[v3,3/3] gdb, zpoint: check for target hardware breakpoint support
Checks
Context |
Check |
Description |
linaro-tcwg-bot/tcwg_gdb_build--master-aarch64 |
success
|
Testing passed
|
linaro-tcwg-bot/tcwg_gdb_build--master-arm |
success
|
Testing passed
|
Commit Message
In 'can_use_hw_breakpoint', check if the target supports
hardware-assisted breakpoints. This will prevent GDB
from trying to insert the hardware breakpoint in case
it is not supported.
Reviewed-By: Guinevere Larsen <blarsen@redhat.com>
---
gdb/remote.c | 28 ++++++++++++++++++++++++----
1 file changed, 24 insertions(+), 4 deletions(-)
Comments
>>>>> Mohamed Bouhaouel <mohamed.bouhaouel@intel.com> writes:
> In 'can_use_hw_breakpoint', check if the target supports
> hardware-assisted breakpoints. This will prevent GDB
> from trying to insert the hardware breakpoint in case
> it is not supported.
Thanks.
One question I have is how this is tested.
Another oddity is that supposedly a remote can reply '' to a z packet,
meaning it isn't supported. However, insert_hw_breakpoint (at least)
doesn't seem to record this response.
> + /* If hw read watchpoints are not supported while hw access are,
> + GDB will try to insert the watchpoint as hw access. */
> + bool access_support = supports_z_point_type (
> + bptype_to_target_hw_bp_type (bp_access_watchpoint));
Formatting looks wrong here.
Tom
Hi,
Thanks Tom for the feedback,
> One question I have is how this is tested.
We would need a remote target that does not support hardware
watchpoints. But I am not sure how can we build such a test.
> Another oddity is that supposedly a remote can reply '' to a z packet,
> meaning it isn't supported. However, insert_hw_breakpoint (at least)
> doesn't seem to record this response.
Could you please elaborate? The remote can or can't replay to z packets?
> > + /* If hw read watchpoints are not supported while hw access are,
> > + GDB will try to insert the watchpoint as hw access. */
> > + bool access_support = supports_z_point_type (
> > + bptype_to_target_hw_bp_type (bp_access_watchpoint));
>
> Formatting looks wrong here.
Hmm, how else can we format it? One suggestion is to use an additional variable
to store the return of " bptype_to_target_hw_bp_type (bp_access_watchpoint)".
Thanks,
Mohamed
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928
>>>>> Bouhaouel, Mohamed <mohamed.bouhaouel@intel.com> writes:
>> > + /* If hw read watchpoints are not supported while hw access are,
>> > + GDB will try to insert the watchpoint as hw access. */
>> > + bool access_support = supports_z_point_type (
>> > + bptype_to_target_hw_bp_type (bp_access_watchpoint));
>>
>> Formatting looks wrong here.
> Hmm, how else can we format it? One suggestion is to use an additional variable
> to store the return of " bptype_to_target_hw_bp_type (bp_access_watchpoint)".
Yeah, that would work. A dangling open paren isn't really gdb style.
Another way would be
bool access_support
= (supports_z_point_type
(bptype_to_target_hw_bp_type (bp_access_watchpoint)));
kind of uglier than a new variable but it's done elsewhere.
Tom
@@ -10982,7 +10982,7 @@ remote_target::remove_breakpoint (struct gdbarch *gdbarch,
}
static enum Z_packet_type
-watchpoint_to_Z_packet (int type)
+hw_bp_to_Z_packet (int type)
{
switch (type)
{
@@ -10995,6 +10995,9 @@ watchpoint_to_Z_packet (int type)
case hw_access:
return Z_PACKET_ACCESS_WP;
break;
+ case hw_execute:
+ return Z_PACKET_HARDWARE_BP;
+ break;
default:
internal_error (_("hw_bp_to_z: bad watchpoint type %d"), type);
}
@@ -11003,7 +11006,7 @@ watchpoint_to_Z_packet (int type)
bool
remote_target::supports_z_point_type (int type)
{
- Z_packet_type packet = watchpoint_to_Z_packet (type);
+ Z_packet_type packet = hw_bp_to_Z_packet (type);
return (m_features.packet_support (PACKET_Z0 + packet) != PACKET_DISABLE);
}
@@ -11014,7 +11017,7 @@ remote_target::insert_watchpoint (CORE_ADDR addr, int len,
struct remote_state *rs = get_remote_state ();
char *endbuf = rs->buf.data () + get_remote_packet_size ();
char *p;
- enum Z_packet_type packet = watchpoint_to_Z_packet (type);
+ enum Z_packet_type packet = hw_bp_to_Z_packet (type);
if (m_features.packet_support ((to_underlying (PACKET_Z0)
+ to_underlying (packet))) == PACKET_DISABLE)
@@ -11064,7 +11067,7 @@ remote_target::remove_watchpoint (CORE_ADDR addr, int len,
struct remote_state *rs = get_remote_state ();
char *endbuf = rs->buf.data () + get_remote_packet_size ();
char *p;
- enum Z_packet_type packet = watchpoint_to_Z_packet (type);
+ enum Z_packet_type packet = hw_bp_to_Z_packet (type);
if (m_features.packet_support ((to_underlying (PACKET_Z0)
+ to_underlying (packet))) == PACKET_DISABLE)
@@ -11118,6 +11121,11 @@ remote_target::can_use_hw_breakpoint (enum bptype type, int cnt, int ot)
{
if (type == bp_hardware_breakpoint)
{
+ /* Check if the target supports hardware-assisted breakpoints.
+ Return 0 if not. */
+ if (!supports_z_point_type (bptype_to_target_hw_bp_type (type)))
+ return 0;
+
if (remote_hw_breakpoint_limit == 0)
return 0;
else if (remote_hw_breakpoint_limit < 0)
@@ -11127,6 +11135,18 @@ remote_target::can_use_hw_breakpoint (enum bptype type, int cnt, int ot)
}
else
{
+ /* Check if the target supports the hardware watchpoint type.
+ Return 0 if not. */
+ if (!supports_z_point_type (bptype_to_target_hw_bp_type (type)))
+ {
+ /* If hw read watchpoints are not supported while hw access are,
+ GDB will try to insert the watchpoint as hw access. */
+ bool access_support = supports_z_point_type (
+ bptype_to_target_hw_bp_type (bp_access_watchpoint));
+ if (!(type == bp_read_watchpoint && access_support))
+ return 0;
+ }
+
if (remote_hw_watchpoint_limit == 0)
return 0;
else if (remote_hw_watchpoint_limit < 0)