This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: warning: Could not load shared library symbols for linux-vdso.so.1.
On 08/19/2016 12:40 PM, Yao Qi wrote:
>> So below's the "ugly" patch. It's functionally exactly the same as
>> the larger one. The switch to linux_so_ops was mainly "cosmetic".
>>
>> Let me know what you prefer.
>>
>
> This patch is good to master and 7.12.
Hmm, wait, wait, wait... I think we can do simpler and better even.
If reading the elf out of core memory works, then that means that
there must be a load segment that covers it. So that means we should
be able to find a PT_LOAD segment for the vDSO, in the core file. E.g.,:
$ readelf -l testsuite/outputs/gdb.base/vdso-warning/vdso-warning.core
Elf file type is CORE (Core file)
Entry point 0x0
There are 18 program headers, starting at offset 64
Program Headers:
Type Offset VirtAddr PhysAddr
FileSiz MemSiz Flags Align
NOTE 0x0000000000000430 0x0000000000000000 0x0000000000000000
0x0000000000000c18 0x0000000000000000 R 1
...
LOAD 0x0000000000036048 0x00007ffff7ffa000 0x0000000000000000
0x0000000000002000 0x0000000000002000 R E 1
...
Hence the patch below. If this doesn't work, then I think the ones
based on updating the vsyscall range from the memory bfd's size wouldn't
either. And I think this one should work even for the cases that led to
commit 5979d6b69b20 ("Handle VDSO section headers past end of page")
too, thus making it even better.
I even wrote ChangeLog entries this time... :-)
WDYT?
>From bdf1cb2aa3ef3b22fb80ce24c3b2d19ac5be9438 Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Fri, 19 Aug 2016 14:09:06 +0100
Subject: [PATCH] Make vDSO detection work with core files
Loading a core dump that was either generated on a system running
pristine glibc master, or on a Fedora/RHEL system with LD_DEBUG=unused
set in the environment, solib-svr4.c:svr4_current_sos fails to filter
out the vDSO, resulting in:
(gdb) core-file corefile.core^M
[New LWP 2362]^M
warning: Could not load shared library symbols for linux-vdso.so.1.^M
Do you need "set solib-search-path" or "set sysroot"?^M
Core was generated by `build-gdb/gdb/testsuite/outputs/gdb.base/corefile/'.^M
...
The problem is that gdbarch_vsyscall_range does not support core
inferiors at all.
When live debugging, we're finding the vDSO's start address with
auxv/AT_SYSINFO_EHDR, and then we find the vDSO's size by look for the
corresponding mapping, by parsing /proc/PID/maps. When debugging a
core dump, we can also determine the starting address from
auxv/AT_SYSINFO_EHDR. However, we obviously can't read the core
mappings out of the host's /proc. But we can instead look for a
corresponding load segment in the core's bfd.
gdb/ChangeLog:
2016-08-19 Pedro Alves <palves@redhat.com>
* linux-tdep.c (linux_vsyscall_range_raw): For core inferiors,
find the vDSO's start address with AT_SYSINFO_EHDR too, and
determine the vDSO's size by finding the PT_LOAD segment that
matches AT_SYSINFO_EHDR.
gdb/testsuite/ChangeLog:
2016-08-19 Pedro Alves <palves@redhat.com>
* gdb.base/vdso-warning.exp: Test core dumps too. Use
with_test_prefix. Factor out bits to ...
(test_no_vdso): ... this new procedure.
---
gdb/linux-tdep.c | 35 ++++++++++++---
gdb/testsuite/gdb.base/vdso-warning.exp | 76 ++++++++++++++++++++++-----------
2 files changed, 82 insertions(+), 29 deletions(-)
diff --git a/gdb/linux-tdep.c b/gdb/linux-tdep.c
index ab110b0..718dc1a 100644
--- a/gdb/linux-tdep.c
+++ b/gdb/linux-tdep.c
@@ -2287,17 +2287,42 @@ linux_vsyscall_range_raw (struct gdbarch *gdbarch, struct mem_range *range)
long pid;
char *data;
- /* Can't access /proc if debugging a core file. */
- if (!target_has_execution)
+ if (target_auxv_search (¤t_target, AT_SYSINFO_EHDR, &range->start) <= 0)
return 0;
+ /* It doesn't make sense to access the host's /proc when debugging a
+ core file. Instead, look for the PT_LOAD segment that matches
+ the vDSO. */
+ if (!target_has_execution)
+ {
+ Elf_Internal_Phdr *phdrs;
+ long phdrs_size;
+ int num_phdrs, i;
+
+ phdrs_size = bfd_get_elf_phdr_upper_bound (core_bfd);
+ if (phdrs_size == -1)
+ return 0;
+
+ phdrs = (Elf_Internal_Phdr *) alloca (phdrs_size);
+ num_phdrs = bfd_get_elf_phdrs (core_bfd, phdrs);
+ if (num_phdrs == -1)
+ return 0;
+
+ for (i = 0; i < num_phdrs; i++)
+ if (phdrs[i].p_type == PT_LOAD
+ && phdrs[i].p_vaddr == range->start)
+ {
+ range->length = phdrs[i].p_memsz;
+ return 1;
+ }
+
+ return 0;
+ }
+
/* We need to know the real target PID to access /proc. */
if (current_inferior ()->fake_pid_p)
return 0;
- if (target_auxv_search (¤t_target, AT_SYSINFO_EHDR, &range->start) <= 0)
- return 0;
-
pid = current_inferior ()->pid;
/* Note that reading /proc/PID/task/PID/maps (1) is much faster than
diff --git a/gdb/testsuite/gdb.base/vdso-warning.exp b/gdb/testsuite/gdb.base/vdso-warning.exp
index af2b2b0..aeb85a2 100644
--- a/gdb/testsuite/gdb.base/vdso-warning.exp
+++ b/gdb/testsuite/gdb.base/vdso-warning.exp
@@ -13,42 +13,70 @@
# 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 that on Linux, we don't warn about not finding the vDSO. E.g.:
+#
+# warning: Could not load shared library symbols for linux-vdso.so.1.
+
standard_testfile
if { [prepare_for_testing "failed to prepare" ${testfile} $srcfile] } {
return -1
}
-gdb_breakpoint "main"
+with_test_prefix "setup" {
+ gdb_breakpoint "main"
-# At least some versions of Fedora/RHEL glibc have local patches that
-# hide the vDSO. This lines re-exposes it. See PR libc/13097,
-# comment 2. There's no support for passing environment variables in
-# the remote protocol, but that's OK -- if we're testing against a
-# glibc that doesn't list the vDSO without this, the test should still
-# pass.
-gdb_test_no_output "set environment LD_DEBUG=unused"
+ # At least some versions of Fedora/RHEL glibc have local patches that
+ # hide the vDSO. This lines re-exposes it. See PR libc/13097,
+ # comment 2. There's no support for passing environment variables in
+ # the remote protocol, but that's OK -- if we're testing against a
+ # glibc that doesn't list the vDSO without this, the test should still
+ # pass.
+ gdb_test_no_output "set environment LD_DEBUG=unused"
+}
-gdb_run_cmd
+proc test_no_vdso {command} {
+ global srcfile
+ global gdb_prompt
-set test "stop without warning"
-gdb_test_multiple "" $test {
- -re "Could not load shared library symbols .*\r\n$gdb_prompt $" {
- fail $test
+ set message "startup"
+ gdb_test_multiple "$command" $message {
+ -re "Could not load shared library symbols .*\r\n$gdb_prompt $" {
+ fail $message
+ }
+ -re "main \\(\\) at .*$srcfile.*\r\n$gdb_prompt $" {
+ pass $message
+ }
}
- -re "\r\nBreakpoint \[0-9\]+, main .*\r\n$gdb_prompt $" {
- pass $test
+
+ # Extra testing in case the warning changes and we miss updating
+ # the above.
+ set test "no vdso without symbols is listed"
+ gdb_test_multiple "info shared" $test {
+ -re "No\[^\r\n\]+linux-(vdso|gate).*$gdb_prompt $" {
+ fail $test
+ }
+ -re "$gdb_prompt $" {
+ pass $test
+ }
}
}
-# Extra testing in case the warning changes and we miss updating the
-# above.
-set test "no vdso without symbols is listed"
-gdb_test_multiple "info shared" $test {
- -re "No\[^\r\n\]+linux-(vdso|gate).*$gdb_prompt $" {
- fail $test
- }
- -re "$gdb_prompt $" {
- pass $test
+# First, try a live process.
+with_test_prefix "run" {
+ gdb_run_cmd
+ test_no_vdso ""
+}
+
+# Now, dump a core, and reload it.
+with_test_prefix "core" {
+ set corefile [standard_output_file $testfile.core]
+ set core_supported [gdb_gcore_cmd "$corefile" "save a corefile"]
+ if {!$core_supported} {
+ return -1
}
+
+ clean_restart ${testfile}
+
+ test_no_vdso "core-file $corefile"
}
--
2.5.5