[PATCH v2 0/2] Various build fixes for NetBSD

[PATCH v2 0/2] Various build fixes for NetBSD

[John Baldwin]

Relative to the first patch series: one patch has been committed,
the amd64 KVM support has been dropped, the _KMEMUSER #define was
moved earlier in bsd-kvm.o, and the x86 debug registers change was
expanded to also add needed object files in the NetBSD/x86 native
targets.  I haven't been able to test remaining patches so am
waiting for testing reports from coypu@.

John Baldwin (2):
  Define _KMEMUSER before including BSD kernel headers.
  Enable support for x86 debug registers on NetBSD.

 gdb/ChangeLog     | 12 ++++++++++++
 gdb/bsd-kvm.c     |  1 +
 gdb/configure     |  1 +
 gdb/configure.ac  |  1 +
 gdb/configure.nat |  8 ++++----
 gdb/x86-bsd-nat.c |  9 +++++++--
 6 files changed, 26 insertions(+), 6 deletions(-)

-- 
2.13.3

[PATCH v2 1/2] Define _KMEMUSER before including BSD kernel headers.

[John Baldwin]

Recent versions of NetBSD hide certain kernel structures needed by the
KVM target from userland unless this macro is defined.

gdb/ChangeLog:

	* bsd-kvm.o: Define _KMEMUSER.
	* configure.ac: Define _KMEMUSER when checking for "struct lwp".
	* configure: Regenerate.
---
 gdb/ChangeLog    | 6 ++++++
 gdb/bsd-kvm.c    | 1 +
 gdb/configure    | 1 +
 gdb/configure.ac | 1 +
 4 files changed, 9 insertions(+)

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index b3f9c7b061..e06712d645 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,5 +1,11 @@
 2017-07-25  John Baldwin  <jhb@FreeBSD.org>
 
+	* bsd-kvm.o: Define _KMEMUSER.
+	* configure.ac: Define _KMEMUSER when checking for "struct lwp".
+	* configure: Regenerate.
+
+2017-07-25  John Baldwin  <jhb@FreeBSD.org>
+
 	* configure.nat: Add "-lkvm" for NetBSD/sparc64 and fix typo.
 
 2017-07-25  Yao Qi  <yao.qi@linaro.org>
diff --git a/gdb/bsd-kvm.c b/gdb/bsd-kvm.c
index 98e242836d..8c2139462d 100644
--- a/gdb/bsd-kvm.c
+++ b/gdb/bsd-kvm.c
@@ -17,6 +17,7 @@
    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
 
+#define _KMEMUSER
 #include "defs.h"
 #include "cli/cli-cmds.h"
 #include "command.h"
diff --git a/gdb/configure b/gdb/configure
index 755e3dccf1..c7746fc60f 100755
--- a/gdb/configure
+++ b/gdb/configure
@@ -13571,6 +13571,7 @@ else
   cat confdefs.h - <<_ACEOF >conftest.$ac_ext
 /* end confdefs.h.  */
 #include <sys/param.h>
+#define _KMEMUSER
 #include <sys/lwp.h>
 int
 main ()
diff --git a/gdb/configure.ac b/gdb/configure.ac
index 10d2d10b12..b4d7a87701 100644
--- a/gdb/configure.ac
+++ b/gdb/configure.ac
@@ -1434,6 +1434,7 @@ AC_CHECK_MEMBERS([struct thread.td_pcb], [], [],
 # See if <sys/lwp.h> defines `struct lwp`.
 AC_CACHE_CHECK([for struct lwp], gdb_cv_struct_lwp,
 [AC_TRY_COMPILE([#include <sys/param.h>
+#define _KMEMUSER
 #include <sys/lwp.h>], [struct lwp l;],
 gdb_cv_struct_lwp=yes, gdb_cv_struct_lwp=no)])
 if test $gdb_cv_struct_lwp = yes; then
-- 
2.13.3

[PATCH v2 2/2] Enable support for x86 debug registers on NetBSD.

[John Baldwin]

NetBSD recently added PT_GETDBREGS and PT_SETDBREGS ptrace operations
that match the existing ones supported by x86-bsd-nat.c.  NetBSD's
headers do not provide the DBREG_DRX helper macro, so define a local
version in x86-bsd-nat.c.  In addition, add the x86-nat.o and x86-dregs.o
object files to the native NetBSD x86 build targets.

gdb/ChangeLog:

	* configure.nat: Add "x86-nat.o x86-dregs.o" for NetBSD/amd64 and
	NetBSD/i386.
	* x86-bsd-nat.c [!DBREG_DRX && __NetBSD__]: Define DBREG_DRX.
---
 gdb/ChangeLog     | 6 ++++++
 gdb/configure.nat | 8 ++++----
 gdb/x86-bsd-nat.c | 9 +++++++--
 3 files changed, 17 insertions(+), 6 deletions(-)

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index e06712d645..e24698ae30 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,5 +1,11 @@
 2017-07-25  John Baldwin  <jhb@FreeBSD.org>
 
+	* configure.nat: Add "x86-nat.o x86-dregs.o" for NetBSD/amd64 and
+	NetBSD/i386.
+	* x86-bsd-nat.c [!DBREG_DRX && __NetBSD__]: Define DBREG_DRX.
+
+2017-07-25  John Baldwin  <jhb@FreeBSD.org>
+
 	* bsd-kvm.o: Define _KMEMUSER.
 	* configure.ac: Define _KMEMUSER when checking for "struct lwp".
 	* configure: Regenerate.
diff --git a/gdb/configure.nat b/gdb/configure.nat
index b32c9aa849..b1dda30852 100644
--- a/gdb/configure.nat
+++ b/gdb/configure.nat
@@ -338,8 +338,8 @@ case ${gdb_host} in
 	case ${gdb_host_cpu} in
 	    i386)
 		# Host: NetBSD/amd64
-		NATDEPFILES="${NATDEPFILES} nbsd-nat.o amd64-nat.o \
-		x86-bsd-nat.o amd64-bsd-nat.o amd64-nbsd-nat.o"
+		NATDEPFILES="${NATDEPFILES} nbsd-nat.o amd64-nat.o x86-nat.o \
+		x86-dregs.o x86-bsd-nat.o amd64-bsd-nat.o amd64-nbsd-nat.o"
 		;;
 	    sparc)
 		# Host: NetBSD/sparc64
@@ -358,8 +358,8 @@ case ${gdb_host} in
 		;;
 	    i386)
 		# Host: NetBSD/i386 ELF
-		NATDEPFILES="${NATDEPFILES} nbsd-nat.o x86-bsd-nat.o \
-		i386-bsd-nat.o i386-nbsd-nat.o bsd-kvm.o"
+		NATDEPFILES="${NATDEPFILES} nbsd-nat.o x86-nat.o x86-dregs.o \
+		x86-bsd-nat.o i386-bsd-nat.o i386-nbsd-nat.o bsd-kvm.o"
 		LOADLIBES='-lkvm'
 		;;
 	    m68k)
diff --git a/gdb/x86-bsd-nat.c b/gdb/x86-bsd-nat.c
index 2ff661c179..1d37b2f05d 100644
--- a/gdb/x86-bsd-nat.c
+++ b/gdb/x86-bsd-nat.c
@@ -51,11 +51,16 @@ x86bsd_mourn_inferior (struct target_ops *ops)
   super_mourn_inferior (ops);
 }
 
-/* Not all versions of FreeBSD/i386 that support the debug registers
-   have this macro.  */
+/* Helper macro to access debug register X.  FreeBSD/amd64 and modern
+   versions of FreeBSD/i386 provide this macro in system headers.  Define
+   a local version for systems that do not provide it.  */
 #ifndef DBREG_DRX
+#ifdef __NetBSD__
+#define DBREG_DRX(d, x) ((d)->dr[x])
+#else
 #define DBREG_DRX(d, x) ((&d->dr0)[x])
 #endif
+#endif
 
 static unsigned long
 x86bsd_dr_get (ptid_t ptid, int regnum)
-- 
2.13.3

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[Pedro Alves]

On 07/25/2017 05:50 PM, John Baldwin wrote:
> Relative to the first patch series: one patch has been committed,
> the amd64 KVM support has been dropped, the _KMEMUSER #define was
> moved earlier in bsd-kvm.o, and the x86 debug registers change was
> expanded to also add needed object files in the NetBSD/x86 native
> targets.  I haven't been able to test remaining patches so am
> waiting for testing reports from coypu@.
> 

This hasn't been pushed yet, right?
FWIW, it looks good to me.

Thanks,
Pedro Alves

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[Kamil Rytarowski]

[-- Attachment #1.1: Type: text/plain, Size: 693 bytes --]

On 04.09.2017 14:50, Pedro Alves wrote:
> On 07/25/2017 05:50 PM, John Baldwin wrote:
>> Relative to the first patch series: one patch has been committed,
>> the amd64 KVM support has been dropped, the _KMEMUSER #define was
>> moved earlier in bsd-kvm.o, and the x86 debug registers change was
>> expanded to also add needed object files in the NetBSD/x86 native
>> targets.  I haven't been able to test remaining patches so am
>> waiting for testing reports from coypu@.
>>
> 
> This hasn't been pushed yet, right?
> FWIW, it looks good to me.
> 
> Thanks,
> Pedro Alves
> 

Hi,

NetBSD patches are still pending. I'm going to readd NetBSD buildbot
after merging them.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[Pedro Alves]

On 09/04/2017 02:04 PM, Kamil Rytarowski wrote:
> On 04.09.2017 14:50, Pedro Alves wrote:
>> On 07/25/2017 05:50 PM, John Baldwin wrote:
>>> Relative to the first patch series: one patch has been committed,
>>> the amd64 KVM support has been dropped, the _KMEMUSER #define was
>>> moved earlier in bsd-kvm.o, and the x86 debug registers change was
>>> expanded to also add needed object files in the NetBSD/x86 native
>>> targets.  I haven't been able to test remaining patches so am
>>> waiting for testing reports from coypu@.
>>>
>>
>> This hasn't been pushed yet, right?
>> FWIW, it looks good to me.
>>
>> Thanks,
>> Pedro Alves
>>
> 
> Hi,
> 
> NetBSD patches are still pending. I'm going to readd NetBSD buildbot
> after merging them.

So, how to break the deadlock?  Can you perhaps try applying the
patches manually (or John could push them to some branch), and do some
quick smoke testing just to be sure things aren't horribly broken?
Otherwise, I'd be fine with just pushing them to master directly
and if the new buildbot reveals breakage we can always back out the
change.

Thanks,
Pedro Alves

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[Kamil Rytarowski]

[-- Attachment #1.1: Type: text/plain, Size: 1759 bytes --]

On 04.09.2017 16:01, Pedro Alves wrote:
> 
> On 09/04/2017 02:04 PM, Kamil Rytarowski wrote:
>> On 04.09.2017 14:50, Pedro Alves wrote:
>>> On 07/25/2017 05:50 PM, John Baldwin wrote:
>>>> Relative to the first patch series: one patch has been committed,
>>>> the amd64 KVM support has been dropped, the _KMEMUSER #define was
>>>> moved earlier in bsd-kvm.o, and the x86 debug registers change was
>>>> expanded to also add needed object files in the NetBSD/x86 native
>>>> targets.  I haven't been able to test remaining patches so am
>>>> waiting for testing reports from coypu@.
>>>>
>>>
>>> This hasn't been pushed yet, right?
>>> FWIW, it looks good to me.
>>>
>>> Thanks,
>>> Pedro Alves
>>>
>>
>> Hi,
>>
>> NetBSD patches are still pending. I'm going to readd NetBSD buildbot
>> after merging them.
> 
> So, how to break the deadlock?  Can you perhaps try applying the
> patches manually (or John could push them to some branch), and do some
> quick smoke testing just to be sure things aren't horribly broken?
> Otherwise, I'd be fine with just pushing them to master directly
> and if the new buildbot reveals breakage we can always back out the
> change.
> 
> Thanks,
> Pedro Alves
> 

I'm right now upstreaming pkgsrc patches to qemu. Once I will be done, I
will switch to gdb.

In general there are around 2 remaining build issues on NetBSD
(<sys/user.h> freebsdism and missing _KMEMUSER). In the past I was
demotivated by that ChangeLog format.

Long term, I'm still working on ptrace(2) correctness (appropriate
support for threads in tracee) and LLDB/NetBSD porting.. I expect to
allocate at some point time - when working on LLDB/kgdb - to improve GDB
support and resurrect execution of tests.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[John Baldwin]

On Monday, September 04, 2017 04:35:58 PM Kamil Rytarowski wrote:
> On 04.09.2017 16:01, Pedro Alves wrote:
> > 
> > On 09/04/2017 02:04 PM, Kamil Rytarowski wrote:
> >> On 04.09.2017 14:50, Pedro Alves wrote:
> >>> On 07/25/2017 05:50 PM, John Baldwin wrote:
> >>>> Relative to the first patch series: one patch has been committed,
> >>>> the amd64 KVM support has been dropped, the _KMEMUSER #define was
> >>>> moved earlier in bsd-kvm.o, and the x86 debug registers change was
> >>>> expanded to also add needed object files in the NetBSD/x86 native
> >>>> targets.  I haven't been able to test remaining patches so am
> >>>> waiting for testing reports from coypu@.
> >>>>
> >>>
> >>> This hasn't been pushed yet, right?
> >>> FWIW, it looks good to me.
> >>>
> >>> Thanks,
> >>> Pedro Alves
> >>>
> >>
> >> Hi,
> >>
> >> NetBSD patches are still pending. I'm going to readd NetBSD buildbot
> >> after merging them.
> > 
> > So, how to break the deadlock?  Can you perhaps try applying the
> > patches manually (or John could push them to some branch), and do some
> > quick smoke testing just to be sure things aren't horribly broken?
> > Otherwise, I'd be fine with just pushing them to master directly
> > and if the new buildbot reveals breakage we can always back out the
> > change.
> > 
> > Thanks,
> > Pedro Alves
> > 
> 
> I'm right now upstreaming pkgsrc patches to qemu. Once I will be done, I
> will switch to gdb.
> 
> In general there are around 2 remaining build issues on NetBSD
> (<sys/user.h> freebsdism and missing _KMEMUSER). In the past I was
> demotivated by that ChangeLog format.
> 
> Long term, I'm still working on ptrace(2) correctness (appropriate
> support for threads in tracee) and LLDB/NetBSD porting.. I expect to
> allocate at some point time - when working on LLDB/kgdb - to improve GDB
> support and resurrect execution of tests.

I'm fine with pushing the current patchset to master if that works best
for you Kamil.  I'm also happy to help shepherd NetBSD patches in when I
can.

-- 
John Baldwin

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[John Baldwin]

On Monday, September 04, 2017 01:50:37 PM Pedro Alves wrote:
> On 07/25/2017 05:50 PM, John Baldwin wrote:
> > Relative to the first patch series: one patch has been committed,
> > the amd64 KVM support has been dropped, the _KMEMUSER #define was
> > moved earlier in bsd-kvm.o, and the x86 debug registers change was
> > expanded to also add needed object files in the NetBSD/x86 native
> > targets.  I haven't been able to test remaining patches so am
> > waiting for testing reports from coypu@.
> > 
> 
> This hasn't been pushed yet, right?
> FWIW, it looks good to me.

I have just been waiting for coypu@ to verify that they tested out ok.

-- 
John Baldwin

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[coypu]

[-- Attachment #1: Type: text/plain, Size: 152 bytes --]

Hi,

I've needed amendments to patch #3, and an additional patch.
The additional patch is needed as kamil had removed sys/user.h in
netbsd.

Thank you.

[-- Attachment #2: 0001-Use-existing-configure-test-to-guard-inclusion-of-sy.patch --]
[-- Type: text/plain, Size: 587 bytes --]

From 147b605f040889110fc47bf6b9d2b00af7317958 Mon Sep 17 00:00:00 2001
From: coypu <coypu@sdf.org>
Date: Mon, 4 Sep 2017 22:59:17 +0300
Subject: [PATCH 1/2] Use existing configure test to guard inclusion of
 sys/user.h

---
 gdb/bsd-kvm.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/gdb/bsd-kvm.c b/gdb/bsd-kvm.c
index f872fc084..67f7e5113 100644
--- a/gdb/bsd-kvm.c
+++ b/gdb/bsd-kvm.c
@@ -37,7 +37,9 @@
 #include "readline/readline.h"
 #include <sys/param.h>
 #include <sys/proc.h>
+#ifdef HAVE_SYS_USER_H
 #include <sys/user.h>
+#endif
 
 #include "bsd-kvm.h"
 
-- 
2.14.1


[-- Attachment #3: 0002-configure.nat-Add-bsd-kvm.o-x86-nat.o-x86-dregs.o.patch --]
[-- Type: text/plain, Size: 890 bytes --]

From 426613ab4883d2edae286a7635298f217bb031c8 Mon Sep 17 00:00:00 2001
From: coypu <coypu@sdf.org>
Date: Mon, 4 Sep 2017 23:00:00 +0300
Subject: [PATCH 2/2] configure.nat: Add "bsd-kvm.o x86-nat.o x86-dregs.o" and
 "-lkvm" for NetBSD/amd64

---
 gdb/configure.nat | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/gdb/configure.nat b/gdb/configure.nat
index b32c9aa84..36c48b10f 100644
--- a/gdb/configure.nat
+++ b/gdb/configure.nat
@@ -338,8 +338,10 @@ case ${gdb_host} in
 	case ${gdb_host_cpu} in
 	    i386)
 		# Host: NetBSD/amd64
-		NATDEPFILES="${NATDEPFILES} nbsd-nat.o amd64-nat.o \
-		x86-bsd-nat.o amd64-bsd-nat.o amd64-nbsd-nat.o"
+		NATDEPFILES="${NATDEPFILES} nbsd-nat.o amd64-nat.o x86-nat.o \
+		x86-bsd-nat.o amd64-bsd-nat.o amd64-nbsd-nat.o bsd-kvm.o \
+		x86-dregs.o"
+		LOADLIBES='-lkvm'
 		;;
 	    sparc)
 		# Host: NetBSD/sparc64
-- 
2.14.1

Re: [PATCH v2 0/2] Various build fixes for NetBSD

[John Baldwin]

On Monday, September 04, 2017 08:04:16 PM coypu@sdf.org wrote:
> Hi,
> 
> I've needed amendments to patch #3, and an additional patch.
> The additional patch is needed as kamil had removed sys/user.h in
> netbsd.

To be clear, you now want bsd-kvm.o for NetBSD/amd64?  Previously I had
included this but then removed it as the current amd64-nbsd-nat.c doesn't
include a call to 'bsd_kvm_add_target' with a suitable callback function
to supply registers from a PCB.  Without that callback, the kvm target
won't be able to walk stacks or switch to other threads making it less
useful.

For now I will push the other fixes without bsd-kvm.o for amd64 but I can
add that in as a followup commit easily enough.

-- 
John Baldwin