[PATCH 0/3] Fix some sanitizer errors

[PATCH 0/3] Fix some sanitizer errors

[Tom Tromey]

I tried a build with the address sanitizer, and another with the
undefined behavior sanitizer.

This series fixes the simplest bugs I found.

There is still one remaining address sanitizer report that I didn't
investigate, coming from gdb.dwarf2/dw2-icc-opaque.exp.  There's also
one remaining undefined behavior report, triggered by
gdb.ada/access_to_packed_array.exp.

Tested on x86-64 Fedora 32.

Tom

[PATCH 1/3] Add virtual destructor to dwarf2_per_cu_data

[Tom Tromey]

Address sanitizer pointed out that the patch to use 'delete' for
dwarf2_per_cu_data introduced a bug -- now it is possible to delete a
signatured_type using a pointer to its base class.  This patch fixes
the problem by introducing a virtual destructor.

gdb/ChangeLog
2021-05-03  Tom Tromey  <tromey@adacore.com>

	* dwarf2/read.h (dwarf2_per_cu_data): Add virtual destructor.
---
 gdb/ChangeLog     | 4 ++++
 gdb/dwarf2/read.h | 2 ++
 2 files changed, 6 insertions(+)

diff --git a/gdb/dwarf2/read.h b/gdb/dwarf2/read.h
index 80f0e3ee13b..f514b8316ec 100644
--- a/gdb/dwarf2/read.h
+++ b/gdb/dwarf2/read.h
@@ -418,6 +418,8 @@ struct dwarf2_per_cu_data
   {
   }
 
+  virtual ~dwarf2_per_cu_data () = default;
+
   /* The start offset and length of this compilation unit.
      NOTE: Unlike comp_unit_head.length, this length includes
      initial_length_size.
-- 
2.26.3

[PATCH 2/3] Fix buffer underflow in add_path

[Tom Tromey]

Address sanitizer pointed out a buglet in source.c:add_path.
In this test, from gdb.base/source-dir.exp:

    (gdb) set directories :/foo:/bar

... 'p[-1]' will result in a buffer underflow.
This patch fixes the bug by introducing a new check.

gdb/ChangeLog
2021-05-03  Tom Tromey  <tromey@adacore.com>

	* source.c (add_path): Check 'p' before using 'p[-1]'.
---
 gdb/ChangeLog | 4 ++++
 gdb/source.c  | 1 +
 2 files changed, 5 insertions(+)

diff --git a/gdb/source.c b/gdb/source.c
index 6fc27ae72f7..b6dab6eb236 100644
--- a/gdb/source.c
+++ b/gdb/source.c
@@ -537,6 +537,7 @@ add_path (const char *dirname, char **which_path, int parse_separators)
       /* On MS-DOS and MS-Windows, h:\ is different from h: */
 	     && !(p == name + 3 && name[1] == ':')		/* "d:/" */
 #endif
+	     && p > name
 	     && IS_DIR_SEPARATOR (p[-1]))
 	/* Sigh.  "foo/" => "foo" */
 	--p;
-- 
2.26.3

[PATCH 3/3] Fix ubsan build

[Tom Tromey]

I tried a build using the undefined behavior sanitizer, and gcc gave
this error:

In file included from /usr/include/string.h:495,
                 from ../gnulib/import/string.h:41,
                 from ../../binutils-gdb/gdb/../gdbsupport/common-defs.h:95,
                 from ../../binutils-gdb/gdb/nat/linux-osdata.c:20:
In function ‘char* strncpy(char*, const char*, size_t)’,
    inlined from ‘void time_from_time_t(char*, int, TIME_T)’ at ../../binutils-gdb/gdb/nat/linux-osdata.c:923:15,
    inlined from ‘void time_from_time_t(char*, int, TIME_T)’ at ../../binutils-gdb/gdb/nat/linux-osdata.c:911:1,
    inlined from ‘void linux_xfer_osdata_sem(buffer*)’ at ../../binutils-gdb/gdb/nat/linux-osdata.c:1082:22:
/usr/include/bits/string_fortified.h:106:34: error: ‘char* __builtin_strncpy(char*, const char*, long unsigned int)’ specified bound 32 equals destination size [-Werror=stringop-truncation]

This patch fixes the problem by subtracting one from the length
parameter to strncpy.

I changed a couple of other similar functions -- gcc does not warn
about these, but I didn't see any substantial difference between the
different cases, and I think these are just latent warnings, to be
triggered in the future by a change to inlining heuristics.

gdb/ChangeLog
2021-05-03  Tom Tromey  <tromey@adacore.com>

	* nat/linux-osdata.c (user_from_uid, time_from_time_t)
	(group_from_gid): Subtract one from strncpy length.
---
 gdb/ChangeLog          | 5 +++++
 gdb/nat/linux-osdata.c | 6 +++---
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/gdb/nat/linux-osdata.c b/gdb/nat/linux-osdata.c
index 7034dd82376..12f66d3c981 100644
--- a/gdb/nat/linux-osdata.c
+++ b/gdb/nat/linux-osdata.c
@@ -212,7 +212,7 @@ user_from_uid (char *user, int maxlen, uid_t uid)
 
   if (pwentry)
     {
-      strncpy (user, pwentry->pw_name, maxlen);
+      strncpy (user, pwentry->pw_name, maxlen - 1);
       /* Ensure that the user name is null-terminated.  */
       user[maxlen - 1] = '\0';
     }
@@ -920,7 +920,7 @@ time_from_time_t (char *time, int maxlen, TIME_T seconds)
 	 characters long.  */
       char buf[30];
       const char *time_str = ctime_r (&t, buf);
-      strncpy (time, time_str, maxlen);
+      strncpy (time, time_str, maxlen - 1);
       time[maxlen - 1] = '\0';
     }
 }
@@ -935,7 +935,7 @@ group_from_gid (char *group, int maxlen, gid_t gid)
 
   if (grentry)
     {
-      strncpy (group, grentry->gr_name, maxlen);
+      strncpy (group, grentry->gr_name, maxlen - 1);
       /* Ensure that the group name is null-terminated.  */
       group[maxlen - 1] = '\0';
     }
-- 
2.26.3

Re: [PATCH 1/3] Add virtual destructor to dwarf2_per_cu_data

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tromey@adacore.com> writes:

Tom> Address sanitizer pointed out that the patch to use 'delete' for
Tom> dwarf2_per_cu_data introduced a bug -- now it is possible to delete a
Tom> signatured_type using a pointer to its base class.  This patch fixes
Tom> the problem by introducing a virtual destructor.

I'm thinking that instead I should add a custom deleter for these
objects that knows to cast to signatured_type when appropriate.  This
would avoid a vtable pointer just to hold the destructor.

Tom

Re: [PATCH 2/3] Fix buffer underflow in add_path

[Tom de Vries]

On 5/3/21 9:32 PM, Tom Tromey wrote:
> Address sanitizer pointed out a buglet in source.c:add_path.
> In this test, from gdb.base/source-dir.exp:
> 
>     (gdb) set directories :/foo:/bar
> 
> ... 'p[-1]' will result in a buffer underflow.
> This patch fixes the bug by introducing a new check.
> 

I also ran into this and came up with the same solution.  LGTM.

Thanks,
- Tom

> gdb/ChangeLog
> 2021-05-03  Tom Tromey  <tromey@adacore.com>
> 
> 	* source.c (add_path): Check 'p' before using 'p[-1]'.
> ---
>  gdb/ChangeLog | 4 ++++
>  gdb/source.c  | 1 +
>  2 files changed, 5 insertions(+)
> 
> diff --git a/gdb/source.c b/gdb/source.c
> index 6fc27ae72f7..b6dab6eb236 100644
> --- a/gdb/source.c
> +++ b/gdb/source.c
> @@ -537,6 +537,7 @@ add_path (const char *dirname, char **which_path, int parse_separators)
>        /* On MS-DOS and MS-Windows, h:\ is different from h: */
>  	     && !(p == name + 3 && name[1] == ':')		/* "d:/" */
>  #endif
> +	     && p > name
>  	     && IS_DIR_SEPARATOR (p[-1]))
>  	/* Sigh.  "foo/" => "foo" */
>  	--p;
> 

Re: [PATCH 1/3] Add virtual destructor to dwarf2_per_cu_data

[Tom Tromey]

Tom> I'm thinking that instead I should add a custom deleter for these
Tom> objects that knows to cast to signatured_type when appropriate.  This
Tom> would avoid a vtable pointer just to hold the destructor.

Here's the patch.

Tom

commit 25e4481af0d85f3c22c824323c42bc23438c1cbf
Author: Tom Tromey <tromey@adacore.com>
Date:   Mon May 3 12:42:58 2021 -0600

    Change how dwarf2_per_cu_data is deleted
    
    Address sanitizer pointed out that the patch to use 'delete' for
    dwarf2_per_cu_data introduced a bug -- now it is possible to delete a
    signatured_type using a pointer to its base class.
    
    This patch fixes the problem by introducing a deleter and a unique_ptr
    specialization.  A virtual destructor would be more ordinary here, but
    it seemed wasteful to add a vtable just for this purpose.  If virtual
    methods are ever needed here, we can revisit this.
    
    2021-05-06  Tom Tromey  <tromey@adacore.com>
    
            * dwarf2/read.h (struct dwarf2_per_cu_data_deleter: New.
            (dwarf2_per_cu_data_up): New typedef.
            (struct dwarf2_per_bfd) <allocate_per_cu>: Change return type.
            <all_comp_units>: Use dwarf2_per_cu_data_up.
            * dwarf2/read.c (dwarf2_per_cu_data::operator()): New function.
            (dwarf2_per_bfd::allocate_per_cu): Return dwarf2_per_cu_data_up.
            (create_cu_from_index_list): Likewise.
            (create_signatured_type_table_from_index)
            (create_cus_from_debug_names_list, add_type_unit)
            (read_comp_units_from_section): Update.
            (dwarf2_find_containing_comp_unit): Change type of all_comp_units.
            (run_test): Update.

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 187249db47e..78a3dadbdd5 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,18 @@
+2021-05-06  Tom Tromey  <tromey@adacore.com>
+
+	* dwarf2/read.h (struct dwarf2_per_cu_data_deleter: New.
+	(dwarf2_per_cu_data_up): New typedef.
+	(struct dwarf2_per_bfd) <allocate_per_cu>: Change return type.
+	<all_comp_units>: Use dwarf2_per_cu_data_up.
+	* dwarf2/read.c (dwarf2_per_cu_data::operator()): New function.
+	(dwarf2_per_bfd::allocate_per_cu): Return dwarf2_per_cu_data_up.
+	(create_cu_from_index_list): Likewise.
+	(create_signatured_type_table_from_index)
+	(create_cus_from_debug_names_list, add_type_unit)
+	(read_comp_units_from_section): Update.
+	(dwarf2_find_containing_comp_unit): Change type of all_comp_units.
+	(run_test): Update.
+
 2021-05-06  Andrew Burgess  <andrew.burgess@embecosm.com>
 
 	* guile/scm-breakpoint.c (bpscm_print_breakpoint_smob): Only print
diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index 5796cf1730b..e1965c193ed 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -1760,6 +1760,17 @@ dwarf2_queue_item::~dwarf2_queue_item ()
     }
 }
 
+/* See dwarf2/read.h.  */
+
+void
+dwarf2_per_cu_data_deleter::operator() (dwarf2_per_cu_data *data)
+{
+  if (data->is_debug_types)
+    delete static_cast<signatured_type *> (data);
+  else
+    delete data;
+}
+
 /* The return type of find_file_and_directory.  Note, the enclosed
    string pointers are only valid while this object is valid.  */
 
@@ -2522,10 +2533,10 @@ dw2_instantiate_symtab (dwarf2_per_cu_data *per_cu,
 
 /* See read.h.  */
 
-std::unique_ptr<dwarf2_per_cu_data>
+dwarf2_per_cu_data_up
 dwarf2_per_bfd::allocate_per_cu ()
 {
-  std::unique_ptr<dwarf2_per_cu_data> result (new dwarf2_per_cu_data);
+  dwarf2_per_cu_data_up result (new dwarf2_per_cu_data);
   result->per_bfd = this;
   result->index = m_num_psymtabs++;
   return result;
@@ -2546,13 +2557,13 @@ dwarf2_per_bfd::allocate_signatured_type ()
 /* Return a new dwarf2_per_cu_data allocated on the per-bfd
    obstack, and constructed with the specified field values.  */
 
-static std::unique_ptr<dwarf2_per_cu_data>
+static dwarf2_per_cu_data_up
 create_cu_from_index_list (dwarf2_per_bfd *per_bfd,
 			   struct dwarf2_section_info *section,
 			   int is_dwz,
 			   sect_offset sect_off, ULONGEST length)
 {
-  std::unique_ptr<dwarf2_per_cu_data> the_cu = per_bfd->allocate_per_cu ();
+  dwarf2_per_cu_data_up the_cu = per_bfd->allocate_per_cu ();
   the_cu->sect_off = sect_off;
   the_cu->length = length;
   the_cu->section = section;
@@ -2580,7 +2591,7 @@ create_cus_from_index_list (dwarf2_per_bfd *per_bfd,
       ULONGEST length = extract_unsigned_integer (cu_list + 8, 8, BFD_ENDIAN_LITTLE);
       cu_list += 2 * 8;
 
-      std::unique_ptr<dwarf2_per_cu_data> per_cu
+      dwarf2_per_cu_data_up per_cu
 	= create_cu_from_index_list (per_bfd, section, is_dwz, sect_off,
 				     length);
       per_bfd->all_comp_units.push_back (std::move (per_cu));
@@ -2647,7 +2658,7 @@ create_signatured_type_table_from_index
       slot = htab_find_slot (sig_types_hash.get (), sig_type.get (), INSERT);
       *slot = sig_type.get ();
 
-      per_bfd->all_comp_units.push_back (std::move (sig_type));
+      per_bfd->all_comp_units.emplace_back (sig_type.release ());
     }
 
   per_bfd->signatured_types = std::move (sig_types_hash);
@@ -2699,7 +2710,7 @@ create_signatured_type_table_from_debug_names
       slot = htab_find_slot (sig_types_hash.get (), sig_type.get (), INSERT);
       *slot = sig_type.get ();
 
-      per_objfile->per_bfd->all_comp_units.push_back (std::move (sig_type));
+      per_objfile->per_bfd->all_comp_units.emplace_back (sig_type.release ());
     }
 
   per_objfile->per_bfd->signatured_types = std::move (sig_types_hash);
@@ -4930,7 +4941,7 @@ create_cus_from_debug_names_list (dwarf2_per_bfd *per_bfd,
 	     of the next CU as end of this CU.  We create the CUs here with
 	     length 0, and in cutu_reader::cutu_reader we'll fill in the
 	     actual length.  */
-	  std::unique_ptr<dwarf2_per_cu_data> per_cu
+	  dwarf2_per_cu_data_up per_cu
 	    = create_cu_from_index_list (per_bfd, &section, is_dwz,
 					 sect_off, 0);
 	  per_bfd->all_comp_units.push_back (std::move (per_cu));
@@ -4955,7 +4966,7 @@ create_cus_from_debug_names_list (dwarf2_per_bfd *per_bfd,
       if (i >= 1)
 	{
 	  const ULONGEST length = sect_off_next - sect_off_prev;
-	  std::unique_ptr<dwarf2_per_cu_data> per_cu
+	  dwarf2_per_cu_data_up per_cu
 	    = create_cu_from_index_list (per_bfd, &section, is_dwz,
 					 sect_off_prev, length);
 	  per_bfd->all_comp_units.push_back (std::move (per_cu));
@@ -6132,7 +6143,8 @@ add_type_unit (dwarf2_per_objfile *per_objfile, ULONGEST sig, void **slot)
 
   per_objfile->resize_symtabs ();
 
-  per_objfile->per_bfd->all_comp_units.push_back (std::move (sig_type_holder));
+  per_objfile->per_bfd->all_comp_units.emplace_back
+    (sig_type_holder.release ());
   sig_type->signature = sig;
   sig_type->is_debug_types = 1;
   if (per_objfile->per_bfd->using_index)
@@ -7711,7 +7723,7 @@ read_comp_units_from_section (dwarf2_per_objfile *per_objfile,
 
   while (info_ptr < section->buffer + section->size)
     {
-      std::unique_ptr<dwarf2_per_cu_data> this_cu;
+      dwarf2_per_cu_data_up this_cu;
 
       sect_offset sect_off = (sect_offset) (info_ptr - section->buffer);
 
@@ -7732,7 +7744,7 @@ read_comp_units_from_section (dwarf2_per_objfile *per_objfile,
 	  signatured_type *sig_ptr = sig_type.get ();
 	  sig_type->signature = cu_header.signature;
 	  sig_type->type_offset_in_tu = cu_header.type_cu_offset_in_tu;
-	  this_cu = std::move (sig_type);
+	  this_cu.reset (sig_type.release ());
 
 	  void **slot = htab_find_slot (types_htab.get (), sig_ptr, INSERT);
 	  gdb_assert (slot != nullptr);
@@ -24581,7 +24593,7 @@ static int
 dwarf2_find_containing_comp_unit
   (sect_offset sect_off,
    unsigned int offset_in_dwz,
-   const std::vector<std::unique_ptr<dwarf2_per_cu_data>> &all_comp_units)
+   const std::vector<dwarf2_per_cu_data_up> &all_comp_units)
 {
   int low, high;
 
@@ -24647,13 +24659,13 @@ namespace find_containing_comp_unit {
 static void
 run_test ()
 {
-  std::unique_ptr<dwarf2_per_cu_data> one (new dwarf2_per_cu_data);
+  dwarf2_per_cu_data_up one (new dwarf2_per_cu_data);
   dwarf2_per_cu_data *one_ptr = one.get ();
-  std::unique_ptr<dwarf2_per_cu_data> two (new dwarf2_per_cu_data);
+  dwarf2_per_cu_data_up two (new dwarf2_per_cu_data);
   dwarf2_per_cu_data *two_ptr = two.get ();
-  std::unique_ptr<dwarf2_per_cu_data> three (new dwarf2_per_cu_data);
+  dwarf2_per_cu_data_up three (new dwarf2_per_cu_data);
   dwarf2_per_cu_data *three_ptr = three.get ();
-  std::unique_ptr<dwarf2_per_cu_data> four (new dwarf2_per_cu_data);
+  dwarf2_per_cu_data_up four (new dwarf2_per_cu_data);
   dwarf2_per_cu_data *four_ptr = four.get ();
 
   one->length = 5;
@@ -24666,7 +24678,7 @@ run_test ()
   four->length = 7;
   four->is_dwz = 1;
 
-  std::vector<std::unique_ptr<dwarf2_per_cu_data>> units;
+  std::vector<dwarf2_per_cu_data_up> units;
   units.push_back (std::move (one));
   units.push_back (std::move (two));
   units.push_back (std::move (three));
diff --git a/gdb/dwarf2/read.h b/gdb/dwarf2/read.h
index 80f0e3ee13b..756d1934ca0 100644
--- a/gdb/dwarf2/read.h
+++ b/gdb/dwarf2/read.h
@@ -74,6 +74,20 @@ struct dwarf2_queue_item
   enum language pretend_language;
 };
 
+/* A deleter for dwarf2_per_cu_data that knows to downcast to
+   signatured_type as appropriate.  This approach lets us avoid a
+   virtual destructor, which saves a bit of space.  */
+
+struct dwarf2_per_cu_data_deleter
+{
+  void operator() (dwarf2_per_cu_data *data);
+};
+
+/* A specialization of unique_ptr for dwarf2_per_cu_data and
+   subclasses.  */
+typedef std::unique_ptr<dwarf2_per_cu_data, dwarf2_per_cu_data_deleter>
+    dwarf2_per_cu_data_up;
+
 /* Some DWARF data can be shared across objfiles who share the same BFD,
    this data is stored in this object.
 
@@ -103,7 +117,7 @@ struct dwarf2_per_bfd
   /* A convenience function to allocate a dwarf2_per_cu_data.  The
      returned object has its "index" field set properly.  The object
      is allocated on the dwarf2_per_bfd obstack.  */
-  std::unique_ptr<dwarf2_per_cu_data> allocate_per_cu ();
+  dwarf2_per_cu_data_up allocate_per_cu ();
 
   /* A convenience function to allocate a signatured_type.  The
      returned object has its "index" field set properly.  The object
@@ -154,7 +168,7 @@ struct dwarf2_per_bfd
 
   /* Table of all the compilation units.  This is used to locate
      the target compilation unit of a particular reference.  */
-  std::vector<std::unique_ptr<dwarf2_per_cu_data>> all_comp_units;
+  std::vector<dwarf2_per_cu_data_up> all_comp_units;
 
   /* Table of struct type_unit_group objects.
      The hash key is the DW_AT_stmt_list value.  */

Re: [PATCH 0/3] Fix some sanitizer errors

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tromey@adacore.com> writes:

Tom> I tried a build with the address sanitizer, and another with the
Tom> undefined behavior sanitizer.

Tom> This series fixes the simplest bugs I found.

I'm checking these in now.

Tom