[PATCH] tools/gdbsx: define format strings for aarch64

[PATCH] tools/gdbsx: define format strings for aarch64

[Olaf Hering]

gx_main.c: In function '_do_qRcmd_req':
gx_main.c:119:13: error: format '%llx' expects argument of type 'long long unsigned int', but argument 3 has type 'uint64_t' [-Werror=format=]
             sprintf(buf1, "pgd3val set to: "XGF64"\n", pgd3val);
             ^
gx_main.c:121:13: error: format '%llx' expects argument of type 'long long unsigned int', but argument 3 has type 'uint64_t' [-Werror=format=]
             sprintf(buf1, "Invalid  pgd3val "XGF64"\n", pgd3val);

Signed-off-by: Olaf Hering <olaf@aepfle.de>
---
 tools/debugger/gdbsx/xg/xg_public.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/debugger/gdbsx/xg/xg_public.h b/tools/debugger/gdbsx/xg/xg_public.h
index 6236d08..046e21b 100644
--- a/tools/debugger/gdbsx/xg/xg_public.h
+++ b/tools/debugger/gdbsx/xg/xg_public.h
@@ -23,7 +23,7 @@
 #define XGTRC1(...)  \
            do {(xgtrc_on==2) ? (xgtrc(__FUNCTION__,__VA_ARGS__)):0;} while (0)
 
-#if defined(__x86_64__)
+#if defined(__x86_64__) || defined(__aarch64__)
     #define  XGFM64  "%lx"
     #define  XGF64   "%016lx"
 #else

Re: [PATCH] tools/gdbsx: define format strings for aarch64

[Ian Campbell]

On Sun, 2014-02-02 at 09:19 +0100, Olaf Hering wrote:
> gx_main.c: In function '_do_qRcmd_req':
> gx_main.c:119:13: error: format '%llx' expects argument of type 'long long unsigned int', but argument 3 has type 'uint64_t' [-Werror=format=]
>              sprintf(buf1, "pgd3val set to: "XGF64"\n", pgd3val);
>              ^
> gx_main.c:121:13: error: format '%llx' expects argument of type 'long long unsigned int', but argument 3 has type 'uint64_t' [-Werror=format=]
>              sprintf(buf1, "Invalid  pgd3val "XGF64"\n", pgd3val);
> 
> Signed-off-by: Olaf Hering <olaf@aepfle.de>

I suppose there is no harm in this, but is there any chance that gdbsx
would actually work on arm64 without significant actual work going into
it?

Also, you forgot to CC the gdbsx maintainer.

(why doesn't this code use stdint.h?)

> ---
>  tools/debugger/gdbsx/xg/xg_public.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/debugger/gdbsx/xg/xg_public.h b/tools/debugger/gdbsx/xg/xg_public.h
> index 6236d08..046e21b 100644
> --- a/tools/debugger/gdbsx/xg/xg_public.h
> +++ b/tools/debugger/gdbsx/xg/xg_public.h
> @@ -23,7 +23,7 @@
>  #define XGTRC1(...)  \
>             do {(xgtrc_on==2) ? (xgtrc(__FUNCTION__,__VA_ARGS__)):0;} while (0)
>  
> -#if defined(__x86_64__)
> +#if defined(__x86_64__) || defined(__aarch64__)
>      #define  XGFM64  "%lx"
>      #define  XGF64   "%016lx"
>  #else

Re: [PATCH] tools/gdbsx: define format strings for aarch64

[Olaf Hering]

On Sun, Feb 02, Ian Campbell wrote:

> I suppose there is no harm in this, but is there any chance that gdbsx
> would actually work on arm64 without significant actual work going into
> it?

I have no idea. But I just noticed its built only due to this line in
our xen.spec:

make -C tools/debugger/gdbsx

Perhaps this should be guarded by a %ifarch x86_64.

Olaf

Re: [PATCH] tools/gdbsx: define format strings for aarch64

[Ian Campbell]

On Sun, 2014-02-02 at 11:28 +0100, Olaf Hering wrote:
> On Sun, Feb 02, Ian Campbell wrote:
> 
> > I suppose there is no harm in this, but is there any chance that gdbsx
> > would actually work on arm64 without significant actual work going into
> > it?
> 
> I have no idea. But I just noticed its built only due to this line in
> our xen.spec:
> 
> make -C tools/debugger/gdbsx
> 
> Perhaps this should be guarded by a %ifarch x86_64.

I suspect that right now that this would be wise (maybe i386 too?) --
hopefully Mukesh can advise.

Ian.

Re: [PATCH] tools/gdbsx: define format strings for aarch64

[Mukesh Rathor]

On Sun, 2 Feb 2014 10:36:16 +0000
Ian Campbell <ian.campbell@citrix.com> wrote:

> On Sun, 2014-02-02 at 11:28 +0100, Olaf Hering wrote:
> > On Sun, Feb 02, Ian Campbell wrote:
> > 
> > > I suppose there is no harm in this, but is there any chance that
> > > gdbsx would actually work on arm64 without significant actual
> > > work going into it?
> > 
> > I have no idea. But I just noticed its built only due to this line
> > in our xen.spec:
> > 
> > make -C tools/debugger/gdbsx
> > 
> > Perhaps this should be guarded by a %ifarch x86_64.
> 
> I suspect that right now that this would be wise (maybe i386 too?) --
> hopefully Mukesh can advise.
> 
> Ian.

It should just work if you can quickly implement 
arch/x86/debug.c for arm. If not, you'd need to "arch it out".

thanks
mukesh

Re: [PATCH] tools/gdbsx: define format strings for aarch64

[Olaf Hering]

On Thu, Feb 06, Mukesh Rathor wrote:

> It should just work if you can quickly implement 
> arch/x86/debug.c for arm. If not, you'd need to "arch it out".

Some Makefile already has something like SUBDIR-$(ARCH), it was just the
hard "make -C tools/gdbsx" in xen.spec which caused the error.

Olaf

Re: [PATCH] tools/gdbsx: define format strings for aarch64

[Ian Campbell]

On Thu, 2014-02-06 at 17:12 -0800, Mukesh Rathor wrote:
> On Sun, 2 Feb 2014 10:36:16 +0000
> Ian Campbell <ian.campbell@citrix.com> wrote:
> 
> > On Sun, 2014-02-02 at 11:28 +0100, Olaf Hering wrote:
> > > On Sun, Feb 02, Ian Campbell wrote:
> > > 
> > > > I suppose there is no harm in this, but is there any chance that
> > > > gdbsx would actually work on arm64 without significant actual
> > > > work going into it?
> > > 
> > > I have no idea. But I just noticed its built only due to this line
> > > in our xen.spec:
> > > 
> > > make -C tools/debugger/gdbsx
> > > 
> > > Perhaps this should be guarded by a %ifarch x86_64.
> > 
> > I suspect that right now that this would be wise (maybe i386 too?) --
> > hopefully Mukesh can advise.
> > 
> > Ian.
> 
> It should just work if you can quickly implement 
> arch/x86/debug.c for arm.

Thanks.

Looks like we'd also need to make a bunch of arch/x86/domctl.c generic
too, or implement the appropriate arm equivalents.

Not a 4.4 thing for sure, if someone wants gdbsx on arm then patches for
4.5 would be welcome as always.

>  If not, you'd need to "arch it out".
> 
> thanks
> mukesh