public inbox for elfutils@sourceware.org
 help / color / mirror / Atom feed
From: Mark Wielaard <mark@klomp.org>
To: elfutils-devel@sourceware.org
Cc: Ulf Hermann <ulf.hermann@qt.io>
Subject: [PATCH 3/5] Add frame pointer unwinding as fallback on x86_64
Date: Tue, 25 Apr 2017 12:55:00 -0000	[thread overview]
Message-ID: <1493124579-21017-3-git-send-email-mark@klomp.org> (raw)
In-Reply-To: <1493124579-21017-1-git-send-email-mark@klomp.org>

From: Ulf Hermann <ulf.hermann@qt.io>

If we don't find any debug information for a given frame, we usually
cannot unwind any further. However, the binary in question might have
been compiled with frame pointers, in which case we can look up the
well known frame pointer locations in the stack snapshot and use them
to bridge the frames without debug information.

The "unwind" hook is the right place for this as it is so far only
used on s390 and called only after trying to unwind with debug
information.

Signed-off-by: Ulf Hermann <ulf.hermann@qt.io>
---
 backends/ChangeLog                    |   6 +++
 backends/Makefile.am                  |   2 +-
 backends/x86_64_init.c                |   1 +
 backends/x86_64_unwind.c              |  86 ++++++++++++++++++++++++++++++++++
 tests/ChangeLog                       |   7 +++
 tests/Makefile.am                     |   3 ++
 tests/backtrace.x86_64.fp.core.bz2    | Bin 0 -> 11072 bytes
 tests/backtrace.x86_64.fp.exec.bz2    | Bin 0 -> 434645 bytes
 tests/run-backtrace-fp-core-x86_64.sh |  29 ++++++++++++
 9 files changed, 133 insertions(+), 1 deletion(-)
 create mode 100644 backends/x86_64_unwind.c
 create mode 100644 tests/backtrace.x86_64.fp.core.bz2
 create mode 100644 tests/backtrace.x86_64.fp.exec.bz2
 create mode 100755 tests/run-backtrace-fp-core-x86_64.sh

diff --git a/backends/ChangeLog b/backends/ChangeLog
index 39390cb..eec9923 100644
--- a/backends/ChangeLog
+++ b/backends/ChangeLog
@@ -1,3 +1,9 @@
+2017-02-09  Ulf Hermann  <ulf.hermann@qt.io>
+
+	* x86_64_unwind.c: New file
+	* Makefile.am (x86_64_SRCS): Add x86_64_unwind.c
+	* x86_64_init.c (x86_64_init): Hook x86_64_unwind
+
 2017-02-15  Mark Wielaard  <mark@klomp.org>
 
 	* ppc64_init.c (ppc64_init): Add check_object_attribute HOOK.
diff --git a/backends/Makefile.am b/backends/Makefile.am
index b553ec3..60917b9 100644
--- a/backends/Makefile.am
+++ b/backends/Makefile.am
@@ -59,7 +59,7 @@ am_libebl_sh_pic_a_OBJECTS = $(sh_SRCS:.c=.os)
 
 x86_64_SRCS = x86_64_init.c x86_64_symbol.c x86_64_corenote.c x86_64_cfi.c \
 	      x86_64_retval.c x86_64_regs.c i386_auxv.c x86_64_syscall.c \
-	      x86_64_initreg.c x32_corenote.c
+	      x86_64_initreg.c x86_64_unwind.c x32_corenote.c
 cpu_x86_64 = ../libcpu/libcpu_x86_64.a
 libebl_x86_64_pic_a_SOURCES = $(x86_64_SRCS)
 am_libebl_x86_64_pic_a_OBJECTS = $(x86_64_SRCS:.c=.os)
diff --git a/backends/x86_64_init.c b/backends/x86_64_init.c
index cfd0158..adfa479 100644
--- a/backends/x86_64_init.c
+++ b/backends/x86_64_init.c
@@ -68,6 +68,7 @@ x86_64_init (Elf *elf __attribute__ ((unused)),
   /* gcc/config/ #define DWARF_FRAME_REGISTERS.  */
   eh->frame_nregs = 17;
   HOOK (eh, set_initial_registers_tid);
+  HOOK (eh, unwind);
 
   return MODVERSION;
 }
diff --git a/backends/x86_64_unwind.c b/backends/x86_64_unwind.c
new file mode 100644
index 0000000..ade64c0
--- /dev/null
+++ b/backends/x86_64_unwind.c
@@ -0,0 +1,86 @@
+/* Get previous frame state for an existing frame state.
+   Copyright (C) 2016 The Qt Company Ltd.
+   This file is part of elfutils.
+
+   This file is free software; you can redistribute it and/or modify
+   it under the terms of either
+
+     * the GNU Lesser General Public License as published by the Free
+       Software Foundation; either version 3 of the License, or (at
+       your option) any later version
+
+   or
+
+     * the GNU General Public License as published by the Free
+       Software Foundation; either version 2 of the License, or (at
+       your option) any later version
+
+   or both in parallel, as here.
+
+   elfutils 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 copies of the GNU General Public License and
+   the GNU Lesser General Public License along with this program.  If
+   not, see <http://www.gnu.org/licenses/>.  */
+
+#ifdef HAVE_CONFIG_H
+# include <config.h>
+#endif
+
+#include <stdlib.h>
+#include <assert.h>
+
+#define BACKEND x86_64_
+#include "libebl_CPU.h"
+
+/* There was no CFI. Maybe we happen to have a frame pointer and can unwind from that?  */
+
+bool
+x86_64_unwind (Ebl *ebl __attribute__ ((unused)),
+               Dwarf_Addr pc __attribute__ ((unused)),
+               ebl_tid_registers_t *setfunc, ebl_tid_registers_get_t *getfunc,
+               ebl_pid_memory_read_t *readfunc, void *arg,
+               bool *signal_framep __attribute__ ((unused)))
+{
+  // Register 6 is supposed to be rbp, thus the conventional frame pointer
+  const int fpReg = 6;
+  const int spReg = 7;
+
+  Dwarf_Word fp;
+  if (!getfunc(fpReg, 1, &fp, arg) || fp == 0)
+    return false;
+
+  // Try to read old sp, so that we can avoid infinite loops below
+  Dwarf_Word sp;
+  if (!getfunc(spReg, 1, &sp, arg))
+    sp = 0;
+
+  Dwarf_Word prev_fp;
+  if (!readfunc(fp, &prev_fp, arg))
+    prev_fp = 0;
+
+  Dwarf_Word ret;
+  if (!readfunc(fp + 8, &ret, arg))
+    return false;
+
+  if (!setfunc(fpReg, 1, &prev_fp, arg))
+    return false;
+
+  fp += 16; // Pop fp and return address and write result to sp
+  if (!setfunc(spReg, 1, &fp, arg))
+    return false;
+
+  if (!setfunc(-1, 1, &ret, arg))
+    return false;
+
+  // If the sp didn't move up we don't actually have a new stack
+  // frame but rather some random data that doesn't include frame
+  // pointers. Break the unwinding then.
+  if (sp >= fp)
+    return false;
+
+  return true;
+}
diff --git a/tests/ChangeLog b/tests/ChangeLog
index 6f5956b..49b6a3f 100644
--- a/tests/ChangeLog
+++ b/tests/ChangeLog
@@ -1,3 +1,10 @@
+2017-02-09  Ulf Hermann  <ulf.hermann@qt.io>
+
+	* Makefile.am: Add test for unwinding with frame pointers on x86_64
+	* backtrace.x86_64.fp.core.bz2: New file
+	* backtrace.x86_64.fp.exec.bz2: New file
+	* run-backtrace-fp-core-x86_64.sh: New file
+
 2017-04-25  Mark Wielaard  <mark@klomp.org>
 
 	* backtrace-subr.sh (check_backtracegen): New function.
diff --git a/tests/Makefile.am b/tests/Makefile.am
index f6d8b0d..b0db19f 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -115,6 +115,7 @@ TESTS = run-arextract.sh run-arsymtest.sh newfile test-nlist \
 	run-backtrace-native.sh run-backtrace-data.sh run-backtrace-dwarf.sh \
 	run-backtrace-native-biarch.sh run-backtrace-native-core.sh \
 	run-backtrace-native-core-biarch.sh run-backtrace-core-x86_64.sh \
+	run-backtrace-fp-core-x86_64.sh \
 	run-backtrace-core-x32.sh \
 	run-backtrace-core-i386.sh run-backtrace-core-ppc.sh \
 	run-backtrace-core-s390x.sh run-backtrace-core-s390.sh \
@@ -293,9 +294,11 @@ EXTRA_DIST = run-arextract.sh run-arsymtest.sh \
 	     run-backtrace-native.sh run-backtrace-native-biarch.sh \
 	     run-backtrace-native-core.sh run-backtrace-native-core-biarch.sh \
 	     run-backtrace-core-x86_64.sh run-backtrace-core-i386.sh \
+	     run-backtrace-fp-core-x86_64.sh \
 	     run-backtrace-core-x32.sh \
 	     backtrace-subr.sh backtrace.i386.core.bz2 backtrace.i386.exec.bz2 \
 	     backtrace.x86_64.core.bz2 backtrace.x86_64.exec.bz2 \
+	     backtrace.x86_64.fp.core.bz2 backtrace.x86_64.fp.exec.bz2 \
 	     backtrace.ppc.core.bz2 backtrace.ppc.exec.bz2 \
 	     run-backtrace-core-ppc.sh testfile66.bz2 testfile66.core.bz2 \
 	     backtrace.s390x.core.bz2 backtrace.s390x.exec.bz2 \

diff --git a/tests/run-backtrace-fp-core-x86_64.sh b/tests/run-backtrace-fp-core-x86_64.sh
new file mode 100755
index 0000000..348eb18
--- /dev/null
+++ b/tests/run-backtrace-fp-core-x86_64.sh
@@ -0,0 +1,29 @@
+#! /bin/bash
+# Copyright (C) 2017 The Qt Company
+# This file is part of elfutils.
+#
+# This file 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.
+#
+# elfutils 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/>.
+
+. $srcdir/backtrace-subr.sh
+
+# The binary is generated by compiling with eh_frame CFI, but with frame
+# pointers.
+#
+# gcc -static -O2 -fno-omit-frame-pointer -fno-asynchronous-unwind-tables \
+#     -D_GNU_SOURCE -pthread -o tests/backtrace.x86_64.fp.exec -I. -Ilib \
+#     tests/backtrace-child.c
+#
+# The core is generated by calling the binary with --gencore
+
+check_core x86_64.fp
-- 
1.8.3.1

  parent reply	other threads:[~2017-04-25 12:50 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-02-15 23:34 frame unwinding patches Mark Wielaard
2017-02-15 23:34 ` [PATCH 2/3] Add frame pointer unwinding as fallback on arm Mark Wielaard
2017-02-15 23:34 ` [PATCH 1/3] Add frame pointer unwinding as fallback on x86_64 Mark Wielaard
2017-02-15 23:34 ` [PATCH 3/3] Add frame pointer unwinding for aarch64 Mark Wielaard
2017-02-16  9:13 ` frame unwinding patches Ulf Hermann
2017-04-03  9:00 ` Milian Wolff
2017-04-03  9:03   ` Ulf Hermann
2017-04-03 21:14     ` Mark Wielaard
2017-04-07 10:27       ` Mark Wielaard
2017-04-11 10:16         ` Ulf Hermann
2017-04-19 19:48           ` Mark Wielaard
2017-04-20  9:26             ` Ulf Hermann
2017-04-25 12:50               ` Mark Wielaard
2017-04-25 12:54                 ` [PATCH 1/5] Revert "Optionally allow unknown symbols in the backtrace tests" Mark Wielaard
2017-04-25 12:50                   ` [PATCH 2/5] tests: Add core backtracegen chec and regenerate ppc32 backtrace test files Mark Wielaard
2017-04-25 13:04                     ` Ulf Hermann
2017-04-25 12:55                   ` Mark Wielaard [this message]
2017-04-25 13:05                     ` [PATCH 3/5] Add frame pointer unwinding as fallback on x86_64 Ulf Hermann
2017-04-25 12:56                   ` [PATCH 1/5] Revert "Optionally allow unknown symbols in the backtrace tests" Ulf Hermann
2017-04-25 12:56                   ` [PATCH 4/5] Add i386 frame pointer unwinder Mark Wielaard
2017-04-25 13:38                     ` Ulf Hermann
2017-04-25 13:11                   ` [PATCH 5/5] Add frame pointer unwinding for aarch64 Mark Wielaard
2017-04-25 21:55                     ` Ulf Hermann
2017-04-25 22:13                     ` Mark Wielaard
2017-04-25 22:23                       ` Ulf Hermann
2017-04-26 15:24                         ` Mark Wielaard
2017-04-27 14:02                           ` Ulf Hermann
2017-04-27 14:29                             ` Mark Wielaard
2017-04-27 14:35                               ` Ulf Hermann
2017-04-27 15:09                                 ` Mark Wielaard
2017-04-27 15:42                                   ` Ulf Hermann
2017-05-03  8:46                                 ` Mark Wielaard
2017-04-26 15:20                 ` frame unwinding patches Ulf Hermann
2017-04-03 21:23   ` Jan Kratochvil
2017-04-04  7:40     ` Milian Wolff
2017-04-04  7:55       ` Jan Kratochvil
2017-04-04  8:25         ` Ulf Hermann

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=1493124579-21017-3-git-send-email-mark@klomp.org \
    --to=mark@klomp.org \
    --cc=elfutils-devel@sourceware.org \
    --cc=ulf.hermann@qt.io \
    /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).