[COMMIT] Add 64-bit support to hppa-hpux-nat.c

[COMMIT] Add 64-bit support to hppa-hpux-nat.c

[Mark Kettenis]

This adds 64-bit support to hppa-hpux-nat.c.  This makes it possible
to use the new inf-ttrace.c for 64-bit HP-UX.  It seems to work rather
well for me on hppa64-hp-hpux11.11.  Of course there are loads of
failures because inferior function calls don't work..  There seems to
be an issue with "hardware" watchpoints however.  I'll investigate
that for a bit, and do the final conversion at the end of the coming
week.

Meanwhile, if you want to check this out, simply remove the
hppa*64*-*-hpux11* line from your configure.host file.  As you can,
this means that there's only a single HP-UX native configuration left!

It also means that the days of hppah-nat.c and infttrace.c are really
counted now.  Don't waste any time on those!  There'll be some fallout
in core GDB too; many of the HP-UX specific hacks can just go now.

The target side still needs some work though.  Currently we can't
build a GDB that supports both 32-bit and 64-bit HP-UX targets, but
the native code should support that just fine.

Mark


Index: ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>
 
	* hppa-hpux-nat.c (ss_mpsfu_high): Remove define.
	(hppa_hpux_save_state_offset): New function, replacing variable
	with the same name.
	(hppa_cannot_fetch_register, hppa_cannot_store_register): Remove
	functions.
	(hppa_hpux_fetch_register): Don't call hppa_cannot_fetch_register.
	Use new hppa_hpux_save_state_offset function.  Add special
	handling for handling 64-bit "flags" register.
	(hppa_hpux_store_register): Likewise.
	(hppa_hpux_child_can_run): Tweak comment.

Index: hppa-hpux-nat.c
===================================================================
RCS file: /cvs/src/src/gdb/hppa-hpux-nat.c,v
retrieving revision 1.3
diff -u -p -r1.3 hppa-hpux-nat.c
--- hppa-hpux-nat.c 23 Nov 2004 21:14:32 -0000 1.3
+++ hppa-hpux-nat.c 11 Dec 2004 22:47:45 -0000
@@ -36,159 +36,48 @@
 #include "inf-ptrace.h"
 #include "inf-ttrace.h"
 
-/* HP-UX 10.20 has a different name than HP-UX 11.00 and later.
-   Apparently, the intended usage changed.  Unfortunately HP didn't
-   care about backwards compatibility.  */
-#ifdef ss_tlsp
-#define ss_mpsfu_high ss_tlsp
-#endif
+/* Non-zero if we should pretend not to be a runnable target.  */
+int child_suppress_run = 0;
 
-int child_suppress_run = 0;     /* Non-zero if we should pretend not to be
-				   a runnable target.  */
+/* Return the offset of register REGNUM within `struct save_state'.
+   The offset returns depends on the flags in the "flags" register and
+   the register size (32-bit or 64-bit).  These are taken from
+   REGCACHE.  */
 
-static int hppa_hpux_save_state_offset[] =
+LONGEST
+hppa_hpux_save_state_offset (struct regcache *regcache, int regnum)
 {
-  ssoff(ss_flags),
-  ssoff(ss_narrow.ss_gr1),
-  ssoff(ss_narrow.ss_rp),
-  ssoff(ss_narrow.ss_gr3),
-  ssoff(ss_narrow.ss_gr4),
-  ssoff(ss_narrow.ss_gr5),
-  ssoff(ss_narrow.ss_gr6),
-  ssoff(ss_narrow.ss_gr7),
-  ssoff(ss_narrow.ss_gr8),
-  ssoff(ss_narrow.ss_gr9),
-  ssoff(ss_narrow.ss_gr10),
-  ssoff(ss_narrow.ss_gr11),
-  ssoff(ss_narrow.ss_gr12),
-  ssoff(ss_narrow.ss_gr13),
-  ssoff(ss_narrow.ss_gr14),
-  ssoff(ss_narrow.ss_gr15),
-  ssoff(ss_narrow.ss_gr16),
-  ssoff(ss_narrow.ss_gr17),
-  ssoff(ss_narrow.ss_gr18),
-  ssoff(ss_narrow.ss_gr19),
-  ssoff(ss_narrow.ss_gr20),
-  ssoff(ss_narrow.ss_gr21),
-  ssoff(ss_narrow.ss_gr22),
-  ssoff(ss_narrow.ss_arg3),
-  ssoff(ss_narrow.ss_arg2),
-  ssoff(ss_narrow.ss_arg1),
-  ssoff(ss_narrow.ss_arg0),
-  ssoff(ss_narrow.ss_dp),
-  ssoff(ss_narrow.ss_ret0),
-  ssoff(ss_narrow.ss_ret1),
-  ssoff(ss_narrow.ss_sp),
-  ssoff(ss_narrow.ss_gr31),
-  ssoff(ss_narrow.ss_cr11),
-  ssoff(ss_narrow.ss_pcoq_head),
-  ssoff(ss_narrow.ss_pcsq_head),
-  ssoff(ss_narrow.ss_pcoq_tail),
-  ssoff(ss_narrow.ss_pcsq_tail),
-  ssoff(ss_narrow.ss_cr15),
-  ssoff(ss_narrow.ss_cr19),
-  ssoff(ss_narrow.ss_cr20),
-  ssoff(ss_narrow.ss_cr21),
-  ssoff(ss_narrow.ss_cr22),
-  ssoff(ss_narrow.ss_cpustate),
-  ssoff(ss_narrow.ss_sr4),
-  ssoff(ss_narrow.ss_sr0),
-  ssoff(ss_narrow.ss_sr1),
-  ssoff(ss_narrow.ss_sr2),
-  ssoff(ss_narrow.ss_sr3),
-  ssoff(ss_narrow.ss_sr5),
-  ssoff(ss_narrow.ss_sr6),
-  ssoff(ss_narrow.ss_sr7),
-  ssoff(ss_narrow.ss_cr0),
-  ssoff(ss_narrow.ss_cr8),
-  ssoff(ss_narrow.ss_cr9),
-  ssoff(ss_narrow.ss_cr10),
-  ssoff(ss_narrow.ss_cr12),
-  ssoff(ss_narrow.ss_cr13),
-  ssoff(ss_narrow.ss_cr24),
-  ssoff(ss_narrow.ss_cr25),
-  ssoff(ss_narrow.ss_cr26),
-  ssoff(ss_narrow.ss_mpsfu_high),
-  ssoff(ss_narrow.ss_mpsfu_low),
-  ssoff(ss_narrow.ss_mpsfu_ovflo),
-  ssoff(ss_pad),
-  ssoff(ss_frstat),
-  ssoff(ss_frexcp1),
-  ssoff(ss_frexcp2),
-  ssoff(ss_frexcp3),
-  ssoff(ss_frexcp4),
-  ssoff(ss_frexcp5),
-  ssoff(ss_frexcp6),
-  ssoff(ss_frexcp7),
-  ssoff(ss_fr4_hi),
-  ssoff(ss_fr4_lo),
-  ssoff(ss_fr5_hi),
-  ssoff(ss_fr5_lo),
-  ssoff(ss_fr6_hi),
-  ssoff(ss_fr6_lo),
-  ssoff(ss_fr7_hi),
-  ssoff(ss_fr7_lo),
-  ssoff(ss_fr8_hi),
-  ssoff(ss_fr8_lo),
-  ssoff(ss_fr9_hi),
-  ssoff(ss_fr9_lo),
-  ssoff(ss_fr10_hi),
-  ssoff(ss_fr10_lo),
-  ssoff(ss_fr11_hi),
-  ssoff(ss_fr11_lo),
-  ssoff(ss_fr12_hi),
-  ssoff(ss_fr12_lo),
-  ssoff(ss_fr13_hi),
-  ssoff(ss_fr13_lo),
-  ssoff(ss_fr14_hi),
-  ssoff(ss_fr14_lo),
-  ssoff(ss_fr15_hi),
-  ssoff(ss_fr15_lo),
-  ssoff(ss_fr16_hi),
-  ssoff(ss_fr16_lo),
-  ssoff(ss_fr17_hi),
-  ssoff(ss_fr17_lo),
-  ssoff(ss_fr18_hi),
-  ssoff(ss_fr18_lo),
-  ssoff(ss_fr19_hi),
-  ssoff(ss_fr19_lo),
-  ssoff(ss_fr20_hi),
-  ssoff(ss_fr20_lo),
-  ssoff(ss_fr21_hi),
-  ssoff(ss_fr21_lo),
-  ssoff(ss_fr22_hi),
-  ssoff(ss_fr22_lo),
-  ssoff(ss_fr23_hi),
-  ssoff(ss_fr23_lo),
-  ssoff(ss_fr24_hi),
-  ssoff(ss_fr24_lo),
-  ssoff(ss_fr25_hi),
-  ssoff(ss_fr25_lo),
-  ssoff(ss_fr26_hi),
-  ssoff(ss_fr26_lo),
-  ssoff(ss_fr27_hi),
-  ssoff(ss_fr27_lo),
-  ssoff(ss_fr28_hi),
-  ssoff(ss_fr28_lo),
-  ssoff(ss_fr29_hi),
-  ssoff(ss_fr29_lo),
-  ssoff(ss_fr30_hi),
-  ssoff(ss_fr30_lo),
-  ssoff(ss_fr31_hi),
-  ssoff(ss_fr31_lo)
-};
+  LONGEST offset;
 
-static int
-hppa_hpux_cannot_fetch_register (int regnum)
-{
-  gdb_assert (regnum >= 0 && regnum < NUM_REGS);
-  return (regnum >= ARRAY_SIZE(hppa_hpux_save_state_offset));
-}
+  if (regnum == HPPA_FLAGS_REGNUM)
+    return ssoff (ss_flags);
 
-static int
-hppa_hpux_cannot_store_register (int regnum)
-{
-  return hppa_hpux_cannot_fetch_register (regnum);
+  if (HPPA_R0_REGNUM < regnum && regnum < HPPA_FP0_REGNUM)
+    {
+      struct gdbarch *arch = get_regcache_arch (regcache);
+      size_t size = register_size (arch, HPPA_R1_REGNUM);
+      ULONGEST flags;
+
+      gdb_assert (size == 4 || size == 8);
+
+      regcache_cooked_read_unsigned (regcache, HPPA_FLAGS_REGNUM, &flags);
+      if (flags & SS_WIDEREGS)
+	offset = ssoff (ss_wide) + (8 - size) + (regnum - HPPA_R0_REGNUM) * 8;
+      else
+	offset = ssoff (ss_narrow) + (regnum - HPPA_R1_REGNUM) * 4;
+    }
+  else
+    {
+      struct gdbarch *arch = get_regcache_arch (regcache);
+      size_t size = register_size (arch, HPPA_FP0_REGNUM);
+
+      gdb_assert (size == 4 || size == 8);
+      gdb_assert (regnum >= HPPA_FP0_REGNUM);
+      offset = ssoff(ss_fpblock) + (regnum - HPPA_FP0_REGNUM) * size;
+    }
+
+  gdb_assert (offset < sizeof (save_state_t));
+  return offset;
 }
 
 /* Just in case a future version of PA-RISC HP-UX won't have ptrace(2)
@@ -206,16 +95,10 @@ hppa_hpux_fetch_register (int regnum)
   pid_t pid;
   int i;
 
-  if (hppa_hpux_cannot_fetch_register (regnum))
-    {
-      regcache_raw_supply (current_regcache, regnum, NULL);
-      return;
-    }
-
   pid = ptid_get_pid (inferior_ptid);
 
-  /* This isn't really an address.  But ptrace thinks of it as one.  */
-  addr = hppa_hpux_save_state_offset[regnum];
+  /* This isn't really an address, but ptrace thinks of it as one.  */
+  addr = hppa_hpux_save_state_offset(current_regcache, regnum);
   size = register_size (current_gdbarch, regnum);
 
   gdb_assert (size == 4 || size == 8);
@@ -247,6 +130,14 @@ hppa_hpux_fetch_register (int regnum)
   }
 #endif
 
+  /* Take care with the "flags" register.  It's stored as an `int' in
+     `struct save_state', even for 64-bit code.  */
+  if (regnum == HPPA_FLAGS_REGNUM && size == 8)
+    {
+      ULONGEST flags = extract_unsigned_integer (buf, 4);
+      store_unsigned_integer (buf, 8, flags);
+    }
+
   regcache_raw_supply (current_regcache, regnum, buf);
 }
 
@@ -270,13 +161,10 @@ hppa_hpux_store_register (int regnum)
   PTRACE_TYPE_RET *buf;
   pid_t pid;
 
-  if (hppa_hpux_cannot_store_register (regnum))
-    return;
-
   pid = ptid_get_pid (inferior_ptid);
 
-  /* This isn't really an address.  But ptrace thinks of it as one.  */
-  addr = hppa_hpux_save_state_offset[regnum];
+  /* This isn't really an address, but ptrace thinks of it as one.  */
+  addr = hppa_hpux_save_state_offset(current_regcache, regnum);
   size = register_size (current_gdbarch, regnum);
 
   gdb_assert (size == 4 || size == 8);
@@ -284,6 +172,15 @@ hppa_hpux_store_register (int regnum)
 
   regcache_raw_collect (current_regcache, regnum, buf);
 
+  /* Take care with the "flags" register.  It's stored as an `int' in
+     `struct save_state', even for 64-bit code.  */
+  if (regnum == HPPA_FLAGS_REGNUM && size == 8)
+    {
+      ULONGEST flags = extract_unsigned_integer (buf, 8);
+      store_unsigned_integer (buf, 4, flags);
+      size = 4;
+    }
+
 #ifdef HAVE_TTRACE
   {
     lwpid_t lwp = ptid_get_lwp (inferior_ptid);
@@ -327,10 +224,9 @@ hppa_hpux_store_inferior_registers (int 
 static int
 hppa_hpux_child_can_run (void)
 {
-  /* This variable is controlled by modules that layer their own process
-     structure atop that provided here.  hpux-thread.c does this because
-     of the HP-UX user-mode level thread model.  */
-
+  /* This variable is controlled by modules that layer their own
+     process structure atop that provided here.  hpux-thread.c does
+     this because of the HP-UX user-mode level thread model.  */
   return !child_suppress_run;
 }
 \f

Re: [COMMIT] Add 64-bit support to hppa-hpux-nat.c

[Randolph Chung]

> This adds 64-bit support to hppa-hpux-nat.c.  This makes it possible
> to use the new inf-ttrace.c for 64-bit HP-UX.  It seems to work rather
> well for me on hppa64-hp-hpux11.11.  Of course there are loads of
> failures because inferior function calls don't work..  There seems to
> be an issue with "hardware" watchpoints however.  I'll investigate
> that for a bit, and do the final conversion at the end of the coming
> week.

thanks. I'm working on the inferior function calls problem (posted some
RFCs this week). Hope to get that working next week.

> Meanwhile, if you want to check this out, simply remove the
> hppa*64*-*-hpux11* line from your configure.host file.  As you can,
> this means that there's only a single HP-UX native configuration left!

yay!!

> The target side still needs some work though.  Currently we can't
> build a GDB that supports both 32-bit and 64-bit HP-UX targets, but
> the native code should support that just fine.

yeah, i've been working on cleaning that up. once the new solib stuff go
in it should be better, but it still doesn't work completely yet.

i'll pull in your changes and try it out; if things are working in
general, is it ok if i enable the new solib support? it's not very
extensively tested but the basics seem to be ok.

i have a bunch of hppa-hpux-tdep.c changes queued up in my tree. i need
to sort those out and submit them. the last time i run the testsuite (on
weds) there were about 580 FAILs (vs 620 posted by Joel in October)

randolph
-- 
Randolph Chung
Debian GNU/Linux Developer, hppa/ia64 ports
http://www.tausq.org/

Re: [COMMIT] Add 64-bit support to hppa-hpux-nat.c

[Daniel Jacobowitz]

On Sun, Dec 12, 2004 at 12:05:26AM +0100, Mark Kettenis wrote:
> This adds 64-bit support to hppa-hpux-nat.c.  This makes it possible
> to use the new inf-ttrace.c for 64-bit HP-UX.  It seems to work rather
> well for me on hppa64-hp-hpux11.11.  Of course there are loads of
> failures because inferior function calls don't work..  There seems to
> be an issue with "hardware" watchpoints however.  I'll investigate
> that for a bit, and do the final conversion at the end of the coming
> week.

Mark, I just want to thank you for all the work you've put into
updating this port.  It's really amazing.

Now that that's said, I don't suppose you want to slaughter the SOM
reader next? :-)

-- 
Daniel Jacobowitz