[PATCH] hurd: unwinding support over signal trampolines

[PATCH] hurd: unwinding support over signal trampolines

[Samuel Thibault]

gdb/ChangeLog:

2020-05-29  Samuel Thibault  <samuel.thibault@ens-lyon.org>

	* i386-gnu-tdep.c: Include "gdbcore.h"
	(gnu_sigtramp_code, i386_gnu_sc_reg_offset): New arrays.
        (GNU_SIGTRAMP_LEN, GNU_SIGTRAMP_TAIL,
        I386_GNU_SIGCONTEXT_THREAD_STATE_OFFSET): New macros
        (i386_gnu_sigtramp_start, i386_gnu_sigtramp_p,
        i386_gnu_sigcontext_addr): New functions
        (i386gnu_init_abi): Register i386_gnu_sigtramp_p,
        i386_gnu_sigcontext_addr, and i386_gnu_sc_reg_offset in the gdbarch
        tdep.

Index: binutils-gdb/gdb/i386-gnu-tdep.c
===================================================================
--- binutils-gdb.orig/gdb/i386-gnu-tdep.c
+++ binutils-gdb/gdb/i386-gnu-tdep.c
@@ -17,11 +17,138 @@
    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
 
 #include "defs.h"
+#include "gdbcore.h"
 #include "osabi.h"
 #include "solib-svr4.h"
 
 #include "i386-tdep.h"
 
+/* Recognizing signal handler frames.  */
+
+/* When the GNU/Hurd libc calls a signal handler, the return address points
+   inside the trampoline assembly snippet.
+
+   If the trampoline function name can not be identified, we resort to reading
+   memory from the process in order to identify it.  */
+
+static const gdb_byte gnu_sigtramp_code[] =
+{
+/* rpc_wait_trampoline: */
+  0xb8, 0xe7, 0xff, 0xff, 0xff,			/* mov    $-25,%eax */
+  0x9a, 0x00, 0x00, 0x00, 0x00, 0x07, 0x00,	/* lcall  $7,$0 */
+  0x89, 0x01,					/* movl   %eax, (%ecx) */
+  0x89, 0xdc,					/* movl   %ebx, %esp */
+
+/* trampoline: */
+  0xff, 0xd2,					/* call   *%edx */
+/* RA HERE */
+  0x83, 0xc4, 0x0c,				/* addl   $12, %esp */
+  0xc3,						/* ret */
+
+/* firewall: */
+  0xf4,						/* hlt */
+};
+
+#define GNU_SIGTRAMP_LEN (sizeof gnu_sigtramp_code)
+#define GNU_SIGTRAMP_TAIL 5			/* length of tail after RA */
+
+/* If THIS_FRAME is a sigtramp routine, return the address of the
+   start of the routine.  Otherwise, return 0.  */
+
+static CORE_ADDR
+i386_gnu_sigtramp_start (struct frame_info *this_frame)
+{
+  CORE_ADDR pc = get_frame_pc (this_frame);
+  gdb_byte buf[GNU_SIGTRAMP_LEN];
+
+  if (!safe_frame_unwind_memory (this_frame,
+				 pc + GNU_SIGTRAMP_TAIL - GNU_SIGTRAMP_LEN,
+				 buf, GNU_SIGTRAMP_LEN))
+    return 0;
+
+  if (memcmp (buf, gnu_sigtramp_code, GNU_SIGTRAMP_LEN) != 0)
+    return 0;
+
+  return pc;
+}
+
+/* Return whether THIS_FRAME corresponds to a GNU/Linux sigtramp
+   routine.  */
+
+static int
+i386_gnu_sigtramp_p (struct frame_info *this_frame)
+{
+  CORE_ADDR pc = get_frame_pc (this_frame);
+  const char *name;
+
+  find_pc_partial_function (pc, &name, NULL, NULL);
+
+  /* If we have a NAME, we can check for the trampoline function */
+  if (name != NULL && strcmp (name, "trampoline") == 0)
+    return 1;
+
+  return i386_gnu_sigtramp_start (this_frame) != 0;
+}
+
+/* Offset to sc_i386_thread_state in sigcontext, from <bits/sigcontext.h>.  */
+#define I386_GNU_SIGCONTEXT_THREAD_STATE_OFFSET 20
+
+/* Assuming THIS_FRAME is a GNU/Linux sigtramp routine, return the
+   address of the associated sigcontext structure.  */
+
+static CORE_ADDR
+i386_gnu_sigcontext_addr (struct frame_info *this_frame)
+{
+  struct gdbarch *gdbarch = get_frame_arch (this_frame);
+  enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
+  CORE_ADDR pc;
+  CORE_ADDR sp;
+  gdb_byte buf[4];
+
+  get_frame_register (this_frame, I386_ESP_REGNUM, buf);
+  sp = extract_unsigned_integer (buf, 4, byte_order);
+
+  pc = i386_gnu_sigtramp_start (this_frame);
+  if (pc)
+    {
+      CORE_ADDR sigcontext_addr;
+
+      /* The sigcontext structure address is passed as the third argument to
+	 the signal handler. */
+      read_memory (sp + 8, buf, 4);
+      sigcontext_addr = extract_unsigned_integer (buf, 4, byte_order);
+      return sigcontext_addr + I386_GNU_SIGCONTEXT_THREAD_STATE_OFFSET;
+    }
+
+  error (_("Couldn't recognize signal trampoline."));
+  return 0;
+}
+
+/* Mapping between the general-purpose registers in `struct
+   sigcontext' format (starting at sc_i386_thread_state)
+   and GDB's register cache layout.  */
+
+/* From <bits/sigcontext.h>.  */
+static int i386_gnu_sc_reg_offset[] =
+{
+  11 * 4,			/* %eax */
+  10 * 4,			/* %ecx */
+  9 * 4,			/* %edx */
+  8 * 4,			/* %ebx */
+  7 * 4,			/* %esp */
+  6 * 4,			/* %ebp */
+  5 * 4,			/* %esi */
+  4 * 4,			/* %edi */
+  12 * 4,			/* %eip */
+  14 * 4,			/* %eflags */
+  13 * 4,			/* %cs */
+  16 * 4,			/* %ss */
+  3 * 4,			/* %ds */
+  2 * 4,			/* %es */
+  1 * 4,			/* %fs */
+  0 * 4				/* %gs */
+};
+
 /* From <sys/ucontext.h>.  */
 static int i386gnu_gregset_reg_offset[] =
 {
@@ -59,6 +186,11 @@ i386gnu_init_abi (struct gdbarch_info in
   tdep->sizeof_gregset = 19 * 4;
 
   tdep->jb_pc_offset = 20;	/* From <bits/setjmp.h>.  */
+
+  tdep->sigtramp_p = i386_gnu_sigtramp_p;
+  tdep->sigcontext_addr = i386_gnu_sigcontext_addr;
+  tdep->sc_reg_offset = i386_gnu_sc_reg_offset;
+  tdep->sc_num_regs = ARRAY_SIZE (i386_gnu_sc_reg_offset);
 }
 
 void _initialize_i386gnu_tdep ();

Re: [PATCH] hurd: unwinding support over signal trampolines

[Simon Marchi]

On 2020-05-29 7:00 p.m., Samuel Thibault wrote:
> gdb/ChangeLog:
> 
> 2020-05-29  Samuel Thibault  <samuel.thibault@ens-lyon.org>
> 
> 	* i386-gnu-tdep.c: Include "gdbcore.h"
> 	(gnu_sigtramp_code, i386_gnu_sc_reg_offset): New arrays.
>         (GNU_SIGTRAMP_LEN, GNU_SIGTRAMP_TAIL,
>         I386_GNU_SIGCONTEXT_THREAD_STATE_OFFSET): New macros
>         (i386_gnu_sigtramp_start, i386_gnu_sigtramp_p,
>         i386_gnu_sigcontext_addr): New functions
>         (i386gnu_init_abi): Register i386_gnu_sigtramp_p,
>         i386_gnu_sigcontext_addr, and i386_gnu_sc_reg_offset in the gdbarch
>         tdep.

Please add a commit message.  It should at least explain why this change is needed / desirable
(what case isn't working, what functionality is missing) and what it does to address that (not
in terms of which functions are added / removed, but how the GDB behavior changes as a
consequence of that change).  If the "how" is complex, it's always good to include some notes
about that to help those who aren't experts of the system you work on.

Simon