1
0
mirror of https://github.com/wine-mirror/wine synced 2024-07-05 17:28:47 +00:00
wine/loader/preloader.c

1527 lines
49 KiB
C
Raw Normal View History

/*
* Preloader for ld.so
*
* Copyright (C) 1995,96,97,98,99,2000,2001,2002 Free Software Foundation, Inc.
* Copyright (C) 2004 Mike McCormack for CodeWeavers
* Copyright (C) 2004 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
*/
/*
* Design notes
*
* The goal of this program is to be a workaround for exec-shield, as used
* by the Linux kernel distributed with Fedora Core and other distros.
*
* To do this, we implement our own shared object loader that reserves memory
* that is important to Wine, and then loads the main binary and its ELF
* interpreter.
*
* We will try to set up the stack and memory area so that the program that
* loads after us (eg. the wine binary) never knows we were here, except that
* areas of memory it needs are already magically reserved.
*
* The following memory areas are important to Wine:
* 0x00000000 - 0x00110000 the DOS area
* 0x80000000 - 0x81000000 the shared heap
* ??? - ??? the PE binary load address (usually starting at 0x00400000)
*
* If this program is used as the shared object loader, the only difference
* that the loaded programs should see is that this loader will be mapped
* into memory when it starts.
*/
/*
* References (things I consulted to understand how ELF loading works):
*
* glibc 2.3.2 elf/dl-load.c
* http://www.gnu.org/directory/glibc.html
*
* Linux 2.6.4 fs/binfmt_elf.c
* ftp://ftp.kernel.org/pub/linux/kernel/v2.6/linux-2.6.4.tar.bz2
*
* Userland exec, by <grugq@hcunix.net>
* http://cert.uni-stuttgart.de/archive/bugtraq/2004/01/msg00002.html
*
* The ELF specification:
* http://www.linuxbase.org/spec/booksets/LSB-Embedded/LSB-Embedded/book387.html
*/
#ifdef __linux__
#include "config.h"
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/mman.h>
#ifdef HAVE_SYS_SYSCALL_H
# include <sys/syscall.h>
#endif
#include <unistd.h>
#ifdef HAVE_ELF_H
# include <elf.h>
#endif
#ifdef HAVE_LINK_H
# include <link.h>
#endif
#ifdef HAVE_SYS_LINK_H
# include <sys/link.h>
#endif
#include "wine/asm.h"
#include "main.h"
loader: Set the visibility of symbols in preloader to "hidden". Today, the preloader is linked with -fPIE in spite of the fact that the preloader is a non-PIE statically linked binary. This is due to a limitation in tools/makedep that makes it difficult to specify CFLAGS for each individual object file's recipe. This can seemingly cause problems with some GOTPCREL(X) relocations inside the preloader. Since preloader does not link to the system library directly, there is no need for a Global Offset Table (GOT). However, a few extern (non-static) function symbols are declared, the use of which makes GCC emit instructions that references those symbols by indirection through GOT. The linker then tries to optimize such instructions to eliminate GOT references, which can fail due to various reasons. This stands in contradiction with Jinoh Kang's suggestion (in bug 55050) that "-fPIE is harmless even when applied to an object linked into non-PIE executables." The claim is theoretically true since position-independent code can in principle be relocated to any address (fixed or dynamic); however, it fails due to some peculiar practical issues, which is arguably a limitation in the linker's implementation (since it can be worked around with -Wl,--no-relax without issues). Fix this by eliminating GOT usage by setting the default visibility of non-static declarations to "hidden". Assuming GCC's medium code model (-mcmodel=medium; default code model for x86_64), this suppresses any unnecessary PLT or GOT relocations for defined symbols, and provides opportunity for GCC to optimize the code better. Fixes: 78ed343842dcd8ffb95c416420953e121959d40d Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55091
2023-06-20 14:01:25 +00:00
#pragma GCC visibility push(hidden)
/* ELF definitions */
#define ELF_PREFERRED_ADDRESS(loader, maplength, mapstartpref) (mapstartpref)
#define ELF_FIXED_ADDRESS(loader, mapstart) ((void) 0)
#define MAP_BASE_ADDR(l) 0
#ifndef MAP_COPY
#define MAP_COPY MAP_PRIVATE
#endif
#ifndef MAP_NORESERVE
#define MAP_NORESERVE 0
#endif
static struct wine_preload_info preload_info[] =
{
#if defined(__i386__) || defined(__arm__)
{ (void *)0x00000000, 0x00010000 }, /* low 64k */
{ (void *)0x00010000, 0x00100000 }, /* DOS area */
{ (void *)0x00110000, 0x67ef0000 }, /* low memory area */
{ (void *)0x7f000000, 0x03000000 }, /* top-down allocations + shared user data + virtual heap */
#else
{ (void *)0x000000010000, 0x00100000 }, /* DOS area */
{ (void *)0x000000110000, 0x67ef0000 }, /* low memory area */
{ (void *)0x00007f000000, 0x00ff0000 }, /* 32-bit top-down allocations + shared user data */
{ (void *)0x7ffffe000000, 0x01ff0000 }, /* top-down allocations + virtual heap */
#endif
{ 0, 0 }, /* PE exe range set with WINEPRELOADRESERVE */
{ 0, 0 } /* end of list */
};
/* debugging */
#undef DUMP_SEGMENTS
#undef DUMP_AUX_INFO
#undef DUMP_SYMS
#undef DUMP_MAPS
/* older systems may not define these */
#ifndef PT_TLS
#define PT_TLS 7
#endif
2004-06-28 20:28:05 +00:00
#ifndef AT_SYSINFO
#define AT_SYSINFO 32
#endif
#ifndef AT_SYSINFO_EHDR
#define AT_SYSINFO_EHDR 33
#endif
#ifndef DT_GNU_HASH
#define DT_GNU_HASH 0x6ffffef5
#endif
static size_t page_size, page_mask;
static char *preloader_start, *preloader_end;
struct wld_link_map {
ElfW(Addr) l_addr;
ElfW(Dyn) *l_ld;
ElfW(Phdr)*l_phdr;
ElfW(Addr) l_entry;
ElfW(Half) l_ldnum;
ElfW(Half) l_phnum;
ElfW(Addr) l_map_start, l_map_end;
ElfW(Addr) l_interp;
};
2013-02-13 15:26:39 +00:00
struct wld_auxv
{
ElfW(Addr) a_type;
union
{
ElfW(Addr) a_val;
} a_un;
};
/*
* The __bb_init_func is an empty function only called when file is
* compiled with gcc flags "-fprofile-arcs -ftest-coverage". This
* function is normally provided by libc's startup files, but since we
* build the preloader with "-nostartfiles -nodefaultlibs", we have to
* provide our own (empty) version, otherwise linker fails.
*/
2006-02-06 12:13:28 +00:00
void __bb_init_func(void) { return; }
#ifdef __i386__
/* data for setting up the glibc-style thread-local storage in %gs */
static int thread_data[256];
struct
{
/* this is the kernel modify_ldt struct */
unsigned int entry_number;
unsigned long base_addr;
unsigned int limit;
unsigned int seg_32bit : 1;
unsigned int contents : 2;
unsigned int read_exec_only : 1;
unsigned int limit_in_pages : 1;
unsigned int seg_not_present : 1;
2008-04-13 23:03:24 +00:00
unsigned int usable : 1;
unsigned int garbage : 25;
} thread_ldt = { -1, (unsigned long)thread_data, 0xfffff, 1, 0, 0, 1, 0, 1, 0 };
/*
* The _start function is the entry and exit point of this program
*
* It calls wld_start, passing a pointer to the args it receives
* then jumps to the address wld_start returns.
*/
2009-10-07 10:24:53 +00:00
void _start(void);
extern char _end[];
__ASM_GLOBAL_FUNC(_start,
__ASM_CFI("\t.cfi_undefined %eip\n")
"\tmovl $243,%eax\n" /* SYS_set_thread_area */
"\tmovl $thread_ldt,%ebx\n"
"\tint $0x80\n" /* allocate gs segment */
"\torl %eax,%eax\n"
"\tjl 1f\n"
"\tmovl thread_ldt,%eax\n" /* thread_ldt.entry_number */
"\tshl $3,%eax\n"
"\torl $3,%eax\n"
"\tmov %ax,%gs\n"
"\tmov %ax,%fs\n" /* set %fs too so libwine can retrieve it later on */
"1:\tmovl %esp,%eax\n"
"\tleal -136(%esp),%esp\n" /* allocate some space for extra aux values */
"\tpushl %eax\n" /* orig stack pointer */
"\tpushl %esp\n" /* ptr to orig stack pointer */
"\tcall wld_start\n"
"\tpopl %ecx\n" /* remove ptr to stack pointer */
"\tpopl %esp\n" /* new stack pointer */
"\tpush %eax\n" /* ELF interpreter entry point */
"\txor %eax,%eax\n"
"\txor %ecx,%ecx\n"
"\txor %edx,%edx\n"
"\tmov %ax,%gs\n" /* clear %gs again */
"\tret\n")
/* wrappers for Linux system calls */
#define SYSCALL_RET(ret) (((ret) < 0 && (ret) > -4096) ? -1 : (ret))
static inline __attribute__((noreturn)) void wld_exit( int code )
{
for (;;) /* avoid warning */
__asm__ __volatile__( "pushl %%ebx; movl %1,%%ebx; int $0x80; popl %%ebx"
: : "a" (1 /* SYS_exit */), "r" (code) );
}
static inline int wld_open( const char *name, int flags )
{
int ret;
__asm__ __volatile__( "pushl %%ebx; movl %2,%%ebx; int $0x80; popl %%ebx"
: "=a" (ret) : "0" (5 /* SYS_open */), "r" (name), "c" (flags) );
return SYSCALL_RET(ret);
}
static inline int wld_close( int fd )
{
int ret;
__asm__ __volatile__( "pushl %%ebx; movl %2,%%ebx; int $0x80; popl %%ebx"
: "=a" (ret) : "0" (6 /* SYS_close */), "r" (fd) );
return SYSCALL_RET(ret);
}
static inline ssize_t wld_read( int fd, void *buffer, size_t len )
{
int ret;
__asm__ __volatile__( "pushl %%ebx; movl %2,%%ebx; int $0x80; popl %%ebx"
: "=a" (ret)
: "0" (3 /* SYS_read */), "r" (fd), "c" (buffer), "d" (len)
: "memory" );
return SYSCALL_RET(ret);
}
static inline ssize_t wld_write( int fd, const void *buffer, size_t len )
{
int ret;
__asm__ __volatile__( "pushl %%ebx; movl %2,%%ebx; int $0x80; popl %%ebx"
: "=a" (ret) : "0" (4 /* SYS_write */), "r" (fd), "c" (buffer), "d" (len) );
return SYSCALL_RET(ret);
}
static inline int wld_mprotect( const void *addr, size_t len, int prot )
{
int ret;
__asm__ __volatile__( "pushl %%ebx; movl %2,%%ebx; int $0x80; popl %%ebx"
: "=a" (ret) : "0" (125 /* SYS_mprotect */), "r" (addr), "c" (len), "d" (prot) );
return SYSCALL_RET(ret);
}
void *wld_mmap( void *start, size_t len, int prot, int flags, int fd, unsigned int offset );
__ASM_GLOBAL_FUNC(wld_mmap,
"\tpushl %ebp\n"
__ASM_CFI(".cfi_adjust_cfa_offset 4\n\t")
"\tpushl %ebx\n"
__ASM_CFI(".cfi_adjust_cfa_offset 4\n\t")
"\tpushl %esi\n"
__ASM_CFI(".cfi_adjust_cfa_offset 4\n\t")
"\tpushl %edi\n"
__ASM_CFI(".cfi_adjust_cfa_offset 4\n\t")
"\tmovl $192,%eax\n" /* SYS_mmap2 */
"\tmovl 20(%esp),%ebx\n" /* start */
"\tmovl 24(%esp),%ecx\n" /* len */
"\tmovl 28(%esp),%edx\n" /* prot */
"\tmovl 32(%esp),%esi\n" /* flags */
"\tmovl 36(%esp),%edi\n" /* fd */
"\tmovl 40(%esp),%ebp\n" /* offset */
"\tshrl $12,%ebp\n"
"\tint $0x80\n"
"\tcmpl $-4096,%eax\n"
"\tjbe 2f\n"
"\tcmpl $-38,%eax\n" /* ENOSYS */
"\tjne 1f\n"
"\tmovl $90,%eax\n" /* SYS_mmap */
"\tleal 20(%esp),%ebx\n"
"\tint $0x80\n"
"\tcmpl $-4096,%eax\n"
"\tjbe 2f\n"
"1:\tmovl $-1,%eax\n"
"2:\tpopl %edi\n"
__ASM_CFI(".cfi_adjust_cfa_offset -4\n\t")
"\tpopl %esi\n"
__ASM_CFI(".cfi_adjust_cfa_offset -4\n\t")
"\tpopl %ebx\n"
__ASM_CFI(".cfi_adjust_cfa_offset -4\n\t")
"\tpopl %ebp\n"
__ASM_CFI(".cfi_adjust_cfa_offset -4\n\t")
"\tret\n" )
static inline int wld_prctl( int code, long arg )
{
int ret;
__asm__ __volatile__( "pushl %%ebx; movl %2,%%ebx; int $0x80; popl %%ebx"
: "=a" (ret) : "0" (172 /* SYS_prctl */), "r" (code), "c" (arg) );
return SYSCALL_RET(ret);
}
#elif defined(__x86_64__)
void *thread_data[256];
/*
* The _start function is the entry and exit point of this program
*
* It calls wld_start, passing a pointer to the args it receives
* then jumps to the address wld_start returns.
*/
void _start(void);
extern char _end[];
__ASM_GLOBAL_FUNC(_start,
__ASM_CFI(".cfi_undefined %rip\n\t")
"movq %rsp,%rax\n\t"
"leaq -144(%rsp),%rsp\n\t" /* allocate some space for extra aux values */
"movq %rax,(%rsp)\n\t" /* orig stack pointer */
"leaq thread_data(%rip),%rsi\n\t"
"movq $0x1002,%rdi\n\t" /* ARCH_SET_FS */
"movq $158,%rax\n\t" /* SYS_arch_prctl */
"syscall\n\t"
"movq %rsp,%rdi\n\t" /* ptr to orig stack pointer */
"call wld_start\n\t"
"movq (%rsp),%rsp\n\t" /* new stack pointer */
"pushq %rax\n\t" /* ELF interpreter entry point */
"xorq %rax,%rax\n\t"
"xorq %rcx,%rcx\n\t"
"xorq %rdx,%rdx\n\t"
"xorq %rsi,%rsi\n\t"
"xorq %rdi,%rdi\n\t"
"xorq %r8,%r8\n\t"
"xorq %r9,%r9\n\t"
"xorq %r10,%r10\n\t"
"xorq %r11,%r11\n\t"
"ret")
#define SYSCALL_FUNC( name, nr ) \
__ASM_GLOBAL_FUNC( name, \
"movq $" #nr ",%rax\n\t" \
"movq %rcx,%r10\n\t" \
"syscall\n\t" \
"leaq 4096(%rax),%rcx\n\t" \
"movq $-1,%rdx\n\t" \
"cmp $4096,%rcx\n\t" \
"cmovb %rdx,%rax\n\t" \
"ret" )
#define SYSCALL_NOERR( name, nr ) \
__ASM_GLOBAL_FUNC( name, \
"movq $" #nr ",%rax\n\t" \
"syscall\n\t" \
"ret" )
void wld_exit( int code ) __attribute__((noreturn));
SYSCALL_NOERR( wld_exit, 60 /* SYS_exit */ );
ssize_t wld_read( int fd, void *buffer, size_t len );
SYSCALL_FUNC( wld_read, 0 /* SYS_read */ );
ssize_t wld_write( int fd, const void *buffer, size_t len );
SYSCALL_FUNC( wld_write, 1 /* SYS_write */ );
int wld_open( const char *name, int flags );
SYSCALL_FUNC( wld_open, 2 /* SYS_open */ );
int wld_close( int fd );
SYSCALL_FUNC( wld_close, 3 /* SYS_close */ );
void *wld_mmap( void *start, size_t len, int prot, int flags, int fd, off_t offset );
SYSCALL_FUNC( wld_mmap, 9 /* SYS_mmap */ );
int wld_mprotect( const void *addr, size_t len, int prot );
SYSCALL_FUNC( wld_mprotect, 10 /* SYS_mprotect */ );
int wld_prctl( int code, long arg );
SYSCALL_FUNC( wld_prctl, 157 /* SYS_prctl */ );
uid_t wld_getuid(void);
SYSCALL_NOERR( wld_getuid, 102 /* SYS_getuid */ );
gid_t wld_getgid(void);
SYSCALL_NOERR( wld_getgid, 104 /* SYS_getgid */ );
uid_t wld_geteuid(void);
SYSCALL_NOERR( wld_geteuid, 107 /* SYS_geteuid */ );
gid_t wld_getegid(void);
SYSCALL_NOERR( wld_getegid, 108 /* SYS_getegid */ );
#elif defined(__aarch64__)
void *thread_data[256];
/*
* The _start function is the entry and exit point of this program
*
* It calls wld_start, passing a pointer to the args it receives
* then jumps to the address wld_start returns.
*/
void _start(void);
extern char _end[];
__ASM_GLOBAL_FUNC(_start,
"mov x0, SP\n\t"
"sub SP, SP, #144\n\t" /* allocate some space for extra aux values */
"str x0, [SP]\n\t" /* orig stack pointer */
"adrp x0, thread_data\n\t"
"add x0, x0, :lo12:thread_data\n\t"
"msr tpidr_el0, x0\n\t"
"mov x0, SP\n\t" /* ptr to orig stack pointer */
"bl wld_start\n\t"
"ldr x1, [SP]\n\t" /* new stack pointer */
"mov SP, x1\n\t"
"mov x30, x0\n\t"
"mov x0, #0\n\t"
"mov x1, #0\n\t"
"mov x2, #0\n\t"
"mov x3, #0\n\t"
"mov x4, #0\n\t"
"mov x5, #0\n\t"
"mov x6, #0\n\t"
"mov x7, #0\n\t"
"mov x8, #0\n\t"
"mov x9, #0\n\t"
"mov x10, #0\n\t"
"mov x11, #0\n\t"
"mov x12, #0\n\t"
"mov x13, #0\n\t"
"mov x14, #0\n\t"
"mov x15, #0\n\t"
"mov x16, #0\n\t"
"mov x17, #0\n\t"
"mov x18, #0\n\t"
"ret")
#define SYSCALL_FUNC( name, nr ) \
__ASM_GLOBAL_FUNC( name, \
"stp x8, x9, [SP, #-16]!\n\t" \
"mov x8, #" #nr "\n\t" \
"svc #0\n\t" \
"ldp x8, x9, [SP], #16\n\t" \
"cmn x0, #1, lsl#12\n\t" \
"cinv x0, x0, hi\n\t" \
"b.hi 1f\n\t" \
"ret\n\t" \
"1: mov x0, #-1\n\t" \
"ret" )
#define SYSCALL_NOERR( name, nr ) \
__ASM_GLOBAL_FUNC( name, \
"stp x8, x9, [SP, #-16]!\n\t" \
"mov x8, #" #nr "\n\t" \
"svc #0\n\t" \
"ldp x8, x9, [SP], #16\n\t" \
"ret" )
void wld_exit( int code ) __attribute__((noreturn));
SYSCALL_NOERR( wld_exit, 93 /* SYS_exit */ );
ssize_t wld_read( int fd, void *buffer, size_t len );
SYSCALL_FUNC( wld_read, 63 /* SYS_read */ );
ssize_t wld_write( int fd, const void *buffer, size_t len );
SYSCALL_FUNC( wld_write, 64 /* SYS_write */ );
int wld_openat( int dirfd, const char *name, int flags );
SYSCALL_FUNC( wld_openat, 56 /* SYS_openat */ );
int wld_open( const char *name, int flags )
{
return wld_openat(-100 /* AT_FDCWD */, name, flags);
}
int wld_close( int fd );
SYSCALL_FUNC( wld_close, 57 /* SYS_close */ );
void *wld_mmap( void *start, size_t len, int prot, int flags, int fd, off_t offset );
SYSCALL_FUNC( wld_mmap, 222 /* SYS_mmap */ );
int wld_mprotect( const void *addr, size_t len, int prot );
SYSCALL_FUNC( wld_mprotect, 226 /* SYS_mprotect */ );
int wld_prctl( int code, long arg );
SYSCALL_FUNC( wld_prctl, 167 /* SYS_prctl */ );
uid_t wld_getuid(void);
SYSCALL_NOERR( wld_getuid, 174 /* SYS_getuid */ );
gid_t wld_getgid(void);
SYSCALL_NOERR( wld_getgid, 176 /* SYS_getgid */ );
uid_t wld_geteuid(void);
SYSCALL_NOERR( wld_geteuid, 175 /* SYS_geteuid */ );
gid_t wld_getegid(void);
SYSCALL_NOERR( wld_getegid, 177 /* SYS_getegid */ );
#elif defined(__arm__)
void *thread_data[256];
/*
* The _start function is the entry and exit point of this program
*
* It calls wld_start, passing a pointer to the args it receives
* then jumps to the address wld_start returns.
*/
void _start(void);
extern char _end[];
__ASM_GLOBAL_FUNC(_start,
ntdll: Add ARM EHABI unwind instructions in assembly functions. On most ELF platforms on ARM, ARM EHABI is the unwind info format normally used, instead of DWARF like on most other platforms. Currently, when unwinding through ELF objects with libunwind, the libraries don't have any .eh_frame section mapped at runtime (since DWARF isn't used for unwinding). Instead, what happens is that libunwind ends up loading .debug_frame from the libraries on disk instead. Therefore, currently, ELF unwinding relies on the .so files not being stripped. This patch adds the necessary EHABI unwinding instructions in the assembly functions that currently have DWARF CFI instructions. EHABI isn't signaled via any specific preprocessor macro, but is signaled by the absence of other unwind mechanisms (such as __ARM_DWARF_EH__ and __SEH__, or maybe SjLj). Mark the asm functions in the preloaders as .cantunwind, to avoid undefined references to __aeabi_unwind_cpp_pr* functions. Also mark other assembly functions as .cantunwind; for signal_exit_thread this is essential if the function is marked with .fnstart/.fnend - otherwise exiting threads does hang. (pthread_exit internally calls _Unwind_ForcedUnwind, which would hang if signal_exit_thread had .fnstart without any matching unwind info). This would, in principle, allow unwinding through these functions with libunwind, for versions of libunwind that can parse the EHABI unwind info - see e.g. https://github.com/libunwind/libunwind/commit/4d779f55c0d0a3814453517e54cd0f7bed69ca74. (This commit isn't yet in any current release AFAIK). Unwinding with EHABI via libunwind would require a few tweaks to the libunwind interface usage in unix/signal_arm.c though, since e.g. the unw_get_proc_info call fails if there's no .eh_frame or .debug_frame available. Signed-off-by: Martin Storsjö <martin@martin.st>
2022-10-20 10:44:24 +00:00
__ASM_EHABI(".cantunwind\n\t")
"mov r0, sp\n\t"
"sub sp, sp, #144\n\t" /* allocate some space for extra aux values */
"str r0, [sp]\n\t" /* orig stack pointer */
"ldr r0, =thread_data\n\t"
"movw r7, #0x0005\n\t" /* __ARM_NR_set_tls */
"movt r7, #0xf\n\t" /* __ARM_NR_set_tls */
"svc #0\n\t"
"mov r0, sp\n\t" /* ptr to orig stack pointer */
"bl wld_start\n\t"
"ldr r1, [sp]\n\t" /* new stack pointer */
"mov sp, r1\n\t"
"mov lr, r0\n\t"
"mov r0, #0\n\t"
"mov r1, #0\n\t"
"mov r2, #0\n\t"
"mov r3, #0\n\t"
"mov r12, #0\n\t"
"bx lr\n\t"
".ltorg\n\t")
#define SYSCALL_FUNC( name, nr ) \
__ASM_GLOBAL_FUNC( name, \
ntdll: Add ARM EHABI unwind instructions in assembly functions. On most ELF platforms on ARM, ARM EHABI is the unwind info format normally used, instead of DWARF like on most other platforms. Currently, when unwinding through ELF objects with libunwind, the libraries don't have any .eh_frame section mapped at runtime (since DWARF isn't used for unwinding). Instead, what happens is that libunwind ends up loading .debug_frame from the libraries on disk instead. Therefore, currently, ELF unwinding relies on the .so files not being stripped. This patch adds the necessary EHABI unwinding instructions in the assembly functions that currently have DWARF CFI instructions. EHABI isn't signaled via any specific preprocessor macro, but is signaled by the absence of other unwind mechanisms (such as __ARM_DWARF_EH__ and __SEH__, or maybe SjLj). Mark the asm functions in the preloaders as .cantunwind, to avoid undefined references to __aeabi_unwind_cpp_pr* functions. Also mark other assembly functions as .cantunwind; for signal_exit_thread this is essential if the function is marked with .fnstart/.fnend - otherwise exiting threads does hang. (pthread_exit internally calls _Unwind_ForcedUnwind, which would hang if signal_exit_thread had .fnstart without any matching unwind info). This would, in principle, allow unwinding through these functions with libunwind, for versions of libunwind that can parse the EHABI unwind info - see e.g. https://github.com/libunwind/libunwind/commit/4d779f55c0d0a3814453517e54cd0f7bed69ca74. (This commit isn't yet in any current release AFAIK). Unwinding with EHABI via libunwind would require a few tweaks to the libunwind interface usage in unix/signal_arm.c though, since e.g. the unw_get_proc_info call fails if there's no .eh_frame or .debug_frame available. Signed-off-by: Martin Storsjö <martin@martin.st>
2022-10-20 10:44:24 +00:00
__ASM_EHABI(".cantunwind\n\t") \
"push {r4-r5,r7,lr}\n\t" \
"ldr r4, [sp, #16]\n\t" \
"ldr r5, [sp, #20]\n\t" \
"mov r7, #" #nr "\n\t" \
"svc #0\n\t" \
"cmn r0, #4096\n\t" \
"it hi\n\t" \
"movhi r0, #-1\n\t" \
"pop {r4-r5,r7,pc}\n\t" )
#define SYSCALL_NOERR( name, nr ) \
__ASM_GLOBAL_FUNC( name, \
ntdll: Add ARM EHABI unwind instructions in assembly functions. On most ELF platforms on ARM, ARM EHABI is the unwind info format normally used, instead of DWARF like on most other platforms. Currently, when unwinding through ELF objects with libunwind, the libraries don't have any .eh_frame section mapped at runtime (since DWARF isn't used for unwinding). Instead, what happens is that libunwind ends up loading .debug_frame from the libraries on disk instead. Therefore, currently, ELF unwinding relies on the .so files not being stripped. This patch adds the necessary EHABI unwinding instructions in the assembly functions that currently have DWARF CFI instructions. EHABI isn't signaled via any specific preprocessor macro, but is signaled by the absence of other unwind mechanisms (such as __ARM_DWARF_EH__ and __SEH__, or maybe SjLj). Mark the asm functions in the preloaders as .cantunwind, to avoid undefined references to __aeabi_unwind_cpp_pr* functions. Also mark other assembly functions as .cantunwind; for signal_exit_thread this is essential if the function is marked with .fnstart/.fnend - otherwise exiting threads does hang. (pthread_exit internally calls _Unwind_ForcedUnwind, which would hang if signal_exit_thread had .fnstart without any matching unwind info). This would, in principle, allow unwinding through these functions with libunwind, for versions of libunwind that can parse the EHABI unwind info - see e.g. https://github.com/libunwind/libunwind/commit/4d779f55c0d0a3814453517e54cd0f7bed69ca74. (This commit isn't yet in any current release AFAIK). Unwinding with EHABI via libunwind would require a few tweaks to the libunwind interface usage in unix/signal_arm.c though, since e.g. the unw_get_proc_info call fails if there's no .eh_frame or .debug_frame available. Signed-off-by: Martin Storsjö <martin@martin.st>
2022-10-20 10:44:24 +00:00
__ASM_EHABI(".cantunwind\n\t") \
"push {r7,lr}\n\t" \
"mov r7, #" #nr "\n\t" \
"svc #0\n\t" \
"pop {r7,pc}\n\t" )
void wld_exit( int code ) __attribute__((noreturn));
SYSCALL_NOERR( wld_exit, 1 /* SYS_exit */ );
ssize_t wld_read( int fd, void *buffer, size_t len );
SYSCALL_FUNC( wld_read, 3 /* SYS_read */ );
ssize_t wld_write( int fd, const void *buffer, size_t len );
SYSCALL_FUNC( wld_write, 4 /* SYS_write */ );
int wld_openat( int dirfd, const char *name, int flags );
SYSCALL_FUNC( wld_openat, 322 /* SYS_openat */ );
int wld_open( const char *name, int flags )
{
return wld_openat(-100 /* AT_FDCWD */, name, flags);
}
int wld_close( int fd );
SYSCALL_FUNC( wld_close, 6 /* SYS_close */ );
void *wld_mmap2( void *start, size_t len, int prot, int flags, int fd, int offset );
SYSCALL_FUNC( wld_mmap2, 192 /* SYS_mmap2 */ );
void *wld_mmap( void *start, size_t len, int prot, int flags, int fd, off_t offset )
{
return wld_mmap2(start, len, prot, flags, fd, offset >> 12);
}
int wld_mprotect( const void *addr, size_t len, int prot );
SYSCALL_FUNC( wld_mprotect, 125 /* SYS_mprotect */ );
int wld_prctl( int code, long arg );
SYSCALL_FUNC( wld_prctl, 172 /* SYS_prctl */ );
uid_t wld_getuid(void);
SYSCALL_NOERR( wld_getuid, 24 /* SYS_getuid */ );
gid_t wld_getgid(void);
SYSCALL_NOERR( wld_getgid, 47 /* SYS_getgid */ );
uid_t wld_geteuid(void);
SYSCALL_NOERR( wld_geteuid, 49 /* SYS_geteuid */ );
gid_t wld_getegid(void);
SYSCALL_NOERR( wld_getegid, 50 /* SYS_getegid */ );
unsigned long long __aeabi_uidivmod(unsigned int num, unsigned int den)
{
unsigned int bit = 1;
unsigned int quota = 0;
if (!den)
wld_exit(1);
while (den < num && !(den & 0x80000000)) {
den <<= 1;
bit <<= 1;
}
do {
if (den <= num) {
quota |= bit;
num -= den;
}
bit >>= 1;
den >>= 1;
} while (bit);
return ((unsigned long long)num << 32) | quota;
}
#else
#error preloader not implemented for this CPU
#endif
/* replacement for libc functions */
static int wld_strcmp( const char *str1, const char *str2 )
{
while (*str1 && (*str1 == *str2)) { str1++; str2++; }
return *str1 - *str2;
}
static int wld_strncmp( const char *str1, const char *str2, size_t len )
{
if (len <= 0) return 0;
while ((--len > 0) && *str1 && (*str1 == *str2)) { str1++; str2++; }
return *str1 - *str2;
}
static inline void *wld_memset( void *dest, int val, size_t len )
{
char *dst = dest;
while (len--) *dst++ = val;
return dest;
}
/*
* wld_printf - just the basics
*
* %x prints a hex number
* %s prints a string
* %p prints a pointer
*/
static int wld_vsprintf(char *buffer, const char *fmt, va_list args )
{
static const char hex_chars[16] = "0123456789abcdef";
const char *p = fmt;
char *str = buffer;
int i;
while( *p )
{
if( *p == '%' )
{
p++;
if( *p == 'x' )
{
unsigned int x = va_arg( args, unsigned int );
for (i = 2*sizeof(x) - 1; i >= 0; i--)
*str++ = hex_chars[(x>>(i*4))&0xf];
}
else if (p[0] == 'l' && p[1] == 'x')
{
unsigned long x = va_arg( args, unsigned long );
for (i = 2*sizeof(x) - 1; i >= 0; i--)
*str++ = hex_chars[(x>>(i*4))&0xf];
p++;
}
else if( *p == 'p' )
{
unsigned long x = (unsigned long)va_arg( args, void * );
for (i = 2*sizeof(x) - 1; i >= 0; i--)
*str++ = hex_chars[(x>>(i*4))&0xf];
}
else if( *p == 's' )
{
char *s = va_arg( args, char * );
while(*s)
*str++ = *s++;
}
else if( *p == 0 )
break;
p++;
}
*str++ = *p++;
}
*str = 0;
return str - buffer;
}
static __attribute__((format(printf,1,2))) void wld_printf(const char *fmt, ... )
{
va_list args;
char buffer[256];
int len;
va_start( args, fmt );
len = wld_vsprintf(buffer, fmt, args );
va_end( args );
wld_write(2, buffer, len);
}
static __attribute__((noreturn,format(printf,1,2))) void fatal_error(const char *fmt, ... )
{
va_list args;
char buffer[256];
int len;
va_start( args, fmt );
len = wld_vsprintf(buffer, fmt, args );
va_end( args );
wld_write(2, buffer, len);
wld_exit(1);
}
/*
* The __stack_chk_* symbols are only used when file is compiled with gcc flags
* "-fstack-protector". This function is normally provided by libc's startup
* files, but since we build the preloader with "-nostartfiles -nodefaultlibs",
* we have to provide our own version to keep the linker happy.
*/
unsigned long __stack_chk_guard = 0;
void __attribute__((noreturn)) __stack_chk_fail(void)
{
static const char message[] = "preloader: stack overrun detected, crashing\n";
/* Avoid using non-syscall functions that can re-enter this function */
wld_write(2, message, sizeof(message) - 1);
/* Deliberate induce crash and possibly dump core */
*(volatile char *)0;
/* Last resort if the zero page turns out to be actually readable */
wld_exit(1);
}
void __attribute__((noreturn)) __stack_chk_fail_local(void)
{
__stack_chk_fail();
}
#ifdef DUMP_AUX_INFO
/*
* Dump interesting bits of the ELF auxv_t structure that is passed
* as the 4th parameter to the _start function
*/
2013-02-13 15:26:39 +00:00
static void dump_auxiliary( struct wld_auxv *av )
{
#define NAME(at) { at, #at }
static const struct { int val; const char *name; } names[] =
{
NAME(AT_BASE),
NAME(AT_CLKTCK),
NAME(AT_EGID),
NAME(AT_ENTRY),
NAME(AT_EUID),
NAME(AT_FLAGS),
NAME(AT_GID),
NAME(AT_HWCAP),
NAME(AT_PAGESZ),
NAME(AT_PHDR),
NAME(AT_PHENT),
NAME(AT_PHNUM),
NAME(AT_PLATFORM),
NAME(AT_SYSINFO),
NAME(AT_SYSINFO_EHDR),
NAME(AT_UID),
{ 0, NULL }
};
#undef NAME
int i;
for ( ; av->a_type != AT_NULL; av++)
{
for (i = 0; names[i].name; i++) if (names[i].val == av->a_type) break;
if (names[i].name) wld_printf("%s = %lx\n", names[i].name, (unsigned long)av->a_un.a_val);
else wld_printf( "%lx = %lx\n", (unsigned long)av->a_type, (unsigned long)av->a_un.a_val );
}
}
#endif
/*
* set_auxiliary_values
*
* Set the new auxiliary values
*/
2013-02-13 15:26:39 +00:00
static void set_auxiliary_values( struct wld_auxv *av, const struct wld_auxv *new_av,
const struct wld_auxv *delete_av, void **stack )
{
int i, j, av_count = 0, new_count = 0, delete_count = 0;
char *src, *dst;
/* count how many aux values we have already */
while (av[av_count].a_type != AT_NULL) av_count++;
/* delete unwanted values */
for (j = 0; delete_av[j].a_type != AT_NULL; j++)
{
for (i = 0; i < av_count; i++) if (av[i].a_type == delete_av[j].a_type)
{
av[i].a_type = av[av_count-1].a_type;
av[i].a_un.a_val = av[av_count-1].a_un.a_val;
av[--av_count].a_type = AT_NULL;
delete_count++;
break;
}
}
/* count how many values we have in new_av that aren't in av */
for (j = 0; new_av[j].a_type != AT_NULL; j++)
2004-06-01 19:47:11 +00:00
{
for (i = 0; i < av_count; i++) if (av[i].a_type == new_av[j].a_type) break;
if (i == av_count) new_count++;
2004-06-01 19:47:11 +00:00
}
src = (char *)*stack;
dst = src - (new_count - delete_count) * sizeof(*av);
dst = (char *)((unsigned long)dst & ~15);
if (dst < src) /* need to make room for the extra values */
{
int len = (char *)(av + av_count + 1) - src;
for (i = 0; i < len; i++) dst[i] = src[i];
}
else if (dst > src) /* get rid of unused values */
{
int len = (char *)(av + av_count + 1) - src;
for (i = len - 1; i >= 0; i--) dst[i] = src[i];
}
*stack = dst;
2013-02-13 15:26:39 +00:00
av = (struct wld_auxv *)((char *)av + (dst - src));
/* now set the values */
for (j = 0; new_av[j].a_type != AT_NULL; j++)
{
for (i = 0; i < av_count; i++) if (av[i].a_type == new_av[j].a_type) break;
if (i < av_count) av[i].a_un.a_val = new_av[j].a_un.a_val;
else
{
av[av_count].a_type = new_av[j].a_type;
av[av_count].a_un.a_val = new_av[j].a_un.a_val;
av_count++;
}
}
#ifdef DUMP_AUX_INFO
wld_printf("New auxiliary info:\n");
dump_auxiliary( av );
#endif
}
/*
* get_auxiliary
*
* Get a field of the auxiliary structure
*/
static ElfW(Addr) get_auxiliary( struct wld_auxv *av, int type, ElfW(Addr) def_val )
{
for ( ; av->a_type != AT_NULL; av++)
if( av->a_type == type ) return av->a_un.a_val;
return def_val;
}
/*
* map_so_lib
*
* modelled after _dl_map_object_from_fd() from glibc-2.3.1/elf/dl-load.c
*
* This function maps the segments from an ELF object, and optionally
* stores information about the mapping into the auxv_t structure.
*/
static void map_so_lib( const char *name, struct wld_link_map *l)
{
int fd;
unsigned char buf[0x800];
ElfW(Ehdr) *header = (ElfW(Ehdr)*)buf;
ElfW(Phdr) *phdr, *ph;
/* Scan the program header table, collecting its load commands. */
struct loadcmd
{
ElfW(Addr) mapstart, mapend, dataend, allocend;
off_t mapoff;
int prot;
} loadcmds[16], *c;
size_t nloadcmds = 0, maplength;
fd = wld_open( name, O_RDONLY );
if (fd == -1) fatal_error("%s: could not open\n", name );
if (wld_read( fd, buf, sizeof(buf) ) != sizeof(buf))
fatal_error("%s: failed to read ELF header\n", name);
phdr = (void*) (((unsigned char*)buf) + header->e_phoff);
if( ( header->e_ident[0] != 0x7f ) ||
( header->e_ident[1] != 'E' ) ||
( header->e_ident[2] != 'L' ) ||
( header->e_ident[3] != 'F' ) )
fatal_error( "%s: not an ELF binary... don't know how to load it\n", name );
#ifdef __i386__
if( header->e_machine != EM_386 )
fatal_error("%s: not an i386 ELF binary... don't know how to load it\n", name );
#elif defined(__x86_64__)
if( header->e_machine != EM_X86_64 )
fatal_error("%s: not an x86-64 ELF binary... don't know how to load it\n", name );
#elif defined(__aarch64__)
if( header->e_machine != EM_AARCH64 )
fatal_error("%s: not an aarch64 ELF binary... don't know how to load it\n", name );
#elif defined(__arm__)
if( header->e_machine != EM_ARM )
fatal_error("%s: not an arm ELF binary... don't know how to load it\n", name );
#endif
if (header->e_phnum > sizeof(loadcmds)/sizeof(loadcmds[0]))
fatal_error( "%s: oops... not enough space for load commands\n", name );
maplength = header->e_phnum * sizeof (ElfW(Phdr));
if (header->e_phoff + maplength > sizeof(buf))
fatal_error( "%s: oops... not enough space for ELF headers\n", name );
l->l_ld = 0;
l->l_addr = 0;
l->l_phdr = 0;
l->l_phnum = header->e_phnum;
l->l_entry = header->e_entry;
l->l_interp = 0;
for (ph = phdr; ph < &phdr[l->l_phnum]; ++ph)
{
#ifdef DUMP_SEGMENTS
wld_printf( "ph = %p\n", ph );
wld_printf( " p_type = %lx\n", (unsigned long)ph->p_type );
wld_printf( " p_flags = %lx\n", (unsigned long)ph->p_flags );
wld_printf( " p_offset = %lx\n", (unsigned long)ph->p_offset );
wld_printf( " p_vaddr = %lx\n", (unsigned long)ph->p_vaddr );
wld_printf( " p_paddr = %lx\n", (unsigned long)ph->p_paddr );
wld_printf( " p_filesz = %lx\n", (unsigned long)ph->p_filesz );
wld_printf( " p_memsz = %lx\n", (unsigned long)ph->p_memsz );
wld_printf( " p_align = %lx\n", (unsigned long)ph->p_align );
#endif
switch (ph->p_type)
{
/* These entries tell us where to find things once the file's
segments are mapped in. We record the addresses it says
verbatim, and later correct for the run-time load address. */
case PT_DYNAMIC:
l->l_ld = (void *) ph->p_vaddr;
l->l_ldnum = ph->p_memsz / sizeof (Elf32_Dyn);
break;
case PT_PHDR:
l->l_phdr = (void *) ph->p_vaddr;
break;
case PT_LOAD:
{
if ((ph->p_align & page_mask) != 0)
fatal_error( "%s: ELF load command alignment not page-aligned\n", name );
if (((ph->p_vaddr - ph->p_offset) & (ph->p_align - 1)) != 0)
fatal_error( "%s: ELF load command address/offset not properly aligned\n", name );
c = &loadcmds[nloadcmds++];
c->mapstart = ph->p_vaddr & ~(ph->p_align - 1);
c->mapend = ((ph->p_vaddr + ph->p_filesz + page_mask) & ~page_mask);
c->dataend = ph->p_vaddr + ph->p_filesz;
c->allocend = ph->p_vaddr + ph->p_memsz;
c->mapoff = ph->p_offset & ~(ph->p_align - 1);
c->prot = 0;
if (ph->p_flags & PF_R)
c->prot |= PROT_READ;
if (ph->p_flags & PF_W)
c->prot |= PROT_WRITE;
if (ph->p_flags & PF_X)
c->prot |= PROT_EXEC;
}
break;
case PT_INTERP:
l->l_interp = ph->p_vaddr;
break;
case PT_TLS:
/*
* We don't need to set anything up because we're
* emulating the kernel, not ld-linux.so.2
* The ELF loader will set up the TLS data itself.
*/
case PT_SHLIB:
case PT_NOTE:
default:
break;
}
}
/* Now process the load commands and map segments into memory. */
if (!nloadcmds)
fatal_error( "%s: no segments to load\n", name );
c = loadcmds;
/* Length of the sections to be loaded. */
maplength = loadcmds[nloadcmds - 1].allocend - c->mapstart;
if( header->e_type == ET_DYN )
{
ElfW(Addr) mappref;
mappref = (ELF_PREFERRED_ADDRESS (loader, maplength, c->mapstart)
- MAP_BASE_ADDR (l));
/* Remember which part of the address space this object uses. */
l->l_map_start = (ElfW(Addr)) wld_mmap ((void *) mappref, maplength,
c->prot, MAP_COPY | MAP_FILE,
fd, c->mapoff);
/* wld_printf("set : offset = %x\n", c->mapoff); */
/* wld_printf("l->l_map_start = %x\n", l->l_map_start); */
l->l_map_end = l->l_map_start + maplength;
l->l_addr = l->l_map_start - c->mapstart;
wld_mprotect ((caddr_t) (l->l_addr + c->mapend),
loadcmds[nloadcmds - 1].allocend - c->mapend,
PROT_NONE);
goto postmap;
}
else
{
/* sanity check */
if ((char *)c->mapstart + maplength > preloader_start &&
(char *)c->mapstart <= preloader_end)
fatal_error( "%s: binary overlaps preloader (%p-%p)\n",
name, (char *)c->mapstart, (char *)c->mapstart + maplength );
ELF_FIXED_ADDRESS (loader, c->mapstart);
}
/* Remember which part of the address space this object uses. */
l->l_map_start = c->mapstart + l->l_addr;
l->l_map_end = l->l_map_start + maplength;
while (c < &loadcmds[nloadcmds])
{
if (c->mapend > c->mapstart)
/* Map the segment contents from the file. */
wld_mmap ((void *) (l->l_addr + c->mapstart),
c->mapend - c->mapstart, c->prot,
MAP_FIXED | MAP_COPY | MAP_FILE, fd, c->mapoff);
postmap:
if (l->l_phdr == 0
&& (ElfW(Off)) c->mapoff <= header->e_phoff
&& ((size_t) (c->mapend - c->mapstart + c->mapoff)
>= header->e_phoff + header->e_phnum * sizeof (ElfW(Phdr))))
/* Found the program header in this segment. */
l->l_phdr = (void *)(unsigned long)(c->mapstart + header->e_phoff - c->mapoff);
if (c->allocend > c->dataend)
{
/* Extra zero pages should appear at the end of this segment,
after the data mapped from the file. */
ElfW(Addr) zero, zeroend, zeropage;
zero = l->l_addr + c->dataend;
zeroend = l->l_addr + c->allocend;
zeropage = (zero + page_mask) & ~page_mask;
/*
* This is different from the dl-load load...
* ld-linux.so.2 relies on the whole page being zero'ed
*/
zeroend = (zeroend + page_mask) & ~page_mask;
if (zeroend < zeropage)
{
/* All the extra data is in the last page of the segment.
We can just zero it. */
zeropage = zeroend;
}
if (zeropage > zero)
{
/* Zero the final part of the last page of the segment. */
if ((c->prot & PROT_WRITE) == 0)
{
/* Dag nab it. */
wld_mprotect ((caddr_t) (zero & ~page_mask), page_size, c->prot|PROT_WRITE);
}
wld_memset ((void *) zero, '\0', zeropage - zero);
if ((c->prot & PROT_WRITE) == 0)
wld_mprotect ((caddr_t) (zero & ~page_mask), page_size, c->prot);
}
if (zeroend > zeropage)
{
/* Map the remaining zero pages in from the zero fill FD. */
wld_mmap ((caddr_t) zeropage, zeroend - zeropage,
c->prot, MAP_ANON|MAP_PRIVATE|MAP_FIXED,
-1, 0);
}
}
++c;
}
if (l->l_phdr == NULL) fatal_error("no program header\n");
l->l_phdr = (void *)((ElfW(Addr))l->l_phdr + l->l_addr);
l->l_entry += l->l_addr;
wld_close( fd );
}
static unsigned int wld_elf_hash( const char *name )
{
unsigned int hi, hash = 0;
while (*name)
{
hash = (hash << 4) + (unsigned char)*name++;
hi = hash & 0xf0000000;
hash ^= hi;
hash ^= hi >> 24;
}
return hash;
}
static unsigned int gnu_hash( const char *name )
{
unsigned int h = 5381;
while (*name) h = h * 33 + (unsigned char)*name++;
return h;
}
/*
* Find a symbol in the symbol table of the executable loaded
*/
static void *find_symbol( const struct wld_link_map *map, const char *var, int type )
{
const ElfW(Dyn) *dyn = NULL;
const ElfW(Phdr) *ph;
const ElfW(Sym) *symtab = NULL;
const Elf32_Word *hashtab = NULL;
const Elf32_Word *gnu_hashtab = NULL;
const char *strings = NULL;
Elf32_Word idx;
/* check the values */
#ifdef DUMP_SYMS
wld_printf("%p %x\n", map->l_phdr, map->l_phnum );
#endif
/* parse the (already loaded) ELF executable's header */
for (ph = map->l_phdr; ph < &map->l_phdr[map->l_phnum]; ++ph)
{
if( PT_DYNAMIC == ph->p_type )
{
dyn = (void *)(ph->p_vaddr + map->l_addr);
break;
}
}
if( !dyn ) return NULL;
while( dyn->d_tag )
{
if( dyn->d_tag == DT_STRTAB )
strings = (const char*)(dyn->d_un.d_ptr + map->l_addr);
if( dyn->d_tag == DT_SYMTAB )
symtab = (const ElfW(Sym) *)(dyn->d_un.d_ptr + map->l_addr);
if( dyn->d_tag == DT_HASH )
hashtab = (const Elf32_Word *)(dyn->d_un.d_ptr + map->l_addr);
if( dyn->d_tag == DT_GNU_HASH )
gnu_hashtab = (const Elf32_Word *)(dyn->d_un.d_ptr + map->l_addr);
#ifdef DUMP_SYMS
wld_printf("%lx %p\n", (unsigned long)dyn->d_tag, (void *)dyn->d_un.d_ptr );
#endif
dyn++;
}
if( (!symtab) || (!strings) ) return NULL;
if (gnu_hashtab) /* new style hash table */
{
const unsigned int hash = gnu_hash(var);
const Elf32_Word nbuckets = gnu_hashtab[0];
const Elf32_Word symbias = gnu_hashtab[1];
const Elf32_Word nwords = gnu_hashtab[2];
const ElfW(Addr) *bitmask = (const ElfW(Addr) *)(gnu_hashtab + 4);
const Elf32_Word *buckets = (const Elf32_Word *)(bitmask + nwords);
const Elf32_Word *chains = buckets + nbuckets - symbias;
if (!(idx = buckets[hash % nbuckets])) return NULL;
do
{
if ((chains[idx] & ~1u) == (hash & ~1u) &&
ELF32_ST_BIND(symtab[idx].st_info) == STB_GLOBAL &&
ELF32_ST_TYPE(symtab[idx].st_info) == type &&
!wld_strcmp( strings + symtab[idx].st_name, var ))
goto found;
} while (!(chains[idx++] & 1u));
}
else if (hashtab) /* old style hash table */
{
const unsigned int hash = wld_elf_hash(var);
const Elf32_Word nbuckets = hashtab[0];
const Elf32_Word *buckets = hashtab + 2;
const Elf32_Word *chains = buckets + nbuckets;
for (idx = buckets[hash % nbuckets]; idx; idx = chains[idx])
{
if (ELF32_ST_BIND(symtab[idx].st_info) == STB_GLOBAL &&
ELF32_ST_TYPE(symtab[idx].st_info) == type &&
!wld_strcmp( strings + symtab[idx].st_name, var ))
goto found;
}
}
return NULL;
found:
#ifdef DUMP_SYMS
wld_printf("Found %s -> %p\n", strings + symtab[idx].st_name, (void *)symtab[idx].st_value );
#endif
return (void *)(symtab[idx].st_value + map->l_addr);
}
/*
* preload_reserve
*
* Reserve a range specified in string format
*/
static void preload_reserve( const char *str )
{
const char *p;
unsigned long result = 0;
void *start = NULL, *end = NULL;
int i, first = 1;
for (p = str; *p; p++)
{
if (*p >= '0' && *p <= '9') result = result * 16 + *p - '0';
else if (*p >= 'a' && *p <= 'f') result = result * 16 + *p - 'a' + 10;
else if (*p >= 'A' && *p <= 'F') result = result * 16 + *p - 'A' + 10;
else if (*p == '-')
{
if (!first) goto error;
start = (void *)(result & ~page_mask);
result = 0;
first = 0;
}
else goto error;
}
if (!first) end = (void *)((result + page_mask) & ~page_mask);
else if (result) goto error; /* single value '0' is allowed */
/* sanity checks */
if (end <= start) start = end = NULL;
else if ((char *)end > preloader_start &&
(char *)start <= preloader_end)
{
wld_printf( "WINEPRELOADRESERVE range %p-%p overlaps preloader %p-%p\n",
start, end, preloader_start, preloader_end );
start = end = NULL;
}
/* check for overlap with low memory areas */
for (i = 0; preload_info[i].size; i++)
{
if ((char *)preload_info[i].addr > (char *)0x00110000) break;
if ((char *)end <= (char *)preload_info[i].addr + preload_info[i].size)
{
start = end = NULL;
break;
}
if ((char *)start < (char *)preload_info[i].addr + preload_info[i].size)
start = (char *)preload_info[i].addr + preload_info[i].size;
}
while (preload_info[i].size) i++;
preload_info[i].addr = start;
preload_info[i].size = (char *)end - (char *)start;
return;
error:
fatal_error( "invalid WINEPRELOADRESERVE value '%s'\n", str );
}
/* check if address is in one of the reserved ranges */
static int is_addr_reserved( const void *addr )
{
int i;
for (i = 0; preload_info[i].size; i++)
{
if ((const char *)addr >= (const char *)preload_info[i].addr &&
(const char *)addr < (const char *)preload_info[i].addr + preload_info[i].size)
return 1;
}
return 0;
}
/* remove a range from the preload list */
static void remove_preload_range( int i )
{
while (preload_info[i].size)
{
preload_info[i].addr = preload_info[i+1].addr;
preload_info[i].size = preload_info[i+1].size;
i++;
}
}
/*
* is_in_preload_range
*
* Check if address of the given aux value is in one of the reserved ranges
*/
2013-02-13 15:26:39 +00:00
static int is_in_preload_range( const struct wld_auxv *av, int type )
{
while (av->a_type != AT_NULL)
{
if (av->a_type == type) return is_addr_reserved( (const void *)av->a_un.a_val );
av++;
}
return 0;
}
/* set the process name if supported */
static void set_process_name( int argc, char *argv[] )
{
2007-03-03 13:30:55 +00:00
int i;
unsigned int off;
char *p, *name, *end;
/* set the process short name */
for (p = name = argv[1]; *p; p++) if (p[0] == '/' && p[1]) name = p + 1;
if (wld_prctl( 15 /* PR_SET_NAME */, (long)name ) == -1) return;
/* find the end of the argv array and move everything down */
end = argv[argc - 1];
while (*end) end++;
off = argv[1] - argv[0];
for (p = argv[1]; p <= end; p++) *(p - off) = *p;
wld_memset( end - off, 0, off );
for (i = 1; i < argc; i++) argv[i] -= off;
}
/*
* wld_start
*
* Repeat the actions the kernel would do when loading a dynamically linked .so
* Load the binary and then its ELF interpreter.
* Note, we assume that the binary is a dynamically linked ELF shared object.
*/
void* wld_start( void **stack )
{
long i, *pargc;
char **argv, **p;
char *interp, *reserve = NULL;
struct wld_auxv new_av[8], delete_av[3], *av;
struct wld_link_map main_binary_map, ld_so_map;
struct wine_preload_info **wine_main_preload_info;
pargc = *stack;
argv = (char **)pargc + 1;
if (*pargc < 2) fatal_error( "Usage: %s wine_binary [args]\n", argv[0] );
/* skip over the parameters */
p = argv + *pargc + 1;
/* skip over the environment */
while (*p)
{
static const char res[] = "WINEPRELOADRESERVE=";
if (!wld_strncmp( *p, res, sizeof(res)-1 )) reserve = *p + sizeof(res) - 1;
p++;
}
2013-02-13 15:26:39 +00:00
av = (struct wld_auxv *)(p+1);
page_size = get_auxiliary( av, AT_PAGESZ, 4096 );
page_mask = page_size - 1;
preloader_start = (char *)((unsigned long)_start & ~page_mask);
preloader_end = (char *)((unsigned long)(_end + page_mask) & ~page_mask);
#ifdef DUMP_AUX_INFO
wld_printf( "stack = %p\n", *stack );
for( i = 0; i < *pargc; i++ ) wld_printf("argv[%lx] = %s\n", i, argv[i]);
dump_auxiliary( av );
#endif
/* reserve memory that Wine needs */
if (reserve) preload_reserve( reserve );
for (i = 0; preload_info[i].size; i++)
{
if ((char *)av >= (char *)preload_info[i].addr &&
(char *)pargc <= (char *)preload_info[i].addr + preload_info[i].size)
{
remove_preload_range( i );
i--;
}
else if (wld_mmap( preload_info[i].addr, preload_info[i].size, PROT_NONE,
MAP_FIXED | MAP_PRIVATE | MAP_ANON | MAP_NORESERVE, -1, 0 ) == (void *)-1)
{
/* don't warn for low 64k */
if (preload_info[i].addr >= (void *)0x10000
#ifdef __aarch64__
&& preload_info[i].addr < (void *)0x7fffffffff /* ARM64 address space might end here*/
#endif
)
wld_printf( "preloader: Warning: failed to reserve range %p-%p\n",
preload_info[i].addr, (char *)preload_info[i].addr + preload_info[i].size );
remove_preload_range( i );
i--;
}
}
/* add an executable page at the top of the address space to defeat
* broken no-exec protections that play with the code selector limit */
if (is_addr_reserved( (char *)0x80000000 - page_size ))
wld_mprotect( (char *)0x80000000 - page_size, page_size, PROT_EXEC | PROT_READ );
/* load the main binary */
map_so_lib( argv[1], &main_binary_map );
/* load the ELF interpreter */
interp = (char *)main_binary_map.l_addr + main_binary_map.l_interp;
map_so_lib( interp, &ld_so_map );
/* store pointer to the preload info into the appropriate main binary variable */
wine_main_preload_info = find_symbol( &main_binary_map, "wine_main_preload_info", STT_OBJECT );
if (wine_main_preload_info) *wine_main_preload_info = preload_info;
else wld_printf( "wine_main_preload_info not found\n" );
#define SET_NEW_AV(n,type,val) new_av[n].a_type = (type); new_av[n].a_un.a_val = (val);
SET_NEW_AV( 0, AT_PHDR, (unsigned long)main_binary_map.l_phdr );
SET_NEW_AV( 1, AT_PHENT, sizeof(ElfW(Phdr)) );
SET_NEW_AV( 2, AT_PHNUM, main_binary_map.l_phnum );
SET_NEW_AV( 3, AT_PAGESZ, page_size );
SET_NEW_AV( 4, AT_BASE, ld_so_map.l_addr );
SET_NEW_AV( 5, AT_FLAGS, get_auxiliary( av, AT_FLAGS, 0 ) );
SET_NEW_AV( 6, AT_ENTRY, main_binary_map.l_entry );
SET_NEW_AV( 7, AT_NULL, 0 );
#undef SET_NEW_AV
i = 0;
/* delete sysinfo values if addresses conflict */
if (is_in_preload_range( av, AT_SYSINFO ) || is_in_preload_range( av, AT_SYSINFO_EHDR ))
{
delete_av[i++].a_type = AT_SYSINFO;
delete_av[i++].a_type = AT_SYSINFO_EHDR;
}
delete_av[i].a_type = AT_NULL;
/* get rid of first argument */
set_process_name( *pargc, argv );
pargc[1] = pargc[0] - 1;
*stack = pargc + 1;
set_auxiliary_values( av, new_av, delete_av, stack );
#ifdef DUMP_AUX_INFO
wld_printf("new stack = %p\n", *stack);
wld_printf("jumping to %p\n", (void *)ld_so_map.l_entry);
#endif
#ifdef DUMP_MAPS
{
char buffer[1024];
int len, fd = wld_open( "/proc/self/maps", O_RDONLY );
if (fd != -1)
{
while ((len = wld_read( fd, buffer, sizeof(buffer) )) > 0) wld_write( 2, buffer, len );
wld_close( fd );
}
}
#endif
return (void *)ld_so_map.l_entry;
}
loader: Set the visibility of symbols in preloader to "hidden". Today, the preloader is linked with -fPIE in spite of the fact that the preloader is a non-PIE statically linked binary. This is due to a limitation in tools/makedep that makes it difficult to specify CFLAGS for each individual object file's recipe. This can seemingly cause problems with some GOTPCREL(X) relocations inside the preloader. Since preloader does not link to the system library directly, there is no need for a Global Offset Table (GOT). However, a few extern (non-static) function symbols are declared, the use of which makes GCC emit instructions that references those symbols by indirection through GOT. The linker then tries to optimize such instructions to eliminate GOT references, which can fail due to various reasons. This stands in contradiction with Jinoh Kang's suggestion (in bug 55050) that "-fPIE is harmless even when applied to an object linked into non-PIE executables." The claim is theoretically true since position-independent code can in principle be relocated to any address (fixed or dynamic); however, it fails due to some peculiar practical issues, which is arguably a limitation in the linker's implementation (since it can be worked around with -Wl,--no-relax without issues). Fix this by eliminating GOT usage by setting the default visibility of non-static declarations to "hidden". Assuming GCC's medium code model (-mcmodel=medium; default code model for x86_64), this suppresses any unnecessary PLT or GOT relocations for defined symbols, and provides opportunity for GCC to optimize the code better. Fixes: 78ed343842dcd8ffb95c416420953e121959d40d Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55091
2023-06-20 14:01:25 +00:00
#pragma GCC visibility pop
#endif /* __linux__ */