NetBSD-Bugs archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
bin/43332: gdbserver for powerpc
>Number: 43332
>Category: bin
>Synopsis: gdbserver for powerpc
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: bin-bug-people
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Fri May 21 07:20:00 +0000 2010
>Originator: Jan Kobler
>Release: 5.0.2
>Organization:
Kobler Systems GmbH
>Environment:
Linux 2.6.29-gentoo-r5 i686
>Description:
gdbserver is not built for netbsd, because gdbserver doesn't support netbsd.
I have added the support for netbsd powerpc to gdbserver.
When you follow the instructions in the sections How-To-Repeat and Fix, you get
a working gdbserver for netbsd powerpc.
>How-To-Repeat:
This section describes how to generate the native files for GDB on a target
platform and
how to build gdb for the target. Some bugs are encountered.
This is also the preparation for the bug fix later on.
Install the cross build environment for netbsd on the linux host:
MY_NETBSD=~/netbsd; export MY_NETBSD # define your own folder here
# create folders
mkdir ${MY_NETBSD} # root of the download, build, tools and
installation folders for netbsd
mkdir ${MY_NETBSD}/system # source sets are extracted into this folder,
contains usr/src and etc, netbsd build tree
mkdir ${MY_NETBSD}/obj # object dir
mkdir ${MY_NETBSD}/dest # destination dir
mkdir ${MY_NETBSD}/tools # tools running on the host
mkdir ${MY_NETBSD}/release # release dir
mkdir ${MY_NETBSD}/log # log files
mkdir ${MY_NETBSD}/download # download folder for the source sets
mkdir ${MY_NETBSD}/obj/evbppc # object dir of arch evbppc
mkdir ${MY_NETBSD}/dest/evbppc # destination dir of arch evbppc
# put source sets into ${MY_NETBSD}/download
# extract source sets into ${MY_NETBSD}/system
cd ${MY_NETBSD}/system
tar -xzf ${MY_NETBSD}/download/gnusrc.tgz
tar -xzf ${MY_NETBSD}/download/sharesrc.tgz
tar -xzf ${MY_NETBSD}/download/src.tgz
tar -xzf ${MY_NETBSD}/download/syssrc.tgz
tar -xzf ${MY_NETBSD}/download/xsrc.tgz
# Use the cross-compiler to generate the native files for GDB on a target
platform
# as described in the file ${MY_NETBSD}/usr/src/tools/gdb/README.mknative
# create the etc/mk.conf file
mkdir ${MY_NETBSD}/system/etc
# create and edit the file ${MY_NETBSD}/system/etc/mk.conf
# fill mk.conf with:
MKMAINTAINERTOOLS=yes
MKUNPRIVED=yes
#define environment variable MAKECONF
MAKECONF=${MY_NETBSD}/system/etc/mk.conf; export MAKECONF
# build tool chain
cd ${MY_NETBSD}/system/usr/src
./build.sh -U -u -m evbppc -D ${MY_NETBSD}/dest/evbppc -O
${MY_NETBSD}/obj/evbppc -R ${MY_NETBSD}/release -T ${MY_NETBSD}/tools -X
${MY_NETBSD}/system/usr/xsrc params tools 2>&1 | tee
${MY_NETBSD}/log/build.sh.tools.out
# toolchain has been built successfully
cd ${MY_NETBSD}/system/usr/src
${MY_NETBSD}/tools/bin/nbmake-evbppc do-distrib-dirs obj includes
cd ${MY_NETBSD}/system/usr/src/gnu/lib/crtstuff4
${MY_NETBSD}/tools/bin/nbmake-evbppc depend all install
cd ${MY_NETBSD}/system/usr/src/lib/csu
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
cd ${MY_NETBSD}/system/usr/src/gnu/lib/libgcc4
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
cd ${MY_NETBSD}/system/usr/src/lib
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
# stops with an error message:
/home/jan/netbsd/tools/bin/../lib/gcc/powerpc--netbsd/4.1.3/../../../../powerpc--netbsd/bin/ld:
cannot find -lc
collect2: ld returned 1 exit status
# reason: usr/src/lib/libc is missing
# bug fix: before building usr/src/lib build usr/src/lib/libc
# I have changed the file ${MY_NETBSD}usr/src/tools/gdb/README.mknative
accordingly
cd ${MY_NETBSD}/system/usr/src/lib/libc
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
# continue
cd ${MY_NETBSD}/system/usr/src/lib
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
# now it is built successfully
cd ${MY_NETBSD}/system/usr/src/tools/gdb
# remove files
${MY_NETBSD}/tools/bin/nbmake-evbppc clean
# create confg files for gdb
${MY_NETBSD}/tools/bin/nbmake-evbppc obj native-gdb 2>&1 | tee
${MY_NETBSD}/log/nbmake.gdb.out
#stops with error message:
/bin/sh:
/home/jan/netbsd/system/usr/src/tools/gdb/../../gnu/dist/gdb6/gdb/observer.sh:
Permission denied
ls -l ${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/observer.sh
# observer.sh is not executable
# this bug has been reported already in PR toolchain/43219
# bug fix:
chmod a+x ${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/observer.sh
# continue
${MY_NETBSD}/tools/bin/nbmake-evbppc obj native-gdb 2>&1 | tee
${MY_NETBSD}/log/nbmake.gdb.out
# writes to
# ${MY_NETBSD}/system/usr/src/tools/gdb/.native
# ${MY_NETBSD}/system/usr/src/gnu
Extracting GDB configury for a native toolchain.
MAKE=/home/jan/netbsd/tools/bin/nbmake /bin/sh
/home/jan/netbsd/system/usr/src/tools/gdb/mknative-gdb gdb
/home/jan/netbsd/obj/evbppc/tools/gdb/.native
/home/jan/netbsd/system/usr/src/tools/gdb/../.. powerpc--netbsd
gnu/usr.bin/gdb6/readline/arch/powerpc/defs.mk changed
# build gdb
cd ${MY_NETBSD}/system/usr/src/gnu/usr.bin/gdb6/
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
# gdb is built in ${MY_NETBSD}/obj/evbppc/gnu/usr.bin/gdb6/gdb
# and installed in ${MY_NETBSD}/dest/evbppc/usr/bin
ls -l ${MY_NETBSD}/obj/evbppc/gnu/usr.bin/gdb6/gdb
ls -l ${MY_NETBSD}/dest/evbppc/usr/bin/gdb*
# gdb is rebuilt successfully, but without gdbserver
ls -l ${MY_NETBSD}/obj/evbppc/gnu/usr.bin/gdb6
>Fix:
I have added the support for netbsd powerpc to gdbserver.
build.sh can build and install gdbserver now.
With the new macro MKGDBSERVER you can select whether gdbserver is built or not.
# apply patches
###############################################################
# start of patches #
###############################################################
diff -r -N -u --exclude='*.orig' --exclude '*~'
old/usr/src/gnu/dist/gdb6/gdb/configure.tgt
new/usr/src/gnu/dist/gdb6/gdb/configure.tgt
--- old/usr/src/gnu/dist/gdb6/gdb/configure.tgt 2007-11-18 13:45:25.000000000
+0100
+++ new/usr/src/gnu/dist/gdb6/gdb/configure.tgt 2010-05-20 01:26:27.000000000
+0200
@@ -148,7 +148,9 @@
mt-*-*) gdb_target=mt ;;
powerpc-*-netbsd* | powerpc-*-knetbsd*-gnu)
- gdb_target=nbsd ;;
+ gdb_target=nbsd
+ build_gdbserver=yes
+ ;;
powerpc-*-openbsd*) gdb_target=obsd ;;
powerpc-*-aix*) gdb_target=aix ;;
powerpc-*-linux*) gdb_target=linux
diff -r -N -u --exclude='*.orig' --exclude '*~'
old/usr/src/gnu/dist/gdb6/gdb/gdbserver/configure.srv
new/usr/src/gnu/dist/gdb6/gdb/gdbserver/configure.srv
--- old/usr/src/gnu/dist/gdb6/gdb/gdbserver/configure.srv 2006-07-02
22:16:23.000000000 +0200
+++ new/usr/src/gnu/dist/gdb6/gdb/gdbserver/configure.srv 2010-05-20
01:26:27.000000000 +0200
@@ -100,6 +100,10 @@
srv_linux_usrregs=yes
srv_linux_thread_db=yes
;;
+ powerpc-*-netbsd*) srv_regobj=reg-ppc.o
+ srv_tgtobj="nbsd-low.o nbsd-ppc-low.o"
+ srv_linux_regsets=yes
+ ;;
*) echo "Error: target not supported by gdbserver."
exit 1
;;
diff -r -N -u --exclude='*.orig' --exclude '*~'
old/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.c
new/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.c
--- old/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.c 1970-01-01
01:00:00.000000000 +0100
+++ new/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.c 2010-05-20
18:09:43.000000000 +0200
@@ -0,0 +1,1666 @@
+/* Low level interface to ptrace, for the remote server for GDB.
+ Copyright (C) 1995, 1996, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,
+ 2006
+ Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street, Fifth Floor,
+ Boston, MA 02110-1301, USA. */
+
+#include "server.h"
+#include "nbsd-low.h"
+
+#include <sys/wait.h>
+#include <stdio.h>
+#include <sys/param.h>
+#include <sys/dir.h>
+#include <sys/ptrace.h>
+#include <sys/user.h>
+#include <signal.h>
+#include <sys/ioctl.h>
+#include <fcntl.h>
+#include <string.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <sys/syscall.h>
+
+#ifdef HAVE_LINUX_USRREGS
+#error HAVE_LINUX_USRREGS is not supported
+#endif
+#define __SIGRTMIN 33
+
+/* ``all_threads'' is keyed by the LWP ID - it should be the thread ID instead,
+ however. This requires changing the ID in place when we go from
!using_threads
+ to using_threads, immediately.
+
+ ``all_processes'' is keyed by the process ID - which on Linux is (presently)
+ the same as the LWP ID. */
+
+struct inferior_list all_processes;
+
+/* FIXME this is a bit of a hack, and could be removed. */
+int stopping_threads;
+
+/* FIXME make into a target method? */
+int using_threads;
+
+static void linux_resume_one_process (struct inferior_list_entry *entry,
+ int step, int signal);
+static void linux_resume (struct thread_resume *resume_info);
+static void stop_all_processes (void);
+static int linux_wait_for_event (struct thread_info *child);
+
+struct pending_signals
+{
+ int signal;
+ struct pending_signals *prev;
+};
+
+#define PTRACE_ARG3_TYPE caddr_t
+#define PTRACE_XFER_TYPE int
+
+/* because CODE_ADDR is long long, to avoid compiler warning */
+#define PTRACE_CODE_ADDR_TO_ARG3 (int)
+#define PTRACE_ARG4_TYPE int
+
+#ifdef HAVE_LINUX_REGSETS
+static int use_regsets_p = 1;
+#endif
+
+int debug_threads = 0;
+
+#define pid_of(proc) ((proc)->head.id)
+
+/* FIXME: Delete eventually. */
+#define inferior_pid (pid_of (get_thread_process (current_inferior)))
+
+/* This function should only be called if the process got a SIGTRAP.
+ The SIGTRAP could mean several things.
+
+ On i386, where decr_pc_after_break is non-zero:
+ If we were single-stepping this process using PTRACE_SINGLESTEP,
+ we will get only the one SIGTRAP (even if the instruction we
+ stepped over was a breakpoint). The value of $eip will be the
+ next instruction.
+ If we continue the process using PTRACE_CONT, we will get a
+ SIGTRAP when we hit a breakpoint. The value of $eip will be
+ the instruction after the breakpoint (i.e. needs to be
+ decremented). If we report the SIGTRAP to GDB, we must also
+ report the undecremented PC. If we cancel the SIGTRAP, we
+ must resume at the decremented PC.
+
+ (Presumably, not yet tested) On a non-decr_pc_after_break machine
+ with hardware or kernel single-step:
+ If we single-step over a breakpoint instruction, our PC will
+ point at the following instruction. If we continue and hit a
+ breakpoint instruction, our PC will point at the breakpoint
+ instruction. */
+
+static CORE_ADDR
+get_stop_pc (void)
+{
+ CORE_ADDR stop_pc = (*the_low_target.get_pc) ();
+
+ if (get_thread_process (current_inferior)->stepping)
+ return stop_pc;
+ else
+ return stop_pc - the_low_target.decr_pc_after_break;
+}
+
+static void *
+add_process (unsigned long pid)
+{
+ struct process_info *process;
+
+ process = (struct process_info *) malloc (sizeof (*process));
+ memset (process, 0, sizeof (*process));
+
+ process->head.id = pid;
+
+ /* Default to tid == lwpid == pid. */
+ process->tid = pid;
+ process->lwpid = pid;
+
+ add_inferior_to_list (&all_processes, &process->head);
+
+ return process;
+}
+
+/* Start an inferior process and returns its pid.
+ ALLARGS is a vector of program-name and args. */
+
+static int
+linux_create_inferior (char *program, char **allargs)
+{
+ void *new_process;
+ int pid;
+
+#if defined(__UCLIBC__) && !defined(__UCLIBC_HAS_MMU__)
+ pid = vfork ();
+#else
+ pid = fork ();
+#endif
+ if (pid < 0)
+ perror_with_name ("fork");
+
+ if (pid == 0)
+ {
+ ptrace (PT_TRACE_ME, 0, 0, 0);
+
+ signal (__SIGRTMIN + 1, SIG_DFL);
+
+ setpgid (0, 0);
+
+ execv (program, allargs);
+
+ fprintf (stderr, "Cannot exec %s: %s.\n", program,
+ strerror (errno));
+ fflush (stderr);
+ _exit (0177);
+ }
+
+ new_process = add_process (pid);
+ add_thread (pid, new_process, pid);
+
+ return pid;
+}
+
+/* Attach to an inferior process. */
+
+void
+linux_attach_lwp (unsigned long pid, unsigned long tid)
+{
+ struct process_info *new_process;
+
+ if (ptrace (PT_ATTACH, pid, 0, 0) != 0)
+ {
+ fprintf (stderr, "Cannot attach to process %ld: %s (%d)\n", pid,
+ strerror (errno), errno);
+ fflush (stderr);
+
+ /* If we fail to attach to an LWP, just return. */
+ if (!using_threads)
+ _exit (0177);
+ return;
+ }
+
+ new_process = (struct process_info *) add_process (pid);
+ add_thread (tid, new_process, pid);
+
+ /* The next time we wait for this LWP we'll see a SIGSTOP as PT_ATTACH
+ brings it to a halt. We should ignore that SIGSTOP and resume the process
+ (unless this is the first process, in which case the flag will be cleared
+ in linux_attach).
+
+ On the other hand, if we are currently trying to stop all threads, we
+ should treat the new thread as if we had sent it a SIGSTOP. This works
+ because we are guaranteed that add_process added us to the end of the
+ list, and so the new thread has not yet reached wait_for_sigstop (but
+ will). */
+ if (! stopping_threads)
+ new_process->stop_expected = 1;
+}
+
+int
+linux_attach (unsigned long pid)
+{
+ struct process_info *process;
+
+ linux_attach_lwp (pid, pid);
+
+ /* Don't ignore the initial SIGSTOP if we just attached to this process. */
+ process = (struct process_info *) find_inferior_id (&all_processes, pid);
+ process->stop_expected = 0;
+
+ return 0;
+}
+
+/* Kill the inferior process. Make us have no inferior. */
+
+static void
+linux_kill_one_process (struct inferior_list_entry *entry)
+{
+ struct thread_info *thread = (struct thread_info *) entry;
+ struct process_info *process = get_thread_process (thread);
+ int wstat;
+
+ /* We avoid killing the first thread here, because of a Linux kernel (at
+ least 2.6.0-test7 through 2.6.8-rc4) bug; if we kill the parent before
+ the children get a chance to be reaped, it will remain a zombie
+ forever. */
+ if (entry == all_threads.head)
+ return;
+
+ do
+ {
+ ptrace (PT_KILL, pid_of (process), 0, 0);
+
+ /* Make sure it died. The loop is most likely unnecessary. */
+ wstat = linux_wait_for_event (thread);
+ } while (WIFSTOPPED (wstat));
+}
+
+static void
+linux_kill (void)
+{
+ struct thread_info *thread = (struct thread_info *) all_threads.head;
+ struct process_info *process = get_thread_process (thread);
+ int wstat;
+
+ for_each_inferior (&all_threads, linux_kill_one_process);
+
+ /* See the comment in linux_kill_one_process. We did not kill the first
+ thread in the list, so do so now. */
+ do
+ {
+ ptrace (PT_KILL, pid_of (process), 0, 0);
+
+ /* Make sure it died. The loop is most likely unnecessary. */
+ wstat = linux_wait_for_event (thread);
+ } while (WIFSTOPPED (wstat));
+}
+
+static void
+linux_detach_one_process (struct inferior_list_entry *entry)
+{
+ struct thread_info *thread = (struct thread_info *) entry;
+ struct process_info *process = get_thread_process (thread);
+
+ ptrace (PT_DETACH, pid_of (process), 0, 0);
+}
+
+static void
+linux_detach (void)
+{
+ for_each_inferior (&all_threads, linux_detach_one_process);
+}
+
+/* Return nonzero if the given thread is still alive. */
+static int
+linux_thread_alive (unsigned long tid)
+{
+ if (find_inferior_id (&all_threads, tid) != NULL)
+ return 1;
+ else
+ return 0;
+}
+
+/* Return nonzero if this process stopped at a breakpoint which
+ no longer appears to be inserted. Also adjust the PC
+ appropriately to resume where the breakpoint used to be. */
+static int
+check_removed_breakpoint (struct process_info *event_child)
+{
+ CORE_ADDR stop_pc;
+ struct thread_info *saved_inferior;
+
+ if (event_child->pending_is_breakpoint == 0)
+ return 0;
+
+ if (debug_threads)
+ fprintf (stderr, "Checking for breakpoint.\n");
+
+ saved_inferior = current_inferior;
+ current_inferior = get_process_thread (event_child);
+
+ stop_pc = get_stop_pc ();
+
+ /* If the PC has changed since we stopped, then we shouldn't do
+ anything. This happens if, for instance, GDB handled the
+ decr_pc_after_break subtraction itself. */
+ if (stop_pc != event_child->pending_stop_pc)
+ {
+ if (debug_threads)
+ fprintf (stderr, "Ignoring, PC was changed.\n");
+
+ event_child->pending_is_breakpoint = 0;
+ current_inferior = saved_inferior;
+ return 0;
+ }
+
+ /* If the breakpoint is still there, we will report hitting it. */
+ if ((*the_low_target.breakpoint_at) (stop_pc))
+ {
+ if (debug_threads)
+ fprintf (stderr, "Ignoring, breakpoint is still present.\n");
+ current_inferior = saved_inferior;
+ return 0;
+ }
+
+ if (debug_threads)
+ fprintf (stderr, "Removed breakpoint.\n");
+
+ /* For decr_pc_after_break targets, here is where we perform the
+ decrement. We go immediately from this function to resuming,
+ and can not safely call get_stop_pc () again. */
+ if (the_low_target.set_pc != NULL)
+ (*the_low_target.set_pc) (stop_pc);
+
+ /* We consumed the pending SIGTRAP. */
+ event_child->pending_is_breakpoint = 0;
+ event_child->status_pending_p = 0;
+ event_child->status_pending = 0;
+
+ current_inferior = saved_inferior;
+ return 1;
+}
+
+/* Return 1 if this process has an interesting status pending. This function
+ may silently resume an inferior process. */
+static int
+status_pending_p (struct inferior_list_entry *entry, void *dummy)
+{
+ struct process_info *process = (struct process_info *) entry;
+
+ if (process->status_pending_p)
+ if (check_removed_breakpoint (process))
+ {
+ /* This thread was stopped at a breakpoint, and the breakpoint
+ is now gone. We were told to continue (or step...) all threads,
+ so GDB isn't trying to single-step past this breakpoint.
+ So instead of reporting the old SIGTRAP, pretend we got to
+ the breakpoint just after it was removed instead of just
+ before; resume the process. */
+ linux_resume_one_process (&process->head, 0, 0);
+ return 0;
+ }
+
+ return process->status_pending_p;
+}
+
+static void
+linux_wait_for_process (struct process_info **childp, int *wstatp)
+{
+ int ret;
+ int to_wait_for = -1;
+
+ if (*childp != NULL)
+ to_wait_for = (*childp)->lwpid;
+
+ while (1)
+ {
+ ret = waitpid (to_wait_for, wstatp, WNOHANG);
+
+ if (ret == -1)
+ {
+ if (errno != ECHILD)
+ perror_with_name ("waitpid");
+ }
+ else if (ret > 0)
+ break;
+
+ ret = waitpid (to_wait_for, wstatp, WNOHANG | __WCLONE);
+
+ if (ret == -1)
+ {
+ if (errno != ECHILD)
+ perror_with_name ("waitpid (WCLONE)");
+ }
+ else if (ret > 0)
+ break;
+
+ usleep (1000);
+ }
+
+ if (debug_threads
+ && (!WIFSTOPPED (*wstatp)
+ || (WSTOPSIG (*wstatp) != 32
+ && WSTOPSIG (*wstatp) != 33)))
+ fprintf (stderr, "Got an event from %d (%x)\n", ret, *wstatp);
+
+ if (to_wait_for == -1)
+ *childp = (struct process_info *) find_inferior_id (&all_processes, ret);
+
+ (*childp)->stopped = 1;
+ (*childp)->pending_is_breakpoint = 0;
+
+ if (debug_threads
+ && WIFSTOPPED (*wstatp))
+ {
+ current_inferior = (struct thread_info *)
+ find_inferior_id (&all_threads, (*childp)->tid);
+ /* For testing only; i386_stop_pc prints out a diagnostic. */
+ if (the_low_target.get_pc != NULL)
+ get_stop_pc ();
+ }
+}
+
+static int
+linux_wait_for_event (struct thread_info *child)
+{
+ CORE_ADDR stop_pc;
+ struct process_info *event_child;
+ int wstat;
+
+ /* Check for a process with a pending status. */
+ /* It is possible that the user changed the pending task's registers since
+ it stopped. We correctly handle the change of PC if we hit a breakpoint
+ (in check_removed_breakpoint); signals should be reported anyway. */
+ if (child == NULL)
+ {
+ event_child = (struct process_info *)
+ find_inferior (&all_processes, status_pending_p, NULL);
+ if (debug_threads && event_child)
+ fprintf (stderr, "Got a pending child %ld\n", event_child->lwpid);
+ }
+ else
+ {
+ event_child = get_thread_process (child);
+ if (event_child->status_pending_p
+ && check_removed_breakpoint (event_child))
+ event_child = NULL;
+ }
+
+ if (event_child != NULL)
+ {
+ if (event_child->status_pending_p)
+ {
+ if (debug_threads)
+ fprintf (stderr, "Got an event from pending child %ld (%04x)\n",
+ event_child->lwpid, event_child->status_pending);
+ wstat = event_child->status_pending;
+ event_child->status_pending_p = 0;
+ event_child->status_pending = 0;
+ current_inferior = get_process_thread (event_child);
+ return wstat;
+ }
+ }
+
+ /* We only enter this loop if no process has a pending wait status. Thus
+ any action taken in response to a wait status inside this loop is
+ responding as soon as we detect the status, not after any pending
+ events. */
+ while (1)
+ {
+ if (child == NULL)
+ event_child = NULL;
+ else
+ event_child = get_thread_process (child);
+
+ linux_wait_for_process (&event_child, &wstat);
+
+ if (event_child == NULL)
+ error ("event from unknown child");
+
+ current_inferior = (struct thread_info *)
+ find_inferior_id (&all_threads, event_child->tid);
+
+ if (using_threads)
+ {
+ /* Check for thread exit. */
+ if (! WIFSTOPPED (wstat))
+ {
+ if (debug_threads)
+ fprintf (stderr, "Thread %ld (LWP %ld) exiting\n",
+ event_child->tid, event_child->head.id);
+
+ /* If the last thread is exiting, just return. */
+ if (all_threads.head == all_threads.tail)
+ return wstat;
+
+ dead_thread_notify (event_child->tid);
+
+ remove_inferior (&all_processes, &event_child->head);
+ free (event_child);
+ remove_thread (current_inferior);
+ current_inferior = (struct thread_info *) all_threads.head;
+
+ /* If we were waiting for this particular child to do something...
+ well, it did something. */
+ if (child != NULL)
+ return wstat;
+
+ /* Wait for a more interesting event. */
+ continue;
+ }
+
+ if (WIFSTOPPED (wstat)
+ && WSTOPSIG (wstat) == SIGSTOP
+ && event_child->stop_expected)
+ {
+ if (debug_threads)
+ fprintf (stderr, "Expected stop.\n");
+ event_child->stop_expected = 0;
+ linux_resume_one_process (&event_child->head,
+ event_child->stepping, 0);
+ continue;
+ }
+
+ /* FIXME drow/2002-06-09: Get signal numbers from the inferior's
+ thread library? */
+ if (WIFSTOPPED (wstat)
+ && (WSTOPSIG (wstat) == __SIGRTMIN
+ || WSTOPSIG (wstat) == __SIGRTMIN + 1))
+ {
+ if (debug_threads)
+ fprintf (stderr, "Ignored signal %d for %ld (LWP %ld).\n",
+ WSTOPSIG (wstat), event_child->tid,
+ event_child->head.id);
+ linux_resume_one_process (&event_child->head,
+ event_child->stepping,
+ WSTOPSIG (wstat));
+ continue;
+ }
+ }
+
+ /* If this event was not handled above, and is not a SIGTRAP, report
+ it. */
+ if (!WIFSTOPPED (wstat) || WSTOPSIG (wstat) != SIGTRAP)
+ return wstat;
+
+ /* If this target does not support breakpoints, we simply report the
+ SIGTRAP; it's of no concern to us. */
+ if (the_low_target.get_pc == NULL)
+ return wstat;
+
+ stop_pc = get_stop_pc ();
+
+ /* bp_reinsert will only be set if we were single-stepping.
+ Notice that we will resume the process after hitting
+ a gdbserver breakpoint; single-stepping to/over one
+ is not supported (yet). */
+ if (event_child->bp_reinsert != 0)
+ {
+ if (debug_threads)
+ fprintf (stderr, "Reinserted breakpoint.\n");
+ reinsert_breakpoint (event_child->bp_reinsert);
+ event_child->bp_reinsert = 0;
+
+ /* Clear the single-stepping flag and SIGTRAP as we resume. */
+ linux_resume_one_process (&event_child->head, 0, 0);
+ continue;
+ }
+
+ if (debug_threads)
+ fprintf (stderr, "Hit a (non-reinsert) breakpoint.\n");
+
+ if (check_breakpoints (stop_pc) != 0)
+ {
+ /* We hit one of our own breakpoints. We mark it as a pending
+ breakpoint, so that check_removed_breakpoint () will do the PC
+ adjustment for us at the appropriate time. */
+ event_child->pending_is_breakpoint = 1;
+ event_child->pending_stop_pc = stop_pc;
+
+ /* Now we need to put the breakpoint back. We continue in the event
+ loop instead of simply replacing the breakpoint right away,
+ in order to not lose signals sent to the thread that hit the
+ breakpoint. Unfortunately this increases the window where another
+ thread could sneak past the removed breakpoint. For the current
+ use of server-side breakpoints (thread creation) this is
+ acceptable; but it needs to be considered before this breakpoint
+ mechanism can be used in more general ways. For some breakpoints
+ it may be necessary to stop all other threads, but that should
+ be avoided where possible.
+
+ If breakpoint_reinsert_addr is NULL, that means that we can
+ use PT_STEP on this platform. Uninsert the breakpoint,
+ mark it for reinsertion, and single-step.
+
+ Otherwise, call the target function to figure out where we need
+ our temporary breakpoint, create it, and continue executing this
+ process. */
+ if (the_low_target.breakpoint_reinsert_addr == NULL)
+ {
+ event_child->bp_reinsert = stop_pc;
+ uninsert_breakpoint (stop_pc);
+ linux_resume_one_process (&event_child->head, 1, 0);
+ }
+ else
+ {
+ reinsert_breakpoint_by_bp
+ (stop_pc, (*the_low_target.breakpoint_reinsert_addr) ());
+ linux_resume_one_process (&event_child->head, 0, 0);
+ }
+
+ continue;
+ }
+
+ /* If we were single-stepping, we definitely want to report the
+ SIGTRAP. The single-step operation has completed, so also
+ clear the stepping flag; in general this does not matter,
+ because the SIGTRAP will be reported to the client, which
+ will give us a new action for this thread, but clear it for
+ consistency anyway. It's safe to clear the stepping flag
+ because the only consumer of get_stop_pc () after this point
+ is check_removed_breakpoint, and pending_is_breakpoint is not
+ set. It might be wiser to use a step_completed flag instead. */
+ if (event_child->stepping)
+ {
+ event_child->stepping = 0;
+ return wstat;
+ }
+
+ /* A SIGTRAP that we can't explain. It may have been a breakpoint.
+ Check if it is a breakpoint, and if so mark the process information
+ accordingly. This will handle both the necessary fiddling with the
+ PC on decr_pc_after_break targets and suppressing extra threads
+ hitting a breakpoint if two hit it at once and then GDB removes it
+ after the first is reported. Arguably it would be better to report
+ multiple threads hitting breakpoints simultaneously, but the current
+ remote protocol does not allow this. */
+ if ((*the_low_target.breakpoint_at) (stop_pc))
+ {
+ event_child->pending_is_breakpoint = 1;
+ event_child->pending_stop_pc = stop_pc;
+ }
+
+ return wstat;
+ }
+
+ /* NOTREACHED */
+ return 0;
+}
+
+/* Wait for process, returns status. */
+
+static unsigned char
+linux_wait (char *status)
+{
+ int w;
+ struct thread_info *child = NULL;
+
+retry:
+ /* If we were only supposed to resume one thread, only wait for
+ that thread - if it's still alive. If it died, however - which
+ can happen if we're coming from the thread death case below -
+ then we need to make sure we restart the other threads. We could
+ pick a thread at random or restart all; restarting all is less
+ arbitrary. */
+ if (cont_thread != 0 && cont_thread != -1)
+ {
+ child = (struct thread_info *) find_inferior_id (&all_threads,
+ cont_thread);
+
+ /* No stepping, no signal - unless one is pending already, of course. */
+ if (child == NULL)
+ {
+ struct thread_resume resume_info;
+ resume_info.thread = -1;
+ resume_info.step = resume_info.sig = resume_info.leave_stopped = 0;
+ linux_resume (&resume_info);
+ }
+ }
+
+ enable_async_io ();
+ unblock_async_io ();
+ w = linux_wait_for_event (child);
+ stop_all_processes ();
+ disable_async_io ();
+
+ /* If we are waiting for a particular child, and it exited,
+ linux_wait_for_event will return its exit status. Similarly if
+ the last child exited. If this is not the last child, however,
+ do not report it as exited until there is a 'thread exited' response
+ available in the remote protocol. Instead, just wait for another event.
+ This should be safe, because if the thread crashed we will already
+ have reported the termination signal to GDB; that should stop any
+ in-progress stepping operations, etc.
+
+ Report the exit status of the last thread to exit. This matches
+ LinuxThreads' behavior. */
+
+ if (all_threads.head == all_threads.tail)
+ {
+ if (WIFEXITED (w))
+ {
+ fprintf (stderr, "\nChild exited with retcode = %x \n", WEXITSTATUS
(w));
+ *status = 'W';
+ clear_inferiors ();
+ free (all_processes.head);
+ all_processes.head = all_processes.tail = NULL;
+ return ((unsigned char) WEXITSTATUS (w));
+ }
+ else if (!WIFSTOPPED (w))
+ {
+ fprintf (stderr, "\nChild terminated with signal = %x \n", WTERMSIG
(w));
+ *status = 'X';
+ clear_inferiors ();
+ free (all_processes.head);
+ all_processes.head = all_processes.tail = NULL;
+ return ((unsigned char) WTERMSIG (w));
+ }
+ }
+ else
+ {
+ if (!WIFSTOPPED (w))
+ goto retry;
+ }
+
+ *status = 'T';
+ return ((unsigned char) WSTOPSIG (w));
+}
+
+/* Send a signal to an LWP. For LinuxThreads, kill is enough; however, if
+ thread groups are in use, we need to use tkill. */
+
+static int
+kill_lwp (unsigned long lwpid, int signo)
+{
+ static int tkill_failed;
+
+ errno = 0;
+
+#ifdef SYS_tkill
+ if (!tkill_failed)
+ {
+ int ret = syscall (SYS_tkill, lwpid, signo);
+ if (errno != ENOSYS)
+ return ret;
+ errno = 0;
+ tkill_failed = 1;
+ }
+#endif
+
+ return kill (lwpid, signo);
+}
+
+static void
+send_sigstop (struct inferior_list_entry *entry)
+{
+ struct process_info *process = (struct process_info *) entry;
+
+ if (process->stopped)
+ return;
+
+ /* If we already have a pending stop signal for this process, don't
+ send another. */
+ if (process->stop_expected)
+ {
+ process->stop_expected = 0;
+ return;
+ }
+
+ if (debug_threads)
+ fprintf (stderr, "Sending sigstop to process %ld\n", process->head.id);
+
+ kill_lwp (process->head.id, SIGSTOP);
+ process->sigstop_sent = 1;
+}
+
+static void
+wait_for_sigstop (struct inferior_list_entry *entry)
+{
+ struct process_info *process = (struct process_info *) entry;
+ struct thread_info *saved_inferior, *thread;
+ int wstat;
+ unsigned long saved_tid;
+
+ if (process->stopped)
+ return;
+
+ saved_inferior = current_inferior;
+ saved_tid = ((struct inferior_list_entry *) saved_inferior)->id;
+ thread = (struct thread_info *) find_inferior_id (&all_threads,
+ process->tid);
+ wstat = linux_wait_for_event (thread);
+
+ /* If we stopped with a non-SIGSTOP signal, save it for later
+ and record the pending SIGSTOP. If the process exited, just
+ return. */
+ if (WIFSTOPPED (wstat)
+ && WSTOPSIG (wstat) != SIGSTOP)
+ {
+ if (debug_threads)
+ fprintf (stderr, "Stopped with non-sigstop signal\n");
+ process->status_pending_p = 1;
+ process->status_pending = wstat;
+ process->stop_expected = 1;
+ }
+
+ if (linux_thread_alive (saved_tid))
+ current_inferior = saved_inferior;
+ else
+ {
+ if (debug_threads)
+ fprintf (stderr, "Previously current thread died.\n");
+
+ /* Set a valid thread as current. */
+ set_desired_inferior (0);
+ }
+}
+
+static void
+stop_all_processes (void)
+{
+ stopping_threads = 1;
+ for_each_inferior (&all_processes, send_sigstop);
+ for_each_inferior (&all_processes, wait_for_sigstop);
+ stopping_threads = 0;
+}
+
+/* Resume execution of the inferior process.
+ If STEP is nonzero, single-step it.
+ If SIGNAL is nonzero, give it that signal. */
+
+static void
+linux_resume_one_process (struct inferior_list_entry *entry,
+ int step, int signal)
+{
+ struct process_info *process = (struct process_info *) entry;
+ struct thread_info *saved_inferior;
+
+ if (process->stopped == 0)
+ return;
+
+ /* If we have pending signals or status, and a new signal, enqueue the
+ signal. Also enqueue the signal if we are waiting to reinsert a
+ breakpoint; it will be picked up again below. */
+ if (signal != 0
+ && (process->status_pending_p || process->pending_signals != NULL
+ || process->bp_reinsert != 0))
+ {
+ struct pending_signals *p_sig;
+ p_sig = malloc (sizeof (*p_sig));
+ p_sig->prev = process->pending_signals;
+ p_sig->signal = signal;
+ process->pending_signals = p_sig;
+ }
+
+ if (process->status_pending_p && !check_removed_breakpoint (process))
+ return;
+
+ saved_inferior = current_inferior;
+ current_inferior = get_process_thread (process);
+
+ if (debug_threads)
+ fprintf (stderr, "Resuming process %ld (%s, signal %d, stop %s)\n",
inferior_pid,
+ step ? "step" : "continue", signal,
+ process->stop_expected ? "expected" : "not expected");
+
+ /* This bit needs some thinking about. If we get a signal that
+ we must report while a single-step reinsert is still pending,
+ we often end up resuming the thread. It might be better to
+ (ew) allow a stack of pending events; then we could be sure that
+ the reinsert happened right away and not lose any signals.
+
+ Making this stack would also shrink the window in which breakpoints are
+ uninserted (see comment in linux_wait_for_process) but not enough for
+ complete correctness, so it won't solve that problem. It may be
+ worthwhile just to solve this one, however. */
+ if (process->bp_reinsert != 0)
+ {
+ if (debug_threads)
+ fprintf (stderr, " pending reinsert at %08lx",
(long)process->bp_reinsert);
+ if (step == 0)
+ fprintf (stderr, "BAD - reinserting but not stepping.\n");
+ step = 1;
+
+ /* Postpone any pending signal. It was enqueued above. */
+ signal = 0;
+ }
+
+ check_removed_breakpoint (process);
+
+ if (debug_threads && the_low_target.get_pc != NULL)
+ {
+ fprintf (stderr, " ");
+ (*the_low_target.get_pc) ();
+ }
+
+ /* If we have pending signals, consume one unless we are trying to reinsert
+ a breakpoint. */
+ if (process->pending_signals != NULL && process->bp_reinsert == 0)
+ {
+ struct pending_signals **p_sig;
+
+ p_sig = &process->pending_signals;
+ while ((*p_sig)->prev != NULL)
+ p_sig = &(*p_sig)->prev;
+
+ signal = (*p_sig)->signal;
+ free (*p_sig);
+ *p_sig = NULL;
+ }
+
+ regcache_invalidate_one ((struct inferior_list_entry *)
+ get_process_thread (process));
+ errno = 0;
+ process->stopped = 0;
+ process->stepping = step;
+ ptrace (step ? PT_STEP : PT_CONTINUE, process->lwpid, (PTRACE_ARG3_TYPE)1,
signal);
+ /* netbsd: PT_CONTINUE: addr = 1: use current pc, addr != 1: use pc = addr */
+ /* linux: PTRACE_CONT: addr = 0: use current pc, addr != 0: use pc = addr */
+
+
+ current_inferior = saved_inferior;
+ if (errno)
+ perror_with_name ("ptrace");
+}
+
+static struct thread_resume *resume_ptr;
+
+/* This function is called once per thread. We look up the thread
+ in RESUME_PTR, and mark the thread with a pointer to the appropriate
+ resume request.
+
+ This algorithm is O(threads * resume elements), but resume elements
+ is small (and will remain small at least until GDB supports thread
+ suspension). */
+static void
+linux_set_resume_request (struct inferior_list_entry *entry)
+{
+ struct process_info *process;
+ struct thread_info *thread;
+ int ndx;
+
+ thread = (struct thread_info *) entry;
+ process = get_thread_process (thread);
+
+ ndx = 0;
+ while (resume_ptr[ndx].thread != -1 && resume_ptr[ndx].thread != entry->id)
+ ndx++;
+
+ process->resume = &resume_ptr[ndx];
+}
+
+/* This function is called once per thread. We check the thread's resume
+ request, which will tell us whether to resume, step, or leave the thread
+ stopped; and what signal, if any, it should be sent. For threads which
+ we aren't explicitly told otherwise, we preserve the stepping flag; this
+ is used for stepping over gdbserver-placed breakpoints. */
+
+static void
+linux_continue_one_thread (struct inferior_list_entry *entry)
+{
+ struct process_info *process;
+ struct thread_info *thread;
+ int step;
+
+ thread = (struct thread_info *) entry;
+ process = get_thread_process (thread);
+
+ if (process->resume->leave_stopped)
+ return;
+
+ if (process->resume->thread == -1)
+ step = process->stepping || process->resume->step;
+ else
+ step = process->resume->step;
+
+ linux_resume_one_process (&process->head, step, process->resume->sig);
+
+ process->resume = NULL;
+}
+
+/* This function is called once per thread. We check the thread's resume
+ request, which will tell us whether to resume, step, or leave the thread
+ stopped; and what signal, if any, it should be sent. We queue any needed
+ signals, since we won't actually resume. We already have a pending event
+ to report, so we don't need to preserve any step requests; they should
+ be re-issued if necessary. */
+
+static void
+linux_queue_one_thread (struct inferior_list_entry *entry)
+{
+ struct process_info *process;
+ struct thread_info *thread;
+
+ thread = (struct thread_info *) entry;
+ process = get_thread_process (thread);
+
+ if (process->resume->leave_stopped)
+ return;
+
+ /* If we have a new signal, enqueue the signal. */
+ if (process->resume->sig != 0)
+ {
+ struct pending_signals *p_sig;
+ p_sig = malloc (sizeof (*p_sig));
+ p_sig->prev = process->pending_signals;
+ p_sig->signal = process->resume->sig;
+ process->pending_signals = p_sig;
+ }
+
+ process->resume = NULL;
+}
+
+/* Set DUMMY if this process has an interesting status pending. */
+static int
+resume_status_pending_p (struct inferior_list_entry *entry, void *flag_p)
+{
+ struct process_info *process = (struct process_info *) entry;
+
+ /* Processes which will not be resumed are not interesting, because
+ we might not wait for them next time through linux_wait. */
+ if (process->resume->leave_stopped)
+ return 0;
+
+ /* If this thread has a removed breakpoint, we won't have any
+ events to report later, so check now. check_removed_breakpoint
+ may clear status_pending_p. We avoid calling check_removed_breakpoint
+ for any thread that we are not otherwise going to resume - this
+ lets us preserve stopped status when two threads hit a breakpoint.
+ GDB removes the breakpoint to single-step a particular thread
+ past it, then re-inserts it and resumes all threads. We want
+ to report the second thread without resuming it in the interim. */
+ if (process->status_pending_p)
+ check_removed_breakpoint (process);
+
+ if (process->status_pending_p)
+ * (int *) flag_p = 1;
+
+ return 0;
+}
+
+static void
+linux_resume (struct thread_resume *resume_info)
+{
+ int pending_flag;
+
+ /* Yes, the use of a global here is rather ugly. */
+ resume_ptr = resume_info;
+
+ for_each_inferior (&all_threads, linux_set_resume_request);
+
+ /* If there is a thread which would otherwise be resumed, which
+ has a pending status, then don't resume any threads - we can just
+ report the pending status. Make sure to queue any signals
+ that would otherwise be sent. */
+ pending_flag = 0;
+ find_inferior (&all_processes, resume_status_pending_p, &pending_flag);
+
+ if (debug_threads)
+ {
+ if (pending_flag)
+ fprintf (stderr, "Not resuming, pending status\n");
+ else
+ fprintf (stderr, "Resuming, no pending status\n");
+ }
+
+ if (pending_flag)
+ for_each_inferior (&all_threads, linux_queue_one_thread);
+ else
+ {
+ block_async_io ();
+ enable_async_io ();
+ for_each_inferior (&all_threads, linux_continue_one_thread);
+ }
+}
+
+#ifdef HAVE_LINUX_USRREGS
+
+int
+register_addr (int regnum)
+{
+ int addr;
+
+ if (regnum < 0 || regnum >= the_low_target.num_regs)
+ error ("Invalid register number %d.", regnum);
+
+ addr = the_low_target.regmap[regnum];
+
+ return addr;
+}
+
+/* Fetch one register. */
+static void
+fetch_register (int regno)
+{
+ CORE_ADDR regaddr;
+ int i, size;
+ char *buf;
+
+ if (regno >= the_low_target.num_regs)
+ return;
+ if ((*the_low_target.cannot_fetch_register) (regno))
+ return;
+
+ regaddr = register_addr (regno);
+ if (regaddr == -1)
+ return;
+ size = (register_size (regno) + sizeof (PTRACE_XFER_TYPE) - 1)
+ & - sizeof (PTRACE_XFER_TYPE);
+ buf = alloca (size);
+ for (i = 0; i < size; i += sizeof (PTRACE_XFER_TYPE))
+ {
+ errno = 0;
+ *(PTRACE_XFER_TYPE *) (buf + i) =
+ ptrace (PTRACE_PEEKUSER, inferior_pid, (PTRACE_ARG3_TYPE)
PTRACE_CODE_ADDR_TO_ARG3 regaddr, 0);
+ regaddr += sizeof (PTRACE_XFER_TYPE);
+ if (errno != 0)
+ {
+ /* Warning, not error, in case we are attached; sometimes the
+ kernel doesn't let us at the registers. */
+ char *err = strerror (errno);
+ char *msg = alloca (strlen (err) + 128);
+ sprintf (msg, "reading register %d: %s", regno, err);
+ error (msg);
+ goto error_exit;
+ }
+ }
+ if (the_low_target.left_pad_xfer
+ && register_size (regno) < sizeof (PTRACE_XFER_TYPE))
+ supply_register (regno, (buf + sizeof (PTRACE_XFER_TYPE)
+ - register_size (regno)));
+ else
+ supply_register (regno, buf);
+
+error_exit:;
+}
+
+/* Fetch all registers, or just one, from the child process. */
+static void
+usr_fetch_inferior_registers (int regno)
+{
+ if (regno == -1 || regno == 0)
+ for (regno = 0; regno < the_low_target.num_regs; regno++)
+ fetch_register (regno);
+ else
+ fetch_register (regno);
+}
+
+/* Store our register values back into the inferior.
+ If REGNO is -1, do this for all registers.
+ Otherwise, REGNO specifies which register (so we can save time). */
+static void
+usr_store_inferior_registers (int regno)
+{
+ CORE_ADDR regaddr;
+ int i, size;
+ char *buf;
+
+ if (regno >= 0)
+ {
+ if (regno >= the_low_target.num_regs)
+ return;
+
+ if ((*the_low_target.cannot_store_register) (regno) == 1)
+ return;
+
+ regaddr = register_addr (regno);
+ if (regaddr == -1)
+ return;
+ errno = 0;
+ size = (register_size (regno) + sizeof (PTRACE_XFER_TYPE) - 1)
+ & - sizeof (PTRACE_XFER_TYPE);
+ buf = alloca (size);
+ memset (buf, 0, size);
+ if (the_low_target.left_pad_xfer
+ && register_size (regno) < sizeof (PTRACE_XFER_TYPE))
+ collect_register (regno, (buf + sizeof (PTRACE_XFER_TYPE)
+ - register_size (regno)));
+ else
+ collect_register (regno, buf);
+ for (i = 0; i < size; i += sizeof (PTRACE_XFER_TYPE))
+ {
+ errno = 0;
+ ptrace (PTRACE_POKEUSER, inferior_pid, (PTRACE_ARG3_TYPE)
PTRACE_CODE_ADDR_TO_ARG3 regaddr,
+ *(PTRACE_XFER_TYPE *) (buf + i));
+ if (errno != 0)
+ {
+ if ((*the_low_target.cannot_store_register) (regno) == 0)
+ {
+ char *err = strerror (errno);
+ char *msg = alloca (strlen (err) + 128);
+ sprintf (msg, "writing register %d: %s",
+ regno, err);
+ error (msg);
+ return;
+ }
+ }
+ regaddr += sizeof (PTRACE_XFER_TYPE);
+ }
+ }
+ else
+ for (regno = 0; regno < the_low_target.num_regs; regno++)
+ usr_store_inferior_registers (regno);
+}
+#endif /* HAVE_LINUX_USRREGS */
+
+
+
+#ifdef HAVE_LINUX_REGSETS
+
+static int
+regsets_fetch_inferior_registers ()
+{
+ struct regset_info *regset;
+ int saw_general_regs = 0;
+
+ regset = target_regsets;
+
+ while (regset->size >= 0)
+ {
+ void *buf;
+ int res;
+
+ if (regset->size == 0)
+ {
+ regset ++;
+ continue;
+ }
+
+ buf = malloc (regset->size);
+ res = ptrace (regset->get_request, inferior_pid, (PTRACE_ARG3_TYPE)buf,
0);
+ /* arg3, arg4: netbsd: buf, 0 */
+ /* arg3, arg4: linux: 0, buf */
+
+ if (res < 0)
+ {
+ if (errno == EIO)
+ {
+ /* If we get EIO on the first regset, do not try regsets again.
+ If we get EIO on a later regset, disable that regset. */
+ if (regset == target_regsets)
+ {
+ use_regsets_p = 0;
+ free(buf);
+ return -1;
+ }
+ else
+ {
+ regset->size = 0;
+ continue;
+ }
+ }
+ else
+ {
+ char s[256];
+ sprintf (s, "ptrace(regsets_fetch_inferior_registers) PID=%ld",
+ inferior_pid);
+ perror (s);
+ }
+ }
+ else if (regset->type == GENERAL_REGS)
+ saw_general_regs = 1;
+ regset->store_function (buf);
+ free(buf);
+ regset ++;
+ }
+ if (saw_general_regs)
+ return 0;
+ else
+ return 1;
+}
+
+static int
+regsets_store_inferior_registers ()
+{
+ struct regset_info *regset;
+ int saw_general_regs = 0;
+
+ regset = target_regsets;
+
+ while (regset->size >= 0)
+ {
+ void *buf;
+ int res;
+
+ if (regset->size == 0)
+ {
+ regset ++;
+ continue;
+ }
+
+ buf = malloc (regset->size);
+
+ /* First fill the buffer with the current register set contents,
+ in case there are any items in the kernel's regset that are
+ not in gdbserver's regcache. */
+ res = ptrace (regset->get_request, inferior_pid, (PTRACE_ARG3_TYPE)buf,
0 );
+ /* arg3, arg4: netbsd: buf, 0 */
+ /* arg3, arg4: linux: 0, buf */
+
+
+ if (res == 0)
+ {
+ /* Then overlay our cached registers on that. */
+ regset->fill_function (buf);
+
+ /* Only now do we write the register set. */
+ res = ptrace (regset->set_request, inferior_pid,
(PTRACE_ARG3_TYPE)buf, 0);
+ /* arg3, arg4: netbsd: buf, 0 */
+ /* arg3, arg4: linux: 0, buf */
+ }
+
+ if (res < 0)
+ {
+ if (errno == EIO)
+ {
+ /* If we get EIO on the first regset, do not try regsets again.
+ If we get EIO on a later regset, disable that regset. */
+ if (regset == target_regsets)
+ {
+ use_regsets_p = 0;
+ free(buf);
+ return -1;
+ }
+ else
+ {
+ regset->size = 0;
+ continue;
+ }
+ }
+ else
+ {
+ perror ("Warning: ptrace(regsets_store_inferior_registers)");
+ }
+ }
+ else if (regset->type == GENERAL_REGS)
+ saw_general_regs = 1;
+ regset ++;
+ free (buf);
+ }
+ if (saw_general_regs)
+ return 0;
+ else
+ return 1;
+ return 0;
+}
+
+#endif /* HAVE_LINUX_REGSETS */
+
+
+void
+linux_fetch_registers (int regno)
+{
+#ifdef HAVE_LINUX_REGSETS
+ if (use_regsets_p)
+ {
+ if (regsets_fetch_inferior_registers () == 0)
+ return;
+ }
+#endif
+#ifdef HAVE_LINUX_USRREGS
+ usr_fetch_inferior_registers (regno);
+#endif
+}
+
+void
+linux_store_registers (int regno)
+{
+#ifdef HAVE_LINUX_REGSETS
+ if (use_regsets_p)
+ {
+ if (regsets_store_inferior_registers () == 0)
+ return;
+ }
+#endif
+#ifdef HAVE_LINUX_USRREGS
+ usr_store_inferior_registers (regno);
+#endif
+}
+
+
+/* Copy LEN bytes from inferior's memory starting at MEMADDR
+ to debugger memory starting at MYADDR. */
+
+static int
+linux_read_memory (CORE_ADDR memaddr, unsigned char *myaddr, int len)
+{
+ register int i;
+ /* Round starting address down to longword boundary. */
+ register CORE_ADDR addr = memaddr & -(CORE_ADDR) sizeof (PTRACE_XFER_TYPE);
+ /* Round ending address up; get number of longwords that makes. */
+ register int count
+ = (((memaddr + len) - addr) + sizeof (PTRACE_XFER_TYPE) - 1)
+ / sizeof (PTRACE_XFER_TYPE);
+ /* Allocate buffer of that many longwords. */
+ register PTRACE_XFER_TYPE *buffer
+ = (PTRACE_XFER_TYPE *) alloca (count * sizeof (PTRACE_XFER_TYPE));
+
+ /* Read all the longwords */
+ for (i = 0; i < count; i++, addr += sizeof (PTRACE_XFER_TYPE))
+ {
+ errno = 0;
+ buffer[i] = ptrace (PT_READ_I, inferior_pid, (PTRACE_ARG3_TYPE)
PTRACE_CODE_ADDR_TO_ARG3 addr, 0);
+ if (errno)
+ return errno;
+ }
+
+ /* Copy appropriate bytes out of the buffer. */
+ memcpy (myaddr, (char *) buffer + (memaddr & (sizeof (PTRACE_XFER_TYPE) -
1)), len);
+
+ return 0;
+}
+
+/* Copy LEN bytes of data from debugger memory at MYADDR
+ to inferior's memory at MEMADDR.
+ On failure (cannot write the inferior)
+ returns the value of errno. */
+
+static int
+linux_write_memory (CORE_ADDR memaddr, const unsigned char *myaddr, int len)
+{
+ register int i;
+ /* Round starting address down to longword boundary. */
+ register CORE_ADDR addr = memaddr & -(CORE_ADDR) sizeof (PTRACE_XFER_TYPE);
+ /* Round ending address up; get number of longwords that makes. */
+ register int count
+ = (((memaddr + len) - addr) + sizeof (PTRACE_XFER_TYPE) - 1) / sizeof
(PTRACE_XFER_TYPE);
+ /* Allocate buffer of that many longwords. */
+ register PTRACE_XFER_TYPE *buffer = (PTRACE_XFER_TYPE *) alloca (count *
sizeof (PTRACE_XFER_TYPE));
+ extern int errno;
+
+ if (debug_threads)
+ {
+ fprintf (stderr, "Writing %02x to %08lx\n", (unsigned)myaddr[0],
(long)memaddr);
+ }
+
+ /* Fill start and end extra bytes of buffer with existing memory data. */
+
+ buffer[0] = ptrace (PT_READ_I, inferior_pid,
+ (PTRACE_ARG3_TYPE) PTRACE_CODE_ADDR_TO_ARG3 addr, 0);
+
+ if (count > 1)
+ {
+ buffer[count - 1]
+ = ptrace (PT_READ_I, inferior_pid,
+ (PTRACE_ARG3_TYPE) PTRACE_CODE_ADDR_TO_ARG3 (addr + (count -
1)
+ * sizeof (PTRACE_XFER_TYPE)),
+ 0);
+ }
+
+ /* Copy data to be written over corresponding part of buffer */
+
+ memcpy ((char *) buffer + (memaddr & (sizeof (PTRACE_XFER_TYPE) - 1)),
myaddr, len);
+
+ /* Write the entire buffer. */
+
+ for (i = 0; i < count; i++, addr += sizeof (PTRACE_XFER_TYPE))
+ {
+ errno = 0;
+ ptrace (PT_WRITE_I, inferior_pid, (PTRACE_ARG3_TYPE)
PTRACE_CODE_ADDR_TO_ARG3 addr, buffer[i]);
+ if (errno)
+ return errno;
+ }
+
+ return 0;
+}
+
+static void
+linux_look_up_symbols (void)
+{
+#ifdef USE_THREAD_DB
+ if (using_threads)
+ return;
+
+ using_threads = thread_db_init ();
+#endif
+}
+
+static void
+linux_send_signal (int signum)
+{
+ extern unsigned long signal_pid;
+
+ if (cont_thread != 0 && cont_thread != -1)
+ {
+ struct process_info *process;
+
+ process = get_thread_process (current_inferior);
+ kill_lwp (process->lwpid, signum);
+ }
+ else
+ kill_lwp (signal_pid, signum);
+}
+
+/* Copy LEN bytes from inferior's auxiliary vector starting at OFFSET
+ to debugger memory starting at MYADDR. */
+
+static int
+linux_read_auxv (CORE_ADDR offset, unsigned char *myaddr, unsigned int len)
+{
+ char filename[PATH_MAX];
+ int fd, n;
+
+ snprintf (filename, sizeof filename, "/proc/%ld/auxv", inferior_pid);
+
+ fd = open (filename, O_RDONLY);
+ if (fd < 0)
+ return -1;
+
+ if (offset != (CORE_ADDR) 0
+ && lseek (fd, (off_t) offset, SEEK_SET) != (off_t) offset)
+ n = -1;
+ else
+ n = read (fd, myaddr, len);
+
+ close (fd);
+
+ return n;
+}
+
+/* These watchpoint related wrapper functions simply pass on the function call
+ if the target has registered a corresponding function. */
+
+static int
+linux_insert_watchpoint (char type, CORE_ADDR addr, int len)
+{
+ if (the_low_target.insert_watchpoint != NULL)
+ return the_low_target.insert_watchpoint (type, addr, len);
+ else
+ /* Unsupported (see target.h). */
+ return 1;
+}
+
+static int
+linux_remove_watchpoint (char type, CORE_ADDR addr, int len)
+{
+ if (the_low_target.remove_watchpoint != NULL)
+ return the_low_target.remove_watchpoint (type, addr, len);
+ else
+ /* Unsupported (see target.h). */
+ return 1;
+}
+
+static int
+linux_stopped_by_watchpoint (void)
+{
+ if (the_low_target.stopped_by_watchpoint != NULL)
+ return the_low_target.stopped_by_watchpoint ();
+ else
+ return 0;
+}
+
+static CORE_ADDR
+linux_stopped_data_address (void)
+{
+ if (the_low_target.stopped_data_address != NULL)
+ return the_low_target.stopped_data_address ();
+ else
+ return 0;
+}
+
+#if defined(__UCLIBC__) && !defined(__UCLIBC_HAS_MMU__)
+#if defined(__mcoldfire__)
+/* These should really be defined in the kernel's ptrace.h header. */
+#define PT_TEXT_ADDR 49*4
+#define PT_DATA_ADDR 50*4
+#define PT_TEXT_END_ADDR 51*4
+#endif
+
+/* Under uClinux, programs are loaded at non-zero offsets, which we need
+ to tell gdb about. */
+
+static int
+linux_read_offsets (CORE_ADDR *text_p, CORE_ADDR *data_p)
+{
+#if defined(PT_TEXT_ADDR) && defined(PT_DATA_ADDR) && defined(PT_TEXT_END_ADDR)
+ unsigned long text, text_end, data;
+ int pid = get_thread_process (current_inferior)->head.id;
+
+ errno = 0;
+
+ text = ptrace (PTRACE_PEEKUSER, pid, (long)PT_TEXT_ADDR, 0);
+ text_end = ptrace (PTRACE_PEEKUSER, pid, (long)PT_TEXT_END_ADDR, 0);
+ data = ptrace (PTRACE_PEEKUSER, pid, (long)PT_DATA_ADDR, 0);
+
+ if (errno == 0)
+ {
+ /* Both text and data offsets produced at compile-time (and so
+ used by gdb) are relative to the beginning of the program,
+ with the data segment immediately following the text segment.
+ However, the actual runtime layout in memory may put the data
+ somewhere else, so when we send gdb a data base-address, we
+ use the real data base address and subtract the compile-time
+ data base-address from it (which is just the length of the
+ text segment). BSS immediately follows data in both
+ cases. */
+ *text_p = text;
+ *data_p = data - (text_end - text);
+
+ return 1;
+ }
+#endif
+ return 0;
+}
+#endif
+
+static struct target_ops linux_target_ops = {
+ linux_create_inferior,
+ linux_attach,
+ linux_kill,
+ linux_detach,
+ linux_thread_alive,
+ linux_resume,
+ linux_wait,
+ linux_fetch_registers,
+ linux_store_registers,
+ linux_read_memory,
+ linux_write_memory,
+ linux_look_up_symbols,
+ linux_send_signal,
+ linux_read_auxv,
+ linux_insert_watchpoint,
+ linux_remove_watchpoint,
+ linux_stopped_by_watchpoint,
+ linux_stopped_data_address,
+#if defined(__UCLIBC__) && !defined(__UCLIBC_HAS_MMU__)
+ linux_read_offsets,
+#endif
+};
+
+static void
+linux_init_signals ()
+{
+ /* FIXME drow/2002-06-09: As above, we should check with LinuxThreads
+ to find what the cancel signal actually is. */
+ signal (__SIGRTMIN+1, SIG_IGN);
+}
+
+void
+initialize_low (void)
+{
+ using_threads = 0;
+ set_target_ops (&linux_target_ops);
+ set_breakpoint_data (the_low_target.breakpoint,
+ the_low_target.breakpoint_len);
+ init_registers ();
+ linux_init_signals ();
+}
diff -r -N -u --exclude='*.orig' --exclude '*~'
old/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.h
new/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.h
--- old/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.h 1970-01-01
01:00:00.000000000 +0100
+++ new/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.h 2010-05-20
01:26:27.000000000 +0200
@@ -0,0 +1,130 @@
+/* Internal interfaces for the GNU/Linux specific target code for gdbserver.
+ Copyright (C) 2002, 2004, 2005 Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street, Fifth Floor,
+ Boston, MA 02110-1301, USA. */
+
+#ifdef HAVE_LINUX_REGSETS
+typedef void (*regset_fill_func) (void *);
+typedef void (*regset_store_func) (const void *);
+enum regset_type {
+ GENERAL_REGS,
+ FP_REGS,
+ EXTENDED_REGS,
+};
+
+struct regset_info
+{
+ int get_request, set_request;
+ int size;
+ enum regset_type type;
+ regset_fill_func fill_function;
+ regset_store_func store_function;
+};
+extern struct regset_info target_regsets[];
+#endif
+
+struct linux_target_ops
+{
+ int num_regs;
+ int *regmap;
+ int (*cannot_fetch_register) (int);
+
+ /* Returns 0 if we can store the register, 1 if we can not
+ store the register, and 2 if failure to store the register
+ is acceptable. */
+ int (*cannot_store_register) (int);
+ CORE_ADDR (*get_pc) (void);
+ void (*set_pc) (CORE_ADDR newpc);
+ const unsigned char *breakpoint;
+ int breakpoint_len;
+ CORE_ADDR (*breakpoint_reinsert_addr) (void);
+
+
+ int decr_pc_after_break;
+ int (*breakpoint_at) (CORE_ADDR pc);
+
+ /* Watchpoint related functions. See target.h for comments. */
+ int (*insert_watchpoint) (char type, CORE_ADDR addr, int len);
+ int (*remove_watchpoint) (char type, CORE_ADDR addr, int len);
+ int (*stopped_by_watchpoint) (void);
+ CORE_ADDR (*stopped_data_address) (void);
+
+ /* Whether to left-pad registers for PEEKUSR/POKEUSR if they are smaller
+ than an xfer unit. */
+ int left_pad_xfer;
+};
+
+extern struct linux_target_ops the_low_target;
+
+#define get_process(inf) ((struct process_info *)(inf))
+#define get_thread_process(thr) (get_process (inferior_target_data (thr)))
+#define get_process_thread(proc) ((struct thread_info *) \
+ find_inferior_id (&all_threads, \
+ get_process (proc)->tid))
+
+struct process_info
+{
+ struct inferior_list_entry head;
+ int thread_known;
+ unsigned long lwpid;
+ unsigned long tid;
+
+ /* If this flag is set, the next SIGSTOP will be ignored (the process will
+ be immediately resumed). */
+ int stop_expected;
+
+ /* If this flag is set, the process is known to be stopped right now (stop
+ event already received in a wait()). */
+ int stopped;
+
+ /* If this flag is set, we have sent a SIGSTOP to this process and are
+ waiting for it to stop. */
+ int sigstop_sent;
+
+ /* If this flag is set, STATUS_PENDING is a waitstatus that has not yet
+ been reported. */
+ int status_pending_p;
+ int status_pending;
+
+ /* If this flag is set, the pending status is a (GDB-placed) breakpoint. */
+ int pending_is_breakpoint;
+ CORE_ADDR pending_stop_pc;
+
+ /* If this is non-zero, it is a breakpoint to be reinserted at our next
+ stop (SIGTRAP stops only). */
+ CORE_ADDR bp_reinsert;
+
+ /* If this flag is set, the last continue operation on this process
+ was a single-step. */
+ int stepping;
+
+ /* If this is non-zero, it points to a chain of signals which need to
+ be delivered to this process. */
+ struct pending_signals *pending_signals;
+
+ /* A link used when resuming. It is initialized from the resume request,
+ and then processed and cleared in linux_resume_one_process. */
+
+ struct thread_resume *resume;
+};
+
+extern struct inferior_list all_processes;
+
+void linux_attach_lwp (unsigned long pid, unsigned long tid);
+
+int thread_db_init (void);
diff -r -N -u --exclude='*.orig' --exclude '*~'
old/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-ppc-low.c
new/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-ppc-low.c
--- old/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-ppc-low.c 1970-01-01
01:00:00.000000000 +0100
+++ new/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-ppc-low.c 2010-05-20
18:05:17.000000000 +0200
@@ -0,0 +1,306 @@
+/* GNU/Linux/PowerPC specific low level interface, for the remote server for
+ GDB.
+ Copyright (C) 1995, 1996, 1998, 1999, 2000, 2001, 2002, 2005
+ Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ 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 2 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, write to the Free Software
+ Foundation, Inc., 51 Franklin Street, Fifth Floor,
+ Boston, MA 02110-1301, USA. */
+
+#include "server.h"
+#include "nbsd-low.h"
+
+#include <sys/ptrace.h>
+
+/* offset in ptrace register buffer, offset in 4byte steps */
+
+/* Data structures declared in netbsd/usr/src/sys/arch/powerpc/include/reg.h */
+
+/* struct reg */
+#define PT_R0 0
+#define PT_R1 1
+#define PT_R2 2
+#define PT_R3 3
+#define PT_R4 4
+#define PT_R5 5
+#define PT_R6 6
+#define PT_R7 7
+#define PT_R8 8
+#define PT_R9 9
+#define PT_R10 10
+#define PT_R11 11
+#define PT_R12 12
+#define PT_R13 13
+#define PT_R14 14
+#define PT_R15 15
+#define PT_R16 16
+#define PT_R17 17
+#define PT_R18 18
+#define PT_R19 19
+#define PT_R20 20
+#define PT_R21 21
+#define PT_R22 22
+#define PT_R23 23
+#define PT_R24 24
+#define PT_R25 25
+#define PT_R26 26
+#define PT_R27 27
+#define PT_R28 28
+#define PT_R29 29
+#define PT_R30 30
+#define PT_R31 31
+
+#define PT_LR 32
+#define PT_CR 33
+#define PT_XER 34
+#define PT_CTR 35
+#define PT_PC 36
+
+/* TODO: MSR is not defined in struct reg */
+#define PT_MSR 37
+
+/* struct fpreg */
+/* f0..f31, size 8bytes */
+#define PT_F0 0
+#define PT_F1 1
+#define PT_F2 2
+#define PT_F3 3
+#define PT_F4 4
+#define PT_F5 5
+#define PT_F6 6
+#define PT_F7 7
+#define PT_F8 8
+#define PT_F9 9
+#define PT_F10 10
+#define PT_F11 11
+#define PT_F12 12
+#define PT_F13 13
+#define PT_F14 14
+#define PT_F15 15
+#define PT_F16 16
+#define PT_F17 17
+#define PT_F18 18
+#define PT_F19 19
+#define PT_F20 20
+#define PT_F21 21
+#define PT_F22 22
+#define PT_F23 23
+#define PT_F24 24
+#define PT_F25 25
+#define PT_F26 26
+#define PT_F27 27
+#define PT_F28 28
+#define PT_F29 29
+#define PT_F30 30
+#define PT_F31 31
+
+/* fpscr: size 8byte, byte 0..3 invalid, byte 4..7: value of fpscr */
+#define PT_FPSCR 32
+/* fpscr is in reg-ppc.c only 4byte */
+
+/* number of regs in reg-ppc.c */
+#define ppc_num_regs 71
+
+#define ppc_size_gregs (38*4)
+/* TODO: sizeof(struct reg) is 37*4, with dummy msr at the end 38*4
+ buf is allocated in regsets_store_inferior_registers,
regsets_fetch_inferior_registers
+*/
+
+#define ppc_size_fpregs (33*8)
+
+/* sequence of registers same as in gdb/regformats/reg-ppc.dat */
+/* offset in ptrace register buffer, offset in byte */
+static int ppc_regmap[] =
+ {
+ /* struct reg */
+ PT_R0 * 4, PT_R1 * 4, PT_R2 * 4, PT_R3 * 4,
+ PT_R4 * 4, PT_R5 * 4, PT_R6 * 4, PT_R7 * 4,
+ PT_R8 * 4, PT_R9 * 4, PT_R10 * 4, PT_R11 * 4,
+ PT_R12 * 4, PT_R13 * 4, PT_R14 * 4, PT_R15 * 4,
+ PT_R16 * 4, PT_R17 * 4, PT_R18 * 4, PT_R19 * 4,
+ PT_R20 * 4, PT_R21 * 4, PT_R22 * 4, PT_R23 * 4,
+ PT_R24 * 4, PT_R25 * 4, PT_R26 * 4, PT_R27 * 4,
+ PT_R28 * 4, PT_R29 * 4, PT_R30 * 4, PT_R31 * 4,
+
+ /* struct fpreg */
+ PT_F0 * 8, PT_F1 * 8, PT_F2 * 8, PT_F3 * 8,
+ PT_F4 * 8, PT_F5 * 8, PT_F6 * 8, PT_F7 * 8,
+ PT_F8 * 8, PT_F9 * 8, PT_F10 * 8, PT_F11 * 8,
+ PT_F12 * 8, PT_F13 * 8, PT_F14 * 8, PT_F15 * 8,
+ PT_F16 * 8, PT_F17 * 8, PT_F18 * 8, PT_F19 * 8,
+ PT_F20 * 8, PT_F21 * 8, PT_F22 * 8, PT_F23 * 8,
+ PT_F24 * 8, PT_F25 * 8, PT_F26 * 8, PT_F27 * 8,
+ PT_F28 * 8, PT_F29 * 8, PT_F30 * 8, PT_F31 * 8,
+
+ /* struct reg */
+ PT_PC * 4, PT_MSR * 4, PT_CR * 4, PT_LR * 4,
+ PT_CTR * 4, PT_XER * 4,
+
+ /* struct fpreg */
+ PT_FPSCR * 8,
+ };
+
+static int
+ppc_cannot_store_register (int regno)
+{
+ /* Some kernels do not allow us to store fpscr. */
+ if (regno == find_regno ("fpscr"))
+ return 2;
+ if( regno == find_regno("msr")) // TODO: try to find msr value, saved as
srr0
+ return 2;
+
+ return 0;
+}
+
+static int
+ppc_cannot_fetch_register (int regno)
+{
+ if( regno == find_regno("msr")) // TODO: try to find msr value, saved as srr0
+ return 1;
+ return 0;
+}
+
+static CORE_ADDR
+ppc_get_pc (void)
+{
+ unsigned long pc;
+
+ collect_register_by_name ("pc", &pc);
+ return (CORE_ADDR) pc;
+}
+
+static void
+ppc_set_pc (CORE_ADDR pc)
+{
+ unsigned long newpc = pc;
+
+ supply_register_by_name ("pc", &newpc);
+}
+
+/* Correct in either endianness. Note that this file is
+ for PowerPC only, not PowerPC64.
+ This instruction is "twge r2, r2", which GDB uses as a software
+ breakpoint. */
+static const unsigned long ppc_breakpoint = 0x7d821008;
+#define ppc_breakpoint_len 4
+
+static int
+ppc_breakpoint_at (CORE_ADDR where)
+{
+ unsigned long insn;
+
+ (*the_target->read_memory) (where, (unsigned char *) &insn, 4);
+ if (insn == ppc_breakpoint)
+ return 1;
+ /* If necessary, recognize more trap instructions here. GDB only uses the
+ one. */
+ return 0;
+}
+
+/* Provide only a fill function for the general register set. ps_lgetregs
+ will use this for NPTL support. */
+
+static void ppc_fill_gregset (void *buf)
+{
+ int i;
+ /* copy from regcache to buf */
+
+ for (i = 0; i < 32; i++) // r0..r31
+ collect_register (i, (char *) buf + ppc_regmap[i]);
+
+
+ for (i = 64; i < 65; i++) // pc
+ collect_register (i, (char *) buf + ppc_regmap[i]);
+
+ // skip i==65: msr is not defined !!
+
+ for (i = 66; i < 70; i++) // cr, lr, ctr, xer
+ collect_register (i, (char *) buf + ppc_regmap[i]);
+
+}
+
+static void ppc_store_gregset (const void *buf)
+{
+ int i;
+
+ // copy from buf to regcache */
+ for (i = 0; i < 32; i++) // r0..r31
+ supply_register (i, (char *) buf + ppc_regmap[i]);
+
+ for (i = 64; i < 65; i++) // pc
+ supply_register (i, (char *) buf + ppc_regmap[i]);
+
+ // skip i==65: msr is not defined !!
+
+ for (i = 66; i < 70; i++) // cr, lr, ctr, xer
+ supply_register (i, (char *) buf + ppc_regmap[i]);
+}
+
+static void ppc_fill_fpregset (void *buf)
+{
+ int i;
+ /* copy from regcache to buf */
+ for (i = 32; i < 64; i++) // f0..f31
+ collect_register (i, (char *) buf + ppc_regmap[i]);
+
+ i=71; // fpscr:
+ collect_register (i, (char *) buf + ppc_regmap[i] + 4);
+
+ // inf buf 8byte: byte 0..3 invalid, byte 4..7
+ // in regcache 4byte: byte 0..3
+
+}
+
+static void ppc_store_fpregset (const void *buf)
+{
+ int i;
+
+ // copy from buf to regcache */
+ for (i = 32; i < 64; i++) // f0..f31
+ supply_register (i, (char *) buf + ppc_regmap[i]);
+
+ i=71; // fpscr:
+ supply_register (i, (char *) buf + ppc_regmap[i] + 4);
+
+ // inf buf 8byte: byte 0..3 invalid, byte 4..7
+ // in regcache 4byte: byte 0..3
+
+}
+
+struct regset_info target_regsets[] = {
+ /* Data structures declared in netbsd/usr/src/sys/arch/powerpc/include/reg.h
*/
+ { PT_GETREGS, PT_SETREGS, ppc_size_gregs, GENERAL_REGS, ppc_fill_gregset,
ppc_store_gregset }, /* struct reg */
+ { PT_GETFPREGS, PT_SETFPREGS, ppc_size_fpregs, FP_REGS, ppc_fill_fpregset,
ppc_store_fpregset }, /* struct fpreg */
+ { 0, 0, -1, -1, NULL, NULL }
+};
+
+struct linux_target_ops the_low_target = {
+ ppc_num_regs,
+ ppc_regmap,
+ ppc_cannot_fetch_register,
+ ppc_cannot_store_register,
+ ppc_get_pc,
+ ppc_set_pc,
+ (const unsigned char *) &ppc_breakpoint,
+ ppc_breakpoint_len,
+ NULL, /* breakpoint_reinsert_addr */
+ 0, /* decr_pc_after_break */
+ ppc_breakpoint_at,
+ /* insert_watchpoint */
+ /* remove_watchpoint */
+ /* stpped_by_watchpoint */
+ /* stopped_data_address */
+};
--- old/usr/src/gnu/usr.bin/gdb6/gdbserver/Makefile 1970-01-01
01:00:00.000000000 +0100
+++ new/usr/src/gnu/usr.bin/gdb6/gdbserver/Makefile 2010-05-20
01:26:27.000000000 +0200
@@ -0,0 +1,36 @@
+# $NetBSD$
+# TODO: is this Makefile is correct, is anything missing
+
+.include "../Makefile.inc"
+
+
+TOP= ${NETBSDSRCDIR}/gnu
+DIST= ${TOP}/dist/gdb6
+
+.include "${.CURDIR}/arch/${MACHINE_ARCH}/defs.mk"
+
+PROG= gdbserver
+
+SRCS= ${G_OBS:.o=.c}
+GCPPFLAGS= ${G_INTERNAL_CFLAGS}
+
+CPPFLAGS+= -DHAVE_CONFIG_H -I${.CURDIR} \
+ -I${.CURDIR}/arch/${MACHINE_ARCH} \
+ -I${DIST}/gdb/gdbserver \
+ -I${DIST}/gdb/regformats \
+ -I${DIST}/include \
+ -I${DIST} \
+ -I${.CURDIR}/../bfd/arch/${MACHINE_ARCH} \
+ -I${DIST}/bfd \
+ ${GCPPFLAGS:M-D*}
+
+# define macro.
+CPPFLAGS.signals.c= -DGDBSERVER
+
+.include "../../Makefile.inc"
+
+.include <bsd.prog.mk>
+
+.PATH: ${DIST}/gdb/gdbserver ${.CURDIR}/arch/${MACHINE_ARCH}
${DIST}/gdb/signals
+
+
--- old/usr/src/gnu/usr.bin/gdb6/Makefile 2006-11-29 16:18:10.000000000
+0100
+++ new/usr/src/gnu/usr.bin/gdb6/Makefile 2010-05-20 01:26:27.000000000
+0200
@@ -14,6 +14,8 @@
SUBDIR+= gdbtui
.endif
-#SUBDIR+= gdbserver
+.if ${MKGDBSERVER:Uno} != "no"
+SUBDIR+= gdbserver
+.endif
.include <bsd.subdir.mk>
--- old/usr/src/tools/gdb/Makefile 2007-02-19 19:26:22.000000000 +0100
+++ new/usr/src/tools/gdb/Makefile 2010-05-20 01:26:27.000000000 +0200
@@ -72,7 +72,8 @@
/usr/bin/env ${MKNATIVE_CONFIGURE_PRESET} \
${MAKE} configure-host) && \
(cd ${.OBJDIR}/.native/bfd && ${MAKE} bfd.h bfdver.h) && \
- (cd ${.OBJDIR}/.native/gdb && ${MAKE} init.c version.c)
+ (cd ${.OBJDIR}/.native/gdb && ${MAKE} init.c version.c) && \
+ (cd ${.OBJDIR}/.native/gdb/gdbserver && ${MAKE} reg-ppc.c
reg-ppc64.c version.c)
@touch $@
clean: clean.native
--- old/usr/src/tools/gdb/mknative-gdb 2007-02-19 19:26:22.000000000 +0100
+++ new/usr/src/tools/gdb/mknative-gdb 2010-05-20 01:26:27.000000000 +0200
@@ -85,9 +85,6 @@
INTERNAL_CFLAGS LIBGDB_OBS SIM_OBS
} | write_mk gnu/usr.bin/gdb6/arch/$MACHINE_ARCH/defs.mk
-# getvars gdb/gdbserver/Makefile \
-# INTERNAL_CFLAGS OBS \
-# | write_mk gnu/usr.bin/gdb/arch/$MACHINE_ARCH/gdbserver.mk
write_c gnu/usr.bin/gdb6/arch/$MACHINE_ARCH/config.h \
< $_TMPDIR/gdb/config.h
@@ -128,6 +125,27 @@
esac
}
+##### gnu/usr.bin/gdb6/gdb/gdbserver #####
+
+get_gdb_gdbserver () {
+
+ case $MACHINE_ARCH in
+ powerpc*)
+ mkdir -p $_TOP/gnu/usr.bin/gdb6/gdbserver/arch/$MACHINE_ARCH
+ { getvars gdb/gdbserver/Makefile \
+ INTERNAL_CFLAGS OBS
+ } | write_mk
gnu/usr.bin/gdb6/gdbserver/arch/$MACHINE_ARCH/defs.mk
+ write_c gnu/usr.bin/gdb6/gdbserver/arch/$MACHINE_ARCH/config.h \
+ < $_TMPDIR/gdb/gdbserver/config.h
+ for f in reg-ppc version ; do
+ write_c
gnu/usr.bin/gdb6/gdbserver/arch/$MACHINE_ARCH/$f.c <$_TMPDIR/gdb/gdbserver/$f.c
+ done
+ ;;
+ *)
+ ;;
+ esac
+}
+
##### main #####
case $1 in
@@ -137,6 +155,7 @@
get_gdb_libopcodes
get_gdb_libiberty
get_gdb_libreadline
+ get_gdb_gdbserver
exit 0
;;
--- old/usr/src/tools/gdb/README.mknative 2006-11-24 00:33:56.000000000
+0100
+++ new/usr/src/tools/gdb/README.mknative 2010-05-20 01:26:27.000000000
+0200
@@ -26,7 +26,7 @@
4. In src/gnu/lib/crtstuff4 do "nbmake-MACHINE depend all install"
-5. In src/lib/csu, src/gnu/lib/libgcc4, and src/lib, do
+5. In src/lib/csu, src/gnu/lib/libgcc4, src/libc, and src/lib, do
"nbmake-MACHINE all install".
6. In src/tools/gdb, do "nbmake-MACHINE obj native-gdb".
--- old/usr/src/share/man/man5/mk.conf.5 2009-01-16 22:41:10.000000000
+0100
+++ new/usr/src/share/man/man5/mk.conf.5 2010-05-20 01:26:27.000000000
+0200
@@ -343,6 +343,15 @@
is built.
.DFLTy
.
+.It Sy MKGDBSERVER
+.YorN
+Indicates whether
+.Xr gdbserver 1
+is built, when
+.Xr gdb 1
+is built.
+.DFLTn
+.
.It Sy MKHESIOD
.YorN
Indicates whether the Hesiod infrastructure
--- old/usr/src/share/mk/bsd.own.mk 2009-03-18 06:39:06.000000000 +0100
+++ new/usr/src/share/mk/bsd.own.mk 2010-05-20 01:26:27.000000000 +0200
@@ -529,6 +529,13 @@
.endif
#
+# gdbserver only on powerpc
+#
+.if ${MACHINE_ARCH} != "powerpc"
+MKGDBSERVER= no
+.endif
+
+#
# Location of the file that contains the major and minor numbers of the
# version of a shared library. If this file exists a shared library
# will be built by <bsd.lib.mk>.
@@ -665,6 +672,7 @@
#
.for var in \
MKCRYPTO_IDEA MKCRYPTO_MDC2 MKCRYPTO_RC5 MKDEBUG MKDEBUGLIB \
+ MKGDBSERVER \
MKMANZ MKMODULAR MKOBJDIRS \
MKPCC MKPCCCMDS \
MKSOFTFLOAT MKSTRIPIDENT \
--- old/usr/src/share/mk/bsd.README 2009-06-07 00:10:49.000000000 +0200
+++ new/usr/src/share/mk/bsd.README 2010-05-20 01:26:27.000000000 +0200
@@ -172,6 +172,9 @@
MKGDB If "no", don't build gdb(1).
Default: yes
+MKGDBSERVER If "no", don't build gdbserver(1), when gdb(1) is built.
+ Default: no
+
MKHESIOD If "no", disables building of Hesiod infrastructure
(libraries and support programs).
Default: yes
--- old/usr/src/distrib/sets/lists/comp/mi 2009-07-14 21:35:22.000000000
+0200
+++ new/usr/src/distrib/sets/lists/comp/mi 2010-05-20 01:26:27.000000000
+0200
@@ -19673,3 +19673,8 @@
./usr/share/misc/operator comp-c-share share
./usr/share/misc/style comp-c-share share
./var/db/libc.tags comp-c-root
+./usr/bin/gdbserver comp-debug-bin
gdb,gdb=6,gdbserver
+./usr/share/man/man1/gdbserver.1 comp-debug-man
gdb,gdb=6,gdbserver,.man
+./usr/share/man/cat1/gdbserver.0 comp-debug-catman
gdb,gdb=6,gdbserver,.cat
+./usr/share/man/html1/gdbserver.html comp-debug-htmlman
gdb,gdb=6,gdbserver,html
+./usr/libdata/debug/usr/bin/gdbserver.debug comp-debug-debug
gdb,gdb=6,gdbserver,debug
--- old/usr/src/distrib/sets/sets.subr 2009-04-08 02:09:27.000000000 +0200
+++ new/usr/src/distrib/sets/sets.subr 2010-05-20 01:26:27.000000000 +0200
@@ -37,6 +37,7 @@
MKGCC \
MKGCCCMDS \
MKGDB \
+ MKGDBSERVER \
MKHESIOD \
MKHTML \
MKINET6 \
@@ -263,6 +264,7 @@
# gcc ${MKGCC} != no
# gcccmds ${MKGCCCMDS} != no
# gdb ${MKGDB} != no
+# gdbserver ${MKGDBSERVER} != no
# hesiod ${MKHESIOD} != no
# html ${MKHTML} != no
# inet6 ${MKINET6} != no
--- old/usr/src/etc/Makefile 2009-10-15 10:21:05.000000000 +0200
+++ new/usr/src/etc/Makefile 2010-05-20 01:26:27.000000000 +0200
@@ -187,6 +187,7 @@
MKCRYPTO MKCRYPTO_IDEA MKCRYPTO_MDC2 MKCRYPTO_RC5 MKCVS \
MKDEBUG MKDEBUGLIB MKDOC MKDYNAMICROOT \
MKGCC MKGCCCMDS MKGDB \
+ MKGDBSERVER \
MKHESIOD MKHOSTOBJ MKHTML MKIEEEFP MKINET6 MKINFO MKIPFILTER \
MKKERBEROS MKLDAP MKLINKLIB MKLINT \
MKMAN MKMANZ MKNLS MKNVI MKOBJ MKOBJDIRS \
--- old/usr/src/build.sh 2009-03-18 06:39:06.000000000 +0100
+++ new/usr/src/build.sh 2010-05-20 01:26:27.000000000 +0200
@@ -1014,6 +1014,11 @@
exit 1
fi
+ # MKGDBSERVER is used in usr/src/gnu/usr.bin/gdb6/Makefile
+ makeenv="${makeenv} MKGDBSERVER"
+ MKGDBSERVER=$(getmakevar MKGDBSERVER)
+ export MKGDBSERVER
+
# Normalise MKOBJDIRS, MKUNPRIVED, and MKUPDATE
# These may be set as build.sh options or in "mk.conf".
# Don't export them as they're only used for tests in build.sh.
###############################################################
# end of patches #
###############################################################
####### new files in patch above ########
The new files are
${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.c
${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.h
${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-ppc-low.c
You can compare these files to the files they are based on:
diff ${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/linux-low.c
${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.c
diff ${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/linux-low.h
${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-low.h
diff ${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/linux-ppc-low.c
${MY_NETBSD}/system/usr/src/gnu/dist/gdb6/gdb/gdbserver/nbsd-ppc-low.c
I have changed as few code as possible in these new files in order to be able
to compare it to the files they are based on.
TODO: In the long run comments and function names should be adjusted to reflect
netbsd.
####### cross-build gdb and gdbserver ########
After applying all the patches above to the files in ${MY_NETBSD}/system
it is possible to build netbsd for evbppc on a x86 linux host.
IMPORTANT: At first do the steps in the How-To-Repeat section as a preparation!
Build gdb as described in the file
${MY_NETBSD}/usr/src/tools/gdb/README.mknative:
# edit file ${MY_NETBSD}/system/etc/mk.conf
# fill mk.conf with:
MKMAINTAINERTOOLS=yes
MKUNPRIVED=yes
MKGDBSERVER=yes
# New macro MKGDBSERVER=yes build gdbserver, when gdb is built.
# update nbmake-evbppc
cd ${MY_NETBSD}/system/usr/src
./build.sh -U -u -m evbppc -D ${MY_NETBSD}/dest/evbppc -O
${MY_NETBSD}/obj/evbppc -R ${MY_NETBSD}/release -T ${MY_NETBSD}/tools -X
${MY_NETBSD}/system/usr/xsrc params
cd ${MY_NETBSD}/system/usr/src/tools/gdb
# remove all files
${MY_NETBSD}/tools/bin/nbmake-evbppc clean
# create confg files for gdb
${MY_NETBSD}/tools/bin/nbmake-evbppc obj native-gdb 2>&1 | tee
${MY_NETBSD}/log/nbmake.gdb.out
# writes to
# ${MY_NETBSD}/obj/evbppc/tools/gdb/.native
# ${MY_NETBSD}/system/usr/src/gnu/usr.bin/gdb6
Extracting GDB configury for a native toolchain.
MAKE=/home/jan/netbsd/tools/bin/nbmake /bin/sh
/home/jan/netbsd/system/usr/src/tools/gdb/mknative-gdb gdb
/home/jan/netbsd/obj/evbppc/tools/gdb/.native
/home/jan/netbsd/system/usr/src/tools/gdb/../.. powerpc--netbsd
gnu/usr.bin/gdb6/gdbserver/arch/powerpc/defs.mk changed
gnu/usr.bin/gdb6/gdbserver/arch/powerpc/config.h changed
gnu/usr.bin/gdb6/gdbserver/arch/powerpc/reg-ppc.c changed
gnu/usr.bin/gdb6/gdbserver/arch/powerpc/version.c changed
# workaround:
${MY_NETBSD}/system/usr/src/gnu/usr.bin/gdb6/Makefile uses MKGDBSERVER to
decide,
if gdbserver should be built or not.
Because ${MY_NETBSD}/system/usr/src/gnu/usr.bin/gdb6/Makefile is not including
${MY_NETBSD}/system/usr/src/share/mk/bsd.own.mk,
the file ${MAKECONF} is also not included.
Even if MKGDBSERVER is defined in $MAKECONF,
${MY_NETBSD}/system/usr/src/gnu/usr.bin/gdb6/Makefile
will not know about it.
Because I don't know, if bsd.own.mk may be included in this Makefile, I have
added MKGDBSERVER to
${MY_NETBSD}/system/usr/src/build.sh instead.
${MY_NETBSD}/tools/bin/nbmake-evbppc will be created with a defintion of
MKGDBSERVER.
# create folder
cd ${MY_NETBSD}/system/usr/src
${MY_NETBSD}/tools/bin/nbmake-evbppc do-distrib-dirs obj includes
# folder ${MY_NETBSD}/obj/evbppc/gnu/usr.bin/gdb6/gdbserver is created
# build gdb and gdbserver
cd ${MY_NETBSD}/system/usr/src/gnu/usr.bin/gdb6/
${MY_NETBSD}/tools/bin/nbmake-evbppc all install
# gdbserver is built in ${MY_NETBSD}/obj/evbppc/gnu/usr.bin/gdb6/gdbserver
# and installed in ${MY_NETBSD}/dest/evbppc/usr/bin
# and ${MY_NETBSD}/dest/evbppc/usr/share/man/man1
ls -l ${MY_NETBSD}/obj/evbppc/gnu/usr.bin/gdb6/gdbserver
ls -l ${MY_NETBSD}/dest/evbppc/usr/bin/gdbserver
ls -l ${MY_NETBSD}/dest/evbppc/usr/share/man/man1/gdbserver.1
# edit file ${MY_NETBSD}/system/etc/mk.conf
# fill mk.conf with:
# MKMAINTAINERTOOLS=yes
MKUNPRIVED=yes
MKGDBSERVER=yes
# build userland
cd ${MY_NETBSD}/system/usr/src
./build.sh -U -u -m evbppc -D ${MY_NETBSD}/dest/evbppc -O
${MY_NETBSD}/obj/evbppc -R ${MY_NETBSD}/release -T ${MY_NETBSD}/tools -X
${MY_NETBSD}/system/usr/xsrc params release 2>&1 | tee
${MY_NETBSD}/log/build.sh.release.out
===> Summary of results:
build.sh command: ./build.sh -U -u -m evbppc -D
/home/jan/netbsd/dest/evbppc -O /home/jan/netbsd/obj/evbppc -R
/home/jan/netbsd/release -T /home/jan/netbsd/tools -X
/home/jan/netbsd/system/usr/xsrc params release
build.sh started: Wed May 19 14:35:15 CEST 2010
NetBSD version: 5.0.2
MACHINE: evbppc
MACHINE_ARCH: powerpc
Build platform: Linux 2.6.29-gentoo-r5 i686
HOST_SH: /bin/sh
TOOLDIR path: /home/jan/netbsd/tools
DESTDIR path: /home/jan/netbsd/dest/evbppc
RELEASEDIR path: /home/jan/netbsd/release
makewrapper: /home/jan/netbsd/tools/bin/nbmake-evbppc
Updated /home/jan/netbsd/tools/bin/nbmake-evbppc
Successful make params
Successful make release
build.sh ended: Wed May 19 18:36:45 CEST 2010
===> .
# After applying the patches in PR port-powerpc/43196 and PR toolchain/43219
too,
# I have been able to start the gdbserver on a powerpc board running netbsd and
debug remotely from a linux host.
Home |
Main Index |
Thread Index |
Old Index