E500 APUinfo fix

E500 APUinfo fix

[Mark Mitchell]

Section 3.6 of the E500 ABI specifies the contents of the APU
Information Section, and requires that offset 4 contains the number of
bytes of APU information.  However, GAS and binutils incorrectly use
the number of APU information slots -- rather than bytes.

Fixed with this patch.

OK to apply?

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com

2003-06-23  Mark Mitchell  <mark@codesourcery.com>

	* elf32-ppc.c (ppc_elf_begin_write_processing): Use bytes to count
	APUinfo slots.
	(ppc_elf_final_write_processing): Likewise.

2003-06-23  Mark Mitchell  <mark@codesourcery.com>

	* config/tc-ppc.c (ppc_cleanup): Use bytes to count APUinfo slots.

Index: bfd/elf32-ppc.c
===================================================================
RCS file: /cvs/src/src/bfd/elf32-ppc.c,v
retrieving revision 1.83
diff -c -5 -p -r1.83 elf32-ppc.c
*** bfd/elf32-ppc.c	31 May 2003 07:55:10 -0000	1.83
--- bfd/elf32-ppc.c	24 Jun 2003 03:30:02 -0000
*************** ppc_elf_begin_write_processing (abfd, li
*** 5719,5740 ****
  	goto fail;
  
        if (strcmp (ptr + 12, APUINFO_LABEL) != 0)
  	goto fail;
  
!       /* Get the number of apuinfo entries.  */
        datum = bfd_get_32 (ibfd, ptr + 4);
!       if ((datum * 4 + 20) != length)
  	goto fail;
  
        /* Make sure that we do not run off the end of the section.  */
        if (offset + length > output_section_size)
  	goto fail;
  
        /* Scan the apuinfo section, building a list of apuinfo numbers.  */
!       for (i = 0; i < datum; i++)
! 	apuinfo_list_add (bfd_get_32 (ibfd, ptr + 20 + (i * 4)));
  
        /* Update the offset.  */
        offset += length;
      }
  
--- 5719,5740 ----
  	goto fail;
  
        if (strcmp (ptr + 12, APUINFO_LABEL) != 0)
  	goto fail;
  
!       /* Get the number of bytes used for apuinfo entries.  */
        datum = bfd_get_32 (ibfd, ptr + 4);
!       if (datum + 20 != length)
  	goto fail;
  
        /* Make sure that we do not run off the end of the section.  */
        if (offset + length > output_section_size)
  	goto fail;
  
        /* Scan the apuinfo section, building a list of apuinfo numbers.  */
!       for (i = 0; i < datum; i += 4)
! 	apuinfo_list_add (bfd_get_32 (ibfd, ptr + 20 + i));
  
        /* Update the offset.  */
        offset += length;
      }
  
*************** ppc_elf_final_write_processing (abfd, li
*** 5806,5816 ****
      }
  
    /* Create the apuinfo header.  */
    num_entries = apuinfo_list_length ();
    bfd_put_32 (abfd, sizeof APUINFO_LABEL, buffer);
!   bfd_put_32 (abfd, num_entries, buffer + 4);
    bfd_put_32 (abfd, 0x2, buffer + 8);
    strcpy (buffer + 12, APUINFO_LABEL);
  
    length = 20;
    for (i = 0; i < num_entries; i++)
--- 5806,5816 ----
      }
  
    /* Create the apuinfo header.  */
    num_entries = apuinfo_list_length ();
    bfd_put_32 (abfd, sizeof APUINFO_LABEL, buffer);
!   bfd_put_32 (abfd, num_entries * 4, buffer + 4);
    bfd_put_32 (abfd, 0x2, buffer + 8);
    strcpy (buffer + 12, APUINFO_LABEL);
  
    length = 20;
    for (i = 0; i < num_entries; i++)
Index: gas/config/tc-ppc.c
===================================================================
RCS file: /cvs/src/src/gas/config/tc-ppc.c,v
retrieving revision 1.75
diff -c -5 -p -r1.75 tc-ppc.c
*** gas/config/tc-ppc.c	18 Jun 2003 23:05:44 -0000	1.75
--- gas/config/tc-ppc.c	24 Jun 2003 03:30:03 -0000
*************** ppc_cleanup ()
*** 1340,1350 ****
  
      p = frag_more (4);
      md_number_to_chars (p, (valueT) 8, 4);
  
      p = frag_more (4);
!     md_number_to_chars (p, (valueT) ppc_apuinfo_num, 4);
  
      p = frag_more (4);
      md_number_to_chars (p, (valueT) 2, 4);
  
      p = frag_more (8);
--- 1340,1350 ----
  
      p = frag_more (4);
      md_number_to_chars (p, (valueT) 8, 4);
  
      p = frag_more (4);
!     md_number_to_chars (p, (valueT) ppc_apuinfo_num * 4, 4);
  
      p = frag_more (4);
      md_number_to_chars (p, (valueT) 2, 4);
  
      p = frag_more (8);

Re: E500 APUinfo fix

[Alan Modra]

On Mon, Jun 23, 2003 at 08:36:51PM -0700, Mark Mitchell wrote:
> 2003-06-23  Mark Mitchell  <mark@codesourcery.com>
> 
> 	* elf32-ppc.c (ppc_elf_begin_write_processing): Use bytes to count
> 	APUinfo slots.
> 	(ppc_elf_final_write_processing): Likewise.
> 
> 2003-06-23  Mark Mitchell  <mark@codesourcery.com>
> 
> 	* config/tc-ppc.c (ppc_cleanup): Use bytes to count APUinfo slots.

OK.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: E500 APUinfo fix

[Alan Modra]

On Mon, Jun 23, 2003 at 08:36:51PM -0700, Mark Mitchell wrote:
> Use bytes to count APUinfo slots.

Tch.

	* ld-powerpc/apuinfo.rd: Update.

Index: ld/testsuite/ld-powerpc/apuinfo.rd
===================================================================
RCS file: /cvs/src/src/ld/testsuite/ld-powerpc/apuinfo.rd,v
retrieving revision 1.1
diff -u -p -r1.1 apuinfo.rd
--- ld/testsuite/ld-powerpc/apuinfo.rd	3 Dec 2002 18:24:33 -0000	1.1
+++ ld/testsuite/ld-powerpc/apuinfo.rd	25 Jun 2003 08:46:22 -0000
@@ -5,6 +5,6 @@
 #target: powerpc-eabi*
 
 Hex dump of section '.PPC.EMB.apuinfo':
-  0x00000000 00000008 00000007 00000002 41505569 ............APUi
+  0x00000000 00000008 0000001c 00000002 41505569 ............APUi
   0x00000010 6e666f00 00420001 00430001 00410001 nfo..B...C...A..
   0x00000020 01020001 01010001 00400001 01000001 .........@......


-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

ppc-opc.c fix for disasm

[Kumar Gala]

I noticed a bug in disasm of dcbt, dcbtst today.  I'm wondering what 
changes I need to make to binutils beyond fixing the opcode/ppc-opc.c 
file to change the mask from XRT_MASK -> X_MASK for the two 
instructions.

Also is there anywhere that talks about how to properly create patches 
for submission?

thanks

- kumar

Re: ppc-opc.c fix for disasm

[Nick Clifton]

Hi Kumar,

> I noticed a bug in disasm of dcbt, dcbtst today.  I'm wondering what
> changes I need to make to binutils beyond fixing the
> opcode/ppc-opc.c file to change the mask from XRT_MASK -> X_MASK for
> the two instructions.

A test case that demonstrates the problem and which can be added to
the ppc specific part of the GAS testsuite would also be very helpful.

Other than that though, there are no other files that you should need
to change.


> Also is there anywhere that talks about how to properly create patches
> for submission?

Binutils generally follows the same rules as gcc, so the documentation
on their web site applies here:

  http://www.gnu.org/software/gcc/contribute.html

Cheers
        Nick