[rfc] 16 bit real-mode for the i386

[rfc] 16 bit real-mode for the i386

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 574 bytes --]

Hello,

The attached modifies i386-tdep.c so that there is a ``set i386 
real-mode'' command (doco to go).

When real-mode is enabled, GDB computes the ``20 bit'' ``stop address'' 
(aka PC but not to be confused with $pc ... :-) from both the $cs and 
$pc registers.  That way, core GDB sees a cannonical PC address that 
(regardless of $cs) will match a ``20 bit'' breakpoint address.

Thoughts?

I'm desperatly trying to come up with a test-case mind.  This is a 
rewrite of an old old patch (that hacked breakpoint.c) and the original 
testcase has been lost :-(

Andrew

[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 4110 bytes --]

2002-08-29  Andrew Cagney  <cagney@redhat.com>

	* i386-tdep.h (CS_REGNUM): Define.
	* i386-tdep.c (_initialize_i386_tdep): Add `set/show i386
	real-mode' command.
	(i386_write_pc): New function.
	(i386_read_pc): New function.
	(i386_gdbarch_init): Set read_pc and write_pc.
	(set_i386_cmd): New function.
	(show_i386_cmd): New function.

Index: i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.85
diff -u -r1.85 i386-tdep.c
--- i386-tdep.c	26 Aug 2002 18:35:25 -0000	1.85
+++ i386-tdep.c	29 Aug 2002 20:22:12 -0000
@@ -1223,6 +1223,40 @@
      to the extended floating-point format used by the FPU.  */
   convert_typed_floating (from, type, to, builtin_type_i387_ext);
 }
+\f
+/* The i386 has a number of addressing modes -- in ``real mode'', a
+   text address is computed using the CS and the PC.  The below
+   performs this computation so that GDB sees a cannonical ``stop
+   address'' (PC).  The breakpoint code needs this so that it can
+   correctly match the ``stop address'' against the breakpoint
+   address..  The ia32 manual also talks about optionally truncating
+   an address to 20 bits.  Take the easy option - don't truncate.  */
+
+static int i386_real_mode_p;
+
+static CORE_ADDR
+i386_read_pc (ptid_t ptid)
+{
+  CORE_ADDR pc = read_register_pid (PC_REGNUM, ptid);
+  if (i386_real_mode_p)
+    {
+      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
+      pc = pc + (cs << 4);
+    }
+  return pc;
+}
+
+static void
+i386_write_pc (CORE_ADDR pc, ptid_t ptid)
+{
+  if (i386_real_mode_p)
+    {
+      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
+      pc = pc - (cs << 4);
+    }
+  write_register_pid (PC_REGNUM, pc, ptid);
+}
+
 \f     
 
 #ifdef STATIC_TRANSFORM_NAME
@@ -1474,6 +1508,8 @@
   set_gdbarch_pc_regnum (gdbarch, 8);
   set_gdbarch_ps_regnum (gdbarch, 9);
   set_gdbarch_fp0_regnum (gdbarch, 16);
+  set_gdbarch_read_pc (gdbarch, i386_read_pc);
+  set_gdbarch_write_pc (gdbarch, i386_write_pc);
 
   /* Use the "default" register numbering scheme for stabs and COFF.  */
   set_gdbarch_stab_reg_to_regnum (gdbarch, i386_stab_reg_to_regnum);
@@ -1596,9 +1632,37 @@
 /* Provide a prototype to silence -Wmissing-prototypes.  */
 void _initialize_i386_tdep (void);
 
+/* Dummy function.  */
+static void
+set_i386_cmd (char *args, int from_tty)
+{
+}
+
+static void
+show_i386_cmd (char *args, int from_tty)
+{
+}
+
+
 void
 _initialize_i386_tdep (void)
 {
+  static struct cmd_list_element *set_i386_cmdlist;
+  static struct cmd_list_element *show_i386_cmdlist;
+  struct cmd_list_element *tmpcmd;
+
+  /* Add an i386 set/show prefix.  */
+  add_prefix_cmd ("i386", class_maintenance, set_i386_cmd, "\
+Set i386 specific variables\n\
+Configure various i386 specific variables such as real-mode",
+		  &set_i386_cmdlist, "set i386 ",
+		  0/*allow-unknown*/, &setlist);
+  add_prefix_cmd ("i386", class_maintenance, show_i386_cmd, "\
+Show i386 specific variables\n\
+Configure various i386 specific variables such as real-mode",
+		  &show_i386_cmdlist, "show i386 ",
+		  0/*allow-unknown*/, &showlist);
+
   register_gdbarch_init (bfd_arch_i386, i386_gdbarch_init);
 
   tm_print_insn = gdb_print_insn_i386;
@@ -1643,4 +1707,11 @@
 			  i386_go32_init_abi);
   gdbarch_register_osabi (bfd_arch_i386, GDB_OSABI_NETWARE,
 			  i386_nw_init_abi);
+
+  /* Add real-mode.  */
+  add_setshow_boolean_cmd ("real-mode", no_class, &i386_real_mode_p,"\
+Set real mode (16-bit operands/addresses) operation.","\
+Show real mode (16-bit operands/addresses) operation.",
+			   NULL, NULL, &set_i386_cmdlist, &show_i386_cmdlist);
+
 }
Index: i386-tdep.h
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.h,v
retrieving revision 1.13
diff -u -r1.13 i386-tdep.h
--- i386-tdep.h	20 Aug 2002 17:59:50 -0000	1.13
+++ i386-tdep.h	29 Aug 2002 20:22:12 -0000
@@ -77,6 +77,11 @@
   int sc_sp_offset;
 };
 
+/* Code segment register.  */
+
+#define CS_REGNUM 10
+
+
 /* Floating-point registers.  */
 
 #define FPU_REG_RAW_SIZE 10

Re: [rfc] 16 bit real-mode for the i386

[Daniel Jacobowitz]

On Thu, Aug 29, 2002 at 04:30:11PM -0400, Andrew Cagney wrote:
> Hello,
> 
> The attached modifies i386-tdep.c so that there is a ``set i386 
> real-mode'' command (doco to go).
> 
> When real-mode is enabled, GDB computes the ``20 bit'' ``stop address'' 
> (aka PC but not to be confused with $pc ... :-) from both the $cs and 
> $pc registers.  That way, core GDB sees a cannonical PC address that 
> (regardless of $cs) will match a ``20 bit'' breakpoint address.
> 
> Thoughts?
> 
> I'm desperatly trying to come up with a test-case mind.  This is a 
> rewrite of an old old patch (that hacked breakpoint.c) and the original 
> testcase has been lost :-(
> 
> Andrew

Now, my i386 knowledge is a bit rusty.  But:

> +static CORE_ADDR
> +i386_read_pc (ptid_t ptid)
> +{
> +  CORE_ADDR pc = read_register_pid (PC_REGNUM, ptid);
> +  if (i386_real_mode_p)
> +    {
> +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> +      pc = pc + (cs << 4);
> +    }
> +  return pc;
> +}
> +
> +static void
> +i386_write_pc (CORE_ADDR pc, ptid_t ptid)
> +{
> +  if (i386_real_mode_p)
> +    {
> +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> +      pc = pc - (cs << 4);
> +    }
> +  write_register_pid (PC_REGNUM, pc, ptid);
> +}
> +

Left shift of _four_?  Surely the PC is more than four bits.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [rfc] 16 bit real-mode for the i386

[Kevin Buettner]

On Aug 29,  5:36pm, Daniel Jacobowitz wrote:

> Now, my i386 knowledge is a bit rusty.  But:
> 
> > +static CORE_ADDR
> > +i386_read_pc (ptid_t ptid)
> > +{
> > +  CORE_ADDR pc = read_register_pid (PC_REGNUM, ptid);
> > +  if (i386_real_mode_p)
> > +    {
> > +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> > +      pc = pc + (cs << 4);
> > +    }
> > +  return pc;
> > +}
> > +
> > +static void
> > +i386_write_pc (CORE_ADDR pc, ptid_t ptid)
> > +{
> > +  if (i386_real_mode_p)
> > +    {
> > +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> > +      pc = pc - (cs << 4);
> > +    }
> > +  write_register_pid (PC_REGNUM, pc, ptid);
> > +}
> > +
> 
> Left shift of _four_?  Surely the PC is more than four bits.

I think that's right.  My (hazy) recollection from my days of programming
x86 before there was an 80386 is that an address was formed by shifting
the segment registers left by four bits and adding some other sixteen bit
register to it.

Kevin

Re: [rfc] 16 bit real-mode for the i386

[Daniel Jacobowitz]

On Thu, Aug 29, 2002 at 02:43:04PM -0700, Kevin Buettner wrote:
> On Aug 29,  5:36pm, Daniel Jacobowitz wrote:
> 
> > Now, my i386 knowledge is a bit rusty.  But:
> > 
> > > +static CORE_ADDR
> > > +i386_read_pc (ptid_t ptid)
> > > +{
> > > +  CORE_ADDR pc = read_register_pid (PC_REGNUM, ptid);
> > > +  if (i386_real_mode_p)
> > > +    {
> > > +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> > > +      pc = pc + (cs << 4);
> > > +    }
> > > +  return pc;
> > > +}
> > > +
> > > +static void
> > > +i386_write_pc (CORE_ADDR pc, ptid_t ptid)
> > > +{
> > > +  if (i386_real_mode_p)
> > > +    {
> > > +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> > > +      pc = pc - (cs << 4);
> > > +    }
> > > +  write_register_pid (PC_REGNUM, pc, ptid);
> > > +}
> > > +
> > 
> > Left shift of _four_?  Surely the PC is more than four bits.
> 
> I think that's right.  My (hazy) recollection from my days of programming
> x86 before there was an 80386 is that an address was formed by shifting
> the segment registers left by four bits and adding some other sixteen bit
> register to it.

That's positively perverse.  Thanks.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [rfc] 16 bit real-mode for the i386

[Andrew Cagney]

> Now, my i386 knowledge is a bit rusty.  But:
> 
> 
>> +static CORE_ADDR
>> +i386_read_pc (ptid_t ptid)
>> +{
>> +  CORE_ADDR pc = read_register_pid (PC_REGNUM, ptid);
>> +  if (i386_real_mode_p)
>> +    {
>> +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
>> +      pc = pc + (cs << 4);
>> +    }
>> +  return pc;
>> +}
>> +
>> +static void
>> +i386_write_pc (CORE_ADDR pc, ptid_t ptid)
>> +{
>> +  if (i386_real_mode_p)
>> +    {
>> +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
>> +      pc = pc - (cs << 4);
>> +    }
>> +  write_register_pid (PC_REGNUM, pc, ptid);
>> +}
>> +
> 
> 
> Left shift of _four_?  Surely the PC is more than four bits.

It is.  From an ia32 manual (www.intel.com):

16.1.1.    Address Translation in Real-Address Mode
In real-address mode, the processor does not interpret segment selectors 
as indexes into a descriptor table; instead, it uses them directly to 
form linear addresses as the 8086 processor does. It shifts the segment 
selector left by 4 bits to form a 20-bit base address (see Figure 16-1).
The offset into a segment is added to the base address to create a 
linear address that maps directly
to the physical address space.

Andrew

Re: [rfc] 16 bit real-mode for the i386

[Eli Zaretskii]

> Date: Thu, 29 Aug 2002 17:36:53 -0400
> From: Daniel Jacobowitz <drow@mvista.com>
> 
> > +static void
> > +i386_write_pc (CORE_ADDR pc, ptid_t ptid)
> > +{
> > +  if (i386_real_mode_p)
> > +    {
> > +      CORE_ADDR cs = read_register_pid (CS_REGNUM, ptid);
> > +      pc = pc - (cs << 4);
> > +    }
> > +  write_register_pid (PC_REGNUM, pc, ptid);
> > +}
> > +
> 
> Left shift of _four_?  Surely the PC is more than four bits.

It's true that PC can be more than 4 bits, but the code above is still
correct.  That's why real mode can only access 20 bit addresses:
16+4=20.

You can look at this issue this way: for each linear 2-bit address,
there are many CS:PC combination that all point to that address.

Re: [rfc] 16 bit real-mode for the i386

[Mark Kettenis]

Andrew Cagney <ac131313@ges.redhat.com> writes:

> Hello,
> 
> The attached modifies i386-tdep.c so that there is a ``set i386 
> real-mode'' command (doco to go).
> 
> When real-mode is enabled, GDB computes the ``20 bit'' ``stop address'' 
> (aka PC but not to be confused with $pc ... :-) from both the $cs and 
> $pc registers.  That way, core GDB sees a cannonical PC address that 
> (regardless of $cs) will match a ``20 bit'' breakpoint address.
> 
> Thoughts?

Looks OK to me.  Just two nits: could you consider
s/CS_REGNUM/I386_CS_REGNUM/?  And I believe the GNU coding standards
say that one shouldn't introduce unecessary whitespace.  Therefore I
have some problems with the extra newlines you're introducing.

Are you planning to add this before we cut the branch?  I'm a little
uncomfortable with adding any new features just before doing so.

Mark

Re: [rfc] 16 bit real-mode for the i386

[Andrew Cagney]

> Andrew Cagney <ac131313@ges.redhat.com> writes:
> 
> 
>> Hello,
>> 
>> The attached modifies i386-tdep.c so that there is a ``set i386 
>> real-mode'' command (doco to go).
>> 
>> When real-mode is enabled, GDB computes the ``20 bit'' ``stop address'' 
>> (aka PC but not to be confused with $pc ... :-) from both the $cs and 
>> $pc registers.  That way, core GDB sees a cannonical PC address that 
>> (regardless of $cs) will match a ``20 bit'' breakpoint address.
>> 
>> Thoughts?
> 
> 
> Looks OK to me.  Just two nits: could you consider
> s/CS_REGNUM/I386_CS_REGNUM/?  And I believe the GNU coding standards
> say that one shouldn't introduce unecessary whitespace.  Therefore I
> have some problems with the extra newlines you're introducing.

I'll change the macro^D^Denum. Many of the new lines are consistent with 
whats around (but I'll prune those that look excessive little).

> Are you planning to add this before we cut the branch?  I'm a little
> uncomfortable with adding any new features just before doing so.

I'm not expecting to really follow this up until after the branch is 
cut.  (And I still need to think up a testcase :-)

Andrew