Sweden-Number/server/mapping.c

1126 lines
40 KiB
C
Raw Normal View History

/*
* Server-side file mapping management
*
* Copyright (C) 1999 Alexandre Julliard
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
*/
#include "config.h"
#include "wine/port.h"
#include <assert.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/stat.h>
#ifdef HAVE_SYS_MMAN_H
# include <sys/mman.h>
#endif
#include <unistd.h>
2005-11-28 17:32:54 +01:00
#include "ntstatus.h"
#define WIN32_NO_STATUS
#include "windef.h"
#include "winternl.h"
#include "ddk/wdm.h"
#include "file.h"
#include "handle.h"
#include "thread.h"
#include "process.h"
#include "request.h"
#include "security.h"
/* list of memory ranges, used to store committed info */
struct ranges
{
struct object obj; /* object header */
unsigned int count; /* number of used ranges */
unsigned int max; /* number of allocated ranges */
struct range
{
file_pos_t start;
file_pos_t end;
} *ranges;
};
static void ranges_dump( struct object *obj, int verbose );
static void ranges_destroy( struct object *obj );
static const struct object_ops ranges_ops =
{
sizeof(struct ranges), /* size */
ranges_dump, /* dump */
no_get_type, /* get_type */
no_add_queue, /* add_queue */
NULL, /* remove_queue */
NULL, /* signaled */
NULL, /* satisfied */
no_signal, /* signal */
no_get_fd, /* get_fd */
no_map_access, /* map_access */
default_get_sd, /* get_sd */
default_set_sd, /* set_sd */
no_lookup_name, /* lookup_name */
no_link_name, /* link_name */
NULL, /* unlink_name */
no_open_file, /* open_file */
no_kernel_obj_list, /* get_kernel_obj_list */
no_close_handle, /* close_handle */
ranges_destroy /* destroy */
};
/* file backing the shared sections of a PE image mapping */
struct shared_map
{
struct object obj; /* object header */
struct fd *fd; /* file descriptor of the mapped PE file */
struct file *file; /* temp file holding the shared data */
struct list entry; /* entry in global shared maps list */
};
static void shared_map_dump( struct object *obj, int verbose );
static void shared_map_destroy( struct object *obj );
static const struct object_ops shared_map_ops =
{
sizeof(struct shared_map), /* size */
shared_map_dump, /* dump */
no_get_type, /* get_type */
no_add_queue, /* add_queue */
NULL, /* remove_queue */
NULL, /* signaled */
NULL, /* satisfied */
no_signal, /* signal */
no_get_fd, /* get_fd */
no_map_access, /* map_access */
default_get_sd, /* get_sd */
default_set_sd, /* set_sd */
no_lookup_name, /* lookup_name */
no_link_name, /* link_name */
NULL, /* unlink_name */
no_open_file, /* open_file */
no_kernel_obj_list, /* get_kernel_obj_list */
no_close_handle, /* close_handle */
shared_map_destroy /* destroy */
};
static struct list shared_map_list = LIST_INIT( shared_map_list );
/* memory view mapped in client address space */
struct memory_view
{
struct list entry; /* entry in per-process view list */
struct fd *fd; /* fd for mapped file */
struct ranges *committed; /* list of committed ranges in this mapping */
struct shared_map *shared; /* temp file for shared PE mapping */
pe_image_info_t image; /* image info (for PE image mapping) */
unsigned int flags; /* SEC_* flags */
client_ptr_t base; /* view base address (in process addr space) */
mem_size_t size; /* view size */
file_pos_t start; /* start offset in mapping */
};
struct mapping
{
2001-04-14 00:38:39 +02:00
struct object obj; /* object header */
mem_size_t size; /* mapping size */
unsigned int flags; /* SEC_* flags */
struct fd *fd; /* fd for mapped file */
pe_image_info_t image; /* image info (for PE image mapping) */
struct ranges *committed; /* list of committed ranges in this mapping */
struct shared_map *shared; /* temp file for shared PE mapping */
};
static void mapping_dump( struct object *obj, int verbose );
static struct object_type *mapping_get_type( struct object *obj );
static struct fd *mapping_get_fd( struct object *obj );
static unsigned int mapping_map_access( struct object *obj, unsigned int access );
static void mapping_destroy( struct object *obj );
static enum server_fd_type mapping_get_fd_type( struct fd *fd );
static const struct object_ops mapping_ops =
{
sizeof(struct mapping), /* size */
mapping_dump, /* dump */
mapping_get_type, /* get_type */
no_add_queue, /* add_queue */
NULL, /* remove_queue */
NULL, /* signaled */
NULL, /* satisfied */
no_signal, /* signal */
mapping_get_fd, /* get_fd */
mapping_map_access, /* map_access */
default_get_sd, /* get_sd */
default_set_sd, /* set_sd */
no_lookup_name, /* lookup_name */
directory_link_name, /* link_name */
default_unlink_name, /* unlink_name */
no_open_file, /* open_file */
no_kernel_obj_list, /* get_kernel_obj_list */
fd_close_handle, /* close_handle */
mapping_destroy /* destroy */
};
static const struct fd_ops mapping_fd_ops =
{
default_fd_get_poll_events, /* get_poll_events */
default_poll_event, /* poll_event */
mapping_get_fd_type, /* get_fd_type */
no_fd_read, /* read */
no_fd_write, /* write */
no_fd_flush, /* flush */
no_fd_get_file_info, /* get_file_info */
no_fd_get_volume_info, /* get_volume_info */
no_fd_ioctl, /* ioctl */
no_fd_queue_async, /* queue_async */
default_fd_reselect_async /* reselect_async */
};
static size_t page_mask;
#define ROUND_SIZE(size) (((size) + page_mask) & ~page_mask)
static void ranges_dump( struct object *obj, int verbose )
{
struct ranges *ranges = (struct ranges *)obj;
fprintf( stderr, "Memory ranges count=%u\n", ranges->count );
}
static void ranges_destroy( struct object *obj )
{
struct ranges *ranges = (struct ranges *)obj;
free( ranges->ranges );
}
static void shared_map_dump( struct object *obj, int verbose )
{
struct shared_map *shared = (struct shared_map *)obj;
fprintf( stderr, "Shared mapping fd=%p file=%p\n", shared->fd, shared->file );
}
static void shared_map_destroy( struct object *obj )
{
struct shared_map *shared = (struct shared_map *)obj;
release_object( shared->fd );
release_object( shared->file );
list_remove( &shared->entry );
}
/* extend a file beyond the current end of file */
static int grow_file( int unix_fd, file_pos_t new_size )
{
static const char zero;
off_t size = new_size;
if (sizeof(new_size) > sizeof(size) && size != new_size)
{
set_error( STATUS_INVALID_PARAMETER );
return 0;
}
/* extend the file one byte beyond the requested size and then truncate it */
/* this should work around ftruncate implementations that can't extend files */
if (pwrite( unix_fd, &zero, 1, size ) != -1)
{
ftruncate( unix_fd, size );
return 1;
}
file_set_error();
return 0;
}
/* check if the current directory allows exec mappings */
static int check_current_dir_for_exec(void)
{
int fd;
char tmpfn[] = "anonmap.XXXXXX";
void *ret = MAP_FAILED;
fd = mkstemps( tmpfn, 0 );
if (fd == -1) return 0;
if (grow_file( fd, 1 ))
{
ret = mmap( NULL, get_page_size(), PROT_READ | PROT_EXEC, MAP_PRIVATE, fd, 0 );
if (ret != MAP_FAILED) munmap( ret, get_page_size() );
}
close( fd );
unlink( tmpfn );
return (ret != MAP_FAILED);
}
/* create a temp file for anonymous mappings */
static int create_temp_file( file_pos_t size )
{
static int temp_dir_fd = -1;
char tmpfn[] = "anonmap.XXXXXX";
int fd;
if (temp_dir_fd == -1)
{
temp_dir_fd = server_dir_fd;
if (!check_current_dir_for_exec())
{
/* the server dir is noexec, try the config dir instead */
fchdir( config_dir_fd );
if (check_current_dir_for_exec())
temp_dir_fd = config_dir_fd;
else /* neither works, fall back to server dir */
fchdir( server_dir_fd );
}
}
else if (temp_dir_fd != server_dir_fd) fchdir( temp_dir_fd );
fd = mkstemps( tmpfn, 0 );
if (fd != -1)
{
if (!grow_file( fd, size ))
{
close( fd );
fd = -1;
}
unlink( tmpfn );
}
else file_set_error();
if (temp_dir_fd != server_dir_fd) fchdir( server_dir_fd );
return fd;
}
/* find a memory view from its base address */
static struct memory_view *find_mapped_view( struct process *process, client_ptr_t base )
{
struct memory_view *view;
LIST_FOR_EACH_ENTRY( view, &process->views, struct memory_view, entry )
if (view->base == base) return view;
set_error( STATUS_NOT_MAPPED_VIEW );
return NULL;
}
static void free_memory_view( struct memory_view *view )
{
if (view->fd) release_object( view->fd );
if (view->committed) release_object( view->committed );
if (view->shared) release_object( view->shared );
list_remove( &view->entry );
free( view );
}
/* free all mapped views at process exit */
void free_mapped_views( struct process *process )
{
struct list *ptr;
while ((ptr = list_head( &process->views )))
free_memory_view( LIST_ENTRY( ptr, struct memory_view, entry ));
}
2001-04-14 00:38:39 +02:00
/* find the shared PE mapping for a given mapping */
static struct shared_map *get_shared_file( struct fd *fd )
2001-04-14 00:38:39 +02:00
{
struct shared_map *ptr;
2001-04-14 00:38:39 +02:00
LIST_FOR_EACH_ENTRY( ptr, &shared_map_list, struct shared_map, entry )
if (is_same_file_fd( ptr->fd, fd ))
return (struct shared_map *)grab_object( ptr );
2001-04-14 00:38:39 +02:00
return NULL;
}
/* return the size of the memory mapping and file range of a given section */
static inline void get_section_sizes( const IMAGE_SECTION_HEADER *sec, size_t *map_size,
off_t *file_start, size_t *file_size )
{
static const unsigned int sector_align = 0x1ff;
if (!sec->Misc.VirtualSize) *map_size = ROUND_SIZE( sec->SizeOfRawData );
else *map_size = ROUND_SIZE( sec->Misc.VirtualSize );
*file_start = sec->PointerToRawData & ~sector_align;
*file_size = (sec->SizeOfRawData + (sec->PointerToRawData & sector_align) + sector_align) & ~sector_align;
if (*file_size > *map_size) *file_size = *map_size;
}
/* add a range to the committed list */
static void add_committed_range( struct memory_view *view, file_pos_t start, file_pos_t end )
{
unsigned int i, j;
struct ranges *committed = view->committed;
struct range *ranges;
if ((start & page_mask) || (end & page_mask) ||
start >= view->size || end >= view->size ||
start >= end)
{
set_error( STATUS_INVALID_PARAMETER );
return;
}
if (!committed) return; /* everything committed already */
start += view->start;
end += view->start;
for (i = 0, ranges = committed->ranges; i < committed->count; i++)
{
if (ranges[i].start > end) break;
if (ranges[i].end < start) continue;
if (ranges[i].start > start) ranges[i].start = start; /* extend downwards */
if (ranges[i].end < end) /* extend upwards and maybe merge with next */
{
for (j = i + 1; j < committed->count; j++)
{
if (ranges[j].start > end) break;
if (ranges[j].end > end) end = ranges[j].end;
}
if (j > i + 1)
{
memmove( &ranges[i + 1], &ranges[j], (committed->count - j) * sizeof(*ranges) );
committed->count -= j - (i + 1);
}
ranges[i].end = end;
}
return;
}
/* now add a new range */
if (committed->count == committed->max)
{
unsigned int new_size = committed->max * 2;
struct range *new_ptr = realloc( committed->ranges, new_size * sizeof(*new_ptr) );
if (!new_ptr) return;
committed->max = new_size;
ranges = committed->ranges = new_ptr;
}
memmove( &ranges[i + 1], &ranges[i], (committed->count - i) * sizeof(*ranges) );
ranges[i].start = start;
ranges[i].end = end;
committed->count++;
}
/* find the range containing start and return whether it's committed */
static int find_committed_range( struct memory_view *view, file_pos_t start, mem_size_t *size )
{
unsigned int i;
struct ranges *committed = view->committed;
struct range *ranges;
if ((start & page_mask) || start >= view->size)
{
set_error( STATUS_INVALID_PARAMETER );
return 0;
}
if (!committed) /* everything is committed */
{
*size = view->size - start;
return 1;
}
for (i = 0, ranges = committed->ranges; i < committed->count; i++)
{
if (ranges[i].start > view->start + start)
{
*size = min( ranges[i].start, view->start + view->size ) - (view->start + start);
return 0;
}
if (ranges[i].end > view->start + start)
{
*size = min( ranges[i].end, view->start + view->size ) - (view->start + start);
return 1;
}
}
*size = view->size - start;
return 0;
}
/* allocate and fill the temp file for a shared PE image mapping */
static int build_shared_mapping( struct mapping *mapping, int fd,
IMAGE_SECTION_HEADER *sec, unsigned int nb_sec )
{
struct shared_map *shared;
struct file *file;
unsigned int i;
mem_size_t total_size;
size_t file_size, map_size, max_size;
off_t shared_pos, read_pos, write_pos;
char *buffer = NULL;
int shared_fd;
long toread;
/* compute the total size of the shared mapping */
total_size = max_size = 0;
for (i = 0; i < nb_sec; i++)
{
if ((sec[i].Characteristics & IMAGE_SCN_MEM_SHARED) &&
(sec[i].Characteristics & IMAGE_SCN_MEM_WRITE))
{
get_section_sizes( &sec[i], &map_size, &read_pos, &file_size );
if (file_size > max_size) max_size = file_size;
total_size += map_size;
}
}
if (!total_size) return 1; /* nothing to do */
if ((mapping->shared = get_shared_file( mapping->fd ))) return 1;
2001-04-14 00:38:39 +02:00
/* create a temp file for the mapping */
if ((shared_fd = create_temp_file( total_size )) == -1) return 0;
if (!(file = create_file_for_fd( shared_fd, FILE_GENERIC_READ|FILE_GENERIC_WRITE, 0 ))) return 0;
if (!(buffer = malloc( max_size ))) goto error;
/* copy the shared sections data into the temp file */
shared_pos = 0;
for (i = 0; i < nb_sec; i++)
{
if (!(sec[i].Characteristics & IMAGE_SCN_MEM_SHARED)) continue;
if (!(sec[i].Characteristics & IMAGE_SCN_MEM_WRITE)) continue;
get_section_sizes( &sec[i], &map_size, &read_pos, &file_size );
write_pos = shared_pos;
shared_pos += map_size;
if (!sec[i].PointerToRawData || !file_size) continue;
toread = file_size;
while (toread)
{
long res = pread( fd, buffer + file_size - toread, toread, read_pos );
if (!res && toread < 0x200) /* partial sector at EOF is not an error */
{
file_size -= toread;
break;
}
if (res <= 0) goto error;
toread -= res;
read_pos += res;
}
if (pwrite( shared_fd, buffer, file_size, write_pos ) != file_size) goto error;
}
if (!(shared = alloc_object( &shared_map_ops ))) goto error;
shared->fd = (struct fd *)grab_object( mapping->fd );
shared->file = file;
list_add_head( &shared_map_list, &shared->entry );
mapping->shared = shared;
free( buffer );
return 1;
error:
release_object( file );
free( buffer );
return 0;
}
/* load the CLR header from its section */
static int load_clr_header( IMAGE_COR20_HEADER *hdr, size_t va, size_t size, int unix_fd,
IMAGE_SECTION_HEADER *sec, unsigned int nb_sec )
{
ssize_t ret;
size_t map_size, file_size;
off_t file_start;
unsigned int i;
if (!va || !size) return 0;
for (i = 0; i < nb_sec; i++)
{
if (va < sec[i].VirtualAddress) continue;
if (sec[i].Misc.VirtualSize && va - sec[i].VirtualAddress >= sec[i].Misc.VirtualSize) continue;
get_section_sizes( &sec[i], &map_size, &file_start, &file_size );
if (size >= map_size) continue;
if (va - sec[i].VirtualAddress >= map_size - size) continue;
file_size = min( file_size, map_size );
size = min( size, sizeof(*hdr) );
ret = pread( unix_fd, hdr, min( size, file_size ), file_start + va - sec[i].VirtualAddress );
if (ret <= 0) break;
if (ret < sizeof(*hdr)) memset( (char *)hdr + ret, 0, sizeof(*hdr) - ret );
return (hdr->MajorRuntimeVersion > COR_VERSION_MAJOR_V2 ||
(hdr->MajorRuntimeVersion == COR_VERSION_MAJOR_V2 &&
hdr->MinorRuntimeVersion >= COR_VERSION_MINOR));
}
return 0;
}
/* retrieve the mapping parameters for an executable (PE) image */
static unsigned int get_image_params( struct mapping *mapping, file_pos_t file_size, int unix_fd )
{
static const char builtin_signature[] = "Wine builtin DLL";
static const char fakedll_signature[] = "Wine placeholder DLL";
IMAGE_COR20_HEADER clr;
IMAGE_SECTION_HEADER sec[96];
struct
{
IMAGE_DOS_HEADER dos;
char buffer[32];
} mz;
struct
{
DWORD Signature;
IMAGE_FILE_HEADER FileHeader;
union
{
IMAGE_OPTIONAL_HEADER32 hdr32;
IMAGE_OPTIONAL_HEADER64 hdr64;
} opt;
} nt;
off_t pos;
int size, opt_size;
size_t mz_size, clr_va, clr_size;
unsigned int i, cpu_mask = get_supported_cpu_mask();
/* load the headers */
if (!file_size) return STATUS_INVALID_FILE_FOR_SECTION;
size = pread( unix_fd, &mz, sizeof(mz), 0 );
if (size < sizeof(mz.dos)) return STATUS_INVALID_IMAGE_NOT_MZ;
if (mz.dos.e_magic != IMAGE_DOS_SIGNATURE) return STATUS_INVALID_IMAGE_NOT_MZ;
mz_size = size;
pos = mz.dos.e_lfanew;
size = pread( unix_fd, &nt, sizeof(nt), pos );
if (size < sizeof(nt.Signature) + sizeof(nt.FileHeader)) return STATUS_INVALID_IMAGE_PROTECT;
/* zero out Optional header in the case it's not present or partial */
opt_size = max( nt.FileHeader.SizeOfOptionalHeader, offsetof( IMAGE_OPTIONAL_HEADER32, CheckSum ));
size = min( size, sizeof(nt.Signature) + sizeof(nt.FileHeader) + opt_size );
if (size < sizeof(nt)) memset( (char *)&nt + size, 0, sizeof(nt) - size );
if (nt.Signature != IMAGE_NT_SIGNATURE)
{
IMAGE_OS2_HEADER *os2 = (IMAGE_OS2_HEADER *)&nt;
if (os2->ne_magic != IMAGE_OS2_SIGNATURE) return STATUS_INVALID_IMAGE_PROTECT;
if (os2->ne_exetyp == 2) return STATUS_INVALID_IMAGE_WIN_16;
if (os2->ne_exetyp == 5) return STATUS_INVALID_IMAGE_PROTECT;
return STATUS_INVALID_IMAGE_NE_FORMAT;
}
switch (nt.opt.hdr32.Magic)
{
case IMAGE_NT_OPTIONAL_HDR32_MAGIC:
switch (nt.FileHeader.Machine)
{
case IMAGE_FILE_MACHINE_I386:
mapping->image.cpu = CPU_x86;
if (cpu_mask & (CPU_FLAG(CPU_x86) | CPU_FLAG(CPU_x86_64))) break;
return STATUS_INVALID_IMAGE_FORMAT;
case IMAGE_FILE_MACHINE_ARM:
case IMAGE_FILE_MACHINE_THUMB:
case IMAGE_FILE_MACHINE_ARMNT:
mapping->image.cpu = CPU_ARM;
if (cpu_mask & (CPU_FLAG(CPU_ARM) | CPU_FLAG(CPU_ARM64))) break;
return STATUS_INVALID_IMAGE_FORMAT;
case IMAGE_FILE_MACHINE_POWERPC:
mapping->image.cpu = CPU_POWERPC;
if (cpu_mask & CPU_FLAG(CPU_POWERPC)) break;
return STATUS_INVALID_IMAGE_FORMAT;
default:
return STATUS_INVALID_IMAGE_FORMAT;
}
clr_va = nt.opt.hdr32.DataDirectory[IMAGE_DIRECTORY_ENTRY_COM_DESCRIPTOR].VirtualAddress;
clr_size = nt.opt.hdr32.DataDirectory[IMAGE_DIRECTORY_ENTRY_COM_DESCRIPTOR].Size;
mapping->image.base = nt.opt.hdr32.ImageBase;
mapping->image.entry_point = nt.opt.hdr32.ImageBase + nt.opt.hdr32.AddressOfEntryPoint;
mapping->image.map_size = ROUND_SIZE( nt.opt.hdr32.SizeOfImage );
mapping->image.stack_size = nt.opt.hdr32.SizeOfStackReserve;
mapping->image.stack_commit = nt.opt.hdr32.SizeOfStackCommit;
mapping->image.subsystem = nt.opt.hdr32.Subsystem;
mapping->image.subsystem_low = nt.opt.hdr32.MinorSubsystemVersion;
mapping->image.subsystem_high = nt.opt.hdr32.MajorSubsystemVersion;
mapping->image.dll_charact = nt.opt.hdr32.DllCharacteristics;
mapping->image.contains_code = (nt.opt.hdr32.SizeOfCode ||
nt.opt.hdr32.AddressOfEntryPoint ||
nt.opt.hdr32.SectionAlignment & page_mask);
mapping->image.header_size = nt.opt.hdr32.SizeOfHeaders;
mapping->image.checksum = nt.opt.hdr32.CheckSum;
mapping->image.image_flags = 0;
if (nt.opt.hdr32.SectionAlignment & page_mask)
mapping->image.image_flags |= IMAGE_FLAGS_ImageMappedFlat;
if ((nt.opt.hdr32.DllCharacteristics & IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE) &&
mapping->image.contains_code && !(clr_va && clr_size))
mapping->image.image_flags |= IMAGE_FLAGS_ImageDynamicallyRelocated;
break;
case IMAGE_NT_OPTIONAL_HDR64_MAGIC:
if (!(cpu_mask & CPU_64BIT_MASK)) return STATUS_INVALID_IMAGE_WIN_64;
switch (nt.FileHeader.Machine)
{
case IMAGE_FILE_MACHINE_AMD64:
mapping->image.cpu = CPU_x86_64;
if (cpu_mask & (CPU_FLAG(CPU_x86) | CPU_FLAG(CPU_x86_64))) break;
return STATUS_INVALID_IMAGE_FORMAT;
case IMAGE_FILE_MACHINE_ARM64:
mapping->image.cpu = CPU_ARM64;
if (cpu_mask & (CPU_FLAG(CPU_ARM) | CPU_FLAG(CPU_ARM64))) break;
return STATUS_INVALID_IMAGE_FORMAT;
default:
return STATUS_INVALID_IMAGE_FORMAT;
}
clr_va = nt.opt.hdr64.DataDirectory[IMAGE_DIRECTORY_ENTRY_COM_DESCRIPTOR].VirtualAddress;
clr_size = nt.opt.hdr64.DataDirectory[IMAGE_DIRECTORY_ENTRY_COM_DESCRIPTOR].Size;
mapping->image.base = nt.opt.hdr64.ImageBase;
mapping->image.entry_point = nt.opt.hdr64.ImageBase + nt.opt.hdr64.AddressOfEntryPoint;
mapping->image.map_size = ROUND_SIZE( nt.opt.hdr64.SizeOfImage );
mapping->image.stack_size = nt.opt.hdr64.SizeOfStackReserve;
mapping->image.stack_commit = nt.opt.hdr64.SizeOfStackCommit;
mapping->image.subsystem = nt.opt.hdr64.Subsystem;
mapping->image.subsystem_low = nt.opt.hdr64.MinorSubsystemVersion;
mapping->image.subsystem_high = nt.opt.hdr64.MajorSubsystemVersion;
mapping->image.dll_charact = nt.opt.hdr64.DllCharacteristics;
mapping->image.contains_code = (nt.opt.hdr64.SizeOfCode ||
nt.opt.hdr64.AddressOfEntryPoint ||
nt.opt.hdr64.SectionAlignment & page_mask);
mapping->image.header_size = nt.opt.hdr64.SizeOfHeaders;
mapping->image.checksum = nt.opt.hdr64.CheckSum;
mapping->image.image_flags = 0;
if (nt.opt.hdr64.SectionAlignment & page_mask)
mapping->image.image_flags |= IMAGE_FLAGS_ImageMappedFlat;
if ((nt.opt.hdr64.DllCharacteristics & IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE) &&
mapping->image.contains_code && !(clr_va && clr_size))
mapping->image.image_flags |= IMAGE_FLAGS_ImageDynamicallyRelocated;
break;
default:
return STATUS_INVALID_IMAGE_FORMAT;
}
mapping->image.image_charact = nt.FileHeader.Characteristics;
mapping->image.machine = nt.FileHeader.Machine;
mapping->image.zerobits = 0; /* FIXME */
mapping->image.gp = 0; /* FIXME */
mapping->image.file_size = file_size;
mapping->image.loader_flags = clr_va && clr_size;
mapping->image.__pad = 0;
if (mz_size == sizeof(mz) && !memcmp( mz.buffer, builtin_signature, sizeof(builtin_signature) ))
mapping->image.image_flags |= IMAGE_FLAGS_WineBuiltin;
else if (mz_size == sizeof(mz) && !memcmp( mz.buffer, fakedll_signature, sizeof(fakedll_signature) ))
mapping->image.image_flags |= IMAGE_FLAGS_WineFakeDll;
/* load the section headers */
pos += sizeof(nt.Signature) + sizeof(nt.FileHeader) + nt.FileHeader.SizeOfOptionalHeader;
if (nt.FileHeader.NumberOfSections > ARRAY_SIZE( sec )) return STATUS_INVALID_IMAGE_FORMAT;
size = sizeof(*sec) * nt.FileHeader.NumberOfSections;
if (!mapping->size) mapping->size = mapping->image.map_size;
else if (mapping->size > mapping->image.map_size) return STATUS_SECTION_TOO_BIG;
if (pos + size > mapping->image.map_size) return STATUS_INVALID_FILE_FOR_SECTION;
if (pos + size > mapping->image.header_size) mapping->image.header_size = pos + size;
if (pread( unix_fd, sec, size, pos ) != size) return STATUS_INVALID_FILE_FOR_SECTION;
for (i = 0; i < nt.FileHeader.NumberOfSections && !mapping->image.contains_code; i++)
if (sec[i].Characteristics & IMAGE_SCN_MEM_EXECUTE) mapping->image.contains_code = 1;
if (load_clr_header( &clr, clr_va, clr_size, unix_fd, sec, nt.FileHeader.NumberOfSections ) &&
(clr.Flags & COMIMAGE_FLAGS_ILONLY))
{
mapping->image.image_flags |= IMAGE_FLAGS_ComPlusILOnly;
if (nt.opt.hdr32.Magic == IMAGE_NT_OPTIONAL_HDR32_MAGIC &&
!(clr.Flags & COMIMAGE_FLAGS_32BITREQUIRED))
{
mapping->image.image_flags |= IMAGE_FLAGS_ComPlusNativeReady;
if (cpu_mask & CPU_FLAG(CPU_x86_64)) mapping->image.cpu = CPU_x86_64;
else if (cpu_mask & CPU_FLAG(CPU_ARM64)) mapping->image.cpu = CPU_ARM64;
}
}
if (!build_shared_mapping( mapping, unix_fd, sec, nt.FileHeader.NumberOfSections ))
return STATUS_INVALID_FILE_FOR_SECTION;
return STATUS_SUCCESS;
}
static struct ranges *create_ranges(void)
{
struct ranges *ranges = alloc_object( &ranges_ops );
if (!ranges) return NULL;
ranges->count = 0;
ranges->max = 8;
if (!(ranges->ranges = mem_alloc( ranges->max * sizeof(*ranges->ranges) )))
{
release_object( ranges );
return NULL;
}
return ranges;
}
static unsigned int get_mapping_flags( obj_handle_t handle, unsigned int flags )
{
switch (flags & (SEC_IMAGE | SEC_RESERVE | SEC_COMMIT | SEC_FILE))
{
case SEC_IMAGE:
if (flags & (SEC_WRITECOMBINE | SEC_LARGE_PAGES)) break;
if (handle) return SEC_FILE | SEC_IMAGE;
set_error( STATUS_INVALID_FILE_FOR_SECTION );
return 0;
case SEC_COMMIT:
if (!handle) return flags;
/* fall through */
case SEC_RESERVE:
if (flags & SEC_LARGE_PAGES) break;
if (handle) return SEC_FILE | (flags & (SEC_NOCACHE | SEC_WRITECOMBINE));
return flags;
}
set_error( STATUS_INVALID_PARAMETER );
return 0;
}
static struct mapping *create_mapping( struct object *root, const struct unicode_str *name,
unsigned int attr, mem_size_t size, unsigned int flags,
obj_handle_t handle, unsigned int file_access,
const struct security_descriptor *sd )
{
struct mapping *mapping;
struct file *file;
struct fd *fd;
int unix_fd;
struct stat st;
if (!page_mask) page_mask = sysconf( _SC_PAGESIZE ) - 1;
if (!(mapping = create_named_object( root, &mapping_ops, name, attr, sd )))
return NULL;
if (get_error() == STATUS_OBJECT_NAME_EXISTS)
return mapping; /* Nothing else to do */
mapping->size = size;
mapping->fd = NULL;
mapping->shared = NULL;
mapping->committed = NULL;
if (!(mapping->flags = get_mapping_flags( handle, flags ))) goto error;
if (handle)
{
const unsigned int sharing = FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE;
unsigned int mapping_access = FILE_MAPPING_ACCESS;
if (!(file = get_file_obj( current->process, handle, file_access ))) goto error;
fd = get_obj_fd( (struct object *)file );
/* file sharing rules for mappings are different so we use magic the access rights */
if (flags & SEC_IMAGE) mapping_access |= FILE_MAPPING_IMAGE;
else if (file_access & FILE_WRITE_DATA) mapping_access |= FILE_MAPPING_WRITE;
if (!(mapping->fd = get_fd_object_for_mapping( fd, mapping_access, sharing )))
{
mapping->fd = dup_fd_object( fd, mapping_access, sharing, FILE_SYNCHRONOUS_IO_NONALERT );
if (mapping->fd) set_fd_user( mapping->fd, &mapping_fd_ops, NULL );
}
release_object( file );
release_object( fd );
if (!mapping->fd) goto error;
if ((unix_fd = get_unix_fd( mapping->fd )) == -1) goto error;
if (fstat( unix_fd, &st ) == -1)
{
file_set_error();
goto error;
}
if (flags & SEC_IMAGE)
{
unsigned int err = get_image_params( mapping, st.st_size, unix_fd );
if (!err) return mapping;
set_error( err );
goto error;
}
if (!mapping->size)
{
if (!(mapping->size = st.st_size))
2004-01-09 01:34:14 +01:00
{
set_error( STATUS_MAPPED_FILE_SIZE_ZERO );
2004-01-09 01:34:14 +01:00
goto error;
}
}
else if (st.st_size < mapping->size)
{
if (!(file_access & FILE_WRITE_DATA))
{
set_error( STATUS_SECTION_TOO_BIG );
goto error;
}
if (!grow_file( unix_fd, mapping->size )) goto error;
}
}
else /* Anonymous mapping (no associated file) */
{
if (!mapping->size)
{
set_error( STATUS_INVALID_PARAMETER );
goto error;
}
if ((flags & SEC_RESERVE) && !(mapping->committed = create_ranges())) goto error;
mapping->size = (mapping->size + page_mask) & ~((mem_size_t)page_mask);
if ((unix_fd = create_temp_file( mapping->size )) == -1) goto error;
if (!(mapping->fd = create_anonymous_fd( &mapping_fd_ops, unix_fd, &mapping->obj,
FILE_SYNCHRONOUS_IO_NONALERT ))) goto error;
allow_fd_caching( mapping->fd );
}
return mapping;
error:
release_object( mapping );
return NULL;
}
static struct mapping *get_mapping_obj( struct process *process, obj_handle_t handle, unsigned int access )
{
return (struct mapping *)get_handle_obj( process, handle, access, &mapping_ops );
}
/* open a new file for the file descriptor backing the mapping */
struct file *get_mapping_file( struct process *process, client_ptr_t base,
unsigned int access, unsigned int sharing )
{
struct memory_view *view = find_mapped_view( process, base );
if (!view || !view->fd) return NULL;
return create_file_for_fd_obj( view->fd, access, sharing );
}
/* get the image info for a SEC_IMAGE mapping */
const pe_image_info_t *get_mapping_image_info( struct process *process, client_ptr_t base )
{
struct memory_view *view = find_mapped_view( process, base );
if (!view || !(view->flags & SEC_IMAGE)) return NULL;
return &view->image;
}
static void mapping_dump( struct object *obj, int verbose )
{
struct mapping *mapping = (struct mapping *)obj;
assert( obj->ops == &mapping_ops );
fprintf( stderr, "Mapping size=%08x%08x flags=%08x fd=%p shared=%p\n",
(unsigned int)(mapping->size >> 32), (unsigned int)mapping->size,
mapping->flags, mapping->fd, mapping->shared );
}
static struct object_type *mapping_get_type( struct object *obj )
{
static const WCHAR name[] = {'S','e','c','t','i','o','n'};
static const struct unicode_str str = { name, sizeof(name) };
return get_object_type( &str );
}
static struct fd *mapping_get_fd( struct object *obj )
{
struct mapping *mapping = (struct mapping *)obj;
return (struct fd *)grab_object( mapping->fd );
}
static unsigned int mapping_map_access( struct object *obj, unsigned int access )
{
if (access & GENERIC_READ) access |= STANDARD_RIGHTS_READ | SECTION_QUERY | SECTION_MAP_READ;
if (access & GENERIC_WRITE) access |= STANDARD_RIGHTS_WRITE | SECTION_MAP_WRITE;
if (access & GENERIC_EXECUTE) access |= STANDARD_RIGHTS_EXECUTE | SECTION_MAP_EXECUTE;
if (access & GENERIC_ALL) access |= SECTION_ALL_ACCESS;
return access & ~(GENERIC_READ | GENERIC_WRITE | GENERIC_EXECUTE | GENERIC_ALL);
}
static void mapping_destroy( struct object *obj )
{
struct mapping *mapping = (struct mapping *)obj;
assert( obj->ops == &mapping_ops );
if (mapping->fd) release_object( mapping->fd );
if (mapping->committed) release_object( mapping->committed );
if (mapping->shared) release_object( mapping->shared );
}
static enum server_fd_type mapping_get_fd_type( struct fd *fd )
{
return FD_TYPE_FILE;
}
1999-11-13 20:49:05 +01:00
int get_page_size(void)
{
if (!page_mask) page_mask = sysconf( _SC_PAGESIZE ) - 1;
1999-11-13 20:49:05 +01:00
return page_mask + 1;
}
struct object *create_user_data_mapping( struct object *root, const struct unicode_str *name,
unsigned int attr, const struct security_descriptor *sd )
{
void *ptr;
struct mapping *mapping;
if (!(mapping = create_mapping( root, name, OBJ_OPENIF, sizeof(KSHARED_USER_DATA),
SEC_COMMIT, 0, FILE_READ_DATA | FILE_WRITE_DATA, NULL ))) return NULL;
ptr = mmap( NULL, mapping->size, PROT_WRITE, MAP_SHARED, get_unix_fd( mapping->fd ), 0 );
if (ptr != MAP_FAILED) user_shared_data = ptr;
return &mapping->obj;
}
/* create a file mapping */
DECL_HANDLER(create_mapping)
{
struct object *root;
struct mapping *mapping;
struct unicode_str name;
const struct security_descriptor *sd;
const struct object_attributes *objattr = get_req_object_attributes( &sd, &name, &root );
if (!objattr) return;
if ((mapping = create_mapping( root, &name, objattr->attributes, req->size, req->flags,
req->file_handle, req->file_access, sd )))
{
if (get_error() == STATUS_OBJECT_NAME_EXISTS)
reply->handle = alloc_handle( current->process, &mapping->obj, req->access, objattr->attributes );
else
reply->handle = alloc_handle_no_access_check( current->process, &mapping->obj,
req->access, objattr->attributes );
release_object( mapping );
}
if (root) release_object( root );
}
/* open a handle to a mapping */
DECL_HANDLER(open_mapping)
{
struct unicode_str name = get_req_unicode_str();
reply->handle = open_object( current->process, req->rootdir, req->access,
&mapping_ops, &name, req->attributes );
}
/* get a mapping information */
DECL_HANDLER(get_mapping_info)
{
struct mapping *mapping;
if (!(mapping = get_mapping_obj( current->process, req->handle, req->access ))) return;
reply->size = mapping->size;
reply->flags = mapping->flags;
if (mapping->flags & SEC_IMAGE)
set_reply_data( &mapping->image, min( sizeof(mapping->image), get_reply_max_size() ));
if (!(req->access & (SECTION_MAP_READ | SECTION_MAP_WRITE))) /* query only */
{
release_object( mapping );
return;
}
if (mapping->shared)
reply->shared_file = alloc_handle( current->process, mapping->shared->file,
GENERIC_READ|GENERIC_WRITE, 0 );
release_object( mapping );
}
/* add a memory view in the current process */
DECL_HANDLER(map_view)
{
struct mapping *mapping = NULL;
struct memory_view *view;
if (!req->size || (req->base & page_mask) || req->base + req->size < req->base) /* overflow */
{
set_error( STATUS_INVALID_PARAMETER );
return;
}
/* make sure we don't already have an overlapping view */
LIST_FOR_EACH_ENTRY( view, &current->process->views, struct memory_view, entry )
{
if (view->base + view->size <= req->base) continue;
if (view->base >= req->base + req->size) continue;
set_error( STATUS_INVALID_PARAMETER );
return;
}
if (!(mapping = get_mapping_obj( current->process, req->mapping, req->access ))) return;
if (mapping->flags & SEC_IMAGE)
{
if (req->start || req->size > mapping->image.map_size)
{
set_error( STATUS_INVALID_PARAMETER );
goto done;
}
}
else if (req->start >= mapping->size ||
req->start + req->size < req->start ||
req->start + req->size > ((mapping->size + page_mask) & ~(mem_size_t)page_mask))
{
set_error( STATUS_INVALID_PARAMETER );
goto done;
}
if ((view = mem_alloc( sizeof(*view) )))
{
view->base = req->base;
view->size = req->size;
view->start = req->start;
view->flags = mapping->flags;
view->fd = !is_fd_removable( mapping->fd ) ? (struct fd *)grab_object( mapping->fd ) : NULL;
view->committed = mapping->committed ? (struct ranges *)grab_object( mapping->committed ) : NULL;
view->shared = mapping->shared ? (struct shared_map *)grab_object( mapping->shared ) : NULL;
if (mapping->flags & SEC_IMAGE)
{
view->image = mapping->image;
if (view->base != mapping->image.base) set_error( STATUS_IMAGE_NOT_AT_BASE );
}
list_add_tail( &current->process->views, &view->entry );
}
done:
release_object( mapping );
}
/* unmap a memory view from the current process */
DECL_HANDLER(unmap_view)
{
struct memory_view *view = find_mapped_view( current->process, req->base );
if (view) free_memory_view( view );
}
/* get a range of committed pages in a file mapping */
DECL_HANDLER(get_mapping_committed_range)
{
struct memory_view *view = find_mapped_view( current->process, req->base );
if (view) reply->committed = find_committed_range( view, req->offset, &reply->size );
}
/* add a range to the committed pages in a file mapping */
DECL_HANDLER(add_mapping_committed_range)
{
struct memory_view *view = find_mapped_view( current->process, req->base );
if (view) add_committed_range( view, req->offset, req->offset + req->size );
}
/* check if two memory maps are for the same file */
DECL_HANDLER(is_same_mapping)
{
struct memory_view *view1 = find_mapped_view( current->process, req->base1 );
struct memory_view *view2 = find_mapped_view( current->process, req->base2 );
if (!view1 || !view2) return;
if (!view1->fd || !view2->fd ||
!(view1->flags & SEC_IMAGE) || !(view2->flags & SEC_IMAGE) ||
!is_same_file_fd( view1->fd, view2->fd ))
set_error( STATUS_NOT_SAME_DEVICE );
}