[rfa:ppc64] Eliminate call_dummy_address

[rfa:ppc64] Eliminate call_dummy_address

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 381 bytes --]

Hello,

This patch does two things:

- It modifies "infcall.c" so that it converts the entry-point address 
into a code address (it might be a descriptor).

- since it's now redundant, for ppc64 GNU/Linux, it stops setting 
entry_point_address

Kevin, ppc64 part ok?  It doesn't immediatly affect test results as this 
GDB still requires a mod to relocate the entry point.

Andrew

[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 2548 bytes --]

2003-08-27  Andrew Cagney  <cagney@redhat.com>

	* ppc-linux-tdep.c (ppc64_call_dummy_address): Delete function.
	(ppc_linux_init_abi): For PPC64, do not set call_dummy_address.
	* infcall.c (call_function_by_hand): Convert the entry point
	address into a code address.
 
Index: infcall.c
===================================================================
RCS file: /cvs/src/src/gdb/infcall.c,v
retrieving revision 1.23
diff -u -r1.23 infcall.c
--- infcall.c	18 Aug 2003 20:04:55 -0000	1.23
+++ infcall.c	27 Aug 2003 20:03:46 -0000
@@ -558,6 +558,9 @@
 	}
       real_pc = funaddr;
       dummy_addr = CALL_DUMMY_ADDRESS ();
+      /* Make certain that the address points at real code, and not a
+         function descriptor.  */
+      dummy_addr = CONVERT_FROM_FUNC_PTR_ADDR (dummy_addr);
       /* A call dummy always consists of just a single breakpoint, so
          it's address is the same as the address of the dummy.  */
       bp_addr = dummy_addr;
@@ -576,6 +579,11 @@
 	  dummy_addr = SYMBOL_VALUE_ADDRESS (sym);
 	else
 	  dummy_addr = entry_point_address ();
+	/* Make certain that the address points at real code, and not
+	   a function descriptor.  */
+	dummy_addr = CONVERT_FROM_FUNC_PTR_ADDR (dummy_addr);
+	/* A call dummy always consists of just a single breakpoint,
+	   so it's address is the same as the address of the dummy.  */
 	bp_addr = dummy_addr;
 	break;
       }
Index: ppc-linux-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/ppc-linux-tdep.c,v
retrieving revision 1.37
diff -u -r1.37 ppc-linux-tdep.c
--- ppc-linux-tdep.c	24 Jun 2003 23:09:22 -0000	1.37
+++ ppc-linux-tdep.c	27 Aug 2003 20:03:51 -0000
@@ -929,18 +929,6 @@
 }
 
 
-/* On 64-bit PowerPC GNU/Linux, the ELF header's e_entry field is the
-   address of a function descriptor for the entry point function, not
-   the actual entry point itself.  So to find the actual address at
-   which execution should begin, we need to fetch the function's entry
-   point from that descriptor.  */
-static CORE_ADDR
-ppc64_call_dummy_address (void)
-{
-  return ppc64_desc_entry_point (entry_point_address ());
-}
-
-
 enum {
   ELF_NGREG = 48,
   ELF_NFPREG = 33,
@@ -1066,8 +1054,6 @@
          function descriptors).  */
       set_gdbarch_convert_from_func_ptr_addr
         (gdbarch, ppc64_linux_convert_from_func_ptr_addr);
-
-      set_gdbarch_call_dummy_address (gdbarch, ppc64_call_dummy_address);
 
       set_gdbarch_in_solib_call_trampoline
         (gdbarch, ppc64_in_solib_call_trampoline);

Re: [rfa:ppc64] Eliminate call_dummy_address

[Kevin Buettner]

On Aug 27,  4:14pm, Andrew Cagney wrote:

> This patch does two things:
> 
> - It modifies "infcall.c" so that it converts the entry-point address 
> into a code address (it might be a descriptor).
> 
> - since it's now redundant, for ppc64 GNU/Linux, it stops setting 
> entry_point_address
> 
> Kevin, ppc64 part ok?  It doesn't immediatly affect test results as this 
> GDB still requires a mod to relocate the entry point.

Sounds reasonable.  Okay.

Kevin

Re: [rfa:ppc64] Eliminate call_dummy_address

[Andrew Cagney]

> On Aug 27,  4:14pm, Andrew Cagney wrote:
> 
> 
>> This patch does two things:
>> 
>> - It modifies "infcall.c" so that it converts the entry-point address 
>> into a code address (it might be a descriptor).
>> 
>> - since it's now redundant, for ppc64 GNU/Linux, it stops setting 
>> entry_point_address
>> 
>> Kevin, ppc64 part ok?  It doesn't immediatly affect test results as this 
>> GDB still requires a mod to relocate the entry point.
> 
> 
> Sounds reasonable.  Okay.

TKS, its in.  Now for the next one.

Andrew