Re: question about ARM watchpoints

[Pedro Alves <palves@redhat.com>]

On 09/05/2014 10:46 AM, Pedro Alves wrote:

> Like this?  This "hopefully" is failing on ARMv7-M currently.
> 
> Passes on x86-64 native and gdbserver.
> 
> Also uploaded to:
> 
>   git@github.com:palves/gdb.git palves/hw_watch_test

Hey guys, I guess I should just push it in (and post it to gdb-patches) ?

> 
> ----------->8-----------
> From e4035b618fdca66038bae8fd51b14cc2913f6298 Mon Sep 17 00:00:00 2001
> From: Pedro Alves <palves@redhat.com>
> Date: Fri, 5 Sep 2014 10:29:08 +0100
> Subject: [PATCH] Add test to make sure GDB knows which "kind" of watchpoint
>  the target has
> 
> This adds a test that makes sure GDB knows whether the target has
> continuable, or non-continuable watchpoints.
> 
> That is, the test confirms that GDB presents a watchpoint value change
> at the first instruction right after the instruction that changes
> memory.
> 
> gdb/testsuite/ChangeLog:
> 
> 2014-09-05  Pedro Alves  <palves@redhat.com>
> 
> 	* gdb.base/watchpoint-stops-at-right-insn.c: New file.
> 	* gdb.base/watchpoint-stops-at-right-insn.exp: New file.
> ---
>  .../gdb.base/watchpoint-stops-at-right-insn.c      |  33 ++++
>  .../gdb.base/watchpoint-stops-at-right-insn.exp    | 177 +++++++++++++++++++++
>  2 files changed, 210 insertions(+)
>  create mode 100644 gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.c
>  create mode 100644 gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.exp
> 
> diff --git a/gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.c b/gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.c
> new file mode 100644
> index 0000000..005d63f
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.c
> @@ -0,0 +1,33 @@
> +/* This testcase is part of GDB, the GNU debugger.
> +
> +   Copyright 2014 Free Software Foundation, Inc.
> +
> +   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 3 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, see <http://www.gnu.org/licenses/>.  */
> +
> +volatile int global;
> +
> +void
> +set_global (int val)
> +{
> +  global = val;
> +}
> +
> +int
> +main (void)
> +{
> +  set_global (1);
> +  set_global (2);
> +
> +  return 0;
> +}
> diff --git a/gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.exp b/gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.exp
> new file mode 100644
> index 0000000..9ad6080
> --- /dev/null
> +++ b/gdb/testsuite/gdb.base/watchpoint-stops-at-right-insn.exp
> @@ -0,0 +1,177 @@
> +# This testcase is part of GDB, the GNU debugger.
> +
> +# Copyright 2014 Free Software Foundation, Inc.
> +
> +# 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 3 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, see <http://www.gnu.org/licenses/>.
> +
> +# This file is part of the gdb testsuite.
> +
> +# Test that GDB presents a hardware watchpoint stop at the first
> +# instruction right after the instruction that changes memory.
> +#
> +# Some targets trigger a hardware watchpoint after the instruction
> +# that wrote memory executes, thus with the memory already changed and
> +# the PC pointing to the instruction after the instruction that wrote
> +# to memory.  These targets are said to have "continuable"
> +# watchpoints, referring to the fact that to make progress after the
> +# watchpoint triggers, GDB just needs to continue the target.
> +#
> +# Other targets trigger a hardware watchpoint at the instruction which
> +# has attempted to write to the piece of memory under control of the
> +# watchpoint, with the instruction actually not executed yet.  To be
> +# able to check whether the watched value changed, GDB needs to
> +# complete the memory write, single-stepping the target once.  These
> +# targets are said to have "non-continuable" watchpoints.
> +#
> +# This test makes sure that GDB knows which kind of watchpoint the
> +# target has, using this sequence of steps:
> +#
> +# 1 - run to main
> +#
> +# 2 - set a software watchpoint
> +#
> +# 3 - continue until watchpoint triggers
> +#
> +# 4 - the PC now points to the instruction right after the instruction
> +#     that actually caused the memory write.  So this is the address a
> +#     hardware watchpoint should present the stop to the user too.
> +#     Store the PC address.
> +#
> +# 5 - replace the software watchpoint by a hardware watchpoint
> +#
> +# 6 - continue until hardware watchpoint triggers
> +#
> +# 7 - the PC must point to the same address the software watchpoint
> +#     triggered at.
> +#
> +# If the target has continuable watchpoints, but GDB thinks it has
> +# non-continuable watchpoints, GDB will stop the inferior two
> +# instructions after the watched value change, rather than at the next
> +# instruction.
> +#
> +# If the target has non-continuable watchpoints, while GDB thinks it
> +# has continuable watchpoints, GDB will see a watchpoint trigger,
> +# notice no value changed, and immediatly continue the target.  Now,
> +# either the target manages to step-over the watchpoint transparently,
> +# and GDB thus fails to present to value change to the user, or, the
> +# watchpoint will keep re-triggering, with the program never making
> +# any progress.
> +
> +standard_testfile
> +
> +# No use testing this if we can't use hardware watchpoints.
> +if {[target_info exists gdb,no_hardware_watchpoints]} {
> +    return -1
> +}
> +
> +if { [prepare_for_testing ${testfile}.exp ${testfile} ${srcfile}] } {
> +    untested ${testfile}.exp
> +    return -1
> +}
> +
> +if { ![runto main] } then {
> +    fail "run to main"
> +    return
> +}
> +
> +# Get the current PC.  TEST is used as test prefix.
> +
> +proc get_pc {test} {
> +    global hex gdb_prompt
> +
> +    set addr ""
> +    gdb_test_multiple "p /x \$pc" "$test" {
> +	-re " = ($hex).*$gdb_prompt $" {
> +	    set addr $expect_out(1,string)
> +	    pass "$test"
> +	}
> +    }
> +
> +    return $addr
> +}
> +
> +# So we get an immediate warning/error if the target doesn't support a
> +# given watchpoint type.
> +gdb_test_no_output "set breakpoint always-inserted on"
> +
> +set hw_watchpoints_supported 0
> +
> +set test "set probe hw watchpoint"
> +gdb_test_multiple "watch global" $test {
> +    -re "You may have requested too many.*$gdb_prompt $" {
> +	pass $test
> +    }
> +    -re "Target does not support.*$gdb_prompt $" {
> +	pass $test
> +    }
> +    -re "$gdb_prompt $" {
> +	pass $test
> +	set hw_watchpoints_supported 1
> +    }
> +}
> +
> +if {!$hw_watchpoints_supported} {
> +    unsupported "no hw watchpoints support"
> +    return
> +}
> +
> +delete_breakpoints
> +
> +proc test {always_inserted} {
> +    global srcfile binfile
> +
> +    with_test_prefix "always-inserted $always_inserted" {
> +
> +	clean_restart $binfile
> +
> +	if { ![runto main] } then {
> +	    fail "run to main"
> +	    return
> +	}
> +
> +	# Force use of software watchpoints.
> +	gdb_test_no_output "set can-use-hw-watchpoints 0"
> +
> +	gdb_test "watch global" \
> +	    "Watchpoint .*: global" \
> +	    "set software watchpoint on global variable"
> +
> +	gdb_test "continue" \
> +	    "Watchpoint .*: global.*Old value = 0.*New value = 1.*set_global \\(val=1\\).*$srcfile.*" \
> +	    "software watchpoint triggers"
> +
> +	set sw_watch_pc [get_pc "get sw watchpoint PC"]
> +
> +	delete_breakpoints
> +
> +	# Allow hardware watchpoints again.
> +	gdb_test_no_output "set can-use-hw-watchpoints 1"
> +
> +	gdb_test "watch global" \
> +	    "Hardware watchpoint .*: global" \
> +	    "set hardware watchpoint on global variable"
> +
> +	gdb_test "continue" \
> +	    "Hardware watchpoint .*: global.*Old value = 1.*New value = 2.*set_global \\(val=2\\).*$srcfile.*" \
> +	    "hardware watchpoint triggers"
> +
> +	set hw_watch_pc [get_pc "get hw watchpoint PC"]
> +
> +	gdb_assert {$sw_watch_pc == $sw_watch_pc} "hw watchpoint stops at right instruction"
> +    }
> +}
> +
> +foreach always_inserted {"off" "on" } {
> +    test $always_inserted
> +}
> 


Thanks,
Pedro Alves