Generate minidumps for 64-bit ARM apps on iOS.

src/google_breakpad/common/minidump_cpu_arm64.h

-/* Copyright (c) 2009, Google Inc.
+/* Copyright 2013, Google Inc.
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are
  * met:
  *
  *     * Redistributions of source code must retain the above copyright
  * notice, this list of conditions and the following disclaimer.
  *     * Redistributions in binary form must reproduce the above
  * copyright notice, this list of conditions and the following disclaimer
@@ skipping 37 matching lines @@
  * zero-length arrays are forbidden by standard C and C++.  In these cases,
  * *_minsize constants are provided to be used in place of sizeof.  For a
  * cleaner interface to these sizes when using C++, see minidump_size.h.
  *
  * These structures are also sufficient to populate minidump files.
  *
  * Because precise data type sizes are crucial for this implementation to
  * function properly and portably, a set of primitive types with known sizes
  * are used as the basis of each structure defined by this file.
  *
- * Author: Julian Seward
+ * Author: Colin Blundell
  */
 
 /*
- * ARM support
+ * ARM64 support
  */
 
-#ifndef GOOGLE_BREAKPAD_COMMON_MINIDUMP_CPU_ARM_H__
-#define GOOGLE_BREAKPAD_COMMON_MINIDUMP_CPU_ARM_H__
+#ifndef GOOGLE_BREAKPAD_COMMON_MINIDUMP_CPU_ARM64_H__
+#define GOOGLE_BREAKPAD_COMMON_MINIDUMP_CPU_ARM64_H__
 
-#define MD_FLOATINGSAVEAREA_ARM_FPR_COUNT 32
-#define MD_FLOATINGSAVEAREA_ARM_FPEXTRA_COUNT 8
+#define MD_FLOATINGSAVEAREA_ARM64_FPR_COUNT 32
+#define MD_FLOATINGSAVEAREA_ARM64_FPEXTRA_COUNT 8

[qsr, 2013/10/28 10:25:38]

Did you check those values?

[blundell, 2013/11/07 16:41:26]

Done.

 
 /*
  * Note that these structures *do not* map directly to the CONTEXT
  * structure defined in WinNT.h in the Windows Mobile SDK. That structure
  * does not accomodate VFPv3, and I'm unsure if it was ever used in the
  * wild anyway, as Windows CE only seems to produce "cedumps" which
  * are not exactly minidumps.
  */
+// TODO(blundell)_: Where can I find the right definition for these?
 typedef struct {
   uint64_t      fpscr;      /* FPU status register */
 
   /* 32 64-bit floating point registers, d0 .. d31. */
-  uint64_t      regs[MD_FLOATINGSAVEAREA_ARM_FPR_COUNT];
+  uint64_t      regs[MD_FLOATINGSAVEAREA_ARM64_FPR_COUNT];
 
   /* Miscellaneous control words */
-  uint32_t     extra[MD_FLOATINGSAVEAREA_ARM_FPEXTRA_COUNT];
-} MDFloatingSaveAreaARM;
+  uint32_t     extra[MD_FLOATINGSAVEAREA_ARM64_FPEXTRA_COUNT];
+} MDFloatingSaveAreaARM64;
 
-#define MD_CONTEXT_ARM_GPR_COUNT 16
+#define MD_CONTEXT_ARM64_GPR_COUNT 33
 
 typedef struct {
   /* The next field determines the layout of the structure, and which parts
    * of it are populated
    */
-  uint32_t      context_flags;
+  uint64_t      context_flags;
 
-  /* 16 32-bit integer registers, r0 .. r15
+  /* 33 64-bit integer registers, x0 .. x31 + the PC
    * Note the following fixed uses:
-   *   r13 is the stack pointer
-   *   r14 is the link register
-   *   r15 is the program counter
+   *   x29 is the frame pointer
+   *   x30 is the link register
+   *   x31 is the stack pointer
+   *   The PC is effectively x32.
    */
-  uint32_t     iregs[MD_CONTEXT_ARM_GPR_COUNT];
+  uint64_t     iregs[MD_CONTEXT_ARM64_GPR_COUNT];
 
   /* CPSR (flags, basically): 32 bits:
         bit 31 - N (negative)
         bit 30 - Z (zero)
         bit 29 - C (carry)
         bit 28 - V (overflow)
         bit 27 - Q (saturation flag, sticky)
      All other fields -- ignore */
   uint32_t    cpsr;
 
-  /* The next field is included with MD_CONTEXT_ARM_FLOATING_POINT */
-  MDFloatingSaveAreaARM float_save;
+  /* The next field is included with MD_CONTEXT64_ARM_FLOATING_POINT */
+  MDFloatingSaveAreaARM64 float_save;
 
-} MDRawContextARM;
+} MDRawContextARM64;
 
 /* Indices into iregs for registers with a dedicated or conventional
  * purpose.
  */
-enum MDARMRegisterNumbers {
-  MD_CONTEXT_ARM_REG_IOS_FP = 7,
-  MD_CONTEXT_ARM_REG_FP     = 11,
-  MD_CONTEXT_ARM_REG_SP     = 13,
-  MD_CONTEXT_ARM_REG_LR     = 14,
-  MD_CONTEXT_ARM_REG_PC     = 15
+enum MDARM64RegisterNumbers {
+  MD_CONTEXT_ARM64_REG_FP     = 29,
+  MD_CONTEXT_ARM64_REG_LR     = 30,
+  MD_CONTEXT_ARM64_REG_SP     = 31,
+  MD_CONTEXT_ARM64_REG_PC     = 32,
 };
 
-/* For (MDRawContextARM).context_flags.  These values indicate the type of
+/* For (MDRawContextARM64).context_flags.  These values indicate the type of
  * context stored in the structure. */
-/* CONTEXT_ARM from the Windows CE 5.0 SDK. This value isn't correct
+/* CONTEXT_ARM64 from the Windows CE 5.0 SDK. This value isn't correct
  * because this bit can be used for flags. Presumably this value was
  * never actually used in minidumps, but only in "CEDumps" which
  * are a whole parallel minidump file format for Windows CE.
  * Therefore, Breakpad defines its own value for ARM CPUs.

[qsr, 2013/10/28 10:25:38]

The comment is completely wrong. Windows CE never handled 64 bits CPU.

[blundell, 2013/11/07 16:41:26]

Done.

  */
-#define MD_CONTEXT_ARM_OLD               0x00000040
+// TODO(blundell): What should these values be?
+#define MD_CONTEXT_ARM64_OLD               0x00000040

[blundell, 2013/10/25 16:07:02]

I don't know what to use for this value or the value of MD_CONTEXT_ARM64 below.

[qsr, 2013/10/28 10:25:38]

 You either need to find documentation, or you need to examine the real context
at a point where you know what it should contain. You then look at its type, and
find out what the bits mean.

[blundell, 2013/11/07 16:41:26]

I believe that that MD_CONTEXT_ARM64 is Breakpad-defined. Updated to reflect
that.

On 2013/10/28 10:25:38, qsr wrote:

 /* This value was chosen to avoid likely conflicts with MD_CONTEXT_*
  * for other CPUs. */
-#define MD_CONTEXT_ARM                   0x40000000
-#define MD_CONTEXT_ARM_INTEGER           (MD_CONTEXT_ARM | 0x00000002)
-#define MD_CONTEXT_ARM_FLOATING_POINT    (MD_CONTEXT_ARM | 0x00000004)
+#define MD_CONTEXT_ARM64                   0x40000000
+#define MD_CONTEXT_ARM64_INTEGER           (MD_CONTEXT_ARM64 | 0x00000002)
+#define MD_CONTEXT_ARM64_FLOATING_POINT    (MD_CONTEXT_ARM64 | 0x00000004)
 
-#define MD_CONTEXT_ARM_FULL              (MD_CONTEXT_ARM_INTEGER | \
-                                          MD_CONTEXT_ARM_FLOATING_POINT)
+#define MD_CONTEXT_ARM64_FULL              (MD_CONTEXT_ARM64_INTEGER | \
+                                          MD_CONTEXT_ARM64_FLOATING_POINT)
 
-#define MD_CONTEXT_ARM_ALL               (MD_CONTEXT_ARM_INTEGER | \
-                                          MD_CONTEXT_ARM_FLOATING_POINT)
+#define MD_CONTEXT_ARM64_ALL               (MD_CONTEXT_ARM64_INTEGER | \
+                                          MD_CONTEXT_ARM64_FLOATING_POINT)
 
-#endif  /* GOOGLE_BREAKPAD_COMMON_MINIDUMP_CPU_ARM_H__ */
+#endif  /* GOOGLE_BREAKPAD_COMMON_MINIDUMP_CPU_ARM64_H__ */