public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: "Achra, Nitika" <Nitika.Achra@amd.com>
To: Tom Tromey <tom@tromey.com>
Cc: "gdb-patches@sourceware.org" <gdb-patches@sourceware.org>,
	"George, Jini Susan" <JiniSusan.George@amd.com>,
	Ali Tamur <tamur@google.com>
Subject: RE: [PATCH] Support for DW_AT_loclists_base and DW_FORM_loclistx.
Date: Fri, 31 Jan 2020 08:59:00 -0000	[thread overview]
Message-ID: <CH2PR12MB3733AFBDED08D5F7FA6A3E0C9A070@CH2PR12MB3733.namprd12.prod.outlook.com> (raw)
In-Reply-To: <87tv4lreki.fsf@tromey.com>

[-- Attachment #1: Type: text/plain, Size: 5583 bytes --]

[AMD Official Use Only - Internal Distribution Only]

Hello Tom,

Thanks for the review. I have incorporated the review comments. Please have a look.

Nitika> *Support for DW_AT_loclists_base and DW_FORM_loclistx.

Tom>Thanks for the patch.

Tom>My comments below are primarily nits.

Nitika> Tested by running the testsuite before and after the patch and 
Nitika> there is no increase in the number of test cases that fails. 
Nitika> Tested with both
Nitika> -gdwarf-4 and -gdwarf-5 flags. Also tested -gslit-dwarf along 
Nitika> with
Nitika> -gdwarf-4 as well as -gdwarf5 flags.

Tom> I assume it fixed some tests with -gdwarf-5?  Or else we'll need a new test case.

Added a new test case. This test checks if the file command passes without any error with -g-dwarf-5 and -gsplit-dwarf. 
Gcc emits DW_FORM_loclistx only with -gsplit-dwarf. I can also check by printing the variable value, but right now printing
the value is throwing  the error- " DWARF ERROR: Corrupted dwarf expression"  which is not related to this patch.  I have submitted 
another patch which will fix that.

Nitika> +/* Size of .debug_loclist section header for 32-bit DWARF 
Nitika> +format. */ #define LOCLIST_HEADER_SIZE32 12;
Nitika> +
Nitika> +/* Size of .debug_loclist section header for 64-bit DWARF 
Nitika> +format. */ #define LOCLIST_HEADER_SIZE64 20;

Tom> The ";" at the end of these defines is weird.
Removed

Nitika> +  /* Header data from the location list section. */  struct 
Nitika> + loclist_header* loclist_header;

Tom> gdb style puts the "*" on the other side of the " " like

Tom> struct loclist_header *loclist_header;

Nitika> +static struct dwarf2_section_info* cu_debug_loc_section (struct 
Nitika> +dwarf2_cu* cu);

Tom> Here too -- there are actually several instances in the patch.

Done

Nitika> +void
Nitika> +read_loclist_header (struct dwarf2_cu* cu, struct 
Nitika> +dwarf2_section_info* section) {

Tom> New functions should have an introductory comment explaining their purpose.
Done

Tom> The "static" should be repeated here, rather than rely on the declaration.  This affects all of the new functions, I think.  Also, there's no need to forward Tom>declare them if the uses come after the definition, so probably some of those declarations can be removed as well.
Tom>Added static in definitions also. 

Done

Tom>It might be nicer if read_loclist_header took a pointer to the loclist_header rather than a dwarf2_cu, and did not allocate.  See below.

Done

Nitika> +ULONGEST
Nitika> +lookup_loclist_base (struct dwarf2_cu* cu) {
Nitika> +  /* For the .dwo unit, the loclist_base points to the first offset following
Nitika> +     the header. The header consists of the following entities-
Nitika> +     1. Unit Length (4 bytes for 32 bit DWARF format, and 12 bytes for the 64 bit format)
Nitika> +     2. version (2 bytes)
Nitika> +     3. address size (1 byte)
Nitika> +     4. segment selector size (1 byte)
Nitika> +     5. offset entry count (4 bytes)
Nitika> +     These sizes are derived as per the DWARFv5 standard. */
Nitika> +  if (cu->dwo_unit)
Nitika> +  {
Nitika> +    if (cu->header.initial_length_size == 4)
Nitika> +      return LOCLIST_HEADER_SIZE32;
Nitika> +    return LOCLIST_HEADER_SIZE64;

Tom>Is there some way to avoid hard-coding sizes here?

I thought of using sizeof(struct loclist_header) in place of LOCLIST_HEADER_SIZE32 and sizeof(struct loclist_header) + cu->initial_length_size in place of 
LOCLIST_HEADER_SIZE_64. But I am not sure if this a correct way of doing this. Some compilers append some padding at the end of structure. So the
size of structure might not be equal to the sum of size of its members. Sizeof() is also compiler dependent. So, right now I cannot think of any other way.

Nitika> +/* Given a DW_FORM_loclistx value loclist_index, fetch the offset from the array
Nitika> +   of offsets in the .debug_loclists section. */ CORE_ADDR 
Nitika> +read_loclist_index (struct dwarf2_cu* cu, ULONGEST 
Nitika> +loclist_index) {
...
Nitika> +  const gdb_byte* info_ptr;

Tom> This can be declared later, when it's first initialized.
Done

Nitika> +  if (section->buffer == NULL)
Nitika> +    error(_("DW_FORM_loclistx used without .debug_loclist section [in module %s]"),
Nitika> +       objfile_name (objfile));

Tom>I wonder whether errors here will really do something good.  The problem is that the DWARF reader, in general, doesn't handle errors very well.
Tom>It *should* -- but it doesn't.  I don't know about this spot, but in other places, calling error will mean that reading all of the debuginfo for the entire file Tom>will be aborted.  (It can even cause worse problems, there's a bug in bugzilla about it ending a remote session.)

Tom>Maybe complaint() and then a fallback would be preferable.
Tom>Or, test the error() to make sure it's ok.

Replaced with complaint()

Nitika> +  delete cu->loclist_header;

Tom>This is created, only to be deleted in the same function.
Tom>I think it would be better to just stack-allocate this.

Done

Tom>Or, should this be cached in the CU -- that is, read once and then reused?
Tom>If so then a different approach should be used.  It wasn't clear to me how often read_loclist_index is called.

Nitika> +    case DW_FORM_loclistx:
Nitika> +    {
Nitika> +      *need_reprocess = true;
Nitika> +      DW_UNSND(attr) = read_unsigned_leb128 (abfd, info_ptr, 
Nitika> + &bytes_read);

Tom>Space before the first "(".
Done

Regards,
Nitika Achra

[-- Attachment #2: 0001-Support-for-DW_AT_loclists_base-and-DW_FORM_loclistx.patch --]
[-- Type: application/octet-stream, Size: 12495 bytes --]

From e643d4f59486790f00fc495f244dab970402fb56 Mon Sep 17 00:00:00 2001
From: nitachra <Nitika.Achra@amd.com>
Date: Mon, 20 Jan 2020 14:07:29 +0530
Subject: [PATCH] Support for DW_AT_loclists_base and DW_FORM_loclistx.

This patch handles DW_AT_loclists_base and DW_FORM_loclistx.
DW_AT_loclists_base is a new attribute added in DWARFv5 which
points to the beginning of the offset table of .debug_loclist
section. Reference to the location list (DW_FORM_loclistx) is
interpreted relative to this base. DW_FORM_loclistx is a new
form added in DWARFv5 which is used to access location list.

Tested by running the testsuite before and after the patch and there
is no increase in the number of test cases that fails. Tested with both
-gdwarf-4 and -gdwarf-5 flags. Also tested -gslit-dwarf along with
-gdwarf-4 as well as -gdwarf5 flags. Used gcc-9.4 for testing.

gdb/ChangeLog:

   *dwarf2read.c (cu_debug_loc_section): Added the declaration for the function.
   (read_loclist_index): New function declaration.
   (lookup_loclist_base): New function declaration.
   (read_loclist_header): New function declaration
   (dwarf2_cu): Added loclist_base and loclist_header field.
   (dwarf2_locate_dwo_sections): Handle .debug_loclist.dwo section.
   (read_full_die_1): Read the value of DW_AT_loclists_base.
   (read_attribute_reprocess): Handle DW_FORM_loclistx.
   (read_attribute_value): Handle DW_FORM_loclistx.
   (skip_one_die): Handle DW_FORM_loclistx.
   (attr_form_is_section_offset): Handle DW_FORM_loclistx.
   (read_loclist_index): Function definition.
   (lookup_loclist_base): Function definition.
   (read_loclist_header): Function definition.
   (loclist_header): New structure declaration.

gdb/testsuite/ChangeLog:

   *gdb.dwarf2/dw5-form-loclistx.exp: New file.
   *gdb.dwarf2/dw5-form-loclistx.c: New file.
---
 gdb/dwarf2read.c                              | 130 +++++++++++++++++-
 gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.c  |  16 +++
 .../gdb.dwarf2/dw5-form-loclistx.exp          |  42 ++++++
 3 files changed, 187 insertions(+), 1 deletion(-)
 create mode 100644 gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.c
 create mode 100644 gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.exp

diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
index dfa2f91d45..f1096570fc 100644
--- a/gdb/dwarf2read.c
+++ b/gdb/dwarf2read.c
@@ -104,6 +104,12 @@ static int dwarf2_loclist_index;
 static int dwarf2_locexpr_block_index;
 static int dwarf2_loclist_block_index;
 
+/* Size of .debug_loclist section header for 32-bit DWARF format. */
+#define LOCLIST_HEADER_SIZE32 12
+
+/* Size of .debug_loclist section header for 64-bit DWARF format. */
+#define LOCLIST_HEADER_SIZE64 20
+
 /* An index into a (C++) symbol name component in a symbol name as
    recorded in the mapped_index's symbol table.  For each C++ symbol
    in the symbol table, we record one entry for the start of each
@@ -373,6 +379,30 @@ struct comp_unit_head
   cu_offset type_cu_offset_in_tu;
 };
 
+/* The location list section (.debug_loclist) begins with a header,
+   which contains the following information. */
+struct loclist_header
+{
+  /* A 4-byte or 12-byte length containing the length of the
+  set of entries for this compilation unit, not including the
+  length field itself. */
+  unsigned int length;
+
+  /* A 2-byte version identifier. */
+  short version;
+
+  /* A 1-byte unsigned integer containing the size in bytes of an address on
+     the target system. */
+  unsigned char addr_size;
+
+  /* A 1-byte unsigned integer containing the size in bytes of a segment selector
+     on the target system. */
+  unsigned char segment_collector_size;
+
+  /* A 4-byte count of the number of offsets that follow the header. */
+  unsigned int offset_entry_count;
+};
+
 /* Type used for delaying computation of method physnames.
    See comments for compute_delayed_physnames.  */
 struct delayed_method_info
@@ -523,6 +553,9 @@ public:
      whether the DW_AT_ranges attribute came from the skeleton or DWO.  */
   ULONGEST ranges_base = 0;
 
+  /* The DW_AT_loclists_base attribute if present. */
+  gdb::optional<ULONGEST> loclist_base;
+
   /* When reading debug info generated by older versions of rustc, we
      have to rewrite some union types to be struct types with a
      variant part.  This rewriting must be done after the CU is fully
@@ -1705,6 +1738,9 @@ static void read_variable (struct die_info *die, struct dwarf2_cu *cu);
 static int dwarf2_ranges_read (unsigned, CORE_ADDR *, CORE_ADDR *,
 			       struct dwarf2_cu *, struct partial_symtab *);
 
+/* Return the .debug_loclist section to use for cu. */
+static struct dwarf2_section_info *cu_debug_loc_section (struct dwarf2_cu *cu);
+
 /* How dwarf2_get_pc_bounds constructed its *LOWPC and *HIGHPC return
    values.  Keep the items ordered with increasing constraints compliance.  */
 enum pc_bounds_kind
@@ -9445,6 +9481,7 @@ skip_one_die (const struct die_reader_specs *reader, const gdb_byte *info_ptr,
 	case DW_FORM_GNU_addr_index:
 	case DW_FORM_GNU_str_index:
 	case DW_FORM_rnglistx:
+	case DW_FORM_loclistx:
 	  info_ptr = safe_skip_leb128 (info_ptr, buffer_end);
 	  break;
 	case DW_FORM_indirect:
@@ -12979,6 +13016,11 @@ dwarf2_locate_dwo_sections (bfd *abfd, asection *sectp, void *dwo_sections_ptr)
       dwo_sections->loc.s.section = sectp;
       dwo_sections->loc.size = bfd_section_size (sectp);
     }
+  else if (section_is_p (sectp->name, &names->loclists_dwo))
+    {
+      dwo_sections->loclists.s.section = sectp;
+      dwo_sections->loclists.size = bfd_section_size (sectp);
+    }
   else if (section_is_p (sectp->name, &names->macinfo_dwo))
     {
       dwo_sections->macinfo.s.section = sectp;
@@ -18453,6 +18495,9 @@ read_full_die_1 (const struct die_reader_specs *reader,
   struct attribute *attr = dwarf2_attr_no_follow (die, DW_AT_str_offsets_base);
   if (attr != nullptr)
     cu->str_offsets_base = DW_UNSND (attr);
+  attr = dwarf2_attr_no_follow (die, DW_AT_loclists_base);
+  if (attr)
+    cu->loclist_base = DW_UNSND (attr);
 
   auto maybe_addr_base = lookup_addr_base(die);
   if (maybe_addr_base.has_value ())
@@ -19409,6 +19454,78 @@ partial_die_info::fixup (struct dwarf2_cu *cu)
   fixup_called = 1;
 }
 
+static void
+read_loclist_header (struct loclist_header *header, struct dwarf2_section_info *section)
+{
+  unsigned int bytes_read;
+  bfd *abfd = get_section_bfd_owner (section);
+  const gdb_byte *info_ptr = section->buffer;
+  header->length = read_initial_length (abfd, info_ptr, &bytes_read);
+  info_ptr += bytes_read;
+  header->version = read_2_bytes (abfd, info_ptr);
+  info_ptr += 2;
+  header->addr_size = read_1_byte (abfd, info_ptr);
+  info_ptr += 1;
+  header->segment_collector_size = read_1_byte (abfd, info_ptr);
+  info_ptr += 1;
+  header->offset_entry_count = read_4_bytes (abfd, info_ptr);
+}
+
+
+static ULONGEST
+lookup_loclist_base (struct dwarf2_cu *cu)
+{
+  /* For the .dwo unit, the loclist_base points to the first offset following
+     the header. The header consists of the following entities-
+     1. Unit Length (4 bytes for 32 bit DWARF format, and 12 bytes for the 64 bit format)
+     2. version (2 bytes)
+     3. address size (1 byte)
+     4. segment selector size (1 byte)
+     5. offset entry count (4 bytes)
+     These sizes are derived as per the DWARFv5 standard. */
+  if (cu->dwo_unit)
+  {
+    if (cu->header.initial_length_size == 4)
+      return LOCLIST_HEADER_SIZE32;
+    return LOCLIST_HEADER_SIZE64;
+  }
+  return *cu->loclist_base;
+}
+
+/* Given a DW_FORM_loclistx value loclist_index, fetch the offset from the array
+   of offsets in the .debug_loclists section. */
+static CORE_ADDR
+read_loclist_index (struct dwarf2_cu *cu, ULONGEST loclist_index)
+{
+  struct dwarf2_per_objfile *dwarf2_per_objfile
+	= cu->per_cu->dwarf2_per_objfile;
+  struct objfile *objfile = dwarf2_per_objfile->objfile;
+  bfd *abfd = objfile->obfd;
+  ULONGEST loclist_base = lookup_loclist_base (cu);
+  struct dwarf2_section_info *section = cu_debug_loc_section (cu);
+  dwarf2_read_section (objfile, section);
+  if (section->buffer == NULL)
+    complaint(_("DW_FORM_loclistx used without .debug_loclist section [in module %s]"),
+	objfile_name (objfile));
+  struct loclist_header header;
+  read_loclist_header (&header, section);
+  if (loclist_index >= header.offset_entry_count)
+    complaint(_("DW_FORM_loclistx pointing outside of "
+	".debug_loclist offset array [in module %s]"),
+	objfile_name (objfile));
+  if (loclist_base + loclist_index * cu->header.offset_size
+	>= section->size)
+    complaint(_("DW_FORM_loclistx pointing outside of "
+	".debug_loclist section [in module %s]"),
+	objfile_name (objfile));
+  const gdb_byte *info_ptr = (section->buffer + loclist_base +
+	loclist_index * cu->header.offset_size);
+  if (cu->header.offset_size == 4)
+    return bfd_get_32 (abfd, info_ptr) + loclist_base;
+  else
+    return bfd_get_64 (abfd, info_ptr) + loclist_base;
+}
+
 /* Process the attributes that had to be skipped in the first round. These
    attributes are the ones that need str_offsets_base or addr_base attributes.
    They could not have been processed in the first round, because at the time
@@ -19423,6 +19540,9 @@ void read_attribute_reprocess (const struct die_reader_specs *reader,
       case DW_FORM_GNU_addr_index:
         DW_ADDR (attr) = read_addr_index (cu, DW_UNSND (attr));
         break;
+      case DW_FORM_loclistx:
+	 DW_UNSND (attr) = read_loclist_index (cu, DW_UNSND (attr));
+	 break;
       case DW_FORM_strx:
       case DW_FORM_strx1:
       case DW_FORM_strx2:
@@ -19526,6 +19646,13 @@ read_attribute_value (const struct die_reader_specs *reader,
       DW_UNSND (attr) = read_offset (abfd, info_ptr, &cu->header, &bytes_read);
       info_ptr += bytes_read;
       break;
+    case DW_FORM_loclistx:
+    {
+      *need_reprocess = true;
+      DW_UNSND (attr) = read_unsigned_leb128 (abfd, info_ptr, &bytes_read);
+      info_ptr += bytes_read;
+    }
+      break;
     case DW_FORM_string:
       DW_STRING (attr) = read_direct_string (abfd, info_ptr, &bytes_read);
       DW_STRING_IS_CANONICAL (attr) = 0;
@@ -25407,7 +25534,8 @@ attr_form_is_section_offset (const struct attribute *attr)
 {
   return (attr->form == DW_FORM_data4
           || attr->form == DW_FORM_data8
-	  || attr->form == DW_FORM_sec_offset);
+	  || attr->form == DW_FORM_sec_offset
+	  || attr->form == DW_FORM_loclistx);
 }
 
 /* Return non-zero if ATTR's value falls in the 'constant' class, or
diff --git a/gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.c b/gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.c
new file mode 100644
index 0000000000..a2a14583be
--- /dev/null
+++ b/gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.c
@@ -0,0 +1,16 @@
+
+
+int foo(int a, int b)
+{
+  a = a+b;
+  b = a-b;
+  a = 2*b;
+  return a+b;
+}
+
+int main()
+{
+  int result,a,b;
+  result = foo(a,b);
+  return result;
+}
diff --git a/gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.exp b/gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.exp
new file mode 100644
index 0000000000..15d9ac13e7
--- /dev/null
+++ b/gdb/testsuite/gdb.dwarf2/dw5-form-loclistx.exp
@@ -0,0 +1,42 @@
+# Copyright 2012-2020 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+load_lib dwarf.exp
+
+# This test can only be run on targets which support DWARF-2 and use gas.
+if {![dwarf2_support]} {
+    return 0  
+}
+
+standard_testfile .c
+
+
+# We can't use prepare_for_testing here because we need to check the
+# 'file' command's output.
+
+if { [build_executable ${testfile}.exp ${testfile} ${srcfile}\
+   {additional_flags=-gdwarf-5 additional_flags=-gsplit-dwarf additional_flags=-O2}] == -1 } {
+    return -1
+}
+
+gdb_exit
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+
+# Check that file command gives no error.
+gdb_test "file $binfile" \
+    "Reading symbols from $binfile\.\.\."
+
+
-- 
2.17.1


  reply	other threads:[~2020-01-31  8:55 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-20 10:36 Achra, Nitika
2020-01-23 22:34 ` Tom Tromey
2020-01-31  8:59   ` Achra, Nitika [this message]
2020-02-02  3:10     ` Ali Tamur via gdb-patches
2020-02-24  8:48       ` Achra, Nitika
2020-03-17 14:13         ` Achra, Nitika
2020-03-18 14:09           ` Achra, Nitika

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=CH2PR12MB3733AFBDED08D5F7FA6A3E0C9A070@CH2PR12MB3733.namprd12.prod.outlook.com \
    --to=nitika.achra@amd.com \
    --cc=JiniSusan.George@amd.com \
    --cc=gdb-patches@sourceware.org \
    --cc=tamur@google.com \
    --cc=tom@tromey.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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).