Hi folks,
This is version #5 of the xen support for qemu patch series.
It has been quiet for a while, mostly due to qemu-xen lagging behind
qemu-upstream. Now it catched up with upstream, so it is time to make
some progress here. Not many changes since the last time I posted the
patches. Some prelimary stuff has been merged and thus is gone, some
bugs have been fixed, some adaptions to recent changes happend.
The console and framebuffer backend drivers are largely based on the
xen code, the other bits are rewritten from scratch. Nevertheless
the code should be functionally identical.
Overview (individual patches have longer descriptions):
#1 -- groundwork: build system, cmd line options, ...
#2 -- xen backend driver infrastructrure
#3 -- xen console backend driver
#4 -- xen framebuffer backend driver
#5 -- xen block backend driver
#6 -- xen nic backend driver
#7 -- allow xen disks and nics being configured via qemu command
line options.
With first next four patches in place upstream qemu can replace xen''s
qemu-dm for paravirtual domains. Due to some small differences in
the command line syntax it can''t work as drop-in replacement though.
The last three patches add full userspace implementations of block and
nic backend drivers using the grant table device (gntdev) and command
line support config support for setting up these backends.
xen support is implemented using another machine type. xen''s qemu-dm
already uses the machine type to switch between paravirtualized and
fully virtualized machines, so this was the natural choice. qemu
gets a new "xenpv" machine type additionally to the "pc" and
"isapc"
ones.
The patches are also available here (Patches 0002 -> 0008):
http://kraxel.fedorapeople.org/patches/qemu-upstream/
There is also a patch queue for qemu-xen, located here:
http://kraxel.fedorapeople.org/patches/qemu-xen/
There also a ''-b'' (no whitespace changes) version for easier
review:
http://kraxel.fedorapeople.org/patches/qemu-xen/no-ws/
The qemu-xen patch series has a slightly different ordering to make the
patch queue more bisect-friendly. Also the command line syntax is
unchanged, so we don''t break qemu-dm <=> xend interaction. It
should
work just fine as drop-in replacement, without regressions, with the
additional bonus that the framebuffer handles a restart cycle correctly
(i.e. pvgrub in graphical mode actually works).
Comments?
cheers,
Gerd
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 12:23 UTC
[Xen-devel] [PATCH 1/7] xen: groundwork for xen support
- configure script and build system changes.
- wind up new machine type.
- add -xen-* command line options.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 8 ++++
configure | 36 +++++++++++++++++
hw/boards.h | 3 +
hw/xen.h | 20 +++++++++
hw/xen_machine_pv.c | 105 +++++++++++++++++++++++++++++++++++++++++++++++++
target-i386/machine.c | 3 +
vl.c | 29 +++++++++++++
7 files changed, 204 insertions(+), 0 deletions(-)
create mode 100644 hw/xen.h
create mode 100644 hw/xen_machine_pv.c
diff --git a/Makefile.target b/Makefile.target
index f9c0b34..5368f6e 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -629,6 +629,14 @@ ifdef CONFIG_BLUEZ
LIBS += $(CONFIG_BLUEZ_LIBS)
endif
+# xen backend driver support
+XEN_OBJS := xen_machine_pv.o
+ifeq ($(CONFIG_XEN), yes)
+ OBJS += $(XEN_OBJS)
+ LIBS += $(XEN_LIBS)
+ $(XEN_OBJS) : CFLAGS += -Wall -Wmissing-prototypes -Wstrict-prototypes
+endif
+
# SCSI layer
OBJS+= lsi53c895a.o esp.o
diff --git a/configure b/configure
index ccc4ae0..b1ab014 100755
--- a/configure
+++ b/configure
@@ -114,6 +114,9 @@ aio="yes"
nptl="yes"
mixemu="no"
bluez="yes"
+signalfd="no"
+eventfd="no"
+xen="yes"
# OS specific
targetos=`uname -s`
@@ -298,6 +301,8 @@ for opt do
;;
--disable-kqemu) kqemu="no"
;;
+ --disable-xen) xen="no"
+ ;;
--disable-brlapi) brlapi="no"
;;
--disable-bluez) bluez="no"
@@ -441,6 +446,7 @@ echo " Available drivers:
$audio_possible_drivers"
echo " --audio-card-list=LIST set list of additional emulated audio
cards"
echo " Available cards: ac97 adlib cs4231a
gus"
echo " --enable-mixemu enable mixer emulation"
+echo " --disable-xen disable xen backend driver support"
echo " --disable-brlapi disable BrlAPI"
echo " --disable-vnc-tls disable TLS encryption for VNC
server"
echo " --disable-curses disable curses output"
@@ -748,6 +754,22 @@ else
fi
##########################################
+# xen probe
+
+if test "$xen" = "yes" ; then
+cat > $TMPC <<EOF
+#include <xenctrl.h>
+#include <xs.h>
+int main(void) { xs_daemon_open; xc_interface_open; }
+EOF
+ if $cc $ARCH_CFLAGS -c -o $TMPO $TMPC -lxenstore -lxenctrl 2> /dev/null ;
then
+ :
+ else
+ xen="no"
+ fi
+fi
+
+##########################################
# SDL probe
sdl_too_old=no
@@ -1026,6 +1048,7 @@ if test -n "$sparc_cpu"; then
echo "Target Sparc Arch $sparc_cpu"
fi
echo "kqemu support $kqemu"
+echo "xen support $xen"
echo "brlapi support $brlapi"
echo "Documentation $build_docs"
[ ! -z "$uname_release" ] && \
@@ -1296,6 +1319,9 @@ if test "$bluez" = "yes" ; then
echo "CONFIG_BLUEZ_LIBS=$bluez_libs" >> $config_mak
echo "#define CONFIG_BLUEZ 1" >> $config_h
fi
+if test "$xen" = "yes" ; then
+ echo "XEN_LIBS=-lxenstore -lxenctrl" >> $config_mak
+fi
if test "$aio" = "yes" ; then
echo "#define CONFIG_AIO 1" >> $config_h
echo "CONFIG_AIO=yes" >> $config_mak
@@ -1417,6 +1443,11 @@ case "$target_cpu" in
then
echo "#define USE_KQEMU 1" >> $config_h
fi
+ if test "$xen" = "yes" -a "$target_softmmu" =
"yes";
+ then
+ echo "CONFIG_XEN=yes" >> $config_mak
+ echo "#define CONFIG_XEN 1" >> $config_h
+ fi
gcc3minver=`$cc --version 2> /dev/null| fgrep "(GCC) 3." | awk
''{ print $3 }'' | cut -f2 -d.`
if test -n "$gcc3minver" && test $gcc3minver -gt 3
then
@@ -1434,6 +1465,11 @@ case "$target_cpu" in
then
echo "#define USE_KQEMU 1" >> $config_h
fi
+ if test "$xen" = "yes" -a "$target_softmmu" =
"yes"
+ then
+ echo "CONFIG_XEN=yes" >> $config_mak
+ echo "#define CONFIG_XEN 1" >> $config_h
+ fi
;;
alpha)
echo "TARGET_ARCH=alpha" >> $config_mak
diff --git a/hw/boards.h b/hw/boards.h
index d30c0fc..9bcfdae 100644
--- a/hw/boards.h
+++ b/hw/boards.h
@@ -32,6 +32,9 @@ extern QEMUMachine bareetraxfs_machine;
extern QEMUMachine pc_machine;
extern QEMUMachine isapc_machine;
+/* xen_machine.c */
+extern QEMUMachine xenpv_machine;
+
/* ppc.c */
extern QEMUMachine prep_machine;
extern QEMUMachine core99_machine;
diff --git a/hw/xen.h b/hw/xen.h
new file mode 100644
index 0000000..4bcc0f1
--- /dev/null
+++ b/hw/xen.h
@@ -0,0 +1,20 @@
+#ifndef QEMU_HW_XEN_H
+#define QEMU_HW_XEN_H 1
+/*
+ * public xen header
+ * stuff needed outside xen-*.c, i.e. interfaces to qemu.
+ * must not depend on any xen headers being present in
+ * /usr/include/xen, so it can be included unconditionally.
+ */
+
+/* xen-machine.c */
+enum xen_mode {
+ XEN_EMULATE = 0, // xen emulation, using xenner (default)
+ XEN_CREATE, // create xen domain
+ XEN_ATTACH // attach to xen domain created by xend
+};
+
+extern int xen_domid;
+extern enum xen_mode xen_mode;
+
+#endif /* QEMU_HW_XEN_H */
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
new file mode 100644
index 0000000..9daab75
--- /dev/null
+++ b/hw/xen_machine_pv.c
@@ -0,0 +1,105 @@
+/*
+ * QEMU Xen PV Machine
+ *
+ * Copyright (c) 2007,08 Red Hat
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the
"Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "hw.h"
+#include "boards.h"
+
+#include "xen.h"
+#include <xenctrl.h>
+
+/* -------------------------------------------------------------------- */
+/* variables */
+
+int xen_domid;
+enum xen_mode xen_mode = XEN_EMULATE;
+
+/* -------------------------------------------------------------------- */
+/* initialization code */
+
+static int xen_init(void)
+{
+ if (!xen_domid) {
+ fprintf(stderr, "%s: no domid specified\n", __FUNCTION__);
+ return -1;
+ }
+
+ if (xen_mode == XEN_EMULATE) {
+ fprintf(stderr, "%s: emulating Xen\n", __FUNCTION__);
+ }
+
+ return 0;
+}
+
+static int xen_init_pv(DisplayState *ds)
+{
+ int rc;
+
+ rc = xen_init();
+ if (rc < 0)
+ return rc;
+
+ return 0;
+}
+
+/* -------------------------------------------------------------------- */
+/* paravirtualized xen machine */
+
+static void xenpv_init(ram_addr_t ram_size, int vga_ram_size,
+ const char *boot_device, DisplayState *ds,
+ const char *kernel_filename,
+ const char *kernel_cmdline,
+ const char *initrd_filename,
+ const char *cpu_model)
+{
+ CPUState *env;
+ int rc;
+
+ rc = xen_init_pv(ds);
+ if (-1 == rc)
+ goto err;
+
+ /* create dummy cpu, halted */
+ if (cpu_model == NULL) {
+#ifdef TARGET_X86_64
+ cpu_model = "qemu64";
+#else
+ cpu_model = "qemu32";
+#endif
+ }
+ env = cpu_init(cpu_model);
+ env->halted = 1;
+
+ return;
+
+err:
+ exit(1);
+}
+
+QEMUMachine xenpv_machine = {
+ .name = "xenpv",
+ .desc = "paravirtualized Xen machine",
+ .init = xenpv_init,
+ .nodisk_ok = 1,
+ .max_cpus = 1,
+};
diff --git a/target-i386/machine.c b/target-i386/machine.c
index 9d440fb..be10cfb 100644
--- a/target-i386/machine.c
+++ b/target-i386/machine.c
@@ -9,6 +9,9 @@ void register_machines(void)
{
qemu_register_machine(&pc_machine);
qemu_register_machine(&isapc_machine);
+#ifdef CONFIG_XEN
+ qemu_register_machine(&xenpv_machine);
+#endif
}
static void cpu_put_seg(QEMUFile *f, SegmentCache *dt)
diff --git a/vl.c b/vl.c
index 0ca7151..0d275bf 100644
--- a/vl.c
+++ b/vl.c
@@ -30,6 +30,7 @@
#include "hw/isa.h"
#include "hw/baum.h"
#include "hw/bt.h"
+#include "hw/xen.h"
#include "net.h"
#include "console.h"
#include "sysemu.h"
@@ -8282,6 +8283,13 @@ static void help(int exitcode)
"-startdate select initial date of the clock\n"
"-icount [N|auto]\n"
" Enable virtual instruction counter with 2^N
clock ticks per instruction\n"
+#ifdef CONFIG_XEN
+ "-xen-domid id specify xen guest domain id\n"
+ "-xen-create create domain using xen hypercalls, bypassing
xend\n"
+ " warning: should not be used when xend is in
use\n"
+ "-xen-attach attach to existing xen domain\n"
+ " xend will use this when starting qemu\n"
+#endif
"\n"
"During emulation, the following keys are useful:\n"
"ctrl-alt-f toggle full screen\n"
@@ -8387,6 +8395,11 @@ enum {
QEMU_OPTION_icount,
QEMU_OPTION_uuid,
QEMU_OPTION_incoming,
+#ifdef CONFIG_XEN
+ QEMU_OPTION_xen_domid,
+ QEMU_OPTION_xen_create,
+ QEMU_OPTION_xen_attach,
+#endif
};
typedef struct QEMUOption {
@@ -8476,6 +8489,11 @@ static const QEMUOption qemu_options[] = {
{ "curses", 0, QEMU_OPTION_curses },
#endif
{ "uuid", HAS_ARG, QEMU_OPTION_uuid },
+#ifdef CONFIG_XEN
+ { "xen-domid", HAS_ARG, QEMU_OPTION_xen_domid },
+ { "xen-create", 0, QEMU_OPTION_xen_create },
+ { "xen-attach", 0, QEMU_OPTION_xen_attach },
+#endif
/* temporary options */
{ "usb", 0, QEMU_OPTION_usb },
@@ -9403,6 +9421,17 @@ int main(int argc, char **argv)
case QEMU_OPTION_incoming:
incoming = optarg;
break;
+#ifdef CONFIG_XEN
+ case QEMU_OPTION_xen_domid:
+ xen_domid = atoi(optarg);
+ break;
+ case QEMU_OPTION_xen_create:
+ xen_mode = XEN_CREATE;
+ break;
+ case QEMU_OPTION_xen_attach:
+ xen_mode = XEN_ATTACH;
+ break;
+#endif
}
}
}
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
This patch adds infrastructure for xen backend drivers living in qemu,
so drivers don''t need to implement common stuff on their own.
It''s
mostly xenbus management stuff: some functions to access xentore,
setting up xenstore watches, callbacks on device discovery and state
changes, handle event channel, ...
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 2 +-
hw/xen_backend.c | 691 +++++++++++++++++++++++++++++++++++++++++++++++++++
hw/xen_backend.h | 86 +++++++
hw/xen_common.h | 34 +++
hw/xen_machine_pv.c | 8 +-
5 files changed, 818 insertions(+), 3 deletions(-)
create mode 100644 hw/xen_backend.c
create mode 100644 hw/xen_backend.h
create mode 100644 hw/xen_common.h
diff --git a/Makefile.target b/Makefile.target
index 5368f6e..c18cbc6 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -630,7 +630,7 @@ LIBS += $(CONFIG_BLUEZ_LIBS)
endif
# xen backend driver support
-XEN_OBJS := xen_machine_pv.o
+XEN_OBJS := xen_machine_pv.o xen_backend.o
ifeq ($(CONFIG_XEN), yes)
OBJS += $(XEN_OBJS)
LIBS += $(XEN_LIBS)
diff --git a/hw/xen_backend.c b/hw/xen_backend.c
new file mode 100644
index 0000000..2678dfa
--- /dev/null
+++ b/hw/xen_backend.c
@@ -0,0 +1,691 @@
+/*
+ * xen backend driver infrastructure
+ * (c) 2008 Gerd Hoffmann <kraxel@redhat.com>
+ *
+ * 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; under version 2 of the License.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+/*
+ * TODO: add some xenbus / xenstore concepts overview here.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <sys/signal.h>
+
+#include <xs.h>
+#include <xenctrl.h>
+#include <xen/grant_table.h>
+
+#include "hw.h"
+#include "qemu-char.h"
+#include "xen_backend.h"
+
+/* ------------------------------------------------------------- */
+
+/* public */
+int xen_xc;
+struct xs_handle *xenstore = NULL;
+
+/* private */
+static TAILQ_HEAD(XenDeviceHead, XenDevice) xendevs =
TAILQ_HEAD_INITIALIZER(xendevs);
+static int debug = 0;
+
+/* ------------------------------------------------------------- */
+
+int xenstore_write_str(const char *base, const char *node, const char *val)
+{
+ char abspath[XEN_BUFSIZE];
+
+ snprintf(abspath, sizeof(abspath), "%s/%s", base, node);
+ if (!xs_write(xenstore, 0, abspath, val, strlen(val)))
+ return -1;
+ return 0;
+}
+
+char *xenstore_read_str(const char *base, const char *node)
+{
+ char abspath[XEN_BUFSIZE];
+ unsigned int len;
+
+ snprintf(abspath, sizeof(abspath), "%s/%s", base, node);
+ return xs_read(xenstore, 0, abspath, &len);
+}
+
+int xenstore_write_int(const char *base, const char *node, int ival)
+{
+ char val[32];
+
+ snprintf(val, sizeof(val), "%d", ival);
+ return xenstore_write_str(base, node, val);
+}
+
+int xenstore_read_int(const char *base, const char *node, int *ival)
+{
+ char *val;
+ int rc = -1;
+
+ val = xenstore_read_str(base, node);
+ if (val && 1 == sscanf(val, "%d", ival))
+ rc = 0;
+ qemu_free(val);
+ return rc;
+}
+
+int xenstore_write_be_str(struct XenDevice *xendev, const char *node, const
char *val)
+{
+ return xenstore_write_str(xendev->be, node, val);
+}
+
+int xenstore_write_be_int(struct XenDevice *xendev, const char *node, int ival)
+{
+ return xenstore_write_int(xendev->be, node, ival);
+}
+
+char *xenstore_read_be_str(struct XenDevice *xendev, const char *node)
+{
+ return xenstore_read_str(xendev->be, node);
+}
+
+int xenstore_read_be_int(struct XenDevice *xendev, const char *node, int *ival)
+{
+ return xenstore_read_int(xendev->be, node, ival);
+}
+
+char *xenstore_read_fe_str(struct XenDevice *xendev, const char *node)
+{
+ return xenstore_read_str(xendev->fe, node);
+}
+
+int xenstore_read_fe_int(struct XenDevice *xendev, const char *node, int *ival)
+{
+ return xenstore_read_int(xendev->fe, node, ival);
+}
+
+/* ------------------------------------------------------------- */
+
+const char *xenbus_strstate(enum xenbus_state state)
+{
+ static const char *const name[] = {
+ [ XenbusStateUnknown ] = "Unknown",
+ [ XenbusStateInitialising ] = "Initialising",
+ [ XenbusStateInitWait ] = "InitWait",
+ [ XenbusStateInitialised ] = "Initialised",
+ [ XenbusStateConnected ] = "Connected",
+ [ XenbusStateClosing ] = "Closing",
+ [ XenbusStateClosed ] = "Closed",
+ };
+ return (state < ARRAY_SIZE(name)) ? name[state] : "INVALID";
+}
+
+int xen_be_set_state(struct XenDevice *xendev, enum xenbus_state state)
+{
+ int rc;
+
+ rc = xenstore_write_be_int(xendev, "state", state);
+ if (rc < 0)
+ return rc;
+ xen_be_printf(xendev, 1, "backend state: %s -> %s\n",
+ xenbus_strstate(xendev->be_state), xenbus_strstate(state));
+ xendev->be_state = state;
+ return 0;
+}
+
+/* ------------------------------------------------------------- */
+
+struct XenDevice *xen_be_find_xendev(const char *type, int dom, int dev)
+{
+ struct XenDevice *xendev;
+
+ TAILQ_FOREACH(xendev, &xendevs, next) {
+ if (xendev->dom != dom)
+ continue;
+ if (xendev->dev != dev)
+ continue;
+ if (0 != strcmp(xendev->type, type))
+ continue;
+ return xendev;
+ }
+ return NULL;
+}
+
+/*
+ * get xen backend device, allocate a new one if it doesn''t exist.
+ */
+static struct XenDevice *xen_be_get_xendev(const char *type, int dom, int dev,
+ struct XenDevOps *ops)
+{
+ struct XenDevice *xendev;
+ char *dom0;
+
+ xendev = xen_be_find_xendev(type, dom, dev);
+ if (xendev)
+ return xendev;
+
+ /* init new xendev */
+ xendev = qemu_mallocz(ops->size);
+ if (!xendev)
+ return NULL;
+ xendev->type = type;
+ xendev->dom = dom;
+ xendev->dev = dev;
+ xendev->ops = ops;
+
+ dom0 = xs_get_domain_path(xenstore, 0);
+ snprintf(xendev->be, sizeof(xendev->be),
"%s/backend/%s/%d/%d",
+ dom0, xendev->type, xendev->dom, xendev->dev);
+ snprintf(xendev->name, sizeof(xendev->name), "%s-%d",
+ xendev->type, xendev->dev);
+ free(dom0);
+
+ xendev->debug = debug;
+ xendev->local_port = -1;
+
+ xendev->evtchndev = xc_evtchn_open();
+ if (xendev->evtchndev < 0) {
+ fprintf(stderr, "can''t open evtchn device\n");
+ qemu_free(xendev);
+ return NULL;
+ }
+ fcntl(xc_evtchn_fd(xendev->evtchndev), F_SETFD, FD_CLOEXEC);
+
+ if (ops->flags & DEVOPS_FLAG_NEED_GNTDEV) {
+ xendev->gnttabdev = xc_gnttab_open();
+ if (xendev->gnttabdev < 0) {
+ fprintf(stderr, "can''t open gnttab device\n");
+ xc_evtchn_close(xendev->evtchndev);
+ qemu_free(xendev);
+ return NULL;
+ }
+ } else {
+ xendev->gnttabdev = -1;
+ }
+
+ TAILQ_INSERT_TAIL(&xendevs, xendev, next);
+
+ if (xendev->ops->alloc)
+ xendev->ops->alloc(xendev);
+
+ return xendev;
+}
+
+/*
+ * release xen backend device.
+ */
+static struct XenDevice *xen_be_del_xendev(int dom, int dev)
+{
+ struct XenDevice *xendev, *xnext;
+
+ /*
+ * This is pretty much like TAILQ_FOREACH(xendev, &xendevs, next) but
+ * we save the next pointer in xnext because we might free xendev.
+ */
+ xnext = xendevs.tqh_first;
+ while (xnext) {
+ xendev = xnext;
+ xnext = xendev->next.tqe_next;
+
+ if (xendev->dom != dom)
+ continue;
+ if (xendev->dev != dev && dev != -1)
+ continue;
+
+ if (xendev->ops->free)
+ xendev->ops->free(xendev);
+
+ if (xendev->fe) {
+ char token[XEN_BUFSIZE];
+ snprintf(token, sizeof(token), "fe:%p", xendev);
+ xs_unwatch(xenstore, xendev->fe, token);
+ qemu_free(xendev->fe);
+ }
+
+ if (xendev->evtchndev >= 0)
+ xc_evtchn_close(xendev->evtchndev);
+ if (xendev->gnttabdev >= 0)
+ xc_gnttab_close(xendev->gnttabdev);
+
+ TAILQ_REMOVE(&xendevs, xendev, next);
+ qemu_free(xendev);
+ }
+ return NULL;
+}
+
+/*
+ * Sync internal data structures on xenstore updates.
+ * Node specifies the changed field. node = NULL means
+ * update all fields (used for initialization).
+ */
+static void xen_be_backend_changed(struct XenDevice *xendev, const char *node)
+{
+ if (NULL == node || 0 == strcmp(node, "online")) {
+ if (-1 == xenstore_read_be_int(xendev, "online",
&xendev->online))
+ xendev->online = 0;
+ }
+
+ if (node) {
+ xen_be_printf(xendev, 2, "backend update: %s\n", node);
+ if (xendev->ops->backend_changed)
+ xendev->ops->backend_changed(xendev, node);
+ }
+}
+
+static void xen_be_frontend_changed(struct XenDevice *xendev, const char *node)
+{
+ int fe_state;
+
+ if (NULL == node || 0 == strcmp(node, "state")) {
+ if (-1 == xenstore_read_fe_int(xendev, "state", &fe_state))
+ fe_state = XenbusStateUnknown;
+ if (xendev->fe_state != fe_state)
+ xen_be_printf(xendev, 1, "frontend state: %s -> %s\n",
+ xenbus_strstate(xendev->fe_state),
+ xenbus_strstate(fe_state));
+ xendev->fe_state = fe_state;
+ }
+ if (NULL == node || 0 == strcmp(node, "protocol")) {
+ qemu_free(xendev->protocol);
+ xendev->protocol = xenstore_read_fe_str(xendev, "protocol");
+ if (xendev->protocol)
+ xen_be_printf(xendev, 1, "frontend protocol: %s\n",
xendev->protocol);
+ }
+
+ if (node) {
+ xen_be_printf(xendev, 2, "frontend update: %s\n", node);
+ if (xendev->ops->frontend_changed)
+ xendev->ops->frontend_changed(xendev, node);
+ }
+}
+
+/* ------------------------------------------------------------- */
+/* Check for possible state transitions and perform them. */
+
+/*
+ * Initial xendev setup. Read frontend path, register watch for it.
+ * Should succeed once xend finished setting up the backend device.
+ *
+ * Also sets initial state (-> Initializing) when done. Which
+ * only affects the xendev->be_state variable as xenbus should
+ * already be put into that state by xend.
+ */
+static int xen_be_try_setup(struct XenDevice *xendev)
+{
+ char token[XEN_BUFSIZE];
+ int be_state;
+
+ if (-1 == xenstore_read_be_int(xendev, "state", &be_state)) {
+ xen_be_printf(xendev, 0, "reading backend state failed\n");
+ return -1;
+ }
+
+ if (be_state != XenbusStateInitialising) {
+ xen_be_printf(xendev, 0, "initial backend state is wrong (%s)\n",
+ xenbus_strstate(be_state));
+ return -1;
+ }
+
+ xendev->fe = xenstore_read_be_str(xendev, "frontend");
+ if (NULL == xendev->fe) {
+ xen_be_printf(xendev, 0, "reading frontend path failed\n");
+ return -1;
+ }
+
+ /* setup frontend watch */
+ snprintf(token, sizeof(token), "fe:%p", xendev);
+ if (!xs_watch(xenstore, xendev->fe, token)) {
+ xen_be_printf(xendev, 0, "watching frontend path (%s) failed\n",
+ xendev->fe);
+ return -1;
+ }
+ xen_be_set_state(xendev, XenbusStateInitialising);
+
+ xen_be_backend_changed(xendev, NULL);
+ xen_be_frontend_changed(xendev, NULL);
+ return 0;
+}
+
+/*
+ * Try initialize xendev. Prepare everything the backend can do
+ * without synchronizing with the frontend. Fakes hotplug-status. No
+ * hotplug involved here because this is about userspace drivers, thus
+ * there are kernel backend devices which could invoke hotplug.
+ *
+ * Goes to InitWait on success.
+ */
+static int xen_be_try_init(struct XenDevice *xendev)
+{
+ int rc = 0;
+
+ if (!xendev->online) {
+ xen_be_printf(xendev, 1, "not online\n");
+ return -1;
+ }
+
+ if (xendev->ops->init)
+ rc = xendev->ops->init(xendev);
+ if (0 != rc) {
+ xen_be_printf(xendev, 1, "init() failed\n");
+ return rc;
+ }
+
+ xenstore_write_be_str(xendev, "hotplug-status",
"connected");
+ xen_be_set_state(xendev, XenbusStateInitWait);
+ return 0;
+}
+
+/*
+ * Try to connect xendev. Depends on the frontend being ready
+ * for it (shared ring and evtchn info in xenstore, state being
+ * Initialised or Connected).
+ *
+ * Goes to Connected on success.
+ */
+static int xen_be_try_connect(struct XenDevice *xendev)
+{
+ int rc = 0;
+
+ if (xendev->fe_state != XenbusStateInitialised &&
+ xendev->fe_state != XenbusStateConnected) {
+ if (xendev->ops->flags & DEVOPS_FLAG_IGNORE_STATE) {
+ xen_be_printf(xendev, 2, "frontend not ready, ignoring\n");
+ } else {
+ xen_be_printf(xendev, 2, "frontend not ready (yet)\n");
+ return -1;
+ }
+ }
+
+ if (xendev->ops->connect)
+ rc = xendev->ops->connect(xendev);
+ if (0 != rc) {
+ xen_be_printf(xendev, 0, "connect() failed\n");
+ return rc;
+ }
+
+ xen_be_set_state(xendev, XenbusStateConnected);
+ return 0;
+}
+
+/*
+ * Teardown connection.
+ *
+ * Goes to Closed when done.
+ */
+static void xen_be_disconnect(struct XenDevice *xendev, enum xenbus_state
state)
+{
+ if (xendev->be_state != XenbusStateClosing &&
+ xendev->be_state != XenbusStateClosed &&
+ xendev->ops->disconnect)
+ xendev->ops->disconnect(xendev);
+ if (xendev->be_state != state)
+ xen_be_set_state(xendev, state);
+}
+
+/*
+ * Try to reset xendev, for reconnection by another frontend instance.
+ */
+static int xen_be_try_reset(struct XenDevice *xendev)
+{
+ if (xendev->fe_state != XenbusStateInitialising)
+ return -1;
+
+ xen_be_printf(xendev, 1, "device reset (for re-connect)\n");
+ xen_be_set_state(xendev, XenbusStateInitialising);
+ return 0;
+}
+
+/*
+ * state change dispatcher function
+ */
+void xen_be_check_state(struct XenDevice *xendev)
+{
+ int rc = 0;
+
+ /* frontend may request shutdown from almost anywhere */
+ if (xendev->fe_state == XenbusStateClosing ||
+ xendev->fe_state == XenbusStateClosed) {
+ xen_be_disconnect(xendev, xendev->fe_state);
+ return;
+ }
+
+ /* check for possible backend state transitions */
+ for (;;) {
+ switch (xendev->be_state) {
+ case XenbusStateUnknown:
+ rc = xen_be_try_setup(xendev);
+ break;
+ case XenbusStateInitialising:
+ rc = xen_be_try_init(xendev);
+ break;
+ case XenbusStateInitWait:
+ rc = xen_be_try_connect(xendev);
+ break;
+ case XenbusStateClosed:
+ rc = xen_be_try_reset(xendev);
+ break;
+ default:
+ rc = -1;
+ }
+ if (0 != rc)
+ break;
+ }
+}
+
+/* ------------------------------------------------------------- */
+
+static int xenstore_scan(const char *type, int dom, struct XenDevOps *ops)
+{
+ struct XenDevice *xendev;
+ char path[XEN_BUFSIZE], token[XEN_BUFSIZE];
+ char **dev = NULL, *dom0;
+ unsigned int cdev, j;
+
+ /* setup watch */
+ dom0 = xs_get_domain_path(xenstore, 0);
+ snprintf(token, sizeof(token), "be:%p:%d:%p", type, dom, ops);
+ snprintf(path, sizeof(path), "%s/backend/%s/%d", dom0, type,
dom);
+ free(dom0);
+ if (!xs_watch(xenstore, path, token)) {
+ fprintf(stderr, "xen be: watching backend path (%s) failed\n",
path);
+ return -1;
+ }
+
+ /* look for backends */
+ dev = xs_directory(xenstore, 0, path, &cdev);
+ if (!dev)
+ return 0;
+ for (j = 0; j < cdev; j++) {
+ xendev = xen_be_get_xendev(type, dom, atoi(dev[j]), ops);
+ if (NULL == xendev)
+ continue;
+ xen_be_check_state(xendev);
+ }
+ qemu_free(dev);
+ return 0;
+}
+
+static void xenstore_update_be(char *watch, char *type, int dom,
+ struct XenDevOps *ops)
+{
+ struct XenDevice *xendev;
+ char path[XEN_BUFSIZE], *dom0;
+ unsigned int len, dev;
+
+ dom0 = xs_get_domain_path(xenstore, 0);
+ len = snprintf(path, sizeof(path), "%s/backend/%s/%d", dom0,
type, dom);
+ free(dom0);
+ if (0 != strncmp(path, watch, len))
+ return;
+ if (2 != sscanf(watch+len, "/%u/%255s", &dev, path)) {
+ strcpy(path, "");
+ if (1 != sscanf(watch+len, "/%u", &dev))
+ dev = -1;
+ }
+ if (-1 == dev)
+ return;
+
+ if (0) {
+ /* FIXME: detect devices being deleted from xenstore ... */
+ xen_be_del_xendev(dom, dev);
+ }
+
+ xendev = xen_be_get_xendev(type, dom, dev, ops);
+ if (NULL != xendev) {
+ xen_be_backend_changed(xendev, path);
+ xen_be_check_state(xendev);
+ }
+}
+
+static void xenstore_update_fe(char *watch, struct XenDevice *xendev)
+{
+ char *node;
+ unsigned int len;
+
+ len = strlen(xendev->fe);
+ if (0 != strncmp(xendev->fe, watch, len))
+ return;
+ if (watch[len] != ''/'')
+ return;
+ node = watch + len + 1;
+
+ xen_be_frontend_changed(xendev, node);
+ xen_be_check_state(xendev);
+}
+
+static void xenstore_update(void *unused)
+{
+ char **vec = NULL;
+ intptr_t type, ops, ptr;
+ unsigned int dom, count;
+
+ vec = xs_read_watch(xenstore, &count);
+ if (NULL == vec)
+ goto cleanup;
+
+ if (3 == sscanf(vec[XS_WATCH_TOKEN], "be:%" PRIxPTR
":%d:%" PRIxPTR,
+ &type, &dom, &ops))
+ xenstore_update_be(vec[XS_WATCH_PATH], (void*)type, dom, (void*)ops);
+ if (1 == sscanf(vec[XS_WATCH_TOKEN], "fe:%" PRIxPTR, &ptr))
+ xenstore_update_fe(vec[XS_WATCH_PATH], (void*)ptr);
+
+cleanup:
+ qemu_free(vec);
+}
+
+static void xen_be_evtchn_event(void *opaque)
+{
+ struct XenDevice *xendev = opaque;
+ evtchn_port_t port;
+
+ port = xc_evtchn_pending(xendev->evtchndev);
+ if (port != xendev->local_port) {
+ xen_be_printf(xendev, 0, "xc_evtchn_pending returned %d (expected
%d)\n",
+ port, xendev->local_port);
+ return;
+ }
+ xc_evtchn_unmask(xendev->evtchndev, port);
+
+ if (xendev->ops->event)
+ xendev->ops->event(xendev);
+}
+
+/* -------------------------------------------------------------------- */
+
+int xen_be_init(void)
+{
+ xenstore = xs_daemon_open();
+ if (!xenstore) {
+ fprintf(stderr, "can''t connect to xenstored\n");
+ return -1;
+ }
+
+ if (qemu_set_fd_handler(xs_fileno(xenstore), xenstore_update, NULL, NULL)
< 0)
+ goto err;
+
+ xen_xc = xc_interface_open();
+ if (-1 == xen_xc) {
+ fprintf(stderr, "can''t open xen interface\n");
+ goto err;
+ }
+ return 0;
+
+err:
+ qemu_set_fd_handler(xs_fileno(xenstore), NULL, NULL, NULL);
+ xs_daemon_close(xenstore);
+ xenstore = NULL;
+
+ return -1;
+}
+
+int xen_be_register(const char *type, struct XenDevOps *ops)
+{
+ return xenstore_scan(type, xen_domid, ops);
+}
+
+int xen_be_bind_evtchn(struct XenDevice *xendev)
+{
+ if (xendev->local_port != -1)
+ return 0;
+ xendev->local_port = xc_evtchn_bind_interdomain
+ (xendev->evtchndev, xendev->dom, xendev->remote_port);
+ if (-1 == xendev->local_port) {
+ xen_be_printf(xendev, 0, "xc_evtchn_bind_interdomain failed\n");
+ return -1;
+ }
+ xen_be_printf(xendev, 2, "bind evtchn port %d\n",
xendev->local_port);
+ qemu_set_fd_handler(xc_evtchn_fd(xendev->evtchndev),
+ xen_be_evtchn_event, NULL, xendev);
+ return 0;
+}
+
+void xen_be_unbind_evtchn(struct XenDevice *xendev)
+{
+ if (xendev->local_port == -1)
+ return;
+ qemu_set_fd_handler(xc_evtchn_fd(xendev->evtchndev), NULL, NULL, NULL);
+ xc_evtchn_unbind(xendev->evtchndev, xendev->local_port);
+ xen_be_printf(xendev, 2, "unbind evtchn port %d\n",
xendev->local_port);
+ xendev->local_port = -1;
+}
+
+int xen_be_send_notify(struct XenDevice *xendev)
+{
+ return xc_evtchn_notify(xendev->evtchndev, xendev->local_port);
+}
+
+/*
+ * msg_level:
+ * 0 == errors.
+ * 1 == informative debug messages.
+ * 2 == noisy debug messages.
+ * 3 == will flood your log.
+ */
+void xen_be_printf(struct XenDevice *xendev, int msg_level, const char *fmt,
...)
+{
+ va_list args;
+
+ if (msg_level > xendev->debug)
+ return;
+ fprintf(stderr, "xen be: %s: ", xendev->name);
+ va_start(args, fmt);
+ vfprintf(stderr, fmt, args);
+ va_end(args);
+}
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
new file mode 100644
index 0000000..a1243f6
--- /dev/null
+++ b/hw/xen_backend.h
@@ -0,0 +1,86 @@
+#ifndef QEMU_HW_XEN_BACKEND_H
+#define QEMU_HW_XEN_BACKEND_H 1
+
+#include "xen_common.h"
+
+/* ------------------------------------------------------------- */
+
+#define XEN_BUFSIZE 1024
+
+struct XenDevice;
+
+/* driver uses grant tables -> open gntdev device (xendev->gnttabdev)
*/
+#define DEVOPS_FLAG_NEED_GNTDEV 1
+/* don''t expect frontend doing correct state transitions (aka console
quirk) */
+#define DEVOPS_FLAG_IGNORE_STATE 2
+
+struct XenDevOps {
+ size_t size;
+ uint32_t flags;
+ void (*alloc)(struct XenDevice *xendev);
+ int (*init)(struct XenDevice *xendev);
+ int (*connect)(struct XenDevice *xendev);
+ void (*event)(struct XenDevice *xendev);
+ void (*disconnect)(struct XenDevice *xendev);
+ int (*free)(struct XenDevice *xendev);
+ void (*backend_changed)(struct XenDevice *xendev, const char *node);
+ void (*frontend_changed)(struct XenDevice *xendev, const char *node);
+};
+
+struct XenDevice {
+ const char *type;
+ int dom;
+ int dev;
+ char name[64];
+ int debug;
+
+ enum xenbus_state be_state;
+ enum xenbus_state fe_state;
+ int online;
+ char be[XEN_BUFSIZE];
+ char *fe;
+ char *protocol;
+ int remote_port;
+ int local_port;
+
+ int evtchndev;
+ int gnttabdev;
+
+ struct XenDevOps *ops;
+ TAILQ_ENTRY(XenDevice) next;
+};
+
+/* ------------------------------------------------------------- */
+
+/* variables */
+extern int xen_xc;
+extern struct xs_handle *xenstore;
+
+/* xenstore helper functions */
+int xenstore_write_str(const char *base, const char *node, const char *val);
+int xenstore_write_int(const char *base, const char *node, int ival);
+char *xenstore_read_str(const char *base, const char *node);
+int xenstore_read_int(const char *base, const char *node, int *ival);
+
+int xenstore_write_be_str(struct XenDevice *xendev, const char *node, const
char *val);
+int xenstore_write_be_int(struct XenDevice *xendev, const char *node, int
ival);
+char *xenstore_read_be_str(struct XenDevice *xendev, const char *node);
+int xenstore_read_be_int(struct XenDevice *xendev, const char *node, int
*ival);
+char *xenstore_read_fe_str(struct XenDevice *xendev, const char *node);
+int xenstore_read_fe_int(struct XenDevice *xendev, const char *node, int
*ival);
+
+const char *xenbus_strstate(enum xenbus_state state);
+struct XenDevice *xen_be_find_xendev(const char *type, int dom, int dev);
+void xen_be_check_state(struct XenDevice *xendev);
+
+/* xen backend driver bits */
+int xen_be_init(void);
+int xen_be_register(const char *type, struct XenDevOps *ops);
+int xen_be_set_state(struct XenDevice *xendev, enum xenbus_state state);
+int xen_be_bind_evtchn(struct XenDevice *xendev);
+void xen_be_unbind_evtchn(struct XenDevice *xendev);
+int xen_be_send_notify(struct XenDevice *xendev);
+void xen_be_printf(struct XenDevice *xendev, int msg_level, const char *fmt,
...)
+ __attribute__ ((format(printf, 3, 4)));
+
+#endif /* QEMU_HW_XEN_BACKEND_H */
diff --git a/hw/xen_common.h b/hw/xen_common.h
new file mode 100644
index 0000000..7562567
--- /dev/null
+++ b/hw/xen_common.h
@@ -0,0 +1,34 @@
+#ifndef QEMU_HW_XEN_COMMON_H
+#define QEMU_HW_XEN_COMMON_H 1
+
+#include <stddef.h>
+#include <inttypes.h>
+
+#include <xenctrl.h>
+#include <xs.h>
+#include <xen/io/xenbus.h>
+
+#include "hw.h"
+#include "xen.h"
+#include "sys-queue.h" /* BSD list implementation */
+
+/*
+ * tweaks needed to build with different xen versions
+ * 0x00030205 -> 3.1.0
+ * 0x00030207 -> 3.2.0
+ * 0x00030208 -> unstable
+ */
+#include <xen/xen-compat.h>
+#if __XEN_LATEST_INTERFACE_VERSION__ < 0x00030205
+# define evtchn_port_or_error_t int
+#endif
+#if __XEN_LATEST_INTERFACE_VERSION__ < 0x00030207
+# define xc_map_foreign_pages xc_map_foreign_batch
+#endif
+#if __XEN_LATEST_INTERFACE_VERSION__ < 0x00030208
+# define xen_mb() mb()
+# define xen_rmb() rmb()
+# define xen_wmb() wmb()
+#endif
+
+#endif /* QEMU_HW_XEN_COMMON_H */
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index 9daab75..67f997f 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -25,8 +25,7 @@
#include "hw.h"
#include "boards.h"
-#include "xen.h"
-#include <xenctrl.h>
+#include "xen_backend.h"
/* -------------------------------------------------------------------- */
/* variables */
@@ -48,6 +47,11 @@ static int xen_init(void)
fprintf(stderr, "%s: emulating Xen\n", __FUNCTION__);
}
+ if (-1 == xen_be_init()) {
+ fprintf(stderr, "%s: xen backend core setup failed\n",
__FUNCTION__);
+ return -1;
+ }
+
return 0;
}
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 12:23 UTC
[Xen-devel] [PATCH 3/7] xen: add console backend driver.
This patch adds a xenconsole backend driver. It it based on current
xen-unstable code. It has been changed to make use of the common
backend driver code.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 1 +
hw/xen_backend.h | 3 +
hw/xen_console.c | 276 +++++++++++++++++++++++++++++++++++++++++++++++++++
hw/xen_machine_pv.c | 3 +
4 files changed, 283 insertions(+), 0 deletions(-)
create mode 100644 hw/xen_console.c
diff --git a/Makefile.target b/Makefile.target
index c18cbc6..50f1dca 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -631,6 +631,7 @@ endif
# xen backend driver support
XEN_OBJS := xen_machine_pv.o xen_backend.o
+XEN_OBJS += xen_console.o
ifeq ($(CONFIG_XEN), yes)
OBJS += $(XEN_OBJS)
LIBS += $(XEN_LIBS)
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
index a1243f6..4744713 100644
--- a/hw/xen_backend.h
+++ b/hw/xen_backend.h
@@ -83,4 +83,7 @@ int xen_be_send_notify(struct XenDevice *xendev);
void xen_be_printf(struct XenDevice *xendev, int msg_level, const char *fmt,
...)
__attribute__ ((format(printf, 3, 4)));
+/* actual backend drivers */
+extern struct XenDevOps xen_console_ops; /* xen_console.c */
+
#endif /* QEMU_HW_XEN_BACKEND_H */
diff --git a/hw/xen_console.c b/hw/xen_console.c
new file mode 100644
index 0000000..c172cf2
--- /dev/null
+++ b/hw/xen_console.c
@@ -0,0 +1,276 @@
+/*
+ * Copyright (C) International Business Machines Corp., 2005
+ * Author(s): Anthony Liguori <aliguori@us.ibm.com>
+ *
+ * Copyright (C) Red Hat 2007
+ *
+ * Xen Console
+ *
+ * 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; under version 2 of the License.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#include <stdlib.h>
+#include <errno.h>
+#include <string.h>
+#include <sys/select.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <termios.h>
+#include <stdarg.h>
+#include <sys/mman.h>
+#include <xs.h>
+#include <xen/io/console.h>
+#include <xenctrl.h>
+
+#include "hw.h"
+#include "sysemu.h"
+#include "qemu-char.h"
+#include "xen_backend.h"
+
+#define dolog(val, fmt, ...) fprintf(stderr, fmt "\n", ##
__VA_ARGS__)
+
+struct buffer {
+ uint8_t *data;
+ size_t consumed;
+ size_t size;
+ size_t capacity;
+ size_t max_capacity;
+};
+
+struct XenConsole {
+ struct XenDevice xendev; /* must be first */
+ struct buffer buffer;
+ char console[XEN_BUFSIZE];
+ int ring_ref;
+ void *sring;
+ CharDriverState *chr;
+ int backlog;
+};
+
+static void buffer_append(struct XenConsole *con)
+{
+ struct buffer *buffer = &con->buffer;
+ XENCONS_RING_IDX cons, prod, size;
+ struct xencons_interface *intf = con->sring;
+
+ cons = intf->out_cons;
+ prod = intf->out_prod;
+ xen_mb();
+
+ size = prod - cons;
+ if ((size == 0) || (size > sizeof(intf->out)))
+ return;
+
+ if ((buffer->capacity - buffer->size) < size) {
+ buffer->capacity += (size + 1024);
+ buffer->data = qemu_realloc(buffer->data, buffer->capacity);
+ if (buffer->data == NULL) {
+ dolog(LOG_ERR, "Memory allocation failed");
+ exit(ENOMEM);
+ }
+ }
+
+ while (cons != prod)
+ buffer->data[buffer->size++] = intf->out[
+ MASK_XENCONS_IDX(cons++, intf->out)];
+
+ xen_mb();
+ intf->out_cons = cons;
+ xen_be_send_notify(&con->xendev);
+
+ if (buffer->max_capacity &&
+ buffer->size > buffer->max_capacity) {
+ /* Discard the middle of the data. */
+
+ size_t over = buffer->size - buffer->max_capacity;
+ uint8_t *maxpos = buffer->data + buffer->max_capacity;
+
+ memmove(maxpos - over, maxpos, over);
+ buffer->data = qemu_realloc(buffer->data, buffer->max_capacity);
+ buffer->size = buffer->capacity = buffer->max_capacity;
+
+ if (buffer->consumed > buffer->max_capacity - over)
+ buffer->consumed = buffer->max_capacity - over;
+ }
+}
+
+static void buffer_advance(struct buffer *buffer, size_t len)
+{
+ buffer->consumed += len;
+ if (buffer->consumed == buffer->size) {
+ buffer->consumed = 0;
+ buffer->size = 0;
+ }
+}
+
+static int ring_free_bytes(struct XenConsole *con)
+{
+ struct xencons_interface *intf = con->sring;
+ XENCONS_RING_IDX cons, prod, space;
+
+ cons = intf->in_cons;
+ prod = intf->in_prod;
+ xen_mb();
+
+ space = prod - cons;
+ if (space > sizeof(intf->in))
+ return 0; /* ring is screwed: ignore it */
+
+ return (sizeof(intf->in) - space);
+}
+
+static int xencons_can_receive(void *opaque)
+{
+ struct XenConsole *con = opaque;
+ return ring_free_bytes(con);
+}
+
+static void xencons_receive(void *opaque, const uint8_t *buf, int len)
+{
+ struct XenConsole *con = opaque;
+ struct xencons_interface *intf = con->sring;
+ XENCONS_RING_IDX prod;
+ int i, max;
+
+ max = ring_free_bytes(con);
+ /* The can_receive() func limits this, but check again anyway */
+ if (max < len)
+ len = max;
+
+ prod = intf->in_prod;
+ for (i = 0; i < len; i++) {
+ intf->in[MASK_XENCONS_IDX(prod++, intf->in)] + buf[i];
+ }
+ xen_wmb();
+ intf->in_prod = prod;
+ xen_be_send_notify(&con->xendev);
+}
+
+static void xencons_send(struct XenConsole *con)
+{
+ ssize_t len, size;
+
+ size = con->buffer.size - con->buffer.consumed;
+ if (con->chr)
+ len = qemu_chr_write(con->chr, con->buffer.data +
con->buffer.consumed,
+ size);
+ else
+ len = size;
+ if (len < 1) {
+ if (!con->backlog) {
+ con->backlog = 1;
+ xen_be_printf(&con->xendev, 1, "backlog piling up, nobody
listening?\n");
+ }
+ } else {
+ buffer_advance(&con->buffer, len);
+ if (con->backlog && len == size) {
+ con->backlog = 0;
+ xen_be_printf(&con->xendev, 1, "backlog is gone\n");
+ }
+ }
+}
+
+/* -------------------------------------------------------------------- */
+
+static int con_init(struct XenDevice *xendev)
+{
+ struct XenConsole *con = container_of(xendev, struct XenConsole, xendev);
+ char *type, *dom;
+
+ /* setup */
+ dom = xs_get_domain_path(xenstore, con->xendev.dom);
+ snprintf(con->console, sizeof(con->console), "%s/console",
dom);
+ free(dom);
+
+ type = xenstore_read_str(con->console, "type");
+ if (!type || 0 != strcmp(type, "ioemu")) {
+ xen_be_printf(xendev, 1, "not for me (type=%s)\n", type);
+ return -1;
+ }
+
+ if (!serial_hds[con->xendev.dev])
+ xen_be_printf(xendev, 1, "WARNING: serial line %d not configured\n",
+ con->xendev.dev);
+ else
+ con->chr = serial_hds[con->xendev.dev];
+
+ return 0;
+}
+
+static int con_connect(struct XenDevice *xendev)
+{
+ struct XenConsole *con = container_of(xendev, struct XenConsole, xendev);
+ int limit;
+
+ if (-1 == xenstore_read_int(con->console, "ring-ref",
&con->ring_ref))
+ return -1;
+ if (-1 == xenstore_read_int(con->console, "port",
&con->xendev.remote_port))
+ return -1;
+ if (0 == xenstore_read_int(con->console, "limit", &limit))
+ con->buffer.max_capacity = limit;
+
+ con->sring = xc_map_foreign_range(xen_xc, con->xendev.dom,
+ XC_PAGE_SIZE,
+ PROT_READ|PROT_WRITE,
+ con->ring_ref);
+ if (!con->sring)
+ return -1;
+
+ xen_be_bind_evtchn(&con->xendev);
+ if (con->chr)
+ qemu_chr_add_handlers(con->chr, xencons_can_receive,
xencons_receive,
+ NULL, con);
+
+ xen_be_printf(xendev, 1, "ring mfn %d, remote port %d, local port %d,
limit %zd\n",
+ con->ring_ref,
+ con->xendev.remote_port,
+ con->xendev.local_port,
+ con->buffer.max_capacity);
+ return 0;
+}
+
+static void con_disconnect(struct XenDevice *xendev)
+{
+ struct XenConsole *con = container_of(xendev, struct XenConsole, xendev);
+
+ if (con->chr)
+ qemu_chr_add_handlers(con->chr, NULL, NULL, NULL, NULL);
+ xen_be_unbind_evtchn(&con->xendev);
+
+ if (con->sring) {
+ munmap(con->sring, XC_PAGE_SIZE);
+ con->sring = NULL;
+ }
+}
+
+static void con_event(struct XenDevice *xendev)
+{
+ struct XenConsole *con = container_of(xendev, struct XenConsole, xendev);
+
+ buffer_append(con);
+ if (con->buffer.size - con->buffer.consumed)
+ xencons_send(con);
+}
+
+/* -------------------------------------------------------------------- */
+
+struct XenDevOps xen_console_ops = {
+ .size = sizeof(struct XenConsole),
+ .flags = DEVOPS_FLAG_IGNORE_STATE,
+ .init = con_init,
+ .connect = con_connect,
+ .event = con_event,
+ .disconnect = con_disconnect,
+};
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index 67f997f..fca2e09 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -63,6 +63,9 @@ static int xen_init_pv(DisplayState *ds)
if (rc < 0)
return rc;
+ /* xenbus backend drivers */
+ xen_be_register("console", &xen_console_ops);
+
return 0;
}
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 12:23 UTC
[Xen-devel] [PATCH 4/7] xen: add framebuffer backend driver
This patch adds a frsamebuffer (and kbd+mouse) backend driver. It
it based on current xen-unstable code. It has been changed to make
use of the common backend driver code. It also has been changed to
compile with xen headers older than unstable (aka son-to-be 3.3).
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 2 +-
hw/xen_backend.h | 4 +
hw/xen_machine_pv.c | 5 +
hw/xenfb.c | 977 +++++++++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 987 insertions(+), 1 deletions(-)
create mode 100644 hw/xenfb.c
diff --git a/Makefile.target b/Makefile.target
index 50f1dca..7606a20 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -631,7 +631,7 @@ endif
# xen backend driver support
XEN_OBJS := xen_machine_pv.o xen_backend.o
-XEN_OBJS += xen_console.o
+XEN_OBJS += xen_console.o xenfb.o
ifeq ($(CONFIG_XEN), yes)
OBJS += $(XEN_OBJS)
LIBS += $(XEN_LIBS)
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
index 4744713..ceeb3aa 100644
--- a/hw/xen_backend.h
+++ b/hw/xen_backend.h
@@ -85,5 +85,9 @@ void xen_be_printf(struct XenDevice *xendev, int msg_level,
const char *fmt, ...
/* actual backend drivers */
extern struct XenDevOps xen_console_ops; /* xen_console.c */
+extern struct XenDevOps xen_kbdmouse_ops; /* xen_framebuffer.c */
+extern struct XenDevOps xen_framebuffer_ops; /* xen_framebuffer.c */
+
+void xen_set_display(int domid, DisplayState *ds);
#endif /* QEMU_HW_XEN_BACKEND_H */
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index fca2e09..38e7335 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -65,6 +65,11 @@ static int xen_init_pv(DisplayState *ds)
/* xenbus backend drivers */
xen_be_register("console", &xen_console_ops);
+ xen_be_register("vkbd", &xen_kbdmouse_ops);
+ xen_be_register("vfb", &xen_framebuffer_ops);
+
+ /* setup framebuffer */
+ xen_set_display(xen_domid, ds);
return 0;
}
diff --git a/hw/xenfb.c b/hw/xenfb.c
new file mode 100644
index 0000000..1b26b95
--- /dev/null
+++ b/hw/xenfb.c
@@ -0,0 +1,977 @@
+/*
+ * xen paravirt framebuffer backend
+ *
+ * Copyright IBM, Corp. 2005-2006
+ * Copyright Red Hat, Inc. 2006-2008
+ *
+ * Authors:
+ * Anthony Liguori <aliguori@us.ibm.com>,
+ * Markus Armbruster <armbru@redhat.com>,
+ * Daniel P. Berrange <berrange@redhat.com>,
+ * Pat Campbell <plc@novell.com>,
+ * Gerd Hoffmann <kraxel@redhat.com>
+ *
+ * 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; under version 2 of the License.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#include <stdarg.h>
+#include <stdlib.h>
+#include <sys/types.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <stdbool.h>
+#include <sys/mman.h>
+#include <errno.h>
+#include <stdio.h>
+#include <string.h>
+#include <time.h>
+
+#include <xs.h>
+#include <xenctrl.h>
+#include <xen/event_channel.h>
+#include <xen/io/xenbus.h>
+#include <xen/io/fbif.h>
+#include <xen/io/kbdif.h>
+#include <xen/io/protocols.h>
+
+#include "hw.h"
+#include "sysemu.h"
+#include "console.h"
+#include "qemu-char.h"
+#include "xen_backend.h"
+
+#ifndef BTN_LEFT
+#define BTN_LEFT 0x110 /* from <linux/input.h> */
+#endif
+
+/* -------------------------------------------------------------------- */
+
+struct common {
+ struct XenDevice xendev; /* must be first */
+ void *page;
+ DisplayState *ds;
+};
+
+struct XenInput {
+ struct common c;
+ int abs_pointer_wanted; /* Whether guest supports absolute pointer */
+ int button_state; /* Last seen pointer button state */
+ int extended;
+ QEMUPutMouseEntry *qmouse;
+};
+
+#define UP_QUEUE 8
+
+struct XenFB {
+ struct common c;
+ size_t fb_len;
+ int row_stride;
+ int depth;
+ int width;
+ int height;
+ int offset;
+ void *pixels;
+ int fbpages;
+ int feature_update;
+ int refresh_period;
+ int bug_trigger;
+ int have_console;
+
+ struct {
+ int x,y,w,h;
+ } up_rects[UP_QUEUE];
+ int up_count;
+ int up_fullscreen;
+};
+
+/* -------------------------------------------------------------------- */
+
+static int common_bind(struct common *c)
+{
+ int mfn;
+
+ if (-1 == xenstore_read_fe_int(&c->xendev, "page-ref",
&mfn))
+ return -1;
+ if (-1 == xenstore_read_fe_int(&c->xendev,
"event-channel", &c->xendev.remote_port))
+ return -1;
+
+ xen_be_bind_evtchn(&c->xendev);
+ xen_be_printf(&c->xendev, 1, "ring mfn %d, remote-port %d,
local-port %d\n",
+ mfn, c->xendev.remote_port, c->xendev.local_port);
+
+ c->page = xc_map_foreign_range(xen_xc, c->xendev.dom,
+ XC_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, mfn);
+ if (NULL == c->page)
+ return -1;
+ return 0;
+}
+
+static void common_unbind(struct common *c)
+{
+ if (c->page) {
+ munmap(c->page, XC_PAGE_SIZE);
+ c->page = NULL;
+ }
+ xen_be_unbind_evtchn(&c->xendev);
+}
+
+/* -------------------------------------------------------------------- */
+
+#if 0
+/*
+ * These two tables are not needed any more, but left in here
+ * intentionally as documentation, to show how scancode2linux[]
+ * was generated.
+ *
+ * Tables to map from scancode to Linux input layer keycode.
+ * Scancodes are hardware-specific. These maps assumes a
+ * standard AT or PS/2 keyboard which is what QEMU feeds us.
+ */
+static const unsigned char atkbd_set2_keycode[512] = {
+
+ 0, 67, 65, 63, 61, 59, 60, 88, 0, 68, 66, 64, 62, 15, 41,117,
+ 0, 56, 42, 93, 29, 16, 2, 0, 0, 0, 44, 31, 30, 17, 3, 0,
+ 0, 46, 45, 32, 18, 5, 4, 95, 0, 57, 47, 33, 20, 19, 6,183,
+ 0, 49, 48, 35, 34, 21, 7,184, 0, 0, 50, 36, 22, 8, 9,185,
+ 0, 51, 37, 23, 24, 11, 10, 0, 0, 52, 53, 38, 39, 25, 12, 0,
+ 0, 89, 40, 0, 26, 13, 0, 0, 58, 54, 28, 27, 0, 43, 0, 85,
+ 0, 86, 91, 90, 92, 0, 14, 94, 0, 79,124, 75, 71,121, 0, 0,
+ 82, 83, 80, 76, 77, 72, 1, 69, 87, 78, 81, 74, 55, 73, 70, 99,
+
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 217,100,255, 0, 97,165, 0, 0,156, 0, 0, 0, 0, 0, 0,125,
+ 173,114, 0,113, 0, 0, 0,126,128, 0, 0,140, 0, 0, 0,127,
+ 159, 0,115, 0,164, 0, 0,116,158, 0,150,166, 0, 0, 0,142,
+ 157, 0, 0, 0, 0, 0, 0, 0,155, 0, 98, 0, 0,163, 0, 0,
+ 226, 0, 0, 0, 0, 0, 0, 0, 0,255, 96, 0, 0, 0,143, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0,107, 0,105,102, 0, 0,112,
+ 110,111,108,112,106,103, 0,119, 0,118,109, 0, 99,104,119, 0,
+
+};
+
+static const unsigned char atkbd_unxlate_table[128] = {
+
+ 0,118, 22, 30, 38, 37, 46, 54, 61, 62, 70, 69, 78, 85,102, 13,
+ 21, 29, 36, 45, 44, 53, 60, 67, 68, 77, 84, 91, 90, 20, 28, 27,
+ 35, 43, 52, 51, 59, 66, 75, 76, 82, 14, 18, 93, 26, 34, 33, 42,
+ 50, 49, 58, 65, 73, 74, 89,124, 17, 41, 88, 5, 6, 4, 12, 3,
+ 11, 2, 10, 1, 9,119,126,108,117,125,123,107,115,116,121,105,
+ 114,122,112,113,127, 96, 97,120, 7, 15, 23, 31, 39, 47, 55, 63,
+ 71, 79, 86, 94, 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 87,111,
+ 19, 25, 57, 81, 83, 92, 95, 98, 99,100,101,103,104,106,109,110
+
+};
+#endif
+
+/*
+ * for (i = 0; i < 128; i++) {
+ * scancode2linux[i] = atkbd_set2_keycode[atkbd_unxlate_table[i]];
+ * scancode2linux[i | 0x80] = atkbd_set2_keycode[atkbd_unxlate_table[i] |
0x80];
+ * }
+ */
+static const unsigned char scancode2linux[512] = {
+ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+ 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47,
+ 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
+ 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79,
+ 80, 81, 82, 83, 99, 0, 86, 87, 88,117, 0, 0, 95,183,184,185,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 93, 0, 0, 89, 0, 0, 85, 91, 90, 92, 0, 94, 0,124,121, 0,
+
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 165, 0, 0, 0, 0, 0, 0, 0, 0,163, 0, 0, 96, 97, 0, 0,
+ 113,140,164, 0,166, 0, 0, 0, 0, 0,255, 0, 0, 0,114, 0,
+ 115, 0,150, 0, 0, 98,255, 99,100, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0,119,119,102,103,104, 0,105,112,106,118,107,
+ 108,109,110,111, 0, 0, 0, 0, 0, 0, 0,125,126,127,116,142,
+ 0, 0, 0,143, 0,217,156,173,128,159,158,157,155,226, 0,112,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+/* Send an event to the keyboard frontend driver */
+static int xenfb_kbd_event(struct XenInput *xenfb,
+ union xenkbd_in_event *event)
+{
+ struct xenkbd_page *page = xenfb->c.page;
+ uint32_t prod;
+
+ if (xenfb->c.xendev.be_state != XenbusStateConnected)
+ return 0;
+ if (!page)
+ return 0;
+
+ prod = page->in_prod;
+ if (prod - page->in_cons == XENKBD_IN_RING_LEN) {
+ errno = EAGAIN;
+ return -1;
+ }
+
+ xen_mb(); /* ensure ring space available */
+ XENKBD_IN_RING_REF(page, prod) = *event;
+ xen_wmb(); /* ensure ring contents visible */
+ page->in_prod = prod + 1;
+ return xen_be_send_notify(&xenfb->c.xendev);
+}
+
+/* Send a keyboard (or mouse button) event */
+static int xenfb_send_key(struct XenInput *xenfb, bool down, int keycode)
+{
+ union xenkbd_in_event event;
+
+ memset(&event, 0, XENKBD_IN_EVENT_SIZE);
+ event.type = XENKBD_TYPE_KEY;
+ event.key.pressed = down ? 1 : 0;
+ event.key.keycode = keycode;
+
+ return xenfb_kbd_event(xenfb, &event);
+}
+
+/* Send a relative mouse movement event */
+static int xenfb_send_motion(struct XenInput *xenfb,
+ int rel_x, int rel_y, int rel_z)
+{
+ union xenkbd_in_event event;
+
+ memset(&event, 0, XENKBD_IN_EVENT_SIZE);
+ event.type = XENKBD_TYPE_MOTION;
+ event.motion.rel_x = rel_x;
+ event.motion.rel_y = rel_y;
+#if __XEN_LATEST_INTERFACE_VERSION__ >= 0x00030207
+ event.motion.rel_z = rel_z;
+#endif
+
+ return xenfb_kbd_event(xenfb, &event);
+}
+
+/* Send an absolute mouse movement event */
+static int xenfb_send_position(struct XenInput *xenfb,
+ int abs_x, int abs_y, int z)
+{
+ union xenkbd_in_event event;
+
+ memset(&event, 0, XENKBD_IN_EVENT_SIZE);
+ event.type = XENKBD_TYPE_POS;
+ event.pos.abs_x = abs_x;
+ event.pos.abs_y = abs_y;
+#if __XEN_LATEST_INTERFACE_VERSION__ == 0x00030207
+ event.pos.abs_z = z;
+#endif
+#if __XEN_LATEST_INTERFACE_VERSION__ >= 0x00030208
+ event.pos.rel_z = z;
+#endif
+
+ return xenfb_kbd_event(xenfb, &event);
+}
+
+/*
+ * Send a key event from the client to the guest OS
+ * QEMU gives us a raw scancode from an AT / PS/2 style keyboard.
+ * We have to turn this into a Linux Input layer keycode.
+ *
+ * Extra complexity from the fact that with extended scancodes
+ * (like those produced by arrow keys) this method gets called
+ * twice, but we only want to send a single event. So we have to
+ * track the ''0xe0'' scancode state & collapse the
extended keys
+ * as needed.
+ *
+ * Wish we could just send scancodes straight to the guest which
+ * already has code for dealing with this...
+ */
+static void xenfb_key_event(void *opaque, int scancode)
+{
+ struct XenInput *xenfb = opaque;
+ int down = 1;
+
+ if (scancode == 0xe0) {
+ xenfb->extended = 1;
+ return;
+ } else if (scancode & 0x80) {
+ scancode &= 0x7f;
+ down = 0;
+ }
+ if (xenfb->extended) {
+ scancode |= 0x80;
+ xenfb->extended = 0;
+ }
+ xenfb_send_key(xenfb, down, scancode2linux[scancode]);
+}
+
+/*
+ * Send a mouse event from the client to the guest OS
+ *
+ * The QEMU mouse can be in either relative, or absolute mode.
+ * Movement is sent separately from button state, which has to
+ * be encoded as virtual key events. We also don''t actually get
+ * given any button up/down events, so have to track changes in
+ * the button state.
+ */
+static void xenfb_mouse_event(void *opaque,
+ int dx, int dy, int dz, int button_state)
+{
+ struct XenInput *xenfb = opaque;
+ int i;
+
+ if (xenfb->abs_pointer_wanted)
+ xenfb_send_position(xenfb,
+ dx * (xenfb->c.ds->width - 1) / 0x7fff,
+ dy * (xenfb->c.ds->height - 1) / 0x7fff,
+ dz);
+ else
+ xenfb_send_motion(xenfb, dx, dy, dz);
+
+ for (i = 0 ; i < 8 ; i++) {
+ int lastDown = xenfb->button_state & (1 << i);
+ int down = button_state & (1 << i);
+ if (down == lastDown)
+ continue;
+
+ if (xenfb_send_key(xenfb, down, BTN_LEFT+i) < 0)
+ return;
+ }
+ xenfb->button_state = button_state;
+}
+
+static int input_init(struct XenDevice *xendev)
+{
+ struct XenInput *in = container_of(xendev, struct XenInput, c.xendev);
+
+ if (!in->c.ds) {
+ /* xen_set_display() below will set that and trigger us again */
+ xen_be_printf(xendev, 1, "ds not set (yet)\n");
+ return -1;
+ }
+
+ xenstore_write_be_int(xendev, "feature-abs-pointer", 1);
+ return 0;
+}
+
+static int input_connect(struct XenDevice *xendev)
+{
+ struct XenInput *in = container_of(xendev, struct XenInput, c.xendev);
+ int rc;
+
+ if (-1 == xenstore_read_fe_int(xendev, "request-abs-pointer",
&in->abs_pointer_wanted))
+ in->abs_pointer_wanted = 0;
+
+ rc = common_bind(&in->c);
+ if (0 != rc)
+ return rc;
+
+ qemu_add_kbd_event_handler(xenfb_key_event, in);
+ in->qmouse = qemu_add_mouse_event_handler(xenfb_mouse_event, in,
+ in->abs_pointer_wanted,
+ "Xen PVFB Mouse");
+ return 0;
+}
+
+static void input_disconnect(struct XenDevice *xendev)
+{
+ struct XenInput *in = container_of(xendev, struct XenInput, c.xendev);
+
+ if (in->qmouse) {
+ qemu_remove_mouse_event_handler(in->qmouse);
+ in->qmouse = NULL;
+ }
+ qemu_add_kbd_event_handler(NULL, NULL);
+ common_unbind(&in->c);
+}
+
+static void input_event(struct XenDevice *xendev)
+{
+ struct XenInput *xenfb = container_of(xendev, struct XenInput, c.xendev);
+ struct xenkbd_page *page = xenfb->c.page;
+
+ /* We don''t understand any keyboard events, so just ignore them.
*/
+ if (page->out_prod == page->out_cons)
+ return;
+ page->out_cons = page->out_prod;
+ xen_be_send_notify(&xenfb->c.xendev);
+}
+
+/* -------------------------------------------------------------------- */
+
+static void xenfb_copy_mfns(int mode, int count, unsigned long *dst, void *src)
+{
+ uint32_t *src32 = src;
+ uint64_t *src64 = src;
+ int i;
+
+ for (i = 0; i < count; i++)
+ dst[i] = (mode == 32) ? src32[i] : src64[i];
+}
+
+static int xenfb_map_fb(struct XenFB *xenfb)
+{
+ struct xenfb_page *page = xenfb->c.page;
+ char *protocol = xenfb->c.xendev.protocol;
+ int n_fbdirs;
+ unsigned long *pgmfns = NULL;
+ unsigned long *fbmfns = NULL;
+ void *map, *pd;
+ int mode, ret = -1;
+
+ /* default to native */
+ pd = page->pd;
+ mode = sizeof(unsigned long) * 8;
+
+ if (!protocol) {
+ /*
+ * Undefined protocol, some guesswork needed.
+ *
+ * Old frontends which don''t set the protocol use
+ * one page directory only, thus pd[1] must be zero.
+ * pd[1] of the 32bit struct layout and the lower
+ * 32 bits of pd[0] of the 64bit struct layout have
+ * the same location, so we can check that ...
+ */
+ uint32_t *ptr32 = NULL;
+ uint32_t *ptr64 = NULL;
+#if defined(__i386__)
+ ptr32 = (void*)page->pd;
+ ptr64 = ((void*)page->pd) + 4;
+#elif defined(__x86_64__)
+ ptr32 = ((void*)page->pd) - 4;
+ ptr64 = (void*)page->pd;
+#endif
+ if (ptr32) {
+ if (0 == ptr32[1]) {
+ mode = 32;
+ pd = ptr32;
+ } else {
+ mode = 64;
+ pd = ptr64;
+ }
+ }
+#if defined(__x86_64__)
+ } else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_X86_32)) {
+ /* 64bit dom0, 32bit domU */
+ mode = 32;
+ pd = ((void*)page->pd) - 4;
+#elif defined(__i386__)
+ } else if (0 == strcmp(protocol, XEN_IO_PROTO_ABI_X86_64)) {
+ /* 32bit dom0, 64bit domU */
+ mode = 64;
+ pd = ((void*)page->pd) + 4;
+#endif
+ }
+
+ if (xenfb->pixels) {
+ munmap(xenfb->pixels, xenfb->fbpages * XC_PAGE_SIZE);
+ xenfb->pixels = NULL;
+ }
+
+ xenfb->fbpages = (xenfb->fb_len + (XC_PAGE_SIZE - 1)) / XC_PAGE_SIZE;
+ n_fbdirs = xenfb->fbpages * mode / 8;
+ n_fbdirs = (n_fbdirs + (XC_PAGE_SIZE - 1)) / XC_PAGE_SIZE;
+
+ pgmfns = qemu_mallocz(sizeof(unsigned long) * n_fbdirs);
+ fbmfns = qemu_mallocz(sizeof(unsigned long) * xenfb->fbpages);
+ if (!pgmfns || !fbmfns)
+ goto out;
+
+ xenfb_copy_mfns(mode, n_fbdirs, pgmfns, pd);
+ map = xc_map_foreign_pages(xen_xc, xenfb->c.xendev.dom,
+ PROT_READ, pgmfns, n_fbdirs);
+ if (map == NULL)
+ goto out;
+ xenfb_copy_mfns(mode, xenfb->fbpages, fbmfns, map);
+ munmap(map, n_fbdirs * XC_PAGE_SIZE);
+
+ xenfb->pixels = xc_map_foreign_pages(xen_xc, xenfb->c.xendev.dom,
+ PROT_READ | PROT_WRITE, fbmfns, xenfb->fbpages);
+ if (xenfb->pixels == NULL)
+ goto out;
+
+ ret = 0; /* all is fine */
+
+out:
+ qemu_free(pgmfns);
+ qemu_free(fbmfns);
+ return ret;
+}
+
+static int xenfb_configure_fb(struct XenFB *xenfb, size_t fb_len_lim,
+ int width, int height, int depth,
+ size_t fb_len, int offset, int row_stride)
+{
+ size_t mfn_sz = sizeof(*((struct xenfb_page *)0)->pd);
+ size_t pd_len = sizeof(((struct xenfb_page *)0)->pd) / mfn_sz;
+ size_t fb_pages = pd_len * XC_PAGE_SIZE / mfn_sz;
+ size_t fb_len_max = fb_pages * XC_PAGE_SIZE;
+ int max_width, max_height;
+
+ if (fb_len_lim > fb_len_max) {
+ xen_be_printf(&xenfb->c.xendev, 0, "fb size limit %zu exceeds %zu,
corrected\n",
+ fb_len_lim, fb_len_max);
+ fb_len_lim = fb_len_max;
+ }
+ if (fb_len_lim && fb_len > fb_len_lim) {
+ xen_be_printf(&xenfb->c.xendev, 0, "frontend fb size %zu limited
to %zu\n",
+ fb_len, fb_len_lim);
+ fb_len = fb_len_lim;
+ }
+ if (depth != 8 && depth != 16 && depth != 24 &&
depth != 32) {
+ xen_be_printf(&xenfb->c.xendev, 0, "can''t handle frontend
fb depth %d\n",
+ depth);
+ return -1;
+ }
+ if (row_stride <= 0 || row_stride > fb_len) {
+ xen_be_printf(&xenfb->c.xendev, 0, "invalid frontend stride
%d\n", row_stride);
+ return -1;
+ }
+ max_width = row_stride / (depth / 8);
+ if (width < 0 || width > max_width) {
+ xen_be_printf(&xenfb->c.xendev, 0, "invalid frontend width %d
limited to %d\n",
+ width, max_width);
+ width = max_width;
+ }
+ if (offset < 0 || offset >= fb_len) {
+ xen_be_printf(&xenfb->c.xendev, 0, "invalid frontend offset %d
(max %zu)\n",
+ offset, fb_len - 1);
+ return -1;
+ }
+ max_height = (fb_len - offset) / row_stride;
+ if (height < 0 || height > max_height) {
+ xen_be_printf(&xenfb->c.xendev, 0, "invalid frontend height %d
limited to %d\n",
+ height, max_height);
+ height = max_height;
+ }
+ xenfb->fb_len = fb_len;
+ xenfb->row_stride = row_stride;
+ xenfb->depth = depth;
+ xenfb->width = width;
+ xenfb->height = height;
+ xenfb->offset = offset;
+ xen_be_printf(&xenfb->c.xendev, 1, "framebuffer %dx%dx%d offset
%d stride %d\n",
+ width, height, depth, offset, row_stride);
+ return 0;
+}
+
+/* A convenient function for munging pixels between different depths */
+#define BLT(SRC_T,DST_T,RSB,GSB,BSB,RDB,GDB,BDB) \
+ for (line = y ; line < (y+h) ; line++) { \
+ SRC_T *src = (SRC_T *)(xenfb->pixels \
+ + xenfb->offset \
+ + (line * xenfb->row_stride) \
+ + (x * xenfb->depth / 8)); \
+ DST_T *dst = (DST_T *)(xenfb->c.ds->data \
+ + (line * xenfb->c.ds->linesize) \
+ + (x * xenfb->c.ds->depth / 8)); \
+ int col; \
+ const int RSS = 32 - (RSB + GSB + BSB); \
+ const int GSS = 32 - (GSB + BSB); \
+ const int BSS = 32 - (BSB); \
+ const uint32_t RSM = (~0U) << (32 - RSB); \
+ const uint32_t GSM = (~0U) << (32 - GSB); \
+ const uint32_t BSM = (~0U) << (32 - BSB); \
+ const int RDS = 32 - (RDB + GDB + BDB); \
+ const int GDS = 32 - (GDB + BDB); \
+ const int BDS = 32 - (BDB); \
+ const uint32_t RDM = (~0U) << (32 - RDB); \
+ const uint32_t GDM = (~0U) << (32 - GDB); \
+ const uint32_t BDM = (~0U) << (32 - BDB); \
+ for (col = x ; col < (x+w) ; col++) { \
+ uint32_t spix = *src; \
+ *dst = (((spix << RSS) & RSM & RDM) >> RDS) | \
+ (((spix << GSS) & GSM & GDM) >> GDS) | \
+ (((spix << BSS) & BSM & BDM) >> BDS); \
+ src = (SRC_T *) ((unsigned long) src + xenfb->depth / 8); \
+ dst = (DST_T *) ((unsigned long) dst + xenfb->c.ds->depth / 8); \
+ } \
+ }
+
+
+/* This copies data from the guest framebuffer region, into QEMU''s
copy
+ * NB. QEMU''s copy is stored in the pixel format of a) the local X
+ * server (SDL case) or b) the current VNC client pixel format.
+ * When shifting between colour depths we preserve the MSB.
+ */
+static void xenfb_guest_copy(struct XenFB *xenfb, int x, int y, int w, int h)
+{
+ int line;
+
+ /*
+ * TODO: xen''s qemu-dm seems to have some patches to
+ * make the qemu display code avoid unneeded
+ * work.
+ * - Port them over.
+ * - Put ds->shared_buf back into use then.
+ */
+ if (1 /* !xenfb->c.ds->shared_buf */) {
+ if (xenfb->depth == xenfb->c.ds->depth) { /* Perfect match can use
fast path */
+ for (line = y ; line < (y+h) ; line++) {
+ memcpy(xenfb->c.ds->data + (line * xenfb->c.ds->linesize) + (x *
xenfb->c.ds->depth / 8),
+ xenfb->pixels + xenfb->offset + (line * xenfb->row_stride) +
(x * xenfb->depth / 8),
+ w * xenfb->depth / 8);
+ }
+ } else { /* Mismatch requires slow pixel munging */
+ /* 8 bit == r:3 g:3 b:2 */
+ /* 16 bit == r:5 g:6 b:5 */
+ /* 24 bit == r:8 g:8 b:8 */
+ /* 32 bit == r:8 g:8 b:8 (padding:8) */
+ if (xenfb->depth == 8) {
+ if (xenfb->c.ds->depth == 16) {
+ BLT(uint8_t, uint16_t, 3, 3, 2, 5, 6, 5);
+ } else if (xenfb->c.ds->depth == 32) {
+ BLT(uint8_t, uint32_t, 3, 3, 2, 8, 8, 8);
+ }
+ } else if (xenfb->depth == 16) {
+ if (xenfb->c.ds->depth == 8) {
+ BLT(uint16_t, uint8_t, 5, 6, 5, 3, 3, 2);
+ } else if (xenfb->c.ds->depth == 32) {
+ BLT(uint16_t, uint32_t, 5, 6, 5, 8, 8, 8);
+ }
+ } else if (xenfb->depth == 24 || xenfb->depth == 32) {
+ if (xenfb->c.ds->depth == 8) {
+ BLT(uint32_t, uint8_t, 8, 8, 8, 3, 3, 2);
+ } else if (xenfb->c.ds->depth == 16) {
+ BLT(uint32_t, uint16_t, 8, 8, 8, 5, 6, 5);
+ } else if (xenfb->c.ds->depth == 32) {
+ BLT(uint32_t, uint32_t, 8, 8, 8, 8, 8, 8);
+ }
+ }
+ }
+ }
+ dpy_update(xenfb->c.ds, x, y, w, h);
+}
+
+#ifdef XENFB_TYPE_REFRESH_PERIOD
+static int xenfb_queue_full(struct XenFB *xenfb)
+{
+ struct xenfb_page *page = xenfb->c.page;
+ uint32_t cons, prod;
+
+ if (!page)
+ return 1;
+
+ prod = page->in_prod;
+ cons = page->in_cons;
+ return prod - cons == XENFB_IN_RING_LEN;
+}
+
+static void xenfb_send_event(struct XenFB *xenfb, union xenfb_in_event *event)
+{
+ uint32_t prod;
+ struct xenfb_page *page = xenfb->c.page;
+
+ prod = page->in_prod;
+ /* caller ensures !xenfb_queue_full() */
+ xen_mb(); /* ensure ring space available */
+ XENFB_IN_RING_REF(page, prod) = *event;
+ xen_wmb(); /* ensure ring contents visible */
+ page->in_prod = prod + 1;
+
+ xen_be_send_notify(&xenfb->c.xendev);
+}
+
+static void xenfb_send_refresh_period(struct XenFB *xenfb, int period)
+{
+ union xenfb_in_event event;
+
+ memset(&event, 0, sizeof(event));
+ event.type = XENFB_TYPE_REFRESH_PERIOD;
+ event.refresh_period.period = period;
+ xenfb_send_event(xenfb, &event);
+}
+#endif
+
+/*
+ * Periodic update of display.
+ * Also transmit the refresh interval to the frontend.
+ *
+ * Never ever do any qemu display operations
+ * (resize, screen update) outside this function.
+ * Our screen might be inactive. When asked for
+ * an update we know it is active.
+ */
+static void xenfb_update(void *opaque)
+{
+ struct XenFB *xenfb = opaque;
+ int i;
+
+ if (xenfb->feature_update) {
+#ifdef XENFB_TYPE_REFRESH_PERIOD
+ int period;
+
+ if (xenfb_queue_full(xenfb))
+ return;
+
+ if (xenfb->c.ds->idle)
+ period = XENFB_NO_REFRESH;
+ else {
+ period = xenfb->c.ds->gui_timer_interval;
+ if (!period)
+ period = GUI_REFRESH_INTERVAL;
+ }
+
+ if (xenfb->refresh_period != period) {
+ xenfb_send_refresh_period(xenfb, period);
+ xenfb->refresh_period = period;
+ }
+#else
+ ; /* nothing */
+#endif
+ } else {
+ /* we don''t get update notifications, thus use the
+ * sledge hammer approach ... */
+ xenfb->up_fullscreen = 1;
+ }
+
+ /* resize if needed */
+ if (xenfb->width != xenfb->c.ds->width || xenfb->height !=
xenfb->c.ds->height) {
+ xen_be_printf(&xenfb->c.xendev, 1, "update: resizing:
%dx%d\n",
+ xenfb->width, xenfb->height);
+ dpy_resize(xenfb->c.ds, xenfb->width, xenfb->height);
+ xenfb->up_fullscreen = 1;
+ }
+
+ /* run queued updates */
+ if (xenfb->up_fullscreen) {
+ xen_be_printf(&xenfb->c.xendev, 3, "update: fullscreen\n");
+ xenfb_guest_copy(xenfb, 0, 0, xenfb->width, xenfb->height);
+ } else if (xenfb->up_count) {
+ xen_be_printf(&xenfb->c.xendev, 3, "update: %d rects\n",
xenfb->up_count);
+ for (i = 0; i < xenfb->up_count; i++)
+ xenfb_guest_copy(xenfb,
+ xenfb->up_rects[i].x,
+ xenfb->up_rects[i].y,
+ xenfb->up_rects[i].w,
+ xenfb->up_rects[i].h);
+ } else {
+ xen_be_printf(&xenfb->c.xendev, 3, "update: nothing\n");
+ }
+ xenfb->up_count = 0;
+ xenfb->up_fullscreen = 0;
+}
+
+/* QEMU display state changed, so refresh the framebuffer copy */
+static void xenfb_invalidate(void *opaque)
+{
+ struct XenFB *xenfb = opaque;
+ xenfb->up_fullscreen = 1;
+}
+
+static void xenfb_handle_events(struct XenFB *xenfb)
+{
+ uint32_t prod, cons;
+ struct xenfb_page *page = xenfb->c.page;
+
+ prod = page->out_prod;
+ if (prod == page->out_cons)
+ return;
+ xen_rmb(); /* ensure we see ring contents up to prod */
+ for (cons = page->out_cons; cons != prod; cons++) {
+ union xenfb_out_event *event = &XENFB_OUT_RING_REF(page, cons);
+ int x, y, w, h;
+
+ switch (event->type) {
+ case XENFB_TYPE_UPDATE:
+ if (xenfb->up_count == UP_QUEUE)
+ xenfb->up_fullscreen = 1;
+ if (xenfb->up_fullscreen)
+ break;
+ x = MAX(event->update.x, 0);
+ y = MAX(event->update.y, 0);
+ w = MIN(event->update.width, xenfb->width - x);
+ h = MIN(event->update.height, xenfb->height - y);
+ if (w < 0 || h < 0) {
+ fprintf(stderr, "xen be: %s: bogus update ignored\n",
+ xenfb->c.xendev.name);
+ break;
+ }
+ if (x != event->update.x || y != event->update.y
+ || w != event->update.width
+ || h != event->update.height) {
+ fprintf(stderr, "xen be: %s: bogus update clipped\n",
+ xenfb->c.xendev.name);
+ }
+ if (w == xenfb->width && h > xenfb->height / 2) {
+ /* scroll detector: updated more than 50% of the lines,
+ * don''t bother keeping track of the rectangles then */
+ xenfb->up_fullscreen = 1;
+ } else {
+ xenfb->up_rects[xenfb->up_count].x = x;
+ xenfb->up_rects[xenfb->up_count].y = y;
+ xenfb->up_rects[xenfb->up_count].w = w;
+ xenfb->up_rects[xenfb->up_count].h = h;
+ xenfb->up_count++;
+ }
+ break;
+#ifdef XENFB_TYPE_RESIZE
+ case XENFB_TYPE_RESIZE:
+ if (xenfb_configure_fb(xenfb, xenfb->fb_len,
+ event->resize.width,
+ event->resize.height,
+ event->resize.depth,
+ xenfb->fb_len,
+ event->resize.offset,
+ event->resize.stride) < 0)
+ break;
+ xenfb_invalidate(xenfb);
+ break;
+#endif
+ }
+ }
+ xen_mb(); /* ensure we''re done with ring contents */
+ page->out_cons = cons;
+}
+
+static int fb_init(struct XenDevice *xendev)
+{
+ struct XenFB *fb = container_of(xendev, struct XenFB, c.xendev);
+
+ if (!fb->c.ds) {
+ /* xen_set_display() below will set that and trigger us again */
+ xen_be_printf(xendev, 1, "ds not set (yet)\n");
+ return -1;
+ }
+
+ fb->refresh_period = -1;
+
+#ifdef XENFB_TYPE_RESIZE
+ xenstore_write_be_int(xendev, "feature-resize", 1);
+#endif
+ return 0;
+}
+
+static int fb_connect(struct XenDevice *xendev)
+{
+ struct XenFB *fb = container_of(xendev, struct XenFB, c.xendev);
+ struct xenfb_page *fb_page;
+ int videoram;
+ int rc;
+
+ if (-1 == xenstore_read_fe_int(xendev, "videoram",
&videoram))
+ videoram = 0;
+
+ rc = common_bind(&fb->c);
+ if (0 != rc)
+ return rc;
+
+ fb_page = fb->c.page;
+ rc = xenfb_configure_fb(fb, videoram * 1024 * 1024U,
+ fb_page->width, fb_page->height, fb_page->depth,
+ fb_page->mem_length, 0, fb_page->line_length);
+ if (0 != rc)
+ return rc;
+
+ rc = xenfb_map_fb(fb);
+ if (0 != rc)
+ return rc;
+
+ if (!fb->have_console) {
+ graphic_console_init(fb->c.ds,
+ xenfb_update,
+ xenfb_invalidate,
+ NULL,
+ NULL,
+ fb);
+ fb->have_console = 1;
+ }
+
+ if (-1 == xenstore_read_fe_int(xendev, "feature-update",
&fb->feature_update))
+ fb->feature_update = 0;
+ if (fb->feature_update)
+ xenstore_write_be_int(xendev, "request-update", 1);
+
+ xen_be_printf(xendev, 1, "feature-update=%d, videoram=%d\n",
+ fb->feature_update, videoram);
+
+ return 0;
+}
+
+static void fb_disconnect(struct XenDevice *xendev)
+{
+ struct XenFB *fb = container_of(xendev, struct XenFB, c.xendev);
+
+ /*
+ * FIXME: qemu can''t un-init gfx display (yet?).
+ * Replacing the framebuffer with anonymous shared memory
+ * instead. This releases the guest pages and keeps qemu happy.
+ */
+ fb->pixels = mmap(fb->pixels, fb->fbpages * XC_PAGE_SIZE,
+ PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANON,
+ -1, 0);
+ common_unbind(&fb->c);
+ fb->feature_update = 0;
+ fb->bug_trigger = 0;
+}
+
+static void fb_frontend_changed(struct XenDevice *xendev, const char *node)
+{
+ struct XenFB *fb = container_of(xendev, struct XenFB, c.xendev);
+
+ /*
+ * Set state to Connected *again* once the frontend switched
+ * to connected. We must trigger the watch a second time to
+ * workaround a frontend bug.
+ */
+ if (0 == fb->bug_trigger && 0 == strcmp(node, "state")
&&
+ xendev->fe_state == XenbusStateConnected &&
+ xendev->be_state == XenbusStateConnected) {
+ xen_be_printf(xendev, 2, "re-trigger connected (frontend
bug)\n");
+ xen_be_set_state(xendev, XenbusStateConnected);
+ fb->bug_trigger = 1; /* only once */
+ }
+}
+
+static void fb_event(struct XenDevice *xendev)
+{
+ struct XenFB *xenfb = container_of(xendev, struct XenFB, c.xendev);
+
+ xenfb_handle_events(xenfb);
+ xen_be_send_notify(&xenfb->c.xendev);
+}
+
+/* -------------------------------------------------------------------- */
+
+struct XenDevOps xen_kbdmouse_ops = {
+ .size = sizeof(struct XenInput),
+ .init = input_init,
+ .connect = input_connect,
+ .disconnect = input_disconnect,
+ .event = input_event,
+};
+
+struct XenDevOps xen_framebuffer_ops = {
+ .size = sizeof(struct XenFB),
+ .init = fb_init,
+ .connect = fb_connect,
+ .disconnect = fb_disconnect,
+ .event = fb_event,
+ .frontend_changed = fb_frontend_changed,
+};
+
+static void xen_set_display_type(int domid, const char *type, DisplayState *ds)
+{
+ struct XenDevice *xendev;
+ struct common *c;
+
+ xendev = xen_be_find_xendev(type, domid, 0);
+ if (!xendev)
+ return;
+ c = container_of(xendev, struct common, xendev);
+ c->ds = ds;
+ xen_be_printf(xendev, 1, "ds is %p\n", ds);
+ /* retry ->init() */
+ xen_be_check_state(xendev);
+}
+
+void xen_set_display(int domid, DisplayState *ds)
+{
+ xen_set_display_type(domid, "vkbd", ds);
+ xen_set_display_type(domid, "vfb", ds);
+}
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 12:23 UTC
[Xen-devel] [PATCH 5/7] xen: add block device backend driver.
This patch adds a block device backend driver to qemu. It is a pure
userspace implemention using the gntdev interface. It uses "qdisk" as
backend name in xenstore so it doesn''t interfere with the other
existing
backends (blkback aka "vbd" and tapdisk aka "tap").
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 2 +-
hw/xen_backend.h | 2 +
hw/xen_blkif.h | 103 ++++++++
hw/xen_disk.c | 694 +++++++++++++++++++++++++++++++++++++++++++++++++++
hw/xen_machine_pv.c | 1 +
sysemu.h | 2 +-
vl.c | 4 +
7 files changed, 806 insertions(+), 2 deletions(-)
create mode 100644 hw/xen_blkif.h
create mode 100644 hw/xen_disk.c
diff --git a/Makefile.target b/Makefile.target
index 7606a20..22bdd69 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -631,7 +631,7 @@ endif
# xen backend driver support
XEN_OBJS := xen_machine_pv.o xen_backend.o
-XEN_OBJS += xen_console.o xenfb.o
+XEN_OBJS += xen_console.o xenfb.o xen_disk.o
ifeq ($(CONFIG_XEN), yes)
OBJS += $(XEN_OBJS)
LIBS += $(XEN_LIBS)
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
index ceeb3aa..4fe9ad0 100644
--- a/hw/xen_backend.h
+++ b/hw/xen_backend.h
@@ -2,6 +2,7 @@
#define QEMU_HW_XEN_BACKEND_H 1
#include "xen_common.h"
+#include "sysemu.h"
/* ------------------------------------------------------------- */
@@ -87,6 +88,7 @@ void xen_be_printf(struct XenDevice *xendev, int msg_level,
const char *fmt, ...
extern struct XenDevOps xen_console_ops; /* xen_console.c */
extern struct XenDevOps xen_kbdmouse_ops; /* xen_framebuffer.c */
extern struct XenDevOps xen_framebuffer_ops; /* xen_framebuffer.c */
+extern struct XenDevOps xen_blkdev_ops; /* xen_disk.c */
void xen_set_display(int domid, DisplayState *ds);
diff --git a/hw/xen_blkif.h b/hw/xen_blkif.h
new file mode 100644
index 0000000..254a5fd
--- /dev/null
+++ b/hw/xen_blkif.h
@@ -0,0 +1,103 @@
+#ifndef __XEN_BLKIF_H__
+#define __XEN_BLKIF_H__
+
+#include <xen/io/ring.h>
+#include <xen/io/blkif.h>
+#include <xen/io/protocols.h>
+
+/* Not a real protocol. Used to generate ring structs which contain
+ * the elements common to all protocols only. This way we get a
+ * compiler-checkable way to use common struct elements, so we can
+ * avoid using switch(protocol) in a number of places. */
+struct blkif_common_request {
+ char dummy;
+};
+struct blkif_common_response {
+ char dummy;
+};
+
+/* i386 protocol version */
+#pragma pack(push, 4)
+struct blkif_x86_32_request {
+ uint8_t operation; /* BLKIF_OP_??? */
+ uint8_t nr_segments; /* number of segments */
+ blkif_vdev_t handle; /* only for read/write requests */
+ uint64_t id; /* private guest value, echoed in resp */
+ blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
+ struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+};
+struct blkif_x86_32_response {
+ uint64_t id; /* copied from request */
+ uint8_t operation; /* copied from request */
+ int16_t status; /* BLKIF_RSP_??? */
+};
+typedef struct blkif_x86_32_request blkif_x86_32_request_t;
+typedef struct blkif_x86_32_response blkif_x86_32_response_t;
+#pragma pack(pop)
+
+/* x86_64 protocol version */
+struct blkif_x86_64_request {
+ uint8_t operation; /* BLKIF_OP_??? */
+ uint8_t nr_segments; /* number of segments */
+ blkif_vdev_t handle; /* only for read/write requests */
+ uint64_t __attribute__((__aligned__(8))) id;
+ blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
+ struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+};
+struct blkif_x86_64_response {
+ uint64_t __attribute__((__aligned__(8))) id;
+ uint8_t operation; /* copied from request */
+ int16_t status; /* BLKIF_RSP_??? */
+};
+typedef struct blkif_x86_64_request blkif_x86_64_request_t;
+typedef struct blkif_x86_64_response blkif_x86_64_response_t;
+
+DEFINE_RING_TYPES(blkif_common, struct blkif_common_request, struct
blkif_common_response);
+DEFINE_RING_TYPES(blkif_x86_32, struct blkif_x86_32_request, struct
blkif_x86_32_response);
+DEFINE_RING_TYPES(blkif_x86_64, struct blkif_x86_64_request, struct
blkif_x86_64_response);
+
+union blkif_back_rings {
+ blkif_back_ring_t native;
+ blkif_common_back_ring_t common;
+ blkif_x86_32_back_ring_t x86_32;
+ blkif_x86_64_back_ring_t x86_64;
+};
+typedef union blkif_back_rings blkif_back_rings_t;
+
+enum blkif_protocol {
+ BLKIF_PROTOCOL_NATIVE = 1,
+ BLKIF_PROTOCOL_X86_32 = 2,
+ BLKIF_PROTOCOL_X86_64 = 3,
+};
+
+static void inline blkif_get_x86_32_req(blkif_request_t *dst,
blkif_x86_32_request_t *src)
+{
+ int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
+
+ dst->operation = src->operation;
+ dst->nr_segments = src->nr_segments;
+ dst->handle = src->handle;
+ dst->id = src->id;
+ dst->sector_number = src->sector_number;
+ if (n > src->nr_segments)
+ n = src->nr_segments;
+ for (i = 0; i < n; i++)
+ dst->seg[i] = src->seg[i];
+}
+
+static void inline blkif_get_x86_64_req(blkif_request_t *dst,
blkif_x86_64_request_t *src)
+{
+ int i, n = BLKIF_MAX_SEGMENTS_PER_REQUEST;
+
+ dst->operation = src->operation;
+ dst->nr_segments = src->nr_segments;
+ dst->handle = src->handle;
+ dst->id = src->id;
+ dst->sector_number = src->sector_number;
+ if (n > src->nr_segments)
+ n = src->nr_segments;
+ for (i = 0; i < n; i++)
+ dst->seg[i] = src->seg[i];
+}
+
+#endif /* __XEN_BLKIF_H__ */
diff --git a/hw/xen_disk.c b/hw/xen_disk.c
new file mode 100644
index 0000000..1141dc1
--- /dev/null
+++ b/hw/xen_disk.c
@@ -0,0 +1,694 @@
+/*
+ * xen paravirt block device backend
+ *
+ * (c) Gerd Hoffmann <kraxel@redhat.com>
+ *
+ * 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; under version 2 of the License.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+/*
+ * FIXME: the code is designed to handle multiple outstanding
+ * requests, which isn''t used right now. Plan is to
+ * switch over to the aio block functions once they got
+ * vector support.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <unistd.h>
+#include <signal.h>
+#include <inttypes.h>
+#include <time.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <sys/uio.h>
+
+#include <xs.h>
+#include <xenctrl.h>
+#include <xen/io/xenbus.h>
+
+#include "hw.h"
+#include "block_int.h"
+#include "qemu-char.h"
+#include "xen_blkif.h"
+#include "xen_backend.h"
+
+/* ------------------------------------------------------------- */
+
+#define BLOCK_SIZE 512
+#define IOCB_COUNT (BLKIF_MAX_SEGMENTS_PER_REQUEST + 2)
+
+struct ioreq {
+ blkif_request_t req;
+ int16_t status;
+
+ /* parsed request */
+ off_t start, end;
+ struct iovec vec[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+ int vecs;
+ int presync;
+ int postsync;
+
+ /* grant mapping */
+ uint32_t domids[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+ uint32_t refs[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+ int prot;
+ void *page[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+ void *pages;
+
+ struct XenBlkDev *blkdev;
+ LIST_ENTRY(ioreq) list;
+};
+
+struct XenBlkDev {
+ struct XenDevice xendev; /* must be first */
+ char *params;
+ char *mode;
+ char *type;
+ char *dev;
+ char *devtype;
+ const char *fileproto;
+ const char *filename;
+ int ring_ref;
+ void *sring;
+ int64_t file_blk;
+ int64_t file_size;
+ int protocol;
+ blkif_back_rings_t rings;
+ int more_work;
+ int cnt_map;
+
+ /* request lists */
+ LIST_HEAD(inflight_head, ioreq) inflight;
+ LIST_HEAD(finished_head, ioreq) finished;
+ LIST_HEAD(freelist_head, ioreq) freelist;
+ int requests;
+
+ /* qemu block driver */
+ int index;
+ BlockDriverState *bs;
+};
+
+static int syncwrite = 0;
+static int batch_maps = 0;
+static int max_requests = 32;
+
+/* ------------------------------------------------------------- */
+
+static struct ioreq *ioreq_start(struct XenBlkDev *blkdev)
+{
+ struct ioreq *ioreq = NULL;
+
+ if (LIST_EMPTY(&blkdev->freelist)) {
+ if (blkdev->requests >= max_requests)
+ goto out;
+ /* allocate new struct */
+ ioreq = qemu_mallocz(sizeof(*ioreq));
+ if (NULL == ioreq)
+ goto out;
+ ioreq->blkdev = blkdev;
+ blkdev->requests++;
+ } else {
+ /* get one from freelist */
+ ioreq = LIST_FIRST(&blkdev->freelist);
+ LIST_REMOVE(ioreq, list);
+ }
+ LIST_INSERT_HEAD(&blkdev->inflight, ioreq, list);
+
+out:
+ return ioreq;
+}
+
+static void ioreq_finish(struct ioreq *ioreq)
+{
+ struct XenBlkDev *blkdev = ioreq->blkdev;
+
+ LIST_REMOVE(ioreq, list);
+ LIST_INSERT_HEAD(&blkdev->finished, ioreq, list);
+}
+
+static void ioreq_release(struct ioreq *ioreq)
+{
+ struct XenBlkDev *blkdev = ioreq->blkdev;
+
+ LIST_REMOVE(ioreq, list);
+ memset(ioreq, 0, sizeof(*ioreq));
+ ioreq->blkdev = blkdev;
+ LIST_INSERT_HEAD(&blkdev->freelist, ioreq, list);
+}
+
+/*
+ * translate request into iovec + start offset + end offset
+ * do sanity checks along the way
+ */
+static int ioreq_parse(struct ioreq *ioreq)
+{
+ struct XenBlkDev *blkdev = ioreq->blkdev;
+ uintptr_t mem;
+ size_t len;
+ int i;
+
+ xen_be_printf(&blkdev->xendev, 3,
+ "op %d, nr %d, handle %d, id %" PRId64 ", sector %"
PRId64 "\n",
+ ioreq->req.operation, ioreq->req.nr_segments,
+ ioreq->req.handle, ioreq->req.id, ioreq->req.sector_number);
+ switch (ioreq->req.operation) {
+ case BLKIF_OP_READ:
+ ioreq->prot = PROT_WRITE; /* to memory */
+ if (BLKIF_OP_READ != ioreq->req.operation && blkdev->mode[0] !=
''w'') {
+ xen_be_printf(&blkdev->xendev, 0, "error: write req for ro
device\n");
+ goto err;
+ }
+ break;
+ case BLKIF_OP_WRITE_BARRIER:
+ if (!syncwrite)
+ ioreq->presync = ioreq->postsync = 1;
+ /* fall through */
+ case BLKIF_OP_WRITE:
+ ioreq->prot = PROT_READ; /* from memory */
+ if (syncwrite)
+ ioreq->postsync = 1;
+ break;
+ default:
+ xen_be_printf(&blkdev->xendev, 0, "error: unknown operation
(%d)\n",
+ ioreq->req.operation);
+ goto err;
+ };
+
+ ioreq->start = ioreq->end = ioreq->req.sector_number *
blkdev->file_blk;
+ for (i = 0; i < ioreq->req.nr_segments; i++) {
+ if (i == BLKIF_MAX_SEGMENTS_PER_REQUEST) {
+ xen_be_printf(&blkdev->xendev, 0, "error: nr_segments too
big\n");
+ goto err;
+ }
+ if (ioreq->req.seg[i].first_sect > ioreq->req.seg[i].last_sect) {
+ xen_be_printf(&blkdev->xendev, 0, "error: first > last
sector\n");
+ goto err;
+ }
+ if (ioreq->req.seg[i].last_sect * BLOCK_SIZE >= XC_PAGE_SIZE) {
+ xen_be_printf(&blkdev->xendev, 0, "error: page
crossing\n");
+ goto err;
+ }
+ len = (ioreq->req.seg[i].last_sect - ioreq->req.seg[i].first_sect + 1) *
blkdev->file_blk;
+
+ ioreq->domids[i] = blkdev->xendev.dom;
+ ioreq->refs[i] = ioreq->req.seg[i].gref;
+ mem = ioreq->req.seg[i].first_sect * blkdev->file_blk;
+
+ ioreq->vec[i].iov_base = (void*)mem;
+ ioreq->vec[i].iov_len = len;
+ ioreq->end += len;
+ }
+ if (ioreq->end > blkdev->file_size) {
+ xen_be_printf(&blkdev->xendev, 0, "error: access beyond end of
file\n");
+ goto err;
+ }
+ ioreq->vecs = i;
+ return 0;
+
+err:
+ ioreq->status = BLKIF_RSP_ERROR;
+ return -1;
+}
+
+static void ioreq_unmap(struct ioreq *ioreq)
+{
+ int gnt = ioreq->blkdev->xendev.gnttabdev;
+ int i;
+
+ if (batch_maps) {
+ if (!ioreq->pages)
+ return;
+ if (0 != xc_gnttab_munmap(gnt, ioreq->pages, ioreq->vecs))
+ xen_be_printf(&ioreq->blkdev->xendev, 0, "xc_gnttab_munmap
failed: %s\n",
+ strerror(errno));
+ ioreq->blkdev->cnt_map -= ioreq->vecs;
+ ioreq->pages = NULL;
+ } else {
+ for (i = 0; i < ioreq->vecs; i++) {
+ if (!ioreq->page[i])
+ continue;
+ if (0 != xc_gnttab_munmap(gnt, ioreq->page[i], 1))
+ xen_be_printf(&ioreq->blkdev->xendev, 0, "xc_gnttab_munmap
failed: %s\n",
+ strerror(errno));
+ ioreq->blkdev->cnt_map--;
+ ioreq->page[i] = NULL;
+ }
+ }
+}
+
+static int ioreq_map(struct ioreq *ioreq)
+{
+ int gnt = ioreq->blkdev->xendev.gnttabdev;
+ int i;
+
+ if (batch_maps) {
+ ioreq->pages = xc_gnttab_map_grant_refs
+ (gnt, ioreq->vecs, ioreq->domids, ioreq->refs, ioreq->prot);
+ if (NULL == ioreq->pages) {
+ xen_be_printf(&ioreq->blkdev->xendev, 0,
+ "can''t map %d grant refs (%s, %d maps)\n",
+ ioreq->vecs, strerror(errno), ioreq->blkdev->cnt_map);
+ return -1;
+ }
+ for (i = 0; i < ioreq->vecs; i++)
+ ioreq->vec[i].iov_base = ioreq->pages + i * XC_PAGE_SIZE +
+ (uintptr_t)ioreq->vec[i].iov_base;
+ ioreq->blkdev->cnt_map += ioreq->vecs;
+ } else {
+ for (i = 0; i < ioreq->vecs; i++) {
+ ioreq->page[i] = xc_gnttab_map_grant_ref
+ (gnt, ioreq->domids[i], ioreq->refs[i], ioreq->prot);
+ if (NULL == ioreq->page[i]) {
+ xen_be_printf(&ioreq->blkdev->xendev, 0,
+ "can''t map grant ref %d (%s, %d maps)\n",
+ ioreq->refs[i], strerror(errno), ioreq->blkdev->cnt_map);
+ ioreq_unmap(ioreq);
+ return -1;
+ }
+ ioreq->vec[i].iov_base = ioreq->page[i] +
(uintptr_t)ioreq->vec[i].iov_base;
+ ioreq->blkdev->cnt_map++;
+ }
+ }
+ return 0;
+}
+
+static int ioreq_runio_qemu(struct ioreq *ioreq)
+{
+ struct XenBlkDev *blkdev = ioreq->blkdev;
+ int i, rc, len = 0;
+ off_t pos;
+
+ if (-1 == ioreq_map(ioreq))
+ goto err;
+ if (ioreq->presync)
+ bdrv_flush(blkdev->bs);
+
+ switch (ioreq->req.operation) {
+ case BLKIF_OP_READ:
+ pos = ioreq->start;
+ for (i = 0; i < ioreq->vecs; i++) {
+ rc = bdrv_read(blkdev->bs, pos / BLOCK_SIZE,
+ ioreq->vec[i].iov_base,
+ ioreq->vec[i].iov_len / BLOCK_SIZE);
+ if (rc != 0) {
+ xen_be_printf(&blkdev->xendev, 0, "rd I/O error (%p, len
%zd)\n",
+ ioreq->vec[i].iov_base,
+ ioreq->vec[i].iov_len);
+ goto err;
+ }
+ len += ioreq->vec[i].iov_len;
+ pos += ioreq->vec[i].iov_len;
+ }
+ break;
+ case BLKIF_OP_WRITE:
+ case BLKIF_OP_WRITE_BARRIER:
+ pos = ioreq->start;
+ for (i = 0; i < ioreq->vecs; i++) {
+ rc = bdrv_write(blkdev->bs, pos / BLOCK_SIZE,
+ ioreq->vec[i].iov_base,
+ ioreq->vec[i].iov_len / BLOCK_SIZE);
+ if (rc != 0) {
+ xen_be_printf(&blkdev->xendev, 0, "wr I/O error (%p, len
%zd)\n",
+ ioreq->vec[i].iov_base,
+ ioreq->vec[i].iov_len);
+ goto err;
+ }
+ len += ioreq->vec[i].iov_len;
+ pos += ioreq->vec[i].iov_len;
+ }
+ break;
+ default:
+ /* unknown operation (shouldn''t happen -- parse catches this) */
+ goto err;
+ }
+
+ if (ioreq->postsync)
+ bdrv_flush(blkdev->bs);
+ ioreq->status = BLKIF_RSP_OKAY;
+
+ ioreq_unmap(ioreq);
+ return 0;
+
+err:
+ ioreq->status = BLKIF_RSP_ERROR;
+ return -1;
+}
+
+static int blk_send_response_one(struct ioreq *ioreq)
+{
+ struct XenBlkDev *blkdev = ioreq->blkdev;
+ int send_notify = 0;
+ int have_requests = 0;
+ blkif_response_t resp;
+ void *dst;
+
+ resp.id = ioreq->req.id;
+ resp.operation = ioreq->req.operation;
+ resp.status = ioreq->status;
+
+ /* Place on the response ring for the relevant domain. */
+ switch (blkdev->protocol) {
+ case BLKIF_PROTOCOL_NATIVE:
+ dst = RING_GET_RESPONSE(&blkdev->rings.native,
blkdev->rings.native.rsp_prod_pvt);
+ break;
+ case BLKIF_PROTOCOL_X86_32:
+ dst = RING_GET_RESPONSE(&blkdev->rings.x86_32,
blkdev->rings.x86_32.rsp_prod_pvt);
+ break;
+ case BLKIF_PROTOCOL_X86_64:
+ dst = RING_GET_RESPONSE(&blkdev->rings.x86_64,
blkdev->rings.x86_64.rsp_prod_pvt);
+ break;
+ default:
+ dst = NULL;
+ }
+ memcpy(dst, &resp, sizeof(resp));
+ blkdev->rings.common.rsp_prod_pvt++;
+
+ RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&blkdev->rings.common,
send_notify);
+ if (blkdev->rings.common.rsp_prod_pvt ==
blkdev->rings.common.req_cons) {
+ /*
+ * Tail check for pending requests. Allows frontend to avoid
+ * notifications if requests are already in flight (lower
+ * overheads and promotes batching).
+ */
+ RING_FINAL_CHECK_FOR_REQUESTS(&blkdev->rings.common, have_requests);
+ } else if (RING_HAS_UNCONSUMED_REQUESTS(&blkdev->rings.common)) {
+ have_requests = 1;
+ }
+
+ if (have_requests)
+ blkdev->more_work++;
+ return send_notify;
+}
+
+/* walk finished list, send outstanding responses, free requests */
+static void blk_send_response_all(struct XenBlkDev *blkdev)
+{
+ struct ioreq *ioreq;
+ int send_notify = 0;
+
+ while (!LIST_EMPTY(&blkdev->finished)) {
+ ioreq = LIST_FIRST(&blkdev->finished);
+ send_notify += blk_send_response_one(ioreq);
+ ioreq_release(ioreq);
+ }
+ if (send_notify)
+ xen_be_send_notify(&blkdev->xendev);
+}
+
+static int blk_get_request(struct XenBlkDev *blkdev, struct ioreq *ioreq,
RING_IDX rc)
+{
+ switch (blkdev->protocol) {
+ case BLKIF_PROTOCOL_NATIVE:
+ memcpy(&ioreq->req, RING_GET_REQUEST(&blkdev->rings.native, rc),
+ sizeof(ioreq->req));
+ break;
+ case BLKIF_PROTOCOL_X86_32:
+ blkif_get_x86_32_req(&ioreq->req,
RING_GET_REQUEST(&blkdev->rings.x86_32, rc));
+ break;
+ case BLKIF_PROTOCOL_X86_64:
+ blkif_get_x86_64_req(&ioreq->req,
RING_GET_REQUEST(&blkdev->rings.x86_64, rc));
+ break;
+ }
+ return 0;
+}
+
+static void blk_handle_requests(struct XenBlkDev *blkdev)
+{
+ RING_IDX rc, rp;
+ struct ioreq *ioreq;
+
+ do {
+ blkdev->more_work = 0;
+
+ rc = blkdev->rings.common.req_cons;
+ rp = blkdev->rings.common.sring->req_prod;
+ xen_rmb(); /* Ensure we see queued requests up to ''rp''. */
+
+ /* Limit #of requests we queue up for I/O so we ack requests
+ * faster if busy. Improves backend/frontend parallelism and
+ * reduces evchn signaling. */
+ if (rp > rc + (max_requests >> 2)) {
+ rp = rc + (max_requests >> 2);
+ blkdev->more_work++;
+ }
+
+ while ((rc != rp)) {
+ /* pull request from ring */
+ if (RING_REQUEST_CONS_OVERFLOW(&blkdev->rings.common, rc))
+ break;
+ ioreq = ioreq_start(blkdev);
+ if (NULL == ioreq) {
+ blkdev->more_work++;
+ break;
+ }
+ blk_get_request(blkdev, ioreq, rc);
+ blkdev->rings.common.req_cons = ++rc;
+
+ /* parse them */
+ if (0 != ioreq_parse(ioreq)) {
+ if (blk_send_response_one(ioreq))
+ xen_be_send_notify(&blkdev->xendev);
+ ioreq_release(ioreq);
+ continue;
+ }
+
+ /* run i/o in qemu mode */
+ ioreq_runio_qemu(ioreq);
+ ioreq_finish(ioreq);
+ }
+ blk_send_response_all(blkdev);
+
+ } while (blkdev->more_work);
+}
+
+/* ------------------------------------------------------------- */
+
+static void blk_alloc(struct XenDevice *xendev)
+{
+ struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+ LIST_INIT(&blkdev->inflight);
+ LIST_INIT(&blkdev->finished);
+ LIST_INIT(&blkdev->freelist);
+}
+
+static int blk_init(struct XenDevice *xendev)
+{
+ struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+ int mode, qflags, have_barriers, info = 0;
+ char *h;
+
+ /* read xenstore entries */
+ if (NULL == blkdev->params) {
+ blkdev->params = xenstore_read_be_str(&blkdev->xendev,
"params");
+ if (NULL != (h = strchr(blkdev->params, '':''))) {
+ blkdev->fileproto = blkdev->params;
+ blkdev->filename = h+1;
+ *h = 0;
+ } else {
+ blkdev->fileproto = "<unset>";
+ blkdev->filename = blkdev->params;
+ }
+ }
+ if (NULL == blkdev->mode)
+ blkdev->mode = xenstore_read_be_str(&blkdev->xendev,
"mode");
+ if (NULL == blkdev->type)
+ blkdev->type = xenstore_read_be_str(&blkdev->xendev,
"type");
+ if (NULL == blkdev->dev)
+ blkdev->dev = xenstore_read_be_str(&blkdev->xendev,
"dev");
+ if (NULL == blkdev->devtype)
+ blkdev->devtype = xenstore_read_be_str(&blkdev->xendev,
"device-type");
+
+ /* do we have all we need? */
+ if (NULL == blkdev->params ||
+ NULL == blkdev->mode ||
+ NULL == blkdev->type ||
+ NULL == blkdev->dev)
+ return -1;
+
+ /* read-only ? */
+ if (0 == strcmp(blkdev->mode, "w")) {
+ mode = O_RDWR;
+ qflags = BDRV_O_RDWR;
+ } else {
+ mode = O_RDONLY;
+ qflags = BDRV_O_RDONLY;
+ info |= VDISK_READONLY;
+ }
+
+ /* cdrom ? */
+ if (blkdev->devtype && !strcmp(blkdev->devtype,
"cdrom"))
+ info |= VDISK_CDROM;
+
+ /* init qemu block driver */
+ blkdev->index = (blkdev->xendev.dev - 202 * 256) / 16;
+ blkdev->index = drive_get_index(IF_XEN, 0, blkdev->index);
+ if (blkdev->index == -1) {
+ /* setup via xenbus -> create new block driver instance */
+ xen_be_printf(&blkdev->xendev, 2, "create new bdrv (xenbus
setup)\n");
+ blkdev->bs = bdrv_new(blkdev->dev);
+ if (blkdev->bs) {
+ if (0 != bdrv_open2(blkdev->bs, blkdev->filename, qflags,
+ bdrv_find_format(blkdev->fileproto))) {
+ bdrv_delete(blkdev->bs);
+ blkdev->bs = NULL;
+ }
+ }
+ if (!blkdev->bs)
+ return -1;
+ } else {
+ /* setup via qemu cmdline -> already setup for us */
+ xen_be_printf(&blkdev->xendev, 2, "get configured bdrv
(cmdline setup)\n");
+ blkdev->bs = drives_table[blkdev->index].bdrv;
+ }
+ blkdev->file_blk = BLOCK_SIZE;
+ blkdev->file_size = bdrv_getlength(blkdev->bs);
+ if (blkdev->file_size < 0) {
+ xen_be_printf(&blkdev->xendev, 1, "bdrv_getlength: %d (%s)
| drv %s\n",
+ (int)blkdev->file_size,
strerror(-blkdev->file_size),
+ blkdev->bs->drv ?
blkdev->bs->drv->format_name : "-");
+ blkdev->file_size = 0;
+ }
+ have_barriers = blkdev->bs->drv &&
blkdev->bs->drv->bdrv_flush ? 1 : 0;
+
+ xen_be_printf(xendev, 1, "type \"%s\", fileproto
\"%s\", filename \"%s\","
+ " size %" PRId64 " (%" PRId64 " MB)\n",
+ blkdev->type, blkdev->fileproto, blkdev->filename,
+ blkdev->file_size, blkdev->file_size >> 20);
+
+ /* fill info */
+ xenstore_write_be_int(&blkdev->xendev, "feature-barrier",
have_barriers);
+ xenstore_write_be_int(&blkdev->xendev, "info",
info);
+ xenstore_write_be_int(&blkdev->xendev, "sector-size",
blkdev->file_blk);
+ xenstore_write_be_int(&blkdev->xendev, "sectors",
+ blkdev->file_size / blkdev->file_blk);
+ return 0;
+}
+
+static int blk_connect(struct XenDevice *xendev)
+{
+ struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+ if (-1 == xenstore_read_fe_int(&blkdev->xendev,
"ring-ref", &blkdev->ring_ref))
+ return -1;
+ if (-1 == xenstore_read_fe_int(&blkdev->xendev,
"event-channel",
+ &blkdev->xendev.remote_port))
+ return -1;
+
+ blkdev->protocol = BLKIF_PROTOCOL_NATIVE;
+ if (blkdev->xendev.protocol) {
+ if (0 == strcmp(blkdev->xendev.protocol, XEN_IO_PROTO_ABI_X86_32))
+ blkdev->protocol = BLKIF_PROTOCOL_X86_32;
+ if (0 == strcmp(blkdev->xendev.protocol, XEN_IO_PROTO_ABI_X86_64))
+ blkdev->protocol = BLKIF_PROTOCOL_X86_64;
+ }
+
+ blkdev->sring = xc_gnttab_map_grant_ref(blkdev->xendev.gnttabdev,
+ blkdev->xendev.dom,
+ blkdev->ring_ref,
+ PROT_READ | PROT_WRITE);
+ if (!blkdev->sring)
+ return -1;
+ blkdev->cnt_map++;
+
+ switch (blkdev->protocol) {
+ case BLKIF_PROTOCOL_NATIVE:
+ {
+ blkif_sring_t *sring_native = blkdev->sring;
+ BACK_RING_INIT(&blkdev->rings.native, sring_native, XC_PAGE_SIZE);
+ break;
+ }
+ case BLKIF_PROTOCOL_X86_32:
+ {
+ blkif_x86_32_sring_t *sring_x86_32 = blkdev->sring;
+ BACK_RING_INIT(&blkdev->rings.x86_32, sring_x86_32, XC_PAGE_SIZE);
+ break;
+ }
+ case BLKIF_PROTOCOL_X86_64:
+ {
+ blkif_x86_64_sring_t *sring_x86_64 = blkdev->sring;
+ BACK_RING_INIT(&blkdev->rings.x86_64, sring_x86_64, XC_PAGE_SIZE);
+ break;
+ }
+ }
+
+ xen_be_bind_evtchn(&blkdev->xendev);
+
+ xen_be_printf(&blkdev->xendev, 1, "ok: proto %s, ring-ref %d,
"
+ "remote port %d, local port %d\n",
+ blkdev->xendev.protocol, blkdev->ring_ref,
+ blkdev->xendev.remote_port, blkdev->xendev.local_port);
+ return 0;
+}
+
+static void blk_disconnect(struct XenDevice *xendev)
+{
+ struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+
+ if (blkdev->bs) {
+ if (blkdev->index == -1) {
+ /* close/delete only if we created it ourself */
+ bdrv_close(blkdev->bs);
+ bdrv_delete(blkdev->bs);
+ }
+ blkdev->bs = NULL;
+ }
+ xen_be_unbind_evtchn(&blkdev->xendev);
+
+ if (blkdev->sring) {
+ xc_gnttab_munmap(blkdev->xendev.gnttabdev, blkdev->sring, 1);
+ blkdev->cnt_map--;
+ blkdev->sring = NULL;
+ }
+}
+
+static int blk_free(struct XenDevice *xendev)
+{
+ struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+ struct ioreq *ioreq;
+
+ while (!LIST_EMPTY(&blkdev->freelist)) {
+ ioreq = LIST_FIRST(&blkdev->freelist);
+ LIST_REMOVE(ioreq, list);
+ qemu_free(ioreq);
+ }
+
+ qemu_free(blkdev->params);
+ qemu_free(blkdev->mode);
+ return 0;
+}
+
+static void blk_event(struct XenDevice *xendev)
+{
+ struct XenBlkDev *blkdev = container_of(xendev, struct XenBlkDev, xendev);
+ blk_handle_requests(blkdev);
+}
+
+struct XenDevOps xen_blkdev_ops = {
+ .size = sizeof(struct XenBlkDev),
+ .flags = DEVOPS_FLAG_NEED_GNTDEV,
+ .alloc = blk_alloc,
+ .init = blk_init,
+ .connect = blk_connect,
+ .disconnect = blk_disconnect,
+ .event = blk_event,
+ .free = blk_free,
+};
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index 38e7335..f750b0b 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -67,6 +67,7 @@ static int xen_init_pv(DisplayState *ds)
xen_be_register("console", &xen_console_ops);
xen_be_register("vkbd", &xen_kbdmouse_ops);
xen_be_register("vfb", &xen_framebuffer_ops);
+ xen_be_register("qdisk", &xen_blkdev_ops);
/* setup framebuffer */
xen_set_display(xen_domid, ds);
diff --git a/sysemu.h b/sysemu.h
index 976fecc..67404c2 100644
--- a/sysemu.h
+++ b/sysemu.h
@@ -123,7 +123,7 @@ extern unsigned int nb_prom_envs;
#endif
typedef enum {
- IF_IDE, IF_SCSI, IF_FLOPPY, IF_PFLASH, IF_MTD, IF_SD
+ IF_IDE, IF_SCSI, IF_FLOPPY, IF_PFLASH, IF_MTD, IF_SD, IF_XEN
} BlockInterfaceType;
typedef struct DriveInfo {
diff --git a/vl.c b/vl.c
index 0d275bf..a28ba86 100644
--- a/vl.c
+++ b/vl.c
@@ -5563,6 +5563,9 @@ static int drive_init(struct drive_opt *arg, int snapshot,
} else if (!strcmp(buf, "sd")) {
type = IF_SD;
max_devs = 0;
+ } else if (!strcmp(buf, "xen")) {
+ type = IF_XEN;
+ max_devs = 0;
} else {
fprintf(stderr, "qemu: ''%s'' unsupported bus
type ''%s''\n", str, buf);
return -1;
@@ -5750,6 +5753,7 @@ static int drive_init(struct drive_opt *arg, int snapshot,
switch(type) {
case IF_IDE:
case IF_SCSI:
+ case IF_XEN:
switch(media) {
case MEDIA_DISK:
if (cyls != 0) {
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
This patch adds a network interface backend driver to qemu. It is a pure
userspace implemention using the gntdev interface. It uses "qnet" as
backend name in xenstore so it doesn''t interfere with the netback
backend (aka "vnif").
The network backend is hooked into the corrosponding qemu vlan, i.e.
vif 0 is hooked into vlan 0. To make the packages actually arrive
somewhere you additionally have to link the vlan to the outside world
using the usual qemu command line options such as "-net tap,...".
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 2 +-
hw/xen_backend.h | 1 +
hw/xen_machine_pv.c | 1 +
hw/xen_nic.c | 386 +++++++++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 389 insertions(+), 1 deletions(-)
create mode 100644 hw/xen_nic.c
diff --git a/Makefile.target b/Makefile.target
index 22bdd69..aa4e2c8 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -631,7 +631,7 @@ endif
# xen backend driver support
XEN_OBJS := xen_machine_pv.o xen_backend.o
-XEN_OBJS += xen_console.o xenfb.o xen_disk.o
+XEN_OBJS += xen_console.o xenfb.o xen_disk.o xen_nic.o
ifeq ($(CONFIG_XEN), yes)
OBJS += $(XEN_OBJS)
LIBS += $(XEN_LIBS)
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
index 4fe9ad0..d676aeb 100644
--- a/hw/xen_backend.h
+++ b/hw/xen_backend.h
@@ -89,6 +89,7 @@ extern struct XenDevOps xen_console_ops; /* xen_console.c
*/
extern struct XenDevOps xen_kbdmouse_ops; /* xen_framebuffer.c */
extern struct XenDevOps xen_framebuffer_ops; /* xen_framebuffer.c */
extern struct XenDevOps xen_blkdev_ops; /* xen_disk.c */
+extern struct XenDevOps xen_netdev_ops; /* xen_nic.c */
void xen_set_display(int domid, DisplayState *ds);
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index f750b0b..9ba2c8a 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -68,6 +68,7 @@ static int xen_init_pv(DisplayState *ds)
xen_be_register("vkbd", &xen_kbdmouse_ops);
xen_be_register("vfb", &xen_framebuffer_ops);
xen_be_register("qdisk", &xen_blkdev_ops);
+ xen_be_register("qnic", &xen_netdev_ops);
/* setup framebuffer */
xen_set_display(xen_domid, ds);
diff --git a/hw/xen_nic.c b/hw/xen_nic.c
new file mode 100644
index 0000000..6772028
--- /dev/null
+++ b/hw/xen_nic.c
@@ -0,0 +1,386 @@
+/*
+ * xen paravirt network card backend
+ *
+ * (c) Gerd Hoffmann <kraxel@redhat.com>
+ *
+ * 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; under version 2 of the License.
+ *
+ * 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, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <unistd.h>
+#include <signal.h>
+#include <inttypes.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <pthread.h>
+#include <sys/socket.h>
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <sys/wait.h>
+#include <linux/if.h>
+#include <linux/if_tun.h>
+
+#include <xs.h>
+#include <xenctrl.h>
+#include <xen/io/xenbus.h>
+#include <xen/io/netif.h>
+
+#include "hw.h"
+#include "net.h"
+#include "qemu-char.h"
+#include "xen_backend.h"
+
+/* ------------------------------------------------------------- */
+
+struct XenNetDev {
+ struct XenDevice xendev; /* must be first */
+ char *mac;
+ int tx_work;
+ int tx_ring_ref;
+ int rx_ring_ref;
+ struct netif_tx_sring *txs;
+ struct netif_rx_sring *rxs;
+ netif_tx_back_ring_t tx_ring;
+ netif_rx_back_ring_t rx_ring;
+ VLANClientState *vs;
+};
+
+/* ------------------------------------------------------------- */
+
+static void net_tx_response(struct XenNetDev *netdev, netif_tx_request_t *txp,
int8_t st)
+{
+ RING_IDX i = netdev->tx_ring.rsp_prod_pvt;
+ netif_tx_response_t *resp;
+ int notify;
+
+ resp = RING_GET_RESPONSE(&netdev->tx_ring, i);
+ resp->id = txp->id;
+ resp->status = st;
+
+#if 0
+ if (txp->flags & NETTXF_extra_info)
+ RING_GET_RESPONSE(&netdev->tx_ring, ++i)->status = NETIF_RSP_NULL;
+#endif
+
+ netdev->tx_ring.rsp_prod_pvt = ++i;
+ RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->tx_ring, notify);
+ if (notify)
+ xen_be_send_notify(&netdev->xendev);
+
+ if (i == netdev->tx_ring.req_cons) {
+ int more_to_do;
+ RING_FINAL_CHECK_FOR_REQUESTS(&netdev->tx_ring, more_to_do);
+ if (more_to_do)
+ netdev->tx_work++;
+ }
+}
+
+static void net_tx_error(struct XenNetDev *netdev, netif_tx_request_t *txp,
RING_IDX end)
+{
+#if 0
+ /*
+ * Hmm, why netback fails everything in the ring?
+ * Should we do that even when not supporting SG and TSO?
+ */
+ RING_IDX cons = netdev->tx_ring.req_cons;
+
+ do {
+ make_tx_response(netif, txp, NETIF_RSP_ERROR);
+ if (cons >= end)
+ break;
+ txp = RING_GET_REQUEST(&netdev->tx_ring, cons++);
+ } while (1);
+ netdev->tx_ring.req_cons = cons;
+ netif_schedule_work(netif);
+ netif_put(netif);
+#else
+ net_tx_response(netdev, txp, NETIF_RSP_ERROR);
+#endif
+}
+
+static void net_tx_packets(struct XenNetDev *netdev)
+{
+ netif_tx_request_t txreq;
+ RING_IDX rc, rp;
+ void *page;
+
+ for (;;) {
+ rc = netdev->tx_ring.req_cons;
+ rp = netdev->tx_ring.sring->req_prod;
+ xen_rmb(); /* Ensure we see queued requests up to ''rp''. */
+
+ while ((rc != rp)) {
+ if (RING_REQUEST_CONS_OVERFLOW(&netdev->tx_ring, rc))
+ break;
+ memcpy(&txreq, RING_GET_REQUEST(&netdev->tx_ring, rc),
sizeof(txreq));
+ netdev->tx_ring.req_cons = ++rc;
+
+#if 1
+ /* should not happen in theory, we don''t announce the *
+ * feature-{sg,gso,whatelse} flags in xenstore (yet?) */
+ if (txreq.flags & NETTXF_extra_info) {
+ xen_be_printf(&netdev->xendev, 0, "FIXME: extra info
flag\n");
+ net_tx_error(netdev, &txreq, rc);
+ continue;
+ }
+ if (txreq.flags & NETTXF_more_data) {
+ xen_be_printf(&netdev->xendev, 0, "FIXME: more data
flag\n");
+ net_tx_error(netdev, &txreq, rc);
+ continue;
+ }
+#endif
+
+ if (txreq.size < 14) {
+ xen_be_printf(&netdev->xendev, 0, "bad packet size: %d\n",
txreq.size);
+ net_tx_error(netdev, &txreq, rc);
+ continue;
+ }
+
+ if ((txreq.offset + txreq.size) > XC_PAGE_SIZE) {
+ xen_be_printf(&netdev->xendev, 0, "error: page crossing\n");
+ net_tx_error(netdev, &txreq, rc);
+ continue;
+ }
+
+ xen_be_printf(&netdev->xendev, 3, "tx packet ref %d, off %d,
len %d, flags 0x%x%s%s%s%s\n",
+ txreq.gref, txreq.offset, txreq.size, txreq.flags,
+ (txreq.flags & NETTXF_csum_blank) ? " csum_blank" :
"",
+ (txreq.flags & NETTXF_data_validated) ? " data_validated" :
"",
+ (txreq.flags & NETTXF_more_data) ? " more_data" :
"",
+ (txreq.flags & NETTXF_extra_info) ? " extra_info" :
"");
+
+ page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
+ netdev->xendev.dom,
+ txreq.gref, PROT_READ);
+ if (NULL == page) {
+ xen_be_printf(&netdev->xendev, 0, "error: tx gref dereference
failed (%d)\n",
+ txreq.gref);
+ net_tx_error(netdev, &txreq, rc);
+ continue;
+ }
+ if (txreq.flags & NETTXF_csum_blank)
+ net_checksum_calculate(page + txreq.offset, txreq.size);
+ qemu_send_packet(netdev->vs, page + txreq.offset, txreq.size);
+ xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
+ net_tx_response(netdev, &txreq, NETIF_RSP_OKAY);
+ }
+ if (!netdev->tx_work)
+ break;
+ netdev->tx_work = 0;
+ }
+}
+
+/* ------------------------------------------------------------- */
+
+static void net_rx_response(struct XenNetDev *netdev,
+ netif_rx_request_t *req, int8_t st,
+ uint16_t offset, uint16_t size,
+ uint16_t flags)
+{
+ RING_IDX i = netdev->rx_ring.rsp_prod_pvt;
+ netif_rx_response_t *resp;
+ int notify;
+
+ resp = RING_GET_RESPONSE(&netdev->rx_ring, i);
+ resp->offset = offset;
+ resp->flags = flags;
+ resp->id = req->id;
+ resp->status = (int16_t)size;
+ if (st < 0)
+ resp->status = (int16_t)st;
+
+ xen_be_printf(&netdev->xendev, 3, "rx response: idx %d, status
%d, flags 0x%x\n",
+ i, resp->status, resp->flags);
+
+ netdev->rx_ring.rsp_prod_pvt = ++i;
+ RING_PUSH_RESPONSES_AND_CHECK_NOTIFY(&netdev->rx_ring, notify);
+ if (notify)
+ xen_be_send_notify(&netdev->xendev);
+}
+
+#define NET_IP_ALIGN 2
+
+static int net_rx_ok(void *opaque)
+{
+ struct XenNetDev *netdev = opaque;
+ RING_IDX rc, rp;
+
+ if (netdev->xendev.be_state != XenbusStateConnected)
+ return 0;
+
+ rc = netdev->rx_ring.req_cons;
+ rp = netdev->rx_ring.sring->req_prod;
+ xen_rmb();
+
+ if (rc == rp || RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
+ xen_be_printf(&netdev->xendev, 2, "%s: no rx buffers
(%d/%d)\n",
+ __FUNCTION__, rc, rp);
+ return 0;
+ }
+ return 1;
+}
+
+static void net_rx_packet(void *opaque, const uint8_t *buf, int size)
+{
+ struct XenNetDev *netdev = opaque;
+ netif_rx_request_t rxreq;
+ RING_IDX rc, rp;
+ void *page;
+
+ if (netdev->xendev.be_state != XenbusStateConnected)
+ return;
+
+ rc = netdev->rx_ring.req_cons;
+ rp = netdev->rx_ring.sring->req_prod;
+ xen_rmb(); /* Ensure we see queued requests up to ''rp''.
*/
+
+ if (rc == rp || RING_REQUEST_CONS_OVERFLOW(&netdev->rx_ring, rc)) {
+ xen_be_printf(&netdev->xendev, 2, "no buffer, drop
packet\n");
+ return;
+ }
+ if (size > XC_PAGE_SIZE - NET_IP_ALIGN) {
+ xen_be_printf(&netdev->xendev, 0, "packet too big (%d >
%ld)",
+ size, XC_PAGE_SIZE - NET_IP_ALIGN);
+ return;
+ }
+
+ memcpy(&rxreq, RING_GET_REQUEST(&netdev->rx_ring, rc),
sizeof(rxreq));
+ netdev->rx_ring.req_cons = ++rc;
+
+ page = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
+ netdev->xendev.dom,
+ rxreq.gref, PROT_WRITE);
+ if (NULL == page) {
+ xen_be_printf(&netdev->xendev, 0, "error: rx gref dereference
failed (%d)\n",
+ rxreq.gref);
+ net_rx_response(netdev, &rxreq, NETIF_RSP_ERROR, 0, 0, 0);
+ return;
+ }
+ memcpy(page + NET_IP_ALIGN, buf, size);
+ xc_gnttab_munmap(netdev->xendev.gnttabdev, page, 1);
+ net_rx_response(netdev, &rxreq, NETIF_RSP_OKAY, NET_IP_ALIGN, size, 0);
+}
+
+/* ------------------------------------------------------------- */
+
+static int net_init(struct XenDevice *xendev)
+{
+ struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
+ VLANState *vlan;
+
+ /* read xenstore entries */
+ if (NULL == netdev->mac)
+ netdev->mac = xenstore_read_be_str(&netdev->xendev,
"mac");
+
+ /* do we have all we need? */
+ if (NULL == netdev->mac)
+ return -1;
+
+ vlan = qemu_find_vlan(netdev->xendev.dev);
+ netdev->vs = qemu_new_vlan_client(vlan, net_rx_packet, net_rx_ok,
netdev);
+ snprintf(netdev->vs->info_str, sizeof(netdev->vs->info_str),
+ "nic: xenbus vif macaddr=%s", netdev->mac);
+
+ /* fill info */
+ xenstore_write_be_int(&netdev->xendev, "feature-rx-copy",
1);
+ xenstore_write_be_int(&netdev->xendev, "feature-rx-flip",
0);
+
+ return 0;
+}
+
+static int net_connect(struct XenDevice *xendev)
+{
+ struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
+ int rx_copy;
+
+ if (-1 == xenstore_read_fe_int(&netdev->xendev,
"tx-ring-ref",
+ &netdev->tx_ring_ref))
+ return -1;
+ if (-1 == xenstore_read_fe_int(&netdev->xendev,
"rx-ring-ref",
+ &netdev->rx_ring_ref))
+ return 1;
+ if (-1 == xenstore_read_fe_int(&netdev->xendev,
"event-channel",
+ &netdev->xendev.remote_port))
+ return -1;
+
+ if (-1 == xenstore_read_fe_int(&netdev->xendev,
"request-rx-copy", &rx_copy))
+ rx_copy = 0;
+ if (0 == rx_copy) {
+ xen_be_printf(&netdev->xendev, 0, "frontend doesn''t
support rx-copy.\n");
+ return -1;
+ }
+
+ netdev->txs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
+ netdev->xendev.dom,
+ netdev->tx_ring_ref,
+ PROT_READ | PROT_WRITE);
+ netdev->rxs = xc_gnttab_map_grant_ref(netdev->xendev.gnttabdev,
+ netdev->xendev.dom,
+ netdev->rx_ring_ref,
+ PROT_READ | PROT_WRITE);
+ if (!netdev->txs || !netdev->rxs)
+ return -1;
+ BACK_RING_INIT(&netdev->tx_ring, netdev->txs, XC_PAGE_SIZE);
+ BACK_RING_INIT(&netdev->rx_ring, netdev->rxs, XC_PAGE_SIZE);
+
+ xen_be_bind_evtchn(&netdev->xendev);
+
+ xen_be_printf(&netdev->xendev, 1, "ok: tx-ring-ref %d,
rx-ring-ref %d, "
+ "remote port %d, local port %d\n",
+ netdev->tx_ring_ref, netdev->rx_ring_ref,
+ netdev->xendev.remote_port, netdev->xendev.local_port);
+ return 0;
+}
+
+static void net_disconnect(struct XenDevice *xendev)
+{
+ struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
+
+ xen_be_unbind_evtchn(&netdev->xendev);
+
+ if (netdev->txs) {
+ xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->txs, 1);
+ netdev->txs = NULL;
+ }
+ if (netdev->rxs) {
+ xc_gnttab_munmap(netdev->xendev.gnttabdev, netdev->rxs, 1);
+ netdev->rxs = NULL;
+ }
+ if (netdev->vs) {
+ qemu_del_vlan_client(netdev->vs);
+ netdev->vs = NULL;
+ }
+}
+
+static void net_event(struct XenDevice *xendev)
+{
+ struct XenNetDev *netdev = container_of(xendev, struct XenNetDev, xendev);
+ net_tx_packets(netdev);
+}
+
+/* ------------------------------------------------------------- */
+
+struct XenDevOps xen_netdev_ops = {
+ .size = sizeof(struct XenNetDev),
+ .flags = DEVOPS_FLAG_NEED_GNTDEV,
+ .init = net_init,
+ .connect = net_connect,
+ .event = net_event,
+ .disconnect = net_disconnect,
+};
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 12:23 UTC
[Xen-devel] [PATCH 7/7] xen: blk & nic configuration via cmd line.
This patch makes qemu create backend and frontend device entries in
xenstore for devices configured on the command line. It will use
qdisk and qnic backend names, so the qemu internal backends will
be used.
Disks can be created using -drive if=xen,file=...
Nics can be created using -net nic,macaddr=...
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
Makefile.target | 2 +-
hw/xen_backend.c | 1 +
hw/xen_backend.h | 8 +++
hw/xen_devconfig.c | 143 +++++++++++++++++++++++++++++++++++++++++++++++++++
hw/xen_machine_pv.c | 19 ++++++-
5 files changed, 171 insertions(+), 2 deletions(-)
create mode 100644 hw/xen_devconfig.c
diff --git a/Makefile.target b/Makefile.target
index aa4e2c8..57ba699 100644
--- a/Makefile.target
+++ b/Makefile.target
@@ -630,7 +630,7 @@ LIBS += $(CONFIG_BLUEZ_LIBS)
endif
# xen backend driver support
-XEN_OBJS := xen_machine_pv.o xen_backend.o
+XEN_OBJS := xen_machine_pv.o xen_backend.o xen_devconfig.o
XEN_OBJS += xen_console.o xenfb.o xen_disk.o xen_nic.o
ifeq ($(CONFIG_XEN), yes)
OBJS += $(XEN_OBJS)
diff --git a/hw/xen_backend.c b/hw/xen_backend.c
index 2678dfa..2bd4433 100644
--- a/hw/xen_backend.c
+++ b/hw/xen_backend.c
@@ -45,6 +45,7 @@
/* public */
int xen_xc;
struct xs_handle *xenstore = NULL;
+const char *xen_protocol;
/* private */
static TAILQ_HEAD(XenDeviceHead, XenDevice) xendevs =
TAILQ_HEAD_INITIALIZER(xendevs);
diff --git a/hw/xen_backend.h b/hw/xen_backend.h
index d676aeb..7d2cf0a 100644
--- a/hw/xen_backend.h
+++ b/hw/xen_backend.h
@@ -3,6 +3,8 @@
#include "xen_common.h"
#include "sysemu.h"
+#include "net.h"
+#include "block_int.h"
/* ------------------------------------------------------------- */
@@ -56,6 +58,7 @@ struct XenDevice {
/* variables */
extern int xen_xc;
extern struct xs_handle *xenstore;
+extern const char *xen_protocol;
/* xenstore helper functions */
int xenstore_write_str(const char *base, const char *node, const char *val);
@@ -93,4 +96,9 @@ extern struct XenDevOps xen_netdev_ops; /* xen_nic.c
*/
void xen_set_display(int domid, DisplayState *ds);
+/* configuration (aka xenbus setup) */
+void xen_config_cleanup(void);
+int xen_config_dev_blk(DriveInfo *disk);
+int xen_config_dev_nic(NICInfo *nic);
+
#endif /* QEMU_HW_XEN_BACKEND_H */
diff --git a/hw/xen_devconfig.c b/hw/xen_devconfig.c
new file mode 100644
index 0000000..885c741
--- /dev/null
+++ b/hw/xen_devconfig.c
@@ -0,0 +1,143 @@
+#include "xen_backend.h"
+
+/* ------------------------------------------------------------- */
+
+struct xs_dirs {
+ char *xs_dir;
+ TAILQ_ENTRY(xs_dirs) list;
+};
+static TAILQ_HEAD(xs_dirs_head, xs_dirs) xs_cleanup =
TAILQ_HEAD_INITIALIZER(xs_cleanup);
+
+static void xen_config_cleanup_dir(char *dir)
+{
+ struct xs_dirs *d;
+
+ d = qemu_malloc(sizeof(*d));
+ if (!d)
+ return;
+ d->xs_dir = dir;
+ TAILQ_INSERT_TAIL(&xs_cleanup, d, list);
+}
+
+void xen_config_cleanup(void)
+{
+ struct xs_dirs *d;
+
+ fprintf(stderr, "xen be: %s\n", __FUNCTION__);
+ TAILQ_FOREACH(d, &xs_cleanup, list) {
+ xs_rm(xenstore, 0, d->xs_dir);
+ }
+}
+
+/* ------------------------------------------------------------- */
+
+static int xen_config_dev_mkdir(char *dev, int p)
+{
+ struct xs_permissions perms = {
+ .id = xen_domid,
+ .perms = p,
+ };
+
+ if (!xs_mkdir(xenstore, 0, dev)) {
+ fprintf(stderr, "xs_mkdir %s: failed\n", dev);
+ return -1;
+ }
+ xen_config_cleanup_dir(qemu_strdup(dev));
+
+ if (!xs_set_permissions(xenstore, 0, dev, &perms, 1)) {
+ fprintf(stderr, "%s: xs_set_permissions failed\n", __FUNCTION__);
+ return -1;
+ }
+ return 0;
+}
+
+static int xen_config_dev_dirs(const char *ftype, const char *btype, int vdev,
+ char *fe, char *be, int len)
+{
+ char *dom;
+
+ dom = xs_get_domain_path(xenstore, xen_domid);
+ snprintf(fe, len, "%s/device/%s/%d", dom, ftype, vdev);
+ free(dom);
+
+ dom = xs_get_domain_path(xenstore, 0);
+ snprintf(be, len, "%s/backend/%s/%d/%d", dom, btype, xen_domid,
vdev);
+ free(dom);
+
+ xen_config_dev_mkdir(fe, XS_PERM_WRITE);
+ xen_config_dev_mkdir(be, XS_PERM_READ);
+ return 0;
+}
+
+static int xen_config_dev_all(char *fe, char *be)
+{
+ /* frontend */
+ if (xen_protocol)
+ xenstore_write_str(fe, "protocol", xen_protocol);
+
+ xenstore_write_int(fe, "state",
XenbusStateInitialising);
+ xenstore_write_int(fe, "backend-id", 0);
+ xenstore_write_str(fe, "backend", be);
+
+ /* backend */
+ xenstore_write_str(be, "domain", qemu_name ? qemu_name :
"no-name");
+ xenstore_write_int(be, "online", 1);
+ xenstore_write_int(be, "state",
XenbusStateInitialising);
+ xenstore_write_int(be, "frontend-id", xen_domid);
+ xenstore_write_str(be, "frontend", fe);
+
+ return 0;
+}
+
+/* ------------------------------------------------------------- */
+
+int xen_config_dev_blk(DriveInfo *disk)
+{
+ char fe[256], be[256];
+ int vdev = 202 * 256 + 16 * disk->unit;
+ int cdrom = disk->bdrv->type == BDRV_TYPE_CDROM;
+ const char *devtype = cdrom ? "cdrom" : "disk";
+ const char *mode = cdrom ? "r" : "w";
+
+ snprintf(disk->bdrv->device_name,
sizeof(disk->bdrv->device_name),
+ "xvd%c", ''a'' + disk->unit);
+ fprintf(stderr, "xen be: config disk %d [%s]: %s\n",
+ disk->unit, disk->bdrv->device_name, disk->bdrv->filename);
+ xen_config_dev_dirs("vbd", "qdisk", vdev, fe, be,
sizeof(fe));
+
+ /* frontend */
+ xenstore_write_int(fe, "virtual-device", vdev);
+ xenstore_write_str(fe, "device-type", devtype);
+
+ /* backend */
+ xenstore_write_str(be, "dev",
disk->bdrv->device_name);
+ xenstore_write_str(be, "type", "file");
+ xenstore_write_str(be, "params",
disk->bdrv->filename);
+ xenstore_write_str(be, "mode", mode);
+
+ /* common stuff */
+ return xen_config_dev_all(fe, be);
+}
+
+int xen_config_dev_nic(NICInfo *nic)
+{
+ char fe[256], be[256];
+ char mac[20];
+
+ snprintf(mac, sizeof(mac), "%02x:%02x:%02x:%02x:%02x:%02x",
+ nic->macaddr[0], nic->macaddr[1], nic->macaddr[2],
+ nic->macaddr[3], nic->macaddr[4], nic->macaddr[5]);
+ fprintf(stderr, "xen be: config nic %d: mac=\"%s\"\n",
nic->vlan->id, mac);
+ xen_config_dev_dirs("vif", "qnic", nic->vlan->id,
fe, be, sizeof(fe));
+
+ /* frontend */
+ xenstore_write_int(fe, "handle", nic->vlan->id);
+ xenstore_write_str(fe, "mac", mac);
+
+ /* backend */
+ xenstore_write_int(be, "handle", nic->vlan->id);
+ xenstore_write_str(be, "mac", mac);
+
+ /* common stuff */
+ return xen_config_dev_all(fe, be);
+}
diff --git a/hw/xen_machine_pv.c b/hw/xen_machine_pv.c
index 9ba2c8a..95ad11e 100644
--- a/hw/xen_machine_pv.c
+++ b/hw/xen_machine_pv.c
@@ -87,7 +87,7 @@ static void xenpv_init(ram_addr_t ram_size, int vga_ram_size,
const char *cpu_model)
{
CPUState *env;
- int rc;
+ int index,i,rc;
rc = xen_init_pv(ds);
if (-1 == rc)
@@ -104,6 +104,23 @@ static void xenpv_init(ram_addr_t ram_size, int
vga_ram_size,
env = cpu_init(cpu_model);
env->halted = 1;
+ /* configure disks */
+ for (i = 0; i < 16; i++) {
+ index = drive_get_index(IF_XEN, 0, i);
+ if (index == -1)
+ continue;
+ xen_config_dev_blk(drives_table + index);
+ }
+
+ /* configure nics */
+ for (i = 0; i < nb_nics; i++) {
+ if (nd_table[i].model && 0 != strcmp(nd_table[i].model,
"xen"))
+ continue;
+ xen_config_dev_nic(nd_table + i);
+ }
+
+ /* config cleanup hook */
+ atexit(xen_config_cleanup);
return;
err:
--
1.5.6.5
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Ian Jackson
2008-Oct-28 14:34 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
People unfamiliar with the context should note that Gerd''s submission
is NOT `merging some Xen bits into qemu''. The code that Gerd is
submitting is VERY SUBSTANTIALLY DIFFERENT from the code which exists
in the Xen version of qemu.
Gerd Hoffmann writes ("[Xen-devel] [PATCH 0/7] merge some xen bits into
qemu"):> The console and framebuffer backend drivers are largely based on the
> xen code, the other bits are rewritten from scratch. Nevertheless
> the code should be functionally identical.
I''m afraid I''m still opposed to the approach you are taking
here.
The correct route for Xen support for qemu is via qemu-xen-unstable,
not directly into upstream qemu.
If you want to generalise the backend driver machinery, which in
qemu-xen-unstable is used only for the framebuffer, you should submit
your generalised backend machinery on qemu-devel. We will review it
and include it (when it is of appropriate quality).
Then when we have the Xen-specific structure and interfaces properly
supported, it would be right to submit parts to qemu upstream. This
will form a good basis for your other backend drivers (which are not
useful in the context of a real Xen setup).
> With first next four patches in place upstream qemu can replace
xen''s
> qemu-dm for paravirtual domains. Due to some small differences in
> the command line syntax it can''t work as drop-in replacement
though.
qemu-dm is hardly used for paravirtual domains so this is not really
very helpful. qemu-dm is not even used at all for most PV domains.
The main purpose of qemu-dm is for HVM domains.
Command-line differences are not very important in the grand scheme of
things but they too need to come through xen-devel so that there is a
sane transition.
> The last three patches add full userspace implementations of block and
> nic backend drivers using the grant table device (gntdev) and command
> line support config support for setting up these backends.
>
> xen support is implemented using another machine type. xen''s
qemu-dm
> already uses the machine type to switch between paravirtualized and
> fully virtualized machines, so this was the natural choice. qemu
> gets a new "xenpv" machine type additionally to the
"pc" and "isapc"
> ones.
For the avoidance of any doubt, Gerd''s machine types are for
_emulating_ a Xen environment on a machine without a real Xen
hypervisor.
Gerd: Why do I keep having to repeat this point ? Your messages
always obscure it.
So in summary: Gerd, please post your suggestions to xen-devel only
and explain what the purpose of each patch is.
If their intended use is for a system with a real Xen hypervisor and
xend, and the patches makes sense, then we will of course take them.
If the intended use is to clean up the code so that it will be easier
for you to fit in your simulated-Xen support then that is fine too,
but the patches must go through xen-devel and be discussed there in
that context. When we have agreed on and committed changes in
qemu-xen-unstable, it will be appropriate to submit those to
qemu-devel.
In any case, we do not want any huge upheavals in this area that are
not dealt with appropriately through the Xen testing and release
processes. We definitely don''t want to have replacements of whole
swathes of our existing code appear in qemu upstream!
Please stop crossposting to qemu-devel; it is just causing confusion.
Ian.
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Anthony Liguori
2008-Oct-28 15:11 UTC
Re: [Qemu-devel] Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Hi Ian, Ian Jackson wrote:> People unfamiliar with the context should note that Gerd''s submission > is NOT `merging some Xen bits into qemu''. The code that Gerd is > submitting is VERY SUBSTANTIALLY DIFFERENT from the code which exists > in the Xen version of qemu. >Here''s how I see things: Gerd has implemented an alternative userspace for Xen PV guests all based on QEMU. The code is well structured and integrates into QEMU in a nice way. On technical merits alone, I have a hard time not merging it into QEMU. It''s not Xend/blktap and while you all haven''t been clear on this point, I don''t think there''s any chance that Xen.org or whatever your formal body is will ever bless this approach. Naturally, this alternative userspace fits well for something like Xenner which emulates a Xen PV guest in QEMU or KVM. This is also something that I don''t think Xen.org will ever bless. Gerd has posted this series on xen-devel and there hasn''t been any real substantial feedback. He''s been pretty patient about it. I think this is somewhat misleading because I don''t think there''s any chance these patches will ever be pulled into xen-unstable. And BTW, any code that has a remote chance of being merged into QEMU eventually should be cross-posted on qemu-devel. Developing on xen-devel and then dropping a big pile of crud onto qemu-devel is not a viable development style. The same is true for other QEMU forks like KVM. All this said, I don''t think we should just pull in Gerd''s patches as-is and forget about the Xen.org blessing. I want to encourage all of the work that''s being done in xen-unstable to be merged back into QEMU upstream and that would certainly discourage this process. So at this point, I think we need some clarification. Ian, if you have no intention of ever merging these patches in any form, I think you should be clear about that. Gerd, I think you need to restructure your patches to do two things. The first is to never mention the name "Xen" as it''s trademarked and owned by Xen.org. There should be no confusion that this functionality is in anyway endorsed or supported by Xen.org. The second thing is that whatever you do has to get easily out of the way of anything that''s in xen-unstable. It should cause no merging pain to Ian. It should also be relatively easy to exclude from the build. If you just keep everything as separate files as you are, then it''s just a matter of not compiling them in. I hope this will make everyone happy and we can stop posting this series and arguing about it. Does anyone have any objections to this? Regards, Anthony Liguori _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 15:53 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Ian Jackson wrote:> People unfamiliar with the context should note that Gerd''s submission > is NOT `merging some Xen bits into qemu''. The code that Gerd is > submitting is VERY SUBSTANTIALLY DIFFERENT from the code which exists > in the Xen version of qemu.Yes, as the the intro text says. You''ve even quoted the paragraph here:> Gerd Hoffmann writes ("[Xen-devel] [PATCH 0/7] merge some xen bits into qemu"): >> The console and framebuffer backend drivers are largely based on the >> xen code, the other bits are rewritten from scratch. Nevertheless >> the code should be functionally identical. > > I''m afraid I''m still opposed to the approach you are taking here. > > The correct route for Xen support for qemu is via qemu-xen-unstable, > not directly into upstream qemu.There are *two* patch series for a reason. The one for upstream qemu, and the one for qemu-xen. After applying both patch sets to both threes they are largely identical. Anthony indicated that he wants to wait for the qemu-xen patches being accepted before merging the patches intended for upstream qemu. So where exactly is your problem?> If you want to generalise the backend driver machinery, which in > qemu-xen-unstable is used only for the framebuffer, you should submit > your generalised backend machinery on qemu-devel.Just done.> qemu-dm is hardly used for paravirtual domains so this is not really > very helpful. qemu-dm is not even used at all for most PV domains.It isn''t mandatory. But every pv domain using the pv framebuffer needs qemu-dm. It also can be used to handle the console instead of xenconsoled.> Command-line differences are not very important in the grand scheme of > things but they too need to come through xen-devel so that there is a > sane transition.As noted in the intro text the qemu-xen patch series don''t change the command line interface. I think it is easier to hash that out later separately and not involve xend changes in this patch series.> For the avoidance of any doubt, Gerd''s machine types are for > _emulating_ a Xen environment on a machine without a real Xen > hypervisor.That is one long term goal, yes. And I want the same backend code being used for both emulation and xen support. This patchset is about xen support only. There are no emulation bits, they will come later.> Gerd: Why do I keep having to repeat this point ? Your messages > always obscure it.When submitting xen emulation patches I will clearly state that.> So in summary: Gerd, please post your suggestions to xen-devel only > and explain what the purpose of each patch is.Not involving qemu-devel here is wrong. Discussing patches for upstream qemu without involving the relevant people isn''t going to work. Each patch comes with a description of the changes.> If their intended use is for a system with a real Xen hypervisor and > xend, and the patches makes sense, then we will of course take them.They are intended for a system with a Xen hypervisor. The disk & nic backends need some windup in xend to be usable though.> In any case, we do not want any huge upheavals in this area that are > not dealt with appropriately through the Xen testing and release > processes. We definitely don''t want to have replacements of whole > swathes of our existing code appear in qemu upstream!The qemu-xen series is bisectable. You can take the patches one by one, run them through the test setup @ XenSource and complain loudly if I broke something. cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Markus Armbruster
2008-Oct-28 16:43 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Ian Jackson <Ian.Jackson@eu.citrix.com> writes: [...]> If their intended use is for a system with a real Xen hypervisor and > xend, and the patches makes sense, then we will of course take them.Refraining from comment on how to get Xen support into upstream qemu; instead let me just point out that Gerd''s framebuffer code is a massive improvement in maintainability over what we have now in qemu-xen-unstable, and I''d very much like to see it merged there. [...]> Please stop crossposting to qemu-devel; it is just causing confusion.I respectfully doubt that. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Jackson
2008-Oct-28 18:31 UTC
Re: [Qemu-devel] Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Anthony Liguori writes ("Re: [Qemu-devel] Re: [Xen-devel] [PATCH 0/7] merge
some xen bits into qemu"):> [stuff]
>
> It''s not Xend/blktap and while you all haven''t been clear
on this point,
> I don''t think there''s any chance that Xen.org or whatever
your formal
> body is will ever bless this approach.
Well, I don''t think it is our (Xen.org''s) role to bless it,
nor is it
our role to deprecate it. Having an alternative way of doing things
is good for everyone really so from a personal point of view I would
like to see it supported.
> [procedural discussion]
>
> So at this point, I think we need some clarification. Ian, if you have
> no intention of ever merging these patches in any form, I think you
> should be clear about that.
I do want to qemu upstream to support Xenner. And as regards qemu-xen
I do want Gerd''s changes to clean up our pv fb backend.
But I agree that I don''t think it makes sense for Gerd''s
_other_
backend drivers (eg, the blkback) to go into qemu-xen-unstable; I''m
happy for them to go into qemu upstream in due course.
> I hope [the things suggested below] will make everyone happy and we
> can stop posting this series and arguing about it. Does anyone have
> any objections to this?
I think perhaps I''m being very defensive because I see a giant merge
doom approaching. Gerd keeps posting his enormous patch series and my
immediate priority is to stop it because it will make my life very
difficult if it is just dumped in as-is.
So I apologise if I''m being overly defensive.
And I suppose you (Anthony) are right that I should not object to
qemu-devel having visibility of these proposed changes. So I
apologise for that.
> Gerd, I think you need to restructure your patches to do two things.
> The first is to never mention the name "Xen" as it''s
trademarked and
> owned by Xen.org. There should be no confusion that this functionality
> is in anyway endorsed or supported by Xen.org. The second thing is that
> whatever you do has to get easily out of the way of anything
that''s in
> xen-unstable. It should cause no merging pain to Ian. It should also
> be relatively easy to exclude from the build. If you just keep
> everything as separate files as you are, then it''s just a matter
of not
> compiling them in.
That kind of approach would make me very happy.
I''d also obviously like to share code with Xenner where this is
technically sensible.
I think what I would really like is to be able to straight away apply
to qemu-xen-unstable the subset of Gerd''s patches which are
- structural improvements to the qemu-dm pv fb backend
including separating out the
- code improvements to the above
but excluding all of the other non-actual-Xen stuff.
When this is done and in xen-unstable and working, we can provide it
to qemu-devel and at that point there won''t be a merge problem because
the code added in qemu upstream will be the same as that in qemu-xen.
_Then_ Gerd should submit his Xenner patches to qemu-devel and I''ll
have no objection (provided as Anthony suggests they don''t interfere
with our real-Xen setups, but I doubt there will be any insoluble
problems).
If we could agree on this as an approach then we could spend our time
dealing with code rather than trying to do politics.
In a spirit of goodwill I''ll probably have a spare hour or two
tomorrow to go through Gerd''s patch series and try to experimentally
separate out the backend-separation part with my tree and review it.
I''m afraid that I probably won''t manage to get that stuff into
qemu-xen-3.4 because I''m going away and have already just dumped a
huge pile of new qemu upstream stuff into qemu-xen-unstable (because I
started the merge late due to being pulled away on another project).
But I''m prepared to maintain another branch of qemu-xen-unstable if
that helps. I''ll discuss this with Keir tomorrow.
In terms of naming, I''m not an official representative of Xen.org or
Citrix on this point, but I think since the name `xenner'' has already
been used to refer to the RH KVM-based Xen-emulation, it might be
sensible to use that name to refer to Gerd''s new arrangements.
So for example, having `xenner_pv'' and `xenner_hvm'' machine
types, and
`-xenner-vif'' or whatever for Xenner''s emulated backends.
Ian.
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Ian Jackson
2008-Oct-28 18:32 UTC
Re: [Qemu-devel] Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
I wrote:> I think what I would really like is to be able to straight away apply > to qemu-xen-unstable the subset of Gerd''s patches which are > - structural improvements to the qemu-dm pv fb backend > including separating out the... common backend machinery> - code improvements to the above > but excluding all of the other non-actual-Xen stuff.Ian. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-28 21:15 UTC
Re: [Qemu-devel] Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Ian Jackson wrote:>> So at this point, I think we need some clarification. Ian, if you have >> no intention of ever merging these patches in any form, I think you >> should be clear about that. > > I do want to qemu upstream to support Xenner. And as regards qemu-xen > I do want Gerd''s changes to clean up our pv fb backend.Great.> But I agree that I don''t think it makes sense for Gerd''s _other_ > backend drivers (eg, the blkback) to go into qemu-xen-unstable; I''m > happy for them to go into qemu upstream in due course.> I think perhaps I''m being very defensive because I see a giant merge > doom approaching. Gerd keeps posting his enormous patch series and my > immediate priority is to stop it because it will make my life very > difficult if it is just dumped in as-is.To avoid that merge disaster I have prepared patches for both upstream and qemu-xen, so you''ll have the files I''m touching in both trees in a (almost) identical state when it comes to the next merge. git should have no trouble to handle it then.>> Gerd, I think you need to restructure your patches to do two things. >> The first is to never mention the name "Xen" as it''s trademarked and >> owned by Xen.org. There should be no confusion that this functionality >> is in anyway endorsed or supported by Xen.org. The second thing is that >> whatever you do has to get easily out of the way of anything that''s in >> xen-unstable. It should cause no merging pain to Ian. It should also >> be relatively easy to exclude from the build. If you just keep >> everything as separate files as you are, then it''s just a matter of not >> compiling them in. > > That kind of approach would make me very happy.I''m trying hard to not break qemu-xen and will continue to do so.> I''d also obviously like to share code with Xenner where this is > technically sensible. > > I think what I would really like is to be able to straight away apply > to qemu-xen-unstable the subset of Gerd''s patches which are > - structural improvements to the qemu-dm pv fb backend > including separating out the > - code improvements to the above > but excluding all of the other non-actual-Xen stuff.That are patches 1-4 (upstream) and patches 1-5 (qemu-xen). Patch #1 for upstream qemu is splitted into two for qemu-xen (#1 and #5) to make the qemu-xen patch series bisectable. The last three patches (5-7 upstream, 6-8 qemu-xen) are block backend, net backend and config bits for these.> When this is done and in xen-unstable and working, we can provide it > to qemu-devel and at that point there won''t be a merge problem because > the code added in qemu upstream will be the same as that in qemu-xen. > > _Then_ Gerd should submit his Xenner patches to qemu-devel and I''ll > have no objection (provided as Anthony suggests they don''t interfere > with our real-Xen setups, but I doubt there will be any insoluble > problems).I want keep both patch series in sync exactly to avoid merge issues. Thats why I''m posting the patches on both lists and take into account review comments from both sides. So when we agree on the final cut of the code it can be merged strait into both trees.> But I''m prepared to maintain another branch of qemu-xen-unstable if > that helps. I''ll discuss this wih Keir tomorrow.That will surely help as it will decouple qemu development from xen release cycles. I''d suggest to branch off qemu-xen-3.4 when xen-unstable freezes for the 3.4 release.> In terms of naming, I''m not an official representative of Xen.org or > Citrix on this point, but I think since the name `xenner'' has already > been used to refer to the RH KVM-based Xen-emulation, it might be > sensible to use that name to refer to Gerd''s new arrangements. > > So for example, having `xenner_pv'' and `xenner_hvm'' machine types, and > `-xenner-vif'' or whatever for Xenner''s emulated backends.The naming convention I''m using right now is prefix ''xen_'' for anything usable when running on the Xen hypervisor. That includes code which will be shared by xen and xenner such as the machine type. That also includes the block and net backends. They *do* work on Xen, and with some glue in xend you should be able to use them. The xen emulation bits are prefixed with ''xenner_''. That includes event channel, grant table and xenstore implementations for example. None of these patches are submitted yet. Right now xenner isn''t a separate compile time option. That should be easy to add though. cheers Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Samuel Thibault
2008-Oct-29 00:04 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Gerd Hoffmann, le Tue 28 Oct 2008 13:23:35 +0100, a écrit :> the framebuffer handles a restart cycle correctly > (i.e. pvgrub in graphical mode actually works).Did you also try to check whether it could perhaps fix xen bug #1365? Samuel _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-29 09:08 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Samuel Thibault wrote:> Gerd Hoffmann, le Tue 28 Oct 2008 13:23:35 +0100, a écrit : >> the framebuffer handles a restart cycle correctly >> (i.e. pvgrub in graphical mode actually works). > > Did you also try to check whether it could perhaps fix xen bug #1365?Chances are pretty good it does. I''ve seen qemu-dm crash too, although it isn''t 100% clear from the description it is actually the same bug. With my fb backend code in place I can boot the domain, connect with a vnc client, see the pvgrub boot menu (framebuffer mode), interactively select a kernel, then watch the guest os come up in framebuffer mode. This all with correct screen display and without qemu-dm crashing along the way. cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-29 10:53 UTC
[Xen-devel] Re: [Qemu-devel] [PATCH 3/7] xen: add console backend driver.
Blue Swirl wrote:> On 10/28/08, Gerd Hoffmann <kraxel@redhat.com> wrote: >> This patch adds a xenconsole backend driver. It it based on current >> xen-unstable code. It has been changed to make use of the common >> backend driver code. > >> + * You should have received a copy of the GNU General Public License >> + * along with this program; if not, write to the Free Software >> + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA > > As if anyone cares: FSF address in this patch and others should probably be > 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USAHmm, would it be ok to just scratch the whole paragraph from the comment, so we avoid stale address info in the first place? Shouldn''t be needed to make clear the source code is GPL, but IANAL ... cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Jackson
2008-Oct-30 12:07 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Gerd Hoffmann writes ("Re: [Qemu-devel] Re: [Xen-devel] [PATCH 0/7] merge
some xen bits into qemu"):> Ian Jackson wrote:
> > I think what I would really like is to be able to straight away apply
> > to qemu-xen-unstable the subset of Gerd''s patches which are
> > - structural improvements to the qemu-dm pv fb backend
> > including separating out the
> > - code improvements to the above
> > but excluding all of the other non-actual-Xen stuff.
>
> That are patches 1-4 (upstream) and patches 1-5 (qemu-xen).
I''ve taken a look at your `qemu-xen'' series. I''m
still keen to have
the new backend structure.
I''m afraid I wasn''t able in the short time I have today to
tease out
the parts which I think are important to go into qemu-xen-unstable now
from the parts which I think should (following discussion) go directly
into qemu upstream.
But as there is stuff here that I definitely want and which ought to
go into qemu-xen-unstable for Xen 3.4, I was wondering if I could ask
you to rework it into a form that I can apply to qemu-xen-unstable
when I get back ?
So on to the details:
Firstly, I should say that this review is purely from the point of
view of `should this stuff go into qemu-xen-unstable''. One reason not
to put things into qemu-xen-unstable is because they should go
directly upstream; another is of course that they still need work :-).
I''ll try to make this distinction clear in my comments.
Re 0002-xen-groundwork-for-xen-support.patch [1]:
* You introduce a new call to register xenpv_machine. But our
existing machinery in qemu-xen-unstable for i386-dm already does
this. So I think this is a mistake ? (I''m not opposed to a change
to the place where qemu-dm does its machine registration if that
would suit upstream better but you should surely remove the old
registration call too?)
* You introduce the QEMU_OPTION_domid the default behaviour of which
is XEN_ATTACH. I don''t think this is correct in the long term
because I think that by default qemu should never attach (or
create). The upstream qemu should always emulate Xen things by
default. In a real Xen environment the toolstack (xend) will know
to pass qemu-dm an option saying `please attach''.
Otherwise people who try to run vanilla qemu in what happens to be
a Xen guest will find all sorts of weird behaviours.
I know that we have a -domid option in qemu-xen-unstable but that
is one of the things that we have deliberately avoided passing
upstream. If upstream qemu is going to grow a -domid option it
needs to default to emulating Xen and we should have a new option
for enabling attach.
* I''m not sure that the new Makefile infrastructure you provide here
is a good idea for the qemu-xen-unstable tree. The stuff in
qemu-xen-unstable''s build system is very unpleasant but it is
specifically designed to avoid merge conflicts when qemu upstream
Makefiles change, as they often do.
I would very much prefer to continue to deal with the Makefiles on
this twin-track approach. So I would prefer it if you would add
your new object files to xen-hooks.mak and xen-setup and so forth.
Then when the corresponding changes go upstream I''ll just remove
them again. This will be much less pain for me than dealing with
<<< >>> in Makefiles.
I''m sorry that this means you need to do the build system parts of
your changes twice. But I think doing it exactly twice is better
than doing a similar amount of work every time there are
textually-nearby changes in upstream''s Makefiles.
Re 0003-xen-backend-driver-core.patch:
* You add xen_backend.o to your new Makefile infrastructure which
isn''t in qemu-xen-unstable yet - see above. But apart from that I
think this is probably OK.
* I did get some compiler warnings eg:
/u/iwj/work/qemu-iwj.git/hw/xenfb.c:810: warning: passing
argument 4 of ''xenfb_xs_printf'' discards qualifiers from
pointer
target type
Maybe you don''t have -Wwrite-strings in your setup ?
Re 0005-xen-add-framebuffer-backend-driver.patch:
* I haven''t reviewed this in detail but I am keen to have your new
framebuffer code. So provided this compiles and appears to work in
a quick test I would like to apply it ASAP.
I only haven''t done so right away because of the comments above.
Re 0004-xen-add-console-backend-driver.patch:
* There are many whitespace changes. Surely some mistake ?
Or is this with a view to getting the coding style to resemble
that in most of the rest of qemu ?
If so then I would prefer (as an exception!) a first separate patch
to re-indent the existing code in hw/xen_console.c: that is, a patch
consisting _only_ of whitespace changes. And then a separate patch
containing no whitespace changes.
As it is I haven''t been able to review this very deeply and I
don''t
have time today to redo the diff myself with -b ...
Re 0006-xen-sync-xen_machine_pv-with-upstream.patch:
* What is the purpose of this patch ? `sync xen_machine_pv with
upstream'' ? Which upstream ? Many of these changes seem
cosmetic.
Re 0007-xen-add-block-device-backend-driver.patch
and 0008-xen-add-net-backend-driver.patch:
* I don''t have any objection to these in principle. However they are
of no use or relevance to qemu-xen-unstable because we do not and
will not use them. So they should go into upstream directly rather
than via me I think ? However that needs to wait for the generic
backend, which definitely should go into qemu-xen-unstable before
it goes into qemu upstream. So I think these should wait for now.
Re 0009-xen-blk-nic-configuration-via-cmd-line.patch:
* I don''t think I fully understand the purpose of this patch. It
appears to be part of the Xenner emulation AFAICT ? When running
under the real Xen toolstack we do not have qemu write these kind
of entries to xenstore.
Instead, xenstore.c (in qemu-xen-unstable) reads the necessary
information from xenstore and sets up the qemu internals
appropriately.
And one final point:
* Have you been testing your tree on a real Xen host under xend ?
> I want keep both patch series in sync exactly to avoid merge issues.
> Thats why I''m posting the patches on both lists and take into
account
> review comments from both sides. So when we agree on the final cut of
> the code it can be merged strait into both trees.
That sounds sensible.
> > But I''m prepared to maintain another branch of
qemu-xen-unstable if
> > that helps. I''ll discuss this wih Keir tomorrow.
>
> That will surely help as it will decouple qemu development from xen
> release cycles. I''d suggest to branch off qemu-xen-3.4 when
> xen-unstable freezes for the 3.4 release.
I was thinking of doing that anyway.
> The naming convention I''m using right now is prefix
''xen_'' for anything
> usable when running on the Xen hypervisor. That includes code which
> will be shared by xen and xenner such as the machine type. That also
> includes the block and net backends. They *do* work on Xen, and with
> some glue in xend you should be able to use them.
Before we accept any incompatible changes into qemu-xen-unstable we
obviously need the corresponding changes to xend too ?
But we are not going to be using the block and net backends in qemu.
In a real Xen system these are provided by the dom0 kernel, which is
a far superior arrangement (in our opinion!)
AIUI from your patch the xen_mode specifies whether we''re doing Xenner
or real Xen ? Am I right in thinking that the modes are these:
* XEN_EMULATE - Xenner (no Xen hypervisor, no xend)
IMO this should be the default for the qemu xen machine types
in upstream qemu.
* XEN_ATTACH - for qemu-dm as invoked by xend
IMO arrangements should be made so that this is done only with
a special command-line option, which is provided by xend.
* XEN_CREATE - Xen hypervisor but no xend
This is an unusual use case. I don''t see any harm in having this
as an option for people who want to do it that way. However,
how do you control whether the block/net backends used are those
in the dom0 kernel or those in qemu ?
I''m afraid that since I''m going away now for three weeks I
won''t be
able to commit any revised versions soon. But as I say I hope to get
this into 3.4 and I look forward to seeing your new proposals when I
get back.
I hope you find this mail of mine suitably constructive. Again, sorry
about being overly defensive earlier.
Ian.
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Gerd Hoffmann
2008-Oct-30 14:04 UTC
Re: [Xen-devel] [PATCH 0/7] merge some xen bits into qemu
Hi,> I''m afraid I wasn''t able in the short time I have today to tease out > the parts which I think are important to go into qemu-xen-unstable now > from the parts which I think should (following discussion) go directly > into qemu upstream.Nevertheless I''m happy to finally see some progress here. Is somebody (Samuel?) working on qemu-xen while you are away?> Re 0002-xen-groundwork-for-xen-support.patch [1]: > > * You introduce a new call to register xenpv_machine. But our > existing machinery in qemu-xen-unstable for i386-dm already does > this. So I think this is a mistake ?It isn''t a mistake. It is in the initialization path for qemu-system-{i386,x86_64}, because for upstream qemu I plan to have this as runtime switch, not as separate binary. The code in question isn''t even compiled for qemu-dm, so the existing register call must stay. I can drop the chunk for qemu-xen though if you find that less confusing. Then you''ll get it with one of the next upstream merges instead.> * You introduce the QEMU_OPTION_domid the default behaviour of which > is XEN_ATTACH. I don''t think this is correct in the long term > because I think that by default qemu should never attach (or > create). The upstream qemu should always emulate Xen things by > default. In a real Xen environment the toolstack (xend) will know > to pass qemu-dm an option saying `please attach''.Yes, that was the agreement we reached last time this was discussed. The patch for upstream qemu implements exactly this behaviour: http://kraxel.fedorapeople.org/patches/qemu-upstream/0002-xen-groundwork-for-xen-support.patch The version for qemu-xen doesn''t change the command line arguments to avoid breaking qemu-dm <=> xend interaction. I don''t want to have xend patches in this series, to avoid making the merge even more complicated than it already is. Adding the new ones while keeping the old ones for xend compatibility should work though. I can do that, with comments making clear what the long-term plan is.> * I''m not sure that the new Makefile infrastructure you provide here > is a good idea for the qemu-xen-unstable tree. The stuff in > qemu-xen-unstable''s build system is very unpleasant but it is > specifically designed to avoid merge conflicts when qemu upstream > Makefiles change, as they often do. > > I would very much prefer to continue to deal with the Makefiles on > this twin-track approach. So I would prefer it if you would add > your new object files to xen-hooks.mak and xen-setup and so forth. > > Then when the corresponding changes go upstream I''ll just remove > them again. This will be much less pain for me than dealing with > <<< >>> in Makefiles.The very same Makefile infrastructure is intended for upstream though. But I can leave Makefile.target alone and just use xen-hooks.mak if you prefer that. Either way is fine for me.> Re 0003-xen-backend-driver-core.patch: > > * I did get some compiler warnings eg: > /u/iwj/work/qemu-iwj.git/hw/xenfb.c:810: warning: passing > argument 4 of ''xenfb_xs_printf'' discards qualifiers from pointer > target type > Maybe you don''t have -Wwrite-strings in your setup ?Will check.> Re 0005-xen-add-framebuffer-backend-driver.patch: > > * I haven''t reviewed this in detail but I am keen to have your new > framebuffer code. So provided this compiles and appears to work in > a quick test I would like to apply it ASAP.Fine ;)> Re 0004-xen-add-console-backend-driver.patch: > > * There are many whitespace changes. Surely some mistake ? > Or is this with a view to getting the coding style to resemble > that in most of the rest of qemu ?Yep, was done to adapt qemu style.> If so then I would prefer (as an exception!) a first separate patch > to re-indent the existing code in hw/xen_console.c: that is, a patch > consisting _only_ of whitespace changes. And then a separate patch > containing no whitespace changes. > > As it is I haven''t been able to review this very deeply and I don''t > have time today to redo the diff myself with -b ...There is a "diff -b" version if the patch series at http://kraxel.fedorapeople.org/patches/qemu-xen/no-ws/ Is that good enougth for review?> Re 0006-xen-sync-xen_machine_pv-with-upstream.patch: > > * What is the purpose of this patch ? `sync xen_machine_pv with > upstream'' ? Which upstream ? Many of these changes seem > cosmetic.Patch 0002 + 0006 for qemu-xen is the same as Patch 0002 for qemu-upstream. It has been splitted in the qemu-xen series for better bisectability: 0002 for qemu-xen does the minimal needed changes only, and 0006 brings xen_machine_pv in sync with the upstream patch. The code has been restructed a bit with the upcoming emulation bits in mind. There should be no functional change for qemu-dm.> Re 0007-xen-add-block-device-backend-driver.patch > and 0008-xen-add-net-backend-driver.patch: > > * I don''t have any objection to these in principle. However they are > of no use or relevance to qemu-xen-unstable because we do not and > will not use them. So they should go into upstream directly rather > than via me I think ?Whatever you prefer.> However that needs to wait for the generic > backend, which definitely should go into qemu-xen-unstable before > it goes into qemu upstream.Sure.> Re 0009-xen-blk-nic-configuration-via-cmd-line.patch: > > * I don''t think I fully understand the purpose of this patch. It > appears to be part of the Xenner emulation AFAICT ? When running > under the real Xen toolstack we do not have qemu write these kind > of entries to xenstore.This is for the use cases where xend doesn''t create these entries in xenstore. Yes, one user is Xenner. There also is the patch to (optionally) makes qemu create the domain, which needs this too.> Instead, xenstore.c (in qemu-xen-unstable) reads the necessary > information from xenstore and sets up the qemu internals > appropriately.The backends within qemu get their config information from xenstore. Whenever the xenstore entries are created by qemu or by xend doesn''t matter, they can deal with both cases.> And one final point: > > * Have you been testing your tree on a real Xen host under xend ?I did a while back, when I did most of the coding work for these patches. I didn''t retest the most recent version though.>> The naming convention I''m using right now is prefix ''xen_'' for anything >> usable when running on the Xen hypervisor. That includes code which >> will be shared by xen and xenner such as the machine type. That also >> includes the block and net backends. They *do* work on Xen, and with >> some glue in xend you should be able to use them. > > Before we accept any incompatible changes into qemu-xen-unstable we > obviously need the corresponding changes to xend too ?There shouldn''t be any incompatible changes (see the command line argument discussion above).> But we are not going to be using the block and net backends in qemu. > In a real Xen system these are provided by the dom0 kernel, which is > a far superior arrangement (in our opinion!)Well, the disk backend within qemu isn''t fundamentally different from blktap. Right now it doesn''t do aio due to some limitations in qemu. But the qemu block layer gets some love to improve the performance right now, so once this is done and aio is enabled in the disk backend I''d expect it to match blktap performance-wise. The net backend proably is much less useful when you can use the dom0 kernel netback driver instead, although might be use cases the qemu network support can handle and the kernel netback driver can''t (-net socket comes to mind).> AIUI from your patch the xen_mode specifies whether we''re doing Xenner > or real Xen ? Am I right in thinking that the modes are these: > > * XEN_EMULATE - Xenner (no Xen hypervisor, no xend) > IMO this should be the default for the qemu xen machine types > in upstream qemu.Yes.> * XEN_ATTACH - for qemu-dm as invoked by xend > IMO arrangements should be made so that this is done only with > a special command-line option, which is provided by xend.Yes.> * XEN_CREATE - Xen hypervisor but no xendYes.> However, > how do you control whether the block/net backends used are those > in the dom0 kernel or those in qemu ?You can''t right now, that it is still on my TODO list. Making XEN_CREATE work needs more patches too, so this isn''t a problem yet ;) My plan is to have an additional option for -drive, might look this way: -drive if=xen,backend=vbd,file=... # blkback backend -drive if=xen,backend=tap,file=... # blktap backend -drive if=xen,backend=qdisk,file=... # qemu backend Likewise for -net. cheers, Gerd _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel