/* Machine-dependent ELF dynamic relocation inline functions.  Sparc64 version.
   Copyright (C) 1997,1998,1999,2000,2001,2002 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C 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.

   The GNU C 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 the GNU C Library; if not, write to the Free
   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
   02111-1307 USA.  */

#define ELF_MACHINE_NAME "sparc64"

#include <string.h>
#include <sys/param.h>
#include <ldsodefs.h>
#include <sysdep.h>

#ifndef VALIDX
# define VALIDX(tag) (DT_NUM + DT_THISPROCNUM + DT_VERSIONTAGNUM \
		      + DT_EXTRANUM + DT_VALTAGIDX (tag))
#endif

#define ELF64_R_TYPE_ID(info)	((info) & 0xff)
#define ELF64_R_TYPE_DATA(info) ((info) >> 8)

/* Return nonzero iff ELF header is compatible with the running host.  */
static inline int
elf_machine_matches_host (const Elf64_Ehdr *ehdr)
{
  return ehdr->e_machine == EM_SPARCV9;
}

/* Return the link-time address of _DYNAMIC.  Conveniently, this is the
   first element of the GOT.  This must be inlined in a function which
   uses global data.  */
static inline Elf64_Addr
elf_machine_dynamic (void)
{
  register Elf64_Addr *elf_pic_register __asm__("%l7");

  return *elf_pic_register;
}

/* Return the run-time load address of the shared object.  */
static inline Elf64_Addr
elf_machine_load_address (void)
{
  register Elf64_Addr *elf_pic_register __asm__("%l7");

  /* We used to utilize the fact that a local .got entry will
     be partially initialized at startup awaiting its RELATIVE
     fixup:

     Elf64_Addr pc, la;

     __asm("sethi %%hi(.Load_address), %1\n"
	   ".Load_address:\n\t"
	   "rd %%pc, %0\n\t"
	   "or %1, %%lo(.Load_address), %1\n\t"
	   : "=r"(pc), "=r"(la));

     return pc - *(Elf64_Addr *)(elf_pic_register + la);

     Unfortunately as binutils tries to work around Solaris
     dynamic linker bug which resolves R_SPARC_RELATIVE as X += B + A
     instead of X = B + A this does not work any longer, since ld
     clears it.

     The following method relies on the fact that sparcv9 ABI maximal
     page length is 1MB and all ELF segments on sparc64 are aligned
     to 1MB.  Also, it relies on _DYNAMIC coming after _GLOBAL_OFFSET_TABLE_
     and assumes that they both fit into the first 1MB of the RW segment.
     This should be true for some time unless ld.so grows too much, at the
     moment the whole stripped ld.so is 128KB and only smaller part of that
     is in the RW segment.  */

  return ((Elf64_Addr)elf_pic_register - *elf_pic_register + 0xfffff)
	 & ~0xfffffUL;
}

/* We have 4 cases to handle.  And we code different code sequences
   for each one.  I love V9 code models...  */
static inline void
sparc64_fixup_plt (struct link_map *map, const Elf64_Rela *reloc,
		   Elf64_Addr *reloc_addr, Elf64_Addr value,
		   Elf64_Addr high, int t)
{
  unsigned int *insns = (unsigned int *) reloc_addr;
  Elf64_Addr plt_vaddr = (Elf64_Addr) reloc_addr;
  Elf64_Sxword disp = value - plt_vaddr;

  /* Now move plt_vaddr up to the call instruction.  */
  plt_vaddr += ((t + 1) * 4);

  /* PLT entries .PLT32768 and above look always the same.  */
  if (__builtin_expect (high, 0) != 0)
    {
      *reloc_addr = value - map->l_addr;
    }
  /* Near destination.  */
  else if (disp >= -0x800000 && disp < 0x800000)
    {
      /* As this is just one instruction, it is thread safe and so
	 we can avoid the unnecessary sethi FOO, %g1.
	 b,a target  */
      insns[0] = 0x30800000 | ((disp >> 2) & 0x3fffff);
      __asm __volatile ("flush %0" : : "r" (insns));
    }
  /* 32-bit Sparc style, the target is in the lower 32-bits of
     address space.  */
  else if (insns += t, (value >> 32) == 0)
    {
      /* sethi	%hi(target), %g1
	 jmpl	%g1 + %lo(target), %g0  */

      insns[1] = 0x81c06000 | (value & 0x3ff);
      __asm __volatile ("flush %0 + 4" : : "r" (insns));

      insns[0] = 0x03000000 | ((unsigned int)(value >> 10));
      __asm __volatile ("flush %0" : : "r" (insns));
    }
  /* We can also get somewhat simple sequences if the distance between
     the target and the PLT entry is within +/- 2GB.  */
  else if ((plt_vaddr > value
	    && ((plt_vaddr - value) >> 31) == 0)
	   || (value > plt_vaddr
	       && ((value - plt_vaddr) >> 31) == 0))
    {
      unsigned int displacement;

      if (plt_vaddr > value)
	displacement = (0 - (plt_vaddr - value));
      else
	displacement = value - plt_vaddr;

      /* mov	%o7, %g1
	 call	displacement
	  mov	%g1, %o7  */

      insns[2] = 0x9e100001;
      __asm __volatile ("flush %0 + 8" : : "r" (insns));

      insns[1] = 0x40000000 | (displacement >> 2);
      __asm __volatile ("flush %0 + 4" : : "r" (insns));

      insns[0] = 0x8210000f;
      __asm __volatile ("flush %0" : : "r" (insns));
    }
  /* Worst case, ho hum...  */
  else
    {
      unsigned int high32 = (value >> 32);
      unsigned int low32 = (unsigned int) value;

      /* ??? Some tricks can be stolen from the sparc64 egcs backend
	     constant formation code I wrote.  -DaveM  */

      if (__builtin_expect (high32 & 0x3ff, 0))
	{
	  /* sethi	%hh(value), %g1
	     sethi	%lm(value), %g5
	     or		%g1, %hm(value), %g1
	     or		%g5, %lo(value), %g5
	     sllx	%g1, 32, %g1
	     jmpl	%g1 + %g5, %g0
	      nop  */

	  insns[5] = 0x81c04005;
	  __asm __volatile ("flush %0 + 20" : : "r" (insns));

	  insns[4] = 0x83287020;
	  __asm __volatile ("flush %0 + 16" : : "r" (insns));

	  insns[3] = 0x8a116000 | (low32 & 0x3ff);
	  __asm __volatile ("flush %0 + 12" : : "r" (insns));

	  insns[2] = 0x82106000 | (high32 & 0x3ff);
	}
      else
	{
	  /* sethi	%hh(value), %g1
	     sethi	%lm(value), %g5
	     sllx	%g1, 32, %g1
	     or		%g5, %lo(value), %g5
	     jmpl	%g1 + %g5, %g0
	      nop  */

	  insns[4] = 0x81c04005;
	  __asm __volatile ("flush %0 + 16" : : "r" (insns));

	  insns[3] = 0x8a116000 | (low32 & 0x3ff);
	  __asm __volatile ("flush %0 + 12" : : "r" (insns));

	  insns[2] = 0x83287020;
	}

      __asm __volatile ("flush %0 + 8" : : "r" (insns));

      insns[1] = 0x0b000000 | (low32 >> 10);
      __asm __volatile ("flush %0 + 4" : : "r" (insns));

      insns[0] = 0x03000000 | (high32 >> 10);
      __asm __volatile ("flush %0" : : "r" (insns));
    }
}

static inline Elf64_Addr
elf_machine_fixup_plt (struct link_map *map, lookup_t t,
		       const Elf64_Rela *reloc,
		       Elf64_Addr *reloc_addr, Elf64_Addr value)
{
  sparc64_fixup_plt (map, reloc, reloc_addr, value + reloc->r_addend,
		     reloc->r_addend, 1);
  return value;
}

/* Return the final value of a plt relocation.  */
static inline Elf64_Addr
elf_machine_plt_value (struct link_map *map, const Elf64_Rela *reloc,
		       Elf64_Addr value)
{
  /* Don't add addend here, but in elf_machine_fixup_plt instead.
     value + reloc->r_addend is the value which should actually be
     stored into .plt data slot.  */
  return value;
}

#ifdef RESOLVE

/* Perform the relocation specified by RELOC and SYM (which is fully resolved).
   MAP is the object containing the reloc.  */

static inline void
elf_machine_rela (struct link_map *map, const Elf64_Rela *reloc,
		  const Elf64_Sym *sym, const struct r_found_version *version,
		  Elf64_Addr *const reloc_addr)
{
  const unsigned long int r_type = ELF64_R_TYPE_ID (reloc->r_info);

#if !defined RTLD_BOOTSTRAP || !defined HAVE_Z_COMBRELOC
  if (__builtin_expect (r_type == R_SPARC_RELATIVE, 0))
    *reloc_addr = map->l_addr + reloc->r_addend;
# ifndef RTLD_BOOTSTRAP
  else if (r_type == R_SPARC_NONE) /* Who is Wilbur? */
    return;
# endif
  else
#endif
    {
#if !defined RTLD_BOOTSTRAP && !defined RESOLVE_CONFLICT_FIND_MAP
      const Elf64_Sym *const refsym = sym;
#endif
      Elf64_Addr value;
#ifndef RESOLVE_CONFLICT_FIND_MAP
      if (sym->st_shndx != SHN_UNDEF &&
	  ELF64_ST_BIND (sym->st_info) == STB_LOCAL)
	value = map->l_addr;
      else
	{
	  value = RESOLVE (&sym, version, r_type);
	  if (sym)
	    value += sym->st_value;
	}
#else
      value = 0;
#endif
      value += reloc->r_addend;	/* Assume copy relocs have zero addend.  */

      switch (r_type)
	{
#if !defined RTLD_BOOTSTRAP && !defined RESOLVE_CONFLICT_FIND_MAP
	case R_SPARC_COPY:
	  if (sym == NULL)
	    /* This can happen in trace mode if an object could not be
	       found.  */
	    break;
	  if (sym->st_size > refsym->st_size
	      || (GL(dl_verbose) && sym->st_size < refsym->st_size))
	    {
	      const char *strtab;

	      strtab = (const void *) D_PTR (map, l_info[DT_STRTAB]);
	      _dl_error_printf ("\
%s: Symbol `%s' has different size in shared object, consider re-linking\n",
				rtld_progname ?: "<program name unknown>",
				strtab + refsym->st_name);
	    }
	  memcpy (reloc_addr, (void *) value, MIN (sym->st_size,
						   refsym->st_size));
	  break;
#endif
	case R_SPARC_64:
	case R_SPARC_GLOB_DAT:
	  *reloc_addr = value;
	  break;
#ifndef RTLD_BOOTSTRAP
	case R_SPARC_8:
	  *(char *) reloc_addr = value;
	  break;
	case R_SPARC_16:
	  *(short *) reloc_addr = value;
	  break;
	case R_SPARC_32:
	  *(unsigned int *) reloc_addr = value;
	  break;
	case R_SPARC_DISP8:
	  *(char *) reloc_addr = (value - (Elf64_Addr) reloc_addr);
	  break;
	case R_SPARC_DISP16:
	  *(short *) reloc_addr = (value - (Elf64_Addr) reloc_addr);
	  break;
	case R_SPARC_DISP32:
	  *(unsigned int *) reloc_addr = (value - (Elf64_Addr) reloc_addr);
	  break;
	case R_SPARC_WDISP30:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & 0xc0000000) |
	     ((value - (Elf64_Addr) reloc_addr) >> 2));
	  break;

	/* MEDLOW code model relocs */
	case R_SPARC_LO10:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & ~0x3ff) |
	     (value & 0x3ff));
	  break;
	case R_SPARC_HI22:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & 0xffc00000) |
	     (value >> 10));
	  break;
	case R_SPARC_OLO10:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & ~0x1fff) |
	     (((value & 0x3ff) + ELF64_R_TYPE_DATA (reloc->r_info)) & 0x1fff));
	  break;

	/* MEDMID code model relocs */
	case R_SPARC_H44:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & 0xffc00000) |
	     (value >> 22));
	  break;
	case R_SPARC_M44:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & ~0x3ff) |
	     ((value >> 12) & 0x3ff));
	  break;
	case R_SPARC_L44:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & ~0xfff) |
	     (value & 0xfff));
	  break;

	/* MEDANY code model relocs */
	case R_SPARC_HH22:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & 0xffc00000) |
	     (value >> 42));
	  break;
	case R_SPARC_HM10:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & ~0x3ff) |
	     ((value >> 32) & 0x3ff));
	  break;
	case R_SPARC_LM22:
	  *(unsigned int *) reloc_addr =
	    ((*(unsigned int *)reloc_addr & 0xffc00000) |
	     ((value >> 10) & 0x003fffff));
	  break;
#endif
	case R_SPARC_JMP_SLOT:
#ifdef RESOLVE_CONFLICT_FIND_MAP
	  /* R_SPARC_JMP_SLOT conflicts against .plt[32768+]
	     relocs should be turned into R_SPARC_64 relocs
	     in .gnu.conflict section.
	     r_addend non-zero does not mean it is a .plt[32768+]
	     reloc, instead it is the actual address of the function
	     to call.  */
	  sparc64_fixup_plt (NULL, reloc, reloc_addr, value, 0, 0);
#else
	  sparc64_fixup_plt (map, reloc, reloc_addr, value,
			     reloc->r_addend, 0);
#endif
	  break;
#ifndef RTLD_BOOTSTRAP
	case R_SPARC_UA16:
	  ((unsigned char *) reloc_addr) [0] = value >> 8;
	  ((unsigned char *) reloc_addr) [1] = value;
	  break;
	case R_SPARC_UA32:
	  ((unsigned char *) reloc_addr) [0] = value >> 24;
	  ((unsigned char *) reloc_addr) [1] = value >> 16;
	  ((unsigned char *) reloc_addr) [2] = value >> 8;
	  ((unsigned char *) reloc_addr) [3] = value;
	  break;
	case R_SPARC_UA64:
	  if (! ((long) reloc_addr & 3))
	    {
	      /* Common in .eh_frame */
	      ((unsigned int *) reloc_addr) [0] = value >> 32;
	      ((unsigned int *) reloc_addr) [1] = value;
	      break;
	    }
	  ((unsigned char *) reloc_addr) [0] = value >> 56;
	  ((unsigned char *) reloc_addr) [1] = value >> 48;
	  ((unsigned char *) reloc_addr) [2] = value >> 40;
	  ((unsigned char *) reloc_addr) [3] = value >> 32;
	  ((unsigned char *) reloc_addr) [4] = value >> 24;
	  ((unsigned char *) reloc_addr) [5] = value >> 16;
	  ((unsigned char *) reloc_addr) [6] = value >> 8;
	  ((unsigned char *) reloc_addr) [7] = value;
	  break;
#endif
#if !defined RTLD_BOOTSTRAP || defined _NDEBUG
	default:
	  _dl_reloc_bad_type (map, r_type, 0);
	  break;
#endif
	}
    }
}

static inline void
elf_machine_rela_relative (Elf64_Addr l_addr, const Elf64_Rela *reloc,
			   Elf64_Addr *const reloc_addr)
{
  *reloc_addr = l_addr + reloc->r_addend;
}

static inline void
elf_machine_lazy_rel (struct link_map *map,
		      Elf64_Addr l_addr, const Elf64_Rela *reloc)
{
  switch (ELF64_R_TYPE (reloc->r_info))
    {
    case R_SPARC_NONE:
      break;
    case R_SPARC_JMP_SLOT:
      break;
    default:
      _dl_reloc_bad_type (map, ELFW(R_TYPE) (reloc->r_info), 1);
      break;
    }
}

#endif	/* RESOLVE */

/* ELF_RTYPE_CLASS_PLT iff TYPE describes relocation of a PLT entry, so
   PLT entries should not be allowed to define the value.
   ELF_RTYPE_CLASS_NOCOPY iff TYPE should not be allowed to resolve to one
   of the main executable's symbols, as for a COPY reloc.  */
#define elf_machine_type_class(type) \
  ((((type) == R_SPARC_JMP_SLOT) * ELF_RTYPE_CLASS_PLT)	\
   | (((type) == R_SPARC_COPY) * ELF_RTYPE_CLASS_COPY))

/* A reloc type used for ld.so cmdline arg lookups to reject PLT entries.  */
#define ELF_MACHINE_JMP_SLOT	R_SPARC_JMP_SLOT

/* The SPARC never uses Elf64_Rel relocations.  */
#define ELF_MACHINE_NO_REL 1

/* The SPARC overlaps DT_RELA and DT_PLTREL.  */
#define ELF_MACHINE_PLTREL_OVERLAP 1

/* Set up the loaded object described by L so its unrelocated PLT
   entries will jump to the on-demand fixup code in dl-runtime.c.  */

static inline int
elf_machine_runtime_setup (struct link_map *l, int lazy, int profile)
{
  if (l->l_info[DT_JMPREL] && lazy)
    {
      extern void _dl_runtime_resolve_0 (void);
      extern void _dl_runtime_resolve_1 (void);
      extern void _dl_runtime_profile_0 (void);
      extern void _dl_runtime_profile_1 (void);
      Elf64_Addr res0_addr, res1_addr;
      unsigned int *plt = (void *) D_PTR (l, l_info[DT_PLTGOT]);
      int i = 0;

      if (! profile)
	{
	  res0_addr = (Elf64_Addr) &_dl_runtime_resolve_0;
	  res1_addr = (Elf64_Addr) &_dl_runtime_resolve_1;
	}
      else
	{
	  res0_addr = (Elf64_Addr) &_dl_runtime_profile_0;
	  res1_addr = (Elf64_Addr) &_dl_runtime_profile_1;
	  if (_dl_name_match_p (GL(dl_profile), l))
	    GL(dl_profile_map) = l;
	}

      /* PLT0 looks like:

	 save	%sp, -192, %sp
	 sethi	%hh(_dl_runtime_{resolve,profile}_0), %l0
	 sethi	%lm(_dl_runtime_{resolve,profile}_0), %l1
	 or	%l0, %hm(_dl_runtime_{resolve,profile}_0), %l0
	 or	%l1, %lo(_dl_runtime_{resolve,profile}_0), %l1
	 sllx	%l0, 32, %l0
	 jmpl	%l0 + %l1, %l6
	  sethi	%hi(0xffc00), %l2
       */

      plt[0] = 0x9de3bf40;
      plt[1] = 0x21000000 | (res0_addr >> (64 - 22));
      plt[2] = 0x23000000 | ((res0_addr >> 10) & 0x003fffff);
      plt[3] = 0xa0142000 | ((res0_addr >> 32) & 0x3ff);
      plt[4] = 0xa2146000 | (res0_addr & 0x3ff);
      plt[5] = 0xa12c3020;
      plt[6] = 0xadc40011;
      plt[7] = 0x250003ff;

      /* PLT1 looks like:

	 save	%sp, -192, %sp
	 sethi	%hh(_dl_runtime_{resolve,profile}_1), %l0
	 sethi	%lm(_dl_runtime_{resolve,profile}_1), %l1
	 or	%l0, %hm(_dl_runtime_{resolve,profile}_1), %l0
	 or	%l1, %lo(_dl_runtime_{resolve,profile}_1), %l1
	 sllx	%l0, 32, %l0
	 jmpl	%l0 + %l1, %l6
	  srlx	%g1, 12, %o1
       */

      plt[8 + 0] = 0x9de3bf40;
      if (__builtin_expect (((res1_addr + 4) >> 32) & 0x3ff, 0))
	i = 1;
      else
	res1_addr += 4;
      plt[8 + 1] = 0x21000000 | (res1_addr >> (64 - 22));
      plt[8 + 2] = 0x23000000 | ((res1_addr >> 10) & 0x003fffff);
      if (__builtin_expect (i, 0))
	plt[8 + 3] = 0xa0142000 | ((res1_addr >> 32) & 0x3ff);
      else
	plt[8 + 3] = 0xa12c3020;
      plt[8 + 4] = 0xa2146000 | (res1_addr & 0x3ff);
      if (__builtin_expect (i, 0))
	plt[8 + 5] = 0xa12c3020;
      plt[8 + 5 + i] = 0xadc40011;
      plt[8 + 6 + i] = 0x9330700c;

      /* Now put the magic cookie at the beginning of .PLT2
	 Entry .PLT3 is unused by this implementation.  */
      *((struct link_map **)(&plt[16 + 0])) = l;

      if (__builtin_expect (l->l_info[VALIDX(DT_GNU_PRELINKED)] != NULL, 0)
	  || __builtin_expect (l->l_info [VALIDX (DT_GNU_LIBLISTSZ)] != NULL, 0))
	{
	  /* Need to reinitialize .plt to undo prelinking.  */
	  Elf64_Rela *rela = (Elf64_Rela *) D_PTR (l, l_info[DT_JMPREL]);
	  Elf64_Rela *relaend
	    = (Elf64_Rela *) ((char *) rela
			      + l->l_info[DT_PLTRELSZ]->d_un.d_val);

	  /* prelink must ensure there are no R_SPARC_NONE relocs left
	     in .rela.plt.  */
	  while (rela < relaend)
	    {
	      if (__builtin_expect (rela->r_addend, 0) != 0)
		{
                  Elf64_Addr slot = ((rela->r_offset + 0x400
				      - (Elf64_Addr) plt)
				     / 0x1400) * 0x1400
				    + (Elf64_Addr) plt - 0x400;
		  /* ldx [%o7 + X], %g1  */
		  unsigned int first_ldx = *(unsigned int *)(slot + 12);
		  Elf64_Addr ptr = slot + (first_ldx & 0xfff) + 4;

		  *(Elf64_Addr *) rela->r_offset
		    = (Elf64_Addr) plt
		      - (slot + ((rela->r_offset - ptr) / 8) * 24 + 4);
		  ++rela;
		  continue;
		}

	      *(unsigned int *) rela->r_offset
		= 0x03000000 | (rela->r_offset - (Elf64_Addr) plt);
	      *(unsigned int *) (rela->r_offset + 4)
		= 0x30680000 | ((((Elf64_Addr) plt + 32
				  - rela->r_offset - 4) >> 2) & 0x7ffff);
	      __asm __volatile ("flush %0" : : "r" (rela->r_offset));
	      __asm __volatile ("flush %0+4" : : "r" (rela->r_offset));
	      ++rela;
	    }
	}
    }

  return lazy;
}

/* This code is used in dl-runtime.c to call the `fixup' function
   and then redirect to the address it returns.  */
#define TRAMPOLINE_TEMPLATE(tramp_name, fixup_name)	\
  asm ("\n"						\
"	.text\n"					\
"	.globl	" #tramp_name "_0\n"			\
"	.type	" #tramp_name "_0, @function\n"		\
"	.align	32\n"					\
"\t" #tramp_name "_0:\n"				\
"	! sethi   %hi(1047552), %l2 - Done in .PLT0\n"	\
"	ldx	[%l6 + 32 + 8], %o0\n"			\
"	sub     %g1, %l6, %l0\n"			\
"	xor     %l2, -1016, %l2\n"			\
"	sethi   %hi(5120), %l3	! 160 * 32\n"		\
"	add     %l0, %l2, %l0\n"			\
"	sethi   %hi(32768), %l4\n"			\
"	udivx   %l0, %l3, %l3\n"			\
"	sllx    %l3, 2, %l1\n"				\
"	add     %l1, %l3, %l1\n"			\
"	sllx    %l1, 10, %l2\n"				\
"	sub	%l4, 4, %l4	! No thanks to Sun for not obeying their own ABI\n" \
"	sllx    %l1, 5, %l1\n"				\
"	sub     %l0, %l2, %l0\n"			\
"	udivx   %l0, 24, %l0\n"				\
"	add     %l0, %l4, %l0\n"			\
"	add     %l1, %l0, %l1\n"			\
"	add     %l1, %l1, %l0\n"			\
"	add     %l0, %l1, %l0\n"			\
"	mov	%i7, %o2\n"				\
"	call	" #fixup_name "\n"			\
"	 sllx    %l0, 3, %o1\n"				\
"	jmp	%o0\n"					\
"	 restore\n"					\
"	.size	" #tramp_name "_0, . - " #tramp_name "_0\n" \
"\n"							\
"	.globl	" #tramp_name "_1\n"			\
"	.type	" #tramp_name "_1, @function\n"		\
"	! tramp_name_1 + 4 needs to be .align 32\n"	\
"\t" #tramp_name "_1:\n"				\
"	sub	%l6, 4, %l6\n"				\
"	! srlx	%g1, 12, %o1 - Done in .PLT1\n"		\
"	ldx	[%l6 + 12], %o0\n"			\
"	add	%o1, %o1, %o3\n"			\
"	sub	%o1, 96, %o1	! No thanks to Sun for not obeying their own ABI\n" \
"	mov	%i7, %o2\n"				\
"	call	" #fixup_name "\n"			\
"	 add	%o1, %o3, %o1\n"			\
"	jmp	%o0\n"					\
"	 restore\n"					\
"	.size	" #tramp_name "_1, . - " #tramp_name "_1\n" \
"	.previous\n");

#ifndef PROF
#define ELF_MACHINE_RUNTIME_TRAMPOLINE			\
  TRAMPOLINE_TEMPLATE (_dl_runtime_resolve, fixup);	\
  TRAMPOLINE_TEMPLATE (_dl_runtime_profile, profile_fixup);
#else
#define ELF_MACHINE_RUNTIME_TRAMPOLINE			\
  TRAMPOLINE_TEMPLATE (_dl_runtime_resolve, fixup);	\
  TRAMPOLINE_TEMPLATE (_dl_runtime_profile, fixup);
#endif

/* The PLT uses Elf64_Rela relocs.  */
#define elf_machine_relplt elf_machine_rela

/* Initial entry point code for the dynamic linker.
   The C function `_dl_start' is the real entry point;
   its return value is the user program's entry point.  */

#define __S1(x)	#x
#define __S(x)	__S1(x)

#define RTLD_START __asm__ ( "\n"					\
"	.text\n"							\
"	.global	_start\n"						\
"	.type	_start, @function\n"					\
"	.align	32\n"							\
"_start:\n"								\
"   /* Make room for functions to drop their arguments on the stack.  */\n" \
"	sub	%sp, 6*8, %sp\n"					\
"   /* Pass pointer to argument block to _dl_start.  */\n"		\
"	call	_dl_start\n"						\
"	 add	 %sp," __S(STACK_BIAS) "+22*8,%o0\n"			\
"	/* FALLTHRU */\n"						\
"	.size _start, .-_start\n"					\
"\n"									\
"	.global	_dl_start_user\n"					\
"	.type	_dl_start_user, @function\n"				\
"_dl_start_user:\n"							\
"   /* Load the GOT register.  */\n"					\
"1:	call	11f\n"							\
"	 sethi	%hi(_GLOBAL_OFFSET_TABLE_-(1b-.)), %l7\n"		\
"11:	or	%l7, %lo(_GLOBAL_OFFSET_TABLE_-(1b-.)), %l7\n"		\
"  /* Store the highest stack address.  */\n"				\
"	sethi	%hi(__libc_stack_end), %g5\n"				\
"	add	%l7, %o7, %l7\n"					\
"	or	%g5, %lo(__libc_stack_end), %g5\n"			\
"   /* Save the user entry point address in %l0.  */\n"			\
"	mov	%o0, %l0\n"						\
"	ldx	[%l7 + %g5], %l1\n"					\
"	sethi	%hi(_dl_skip_args), %g5\n"				\
"	add	%sp, 6*8, %l2\n"					\
"   /* See if we were run as a command with the executable file name as an\n" \
"      extra leading argument.  If so, we must shift things around since we\n" \
"      must keep the stack doubleword aligned.  */\n"			\
"	or	%g5, %lo(_dl_skip_args), %g5\n"				\
"	stx	%l2, [%l1]\n"						\
"	ldx	[%l7 + %g5], %i0\n"					\
"	ld	[%i0], %i0\n"						\
"	brz,pt	%i0, 2f\n"						\
"	 ldx	[%sp + " __S(STACK_BIAS) " + 22*8], %i5\n"		\
"	/* Find out how far to shift.  */\n"				\
"	sethi	%hi(_dl_argv), %l4\n"					\
"	sub	%i5, %i0, %i5\n"					\
"	or	%l4, %lo(_dl_argv), %l4\n"				\
"	sllx	%i0, 3, %l6\n"						\
"	ldx	[%l7 + %l4], %l4\n"					\
"	stx	%i5, [%sp + " __S(STACK_BIAS) " + 22*8]\n"		\
"	add	%sp, " __S(STACK_BIAS) " + 23*8, %i1\n"			\
"	add	%i1, %l6, %i2\n"					\
"	ldx	[%l4], %l5\n"						\
"	/* Copy down argv.  */\n"					\
"12:	ldx	[%i2], %i3\n"						\
"	add	%i2, 8, %i2\n"						\
"	stx	%i3, [%i1]\n"						\
"	brnz,pt	%i3, 12b\n"						\
"	 add	%i1, 8, %i1\n"						\
"	sub	%l5, %l6, %l5\n"					\
"	/* Copy down envp.  */\n"					\
"13:	ldx	[%i2], %i3\n"						\
"	add	%i2, 8, %i2\n"						\
"	stx	%i3, [%i1]\n"						\
"	brnz,pt	%i3, 13b\n"						\
"	 add	%i1, 8, %i1\n"						\
"	/* Copy down auxiliary table.  */\n"				\
"14:	ldx	[%i2], %i3\n"						\
"	ldx	[%i2 + 8], %i4\n"					\
"	add	%i2, 16, %i2\n"						\
"	stx	%i3, [%i1]\n"						\
"	stx	%i4, [%i1 + 8]\n"					\
"	brnz,pt	%i3, 14b\n"						\
"	 add	%i1, 16, %i1\n"						\
"	stx	%l5, [%l4]\n"						\
"  /* %o0 = _dl_loaded, %o1 = argc, %o2 = argv, %o3 = envp.  */\n"	\
"2:	sethi	%hi(_rtld_local), %o0\n"				\
"	add	%sp, " __S(STACK_BIAS) " + 23*8, %o2\n"			\
"	orcc	%o0, %lo(_rtld_local), %o0\n"				\
"	sllx	%i5, 3, %o3\n"						\
"	ldx	[%l7 + %o0], %o0\n"					\
"	add	%o3, 8, %o3\n"						\
"	mov	%i5, %o1\n"						\
"	add	%o2, %o3, %o3\n"					\
"	call	_dl_init_internal\n"					\
"	 ldx	[%o0], %o0\n"						\
"   /* Pass our finalizer function to the user in %g1.  */\n"		\
"	sethi	%hi(_dl_fini), %g1\n"					\
"	or	%g1, %lo(_dl_fini), %g1\n"				\
"	ldx	[%l7 + %g1], %g1\n"					\
"  /* Jump to the user's entry point and deallocate the extra stack we got.  */\n" \
"	jmp	%l0\n"							\
"	 add	%sp, 6*8, %sp\n"					\
"	.size	_dl_start_user, . - _dl_start_user\n"			\
"	.previous\n");