libc/winsup/cygwin/mmap.cc

/* mmap.cc

   Copyright 1996, 1997, 1998, 2000, 2001, 2002, 2003, 2004, 2005
   Red Hat, Inc.

This file is part of Cygwin.

This software is a copyrighted work licensed under the terms of the
Cygwin license.  Please consult the file "CYGWIN_LICENSE" for
details. */

#include "winsup.h"
#include <unistd.h>
#include <stdlib.h>
#include <stddef.h>
#include <sys/mman.h>
#include "cygerrno.h"
#include "security.h"
#include "path.h"
#include "fhandler.h"
#include "dtable.h"
#include "cygheap.h"
#include "pinfo.h"
#include "sys/cygwin.h"
#include "ntdll.h"

#define PAGE_CNT(bytes) howmany((bytes),getpagesize())

#define PGBITS		(sizeof (DWORD)*8)
#define MAPSIZE(pages)	howmany ((pages), PGBITS)

#define MAP_SET(n)	(page_map_[(n)/PGBITS] |= (1L << ((n) % PGBITS)))
#define MAP_CLR(n)	(page_map_[(n)/PGBITS] &= ~(1L << ((n) % PGBITS)))
#define MAP_ISSET(n)	(page_map_[(n)/PGBITS] & (1L << ((n) % PGBITS)))

/* Used for accessing the page file (anonymous mmaps). */
static fhandler_disk_file fh_paging_file;

/* Class structure used to keep a record of all current mmap areas
   in a process.  Needed for bookkeeping all mmaps in a process and
   for duplicating all mmaps after fork() since mmaps are not propagated
   to child processes by Windows.  All information must be duplicated
   by hand, see fixup_mmaps_after_fork().

   The class structure:

   One member of class map per process, global variable mmapped_areas.
   Contains a dynamic class list array.  Each list entry represents all
   mapping to a file, keyed by file descriptor and file name hash.
   Each list entry contains a dynamic class mmap_record array.  Each
   mmap_record represents exactly one mapping.  For each mapping, there's
   an additional so called `page_map'.  It's an array of bits, one bit
   per mapped memory page.  The bit is set if the page is accessible,
   unset otherwise. */

class mmap_record
{
  private:
    int fdesc_;
    HANDLE mapping_handle_;
    DWORD access_mode_;
    int flags_;
    _off64_t offset_;
    DWORD size_to_map_;
    caddr_t base_address_;
    DWORD *page_map_;
    device dev;

  public:
    mmap_record (int fd, HANDLE h, DWORD ac, int f, _off64_t o, DWORD s,
    		 caddr_t b) :
       fdesc_ (fd),
       mapping_handle_ (h),
       access_mode_ (ac),
       flags_ (f),
       offset_ (o),
       size_to_map_ (s),
       base_address_ (b),
       page_map_ (NULL)
      {
	dev.devn = 0;
	if (fd >= 0 && !cygheap->fdtab.not_open (fd))
	  dev = cygheap->fdtab[fd]->dev ();
      }

    int get_fd () const { return fdesc_; }
    HANDLE get_handle () const { return mapping_handle_; }
    device& get_device () { return dev; }
    DWORD get_access () const { return access_mode_; }
    DWORD get_flags () const { return flags_; }
    _off64_t get_offset () const { return offset_; }
    DWORD get_size () const { return size_to_map_; }
    caddr_t get_address () const { return base_address_; }

    bool alloc_page_map (_off64_t off, DWORD len);
    void free_page_map () { if (page_map_) cfree (page_map_); }
    void fixup_page_map ();

    DWORD find_unused_pages (DWORD pages);
    _off64_t map_pages (_off64_t off, DWORD len);
    bool map_pages (caddr_t addr, DWORD len);
    bool unmap_pages (caddr_t addr, DWORD len);
    int access (caddr_t address);

    fhandler_base *alloc_fh ();
    void free_fh (fhandler_base *fh);
};

class list
{
  private:
    mmap_record *recs;
    int nrecs, maxrecs;
    int fd;
    DWORD hash;

  public:
    int get_fd () const { return fd; }
    DWORD get_hash () const { return hash; }
    mmap_record *get_record (int i) { return i >= nrecs ? NULL : recs + i; }

    void set (int nfd);
    mmap_record *add_record (mmap_record r, _off64_t off, DWORD len);
    bool del_record (int i);
    void free_recs () { if (recs) cfree (recs); }
    mmap_record *search_record (_off64_t off, DWORD len);
    long search_record (caddr_t addr, DWORD len, caddr_t &m_addr, DWORD &m_len,
		long start);
};

class map
{
  private:
    list *lists;
    int nlists, maxlists;
    caddr_t next_anon_addr;

  public:
    list *get_list (int i) { return i >= nlists ? NULL : lists + i; }
    list *get_list_by_fd (int fd);
    list *add_list (int fd);
    void del_list (int i);
    caddr_t get_next_anon_addr () { return next_anon_addr; }
    void set_next_anon_addr (caddr_t addr) { next_anon_addr = addr; }
};

/* This is the global map structure pointer.  It's allocated once on the
   first call to mmap64(). */
static map *mmapped_areas;

DWORD
mmap_record::find_unused_pages (DWORD pages)
{
  DWORD mapped_pages = PAGE_CNT (size_to_map_);
  DWORD start;

  if (pages > mapped_pages)
    return (DWORD)-1;
  for (start = 0; start <= mapped_pages - pages; ++start)
    if (!MAP_ISSET (start))
      {
	DWORD cnt;
	for (cnt = 0; cnt < pages; ++cnt)
	  if (MAP_ISSET (start + cnt))
	    break;
	if (cnt >= pages)
	  return start;
      }
  return (DWORD)-1;
}

bool
mmap_record::alloc_page_map (_off64_t off, DWORD len)
{
  /* Allocate one bit per page */
  if (!(page_map_ = (DWORD *) ccalloc (HEAP_MMAP,
				       MAPSIZE (PAGE_CNT (size_to_map_)),
				       sizeof (DWORD))))
    return false;

  off -= offset_;
  len = PAGE_CNT (len);

  if (wincap.virtual_protect_works_on_shared_pages ())
    {
      DWORD old_prot;
      DWORD vlen = len * getpagesize ();

      if (off > 0 &&
	  !VirtualProtect (base_address_, off, PAGE_NOACCESS, &old_prot))
	syscall_printf ("VirtualProtect(%x,%D) failed, %E", base_address_, off);
      if (off + vlen < size_to_map_
	  && !VirtualProtect (base_address_ + off + vlen,
			      size_to_map_ - vlen - off,
			      PAGE_NOACCESS, &old_prot))
	syscall_printf ("VirtualProtect(%x,%D) failed, %E",
			base_address_ + off + vlen, size_to_map_ - vlen - off);
    }

  off /= getpagesize ();

  while (len-- > 0)
    MAP_SET (off + len);
  return true;
}

_off64_t
mmap_record::map_pages (_off64_t off, DWORD len)
{
  /* Used ONLY if this mapping matches into the chunk of another already
     performed mapping in a special case of MAP_ANON|MAP_PRIVATE.

     Otherwise it's job is now done by alloc_page_map(). */
  DWORD prot, old_prot;
  switch (access_mode_)
    {
    case FILE_MAP_WRITE:
      prot = PAGE_READWRITE;
      break;
    case FILE_MAP_READ:
      prot = PAGE_READONLY;
      break;
    default:
      prot = PAGE_WRITECOPY;
      break;
    }

  debug_printf ("map_pages (fd=%d, off=%D, len=%u)", fdesc_, off, len);
  len = PAGE_CNT (len);

  if ((off = find_unused_pages (len)) == (DWORD)-1)
    return 0L;
  if (wincap.virtual_protect_works_on_shared_pages ()
      && !VirtualProtect (base_address_ + off * getpagesize (),
			  len * getpagesize (), prot, &old_prot))
    {
      __seterrno ();
      return (_off64_t)-1;
    }

  while (len-- > 0)
    MAP_SET (off + len);
  return off * getpagesize ();
}

bool
mmap_record::map_pages (caddr_t addr, DWORD len)
{
  debug_printf ("map_pages (addr=%x, len=%u)", addr, len);
  DWORD prot, old_prot;
  DWORD off = addr - base_address_;
  off /= getpagesize ();
  len = PAGE_CNT (len);
  /* First check if the area is unused right now. */
  for (DWORD l = 0; l < len; ++l)
    if (MAP_ISSET (off + l))
      {
	set_errno (EINVAL);
	return false;
      }
  switch (access_mode_)
    {
    case FILE_MAP_WRITE:
      prot = PAGE_READWRITE;
      break;
    case FILE_MAP_READ:
      prot = PAGE_READONLY;
      break;
    default:
      prot = PAGE_WRITECOPY;
      break;
    }
  if (wincap.virtual_protect_works_on_shared_pages ()
      && !VirtualProtect (base_address_ + off * getpagesize (),
			  len * getpagesize (), prot, &old_prot))
    {
      __seterrno ();
      return false;
    }
  for (; len-- > 0; ++off)
    MAP_SET (off);
  return true;
}

bool
mmap_record::unmap_pages (caddr_t addr, DWORD len)
{
  DWORD old_prot;
  DWORD off = addr - base_address_;
  off /= getpagesize ();
  len = PAGE_CNT (len);
  if (wincap.virtual_protect_works_on_shared_pages ()
      && !VirtualProtect (base_address_ + off * getpagesize (),
			  len * getpagesize (), PAGE_NOACCESS, &old_prot))
    syscall_printf ("-1 = unmap_pages (), %E");

  for (; len-- > 0; ++off)
    MAP_CLR (off);
  /* Return TRUE if all pages are free'd which may result in unmapping
     the whole chunk. */
  for (len = MAPSIZE (PAGE_CNT (size_to_map_)); len > 0; )
    if (page_map_[--len])
      return false;
  return true;
}

void
mmap_record::fixup_page_map ()
{
  if (!wincap.virtual_protect_works_on_shared_pages ())
    return;

  DWORD prot, old_prot;
  switch (access_mode_)
    {
    case FILE_MAP_WRITE:
      prot = PAGE_READWRITE;
      break;
    case FILE_MAP_READ:
      prot = PAGE_READONLY;
      break;
    default:
      prot = PAGE_WRITECOPY;
      break;
    }

  for (DWORD off = PAGE_CNT (size_to_map_); off > 0; --off)
    VirtualProtect (base_address_ + (off - 1) * getpagesize (), getpagesize (),
		    MAP_ISSET (off - 1) ? prot : PAGE_NOACCESS, &old_prot);
}

int
mmap_record::access (caddr_t address)
{
  if (address < base_address_ || address >= base_address_ + size_to_map_)
    return 0;
  DWORD off = (address - base_address_) / getpagesize ();
  return MAP_ISSET (off);
}

fhandler_base *
mmap_record::alloc_fh ()
{
  if (get_fd () == -1)
    {
      fh_paging_file.set_io_handle (INVALID_HANDLE_VALUE);
      return &fh_paging_file;
    }

  /* The file descriptor could have been closed or, even
     worse, could have been reused for another file before
     the call to fork(). This requires creating a fhandler
     of the correct type to be sure to call the method of the
     correct class. */
  return build_fh_dev (get_device ());
}

void
mmap_record::free_fh (fhandler_base *fh)
{
  if (get_fd () != -1)
    cfree (fh);
}

mmap_record *
list::add_record (mmap_record r, _off64_t off, DWORD len)
{
  if (nrecs == maxrecs)
    {
      mmap_record *new_recs;
      if (maxrecs == 0)
	new_recs = (mmap_record *)
			cmalloc (HEAP_MMAP, 5 * sizeof (mmap_record));
      else
	new_recs = (mmap_record *)
			crealloc (recs, (maxrecs + 5) * sizeof (mmap_record));
      if (!new_recs)
	return NULL;
      maxrecs += 5;
      recs = new_recs;
    }
  recs[nrecs] = r;
  if (!recs[nrecs].alloc_page_map (off, len))
    return NULL;
  return recs + nrecs++;
}

/* Used in mmap() */
mmap_record *
list::search_record (_off64_t off, DWORD len)
{
  if (fd == -1 && !off)
    {
      len = PAGE_CNT (len);
      for (int i = 0; i < nrecs; ++i)
	if (recs[i].find_unused_pages (len) != (DWORD)-1)
	  return recs + i;
    }
  else
    {
      for (int i = 0; i < nrecs; ++i)
	if (off >= recs[i].get_offset ()
	    && off + len <= recs[i].get_offset ()
			 + (PAGE_CNT (recs[i].get_size ()) * getpagesize ()))
	  return recs + i;
    }
  return NULL;
}

/* Used in munmap() */
long
list::search_record (caddr_t addr, DWORD len, caddr_t &m_addr, DWORD &m_len,
		     long start)
{
  caddr_t low, high;

  for (long i = start + 1; i < nrecs; ++i)
    {
      low = (addr >= recs[i].get_address ()) ? addr : recs[i].get_address ();
      high = recs[i].get_address ()
	     + (PAGE_CNT (recs[i].get_size ()) * getpagesize ());
      high = (addr + len < high) ? addr + len : high;
      if (low < high)
	{
	  m_addr = low;
	  m_len = high - low;
	  return i;
	}
    }
  return -1;
}

void
list::set (int nfd)
{
  if ((fd = nfd) != -1)
    hash = cygheap->fdtab[fd]->get_namehash ();
  nrecs = maxrecs = 0;
  recs = NULL;
}

bool
list::del_record (int i)
{
  if (i < nrecs)
    {
      recs[i].free_page_map ();
      for (; i < nrecs - 1; i++)
	recs[i] = recs[i + 1];
      nrecs--;
    }
  /* Return true if the list is empty which allows the caller to remove
     this list from the list array. */
  return !nrecs;
}

list *
map::get_list_by_fd (int fd)
{
  int i;
  for (i=0; i<nlists; i++)
    /* The fd isn't sufficient since it could already be the fd of another
       file.  So we use the name hash value to identify the file unless
       it's an anonymous mapping in which case the fd (-1) is sufficient. */
    if ((fd == -1 && lists[i].get_fd () == -1)
	|| (fd != -1
	    && lists[i].get_hash () == cygheap->fdtab[fd]->get_namehash ()))
      return lists + i;
  return 0;
}

list *
map::add_list (int fd)
{
  if (nlists == maxlists)
    {
      list *new_lists;
      if (maxlists == 0)
	new_lists = (list *) cmalloc (HEAP_MMAP, 5 * sizeof (list));
      else
	new_lists = (list *) crealloc (lists, (maxlists + 5) * sizeof (list));
      if (!new_lists)
	return NULL;
      maxlists += 5;
      lists = new_lists;
    }
  lists[nlists].set (fd);
  return lists + nlists++;
}

void
map::del_list (int i)
{
  if (i < nlists)
    {
      lists[i].free_recs ();
      for (; i < nlists - 1; i++)
	lists[i] = lists[i + 1];
      nlists--;
    }
}

extern "C" void *
mmap64 (void *addr, size_t len, int prot, int flags, int fd, _off64_t off)
{
  syscall_printf ("addr %x, len %u, prot %x, flags %x, fd %d, off %D",
		  addr, len, prot, flags, fd, off);

  DWORD granularity = getshmlba ();

  /* Error conditions according to SUSv2 */
  if (off % getpagesize ()
      || (!(flags & MAP_SHARED) && !(flags & MAP_PRIVATE))
      || ((flags & MAP_SHARED) && (flags & MAP_PRIVATE))
      || ((flags & MAP_FIXED) && ((DWORD)addr % getpagesize ()))
      || !len)
    {
      set_errno (EINVAL);
      syscall_printf ("-1 = mmap(): EINVAL");
      return MAP_FAILED;
    }

  SetResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");

  if (mmapped_areas == NULL)
    {
      /* First mmap call, create STL map */
      mmapped_areas = (map *) ccalloc (HEAP_MMAP, 1, sizeof (map));
      if (mmapped_areas == NULL)
	{
	  set_errno (ENOMEM);
	  syscall_printf ("-1 = mmap(): ENOMEM");
	  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");
	  return MAP_FAILED;
	}
    }

  if (flags & MAP_ANONYMOUS)
    fd = -1;

  fhandler_base *fh = NULL;

  /* Get fhandler and convert /dev/zero mapping to MAP_ANONYMOUS mapping. */
  if (fd != -1)
    {
      /* Ensure that fd is open */
      cygheap_fdget cfd (fd);
      if (cfd < 0)
	{
	  syscall_printf ("-1 = mmap(): EBADF");
	  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");
	  return MAP_FAILED;
	}
      fh = cfd;
      if (fh->get_device () == FH_ZERO)
	{
	  /* mmap /dev/zero is like MAP_ANONYMOUS. */
	  fd = -1;
	  flags |= MAP_ANONYMOUS;
	}
    }
  if (fd == -1)
    {
      fh_paging_file.set_io_handle (INVALID_HANDLE_VALUE);
      fh = &fh_paging_file;
    }

  /* 9x only: If MAP_FIXED is requested on a non-granularity boundary,
     change request so that this looks like a request with offset
     addr % granularity. */
  if (wincap.share_mmaps_only_by_name () && fd == -1 && (flags & MAP_FIXED)
      && ((DWORD)addr % granularity) && !off)
    off = (DWORD)addr % granularity;
  /* Map always in multipliers of `granularity'-sized chunks.
     Not necessary for anonymous maps on NT. */
  _off64_t gran_off = off;
  DWORD gran_len = len;
  if (wincap.share_mmaps_only_by_name () || fd != -1)
    {
      gran_off = off & ~(granularity - 1);
      gran_len = howmany (off + len, granularity) * granularity - gran_off;
    }

  /* File mappings needs some extra care. */
  if (fd != -1 && fh->get_device () == FH_FS)
    {
      DWORD high;
      DWORD low = GetFileSize (fh->get_handle (), &high);
      _off64_t fsiz = ((_off64_t)high << 32) + low;

      /* Don't allow mappings beginning beyond EOF since Windows can't
	 handle that POSIX like, unless MAP_AUTOGROW flag is set, which
	 mimics Windows behaviour.  FIXME: Still looking for a good idea
	 to allow that under POSIX rules. */
      if (gran_off >= fsiz && !(flags & MAP_AUTOGROW))
	{
	  set_errno (ENXIO);
	  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK,
			       "mmap");
	  return MAP_FAILED;
	}
      /* Don't map beyond EOF.  Windows would change the file to the
	 new length otherwise, in contrast to POSIX.  Allow mapping
	 beyon EOF if MAP_AUTOGROW flag is set. */
      fsiz -= gran_off;
      if (gran_len > fsiz)
	{
	  if ((flags & MAP_AUTOGROW) && (off - gran_off) + len > fsiz)
	    {
	      /* Check if file has been opened for writing. */
	      if (!(fh->get_access () & GENERIC_WRITE))
		{
		  set_errno (EINVAL);
		  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK,
				       "mmap");
		  return MAP_FAILED;
		}
	      gran_len = (off - gran_off) + len;
	    }
	  else
	    gran_len = fsiz;
	}
      /* If the requested len is <= file size, drop the MAP_AUTOGROW flag.
	 This simplifes fhandler::mmap's job. */
      if ((flags & MAP_AUTOGROW) && gran_len <= fsiz)
	flags &= ~MAP_AUTOGROW;
    }

  DWORD access = (prot & PROT_WRITE) ? FILE_MAP_WRITE : FILE_MAP_READ;
  /* copy-on-write doesn't work at all on 9x using anonymous maps.
     Workaround: Anonymous mappings always use normal READ or WRITE
		 access and don't use named file mapping.
     copy-on-write doesn't also work properly on 9x with real files.
     While the changes are not propagated to the file, they are
     visible to other processes sharing the same file mapping object.
     Workaround: Don't use named file mapping.  That should work since
		 sharing file mappings only works reliable using named
		 file mapping on 9x.
  */
  if ((flags & MAP_PRIVATE)
      && (wincap.has_working_copy_on_write () || fd != -1))
    access = FILE_MAP_COPY;

  list *map_list = mmapped_areas->get_list_by_fd (fd);

  /* A bit of memory munging on 9x. */
  if (map_list && fd == -1 && wincap.share_mmaps_only_by_name ())
    {
      /* First check if this mapping matches into the chunk of another
	 already performed mapping. Only valid for MAP_ANON in a special
	 case of MAP_PRIVATE. */
      if (off == 0 && !(flags & MAP_FIXED))
	{
	  mmap_record *rec;
	  if ((rec = map_list->search_record (off, len)) != NULL
	      && rec->get_access () == access)
	    {
	      if ((off = rec->map_pages (off, len)) == (_off64_t)-1)
		{
		  syscall_printf ("-1 = mmap()");
		  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK|WRITE_LOCK,
				       "mmap");
		  return MAP_FAILED;
		}
	      caddr_t ret = rec->get_address () + off;
	      syscall_printf ("%x = mmap() succeeded", ret);
	      ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK,
				   "mmap");
	      return ret;
	    }
	}
      if ((flags & MAP_FIXED))
	{
	  caddr_t u_addr;
	  DWORD u_len;
	  long record_idx = -1;
	  if ((record_idx = map_list->search_record ((caddr_t)addr, len,
						     u_addr, u_len,
						     record_idx)) >= 0)
	    {
	      mmap_record *rec = map_list->get_record (record_idx);
	      if (u_addr > (caddr_t)addr || u_addr + len < (caddr_t)addr + len
		  || rec->get_access () != access)
		{
		  /* Partial match only, or access mode doesn't match. */
		  /* FIXME: Handle partial mappings gracefully if adjacent
		     memory is available. */
		  set_errno (EINVAL);
		  syscall_printf ("-1 = mmap()");
		  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK,
				       "mmap");
		  return MAP_FAILED;
		}
	      if (!rec->map_pages ((caddr_t)addr, len))
		{
		  syscall_printf ("-1 = mmap()");
		  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK,
				       "mmap");
		  return MAP_FAILED;
		}
	      caddr_t ret = (caddr_t)addr;
	      syscall_printf ("%x = mmap() succeeded", ret);
	      ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK,
				   "mmap");
	      return ret;
	    }
	}
    }

  caddr_t base = (caddr_t)addr;
  /* This shifts the base address to the next lower 64K boundary.
     The offset is re-added when evaluating the return value. */
  if (base)
    base -= off - gran_off;

  HANDLE h = fh->mmap (&base, gran_len, access, flags, gran_off);

  if (h == INVALID_HANDLE_VALUE)
    {
      ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");
      return MAP_FAILED;
    }

  /* At this point we should have a successfully mmapped area.
     Now it's time for bookkeeping stuff. */
  if (fd == -1)
    gran_len = PAGE_CNT (gran_len) * getpagesize ();
  mmap_record mmap_rec (fd, h, access, flags, gran_off, gran_len, base);

  /* Get list of mmapped areas for this fd, create a new one if
     one does not exist yet.
  */
  if (!map_list)
    {
      /* Create a new one */
      map_list = mmapped_areas->add_list (fd);
      if (!map_list)
	{
	  fh->munmap (h, base, gran_len);
	  set_errno (ENOMEM);
	  syscall_printf ("-1 = mmap(): ENOMEM");
	  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");
	  return MAP_FAILED;
	}
  }

  /* Insert into the list */
  mmap_record *rec = map_list->add_record (mmap_rec, off,
					   len > gran_len ? gran_len : len);
  if (!rec)
    {
      fh->munmap (h, base, gran_len);
      set_errno (ENOMEM);
      syscall_printf ("-1 = mmap(): ENOMEM");
      ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");
      return MAP_FAILED;
    }

  caddr_t ret = rec->get_address () + (off - gran_off);
  syscall_printf ("%x = mmap() succeeded", ret);
  ReleaseResourceLock (LOCK_MMAP_LIST, READ_LOCK | WRITE_LOCK, "mmap");
  return ret;
}

extern "C" void *
mmap (void *addr, size_t len, int prot, int flags, int fd, _off_t off)
{
  return mmap64 (addr, len, prot, flags, fd, (_off64_t)off);
}

/* munmap () removes all mmapped pages between addr and addr+len. */

extern "C" int
munmap (void *addr, size_t len)
{
  syscall_printf ("munmap (addr %x, len %u)", addr, len);

  /* Error conditions according to SUSv3 */
  if (!addr || ((DWORD)addr % getpagesize ()) || !len
      || check_invalid_virtual_addr (addr, len))
    {
      set_errno (EINVAL);
      syscall_printf ("-1 = munmap(): Invalid parameters");
      return -1;
    }

  SetResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK, "munmap");
  if (mmapped_areas == NULL)
    {
      syscall_printf ("-1 = munmap(): mmapped_areas == NULL");
      ReleaseResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK, "munmap");
      return 0;
    }

  /* Iterate through the map, unmap pages between addr and addr+len
     in all maps. */
  list *map_list;
  for (int list_idx = 0;
       (map_list = mmapped_areas->get_list (list_idx));
       ++list_idx)
    {
      long record_idx = -1;
      caddr_t u_addr;
      DWORD u_len;

      while ((record_idx = map_list->search_record((caddr_t)addr, len, u_addr,
						   u_len, record_idx)) >= 0)
	{
	  mmap_record *rec = map_list->get_record (record_idx);
	  if (rec->unmap_pages (u_addr, u_len))
	    {
	      /* The whole record has been unmapped, so we now actually
		 unmap it from the system in full length... */
	      fhandler_base *fh = rec->alloc_fh ();
	      fh->munmap (rec->get_handle (),
			  rec->get_address (),
			  rec->get_size ());
	      rec->free_fh (fh);

	      /* ...and delete the record. */
	      if (map_list->del_record (record_idx--))
		{
		  /* Yay, the last record has been removed from the list,
		     we can remove the list now, too. */
		  mmapped_areas->del_list (list_idx--);
		  break;
		}
	    }
	}
    }

  ReleaseResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK, "munmap");
  syscall_printf ("0 = munmap(): %x", addr);
  return 0;
}

/* Sync file with memory. Ignore flags for now. */

extern "C" int
msync (void *addr, size_t len, int flags)
{
  syscall_printf ("addr %x, len %u, flags %x", addr, len, flags);

  /* However, check flags for validity. */
  if ((flags & ~(MS_ASYNC | MS_SYNC | MS_INVALIDATE))
      || ((flags & MS_ASYNC) && (flags & MS_SYNC)))
    {
      syscall_printf ("-1 = msync(): Invalid flags");
      set_errno (EINVAL);
      return -1;
    }

  SetResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK, "msync");
  /* Check if a mmap'ed area was ever created */
  if (mmapped_areas == NULL)
    {
      syscall_printf ("-1 = msync(): mmapped_areas == NULL");
      set_errno (EINVAL);
      ReleaseResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK, "msync");
      return -1;
    }

  /* Iterate through the map, looking for the mmapped area.
     Error if not found. */

  list *map_list;
  for (int list_idx = 0;
       (map_list = mmapped_areas->get_list (list_idx));
       ++list_idx)
    {
      mmap_record *rec;
      for (int record_idx = 0;
	   (rec = map_list->get_record (record_idx));
	   ++record_idx)
	{
	  if (rec->access ((caddr_t)addr))
	    {
	      /* Check whole area given by len. */
	      for (DWORD i = getpagesize (); i < len; ++i)
		if (!rec->access ((caddr_t)addr + i))
		  goto invalid_address_range;
	      fhandler_base *fh = rec->alloc_fh ();
	      int ret = fh->msync (rec->get_handle (), (caddr_t)addr, len,
				   flags);
	      rec->free_fh (fh);

	      if (ret)
		syscall_printf ("%d = msync(), %E", ret);
	      else
		syscall_printf ("0 = msync()");

	      ReleaseResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK,
				   "msync");
	      return 0;
	    }
	}
    }

invalid_address_range:
  /* SUSv2: Return code if indicated memory was not mapped is ENOMEM. */
  set_errno (ENOMEM);
  syscall_printf ("-1 = msync(): ENOMEM");

  ReleaseResourceLock (LOCK_MMAP_LIST, WRITE_LOCK | READ_LOCK, "msync");
  return -1;
}

/* Set memory protection */

extern "C" int
mprotect (void *addr, size_t len, int prot)
{
  DWORD old_prot;
  DWORD new_prot = 0;

  syscall_printf ("mprotect (addr %x, len %u, prot %x)", addr, len, prot);

  if (!wincap.virtual_protect_works_on_shared_pages ()
      && addr >= (caddr_t)0x80000000 && addr <= (caddr_t)0xBFFFFFFF)
    {
      syscall_printf ("0 = mprotect (9x: No VirtualProtect on shared memory)");
      return 0;
    }

  /* If write protection is requested, check if the page was
     originally protected writecopy.  In this case call VirtualProtect
     requesting PAGE_WRITECOPY, otherwise the VirtualProtect will fail
     on NT version >= 5.0 */
  bool writecopy = false;
  if (prot & PROT_WRITE)
    {
      MEMORY_BASIC_INFORMATION mbi;
      if (VirtualQuery (addr, &mbi, sizeof mbi))
	{
	  if (mbi.AllocationProtect == PAGE_WRITECOPY
	      || mbi.AllocationProtect == PAGE_EXECUTE_WRITECOPY)
	    writecopy = true;
	}
    }

  switch (prot)
    {
      case PROT_READ | PROT_WRITE | PROT_EXEC:
      case PROT_WRITE | PROT_EXEC:
	new_prot = writecopy ? PAGE_EXECUTE_WRITECOPY : PAGE_EXECUTE_READWRITE;
	break;
      case PROT_READ | PROT_WRITE:
      case PROT_WRITE:
	new_prot = writecopy ? PAGE_WRITECOPY : PAGE_READWRITE;
	break;
      case PROT_READ | PROT_EXEC:
	new_prot = PAGE_EXECUTE_READ;
	break;
      case PROT_READ:
	new_prot = PAGE_READONLY;
	break;
      case PROT_EXEC:
	new_prot = PAGE_EXECUTE;
	break;
      case PROT_NONE:
	new_prot = PAGE_NOACCESS;
	break;
      default:
	syscall_printf ("-1 = mprotect (): invalid prot value");
	set_errno (EINVAL);
	return -1;
     }

  if (VirtualProtect (addr, len, new_prot, &old_prot) == 0)
    {
      __seterrno ();
      syscall_printf ("-1 = mprotect (), %E");
      return -1;
    }

  syscall_printf ("0 = mprotect ()");
  return 0;
}

/*
 * Base implementation:
 *
 * `mmap' returns ENODEV as documented in SUSv2.
 * In contrast to the global function implementation, the member function
 * `mmap' has to return the mapped base address in `addr' and the handle to
 * the mapping object as return value. In case of failure, the fhandler
 * mmap has to close that handle by itself and return INVALID_HANDLE_VALUE.
 *
 * `munmap' and `msync' get the handle to the mapping object as first parameter
 * additionally.
*/
HANDLE
fhandler_base::mmap (caddr_t *addr, size_t len, DWORD access,
		     int flags, _off64_t off)
{
  set_errno (ENODEV);
  return INVALID_HANDLE_VALUE;
}

int
fhandler_base::munmap (HANDLE h, caddr_t addr, size_t len)
{
  set_errno (ENODEV);
  return -1;
}

int
fhandler_base::msync (HANDLE h, caddr_t addr, size_t len, int flags)
{
  set_errno (ENODEV);
  return -1;
}

bool
fhandler_base::fixup_mmap_after_fork (HANDLE h, DWORD access, int flags,
				      _off64_t offset, DWORD size,
				      void *address)
{
  set_errno (ENODEV);
  return -1;
}

/* Implementation for disk files. */
HANDLE
fhandler_disk_file::mmap (caddr_t *addr, size_t len, DWORD access,
			  int flags, _off64_t off)
{
  DWORD protect;

  switch (access)
    {
      case FILE_MAP_WRITE:
	protect = PAGE_READWRITE;
	break;
      case FILE_MAP_READ:
	protect = PAGE_READONLY;
	break;
      default:
	protect = PAGE_WRITECOPY;
	break;
    }

  HANDLE h;
  DWORD high, low;

  /* On 9x/ME try first to open the mapping by name when opening a
     shared file object. This is needed since 9x/ME only shares
     objects between processes by name. What a mess... */
  if (wincap.share_mmaps_only_by_name ()
      && get_handle () != INVALID_HANDLE_VALUE
      && !(access & FILE_MAP_COPY))
    {
      /* Grrr, the whole stuff is just needed to try to get a reliable
	 mapping of the same file. Even that uprising isn't bullet
	 proof but it does it's best... */
      char namebuf[CYG_MAX_PATH];
      cygwin_conv_to_full_posix_path (get_name (), namebuf);
      for (int i = strlen (namebuf) - 1; i >= 0; --i)
	namebuf[i] = cyg_tolower (namebuf [i]);

      debug_printf ("named sharing");
      if (!(h = OpenFileMapping (access, TRUE, namebuf)))
	h = CreateFileMapping (get_handle (), &sec_none, protect, 0, 0,
			       namebuf);
    }
  else if (get_handle () == INVALID_HANDLE_VALUE)
    {
      /* Standard anonymous mapping needs non-zero len. */
      h = CreateFileMapping (get_handle (), &sec_none, protect, 0, len, NULL);
    }
  else if (flags & MAP_AUTOGROW)
    {
      high = (off + len) >> 32;
      low = (off + len) & UINT32_MAX;
      /* Auto-grow in CreateFileMapping only works if the protection is
	 PAGE_READWRITE.  So, first we call CreateFileMapping with
	 PAGE_READWRITE, then, if the requested protection is different, we
	 close the mapping and reopen it again with the correct protection,
	 *iff* auto-grow worked. */
      h = CreateFileMapping (get_handle (), &sec_none, PAGE_READWRITE,
			     high, low, NULL);
      if (h && protect != PAGE_READWRITE)
	{
	  CloseHandle (h);
	  h = CreateFileMapping (get_handle (), &sec_none, protect,
				 high, low, NULL);
	}
    }
  else
    {
      /* Zero len creates mapping for whole file. */
      h = CreateFileMapping (get_handle (), &sec_none, protect, 0, 0, NULL);
    }
  if (!h)
    {
      __seterrno ();
      syscall_printf ("-1 = mmap(): CreateFileMapping failed with %E");
      return INVALID_HANDLE_VALUE;
    }

  high = off >> 32;
  low = off & UINT32_MAX;
  void *base = NULL;
  /* If a non-zero address is given, try mapping using the given address first.
     If it fails and flags is not MAP_FIXED, try again with NULL address. */
  if (!wincap.share_mmaps_only_by_name ()
      && get_handle () == INVALID_HANDLE_VALUE)
    {
      PHYSICAL_ADDRESS phys;
      phys.QuadPart = (ULONGLONG) off;
      ULONG ulen = len;
      base = *addr ?: (void *) mmapped_areas->get_next_anon_addr ();
      NTSTATUS ret = NtMapViewOfSection (h, INVALID_HANDLE_VALUE, &base, 0L,
					 ulen, &phys, &ulen, ViewShare,
					 base ? AT_ROUND_TO_PAGE : 0, protect);
      if (ret != STATUS_SUCCESS)
	{
	  __seterrno_from_nt_status (ret);
	  base = NULL;
	}
      else
	mmapped_areas->set_next_anon_addr ((caddr_t) base + len);
    }
  else if (*addr)
    base = MapViewOfFileEx (h, access, high, low, len, *addr);
  if (!base && !(flags & MAP_FIXED))
    base = MapViewOfFileEx (h, access, high, low, len, NULL);
  debug_printf ("%x = MapViewOfFileEx (h:%x, access:%x, 0, off:%D, "
		"len:%u, addr:%x)", base, h, access, off, len, *addr);
  if (!base || ((flags & MAP_FIXED) && base != *addr))
    {
      if (!base)
	{
	  __seterrno ();
	  syscall_printf ("-1 = mmap(): MapViewOfFileEx failed with %E");
	}
      else
	{
	  set_errno (EINVAL);
	  syscall_printf ("-1 = mmap(): address shift with MAP_FIXED given");
	}
      CloseHandle (h);
      return INVALID_HANDLE_VALUE;
    }

  *addr = (caddr_t) base;
  return h;
}

int
fhandler_disk_file::munmap (HANDLE h, caddr_t addr, size_t len)
{
  UnmapViewOfFile (addr);
  CloseHandle (h);
  return 0;
}

int
fhandler_disk_file::msync (HANDLE h, caddr_t addr, size_t len, int flags)
{
  if (FlushViewOfFile (addr, len) == 0)
    {
      __seterrno ();
      return -1;
    }
  return 0;
}

bool
fhandler_disk_file::fixup_mmap_after_fork (HANDLE h, DWORD access, int flags,
					   _off64_t offset, DWORD size,
					   void *address)
{
  /* Re-create the MapViewOfFileEx call */
  void *base;
  if (!wincap.share_mmaps_only_by_name () && (flags & MAP_ANONYMOUS))
    {
      PHYSICAL_ADDRESS phys;
      phys.QuadPart = (ULONGLONG) offset;
      ULONG ulen = size;
      base = address;
      DWORD protect;
      switch (access)
	{
	  case FILE_MAP_WRITE:
	    protect = PAGE_READWRITE;
	    break;
	  case FILE_MAP_READ:
	    protect = PAGE_READONLY;
	    break;
	  default:
	    protect = PAGE_WRITECOPY;
	    break;
	}
      NTSTATUS ret = NtMapViewOfSection (h, INVALID_HANDLE_VALUE, &base, 0L,
					 ulen, &phys, &ulen, ViewShare,
					 AT_ROUND_TO_PAGE, protect);
      if (ret != STATUS_SUCCESS)
	__seterrno_from_nt_status (ret);
    }
  else
    base = MapViewOfFileEx (h, access, 0, offset, size, address);
  if (base != address)
    {
      MEMORY_BASIC_INFORMATION m;
      VirtualQuery (address, &m, sizeof (m));
      system_printf ("requested %p != %p mem alloc base %p, state %p, size %d, %E",
		     address, base, m.AllocationBase, m.State, m.RegionSize);
    }
  return base == address;
}

/*
 * Call to re-create all the file mappings in a forked
 * child. Called from the child in initialization. At this
 * point we are passed a valid mmapped_areas map, and all the
 * HANDLE's are valid for the child, but none of the
 * mapped areas are in our address space. We need to iterate
 * through the map, doing the MapViewOfFile calls.
 */

int __stdcall
fixup_mmaps_after_fork (HANDLE parent)
{

  debug_printf ("recreate_mmaps_after_fork, mmapped_areas %p", mmapped_areas);

  /* Check if a mmapped area was ever created */
  if (mmapped_areas == NULL)
    return 0;

  /* Iterate through the map */
  list *map_list;
  for (int list_idx = 0;
       (map_list = mmapped_areas->get_list (list_idx));
       ++list_idx)
    {
      mmap_record *rec;
      for (int record_idx = 0;
	   (rec = map_list->get_record (record_idx));
	   ++record_idx)
	{

	  debug_printf ("fd %d, h %x, access %x, offset %D, size %u, address %p",
	      rec->get_fd (), rec->get_handle (), rec->get_access (),
	      rec->get_offset (), rec->get_size (), rec->get_address ());

	  fhandler_base *fh = rec->alloc_fh ();
	  bool ret = fh->fixup_mmap_after_fork (rec->get_handle (),
						rec->get_access (),
						rec->get_flags (),
						rec->get_offset (),
						rec->get_size (),
						rec->get_address ());
	  rec->free_fh (fh);

	  if (!ret)
	    return -1;
	  if (rec->get_access () == FILE_MAP_COPY)
	    {
	      for (char *address = rec->get_address ();
		   address < rec->get_address () + rec->get_size ();
		   address += getpagesize ())
		if (rec->access (address)
		    && !ReadProcessMemory (parent, address, address,
					   getpagesize (), NULL))
		  {
		    DWORD old_prot;
		    DWORD last_error = GetLastError ();

		    if (last_error != ERROR_PARTIAL_COPY
			&& last_error != ERROR_NOACCESS
			|| !wincap.virtual_protect_works_on_shared_pages ())
		      {
			system_printf ("ReadProcessMemory failed for "
				       "MAP_PRIVATE address %p, %E",
				       rec->get_address ());
			return -1;
		      }
		    if (!VirtualProtectEx (parent,
					   address, getpagesize (),
					   PAGE_READONLY, &old_prot))
		      {
			system_printf ("VirtualProtectEx failed for "
				       "MAP_PRIVATE address %p, %E",
				       rec->get_address ());
			return -1;
		      }
		    else
		      {
			BOOL ret;
			DWORD dummy_prot;

			ret = ReadProcessMemory (parent, address, address,
						 getpagesize (), NULL);
			if (!VirtualProtectEx(parent,
					      address, getpagesize (),
					      old_prot, &dummy_prot))
			  system_printf ("WARNING: VirtualProtectEx to "
					 "return to previous state "
					 "in parent failed for "
					 "MAP_PRIVATE address %p, %E",
					 rec->get_address ());
			if (!VirtualProtect (address, getpagesize (),
					     old_prot, &dummy_prot))
			  system_printf ("WARNING: VirtualProtect to copy "
					 "protection to child failed for"
					 "MAP_PRIVATE address %p, %E",
					 rec->get_address ());
			if (!ret)
			  {
			    system_printf ("ReadProcessMemory (2nd try) "
					   "failed for "
					   "MAP_PRIVATE address %p, %E",
					   rec->get_address ());
			    return -1;
			  }
		      }
		  }
	    }
	  rec->fixup_page_map ();
	}
    }

  debug_printf ("succeeded");
  return 0;
}