gdb/sparc-tdep.c

gdb/sparc-tdep.c - gdb

Source code

/* Target-dependent code for SPARC.



   Copyright (C) 2003-2015 Free Software Foundation, Inc.



   This file is part of GDB.



   This program is free software; you can redistribute it and/or modify

   it under the terms of the GNU General Public License as published by

   the Free Software Foundation; either version 3 of the License, or

   (at your option) any later version.



   This program 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 General Public License for more details.



   You should have received a copy of the GNU General Public License

   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */



#include "defs.h"

#include "arch-utils.h"

#include "dis-asm.h"

#include "dwarf2-frame.h"

#include "floatformat.h"

#include "frame.h"

#include "frame-base.h"

#include "frame-unwind.h"

#include "gdbcore.h"

#include "gdbtypes.h"

#include "inferior.h"

#include "symtab.h"

#include "objfiles.h"

#include "osabi.h"

#include "regcache.h"

#include "target.h"

#include "value.h"



#include "sparc-tdep.h"

#include "sparc-ravenscar-thread.h"



struct regset;



/* This file implements the SPARC 32-bit ABI as defined by the section

   "Low-Level System Information" of the SPARC Compliance Definition

   (SCD) 2.4.1, which is the 32-bit System V psABI for SPARC.  The SCD

   lists changes with respect to the original 32-bit psABI as defined

   in the "System V ABI, SPARC Processor Supplement".



   Note that if we talk about SunOS, we mean SunOS 4.x, which was

   BSD-based, which is sometimes (retroactively?) referred to as

   Solaris 1.x.  If we talk about Solaris we mean Solaris 2.x and

   above (Solaris 7, 8 and 9 are nothing but Solaris 2.7, 2.8 and 2.9

   suffering from severe version number inflation).  Solaris 2.x is

   also known as SunOS 5.x, since that's what uname(1) says.  Solaris

   2.x is SVR4-based.  */



/* Please use the sparc32_-prefix for 32-bit specific code, the

   sparc64_-prefix for 64-bit specific code and the sparc_-prefix for

   code that can handle both.  The 64-bit specific code lives in

   sparc64-tdep.c; don't add any here.  */



/* The SPARC Floating-Point Quad-Precision format is similar to

   big-endian IA-64 Quad-Precision format.  */

‌#define floatformats_sparc_quad floatformats_ia64_quad



/* The stack pointer is offset from the stack frame by a BIAS of 2047

   (0x7ff) for 64-bit code.  BIAS is likely to be defined on SPARC

   hosts, so undefine it first.  */

‌#undef BIAS

‌#define BIAS 2047



/* Macros to extract fields from SPARC instructions.  */

‌#define X_OP(i) (((i) >> 30) & 0x3)

‌#define X_RD(i) (((i) >> 25) & 0x1f)

‌#define X_A(i) (((i) >> 29) & 1)

‌#define X_COND(i) (((i) >> 25) & 0xf)

‌#define X_OP2(i) (((i) >> 22) & 0x7)

‌#define X_IMM22(i) ((i) & 0x3fffff)

‌#define X_OP3(i) (((i) >> 19) & 0x3f)

‌#define X_RS1(i) (((i) >> 14) & 0x1f)

‌#define X_RS2(i) ((i) & 0x1f)

‌#define X_I(i) (((i) >> 13) & 1)

/* Sign extension macros.  */

‌#define X_DISP22(i) ((X_IMM22 (i) ^ 0x200000) - 0x200000)

‌#define X_DISP19(i) ((((i) & 0x7ffff) ^ 0x40000) - 0x40000)

‌#define X_DISP10(i) ((((((i) >> 11) && 0x300) | (((i) >> 5) & 0xff)) ^ 0x200) - 0x200)

‌#define X_SIMM13(i) ((((i) & 0x1fff) ^ 0x1000) - 0x1000)

/* Macros to identify some instructions.  */

/* RETURN (RETT in V8) */

‌#define X_RETTURN(i) ((X_OP (i) == 0x2) && (X_OP3 (i) == 0x39))



/* Fetch the instruction at PC.  Instructions are always big-endian

   even if the processor operates in little-endian mode.  */



unsigned long

‌sparc_fetch_instruction (CORE_ADDR pc)

{

  gdb_byte buf[4];

  unsigned long insn;

  int i;



  /* If we can't read the instruction at PC, return zero.  */

  if (target_read_memory (pc, buf, sizeof (buf)))

    return 0;



  insn = 0;

  for (i = 0; i < sizeof (buf); i++)

    insn = (insn << 8) | buf[i];

  return insn;

}





/* Return non-zero if the instruction corresponding to PC is an "unimp"

   instruction.  */



static int

‌sparc_is_unimp_insn (CORE_ADDR pc)

{

  const unsigned long insn = sparc_fetch_instruction (pc);



  return ((insn & 0xc1c00000) == 0);

}



/* Return non-zero if the instruction corresponding to PC is an

   "annulled" branch, i.e. the annul bit is set.  */



int

‌sparc_is_annulled_branch_insn (CORE_ADDR pc)

{

  /* The branch instructions featuring an annul bit can be identified

     by the following bit patterns:



     OP=0

      OP2=1: Branch on Integer Condition Codes with Prediction (BPcc).

      OP2=2: Branch on Integer Condition Codes (Bcc).

      OP2=5: Branch on FP Condition Codes with Prediction (FBfcc).

      OP2=6: Branch on FP Condition Codes (FBcc).

      OP2=3 && Bit28=0:

             Branch on Integer Register with Prediction (BPr).



     This leaves out ILLTRAP (OP2=0), SETHI/NOP (OP2=4) and the V8

     coprocessor branch instructions (Op2=7).  */



  const unsigned long insn = sparc_fetch_instruction (pc);

  const unsigned op2 = X_OP2 (insn);



  if ((X_OP (insn) == 0)

      && ((op2 == 1) || (op2 == 2) || (op2 == 5) || (op2 == 6)

          || ((op2 == 3) && ((insn & 0x10000000) == 0))))

    return X_A (insn);

  else

    return 0;

}



/* OpenBSD/sparc includes StackGhost, which according to the author's

   website http://stackghost.cerias.purdue.edu "... transparently and

   automatically protects applications' stack frames; more

   specifically, it guards the return pointers.  The protection

   mechanisms require no application source or binary modification and

   imposes only a negligible performance penalty."



   The same website provides the following description of how

   StackGhost works:



   "StackGhost interfaces with the kernel trap handler that would

   normally write out registers to the stack and the handler that

   would read them back in.  By XORing a cookie into the

   return-address saved in the user stack when it is actually written

   to the stack, and then XOR it out when the return-address is pulled

   from the stack, StackGhost can cause attacker corrupted return

   pointers to behave in a manner the attacker cannot predict.

   StackGhost can also use several unused bits in the return pointer

   to detect a smashed return pointer and abort the process."



   For GDB this means that whenever we're reading %i7 from a stack

   frame's window save area, we'll have to XOR the cookie.



   More information on StackGuard can be found on in:



   Mike Frantzen and Mike Shuey.  "StackGhost: Hardware Facilitated

   Stack Protection."  2001.  Published in USENIX Security Symposium

   '01.  */



/* Fetch StackGhost Per-Process XOR cookie.  */



ULONGEST

‌sparc_fetch_wcookie (struct gdbarch *gdbarch)

{

  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);

  struct target_ops *ops = &current_target;

  gdb_byte buf[8];

  int len;



  len = target_read (ops, TARGET_OBJECT_WCOOKIE, NULL, buf, 0, 8);

  if (len == -1)

    return 0;



  /* We should have either an 32-bit or an 64-bit cookie.  */

  gdb_assert (len == 4 || len == 8);



  return extract_unsigned_integer (buf, len, byte_order);

}





/* The functions on this page are intended to be used to classify

   function arguments.  */



/* Check whether TYPE is "Integral or Pointer".  */



static int

‌sparc_integral_or_pointer_p (const struct type *type)

{

  int len = TYPE_LENGTH (type);



  switch (TYPE_CODE (type))

    {

    case TYPE_CODE_INT:

    case TYPE_CODE_BOOL:

    case TYPE_CODE_CHAR:

    case TYPE_CODE_ENUM:

    case TYPE_CODE_RANGE:

      /* We have byte, half-word, word and extended-word/doubleword

         integral types.  The doubleword is an extension to the

         original 32-bit ABI by the SCD 2.4.x.  */

      return (len == 1 || len == 2 || len == 4 || len == 8);

    case TYPE_CODE_PTR:

    case TYPE_CODE_REF:

      /* Allow either 32-bit or 64-bit pointers.  */

      return (len == 4 || len == 8);

    default:

      break;

    }



  return 0;

}



/* Check whether TYPE is "Floating".  */



static int

‌sparc_floating_p (const struct type *type)

{

  switch (TYPE_CODE (type))

    {

    case TYPE_CODE_FLT:

      {

        int len = TYPE_LENGTH (type);

        return (len == 4 || len == 8 || len == 16);

      }

    default:

      break;

    }



  return 0;

}



/* Check whether TYPE is "Complex Floating".  */



static int

‌sparc_complex_floating_p (const struct type *type)

{

  switch (TYPE_CODE (type))

    {

    case TYPE_CODE_COMPLEX:

      {

        int len = TYPE_LENGTH (type);

        return (len == 8 || len == 16 || len == 32);

      }

    default:

      break;

    }



  return 0;

}



/* Check whether TYPE is "Structure or Union".



   In terms of Ada subprogram calls, arrays are treated the same as

   struct and union types.  So this function also returns non-zero

   for array types.  */



static int

‌sparc_structure_or_union_p (const struct type *type)

{

  switch (TYPE_CODE (type))

    {

    case TYPE_CODE_STRUCT:

    case TYPE_CODE_UNION:

    case TYPE_CODE_ARRAY:

      return 1;

    default:

      break;

    }



  return 0;

}



/* Register information.  */



‌static const char *sparc32_register_names[] =

{

  "g0", "g1", "g2", "g3", "g4", "g5", "g6", "g7",

  "o0", "o1", "o2", "o3", "o4", "o5", "sp", "o7",

  "l0", "l1", "l2", "l3", "l4", "l5", "l6", "l7",

  "i0", "i1", "i2", "i3", "i4", "i5", "fp", "i7",



  "f0", "f1", "f2", "f3", "f4", "f5", "f6", "f7",

  "f8", "f9", "f10", "f11", "f12", "f13", "f14", "f15",

  "f16", "f17", "f18", "f19", "f20", "f21", "f22", "f23",

  "f24", "f25", "f26", "f27", "f28", "f29", "f30", "f31",



  "y", "psr", "wim", "tbr", "pc", "npc", "fsr", "csr"

};



/* Total number of registers.  */

‌#define SPARC32_NUM_REGS ARRAY_SIZE (sparc32_register_names)



/* We provide the aliases %d0..%d30 for the floating registers as

   "psuedo" registers.  */



‌static const char *sparc32_pseudo_register_names[] =

{

  "d0", "d2", "d4", "d6", "d8", "d10", "d12", "d14",

  "d16", "d18", "d20", "d22", "d24", "d26", "d28", "d30"

};



/* Total number of pseudo registers.  */

‌#define SPARC32_NUM_PSEUDO_REGS ARRAY_SIZE (sparc32_pseudo_register_names)



/* Return the name of register REGNUM.  */



static const char *

‌sparc32_register_name (struct gdbarch *gdbarch, int regnum)

{

  if (regnum >= 0 && regnum < SPARC32_NUM_REGS)

    return sparc32_register_names[regnum];



  if (regnum < SPARC32_NUM_REGS + SPARC32_NUM_PSEUDO_REGS)

    return sparc32_pseudo_register_names[regnum - SPARC32_NUM_REGS];



  return NULL;

}



/* Construct types for ISA-specific registers.  */



static struct type *

‌sparc_psr_type (struct gdbarch *gdbarch)

{

  struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);



  if (!tdep->sparc_psr_type)

    {

      struct type *type;



      type = arch_flags_type (gdbarch, "builtin_type_sparc_psr", 4);

      append_flags_type_flag (type, 5, "ET");

      append_flags_type_flag (type, 6, "PS");

      append_flags_type_flag (type, 7, "S");

      append_flags_type_flag (type, 12, "EF");

      append_flags_type_flag (type, 13, "EC");



      tdep->sparc_psr_type = type;

    }



  return tdep->sparc_psr_type;

}



static struct type *

‌sparc_fsr_type (struct gdbarch *gdbarch)

{

  struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);



  if (!tdep->sparc_fsr_type)

    {

      struct type *type;



      type = arch_flags_type (gdbarch, "builtin_type_sparc_fsr", 4);

      append_flags_type_flag (type, 0, "NXA");

      append_flags_type_flag (type, 1, "DZA");

      append_flags_type_flag (type, 2, "UFA");

      append_flags_type_flag (type, 3, "OFA");

      append_flags_type_flag (type, 4, "NVA");

      append_flags_type_flag (type, 5, "NXC");

      append_flags_type_flag (type, 6, "DZC");

      append_flags_type_flag (type, 7, "UFC");

      append_flags_type_flag (type, 8, "OFC");

      append_flags_type_flag (type, 9, "NVC");

      append_flags_type_flag (type, 22, "NS");

      append_flags_type_flag (type, 23, "NXM");

      append_flags_type_flag (type, 24, "DZM");

      append_flags_type_flag (type, 25, "UFM");

      append_flags_type_flag (type, 26, "OFM");

      append_flags_type_flag (type, 27, "NVM");



      tdep->sparc_fsr_type = type;

    }



  return tdep->sparc_fsr_type;

}



/* Return the GDB type object for the "standard" data type of data in

   register REGNUM.  */



static struct type *

‌sparc32_register_type (struct gdbarch *gdbarch, int regnum)

{

  if (regnum >= SPARC_F0_REGNUM && regnum <= SPARC_F31_REGNUM)

    return builtin_type (gdbarch)->builtin_float;



  if (regnum >= SPARC32_D0_REGNUM && regnum <= SPARC32_D30_REGNUM)

    return builtin_type (gdbarch)->builtin_double;



  if (regnum == SPARC_SP_REGNUM || regnum == SPARC_FP_REGNUM)

    return builtin_type (gdbarch)->builtin_data_ptr;



  if (regnum == SPARC32_PC_REGNUM || regnum == SPARC32_NPC_REGNUM)

    return builtin_type (gdbarch)->builtin_func_ptr;



  if (regnum == SPARC32_PSR_REGNUM)

    return sparc_psr_type (gdbarch);



  if (regnum == SPARC32_FSR_REGNUM)

    return sparc_fsr_type (gdbarch);



  return builtin_type (gdbarch)->builtin_int32;

}



static enum register_status

‌sparc32_pseudo_register_read (struct gdbarch *gdbarch,

                              struct regcache *regcache,

                              int regnum, gdb_byte *buf)

{

  enum register_status status;



  gdb_assert (regnum >= SPARC32_D0_REGNUM && regnum <= SPARC32_D30_REGNUM);



  regnum = SPARC_F0_REGNUM + 2 * (regnum - SPARC32_D0_REGNUM);

  status = regcache_raw_read (regcache, regnum, buf);

  if (status == REG_VALID)

    status = regcache_raw_read (regcache, regnum + 1, buf + 4);

  return status;

}



static void

‌sparc32_pseudo_register_write (struct gdbarch *gdbarch,

                               struct regcache *regcache,

                               int regnum, const gdb_byte *buf)

{

  gdb_assert (regnum >= SPARC32_D0_REGNUM && regnum <= SPARC32_D30_REGNUM);



  regnum = SPARC_F0_REGNUM + 2 * (regnum - SPARC32_D0_REGNUM);

  regcache_raw_write (regcache, regnum, buf);

  regcache_raw_write (regcache, regnum + 1, buf + 4);

}



/* Implement "in_function_epilogue_p".  */



int

‌sparc_in_function_epilogue_p (struct gdbarch *gdbarch, CORE_ADDR pc)

{

  /* This function must return true if we are one instruction after an

     instruction that destroyed the stack frame of the current

     function.  The SPARC instructions used to restore the callers

     stack frame are RESTORE and RETURN/RETT.



     Of these RETURN/RETT is a branch instruction and thus we return

     true if we are in its delay slot.



     RESTORE is almost always found in the delay slot of a branch

     instruction that transfers control to the caller, such as JMPL.

     Thus the next instruction is in the caller frame and we don't

     need to do anything about it.  */



  unsigned int insn = sparc_fetch_instruction (pc - 4);



  return X_RETTURN (insn);

}





static CORE_ADDR

‌sparc32_frame_align (struct gdbarch *gdbarch, CORE_ADDR address)

{

  /* The ABI requires double-word alignment.  */

  return address & ~0x7;

}



static CORE_ADDR

‌sparc32_push_dummy_code (struct gdbarch *gdbarch, CORE_ADDR sp,

                         CORE_ADDR funcaddr,

                         struct value **args, int nargs,

                         struct type *value_type,

                         CORE_ADDR *real_pc, CORE_ADDR *bp_addr,

                         struct regcache *regcache)

{

  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);



  *bp_addr = sp - 4;

  *real_pc = funcaddr;



  if (using_struct_return (gdbarch, NULL, value_type))

    {

      gdb_byte buf[4];



      /* This is an UNIMP instruction.  */

      store_unsigned_integer (buf, 4, byte_order,

                              TYPE_LENGTH (value_type) & 0x1fff);

      write_memory (sp - 8, buf, 4);

      return sp - 8;

    }



  return sp - 4;

}



static CORE_ADDR

‌sparc32_store_arguments (struct regcache *regcache, int nargs,

                         struct value **args, CORE_ADDR sp,

                         int struct_return, CORE_ADDR struct_addr)

{

  struct gdbarch *gdbarch = get_regcache_arch (regcache);

  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);

  /* Number of words in the "parameter array".  */

  int num_elements = 0;

  int element = 0;

  int i;



  for (i = 0; i < nargs; i++)

    {

      struct type *type = value_type (args[i]);

      int len = TYPE_LENGTH (type);



      if (sparc_structure_or_union_p (type)

          || (sparc_floating_p (type) && len == 16)

          || sparc_complex_floating_p (type))

        {

          /* Structure, Union and Quad-Precision Arguments.  */

          sp -= len;



          /* Use doubleword alignment for these values.  That's always

             correct, and wasting a few bytes shouldn't be a problem.  */

          sp &= ~0x7;



          write_memory (sp, value_contents (args[i]), len);

          args[i] = value_from_pointer (lookup_pointer_type (type), sp);

          num_elements++;

        }

      else if (sparc_floating_p (type))

        {

          /* Floating arguments.  */

          gdb_assert (len == 4 || len == 8);

          num_elements += (len / 4);

        }

      else

        {

          /* Integral and pointer arguments.  */

          gdb_assert (sparc_integral_or_pointer_p (type));



          if (len < 4)

            args[i] = value_cast (builtin_type (gdbarch)->builtin_int32,

                                  args[i]);

          num_elements += ((len + 3) / 4);

        }

    }



  /* Always allocate at least six words.  */

  sp -= max (6, num_elements) * 4;



  /* The psABI says that "Software convention requires space for the

     struct/union return value pointer, even if the word is unused."  */

  sp -= 4;



  /* The psABI says that "Although software convention and the

     operating system require every stack frame to be doubleword

     aligned."  */

  sp &= ~0x7;



  for (i = 0; i < nargs; i++)

    {

      const bfd_byte *valbuf = value_contents (args[i]);

      struct type *type = value_type (args[i]);

      int len = TYPE_LENGTH (type);



      gdb_assert (len == 4 || len == 8);



      if (element < 6)

        {

          int regnum = SPARC_O0_REGNUM + element;



          regcache_cooked_write (regcache, regnum, valbuf);

          if (len > 4 && element < 5)

            regcache_cooked_write (regcache, regnum + 1, valbuf + 4);

        }



      /* Always store the argument in memory.  */

      write_memory (sp + 4 + element * 4, valbuf, len);

      element += len / 4;

    }



  gdb_assert (element == num_elements);



  if (struct_return)

    {

      gdb_byte buf[4];



      store_unsigned_integer (buf, 4, byte_order, struct_addr);

      write_memory (sp, buf, 4);

    }



  return sp;

}



static CORE_ADDR

‌sparc32_push_dummy_call (struct gdbarch *gdbarch, struct value *function,

                         struct regcache *regcache, CORE_ADDR bp_addr,

                         int nargs, struct value **args, CORE_ADDR sp,

                         int struct_return, CORE_ADDR struct_addr)

{

  CORE_ADDR call_pc = (struct_return ? (bp_addr - 12) : (bp_addr - 8));



  /* Set return address.  */

  regcache_cooked_write_unsigned (regcache, SPARC_O7_REGNUM, call_pc);



  /* Set up function arguments.  */

  sp = sparc32_store_arguments (regcache, nargs, args, sp,

                                struct_return, struct_addr);



  /* Allocate the 16-word window save area.  */

  sp -= 16 * 4;



  /* Stack should be doubleword aligned at this point.  */

  gdb_assert (sp % 8 == 0);



  /* Finally, update the stack pointer.  */

  regcache_cooked_write_unsigned (regcache, SPARC_SP_REGNUM, sp);



  return sp;

}





/* Use the program counter to determine the contents and size of a

   breakpoint instruction.  Return a pointer to a string of bytes that

   encode a breakpoint instruction, store the length of the string in

   *LEN and optionally adjust *PC to point to the correct memory

   location for inserting the breakpoint.  */



static const gdb_byte *

‌sparc_breakpoint_from_pc (struct gdbarch *gdbarch, CORE_ADDR *pc, int *len)

{

  static const gdb_byte break_insn[] = { 0x91, 0xd0, 0x20, 0x01 };



  *len = sizeof (break_insn);

  return break_insn;

}





/* Allocate and initialize a frame cache.  */



static struct sparc_frame_cache *

‌sparc_alloc_frame_cache (void)

{

  struct sparc_frame_cache *cache;



  cache = FRAME_OBSTACK_ZALLOC (struct sparc_frame_cache);



  /* Base address.  */

  cache->base = 0;

  cache->pc = 0;



  /* Frameless until proven otherwise.  */

  cache->frameless_p = 1;

  cache->frame_offset = 0;

  cache->saved_regs_mask = 0;

  cache->copied_regs_mask = 0;

  cache->struct_return_p = 0;



  return cache;

}



/* GCC generates several well-known sequences of instructions at the begining

   of each function prologue when compiling with -fstack-check.  If one of

   such sequences starts at START_PC, then return the address of the

   instruction immediately past this sequence.  Otherwise, return START_PC.  */



static CORE_ADDR

‌sparc_skip_stack_check (const CORE_ADDR start_pc)

{

  CORE_ADDR pc = start_pc;

  unsigned long insn;

  int offset_stack_checking_sequence = 0;

  int probing_loop = 0;



  /* With GCC, all stack checking sequences begin with the same two

     instructions, plus an optional one in the case of a probing loop:



         sethi <some immediate>, %g1

         sub %sp, %g1, %g1



     or:



         sethi <some immediate>, %g1

         sethi <some immediate>, %g4

         sub %sp, %g1, %g1



     or:



         sethi <some immediate>, %g1

         sub %sp, %g1, %g1

         sethi <some immediate>, %g4



     If the optional instruction is found (setting g4), assume that a

     probing loop will follow.  */



  /* sethi <some immediate>, %g1 */

  insn = sparc_fetch_instruction (pc);

  pc = pc + 4;

  if (!(X_OP (insn) == 0 && X_OP2 (insn) == 0x4 && X_RD (insn) == 1))

    return start_pc;



  /* optional: sethi <some immediate>, %g4 */

  insn = sparc_fetch_instruction (pc);

  pc = pc + 4;

  if (X_OP (insn) == 0 && X_OP2 (insn) == 0x4 && X_RD (insn) == 4)

    {

      probing_loop = 1;

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

    }



  /* sub %sp, %g1, %g1 */

  if (!(X_OP (insn) == 2 && X_OP3 (insn) == 0x4 && !X_I(insn)

        && X_RD (insn) == 1 && X_RS1 (insn) == 14 && X_RS2 (insn) == 1))

    return start_pc;



  insn = sparc_fetch_instruction (pc);

  pc = pc + 4;



  /* optional: sethi <some immediate>, %g4 */

  if (X_OP (insn) == 0 && X_OP2 (insn) == 0x4 && X_RD (insn) == 4)

    {

      probing_loop = 1;

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

    }



  /* First possible sequence:

         [first two instructions above]

         clr [%g1 - some immediate]  */



  /* clr [%g1 - some immediate]  */

  if (X_OP (insn) == 3 && X_OP3(insn) == 0x4 && X_I(insn)

      && X_RS1 (insn) == 1 && X_RD (insn) == 0)

    {

      /* Valid stack-check sequence, return the new PC.  */

      return pc;

    }



  /* Second possible sequence: A small number of probes.

         [first two instructions above]

         clr [%g1]

         add   %g1, -<some immediate>, %g1

         clr [%g1]

         [repeat the two instructions above any (small) number of times]

         clr [%g1 - some immediate]  */



  /* clr [%g1] */

  else if (X_OP (insn) == 3 && X_OP3(insn) == 0x4 && !X_I(insn)

      && X_RS1 (insn) == 1 && X_RD (insn) == 0)

    {

      while (1)

        {

          /* add %g1, -<some immediate>, %g1 */

          insn = sparc_fetch_instruction (pc);

          pc = pc + 4;

          if (!(X_OP (insn) == 2  && X_OP3(insn) == 0 && X_I(insn)

                && X_RS1 (insn) == 1 && X_RD (insn) == 1))

            break;



          /* clr [%g1] */

          insn = sparc_fetch_instruction (pc);

          pc = pc + 4;

          if (!(X_OP (insn) == 3 && X_OP3(insn) == 0x4 && !X_I(insn)

                && X_RD (insn) == 0 && X_RS1 (insn) == 1))

            return start_pc;

        }



      /* clr [%g1 - some immediate] */

      if (!(X_OP (insn) == 3 && X_OP3(insn) == 0x4 && X_I(insn)

            && X_RS1 (insn) == 1 && X_RD (insn) == 0))

        return start_pc;



      /* We found a valid stack-check sequence, return the new PC.  */

      return pc;

    }



  /* Third sequence: A probing loop.

         [first three instructions above]

         sub  %g1, %g4, %g4

         cmp  %g1, %g4

         be  <disp>

         add  %g1, -<some immediate>, %g1

         ba  <disp>

         clr  [%g1]



     And an optional last probe for the remainder:



         clr [%g4 - some immediate]  */



  if (probing_loop)

    {

      /* sub  %g1, %g4, %g4 */

      if (!(X_OP (insn) == 2 && X_OP3 (insn) == 0x4 && !X_I(insn)

            && X_RD (insn) == 4 && X_RS1 (insn) == 1 && X_RS2 (insn) == 4))

        return start_pc;



      /* cmp  %g1, %g4 */

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

      if (!(X_OP (insn) == 2 && X_OP3 (insn) == 0x14 && !X_I(insn)

            && X_RD (insn) == 0 && X_RS1 (insn) == 1 && X_RS2 (insn) == 4))

        return start_pc;



      /* be  <disp> */

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

      if (!(X_OP (insn) == 0 && X_COND (insn) == 0x1))

        return start_pc;



      /* add  %g1, -<some immediate>, %g1 */

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

      if (!(X_OP (insn) == 2  && X_OP3(insn) == 0 && X_I(insn)

            && X_RS1 (insn) == 1 && X_RD (insn) == 1))

        return start_pc;



      /* ba  <disp> */

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

      if (!(X_OP (insn) == 0 && X_COND (insn) == 0x8))

        return start_pc;



      /* clr  [%g1] (st %g0, [%g1] or st %g0, [%g1+0]) */

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

      if (!(X_OP (insn) == 3 && X_OP3(insn) == 0x4

            && X_RD (insn) == 0 && X_RS1 (insn) == 1

            && (!X_I(insn) || X_SIMM13 (insn) == 0)))

        return start_pc;



      /* We found a valid stack-check sequence, return the new PC.  */



      /* optional: clr [%g4 - some immediate]  */

      insn = sparc_fetch_instruction (pc);

      pc = pc + 4;

      if (!(X_OP (insn) == 3 && X_OP3(insn) == 0x4 && X_I(insn)

            && X_RS1 (insn) == 4 && X_RD (insn) == 0))

        return pc - 4;

      else

        return pc;

    }



  /* No stack check code in our prologue, return the start_pc.  */

  return start_pc;

}



/* Record the effect of a SAVE instruction on CACHE.  */



void

‌sparc_record_save_insn (struct sparc_frame_cache *cache)

{

  /* The frame is set up.  */

  cache->frameless_p = 0;



  /* The frame pointer contains the CFA.  */

  cache->frame_offset = 0;



  /* The `local' and `in' registers are all saved.  */

  cache->saved_regs_mask = 0xffff;



  /* The `out' registers are all renamed.  */

  cache->copied_regs_mask = 0xff;

}



/* Do a full analysis of the prologue at PC and update CACHE accordingly.

   Bail out early if CURRENT_PC is reached.  Return the address where

   the analysis stopped.



   We handle both the traditional register window model and the single

   register window (aka flat) model.  */



CORE_ADDR

‌sparc_analyze_prologue (struct gdbarch *gdbarch, CORE_ADDR pc,

                        CORE_ADDR current_pc, struct sparc_frame_cache *cache)

{

  struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);

  unsigned long insn;

  int offset = 0;

  int dest = -1;



  pc = sparc_skip_stack_check (pc);



  if (current_pc <= pc)

    return current_pc;



  /* We have to handle to "Procedure Linkage Table" (PLT) special.  On

     SPARC the linker usually defines a symbol (typically

     _PROCEDURE_LINKAGE_TABLE_) at the start of the .plt section.

     This symbol makes us end up here with PC pointing at the start of

     the PLT and CURRENT_PC probably pointing at a PLT entry.  If we

     would do our normal prologue analysis, we would probably conclude

     that we've got a frame when in reality we don't, since the

     dynamic linker patches up the first PLT with some code that

     starts with a SAVE instruction.  Patch up PC such that it points

     at the start of our PLT entry.  */

  if (tdep->plt_entry_size > 0 && in_plt_section (current_pc))

    pc = current_pc - ((current_pc - pc) % tdep->plt_entry_size);



  insn = sparc_fetch_instruction (pc);



  /* Recognize store insns and record their sources.  */

  while (X_OP (insn) == 3

         && (X_OP3 (insn) == 0x4     /* stw */

             || X_OP3 (insn) == 0x7  /* std */

             || X_OP3 (insn) == 0xe) /* stx */

         && X_RS1 (insn) == SPARC_SP_REGNUM)

    {

      int regnum = X_RD (insn);



      /* Recognize stores into the corresponding stack slots.  */

      if (regnum >= SPARC_L0_REGNUM && regnum <= SPARC_I7_REGNUM

          && ((X_I (insn)

               && X_SIMM13 (insn) == (X_OP3 (insn) == 0xe

                                      ? (regnum - SPARC_L0_REGNUM) * 8 + BIAS

                                      : (regnum - SPARC_L0_REGNUM) * 4))

              || (!X_I (insn) && regnum == SPARC_L0_REGNUM)))

        {

          cache->saved_regs_mask |= (1 << (regnum - SPARC_L0_REGNUM));

          if (X_OP3 (insn) == 0x7)

            cache->saved_regs_mask |= (1 << (regnum + 1 - SPARC_L0_REGNUM));

        }



      offset += 4;



      insn = sparc_fetch_instruction (pc + offset);

    }



  /* Recognize a SETHI insn and record its destination.  */

  if (X_OP (insn) == 0 && X_OP2 (insn) == 0x04)

    {

      dest = X_RD (insn);

      offset += 4;



      insn = sparc_fetch_instruction (pc + offset);

    }



  /* Allow for an arithmetic operation on DEST or %g1.  */

  if (X_OP (insn) == 2 && X_I (insn)

      && (X_RD (insn) == 1 || X_RD (insn) == dest))

    {

      offset += 4;



      insn = sparc_fetch_instruction (pc + offset);

    }



  /* Check for the SAVE instruction that sets up the frame.  */

  if (X_OP (insn) == 2 && X_OP3 (insn) == 0x3c)

    {

      sparc_record_save_insn (cache);

      offset += 4;

      return pc + offset;

    }



  /* Check for an arithmetic operation on %sp.  */

  if (X_OP (insn) == 2

      && (X_OP3 (insn) == 0 || X_OP3 (insn) == 0x4)

      && X_RS1 (insn) == SPARC_SP_REGNUM

      && X_RD (insn) == SPARC_SP_REGNUM)

    {

      if (X_I (insn))

        {

          cache->frame_offset = X_SIMM13 (insn);

          if (X_OP3 (insn) == 0)

            cache->frame_offset = -cache->frame_offset;

        }

      offset += 4;



      insn = sparc_fetch_instruction (pc + offset);



      /* Check for an arithmetic operation that sets up the frame.  */

      if (X_OP (insn) == 2

          && (X_OP3 (insn) == 0 || X_OP3 (insn) == 0x4)

          && X_RS1 (insn) == SPARC_SP_REGNUM

          && X_RD (insn) == SPARC_FP_REGNUM)

        {

          cache->frameless_p = 0;

          cache->frame_offset = 0;

          /* We could check that the amount subtracted to %sp above is the

             same as the one added here, but this seems superfluous.  */

          cache->copied_regs_mask |= 0x40;

          offset += 4;



          insn = sparc_fetch_instruction (pc + offset);

        }



      /* Check for a move (or) operation that copies the return register.  */

      if (X_OP (insn) == 2

          && X_OP3 (insn) == 0x2

          && !X_I (insn)

          && X_RS1 (insn) == SPARC_G0_REGNUM

          && X_RS2 (insn) == SPARC_O7_REGNUM

          && X_RD (insn) == SPARC_I7_REGNUM)

        {

           cache->copied_regs_mask |= 0x80;

           offset += 4;

        }



      return pc + offset;

    }



  return pc;

}



static CORE_ADDR

‌sparc_unwind_pc (struct gdbarch *gdbarch, struct frame_info *this_frame)

{

  struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);

  return frame_unwind_register_unsigned (this_frame, tdep->pc_regnum);

}



/* Return PC of first real instruction of the function starting at

   START_PC.  */



static CORE_ADDR

‌sparc32_skip_prologue (struct gdbarch *gdbarch, CORE_ADDR start_pc)

{

  struct symtab_and_line sal;

  CORE_ADDR func_start, func_end;

  struct sparc_frame_cache cache;



  /* This is the preferred method, find the end of the prologue by

     using the debugging information.  */

  if (find_pc_partial_function (start_pc, NULL, &func_start, &func_end))

    {

      sal = find_pc_line (func_start, 0);



      if (sal.end < func_end

          && start_pc <= sal.end)

        return sal.end;

    }



  start_pc = sparc_analyze_prologue (gdbarch, start_pc, 0xffffffffUL, &cache);



  /* The psABI says that "Although the first 6 words of arguments

     reside in registers, the standard stack frame reserves space for

     them.".  It also suggests that a function may use that space to

     "write incoming arguments 0 to 5" into that space, and that's

     indeed what GCC seems to be doing.  In that case GCC will

     generate debug information that points to the stack slots instead

     of the registers, so we should consider the instructions that

     write out these incoming arguments onto the stack.  */



  while (1)

    {

      unsigned long insn = sparc_fetch_instruction (start_pc);



      /* Recognize instructions that store incoming arguments into the

         corresponding stack slots.  */

      if (X_OP (insn) == 3 && (X_OP3 (insn) & 0x3c) == 0x04

          && X_I (insn) && X_RS1 (insn) == SPARC_FP_REGNUM)

        {

          int regnum = X_RD (insn);



          /* Case of arguments still in %o[0..5].  */

          if (regnum >= SPARC_O0_REGNUM && regnum <= SPARC_O5_REGNUM

              && !(cache.copied_regs_mask & (1 << (regnum - SPARC_O0_REGNUM)))

              && X_SIMM13 (insn) == 68 + (regnum - SPARC_O0_REGNUM) * 4)

            {

              start_pc += 4;

              continue;

            }



          /* Case of arguments copied into %i[0..5].  */

          if (regnum >= SPARC_I0_REGNUM && regnum <= SPARC_I5_REGNUM

              && (cache.copied_regs_mask & (1 << (regnum - SPARC_I0_REGNUM)))

              && X_SIMM13 (insn) == 68 + (regnum - SPARC_I0_REGNUM) * 4)

            {

              start_pc += 4;

              continue;

            }

        }



      break;

    }



  return start_pc;

}



/* Normal frames.  */



struct sparc_frame_cache *

‌sparc_frame_cache (struct frame_info *this_frame, void **this_cache)

{

  struct sparc_frame_cache *cache;



  if (*this_cache)

    return *this_cache;



  cache = sparc_alloc_frame_cache ();

  *this_cache = cache;



  cache->pc = get_frame_func (this_frame);

  if (cache->pc != 0)

    sparc_analyze_prologue (get_frame_arch (this_frame), cache->pc,

                            get_frame_pc (this_frame), cache);



  if (cache->frameless_p)

    {

      /* This function is frameless, so %fp (%i6) holds the frame

         pointer for our calling frame.  Use %sp (%o6) as this frame's

         base address.  */

      cache->base =

        get_frame_register_unsigned (this_frame, SPARC_SP_REGNUM);

    }

  else

    {

      /* For normal frames, %fp (%i6) holds the frame pointer, the

         base address for the current stack frame.  */

      cache->base =

        get_frame_register_unsigned (this_frame, SPARC_FP_REGNUM);

    }



  cache->base += cache->frame_offset;



  if (cache->base & 1)

    cache->base += BIAS;



  return cache;

}



static int

‌sparc32_struct_return_from_sym (struct symbol *sym)

{

  struct type *type = check_typedef (SYMBOL_TYPE (sym));

  enum type_code code = TYPE_CODE (type);



  if (code == TYPE_CODE_FUNC || code == TYPE_CODE_METHOD)

    {

      type = check_typedef (TYPE_TARGET_TYPE (type));

      if (sparc_structure_or_union_p (type)

          || (sparc_floating_p (type) && TYPE_LENGTH (type) == 16))

        return 1;

    }



  return 0;

}



struct sparc_frame_cache *

‌sparc32_frame_cache (struct frame_info *this_frame, void **this_cache)

{

  struct sparc_frame_cache *cache;

  struct symbol *sym;



  if (*this_cache)

    return *this_cache;



  cache = sparc_frame_cache (this_frame, this_cache);



  sym = find_pc_function (cache->pc);

  if (sym)

    {

      cache->struct_return_p = sparc32_struct_return_from_sym (sym);

    }

  else

    {

      /* There is no debugging information for this function to

         help us determine whether this function returns a struct

         or not.  So we rely on another heuristic which is to check

         the instruction at the return address and see if this is

         an "unimp" instruction.  If it is, then it is a struct-return

         function.  */

      CORE_ADDR pc;

      int regnum =

        (cache->copied_regs_mask & 0x80) ? SPARC_I7_REGNUM : SPARC_O7_REGNUM;



      pc = get_frame_register_unsigned (this_frame, regnum) + 8;

      if (sparc_is_unimp_insn (pc))

        cache->struct_return_p = 1;

    }



  return cache;

}



static void

‌sparc32_frame_this_id (struct frame_info *this_frame, void **this_cache,

                       struct frame_id *this_id)

{

  struct sparc_frame_cache *cache =

    sparc32_frame_cache (this_frame, this_cache);



  /* This marks the outermost frame.  */

  if (cache->base == 0)

    return;



  (*this_id) = frame_id_build (cache->base, cache->pc);

}



static struct value *

‌sparc32_frame_prev_register (struct frame_info *this_frame,

                             void **this_cache, int regnum)

{

  struct gdbarch *gdbarch = get_frame_arch (this_frame);

  struct sparc_frame_cache *cache =

    sparc32_frame_cache (this_frame, this_cache);



  if (regnum == SPARC32_PC_REGNUM || regnum == SPARC32_NPC_REGNUM)

    {

      CORE_ADDR pc = (regnum == SPARC32_NPC_REGNUM) ? 4 : 0;



      /* If this functions has a Structure, Union or Quad-Precision

         return value, we have to skip the UNIMP instruction that encodes

         the size of the structure.  */

      if (cache->struct_return_p)

        pc += 4;



      regnum =

        (cache->copied_regs_mask & 0x80) ? SPARC_I7_REGNUM : SPARC_O7_REGNUM;

      pc += get_frame_register_unsigned (this_frame, regnum) + 8;

      return frame_unwind_got_constant (this_frame, regnum, pc);

    }



  /* Handle StackGhost.  */

  {

    ULONGEST wcookie = sparc_fetch_wcookie (gdbarch);



    if (wcookie != 0 && !cache->frameless_p && regnum == SPARC_I7_REGNUM)

      {

        CORE_ADDR addr = cache->base + (regnum - SPARC_L0_REGNUM) * 4;

        ULONGEST i7;



        /* Read the value in from memory.  */

        i7 = get_frame_memory_unsigned (this_frame, addr, 4);

        return frame_unwind_got_constant (this_frame, regnum, i7 ^ wcookie);

      }

  }



  /* The previous frame's `local' and `in' registers may have been saved

     in the register save area.  */

  if (regnum >= SPARC_L0_REGNUM && regnum <= SPARC_I7_REGNUM

      && (cache->saved_regs_mask & (1 << (regnum - SPARC_L0_REGNUM))))

    {

      CORE_ADDR addr = cache->base + (regnum - SPARC_L0_REGNUM) * 4;



      return frame_unwind_got_memory (this_frame, regnum, addr);

    }



  /* The previous frame's `out' registers may be accessible as the current

     frame's `in' registers.  */

  if (regnum >= SPARC_O0_REGNUM && regnum <= SPARC_O7_REGNUM

      && (cache->copied_regs_mask & (1 << (regnum - SPARC_O0_REGNUM))))

    regnum += (SPARC_I0_REGNUM - SPARC_O0_REGNUM);



  return frame_unwind_got_register (this_frame, regnum, regnum);

}



‌static const struct frame_unwind sparc32_frame_unwind =

{

  NORMAL_FRAME,

  default_frame_unwind_stop_reason,

  sparc32_frame_this_id,

  sparc32_frame_prev_register,

  NULL,

  default_frame_sniffer

};





static CORE_ADDR

‌sparc32_frame_base_address (struct frame_info *this_frame, void **this_cache)

{

  struct sparc_frame_cache *cache =

    sparc32_frame_cache (this_frame, this_cache);



  return cache->base;

}



‌static const struct frame_base sparc32_frame_base =

{

  &sparc32_frame_unwind,

  sparc32_frame_base_address,

  sparc32_frame_base_address,

  sparc32_frame_base_address

};



static struct frame_id

‌sparc_dummy_id (struct gdbarch *gdbarch, struct frame_info *this_frame)

{

  CORE_ADDR sp;



  sp = get_frame_register_unsigned (this_frame, SPARC_SP_REGNUM);

  if (sp & 1)

    sp += BIAS;

  return frame_id_build (sp, get_frame_pc (this_frame));

}





/* Extract a function return value of TYPE from REGCACHE, and copy

   that into VALBUF.  */



static void

‌sparc32_extract_return_value (struct type *type, struct regcache *regcache,

                              gdb_byte *valbuf)

{

  int len = TYPE_LENGTH (type);

  gdb_byte buf[32];



  gdb_assert (!sparc_structure_or_union_p (type));

  gdb_assert (!(sparc_floating_p (type) && len == 16));



  if (sparc_floating_p (type) || sparc_complex_floating_p (type))

    {

      /* Floating return values.  */

      regcache_cooked_read (regcache, SPARC_F0_REGNUM, buf);

      if (len > 4)

        regcache_cooked_read (regcache, SPARC_F1_REGNUM, buf + 4);

      if (len > 8)

        {

          regcache_cooked_read (regcache, SPARC_F2_REGNUM, buf + 8);

          regcache_cooked_read (regcache, SPARC_F3_REGNUM, buf + 12);

        }

      if (len > 16)

        {

          regcache_cooked_read (regcache, SPARC_F4_REGNUM, buf + 16);

          regcache_cooked_read (regcache, SPARC_F5_REGNUM, buf + 20);

          regcache_cooked_read (regcache, SPARC_F6_REGNUM, buf + 24);

          regcache_cooked_read (regcache, SPARC_F7_REGNUM, buf + 28);

        }

      memcpy (valbuf, buf, len);

    }

  else

    {

      /* Integral and pointer return values.  */

      gdb_assert (sparc_integral_or_pointer_p (type));



      regcache_cooked_read (regcache, SPARC_O0_REGNUM, buf);

      if (len > 4)

        {

          regcache_cooked_read (regcache, SPARC_O1_REGNUM, buf + 4);

          gdb_assert (len == 8);

          memcpy (valbuf, buf, 8);

        }

      else

        {

          /* Just stripping off any unused bytes should preserve the

             signed-ness just fine.  */

          memcpy (valbuf, buf + 4 - len, len);

        }

    }

}



/* Store the function return value of type TYPE from VALBUF into

   REGCACHE.  */



static void

‌sparc32_store_return_value (struct type *type, struct regcache *regcache,

                            const gdb_byte *valbuf)

{

  int len = TYPE_LENGTH (type);

  gdb_byte buf[8];



  gdb_assert (!sparc_structure_or_union_p (type));

  gdb_assert (!(sparc_floating_p (type) && len == 16));

  gdb_assert (len <= 8);



  if (sparc_floating_p (type) || sparc_complex_floating_p (type))

    {

      /* Floating return values.  */

      memcpy (buf, valbuf, len);

      regcache_cooked_write (regcache, SPARC_F0_REGNUM, buf);

      if (len > 4)

        regcache_cooked_write (regcache, SPARC_F1_REGNUM, buf + 4);

      if (len > 8)

        {

          regcache_cooked_write (regcache, SPARC_F2_REGNUM, buf + 8);

          regcache_cooked_write (regcache, SPARC_F3_REGNUM, buf + 12);

        }

      if (len > 16)

        {

          regcache_cooked_write (regcache, SPARC_F4_REGNUM, buf + 16);

          regcache_cooked_write (regcache, SPARC_F5_REGNUM, buf + 20);

          regcache_cooked_write (regcache, SPARC_F6_REGNUM, buf + 24);

          regcache_cooked_write (regcache, SPARC_F7_REGNUM, buf + 28);

        }

    }

  else

    {

      /* Integral and pointer return values.  */

      gdb_assert (sparc_integral_or_pointer_p (type));



      if (len > 4)

        {

          gdb_assert (len == 8);

          memcpy (buf, valbuf, 8);

          regcache_cooked_write (regcache, SPARC_O1_REGNUM, buf + 4);

        }

      else

        {

          /* ??? Do we need to do any sign-extension here?  */

          memcpy (buf + 4 - len, valbuf, len);

        }

      regcache_cooked_write (regcache, SPARC_O0_REGNUM, buf);

    }

}



static enum return_value_convention

‌sparc32_return_value (struct gdbarch *gdbarch, struct value *function,

                      struct type *type, struct regcache *regcache,

                      gdb_byte *readbuf, const gdb_byte *writebuf)

{

  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);



  /* The psABI says that "...every stack frame reserves the word at

     %fp+64.  If a function returns a structure, union, or

     quad-precision value, this word should hold the address of the

     object into which the return value should be copied."  This

     guarantees that we can always find the return value, not just

     before the function returns.  */



  if (sparc_structure_or_union_p (type)

      || (sparc_floating_p (type) && TYPE_LENGTH (type) == 16))

    {

      ULONGEST sp;

      CORE_ADDR addr;



      if (readbuf)

        {

          regcache_cooked_read_unsigned (regcache, SPARC_SP_REGNUM, &sp);

          addr = read_memory_unsigned_integer (sp + 64, 4, byte_order);

          read_memory (addr, readbuf, TYPE_LENGTH (type));

        }

      if (writebuf)

        {

          regcache_cooked_read_unsigned (regcache, SPARC_SP_REGNUM, &sp);

          addr = read_memory_unsigned_integer (sp + 64, 4, byte_order);

          write_memory (addr, writebuf, TYPE_LENGTH (type));

        }



      return RETURN_VALUE_ABI_PRESERVES_ADDRESS;

    }



  if (readbuf)

    sparc32_extract_return_value (type, regcache, readbuf);

  if (writebuf)

    sparc32_store_return_value (type, regcache, writebuf);



  return RETURN_VALUE_REGISTER_CONVENTION;

}



static int

‌sparc32_stabs_argument_has_addr (struct gdbarch *gdbarch, struct type *type)

{

  return (sparc_structure_or_union_p (type)

          || (sparc_floating_p (type) && TYPE_LENGTH (type) == 16)

          || sparc_complex_floating_p (type));

}



static int

‌sparc32_dwarf2_struct_return_p (struct frame_info *this_frame)

{

  CORE_ADDR pc = get_frame_address_in_block (this_frame);

  struct symbol *sym = find_pc_function (pc);



  if (sym)

    return sparc32_struct_return_from_sym (sym);

  return 0;

}



static void

‌sparc32_dwarf2_frame_init_reg (struct gdbarch *gdbarch, int regnum,

                               struct dwarf2_frame_state_reg *reg,

                               struct frame_info *this_frame)

{

  int off;



  switch (regnum)

    {

    case SPARC_G0_REGNUM:

      /* Since %g0 is always zero, there is no point in saving it, and

         people will be inclined omit it from the CFI.  Make sure we

         don't warn about that.  */

      reg->how = DWARF2_FRAME_REG_SAME_VALUE;

      break;

    case SPARC_SP_REGNUM:

      reg->how = DWARF2_FRAME_REG_CFA;

      break;

    case SPARC32_PC_REGNUM:

    case SPARC32_NPC_REGNUM:

      reg->how = DWARF2_FRAME_REG_RA_OFFSET;

      off = 8;

      if (sparc32_dwarf2_struct_return_p (this_frame))

        off += 4;

      if (regnum == SPARC32_NPC_REGNUM)

        off += 4;

      reg->loc.offset = off;

      break;

    }

}





/* The SPARC Architecture doesn't have hardware single-step support,

   and most operating systems don't implement it either, so we provide

   software single-step mechanism.  */



static CORE_ADDR

‌sparc_analyze_control_transfer (struct frame_info *frame,

                                CORE_ADDR pc, CORE_ADDR *npc)

{

  unsigned long insn = sparc_fetch_instruction (pc);

  int conditional_p = X_COND (insn) & 0x7;

  int branch_p = 0, fused_p = 0;

  long offset = 0;                        /* Must be signed for sign-extend.  */



  if (X_OP (insn) == 0 && X_OP2 (insn) == 3)

    {

      if ((insn & 0x10000000) == 0)

        {

          /* Branch on Integer Register with Prediction (BPr).  */

          branch_p = 1;

          conditional_p = 1;

        }

      else

        {

          /* Compare and Branch  */

          branch_p = 1;

          fused_p = 1;

          offset = 4 * X_DISP10 (insn);

        }

    }

  else if (X_OP (insn) == 0 && X_OP2 (insn) == 6)

    {

      /* Branch on Floating-Point Condition Codes (FBfcc).  */

      branch_p = 1;

      offset = 4 * X_DISP22 (insn);

    }

  else if (X_OP (insn) == 0 && X_OP2 (insn) == 5)

    {

      /* Branch on Floating-Point Condition Codes with Prediction

         (FBPfcc).  */

      branch_p = 1;

      offset = 4 * X_DISP19 (insn);

    }

  else if (X_OP (insn) == 0 && X_OP2 (insn) == 2)

    {

      /* Branch on Integer Condition Codes (Bicc).  */

      branch_p = 1;

      offset = 4 * X_DISP22 (insn);

    }

  else if (X_OP (insn) == 0 && X_OP2 (insn) == 1)

    {

      /* Branch on Integer Condition Codes with Prediction (BPcc).  */

      branch_p = 1;

      offset = 4 * X_DISP19 (insn);

    }

  else if (X_OP (insn) == 2 && X_OP3 (insn) == 0x3a)

    {

      /* Trap instruction (TRAP).  */

      return gdbarch_tdep (get_frame_arch (frame))->step_trap (frame, insn);

    }



  /* FIXME: Handle DONE and RETRY instructions.  */



  if (branch_p)

    {

      if (fused_p)

        {

          /* Fused compare-and-branch instructions are non-delayed,

             and do not have an annuling capability.  So we need to

             always set a breakpoint on both the NPC and the branch

             target address.  */

          gdb_assert (offset != 0);

          return pc + offset;

        }

      else if (conditional_p)

        {

          /* For conditional branches, return nPC + 4 iff the annul

             bit is 1.  */

          return (X_A (insn) ? *npc + 4 : 0);

        }

      else

        {

          /* For unconditional branches, return the target if its

             specified condition is "always" and return nPC + 4 if the

             condition is "never".  If the annul bit is 1, set *NPC to

             zero.  */

          if (X_COND (insn) == 0x0)

            pc = *npc, offset = 4;

          if (X_A (insn))

            *npc = 0;



          return pc + offset;

        }

    }



  return 0;

}



static CORE_ADDR

‌sparc_step_trap (struct frame_info *frame, unsigned long insn)

{

  return 0;

}



int

‌sparc_software_single_step (struct frame_info *frame)

{

  struct gdbarch *arch = get_frame_arch (frame);

  struct gdbarch_tdep *tdep = gdbarch_tdep (arch);

  struct address_space *aspace = get_frame_address_space (frame);

  CORE_ADDR npc, nnpc;



  CORE_ADDR pc, orig_npc;



  pc = get_frame_register_unsigned (frame, tdep->pc_regnum);

  orig_npc = npc = get_frame_register_unsigned (frame, tdep->npc_regnum);



  /* Analyze the instruction at PC.  */

  nnpc = sparc_analyze_control_transfer (frame, pc, &npc);

  if (npc != 0)

    insert_single_step_breakpoint (arch, aspace, npc);



  if (nnpc != 0)

    insert_single_step_breakpoint (arch, aspace, nnpc);



  /* Assert that we have set at least one breakpoint, and that

     they're not set at the same spot - unless we're going

     from here straight to NULL, i.e. a call or jump to 0.  */

  gdb_assert (npc != 0 || nnpc != 0 || orig_npc == 0);

  gdb_assert (nnpc != npc || orig_npc == 0);



  return 1;

}



static void

‌sparc_write_pc (struct regcache *regcache, CORE_ADDR pc)

{

  struct gdbarch_tdep *tdep = gdbarch_tdep (get_regcache_arch (regcache));



  regcache_cooked_write_unsigned (regcache, tdep->pc_regnum, pc);

  regcache_cooked_write_unsigned (regcache, tdep->npc_regnum, pc + 4);

}





/* Iterate over core file register note sections.  */



static void

‌sparc_iterate_over_regset_sections (struct gdbarch *gdbarch,

                                    iterate_over_regset_sections_cb *cb,

                                    void *cb_data,

                                    const struct regcache *regcache)

{

  struct gdbarch_tdep *tdep = gdbarch_tdep (gdbarch);



  cb (".reg", tdep->sizeof_gregset, tdep->gregset, NULL, cb_data);

  cb (".reg2", tdep->sizeof_fpregset, tdep->fpregset, NULL, cb_data);

}





static struct gdbarch *

‌sparc32_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)

{

  struct gdbarch_tdep *tdep;

  struct gdbarch *gdbarch;



  /* If there is already a candidate, use it.  */

  arches = gdbarch_list_lookup_by_info (arches, &info);

  if (arches != NULL)

    return arches->gdbarch;



  /* Allocate space for the new architecture.  */

  tdep = XCNEW (struct gdbarch_tdep);

  gdbarch = gdbarch_alloc (&info, tdep);



  tdep->pc_regnum = SPARC32_PC_REGNUM;

  tdep->npc_regnum = SPARC32_NPC_REGNUM;

  tdep->step_trap = sparc_step_trap;



  set_gdbarch_long_double_bit (gdbarch, 128);

  set_gdbarch_long_double_format (gdbarch, floatformats_sparc_quad);



  set_gdbarch_num_regs (gdbarch, SPARC32_NUM_REGS);

  set_gdbarch_register_name (gdbarch, sparc32_register_name);

  set_gdbarch_register_type (gdbarch, sparc32_register_type);

  set_gdbarch_num_pseudo_regs (gdbarch, SPARC32_NUM_PSEUDO_REGS);

  set_gdbarch_pseudo_register_read (gdbarch, sparc32_pseudo_register_read);

  set_gdbarch_pseudo_register_write (gdbarch, sparc32_pseudo_register_write);



  /* Register numbers of various important registers.  */

  set_gdbarch_sp_regnum (gdbarch, SPARC_SP_REGNUM); /* %sp */

  set_gdbarch_pc_regnum (gdbarch, SPARC32_PC_REGNUM); /* %pc */

  set_gdbarch_fp0_regnum (gdbarch, SPARC_F0_REGNUM); /* %f0 */



  /* Call dummy code.  */

  set_gdbarch_frame_align (gdbarch, sparc32_frame_align);

  set_gdbarch_call_dummy_location (gdbarch, ON_STACK);

  set_gdbarch_push_dummy_code (gdbarch, sparc32_push_dummy_code);

  set_gdbarch_push_dummy_call (gdbarch, sparc32_push_dummy_call);



  set_gdbarch_return_value (gdbarch, sparc32_return_value);

  set_gdbarch_stabs_argument_has_addr

    (gdbarch, sparc32_stabs_argument_has_addr);



  set_gdbarch_skip_prologue (gdbarch, sparc32_skip_prologue);



  /* Stack grows downward.  */

  set_gdbarch_inner_than (gdbarch, core_addr_lessthan);



  set_gdbarch_breakpoint_from_pc (gdbarch, sparc_breakpoint_from_pc);



  set_gdbarch_frame_args_skip (gdbarch, 8);



  set_gdbarch_print_insn (gdbarch, print_insn_sparc);



  set_gdbarch_software_single_step (gdbarch, sparc_software_single_step);

  set_gdbarch_write_pc (gdbarch, sparc_write_pc);



  set_gdbarch_dummy_id (gdbarch, sparc_dummy_id);



  set_gdbarch_unwind_pc (gdbarch, sparc_unwind_pc);



  frame_base_set_default (gdbarch, &sparc32_frame_base);



  /* Hook in the DWARF CFI frame unwinder.  */

  dwarf2_frame_set_init_reg (gdbarch, sparc32_dwarf2_frame_init_reg);

  /* FIXME: kettenis/20050423: Don't enable the unwinder until the

     StackGhost issues have been resolved.  */



  /* Hook in ABI-specific overrides, if they have been registered.  */

  gdbarch_init_osabi (info, gdbarch);



  frame_unwind_append_unwinder (gdbarch, &sparc32_frame_unwind);



  /* If we have register sets, enable the generic core file support.  */

  if (tdep->gregset)

    set_gdbarch_iterate_over_regset_sections

      (gdbarch, sparc_iterate_over_regset_sections);



  register_sparc_ravenscar_ops (gdbarch);



  return gdbarch;

}



/* Helper functions for dealing with register windows.  */



void

‌sparc_supply_rwindow (struct regcache *regcache, CORE_ADDR sp, int regnum)

{

  struct gdbarch *gdbarch = get_regcache_arch (regcache);

  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);

  int offset = 0;

  gdb_byte buf[8];

  int i;



  if (sp & 1)

    {

      /* Registers are 64-bit.  */

      sp += BIAS;



      for (i = SPARC_L0_REGNUM; i <= SPARC_I7_REGNUM; i++)

        {

          if (regnum == i || regnum == -1)

            {

              target_read_memory (sp + ((i - SPARC_L0_REGNUM) * 8), buf, 8);



              /* Handle StackGhost.  */

              if (i == SPARC_I7_REGNUM)

                {

                  ULONGEST wcookie = sparc_fetch_wcookie (gdbarch);

                  ULONGEST i7;



                  i7 = extract_unsigned_integer (buf + offset, 8, byte_order);

                  store_unsigned_integer (buf + offset, 8, byte_order,

                                          i7 ^ wcookie);

                }



              regcache_raw_supply (regcache, i, buf);

            }

        }

    }

  else

    {

      /* Registers are 32-bit.  Toss any sign-extension of the stack

         pointer.  */

      sp &= 0xffffffffUL;



      /* Clear out the top half of the temporary buffer, and put the

         register value in the bottom half if we're in 64-bit mode.  */

      if (gdbarch_ptr_bit (get_regcache_arch (regcache)) == 64)

        {

          memset (buf, 0, 4);

          offset = 4;

        }



      for (i = SPARC_L0_REGNUM; i <= SPARC_I7_REGNUM; i++)

        {

          if (regnum == i || regnum == -1)

            {

              target_read_memory (sp + ((i - SPARC_L0_REGNUM) * 4),

                                  buf + offset, 4);



              /* Handle StackGhost.  */

              if (i == SPARC_I7_REGNUM)

                {

                  ULONGEST wcookie = sparc_fetch_wcookie (gdbarch);

                  ULONGEST i7;



                  i7 = extract_unsigned_integer (buf + offset, 4, byte_order);

                  store_unsigned_integer (buf + offset, 4, byte_order,

                                          i7 ^ wcookie);

                }



              regcache_raw_supply (regcache, i, buf);

            }

        }

    }

}



void

‌sparc_collect_rwindow (const struct regcache *regcache,

                       CORE_ADDR sp, int regnum)

{

  struct gdbarch *gdbarch = get_regcache_arch (regcache);

  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);

  int offset = 0;

  gdb_byte buf[8];

  int i;



  if (sp & 1)

    {

      /* Registers are 64-bit.  */

      sp += BIAS;



      for (i = SPARC_L0_REGNUM; i <= SPARC_I7_REGNUM; i++)

        {

          if (regnum == -1 || regnum == SPARC_SP_REGNUM || regnum == i)

            {

              regcache_raw_collect (regcache, i, buf);



              /* Handle StackGhost.  */

              if (i == SPARC_I7_REGNUM)

                {

                  ULONGEST wcookie = sparc_fetch_wcookie (gdbarch);

                  ULONGEST i7;



                  i7 = extract_unsigned_integer (buf + offset, 8, byte_order);

                  store_unsigned_integer (buf, 8, byte_order, i7 ^ wcookie);

                }



              target_write_memory (sp + ((i - SPARC_L0_REGNUM) * 8), buf, 8);

            }

        }

    }

  else

    {

      /* Registers are 32-bit.  Toss any sign-extension of the stack

         pointer.  */

      sp &= 0xffffffffUL;



      /* Only use the bottom half if we're in 64-bit mode.  */

      if (gdbarch_ptr_bit (get_regcache_arch (regcache)) == 64)

        offset = 4;



      for (i = SPARC_L0_REGNUM; i <= SPARC_I7_REGNUM; i++)

        {

          if (regnum == -1 || regnum == SPARC_SP_REGNUM || regnum == i)

            {

              regcache_raw_collect (regcache, i, buf);



              /* Handle StackGhost.  */

              if (i == SPARC_I7_REGNUM)

                {

                  ULONGEST wcookie = sparc_fetch_wcookie (gdbarch);

                  ULONGEST i7;



                  i7 = extract_unsigned_integer (buf + offset, 4, byte_order);

                  store_unsigned_integer (buf + offset, 4, byte_order,

                                          i7 ^ wcookie);

                }



              target_write_memory (sp + ((i - SPARC_L0_REGNUM) * 4),

                                   buf + offset, 4);

            }

        }

    }

}



/* Helper functions for dealing with register sets.  */



void

‌sparc32_supply_gregset (const struct sparc_gregmap *gregmap,

                        struct regcache *regcache,

                        int regnum, const void *gregs)

{

  const gdb_byte *regs = gregs;

  gdb_byte zero[4] = { 0 };

  int i;



  if (regnum == SPARC32_PSR_REGNUM || regnum == -1)

    regcache_raw_supply (regcache, SPARC32_PSR_REGNUM,

                         regs + gregmap->r_psr_offset);



  if (regnum == SPARC32_PC_REGNUM || regnum == -1)

    regcache_raw_supply (regcache, SPARC32_PC_REGNUM,

                         regs + gregmap->r_pc_offset);



  if (regnum == SPARC32_NPC_REGNUM || regnum == -1)

    regcache_raw_supply (regcache, SPARC32_NPC_REGNUM,

                         regs + gregmap->r_npc_offset);



  if (regnum == SPARC32_Y_REGNUM || regnum == -1)

    regcache_raw_supply (regcache, SPARC32_Y_REGNUM,

                         regs + gregmap->r_y_offset);



  if (regnum == SPARC_G0_REGNUM || regnum == -1)

    regcache_raw_supply (regcache, SPARC_G0_REGNUM, &zero);



  if ((regnum >= SPARC_G1_REGNUM && regnum <= SPARC_O7_REGNUM) || regnum == -1)

    {

      int offset = gregmap->r_g1_offset;



      for (i = SPARC_G1_REGNUM; i <= SPARC_O7_REGNUM; i++)

        {

          if (regnum == i || regnum == -1)

            regcache_raw_supply (regcache, i, regs + offset);

          offset += 4;

        }

    }



  if ((regnum >= SPARC_L0_REGNUM && regnum <= SPARC_I7_REGNUM) || regnum == -1)

    {

      /* Not all of the register set variants include Locals and

         Inputs.  For those that don't, we read them off the stack.  */

      if (gregmap->r_l0_offset == -1)

        {

          ULONGEST sp;



          regcache_cooked_read_unsigned (regcache, SPARC_SP_REGNUM, &sp);

          sparc_supply_rwindow (regcache, sp, regnum);

        }

      else

        {

          int offset = gregmap->r_l0_offset;



          for (i = SPARC_L0_REGNUM; i <= SPARC_I7_REGNUM; i++)

            {

              if (regnum == i || regnum == -1)

                regcache_raw_supply (regcache, i, regs + offset);

              offset += 4;

            }

        }

    }

}



void

‌sparc32_collect_gregset (const struct sparc_gregmap *gregmap,

                         const struct regcache *regcache,

                         int regnum, void *gregs)

{

  gdb_byte *regs = gregs;

  int i;



  if (regnum == SPARC32_PSR_REGNUM || regnum == -1)

    regcache_raw_collect (regcache, SPARC32_PSR_REGNUM,

                          regs + gregmap->r_psr_offset);



  if (regnum == SPARC32_PC_REGNUM || regnum == -1)

    regcache_raw_collect (regcache, SPARC32_PC_REGNUM,

                          regs + gregmap->r_pc_offset);



  if (regnum == SPARC32_NPC_REGNUM || regnum == -1)

    regcache_raw_collect (regcache, SPARC32_NPC_REGNUM,

                          regs + gregmap->r_npc_offset);



  if (regnum == SPARC32_Y_REGNUM || regnum == -1)

    regcache_raw_collect (regcache, SPARC32_Y_REGNUM,

                          regs + gregmap->r_y_offset);



  if ((regnum >= SPARC_G1_REGNUM && regnum <= SPARC_O7_REGNUM) || regnum == -1)

    {

      int offset = gregmap->r_g1_offset;



      /* %g0 is always zero.  */

      for (i = SPARC_G1_REGNUM; i <= SPARC_O7_REGNUM; i++)

        {

          if (regnum == i || regnum == -1)

            regcache_raw_collect (regcache, i, regs + offset);

          offset += 4;

        }

    }



  if ((regnum >= SPARC_L0_REGNUM && regnum <= SPARC_I7_REGNUM) || regnum == -1)

    {

      /* Not all of the register set variants include Locals and

         Inputs.  For those that don't, we read them off the stack.  */

      if (gregmap->r_l0_offset != -1)

        {

          int offset = gregmap->r_l0_offset;



          for (i = SPARC_L0_REGNUM; i <= SPARC_I7_REGNUM; i++)

            {

              if (regnum == i || regnum == -1)

                regcache_raw_collect (regcache, i, regs + offset);

              offset += 4;

            }

        }

    }

}



void

‌sparc32_supply_fpregset (const struct sparc_fpregmap *fpregmap,

                         struct regcache *regcache,

                         int regnum, const void *fpregs)

{

  const gdb_byte *regs = fpregs;

  int i;



  for (i = 0; i < 32; i++)

    {

      if (regnum == (SPARC_F0_REGNUM + i) || regnum == -1)

        regcache_raw_supply (regcache, SPARC_F0_REGNUM + i,

                             regs + fpregmap->r_f0_offset + (i * 4));

    }



  if (regnum == SPARC32_FSR_REGNUM || regnum == -1)

    regcache_raw_supply (regcache, SPARC32_FSR_REGNUM,

                         regs + fpregmap->r_fsr_offset);

}



void

‌sparc32_collect_fpregset (const struct sparc_fpregmap *fpregmap,

                          const struct regcache *regcache,

                          int regnum, void *fpregs)

{

  gdb_byte *regs = fpregs;

  int i;



  for (i = 0; i < 32; i++)

    {

      if (regnum == (SPARC_F0_REGNUM + i) || regnum == -1)

        regcache_raw_collect (regcache, SPARC_F0_REGNUM + i,

                              regs + fpregmap->r_f0_offset + (i * 4));

    }



  if (regnum == SPARC32_FSR_REGNUM || regnum == -1)

    regcache_raw_collect (regcache, SPARC32_FSR_REGNUM,

                          regs + fpregmap->r_fsr_offset);

}





/* SunOS 4.  */



/* From <machine/reg.h>.  */

‌const struct sparc_gregmap sparc32_sunos4_gregmap =

{

  0 * 4,                        /* %psr */

  1 * 4,                        /* %pc */

  2 * 4,                        /* %npc */

  3 * 4,                        /* %y */

  -1,                                /* %wim */

  -1,                                /* %tbr */

  4 * 4,                        /* %g1 */

  -1                                /* %l0 */

};



‌const struct sparc_fpregmap sparc32_sunos4_fpregmap =

{

  0 * 4,                        /* %f0 */

  33 * 4,                        /* %fsr */

};



‌const struct sparc_fpregmap sparc32_bsd_fpregmap =

{

  0 * 4,                        /* %f0 */

  32 * 4,                        /* %fsr */

};





/* Provide a prototype to silence -Wmissing-prototypes.  */

void _initialize_sparc_tdep (void);



void

‌_initialize_sparc_tdep (void)

{

  register_gdbarch_init (bfd_arch_sparc, sparc32_gdbarch_init);

}
gdb/sparc-tdep.c - gdb

Global variables defined

Functions defined

Macros defined

Source code
