[Bug binutils/25870] New: objdump: unable to disassemble a project with large bss section
https://sourceware.org/bugzilla/show_bug.cgi?id=25870
Bug ID: 25870
Summary: objdump: unable to disassemble a project with large
bss section
Product: binutils
Version: 2.34
Status: UNCONFIRMED
Severity: normal
Priority: P2
Component: binutils
Assignee: unassigned at sourceware dot org
Reporter: zhongyunde at huawei dot com
Target Milestone: ---
base no the New released binutils 6.34, when you try objdump -SD a.out > a.dump
, you'll get an error similar with follow info, so I think the check in
function bfd_get_full_section_contents need more constraint.
error: a.out(.bss) section size (0x400038 bytes) is larger than file size
(0x184c bytes)
Reading section .bss failed because: memory exhausted
= test case ==
int garray[0x10]; /* define a large bss section */
int main ()
{
return 0;
}
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug binutils/25870] objdump: unable to disassemble a project with large bss section
https://sourceware.org/bugzilla/show_bug.cgi?id=25870 Nick Clifton changed: What|Removed |Added CC||nickc at redhat dot com --- Comment #1 from Nick Clifton --- Hi Zhongyunde, Please could you upload the a.out file that has this problem ? The code (in bfd/compress.c:bdf_get_full_section_contents) that is producing this error message already has a check for .bss sections: /* PR 24753: Sections which have no content should also be excluded as they contain no size on disk. */ && (bfd_section_flags (sec) & SEC_HAS_CONTENTS) != 0 So in theory this error should not happen. Cheers Nick -- You are receiving this mail because: You are on the CC list for the bug.
[Bug binutils/25870] objdump: unable to disassemble a project with large bss section
https://sourceware.org/bugzilla/show_bug.cgi?id=25870 --- Comment #2 from zhongyunde at huawei dot com --- Created attachment 12488 --> https://sourceware.org/bugzilla/attachment.cgi?id=12488&action=edit ./bin/armeb-linux-gnueabi-gcc main.c -o a.out -- You are receiving this mail because: You are on the CC list for the bug.
[Bug binutils/25870] objdump: unable to disassemble a project with large bss section
https://sourceware.org/bugzilla/show_bug.cgi?id=25870 Nick Clifton changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |DUPLICATE --- Comment #3 from Nick Clifton --- Hi zhondyunde, Thanks for the upload. With that I was able to reproduce the problem and also find that it has already been fixed. As a result of PR 24753. Which I actually mentioned in my previous post but forgot to check. Doh! Cheers Nick *** This bug has been marked as a duplicate of bug 24753 *** -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ld/24753] [2.33 Regression] ld: section size (0x1e50 bytes) is larger than file size (0x5a0 bytes), can not size stub section: memory exhausted
https://sourceware.org/bugzilla/show_bug.cgi?id=24753 Nick Clifton changed: What|Removed |Added CC||zhongyunde at huawei dot com --- Comment #18 from Nick Clifton --- *** Bug 25870 has been marked as a duplicate of this bug. *** -- You are receiving this mail because: You are on the CC list for the bug.
[Bug gold/25872] New: ifuncmain7.c:47: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o
https://sourceware.org/bugzilla/show_bug.cgi?id=25872 Bug ID: 25872 Summary: ifuncmain7.c:47: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o Product: binutils Version: 2.35 (HEAD) Status: NEW Severity: normal Priority: P2 Component: gold Assignee: ccoutant at gmail dot com Reporter: hjl.tools at gmail dot com CC: ian at airs dot com Target Milestone: --- Target: x86-64 Compile gold with GCC 9 plus -fcf-protection on Linux/x86-64, I got /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:47: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:55: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:57: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:61: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:63: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:66: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:68: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o /export/gnu/import/git/gitlab/x86-binutils/gold/testsuite/ifuncmain7.c:70: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o ld works on the same inputs. -- You are receiving this mail because: You are on the CC list for the bug.
[Bug gold/25872] ifuncmain7.c:47: error: relocation overflow: reference to local symbol 5 in ifuncmain7.o
https://sourceware.org/bugzilla/show_bug.cgi?id=25872
--- Comment #1 from H.J. Lu ---
template
uint64_t
Output_data_plt_x86_64_ibt::do_address_for_local(const Relobj* object,
unsigned int r_sym)
{
// Convert the PLT offset into an APLT offset.
unsigned int plt_offset = ((object->local_plt_offset(r_sym) - plt_entry_size)
/ (plt_entry_size / aplt_entry_size));
return (this->address()
+ this->aplt_offset_
+ this->regular_count() * aplt_entry_size
+ plt_offset);
}
(gdb) p r_sym
$27 = 5
(gdb) p object->local_plt_offset(r_sym)
$28 = 0
(gdb)
((object->local_plt_offset(r_sym) - plt_entry_size) is negative.
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug binutils/25870] objdump: unable to disassemble a project with large bss section
https://sourceware.org/bugzilla/show_bug.cgi?id=25870 --- Comment #4 from zhongyunde at huawei dot com --- (In reply to Nick Clifton from comment #3) > Hi zhondyunde, > > Thanks for the upload. With that I was able to reproduce the problem > and also find that it has already been fixed. As a result of PR 24753. > Which I actually mentioned in my previous post but forgot to check. > Doh! > > Cheers > Nick > > *** This bug has been marked as a duplicate of bug 24753 *** Yeah, I see. As the issue is fixed at 21 Apr 2020, and binutils 2.34 is released at 1 Feb 2020 before that day, so it is natural that it doesn't fix in such version. -- You are receiving this mail because: You are on the CC list for the bug.
[Bug binutils/25870] objdump: unable to disassemble a project with large bss section
https://sourceware.org/bugzilla/show_bug.cgi?id=25870 zhongyunde at huawei dot com changed: What|Removed |Added CC||zhongyunde at tom dot com -- You are receiving this mail because: You are on the CC list for the bug.
[Bug gold/25872] Gold doesn't properly support CET PLT
https://sourceware.org/bugzilla/show_bug.cgi?id=25872
H.J. Lu changed:
What|Removed |Added
Summary|ifuncmain7.c:47: error: |Gold doesn't properly
|relocation overflow:|support CET PLT
|reference to local symbol 5 |
|in ifuncmain7.o |
--- Comment #2 from H.J. Lu ---
template
uint64_t
Output_data_plt_x86_64_ibt::do_address_for_global(const Symbol* gsym)
{
uint64_t offset = this->aplt_offset_;
// Convert the PLT offset into an APLT offset.
unsigned int plt_offset = gsym->plt_offset();
if (gsym->type() == elfcpp::STT_GNU_IFUNC
&& gsym->can_use_relative_reloc(false))
offset += this->regular_count() * aplt_entry_size;
else
plt_offset -= plt_entry_size;
plt_offset = plt_offset / (plt_entry_size / aplt_entry_size);
return this->address() + offset + plt_offset;
}
checks STT_GNU_IFUNC. But do_address_for_local doesn't. Even with
diff --git a/gold/x86_64.cc b/gold/x86_64.cc
index 1442a009d7..284809c719 100644
--- a/gold/x86_64.cc
+++ b/gold/x86_64.cc
@@ -2288,7 +2288,12 @@
Output_data_plt_x86_64_ibt::do_address_for_local(const Relobj* object,
unsigned int r_sym)
{
// Convert the PLT offset into an APLT offset.
- unsigned int plt_offset = ((object->local_plt_offset(r_sym) -
plt_entry_size)
+ const Sized_relobj_file* sized_relobj =
+static_cast*>(object);
+ const Symbol_value* psymval = sized_relobj->local_symbol(r_sym);
+ unsigned int plt_offset = ((object->local_plt_offset(r_sym)
+ - (psymval->is_ifunc_symbol()
+? 0 : plt_entry_size))
/ (plt_entry_size / aplt_entry_size));
return (this->address()
+ this->aplt_offset_
I got
/export/gnu/import/git/gitlab/x86-binutils/gold/../test-driver: line 107:
299279 Segmentation fault (core dumped) "$@" > $log_file 2>&1
FAIL: tls_shared_gnu2_test
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug gold/25872] Gold doesn't properly support IBT PLT
https://sourceware.org/bugzilla/show_bug.cgi?id=25872 H.J. Lu changed: What|Removed |Added Summary|Gold doesn't properly |Gold doesn't properly |support CET PLT |support IBT PLT -- You are receiving this mail because: You are on the CC list for the bug.
[Bug binutils/25873] New: ar segfaults when liblto_plugin.so.0.0.0 is a symlink in /usr/lib/bfd-plugins
https://sourceware.org/bugzilla/show_bug.cgi?id=25873
Bug ID: 25873
Summary: ar segfaults when liblto_plugin.so.0.0.0 is a symlink
in /usr/lib/bfd-plugins
Product: binutils
Version: 2.34
Status: NEW
Severity: normal
Priority: P2
Component: binutils
Assignee: unassigned at sourceware dot org
Reporter: tstellar at redhat dot com
Target Milestone: ---
ar segfaults when using a plugin from /usr/lib/bfd-plugins that is a symlink.
Here is how to reproduce:
echo "int foo(){}" | gcc -flto -O2 -c -x c -o foo.o -
echo "int bar(){}" | gcc -flto -O2 -c -x c -o bar.o -
ln -s /usr/libexec/gcc/x86_64-redhat-linux/10/liblto_plugin.so.0.0.0
/usr/lib/bfd-plugins/
ar crs foo.a foo.o bar.o
Segmentation fault (core dumped)
If the plugin is copied directly and not a symlink, then there is no segfault:
rm /usr/lib/bfd-plugins/liblto_plugin.so.0.0.0
cp /usr/libexec/gcc/x86_64-redhat-linux/10/liblto_plugin.so.0.0.0
/usr/lib/bfd-plugins/
ar crs foo.a foo.o bar.o
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug binutils/25873] ar segfaults when liblto_plugin.so.0.0.0 is a symlink in /usr/lib/bfd-plugins
https://sourceware.org/bugzilla/show_bug.cgi?id=25873
--- Comment #1 from Tom Stellard ---
It also seems like the LLVMgold.so plugin does not have the same problem. This
similar sequence works with this plugin:
echo "int foo(){}" | clang -flto -O2 -c -x c -o foo.o -
echo "int bar(){}" | clang -flto -O2 -c -x c -o bar.o -
ln -s /usr/lib64/LLVMgold.so /usr/lib/bfd-plugins/
ar crs foo.a foo.o bar.o
--
You are receiving this mail because:
You are on the CC list for the bug.
