1 /* 2 * Copyright (c) 1996 John S. Dyson 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice immediately at the beginning of the file, without modification, 10 * this list of conditions, and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. Absolutely no warranty of function or purpose is made by the author 15 * John S. Dyson. 16 * 4. Modifications may be freely made to this file if the above conditions 17 * are met. 18 * 19 * $FreeBSD: src/sys/kern/sys_pipe.c,v 1.60.2.13 2002/08/05 15:05:15 des Exp $ 20 * $DragonFly: src/sys/kern/sys_pipe.c,v 1.50 2008/09/09 04:06:13 dillon Exp $ 21 */ 22 23 /* 24 * This file contains a high-performance replacement for the socket-based 25 * pipes scheme originally used in FreeBSD/4.4Lite. It does not support 26 * all features of sockets, but does do everything that pipes normally 27 * do. 28 */ 29 #include <sys/param.h> 30 #include <sys/systm.h> 31 #include <sys/kernel.h> 32 #include <sys/proc.h> 33 #include <sys/fcntl.h> 34 #include <sys/file.h> 35 #include <sys/filedesc.h> 36 #include <sys/filio.h> 37 #include <sys/ttycom.h> 38 #include <sys/stat.h> 39 #include <sys/signalvar.h> 40 #include <sys/sysproto.h> 41 #include <sys/pipe.h> 42 #include <sys/vnode.h> 43 #include <sys/uio.h> 44 #include <sys/event.h> 45 #include <sys/globaldata.h> 46 #include <sys/module.h> 47 #include <sys/malloc.h> 48 #include <sys/sysctl.h> 49 #include <sys/socket.h> 50 51 #include <vm/vm.h> 52 #include <vm/vm_param.h> 53 #include <sys/lock.h> 54 #include <vm/vm_object.h> 55 #include <vm/vm_kern.h> 56 #include <vm/vm_extern.h> 57 #include <vm/pmap.h> 58 #include <vm/vm_map.h> 59 #include <vm/vm_page.h> 60 #include <vm/vm_zone.h> 61 62 #include <sys/file2.h> 63 #include <sys/signal2.h> 64 #include <sys/mplock2.h> 65 66 #include <machine/cpufunc.h> 67 68 /* 69 * interfaces to the outside world 70 */ 71 static int pipe_read (struct file *fp, struct uio *uio, 72 struct ucred *cred, int flags); 73 static int pipe_write (struct file *fp, struct uio *uio, 74 struct ucred *cred, int flags); 75 static int pipe_close (struct file *fp); 76 static int pipe_shutdown (struct file *fp, int how); 77 static int pipe_kqfilter (struct file *fp, struct knote *kn); 78 static int pipe_stat (struct file *fp, struct stat *sb, struct ucred *cred); 79 static int pipe_ioctl (struct file *fp, u_long cmd, caddr_t data, 80 struct ucred *cred, struct sysmsg *msg); 81 82 static struct fileops pipeops = { 83 .fo_read = pipe_read, 84 .fo_write = pipe_write, 85 .fo_ioctl = pipe_ioctl, 86 .fo_kqfilter = pipe_kqfilter, 87 .fo_stat = pipe_stat, 88 .fo_close = pipe_close, 89 .fo_shutdown = pipe_shutdown 90 }; 91 92 static void filt_pipedetach(struct knote *kn); 93 static int filt_piperead(struct knote *kn, long hint); 94 static int filt_pipewrite(struct knote *kn, long hint); 95 96 static struct filterops pipe_rfiltops = 97 { FILTEROP_ISFD, NULL, filt_pipedetach, filt_piperead }; 98 static struct filterops pipe_wfiltops = 99 { FILTEROP_ISFD, NULL, filt_pipedetach, filt_pipewrite }; 100 101 MALLOC_DEFINE(M_PIPE, "pipe", "pipe structures"); 102 103 /* 104 * Default pipe buffer size(s), this can be kind-of large now because pipe 105 * space is pageable. The pipe code will try to maintain locality of 106 * reference for performance reasons, so small amounts of outstanding I/O 107 * will not wipe the cache. 108 */ 109 #define MINPIPESIZE (PIPE_SIZE/3) 110 #define MAXPIPESIZE (2*PIPE_SIZE/3) 111 112 /* 113 * Limit the number of "big" pipes 114 */ 115 #define LIMITBIGPIPES 64 116 #define PIPEQ_MAX_CACHE 16 /* per-cpu pipe structure cache */ 117 118 static int pipe_maxbig = LIMITBIGPIPES; 119 static int pipe_maxcache = PIPEQ_MAX_CACHE; 120 static int pipe_bigcount; 121 static int pipe_nbig; 122 static int pipe_bcache_alloc; 123 static int pipe_bkmem_alloc; 124 static int pipe_rblocked_count; 125 static int pipe_wblocked_count; 126 127 SYSCTL_NODE(_kern, OID_AUTO, pipe, CTLFLAG_RW, 0, "Pipe operation"); 128 SYSCTL_INT(_kern_pipe, OID_AUTO, nbig, 129 CTLFLAG_RD, &pipe_nbig, 0, "numer of big pipes allocated"); 130 SYSCTL_INT(_kern_pipe, OID_AUTO, bigcount, 131 CTLFLAG_RW, &pipe_bigcount, 0, "number of times pipe expanded"); 132 SYSCTL_INT(_kern_pipe, OID_AUTO, rblocked, 133 CTLFLAG_RW, &pipe_rblocked_count, 0, "number of times pipe expanded"); 134 SYSCTL_INT(_kern_pipe, OID_AUTO, wblocked, 135 CTLFLAG_RW, &pipe_wblocked_count, 0, "number of times pipe expanded"); 136 SYSCTL_INT(_kern_pipe, OID_AUTO, maxcache, 137 CTLFLAG_RW, &pipe_maxcache, 0, "max pipes cached per-cpu"); 138 SYSCTL_INT(_kern_pipe, OID_AUTO, maxbig, 139 CTLFLAG_RW, &pipe_maxbig, 0, "max number of big pipes"); 140 #ifdef SMP 141 static int pipe_delay = 5000; /* 5uS default */ 142 SYSCTL_INT(_kern_pipe, OID_AUTO, delay, 143 CTLFLAG_RW, &pipe_delay, 0, "SMP delay optimization in ns"); 144 static int pipe_mpsafe = 1; 145 SYSCTL_INT(_kern_pipe, OID_AUTO, mpsafe, 146 CTLFLAG_RW, &pipe_mpsafe, 0, ""); 147 #endif 148 #if !defined(NO_PIPE_SYSCTL_STATS) 149 SYSCTL_INT(_kern_pipe, OID_AUTO, bcache_alloc, 150 CTLFLAG_RW, &pipe_bcache_alloc, 0, "pipe buffer from pcpu cache"); 151 SYSCTL_INT(_kern_pipe, OID_AUTO, bkmem_alloc, 152 CTLFLAG_RW, &pipe_bkmem_alloc, 0, "pipe buffer from kmem"); 153 #endif 154 155 static void pipeclose (struct pipe *cpipe); 156 static void pipe_free_kmem (struct pipe *cpipe); 157 static int pipe_create (struct pipe **cpipep); 158 static __inline void pipewakeup (struct pipe *cpipe); 159 static int pipespace (struct pipe *cpipe, int size); 160 161 static __inline void 162 pipewakeup(struct pipe *cpipe) 163 { 164 if ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) { 165 get_mplock(); 166 pgsigio(cpipe->pipe_sigio, SIGIO, 0); 167 rel_mplock(); 168 } 169 KNOTE(&cpipe->pipe_kq.ki_note, 0); 170 } 171 172 /* 173 * These routines are called before and after a UIO. The UIO 174 * may block, causing our held tokens to be lost temporarily. 175 * 176 * We use these routines to serialize reads against other reads 177 * and writes against other writes. 178 * 179 * The read token is held on entry so *ipp does not race. 180 */ 181 static __inline int 182 pipe_start_uio(struct pipe *cpipe, int *ipp) 183 { 184 int error; 185 186 while (*ipp) { 187 *ipp = -1; 188 error = tsleep(ipp, PCATCH, "pipexx", 0); 189 if (error) 190 return (error); 191 } 192 *ipp = 1; 193 return (0); 194 } 195 196 static __inline void 197 pipe_end_uio(struct pipe *cpipe, int *ipp) 198 { 199 if (*ipp < 0) { 200 *ipp = 0; 201 wakeup(ipp); 202 } else { 203 KKASSERT(*ipp > 0); 204 *ipp = 0; 205 } 206 } 207 208 static __inline void 209 pipe_get_mplock(int *save) 210 { 211 #ifdef SMP 212 if (pipe_mpsafe == 0) { 213 get_mplock(); 214 *save = 1; 215 } else 216 #endif 217 { 218 *save = 0; 219 } 220 } 221 222 static __inline void 223 pipe_rel_mplock(int *save) 224 { 225 #ifdef SMP 226 if (*save) 227 rel_mplock(); 228 #endif 229 } 230 231 232 /* 233 * The pipe system call for the DTYPE_PIPE type of pipes 234 * 235 * pipe_args(int dummy) 236 * 237 * MPSAFE 238 */ 239 int 240 sys_pipe(struct pipe_args *uap) 241 { 242 struct thread *td = curthread; 243 struct filedesc *fdp = td->td_proc->p_fd; 244 struct file *rf, *wf; 245 struct pipe *rpipe, *wpipe; 246 int fd1, fd2, error; 247 248 rpipe = wpipe = NULL; 249 if (pipe_create(&rpipe) || pipe_create(&wpipe)) { 250 pipeclose(rpipe); 251 pipeclose(wpipe); 252 return (ENFILE); 253 } 254 255 error = falloc(td->td_lwp, &rf, &fd1); 256 if (error) { 257 pipeclose(rpipe); 258 pipeclose(wpipe); 259 return (error); 260 } 261 uap->sysmsg_fds[0] = fd1; 262 263 /* 264 * Warning: once we've gotten past allocation of the fd for the 265 * read-side, we can only drop the read side via fdrop() in order 266 * to avoid races against processes which manage to dup() the read 267 * side while we are blocked trying to allocate the write side. 268 */ 269 rf->f_type = DTYPE_PIPE; 270 rf->f_flag = FREAD | FWRITE; 271 rf->f_ops = &pipeops; 272 rf->f_data = rpipe; 273 error = falloc(td->td_lwp, &wf, &fd2); 274 if (error) { 275 fsetfd(fdp, NULL, fd1); 276 fdrop(rf); 277 /* rpipe has been closed by fdrop(). */ 278 pipeclose(wpipe); 279 return (error); 280 } 281 wf->f_type = DTYPE_PIPE; 282 wf->f_flag = FREAD | FWRITE; 283 wf->f_ops = &pipeops; 284 wf->f_data = wpipe; 285 uap->sysmsg_fds[1] = fd2; 286 287 rpipe->pipe_slock = kmalloc(sizeof(struct lock), 288 M_PIPE, M_WAITOK|M_ZERO); 289 wpipe->pipe_slock = rpipe->pipe_slock; 290 rpipe->pipe_peer = wpipe; 291 wpipe->pipe_peer = rpipe; 292 lockinit(rpipe->pipe_slock, "pipecl", 0, 0); 293 294 /* 295 * Once activated the peer relationship remains valid until 296 * both sides are closed. 297 */ 298 fsetfd(fdp, rf, fd1); 299 fsetfd(fdp, wf, fd2); 300 fdrop(rf); 301 fdrop(wf); 302 303 return (0); 304 } 305 306 /* 307 * Allocate kva for pipe circular buffer, the space is pageable 308 * This routine will 'realloc' the size of a pipe safely, if it fails 309 * it will retain the old buffer. 310 * If it fails it will return ENOMEM. 311 */ 312 static int 313 pipespace(struct pipe *cpipe, int size) 314 { 315 struct vm_object *object; 316 caddr_t buffer; 317 int npages, error; 318 319 npages = round_page(size) / PAGE_SIZE; 320 object = cpipe->pipe_buffer.object; 321 322 /* 323 * [re]create the object if necessary and reserve space for it 324 * in the kernel_map. The object and memory are pageable. On 325 * success, free the old resources before assigning the new 326 * ones. 327 */ 328 if (object == NULL || object->size != npages) { 329 get_mplock(); 330 object = vm_object_allocate(OBJT_DEFAULT, npages); 331 buffer = (caddr_t)vm_map_min(&kernel_map); 332 333 error = vm_map_find(&kernel_map, object, 0, 334 (vm_offset_t *)&buffer, 335 size, PAGE_SIZE, 336 1, VM_MAPTYPE_NORMAL, 337 VM_PROT_ALL, VM_PROT_ALL, 338 0); 339 340 if (error != KERN_SUCCESS) { 341 vm_object_deallocate(object); 342 rel_mplock(); 343 return (ENOMEM); 344 } 345 pipe_free_kmem(cpipe); 346 rel_mplock(); 347 cpipe->pipe_buffer.object = object; 348 cpipe->pipe_buffer.buffer = buffer; 349 cpipe->pipe_buffer.size = size; 350 ++pipe_bkmem_alloc; 351 } else { 352 ++pipe_bcache_alloc; 353 } 354 cpipe->pipe_buffer.rindex = 0; 355 cpipe->pipe_buffer.windex = 0; 356 return (0); 357 } 358 359 /* 360 * Initialize and allocate VM and memory for pipe, pulling the pipe from 361 * our per-cpu cache if possible. For now make sure it is sized for the 362 * smaller PIPE_SIZE default. 363 */ 364 static int 365 pipe_create(struct pipe **cpipep) 366 { 367 globaldata_t gd = mycpu; 368 struct pipe *cpipe; 369 int error; 370 371 if ((cpipe = gd->gd_pipeq) != NULL) { 372 gd->gd_pipeq = cpipe->pipe_peer; 373 --gd->gd_pipeqcount; 374 cpipe->pipe_peer = NULL; 375 cpipe->pipe_wantwcnt = 0; 376 } else { 377 cpipe = kmalloc(sizeof(struct pipe), M_PIPE, M_WAITOK|M_ZERO); 378 } 379 *cpipep = cpipe; 380 if ((error = pipespace(cpipe, PIPE_SIZE)) != 0) 381 return (error); 382 vfs_timestamp(&cpipe->pipe_ctime); 383 cpipe->pipe_atime = cpipe->pipe_ctime; 384 cpipe->pipe_mtime = cpipe->pipe_ctime; 385 lwkt_token_init(&cpipe->pipe_rlock, 1); 386 lwkt_token_init(&cpipe->pipe_wlock, 1); 387 return (0); 388 } 389 390 /* 391 * MPALMOSTSAFE (acquires mplock) 392 */ 393 static int 394 pipe_read(struct file *fp, struct uio *uio, struct ucred *cred, int fflags) 395 { 396 struct pipe *rpipe; 397 int error; 398 size_t nread = 0; 399 int nbio; 400 u_int size; /* total bytes available */ 401 u_int nsize; /* total bytes to read */ 402 u_int rindex; /* contiguous bytes available */ 403 int notify_writer; 404 int mpsave; 405 int bigread; 406 int bigcount; 407 408 if (uio->uio_resid == 0) 409 return(0); 410 411 /* 412 * Setup locks, calculate nbio 413 */ 414 pipe_get_mplock(&mpsave); 415 rpipe = (struct pipe *)fp->f_data; 416 lwkt_gettoken(&rpipe->pipe_rlock); 417 418 if (fflags & O_FBLOCKING) 419 nbio = 0; 420 else if (fflags & O_FNONBLOCKING) 421 nbio = 1; 422 else if (fp->f_flag & O_NONBLOCK) 423 nbio = 1; 424 else 425 nbio = 0; 426 427 /* 428 * Reads are serialized. Note however that pipe_buffer.buffer and 429 * pipe_buffer.size can change out from under us when the number 430 * of bytes in the buffer are zero due to the write-side doing a 431 * pipespace(). 432 */ 433 error = pipe_start_uio(rpipe, &rpipe->pipe_rip); 434 if (error) { 435 pipe_rel_mplock(&mpsave); 436 lwkt_reltoken(&rpipe->pipe_rlock); 437 return (error); 438 } 439 notify_writer = 0; 440 441 bigread = (uio->uio_resid > 10 * 1024 * 1024); 442 bigcount = 10; 443 444 while (uio->uio_resid) { 445 /* 446 * Don't hog the cpu. 447 */ 448 if (bigread && --bigcount == 0) { 449 lwkt_user_yield(); 450 bigcount = 10; 451 if (CURSIG(curthread->td_lwp)) { 452 error = EINTR; 453 break; 454 } 455 } 456 457 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex; 458 cpu_lfence(); 459 if (size) { 460 rindex = rpipe->pipe_buffer.rindex & 461 (rpipe->pipe_buffer.size - 1); 462 nsize = size; 463 if (nsize > rpipe->pipe_buffer.size - rindex) 464 nsize = rpipe->pipe_buffer.size - rindex; 465 nsize = szmin(nsize, uio->uio_resid); 466 467 error = uiomove(&rpipe->pipe_buffer.buffer[rindex], 468 nsize, uio); 469 if (error) 470 break; 471 cpu_mfence(); 472 rpipe->pipe_buffer.rindex += nsize; 473 nread += nsize; 474 475 /* 476 * If the FIFO is still over half full just continue 477 * and do not try to notify the writer yet. 478 */ 479 if (size - nsize >= (rpipe->pipe_buffer.size >> 1)) { 480 notify_writer = 0; 481 continue; 482 } 483 484 /* 485 * When the FIFO is less then half full notify any 486 * waiting writer. WANTW can be checked while 487 * holding just the rlock. 488 */ 489 notify_writer = 1; 490 if ((rpipe->pipe_state & PIPE_WANTW) == 0) 491 continue; 492 } 493 494 /* 495 * If the "write-side" was blocked we wake it up. This code 496 * is reached either when the buffer is completely emptied 497 * or if it becomes more then half-empty. 498 * 499 * Pipe_state can only be modified if both the rlock and 500 * wlock are held. 501 */ 502 if (rpipe->pipe_state & PIPE_WANTW) { 503 lwkt_gettoken(&rpipe->pipe_wlock); 504 if (rpipe->pipe_state & PIPE_WANTW) { 505 notify_writer = 0; 506 rpipe->pipe_state &= ~PIPE_WANTW; 507 lwkt_reltoken(&rpipe->pipe_wlock); 508 wakeup(rpipe); 509 } else { 510 lwkt_reltoken(&rpipe->pipe_wlock); 511 } 512 } 513 514 /* 515 * Pick up our copy loop again if the writer sent data to 516 * us while we were messing around. 517 * 518 * On a SMP box poll up to pipe_delay nanoseconds for new 519 * data. Typically a value of 2000 to 4000 is sufficient 520 * to eradicate most IPIs/tsleeps/wakeups when a pipe 521 * is used for synchronous communications with small packets, 522 * and 8000 or so (8uS) will pipeline large buffer xfers 523 * between cpus over a pipe. 524 * 525 * For synchronous communications a hit means doing a 526 * full Awrite-Bread-Bwrite-Aread cycle in less then 2uS, 527 * where as miss requiring a tsleep/wakeup sequence 528 * will take 7uS or more. 529 */ 530 if (rpipe->pipe_buffer.windex != rpipe->pipe_buffer.rindex) 531 continue; 532 533 #if defined(SMP) && defined(_RDTSC_SUPPORTED_) 534 if (pipe_delay) { 535 int64_t tsc_target; 536 int good = 0; 537 538 tsc_target = tsc_get_target(pipe_delay); 539 while (tsc_test_target(tsc_target) == 0) { 540 if (rpipe->pipe_buffer.windex != 541 rpipe->pipe_buffer.rindex) { 542 good = 1; 543 break; 544 } 545 } 546 if (good) 547 continue; 548 } 549 #endif 550 551 /* 552 * Detect EOF condition, do not set error. 553 */ 554 if (rpipe->pipe_state & PIPE_REOF) 555 break; 556 557 /* 558 * Break if some data was read, or if this was a non-blocking 559 * read. 560 */ 561 if (nread > 0) 562 break; 563 564 if (nbio) { 565 error = EAGAIN; 566 break; 567 } 568 569 /* 570 * Last chance, interlock with WANTR. 571 */ 572 lwkt_gettoken(&rpipe->pipe_wlock); 573 size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex; 574 if (size) { 575 lwkt_reltoken(&rpipe->pipe_wlock); 576 continue; 577 } 578 579 /* 580 * Retest EOF - acquiring a new token can temporarily release 581 * tokens already held. 582 */ 583 if (rpipe->pipe_state & PIPE_REOF) { 584 lwkt_reltoken(&rpipe->pipe_wlock); 585 break; 586 } 587 588 /* 589 * If there is no more to read in the pipe, reset its 590 * pointers to the beginning. This improves cache hit 591 * stats. 592 * 593 * We need both locks to modify both pointers, and there 594 * must also not be a write in progress or the uiomove() 595 * in the write might block and temporarily release 596 * its wlock, then reacquire and update windex. We are 597 * only serialized against reads, not writes. 598 * 599 * XXX should we even bother resetting the indices? It 600 * might actually be more cache efficient not to. 601 */ 602 if (rpipe->pipe_buffer.rindex == rpipe->pipe_buffer.windex && 603 rpipe->pipe_wip == 0) { 604 rpipe->pipe_buffer.rindex = 0; 605 rpipe->pipe_buffer.windex = 0; 606 } 607 608 /* 609 * Wait for more data. 610 * 611 * Pipe_state can only be set if both the rlock and wlock 612 * are held. 613 */ 614 rpipe->pipe_state |= PIPE_WANTR; 615 tsleep_interlock(rpipe, PCATCH); 616 lwkt_reltoken(&rpipe->pipe_wlock); 617 error = tsleep(rpipe, PCATCH | PINTERLOCKED, "piperd", 0); 618 ++pipe_rblocked_count; 619 if (error) 620 break; 621 } 622 pipe_end_uio(rpipe, &rpipe->pipe_rip); 623 624 /* 625 * Uptime last access time 626 */ 627 if (error == 0 && nread) 628 vfs_timestamp(&rpipe->pipe_atime); 629 630 /* 631 * If we drained the FIFO more then half way then handle 632 * write blocking hysteresis. 633 * 634 * Note that PIPE_WANTW cannot be set by the writer without 635 * it holding both rlock and wlock, so we can test it 636 * while holding just rlock. 637 */ 638 if (notify_writer) { 639 if (rpipe->pipe_state & PIPE_WANTW) { 640 lwkt_gettoken(&rpipe->pipe_wlock); 641 if (rpipe->pipe_state & PIPE_WANTW) { 642 rpipe->pipe_state &= ~PIPE_WANTW; 643 lwkt_reltoken(&rpipe->pipe_wlock); 644 wakeup(rpipe); 645 } else { 646 lwkt_reltoken(&rpipe->pipe_wlock); 647 } 648 } 649 lwkt_gettoken(&rpipe->pipe_wlock); 650 pipewakeup(rpipe); 651 lwkt_reltoken(&rpipe->pipe_wlock); 652 } 653 /*size = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex;*/ 654 lwkt_reltoken(&rpipe->pipe_rlock); 655 656 pipe_rel_mplock(&mpsave); 657 return (error); 658 } 659 660 /* 661 * MPALMOSTSAFE - acquires mplock 662 */ 663 static int 664 pipe_write(struct file *fp, struct uio *uio, struct ucred *cred, int fflags) 665 { 666 int error; 667 int orig_resid; 668 int nbio; 669 struct pipe *wpipe, *rpipe; 670 u_int windex; 671 u_int space; 672 u_int wcount; 673 int mpsave; 674 int bigwrite; 675 int bigcount; 676 677 pipe_get_mplock(&mpsave); 678 679 /* 680 * Writes go to the peer. The peer will always exist. 681 */ 682 rpipe = (struct pipe *) fp->f_data; 683 wpipe = rpipe->pipe_peer; 684 lwkt_gettoken(&wpipe->pipe_wlock); 685 if (wpipe->pipe_state & PIPE_WEOF) { 686 pipe_rel_mplock(&mpsave); 687 lwkt_reltoken(&wpipe->pipe_wlock); 688 return (EPIPE); 689 } 690 691 /* 692 * Degenerate case (EPIPE takes prec) 693 */ 694 if (uio->uio_resid == 0) { 695 pipe_rel_mplock(&mpsave); 696 lwkt_reltoken(&wpipe->pipe_wlock); 697 return(0); 698 } 699 700 /* 701 * Writes are serialized (start_uio must be called with wlock) 702 */ 703 error = pipe_start_uio(wpipe, &wpipe->pipe_wip); 704 if (error) { 705 pipe_rel_mplock(&mpsave); 706 lwkt_reltoken(&wpipe->pipe_wlock); 707 return (error); 708 } 709 710 if (fflags & O_FBLOCKING) 711 nbio = 0; 712 else if (fflags & O_FNONBLOCKING) 713 nbio = 1; 714 else if (fp->f_flag & O_NONBLOCK) 715 nbio = 1; 716 else 717 nbio = 0; 718 719 /* 720 * If it is advantageous to resize the pipe buffer, do 721 * so. We are write-serialized so we can block safely. 722 */ 723 if ((wpipe->pipe_buffer.size <= PIPE_SIZE) && 724 (pipe_nbig < pipe_maxbig) && 725 wpipe->pipe_wantwcnt > 4 && 726 (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) { 727 /* 728 * Recheck after lock. 729 */ 730 lwkt_gettoken(&wpipe->pipe_rlock); 731 if ((wpipe->pipe_buffer.size <= PIPE_SIZE) && 732 (pipe_nbig < pipe_maxbig) && 733 (wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex)) { 734 atomic_add_int(&pipe_nbig, 1); 735 if (pipespace(wpipe, BIG_PIPE_SIZE) == 0) 736 ++pipe_bigcount; 737 else 738 atomic_subtract_int(&pipe_nbig, 1); 739 } 740 lwkt_reltoken(&wpipe->pipe_rlock); 741 } 742 743 orig_resid = uio->uio_resid; 744 wcount = 0; 745 746 bigwrite = (uio->uio_resid > 10 * 1024 * 1024); 747 bigcount = 10; 748 749 while (uio->uio_resid) { 750 if (wpipe->pipe_state & PIPE_WEOF) { 751 error = EPIPE; 752 break; 753 } 754 755 /* 756 * Don't hog the cpu. 757 */ 758 if (bigwrite && --bigcount == 0) { 759 lwkt_user_yield(); 760 bigcount = 10; 761 if (CURSIG(curthread->td_lwp)) { 762 error = EINTR; 763 break; 764 } 765 } 766 767 windex = wpipe->pipe_buffer.windex & 768 (wpipe->pipe_buffer.size - 1); 769 space = wpipe->pipe_buffer.size - 770 (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex); 771 cpu_lfence(); 772 773 /* Writes of size <= PIPE_BUF must be atomic. */ 774 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF)) 775 space = 0; 776 777 /* 778 * Write to fill, read size handles write hysteresis. Also 779 * additional restrictions can cause select-based non-blocking 780 * writes to spin. 781 */ 782 if (space > 0) { 783 u_int segsize; 784 785 /* 786 * Transfer size is minimum of uio transfer 787 * and free space in pipe buffer. 788 * 789 * Limit each uiocopy to no more then PIPE_SIZE 790 * so we can keep the gravy train going on a 791 * SMP box. This doubles the performance for 792 * write sizes > 16K. Otherwise large writes 793 * wind up doing an inefficient synchronous 794 * ping-pong. 795 */ 796 space = szmin(space, uio->uio_resid); 797 if (space > PIPE_SIZE) 798 space = PIPE_SIZE; 799 800 /* 801 * First segment to transfer is minimum of 802 * transfer size and contiguous space in 803 * pipe buffer. If first segment to transfer 804 * is less than the transfer size, we've got 805 * a wraparound in the buffer. 806 */ 807 segsize = wpipe->pipe_buffer.size - windex; 808 if (segsize > space) 809 segsize = space; 810 811 #ifdef SMP 812 /* 813 * If this is the first loop and the reader is 814 * blocked, do a preemptive wakeup of the reader. 815 * 816 * On SMP the IPI latency plus the wlock interlock 817 * on the reader side is the fastest way to get the 818 * reader going. (The scheduler will hard loop on 819 * lock tokens). 820 * 821 * NOTE: We can't clear WANTR here without acquiring 822 * the rlock, which we don't want to do here! 823 */ 824 if ((wpipe->pipe_state & PIPE_WANTR) && pipe_mpsafe > 1) 825 wakeup(wpipe); 826 #endif 827 828 /* 829 * Transfer segment, which may include a wrap-around. 830 * Update windex to account for both all in one go 831 * so the reader can read() the data atomically. 832 */ 833 error = uiomove(&wpipe->pipe_buffer.buffer[windex], 834 segsize, uio); 835 if (error == 0 && segsize < space) { 836 segsize = space - segsize; 837 error = uiomove(&wpipe->pipe_buffer.buffer[0], 838 segsize, uio); 839 } 840 if (error) 841 break; 842 cpu_mfence(); 843 wpipe->pipe_buffer.windex += space; 844 wcount += space; 845 continue; 846 } 847 848 /* 849 * We need both the rlock and the wlock to interlock against 850 * the EOF, WANTW, and size checks, and to modify pipe_state. 851 * 852 * These are token locks so we do not have to worry about 853 * deadlocks. 854 */ 855 lwkt_gettoken(&wpipe->pipe_rlock); 856 857 /* 858 * If the "read-side" has been blocked, wake it up now 859 * and yield to let it drain synchronously rather 860 * then block. 861 */ 862 if (wpipe->pipe_state & PIPE_WANTR) { 863 wpipe->pipe_state &= ~PIPE_WANTR; 864 wakeup(wpipe); 865 } 866 867 /* 868 * don't block on non-blocking I/O 869 */ 870 if (nbio) { 871 lwkt_reltoken(&wpipe->pipe_rlock); 872 error = EAGAIN; 873 break; 874 } 875 876 /* 877 * re-test whether we have to block in the writer after 878 * acquiring both locks, in case the reader opened up 879 * some space. 880 */ 881 space = wpipe->pipe_buffer.size - 882 (wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex); 883 cpu_lfence(); 884 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF)) 885 space = 0; 886 887 /* 888 * Retest EOF - acquiring a new token can temporarily release 889 * tokens already held. 890 */ 891 if (wpipe->pipe_state & PIPE_WEOF) { 892 lwkt_reltoken(&wpipe->pipe_rlock); 893 error = EPIPE; 894 break; 895 } 896 897 /* 898 * We have no more space and have something to offer, 899 * wake up select/poll/kq. 900 */ 901 if (space == 0) { 902 wpipe->pipe_state |= PIPE_WANTW; 903 ++wpipe->pipe_wantwcnt; 904 pipewakeup(wpipe); 905 if (wpipe->pipe_state & PIPE_WANTW) 906 error = tsleep(wpipe, PCATCH, "pipewr", 0); 907 ++pipe_wblocked_count; 908 } 909 lwkt_reltoken(&wpipe->pipe_rlock); 910 911 /* 912 * Break out if we errored or the read side wants us to go 913 * away. 914 */ 915 if (error) 916 break; 917 if (wpipe->pipe_state & PIPE_WEOF) { 918 error = EPIPE; 919 break; 920 } 921 } 922 pipe_end_uio(wpipe, &wpipe->pipe_wip); 923 924 /* 925 * If we have put any characters in the buffer, we wake up 926 * the reader. 927 * 928 * Both rlock and wlock are required to be able to modify pipe_state. 929 */ 930 if (wpipe->pipe_buffer.windex != wpipe->pipe_buffer.rindex) { 931 if (wpipe->pipe_state & PIPE_WANTR) { 932 lwkt_gettoken(&wpipe->pipe_rlock); 933 if (wpipe->pipe_state & PIPE_WANTR) { 934 wpipe->pipe_state &= ~PIPE_WANTR; 935 lwkt_reltoken(&wpipe->pipe_rlock); 936 wakeup(wpipe); 937 } else { 938 lwkt_reltoken(&wpipe->pipe_rlock); 939 } 940 } 941 lwkt_gettoken(&wpipe->pipe_rlock); 942 pipewakeup(wpipe); 943 lwkt_reltoken(&wpipe->pipe_rlock); 944 } 945 946 /* 947 * Don't return EPIPE if I/O was successful 948 */ 949 if ((wpipe->pipe_buffer.rindex == wpipe->pipe_buffer.windex) && 950 (uio->uio_resid == 0) && 951 (error == EPIPE)) { 952 error = 0; 953 } 954 955 if (error == 0) 956 vfs_timestamp(&wpipe->pipe_mtime); 957 958 /* 959 * We have something to offer, 960 * wake up select/poll/kq. 961 */ 962 /*space = wpipe->pipe_buffer.windex - wpipe->pipe_buffer.rindex;*/ 963 lwkt_reltoken(&wpipe->pipe_wlock); 964 pipe_rel_mplock(&mpsave); 965 return (error); 966 } 967 968 /* 969 * MPALMOSTSAFE - acquires mplock 970 * 971 * we implement a very minimal set of ioctls for compatibility with sockets. 972 */ 973 int 974 pipe_ioctl(struct file *fp, u_long cmd, caddr_t data, 975 struct ucred *cred, struct sysmsg *msg) 976 { 977 struct pipe *mpipe; 978 int error; 979 int mpsave; 980 981 pipe_get_mplock(&mpsave); 982 mpipe = (struct pipe *)fp->f_data; 983 984 lwkt_gettoken(&mpipe->pipe_rlock); 985 lwkt_gettoken(&mpipe->pipe_wlock); 986 987 switch (cmd) { 988 case FIOASYNC: 989 if (*(int *)data) { 990 mpipe->pipe_state |= PIPE_ASYNC; 991 } else { 992 mpipe->pipe_state &= ~PIPE_ASYNC; 993 } 994 error = 0; 995 break; 996 case FIONREAD: 997 *(int *)data = mpipe->pipe_buffer.windex - 998 mpipe->pipe_buffer.rindex; 999 error = 0; 1000 break; 1001 case FIOSETOWN: 1002 get_mplock(); 1003 error = fsetown(*(int *)data, &mpipe->pipe_sigio); 1004 rel_mplock(); 1005 break; 1006 case FIOGETOWN: 1007 *(int *)data = fgetown(mpipe->pipe_sigio); 1008 error = 0; 1009 break; 1010 case TIOCSPGRP: 1011 /* This is deprecated, FIOSETOWN should be used instead. */ 1012 get_mplock(); 1013 error = fsetown(-(*(int *)data), &mpipe->pipe_sigio); 1014 rel_mplock(); 1015 break; 1016 1017 case TIOCGPGRP: 1018 /* This is deprecated, FIOGETOWN should be used instead. */ 1019 *(int *)data = -fgetown(mpipe->pipe_sigio); 1020 error = 0; 1021 break; 1022 default: 1023 error = ENOTTY; 1024 break; 1025 } 1026 lwkt_reltoken(&mpipe->pipe_wlock); 1027 lwkt_reltoken(&mpipe->pipe_rlock); 1028 pipe_rel_mplock(&mpsave); 1029 1030 return (error); 1031 } 1032 1033 /* 1034 * MPSAFE 1035 */ 1036 static int 1037 pipe_stat(struct file *fp, struct stat *ub, struct ucred *cred) 1038 { 1039 struct pipe *pipe; 1040 int mpsave; 1041 1042 pipe_get_mplock(&mpsave); 1043 pipe = (struct pipe *)fp->f_data; 1044 1045 bzero((caddr_t)ub, sizeof(*ub)); 1046 ub->st_mode = S_IFIFO; 1047 ub->st_blksize = pipe->pipe_buffer.size; 1048 ub->st_size = pipe->pipe_buffer.windex - pipe->pipe_buffer.rindex; 1049 ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize; 1050 ub->st_atimespec = pipe->pipe_atime; 1051 ub->st_mtimespec = pipe->pipe_mtime; 1052 ub->st_ctimespec = pipe->pipe_ctime; 1053 /* 1054 * Left as 0: st_dev, st_ino, st_nlink, st_uid, st_gid, st_rdev, 1055 * st_flags, st_gen. 1056 * XXX (st_dev, st_ino) should be unique. 1057 */ 1058 pipe_rel_mplock(&mpsave); 1059 return (0); 1060 } 1061 1062 /* 1063 * MPALMOSTSAFE - acquires mplock 1064 */ 1065 static int 1066 pipe_close(struct file *fp) 1067 { 1068 struct pipe *cpipe; 1069 1070 get_mplock(); 1071 cpipe = (struct pipe *)fp->f_data; 1072 fp->f_ops = &badfileops; 1073 fp->f_data = NULL; 1074 funsetown(cpipe->pipe_sigio); 1075 pipeclose(cpipe); 1076 rel_mplock(); 1077 return (0); 1078 } 1079 1080 /* 1081 * Shutdown one or both directions of a full-duplex pipe. 1082 * 1083 * MPALMOSTSAFE - acquires mplock 1084 */ 1085 static int 1086 pipe_shutdown(struct file *fp, int how) 1087 { 1088 struct pipe *rpipe; 1089 struct pipe *wpipe; 1090 int error = EPIPE; 1091 int mpsave; 1092 1093 pipe_get_mplock(&mpsave); 1094 rpipe = (struct pipe *)fp->f_data; 1095 wpipe = rpipe->pipe_peer; 1096 1097 /* 1098 * We modify pipe_state on both pipes, which means we need 1099 * all four tokens! 1100 */ 1101 lwkt_gettoken(&rpipe->pipe_rlock); 1102 lwkt_gettoken(&rpipe->pipe_wlock); 1103 lwkt_gettoken(&wpipe->pipe_rlock); 1104 lwkt_gettoken(&wpipe->pipe_wlock); 1105 1106 switch(how) { 1107 case SHUT_RDWR: 1108 case SHUT_RD: 1109 rpipe->pipe_state |= PIPE_REOF; /* my reads */ 1110 rpipe->pipe_state |= PIPE_WEOF; /* peer writes */ 1111 if (rpipe->pipe_state & PIPE_WANTR) { 1112 rpipe->pipe_state &= ~PIPE_WANTR; 1113 wakeup(rpipe); 1114 } 1115 if (rpipe->pipe_state & PIPE_WANTW) { 1116 rpipe->pipe_state &= ~PIPE_WANTW; 1117 wakeup(rpipe); 1118 } 1119 error = 0; 1120 if (how == SHUT_RD) 1121 break; 1122 /* fall through */ 1123 case SHUT_WR: 1124 wpipe->pipe_state |= PIPE_REOF; /* peer reads */ 1125 wpipe->pipe_state |= PIPE_WEOF; /* my writes */ 1126 if (wpipe->pipe_state & PIPE_WANTR) { 1127 wpipe->pipe_state &= ~PIPE_WANTR; 1128 wakeup(wpipe); 1129 } 1130 if (wpipe->pipe_state & PIPE_WANTW) { 1131 wpipe->pipe_state &= ~PIPE_WANTW; 1132 wakeup(wpipe); 1133 } 1134 error = 0; 1135 break; 1136 } 1137 pipewakeup(rpipe); 1138 pipewakeup(wpipe); 1139 1140 lwkt_reltoken(&wpipe->pipe_wlock); 1141 lwkt_reltoken(&wpipe->pipe_rlock); 1142 lwkt_reltoken(&rpipe->pipe_wlock); 1143 lwkt_reltoken(&rpipe->pipe_rlock); 1144 1145 pipe_rel_mplock(&mpsave); 1146 return (error); 1147 } 1148 1149 static void 1150 pipe_free_kmem(struct pipe *cpipe) 1151 { 1152 if (cpipe->pipe_buffer.buffer != NULL) { 1153 if (cpipe->pipe_buffer.size > PIPE_SIZE) 1154 atomic_subtract_int(&pipe_nbig, 1); 1155 kmem_free(&kernel_map, 1156 (vm_offset_t)cpipe->pipe_buffer.buffer, 1157 cpipe->pipe_buffer.size); 1158 cpipe->pipe_buffer.buffer = NULL; 1159 cpipe->pipe_buffer.object = NULL; 1160 } 1161 } 1162 1163 /* 1164 * Close the pipe. The slock must be held to interlock against simultanious 1165 * closes. The rlock and wlock must be held to adjust the pipe_state. 1166 */ 1167 static void 1168 pipeclose(struct pipe *cpipe) 1169 { 1170 globaldata_t gd; 1171 struct pipe *ppipe; 1172 1173 if (cpipe == NULL) 1174 return; 1175 1176 /* 1177 * The slock may not have been allocated yet (close during 1178 * initialization) 1179 * 1180 * We need both the read and write tokens to modify pipe_state. 1181 */ 1182 if (cpipe->pipe_slock) 1183 lockmgr(cpipe->pipe_slock, LK_EXCLUSIVE); 1184 lwkt_gettoken(&cpipe->pipe_rlock); 1185 lwkt_gettoken(&cpipe->pipe_wlock); 1186 1187 /* 1188 * Set our state, wakeup anyone waiting in select/poll/kq, and 1189 * wakeup anyone blocked on our pipe. 1190 */ 1191 cpipe->pipe_state |= PIPE_CLOSED | PIPE_REOF | PIPE_WEOF; 1192 pipewakeup(cpipe); 1193 if (cpipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) { 1194 cpipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW); 1195 wakeup(cpipe); 1196 } 1197 1198 /* 1199 * Disconnect from peer. 1200 */ 1201 if ((ppipe = cpipe->pipe_peer) != NULL) { 1202 lwkt_gettoken(&ppipe->pipe_rlock); 1203 lwkt_gettoken(&ppipe->pipe_wlock); 1204 ppipe->pipe_state |= PIPE_REOF | PIPE_WEOF; 1205 pipewakeup(ppipe); 1206 if (ppipe->pipe_state & (PIPE_WANTR | PIPE_WANTW)) { 1207 ppipe->pipe_state &= ~(PIPE_WANTR | PIPE_WANTW); 1208 wakeup(ppipe); 1209 } 1210 if (SLIST_FIRST(&ppipe->pipe_kq.ki_note)) 1211 KNOTE(&ppipe->pipe_kq.ki_note, 0); 1212 lwkt_reltoken(&ppipe->pipe_wlock); 1213 lwkt_reltoken(&ppipe->pipe_rlock); 1214 } 1215 1216 /* 1217 * If the peer is also closed we can free resources for both 1218 * sides, otherwise we leave our side intact to deal with any 1219 * races (since we only have the slock). 1220 */ 1221 if (ppipe && (ppipe->pipe_state & PIPE_CLOSED)) { 1222 cpipe->pipe_peer = NULL; 1223 ppipe->pipe_peer = NULL; 1224 ppipe->pipe_slock = NULL; /* we will free the slock */ 1225 pipeclose(ppipe); 1226 ppipe = NULL; 1227 } 1228 1229 lwkt_reltoken(&cpipe->pipe_wlock); 1230 lwkt_reltoken(&cpipe->pipe_rlock); 1231 if (cpipe->pipe_slock) 1232 lockmgr(cpipe->pipe_slock, LK_RELEASE); 1233 1234 /* 1235 * If we disassociated from our peer we can free resources 1236 */ 1237 if (ppipe == NULL) { 1238 gd = mycpu; 1239 if (cpipe->pipe_slock) { 1240 kfree(cpipe->pipe_slock, M_PIPE); 1241 cpipe->pipe_slock = NULL; 1242 } 1243 if (gd->gd_pipeqcount >= pipe_maxcache || 1244 cpipe->pipe_buffer.size != PIPE_SIZE 1245 ) { 1246 pipe_free_kmem(cpipe); 1247 kfree(cpipe, M_PIPE); 1248 } else { 1249 cpipe->pipe_state = 0; 1250 cpipe->pipe_peer = gd->gd_pipeq; 1251 gd->gd_pipeq = cpipe; 1252 ++gd->gd_pipeqcount; 1253 } 1254 } 1255 } 1256 1257 /* 1258 * MPALMOSTSAFE - acquires mplock 1259 */ 1260 static int 1261 pipe_kqfilter(struct file *fp, struct knote *kn) 1262 { 1263 struct pipe *cpipe; 1264 1265 cpipe = (struct pipe *)kn->kn_fp->f_data; 1266 1267 switch (kn->kn_filter) { 1268 case EVFILT_READ: 1269 kn->kn_fop = &pipe_rfiltops; 1270 break; 1271 case EVFILT_WRITE: 1272 kn->kn_fop = &pipe_wfiltops; 1273 if (cpipe->pipe_peer == NULL) { 1274 /* other end of pipe has been closed */ 1275 return (EPIPE); 1276 } 1277 break; 1278 default: 1279 return (EOPNOTSUPP); 1280 } 1281 kn->kn_hook = (caddr_t)cpipe; 1282 1283 knote_insert(&cpipe->pipe_kq.ki_note, kn); 1284 1285 return (0); 1286 } 1287 1288 static void 1289 filt_pipedetach(struct knote *kn) 1290 { 1291 struct pipe *cpipe = (struct pipe *)kn->kn_hook; 1292 1293 knote_remove(&cpipe->pipe_kq.ki_note, kn); 1294 } 1295 1296 /*ARGSUSED*/ 1297 static int 1298 filt_piperead(struct knote *kn, long hint) 1299 { 1300 struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data; 1301 int ready = 0; 1302 1303 lwkt_gettoken(&rpipe->pipe_rlock); 1304 lwkt_gettoken(&rpipe->pipe_wlock); 1305 1306 kn->kn_data = rpipe->pipe_buffer.windex - rpipe->pipe_buffer.rindex; 1307 if (rpipe->pipe_state & PIPE_REOF) { 1308 kn->kn_flags |= EV_EOF; 1309 ready = 1; 1310 } 1311 1312 lwkt_reltoken(&rpipe->pipe_wlock); 1313 lwkt_reltoken(&rpipe->pipe_rlock); 1314 1315 if (!ready) 1316 ready = kn->kn_data > 0; 1317 1318 return (ready); 1319 } 1320 1321 /*ARGSUSED*/ 1322 static int 1323 filt_pipewrite(struct knote *kn, long hint) 1324 { 1325 struct pipe *rpipe = (struct pipe *)kn->kn_fp->f_data; 1326 struct pipe *wpipe = rpipe->pipe_peer; 1327 int ready = 0; 1328 1329 kn->kn_data = 0; 1330 if (wpipe == NULL) { 1331 kn->kn_flags |= EV_EOF; 1332 return (1); 1333 } 1334 1335 lwkt_gettoken(&wpipe->pipe_rlock); 1336 lwkt_gettoken(&wpipe->pipe_wlock); 1337 1338 if (wpipe->pipe_state & PIPE_WEOF) { 1339 kn->kn_flags |= EV_EOF; 1340 ready = 1; 1341 } 1342 1343 if (!ready) 1344 kn->kn_data = wpipe->pipe_buffer.size - 1345 (wpipe->pipe_buffer.windex - 1346 wpipe->pipe_buffer.rindex); 1347 1348 lwkt_reltoken(&wpipe->pipe_wlock); 1349 lwkt_reltoken(&wpipe->pipe_rlock); 1350 1351 if (!ready) 1352 ready = kn->kn_data >= PIPE_BUF; 1353 1354 return (ready); 1355 } 1356