[Qemu-devel] arm vfp support in gdbstub

[Qemu-devel] arm vfp support in gdbstub

[Vladimir Vukicevic]

[-- Attachment #1: Type: text/plain, Size: 690 bytes --]

Howdy,

gdb doesn't seem to know what to do with vfp registers on arm, and  
qemu wasn't sending them anyway.  The attached patch and gdb tdesc  
files let you examine fp state via the debugger.  What's the right way  
to do this?  Ideally it would be selectable at runtime in some way,  
and gdb should be told what the processor desc is without having to  
manually do it.  Note that I left the FPA hole because  
org.gnu.gdb.arm.core puts cpsr at regnum 25, so there would end up  
being a hole there anyway.

To use, load the tdesc in gdb via 'set tdesc filename arm-with- 
vfp.xml', then target remote qemuhost:port as normal with a qemu built  
with the attached patch.

     - Vlad


[-- Attachment #2: qemu-gdbstubs-vfp.patch --]
[-- Type: application/octet-stream, Size: 1280 bytes --]

Index: gdbstub.c
===================================================================
--- gdbstub.c	(revision 5091)
+++ gdbstub.c	(working copy)
@@ -608,6 +608,21 @@
     /* CPSR (4 bytes).  */
     *(uint32_t *)ptr = tswapl (cpsr_read(env));
     ptr += 4;
+#if 1
+    /* 16 VFP registers (16 bytes each) */
+    /* Note that there are 32 regs available, but we only send along 16 of them -- the other 16 are optional in arm hw. */
+    for (i = 0; i < 16; i++)
+      {
+        memcpy(ptr, &env->vfp.regs[i], 8);
+        ptr += 8;
+      }
+    /* fpsid, fpscr, mvfr1, mvfr0, fpexec, fpinst, fpinst2 (4 bytes each) */
+    for (i = 0; i < 7; i++)
+      {
+        *(uint32_t *)ptr = tswapl(env->vfp.xregs[i]);
+        ptr += 4;
+      }
+#endif
 
     return ptr - mem_buf;
 }
@@ -626,7 +641,20 @@
       }
     /* Ignore FPA regs and scr.  */
     ptr += 8 * 12 + 4;
+
     cpsr_write (env, tswapl(*(uint32_t *)ptr), 0xffffffff);
+    ptr += 4;
+
+#if 1
+    /* 16 VFP registers. */
+    for (i = 0; i < 16; i++)
+      {
+        env->vfp.regs[i] = *(double *)ptr;
+        ptr += 8;
+      }
+    /* Ignore 7 VFP status registers -- XXX fixme. */
+    ptr += 7*4;
+#endif
 }
 #elif defined (TARGET_M68K)
 static int cpu_gdb_read_registers(CPUState *env, uint8_t *mem_buf)

[-- Attachment #3: arm-with-vfp.xml --]
[-- Type: text/xml, Size: 3254 bytes --]

<?xml version="1.0"?>
<!-- Copyright (C) 2007, 2008 Free Software Foundation, Inc.

     Copying and distribution of this file, with or without modification,
     are permitted in any medium without royalty provided the copyright
     notice and this notice are preserved.  -->

<!DOCTYPE target SYSTEM "gdb-target.dtd">
<target>
  <architecture>armv5te</architecture>
<!--  <xi:include href="arm-core.xml"/>-->
<feature name="org.gnu.gdb.arm.core">
  <reg name="r0" bitsize="32"/>
  <reg name="r1" bitsize="32"/>
  <reg name="r2" bitsize="32"/>
  <reg name="r3" bitsize="32"/>
  <reg name="r4" bitsize="32"/>
  <reg name="r5" bitsize="32"/>
  <reg name="r6" bitsize="32"/>
  <reg name="r7" bitsize="32"/>
  <reg name="r8" bitsize="32"/>
  <reg name="r9" bitsize="32"/>
  <reg name="r10" bitsize="32"/>
  <reg name="r11" bitsize="32"/>
  <reg name="r12" bitsize="32"/>
  <reg name="sp" bitsize="32" type="data_ptr"/>
  <reg name="lr" bitsize="32"/>
  <reg name="pc" bitsize="32" type="code_ptr"/>

  <!-- The CPSR is register 25, rather than register 16, because
       the FPA registers historically were placed between the PC
       and the CPSR in the "g" packet.  -->
  <reg name="cpsr" bitsize="32" regnum="25"/>
</feature>
<!--  <xi:include href="arm-fpa.xml"/>-->
<feature name="org.gnu.gdb.arm.fpa">
  <!-- f0's regnum is set explicitly, because the FPA registers
       historically were placed between the PC and the CPSR in the "g"
       packet - in the middle of org.gnu.gdb.arm.core.  -->
  <reg name="f0" bitsize="96" type="arm_fpa_ext" regnum="16"/>
  <reg name="f1" bitsize="96" type="arm_fpa_ext"/>
  <reg name="f2" bitsize="96" type="arm_fpa_ext"/>
  <reg name="f3" bitsize="96" type="arm_fpa_ext"/>
  <reg name="f4" bitsize="96" type="arm_fpa_ext"/>
  <reg name="f5" bitsize="96" type="arm_fpa_ext"/>
  <reg name="f6" bitsize="96" type="arm_fpa_ext"/>
  <reg name="f7" bitsize="96" type="arm_fpa_ext"/>

  <reg name="fps" bitsize="32"/>
</feature>
<!--  <xi:include href="arm-vfp.xml"/>-->
<feature name="org.gnu.gdb.arm.vfp">
  <reg name="d0" bitsize="128" type="ieee_double"/>
  <reg name="d1" bitsize="128" type="ieee_double"/>
  <reg name="d2" bitsize="128" type="ieee_double"/>
  <reg name="d3" bitsize="128" type="ieee_double"/>
  <reg name="d4" bitsize="128" type="ieee_double"/>
  <reg name="d5" bitsize="128" type="ieee_double"/>
  <reg name="d6" bitsize="128" type="ieee_double"/>
  <reg name="d7" bitsize="128" type="ieee_double"/>
  <reg name="d8" bitsize="128" type="ieee_double"/>
  <reg name="d9" bitsize="128" type="ieee_double"/>
  <reg name="d10" bitsize="128" type="ieee_double"/>
  <reg name="d11" bitsize="128" type="ieee_double"/>
  <reg name="d12" bitsize="128" type="ieee_double"/>
  <reg name="d13" bitsize="128" type="ieee_double"/>
  <reg name="d14" bitsize="128" type="ieee_double"/>
  <reg name="d15" bitsize="128" type="ieee_double"/>

  <reg name="fpsid" bitsize="32" type="uint32"/>
  <reg name="fpscr" bitsize="32" type="uint32"/>
  <reg name="mvfr1" bitsize="32" type="uint32"/>
  <reg name="mvfr0" bitsize="32" type="uint32"/>
  <reg name="fpexec" bitsize="32" type="uint32"/>
  <reg name="fpinst" bitsize="32" type="uint32"/>
  <reg name="fpinst2" bitsize="32" type="uint32"/>
</feature>
</target>

Re: [Qemu-devel] arm vfp support in gdbstub

[Paul Brook]

On Thursday 28 August 2008, Vladimir Vukicevic wrote:
> Howdy,
>
> gdb doesn't seem to know what to do with vfp registers on arm, and
> qemu wasn't sending them anyway.

I've got patches to do this properly. I'll try to get them merged soon.

Paul