The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/sys_pipe.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    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 1.21.2.1 1999/09/05 08:15:17 peter Exp $
   20  */
   21 
   22 #ifndef OLD_PIPE
   23 
   24 /*
   25  * This file contains a high-performance replacement for the socket-based
   26  * pipes scheme originally used in FreeBSD/4.4Lite.  It does not support
   27  * all features of sockets, but does do everything that pipes normally
   28  * do.
   29  */
   30 
   31 /*
   32  * This code has two modes of operation, a small write mode and a large
   33  * write mode.  The small write mode acts like conventional pipes with
   34  * a kernel buffer.  If the buffer is less than PIPE_MINDIRECT, then the
   35  * "normal" pipe buffering is done.  If the buffer is between PIPE_MINDIRECT
   36  * and PIPE_SIZE in size, it is fully mapped and wired into the kernel, and
   37  * the receiving process can copy it directly from the pages in the sending
   38  * process.
   39  *
   40  * If the sending process receives a signal, it is possible that it will
   41  * go away, and certainly its address space can change, because control
   42  * is returned back to the user-mode side.  In that case, the pipe code
   43  * arranges to copy the buffer supplied by the user process, to a pageable
   44  * kernel buffer, and the receiving process will grab the data from the
   45  * pageable kernel buffer.  Since signals don't happen all that often,
   46  * the copy operation is normally eliminated.
   47  *
   48  * The constant PIPE_MINDIRECT is chosen to make sure that buffering will
   49  * happen for small transfers so that the system will not spend all of
   50  * its time context switching.  PIPE_SIZE is constrained by the
   51  * amount of kernel virtual memory.
   52  */
   53 
   54 #include <sys/param.h>
   55 #include <sys/systm.h>
   56 #include <sys/proc.h>
   57 #include <sys/file.h>
   58 #include <sys/protosw.h>
   59 #include <sys/stat.h>
   60 #include <sys/filedesc.h>
   61 #include <sys/malloc.h>
   62 #include <sys/ioctl.h>
   63 #include <sys/stat.h>
   64 #include <sys/select.h>
   65 #include <sys/signalvar.h>
   66 #include <sys/errno.h>
   67 #include <sys/queue.h>
   68 #include <sys/vmmeter.h>
   69 #include <sys/kernel.h>
   70 #include <sys/sysproto.h>
   71 #include <sys/pipe.h>
   72 
   73 #include <vm/vm.h>
   74 #include <vm/vm_prot.h>
   75 #include <vm/vm_param.h>
   76 #include <vm/lock.h>
   77 #include <vm/vm_object.h>
   78 #include <vm/vm_kern.h>
   79 #include <vm/vm_extern.h>
   80 #include <vm/pmap.h>
   81 #include <vm/vm_map.h>
   82 #include <vm/vm_page.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  */
   94 static int pipe_read __P((struct file *fp, struct uio *uio, 
   95                 struct ucred *cred));
   96 static int pipe_write __P((struct file *fp, struct uio *uio, 
   97                 struct ucred *cred));
   98 static int pipe_close __P((struct file *fp, struct proc *p));
   99 static int pipe_select __P((struct file *fp, int which, struct proc *p));
  100 static int pipe_ioctl __P((struct file *fp, int cmd, caddr_t data, struct proc *p));
  101 
  102 static struct fileops pipeops =
  103     { pipe_read, pipe_write, pipe_ioctl, pipe_select, pipe_close };
  104 
  105 /*
  106  * Default pipe buffer size(s), this can be kind-of large now because pipe
  107  * space is pageable.  The pipe code will try to maintain locality of
  108  * reference for performance reasons, so small amounts of outstanding I/O
  109  * will not wipe the cache.
  110  */
  111 #define MINPIPESIZE (PIPE_SIZE/3)
  112 #define MAXPIPESIZE (2*PIPE_SIZE/3)
  113 
  114 /*
  115  * Maximum amount of kva for pipes -- this is kind-of a soft limit, but
  116  * is there so that on large systems, we don't exhaust it.
  117  */
  118 #define MAXPIPEKVA (8*1024*1024)
  119 
  120 /*
  121  * Limit for direct transfers, we cannot, of course limit
  122  * the amount of kva for pipes in general though.
  123  */
  124 #define LIMITPIPEKVA (16*1024*1024)
  125 
  126 /*
  127  * Limit the number of "big" pipes
  128  */
  129 #define LIMITBIGPIPES   32
  130 int nbigpipe;
  131 
  132 static int amountpipekva;
  133 
  134 static void pipeclose __P((struct pipe *cpipe));
  135 static void pipeinit __P((struct pipe *cpipe));
  136 static __inline int pipelock __P((struct pipe *cpipe, int catch));
  137 static __inline void pipeunlock __P((struct pipe *cpipe));
  138 static __inline void pipeselwakeup __P((struct pipe *cpipe));
  139 #ifndef PIPE_NODIRECT
  140 static int pipe_build_write_buffer __P((struct pipe *wpipe, struct uio *uio));
  141 static void pipe_destroy_write_buffer __P((struct pipe *wpipe));
  142 static int pipe_direct_write __P((struct pipe *wpipe, struct uio *uio));
  143 static void pipe_clone_write_buffer __P((struct pipe *wpipe));
  144 #endif
  145 static void pipespace __P((struct pipe *cpipe));
  146 
  147 /*
  148  * The pipe system call for the DTYPE_PIPE type of pipes
  149  */
  150 
  151 /* ARGSUSED */
  152 int
  153 pipe(p, uap, retval)
  154         struct proc *p;
  155         struct pipe_args /* {
  156                 int     dummy;
  157         } */ *uap;
  158         int retval[];
  159 {
  160         register struct filedesc *fdp = p->p_fd;
  161         struct file *rf, *wf;
  162         struct pipe *rpipe, *wpipe;
  163         int fd, error;
  164 
  165         rpipe = malloc( sizeof (*rpipe), M_TEMP, M_WAITOK);
  166         pipeinit(rpipe);
  167         rpipe->pipe_state |= PIPE_DIRECTOK;
  168         wpipe = malloc( sizeof (*wpipe), M_TEMP, M_WAITOK);
  169         pipeinit(wpipe);
  170         wpipe->pipe_state |= PIPE_DIRECTOK;
  171 
  172         error = falloc(p, &rf, &fd);
  173         if (error)
  174                 goto free2;
  175         retval[0] = fd;
  176         rf->f_flag = FREAD | FWRITE;
  177         rf->f_type = DTYPE_PIPE;
  178         rf->f_ops = &pipeops;
  179         rf->f_data = (caddr_t)rpipe;
  180         error = falloc(p, &wf, &fd);
  181         if (error)
  182                 goto free3;
  183         wf->f_flag = FREAD | FWRITE;
  184         wf->f_type = DTYPE_PIPE;
  185         wf->f_ops = &pipeops;
  186         wf->f_data = (caddr_t)wpipe;
  187         retval[1] = fd;
  188 
  189         rpipe->pipe_peer = wpipe;
  190         wpipe->pipe_peer = rpipe;
  191 
  192         return (0);
  193 free3:
  194         ffree(rf);
  195         fdp->fd_ofiles[retval[0]] = 0;
  196 free2:
  197         (void)pipeclose(wpipe);
  198         (void)pipeclose(rpipe);
  199         return (error);
  200 }
  201 
  202 /*
  203  * Allocate kva for pipe circular buffer, the space is pageable
  204  */
  205 static void
  206 pipespace(cpipe)
  207         struct pipe *cpipe;
  208 {
  209         int npages, error;
  210 
  211         npages = round_page(cpipe->pipe_buffer.size)/PAGE_SIZE;
  212         /*
  213          * Create an object, I don't like the idea of paging to/from
  214          * kernel_object.
  215          * XXX -- minor change needed here for NetBSD/OpenBSD VM systems.
  216          */
  217         cpipe->pipe_buffer.object = vm_object_allocate(OBJT_DEFAULT, npages);
  218         cpipe->pipe_buffer.buffer = (caddr_t) vm_map_min(kernel_map);
  219 
  220         /*
  221          * Insert the object into the kernel map, and allocate kva for it.
  222          * The map entry is, by default, pageable.
  223          * XXX -- minor change needed here for NetBSD/OpenBSD VM systems.
  224          */
  225         error = vm_map_find(kernel_map, cpipe->pipe_buffer.object, 0,
  226                 (vm_offset_t *) &cpipe->pipe_buffer.buffer, 
  227                 cpipe->pipe_buffer.size, 1,
  228                 VM_PROT_ALL, VM_PROT_ALL, 0);
  229 
  230         if (error != KERN_SUCCESS)
  231                 panic("pipeinit: cannot allocate pipe -- out of kvm -- code = %d", error);
  232         amountpipekva += cpipe->pipe_buffer.size;
  233 }
  234 
  235 /*
  236  * initialize and allocate VM and memory for pipe
  237  */
  238 static void
  239 pipeinit(cpipe)
  240         struct pipe *cpipe;
  241 {
  242         int s;
  243 
  244         cpipe->pipe_buffer.in = 0;
  245         cpipe->pipe_buffer.out = 0;
  246         cpipe->pipe_buffer.cnt = 0;
  247         cpipe->pipe_buffer.size = PIPE_SIZE;
  248 
  249         /* Buffer kva gets dynamically allocated */
  250         cpipe->pipe_buffer.buffer = NULL;
  251         /* cpipe->pipe_buffer.object = invalid */
  252 
  253         cpipe->pipe_state = 0;
  254         cpipe->pipe_peer = NULL;
  255         cpipe->pipe_busy = 0;
  256         s = splhigh();
  257         cpipe->pipe_ctime = time;
  258         cpipe->pipe_atime = time;
  259         cpipe->pipe_mtime = time;
  260         splx(s);
  261         bzero(&cpipe->pipe_sel, sizeof cpipe->pipe_sel);
  262         cpipe->pipe_pgid = NO_PID;
  263 
  264 #ifndef PIPE_NODIRECT
  265         /*
  266          * pipe data structure initializations to support direct pipe I/O
  267          */
  268         cpipe->pipe_map.cnt = 0;
  269         cpipe->pipe_map.kva = 0;
  270         cpipe->pipe_map.pos = 0;
  271         cpipe->pipe_map.npages = 0;
  272         /* cpipe->pipe_map.ms[] = invalid */
  273 #endif
  274 }
  275 
  276 
  277 /*
  278  * lock a pipe for I/O, blocking other access
  279  */
  280 static __inline int
  281 pipelock(cpipe, catch)
  282         struct pipe *cpipe;
  283         int catch;
  284 {
  285         int error;
  286         while (cpipe->pipe_state & PIPE_LOCK) {
  287                 cpipe->pipe_state |= PIPE_LWANT;
  288                 if (error = tsleep( cpipe,
  289                         catch?(PRIBIO|PCATCH):PRIBIO, "pipelk", 0)) {
  290                         return error;
  291                 }
  292         }
  293         cpipe->pipe_state |= PIPE_LOCK;
  294         return 0;
  295 }
  296 
  297 /*
  298  * unlock a pipe I/O lock
  299  */
  300 static __inline void
  301 pipeunlock(cpipe)
  302         struct pipe *cpipe;
  303 {
  304         cpipe->pipe_state &= ~PIPE_LOCK;
  305         if (cpipe->pipe_state & PIPE_LWANT) {
  306                 cpipe->pipe_state &= ~PIPE_LWANT;
  307                 wakeup(cpipe);
  308         }
  309 }
  310 
  311 static __inline void
  312 pipeselwakeup(cpipe)
  313         struct pipe *cpipe;
  314 {
  315         struct proc *p;
  316 
  317         if (cpipe->pipe_state & PIPE_SEL) {
  318                 cpipe->pipe_state &= ~PIPE_SEL;
  319                 selwakeup(&cpipe->pipe_sel);
  320         }
  321         if (cpipe->pipe_state & PIPE_ASYNC) {
  322                 if (cpipe->pipe_pgid < 0)
  323                         gsignal(-cpipe->pipe_pgid, SIGIO);
  324                 else if ((p = pfind(cpipe->pipe_pgid)) != NULL)
  325                         psignal(p, SIGIO);
  326         }
  327 }
  328 
  329 /* ARGSUSED */
  330 static int
  331 pipe_read(fp, uio, cred)
  332         struct file *fp;
  333         struct uio *uio;
  334         struct ucred *cred;
  335 {
  336 
  337         struct pipe *rpipe = (struct pipe *) fp->f_data;
  338         int error = 0;
  339         int nread = 0;
  340         u_int size;
  341 
  342         ++rpipe->pipe_busy;
  343         while (uio->uio_resid) {
  344                 /*
  345                  * normal pipe buffer receive
  346                  */
  347                 if (rpipe->pipe_buffer.cnt > 0) {
  348                         size = rpipe->pipe_buffer.size - rpipe->pipe_buffer.out;
  349                         if (size > rpipe->pipe_buffer.cnt)
  350                                 size = rpipe->pipe_buffer.cnt;
  351                         if (size > (u_int) uio->uio_resid)
  352                                 size = (u_int) uio->uio_resid;
  353                         if ((error = pipelock(rpipe,1)) == 0) {
  354                                 error = uiomove( &rpipe->pipe_buffer.buffer[rpipe->pipe_buffer.out], 
  355                                         size, uio);
  356                                 pipeunlock(rpipe);
  357                         }
  358                         if (error) {
  359                                 break;
  360                         }
  361                         rpipe->pipe_buffer.out += size;
  362                         if (rpipe->pipe_buffer.out >= rpipe->pipe_buffer.size)
  363                                 rpipe->pipe_buffer.out = 0;
  364 
  365                         rpipe->pipe_buffer.cnt -= size;
  366                         nread += size;
  367 #ifndef PIPE_NODIRECT
  368                 /*
  369                  * Direct copy, bypassing a kernel buffer.
  370                  */
  371                 } else if ((size = rpipe->pipe_map.cnt) &&
  372                         (rpipe->pipe_state & PIPE_DIRECTW)) {
  373                         caddr_t va;
  374                         if (size > (u_int) uio->uio_resid)
  375                                 size = (u_int) uio->uio_resid;
  376                         if ((error = pipelock(rpipe,1)) == 0) {
  377                                 va = (caddr_t) rpipe->pipe_map.kva + rpipe->pipe_map.pos;
  378                                 error = uiomove(va, size, uio);
  379                                 pipeunlock(rpipe);
  380                         }
  381                         if (error)
  382                                 break;
  383                         nread += size;
  384                         rpipe->pipe_map.pos += size;
  385                         rpipe->pipe_map.cnt -= size;
  386                         if (rpipe->pipe_map.cnt == 0) {
  387                                 rpipe->pipe_state &= ~PIPE_DIRECTW;
  388                                 wakeup(rpipe);
  389                         }
  390 #endif
  391                 } else {
  392                         /*
  393                          * detect EOF condition
  394                          */
  395                         if (rpipe->pipe_state & PIPE_EOF) {
  396                                 /* XXX error = ? */
  397                                 break;
  398                         }
  399                         /*
  400                          * If the "write-side" has been blocked, wake it up now.
  401                          */
  402                         if (rpipe->pipe_state & PIPE_WANTW) {
  403                                 rpipe->pipe_state &= ~PIPE_WANTW;
  404                                 wakeup(rpipe);
  405                         }
  406                         if (nread > 0)
  407                                 break;
  408 
  409                         if (fp->f_flag & FNONBLOCK) {
  410                                 error = EAGAIN;
  411                                 break;
  412                         }
  413 
  414                         /*
  415                          * If there is no more to read in the pipe, reset
  416                          * its pointers to the beginning.  This improves
  417                          * cache hit stats.
  418                          */
  419                 
  420                         if ((error = pipelock(rpipe,1)) == 0) {
  421                                 if (rpipe->pipe_buffer.cnt == 0) {
  422                                         rpipe->pipe_buffer.in = 0;
  423                                         rpipe->pipe_buffer.out = 0;
  424                                 }
  425                                 pipeunlock(rpipe);
  426                         } else {
  427                                 break;
  428                         }
  429 
  430                         if (rpipe->pipe_state & PIPE_WANTW) {
  431                                 rpipe->pipe_state &= ~PIPE_WANTW;
  432                                 wakeup(rpipe);
  433                         }
  434 
  435                         rpipe->pipe_state |= PIPE_WANTR;
  436                         if (error = tsleep(rpipe, PRIBIO|PCATCH, "piperd", 0)) {
  437                                 break;
  438                         }
  439                 }
  440         }
  441 
  442         if (error == 0) {
  443                 int s = splhigh();
  444                 rpipe->pipe_atime = time;
  445                 splx(s);
  446         }
  447 
  448         --rpipe->pipe_busy;
  449         if ((rpipe->pipe_busy == 0) && (rpipe->pipe_state & PIPE_WANT)) {
  450                 rpipe->pipe_state &= ~(PIPE_WANT|PIPE_WANTW);
  451                 wakeup(rpipe);
  452         } else if (rpipe->pipe_buffer.cnt < MINPIPESIZE) {
  453                 /*
  454                  * If there is no more to read in the pipe, reset
  455                  * its pointers to the beginning.  This improves
  456                  * cache hit stats.
  457                  */
  458                 if (rpipe->pipe_buffer.cnt == 0) {
  459                         if ((error == 0) && (error = pipelock(rpipe,1)) == 0) {
  460                                 rpipe->pipe_buffer.in = 0;
  461                                 rpipe->pipe_buffer.out = 0;
  462                                 pipeunlock(rpipe);
  463                         }
  464                 }
  465 
  466                 /*
  467                  * If the "write-side" has been blocked, wake it up now.
  468                  */
  469                 if (rpipe->pipe_state & PIPE_WANTW) {
  470                         rpipe->pipe_state &= ~PIPE_WANTW;
  471                         wakeup(rpipe);
  472                 }
  473         }
  474 
  475         if ((rpipe->pipe_buffer.size - rpipe->pipe_buffer.cnt) >= PIPE_BUF)
  476                 pipeselwakeup(rpipe);
  477 
  478         return error;
  479 }
  480 
  481 #ifndef PIPE_NODIRECT
  482 /*
  483  * Map the sending processes' buffer into kernel space and wire it.
  484  * This is similar to a physical write operation.
  485  */
  486 static int
  487 pipe_build_write_buffer(wpipe, uio)
  488         struct pipe *wpipe;
  489         struct uio *uio;
  490 {
  491         u_int size;
  492         int i;
  493         vm_offset_t addr, endaddr, paddr;
  494 
  495         size = (u_int) uio->uio_iov->iov_len;
  496         if (size > wpipe->pipe_buffer.size)
  497                 size = wpipe->pipe_buffer.size;
  498 
  499         endaddr = round_page(uio->uio_iov->iov_base + size);
  500         for(i = 0, addr = trunc_page(uio->uio_iov->iov_base);
  501                 addr < endaddr;
  502                 addr += PAGE_SIZE, i+=1) {
  503 
  504                 vm_page_t m;
  505 
  506                 vm_fault_quick( (caddr_t) addr, VM_PROT_READ);
  507                 paddr = pmap_kextract(addr);
  508                 if (!paddr) {
  509                         int j;
  510                         for(j=0;j<i;j++)
  511                                 vm_page_unwire(wpipe->pipe_map.ms[j]);
  512                         return EFAULT;
  513                 }
  514 
  515                 m = PHYS_TO_VM_PAGE(paddr);
  516                 vm_page_wire(m);
  517                 wpipe->pipe_map.ms[i] = m;
  518         }
  519 
  520 /*
  521  * set up the control block
  522  */
  523         wpipe->pipe_map.npages = i;
  524         wpipe->pipe_map.pos = ((vm_offset_t) uio->uio_iov->iov_base) & PAGE_MASK;
  525         wpipe->pipe_map.cnt = size;
  526 
  527 /*
  528  * and map the buffer
  529  */
  530         if (wpipe->pipe_map.kva == 0) {
  531                 /*
  532                  * We need to allocate space for an extra page because the
  533                  * address range might (will) span pages at times.
  534                  */
  535                 wpipe->pipe_map.kva = kmem_alloc_pageable(kernel_map,
  536                         wpipe->pipe_buffer.size + PAGE_SIZE);
  537                 amountpipekva += wpipe->pipe_buffer.size + PAGE_SIZE;
  538         }
  539         pmap_qenter(wpipe->pipe_map.kva, wpipe->pipe_map.ms,
  540                 wpipe->pipe_map.npages);
  541 
  542 /*
  543  * and update the uio data
  544  */
  545 
  546         uio->uio_iov->iov_len -= size;
  547         uio->uio_iov->iov_base += size;
  548         if (uio->uio_iov->iov_len == 0)
  549                 uio->uio_iov++;
  550         uio->uio_resid -= size;
  551         uio->uio_offset += size;
  552         return 0;
  553 }
  554 
  555 /*
  556  * unmap and unwire the process buffer
  557  */
  558 static void
  559 pipe_destroy_write_buffer(wpipe)
  560 struct pipe *wpipe;
  561 {
  562         int i;
  563         if (wpipe->pipe_map.kva) {
  564                 pmap_qremove(wpipe->pipe_map.kva, wpipe->pipe_map.npages);
  565 
  566                 if (amountpipekva > MAXPIPEKVA) {
  567                         vm_offset_t kva = wpipe->pipe_map.kva;
  568                         wpipe->pipe_map.kva = 0;
  569                         kmem_free(kernel_map, kva,
  570                                 wpipe->pipe_buffer.size + PAGE_SIZE);
  571                         amountpipekva -= wpipe->pipe_buffer.size + PAGE_SIZE;
  572                 }
  573         }
  574         for (i=0;i<wpipe->pipe_map.npages;i++)
  575                 vm_page_unwire(wpipe->pipe_map.ms[i]);
  576 }
  577 
  578 /*
  579  * In the case of a signal, the writing process might go away.  This
  580  * code copies the data into the circular buffer so that the source
  581  * pages can be freed without loss of data.
  582  */
  583 static void
  584 pipe_clone_write_buffer(wpipe)
  585 struct pipe *wpipe;
  586 {
  587         int size;
  588         int pos;
  589 
  590         size = wpipe->pipe_map.cnt;
  591         pos = wpipe->pipe_map.pos;
  592         bcopy((caddr_t) wpipe->pipe_map.kva+pos,
  593                         (caddr_t) wpipe->pipe_buffer.buffer,
  594                         size);
  595 
  596         wpipe->pipe_buffer.in = size;
  597         wpipe->pipe_buffer.out = 0;
  598         wpipe->pipe_buffer.cnt = size;
  599         wpipe->pipe_state &= ~PIPE_DIRECTW;
  600 
  601         pipe_destroy_write_buffer(wpipe);
  602 }
  603 
  604 /*
  605  * This implements the pipe buffer write mechanism.  Note that only
  606  * a direct write OR a normal pipe write can be pending at any given time.
  607  * If there are any characters in the pipe buffer, the direct write will
  608  * be deferred until the receiving process grabs all of the bytes from
  609  * the pipe buffer.  Then the direct mapping write is set-up.
  610  */
  611 static int
  612 pipe_direct_write(wpipe, uio)
  613         struct pipe *wpipe;
  614         struct uio *uio;
  615 {
  616         int error;
  617 retry:
  618         while (wpipe->pipe_state & PIPE_DIRECTW) {
  619                 if ( wpipe->pipe_state & PIPE_WANTR) {
  620                         wpipe->pipe_state &= ~PIPE_WANTR;
  621                         wakeup(wpipe);
  622                 }
  623                 wpipe->pipe_state |= PIPE_WANTW;
  624                 error = tsleep(wpipe,
  625                                 PRIBIO|PCATCH, "pipdww", 0);
  626                 if (error)
  627                         goto error1;
  628                 if (wpipe->pipe_state & PIPE_EOF) {
  629                         error = EPIPE;
  630                         goto error1;
  631                 }
  632         }
  633         wpipe->pipe_map.cnt = 0;        /* transfer not ready yet */
  634         if (wpipe->pipe_buffer.cnt > 0) {
  635                 if ( wpipe->pipe_state & PIPE_WANTR) {
  636                         wpipe->pipe_state &= ~PIPE_WANTR;
  637                         wakeup(wpipe);
  638                 }
  639                         
  640                 wpipe->pipe_state |= PIPE_WANTW;
  641                 error = tsleep(wpipe,
  642                                 PRIBIO|PCATCH, "pipdwc", 0);
  643                 if (error)
  644                         goto error1;
  645                 if (wpipe->pipe_state & PIPE_EOF) {
  646                         error = EPIPE;
  647                         goto error1;
  648                 }
  649                 goto retry;
  650         }
  651 
  652         wpipe->pipe_state |= PIPE_DIRECTW;
  653 
  654         error = pipe_build_write_buffer(wpipe, uio);
  655         if (error) {
  656                 wpipe->pipe_state &= ~PIPE_DIRECTW;
  657                 goto error1;
  658         }
  659 
  660         error = 0;
  661         while (!error && (wpipe->pipe_state & PIPE_DIRECTW)) {
  662                 if (wpipe->pipe_state & PIPE_EOF) {
  663                         pipelock(wpipe, 0);
  664                         pipe_destroy_write_buffer(wpipe);
  665                         pipeunlock(wpipe);
  666                         pipeselwakeup(wpipe);
  667                         error = EPIPE;
  668                         goto error1;
  669                 }
  670                 if (wpipe->pipe_state & PIPE_WANTR) {
  671                         wpipe->pipe_state &= ~PIPE_WANTR;
  672                         wakeup(wpipe);
  673                 }
  674                 pipeselwakeup(wpipe);
  675                 error = tsleep(wpipe, PRIBIO|PCATCH, "pipdwt", 0);
  676         }
  677 
  678         pipelock(wpipe,0);
  679         if (wpipe->pipe_state & PIPE_DIRECTW) {
  680                 /*
  681                  * this bit of trickery substitutes a kernel buffer for
  682                  * the process that might be going away.
  683                  */
  684                 pipe_clone_write_buffer(wpipe);
  685         } else {
  686                 pipe_destroy_write_buffer(wpipe);
  687         }
  688         pipeunlock(wpipe);
  689         return error;
  690 
  691 error1:
  692         wakeup(wpipe);
  693         return error;
  694 }
  695 #endif
  696         
  697 static int
  698 pipe_write(fp, uio, cred)
  699         struct file *fp;
  700         struct uio *uio;
  701         struct ucred *cred;
  702 {
  703         int error = 0;
  704         int orig_resid;
  705 
  706         struct pipe *wpipe, *rpipe;
  707 
  708         rpipe = (struct pipe *) fp->f_data;
  709         wpipe = rpipe->pipe_peer;
  710 
  711         /*
  712          * detect loss of pipe read side, issue SIGPIPE if lost.
  713          */
  714         if ((wpipe == NULL) || (wpipe->pipe_state & PIPE_EOF)) {
  715                 return EPIPE;
  716         }
  717 
  718         /*
  719          * If it is advantageous to resize the pipe buffer, do
  720          * so.
  721          */
  722         if ((uio->uio_resid > PIPE_SIZE) &&
  723                 (nbigpipe < LIMITBIGPIPES) &&
  724                 (wpipe->pipe_state & PIPE_DIRECTW) == 0 &&
  725                 (wpipe->pipe_buffer.size <= PIPE_SIZE) &&
  726                 (wpipe->pipe_buffer.cnt == 0)) {
  727 
  728                 if (wpipe->pipe_buffer.buffer) {
  729                         amountpipekva -= wpipe->pipe_buffer.size;
  730                         kmem_free(kernel_map,
  731                                 (vm_offset_t)wpipe->pipe_buffer.buffer,
  732                                 wpipe->pipe_buffer.size);
  733                 }
  734 
  735 #ifndef PIPE_NODIRECT
  736                 if (wpipe->pipe_map.kva) {
  737                         amountpipekva -= wpipe->pipe_buffer.size + PAGE_SIZE;
  738                         kmem_free(kernel_map,
  739                                 wpipe->pipe_map.kva,
  740                                 wpipe->pipe_buffer.size + PAGE_SIZE);
  741                 }
  742 #endif
  743 
  744                 wpipe->pipe_buffer.in = 0;
  745                 wpipe->pipe_buffer.out = 0;
  746                 wpipe->pipe_buffer.cnt = 0;
  747                 wpipe->pipe_buffer.size = BIG_PIPE_SIZE;
  748                 wpipe->pipe_buffer.buffer = NULL;
  749                 ++nbigpipe;
  750 
  751 #ifndef PIPE_NODIRECT
  752                 wpipe->pipe_map.cnt = 0;
  753                 wpipe->pipe_map.kva = 0;
  754                 wpipe->pipe_map.pos = 0;
  755                 wpipe->pipe_map.npages = 0;
  756 #endif
  757 
  758         }
  759                 
  760 
  761         if( wpipe->pipe_buffer.buffer == NULL) {
  762                 if ((error = pipelock(wpipe,1)) == 0) {
  763                         pipespace(wpipe);
  764                         pipeunlock(wpipe);
  765                 } else {
  766                         return error;
  767                 }
  768         }
  769 
  770         ++wpipe->pipe_busy;
  771         orig_resid = uio->uio_resid;
  772         while (uio->uio_resid) {
  773                 int space;
  774 #ifndef PIPE_NODIRECT
  775                 /*
  776                  * If the transfer is large, we can gain performance if
  777                  * we do process-to-process copies directly.
  778                  * If the write is non-blocking, we don't use the
  779                  * direct write mechanism.
  780                  */
  781                 if ((uio->uio_iov->iov_len >= PIPE_MINDIRECT) &&
  782                     (fp->f_flag & FNONBLOCK) == 0 &&
  783                         (wpipe->pipe_map.kva || (amountpipekva < LIMITPIPEKVA)) &&
  784                         (uio->uio_iov->iov_len >= PIPE_MINDIRECT)) {
  785                         error = pipe_direct_write( wpipe, uio);
  786                         if (error) {
  787                                 break;
  788                         }
  789                         continue;
  790                 }
  791 #endif
  792 
  793                 /*
  794                  * Pipe buffered writes cannot be coincidental with
  795                  * direct writes.  We wait until the currently executing
  796                  * direct write is completed before we start filling the
  797                  * pipe buffer.
  798                  */
  799         retrywrite:
  800                 while (wpipe->pipe_state & PIPE_DIRECTW) {
  801                         if (wpipe->pipe_state & PIPE_WANTR) {
  802                                 wpipe->pipe_state &= ~PIPE_WANTR;
  803                                 wakeup(wpipe);
  804                         }
  805                         error = tsleep(wpipe,
  806                                         PRIBIO|PCATCH, "pipbww", 0);
  807                         if (error)
  808                                 break;
  809                 }
  810 
  811                 space = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt;
  812 
  813                 /* Writes of size <= PIPE_BUF must be atomic. */
  814                 /* XXX perhaps they need to be contiguous to be atomic? */
  815                 if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
  816                         space = 0;
  817 
  818                 if (space > 0 && (wpipe->pipe_buffer.cnt < PIPE_SIZE)) {
  819                         /*
  820                          * This set the maximum transfer as a segment of
  821                          * the buffer.
  822                          */
  823                         int size = wpipe->pipe_buffer.size - wpipe->pipe_buffer.in;
  824                         /*
  825                          * space is the size left in the buffer
  826                          */
  827                         if (size > space)
  828                                 size = space;
  829                         /*
  830                          * now limit it to the size of the uio transfer
  831                          */
  832                         if (size > uio->uio_resid)
  833                                 size = uio->uio_resid;
  834                         if ((error = pipelock(wpipe,1)) == 0) {
  835                                 /*
  836                                  * It is possible for a direct write to
  837                                  * slip in on us... handle it here...
  838                                  */
  839                                 if (wpipe->pipe_state & PIPE_DIRECTW) {
  840                                         pipeunlock(wpipe);
  841                                         goto retrywrite;
  842                                 }
  843                                 error = uiomove( &wpipe->pipe_buffer.buffer[wpipe->pipe_buffer.in], 
  844                                         size, uio);
  845                                 pipeunlock(wpipe);
  846                         }
  847                         if (error)
  848                                 break;
  849 
  850                         wpipe->pipe_buffer.in += size;
  851                         if (wpipe->pipe_buffer.in >= wpipe->pipe_buffer.size)
  852                                 wpipe->pipe_buffer.in = 0;
  853 
  854                         wpipe->pipe_buffer.cnt += size;
  855                 } else {
  856                         /*
  857                          * If the "read-side" has been blocked, wake it up now.
  858                          */
  859                         if (wpipe->pipe_state & PIPE_WANTR) {
  860                                 wpipe->pipe_state &= ~PIPE_WANTR;
  861                                 wakeup(wpipe);
  862                         }
  863 
  864                         /*
  865                          * don't block on non-blocking I/O
  866                          */
  867                         if (fp->f_flag & FNONBLOCK) {
  868                                 error = EAGAIN;
  869                                 break;
  870                         }
  871 
  872                         /*
  873                          * We have no more space and have something to offer,
  874                          * wake up selects.
  875                          */
  876                         pipeselwakeup(wpipe);
  877 
  878                         wpipe->pipe_state |= PIPE_WANTW;
  879                         if (error = tsleep(wpipe, (PRIBIO+1)|PCATCH, "pipewr", 0)) {
  880                                 break;
  881                         }
  882                         /*
  883                          * If read side wants to go away, we just issue a signal
  884                          * to ourselves.
  885                          */
  886                         if (wpipe->pipe_state & PIPE_EOF) {
  887                                 error = EPIPE;
  888                                 break;
  889                         }       
  890                 }
  891         }
  892 
  893         --wpipe->pipe_busy;
  894         if ((wpipe->pipe_busy == 0) &&
  895                 (wpipe->pipe_state & PIPE_WANT)) {
  896                 wpipe->pipe_state &= ~(PIPE_WANT|PIPE_WANTR);
  897                 wakeup(wpipe);
  898         } else if (wpipe->pipe_buffer.cnt > 0) {
  899                 /*
  900                  * If we have put any characters in the buffer, we wake up
  901                  * the reader.
  902                  */
  903                 if (wpipe->pipe_state & PIPE_WANTR) {
  904                         wpipe->pipe_state &= ~PIPE_WANTR;
  905                         wakeup(wpipe);
  906                 }
  907         }
  908 
  909         /*
  910          * Don't return EPIPE if I/O was successful
  911          */
  912         if ((wpipe->pipe_buffer.cnt == 0) &&
  913                 (uio->uio_resid == 0) &&
  914                 (error == EPIPE))
  915                 error = 0;
  916 
  917         if (error == 0) {
  918                 int s = splhigh();
  919                 wpipe->pipe_mtime = time;
  920                 splx(s);
  921         }
  922         /*
  923          * We have something to offer,
  924          * wake up select.
  925          */
  926         if (wpipe->pipe_buffer.cnt)
  927                 pipeselwakeup(wpipe);
  928 
  929         return error;
  930 }
  931 
  932 /*
  933  * we implement a very minimal set of ioctls for compatibility with sockets.
  934  */
  935 int
  936 pipe_ioctl(fp, cmd, data, p)
  937         struct file *fp;
  938         int cmd;
  939         register caddr_t data;
  940         struct proc *p;
  941 {
  942         register struct pipe *mpipe = (struct pipe *)fp->f_data;
  943 
  944         switch (cmd) {
  945 
  946         case FIONBIO:
  947                 return (0);
  948 
  949         case FIOASYNC:
  950                 if (*(int *)data) {
  951                         mpipe->pipe_state |= PIPE_ASYNC;
  952                 } else {
  953                         mpipe->pipe_state &= ~PIPE_ASYNC;
  954                 }
  955                 return (0);
  956 
  957         case FIONREAD:
  958                 if (mpipe->pipe_state & PIPE_DIRECTW)
  959                         *(int *)data = mpipe->pipe_map.cnt;
  960                 else
  961                         *(int *)data = mpipe->pipe_buffer.cnt;
  962                 return (0);
  963 
  964         case TIOCSPGRP:
  965                 mpipe->pipe_pgid = *(int *)data;
  966                 return (0);
  967 
  968         case TIOCGPGRP:
  969                 *(int *)data = mpipe->pipe_pgid;
  970                 return (0);
  971 
  972         }
  973         return (ENOTTY);
  974 }
  975 
  976 int
  977 pipe_select(fp, which, p)
  978         struct file *fp;
  979         int which;
  980         struct proc *p;
  981 {
  982         register struct pipe *rpipe = (struct pipe *)fp->f_data;
  983         struct pipe *wpipe;
  984 
  985         wpipe = rpipe->pipe_peer;
  986         switch (which) {
  987 
  988         case FREAD:
  989                 if ( (rpipe->pipe_state & PIPE_DIRECTW) ||
  990                         (rpipe->pipe_buffer.cnt > 0) ||
  991                         (rpipe->pipe_state & PIPE_EOF)) {
  992                         return (1);
  993                 }
  994                 selrecord(p, &rpipe->pipe_sel);
  995                 rpipe->pipe_state |= PIPE_SEL;
  996                 break;
  997 
  998         case FWRITE:
  999                 if ((wpipe == NULL) ||
 1000                         (wpipe->pipe_state & PIPE_EOF) ||
 1001                         (((wpipe->pipe_state & PIPE_DIRECTW) == 0) &&
 1002                          (wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt) >= PIPE_BUF)) {
 1003                         return (1);
 1004                 }
 1005                 selrecord(p, &wpipe->pipe_sel);
 1006                 wpipe->pipe_state |= PIPE_SEL;
 1007                 break;
 1008 
 1009         case 0:
 1010                 if ((rpipe->pipe_state & PIPE_EOF) ||
 1011                         (wpipe == NULL) ||
 1012                         (wpipe->pipe_state & PIPE_EOF)) {
 1013                         return (1);
 1014                 }
 1015                         
 1016                 selrecord(p, &rpipe->pipe_sel);
 1017                 rpipe->pipe_state |= PIPE_SEL;
 1018                 break;
 1019         }
 1020         return (0);
 1021 }
 1022 
 1023 int
 1024 pipe_stat(pipe, ub)
 1025         register struct pipe *pipe;
 1026         register struct stat *ub;
 1027 {
 1028         bzero((caddr_t)ub, sizeof (*ub));
 1029         ub->st_mode = S_IFIFO;
 1030         ub->st_blksize = pipe->pipe_buffer.size;
 1031         ub->st_size = pipe->pipe_buffer.cnt;
 1032         ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize;
 1033         TIMEVAL_TO_TIMESPEC(&pipe->pipe_atime, &ub->st_atimespec);
 1034         TIMEVAL_TO_TIMESPEC(&pipe->pipe_mtime, &ub->st_mtimespec);
 1035         TIMEVAL_TO_TIMESPEC(&pipe->pipe_ctime, &ub->st_ctimespec);
 1036         /*
 1037          * Left as 0: st_dev, st_ino, st_nlink, st_uid, st_gid, st_rdev,
 1038          * st_flags, st_gen.
 1039          * XXX (st_dev, st_ino) should be unique.
 1040          */
 1041         return 0;
 1042 }
 1043 
 1044 /* ARGSUSED */
 1045 static int
 1046 pipe_close(fp, p)
 1047         struct file *fp;
 1048         struct proc *p;
 1049 {
 1050         struct pipe *cpipe = (struct pipe *)fp->f_data;
 1051 
 1052         pipeclose(cpipe);
 1053         fp->f_data = NULL;
 1054         return 0;
 1055 }
 1056 
 1057 /*
 1058  * shutdown the pipe
 1059  */
 1060 static void
 1061 pipeclose(cpipe)
 1062         struct pipe *cpipe;
 1063 {
 1064         struct pipe *ppipe;
 1065         if (cpipe) {
 1066                 
 1067                 pipeselwakeup(cpipe);
 1068 
 1069                 /*
 1070                  * If the other side is blocked, wake it up saying that
 1071                  * we want to close it down.
 1072                  */
 1073                 while (cpipe->pipe_busy) {
 1074                         wakeup(cpipe);
 1075                         cpipe->pipe_state |= PIPE_WANT|PIPE_EOF;
 1076                         tsleep(cpipe, PRIBIO, "pipecl", 0);
 1077                 }
 1078 
 1079                 /*
 1080                  * Disconnect from peer
 1081                  */
 1082                 if (ppipe = cpipe->pipe_peer) {
 1083                         pipeselwakeup(ppipe);
 1084 
 1085                         ppipe->pipe_state |= PIPE_EOF;
 1086                         wakeup(ppipe);
 1087                         ppipe->pipe_peer = NULL;
 1088                 }
 1089 
 1090                 /*
 1091                  * free resources
 1092                  */
 1093                 if (cpipe->pipe_buffer.buffer) {
 1094                         if (cpipe->pipe_buffer.size > PIPE_SIZE)
 1095                                 --nbigpipe;
 1096                         amountpipekva -= cpipe->pipe_buffer.size;
 1097                         kmem_free(kernel_map,
 1098                                 (vm_offset_t)cpipe->pipe_buffer.buffer,
 1099                                 cpipe->pipe_buffer.size);
 1100                 }
 1101 #ifndef PIPE_NODIRECT
 1102                 if (cpipe->pipe_map.kva) {
 1103                         amountpipekva -= cpipe->pipe_buffer.size + PAGE_SIZE;
 1104                         kmem_free(kernel_map,
 1105                                 cpipe->pipe_map.kva,
 1106                                 cpipe->pipe_buffer.size + PAGE_SIZE);
 1107                 }
 1108 #endif
 1109                 free(cpipe, M_TEMP);
 1110         }
 1111 }
 1112 #endif

Cache object: 9c45dfca677ebc517323ebe9f3a594c8


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.