IO-AIO/AIO.xs

#define _REENTRANT 1
#include <errno.h>

#include "EXTERN.h"
#include "perl.h"
#include "XSUB.h"

#include "autoconf/config.h"

#include <pthread.h>

#include <sys/types.h>
#include <sys/stat.h>

#include <unistd.h>
#include <fcntl.h>
#include <signal.h>
#include <sched.h>

#if HAVE_SENDFILE
# if __linux
#  include <sys/sendfile.h>
# elif __freebsd
#  include <sys/socket.h>
#  include <sys/uio.h>
# elif __hpux
#  include <sys/socket.h>
# elif __solaris /* not yet */
#  include <sys/sendfile.h>
# else
#  error sendfile support requested but not available
# endif
#endif

#if __ia64
# define STACKSIZE 65536
#else
# define STACKSIZE  4096
#endif

enum {
  REQ_QUIT,
  REQ_OPEN, REQ_CLOSE,
  REQ_READ, REQ_WRITE, REQ_READAHEAD,
  REQ_SENDFILE,
  REQ_STAT, REQ_LSTAT, REQ_FSTAT,
  REQ_FSYNC, REQ_FDATASYNC,
  REQ_UNLINK, REQ_RMDIR,
  REQ_SYMLINK,
};

typedef struct aio_cb {
  struct aio_cb *volatile next;

  int type;

  int fd, fd2;
  off_t offset;
  size_t length;
  ssize_t result;
  mode_t mode; /* open */
  int errorno;
  SV *data, *callback;
  SV *fh, *fh2;
  void *dataptr, *data2ptr;
  STRLEN dataoffset;

  Stat_t *statdata;
} aio_cb;

typedef aio_cb *aio_req;

static int started, wanted;
static volatile int nreqs;
static int max_outstanding = 1<<30;
static int respipe [2];

static pthread_mutex_t reslock = PTHREAD_MUTEX_INITIALIZER;
static pthread_mutex_t reqlock = PTHREAD_MUTEX_INITIALIZER;
static pthread_cond_t  reqwait = PTHREAD_COND_INITIALIZER;

static volatile aio_req reqs, reqe; /* queue start, queue end */
static volatile aio_req ress, rese; /* queue start, queue end */

static void free_req (aio_req req)
{
  if (req->data)
    SvREFCNT_dec (req->data);

  if (req->fh)
    SvREFCNT_dec (req->fh);

  if (req->fh2)
    SvREFCNT_dec (req->fh2);

  if (req->statdata)
    Safefree (req->statdata);

  if (req->callback)
    SvREFCNT_dec (req->callback);

  Safefree (req);
}

static void
poll_wait ()
{
  if (nreqs && !ress)
    {
      fd_set rfd;
      FD_ZERO(&rfd);
      FD_SET(respipe [0], &rfd);

      select (respipe [0] + 1, &rfd, 0, 0, 0);
    }
}

static int
poll_cb ()
{
  dSP;
  int count = 0;
  int do_croak = 0;
  aio_req req;

  for (;;)
    {
      pthread_mutex_lock (&reslock);
      req = ress;

      if (req)
        {
          ress = req->next;

          if (!ress)
            {
              /* read any signals sent by the worker threads */
              char buf [32];
              while (read (respipe [0], buf, 32) == 32)
                ;

              rese = 0;
            }
        }

      pthread_mutex_unlock (&reslock);

      if (!req)
        break;

      nreqs--;

      if (req->type == REQ_QUIT)
        started--;
      else
        {
          int errorno = errno;
          errno = req->errorno;

          if (req->type == REQ_READ)
            SvCUR_set (req->data, req->dataoffset + (req->result > 0 ? req->result : 0));

          if (req->data2ptr && (req->type == REQ_READ || req->type == REQ_WRITE))
            SvREADONLY_off (req->data);

          if (req->statdata)
            {
              PL_laststype   = req->type == REQ_LSTAT ? OP_LSTAT : OP_STAT;
              PL_laststatval = req->result;
              PL_statcache   = *(req->statdata);
            }

          ENTER;
          PUSHMARK (SP);
          XPUSHs (sv_2mortal (newSViv (req->result)));

          if (req->type == REQ_OPEN)
            {
              /* convert fd to fh */
              SV *fh;

              PUTBACK;
              call_pv ("IO::AIO::_fd2fh", G_SCALAR | G_EVAL);
              SPAGAIN;

              fh = SvREFCNT_inc (POPs);

              PUSHMARK (SP);
              XPUSHs (sv_2mortal (fh));
            }

          if (SvOK (req->callback))
            {
              PUTBACK;
              call_sv (req->callback, G_VOID | G_EVAL);
              SPAGAIN;

              if (SvTRUE (ERRSV))
                {
                  free_req (req);
                  croak (0);
                }
            }

          LEAVE;

          errno = errorno;
          count++;
        }

      free_req (req);
    }

  return count;
}

static void *aio_proc(void *arg);

static void
start_thread (void)
{
  sigset_t fullsigset, oldsigset;
  pthread_t tid;
  pthread_attr_t attr;

  pthread_attr_init (&attr);
  pthread_attr_setstacksize (&attr, STACKSIZE);
  pthread_attr_setdetachstate (&attr, PTHREAD_CREATE_DETACHED);

  sigfillset (&fullsigset);
  sigprocmask (SIG_SETMASK, &fullsigset, &oldsigset);

  if (pthread_create (&tid, &attr, aio_proc, 0) == 0)
    started++;

  sigprocmask (SIG_SETMASK, &oldsigset, 0);
}

static void
send_req (aio_req req)
{
  while (started < wanted && nreqs >= started)
    start_thread ();

  nreqs++;

  pthread_mutex_lock (&reqlock);

  req->next = 0;

  if (reqe)
    {
      reqe->next = req;
      reqe = req;
    }
  else
    reqe = reqs = req;

  pthread_cond_signal (&reqwait);
  pthread_mutex_unlock (&reqlock);

  if (nreqs > max_outstanding)
    for (;;)
      {
        poll_cb ();

        if (nreqs <= max_outstanding)
          break;

        poll_wait ();
      }
}

static void
end_thread (void)
{
  aio_req req;
  Newz (0, req, 1, aio_cb);
  req->type = REQ_QUIT;

  send_req (req);
}

static void min_parallel (int nthreads)
{
  if (wanted < nthreads)
    wanted = nthreads;
}

static void max_parallel (int nthreads)
{
  int cur = started;

  if (wanted > nthreads)
    wanted = nthreads;

  while (cur > wanted)
    {
      end_thread ();
      cur--;
    }

  while (started > wanted)
    {
      poll_wait ();
      poll_cb ();
    }
}

static void create_pipe ()
{
  if (pipe (respipe))
    croak ("unable to initialize result pipe");

  if (fcntl (respipe [0], F_SETFL, O_NONBLOCK))
    croak ("cannot set result pipe to nonblocking mode");

  if (fcntl (respipe [1], F_SETFL, O_NONBLOCK))
    croak ("cannot set result pipe to nonblocking mode");
}

static void atfork_prepare (void)
{
  pthread_mutex_lock (&reqlock);
  pthread_mutex_lock (&reslock);
}

static void atfork_parent (void)
{
  pthread_mutex_unlock (&reslock);
  pthread_mutex_unlock (&reqlock);
}

static void atfork_child (void)
{
  aio_req prv;

  started = 0;

  while (reqs)
    {
      prv = reqs;
      reqs = prv->next;
      free_req (prv);
    }

  reqs = reqe = 0;
      
  while (ress)
    {
      prv = ress;
      ress = prv->next;
      free_req (prv);
    }
      
  ress = rese = 0;

  close (respipe [0]);
  close (respipe [1]);
  create_pipe ();

  atfork_parent ();
}

/* currently noops */
#define LOCK_FD(fd)   do { } while (0)
#define UNLOCK_FD(fd) do { } while (0)

/*****************************************************************************/
/* work around various missing functions */

#if !HAVE_PREADWRITE
# define pread  aio_pread
# define pwrite aio_pwrite

/*
 * make our pread/pwrite safe against themselves, but not against
 * normal read/write by using a mutex. slows down execution a lot,
 * but that's your problem, not mine.
 */
static pthread_mutex_t iolock = PTHREAD_MUTEX_INITIALIZER;

static ssize_t
pread (int fd, void *buf, size_t count, off_t offset)
{
  ssize_t res;
  off_t ooffset;

  LOCK_FD (fd);
  pthread_mutex_lock (&iolock); /* replace by LOCK_FD and private buffer */
  ooffset = lseek (fd, 0, SEEK_CUR);
  lseek (fd, offset, SEEK_SET);
  res = read (fd, buf, count);
  lseek (fd, ooffset, SEEK_SET);
  pthread_mutex_unlock (&iolock);
  UNLOCK_FD (d);

  return res;
}

static ssize_t
pwrite (int fd, void *buf, size_t count, off_t offset)
{
  ssize_t res;
  off_t ooffset;

  LOCK_FD (fd);
  pthread_mutex_lock (&iolock); /* replace by LOCK_FD and private buffer */
  ooffset = lseek (fd, 0, SEEK_CUR);
  lseek (fd, offset, SEEK_SET);
  res = write (fd, buf, count);
  lseek (fd, offset, SEEK_SET);
  pthread_mutex_unlock (&iolock);
  UNLOCK_FD (d);

  return res;
}
#endif

#if !HAVE_FDATASYNC
# define fdatasync fsync
#endif

#if !HAVE_READAHEAD
# define readahead aio_readahead

static char readahead_buf[4096];

static ssize_t
readahead (int fd, off_t offset, size_t count)
{
  while (count > 0)
    {
      size_t len = count < sizeof (readahead_buf) ? count : sizeof (readahead_buf);

      pread (fd, readahead_buf, len, offset);
      offset += len;
      count  -= len;
    }

  errno = 0;
}
#endif

/* sendfile always needs emulation */
static ssize_t
sendfile_ (int ofd, int ifd, off_t offset, size_t count)
{
   ssize_t res;

   if (!count)
     return 0;

   LOCK_FD (ofd);

#if HAVE_SENDFILE
# if __linux
   res = sendfile (ofd, ifd, &offset, count);

# elif __freebsd
   /*
    * Of course, the freebsd sendfile is a dire hack with no thoughts
    * wasted on making it similar to other I/O functions.
    */
   {
     off_t sbytes;
     res = sendfile (ifd, ofd, offset, count, 0, &sbytes, 0);

     if (res < 0 && sbytes)
       /* maybe only on EAGAIN only: as usual, the manpage leaves you guessing */
       res = sbytes;
   }

# elif __hpux
   res = sendfile (ofd, ifd, offset, count, 0, 0);

# elif __solaris
   {
     struct sendfilevec vec;
     size_t sbytes;

     vec.sfv_fd   = ifd;
     vec.sfv_flag = 0;
     vec.sfv_off  = offset;
     vec.sfv_len  = count;

     res = sendfilev (ofd, &vec, 1, &sbytes);

     if (res < 0 && sbytes)
       res = sbytes;
   }

# else
   res = -1;
   errno = ENOSYS;
# endif
#endif

   if (res <  0
       && (errno == ENOSYS || errno == EINVAL || errno == ENOTSOCK
#if __solaris
           || errno == EAFNOSUPPORT || errno == EPROTOTYPE
#endif
          )
       )
     {
       /* emulate sendfile. this is a major pain in the ass */
       char *buf = malloc (4096);
       res = 0;

       for (;;)
         {
           ssize_t cnt;
           
           cnt = pread (ifd, buf, 4096, offset);

           if (cnt <= 0)
             {
               if (cnt && !res) res = -1;
               break;
             }

           cnt = write (ofd, buf, cnt);

           if (cnt <= 0)
             {
               if (cnt && !res) res = -1;
               break;
             }

           offset += cnt;
           res    += cnt;
         }

       {
         int errorno = errno;
         free (buf);
         errno = errorno;
       }
     }

   UNLOCK_FD (ofd);

   return res;
}

/*****************************************************************************/

static void *
aio_proc (void *thr_arg)
{
  aio_req req;
  int type;

  do
    {
      pthread_mutex_lock (&reqlock);

      for (;;)
        {
          req = reqs;

          if (reqs)
            {
              reqs = reqs->next;
              if (!reqs) reqe = 0;
            }

          if (req)
            break;

          pthread_cond_wait (&reqwait, &reqlock);
        }

      pthread_mutex_unlock (&reqlock);
     
      errno = 0; /* strictly unnecessary */

      type = req->type;

      switch (type)
        {
          case REQ_READ:      req->result = pread     (req->fd, req->dataptr, req->length, req->offset); break;
          case REQ_WRITE:     req->result = pwrite    (req->fd, req->dataptr, req->length, req->offset); break;

          case REQ_READAHEAD: req->result = readahead (req->fd, req->offset, req->length); break;
          case REQ_SENDFILE:  req->result = sendfile_ (req->fd, req->fd2, req->offset, req->length); break;

          case REQ_STAT:      req->result = stat      (req->dataptr, req->statdata); break;
          case REQ_LSTAT:     req->result = lstat     (req->dataptr, req->statdata); break;
          case REQ_FSTAT:     req->result = fstat     (req->fd     , req->statdata); break;

          case REQ_OPEN:      req->result = open      (req->dataptr, req->fd, req->mode); break;
          case REQ_CLOSE:     req->result = close     (req->fd); break;
          case REQ_UNLINK:    req->result = unlink    (req->dataptr); break;
          case REQ_RMDIR:     req->result = rmdir     (req->dataptr); break;
          case REQ_SYMLINK:   req->result = symlink   (req->data2ptr, req->dataptr); break;

          case REQ_FDATASYNC: req->result = fdatasync (req->fd); break;
          case REQ_FSYNC:     req->result = fsync     (req->fd); break;

          case REQ_QUIT:
            break;

          default:
            req->result = ENOSYS;
            break;
        }

      req->errorno = errno;

      pthread_mutex_lock (&reslock);

      req->next = 0;

      if (rese)
        {
          rese->next = req;
          rese = req;
        }
      else
        {
          rese = ress = req;

          /* write a dummy byte to the pipe so fh becomes ready */
          write (respipe [1], &respipe, 1);
        }

      pthread_mutex_unlock (&reslock);
    }
  while (type != REQ_QUIT);

  return 0;
}

#define dREQ                                                    \
  aio_req req;                                                  \
                                                                \
  if (SvOK (callback) && !SvROK (callback))                     \
    croak ("clalback must be undef or of reference type");      \
                                                                \
  Newz (0, req, 1, aio_cb);                                     \
  if (!req)                                                     \
    croak ("out of memory during aio_req allocation");          \
                                                                \
  req->callback = newSVsv (callback);
        
MODULE = IO::AIO                PACKAGE = IO::AIO

PROTOTYPES: ENABLE

BOOT:
{
        create_pipe ();
        pthread_atfork (atfork_prepare, atfork_parent, atfork_child);
}

void
min_parallel(nthreads)
        int     nthreads
        PROTOTYPE: $

void
max_parallel(nthreads)
        int     nthreads
        PROTOTYPE: $

int
max_outstanding(nreqs)
        int nreqs
        PROTOTYPE: $
        CODE:
        RETVAL = max_outstanding;
        max_outstanding = nreqs;

void
aio_open(pathname,flags,mode,callback=&PL_sv_undef)
        SV *    pathname
        int     flags
        int     mode
        SV *    callback
        PROTOTYPE: $$$;$
        CODE:
{
        dREQ;

        req->type = REQ_OPEN;
        req->data = newSVsv (pathname);
        req->dataptr = SvPVbyte_nolen (req->data);
        req->fd = flags;
        req->mode = mode;

        send_req (req);
}

void
aio_close(fh,callback=&PL_sv_undef)
        SV *    fh
        SV *    callback
        PROTOTYPE: $;$
        ALIAS:
           aio_close     = REQ_CLOSE
           aio_fsync     = REQ_FSYNC
           aio_fdatasync = REQ_FDATASYNC
        CODE:
{
        dREQ;

        req->type = ix;
        req->fh = newSVsv (fh);
        req->fd = PerlIO_fileno (IoIFP (sv_2io (fh)));

        send_req (req);
}

void
aio_read(fh,offset,length,data,dataoffset,callback=&PL_sv_undef)
        SV *    fh
        UV      offset
        UV      length
        SV *    data
        UV      dataoffset
        SV *    callback
        ALIAS:
           aio_read  = REQ_READ
           aio_write = REQ_WRITE
        PROTOTYPE: $$$$$;$
        CODE:
{
        aio_req req;
        STRLEN svlen;
        char *svptr = SvPVbyte (data, svlen);

        SvUPGRADE (data, SVt_PV);
        SvPOK_on (data);

        if (dataoffset < 0)
          dataoffset += svlen;

        if (dataoffset < 0 || dataoffset > svlen)
          croak ("data offset outside of string");

        if (ix == REQ_WRITE)
          {
            /* write: check length and adjust. */
            if (length < 0 || length + dataoffset > svlen)
              length = svlen - dataoffset;
          }
        else
          {
            /* read: grow scalar as necessary */
            svptr = SvGROW (data, length + dataoffset);
          }

        if (length < 0)
          croak ("length must not be negative");

        {
          dREQ;

          req->type = ix;
          req->fh = newSVsv (fh);
          req->fd = PerlIO_fileno (ix == REQ_READ ? IoIFP (sv_2io (fh))
                                                  : IoOFP (sv_2io (fh)));
          req->offset = offset;
          req->length = length;
          req->data = SvREFCNT_inc (data);
          req->dataptr = (char *)svptr + dataoffset;

          if (!SvREADONLY (data))
            {
              SvREADONLY_on (data);
              req->data2ptr = (void *)data;
            }

          send_req (req);
        }
}

void
aio_sendfile(out_fh,in_fh,in_offset,length,callback=&PL_sv_undef)
        SV *    out_fh
        SV *    in_fh
        UV      in_offset
        UV      length
        SV *    callback
        PROTOTYPE: $$$$;$
        CODE:
{
        dREQ;

        req->type = REQ_SENDFILE;
        req->fh = newSVsv (out_fh);
        req->fd = PerlIO_fileno (IoIFP (sv_2io (out_fh)));
        req->fh2 = newSVsv (in_fh);
        req->fd2 = PerlIO_fileno (IoIFP (sv_2io (in_fh)));
        req->offset = in_offset;
        req->length = length;

        send_req (req);
}

void
aio_readahead(fh,offset,length,callback=&PL_sv_undef)
        SV *    fh
        UV      offset
        IV      length
        SV *    callback
        PROTOTYPE: $$$;$
        CODE:
{
        dREQ;

        req->type = REQ_READAHEAD;
        req->fh = newSVsv (fh);
        req->fd = PerlIO_fileno (IoIFP (sv_2io (fh)));
        req->offset = offset;
        req->length = length;

        send_req (req);
}

void
aio_stat(fh_or_path,callback=&PL_sv_undef)
        SV *            fh_or_path
        SV *            callback
        ALIAS:
           aio_stat  = REQ_STAT
           aio_lstat = REQ_LSTAT
        CODE:
{
        dREQ;

        New (0, req->statdata, 1, Stat_t);
        if (!req->statdata)
          {
            free_req (req);
            croak ("out of memory during aio_req->statdata allocation");
          }

        if (SvPOK (fh_or_path))
          {
            req->type = ix;
            req->data = newSVsv (fh_or_path);
            req->dataptr = SvPVbyte_nolen (req->data);
          }
        else
          {
            req->type = REQ_FSTAT;
            req->fh = newSVsv (fh_or_path);
            req->fd = PerlIO_fileno (IoIFP (sv_2io (fh_or_path)));
          }

        send_req (req);
}

void
aio_unlink(pathname,callback=&PL_sv_undef)
        SV * pathname
        SV * callback
        ALIAS:
           aio_unlink = REQ_UNLINK
           aio_rmdir  = REQ_RMDIR
        CODE:
{
        dREQ;
        
        req->type = ix;
        req->data = newSVsv (pathname);
        req->dataptr = SvPVbyte_nolen (req->data);
        
        send_req (req);
}

void
aio_symlink(oldpath,newpath,callback=&PL_sv_undef)
        SV * oldpath
        SV * newpath
        SV * callback
        CODE:
{
        dREQ;
        
        req->type = REQ_SYMLINK;
        req->fh = newSVsv (oldpath);
        req->data2ptr = SvPVbyte_nolen (req->fh);
        req->data = newSVsv (newpath);
        req->dataptr = SvPVbyte_nolen (req->data);
        
        send_req (req);
}

void
flush()
        PROTOTYPE:
        CODE:
        while (nreqs)
          {
            poll_wait ();
            poll_cb ();
          }

void
poll()
        PROTOTYPE:
        CODE:
        if (nreqs)
          {
            poll_wait ();
            poll_cb ();
          }

int
poll_fileno()
        PROTOTYPE:
        CODE:
        RETVAL = respipe [0];
        OUTPUT:
        RETVAL

int
poll_cb(...)
        PROTOTYPE:
        CODE:
        RETVAL = poll_cb ();
        OUTPUT:
        RETVAL

void
poll_wait()
        PROTOTYPE:
        CODE:
        if (nreqs)
          poll_wait ();

int
nreqs()
        PROTOTYPE:
        CODE:
        RETVAL = nreqs;
        OUTPUT:
        RETVAL

Revision:	1.35
Committed:	Tue Aug 23 01:16:50 2005 UTC (18 years, 8 months ago) by root
Branch:	MAIN
Changes since 1.34:	+48 -10 lines
Log Message:	* empty log message *
#	User	Rev	Content
1	root	1.20	#define _REENTRANT 1
2	root	1.19	#include <errno.h>
3
4	root	1.15	#include "EXTERN.h"
5	root	1.1	#include "perl.h"
6			#include "XSUB.h"
7
8	root	1.16	#include "autoconf/config.h"
9
10	root	1.32	#include <pthread.h>
11
12	root	1.1	#include <sys/types.h>
13			#include <sys/stat.h>
14	root	1.10
15	root	1.1	#include <unistd.h>
16			#include <fcntl.h>
17			#include <signal.h>
18			#include <sched.h>
19
20	root	1.32	#if HAVE_SENDFILE
21			# if __linux
22			# include <sys/sendfile.h>
23			# elif __freebsd
24			# include <sys/socket.h>
25			# include <sys/uio.h>
26			# elif __hpux
27			# include <sys/socket.h>
28	root	1.35	# elif __solaris /* not yet */
29			# include <sys/sendfile.h>
30	root	1.34	# else
31			# error sendfile support requested but not available
32	root	1.32	# endif
33			#endif
34	root	1.1
35	root	1.4	#if __ia64
36			# define STACKSIZE 65536
37	root	1.1	#else
38	root	1.4	# define STACKSIZE 4096
39	root	1.1	#endif
40
41			enum {
42			REQ_QUIT,
43			REQ_OPEN, REQ_CLOSE,
44			REQ_READ, REQ_WRITE, REQ_READAHEAD,
45	root	1.32	REQ_SENDFILE,
46	root	1.22	REQ_STAT, REQ_LSTAT, REQ_FSTAT,
47	root	1.1	REQ_FSYNC, REQ_FDATASYNC,
48	root	1.23	REQ_UNLINK, REQ_RMDIR,
49	root	1.22	REQ_SYMLINK,
50	root	1.1	};
51
52			typedef struct aio_cb {
53	root	1.3	struct aio_cb *volatile next;
54	root	1.1
55			int type;
56
57	root	1.32	int fd, fd2;
58	root	1.1	off_t offset;
59			size_t length;
60			ssize_t result;
61			mode_t mode; /* open */
62			int errorno;
63	root	1.32	SV data, callback;
64			SV fh, fh2;
65	root	1.22	void dataptr, data2ptr;
66	root	1.1	STRLEN dataoffset;
67
68			Stat_t *statdata;
69			} aio_cb;
70
71			typedef aio_cb *aio_req;
72
73	root	1.30	static int started, wanted;
74	root	1.11	static volatile int nreqs;
75	root	1.4	static int max_outstanding = 1<<30;
76	root	1.3	static int respipe [2];
77	root	1.1
78	root	1.3	static pthread_mutex_t reslock = PTHREAD_MUTEX_INITIALIZER;
79			static pthread_mutex_t reqlock = PTHREAD_MUTEX_INITIALIZER;
80			static pthread_cond_t reqwait = PTHREAD_COND_INITIALIZER;
81
82			static volatile aio_req reqs, reqe; /* queue start, queue end */
83			static volatile aio_req ress, rese; /* queue start, queue end */
84	root	1.1
85	root	1.26	static void free_req (aio_req req)
86			{
87			if (req->data)
88			SvREFCNT_dec (req->data);
89
90			if (req->fh)
91			SvREFCNT_dec (req->fh);
92
93	root	1.32	if (req->fh2)
94			SvREFCNT_dec (req->fh2);
95
96	root	1.27	if (req->statdata)
97	root	1.26	Safefree (req->statdata);
98
99			if (req->callback)
100			SvREFCNT_dec (req->callback);
101
102			Safefree (req);
103			}
104
105	root	1.1	static void
106			poll_wait ()
107			{
108	root	1.11	if (nreqs && !ress)
109			{
110			fd_set rfd;
111			FD_ZERO(&rfd);
112			FD_SET(respipe [0], &rfd);
113	root	1.3
114	root	1.11	select (respipe [0] + 1, &rfd, 0, 0, 0);
115			}
116	root	1.1	}
117
118			static int
119	root	1.5	poll_cb ()
120	root	1.1	{
121			dSP;
122			int count = 0;
123	root	1.24	int do_croak = 0;
124	root	1.27	aio_req req;
125
126			for (;;)
127			{
128			pthread_mutex_lock (&reslock);
129			req = ress;
130
131			if (req)
132			{
133			ress = req->next;
134	root	1.11
135	root	1.27	if (!ress)
136			{
137			/* read any signals sent by the worker threads */
138			char buf [32];
139			while (read (respipe [0], buf, 32) == 32)
140			;
141	root	1.30
142			rese = 0;
143	root	1.27	}
144			}
145	root	1.1
146	root	1.27	pthread_mutex_unlock (&reslock);
147	root	1.3
148	root	1.27	if (!req)
149			break;
150	root	1.3
151	root	1.1	nreqs--;
152
153			if (req->type == REQ_QUIT)
154			started--;
155			else
156			{
157			int errorno = errno;
158			errno = req->errorno;
159
160			if (req->type == REQ_READ)
161	root	1.27	SvCUR_set (req->data, req->dataoffset + (req->result > 0 ? req->result : 0));
162	root	1.1
163	root	1.28	if (req->data2ptr && (req->type == REQ_READ \|\| req->type == REQ_WRITE))
164			SvREADONLY_off (req->data);
165
166	root	1.27	if (req->statdata)
167	root	1.1	{
168			PL_laststype = req->type == REQ_LSTAT ? OP_LSTAT : OP_STAT;
169			PL_laststatval = req->result;
170			PL_statcache = *(req->statdata);
171			}
172
173	root	1.13	ENTER;
174	root	1.1	PUSHMARK (SP);
175			XPUSHs (sv_2mortal (newSViv (req->result)));
176	root	1.2
177			if (req->type == REQ_OPEN)
178			{
179			/* convert fd to fh */
180			SV *fh;
181
182			PUTBACK;
183			call_pv ("IO::AIO::_fd2fh", G_SCALAR \| G_EVAL);
184			SPAGAIN;
185
186	root	1.13	fh = SvREFCNT_inc (POPs);
187	root	1.2
188			PUSHMARK (SP);
189	root	1.13	XPUSHs (sv_2mortal (fh));
190	root	1.2	}
191
192	root	1.8	if (SvOK (req->callback))
193			{
194			PUTBACK;
195			call_sv (req->callback, G_VOID \| G_EVAL);
196			SPAGAIN;
197	root	1.27
198			if (SvTRUE (ERRSV))
199			{
200			free_req (req);
201			croak (0);
202			}
203	root	1.8	}
204	root	1.13
205	root	1.26	LEAVE;
206
207	root	1.1	errno = errorno;
208			count++;
209			}
210
211	root	1.27	free_req (req);
212	root	1.1	}
213
214			return count;
215			}
216
217	root	1.4	static void aio_proc(void arg);
218
219			static void
220			start_thread (void)
221			{
222			sigset_t fullsigset, oldsigset;
223			pthread_t tid;
224			pthread_attr_t attr;
225
226			pthread_attr_init (&attr);
227			pthread_attr_setstacksize (&attr, STACKSIZE);
228	root	1.31	pthread_attr_setdetachstate (&attr, PTHREAD_CREATE_DETACHED);
229	root	1.4
230			sigfillset (&fullsigset);
231			sigprocmask (SIG_SETMASK, &fullsigset, &oldsigset);
232
233			if (pthread_create (&tid, &attr, aio_proc, 0) == 0)
234			started++;
235
236			sigprocmask (SIG_SETMASK, &oldsigset, 0);
237			}
238
239			static void
240			send_req (aio_req req)
241			{
242	root	1.30	while (started < wanted && nreqs >= started)
243			start_thread ();
244
245	root	1.4	nreqs++;
246
247			pthread_mutex_lock (&reqlock);
248
249			req->next = 0;
250
251			if (reqe)
252			{
253			reqe->next = req;
254			reqe = req;
255			}
256			else
257			reqe = reqs = req;
258
259			pthread_cond_signal (&reqwait);
260			pthread_mutex_unlock (&reqlock);
261
262	root	1.27	if (nreqs > max_outstanding)
263			for (;;)
264			{
265			poll_cb ();
266
267			if (nreqs <= max_outstanding)
268			break;
269
270			poll_wait ();
271			}
272	root	1.4	}
273
274			static void
275			end_thread (void)
276			{
277			aio_req req;
278	root	1.26	Newz (0, req, 1, aio_cb);
279	root	1.4	req->type = REQ_QUIT;
280
281			send_req (req);
282			}
283
284	root	1.22	static void min_parallel (int nthreads)
285			{
286	root	1.30	if (wanted < nthreads)
287			wanted = nthreads;
288	root	1.22	}
289
290			static void max_parallel (int nthreads)
291			{
292			int cur = started;
293	root	1.27
294	root	1.30	if (wanted > nthreads)
295			wanted = nthreads;
296
297			while (cur > wanted)
298			{
299	root	1.22	end_thread ();
300			cur--;
301			}
302
303	root	1.30	while (started > wanted)
304	root	1.22	{
305			poll_wait ();
306			poll_cb ();
307			}
308			}
309
310	root	1.26	static void create_pipe ()
311	root	1.22	{
312	root	1.26	if (pipe (respipe))
313			croak ("unable to initialize result pipe");
314	root	1.22
315	root	1.26	if (fcntl (respipe [0], F_SETFL, O_NONBLOCK))
316			croak ("cannot set result pipe to nonblocking mode");
317	root	1.22
318	root	1.26	if (fcntl (respipe [1], F_SETFL, O_NONBLOCK))
319			croak ("cannot set result pipe to nonblocking mode");
320			}
321	root	1.22
322	root	1.26	static void atfork_prepare (void)
323			{
324	root	1.28	pthread_mutex_lock (&reqlock);
325	root	1.22	pthread_mutex_lock (&reslock);
326			}
327
328			static void atfork_parent (void)
329			{
330			pthread_mutex_unlock (&reslock);
331	root	1.26	pthread_mutex_unlock (&reqlock);
332			}
333
334			static void atfork_child (void)
335			{
336	root	1.28	aio_req prv;
337
338	root	1.26	started = 0;
339
340	root	1.28	while (reqs)
341			{
342			prv = reqs;
343			reqs = prv->next;
344			free_req (prv);
345			}
346
347			reqs = reqe = 0;
348
349			while (ress)
350			{
351			prv = ress;
352			ress = prv->next;
353			free_req (prv);
354			}
355
356			ress = rese = 0;
357
358	root	1.29	close (respipe [0]);
359			close (respipe [1]);
360			create_pipe ();
361
362	root	1.26	atfork_parent ();
363	root	1.22	}
364
365	root	1.35	/* currently noops */
366			#define LOCK_FD(fd) do { } while (0)
367			#define UNLOCK_FD(fd) do { } while (0)
368
369	root	1.22	/*****************************************************************************/
370	root	1.17	/* work around various missing functions */
371
372			#if !HAVE_PREADWRITE
373			# define pread aio_pread
374			# define pwrite aio_pwrite
375
376			/*
377			* make our pread/pwrite safe against themselves, but not against
378			* normal read/write by using a mutex. slows down execution a lot,
379			* but that's your problem, not mine.
380			*/
381			static pthread_mutex_t iolock = PTHREAD_MUTEX_INITIALIZER;
382
383			static ssize_t
384			pread (int fd, void *buf, size_t count, off_t offset)
385			{
386			ssize_t res;
387			off_t ooffset;
388
389	root	1.35	LOCK_FD (fd);
390			pthread_mutex_lock (&iolock); /* replace by LOCK_FD and private buffer */
391	root	1.17	ooffset = lseek (fd, 0, SEEK_CUR);
392			lseek (fd, offset, SEEK_SET);
393			res = read (fd, buf, count);
394			lseek (fd, ooffset, SEEK_SET);
395			pthread_mutex_unlock (&iolock);
396	root	1.35	UNLOCK_FD (d);
397	root	1.17
398			return res;
399			}
400
401			static ssize_t
402			pwrite (int fd, void *buf, size_t count, off_t offset)
403			{
404			ssize_t res;
405			off_t ooffset;
406
407	root	1.35	LOCK_FD (fd);
408			pthread_mutex_lock (&iolock); /* replace by LOCK_FD and private buffer */
409	root	1.17	ooffset = lseek (fd, 0, SEEK_CUR);
410			lseek (fd, offset, SEEK_SET);
411			res = write (fd, buf, count);
412			lseek (fd, offset, SEEK_SET);
413			pthread_mutex_unlock (&iolock);
414	root	1.35	UNLOCK_FD (d);
415	root	1.17
416			return res;
417			}
418			#endif
419
420			#if !HAVE_FDATASYNC
421			# define fdatasync fsync
422			#endif
423
424			#if !HAVE_READAHEAD
425			# define readahead aio_readahead
426
427			static char readahead_buf[4096];
428
429			static ssize_t
430			readahead (int fd, off_t offset, size_t count)
431			{
432			while (count > 0)
433			{
434			size_t len = count < sizeof (readahead_buf) ? count : sizeof (readahead_buf);
435
436			pread (fd, readahead_buf, len, offset);
437			offset += len;
438			count -= len;
439			}
440
441			errno = 0;
442			}
443			#endif
444
445	root	1.32	/* sendfile always needs emulation */
446			static ssize_t
447			sendfile_ (int ofd, int ifd, off_t offset, size_t count)
448			{
449			ssize_t res;
450
451			if (!count)
452			return 0;
453
454	root	1.35	LOCK_FD (ofd);
455
456			#if HAVE_SENDFILE
457			# if __linux
458	root	1.32	res = sendfile (ofd, ifd, &offset, count);
459
460	root	1.35	# elif __freebsd
461	root	1.32	/*
462			* Of course, the freebsd sendfile is a dire hack with no thoughts
463	root	1.35	* wasted on making it similar to other I/O functions.
464	root	1.32	*/
465			{
466			off_t sbytes;
467			res = sendfile (ifd, ofd, offset, count, 0, &sbytes, 0);
468
469	root	1.35	if (res < 0 && sbytes)
470			/* maybe only on EAGAIN only: as usual, the manpage leaves you guessing */
471	root	1.32	res = sbytes;
472			}
473
474	root	1.35	# elif __hpux
475	root	1.32	res = sendfile (ofd, ifd, offset, count, 0, 0);
476
477	root	1.35	# elif __solaris
478			{
479			struct sendfilevec vec;
480			size_t sbytes;
481
482			vec.sfv_fd = ifd;
483			vec.sfv_flag = 0;
484			vec.sfv_off = offset;
485			vec.sfv_len = count;
486
487			res = sendfilev (ofd, &vec, 1, &sbytes);
488
489			if (res < 0 && sbytes)
490			res = sbytes;
491			}
492
493			# else
494	root	1.32	res = -1;
495			errno = ENOSYS;
496	root	1.35	# endif
497	root	1.32	#endif
498
499	root	1.35	if (res < 0
500			&& (errno == ENOSYS \|\| errno == EINVAL \|\| errno == ENOTSOCK
501			#if __solaris
502			\|\| errno == EAFNOSUPPORT \|\| errno == EPROTOTYPE
503			#endif
504			)
505			)
506	root	1.32	{
507			/* emulate sendfile. this is a major pain in the ass */
508			char *buf = malloc (4096);
509			res = 0;
510
511			for (;;)
512			{
513			ssize_t cnt;
514
515			cnt = pread (ifd, buf, 4096, offset);
516
517			if (cnt <= 0)
518			{
519			if (cnt && !res) res = -1;
520			break;
521			}
522
523			cnt = write (ofd, buf, cnt);
524
525			if (cnt <= 0)
526			{
527			if (cnt && !res) res = -1;
528			break;
529			}
530
531			offset += cnt;
532			res += cnt;
533			}
534
535			{
536			int errorno = errno;
537			free (buf);
538			errno = errorno;
539			}
540			}
541
542	root	1.35	UNLOCK_FD (ofd);
543
544	root	1.32	return res;
545			}
546
547	root	1.22	/*****************************************************************************/
548
549	root	1.1	static void *
550			aio_proc (void *thr_arg)
551			{
552			aio_req req;
553	root	1.3	int type;
554	root	1.1
555	root	1.3	do
556	root	1.1	{
557	root	1.3	pthread_mutex_lock (&reqlock);
558
559			for (;;)
560			{
561			req = reqs;
562
563			if (reqs)
564			{
565			reqs = reqs->next;
566			if (!reqs) reqe = 0;
567			}
568
569			if (req)
570			break;
571
572			pthread_cond_wait (&reqwait, &reqlock);
573			}
574
575			pthread_mutex_unlock (&reqlock);
576
577	root	1.1	errno = 0; /* strictly unnecessary */
578
579	root	1.3	type = req->type;
580
581			switch (type)
582	root	1.1	{
583	root	1.10	case REQ_READ: req->result = pread (req->fd, req->dataptr, req->length, req->offset); break;
584			case REQ_WRITE: req->result = pwrite (req->fd, req->dataptr, req->length, req->offset); break;
585	root	1.16
586	root	1.1	case REQ_READAHEAD: req->result = readahead (req->fd, req->offset, req->length); break;
587	root	1.32	case REQ_SENDFILE: req->result = sendfile_ (req->fd, req->fd2, req->offset, req->length); break;
588	root	1.1
589			case REQ_STAT: req->result = stat (req->dataptr, req->statdata); break;
590			case REQ_LSTAT: req->result = lstat (req->dataptr, req->statdata); break;
591			case REQ_FSTAT: req->result = fstat (req->fd , req->statdata); break;
592
593			case REQ_OPEN: req->result = open (req->dataptr, req->fd, req->mode); break;
594			case REQ_CLOSE: req->result = close (req->fd); break;
595			case REQ_UNLINK: req->result = unlink (req->dataptr); break;
596	root	1.22	case REQ_RMDIR: req->result = rmdir (req->dataptr); break;
597			case REQ_SYMLINK: req->result = symlink (req->data2ptr, req->dataptr); break;
598	root	1.1
599	root	1.17	case REQ_FDATASYNC: req->result = fdatasync (req->fd); break;
600	root	1.1	case REQ_FSYNC: req->result = fsync (req->fd); break;
601
602			case REQ_QUIT:
603	root	1.3	break;
604	root	1.1
605			default:
606			req->result = ENOSYS;
607			break;
608			}
609
610			req->errorno = errno;
611	root	1.3
612			pthread_mutex_lock (&reslock);
613
614			req->next = 0;
615
616			if (rese)
617			{
618			rese->next = req;
619			rese = req;
620			}
621			else
622			{
623			rese = ress = req;
624
625			/* write a dummy byte to the pipe so fh becomes ready */
626			write (respipe [1], &respipe, 1);
627			}
628
629			pthread_mutex_unlock (&reslock);
630	root	1.1	}
631	root	1.3	while (type != REQ_QUIT);
632	root	1.1
633			return 0;
634			}
635
636	root	1.22	#define dREQ \
637			aio_req req; \
638			\
639			if (SvOK (callback) && !SvROK (callback)) \
640			croak ("clalback must be undef or of reference type"); \
641			\
642			Newz (0, req, 1, aio_cb); \
643			if (!req) \
644			croak ("out of memory during aio_req allocation"); \
645			\
646	root	1.27	req->callback = newSVsv (callback);
647	root	1.22
648	root	1.1	MODULE = IO::AIO PACKAGE = IO::AIO
649
650	root	1.8	PROTOTYPES: ENABLE
651
652	root	1.1	BOOT:
653			{
654	root	1.26	create_pipe ();
655	root	1.22	pthread_atfork (atfork_prepare, atfork_parent, atfork_child);
656	root	1.1	}
657
658			void
659			min_parallel(nthreads)
660			int nthreads
661			PROTOTYPE: $
662
663			void
664			max_parallel(nthreads)
665			int nthreads
666			PROTOTYPE: $
667
668	root	1.4	int
669			max_outstanding(nreqs)
670			int nreqs
671			PROTOTYPE: $
672			CODE:
673			RETVAL = max_outstanding;
674			max_outstanding = nreqs;
675
676	root	1.1	void
677	root	1.8	aio_open(pathname,flags,mode,callback=&PL_sv_undef)
678	root	1.1	SV * pathname
679			int flags
680			int mode
681			SV * callback
682	root	1.8	PROTOTYPE: $$$;$
683	root	1.1	CODE:
684			{
685	root	1.22	dREQ;
686	root	1.1
687			req->type = REQ_OPEN;
688			req->data = newSVsv (pathname);
689	root	1.22	req->dataptr = SvPVbyte_nolen (req->data);
690	root	1.1	req->fd = flags;
691			req->mode = mode;
692
693			send_req (req);
694			}
695
696			void
697	root	1.8	aio_close(fh,callback=&PL_sv_undef)
698	root	1.13	SV * fh
699			SV * callback
700	root	1.8	PROTOTYPE: $;$
701	root	1.1	ALIAS:
702			aio_close = REQ_CLOSE
703			aio_fsync = REQ_FSYNC
704			aio_fdatasync = REQ_FDATASYNC
705			CODE:
706			{
707	root	1.22	dREQ;
708	root	1.1
709			req->type = ix;
710	root	1.13	req->fh = newSVsv (fh);
711			req->fd = PerlIO_fileno (IoIFP (sv_2io (fh)));
712	root	1.1
713			send_req (req);
714			}
715
716			void
717	root	1.8	aio_read(fh,offset,length,data,dataoffset,callback=&PL_sv_undef)
718	root	1.13	SV * fh
719			UV offset
720	root	1.32	UV length
721	root	1.13	SV * data
722	root	1.32	UV dataoffset
723	root	1.13	SV * callback
724			ALIAS:
725			aio_read = REQ_READ
726			aio_write = REQ_WRITE
727	root	1.8	PROTOTYPE: $$$$$;$
728	root	1.1	CODE:
729	root	1.13	{
730			aio_req req;
731			STRLEN svlen;
732	root	1.21	char *svptr = SvPVbyte (data, svlen);
733	root	1.13
734			SvUPGRADE (data, SVt_PV);
735			SvPOK_on (data);
736	root	1.1
737	root	1.13	if (dataoffset < 0)
738			dataoffset += svlen;
739
740			if (dataoffset < 0 \|\| dataoffset > svlen)
741			croak ("data offset outside of string");
742
743			if (ix == REQ_WRITE)
744			{
745			/* write: check length and adjust. */
746			if (length < 0 \|\| length + dataoffset > svlen)
747			length = svlen - dataoffset;
748			}
749			else
750			{
751			/* read: grow scalar as necessary */
752			svptr = SvGROW (data, length + dataoffset);
753			}
754
755			if (length < 0)
756			croak ("length must not be negative");
757
758	root	1.22	{
759			dREQ;
760	root	1.13
761	root	1.22	req->type = ix;
762			req->fh = newSVsv (fh);
763			req->fd = PerlIO_fileno (ix == REQ_READ ? IoIFP (sv_2io (fh))
764			: IoOFP (sv_2io (fh)));
765			req->offset = offset;
766			req->length = length;
767			req->data = SvREFCNT_inc (data);
768			req->dataptr = (char *)svptr + dataoffset;
769	root	1.13
770	root	1.28	if (!SvREADONLY (data))
771			{
772			SvREADONLY_on (data);
773			req->data2ptr = (void *)data;
774			}
775
776	root	1.22	send_req (req);
777			}
778	root	1.13	}
779	root	1.1
780			void
781	root	1.32	aio_sendfile(out_fh,in_fh,in_offset,length,callback=&PL_sv_undef)
782			SV * out_fh
783			SV * in_fh
784			UV in_offset
785			UV length
786			SV * callback
787			PROTOTYPE: $$$$;$
788			CODE:
789			{
790			dREQ;
791
792			req->type = REQ_SENDFILE;
793			req->fh = newSVsv (out_fh);
794			req->fd = PerlIO_fileno (IoIFP (sv_2io (out_fh)));
795			req->fh2 = newSVsv (in_fh);
796			req->fd2 = PerlIO_fileno (IoIFP (sv_2io (in_fh)));
797			req->offset = in_offset;
798			req->length = length;
799
800			send_req (req);
801			}
802
803			void
804	root	1.8	aio_readahead(fh,offset,length,callback=&PL_sv_undef)
805	root	1.13	SV * fh
806			UV offset
807			IV length
808			SV * callback
809	root	1.8	PROTOTYPE: $$$;$
810	root	1.1	CODE:
811			{
812	root	1.22	dREQ;
813	root	1.1
814			req->type = REQ_READAHEAD;
815	root	1.13	req->fh = newSVsv (fh);
816			req->fd = PerlIO_fileno (IoIFP (sv_2io (fh)));
817	root	1.1	req->offset = offset;
818			req->length = length;
819
820			send_req (req);
821			}
822
823			void
824	root	1.8	aio_stat(fh_or_path,callback=&PL_sv_undef)
825	root	1.1	SV * fh_or_path
826			SV * callback
827			ALIAS:
828	root	1.8	aio_stat = REQ_STAT
829			aio_lstat = REQ_LSTAT
830	root	1.1	CODE:
831			{
832	root	1.22	dREQ;
833	root	1.1
834			New (0, req->statdata, 1, Stat_t);
835			if (!req->statdata)
836	root	1.27	{
837			free_req (req);
838			croak ("out of memory during aio_req->statdata allocation");
839			}
840	root	1.1
841			if (SvPOK (fh_or_path))
842			{
843	root	1.8	req->type = ix;
844	root	1.1	req->data = newSVsv (fh_or_path);
845	root	1.22	req->dataptr = SvPVbyte_nolen (req->data);
846	root	1.1	}
847			else
848			{
849			req->type = REQ_FSTAT;
850	root	1.13	req->fh = newSVsv (fh_or_path);
851	root	1.1	req->fd = PerlIO_fileno (IoIFP (sv_2io (fh_or_path)));
852			}
853
854			send_req (req);
855			}
856
857			void
858	root	1.8	aio_unlink(pathname,callback=&PL_sv_undef)
859	root	1.1	SV * pathname
860			SV * callback
861	root	1.22	ALIAS:
862			aio_unlink = REQ_UNLINK
863			aio_rmdir = REQ_RMDIR
864	root	1.1	CODE:
865			{
866	root	1.22	dREQ;
867	root	1.1
868	root	1.22	req->type = ix;
869			req->data = newSVsv (pathname);
870			req->dataptr = SvPVbyte_nolen (req->data);
871	root	1.1
872	root	1.22	send_req (req);
873			}
874
875			void
876			aio_symlink(oldpath,newpath,callback=&PL_sv_undef)
877			SV * oldpath
878			SV * newpath
879			SV * callback
880			CODE:
881			{
882			dREQ;
883	root	1.1
884	root	1.22	req->type = REQ_SYMLINK;
885			req->fh = newSVsv (oldpath);
886			req->data2ptr = SvPVbyte_nolen (req->fh);
887			req->data = newSVsv (newpath);
888			req->dataptr = SvPVbyte_nolen (req->data);
889	root	1.1
890			send_req (req);
891			}
892
893	root	1.6	void
894			flush()
895			PROTOTYPE:
896			CODE:
897			while (nreqs)
898			{
899			poll_wait ();
900			poll_cb ();
901			}
902
903	root	1.7	void
904			poll()
905			PROTOTYPE:
906			CODE:
907			if (nreqs)
908			{
909			poll_wait ();
910			poll_cb ();
911			}
912
913	root	1.1	int
914			poll_fileno()
915			PROTOTYPE:
916			CODE:
917	root	1.3	RETVAL = respipe [0];
918	root	1.1	OUTPUT:
919			RETVAL
920
921			int
922			poll_cb(...)
923			PROTOTYPE:
924			CODE:
925	root	1.5	RETVAL = poll_cb ();
926	root	1.1	OUTPUT:
927			RETVAL
928
929			void
930			poll_wait()
931			PROTOTYPE:
932			CODE:
933	root	1.3	if (nreqs)
934			poll_wait ();
935	root	1.1
936			int
937			nreqs()
938			PROTOTYPE:
939			CODE:
940			RETVAL = nreqs;
941			OUTPUT:
942			RETVAL
943