x86: fix unw_is_signal_frame() and unwinding across signal handlers

(Logical change 1.109)

src/x86/Gis_signal_frame-x86.c

@@ -36,14 +36,25 @@ unw_is_signal_frame (unw_cursor_t *cursor)
 
   /* Check if EIP points at sigreturn() sequence.  On Linux, this is:
 
+    __restore:
 	0x58				pop %eax
 	0xb8 0x77 0x00 0x00 0x00	movl 0x77,%eax
 	0xcd 0x80			int 0x80
+
+     without SA_SIGINFO, and
+
+    __restore_rt:
+       0xb8 0xad 0x00 0x00 0x00        movl 0x80,%eax
+       0xcd 0x80                       int 0x80
+       0x90                            nop
+
+     if SA_SIGINFO is specified.
   */
   if ((ret = (*a->access_mem) (c->as, c->eip, &w0, 0, c->as_arg)) < 0
       || (ret = (*a->access_mem) (c->as, c->eip + 4, &w1, 0, c->as_arg)) < 0)
     return ret;
-  return (w0 == 0x0077b858) && (w1 == 0x80cd0000);
+  return (w0 == 0x0077b858) && (w1 == 0x80cd0000)
+         || (w0 == 0x0000adb8) && (w1 == 0x9080cd00);
 #else
   printf ("%s: implement me\n", __FUNCTION__);
 #endif

src/x86/Gstep-x86.c

@@ -0,0 +1,189 @@
+/* libunwind - a platform-independent unwind library
+   Copyright (C) 2002 Hewlett-Packard Co
+	Contributed by David Mosberger-Tang <davidm@hpl.hp.com>
+
+This file is part of libunwind.
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
+
+#include "unwind_i.h"
+#include "offsets.h"
+
+static inline int
+update_frame_state (struct cursor *c)
+{
+#if 0
+  unw_word_t prev_ip, prev_sp, prev_bsp, ip, pr, num_regs, cfm;
+  int ret;
+
+  prev_ip = c->ip;
+  prev_sp = c->sp;
+  prev_bsp = c->bsp;
+
+  c->cfm_loc = c->pfs_loc;
+
+  num_regs = 0;
+  if (c->is_signal_frame)
+    {
+      ret = ia64_get (c, c->sp + 0x10 + SIGFRAME_ARG2_OFF, &c->sigcontext_loc);
+      debug (100, "%s: sigcontext_loc=%lx (ret=%d)\n",
+	     __FUNCTION__, c->sigcontext_loc, ret);
+      if (ret < 0)
+	return ret;
+
+      if (c->ip_loc == c->sigcontext_loc + SIGCONTEXT_BR_OFF + 0*8)
+	{
+	  /* Earlier kernels (before 2.4.19 and 2.5.10) had buggy
+	     unwind info for sigtramp.  Fix it up here.  */
+	  c->ip_loc  = (c->sigcontext_loc + SIGCONTEXT_IP_OFF);
+	  c->cfm_loc = (c->sigcontext_loc + SIGCONTEXT_CFM_OFF);
+	}
+
+      /* do what can't be described by unwind directives: */
+      c->pfs_loc = (c->sigcontext_loc + SIGCONTEXT_AR_PFS_OFF);
+
+      ret = ia64_get (c, c->cfm_loc, &cfm);
+      if (ret < 0)
+	return ret;
+
+      num_regs = cfm & 0x7f;		/* size of frame */
+    }
+  else
+    {
+      ret = ia64_get (c, c->cfm_loc, &cfm);
+      if (ret < 0)
+	return ret;
+      num_regs = (cfm >> 7) & 0x7f;	/* size of locals */
+    }
+  c->bsp = ia64_rse_skip_regs (c->bsp, -num_regs);
+
+  /* update the IP cache: */
+  ret = ia64_get (c, c->ip_loc, &ip);
+  if (ret < 0)
+    return ret;
+  c->ip = ip;
+
+  if ((ip & 0xc) != 0)
+    {
+      /* don't let obviously bad addresses pollute the cache */
+      debug (1, "%s: rejecting bad ip=0x%lx\n",  __FUNCTION__, (long) c->ip);
+      return -UNW_EINVALIDIP;
+    }
+  if (ip == 0)
+    /* end of frame-chain reached */
+    return 0;
+
+  pr = c->pr;
+  c->sp = c->psp;
+  c->is_signal_frame = 0;
+
+  if (c->ip == prev_ip && c->sp == prev_sp && c->bsp == prev_bsp)
+    {
+      dprintf ("%s: ip, sp, and bsp unchanged; stopping here (ip=0x%lx)\n",
+	       __FUNCTION__, (long) ip);
+      return -UNW_EBADFRAME;
+    }
+
+  /* as we unwind, the saved ar.unat becomes the primary unat: */
+  c->pri_unat_loc = c->unat_loc;
+
+  /* restore the predicates: */
+  ret = ia64_get (c, c->pr_loc, &c->pr);
+  if (ret < 0)
+    return ret;
+
+  c->pi_valid = 0;
+#endif
+  return 0;
+}
+
+
+int
+unw_step (unw_cursor_t *cursor)
+{
+  struct cursor *c = (struct cursor *) cursor;
+  int ret;
+
+
+  if (unw_is_signal_frame(cursor))
+    {
+      /* Without SA_SIGINFO,                             */
+      /* c -> esp points at the arguments to the handler. */
+      /* This consists of a signal number followed by a          */
+      /* struct sigcontext.                              */
+      /* With SA_SIGINFO, the arguments consists of the          */
+      /* signal number, a siginfo *, and a ucontext * .          */
+      unw_word_t siginfo_ptr_addr = c->esp + 4;
+      unw_word_t sigcontext_ptr_addr = c->esp + 8;
+      unw_word_t siginfo_ptr, sigcontext_ptr;
+      struct x86_loc esp_loc, siginfo_ptr_loc, sigcontext_ptr_loc;
+
+      siginfo_ptr_loc = X86_LOC (siginfo_ptr_addr, 0);
+      sigcontext_ptr_loc = X86_LOC (sigcontext_ptr_addr, 0);
+      ret = (x86_get (c, siginfo_ptr_loc, &siginfo_ptr)
+            | x86_get (c, sigcontext_ptr_loc, &sigcontext_ptr));
+      if (ret < 0)
+       return 0;
+      if (siginfo_ptr < c -> esp || siginfo_ptr > c -> esp + 256
+         || sigcontext_ptr < c -> esp || sigcontext_ptr > c -> esp + 256)
+        {
+         /* Not plausible for SA_SIGINFO signal */
+          unw_word_t sigcontext_addr = c->esp + 4;
+          esp_loc = X86_LOC (sigcontext_addr + LINUX_SC_ESP_OFF, 0);
+          c->ebp_loc = X86_LOC (sigcontext_addr + LINUX_SC_EBP_OFF, 0);
+          c->eip_loc = X86_LOC (sigcontext_addr + LINUX_SC_EIP_OFF, 0);
+        }
+      else
+       {
+         /* If SA_SIGINFO were not specified, we actually read         */
+         /* various segment pointers instead.  We believe that at      */
+         /* least fs and _fsh are always zero for linux, so it is      */
+         /* is not just unlikely, but impossible that we would end up  */
+         /* here.                                                      */
+         esp_loc = X86_LOC (sigcontext_ptr + LINUX_UC_ESP_OFF, 0);
+          c->ebp_loc = X86_LOC (sigcontext_ptr + LINUX_UC_EBP_OFF, 0);
+          c->eip_loc = X86_LOC (sigcontext_ptr + LINUX_UC_EIP_OFF, 0);
+       }
+      ret = x86_get (c, esp_loc, &c->esp);
+      if (ret < 0)
+       return 0;
+    }
+  else
+    {
+      ret = x86_get (c, c->ebp_loc, &c->esp);
+      if (ret < 0)
+        return ret;
+
+      c->ebp_loc = X86_LOC (c->esp, 0);
+      c->eip_loc = X86_LOC (c->esp + 4, 0);
+      c->esp += 8;
+    }
+
+  if (X86_GET_LOC (c->ebp_loc))
+    {
+      ret = x86_get (c, c->eip_loc, &c->eip);
+      if (ret < 0)
+	return ret;
+    }
+  else
+    c->eip = 0;
+
+  return (c->eip == 0) ? 0 : 1;
+}

src/x86/offsets.h

@@ -0,0 +1,17 @@
+/* Linux-specific definitions: */
+
+/* Define various structure offsets to simplify cross-compilation.  */
+
+/* The struct sigcontext is located at an offset of 4
+   from the stack pointer in the signal frame.         */
+
+#define LINUX_SC_ESP_OFF       0x1c
+#define LINUX_SC_EBP_OFF       0x18
+#define LINUX_SC_EIP_OFF       0x38
+
+/* With SA_SIGINFO set, we believe that basically the same
+   layout is used for ucontext_t, except that 20 bytes are added
+   at the beginning. */
+#define LINUX_UC_ESP_OFF       (LINUX_SC_ESP_OFF+20)
+#define LINUX_UC_EBP_OFF       (LINUX_SC_EBP_OFF+20)
+#define LINUX_UC_EIP_OFF       (LINUX_SC_EIP_OFF+20)