* [delayed-symfile] handle incorrect aranges
@ 2009-08-13 19:52 Tom Tromey
2009-08-14 12:28 ` Jan Kratochvil
0 siblings, 1 reply; 4+ messages in thread
From: Tom Tromey @ 2009-08-13 19:52 UTC (permalink / raw)
To: Project Archer
I'm checking this in on the delayed-symfile branch.
Jan, I think you probably ought to merge this.
It turns out that older versions of GCC could generate a bogus
.debug_aranges section. E.g., I have an oldish library that has:
Contents of the .debug_aranges section:
Length: 1516
Version: 2
Offset into .debug_info: 0x0
Pointer Size: 8
Segment Size: 0
Address Length
0000000000000000 0000000000000000
000000000004f730 0000000000000008
000000000004f740 0000000000000008
[...]
This is bogus because a (0,0) entry marks the end of the section.
The bug is that a section like this would cause gdb to hang.
This patch adds some more sanity checking while we read .debug_aranges.
If anything is amiss, we just discard what we've read and return. This
is safe, it just means we're falling back to the old, slow path.
Tom
2009-08-13 Tom Tromey <tromey@redhat.com>
* dwarf2read.c (dwarf2_create_quick_addrmap): Sanity-check aranges
section.
diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
index 69328f4..2272259 100644
--- a/gdb/dwarf2read.c
+++ b/gdb/dwarf2read.c
@@ -1421,9 +1421,21 @@ dwarf2_create_quick_addrmap (struct objfile *objfile)
unsigned int bytes_read, segment_size, delta;
LONGEST info_offset;
struct dwarf2_cu cu;
+ char *end_ptr;
cu_header.initial_length_size = 0;
+ end_ptr = aranges_ptr;
aranges_ptr = read_comp_unit_head (&cu_header, aranges_ptr, abfd);
+ end_ptr += cu_header.initial_length_size + cu_header.length;
+
+ /* Sanity check. */
+ if (end_ptr - aranges_ptr >= dwarf2_per_objfile->aranges.size)
+ {
+ do_cleanups (old);
+ complaint (&symfile_complaints,
+ _("aranges entry runs off end of `.debug_aranges' section, ignored"));
+ return;
+ }
segment_size = read_1_byte (abfd, aranges_ptr);
aranges_ptr += 1;
@@ -1454,6 +1466,18 @@ dwarf2_create_quick_addrmap (struct objfile *objfile)
addrmap_set_empty (mutable_map, address, address + length, objfile);
}
+
+ /* Some older versions of GCC incorrectly started the arange
+ with a (0,0) pair. If we encounter any oddity while reading
+ the section, just abandon the attempt; falling back to the
+ slower code is always safe. */
+ if (aranges_ptr != end_ptr)
+ {
+ do_cleanups (old);
+ complaint (&symfile_complaints,
+ _("aranges entry ends early, ignored"));
+ return;
+ }
}
objfile->quick_addrmap = addrmap_create_fixed (mutable_map,
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [delayed-symfile] handle incorrect aranges
2009-08-13 19:52 [delayed-symfile] handle incorrect aranges Tom Tromey
@ 2009-08-14 12:28 ` Jan Kratochvil
2009-08-14 17:13 ` Tom Tromey
0 siblings, 1 reply; 4+ messages in thread
From: Jan Kratochvil @ 2009-08-14 12:28 UTC (permalink / raw)
To: Tom Tromey; +Cc: Project Archer
On Thu, 13 Aug 2009 21:52:42 +0200, Tom Tromey wrote:
> Jan, I think you probably ought to merge this.
Could you please also fix the address_size == 0 problem bugreported inside RH?
Testcase attached.
readelf:
Contents of the .debug_aranges section:
Length: 8
Version: 2
Offset into .debug_info: 0x0
Pointer Size: 0
Segment Size: 0
Address Length
Floating point exception
Testcase approval for [delayed-symfile] requested.
Thanks,
Jan
--- /dev/null
+++ b/gdb/testsuite/gdb.dwarf2/dw2-aranges.S
@@ -0,0 +1,140 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2004, 2007, 2008, 2009 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/>. */
+
+/* Test .debug_aranges containing zero address_size. */
+
+/* Dummy function to provide debug information for. */
+
+ .text
+.Lbegin_text1:
+ .globl main
+ .type main, %function
+main:
+.Lbegin_main:
+ .int 0
+.Lend_main:
+ .size main, .-main
+.Lend_text1:
+
+/* Debug information */
+
+ .section .debug_info
+.Lcu1_begin:
+ /* CU header */
+ .4byte .Lcu1_end - .Lcu1_start /* Length of Compilation Unit */
+.Lcu1_start:
+ .2byte 2 /* DWARF Version */
+ .4byte .Labbrev1_begin /* Offset into abbrev section */
+ .byte 4 /* Pointer size */
+
+ /* CU die */
+ .uleb128 1 /* Abbrev: DW_TAG_compile_unit */
+ .4byte .Lend_text1 /* DW_AT_high_pc */
+ .4byte .Lbegin_text1 /* DW_AT_low_pc */
+ .ascii "file1.txt\0" /* DW_AT_name */
+ .ascii "GNU C 3.3.3\0" /* DW_AT_producer */
+ .byte 1 /* DW_AT_language (C) */
+
+ /* main */
+ .uleb128 2 /* Abbrev: DW_TAG_subprogram */
+ .byte 1 /* DW_AT_external */
+ .byte 1 /* DW_AT_decl_file */
+ .byte 2 /* DW_AT_decl_line */
+ .ascii "main\0" /* DW_AT_name */
+ .4byte .Ltype_int-.Lcu1_begin /* DW_AT_type */
+ .4byte .Lbegin_main /* DW_AT_low_pc */
+ .4byte .Lend_main /* DW_AT_high_pc */
+ .byte 1 /* DW_AT_frame_base: length */
+ .byte 0x55 /* DW_AT_frame_base: DW_OP_reg5 */
+
+.Ltype_int:
+ .uleb128 3 /* Abbrev: DW_TAG_base_type */
+ .ascii "int\0" /* DW_AT_name */
+ .byte 4 /* DW_AT_byte_size */
+ .byte 5 /* DW_AT_encoding */
+
+ .byte 0 /* End of children of CU */
+
+.Lcu1_end:
+
+/* Abbrev table */
+ .section .debug_abbrev
+.Labbrev1_begin:
+ .uleb128 1 /* Abbrev code */
+ .uleb128 0x11 /* DW_TAG_compile_unit */
+ .byte 1 /* has_children */
+ .uleb128 0x12 /* DW_AT_high_pc */
+ .uleb128 0x1 /* DW_FORM_addr */
+ .uleb128 0x11 /* DW_AT_low_pc */
+ .uleb128 0x1 /* DW_FORM_addr */
+ .uleb128 0x3 /* DW_AT_name */
+ .uleb128 0x8 /* DW_FORM_string */
+ .uleb128 0x25 /* DW_AT_producer */
+ .uleb128 0x8 /* DW_FORM_string */
+ .uleb128 0x13 /* DW_AT_language */
+ .uleb128 0xb /* DW_FORM_data1 */
+ .byte 0x0 /* Terminator */
+ .byte 0x0 /* Terminator */
+
+ .uleb128 2 /* Abbrev code */
+ .uleb128 0x2e /* DW_TAG_subprogram */
+ .byte 0 /* has_children */
+ .uleb128 0x3f /* DW_AT_external */
+ .uleb128 0xc /* DW_FORM_flag */
+ .uleb128 0x3a /* DW_AT_decl_file */
+ .uleb128 0xb /* DW_FORM_data1 */
+ .uleb128 0x3b /* DW_AT_decl_line */
+ .uleb128 0xb /* DW_FORM_data1 */
+ .uleb128 0x3 /* DW_AT_name */
+ .uleb128 0x8 /* DW_FORM_string */
+ .uleb128 0x49 /* DW_AT_type */
+ .uleb128 0x13 /* DW_FORM_ref4 */
+ .uleb128 0x11 /* DW_AT_low_pc */
+ .uleb128 0x1 /* DW_FORM_addr */
+ .uleb128 0x12 /* DW_AT_high_pc */
+ .uleb128 0x1 /* DW_FORM_addr */
+ .uleb128 0x40 /* DW_AT_frame_base */
+ .uleb128 0xa /* DW_FORM_block1 */
+ .byte 0x0 /* Terminator */
+ .byte 0x0 /* Terminator */
+
+ .uleb128 3 /* Abbrev code */
+ .uleb128 0x24 /* DW_TAG_base_type */
+ .byte 0 /* has_children */
+ .uleb128 0x3 /* DW_AT_name */
+ .uleb128 0x8 /* DW_FORM_string */
+ .uleb128 0xb /* DW_AT_byte_size */
+ .uleb128 0xb /* DW_FORM_data1 */
+ .uleb128 0x3e /* DW_AT_encoding */
+ .uleb128 0xb /* DW_FORM_data1 */
+ .byte 0x0 /* Terminator */
+ .byte 0x0 /* Terminator */
+
+ .byte 0x0 /* Terminator */
+ .byte 0x0 /* Terminator */
+
+/* aranges table */
+ .section .debug_aranges
+ .long .Laranges_end - 1f
+1:
+ .2byte 2 /* aranges Version */
+ .4byte .Lcu1_begin - .debug_info /* Offset into .debug_info section */
+ /* The GDB crasher is this zero value. */
+ .byte 0 /* aranges address_size */
+ .byte 0 /* aranges segment_size */
+
+.Laranges_end:
--- /dev/null
+++ b/gdb/testsuite/gdb.dwarf2/dw2-aranges.exp
@@ -0,0 +1,40 @@
+# Copyright 2004, 2005, 2007, 2008, 2009 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/>.
+
+# Test .debug_aranges containing zero address_size.
+
+# This test can only be run on targets which support DWARF-2 and use gas.
+# For now pick a sampling of likely targets.
+if {![istarget *-*-linux*]
+ && ![istarget *-*-gnu*]
+ && ![istarget *-*-elf*]
+ && ![istarget *-*-openbsd*]
+ && ![istarget arm-*-eabi*]
+ && ![istarget powerpc-*-eabi*]} {
+ return 0
+}
+
+set testfile "dw2-aranges"
+set srcfile ${testfile}.S
+set binfile ${objdir}/${subdir}/${testfile}
+
+if { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable {nodebug}] != "" } {
+ return -1
+}
+
+clean_restart $testfile
+
+# Failed gdb_load would abort the testcase execution earlier.
+pass "file loaded"
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [delayed-symfile] handle incorrect aranges
2009-08-14 12:28 ` Jan Kratochvil
@ 2009-08-14 17:13 ` Tom Tromey
2009-08-15 16:10 ` Jan Kratochvil
0 siblings, 1 reply; 4+ messages in thread
From: Tom Tromey @ 2009-08-14 17:13 UTC (permalink / raw)
To: Jan Kratochvil; +Cc: Project Archer
>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:
Jan> Could you please also fix the address_size == 0 problem bugreported
Jan> inside RH? Testcase attached.
Sure.
FWIW, I started with the original test case. What happened there was
that the aranges reader stopped early, then treated some other bytes
from the middle of an aranges section as a new CU header. This is how
we ended up with address_size==0. If you 'readelf' the original
binaries you'll see there is nothing wrong except that bogus (0,0)
entry.
I'm checking in the appended. It adds a check for the zero address
size, and it also adds a check that the length of an entry is non-zero,
a constraint mentioned in the DWARF standard (and which, I think,
addrmap probably does not handle well).
Finally, this patch fixes a small bug when computing the end of the
range. addrmap uses end-inclusive, so we have to subtract one.
Jan> Testcase approval for [delayed-symfile] requested.
Yes, thank you very much.
I'll wait until you push before committing, so I can test against your
test case.
Tom
2009-08-14 Tom Tromey <tromey@redhat.com>
* dwarf2read.c (dwarf2_create_quick_addrmap): Ignore zero
addr_size and entries with zero length. Properly compute end of
range.
diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
index 2272259..4bae7cb 100644
--- a/gdb/dwarf2read.c
+++ b/gdb/dwarf2read.c
@@ -1436,6 +1436,13 @@ dwarf2_create_quick_addrmap (struct objfile *objfile)
_("aranges entry runs off end of `.debug_aranges' section, ignored"));
return;
}
+ if (cu_header.addr_size == 0)
+ {
+ do_cleanups (old);
+ complaint (&symfile_complaints,
+ _("aranges entry has zero addr_size, ignored"));
+ return;
+ }
segment_size = read_1_byte (abfd, aranges_ptr);
aranges_ptr += 1;
@@ -1462,9 +1469,18 @@ dwarf2_create_quick_addrmap (struct objfile *objfile)
if (address == 0 && length == 0)
break;
+ if (length == 0)
+ {
+ do_cleanups (old);
+ complaint (&symfile_complaints,
+ _("aranges entry has zero length, ignored"));
+ return;
+ }
+
address += baseaddr;
- addrmap_set_empty (mutable_map, address, address + length, objfile);
+ addrmap_set_empty (mutable_map, address, address + length - 1,
+ objfile);
}
/* Some older versions of GCC incorrectly started the arange
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [delayed-symfile] handle incorrect aranges
2009-08-14 17:13 ` Tom Tromey
@ 2009-08-15 16:10 ` Jan Kratochvil
0 siblings, 0 replies; 4+ messages in thread
From: Jan Kratochvil @ 2009-08-15 16:10 UTC (permalink / raw)
To: Tom Tromey; +Cc: Project Archer
On Fri, 14 Aug 2009 19:13:02 +0200, Tom Tromey wrote:
> FWIW, I started with the original test case. What happened there was
> that the aranges reader stopped early, then treated some other bytes
> from the middle of an aranges section as a new CU header. This is how
> we ended up with address_size==0. If you 'readelf' the original
> binaries you'll see there is nothing wrong except that bogus (0,0)
> entry.
Oh, understood now.
> I'll wait until you push before committing, so I can test against your
> test case.
Checked the testcase in as: 77fa7778a37b0d28a7e4e5235f074a10ecf1815d
Fedora 11: https://admin.fedoraproject.org/updates/gdb-6.8.50.20090302-37.fc11
Thanks,
Jan
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2009-08-15 16:10 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-08-13 19:52 [delayed-symfile] handle incorrect aranges Tom Tromey
2009-08-14 12:28 ` Jan Kratochvil
2009-08-14 17:13 ` Tom Tromey
2009-08-15 16:10 ` Jan Kratochvil
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).