* avr and frame unwinding
@ 2003-06-06 0:35 Theodore A. Roth
2003-06-06 2:01 ` Andrew Cagney
0 siblings, 1 reply; 7+ messages in thread
From: Theodore A. Roth @ 2003-06-06 0:35 UTC (permalink / raw)
To: gdb
[-- Attachment #1: Type: TEXT/PLAIN, Size: 2295 bytes --]
Hi,
I've pretty much hit a wall on this, so I will have to ask for help
now.
The attached is what I have so far in my attempt to frame-ify
avr-tdep.c. As it stands, things kinda work. I can 'step' around a
program fine.
Things go wrong as soon as I try to 'next' over a function call, gdb
goes into an infinite loop. If I 'step' down into a function called
from main(), a backtrace never gets me back to main. Here's a dump of
the backtrace output (I'll forego the infinite loop output to save
some bandwidth 8-):
func1 () at tst.c:23
23 struct cc var = {0}; /* This uses memset()! */
(gdb) bt
#0 func1 () at tst.c:23
{ get_prev_frame (this_frame=0) { frame_id_p
(l={stack=0x8010f9,code=0xd0}) -> 1 }
{ frame_register_unwind (frame=0,regnum="PC",...) {
frame_register_unwind (frame=-1,regnum="PC",...) -> *optimizedp=0
*lvalp=2 *addrp=0x23 *bufferp=[fffffff0000000] }
-> *optimizedp=0 *lvalp=2 *addrp=0x23 *bufferp=[fffffff0000000] }
{ frame_pc_unwind (this_frame=0) -> 0xf0 }
->
{level=1,type=UNKNOWN_FRAME,unwind=<unknown>,pc=0xf0,id=<unknown>,func=<unknown>}
}
#1 0x000000f0 in func1 () at tst.c:22
{ get_prev_frame (this_frame=1) { get_frame_id (fi=1) {
frame_func_unwind (fi=0) -> 0xd0 }
Sending packet: $md0,27#96...Ack
Packet received:
cf92df92ef92ff920f931f93cf93df93cdb7deb726970fb6f894debf0fbecdbf86e0fe01319611
Sending packet: $mf7,11#98...Ack
Packet received: 928a95e9f719821a82ed80fe80cc24dd24
{ frame_register_unwind (frame=0,regnum="r28",...) Sending packet:
$m801100,1#f4...Ack
Packet received: 00
-> *optimizedp=0 *lvalp=1 *addrp=0x801100 *bufferp=[00] }
{ frame_register_unwind (frame=0,regnum="r29",...) Sending packet:
$m801101,1#f5...Ack
Packet received: 00
-> *optimizedp=0 *lvalp=1 *addrp=0x801101 *bufferp=[00] }
{ frame_id_eq
(l={stack=0x8010f9,code=0xd0},r={stack=0x80000e,code=0xd0}) -> 0 }
-> {stack=0x80000e,code=0xd0} }
{ frame_id_p (l={stack=0x80000e,code=0xd0}) -> 1 }
{ frame_id_inner
(l={stack=0x80000e,code=0xd0},r={stack=0x8010f9,code=0xd0}) -> 1 }
This frame inner-to next frame (corrupt stack?)
(gdb)
I'm pretty confused as to how frames are getting chained up and
unwound. I suspose that I'm not thinking about the the SP in the
proper manner in relation to the frames.
Could some kind soul lend me some insight?
Thanks.
Ted Roth
[-- Attachment #2: Type: TEXT/PLAIN, Size: 38756 bytes --]
/* Target-dependent code for Atmel AVR, for GDB.
Copyright 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003
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 2 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, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330,
Boston, MA 02111-1307, USA. */
/* Contributed by Theodore A. Roth, troth@openavr.org */
/* Portions of this file were taken from the original gdb-4.18 patch developed
by Denis Chertykov, denisc@overta.ru */
#include "defs.h"
#include "frame.h"
#include "frame-unwind.h"
#include "frame-base.h"
#include "gdbcmd.h"
#include "gdbcore.h"
#include "inferior.h"
#include "symfile.h"
#include "arch-utils.h"
#include "regcache.h"
#include "gdb_string.h"
/* AVR Background:
(AVR micros are pure Harvard Architecture processors.)
The AVR family of microcontrollers have three distinctly different memory
spaces: flash, sram and eeprom. The flash is 16 bits wide and is used for
the most part to store program instructions. The sram is 8 bits wide and is
used for the stack and the heap. Some devices lack sram and some can have
an additional external sram added on as a peripheral.
The eeprom is 8 bits wide and is used to store data when the device is
powered down. Eeprom is not directly accessible, it can only be accessed
via io-registers using a special algorithm. Accessing eeprom via gdb's
remote serial protocol ('m' or 'M' packets) looks difficult to do and is
not included at this time.
[The eeprom could be read manually via ``x/b <eaddr + AVR_EMEM_START>'' or
written using ``set {unsigned char}<eaddr + AVR_EMEM_START>''. For this to
work, the remote target must be able to handle eeprom accesses and perform
the address translation.]
All three memory spaces have physical addresses beginning at 0x0. In
addition, the flash is addressed by gcc/binutils/gdb with respect to 8 bit
bytes instead of the 16 bit wide words used by the real device for the
Program Counter.
In order for remote targets to work correctly, extra bits must be added to
addresses before they are send to the target or received from the target
via the remote serial protocol. The extra bits are the MSBs and are used to
decode which memory space the address is referring to. */
#undef XMALLOC
#define XMALLOC(TYPE) ((TYPE*) xmalloc (sizeof (TYPE)))
#undef EXTRACT_INSN
#define EXTRACT_INSN(addr) extract_unsigned_integer(addr,2)
/* Constants: prefixed with AVR_ to avoid name space clashes */
enum
{
AVR_REG_W = 24,
AVR_REG_X = 26,
AVR_REG_Y = 28,
AVR_FP_REGNUM = 28,
AVR_REG_Z = 30,
AVR_SREG_REGNUM = 32,
AVR_SP_REGNUM = 33,
AVR_PC_REGNUM = 34,
AVR_NUM_REGS = 32 + 1 /*SREG*/ + 1 /*SP*/ + 1 /*PC*/,
AVR_NUM_REG_BYTES = 32 + 1 /*SREG*/ + 2 /*SP*/ + 4 /*PC*/,
AVR_PC_REG_INDEX = 35, /* index into array of registers */
AVR_MAX_PROLOGUE_SIZE = 56, /* bytes */
/* Count of pushed registers. From r2 to r17 (inclusively), r28, r29 */
AVR_MAX_PUSHES = 18,
/* Number of the last pushed register. r17 for current avr-gcc */
AVR_LAST_PUSHED_REGNUM = 17,
/* FIXME: TRoth/2002-01-??: Can we shift all these memory masks left 8
bits? Do these have to match the bfd vma values?. It sure would make
things easier in the future if they didn't need to match.
Note: I chose these values so as to be consistent with bfd vma
addresses.
TRoth/2002-04-08: There is already a conflict with very large programs
in the mega128. The mega128 has 128K instruction bytes (64K words),
thus the Most Significant Bit is 0x10000 which gets masked off my
AVR_MEM_MASK.
The problem manifests itself when trying to set a breakpoint in a
function which resides in the upper half of the instruction space and
thus requires a 17-bit address.
For now, I've just removed the EEPROM mask and changed AVR_MEM_MASK
from 0x00ff0000 to 0x00f00000. Eeprom is not accessible from gdb yet,
but could be for some remote targets by just adding the correct offset
to the address and letting the remote target handle the low-level
details of actually accessing the eeprom. */
AVR_IMEM_START = 0x00000000, /* INSN memory */
AVR_SMEM_START = 0x00800000, /* SRAM memory */
#if 1
/* No eeprom mask defined */
AVR_MEM_MASK = 0x00f00000, /* mask to determine memory space */
#else
AVR_EMEM_START = 0x00810000, /* EEPROM memory */
AVR_MEM_MASK = 0x00ff0000, /* mask to determine memory space */
#endif
};
/* Prologue types:
NORMAL and CALL are the typical types (interchangeable with the
-mcall-prologues gcc option.
MAIN, INTR and SIG are invariant as far as I can tell. */
enum {
AVR_PROLOGUE_NONE, /* No prologue */
AVR_PROLOGUE_NORMAL,
AVR_PROLOGUE_CALL, /* -mcall-prologues */
AVR_PROLOGUE_MAIN,
AVR_PROLOGUE_INTR, /* interrupt handler */
AVR_PROLOGUE_SIG, /* signal handler */
};
struct gdbarch_tdep
{
/* FIXME: TRoth: is there anything to put here? */
int foo;
};
/* Any function with a frame looks like this
....... <-SP POINTS HERE
LOCALS1 <-FP POINTS HERE
LOCALS0
SAVED FP
SAVED R3
SAVED R2
RET PC
FIRST ARG
SECOND ARG */
struct avr_unwind_cache
{
/* CORE_ADDR return_pc; */
/* The previous frame's inner most stack address. Used as this
frame ID's stack_addr. */
CORE_ADDR prev_sp;
/* The frame's base, optionally used by the high-level debug info. */
CORE_ADDR base;
int size;
CORE_ADDR *saved_regs;
/* How far the SP has been offset from the start of the stack frame (as
defined by the previous frame's stack pointer). */
LONGEST sp_offset;
int prologue_type;
void **regs;
};
/* Lookup the name of a register given it's number. */
static const char *
avr_register_name (int regnum)
{
static char *register_names[] = {
"r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
"r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15",
"r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
"r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31",
"SREG", "SP", "PC"
};
if (regnum < 0)
return NULL;
if (regnum >= (sizeof (register_names) / sizeof (*register_names)))
return NULL;
return register_names[regnum];
}
/* Return the GDB type object for the "standard" data type
of data in register N. */
static struct type *
avr_register_type (struct gdbarch *gdbarch, int reg_nr)
{
if (reg_nr == AVR_PC_REGNUM)
return builtin_type_unsigned_long;
if (reg_nr == AVR_SP_REGNUM)
return builtin_type_void_data_ptr;
else
return builtin_type_uint8;
}
/* Instruction address checks and convertions. */
static CORE_ADDR
avr_make_iaddr (CORE_ADDR x)
{
return ((x) | AVR_IMEM_START);
}
static int
avr_iaddr_p (CORE_ADDR x)
{
return (((x) & AVR_MEM_MASK) == AVR_IMEM_START);
}
/* FIXME: TRoth: Really need to use a larger mask for instructions. Some
devices are already up to 128KBytes of flash space.
TRoth/2002-04-8: See comment above where AVR_IMEM_START is defined. */
static CORE_ADDR
avr_convert_iaddr_to_raw (CORE_ADDR x)
{
return ((x) & 0xffffffff);
}
/* SRAM address checks and convertions. */
static CORE_ADDR
avr_make_saddr (CORE_ADDR x)
{
return ((x) | AVR_SMEM_START);
}
static int
avr_saddr_p (CORE_ADDR x)
{
return (((x) & AVR_MEM_MASK) == AVR_SMEM_START);
}
static CORE_ADDR
avr_convert_saddr_to_raw (CORE_ADDR x)
{
return ((x) & 0xffffffff);
}
/* EEPROM address checks and convertions. I don't know if these will ever
actually be used, but I've added them just the same. TRoth */
/* TRoth/2002-04-08: Commented out for now to allow fix for problem with large
programs in the mega128. */
/* static CORE_ADDR */
/* avr_make_eaddr (CORE_ADDR x) */
/* { */
/* return ((x) | AVR_EMEM_START); */
/* } */
/* static int */
/* avr_eaddr_p (CORE_ADDR x) */
/* { */
/* return (((x) & AVR_MEM_MASK) == AVR_EMEM_START); */
/* } */
/* static CORE_ADDR */
/* avr_convert_eaddr_to_raw (CORE_ADDR x) */
/* { */
/* return ((x) & 0xffffffff); */
/* } */
/* Convert from address to pointer and vice-versa. */
static void
avr_address_to_pointer (struct type *type, void *buf, CORE_ADDR addr)
{
/* Is it a code address? */
if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_FUNC
|| TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_METHOD)
{
store_unsigned_integer (buf, TYPE_LENGTH (type),
avr_convert_iaddr_to_raw (addr));
}
else
{
/* Strip off any upper segment bits. */
store_unsigned_integer (buf, TYPE_LENGTH (type),
avr_convert_saddr_to_raw (addr));
}
}
static CORE_ADDR
avr_pointer_to_address (struct type *type, const void *buf)
{
CORE_ADDR addr = extract_unsigned_integer (buf, TYPE_LENGTH (type));
if (TYPE_CODE_SPACE (TYPE_TARGET_TYPE (type)))
{
fprintf_unfiltered (gdb_stderr, "CODE_SPACE ---->> ptr->addr: 0x%lx\n",
addr);
fprintf_unfiltered (gdb_stderr,
"+++ If you see this, please send me an "
"email <troth@openavr.org>\n");
}
/* Is it a code address? */
if (TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_FUNC
|| TYPE_CODE (TYPE_TARGET_TYPE (type)) == TYPE_CODE_METHOD
|| TYPE_CODE_SPACE (TYPE_TARGET_TYPE (type)))
return avr_make_iaddr (addr);
else
return avr_make_saddr (addr);
}
static CORE_ADDR
avr_read_pc (ptid_t ptid)
{
ptid_t save_ptid;
CORE_ADDR pc;
CORE_ADDR retval;
save_ptid = inferior_ptid;
inferior_ptid = ptid;
pc = (int) read_register (AVR_PC_REGNUM);
inferior_ptid = save_ptid;
retval = avr_make_iaddr (pc);
return retval;
}
static void
avr_write_pc (CORE_ADDR val, ptid_t ptid)
{
ptid_t save_ptid;
save_ptid = inferior_ptid;
inferior_ptid = ptid;
write_register (AVR_PC_REGNUM, avr_convert_iaddr_to_raw (val));
inferior_ptid = save_ptid;
}
static CORE_ADDR
avr_read_sp (void)
{
return (avr_make_saddr (read_register (AVR_SP_REGNUM)));
}
static void
avr_write_sp (CORE_ADDR val)
{
write_register (AVR_SP_REGNUM, avr_convert_saddr_to_raw (val));
}
/* Translate a GDB virtual ADDR/LEN into a format the remote target
understands. Returns number of bytes that can be transfered
starting at TARG_ADDR. Return ZERO if no bytes can be transfered
(segmentation fault).
TRoth/2002-04-08: Could this be used to check for dereferencing an invalid
pointer? */
static void
avr_remote_translate_xfer_address (struct gdbarch *gdbarch,
struct regcache *regcache,
CORE_ADDR memaddr, int nr_bytes,
CORE_ADDR *targ_addr, int *targ_len)
{
long out_addr;
long out_len;
/* FIXME: TRoth: Do nothing for now. Will need to examine memaddr at this
point and see if the high bit are set with the masks that we want. */
*targ_addr = memaddr;
*targ_len = nr_bytes;
}
/* Function pointers obtained from the target are half of what gdb expects so
multiply by 2. */
static CORE_ADDR
avr_convert_from_func_ptr_addr (CORE_ADDR addr)
{
return addr * 2;
}
/* Returns the return address for a dummy. */
static CORE_ADDR
avr_call_dummy_address (void)
{
return entry_point_address ();
}
/* Function: avr_scan_prologue
This function decodes an AVR function prologue to determine:
1) the size of the stack frame
2) which registers are saved on it
3) the offsets of saved regs
This information is stored in the avr_unwind_cache structure.
Some devices lack the sbiw instruction, so on those replace this:
sbiw r28, XX
with this:
subi r28,lo8(XX)
sbci r29,hi8(XX)
A typical AVR function prologue with a frame pointer might look like this:
push rXX ; saved regs
...
push r28
push r29
in r28,__SP_L__
in r29,__SP_H__
sbiw r28,<LOCALS_SIZE>
in __tmp_reg__,__SREG__
cli
out __SP_L__,r28
out __SREG__,__tmp_reg__
out __SP_H__,r29
A typical AVR function prologue without a frame pointer might look like
this:
push rXX ; saved regs
...
A main function prologue looks like this:
ldi r28,lo8(<RAM_ADDR> - <LOCALS_SIZE>)
ldi r29,hi8(<RAM_ADDR> - <LOCALS_SIZE>)
out __SP_H__,r29
out __SP_L__,r28
A signal handler prologue looks like this:
push __zero_reg__
push __tmp_reg__
in __tmp_reg__, __SREG__
push __tmp_reg__
clr __zero_reg__
push rXX ; save registers r18:r27, r30:r31
...
push r28 ; save frame pointer
push r29
in r28, __SP_L__
in r29, __SP_H__
sbiw r28, <LOCALS_SIZE>
out __SP_H__, r29
out __SP_L__, r28
A interrupt handler prologue looks like this:
sei
push __zero_reg__
push __tmp_reg__
in __tmp_reg__, __SREG__
push __tmp_reg__
clr __zero_reg__
push rXX ; save registers r18:r27, r30:r31
...
push r28 ; save frame pointer
push r29
in r28, __SP_L__
in r29, __SP_H__
sbiw r28, <LOCALS_SIZE>
cli
out __SP_H__, r29
sei
out __SP_L__, r28
A `-mcall-prologues' prologue looks like this (Note that the megas use a
jmp instead of a rjmp, thus the prologue is one word larger since jmp is a
32 bit insn and rjmp is a 16 bit insn):
ldi r26,lo8(<LOCALS_SIZE>)
ldi r27,hi8(<LOCALS_SIZE>)
ldi r30,pm_lo8(.L_foo_body)
ldi r31,pm_hi8(.L_foo_body)
rjmp __prologue_saves__+RRR
.L_foo_body: */
static CORE_ADDR
avr_scan_prologue (CORE_ADDR pc, struct avr_unwind_cache *info)
{
int i;
unsigned short insn;
int scan_stage = 0;
struct minimal_symbol *msymbol;
unsigned char prologue[AVR_MAX_PROLOGUE_SIZE];
int vpc = 0;
read_memory (pc, prologue, AVR_MAX_PROLOGUE_SIZE);
/* Scanning main()'s prologue
ldi r28,lo8(<RAM_ADDR> - <LOCALS_SIZE>)
ldi r29,hi8(<RAM_ADDR> - <LOCALS_SIZE>)
out __SP_H__,r29
out __SP_L__,r28 */
if (1)
{
CORE_ADDR locals;
unsigned char img[] = {
0xde, 0xbf, /* out __SP_H__,r29 */
0xcd, 0xbf /* out __SP_L__,r28 */
};
insn = EXTRACT_INSN (&prologue[vpc]);
/* ldi r28,lo8(<RAM_ADDR> - <LOCALS_SIZE>) */
if ((insn & 0xf0f0) == 0xe0c0)
{
locals = (insn & 0xf) | ((insn & 0x0f00) >> 4);
insn = EXTRACT_INSN (&prologue[vpc + 2]);
/* ldi r29,hi8(<RAM_ADDR> - <LOCALS_SIZE>) */
if ((insn & 0xf0f0) == 0xe0d0)
{
locals |= ((insn & 0xf) | ((insn & 0x0f00) >> 4)) << 8;
if (memcmp (prologue + vpc + 4, img, sizeof (img)) == 0)
{
info->prologue_type = AVR_PROLOGUE_MAIN;
info->base = locals;
return pc + 4;
}
}
}
}
/* Scanning `-mcall-prologues' prologue
FIXME: mega prologue is a 12 bytes long */
while (1) /* Use while to avoid many goto's */
{
int loc_size;
int body_addr;
unsigned num_pushes;
int pc_offset = 0;
insn = EXTRACT_INSN (&prologue[vpc]);
/* ldi r26,<LOCALS_SIZE> */
if ((insn & 0xf0f0) != 0xe0a0)
break;
loc_size = (insn & 0xf) | ((insn & 0x0f00) >> 4);
pc_offset += 2;
insn = EXTRACT_INSN (&prologue[vpc + 2]);
/* ldi r27,<LOCALS_SIZE> / 256 */
if ((insn & 0xf0f0) != 0xe0b0)
break;
loc_size |= ((insn & 0xf) | ((insn & 0x0f00) >> 4)) << 8;
pc_offset += 2;
insn = EXTRACT_INSN (&prologue[vpc + 4]);
/* ldi r30,pm_lo8(.L_foo_body) */
if ((insn & 0xf0f0) != 0xe0e0)
break;
body_addr = (insn & 0xf) | ((insn & 0x0f00) >> 4);
pc_offset += 2;
insn = EXTRACT_INSN (&prologue[vpc + 6]);
/* ldi r31,pm_hi8(.L_foo_body) */
if ((insn & 0xf0f0) != 0xe0f0)
break;
body_addr |= ((insn & 0xf) | ((insn & 0x0f00) >> 4)) << 8;
pc_offset += 2;
msymbol = lookup_minimal_symbol ("__prologue_saves__", NULL, NULL);
if (!msymbol)
break;
insn = EXTRACT_INSN (&prologue[vpc + 8]);
/* rjmp __prologue_saves__+RRR */
if ((insn & 0xf000) == 0xc000)
{
/* Extract PC relative offset from RJMP */
i = (insn & 0xfff) | (insn & 0x800 ? (-1 ^ 0xfff) : 0);
/* Convert offset to byte addressable mode */
i *= 2;
/* Destination address */
i += pc_offset + pc + 10;
if (body_addr != (pc + 10)/2)
break;
pc_offset += 2;
}
else if ((insn & 0xfe0e) == 0x940c)
{
/* Extract absolute PC address from JMP */
i = (((insn & 0x1) | ((insn & 0x1f0) >> 3) << 16)
| (EXTRACT_INSN (&prologue[vpc + 10]) & 0xffff));
/* Convert address to byte addressable mode */
i *= 2;
if (body_addr != (pc + 12)/2)
break;
pc_offset += 4;
}
else
break;
/* Resolve offset (in words) from __prologue_saves__ symbol.
Which is a pushes count in `-mcall-prologues' mode */
num_pushes = AVR_MAX_PUSHES - (i - SYMBOL_VALUE_ADDRESS (msymbol)) / 2;
if (num_pushes > AVR_MAX_PUSHES)
{
fprintf_unfiltered (gdb_stderr, "Num pushes too large: %d\n",
num_pushes);
num_pushes = 0;
}
if (num_pushes)
{
int from;
info->saved_regs[AVR_FP_REGNUM + 1] = num_pushes;
if (num_pushes >= 2)
info->saved_regs[AVR_FP_REGNUM] = num_pushes - 1;
i = 0;
for (from = AVR_LAST_PUSHED_REGNUM + 1 - (num_pushes - 2);
from <= AVR_LAST_PUSHED_REGNUM; ++from)
info->saved_regs [from] = ++i;
}
info->size = loc_size + num_pushes;
info->prologue_type = AVR_PROLOGUE_CALL;
return pc + pc_offset + num_pushes*2;
}
/* Scan for the beginning of the prologue for an interrupt or signal
function */
if (1)
{
unsigned char img[] = {
0x78, 0x94, /* sei */
0x1f, 0x92, /* push r1 */
0x0f, 0x92, /* push r0 */
0x0f, 0xb6, /* in r0,0x3f SREG */
0x0f, 0x92, /* push r0 */
0x11, 0x24 /* clr r1 */
};
if (memcmp (prologue, img, sizeof (img)) == 0)
{
vpc += sizeof (img);
info->saved_regs[0] = 2;
info->saved_regs[1] = 1;
info->size += 3;
}
else if (memcmp (img + 2, prologue, sizeof (img) - 2) == 0)
{
vpc += sizeof (img) - 2;
info->saved_regs[0] = 2;
info->saved_regs[1] = 1;
info->size += 3;
}
}
/* First stage of the prologue scanning.
Scan pushes (saved registers) */
for (; vpc < AVR_MAX_PROLOGUE_SIZE; vpc += 2)
{
insn = EXTRACT_INSN (&prologue[vpc]);
if ((insn & 0xfe0f) == 0x920f) /* push rXX */
{
/* Bits 4-9 contain a mask for registers R0-R32. */
int regno = (insn & 0x1f0) >> 4;
info->size++;
info->saved_regs[regno] = info->size;
scan_stage = 1;
}
else
break;
}
if (vpc >= AVR_MAX_PROLOGUE_SIZE)
fprintf_unfiltered (gdb_stderr,
"Hit end of prologue while scanning pushes\n");
/* Second stage of the prologue scanning.
Scan:
in r28,__SP_L__
in r29,__SP_H__ */
if (scan_stage == 1 && vpc < AVR_MAX_PROLOGUE_SIZE)
{
unsigned char img[] = {
0xcd, 0xb7, /* in r28,__SP_L__ */
0xde, 0xb7 /* in r29,__SP_H__ */
};
unsigned short insn1;
if (memcmp (prologue + vpc, img, sizeof (img)) == 0)
{
vpc += 4;
scan_stage = 2;
}
}
/* Third stage of the prologue scanning. (Really two stages)
Scan for:
sbiw r28,XX or subi r28,lo8(XX)
sbci r29,hi8(XX)
in __tmp_reg__,__SREG__
cli
out __SP_H__,r29
out __SREG__,__tmp_reg__
out __SP_L__,r2* */
if (scan_stage == 2 && vpc < AVR_MAX_PROLOGUE_SIZE)
{
int locals_size = 0;
unsigned char img[] = {
0x0f, 0xb6, /* in r0,0x3f ; store SREG in r0 */
0xf8, 0x94, /* cli */
0xde, 0xbf, /* out 0x3e,r29 ; SPH */
0x0f, 0xbe, /* out 0x3f,r0 ; SREG */
0xcd, 0xbf /* out 0x3d,r28 ; SPL */
};
unsigned char img_sig[] = {
0xde, 0xbf, /* out 0x3e,r29 ; SPH */
0xcd, 0xbf /* out 0x3d,r28 ; SPL */
};
unsigned char img_intr[] = {
0xf8, 0x94, /* cli */
0xde, 0xbf, /* out 0x3e,r29 ; SPH */
0x78, 0x94, /* sei */
0xcd, 0xbf /* out 0x3d,r28 ; SPL */
};
insn = EXTRACT_INSN (&prologue[vpc]);
vpc += 2;
if ((insn & 0xff30) == 0x9720) /* sbiw r28,XXX */
locals_size = (insn & 0xf) | ((insn & 0xc0) >> 2);
else if ((insn & 0xf0f0) == 0x50c0) /* subi r28,lo8(XX) */
{
locals_size = (insn & 0xf) | ((insn & 0xf00) >> 4);
insn = EXTRACT_INSN (&prologue[vpc]);
vpc += 2;
locals_size += ((insn & 0xf) | ((insn & 0xf00) >> 4) << 8);
}
else
return pc + vpc;
/* Scan the last part of the prologue. */
if (memcmp (prologue + vpc, img_sig, sizeof (img_sig)) == 0)
{
info->prologue_type = AVR_PROLOGUE_SIG;
vpc += sizeof (img_sig);
}
else if (memcmp (prologue + vpc, img_intr, sizeof (img_intr)) == 0)
{
info->prologue_type = AVR_PROLOGUE_INTR;
vpc += sizeof (img_intr);
}
if (memcmp (prologue + vpc, img, sizeof (img)) == 0)
{
info->prologue_type = AVR_PROLOGUE_NORMAL;
vpc += sizeof (img);
}
info->size += locals_size;
return pc + vpc;
}
/* If we got this far, we could not scan the prologue, so just return the pc
of the frame. */
return pc;
}
static CORE_ADDR
avr_skip_prologue (CORE_ADDR pc)
{
CORE_ADDR func_addr, func_end;
CORE_ADDR prologue_end = pc;
/* See what the symbol table says */
if (find_pc_partial_function (pc, NULL, &func_addr, &func_end))
{
struct symtab_and_line sal;
struct avr_unwind_cache info = {0};
CORE_ADDR saved_regs[AVR_NUM_REGS] = {0};
info.saved_regs = saved_regs;
/* Need to run the prologue scanner to figure out if the function has a
prologue and possibly skip over moving data around some registers. */
prologue_end = avr_scan_prologue (pc, &info);
if (info.prologue_type != AVR_PROLOGUE_NONE)
{
sal = find_pc_line (func_addr, 0);
if (sal.line != 0 && sal.end < func_end)
return sal.end;
}
}
/* Either we didn't find the start of this function (nothing we can do),
or there's no line info, or the line after the prologue is after
the end of the function (there probably isn't a prologue). */
return prologue_end;
}
static CORE_ADDR
avr_frame_address (struct frame_info *fi)
{
return avr_make_saddr (get_frame_base (fi));
}
/* Not all avr devices support the BREAK insn. Those that don't should treat
it as a NOP. Thus, it should be ok. Since the avr is currently a remote
only target, this shouldn't be a problem (I hope). TRoth/2003-05-14 */
const unsigned char *
avr_breakpoint_from_pc (CORE_ADDR * pcptr, int *lenptr)
{
static unsigned char avr_break_insn [] = { 0x98, 0x95 };
*lenptr = sizeof (avr_break_insn);
return avr_break_insn;
}
static void
avr_saved_regs_unwinder (struct frame_info *next_frame,
CORE_ADDR *this_saved_regs,
int regnum, int *optimizedp,
enum lval_type *lvalp, CORE_ADDR *addrp,
int *realnump, void *bufferp)
{
if (this_saved_regs[regnum] != 0)
{
*optimizedp = 0;
*lvalp = lval_memory;
*addrp = this_saved_regs[regnum];
*realnump = -1;
if (bufferp != NULL)
{
/* Read the value in from memory. */
read_memory (this_saved_regs[regnum], bufferp,
register_size (current_gdbarch, regnum));
}
return;
}
/* No luck, assume this and the next frame have the same register
value. If a value is needed, pass the request on down the chain;
otherwise just return an indication that the value is in the same
register as the next frame. */
frame_register_unwind (next_frame, regnum, optimizedp, lvalp, addrp,
realnump, bufferp);
}
/* Put here the code to store, into fi->saved_regs, the addresses of
the saved registers of frame described by FRAME_INFO. This
includes special registers such as pc and fp saved in special ways
in the stack frame. sp is even more special: the address we return
for it IS the sp for the next frame. */
struct avr_unwind_cache *
avr_frame_unwind_cache (struct frame_info *next_frame,
void **this_prologue_cache)
{
CORE_ADDR pc;
ULONGEST prev_sp;
ULONGEST this_base;
struct avr_unwind_cache *info;
/* ULONGEST return_pc; */
int i;
if ((*this_prologue_cache))
return (*this_prologue_cache);
info = FRAME_OBSTACK_ZALLOC (struct avr_unwind_cache);
(*this_prologue_cache) = info;
info->saved_regs = FRAME_OBSTACK_CALLOC (NUM_REGS, CORE_ADDR);
info->size = 0;
/* Return PC is either the first 2 bytes of the this frame or the last two
bytes of the previsous frame. Not quite sure where the boundary is
yet. */
/* info->return_pc = 0; */
info->sp_offset = 0;
info->prologue_type = AVR_PROLOGUE_NONE;
pc = frame_func_unwind (next_frame);
avr_scan_prologue (pc, info);
if (info->prologue_type != AVR_PROLOGUE_NONE)
{
ULONGEST high_base; /* High byte of FP */
/* The SP was moved to the FP. This indicates that a new frame
was created. Get THIS frame's FP value by unwinding it from
the next frame. */
frame_unwind_unsigned_register (next_frame, AVR_FP_REGNUM, &this_base);
frame_unwind_unsigned_register (next_frame, AVR_FP_REGNUM+1, &high_base);
this_base += (high_base << 8);
/* The FP points at the last saved register. Adjust the FP back
to before the first saved register giving the SP. */
prev_sp = this_base + info->size;
}
else
{
/* Assume that the FP is this frame's SP but with that pushed
stack space added back. */
frame_unwind_unsigned_register (next_frame, AVR_SP_REGNUM, &this_base);
prev_sp = this_base + info->size;
}
info->base = avr_make_saddr (this_base);
info->prev_sp = avr_make_saddr (prev_sp);
/* Adjust all the saved registers so that they contain addresses and
not offsets. */
for (i = 0; i < NUM_REGS - 1; i++)
if (info->saved_regs[i])
{
info->saved_regs[i] = (info->prev_sp + info->saved_regs[i]);
}
#if 0
/* Calculate the return PC. Main has no return value on the stack, so ignore
that case. */
if (info->prologue_type != AVR_PROLOGUE_MAIN)
{
frame_unwind_unsigned_register (next_frame, AVR_SP_REGNUM, &return_pc);
info->return_pc = avr_make_iaddr (return_pc);
}
#endif
return info;
}
static CORE_ADDR
avr_unwind_pc (struct gdbarch *gdbarch, struct frame_info *next_frame)
{
ULONGEST pc;
frame_unwind_unsigned_register (next_frame, AVR_PC_REGNUM, &pc);
return avr_make_iaddr (pc);
}
/* Given a GDB frame, determine the address of the calling function's
frame. This will be used to create a new GDB frame struct. */
static void
avr_frame_this_id (struct frame_info *next_frame,
void **this_prologue_cache,
struct frame_id *this_id)
{
struct avr_unwind_cache *info
= avr_frame_unwind_cache (next_frame, this_prologue_cache);
CORE_ADDR base;
CORE_ADDR func;
struct frame_id id;
/* The FUNC is easy. */
func = frame_func_unwind (next_frame);
/* This is meant to halt the backtrace at "_start". Make sure we
don't halt it at a generic dummy frame. */
if (inside_entry_file (func))
return;
/* Hopefully the prologue analysis either correctly determined the
frame's base (which is the SP from the previous frame), or set
that base to "NULL". */
base = info->prev_sp;
if (base == 0)
return;
id = frame_id_build (base, func);
/* Check that we're not going round in circles with the same frame
ID (but avoid applying the test to sentinel frames which do go
round in circles). Can't use frame_id_eq() as that doesn't yet
compare the frame's PC value. */
if (frame_relative_level (next_frame) >= 0
&& get_frame_type (next_frame) != DUMMY_FRAME
&& frame_id_eq (get_frame_id (next_frame), id))
return;
(*this_id) = id;
}
static void
avr_frame_prev_register (struct frame_info *next_frame,
void **this_prologue_cache,
int regnum, int *optimizedp,
enum lval_type *lvalp, CORE_ADDR *addrp,
int *realnump, void *bufferp)
{
struct avr_unwind_cache *info
= avr_frame_unwind_cache (next_frame, this_prologue_cache);
avr_saved_regs_unwinder (next_frame, info->saved_regs, regnum, optimizedp,
lvalp, addrp, realnump, bufferp);
}
static const struct frame_unwind avr_frame_unwind = {
NORMAL_FRAME,
avr_frame_this_id,
avr_frame_prev_register
};
const struct frame_unwind *
avr_frame_p (CORE_ADDR pc)
{
return &avr_frame_unwind;
}
static CORE_ADDR
avr_frame_base_address (struct frame_info *next_frame, void **this_cache)
{
struct avr_unwind_cache *info
= avr_frame_unwind_cache (next_frame, this_cache);
return info->base;
}
static const struct frame_base avr_frame_base = {
&avr_frame_unwind,
avr_frame_base_address,
avr_frame_base_address,
avr_frame_base_address
};
/* Assuming NEXT_FRAME->prev is a dummy, return the frame ID of that
dummy frame. The frame ID's base needs to match the TOS value
saved by save_dummy_frame_tos(), and the PC match the dummy frame's
breakpoint. */
static struct frame_id
avr_unwind_dummy_id (struct gdbarch *gdbarch, struct frame_info *next_frame)
{
ULONGEST base;
frame_unwind_unsigned_register (next_frame, AVR_SP_REGNUM, &base);
return frame_id_build (avr_make_saddr (base), frame_pc_unwind (next_frame));
}
/* Initialize the gdbarch structure for the AVR's. */
static struct gdbarch *
avr_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
{
struct gdbarch *gdbarch;
struct gdbarch_tdep *tdep;
/* Find a candidate among the list of pre-declared architectures. */
arches = gdbarch_list_lookup_by_info (arches, &info);
if (arches != NULL)
return arches->gdbarch;
/* None found, create a new architecture from the information provided. */
tdep = XMALLOC (struct gdbarch_tdep);
gdbarch = gdbarch_alloc (&info, tdep);
/* If we ever need to differentiate the device types, do it here. */
switch (info.bfd_arch_info->mach)
{
case bfd_mach_avr1:
case bfd_mach_avr2:
case bfd_mach_avr3:
case bfd_mach_avr4:
case bfd_mach_avr5:
break;
}
set_gdbarch_short_bit (gdbarch, 2 * TARGET_CHAR_BIT);
set_gdbarch_int_bit (gdbarch, 2 * TARGET_CHAR_BIT);
set_gdbarch_long_bit (gdbarch, 4 * TARGET_CHAR_BIT);
set_gdbarch_long_long_bit (gdbarch, 8 * TARGET_CHAR_BIT);
set_gdbarch_ptr_bit (gdbarch, 2 * TARGET_CHAR_BIT);
set_gdbarch_addr_bit (gdbarch, 4 * TARGET_CHAR_BIT);
set_gdbarch_float_bit (gdbarch, 4 * TARGET_CHAR_BIT);
set_gdbarch_double_bit (gdbarch, 4 * TARGET_CHAR_BIT);
set_gdbarch_long_double_bit (gdbarch, 4 * TARGET_CHAR_BIT);
set_gdbarch_float_format (gdbarch, &floatformat_ieee_single_little);
set_gdbarch_double_format (gdbarch, &floatformat_ieee_single_little);
set_gdbarch_long_double_format (gdbarch, &floatformat_ieee_single_little);
set_gdbarch_read_pc (gdbarch, avr_read_pc);
set_gdbarch_write_pc (gdbarch, avr_write_pc);
set_gdbarch_read_sp (gdbarch, avr_read_sp);
set_gdbarch_num_regs (gdbarch, AVR_NUM_REGS);
set_gdbarch_sp_regnum (gdbarch, AVR_SP_REGNUM);
set_gdbarch_pc_regnum (gdbarch, AVR_PC_REGNUM);
set_gdbarch_register_name (gdbarch, avr_register_name);
set_gdbarch_register_type (gdbarch, avr_register_type);
set_gdbarch_print_insn (gdbarch, print_insn_avr);
set_gdbarch_call_dummy_address (gdbarch, avr_call_dummy_address);
set_gdbarch_address_to_pointer (gdbarch, avr_address_to_pointer);
set_gdbarch_pointer_to_address (gdbarch, avr_pointer_to_address);
set_gdbarch_use_struct_convention (gdbarch, generic_use_struct_convention);
set_gdbarch_skip_prologue (gdbarch, avr_skip_prologue);
set_gdbarch_inner_than (gdbarch, core_addr_lessthan);
set_gdbarch_decr_pc_after_break (gdbarch, 0);
set_gdbarch_breakpoint_from_pc (gdbarch, avr_breakpoint_from_pc);
set_gdbarch_function_start_offset (gdbarch, 0);
set_gdbarch_remote_translate_xfer_address (gdbarch,
avr_remote_translate_xfer_address);
set_gdbarch_frame_args_skip (gdbarch, 0);
set_gdbarch_frameless_function_invocation (gdbarch,
frameless_look_for_prologue);
set_gdbarch_frame_args_address (gdbarch, avr_frame_address);
set_gdbarch_frame_locals_address (gdbarch, avr_frame_address);
set_gdbarch_frame_num_args (gdbarch, frame_num_args_unknown);
set_gdbarch_convert_from_func_ptr_addr (gdbarch,
avr_convert_from_func_ptr_addr);
frame_unwind_append_predicate (gdbarch, avr_frame_p);
frame_base_set_default (gdbarch, &avr_frame_base);
set_gdbarch_unwind_dummy_id (gdbarch, avr_unwind_dummy_id);
set_gdbarch_save_dummy_frame_tos (gdbarch, generic_save_dummy_frame_tos);
set_gdbarch_unwind_pc (gdbarch, avr_unwind_pc);
return gdbarch;
}
/* Send a query request to the avr remote target asking for values of the io
registers. If args parameter is not NULL, then the user has requested info
on a specific io register [This still needs implemented and is ignored for
now]. The query string should be one of these forms:
"Ravr.io_reg" -> reply is "NN" number of io registers
"Ravr.io_reg:addr,len" where addr is first register and len is number of
registers to be read. The reply should be "<NAME>,VV;" for each io register
where, <NAME> is a string, and VV is the hex value of the register.
All io registers are 8-bit. */
static void
avr_io_reg_read_command (char *args, int from_tty)
{
int bufsiz = 0;
char buf[400];
char query[400];
char *p;
unsigned int nreg = 0;
unsigned int val;
int i, j, k, step;
/* fprintf_unfiltered (gdb_stderr, "DEBUG: avr_io_reg_read_command " */
/* "(\"%s\", %d)\n", args, from_tty); */
if (!current_target.to_query)
{
fprintf_unfiltered (gdb_stderr,
"ERR: info io_registers NOT supported by current "
"target\n");
return;
}
/* Just get the maximum buffer size. */
target_query ((int) 'R', 0, 0, &bufsiz);
if (bufsiz > sizeof (buf))
bufsiz = sizeof (buf);
/* Find out how many io registers the target has. */
strcpy (query, "avr.io_reg");
target_query ((int) 'R', query, buf, &bufsiz);
if (strncmp (buf, "", bufsiz) == 0)
{
fprintf_unfiltered (gdb_stderr,
"info io_registers NOT supported by target\n");
return;
}
if (sscanf (buf, "%x", &nreg) != 1)
{
fprintf_unfiltered (gdb_stderr,
"Error fetching number of io registers\n");
return;
}
reinitialize_more_filter ();
printf_unfiltered ("Target has %u io registers:\n\n", nreg);
/* only fetch up to 8 registers at a time to keep the buffer small */
step = 8;
for (i = 0; i < nreg; i += step)
{
/* how many registers this round? */
j = step;
if ((i+j) >= nreg)
j = nreg - i; /* last block is less than 8 registers */
snprintf (query, sizeof (query) - 1, "avr.io_reg:%x,%x", i, j);
target_query ((int) 'R', query, buf, &bufsiz);
p = buf;
for (k = i; k < (i + j); k++)
{
if (sscanf (p, "%[^,],%x;", query, &val) == 2)
{
printf_filtered ("[%02x] %-15s : %02x\n", k, query, val);
while ((*p != ';') && (*p != '\0'))
p++;
p++; /* skip over ';' */
if (*p == '\0')
break;
}
}
}
}
void
_initialize_avr_tdep (void)
{
register_gdbarch_init (bfd_arch_avr, avr_gdbarch_init);
/* Add a new command to allow the user to query the avr remote target for
the values of the io space registers in a saner way than just using
`x/NNNb ADDR`. */
/* FIXME: TRoth/2002-02-18: This should probably be changed to 'info avr
io_registers' to signify it is not available on other platforms. */
add_cmd ("io_registers", class_info, avr_io_reg_read_command,
"query remote avr target for io space register values", &infolist);
}
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: avr and frame unwinding
2003-06-06 0:35 avr and frame unwinding Theodore A. Roth
@ 2003-06-06 2:01 ` Andrew Cagney
2003-06-07 20:23 ` Theodore A. Roth
0 siblings, 1 reply; 7+ messages in thread
From: Andrew Cagney @ 2003-06-06 2:01 UTC (permalink / raw)
To: Theodore A. Roth; +Cc: gdb
Can you try the exact same operation with a GDB that doesn't have the
changes? That way it should be possible to compare the two traces
side-by-side and see where things are going wrong.
A guess is that it is getting the unwound PC value wrong. For the d10v
I had to make two tweaks:
- abort the prologue scanner when it reached PC if that is before the
end of the prologue
This stopped the prologue getting the unwound PC's location wrong. It
may not have yet executed the save PC instruction.
- track the register that the PC is in before it is saved
The code needs to be able to unwind the PC value in the prologue before
it has been saved on the stack.
Andrew
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: avr and frame unwinding
2003-06-06 2:01 ` Andrew Cagney
@ 2003-06-07 20:23 ` Theodore A. Roth
2003-06-07 22:37 ` Andrew Cagney
0 siblings, 1 reply; 7+ messages in thread
From: Theodore A. Roth @ 2003-06-07 20:23 UTC (permalink / raw)
To: Andrew Cagney; +Cc: gdb
On Thu, 5 Jun 2003, Andrew Cagney wrote:
:)Can you try the exact same operation with a GDB that doesn't have the
:)changes? That way it should be possible to compare the two traces
:)side-by-side and see where things are going wrong.
:)
:)A guess is that it is getting the unwound PC value wrong. For the d10v
That was the problem. I've got it working, but the solution was ugly (see
below).
:)I had to make two tweaks:
:)
:)- abort the prologue scanner when it reached PC if that is before the
:)end of the prologue
:)This stopped the prologue getting the unwound PC's location wrong. It
:)may not have yet executed the save PC instruction.
This seems to also hold for the avr. Adding the extra pc check before
prologue scanning like the d10v does got things working as far as stepping
and backtracing go.
:)
:)- track the register that the PC is in before it is saved
:)The code needs to be able to unwind the PC value in the prologue before
:)it has been saved on the stack.
:)
:)Andrew
:)
Thanks for the nudge Andrew.
Any idea how to make this piece of code sane? I'm not thrilled with what I
had to do to get the PC read and converted then finally stored in bufferp.
This function is analogous to saved_regs_unwinder() in d10v-tdep.c.
static void
avr_saved_regs_unwinder (struct frame_info *next_frame,
CORE_ADDR *this_saved_regs,
int regnum, int *optimizedp,
enum lval_type *lvalp, CORE_ADDR *addrp,
int *realnump, void *bufferp)
{
if (this_saved_regs[regnum] != 0)
{
*optimizedp = 0;
*lvalp = lval_memory;
*addrp = this_saved_regs[regnum];
*realnump = -1;
if (bufferp != NULL)
{
/* Read the value in from memory. */
if (regnum == AVR_PC_REGNUM)
{
/* Reading the return PC from the PC register is slightly
abnormal. register_size(AVR_PC_REGNUM) says it is 4 bytes,
but in reality, only two bytes (3 in upcoming mega256) are
stored on the stack.
Also, note that the value on the stack is an addr to a word
not a byte, so we will need to multiply it by two at some
point. */
ULONGEST pc;
unsigned char buf[2];
read_memory (this_saved_regs[regnum], buf, sizeof (buf));
pc = (extract_unsigned_integer (buf, 2) * 2) >> 8;
memcpy (bufferp, &pc, sizeof(pc));
}
else
{
read_memory (this_saved_regs[regnum], bufferp,
register_size (current_gdbarch, regnum));
}
}
return;
}
/* No luck, assume this and the next frame have the same register
value. If a value is needed, pass the request on down the chain;
otherwise just return an indication that the value is in the same
register as the next frame. */
frame_register_unwind (next_frame, regnum, optimizedp, lvalp, addrp,
realnump, bufferp);
}
The last problem I need to solve I think is also related to register
unwinding. If I step down into a function, then 'up' to move up the stack
frame, examining a local variable gives the wrong value. I need to do some
more research into why this is happening though. Am I on the right track
with thinking the register unwinding could be the problem? Maybe I'm setting
the frame base incorrectly since it doesn't seem to resolve the variable's
memory address?
Ted Roth
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: avr and frame unwinding
2003-06-07 20:23 ` Theodore A. Roth
@ 2003-06-07 22:37 ` Andrew Cagney
2003-06-08 0:26 ` Theodore A. Roth
0 siblings, 1 reply; 7+ messages in thread
From: Andrew Cagney @ 2003-06-07 22:37 UTC (permalink / raw)
To: Theodore A. Roth; +Cc: gdb
> On Thu, 5 Jun 2003, Andrew Cagney wrote:
>
> :)Can you try the exact same operation with a GDB that doesn't have the
> :)changes? That way it should be possible to compare the two traces
> :)side-by-side and see where things are going wrong.
> :)
> :)A guess is that it is getting the unwound PC value wrong. For the d10v
>
> That was the problem. I've got it working, but the solution was ugly (see
> below).
>
> :)I had to make two tweaks:
> :)
> :)- abort the prologue scanner when it reached PC if that is before the
> :)end of the prologue
> :)This stopped the prologue getting the unwound PC's location wrong. It
> :)may not have yet executed the save PC instruction.
>
> This seems to also hold for the avr. Adding the extra pc check before
> prologue scanning like the d10v does got things working as far as stepping
> and backtracing go.
>
> :)
> :)- track the register that the PC is in before it is saved
> :)The code needs to be able to unwind the PC value in the prologue before
> :)it has been saved on the stack.
> :)
> :)Andrew
> :)
>
> Thanks for the nudge Andrew.
>
> Any idea how to make this piece of code sane? I'm not thrilled with what I
> had to do to get the PC read and converted then finally stored in bufferp.
> This function is analogous to saved_regs_unwinder() in d10v-tdep.c.
>
> static void
> avr_saved_regs_unwinder (struct frame_info *next_frame,
> CORE_ADDR *this_saved_regs,
> int regnum, int *optimizedp,
> enum lval_type *lvalp, CORE_ADDR *addrp,
> int *realnump, void *bufferp)
> {
> if (this_saved_regs[regnum] != 0)
> {
> *optimizedp = 0;
> *lvalp = lval_memory;
> *addrp = this_saved_regs[regnum];
> *realnump = -1;
> if (bufferp != NULL)
> {
> /* Read the value in from memory. */
>
> if (regnum == AVR_PC_REGNUM)
> {
> /* Reading the return PC from the PC register is slightly
> abnormal. register_size(AVR_PC_REGNUM) says it is 4 bytes,
> but in reality, only two bytes (3 in upcoming mega256) are
> stored on the stack.
Is the PC 2 bytes, or that only two bytes are saved? You don't have a
d10v PC which is 2 bytes but addresses words?
> Also, note that the value on the stack is an addr to a word
> not a byte, so we will need to multiply it by two at some
> point. */
>
> ULONGEST pc;
> unsigned char buf[2];
>
> read_memory (this_saved_regs[regnum], buf, sizeof (buf));
Is this arrithmetic correct - I understand the ``* 2'' but not the ``>>8''.
> pc = (extract_unsigned_integer (buf, 2) * 2) >> 8;
this memcpy will need to be a
store_unsigned_integer (bufferp, pc, SIZEOF_AVR_PC);
> memcpy (bufferp, &pc, sizeof(pc));
> }
> else
> {
> read_memory (this_saved_regs[regnum], bufferp,
> register_size (current_gdbarch, regnum));
> }
> The last problem I need to solve I think is also related to register
> unwinding. If I step down into a function, then 'up' to move up the stack
> frame, examining a local variable gives the wrong value. I need to do some
> more research into why this is happening though. Am I on the right track
> with thinking the register unwinding could be the problem? Maybe I'm setting
> the frame base incorrectly since it doesn't seem to resolve the variable's
> memory address?
Sounds like you're on the right track. The new unwind code typically
has different:
frame ID.stack_addr
frame {base,locals,args} address
the former is the address of the previous frame's inner-most stack
address while the latter is the clasic frame pointer.
Andrew
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: avr and frame unwinding
2003-06-07 22:37 ` Andrew Cagney
@ 2003-06-08 0:26 ` Theodore A. Roth
2003-06-08 17:51 ` Theodore A. Roth
0 siblings, 1 reply; 7+ messages in thread
From: Theodore A. Roth @ 2003-06-08 0:26 UTC (permalink / raw)
To: Andrew Cagney; +Cc: gdb
On Sat, 7 Jun 2003, Andrew Cagney wrote:
:)> Any idea how to make this piece of code sane? I'm not thrilled with what I
:)> had to do to get the PC read and converted then finally stored in bufferp.
:)> This function is analogous to saved_regs_unwinder() in d10v-tdep.c.
:)>
:)> static void
:)> avr_saved_regs_unwinder (struct frame_info *next_frame,
:)> CORE_ADDR *this_saved_regs,
:)> int regnum, int *optimizedp,
:)> enum lval_type *lvalp, CORE_ADDR *addrp,
:)> int *realnump, void *bufferp)
:)> {
:)> if (this_saved_regs[regnum] != 0)
:)> {
:)> *optimizedp = 0;
:)> *lvalp = lval_memory;
:)> *addrp = this_saved_regs[regnum];
:)> *realnump = -1;
:)> if (bufferp != NULL)
:)> {
:)> /* Read the value in from memory. */
:)>
:)> if (regnum == AVR_PC_REGNUM)
:)> {
:)> /* Reading the return PC from the PC register is slightly
:)> abnormal. register_size(AVR_PC_REGNUM) says it is 4 bytes,
:)> but in reality, only two bytes (3 in upcoming mega256) are
:)> stored on the stack.
:)
:)Is the PC 2 bytes, or that only two bytes are saved? You don't have a
:)d10v PC which is 2 bytes but addresses words?
The PC is stored on the stack as an unsigned 16-bit value which addresses
64K of 16-bit wide instructions. The problem is our remote targets perform
the word to byte addressing convertion before sending the PC register value
over the RSP. In this case, I have to read the PC from memory, so the remote
target only thinks it is making a memory read (which by the way uses a byte
addressing scheme). After reading the PC from memory, I need to convert it
to a byte address (multiply by 2).
:)
:)> Also, note that the value on the stack is an addr to a word
:)> not a byte, so we will need to multiply it by two at some
:)> point. */
:)>
:)> ULONGEST pc;
:)> unsigned char buf[2];
:)>
:)> read_memory (this_saved_regs[regnum], buf, sizeof (buf));
:)
:)Is this arrithmetic correct - I understand the ``* 2'' but not the ``>>8''.
:)
:)> pc = (extract_unsigned_integer (buf, 2) * 2) >> 8;
That's the ugly part I don't understand. It seems to give the correct
result, but now that I think about it more, it could mean that my memory
address is off by 1. I will have to re-examine that.
:)
:)this memcpy will need to be a
:)
:) store_unsigned_integer (bufferp, pc, SIZEOF_AVR_PC);
:)
I tried that, but it performed a endian byte swap and the PC came out wrong.
I dug around and saw what looked to be too many byte swaps.
:)> memcpy (bufferp, &pc, sizeof(pc));
:)> }
:)> else
:)> {
:)> read_memory (this_saved_regs[regnum], bufferp,
:)> register_size (current_gdbarch, regnum));
:)> }
:)
:)> The last problem I need to solve I think is also related to register
:)> unwinding. If I step down into a function, then 'up' to move up the stack
:)> frame, examining a local variable gives the wrong value. I need to do some
:)> more research into why this is happening though. Am I on the right track
:)> with thinking the register unwinding could be the problem? Maybe I'm setting
:)> the frame base incorrectly since it doesn't seem to resolve the variable's
:)> memory address?
:)
:)Sounds like you're on the right track. The new unwind code typically
:)has different:
:)
:) frame ID.stack_addr
:) frame {base,locals,args} address
:)
:)the former is the address of the previous frame's inner-most stack
:)address while the latter is the clasic frame pointer.
Thanks.
Now back to getting my brain wrapped around this.
Ted Roth
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: avr and frame unwinding
2003-06-08 0:26 ` Theodore A. Roth
@ 2003-06-08 17:51 ` Theodore A. Roth
2003-06-08 19:15 ` Andrew Cagney
0 siblings, 1 reply; 7+ messages in thread
From: Theodore A. Roth @ 2003-06-08 17:51 UTC (permalink / raw)
To: Andrew Cagney; +Cc: gdb
On Sat, 7 Jun 2003, Theodore A. Roth wrote:
:)On Sat, 7 Jun 2003, Andrew Cagney wrote:
:)
:):)> Any idea how to make this piece of code sane? I'm not thrilled with what I
:):)> had to do to get the PC read and converted then finally stored in bufferp.
:):)> This function is analogous to saved_regs_unwinder() in d10v-tdep.c.
:):)>
:):)> static void
:):)> avr_saved_regs_unwinder (struct frame_info *next_frame,
:):)> CORE_ADDR *this_saved_regs,
:):)> int regnum, int *optimizedp,
:):)> enum lval_type *lvalp, CORE_ADDR *addrp,
:):)> int *realnump, void *bufferp)
:):)> {
:):)> if (this_saved_regs[regnum] != 0)
:):)> {
:):)> *optimizedp = 0;
:):)> *lvalp = lval_memory;
:):)> *addrp = this_saved_regs[regnum];
:):)> *realnump = -1;
:):)> if (bufferp != NULL)
:):)> {
:):)> /* Read the value in from memory. */
:):)>
:):)> if (regnum == AVR_PC_REGNUM)
:):)> {
:):)> /* Reading the return PC from the PC register is slightly
:):)> abnormal. register_size(AVR_PC_REGNUM) says it is 4 bytes,
:):)> but in reality, only two bytes (3 in upcoming mega256) are
:):)> stored on the stack.
:):)
:):)Is the PC 2 bytes, or that only two bytes are saved? You don't have a
:):)d10v PC which is 2 bytes but addresses words?
:)
:)The PC is stored on the stack as an unsigned 16-bit value which addresses
:)64K of 16-bit wide instructions. The problem is our remote targets perform
:)the word to byte addressing convertion before sending the PC register value
:)over the RSP. In this case, I have to read the PC from memory, so the remote
:)target only thinks it is making a memory read (which by the way uses a byte
:)addressing scheme). After reading the PC from memory, I need to convert it
:)to a byte address (multiply by 2).
:)
:):)
:):)> Also, note that the value on the stack is an addr to a word
:):)> not a byte, so we will need to multiply it by two at some
:):)> point. */
:):)>
:):)> ULONGEST pc;
:):)> unsigned char buf[2];
:):)>
:):)> read_memory (this_saved_regs[regnum], buf, sizeof (buf));
:):)
:):)Is this arrithmetic correct - I understand the ``* 2'' but not the ``>>8''.
:):)
:):)> pc = (extract_unsigned_integer (buf, 2) * 2) >> 8;
:)
:)That's the ugly part I don't understand. It seems to give the correct
:)result, but now that I think about it more, it could mean that my memory
:)address is off by 1. I will have to re-examine that.
:)
:):)
:):)this memcpy will need to be a
:):)
:):) store_unsigned_integer (bufferp, pc, SIZEOF_AVR_PC);
:):)
:)
:)I tried that, but it performed a endian byte swap and the PC came out wrong.
:)I dug around and saw what looked to be too many byte swaps.
:)
:):)> memcpy (bufferp, &pc, sizeof(pc));
:):)> }
:):)> else
:):)> {
:):)> read_memory (this_saved_regs[regnum], bufferp,
:):)> register_size (current_gdbarch, regnum));
:):)> }
I think found the root of the ugliness. When the avr performs a call
instruction the PC is pushed onto the stack, but it turns out that it is
pushed in big endian order. For the most part though, the avr is little
endian.
Ted Roth
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: avr and frame unwinding
2003-06-08 17:51 ` Theodore A. Roth
@ 2003-06-08 19:15 ` Andrew Cagney
0 siblings, 0 replies; 7+ messages in thread
From: Andrew Cagney @ 2003-06-08 19:15 UTC (permalink / raw)
To: Theodore A. Roth; +Cc: gdb
> :):)Is this arrithmetic correct - I understand the ``* 2'' but not the ``>>8''.
> :):)
> :):)> pc = (extract_unsigned_integer (buf, 2) * 2) >> 8;
> :)
> :)That's the ugly part I don't understand. It seems to give the correct
> :)result, but now that I think about it more, it could mean that my memory
> :)address is off by 1. I will have to re-examine that.
> :)
> :):)
> :):)this memcpy will need to be a
> :):)
> :):) store_unsigned_integer (bufferp, pc, SIZEOF_AVR_PC);
> :):)
> :)
> :)I tried that, but it performed a endian byte swap and the PC came out wrong.
> :)I dug around and saw what looked to be too many byte swaps.
> :)
> :):)> memcpy (bufferp, &pc, sizeof(pc));
> :):)> }
> :):)> else
> :):)> {
> :):)> read_memory (this_saved_regs[regnum], bufferp,
> :):)> register_size (current_gdbarch, regnum));
> :):)> }
>
> I think found the root of the ugliness. When the avr performs a call
> instruction the PC is pushed onto the stack, but it turns out that it is
> pushed in big endian order. For the most part though, the avr is little
> endian.
That would explain it. I guess it needs an explicit big endian extract
followed by a little endian store (via store_unsigned_integer).
Andrew
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2003-06-08 19:15 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-06-06 0:35 avr and frame unwinding Theodore A. Roth
2003-06-06 2:01 ` Andrew Cagney
2003-06-07 20:23 ` Theodore A. Roth
2003-06-07 22:37 ` Andrew Cagney
2003-06-08 0:26 ` Theodore A. Roth
2003-06-08 17:51 ` Theodore A. Roth
2003-06-08 19:15 ` Andrew Cagney
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).