This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
[patch rfc; rfa:doco] frame_sp_unwind
- From: Andrew Cagney <ac131313 at redhat dot com>
- To: gdb-patches at sources dot redhat dot com
- Date: Fri, 06 Jun 2003 15:34:40 -0400
- Subject: [patch rfc; rfa:doco] frame_sp_unwind
Hello,
This patch adds a per-architecture gdbarch_unwind_sp and the frame
wrapper methods:
get_frame_sp()
and
frame_sp_unwind()
as replacements for TARGET_READ_SP. The key advantage of this new
method is that it takes a ``struct frame_info'' parameter and hence
avoids any global state.
It also modifies read_sp() so that it will use get_frame_sp and hence
provide a certain level of forward compatibility. Similarly,
frame_sp_unwind() has been written in a way that provides a certain
level of backward compatibility.
For the d10v, sets UNWIND_SP instead of READ_SP and eliminates another
global regcache/target reference.
Follow-on changes will include things like replace read_sp() with
get_frame_sp(), eliminate generic_target_read_sp(), and deprecate
TARGET_READ_SP.
Eli, note the doco change.
I'll look to commit this in a few days.
Andrew
2003-06-06 Andrew Cagney <cagney@redhat.com>
* gdbarch.sh (UNWIND_SP): Add.
(TARGET_READ_SP): Change to function with predicate.
* gdbarch.h, gdbarch.c: Re-generate.
* frame.c (frame_sp_unwind): New function.
(get_frame_sp): New function.
* frame.h (get_frame_sp, frame_sp_unwind): Declare.
* regcache.c (read_sp): Rewrite, try each of TARGET_READ_SP,
gdbarch_unwind_sp and SP_REGNUM when looking for the SP register
value.
* d10v-tdep.c (d10v_unwind_sp): Replace d10v_read_sp.
(d10v_gdbarch_init): Set unwind_sp instead of read_sp.
Index: doc/ChangeLog
2003-06-06 Andrew Cagney <cagney@redhat.com>
* gdbint.texinfo (Target Architecture Definition): Document
"unwind_sp". Cross reference "unwind_sp" and TARGET_READ_SP.
Index: d10v-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/d10v-tdep.c,v
retrieving revision 1.118
diff -u -r1.118 d10v-tdep.c
--- d10v-tdep.c 3 Jun 2003 18:53:37 -0000 1.118
+++ d10v-tdep.c 6 Jun 2003 19:18:57 -0000
@@ -101,8 +101,6 @@
extern void _initialize_d10v_tdep (void);
-static CORE_ADDR d10v_read_sp (void);
-
static void d10v_eva_prepare_to_trace (void);
static void d10v_eva_get_trace_data (void);
@@ -918,9 +916,11 @@
}
static CORE_ADDR
-d10v_read_sp (void)
+d10v_unwind_sp (struct gdbarch *gdbarch, struct frame_info *next_frame)
{
- return (d10v_make_daddr (read_register (D10V_SP_REGNUM)));
+ ULONGEST sp;
+ frame_unwind_unsigned_register (next_frame, D10V_SP_REGNUM, &sp);
+ return d10v_make_daddr (sp);
}
/* When arguments must be pushed onto the stack, they go on in reverse
@@ -1604,7 +1604,7 @@
set_gdbarch_read_pc (gdbarch, d10v_read_pc);
set_gdbarch_write_pc (gdbarch, d10v_write_pc);
- set_gdbarch_read_sp (gdbarch, d10v_read_sp);
+ set_gdbarch_unwind_sp (gdbarch, d10v_unwind_sp);
set_gdbarch_num_regs (gdbarch, d10v_num_regs);
set_gdbarch_sp_regnum (gdbarch, D10V_SP_REGNUM);
Index: frame.c
===================================================================
RCS file: /cvs/src/src/gdb/frame.c,v
retrieving revision 1.117
diff -u -r1.117 frame.c
--- frame.c 3 Jun 2003 18:53:37 -0000 1.117
+++ frame.c 6 Jun 2003 19:18:57 -0000
@@ -2250,6 +2250,37 @@
return current_gdbarch;
}
+/* Stack pointer methods. */
+
+CORE_ADDR
+get_frame_sp (struct frame_info *this_frame)
+{
+ return frame_sp_unwind (this_frame->next);
+}
+
+CORE_ADDR
+frame_sp_unwind (struct frame_info *next_frame)
+{
+ /* Normality, an architecture that provides a way of obtaining any
+ frame inner-most address. */
+ if (gdbarch_unwind_sp_p (current_gdbarch))
+ return gdbarch_unwind_sp (current_gdbarch, next_frame);
+ /* Things are looking grim. If it's the inner-most frame and there
+ is a TARGET_READ_SP then that can be used. */
+ if (next_frame->level < 0 && TARGET_READ_SP_P ())
+ return TARGET_READ_SP ();
+ /* Now things are really are grim. Hope that the value returned by
+ the SP_REGNUM register is meaningful. */
+ if (SP_REGNUM >= 0)
+ {
+ ULONGEST sp;
+ frame_unwind_unsigned_register (next_frame, SP_REGNUM, &sp);
+ return sp;
+ }
+ internal_error (__FILE__, __LINE__, "Missing unwind SP method");
+}
+
+
int
legacy_frame_p (struct gdbarch *current_gdbarch)
{
Index: frame.h
===================================================================
RCS file: /cvs/src/src/gdb/frame.h,v
retrieving revision 1.96
diff -u -r1.96 frame.h
--- frame.h 3 Jun 2003 18:53:37 -0000 1.96
+++ frame.h 6 Jun 2003 19:19:00 -0000
@@ -166,6 +166,13 @@
This replaced: frame->pc; */
extern CORE_ADDR get_frame_pc (struct frame_info *);
+/* The frame's inner-most bound. AKA the stack-pointer. Confusingly
+ known as top-of-stack. */
+
+extern CORE_ADDR get_frame_sp (struct frame_info *);
+extern CORE_ADDR frame_sp_unwind (struct frame_info *);
+
+
/* Following on from the `resume' address. Return the entry point
address of the function containing that resume address, or zero if
that function isn't known. */
Index: gdbarch.sh
===================================================================
RCS file: /cvs/src/src/gdb/gdbarch.sh,v
retrieving revision 1.240
diff -u -r1.240 gdbarch.sh
--- gdbarch.sh 2 Jun 2003 02:54:34 -0000 1.240
+++ gdbarch.sh 6 Jun 2003 19:19:07 -0000
@@ -431,7 +431,8 @@
# This is simply not needed. See value_of_builtin_frame_fp_reg and
# call_function_by_hand.
F::DEPRECATED_TARGET_READ_FP:CORE_ADDR:deprecated_target_read_fp:void
-f:2:TARGET_READ_SP:CORE_ADDR:read_sp:void:::0:generic_target_read_sp::0
+# UNWIND_SP is a direct replacement for TARGET_READ_SP.
+F:2:TARGET_READ_SP:CORE_ADDR:read_sp:void
# The dummy call frame SP should be set by push_dummy_call.
F:2:DEPRECATED_DUMMY_WRITE_SP:void:deprecated_dummy_write_sp:CORE_ADDR val:val
# Function for getting target's idea of a frame pointer. FIXME: GDB's
@@ -452,6 +453,7 @@
# GDB's standard (or well known) register numbers. These can map onto
# a real register or a pseudo (computed) register or not be defined at
# all (-1).
+# SP_REGNUM will hopefully be replaced by UNWIND_SP.
v:2:SP_REGNUM:int:sp_regnum::::-1:-1::0
# This is simply not needed. See value_of_builtin_frame_fp_reg and
# call_function_by_hand.
@@ -609,6 +611,7 @@
# interfaces they have very different underlying implementations.
F:2:DEPRECATED_FRAME_SAVED_PC:CORE_ADDR:deprecated_frame_saved_pc:struct frame_info *fi:fi::0:0
M::UNWIND_PC:CORE_ADDR:unwind_pc:struct frame_info *next_frame:next_frame:
+M::UNWIND_SP:CORE_ADDR:unwind_sp:struct frame_info *next_frame:next_frame:
f:2:FRAME_ARGS_ADDRESS:CORE_ADDR:frame_args_address:struct frame_info *fi:fi::0:get_frame_base::0
f:2:FRAME_LOCALS_ADDRESS:CORE_ADDR:frame_locals_address:struct frame_info *fi:fi::0:get_frame_base::0
F::DEPRECATED_SAVED_PC_AFTER_CALL:CORE_ADDR:deprecated_saved_pc_after_call:struct frame_info *frame:frame
Index: regcache.c
===================================================================
RCS file: /cvs/src/src/gdb/regcache.c,v
retrieving revision 1.82
diff -u -r1.82 regcache.c
--- regcache.c 17 May 2003 05:59:58 -0000 1.82
+++ regcache.c 6 Jun 2003 19:19:09 -0000
@@ -1375,7 +1375,26 @@
CORE_ADDR
read_sp (void)
{
- return TARGET_READ_SP ();
+ /* Try TARGET_READ_SP first: that way code that provides
+ TARGET_READ_SP method preserves its current behavior. */
+ if (TARGET_READ_SP_P ())
+ return TARGET_READ_SP ();
+ /* Try UNWIND_SP second: ensures that architectures which need to
+ unwind/unpack SP in weird and wonderful ways always override
+ SP_REGNUM. The d10v, for instance needs to convert the SP
+ pointer register into a CORE_ADDR. The MIPS needs to sign extend
+ its SP register. */
+ if (gdbarch_unwind_sp_p (current_gdbarch))
+ return get_frame_sp (get_current_frame ());
+ /* Try SP_REGNUM last: this makes all sorts of [wrong] assumptions
+ about the architecture so put it at the end. */
+ if (SP_REGNUM >= 0)
+ {
+ ULONGEST sp;
+ regcache_cooked_read_unsigned (current_regcache, SP_REGNUM, &sp);
+ return sp;
+ }
+ internal_error (__FILE__, __LINE__, "read_sp");
}
void
Index: doc/gdbint.texinfo
===================================================================
RCS file: /cvs/src/src/gdb/doc/gdbint.texinfo,v
retrieving revision 1.148
diff -u -r1.148 gdbint.texinfo
--- doc/gdbint.texinfo 1 Jun 2003 23:05:43 -0000 1.148
+++ doc/gdbint.texinfo 6 Jun 2003 19:19:14 -0000
@@ -3291,6 +3291,23 @@
@noindent
@xref{DEPRECATED_FRAME_SAVED_PC}, which this method replaces.
+@item CORE_ADDR unwind_sp (struct frame_info *@var{this_frame})
+@findex unwind_sp
+@anchor{unwind_sp} Return the frame's inner most stack address. This is
+commonly refered to as the frame's stack pointer.
+
+The implementation, which must be frame agnostic (work with any frame),
+is typically no more than:
+
+@smallexample
+ULONGEST sp;
+frame_unwind_unsigned_register (this_frame, D10V_SP_REGNUM, &sp);
+return d10v_make_daddr (sp);
+@end smallexample
+
+@noindent
+@xref{TARGET_READ_SP}, which this method replaces.
+
@item FUNCTION_EPILOGUE_SIZE
@findex FUNCTION_EPILOGUE_SIZE
For some COFF targets, the @code{x_sym.x_misc.x_fsize} field of the
@@ -3890,14 +3907,16 @@
@findex write_pc
@findex read_sp
@findex read_fp
-These change the behavior of @code{read_pc}, @code{write_pc},
-@code{read_sp} and @code{deprecated_read_fp}. For most targets, these
-may be left undefined. @value{GDBN} will call the read and write
-register functions with the relevant @code{_REGNUM} argument.
+@anchor{TARGET_READ_SP} These change the behavior of @code{read_pc},
+@code{write_pc}, @code{read_sp} and @code{deprecated_read_fp}. For most
+targets, these may be left undefined. @value{GDBN} will call the read
+and write register functions with the relevant @code{_REGNUM} argument.
These macros are useful when a target keeps one of these registers in a
hard to get at place; for example, part in a segment register and part
in an ordinary register.
+
+@xref{unwind_sp}, which replaces @code{TARGET_READ_SP}.
@item TARGET_VIRTUAL_FRAME_POINTER(@var{pc}, @var{regp}, @var{offsetp})
@findex TARGET_VIRTUAL_FRAME_POINTER