d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

rtld support for sparc64.

Browse Source 
Largely obtained from:	netbsd
Submitted by:	jake, tmm
This commit is contained in:
jake 2002-03-13 02:40:39 +00:00
parent bceec52b62
commit b71e3c8c82
5 changed files with 1073 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
libexec/rtld-elf/sparc64/Makefile.inc Normal file
View File

@ -0,0 +1 @@
# $FreeBSD$

176
libexec/rtld-elf/sparc64/lockdflt.c Normal file
View File

@ -0,0 +1,176 @@
/*-
* Copyright 1999, 2000 John D. Polstra.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* from: FreeBSD: src/libexec/rtld-elf/alpha/lockdflt.c,v 1.6 2000/07/17
* $FreeBSD$
*/
/*
* Thread locking implementation for the dynamic linker.
*
* We use the "simple, non-scalable reader-preference lock" from:
*
* J. M. Mellor-Crummey and M. L. Scott. "Scalable Reader-Writer
* Synchronization for Shared-Memory Multiprocessors." 3rd ACM Symp. on
* Principles and Practice of Parallel Programming, April 1991.
*
* In this algorithm the lock is a single word. Its low-order bit is
* set when a writer holds the lock. The remaining high-order bits
* contain a count of readers desiring the lock. The algorithm requires
* atomic "compare_and_store" and "add" operations, which we implement
* using assembly language sequences in "rtld_start.S".
*
* These are spinlocks. When spinning we call nanosleep() for 1
* microsecond each time around the loop. This will most likely yield
* the CPU to other threads (including, we hope, the lockholder) allowing
* them to make some progress.
*/
#include <signal.h>
#include <stdlib.h>
#include <time.h>
#include <machine/atomic.h>
#include "debug.h"
#include "rtld.h"
#include "rtld_machdep.h"
#define WAFLAG 0x1 /* A writer holds the lock */
#define RC_INCR 0x2 /* Adjusts count of readers desiring lock */
typedef struct Struct_Lock {
volatile int lock;
void *base;
} Lock;
static const struct timespec usec = { 0, 1000 }; /* 1 usec. */
static sigset_t fullsigmask, oldsigmask;
static void *
lock_create(void *context)
{
void *base;
char *p;
uintptr_t r;
Lock *l;
/*
* Arrange for the lock to occupy its own cache line. First, we
* optimistically allocate just a cache line, hoping that malloc
* will give us a well-aligned block of memory. If that doesn't
* work, we allocate a larger block and take a well-aligned cache
* line from it.
*/
base = xmalloc(CACHE_LINE_SIZE);
p = (char *)base;
if ((uintptr_t)p % CACHE_LINE_SIZE != 0) {
free(base);
base = xmalloc(2 * CACHE_LINE_SIZE);
p = (char *)base;
if ((r = (uintptr_t)p % CACHE_LINE_SIZE) != 0)
p += CACHE_LINE_SIZE - r;
}
l = (Lock *)p;
l->base = base;
l->lock = 0;
return l;
}
static void
lock_destroy(void *lock)
{
Lock *l = (Lock *)lock;
free(l->base);
}
static void
rlock_acquire(void *lock)
{
Lock *l = (Lock *)lock;
atomic_add_acq_int(&l->lock, RC_INCR);
while (l->lock & WAFLAG)
nanosleep(&usec, NULL);
}
static void
wlock_acquire(void *lock)
{
Lock *l = (Lock *)lock;
sigset_t tmp_oldsigmask;
for ( ; ; ) {
sigprocmask(SIG_BLOCK, &fullsigmask, &tmp_oldsigmask);
if (atomic_cmpset_acq_int(&l->lock, 0, WAFLAG))
break;
sigprocmask(SIG_SETMASK, &tmp_oldsigmask, NULL);
nanosleep(&usec, NULL);
}
oldsigmask = tmp_oldsigmask;
}
static void
rlock_release(void *lock)
{
Lock *l = (Lock *)lock;
atomic_add_rel_int(&l->lock, -RC_INCR);
}
static void
wlock_release(void *lock)
{
Lock *l = (Lock *)lock;
atomic_add_rel_int(&l->lock, -WAFLAG);
sigprocmask(SIG_SETMASK, &oldsigmask, NULL);
}
void
lockdflt_init(LockInfo *li)
{
li->context = NULL;
li->lock_create = lock_create;
li->rlock_acquire = rlock_acquire;
li->wlock_acquire = wlock_acquire;
li->rlock_release = rlock_release;
li->wlock_release = wlock_release;
li->lock_destroy = lock_destroy;
li->context_destroy = NULL;
/*
* Construct a mask to block all signals except traps which might
* conceivably be generated within the dynamic linker itself.
*/
sigfillset(&fullsigmask);
sigdelset(&fullsigmask, SIGILL);
sigdelset(&fullsigmask, SIGTRAP);
sigdelset(&fullsigmask, SIGABRT);
sigdelset(&fullsigmask, SIGEMT);
sigdelset(&fullsigmask, SIGFPE);
sigdelset(&fullsigmask, SIGBUS);
sigdelset(&fullsigmask, SIGSEGV);
sigdelset(&fullsigmask, SIGSYS);
}

693
libexec/rtld-elf/sparc64/reloc.c Normal file
View File

@ -0,0 +1,693 @@
/* $NetBSD: mdreloc.c,v 1.5 2001/04/25 12:24:51 kleink Exp $ */
/*-
* Copyright (c) 2000 Eduardo Horvath.
* Copyright (c) 1999 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Paul Kranenburg.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <errno.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <sys/stat.h>
#include "debug.h"
#include "rtld.h"
/*
* The following table holds for each relocation type:
* - the width in bits of the memory location the relocation
* applies to (not currently used)
* - the number of bits the relocation value must be shifted to the
* right (i.e. discard least significant bits) to fit into
* the appropriate field in the instruction word.
* - flags indicating whether
* * the relocation involves a symbol
* * the relocation is relative to the current position
* * the relocation is for a GOT entry
* * the relocation is relative to the load address
*
*/
#define _RF_S 0x80000000 /* Resolve symbol */
#define _RF_A 0x40000000 /* Use addend */
#define _RF_P 0x20000000 /* Location relative */
#define _RF_G 0x10000000 /* GOT offset */
#define _RF_B 0x08000000 /* Load address relative */
#define _RF_U 0x04000000 /* Unaligned */
#define _RF_SZ(s) (((s) & 0xff) << 8) /* memory target size */
#define _RF_RS(s) ( (s) & 0xff) /* right shift */
static int reloc_target_flags[] = {
0, /* NONE */
_RF_S|_RF_A| _RF_SZ(8) | _RF_RS(0), /* RELOC_8 */
_RF_S|_RF_A| _RF_SZ(16) | _RF_RS(0), /* RELOC_16 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* RELOC_32 */
_RF_S|_RF_A|_RF_P| _RF_SZ(8) | _RF_RS(0), /* DISP_8 */
_RF_S|_RF_A|_RF_P| _RF_SZ(16) | _RF_RS(0), /* DISP_16 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(0), /* DISP_32 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(2), /* WDISP_30 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(2), /* WDISP_22 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(10), /* HI22 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 22 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 13 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* LO10 */
_RF_G| _RF_SZ(32) | _RF_RS(0), /* GOT10 */
_RF_G| _RF_SZ(32) | _RF_RS(0), /* GOT13 */
_RF_G| _RF_SZ(32) | _RF_RS(10), /* GOT22 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(0), /* PC10 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(10), /* PC22 */
_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(2), /* WPLT30 */
_RF_SZ(32) | _RF_RS(0), /* COPY */
_RF_S|_RF_A| _RF_SZ(64) | _RF_RS(0), /* GLOB_DAT */
_RF_SZ(32) | _RF_RS(0), /* JMP_SLOT */
_RF_A| _RF_B| _RF_SZ(64) | _RF_RS(0), /* RELATIVE */
_RF_S|_RF_A| _RF_U| _RF_SZ(32) | _RF_RS(0), /* UA_32 */
_RF_A| _RF_SZ(32) | _RF_RS(0), /* PLT32 */
_RF_A| _RF_SZ(32) | _RF_RS(10), /* HIPLT22 */
_RF_A| _RF_SZ(32) | _RF_RS(0), /* LOPLT10 */
_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(0), /* PCPLT32 */
_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(10), /* PCPLT22 */
_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(0), /* PCPLT10 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 10 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 11 */
_RF_S|_RF_A| _RF_SZ(64) | _RF_RS(0), /* 64 */
_RF_S|_RF_A|/*extra*/ _RF_SZ(32) | _RF_RS(0), /* OLO10 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(42), /* HH22 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(32), /* HM10 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(10), /* LM22 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(42), /* PC_HH22 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(32), /* PC_HM10 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(10), /* PC_LM22 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(2), /* WDISP16 */
_RF_S|_RF_A|_RF_P| _RF_SZ(32) | _RF_RS(2), /* WDISP19 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* GLOB_JMP */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 7 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 5 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* 6 */
_RF_S|_RF_A|_RF_P| _RF_SZ(64) | _RF_RS(0), /* DISP64 */
_RF_A| _RF_SZ(64) | _RF_RS(0), /* PLT64 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(10), /* HIX22 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* LOX10 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(22), /* H44 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(12), /* M44 */
_RF_S|_RF_A| _RF_SZ(32) | _RF_RS(0), /* L44 */
_RF_S|_RF_A| _RF_SZ(64) | _RF_RS(0), /* REGISTER */
_RF_S|_RF_A| _RF_U| _RF_SZ(64) | _RF_RS(0), /* UA64 */
_RF_S|_RF_A| _RF_U| _RF_SZ(16) | _RF_RS(0), /* UA16 */
};
#if 0
static const char *reloc_names[] = {
"NONE", "RELOC_8", "RELOC_16", "RELOC_32", "DISP_8",
"DISP_16", "DISP_32", "WDISP_30", "WDISP_22", "HI22",
"22", "13", "LO10", "GOT10", "GOT13",
"GOT22", "PC10", "PC22", "WPLT30", "COPY",
"GLOB_DAT", "JMP_SLOT", "RELATIVE", "UA_32", "PLT32",
"HIPLT22", "LOPLT10", "LOPLT10", "PCPLT22", "PCPLT32",
"10", "11", "64", "OLO10", "HH22",
"HM10", "LM22", "PC_HH22", "PC_HM10", "PC_LM22",
"WDISP16", "WDISP19", "GLOB_JMP", "7", "5", "6",
"DISP64", "PLT64", "HIX22", "LOX10", "H44", "M44",
"L44", "REGISTER", "UA64", "UA16"
};
#endif
#define RELOC_RESOLVE_SYMBOL(t) ((reloc_target_flags[t] & _RF_S) != 0)
#define RELOC_PC_RELATIVE(t) ((reloc_target_flags[t] & _RF_P) != 0)
#define RELOC_BASE_RELATIVE(t) ((reloc_target_flags[t] & _RF_B) != 0)
#define RELOC_UNALIGNED(t) ((reloc_target_flags[t] & _RF_U) != 0)
#define RELOC_USE_ADDEND(t) ((reloc_target_flags[t] & _RF_A) != 0)
#define RELOC_TARGET_SIZE(t) ((reloc_target_flags[t] >> 8) & 0xff)
#define RELOC_VALUE_RIGHTSHIFT(t) (reloc_target_flags[t] & 0xff)
static long reloc_target_bitmask[] = {
#define _BM(x) (~(-(1ULL << (x))))
0, /* NONE */
_BM(8), _BM(16), _BM(32), /* RELOC_8, _16, _32 */
_BM(8), _BM(16), _BM(32), /* DISP8, DISP16, DISP32 */
_BM(30), _BM(22), /* WDISP30, WDISP22 */
_BM(22), _BM(22), /* HI22, _22 */
_BM(13), _BM(10), /* RELOC_13, _LO10 */
_BM(10), _BM(13), _BM(22), /* GOT10, GOT13, GOT22 */
_BM(10), _BM(22), /* _PC10, _PC22 */
_BM(30), 0, /* _WPLT30, _COPY */
_BM(32), _BM(32), _BM(32), /* _GLOB_DAT, JMP_SLOT, _RELATIVE */
_BM(32), _BM(32), /* _UA32, PLT32 */
_BM(22), _BM(10), /* _HIPLT22, LOPLT10 */
_BM(32), _BM(22), _BM(10), /* _PCPLT32, _PCPLT22, _PCPLT10 */
_BM(10), _BM(11), -1, /* _10, _11, _64 */
_BM(10), _BM(22), /* _OLO10, _HH22 */
_BM(10), _BM(22), /* _HM10, _LM22 */
_BM(22), _BM(10), _BM(22), /* _PC_HH22, _PC_HM10, _PC_LM22 */
_BM(16), _BM(19), /* _WDISP16, _WDISP19 */
-1, /* GLOB_JMP */
_BM(7), _BM(5), _BM(6) /* _7, _5, _6 */
-1, -1, /* DISP64, PLT64 */
_BM(22), _BM(13), /* HIX22, LOX10 */
_BM(22), _BM(10), _BM(13), /* H44, M44, L44 */
-1, -1, _BM(16), /* REGISTER, UA64, UA16 */
#undef _BM
};
#define RELOC_VALUE_BITMASK(t) (reloc_target_bitmask[t])
#undef flush
#define flush(va, offs) \
__asm __volatile("flush %0 + %1" : : "r" (va), "I" (offs));
static int reloc_nonplt_object(Obj_Entry *obj, const Elf_Rela *rela,
SymCache *cache);
static void install_plt(Elf_Half *pltgot, Elf_Addr proc);
extern char _rtld_bind_start_0[];
extern char _rtld_bind_start_1[];
int
do_copy_relocations(Obj_Entry *dstobj)
{
const Elf_Rela *relalim;
const Elf_Rela *rela;
const Elf_Sym *dstsym;
const Elf_Sym *srcsym;
void *dstaddr;
const void *srcaddr;
Obj_Entry *srcobj;
unsigned long hash;
const char *name;
size_t size;
assert(dstobj->mainprog); /* COPY relocations are invalid elsewhere */
relalim = (const Elf_Rela *)((caddr_t)dstobj->rela + dstobj->relasize);
for (rela = dstobj->rela; rela < relalim; rela++) {
if (ELF_R_TYPE(rela->r_info) == R_SPARC_COPY) {
dstaddr = (void *)(dstobj->relocbase + rela->r_offset);
dstsym = dstobj->symtab + ELF_R_SYM(rela->r_info);
name = dstobj->strtab + dstsym->st_name;
hash = elf_hash(name);
size = dstsym->st_size;
for (srcobj = dstobj->next; srcobj != NULL;
srcobj = srcobj->next)
if ((srcsym = symlook_obj(name, hash, srcobj,
false)) != NULL)
break;
if (srcobj == NULL) {
_rtld_error("Undefined symbol \"%s\""
"referenced from COPY relocation"
"in %s", name, dstobj->path);
return (-1);
}
srcaddr = (const void *)(srcobj->relocbase +
srcsym->st_value);
memcpy(dstaddr, srcaddr, size);
}
}
return (0);
}
int
reloc_non_plt(Obj_Entry *obj, Obj_Entry *obj_rtld)
{
const Elf_Rela *relalim;
const Elf_Rela *rela;
SymCache *cache;
cache = (SymCache *)alloca(obj->nchains * sizeof(SymCache));
if (cache != NULL)
memset(cache, 0, obj->nchains * sizeof(SymCache));
relalim = (const Elf_Rela *)((caddr_t)obj->rela + obj->relasize);
for (rela = obj->rela; rela < relalim; rela++) {
if (reloc_nonplt_object(obj, rela, cache) < 0)
return (-1);
}
return (0);
}
static int
reloc_nonplt_object(Obj_Entry *obj, const Elf_Rela *rela, SymCache *cache)
{
const Obj_Entry *defobj;
const Elf_Sym *def;
Elf_Addr *where;
Elf_Half *where32;
Elf_Word type;
Elf_Addr value;
Elf_Addr mask;
where = (Elf_Addr *)(obj->relocbase + rela->r_offset);
where32 = (Elf_Half *)where;
defobj = NULL;
def = NULL;
type = ELF_R_TYPE(rela->r_info);
if (type == R_SPARC_NONE)
return (0);
/* We do JMP_SLOTs below */
if (type == R_SPARC_JMP_SLOT)
return (0);
/* COPY relocs are also handled elsewhere */
if (type == R_SPARC_COPY)
return (0);
/*
* Note: R_SPARC_UA16 must be numerically largest relocation type.
*/
if (type > R_SPARC_UA16)
return (-1);
value = rela->r_addend;
/*
* Handle relative relocs here, because we might not
* be able to access globals yet.
*/
if (type == R_SPARC_RELATIVE) {
/* XXXX -- apparently we ignore the preexisting value */
*where = (Elf_Addr)(obj->relocbase + value);
return (0);
}
/*
* If we get here while relocating rtld itself, we will crash because
* a non-local variable is accessed.
*/
if (RELOC_RESOLVE_SYMBOL(type)) {
/* Find the symbol */
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
false, cache);
if (def == NULL)
return (-1);
/* Add in the symbol's absolute address */
value += (Elf_Addr)(defobj->relocbase + def->st_value);
}
if (RELOC_PC_RELATIVE(type))
value -= (Elf_Addr)where;
if (RELOC_BASE_RELATIVE(type)) {
/*
* Note that even though sparcs use `Elf_rela' exclusively
* we still need the implicit memory addend in relocations
* referring to GOT entries. Undoubtedly, someone f*cked
* this up in the distant past, and now we're stuck with
* it in the name of compatibility for all eternity..
*
* In any case, the implicit and explicit should be mutually
* exclusive. We provide a check for that here.
*/
/* XXXX -- apparently we ignore the preexisting value */
value += (Elf_Addr)(obj->relocbase);
}
mask = RELOC_VALUE_BITMASK(type);
value >>= RELOC_VALUE_RIGHTSHIFT(type);
value &= mask;
if (RELOC_UNALIGNED(type)) {
/* Handle unaligned relocations. */
Elf_Addr tmp;
char *ptr;
int size;
int i;
size = RELOC_TARGET_SIZE(type) / 8;
ptr = (char *)where;
tmp = 0;
/* Read it in one byte at a time. */
for (i = 0; i < size; i++)
tmp = (tmp << 8) | ptr[i];
tmp &= ~mask;
tmp |= value;
/* Write it back out. */
for (i = 0; i < size; i++)
ptr[i] = ((tmp >> (8 * i)) & 0xff);
} else if (RELOC_TARGET_SIZE(type) > 32) {
*where &= ~mask;
*where |= value;
} else {
*where32 &= ~mask;
*where32 |= value;
}
return (0);
}
int
reloc_plt(Obj_Entry *obj)
{
#if 0
const Obj_Entry *defobj;
const Elf_Rela *relalim;
const Elf_Rela *rela;
const Elf_Sym *def;
Elf_Addr *where;
Elf_Addr value;
relalim = (const Elf_Rela *)((char *)obj->pltrela + obj->pltrelasize);
for (rela = obj->pltrela; rela < relalim; rela++) {
if (rela->r_addend == 0)
continue;
assert(ELF_R_TYPE(rela->r_info) == R_SPARC_JMP_SLOT);
where = (Elf_Addr *)(obj->relocbase + rela->r_offset);
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
true, NULL);
value = (Elf_Addr)(defobj->relocbase + def->st_value);
*where = value;
}
#endif
return (0);
}
/*
* Instruction templates:
*/
#define BAA 0x10400000 /* ba,a %xcc, 0 */
#define SETHI 0x03000000 /* sethi %hi(0), %g1 */
#define JMP 0x81c06000 /* jmpl %g1+%lo(0), %g0 */
#define NOP 0x01000000 /* sethi %hi(0), %g0 */
#define OR 0x82806000 /* or %g1, 0, %g1 */
#define XOR 0x82c06000 /* xor %g1, 0, %g1 */
#define MOV71 0x8283a000 /* or %o7, 0, %g1 */
#define MOV17 0x9c806000 /* or %g1, 0, %o7 */
#define CALL 0x40000000 /* call 0 */
#define SLLX 0x8b407000 /* sllx %g1, 0, %g1 */
#define SETHIG5 0x0b000000 /* sethi %hi(0), %g5 */
#define ORG5 0x82804005 /* or %g1, %g5, %g1 */
/* %hi(v) with variable shift */
#define HIVAL(v, s) (((v) >> (s)) & 0x003fffff)
#define LOVAL(v) ((v) & 0x000003ff)
int
reloc_jmpslots(Obj_Entry *obj)
{
const Obj_Entry *defobj;
const Elf_Rela *relalim;
const Elf_Rela *rela;
const Elf_Sym *def;
Elf_Addr *where;
Elf_Addr target;
relalim = (const Elf_Rela *)((char *)obj->pltrela + obj->pltrelasize);
for (rela = obj->pltrela; rela < relalim; rela++) {
assert(ELF_R_TYPE(rela->r_info) == R_SPARC_JMP_SLOT);
where = (Elf_Addr *)(obj->relocbase + rela->r_offset);
def = find_symdef(ELF_R_SYM(rela->r_info), obj, &defobj,
true, NULL);
target = (Elf_Addr)(defobj->relocbase + def->st_value);
reloc_jmpslot(where, target, defobj);
}
obj->jmpslots_done = true;
return (0);
}
Elf_Addr
reloc_jmpslot(Elf_Addr *wherep, Elf_Addr target, const Obj_Entry *obj)
{
Elf_Addr offset;
Elf_Half *where;
/*
* At the PLT entry pointed at by `where', we now construct
* a direct transfer to the now fully resolved function
* address.
*
* A PLT entry is supposed to start by looking like this:
*
* sethi %hi(. - .PLT0), %g1
* ba,a %xcc, .PLT1
* nop
* nop
* nop
* nop
* nop
* nop
*
* When we replace these entries we start from the second
* entry and do it in reverse order so the last thing we
* do is replace the branch. That allows us to change this
* atomically.
*
* We now need to find out how far we need to jump. We
* have a choice of several different relocation techniques
* which are increasingly expensive.
*/
where = (Elf_Half *)wherep;
offset = ((Elf_Addr)where) - target;
if (offset <= (1L<<20) && offset >= -(1L<<20)) {
/*
* We're within 1MB -- we can use a direct branch insn.
*
* We can generate this pattern:
*
* sethi %hi(. - .PLT0), %g1
* ba,a %xcc, addr
* nop
* nop
* nop
* nop
* nop
* nop
*
*/
where[1] = BAA | ((offset >> 2) &0x3fffff);
flush(where, 4);
} else if (target >= 0 && target < (1L<<32)) {
/*
* We're withing 32-bits of address zero.
*
* The resulting code in the jump slot is:
*
* sethi %hi(. - .PLT0), %g1
* sethi %hi(addr), %g1
* jmp %g1+%lo(addr)
* nop
* nop
* nop
* nop
* nop
*
*/
where[2] = JMP | LOVAL(target);
flush(where, 8);
where[1] = SETHI | HIVAL(target, 10);
flush(where, 4);
} else if (target <= 0 && target > -(1L<<32)) {
/*
* We're withing 32-bits of address -1.
*
* The resulting code in the jump slot is:
*
* sethi %hi(. - .PLT0), %g1
* sethi %hix(addr), %g1
* xor %g1, %lox(addr), %g1
* jmp %g1
* nop
* nop
* nop
* nop
*
*/
where[3] = JMP;
flush(where, 12);
where[2] = XOR | ((~target) & 0x00001fff);
flush(where, 8);
where[1] = SETHI | HIVAL(~target, 10);
flush(where, 4);
} else if (offset <= (1L<<32) && offset >= -((1L<<32) - 4)) {
/*
* We're withing 32-bits -- we can use a direct call insn
*
* The resulting code in the jump slot is:
*
* sethi %hi(. - .PLT0), %g1
* mov %o7, %g1
* call (.+offset)
* mov %g1, %o7
* nop
* nop
* nop
* nop
*
*/
where[3] = MOV17;
flush(where, 12);
where[2] = CALL | ((offset >> 4) & 0x3fffffff);
flush(where, 8);
where[1] = MOV71;
flush(where, 4);
} else if (offset >= 0 && offset < (1L<<44)) {
/*
* We're withing 44 bits. We can generate this pattern:
*
* The resulting code in the jump slot is:
*
* sethi %hi(. - .PLT0), %g1
* sethi %h44(addr), %g1
* or %g1, %m44(addr), %g1
* sllx %g1, 12, %g1
* jmp %g1+%l44(addr)
* nop
* nop
* nop
*
*/
where[4] = JMP | LOVAL(offset);
flush(where, 16);
where[3] = SLLX | 12;
flush(where, 12);
where[2] = OR | (((offset) >> 12) & 0x00001fff);
flush(where, 8);
where[1] = SETHI | HIVAL(offset, 22);
flush(where, 4);
} else if (offset < 0 && offset > -(1L<<44)) {
/*
* We're withing 44 bits. We can generate this pattern:
*
* The resulting code in the jump slot is:
*
* sethi %hi(. - .PLT0), %g1
* sethi %h44(-addr), %g1
* xor %g1, %m44(-addr), %g1
* sllx %g1, 12, %g1
* jmp %g1+%l44(addr)
* nop
* nop
* nop
*
*/
where[4] = JMP | LOVAL(offset);
flush(where, 16);
where[3] = SLLX | 12;
flush(where, 12);
where[2] = XOR | (((~offset) >> 12) & 0x00001fff);
flush(where, 8);
where[1] = SETHI | HIVAL(~offset, 22);
flush(where, 4);
} else {
/*
* We need to load all 64-bits
*
* The resulting code in the jump slot is:
*
* sethi %hi(. - .PLT0), %g1
* sethi %hh(addr), %g1
* sethi %lm(addr), %g5
* or %g1, %hm(addr), %g1
* sllx %g1, 32, %g1
* or %g1, %g5, %g1
* jmp %g1+%lo(addr)
* nop
*
*/
where[6] = JMP | LOVAL(target);
flush(where, 24);
where[5] = ORG5;
flush(where, 20);
where[4] = SLLX | 12;
flush(where, 16);
where[3] = OR | LOVAL((target) >> 32);
flush(where, 12);
where[2] = SETHIG5 | HIVAL(target, 10);
flush(where, 8);
where[1] = SETHI | HIVAL(target, 42);
flush(where, 4);
}
return (target);
}
/*
* Install rtld function call into this PLT slot.
*/
#define SAVE 0x9de3bf50
#define SETHI_l0 0x21000000
#define SETHI_l1 0x23000000
#define OR_l0_l0 0xa0142000
#define SLLX_l0_32_l0 0xa12c3020
#define OR_l0_l1_l0 0xa0140011
#define JMPL_l0_o1 0x93c42000
#define MOV_g1_o0 0x90100001
void
init_pltgot(Obj_Entry *obj)
{
Elf_Half *entry;
if (obj->pltgot != NULL) {
entry = (Elf_Half *)obj->pltgot;
install_plt(&entry[0], (Elf_Addr)_rtld_bind_start_0);
install_plt(&entry[8], (Elf_Addr)_rtld_bind_start_1);
obj->pltgot[8] = (Elf_Addr)obj;
}
}
static void
install_plt(Elf_Half *pltgot, Elf_Addr proc)
{
pltgot[0] = SAVE;
flush(pltgot, 0);
pltgot[1] = SETHI_l0 | HIVAL(proc, 42);
flush(pltgot, 4);
pltgot[2] = SETHI_l1 | HIVAL(proc, 10);
flush(pltgot, 8);
pltgot[3] = OR_l0_l0 | LOVAL((proc) >> 32);
flush(pltgot, 12);
pltgot[4] = SLLX_l0_32_l0;
flush(pltgot, 16);
pltgot[5] = OR_l0_l1_l0;
flush(pltgot, 20);
pltgot[6] = JMPL_l0_o1 | LOVAL(proc);
flush(pltgot, 24);
pltgot[7] = MOV_g1_o0;
flush(pltgot, 28);
}

60
libexec/rtld-elf/sparc64/rtld_machdep.h Normal file
View File

@ -0,0 +1,60 @@
/*-
* Copyright (c) 1999, 2000 John D. Polstra.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#ifndef RTLD_MACHDEP_H
#define RTLD_MACHDEP_H 1
#include <machine/atomic.h>
#define atomic_incr_int(p) atomic_add_int((p), 1)
#define atomic_decr_int(p) atomic_subtract_int((p), 1)
/*
* This value of CACHE_LINE_SIZE is conservative. The actual size
* is 32 on the 21064, 21064A, 21066, 21066A, and 21164. It is 64
* on the 21264. Compaq recommends sequestering each lock in its own
* 128-byte block to allow for future implementations with larger
* cache lines.
*/
#define CACHE_LINE_SIZE 128
struct Struct_Obj_Entry;
/* Return the address of the .dynamic section in the dynamic linker. */
#define rtld_dynamic(obj) ((Elf_Dyn *)(((char *)&_DYNAMIC) + (vm_offset_t)(obj)->relocbase))
Elf_Addr reloc_jmpslot(Elf_Addr *, Elf_Addr,
const struct Struct_Obj_Entry *obj);
#define make_function_pointer(def, defobj) \
((defobj)->relocbase + (def)->st_value)
#define call_initfini_pointer(obj, target) \
(((InitFunc)(target))())
#endif

143
libexec/rtld-elf/sparc64/rtld_start.S Normal file
View File

@ -0,0 +1,143 @@
/* $NetBSD: rtld_start.S,v 1.5 2001/08/14 22:17:48 eeh Exp $ */
/*-
* Copyright (c) 2001 Jake Burkholder.
* Copyright (c) 2000 Eduardo Horvath.
* Copyright (c) 1999 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Christos Zoulas and Paul Kranenburg.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <machine/asm.h>
/*
* ELF:
* On startup the stack should contain 16 extended word register save
* area, followed by the arg count, etc.
*/
ENTRY(.rtld_start)
clr %fp
mov %o0, %l0
mov %o3, %l1
sub %sp, 16, %sp
add %sp, SPOFF + CCFSZ + 0x0, %o1
call _rtld
add %sp, SPOFF + CCFSZ + 0x8, %o2
ldx [%sp + SPOFF + CCFSZ + 0x0], %o1
ldx [%sp + SPOFF + CCFSZ + 0x8], %o2
add %sp, 16, %sp
mov %l1, %o3
jmp %o0
mov %l0, %o0
/*
* We have two separate entry points to the runtime linker.
* I'm implementing this following the SPARC v9 ABI spec.
*
* _rtld_bind_start_0(x, y) is called from .PLT0, and is used for
* PLT entries above 32768.
*
* _rtld_bind_start_1(x, y) is called from .PLT1, and is used for
* PLT entries below 32768.
*
* The first two entries of PLT2 contain the xword object pointer.
*
* These routines are called with two longword arguments,
* x and y. To calculate the address of the entry,
* _rtld_bind_start_1(x, y) does:
*
* n = x >> 15;
*
* and _rtld_bind_start_0(x, y) does:
*
* i = x - y + 1048596;
* n = 32768 + (i/5120)*160 + (i%5120)/24;
*
* Neither routine needs to issue a save since it's already been
* done in the PLT entry.
*/
ENTRY(_rtld_bind_start_0)
sethi %hi(1048596), %l1
sub %o0, %o1, %l0 /* x - y */
or %l1, %lo(1048596), %l1
add %l0, %l1, %l0 /* x - y + 1048596 */
sdivx %l0, 5120, %l1 /* Calculate i/5120 */
ldx [%o1 + (10*4)], %o0 /* Load object pointer from PLT2 */
sub %l0, %l1, %l2 /* And i%5120 */
/* Let the division churn for a bit. */
sdivx %l2, 14, %l4 /* (i%5120)/24 */
/* 160 is (32 * 5) or (32 * (4 + 1)) */
sllx %l1, 2, %l3 /* 4 * (i/5120) */
add %l1, %l3, %l3 /* 5 * (i/5120) */
sllx %l3, 5, %l3 /* 32 * 5 * (i/5120) */
sethi %hi(32768), %l6
add %l3, %l4, %l5 /* %l5 = (i/5120)*160 + (i%5120)/24; */
add %l5, %l6, %l5
sllx %l5, 1, %l6 /* Each element is an Elf_Rela which */
add %l6, %l5, %l5 /* is 3 longwords or 24 bytes. */
sllx %l5, 3, %l5 /* So multiply by 24. */
call _rtld_bind /* Call _rtld_bind(obj, offset) */
mov %l5, %o1
jmp %o0 /* return value == function address */
restore /* Dump our stack frame */
ENTRY(_rtld_bind_start_1)
srax %o0, 15, %o2 /* %o0 is the index to our PLT slot */
sub %o2, 4, %o2 /* XXX: 4 entries are reserved */
sllx %o2, 1, %o3 /* Each element is an Elf_Rela which */
add %o3, %o2, %o2 /* is 3 longwords or 24 bytes. */
sllx %o2, 3, %o2 /* So multiply by 24. */
ldx [%o1 + 8], %o0 /* The object pointer is at [%o1 + 8] */
call _rtld_bind /* Call _rtld_bind(obj, offset) */
mov %o2, %o1
jmp %o0 /* return value == function address */
restore /* Dump our stack frame */