 981 */ 982 segsize = wpipe->pipe_buffer.size - 983 wpipe->pipe_buffer.in; 984 if (segsize > size) 985 segsize = size; 986 987 /* Transfer first segment */ 988 989 PIPE_UNLOCK(rpipe); 990 error = uiomove(&wpipe->pipe_buffer.buffer[wpipe->pipe_buffer.in], 991 segsize, uio); 992 PIPE_LOCK(rpipe); 993 994 if (error == 0 && segsize < size) { 995 /* 996 * Transfer remaining part now, to 997 * support atomic writes. Wraparound 998 * happened. 999 */ 1000 if (wpipe->pipe_buffer.in + segsize != 1001 wpipe->pipe_buffer.size) 1002 panic("Expected pipe buffer wraparound disappeared"); 1003 1004 PIPE_UNLOCK(rpipe); 1005 error = uiomove(&wpipe->pipe_buffer.buffer[0], 1006 size - segsize, uio); 1007 PIPE_LOCK(rpipe); 1008 } 1009 if (error == 0) { 1010 wpipe->pipe_buffer.in += size; 1011 if (wpipe->pipe_buffer.in >= 1012 wpipe->pipe_buffer.size) { 1013 if (wpipe->pipe_buffer.in != size - segsize + wpipe->pipe_buffer.size) 1014 panic("Expected wraparound bad"); 1015 wpipe->pipe_buffer.in = size - segsize; 1016 } 1017 1018 wpipe->pipe_buffer.cnt += size; 1019 if (wpipe->pipe_buffer.cnt > wpipe->pipe_buffer.size) 1020 panic("Pipe buffer overflow"); 1021 1022 } 1023 pipeunlock(wpipe); 1024 } 1025 if (error) 1026 break; 1027 1028 } else { 1029 /* 1030 * If the "read-side" has been blocked, wake it up now. 1031 */ 1032 if (wpipe->pipe_state & PIPE_WANTR) { 1033 wpipe->pipe_state &= ~PIPE_WANTR; 1034 wakeup(wpipe); 1035 } 1036 1037 /* 1038 * don't block on non-blocking I/O 1039 */ 1040 if (fp->f_flag & FNONBLOCK) { 1041 error = EAGAIN; 1042 break; 1043 } 1044 1045 /* 1046 * We have no more space and have something to offer, 1047 * wake up select/poll. 1048 */ 1049 pipeselwakeup(wpipe); 1050 1051 wpipe->pipe_state |= PIPE_WANTW; 1052 error = msleep(wpipe, PIPE_MTX(rpipe), 1053 PRIBIO | PCATCH, "pipewr", 0); 1054 if (error != 0) 1055 break; 1056 /* 1057 * If read side wants to go away, we just issue a signal 1058 * to ourselves. 1059 */ 1060 if (wpipe->pipe_state & PIPE_EOF) { 1061 error = EPIPE; 1062 break; 1063 } 1064 } 1065 } 1066 1067 --wpipe->pipe_busy; 1068 1069 if ((wpipe->pipe_busy == 0) && (wpipe->pipe_state & PIPE_WANT)) { 1070 wpipe->pipe_state &= ~(PIPE_WANT | PIPE_WANTR); 1071 wakeup(wpipe); 1072 } else if (wpipe->pipe_buffer.cnt > 0) { 1073 /* 1074 * If we have put any characters in the buffer, we wake up 1075 * the reader. 1076 */ 1077 if (wpipe->pipe_state & PIPE_WANTR) { 1078 wpipe->pipe_state &= ~PIPE_WANTR; 1079 wakeup(wpipe); 1080 } 1081 } 1082 1083 /* 1084 * Don't return EPIPE if I/O was successful 1085 */ 1086 if ((wpipe->pipe_buffer.cnt == 0) && 1087 (uio->uio_resid == 0) && 1088 (error == EPIPE)) { 1089 error = 0; 1090 } 1091 1092 if (error == 0) 1093 vfs_timestamp(&wpipe->pipe_mtime); 1094 1095 /* 1096 * We have something to offer, 1097 * wake up select/poll. 1098 */ 1099 if (wpipe->pipe_buffer.cnt) 1100 pipeselwakeup(wpipe); 1101 1102 PIPE_UNLOCK(rpipe); 1103 return (error); 1104} 1105 1106/* 1107 * we implement a very minimal set of ioctls for compatibility with sockets. 1108 */ 1109int 1110pipe_ioctl(fp, cmd, data, td) 1111 struct file *fp; 1112 u_long cmd; 1113 caddr_t data; 1114 struct thread *td; 1115{ 1116 struct pipe *mpipe = (struct pipe *)fp->f_data; 1117 1118 switch (cmd) { 1119 1120 case FIONBIO: 1121 return (0); 1122 1123 case FIOASYNC: 1124 PIPE_LOCK(mpipe); 1125 if (*(int *)data) { 1126 mpipe->pipe_state |= PIPE_ASYNC; 1127 } else { 1128 mpipe->pipe_state &= ~PIPE_ASYNC; 1129 } 1130 PIPE_UNLOCK(mpipe); 1131 return (0); 1132 1133 case FIONREAD: 1134 PIPE_LOCK(mpipe); 1135 if (mpipe->pipe_state & PIPE_DIRECTW) 1136 *(int *)data = mpipe->pipe_map.cnt; 1137 else 1138 *(int *)data = mpipe->pipe_buffer.cnt; 1139 PIPE_UNLOCK(mpipe); 1140 return (0); 1141 1142 case FIOSETOWN: 1143 return (fsetown(*(int *)data, &mpipe->pipe_sigio)); 1144 1145 case FIOGETOWN: 1146 *(int *)data = fgetown(mpipe->pipe_sigio); 1147 return (0); 1148 1149 /* This is deprecated, FIOSETOWN should be used instead. */ 1150 case TIOCSPGRP: 1151 return (fsetown(-(*(int *)data), &mpipe->pipe_sigio)); 1152 1153 /* This is deprecated, FIOGETOWN should be used instead. */ 1154 case TIOCGPGRP: 1155 *(int *)data = -fgetown(mpipe->pipe_sigio); 1156 return (0); 1157 1158 } 1159 return (ENOTTY); 1160} 1161 1162int 1163pipe_poll(fp, events, cred, td) 1164 struct file *fp; 1165 int events; 1166 struct ucred *cred; 1167 struct thread *td; 1168{ 1169 struct pipe *rpipe = (struct pipe *)fp->f_data; 1170 struct pipe *wpipe; 1171 int revents = 0; 1172 1173 wpipe = rpipe->pipe_peer; 1174 PIPE_LOCK(rpipe); 1175 if (events & (POLLIN | POLLRDNORM)) 1176 if ((rpipe->pipe_state & PIPE_DIRECTW) || 1177 (rpipe->pipe_buffer.cnt > 0) || 1178 (rpipe->pipe_state & PIPE_EOF)) 1179 revents |= events & (POLLIN | POLLRDNORM); 1180 1181 if (events & (POLLOUT | POLLWRNORM)) 1182 if (wpipe == NULL || (wpipe->pipe_state & PIPE_EOF) || 1183 (((wpipe->pipe_state & PIPE_DIRECTW) == 0) && 1184 (wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt) >= PIPE_BUF)) 1185 revents |= events & (POLLOUT | POLLWRNORM); 1186 1187 if ((rpipe->pipe_state & PIPE_EOF) || 1188 (wpipe == NULL) || 1189 (wpipe->pipe_state & PIPE_EOF)) 1190 revents |= POLLHUP; 1191 1192 if (revents == 0) { 1193 if (events & (POLLIN | POLLRDNORM)) { 1194 selrecord(td, &rpipe->pipe_sel); 1195 rpipe->pipe_state |= PIPE_SEL; 1196 } 1197 1198 if (events & (POLLOUT | POLLWRNORM)) { 1199 selrecord(td, &wpipe->pipe_sel); 1200 wpipe->pipe_state |= PIPE_SEL; 1201 } 1202 } 1203 PIPE_UNLOCK(rpipe); 1204 1205 return (revents); 1206} 1207 1208static int 1209pipe_stat(fp, ub, td) 1210 struct file *fp; 1211 struct stat *ub; 1212 struct thread *td; 1213{ 1214 struct pipe *pipe = (struct pipe *)fp->f_data; 1215 1216 bzero((caddr_t)ub, sizeof(*ub)); 1217 ub->st_mode = S_IFIFO; 1218 ub->st_blksize = pipe->pipe_buffer.size; 1219 ub->st_size = pipe->pipe_buffer.cnt; 1220 ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize; 1221 ub->st_atimespec = pipe->pipe_atime; 1222 ub->st_mtimespec = pipe->pipe_mtime; 1223 ub->st_ctimespec = pipe->pipe_ctime; 1224 ub->st_uid = fp->f_cred->cr_uid; 1225 ub->st_gid = fp->f_cred->cr_gid; 1226 /* 1227 * Left as 0: st_dev, st_ino, st_nlink, st_rdev, st_flags, st_gen. 1228 * XXX (st_dev, st_ino) should be unique. 1229 */ 1230 return (0); 1231} 1232 1233/* ARGSUSED */ 1234static int 1235pipe_close(fp, td) 1236 struct file *fp; 1237 struct thread *td; 1238{ 1239 struct pipe *cpipe = (struct pipe *)fp->f_data; 1240 1241 fp->f_ops = &badfileops; 1242 fp->f_data = NULL; 1243 funsetown(cpipe->pipe_sigio); 1244 pipeclose(cpipe); 1245 return (0); 1246} 1247 1248static void 1249pipe_free_kmem(cpipe) 1250 struct pipe *cpipe; 1251{ 1252 1253 GIANT_REQUIRED; 1254 KASSERT(cpipe->pipe_mtxp == NULL || !mtx_owned(PIPE_MTX(cpipe)), 1255 ("pipespace: pipe mutex locked")); 1256 1257 if (cpipe->pipe_buffer.buffer != NULL) { 1258 if (cpipe->pipe_buffer.size > PIPE_SIZE) 1259 --nbigpipe; 1260 amountpipekva -= cpipe->pipe_buffer.size; 1261 kmem_free(kernel_map, 1262 (vm_offset_t)cpipe->pipe_buffer.buffer, 1263 cpipe->pipe_buffer.size); 1264 cpipe->pipe_buffer.buffer = NULL; 1265 } 1266#ifndef PIPE_NODIRECT 1267 if (cpipe->pipe_map.kva != NULL) { 1268 amountpipekva -= cpipe->pipe_buffer.size + PAGE_SIZE; 1269 kmem_free(kernel_map, 1270 cpipe->pipe_map.kva, 1271 cpipe->pipe_buffer.size + PAGE_SIZE); 1272 cpipe->pipe_map.cnt = 0; 1273 cpipe->pipe_map.kva = 0; 1274 cpipe->pipe_map.pos = 0; 1275 cpipe->pipe_map.npages = 0; 1276 } 1277#endif 1278} 1279 1280/* 1281 * shutdown the pipe 1282 */ 1283static void 1284pipeclose(cpipe) 1285 struct pipe *cpipe; 1286{ 1287 struct pipe *ppipe; 1288 int hadpeer; 1289 1290 if (cpipe == NULL) 1291 return; 1292 1293 hadpeer = 0; 1294 1295 /* partially created pipes won't have a valid mutex. */ 1296 if (PIPE_MTX(cpipe) != NULL) 1297 PIPE_LOCK(cpipe); 1298 1299 pipeselwakeup(cpipe); 1300 1301 /* 1302 * If the other side is blocked, wake it up saying that 1303 * we want to close it down. 1304 */ 1305 while (cpipe->pipe_busy) { 1306 wakeup(cpipe); 1307 cpipe->pipe_state |= PIPE_WANT | PIPE_EOF; 1308 msleep(cpipe, PIPE_MTX(cpipe), PRIBIO, "pipecl", 0); 1309 } 1310 1311 /* 1312 * Disconnect from peer 1313 */ 1314 if ((ppipe = cpipe->pipe_peer) != NULL) { 1315 hadpeer++; 1316 pipeselwakeup(ppipe); 1317 1318 ppipe->pipe_state |= PIPE_EOF; 1319 wakeup(ppipe); 1320 KNOTE(&ppipe->pipe_sel.si_note, 0); 1321 ppipe->pipe_peer = NULL; 1322 } 1323 /* 1324 * free resources 1325 */ 1326 if (PIPE_MTX(cpipe) != NULL) { 1327 PIPE_UNLOCK(cpipe); 1328 if (!hadpeer) { 1329 mtx_destroy(PIPE_MTX(cpipe)); 1330 free(PIPE_MTX(cpipe), M_TEMP); 1331 } 1332 } 1333 mtx_lock(&Giant); 1334 pipe_free_kmem(cpipe); 1335 zfree(pipe_zone, cpipe); 1336 mtx_unlock(&Giant); 1337} 1338 1339/*ARGSUSED*/ 1340static int 1341pipe_kqfilter(struct file *fp, struct knote *kn) 1342{ 1343 struct pipe *cpipe; 1344 1345 cpipe = (struct pipe *)kn->kn_fp->f_data; 1346 switch (kn->kn_filter) { 1347 case EVFILT_READ: 1348 kn->kn_fop = &pipe_rfiltops; 1349 break; 1350 case EVFILT_WRITE: 1351 kn->kn_fop = &pipe_wfiltops; 1352 cpipe = cpipe->pipe_peer; 1353 break; 1354 default: 1355 return (1); 1356 } 1357 kn->kn_hook = (caddr_t)cpipe; 1358 1359 PIPE_LOCK(cpipe); 1360 SLIST_INSERT_HEAD(&cpipe->pipe_sel.si_note, kn, kn_selnext); 1361 PIPE_UNLOCK(cpipe); 1362 return (0); 1363} 1364 1365static void 1366filt_pipedetach(struct knote *kn) 1367{ 1368 struct pipe *cpipe = (struct pipe *)kn->kn_hook; 1369 1370 PIPE_LOCK(cpipe); 1371 SLIST_REMOVE(&cpipe->pipe_sel.si_note, kn, knote, kn_selnext); 1372 PIPE_UNLOCK(cpipe); 1373} 1374 1375/*ARGSUSED*/ 1376static int 1377filt_piperead(struct knote *kn, long hint) 1378{ 1379 struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data; 1380 struct pipe *wpipe = rpipe->pipe_peer; 1381 1382 PIPE_LOCK(rpipe); 1383 kn->kn_data = rpipe->pipe_buffer.cnt; 1384 if ((kn->kn_data == 0) && (rpipe->pipe_state & PIPE_DIRECTW)) 1385 kn->kn_data = rpipe->pipe_map.cnt; 1386 1387 if ((rpipe->pipe_state & PIPE_EOF) || 1388 (wpipe == NULL) || (wpipe->pipe_state & PIPE_EOF)) { 1389 kn->kn_flags |= EV_EOF; 1390 PIPE_UNLOCK(rpipe); 1391 return (1); 1392 } 1393 PIPE_UNLOCK(rpipe); 1394 return (kn->kn_data > 0); 1395} 1396 1397/*ARGSUSED*/ 1398static int 1399filt_pipewrite(struct knote *kn, long hint) 1400{ 1401 struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data; 1402 struct pipe *wpipe = rpipe->pipe_peer; 1403 1404 PIPE_LOCK(rpipe); 1405 if ((wpipe == NULL) || (wpipe->pipe_state & PIPE_EOF)) { 1406 kn->kn_data = 0; 1407 kn->kn_flags |= EV_EOF; 1408 PIPE_UNLOCK(rpipe); 1409 return (1); 1410 } 1411 kn->kn_data = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt; 1412 if (wpipe->pipe_state & PIPE_DIRECTW) 1413 kn->kn_data = 0; 1414 1415 PIPE_UNLOCK(rpipe); 1416 return (kn->kn_data >= PIPE_BUF); 1417} 1418