[PATCH] Fix setting-breakpoints regression on PPC64 (function descriptors) (was: Re: ppc64 regression: [PATCH 1/2] Fix "list ambiguous_variable")

[Pedro Alves <palves at redhat dot com>]

On 11/26/2017 04:37 PM, Ulrich Weigand wrote:

> I now see this as well on Cell/B.E.  This is a serious regression that
> causes "start" to always fail for me ...
> 
> The problem seems to be that GDB sets a breakpoint into the function
> descriptor for main, which is not a good idea.
> 
> Looking at the commit identified above, it seems that GDB now only
> runs the address through gdbarch_convert_from_func_ptr_addr if
> msymbol_is_text returns true.  However, if the symbol points to
> a function descriptor, msymbol_is_text would be false since this
> is in fact outside the text section.
> 
> So I think probably we need to still run the address through
> gdbarch_convert_from_func_ptr_addr, and if that detects that it
> was indeed a function descriptor, always treat the resulting
> address as a function.

Thanks for the suggestion.  Here's a patch that implements
something like that.  WDYT?

I've pushed this to the users/palves/fix-ppc64-func-descriptors
branch as well.

>From d1e2a9dc61a645ff29606f01f96a02b477761c23 Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Wed, 29 Nov 2017 18:06:33 +0000
Subject: [PATCH] Fix setting-breakpoints regression on PPC64 (function
 descriptors)

The recent-ish commit e5f25bc5d6db ('Fix "list ambiguous_variable"')
caused a serious regression on PPC64.  See
<https://sourceware.org/ml/gdb-patches/2017-11/msg00666.html>.

Basically, after that patch, GDB sets breakpoints in function
descriptors instead of where the descriptors point to, which is
incorrect.

The problem is that GDB now only runs a minsym's address through
gdbarch_convert_from_func_ptr_addr if msymbol_is_text returns true.
However, if the symbol points to a function descriptor,
msymbol_is_text is false since function descriptors are in fact
outside the text section.

The fix is to also run a non-text address through
gdbarch_convert_from_func_ptr_addr, and if that detects that it was
indeed a function descriptor, treat the resulting address as a
function.

While implementing that directly in linespec.c:minsym_found (where the
bad msymbol_is_text check is) fixes the issue, I noticed that
linespec.c:add_minsym has some code that also basically needs to do
the same checks, however it's implemented differently.  Also,
add_minsym is calling find_pc_sect_line on non-function symbols, which
also doesn't look right.

So I introduced msymbol_is_function, so that we have a simple place to
consider minsyms and function descriptors.

And then, the only other use of msymbol_is_text is in
find_function_alias_target, which turns out to also be incorrect.
Changing that one to use msymbol_is_function, i.e., to consider
function descriptors too fixes (on PPC64):

  -FAIL: gdb.base/symbol-alias.exp: p func_alias
  -FAIL: gdb.base/symbol-alias.exp: p *func_alias()
  +PASS: gdb.base/symbol-alias.exp: p func_alias
  +PASS: gdb.base/symbol-alias.exp: p *func_alias()

And then after that, msymbol_is_text is no longer used anywhere, so it
can be removed.

Tested on x86_64 GNU/Linux, no regressions.  Tested on PPC64 GNU/Linux
and results compared to a testrun of e5f25bc5d6db^ (before the
offending commit), also no regressions.  (there's a couple new FAILs
and some new symbol name matching unit tests are crashing, but that
looks unrelated).

gdb/ChangeLog:
yyyy-mm-dd  Pedro Alves  <palves@redhat.com>

	* linespec.c (minsym_found, add_minsym): Use msymbol_is_function.
	* minsyms.c (msymbol_is_text): Delete.
	(msymbol_is_function): New function.
	* minsyms.h (msymbol_is_text): Delete.
	(msymbol_is_function): New declaration.
	* symtab.c (find_function_alias_target): Use msymbol_is_function.
---
 gdb/linespec.c | 83 ++++++++++++++++------------------------------------------
 gdb/minsyms.c  | 34 ++++++++++++++++++------
 gdb/minsyms.h  | 11 +++++---
 gdb/symtab.c   |  8 +++---
 4 files changed, 61 insertions(+), 75 deletions(-)

diff --git a/gdb/linespec.c b/gdb/linespec.c
index 3f7f171..83f9bbf 100644
--- a/gdb/linespec.c
+++ b/gdb/linespec.c
@@ -4309,22 +4309,16 @@ minsym_found (struct linespec_state *self, struct objfile *objfile,
 	      struct minimal_symbol *msymbol,
 	      std::vector<symtab_and_line> *result)
 {
-  struct gdbarch *gdbarch = get_objfile_arch (objfile);
-  CORE_ADDR pc;
   struct symtab_and_line sal;
 
-  if (msymbol_is_text (msymbol))
-    {
-      sal = find_pc_sect_line (MSYMBOL_VALUE_ADDRESS (objfile, msymbol),
-			       (struct obj_section *) 0, 0);
-      sal.section = MSYMBOL_OBJ_SECTION (objfile, msymbol);
+  /* The minimal symbol might point to a function descriptor, which is
+     not a text symbol; try resolving it to the actual code
+     address.  */
 
-      /* The minimal symbol might point to a function descriptor;
-	 resolve it to the actual code address instead.  */
-      pc = gdbarch_convert_from_func_ptr_addr (gdbarch, sal.pc,
-					       &current_target);
-      if (pc != sal.pc)
-	sal = find_pc_sect_line (pc, NULL, 0);
+  CORE_ADDR func_addr;
+  if (msymbol_is_function (objfile, msymbol, &func_addr))
+    {
+      sal = find_pc_sect_line (func_addr, NULL, 0);
 
       if (self->funfirstline)
 	{
@@ -4332,14 +4326,9 @@ minsym_found (struct linespec_state *self, struct objfile *objfile,
 	      && (COMPUNIT_LOCATIONS_VALID (SYMTAB_COMPUNIT (sal.symtab))
 		  || SYMTAB_LANGUAGE (sal.symtab) == language_asm))
 	    {
-	      /* If gdbarch_convert_from_func_ptr_addr does not apply then
-		 sal.SECTION, sal.LINE&co. will stay correct from above.
-		 If gdbarch_convert_from_func_ptr_addr applies then
-		 sal.SECTION is cleared from above and sal.LINE&co. will
-		 stay correct from the last find_pc_sect_line above.  */
-	      sal.pc = MSYMBOL_VALUE_ADDRESS (objfile, msymbol);
-	      sal.pc = gdbarch_convert_from_func_ptr_addr (gdbarch, sal.pc,
-							   &current_target);
+	      struct gdbarch *gdbarch = get_objfile_arch (objfile);
+
+	      sal.pc = func_addr;
 	      if (gdbarch_skip_entrypoint_p (gdbarch))
 		sal.pc = gdbarch_skip_entrypoint (gdbarch, sal.pc);
 	    }
@@ -4424,52 +4413,26 @@ static void
 add_minsym (struct minimal_symbol *minsym, void *d)
 {
   struct collect_minsyms *info = (struct collect_minsyms *) d;
-  bound_minimal_symbol_d mo;
-
-  mo.minsym = minsym;
-  mo.objfile = info->objfile;
 
   if (info->symtab != NULL)
     {
-      CORE_ADDR pc;
-      struct symtab_and_line sal;
-      struct gdbarch *gdbarch = get_objfile_arch (info->objfile);
-
-      sal = find_pc_sect_line (MSYMBOL_VALUE_ADDRESS (info->objfile, minsym),
-			       NULL, 0);
-      sal.section = MSYMBOL_OBJ_SECTION (info->objfile, minsym);
-      pc
-	= gdbarch_convert_from_func_ptr_addr (gdbarch, sal.pc, &current_target);
-      if (pc != sal.pc)
-	sal = find_pc_sect_line (pc, NULL, 0);
+      /* We're looking for a label for which we don't have debug
+	 info.  */
+      CORE_ADDR func_addr;
+      if (msymbol_is_function (info->objfile, minsym, &func_addr))
+	{
+	  symtab_and_line sal = find_pc_sect_line (func_addr, NULL, 0);
 
-      if (info->symtab != sal.symtab)
-	return;
+	  if (info->symtab != sal.symtab)
+	    return;
+	}
     }
 
-  /* Exclude data symbols when looking for breakpoint locations.   */
-  if (!info->list_mode)
-    switch (minsym->type)
-      {
-	case mst_slot_got_plt:
-	case mst_data:
-	case mst_bss:
-	case mst_abs:
-	case mst_file_data:
-	case mst_file_bss:
-	  {
-	    /* Make sure this minsym is not a function descriptor
-	       before we decide to discard it.  */
-	    struct gdbarch *gdbarch = get_objfile_arch (info->objfile);
-	    CORE_ADDR addr = gdbarch_convert_from_func_ptr_addr
-			       (gdbarch, BMSYMBOL_VALUE_ADDRESS (mo),
-				&current_target);
-
-	    if (addr == BMSYMBOL_VALUE_ADDRESS (mo))
-	      return;
-	  }
-      }
+  /* Exclude data symbols when looking for breakpoint locations.  */
+  if (!info->list_mode && !msymbol_is_function (info->objfile, minsym))
+    return;
 
+  bound_minimal_symbol_d mo = {minsym, info->objfile};
   VEC_safe_push (bound_minimal_symbol_d, info->msyms, &mo);
 }
 
diff --git a/gdb/minsyms.c b/gdb/minsyms.c
index 4898da1..d68fb65 100644
--- a/gdb/minsyms.c
+++ b/gdb/minsyms.c
@@ -57,17 +57,35 @@
 /* See minsyms.h.  */
 
 bool
-msymbol_is_text (minimal_symbol *msymbol)
+msymbol_is_function (struct objfile *objfile, minimal_symbol *minsym,
+		     CORE_ADDR *func_address_p)
 {
-  switch (MSYMBOL_TYPE (msymbol))
+  CORE_ADDR msym_addr = MSYMBOL_VALUE_ADDRESS (objfile, minsym);
+
+  switch (minsym->type)
     {
-    case mst_text:
-    case mst_text_gnu_ifunc:
-    case mst_solib_trampoline:
-    case mst_file_text:
-      return true;
+    case mst_slot_got_plt:
+    case mst_data:
+    case mst_bss:
+    case mst_abs:
+    case mst_file_data:
+    case mst_file_bss:
+      {
+	struct gdbarch *gdbarch = get_objfile_arch (objfile);
+	CORE_ADDR pc = gdbarch_convert_from_func_ptr_addr (gdbarch, msym_addr,
+							   &current_target);
+	if (pc != msym_addr)
+	  {
+	    if (func_address_p != NULL)
+	      *func_address_p = pc;
+	    return true;
+	  }
+	return false;
+      }
     default:
-      return false;
+      if (func_address_p != NULL)
+	*func_address_p = msym_addr;
+      return true;
     }
 }
 
diff --git a/gdb/minsyms.h b/gdb/minsyms.h
index 5c0dde4..baa87f0 100644
--- a/gdb/minsyms.h
+++ b/gdb/minsyms.h
@@ -159,9 +159,14 @@ void terminate_minimal_symbol_table (struct objfile *objfile);
 
 
 
-/* Return whether MSYMBOL is a function/method.  */
-
-bool msymbol_is_text (minimal_symbol *msymbol);
+/* Return whether MSYMBOL is a function/method.  If FUNC_ADDRESS_P is
+   non-NULL, and the MSYMBOL is a function, then *FUNC_ADDRESS_P is
+   set to the function's address, already resolved if MINSYM points to
+   a function descriptor.  */
+
+bool msymbol_is_function (struct objfile *objfile,
+			  minimal_symbol *minsym,
+			  CORE_ADDR *func_address_p = NULL);
 
 /* Compute a hash code for the string argument.  */
 
diff --git a/gdb/symtab.c b/gdb/symtab.c
index 3d59367..a249a8d 100644
--- a/gdb/symtab.c
+++ b/gdb/symtab.c
@@ -3927,14 +3927,14 @@ skip_prologue_using_sal (struct gdbarch *gdbarch, CORE_ADDR func_addr)
 symbol *
 find_function_alias_target (bound_minimal_symbol msymbol)
 {
-  if (!msymbol_is_text (msymbol.minsym))
+  CORE_ADDR func_addr;
+  if (!msymbol_is_function (msymbol.objfile, msymbol.minsym, &func_addr))
     return NULL;
 
-  CORE_ADDR addr = BMSYMBOL_VALUE_ADDRESS (msymbol);
-  symbol *sym = find_pc_function (addr);
+  symbol *sym = find_pc_function (func_addr);
   if (sym != NULL
       && SYMBOL_CLASS (sym) == LOC_BLOCK
-      && BLOCK_START (SYMBOL_BLOCK_VALUE (sym)) == addr)
+      && BLOCK_START (SYMBOL_BLOCK_VALUE (sym)) == func_addr)
     return sym;
 
   return NULL;
-- 
2.5.5