old-cross-binutils/gdb/testsuite/gdb.base/watchpoint-reuse-slot.exp

252 lines
6.8 KiB
Text
Raw Normal View History

# Copyright 2014-2016 Free Software Foundation, Inc.
x86 Linux watchpoints: Couldn't write debug register: Invalid argument. This patch fixes this on x86 Linux: (gdb) watch *buf@2 Hardware watchpoint 8: *buf@2 (gdb) si 0x00000000004005a7 34 for (i = 0; i < 100000; i++); /* stepi line */ (gdb) del Delete all breakpoints? (y or n) y (gdb) watch *(buf+1)@1 Hardware watchpoint 9: *(buf+1)@1 (gdb) si 0x00000000004005a7 in main () at ../../../src/gdb/testsuite/gdb.base/watchpoint-reuse-slot.c:34 34 for (i = 0; i < 100000; i++); /* stepi line */ Couldn't write debug register: Invalid argument. (gdb) In the example above the debug registers are being switched from this state: CONTROL (DR7): 0000000000050101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601040, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 to this: CONTROL (DR7): 0000000000010101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601041, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 That is, before, DR7 was setup for watching a 2 byte region starting at what's in DR0 (0x601040). And after, DR7 is setup for watching a 1 byte region starting at what's in DR0 (0x601041). We always write DR0..DR3 before DR7, because if we enable a slot's bits in DR7, you need to have already written the corresponding DR0..DR3 registers -- the kernel rejects the DR7 write with EINVAL otherwise. The error shown above is the opposite scenario. When we try to write 0x601041 to DR0, DR7's bits still indicate intent of watching a 2-byte region. That DR0/DR7 combination is invalid, because 0x601041 is unaligned. To watch two bytes, we'd have to use two slots. So the kernel errors out with EINVAL. Fix this by always first clearing DR7, then writing DR0..DR3, and then setting DR7's bits. A little optimization -- if we're disabling the last watchpoint, then we can clear DR7 just once. The changes to nat/i386-dregs.c make that easier to detect, and as bonus, they make it a little easier to make sense of DR7 in the debug logs, as we no longer need to remember we're seeing stale bits. Tested on x86_64 Fedora 20, native and GDBserver. This adds an exhaustive test that switches between many different combinations of watchpoint types and addresses and widths. gdb/ 2014-06-23 Pedro Alves <palves@redhat.com> * amd64-linux-nat.c (amd64_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. * i386-linux-nat.c (i386_linux_prepare_to_resume): Likewise. * nat/i386-dregs.c (i386_remove_aligned_watchpoint): Clear all bits of DR_CONTROL related to the debug register slot being disabled. If all slots are vacant, clear local slowdown as well, and assert DR_CONTROL is 0. gdb/gdbserver/ 2014-06-23 Pedro Alves <palves@redhat.com> * linux-x86-low.c (x86_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. gdb/testsuite/ 2014-06-23 Pedro Alves <palves@redhat.com> * gdb.base/watchpoint-reuse-slot.c: New file. * gdb.base/watchpoint-reuse-slot.exp: New file.
2014-06-23 15:44:04 +00:00
# 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/>.
# Test alternating between watchpoint types, watching a sliding window
# of addresses (thus alternating between aligned and unaligned
# addresses). Only a single watchpoint exists at any given time. On
# targets that only update the debug registers on resume, this
# stresses the debug register setup code, both in GDB and in the
# target/kernel as one watchpoint replaces the other in a single
# operation. (Note that we don't have any of these watchpoints
# trigger.)
if [target_info exists gdb,no_hardware_watchpoints] {
unsupported "no target support"
return
}
standard_testfile
if {[prepare_for_testing "failed to prepare" $testfile $srcfile debug]} {
return -1
}
if ![runto_main] then {
fail "Can't run to main"
return 0
}
# The line we'll be stepping.
set srcline [gdb_get_line_number "stepi line"]
# The address the program is stopped at currently.
set cur_addr ""
# Get the current PC.
proc get_pc {} {
global hex gdb_prompt
set addr ""
set test "get PC"
gdb_test_multiple "p /x \$pc" "$test" {
-re " = ($hex).*$gdb_prompt $" {
set addr $expect_out(1,string)
pass "$test"
}
}
return $addr
}
# Issue a stepi, and make sure the program advanced past the current
# instruction (stored in the CUR_ADDR global).
proc stepi {} {
global hex gdb_prompt cur_addr
set srcline " for (i = 0; i < 100000; i++); /* stepi line */"
set test "stepi advanced"
gdb_test_multiple "stepi" $test {
-re "($hex).*[string_to_regexp $srcline]\r\n$gdb_prompt $" {
set addr $expect_out(1,string)
if {$addr != $cur_addr} {
pass $test
} else {
fail $test
}
set cur_addr addr
}
}
}
gdb_breakpoint $srcline
gdb_continue_to_breakpoint "stepi line"
# The test tries various sequences of different types of watchpoints.
# Probe for support first.
# 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"
# The list of supported commands. Below we'll probe for support and
# add elements to this list.
set cmds {}
foreach cmd {"watch" "awatch" "rwatch"} {
set test $cmd
gdb_test_multiple "$cmd buf.byte\[0\]" $test {
-re "You may have requested too many.*$gdb_prompt $" {
unsupported $test
}
-re "Target does not support.*$gdb_prompt $" {
unsupported $test
}
x86 Linux watchpoints: Couldn't write debug register: Invalid argument. This patch fixes this on x86 Linux: (gdb) watch *buf@2 Hardware watchpoint 8: *buf@2 (gdb) si 0x00000000004005a7 34 for (i = 0; i < 100000; i++); /* stepi line */ (gdb) del Delete all breakpoints? (y or n) y (gdb) watch *(buf+1)@1 Hardware watchpoint 9: *(buf+1)@1 (gdb) si 0x00000000004005a7 in main () at ../../../src/gdb/testsuite/gdb.base/watchpoint-reuse-slot.c:34 34 for (i = 0; i < 100000; i++); /* stepi line */ Couldn't write debug register: Invalid argument. (gdb) In the example above the debug registers are being switched from this state: CONTROL (DR7): 0000000000050101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601040, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 to this: CONTROL (DR7): 0000000000010101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601041, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 That is, before, DR7 was setup for watching a 2 byte region starting at what's in DR0 (0x601040). And after, DR7 is setup for watching a 1 byte region starting at what's in DR0 (0x601041). We always write DR0..DR3 before DR7, because if we enable a slot's bits in DR7, you need to have already written the corresponding DR0..DR3 registers -- the kernel rejects the DR7 write with EINVAL otherwise. The error shown above is the opposite scenario. When we try to write 0x601041 to DR0, DR7's bits still indicate intent of watching a 2-byte region. That DR0/DR7 combination is invalid, because 0x601041 is unaligned. To watch two bytes, we'd have to use two slots. So the kernel errors out with EINVAL. Fix this by always first clearing DR7, then writing DR0..DR3, and then setting DR7's bits. A little optimization -- if we're disabling the last watchpoint, then we can clear DR7 just once. The changes to nat/i386-dregs.c make that easier to detect, and as bonus, they make it a little easier to make sense of DR7 in the debug logs, as we no longer need to remember we're seeing stale bits. Tested on x86_64 Fedora 20, native and GDBserver. This adds an exhaustive test that switches between many different combinations of watchpoint types and addresses and widths. gdb/ 2014-06-23 Pedro Alves <palves@redhat.com> * amd64-linux-nat.c (amd64_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. * i386-linux-nat.c (i386_linux_prepare_to_resume): Likewise. * nat/i386-dregs.c (i386_remove_aligned_watchpoint): Clear all bits of DR_CONTROL related to the debug register slot being disabled. If all slots are vacant, clear local slowdown as well, and assert DR_CONTROL is 0. gdb/gdbserver/ 2014-06-23 Pedro Alves <palves@redhat.com> * linux-x86-low.c (x86_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. gdb/testsuite/ 2014-06-23 Pedro Alves <palves@redhat.com> * gdb.base/watchpoint-reuse-slot.c: New file. * gdb.base/watchpoint-reuse-slot.exp: New file.
2014-06-23 15:44:04 +00:00
-re "$gdb_prompt $" {
pass $test
lappend cmds $cmd
}
}
delete_breakpoints
}
set test "hbreak"
gdb_test_multiple "hbreak main" $test {
-re "You may have requested too many.*$gdb_prompt $" {
unsupported $test
}
-re "No hardware breakpoint support.*$gdb_prompt $" {
unsupported $test
x86 Linux watchpoints: Couldn't write debug register: Invalid argument. This patch fixes this on x86 Linux: (gdb) watch *buf@2 Hardware watchpoint 8: *buf@2 (gdb) si 0x00000000004005a7 34 for (i = 0; i < 100000; i++); /* stepi line */ (gdb) del Delete all breakpoints? (y or n) y (gdb) watch *(buf+1)@1 Hardware watchpoint 9: *(buf+1)@1 (gdb) si 0x00000000004005a7 in main () at ../../../src/gdb/testsuite/gdb.base/watchpoint-reuse-slot.c:34 34 for (i = 0; i < 100000; i++); /* stepi line */ Couldn't write debug register: Invalid argument. (gdb) In the example above the debug registers are being switched from this state: CONTROL (DR7): 0000000000050101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601040, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 to this: CONTROL (DR7): 0000000000010101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601041, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 That is, before, DR7 was setup for watching a 2 byte region starting at what's in DR0 (0x601040). And after, DR7 is setup for watching a 1 byte region starting at what's in DR0 (0x601041). We always write DR0..DR3 before DR7, because if we enable a slot's bits in DR7, you need to have already written the corresponding DR0..DR3 registers -- the kernel rejects the DR7 write with EINVAL otherwise. The error shown above is the opposite scenario. When we try to write 0x601041 to DR0, DR7's bits still indicate intent of watching a 2-byte region. That DR0/DR7 combination is invalid, because 0x601041 is unaligned. To watch two bytes, we'd have to use two slots. So the kernel errors out with EINVAL. Fix this by always first clearing DR7, then writing DR0..DR3, and then setting DR7's bits. A little optimization -- if we're disabling the last watchpoint, then we can clear DR7 just once. The changes to nat/i386-dregs.c make that easier to detect, and as bonus, they make it a little easier to make sense of DR7 in the debug logs, as we no longer need to remember we're seeing stale bits. Tested on x86_64 Fedora 20, native and GDBserver. This adds an exhaustive test that switches between many different combinations of watchpoint types and addresses and widths. gdb/ 2014-06-23 Pedro Alves <palves@redhat.com> * amd64-linux-nat.c (amd64_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. * i386-linux-nat.c (i386_linux_prepare_to_resume): Likewise. * nat/i386-dregs.c (i386_remove_aligned_watchpoint): Clear all bits of DR_CONTROL related to the debug register slot being disabled. If all slots are vacant, clear local slowdown as well, and assert DR_CONTROL is 0. gdb/gdbserver/ 2014-06-23 Pedro Alves <palves@redhat.com> * linux-x86-low.c (x86_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. gdb/testsuite/ 2014-06-23 Pedro Alves <palves@redhat.com> * gdb.base/watchpoint-reuse-slot.c: New file. * gdb.base/watchpoint-reuse-slot.exp: New file.
2014-06-23 15:44:04 +00:00
}
-re "$gdb_prompt $" {
pass $test
lappend cmds "hbreak"
}
}
delete_breakpoints
set cur_addr [get_pc]
watchpoint-reuse-slot.exp: skip setting HW breakpoints on some address We see some fails in watchpoint-reuse-slot.exp on aarch64-linux, because it sets some HW breakpoint on some address doesn't meet the alignment requirements by kernel, kernel will reject the ptrace (PTRACE_SETHBPREGS) call, and some fails are caused, for example: (gdb) PASS: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 0: delete $bpnum hbreak *(buf.byte + 0 + 1)^M Hardware assisted breakpoint 80 at 0x410a61^M (gdb) PASS: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 1: hbreak *(buf.byte + 0 + 1) stepi^M Warning:^M Cannot insert hardware breakpoint 80.^M Could not insert hardware breakpoints:^M You may have requested too many hardware breakpoints/watchpoints.^M ^M (gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 1: stepi advanced hbreak *(buf.byte + 0 + 1)^M Hardware assisted breakpoint 440 at 0x410a61^M Warning:^M Cannot insert hardware breakpoint 440.^M Could not insert hardware breakpoints:^M You may have requested too many hardware breakpoints/watchpoints.^M ^M (gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted on: watch x hbreak: : width 1, iter 0: base + 1: hbreak *(buf.byte + 0 + 1) This patch is to skip some tests by checking proc valid_addr_p. We can handle other targets in valid_addr_p too. gdb/testsuite: 2015-03-16 Yao Qi <yao.qi@linaro.org> * gdb.base/watchpoint-reuse-slot.exp (valid_addr_p): New proc. (top level): Skip tests if valid_addr_p returns false for $cmd1 or $cmd2.
2015-03-13 10:56:01 +00:00
# Return true if the memory range [buf.byte + OFFSET, +WIDTH] can be
# monitored by CMD, otherwise return false.
proc valid_addr_p {cmd offset width} {
if { [istarget "aarch64*-*-linux*"] } {
# The aarch64 Linux kernel port only accepts 4-byte aligned addresses
# for hardware breakpoints and 8-byte aligned addresses for hardware
# watchpoints. However, both GDB and GDBserver support unaligned
# watchpoints by using more than one properly aligned watchpoint
# registers to represent the whole unaligned region. Breakpoint
# addresses must still be aligned though.
if {$cmd == "hbreak" } {
if { [expr ($offset) % 4] != 0 } {
return 0
}
}
} elseif { [istarget "arm*-*-linux*"] } {
if { $cmd == "hbreak" } {
# Breakpoints must be of length 2 (thumb) or 4 (ARM) bytes.
if { $width != 2 && $width != 4 } {
return 0
}
} else {
# Watchpoints can be of length 1, 2, 4 or 8 bytes.
if { [expr $width % 2] != 0 } {
return 0
}
}
if { [expr ($offset) % 8] == 0 && $width == 8 } {
# If WIDTH is 8 byte, the address should be 8-byte aligned.
return 1
} elseif { [expr ($offset) % 4] == 0 } {
return 1
} elseif { [expr ($offset) % 4] == 2 && $width == 2 } {
# Halfword watchpoints and breakpoints.
return 1
} elseif { [expr ($offset) % 4] == 1 && $width == 1 && $cmd != "hbreak" } {
# Single byte watchpoints.
return 1
} else {
return 0
}
watchpoint-reuse-slot.exp: skip setting HW breakpoints on some address We see some fails in watchpoint-reuse-slot.exp on aarch64-linux, because it sets some HW breakpoint on some address doesn't meet the alignment requirements by kernel, kernel will reject the ptrace (PTRACE_SETHBPREGS) call, and some fails are caused, for example: (gdb) PASS: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 0: delete $bpnum hbreak *(buf.byte + 0 + 1)^M Hardware assisted breakpoint 80 at 0x410a61^M (gdb) PASS: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 1: hbreak *(buf.byte + 0 + 1) stepi^M Warning:^M Cannot insert hardware breakpoint 80.^M Could not insert hardware breakpoints:^M You may have requested too many hardware breakpoints/watchpoints.^M ^M (gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 1: stepi advanced hbreak *(buf.byte + 0 + 1)^M Hardware assisted breakpoint 440 at 0x410a61^M Warning:^M Cannot insert hardware breakpoint 440.^M Could not insert hardware breakpoints:^M You may have requested too many hardware breakpoints/watchpoints.^M ^M (gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted on: watch x hbreak: : width 1, iter 0: base + 1: hbreak *(buf.byte + 0 + 1) This patch is to skip some tests by checking proc valid_addr_p. We can handle other targets in valid_addr_p too. gdb/testsuite: 2015-03-16 Yao Qi <yao.qi@linaro.org> * gdb.base/watchpoint-reuse-slot.exp (valid_addr_p): New proc. (top level): Skip tests if valid_addr_p returns false for $cmd1 or $cmd2.
2015-03-13 10:56:01 +00:00
}
return 1
}
x86 Linux watchpoints: Couldn't write debug register: Invalid argument. This patch fixes this on x86 Linux: (gdb) watch *buf@2 Hardware watchpoint 8: *buf@2 (gdb) si 0x00000000004005a7 34 for (i = 0; i < 100000; i++); /* stepi line */ (gdb) del Delete all breakpoints? (y or n) y (gdb) watch *(buf+1)@1 Hardware watchpoint 9: *(buf+1)@1 (gdb) si 0x00000000004005a7 in main () at ../../../src/gdb/testsuite/gdb.base/watchpoint-reuse-slot.c:34 34 for (i = 0; i < 100000; i++); /* stepi line */ Couldn't write debug register: Invalid argument. (gdb) In the example above the debug registers are being switched from this state: CONTROL (DR7): 0000000000050101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601040, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 to this: CONTROL (DR7): 0000000000010101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601041, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 That is, before, DR7 was setup for watching a 2 byte region starting at what's in DR0 (0x601040). And after, DR7 is setup for watching a 1 byte region starting at what's in DR0 (0x601041). We always write DR0..DR3 before DR7, because if we enable a slot's bits in DR7, you need to have already written the corresponding DR0..DR3 registers -- the kernel rejects the DR7 write with EINVAL otherwise. The error shown above is the opposite scenario. When we try to write 0x601041 to DR0, DR7's bits still indicate intent of watching a 2-byte region. That DR0/DR7 combination is invalid, because 0x601041 is unaligned. To watch two bytes, we'd have to use two slots. So the kernel errors out with EINVAL. Fix this by always first clearing DR7, then writing DR0..DR3, and then setting DR7's bits. A little optimization -- if we're disabling the last watchpoint, then we can clear DR7 just once. The changes to nat/i386-dregs.c make that easier to detect, and as bonus, they make it a little easier to make sense of DR7 in the debug logs, as we no longer need to remember we're seeing stale bits. Tested on x86_64 Fedora 20, native and GDBserver. This adds an exhaustive test that switches between many different combinations of watchpoint types and addresses and widths. gdb/ 2014-06-23 Pedro Alves <palves@redhat.com> * amd64-linux-nat.c (amd64_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. * i386-linux-nat.c (i386_linux_prepare_to_resume): Likewise. * nat/i386-dregs.c (i386_remove_aligned_watchpoint): Clear all bits of DR_CONTROL related to the debug register slot being disabled. If all slots are vacant, clear local slowdown as well, and assert DR_CONTROL is 0. gdb/gdbserver/ 2014-06-23 Pedro Alves <palves@redhat.com> * linux-x86-low.c (x86_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. gdb/testsuite/ 2014-06-23 Pedro Alves <palves@redhat.com> * gdb.base/watchpoint-reuse-slot.c: New file. * gdb.base/watchpoint-reuse-slot.exp: New file.
2014-06-23 15:44:04 +00:00
# Watch WIDTH bytes at BASE + OFFSET. CMD specifices the specific
# type of watchpoint to use. If CMD is "hbreak", WIDTH is ignored.
proc watch_command {cmd base offset width} {
global srcfile srcline hex
if {$cmd == "hbreak"} {
set expr "*(buf.byte + $base + $offset)"
gdb_test "hbreak $expr" "Hardware assisted breakpoint \[0-9\]+ at $hex"
} elseif {$cmd == "watch"} {
set expr "*(buf.byte + $base + $offset)@$width"
gdb_test "$cmd $expr" \
"Hardware watchpoint \[0-9\]+: [string_to_regexp $expr]"
} elseif {$cmd == "awatch"} {
set expr "*(buf.byte + $base + $offset)@$width"
gdb_test "$cmd $expr" \
"Hardware access \\(read/write\\) watchpoint \[0-9\]+: [string_to_regexp $expr]"
} elseif {$cmd == "rwatch"} {
set expr "*(buf.byte + $base + $offset)@$width"
gdb_test "$cmd $expr" \
"Hardware read watchpoint \[0-9\]+: [string_to_regexp $expr]"
}
}
# Run test proper. See intro for description.
foreach always_inserted {"off" "on" } {
gdb_test_no_output "set breakpoint always-inserted $always_inserted"
foreach cmd1 $cmds {
foreach cmd2 $cmds {
for {set width 1} {$width < 4} {incr width} {
if {$cmd1 == "hbreak" && $cmd2 == "hbreak" && $width > 1} {
# hbreak ignores WIDTH, no use testing more than
# once.
continue
}
for {set x 0} {$x < 4} {incr x} {
watchpoint-reuse-slot.exp: skip setting HW breakpoints on some address We see some fails in watchpoint-reuse-slot.exp on aarch64-linux, because it sets some HW breakpoint on some address doesn't meet the alignment requirements by kernel, kernel will reject the ptrace (PTRACE_SETHBPREGS) call, and some fails are caused, for example: (gdb) PASS: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 0: delete $bpnum hbreak *(buf.byte + 0 + 1)^M Hardware assisted breakpoint 80 at 0x410a61^M (gdb) PASS: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 1: hbreak *(buf.byte + 0 + 1) stepi^M Warning:^M Cannot insert hardware breakpoint 80.^M Could not insert hardware breakpoints:^M You may have requested too many hardware breakpoints/watchpoints.^M ^M (gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x hbreak: : width 1, iter 0: base + 1: stepi advanced hbreak *(buf.byte + 0 + 1)^M Hardware assisted breakpoint 440 at 0x410a61^M Warning:^M Cannot insert hardware breakpoint 440.^M Could not insert hardware breakpoints:^M You may have requested too many hardware breakpoints/watchpoints.^M ^M (gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted on: watch x hbreak: : width 1, iter 0: base + 1: hbreak *(buf.byte + 0 + 1) This patch is to skip some tests by checking proc valid_addr_p. We can handle other targets in valid_addr_p too. gdb/testsuite: 2015-03-16 Yao Qi <yao.qi@linaro.org> * gdb.base/watchpoint-reuse-slot.exp (valid_addr_p): New proc. (top level): Skip tests if valid_addr_p returns false for $cmd1 or $cmd2.
2015-03-13 10:56:01 +00:00
if { ![valid_addr_p $cmd1 $x $width]
|| ![valid_addr_p $cmd2 $x+1 $width] } {
# Skip tests if requested address or length
# of breakpoint or watchpoint don't meet
# target or kernel requirements.
continue
}
x86 Linux watchpoints: Couldn't write debug register: Invalid argument. This patch fixes this on x86 Linux: (gdb) watch *buf@2 Hardware watchpoint 8: *buf@2 (gdb) si 0x00000000004005a7 34 for (i = 0; i < 100000; i++); /* stepi line */ (gdb) del Delete all breakpoints? (y or n) y (gdb) watch *(buf+1)@1 Hardware watchpoint 9: *(buf+1)@1 (gdb) si 0x00000000004005a7 in main () at ../../../src/gdb/testsuite/gdb.base/watchpoint-reuse-slot.c:34 34 for (i = 0; i < 100000; i++); /* stepi line */ Couldn't write debug register: Invalid argument. (gdb) In the example above the debug registers are being switched from this state: CONTROL (DR7): 0000000000050101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601040, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 to this: CONTROL (DR7): 0000000000010101 STATUS (DR6): 0000000000000000 DR0: addr=0x0000000000601041, ref.count=1 DR1: addr=0x0000000000000000, ref.count=0 DR2: addr=0x0000000000000000, ref.count=0 DR3: addr=0x0000000000000000, ref.count=0 That is, before, DR7 was setup for watching a 2 byte region starting at what's in DR0 (0x601040). And after, DR7 is setup for watching a 1 byte region starting at what's in DR0 (0x601041). We always write DR0..DR3 before DR7, because if we enable a slot's bits in DR7, you need to have already written the corresponding DR0..DR3 registers -- the kernel rejects the DR7 write with EINVAL otherwise. The error shown above is the opposite scenario. When we try to write 0x601041 to DR0, DR7's bits still indicate intent of watching a 2-byte region. That DR0/DR7 combination is invalid, because 0x601041 is unaligned. To watch two bytes, we'd have to use two slots. So the kernel errors out with EINVAL. Fix this by always first clearing DR7, then writing DR0..DR3, and then setting DR7's bits. A little optimization -- if we're disabling the last watchpoint, then we can clear DR7 just once. The changes to nat/i386-dregs.c make that easier to detect, and as bonus, they make it a little easier to make sense of DR7 in the debug logs, as we no longer need to remember we're seeing stale bits. Tested on x86_64 Fedora 20, native and GDBserver. This adds an exhaustive test that switches between many different combinations of watchpoint types and addresses and widths. gdb/ 2014-06-23 Pedro Alves <palves@redhat.com> * amd64-linux-nat.c (amd64_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. * i386-linux-nat.c (i386_linux_prepare_to_resume): Likewise. * nat/i386-dregs.c (i386_remove_aligned_watchpoint): Clear all bits of DR_CONTROL related to the debug register slot being disabled. If all slots are vacant, clear local slowdown as well, and assert DR_CONTROL is 0. gdb/gdbserver/ 2014-06-23 Pedro Alves <palves@redhat.com> * linux-x86-low.c (x86_linux_prepare_to_resume): Clear DR_CONTROL before setting DR0..DR3. gdb/testsuite/ 2014-06-23 Pedro Alves <palves@redhat.com> * gdb.base/watchpoint-reuse-slot.c: New file. * gdb.base/watchpoint-reuse-slot.exp: New file.
2014-06-23 15:44:04 +00:00
set prefix "always-inserted $always_inserted: "
append prefix "$cmd1 x $cmd2: "
with_test_prefix "$prefix: width $width, iter $x" {
with_test_prefix "base + 0" {
watch_command $cmd1 $x 0 $width
stepi
gdb_test_no_output "delete \$bpnum"
}
with_test_prefix "base + 1" {
watch_command $cmd2 $x 1 $width
stepi
gdb_test_no_output "delete \$bpnum"
}
}
}
}
}
}
}