Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Longjun Luo <luolongjuna@gmail.com>
To: gdb-patches@sourceware.org
Cc: Kevin Buettner <kevinb@redhat.com>, Longjun Luo <luolongjuna@gmail.com>
Subject: [PATCH v3] gdb/dwarf: filter discarded zero-address variables after relocation
Date: Tue, 27 Oct 2026 16:27:17 +0800	[thread overview]
Message-ID: <20261027082717.3945887-1-luolongjuna@gmail.com> (raw)
In-Reply-To: <20261016040017.1155012-1-luolongjuna@gmail.com>

A variable discarded by section GC can retain a DIE with a zero DWARF
address. var_decode_location adds the section offset before new_symbol
checks that address. If the objfile has already been relocated, the
symbol survives the zero-address check and can read the ELF header as
data.

This affects PIE executables when the CU is first expanded after
starting the program. It also affects normally loaded shared libraries,
including with --readnow, because their objfiles already have a load
offset.

Have var_decode_location return the address it decoded before
relocation, and check that address in new_symbol, as the DWARF reader
already does for discarded functions and address ranges. Preserve the
existing exception for objects with allocated sections at zero.

Extend gdb.dwarf2/dw2-var-zero-addr.exp with a PIE whose CU is first
expanded after starting the program, and with a loaded shared library,
with and without --readnow. Use a distinctive variable name to avoid
clashes with shared-library debug symbols after the program starts.

Tested on x86_64 GNU/Linux with GCC. Without the fix, the three new
cases fail; with it, the test passes. GDB selftests pass.

Signed-off-by: Longjun Luo <luolongjuna@gmail.com>
---
Changes in v3:
- Document the std::nullopt return value in the comment above
  var_decode_location. Leave the two callers that ignore the return
  value as-is, as they only populate the symbol's location.

Regarding the copyright assignment: I looked into the process, but I am
currently unable to complete the paperwork. I understand the full patch
(especially with the new test cases) exceeds the ~15-line threshold for
legally significant changes.

Would it be acceptable to take only the minimal fix in gdb/dwarf2/read.c
under the small-change rule, or should we hold off on this patch until
GDB's DCO policy is finalized? I will hold off on sending further patches
in the meantime.

v2: https://sourceware.org/pipermail/gdb-patches/2026-October/231127.html

 gdb/dwarf2/read.c                             | 27 +++++----
 gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.S  |  2 +-
 .../gdb.dwarf2/dw2-var-zero-addr.exp          | 60 ++++++++++++++++++-
 3 files changed, 76 insertions(+), 13 deletions(-)

diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index 1fc6d7ffb4f..77a74c49b35 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -514,9 +514,8 @@ show_dwarf_synchronous (struct ui_file *file, int from_tty,
 \f
 /* local function prototypes */
 
-static void var_decode_location (struct attribute *attr,
-				 struct symbol *sym,
-				 struct dwarf2_cu *cu);
+static std::optional<unrelocated_addr> var_decode_location
+  (struct attribute *attr, struct symbol *sym, struct dwarf2_cu *cu);
 
 static unsigned int peek_abbrev_code (bfd *, const gdb_byte *);
 
@@ -15160,7 +15159,11 @@ dwarf_decode_line_header (sect_offset sect_off, struct dwarf2_cu *cu,
 				   &cu->header, comp_dir);
 }
 
-static void
+/* Decode the DW_AT_location attribute ATTR of SYM.  If ATTR is a single
+   DW_OP_addr, DW_OP_addrx or DW_OP_GNU_addr_index, return its address
+   before relocation.  Otherwise, return std::nullopt.  */
+
+static std::optional<unrelocated_addr>
 var_decode_location (struct attribute *attr, struct symbol *sym,
 		     struct dwarf2_cu *cu)
 {
@@ -15182,7 +15185,7 @@ var_decode_location (struct attribute *attr, struct symbol *sym,
   if (attr->form_is_block () && attr->as_block ()->size == 0)
     {
       sym->set_loc_class_index (LOC_OPTIMIZED_OUT);
-      return;
+      return {};
     }
 
   /* Handle one degenerate form of location expression specially, to
@@ -15216,7 +15219,7 @@ var_decode_location (struct attribute *attr, struct symbol *sym,
 	  sym->set_value_address
 	    (sym->value_address ()
 	     + objfile->section_offsets[sym->section_index ()]);
-	  return;
+	  return tem;
 	}
     }
 
@@ -15231,6 +15234,8 @@ var_decode_location (struct attribute *attr, struct symbol *sym,
 
   if (sym->computed_ops ()->location_has_loclist)
     cu->has_loclist = true;
+
+  return {};
 }
 
 /* A helper function to add an "export" symbol.  The new symbol starts
@@ -15482,7 +15487,8 @@ new_symbol (struct die_info *die, struct dwarf2_cu *cu, struct symbol *sym,
       attr = dwarf2_attr (die, DW_AT_location, cu);
       if (attr != nullptr)
 	{
-	  var_decode_location (attr, sym, cu);
+	  std::optional<unrelocated_addr> unrelocated
+	    = var_decode_location (attr, sym, cu);
 	  attr2 = dwarf2_attr (die, DW_AT_external, cu);
 
 	  /* Fortran explicitly imports any global symbols to the local
@@ -15491,14 +15497,15 @@ new_symbol (struct die_info *die, struct dwarf2_cu *cu, struct symbol *sym,
 	      && die->parent->tag == DW_TAG_common_block)
 	    attr2 = nullptr;
 
-	  if (sym->loc_class () == LOC_STATIC
-	      && sym->value_address () == 0
+	  if (unrelocated == (unrelocated_addr) 0
 	      && !per_objfile->per_bfd->has_section_at_zero)
 	    {
 	      /* When a static variable is eliminated by the linker,
 		 the corresponding debug information is not stripped
 		 out, but the variable address is set to null;
-		 do not add such variables into symbol table.  */
+		 do not add such variables into symbol table.  Check
+		 the address before relocation, which adds the load
+		 offset of a PIE or shared library.  */
 	    }
 	  else if (attr2 != nullptr && attr2->as_boolean ())
 	    {
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.S b/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.S
index f7394f33946..233bd6ce79e 100644
--- a/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.S
+++ b/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.S
@@ -39,7 +39,7 @@
 	.byte		5			/* DW_AT_encoding */
 
 	.uleb128	3			/* Abbrev: DW_TAG_variable (location) */
-	.ascii		"var\0"			/* DW_AT_name */
+	.ascii		"dw2_discarded_zero_addr_var\0"	/* DW_AT_name */
 	.byte		2f - 1f			/* DW_AT_location */
 1:	.byte		3			/*   DW_OP_addr */
 	.4byte		0			/*   <addr> */
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.exp b/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.exp
index 04a2aa06ea9..360894661e5 100644
--- a/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.exp
+++ b/gdb/testsuite/gdb.dwarf2/dw2-var-zero-addr.exp
@@ -25,9 +25,12 @@ if {[prepare_for_testing "failed to prepare" ${testfile} \
 }
 
 proc test { } {
-    # FAIL was: = (int *) 0x0
+    # FAIL was: = (int *) 0x0, or the load address in a relocated objfile.
     # Such DIE record can be produced using: gcc -fdata-sections -Wl,-gc-sections
-    gdb_test "print &var" {No symbol "var" in current context\.}
+    # Use a distinctive name to avoid clashes with shared-library debug
+    # symbols once the program runs.
+    gdb_test "print &dw2_discarded_zero_addr_var" \
+	{No symbol "dw2_discarded_zero_addr_var" in current context\.}
 }
 
 test
@@ -40,3 +43,56 @@ save_vars { GDBFLAGS } {
 with_test_prefix "readnow" {
     test
 }
+
+# Expand the CU for the first time after the PIE has been relocated.
+proc_with_prefix test_pie_after_run {} {
+    set executable ${::testfile}-pie
+    if {[build_executable "failed to prepare" $executable \
+	     [list $::srcfile $::srcfile2] {nodebug pie}]} {
+	return
+    }
+
+    clean_restart $executable
+    if {[runto_main]} {
+	test
+    }
+}
+
+# A shared library is relocated before its CUs are expanded, even with
+# --readnow.
+proc_with_prefix test_shlib {} {
+    set lib_path [standard_output_file ${::testfile}-shlib.so]
+    if {[gdb_compile_shlib $::srcdir/$::subdir/$::srcfile $lib_path \
+	     {nodebug}] != ""} {
+	untested "failed to compile shared library"
+	return
+    }
+
+    set executable ${::testfile}-shlib
+    if {[build_executable "failed to prepare" $executable $::srcfile2 \
+	     [list debug shlib=$lib_path]]} {
+	return
+    }
+
+    foreach_with_prefix readnow {false true} {
+	save_vars { ::GDBFLAGS } {
+	    if {$readnow} {
+		append ::GDBFLAGS " --readnow"
+	    }
+	    clean_restart $executable
+	}
+	gdb_load_shlib $lib_path
+	if {[runto_main]} {
+	    set lib_name [file tail $lib_path]
+	    gdb_test "info sharedlibrary $lib_name" \
+		".*Yes.*[string_to_regexp $lib_name]" \
+		"shared library is loaded"
+	    test
+	}
+    }
+}
+
+test_pie_after_run
+
+require allow_shlib_tests
+test_shlib
-- 
2.53.0


      parent reply	other threads:[~2026-10-27  8:28 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-09 17:46 [PATCH] " Longjun Luo
2026-10-16  4:00 ` [PATCH v2] " Longjun Luo
2026-10-18 23:29   ` Kevin Buettner
2026-10-27  8:27   ` Longjun Luo [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261027082717.3945887-1-luolongjuna@gmail.com \
    --to=luolongjuna@gmail.com \
    --cc=gdb-patches@sourceware.org \
    --cc=kevinb@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox