Blame SOURCES/gdb-rhbz1842691-corefile-mem-access-14of15.patch

be07d7
From FEDORA_PATCHES Mon Sep 17 00:00:00 2001
be07d7
From: Keith Seitz <keiths@redhat.com>
be07d7
Date: Tue, 28 Jul 2020 09:46:44 -0400
be07d7
Subject: gdb-rhbz1842691-corefile-mem-access-14of15.patch
be07d7
be07d7
;; New core file tests with mappings over existing program memory
be07d7
;; Kevin Buettner, RH BZ 1842961
be07d7
be07d7
   Author: Kevin Buettner <kevinb@redhat.com>
be07d7
   Date:   Wed Jun 17 19:25:47 2020 -0700
be07d7
be07d7
    New core file tests with mappings over existing program memory
be07d7
be07d7
    This test case was inspired by Pedro's demonstration of a problem
be07d7
    with my v2 patches.  It can be found here:
be07d7
be07d7
        https://sourceware.org/pipermail/gdb-patches/2020-May/168826.html
be07d7
be07d7
    In a nutshell, my earlier patches could not handle the case in
be07d7
    which a read-only mapping created with mmap() was created at
be07d7
    an address used by other file-backed read-only memory in use by
be07d7
    the process.
be07d7
be07d7
    This problem has been fixed (for Linux, anyway) by the commit "Use
be07d7
    NT_FILE note section for reading core target memory".
be07d7
be07d7
    When I run this test without any of my recent corefile patches,
be07d7
    I see these failures:
be07d7
be07d7
    FAIL: gdb.base/corefile2.exp: kernel core: print/x mbuf_ro[0]@4
be07d7
    FAIL: gdb.base/corefile2.exp: kernel core: print/x mbuf_ro[pagesize-4]@4
be07d7
    FAIL: gdb.base/corefile2.exp: kernel core: print/x mbuf_ro[-3]@6
be07d7
    FAIL: gdb.base/corefile2.exp: kernel core: print/x mbuf_rw[pagesize-3]@6
be07d7
    FAIL: gdb.base/corefile2.exp: kernel core: print/x mbuf_ro[pagesize-3]@6
be07d7
    FAIL: gdb.base/corefile2.exp: maint print core-file-backed-mappings
be07d7
    FAIL: gdb.base/corefile2.exp: gcore core: print/x mbuf_ro[-3]@6
be07d7
be07d7
    The ones involving mbuf_ro will almost certainly fail when run on
be07d7
    non-Linux systems; I've used setup_xfail on those tests to prevent
be07d7
    them from outright FAILing when not run on Linux.  For a time, I
be07d7
    had considered skipping these tests altogether when not run on
be07d7
    Linux, but I changed my mind due to this failure...
be07d7
be07d7
    FAIL: gdb.base/corefile2.exp: print/x mbuf_rw[pagesize-3]@6
be07d7
be07d7
    I think it *should* pass without my recent corefile patches.  The fact
be07d7
    that it doesn't is likely due to a bug in GDB.  The following
be07d7
    interaction with GDB demonstrates the problem:
be07d7
be07d7
    (gdb) print/x mbuf_rw[pagesize-3]@6
be07d7
    $1 = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0}
be07d7
    (gdb) print/x mbuf_rw[pagesize]@3
be07d7
    $2 = {0x6b, 0x6b, 0x6b}
be07d7
be07d7
    The last three values in display of $1 should be the same as those
be07d7
    shown by $2.  Like this...
be07d7
be07d7
    (gdb) print/x mbuf_rw[pagesize-3]@6
be07d7
    $1 = {0x0, 0x0, 0x0, 0x6b, 0x6b, 0x6b}
be07d7
    (gdb) print/x mbuf_rw[pagesize]@3
be07d7
    $2 = {0x6b, 0x6b, 0x6b}
be07d7
be07d7
    That latter output was obtained with the use of all of my current
be07d7
    corefile patches.  I see no failures on Linux when running this test
be07d7
    with my current set of corefile patches.  I tested 3 architectures:
be07d7
    x86_64, s390x, and aarch64.
be07d7
be07d7
    I also tested on FreeBSD 12.1-RELEASE.  I see the following results
be07d7
    both with and without the current set of core file patches:
be07d7
be07d7
        # of expected passes		26
be07d7
        # of expected failures		8
be07d7
be07d7
    Of particular interest is that I did *not* see the problematic mbuf_rw
be07d7
    failure noted earlier (both with and without the core file patches).
be07d7
    I still don't have an explanation for why this failure occurred on
be07d7
    Linux.  Prior to running the tests, I had hypothesized that I'd see
be07d7
    this failure on FreeBSD too, but testing shows that this is not the
be07d7
    case.
be07d7
be07d7
    Also of importance is that we see no FAILs with this test on FreeBSD
be07d7
    which indicates that I XFAILed the correct tests.
be07d7
be07d7
    This version runs the interesting tests twice, once with a kernel
be07d7
    created core file and another time with a gcore created core file.
be07d7
be07d7
    It also does a very minimal test of the new command "maint print
be07d7
    core-file-backed-mappings".
be07d7
be07d7
    gdb/testsuite/ChangeLog:
be07d7
be07d7
    	* gdb.base/corefile2.exp: New file.
be07d7
    	* gdb.base/coremaker2.exp: New file.
be07d7
be07d7
diff --git a/gdb/testsuite/gdb.base/corefile2.exp b/gdb/testsuite/gdb.base/corefile2.exp
be07d7
new file mode 100644
be07d7
--- /dev/null
be07d7
+++ b/gdb/testsuite/gdb.base/corefile2.exp
be07d7
@@ -0,0 +1,185 @@
be07d7
+# Copyright 2020 Free Software Foundation, Inc.
be07d7
+
be07d7
+# This program is free software; you can redistribute it and/or modify
be07d7
+# it under the terms of the GNU General Public License as published by
be07d7
+# the Free Software Foundation; either version 3 of the License, or
be07d7
+# (at your option) any later version.
be07d7
+#
be07d7
+# This program is distributed in the hope that it will be useful,
be07d7
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
be07d7
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
be07d7
+# GNU General Public License for more details.
be07d7
+#
be07d7
+# You should have received a copy of the GNU General Public License
be07d7
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
be07d7
+
be07d7
+# Tests of core file memory accesses when mmap() has been used to
be07d7
+# create a "hole" of zeroes over pre-existing memory regions.  See
be07d7
+# coremaker2.c for details.
be07d7
+
be07d7
+# are we on a target board
be07d7
+if ![isnative] then {
be07d7
+    return
be07d7
+}
be07d7
+
be07d7
+# Some of these tests will only work on GNU/Linux due to the
be07d7
+# fact that Linux core files includes a section describing
be07d7
+# memory address to file mappings.  We'll use set_up_xfail for the
be07d7
+# affected tests.  As other targets become supported, the condition
be07d7
+# can be changed accordingly.
be07d7
+
be07d7
+set xfail 0
be07d7
+if { ![istarget *-linux*] } {
be07d7
+    set xfail 1
be07d7
+}
be07d7
+
be07d7
+standard_testfile coremaker2.c
be07d7
+
be07d7
+if {[build_executable $testfile.exp $testfile $srcfile debug] == -1} {
be07d7
+    untested "failed to compile"
be07d7
+    return -1
be07d7
+}
be07d7
+
be07d7
+set corefile [core_find $binfile {}]
be07d7
+if {$corefile == ""} {
be07d7
+    return 0
be07d7
+}
be07d7
+
be07d7
+gdb_start
be07d7
+gdb_reinitialize_dir $srcdir/$subdir
be07d7
+gdb_load ${binfile}
be07d7
+
be07d7
+# Attempt to load the core file.
be07d7
+
be07d7
+gdb_test_multiple "core-file $corefile" "core-file command" {
be07d7
+    -re ".* program is being debugged already.*y or n. $" {
be07d7
+	# gdb_load may connect us to a gdbserver.
be07d7
+	send_gdb "y\n"
be07d7
+	exp_continue
be07d7
+    }
be07d7
+    -re "Core was generated by .*corefile.*\r\n\#0  .*\(\).*\r\n$gdb_prompt $" {
be07d7
+	pass "core-file command"
be07d7
+    }
be07d7
+    -re "Core was generated by .*\r\n\#0  .*\(\).*\r\n$gdb_prompt $" {
be07d7
+	pass "core-file command (with bad program name)"
be07d7
+    }
be07d7
+    -re ".*registers from core file: File in wrong format.* $" {
be07d7
+	fail "core-file command (could not read registers from core file)"
be07d7
+    }
be07d7
+}
be07d7
+
be07d7
+# Perform the "interesting" tests which check the contents of certain
be07d7
+# memory regions.
be07d7
+
be07d7
+proc do_tests { } {
be07d7
+    global xfail
be07d7
+
be07d7
+    # Check contents of beginning of buf_rw and buf_ro.
be07d7
+
be07d7
+    gdb_test {print/x buf_rw[0]@4} {\{0x6b, 0x6b, 0x6b, 0x6b\}}
be07d7
+    gdb_test {print/x buf_ro[0]@4} {\{0xc5, 0xc5, 0xc5, 0xc5\}}
be07d7
+
be07d7
+    # Check for correct contents at beginning of mbuf_rw and mbuf_ro.
be07d7
+
be07d7
+    gdb_test {print/x mbuf_rw[0]@4} {\{0x0, 0x0, 0x0, 0x0\}}
be07d7
+
be07d7
+    if { $xfail } { setup_xfail "*-*-*" }
be07d7
+    gdb_test {print/x mbuf_ro[0]@4} {\{0x0, 0x0, 0x0, 0x0\}}
be07d7
+
be07d7
+    # Check contents of mbuf_rw and mbuf_ro at the end of these regions.
be07d7
+
be07d7
+    gdb_test {print/x mbuf_rw[pagesize-4]@4} {\{0x0, 0x0, 0x0, 0x0\}}
be07d7
+
be07d7
+    if { $xfail } { setup_xfail "*-*-*" }
be07d7
+    gdb_test {print/x mbuf_ro[pagesize-4]@4} {\{0x0, 0x0, 0x0, 0x0\}}
be07d7
+
be07d7
+    # Check contents of mbuf_rw and mbuf_ro, right before the hole,
be07d7
+    # overlapping into the beginning of these mmap'd regions.
be07d7
+
be07d7
+    gdb_test {print/x mbuf_rw[-3]@6} {\{0x6b, 0x6b, 0x6b, 0x0, 0x0, 0x0\}}
be07d7
+
be07d7
+    if { $xfail } { setup_xfail "*-*-*" }
be07d7
+    gdb_test {print/x mbuf_ro[-3]@6} {\{0xc5, 0xc5, 0xc5, 0x0, 0x0, 0x0\}}
be07d7
+
be07d7
+    # Likewise, at the end of the mbuf_rw and mbuf_ro, with overlap.
be07d7
+
be07d7
+    # If this test FAILs, it's probably a genuine bug unrelated to whether
be07d7
+    # the core file includes a section describing memory address to file
be07d7
+    # mappings or not.  (So don't xfail it!)
be07d7
+    gdb_test {print/x mbuf_rw[pagesize-3]@6} {\{0x0, 0x0, 0x0, 0x6b, 0x6b, 0x6b\}}
be07d7
+
be07d7
+    if { $xfail } { setup_xfail "*-*-*" }
be07d7
+    gdb_test {print/x mbuf_ro[pagesize-3]@6} {\{0x0, 0x0, 0x0, 0xc5, 0xc5, 0xc5\}}
be07d7
+
be07d7
+    # Check contents of (what should be) buf_rw and buf_ro immediately after
be07d7
+    # mbuf_rw and mbuf_ro holes.
be07d7
+
be07d7
+    gdb_test {print/x mbuf_rw[pagesize]@4} {\{0x6b, 0x6b, 0x6b, 0x6b\}}
be07d7
+    gdb_test {print/x mbuf_ro[pagesize]@4} {\{0xc5, 0xc5, 0xc5, 0xc5\}}
be07d7
+
be07d7
+    # Check contents at ends of buf_rw and buf_rw.
be07d7
+
be07d7
+    gdb_test {print/x buf_rw[sizeof(buf_rw)-4]@4} {\{0x6b, 0x6b, 0x6b, 0x6b\}}
be07d7
+    gdb_test {print/x buf_ro[sizeof(buf_ro)-4]@4} {\{0xc5, 0xc5, 0xc5, 0xc5\}}
be07d7
+}
be07d7
+
be07d7
+# Run tests with kernel-produced core file.
be07d7
+
be07d7
+with_test_prefix "kernel core" {
be07d7
+    do_tests
be07d7
+}
be07d7
+
be07d7
+# Verify that "maint print core-file-backed-mappings" exists and does
be07d7
+# not crash GDB.  If it produces any output at all, make sure that
be07d7
+# that output at least mentions binfile.
be07d7
+
be07d7
+set test "maint print core-file-backed-mappings"
be07d7
+gdb_test_multiple $test "" {
be07d7
+    -re ".*$binfile.*$gdb_prompt $" {
be07d7
+	pass $test
be07d7
+    }
be07d7
+    -re "^$test\[\r\n\]*$gdb_prompt $" {
be07d7
+	pass "$test (no output)"
be07d7
+    }
be07d7
+}
be07d7
+
be07d7
+# Restart and run to the abort call.
be07d7
+
be07d7
+clean_restart $binfile
be07d7
+
be07d7
+if ![runto_main] then {
be07d7
+    fail "can't run to main"
be07d7
+    return
be07d7
+}
be07d7
+
be07d7
+gdb_breakpoint [gdb_get_line_number "abort"]
be07d7
+gdb_continue_to_breakpoint "at abort"
be07d7
+
be07d7
+# Do not execute abort call; instead, invoke gcore command to make a
be07d7
+# gdb-produced core file.
be07d7
+
be07d7
+set corefile [standard_output_file gcore.test]
be07d7
+set core_supported [gdb_gcore_cmd "$corefile" "save a corefile"]
be07d7
+if {!$core_supported} {
be07d7
+  return
be07d7
+}
be07d7
+
be07d7
+# maint print-core-file-backed-mappings shouldn't produce any output
be07d7
+# when not debugging a core file.
be07d7
+
be07d7
+gdb_test_no_output "maint print core-file-backed-mappings" \
be07d7
+    "maint print core-file-backed-mapping with no core file"
be07d7
+
be07d7
+clean_restart $binfile
be07d7
+
be07d7
+set core_loaded [gdb_core_cmd "$corefile" "re-load generated corefile"]
be07d7
+if { $core_loaded == -1 } {
be07d7
+    # No use proceeding from here.
be07d7
+    return
be07d7
+}
be07d7
+
be07d7
+# Run tests using gcore-produced core file.
be07d7
+
be07d7
+with_test_prefix "gcore core" {
be07d7
+    do_tests
be07d7
+}
be07d7
diff --git a/gdb/testsuite/gdb.base/coremaker2.c b/gdb/testsuite/gdb.base/coremaker2.c
be07d7
new file mode 100644
be07d7
--- /dev/null
be07d7
+++ b/gdb/testsuite/gdb.base/coremaker2.c
be07d7
@@ -0,0 +1,150 @@
be07d7
+/* Copyright 1992-2020 Free Software Foundation, Inc.
be07d7
+
be07d7
+   This file is part of GDB.
be07d7
+
be07d7
+   This program is free software; you can redistribute it and/or modify
be07d7
+   it under the terms of the GNU General Public License as published by
be07d7
+   the Free Software Foundation; either version 3 of the License, or
be07d7
+   (at your option) any later version.
be07d7
+
be07d7
+   This program is distributed in the hope that it will be useful,
be07d7
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
be07d7
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
be07d7
+   GNU General Public License for more details.
be07d7
+
be07d7
+   You should have received a copy of the GNU General Public License
be07d7
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
be07d7
+
be07d7
+/*  This test has two large memory areas buf_rw and buf_ro. 
be07d7
+
be07d7
+    buf_rw is written to by the program while buf_ro is initialized at
be07d7
+    compile / load time.  Thus, when a core file is created, buf_rw's
be07d7
+    memory should reside in the core file, but buf_ro probably won't be.
be07d7
+    Instead, the contents of buf_ro are available from the executable.
be07d7
+
be07d7
+    Now, for the wrinkle:  We create a one page read-only mapping over
be07d7
+    both of these areas.  This will create a one page "hole" of all
be07d7
+    zeros in each area.
be07d7
+
be07d7
+    Will GDB be able to correctly read memory from each of the four
be07d7
+    (or six, if you count the regions on the other side of each hole)
be07d7
+    memory regions?  */
be07d7
+
be07d7
+#include <stdio.h>
be07d7
+#include <sys/types.h>
be07d7
+#include <fcntl.h>
be07d7
+#include <sys/mman.h>
be07d7
+#include <signal.h>
be07d7
+#include <stdlib.h>
be07d7
+#include <unistd.h>
be07d7
+#include <string.h>
be07d7
+#include <errno.h>
be07d7
+
be07d7
+/* These are globals so that we can find them easily when debugging
be07d7
+   the core file.  */
be07d7
+long pagesize;
be07d7
+unsigned long long addr;
be07d7
+char *mbuf_ro;
be07d7
+char *mbuf_rw;
be07d7
+
be07d7
+/* 24 KiB buffer.  */
be07d7
+char buf_rw[24 * 1024];
be07d7
+
be07d7
+/* 24 KiB worth of data.  For this test case, we can't allocate a
be07d7
+   buffer and then fill it; we want GDB to have to read this data
be07d7
+   from the executable; it should NOT find it in the core file.  */
be07d7
+
be07d7
+#define C5_16 \
be07d7
+  0xc5, 0xc5, 0xc5, 0xc5, \
be07d7
+  0xc5, 0xc5, 0xc5, 0xc5, \
be07d7
+  0xc5, 0xc5, 0xc5, 0xc5, \
be07d7
+  0xc5, 0xc5, 0xc5, 0xc5
be07d7
+
be07d7
+#define C5_256 \
be07d7
+  C5_16, C5_16, C5_16, C5_16, \
be07d7
+  C5_16, C5_16, C5_16, C5_16, \
be07d7
+  C5_16, C5_16, C5_16, C5_16, \
be07d7
+  C5_16, C5_16, C5_16, C5_16
be07d7
+
be07d7
+#define C5_1k \
be07d7
+  C5_256, C5_256, C5_256, C5_256
be07d7
+
be07d7
+#define C5_24k \
be07d7
+  C5_1k, C5_1k, C5_1k, C5_1k, \
be07d7
+  C5_1k, C5_1k, C5_1k, C5_1k, \
be07d7
+  C5_1k, C5_1k, C5_1k, C5_1k, \
be07d7
+  C5_1k, C5_1k, C5_1k, C5_1k, \
be07d7
+  C5_1k, C5_1k, C5_1k, C5_1k, \
be07d7
+  C5_1k, C5_1k, C5_1k, C5_1k
be07d7
+
be07d7
+const char buf_ro[] = { C5_24k };
be07d7
+
be07d7
+int
be07d7
+main (int argc, char **argv)
be07d7
+{
be07d7
+  int i, bitcount;
be07d7
+
be07d7
+#ifdef _SC_PAGESIZE
be07d7
+  pagesize = sysconf (_SC_PAGESIZE);
be07d7
+#else
be07d7
+  pagesize = 8192;
be07d7
+#endif
be07d7
+
be07d7
+  /* Verify that pagesize is a power of 2.  */
be07d7
+  bitcount = 0;
be07d7
+  for (i = 0; i < 4 * sizeof (pagesize); i++)
be07d7
+    if (pagesize & (1 << i))
be07d7
+      bitcount++;
be07d7
+
be07d7
+  if (bitcount != 1)
be07d7
+    {
be07d7
+      fprintf (stderr, "pagesize is not a power of 2.\n");
be07d7
+      exit (1);
be07d7
+    }
be07d7
+
be07d7
+  /* Compute an address that should be within buf_ro.  Complain if not.  */
be07d7
+  addr = ((unsigned long long) buf_ro + pagesize) & ~(pagesize - 1);
be07d7
+
be07d7
+  if (addr <= (unsigned long long) buf_ro
be07d7
+      || addr >= (unsigned long long) buf_ro + sizeof (buf_ro))
be07d7
+    {
be07d7
+      fprintf (stderr, "Unable to compute a suitable address within buf_ro.\n");
be07d7
+      exit (1);
be07d7
+    }
be07d7
+
be07d7
+  mbuf_ro = mmap ((void *) addr, pagesize, PROT_READ,
be07d7
+               MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED, -1, 0);
be07d7
+
be07d7
+  if (mbuf_ro == MAP_FAILED)
be07d7
+    {
be07d7
+      fprintf (stderr, "mmap #1 failed: %s.\n", strerror (errno));
be07d7
+      exit (1);
be07d7
+    }
be07d7
+
be07d7
+  /* Write (and fill) the R/W region.  */
be07d7
+  for (i = 0; i < sizeof (buf_rw); i++)
be07d7
+    buf_rw[i] = 0x6b;
be07d7
+
be07d7
+  /* Compute an mmap address within buf_rw.  Complain if it's somewhere
be07d7
+     else.  */
be07d7
+  addr = ((unsigned long long) buf_rw + pagesize) & ~(pagesize - 1);
be07d7
+
be07d7
+  if (addr <= (unsigned long long) buf_rw
be07d7
+      || addr >= (unsigned long long) buf_rw + sizeof (buf_rw))
be07d7
+    {
be07d7
+      fprintf (stderr, "Unable to compute a suitable address within buf_rw.\n");
be07d7
+      exit (1);
be07d7
+    }
be07d7
+
be07d7
+  mbuf_rw = mmap ((void *) addr, pagesize, PROT_READ,
be07d7
+               MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED, -1, 0);
be07d7
+
be07d7
+  if (mbuf_rw == MAP_FAILED)
be07d7
+    {
be07d7
+      fprintf (stderr, "mmap #2 failed: %s.\n", strerror (errno));
be07d7
+      exit (1);
be07d7
+    }
be07d7
+
be07d7
+  /* With correct ulimit, etc. this should cause a core dump.  */
be07d7
+  abort ();
be07d7
+}