Re: [Xen-users] XEN/arm XENFB support

["peter" <peter@perkbv.com>]

Thank you for the info/patch Ian.
With some modifications to Qemu, I managed to cross compile
qemu-system-arm with XEN support.
Based on version: 1c514a7734b7f98625a0d18d5e8ee7581f26e50c (Merge remote
branch 'perard/cpu-hotplug-port-v2' into xen-staging-master-7).
And got my Qemu backend working.

I can see that INPUT_XEN_KBDDEV_FRONTEND is working
(d9) 6input: Xen Virtual Keyboard as /devices/virtual/input/input0
(d9) 6input: Xen Virtual Pointer as /devices/virtual/input/input1

And that XEN bus is prompting for VFB
(d9) 6xenbus_probe_frontend: Device with no driver: device/vfb/0

But it seems that the xen-fbfront only supports PV and not PVH. (Linux
kernel 3.12-RC5)
Because the init code is returning on:

	if (!xen_pv_domain())
		return -ENODEV;

When I comment this code out.
I do get these errors:

[ 1698.899496] Failed to unmap pfn:5c081 rc:-2

(XEN) dom10 IPA 0x0000000000000000
(XEN) P2M @ 021b8300 mfn:0x4dc18
(XEN) 1ST[0x0] = 0x000000004da276ff
(XEN) 2ND[0x0] = 0x000000006b6856ff
(XEN) 3RD[0x0] = 0x0000000000000000
[ 1698.933034] Failed to map pfn to mfn rc:0:-22 pfn:5c349 mfn:0

(XEN) dom10 IPA 0x000000002777c000
(XEN) P2M @ 021b8300 mfn:0x4dc18
(XEN) 1ST[0x0] = 0x000000004da276ff
(XEN) 2ND[0x13b] = 0x0000000000000000
[ 1698.951353] Failed to map pfn to mfn rc:0:-22 pfn:5c348 mfn:b6b2777c

Yes lots of them, and my VFB is not working tough.

So I've searched on the internet if someone got VFB on PVH working.
But I couldn't find anything about, so is there someone who got VFB
working on PVH (x86) ?


-----Original Message-----
From: Ian Campbell <Ian.Campbell@citrix.com>
To: peter <peter@perkbv.com>
Cc: "xen-users@lists.xen.org" <xen-users@lists.xen.org>,
"xen-devel@lists.xen.org" <xen-devel@lists.xen.org>, Lars Kurth
<lars.kurth.xen@gmail.com>
Date: Mon, 16 Dec 2013 09:59:09 +0000
Subject: Re: [Xen-devel] [Xen-users] XEN/arm XENFB support

> On Mon, 2013-12-16 at 08:45 +0000, Lars Kurth wrote:
> 
> > On Mon, Dec 16, 2013 at 8:16 AM, peter <peter@perkbv.com> wrote:
> >         I'm currently playing with XEN/arm on my Allwinner A20
> (cubieboard2)
> >         I would like to get the XENFB driver working on domU.
> >         But currently in xen/arm there's no support for VFB, atleast
> qemu is not
> >         supported.
> >         But this video http://www.youtube.com/watch?v=po1IeElg8tg and
> this one
> >         http://www.youtube.com/watch?v=Km6gBnIqaWo is showing a
> working framebuffer.
> >         So there are people which got a framebuffer working on domU.
> >         But still i couldn't find anything on the internet how to do
> this.
> >         Is there anyone with experience with VFB on xen/arm ?
> 
> I'm not sure what the first link used but the second was using a form
> of
> hybrid GPU passthrough/paravirtualisation for which the code has not
> been reeleased and AFAIK has nothing to do with VFB.
> 
> I've not tried VFB on ARM yet (busy with other things) but I suspect
> that if you were to build upstream qemu on ARM with Xen support enabled
> then it would just work. I would recommend giving that a go -- please
> let us know how you get on.
> 
> Below is a patch which I wrote ages ago that I have just rebased onto
> current xen staging *without retesting*. It is supposed to DTWT and
> enable the qemu build for ARM (as well as providing a mechanism to
> start
> the correct dom0 qemu, which you don't need for vfb).
> 
> As I say I haven't tested, or even built, and in particular I never
> tried PVFB with it even what I wrote it, so YMMV.
> 
> Ian.
> 
> 
> commit 7b5d54c9a5d09c4138bec905c9accea34173ba77
> Author: Ian Campbell <ian.campbell@citrix.com>
> Date:   Wed May 15 16:34:55 2013 +0100
> 
>     tools: build and launch correct qemu for architecture
>     
>     xl now provides a launch-dom0-qemu command which avoids the need to
> have the
>     initscripts be aware of the specific qermu binary name, which
> differs by
>     architecture and which also may have been specified by the user via
> the
>     --with-system-qemu=PATH option to configure.
>     
>     Perhaps this should be a separate binary hidden in libexec?
>     
>     Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
> 
> diff --git a/config/arm32.mk b/config/arm32.mk
> index aa79d22..f3599a3 100644
> --- a/config/arm32.mk
> +++ b/config/arm32.mk
> @@ -3,6 +3,8 @@ CONFIG_ARM_32 := y
>  CONFIG_ARM_$(XEN_OS) := y
>  
>  CONFIG_XEN_INSTALL_SUFFIX :=
> +CONFIG_QEMU_ARCH := arm
> +CONFIG_QEMU_TARGET := arm-softmmu
>  
>  # -march= -mcpu=
>  
> diff --git a/config/arm64.mk b/config/arm64.mk
> index 15b57a4..4ff15e0 100644
> --- a/config/arm64.mk
> +++ b/config/arm64.mk
> @@ -3,6 +3,8 @@ CONFIG_ARM_64 := y
>  CONFIG_ARM_$(XEN_OS) := y
>  
>  CONFIG_XEN_INSTALL_SUFFIX :=
> +CONFIG_QEMU_ARCH := aarch64
> +CONFIG_QEMU_TARGET := arm-softmmu
>  
>  CFLAGS += #-marm -march= -mcpu= etc
>  
> diff --git a/config/x86_32.mk b/config/x86_32.mk
> index 7f76b25..da3111d 100644
> --- a/config/x86_32.mk
> +++ b/config/x86_32.mk
> @@ -2,6 +2,9 @@ CONFIG_X86 := y
>  CONFIG_X86_32 := y
>  CONFIG_X86_$(XEN_OS) := y
>  
> +CONFIG_QEMU_ARCH := i386
> +CONFIG_QEMU_TARGET := i386-softmmu
> +
>  CONFIG_HVM := y
>  CONFIG_MIGRATE := y
>  CONFIG_XCUTILS := y
> diff --git a/config/x86_64.mk b/config/x86_64.mk
> index 11104bd..f59e36d 100644
> --- a/config/x86_64.mk
> +++ b/config/x86_64.mk
> @@ -2,6 +2,9 @@ CONFIG_X86 := y
>  CONFIG_X86_64 := y
>  CONFIG_X86_$(XEN_OS) := y
>  
> +CONFIG_QEMU_ARCH := x86_64
> +CONFIG_QEMU_ARCH := i386-softmmu
> +
>  CONFIG_COMPAT := y
>  CONFIG_HVM := y
>  CONFIG_MIGRATE := y
> diff --git a/tools/Makefile b/tools/Makefile
> index 00c69ee..250b931 100644
> --- a/tools/Makefile
> +++ b/tools/Makefile
> @@ -107,7 +107,7 @@ distclean: subdirs-distclean
>  		config.cache autom4te.cache
>  
>  ifneq ($(XEN_COMPILE_ARCH),$(XEN_TARGET_ARCH))
> -IOEMU_CONFIGURE_CROSS ?= --cpu=$(XEN_TARGET_ARCH) \
> +IOEMU_CONFIGURE_CROSS ?= --cpu=$(CONFIG_QEMU_ARCH) \
>  			 --cross-prefix=$(CROSS_COMPILE) \
>  			 --interp-prefix=$(CROSS_SYS_ROOT)
>  endif
> @@ -186,8 +186,7 @@ subdir-all-qemu-xen-dir: qemu-xen-dir-find
>  		source=.; \
>  	fi; \
>  	cd qemu-xen-dir; \
> -	$$source/configure --enable-xen --target-list=i386-softmmu \
> -		$(QEMU_XEN_ENABLE_DEBUG) \
> +	$$source/configure --enable-xen --target-list=$(CONFIG_QEMU_TARGET) \
>  		--prefix=$(PREFIX) \
>  		--source-path=$$source \
>  		--extra-cflags="-I$(XEN_ROOT)/tools/include \
> diff --git a/tools/hotplug/Linux/init.d/xencommons
> b/tools/hotplug/Linux/init.d/xencommons
> index 4ebd636..f568085 100644
> --- a/tools/hotplug/Linux/init.d/xencommons
> +++ b/tools/hotplug/Linux/init.d/xencommons
> @@ -116,11 +116,7 @@ do_start () {
>  	echo Starting xenconsoled...
>  	test -z "$XENCONSOLED_TRACE" || XENCONSOLED_ARGS="
> --log=$XENCONSOLED_TRACE"
>  	${SBINDIR}/xenconsoled --pid-file=$XENCONSOLED_PIDFILE
> $XENCONSOLED_ARGS
> -	echo Starting QEMU as disk backend for dom0
> -	test -z "$QEMU_XEN" && QEMU_XEN="${LIBEXEC}/qemu-system-i386"
> -	$QEMU_XEN -xen-domid 0 -xen-attach -name dom0 -nographic -M xenpv
> -daemonize \
> -		-monitor /dev/null -serial /dev/null -parallel /dev/null \
> -		-pidfile $QEMU_PIDFILE
> +	xl launch-dom0-qemu $QEMU_XEN
>  }
>  do_stop () {
>          echo Stopping xenconsoled
> diff --git a/tools/libxl/Makefile b/tools/libxl/Makefile
> index d8495bb..d8b6a5c 100644
> --- a/tools/libxl/Makefile
> +++ b/tools/libxl/Makefile
> @@ -27,6 +27,7 @@ CFLAGS_LIBXL += $(CFLAGS_libxenguest)
>  CFLAGS_LIBXL += $(CFLAGS_libxenstore)
>  CFLAGS_LIBXL += $(CFLAGS_libblktapctl) 
>  CFLAGS_LIBXL += -Wshadow
> +CFLAGS_LIBXL += -DCONFIG_QEMU_ARCH=\"$(CONFIG_QEMU_ARCH)\"
>  
>  LIBXL_LIBS-$(CONFIG_ARM) += -lfdt
>  
> diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
> index 12d6c31..0287a35 100644
> --- a/tools/libxl/libxl.h
> +++ b/tools/libxl/libxl.h
> @@ -605,6 +605,9 @@ int libxl_ctx_alloc(libxl_ctx **pctx, int version,
>                      xentoollog_logger *lg);
>  int libxl_ctx_free(libxl_ctx *ctx /* 0 is OK */);
>  
> +/* exec's device model for dom0 and does not return. */
> +void libxl_launch_dom0_qemu(libxl_ctx *ctx, const char *qemu_path,
> const char *pidfile);
> +
>  /* domain related functions */
>  
>  int libxl_domain_create_new(libxl_ctx *ctx, libxl_domain_config
> *d_config,
> diff --git a/tools/libxl/libxl_dm.c b/tools/libxl/libxl_dm.c
> index f6f7bbd..f0b13a9 100644
> --- a/tools/libxl/libxl_dm.c
> +++ b/tools/libxl/libxl_dm.c
> @@ -38,7 +38,7 @@ static const char *qemu_xen_path(libxl__gc *gc)
>  #ifdef QEMU_XEN_PATH
>      return QEMU_XEN_PATH;
>  #else
> -    return libxl__abs_path(gc, "qemu-system-i386",
> libxl__libexec_path());
> +    return libxl__abs_path(gc, "qemu-system-" CONFIG_QEMU_ARCH,
> libxl__libexec_path());
>  #endif
>  }
>  
> @@ -1560,6 +1560,37 @@ out:
>      return ret;
>  }
>  
> +void libxl_launch_dom0_qemu(libxl_ctx *ctx, const char *qemu_path,
> const char *pidfile)
> +{
> +    GC_INIT(ctx);
> +
> +    flexarray_t *dm_args = flexarray_make(gc, 16, 1);
> +
> +    if (qemu_path == NULL)
> +        qemu_path = qemu_xen_path(gc);
> +
> +    flexarray_vappend(dm_args,
> +                      "-xen-domid", "0",
> +                      "-xen-attach",
> +                      "-name", "dom0",
> +                      "-nographic",
> +                      "-M" "xenpv",
> +                      "-daemonize",
> +                      "-monitor", "/dev/null",
> +                      "-serial", "/dev/null",
> +                      "-parallel", "/dev/null",
> +                      NULL);
> +    if (pidfile)
> +        flexarray_append_pair(dm_args, "-pidfile", libxl__strdup(gc,
> pidfile));
> +
> +    libxl__exec(gc, -1, -1, -1,
> +                qemu_path, (char **)flexarray_contents(dm_args),
> NULL);
> +
> +    /* Shouldn't return */
> +    LOG(CRITICAL, "Failed to exec dom0 device model");
> +    GC_FREE;
> +}
> +
>  /*
>   * Local variables:
>   * mode: C
> diff --git a/tools/libxl/xl.h b/tools/libxl/xl.h
> index c876a33..1d7602c 100644
> --- a/tools/libxl/xl.h
> +++ b/tools/libxl/xl.h
> @@ -106,6 +106,7 @@ int main_setenforce(int argc, char **argv);
>  int main_loadpolicy(int argc, char **argv);
>  int main_remus(int argc, char **argv);
>  int main_devd(int argc, char **argv);
> +int main_launch_dom0_qemu(int argc, char **argv);
>  
>  void help(const char *command);
>  
> diff --git a/tools/libxl/xl_cmdimpl.c b/tools/libxl/xl_cmdimpl.c
> index bd26bcc..108dfac 100644
> --- a/tools/libxl/xl_cmdimpl.c
> +++ b/tools/libxl/xl_cmdimpl.c
> @@ -7313,6 +7313,31 @@ out:
>      return ret;
>  }
>  
> +int main_launch_dom0_qemu(int argc, char **argv)
> +{
> +    int opt;
> +    const char *qemu = NULL;
> +    const char *pidfile = NULL;
> +
> +    SWITCH_FOREACH_OPT(opt, "p:", NULL, "launch-dom-qemu", 0) {
> +    case 'p':
> +        pidfile = optarg;
> +        break;
> +        /* No options */
> +    }
> +    if (optind < argc)
> +        qemu = argv[optind];
> +
> +    fprintf(stderr, "argc %d\n", argc);
> +    fprintf(stderr, "qemu = %s", qemu ? : "<default>");
> +
> +    fprintf(stdout, "Starting QEMU as disk backend for dom0");
> +
> +    libxl_launch_dom0_qemu(ctx, qemu, pidfile);
> +
> +    return 0;
> +}
> +
>  /*
>   * Local variables:
>   * mode: C
> diff --git a/tools/libxl/xl_cmdtable.c b/tools/libxl/xl_cmdtable.c
> index ebe0220..ab4d56c 100644
> --- a/tools/libxl/xl_cmdtable.c
> +++ b/tools/libxl/xl_cmdtable.c
> @@ -494,6 +494,12 @@ struct cmd_spec cmd_table[] = {
>        "[options]",
>        "-F                      Run in the foreground",
>      },
> +    { "launch-dom0-qemu",
> +      &main_launch_dom0_qemu, 0, 1,
> +      "Start qemu process to service dom0 disk backends",
> +      "[options] [QEMU_PATH]",
> +      "-p PIDFILE              Write a PIDFILE\n",
> +    },
>  };
>  
>  int cmdtable_len = sizeof(cmd_table)/sizeof(struct cmd_spec);