[PATCH v2 0/5] Build libxc on rump kernel

[PATCH v2 0/5] Build libxc on rump kernel

[Wei Liu]

Wei Liu (5):
  NetBSDRump: provide evtchn.h
  libxc: split off xc_minios_common.c
  libxc: split off xc_netbsd_evtchn.c
  libxc: minios: introduce abstraction for files[]
  libxc: rumpxen: provide xc_osdep_info

 tools/include/xen-sys/NetBSDRump/evtchn.h |  86 ++++++++
 tools/libxc/Makefile                      |   6 +-
 tools/libxc/xc_minios.c                   | 243 +---------------------
 tools/libxc/xc_minios_common.c            | 322 ++++++++++++++++++++++++++++++
 tools/libxc/xc_netbsd.c                   | 168 +---------------
 tools/libxc/xc_netbsd_evtchn.c            | 196 ++++++++++++++++++
 tools/libxc/xc_netbsd_rumpkern.c          |  59 ++++++
 tools/libxc/xc_private.h                  |   3 +
 8 files changed, 675 insertions(+), 408 deletions(-)
 create mode 100644 tools/include/xen-sys/NetBSDRump/evtchn.h
 create mode 100644 tools/libxc/xc_minios_common.c
 create mode 100644 tools/libxc/xc_netbsd_evtchn.c
 create mode 100644 tools/libxc/xc_netbsd_rumpkern.c

-- 
1.9.1

[PATCH v2 1/5] NetBSDRump: provide evtchn.h

[Wei Liu]

Xen's build system has a target for rump kernel called NetBSDRump. We
want to build libxc against rump kernel, so we need to copy NetBSD's
evtchn.h to NetBSDRump. This copy is not very likely to diverge from
NetBSD's copy, but we don't preclude such possibility.

The actual implementation of evtch code will be factored out from
libxc's NetBSD osdep code.

Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
---
Changes in v2:
1. Better commit message.
---
 tools/include/xen-sys/NetBSDRump/evtchn.h | 86 +++++++++++++++++++++++++++++++
 1 file changed, 86 insertions(+)
 create mode 100644 tools/include/xen-sys/NetBSDRump/evtchn.h

diff --git a/tools/include/xen-sys/NetBSDRump/evtchn.h b/tools/include/xen-sys/NetBSDRump/evtchn.h
new file mode 100644
index 0000000..2d8a1f9
--- /dev/null
+++ b/tools/include/xen-sys/NetBSDRump/evtchn.h
@@ -0,0 +1,86 @@
+/* $NetBSD: evtchn.h,v 1.1.1.1 2007/06/14 19:39:45 bouyer Exp $ */
+/******************************************************************************
+ * evtchn.h
+ * 
+ * Interface to /dev/xen/evtchn.
+ * 
+ * Copyright (c) 2003-2005, K A Fraser
+ * 
+ * This file may be distributed separately from the Linux kernel, or
+ * incorporated into other software packages, subject to the following license:
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this source file (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.
+ */
+
+#ifndef __NetBSD_EVTCHN_H__
+#define __NetBSD_EVTCHN_H__
+
+/*
+ * Bind a fresh port to VIRQ @virq.
+ */
+#define IOCTL_EVTCHN_BIND_VIRQ				\
+	_IOWR('E', 4, struct ioctl_evtchn_bind_virq)
+struct ioctl_evtchn_bind_virq {
+	unsigned int virq;
+	unsigned int port;
+};
+
+/*
+ * Bind a fresh port to remote <@remote_domain, @remote_port>.
+ */
+#define IOCTL_EVTCHN_BIND_INTERDOMAIN			\
+	_IOWR('E', 5, struct ioctl_evtchn_bind_interdomain)
+struct ioctl_evtchn_bind_interdomain {
+	unsigned int remote_domain, remote_port;
+	unsigned int port;
+};
+
+/*
+ * Allocate a fresh port for binding to @remote_domain.
+ */
+#define IOCTL_EVTCHN_BIND_UNBOUND_PORT			\
+	_IOWR('E', 6, struct ioctl_evtchn_bind_unbound_port)
+struct ioctl_evtchn_bind_unbound_port {
+	unsigned int remote_domain;
+	unsigned int port;
+};
+
+/*
+ * Unbind previously allocated @port.
+ */
+#define IOCTL_EVTCHN_UNBIND				\
+	_IOW('E', 7, struct ioctl_evtchn_unbind)
+struct ioctl_evtchn_unbind {
+	unsigned int port;
+};
+
+/*
+ * Send event to previously allocated @port.
+ */
+#define IOCTL_EVTCHN_NOTIFY				\
+	_IOW('E', 8, struct ioctl_evtchn_notify)
+struct ioctl_evtchn_notify {
+	unsigned int port;
+};
+
+/* Clear and reinitialise the event buffer. Clear error condition. */
+#define IOCTL_EVTCHN_RESET				\
+	_IO('E', 9)
+
+#endif /* __NetBSD_EVTCHN_H__ */
-- 
1.9.1

[PATCH v2 2/5] libxc: split off xc_minios_common.c

[Wei Liu]

Rump kernel is going to use some but not all of the machinery previously
in xc_minios.c.  Split the privcmd and gnttab code into its own file.
This part is pure code motion.

But we also have to:

 - Alter the Makefile to build and link xc_minios_common.c too.

 - Rename some of the minios_*_ops symbols to have proper namespaceing
   and make them have external linkage, so that the init code (which
   remains in xc_minios.c) can reference them.

 - Call these *_ops symbols xc_*_ops so that we can mix and match in
   the future.  This does not impede the existing mechanisms for
   run-time overriding.  (But leave a comment next to the new
   declarations in xc_private.h saying not to use these.)

 - Change map_frames_ex to minios_map_frames_ex if compiling on rump
   kernel.

Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
---
Changes in v2:
1. xc_minios_privcmd.c -> xc_minios_common.c.
---
 tools/libxc/Makefile           |   2 +-
 tools/libxc/xc_minios.c        | 243 +---------------------------------
 tools/libxc/xc_minios_common.c | 291 +++++++++++++++++++++++++++++++++++++++++
 tools/libxc/xc_private.h       |   3 +
 4 files changed, 299 insertions(+), 240 deletions(-)
 create mode 100644 tools/libxc/xc_minios_common.c

diff --git a/tools/libxc/Makefile b/tools/libxc/Makefile
index 6fa88c7..3a0ff22 100644
--- a/tools/libxc/Makefile
+++ b/tools/libxc/Makefile
@@ -47,7 +47,7 @@ CTRL_SRCS-$(CONFIG_Linux) += xc_linux.c xc_linux_osdep.c
 CTRL_SRCS-$(CONFIG_FreeBSD) += xc_freebsd.c xc_freebsd_osdep.c
 CTRL_SRCS-$(CONFIG_SunOS) += xc_solaris.c
 CTRL_SRCS-$(CONFIG_NetBSD) += xc_netbsd.c
-CTRL_SRCS-$(CONFIG_MiniOS) += xc_minios.c
+CTRL_SRCS-$(CONFIG_MiniOS) += xc_minios.c xc_minios_common.c
 
 GUEST_SRCS-y :=
 GUEST_SRCS-y += xg_private.c xc_suspend.c
diff --git a/tools/libxc/xc_minios.c b/tools/libxc/xc_minios.c
index e703684..90e3363 100644
--- a/tools/libxc/xc_minios.c
+++ b/tools/libxc/xc_minios.c
@@ -41,164 +41,10 @@
 
 #include "xc_private.h"
 
-void minios_interface_close_fd(int fd);
 void minios_evtchn_close_fd(int fd);
-void minios_gnttab_close_fd(int fd);
-
-extern void minios_interface_close_fd(int fd);
-extern void minios_evtchn_close_fd(int fd);
 
 extern struct wait_queue_head event_queue;
 
-static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
-{
-    int fd = alloc_fd(FTYPE_XC);
-
-    if ( fd == -1)
-        return XC_OSDEP_OPEN_ERROR;
-
-    return (xc_osdep_handle)fd;
-}
-
-static int minios_privcmd_close(xc_interface *xch, xc_osdep_handle h)
-{
-    int fd = (int)h;
-    return close(fd);
-}
-
-void minios_interface_close_fd(int fd)
-{
-    files[fd].type = FTYPE_NONE;
-}
-
-static void *minios_privcmd_alloc_hypercall_buffer(xc_interface *xch, xc_osdep_handle h, int npages)
-{
-    return xc_memalign(xch, PAGE_SIZE, npages * PAGE_SIZE);
-}
-
-static void minios_privcmd_free_hypercall_buffer(xc_interface *xch, xc_osdep_handle h, void *ptr, int npages)
-{
-    free(ptr);
-}
-
-static int minios_privcmd_hypercall(xc_interface *xch, xc_osdep_handle h, privcmd_hypercall_t *hypercall)
-{
-    multicall_entry_t call;
-    int i, ret;
-
-    call.op = hypercall->op;
-    for (i = 0; i < ARRAY_SIZE(hypercall->arg); i++)
-	call.args[i] = hypercall->arg[i];
-
-    ret = HYPERVISOR_multicall(&call, 1);
-
-    if (ret < 0) {
-	errno = -ret;
-	return -1;
-    }
-    if ((long) call.result < 0) {
-        errno = - (long) call.result;
-        return -1;
-    }
-    return call.result;
-}
-
-static void *minios_privcmd_map_foreign_bulk(xc_interface *xch, xc_osdep_handle h,
-                                             uint32_t dom, int prot,
-                                             const xen_pfn_t *arr, int *err, unsigned int num)
-{
-    unsigned long pt_prot = 0;
-    if (prot & PROT_READ)
-	pt_prot = L1_PROT_RO;
-    if (prot & PROT_WRITE)
-	pt_prot = L1_PROT;
-    return map_frames_ex(arr, num, 1, 0, 1, dom, err, pt_prot);    
-}
-
-static void *minios_privcmd_map_foreign_batch(xc_interface *xch,  xc_osdep_handle h,
-                                              uint32_t dom, int prot,
-                                              xen_pfn_t *arr, int num)
-{
-    unsigned long pt_prot = 0;
-    int err[num];
-    int i;
-    unsigned long addr;
-
-    if (prot & PROT_READ)
-	pt_prot = L1_PROT_RO;
-    if (prot & PROT_WRITE)
-	pt_prot = L1_PROT;
-
-    addr = (unsigned long) map_frames_ex(arr, num, 1, 0, 1, dom, err, pt_prot);
-    for (i = 0; i < num; i++) {
-        if (err[i])
-            arr[i] |= 0xF0000000;
-    }
-    return (void *) addr;
-}
-
-static void *minios_privcmd_map_foreign_range(xc_interface *xch, xc_osdep_handle h,
-                                              uint32_t dom,
-                                              int size, int prot,
-                                              unsigned long mfn)
-{
-    unsigned long pt_prot = 0;
-
-    if (prot & PROT_READ)
-	pt_prot = L1_PROT_RO;
-    if (prot & PROT_WRITE)
-	pt_prot = L1_PROT;
-
-    assert(!(size % getpagesize()));
-    return map_frames_ex(&mfn, size / getpagesize(), 0, 1, 1, dom, NULL, pt_prot);
-}
-
-static void *minios_privcmd_map_foreign_ranges(xc_interface *xch, xc_osdep_handle h,
-                                               uint32_t dom,
-                                               size_t size, int prot, size_t chunksize,
-                                               privcmd_mmap_entry_t entries[], int nentries)
-{
-    unsigned long *mfns;
-    int i, j, n;
-    unsigned long pt_prot = 0;
-    void *ret;
-
-    if (prot & PROT_READ)
-	pt_prot = L1_PROT_RO;
-    if (prot & PROT_WRITE)
-	pt_prot = L1_PROT;
-
-    mfns = malloc((size / XC_PAGE_SIZE) * sizeof(*mfns));
-
-    n = 0;
-    for (i = 0; i < nentries; i++)
-        for (j = 0; j < chunksize / XC_PAGE_SIZE; j++)
-            mfns[n++] = entries[i].mfn + j;
-
-    ret = map_frames_ex(mfns, n, 1, 0, 1, dom, NULL, pt_prot);
-    free(mfns);
-    return ret;
-}
-
-
-static struct xc_osdep_ops minios_privcmd_ops = {
-    .open = &minios_privcmd_open,
-    .close = &minios_privcmd_close,
-
-    .u.privcmd = {
-        .alloc_hypercall_buffer = &minios_privcmd_alloc_hypercall_buffer,
-        .free_hypercall_buffer = &minios_privcmd_free_hypercall_buffer,
-
-        .hypercall = &minios_privcmd_hypercall,
-
-        .map_foreign_batch = &minios_privcmd_map_foreign_batch,
-        .map_foreign_bulk = &minios_privcmd_map_foreign_bulk,
-        .map_foreign_range = &minios_privcmd_map_foreign_range,
-        .map_foreign_ranges = &minios_privcmd_map_foreign_ranges,
-    },
-};
-
-
 /* XXX Note: This is not threadsafe */
 static struct evtchn_port_info* port_alloc(int fd) {
     struct evtchn_port_info *port_info;
@@ -409,7 +255,7 @@ static int minios_evtchn_unmask(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t
     return 0;
 }
 
-static struct xc_osdep_ops minios_evtchn_ops = {
+struct xc_osdep_ops xc_evtchn_ops = {
     .open = &minios_evtchn_open,
     .close = &minios_evtchn_close,
 
@@ -437,97 +283,16 @@ void *xc_memalign(xc_interface *xch, size_t alignment, size_t size)
     return memalign(alignment, size);
 }
 
-static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
-{
-    int fd = alloc_fd(FTYPE_GNTMAP);
-    if ( fd == -1 )
-        return XC_OSDEP_OPEN_ERROR;
-    gntmap_init(&files[fd].gntmap);
-    return (xc_osdep_handle)fd;
-}
-
-static int minios_gnttab_close(xc_gnttab *xcg, xc_osdep_handle h)
-{
-    int fd = (int)h;
-    return close(fd);
-}
-
-void minios_gnttab_close_fd(int fd)
-{
-    gntmap_fini(&files[fd].gntmap);
-    files[fd].type = FTYPE_NONE;
-}
-
-static void *minios_gnttab_grant_map(xc_gnttab *xcg, xc_osdep_handle h,
-                                     uint32_t count, int flags, int prot,
-                                     uint32_t *domids, uint32_t *refs,
-                                     uint32_t notify_offset,
-                                     evtchn_port_t notify_port)
-{
-    int fd = (int)h;
-    int stride = 1;
-    if (flags & XC_GRANT_MAP_SINGLE_DOMAIN)
-        stride = 0;
-    if (notify_offset != -1 || notify_port != -1) {
-        errno = ENOSYS;
-        return NULL;
-    }
-    return gntmap_map_grant_refs(&files[fd].gntmap,
-                                 count, domids, stride,
-                                 refs, prot & PROT_WRITE);
-}
-
-static int minios_gnttab_munmap(xc_gnttab *xcg, xc_osdep_handle h,
-                                void *start_address,
-                                uint32_t count)
-{
-    int fd = (int)h;
-    int ret;
-    ret = gntmap_munmap(&files[fd].gntmap,
-                        (unsigned long) start_address,
-                        count);
-    if (ret < 0) {
-        errno = -ret;
-        return -1;
-    }
-    return ret;
-}
-
-static int minios_gnttab_set_max_grants(xc_gnttab *xcg, xc_osdep_handle h,
-                             uint32_t count)
-{
-    int fd = (int)h;
-    int ret;
-    ret = gntmap_set_max_grants(&files[fd].gntmap,
-                                count);
-    if (ret < 0) {
-        errno = -ret;
-        return -1;
-    }
-    return ret;
-}
-
-static struct xc_osdep_ops minios_gnttab_ops = {
-    .open = &minios_gnttab_open,
-    .close = &minios_gnttab_close,
-
-    .u.gnttab = {
-        .grant_map = &minios_gnttab_grant_map,
-        .munmap = &minios_gnttab_munmap,
-        .set_max_grants = &minios_gnttab_set_max_grants,
-    },
-};
-
 static struct xc_osdep_ops *minios_osdep_init(xc_interface *xch, enum xc_osdep_type type)
 {
     switch ( type )
     {
     case XC_OSDEP_PRIVCMD:
-        return &minios_privcmd_ops;
+        return &xc_privcmd_ops;
     case XC_OSDEP_EVTCHN:
-        return &minios_evtchn_ops;
+        return &xc_evtchn_ops;
     case XC_OSDEP_GNTTAB:
-        return &minios_gnttab_ops;
+        return &xc_gnttab_ops;
     default:
         return NULL;
     }
diff --git a/tools/libxc/xc_minios_common.c b/tools/libxc/xc_minios_common.c
new file mode 100644
index 0000000..8b08cde
--- /dev/null
+++ b/tools/libxc/xc_minios_common.c
@@ -0,0 +1,291 @@
+/******************************************************************************
+ *
+ * Copyright 2007-2008 Samuel Thibault <samuel.thibault@eu.citrix.com>.
+ * All rights reserved.
+ * Use is subject to license terms.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation;
+ * version 2.1 of the License.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+#undef NDEBUG
+#include "xen-external/bsd-sys-queue.h"
+#include <mini-os/types.h>
+#include <mini-os/os.h>
+#include <mini-os/mm.h>
+#include <mini-os/lib.h>
+#include <mini-os/gntmap.h>
+#include <mini-os/events.h>
+#include <mini-os/wait.h>
+#include <sys/mman.h>
+
+#include <xen/memory.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <assert.h>
+#include <stdint.h>
+#include <inttypes.h>
+#include <malloc.h>
+
+#include "xc_private.h"
+
+#ifdef __RUMPRUN__
+# define map_frames_ex minios_map_frames_ex
+#endif /* __RUMPRUN__ */
+
+void minios_interface_close_fd(int fd);
+void minios_gnttab_close_fd(int fd);
+
+static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
+{
+    int fd = alloc_fd(FTYPE_XC);
+
+    if ( fd == -1)
+        return XC_OSDEP_OPEN_ERROR;
+
+    return (xc_osdep_handle)fd;
+}
+
+static int minios_privcmd_close(xc_interface *xch, xc_osdep_handle h)
+{
+    int fd = (int)h;
+    return close(fd);
+}
+
+void minios_interface_close_fd(int fd)
+{
+    files[fd].type = FTYPE_NONE;
+}
+
+static void *minios_privcmd_alloc_hypercall_buffer(xc_interface *xch,
+						   xc_osdep_handle h,
+						   int npages)
+{
+    return xc_memalign(xch, PAGE_SIZE, npages * PAGE_SIZE);
+}
+
+static void minios_privcmd_free_hypercall_buffer(xc_interface *xch,
+						 xc_osdep_handle h,
+						 void *ptr, int npages)
+{
+    free(ptr);
+}
+
+static int minios_privcmd_hypercall(xc_interface *xch, xc_osdep_handle h,
+				    privcmd_hypercall_t *hypercall)
+{
+    multicall_entry_t call;
+    int i, ret;
+
+    call.op = hypercall->op;
+    for (i = 0; i < ARRAY_SIZE(hypercall->arg); i++)
+	call.args[i] = hypercall->arg[i];
+
+    ret = HYPERVISOR_multicall(&call, 1);
+
+    if (ret < 0) {
+	errno = -ret;
+	return -1;
+    }
+    if ((long) call.result < 0) {
+        errno = - (long) call.result;
+        return -1;
+    }
+    return call.result;
+}
+
+static void *minios_privcmd_map_foreign_bulk(xc_interface *xch,
+					     xc_osdep_handle h,
+                                             uint32_t dom, int prot,
+                                             const xen_pfn_t *arr,
+					     int *err, unsigned int num)
+{
+    unsigned long pt_prot = 0;
+    if (prot & PROT_READ)
+	pt_prot = L1_PROT_RO;
+    if (prot & PROT_WRITE)
+	pt_prot = L1_PROT;
+    return map_frames_ex(arr, num, 1, 0, 1, dom, err, pt_prot);    
+}
+
+static void *minios_privcmd_map_foreign_batch(xc_interface *xch,
+					      xc_osdep_handle h,
+                                              uint32_t dom, int prot,
+                                              xen_pfn_t *arr, int num)
+{
+    unsigned long pt_prot = 0;
+    int err[num];
+    int i;
+    unsigned long addr;
+
+    if (prot & PROT_READ)
+	pt_prot = L1_PROT_RO;
+    if (prot & PROT_WRITE)
+	pt_prot = L1_PROT;
+
+    addr = (unsigned long) map_frames_ex(arr, num, 1, 0, 1, dom, err, pt_prot);
+    for (i = 0; i < num; i++) {
+        if (err[i])
+            arr[i] |= 0xF0000000;
+    }
+    return (void *) addr;
+}
+
+static void *minios_privcmd_map_foreign_range(xc_interface *xch,
+					      xc_osdep_handle h,
+                                              uint32_t dom,
+                                              int size, int prot,
+                                              unsigned long mfn)
+{
+    unsigned long pt_prot = 0;
+
+    if (prot & PROT_READ)
+	pt_prot = L1_PROT_RO;
+    if (prot & PROT_WRITE)
+	pt_prot = L1_PROT;
+
+    assert(!(size % getpagesize()));
+    return map_frames_ex(&mfn, size / getpagesize(), 0, 1, 1, dom,
+			 NULL, pt_prot);
+}
+
+static void *minios_privcmd_map_foreign_ranges(xc_interface *xch,
+					       xc_osdep_handle h,
+                                               uint32_t dom,
+                                               size_t size, int prot,
+					       size_t chunksize,
+                                               privcmd_mmap_entry_t entries[],
+					       int nentries)
+{
+    unsigned long *mfns;
+    int i, j, n;
+    unsigned long pt_prot = 0;
+    void *ret;
+
+    if (prot & PROT_READ)
+	pt_prot = L1_PROT_RO;
+    if (prot & PROT_WRITE)
+	pt_prot = L1_PROT;
+
+    mfns = malloc((size / XC_PAGE_SIZE) * sizeof(*mfns));
+
+    n = 0;
+    for (i = 0; i < nentries; i++)
+        for (j = 0; j < chunksize / XC_PAGE_SIZE; j++)
+            mfns[n++] = entries[i].mfn + j;
+
+    ret = map_frames_ex(mfns, n, 1, 0, 1, dom, NULL, pt_prot);
+    free(mfns);
+    return ret;
+}
+
+struct xc_osdep_ops xc_privcmd_ops = {
+    .open = &minios_privcmd_open,
+    .close = &minios_privcmd_close,
+
+    .u.privcmd = {
+        .alloc_hypercall_buffer = &minios_privcmd_alloc_hypercall_buffer,
+        .free_hypercall_buffer = &minios_privcmd_free_hypercall_buffer,
+
+        .hypercall = &minios_privcmd_hypercall,
+
+        .map_foreign_batch = &minios_privcmd_map_foreign_batch,
+        .map_foreign_bulk = &minios_privcmd_map_foreign_bulk,
+        .map_foreign_range = &minios_privcmd_map_foreign_range,
+        .map_foreign_ranges = &minios_privcmd_map_foreign_ranges,
+    },
+};
+
+static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
+{
+    int fd = alloc_fd(FTYPE_GNTMAP);
+    if ( fd == -1 )
+        return XC_OSDEP_OPEN_ERROR;
+    gntmap_init(&files[fd].gntmap);
+    return (xc_osdep_handle)fd;
+}
+
+static int minios_gnttab_close(xc_gnttab *xcg, xc_osdep_handle h)
+{
+    int fd = (int)h;
+    return close(fd);
+}
+
+void minios_gnttab_close_fd(int fd)
+{
+    gntmap_fini(&files[fd].gntmap);
+    files[fd].type = FTYPE_NONE;
+}
+
+static void *minios_gnttab_grant_map(xc_gnttab *xcg, xc_osdep_handle h,
+                                     uint32_t count, int flags, int prot,
+                                     uint32_t *domids, uint32_t *refs,
+                                     uint32_t notify_offset,
+                                     evtchn_port_t notify_port)
+{
+    int fd = (int)h;
+    int stride = 1;
+    if (flags & XC_GRANT_MAP_SINGLE_DOMAIN)
+        stride = 0;
+    if (notify_offset != -1 || notify_port != -1) {
+        errno = ENOSYS;
+        return NULL;
+    }
+    return gntmap_map_grant_refs(&files[fd].gntmap,
+                                 count, domids, stride,
+                                 refs, prot & PROT_WRITE);
+}
+
+static int minios_gnttab_munmap(xc_gnttab *xcg, xc_osdep_handle h,
+                                void *start_address,
+                                uint32_t count)
+{
+    int fd = (int)h;
+    int ret;
+    ret = gntmap_munmap(&files[fd].gntmap,
+                        (unsigned long) start_address,
+                        count);
+    if (ret < 0) {
+        errno = -ret;
+        return -1;
+    }
+    return ret;
+}
+
+static int minios_gnttab_set_max_grants(xc_gnttab *xcg, xc_osdep_handle h,
+                             uint32_t count)
+{
+    int fd = (int)h;
+    int ret;
+    ret = gntmap_set_max_grants(&files[fd].gntmap,
+                                count);
+    if (ret < 0) {
+        errno = -ret;
+        return -1;
+    }
+    return ret;
+}
+
+struct xc_osdep_ops xc_gnttab_ops = {
+    .open = &minios_gnttab_open,
+    .close = &minios_gnttab_close,
+
+    .u.gnttab = {
+        .grant_map = &minios_gnttab_grant_map,
+        .munmap = &minios_gnttab_munmap,
+        .set_max_grants = &minios_gnttab_set_max_grants,
+    },
+};
+
diff --git a/tools/libxc/xc_private.h b/tools/libxc/xc_private.h
index 45b8644..152465f 100644
--- a/tools/libxc/xc_private.h
+++ b/tools/libxc/xc_private.h
@@ -111,6 +111,9 @@ struct xc_interface_core {
     xc_osdep_handle  ops_handle; /* opaque data for xc_osdep_ops */
 };
 
+/* Do not use these directly; go via the handle you already have. */
+extern struct xc_osdep_ops xc_privcmd_ops, xc_evtchn_ops, xc_gnttab_ops;
+
 void xc_report_error(xc_interface *xch, int code, const char *fmt, ...)
     __attribute__((format(printf,3,4)));
 void xc_reportv(xc_interface *xch, xentoollog_logger *lg, xentoollog_level,
-- 
1.9.1

[PATCH v2 3/5] libxc: split off xc_netbsd_evtchn.c

[Wei Liu]

Rump kernel is going to use some but not all of the machinery previously
in xc_netbsd.c. Split the evtchn and ancillary code into its own file.
This part is pure code motion.

But we also have to alter the Makefile, and rename some symbols, as
with xc_minios*.c.

Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
---
Changes in v2:
1. xc_netbsd_user.c -> xc_netbsd_evtchn.c.
---
 tools/libxc/Makefile           |   2 +-
 tools/libxc/xc_netbsd.c        | 168 +----------------------------------
 tools/libxc/xc_netbsd_evtchn.c | 196 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 198 insertions(+), 168 deletions(-)
 create mode 100644 tools/libxc/xc_netbsd_evtchn.c

diff --git a/tools/libxc/Makefile b/tools/libxc/Makefile
index 3a0ff22..8bf7344 100644
--- a/tools/libxc/Makefile
+++ b/tools/libxc/Makefile
@@ -46,7 +46,7 @@ CTRL_SRCS-$(CONFIG_X86) += xc_pagetab.c
 CTRL_SRCS-$(CONFIG_Linux) += xc_linux.c xc_linux_osdep.c
 CTRL_SRCS-$(CONFIG_FreeBSD) += xc_freebsd.c xc_freebsd_osdep.c
 CTRL_SRCS-$(CONFIG_SunOS) += xc_solaris.c
-CTRL_SRCS-$(CONFIG_NetBSD) += xc_netbsd.c
+CTRL_SRCS-$(CONFIG_NetBSD) += xc_netbsd.c xc_netbsd_evtchn.c
 CTRL_SRCS-$(CONFIG_MiniOS) += xc_minios.c xc_minios_common.c
 
 GUEST_SRCS-y :=
diff --git a/tools/libxc/xc_netbsd.c b/tools/libxc/xc_netbsd.c
index 8a90ef3..f940607 100644
--- a/tools/libxc/xc_netbsd.c
+++ b/tools/libxc/xc_netbsd.c
@@ -224,172 +224,6 @@ static struct xc_osdep_ops netbsd_privcmd_ops = {
     },
 };
 
-#define EVTCHN_DEV_NAME  "/dev/xenevt"
-
-static xc_osdep_handle netbsd_evtchn_open(xc_evtchn *xce)
-{
-    int fd = open(EVTCHN_DEV_NAME, O_NONBLOCK|O_RDWR);
-    if ( fd == -1 )
-        return XC_OSDEP_OPEN_ERROR;
-
-    return (xc_osdep_handle)fd;
-}
-
-static int netbsd_evtchn_close(xc_evtchn *xce, xc_osdep_handle h)
-{
-    int fd = (int)h;
-    return close(fd);
-}
-
-static int netbsd_evtchn_fd(xc_evtchn *xce, xc_osdep_handle h)
-{
-    return (int)h;
-}
-
-static int netbsd_evtchn_notify(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t port)
-{
-    int fd = (int)h;
-    struct ioctl_evtchn_notify notify;
-
-    notify.port = port;
-
-    return ioctl(fd, IOCTL_EVTCHN_NOTIFY, &notify);
-}
-
-static evtchn_port_or_error_t
-netbsd_evtchn_bind_unbound_port(xc_evtchn * xce, xc_osdep_handle h, int domid)
-{
-    int fd = (int)h;
-    struct ioctl_evtchn_bind_unbound_port bind;
-    int ret;
-
-    bind.remote_domain = domid;
-
-    ret = ioctl(fd, IOCTL_EVTCHN_BIND_UNBOUND_PORT, &bind);
-    if (ret == 0)
-	return bind.port;
-    else
-	return -1;
-}
-
-static evtchn_port_or_error_t
-netbsd_evtchn_bind_interdomain(xc_evtchn *xce, xc_osdep_handle h, int domid,
-                               evtchn_port_t remote_port)
-{
-    int fd = (int)h;
-    struct ioctl_evtchn_bind_interdomain bind;
-    int ret;
-
-    bind.remote_domain = domid;
-    bind.remote_port = remote_port;
-
-    ret = ioctl(fd, IOCTL_EVTCHN_BIND_INTERDOMAIN, &bind);
-    if (ret == 0)
-	return bind.port;
-    else
-	return -1;
-}
-
-static int netbsd_evtchn_unbind(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t port)
-{
-    int fd = (int)h;
-    struct ioctl_evtchn_unbind unbind;
-
-    unbind.port = port;
-
-    return ioctl(fd, IOCTL_EVTCHN_UNBIND, &unbind);
-}
-
-static evtchn_port_or_error_t
-netbsd_evtchn_bind_virq(xc_evtchn *xce, xc_osdep_handle h, unsigned int virq)
-{
-    int fd = (int)h;
-    struct ioctl_evtchn_bind_virq bind;
-    int err;
-
-    bind.virq = virq;
-
-    err = ioctl(fd, IOCTL_EVTCHN_BIND_VIRQ, &bind);
-    if (err)
-	return -1;
-    else
-	return bind.port;
-}
-
-static evtchn_port_or_error_t
-netbsd_evtchn_pending(xc_evtchn *xce, xc_osdep_handle h)
-{
-    int fd = (int)h;
-    evtchn_port_t port;
-
-    if ( read_exact(fd, (char *)&port, sizeof(port)) == -1 )
-        return -1;
-
-    return port;
-}
-
-static int netbsd_evtchn_unmask(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t port)
-{
-    int fd = (int)h;
-    return write_exact(fd, (char *)&port, sizeof(port));
-}
-
-static struct xc_osdep_ops netbsd_evtchn_ops = {
-    .open = &netbsd_evtchn_open,
-    .close = &netbsd_evtchn_close,
-
-    .u.evtchn = {
-         .fd = &netbsd_evtchn_fd,
-         .notify = &netbsd_evtchn_notify,
-         .bind_unbound_port = &netbsd_evtchn_bind_unbound_port,
-         .bind_interdomain = &netbsd_evtchn_bind_interdomain,
-         .bind_virq = &netbsd_evtchn_bind_virq,
-         .unbind = &netbsd_evtchn_unbind,
-         .pending = &netbsd_evtchn_pending,
-         .unmask = &netbsd_evtchn_unmask,
-    },
-};
-
-/* Optionally flush file to disk and discard page cache */
-void discard_file_cache(xc_interface *xch, int fd, int flush) 
-{
-    off_t cur = 0;
-    int saved_errno = errno;
-
-    if ( flush && (fsync(fd) < 0) )
-    {
-        /*PERROR("Failed to flush file: %s", strerror(errno));*/
-        goto out;
-    }
-
-    /*
-     * Calculate last page boundry of amount written so far
-     * unless we are flushing in which case entire cache
-     * is discarded.
-     */
-    if ( !flush )
-    {
-        if ( ( cur = lseek(fd, 0, SEEK_CUR)) == (off_t)-1 )
-            cur = 0;
-        cur &= ~(PAGE_SIZE - 1);
-    }
-
-    /* Discard from the buffer cache. */
-    if ( posix_fadvise(fd, 0, cur, POSIX_FADV_DONTNEED) < 0 )
-    {
-        /*PERROR("Failed to discard cache: %s", strerror(errno));*/
-        goto out;
-    }
-
- out:
-    errno = saved_errno;
-}
-
-void *xc_memalign(xc_interface *xch, size_t alignment, size_t size)
-{
-    return valloc(size);
-}
-
 static struct xc_osdep_ops *netbsd_osdep_init(xc_interface *xch, enum xc_osdep_type type)
 {
     switch ( type )
@@ -397,7 +231,7 @@ static struct xc_osdep_ops *netbsd_osdep_init(xc_interface *xch, enum xc_osdep_t
     case XC_OSDEP_PRIVCMD:
         return &netbsd_privcmd_ops;
     case XC_OSDEP_EVTCHN:
-        return &netbsd_evtchn_ops;
+        return &xc_evtchn_ops;
     default:
         return NULL;
     }
diff --git a/tools/libxc/xc_netbsd_evtchn.c b/tools/libxc/xc_netbsd_evtchn.c
new file mode 100644
index 0000000..b5c2491
--- /dev/null
+++ b/tools/libxc/xc_netbsd_evtchn.c
@@ -0,0 +1,196 @@
+/******************************************************************************
+ *
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ *
+ * xc_gnttab functions:
+ * Copyright (c) 2007-2008, D G Murray <Derek.Murray@cl.cam.ac.uk>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation;
+ * version 2.1 of the License.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+#include "xc_private.h"
+
+#include <xen/sys/evtchn.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <malloc.h>
+#include <sys/mman.h>
+
+#define EVTCHN_DEV_NAME  "/dev/xenevt"
+
+static xc_osdep_handle netbsd_evtchn_open(xc_evtchn *xce)
+{
+    int fd = open(EVTCHN_DEV_NAME, O_NONBLOCK|O_RDWR);
+    if ( fd == -1 )
+        return XC_OSDEP_OPEN_ERROR;
+
+    return (xc_osdep_handle)fd;
+}
+
+static int netbsd_evtchn_close(xc_evtchn *xce, xc_osdep_handle h)
+{
+    int fd = (int)h;
+    return close(fd);
+}
+
+static int netbsd_evtchn_fd(xc_evtchn *xce, xc_osdep_handle h)
+{
+    return (int)h;
+}
+
+static int netbsd_evtchn_notify(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t port)
+{
+    int fd = (int)h;
+    struct ioctl_evtchn_notify notify;
+
+    notify.port = port;
+
+    return ioctl(fd, IOCTL_EVTCHN_NOTIFY, &notify);
+}
+
+static evtchn_port_or_error_t
+netbsd_evtchn_bind_unbound_port(xc_evtchn * xce, xc_osdep_handle h, int domid)
+{
+    int fd = (int)h;
+    struct ioctl_evtchn_bind_unbound_port bind;
+    int ret;
+
+    bind.remote_domain = domid;
+
+    ret = ioctl(fd, IOCTL_EVTCHN_BIND_UNBOUND_PORT, &bind);
+    if (ret == 0)
+	return bind.port;
+    else
+	return -1;
+}
+
+static evtchn_port_or_error_t
+netbsd_evtchn_bind_interdomain(xc_evtchn *xce, xc_osdep_handle h, int domid,
+                               evtchn_port_t remote_port)
+{
+    int fd = (int)h;
+    struct ioctl_evtchn_bind_interdomain bind;
+    int ret;
+
+    bind.remote_domain = domid;
+    bind.remote_port = remote_port;
+
+    ret = ioctl(fd, IOCTL_EVTCHN_BIND_INTERDOMAIN, &bind);
+    if (ret == 0)
+	return bind.port;
+    else
+	return -1;
+}
+
+static int netbsd_evtchn_unbind(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t port)
+{
+    int fd = (int)h;
+    struct ioctl_evtchn_unbind unbind;
+
+    unbind.port = port;
+
+    return ioctl(fd, IOCTL_EVTCHN_UNBIND, &unbind);
+}
+
+static evtchn_port_or_error_t
+netbsd_evtchn_bind_virq(xc_evtchn *xce, xc_osdep_handle h, unsigned int virq)
+{
+    int fd = (int)h;
+    struct ioctl_evtchn_bind_virq bind;
+    int err;
+
+    bind.virq = virq;
+
+    err = ioctl(fd, IOCTL_EVTCHN_BIND_VIRQ, &bind);
+    if (err)
+	return -1;
+    else
+	return bind.port;
+}
+
+static evtchn_port_or_error_t
+netbsd_evtchn_pending(xc_evtchn *xce, xc_osdep_handle h)
+{
+    int fd = (int)h;
+    evtchn_port_t port;
+
+    if ( read_exact(fd, (char *)&port, sizeof(port)) == -1 )
+        return -1;
+
+    return port;
+}
+
+static int netbsd_evtchn_unmask(xc_evtchn *xce, xc_osdep_handle h, evtchn_port_t port)
+{
+    int fd = (int)h;
+    return write_exact(fd, (char *)&port, sizeof(port));
+}
+
+struct xc_osdep_ops xc_evtchn_ops = {
+    .open = &netbsd_evtchn_open,
+    .close = &netbsd_evtchn_close,
+
+    .u.evtchn = {
+         .fd = &netbsd_evtchn_fd,
+         .notify = &netbsd_evtchn_notify,
+         .bind_unbound_port = &netbsd_evtchn_bind_unbound_port,
+         .bind_interdomain = &netbsd_evtchn_bind_interdomain,
+         .bind_virq = &netbsd_evtchn_bind_virq,
+         .unbind = &netbsd_evtchn_unbind,
+         .pending = &netbsd_evtchn_pending,
+         .unmask = &netbsd_evtchn_unmask,
+    },
+};
+
+/* Optionally flush file to disk and discard page cache */
+void discard_file_cache(xc_interface *xch, int fd, int flush) 
+{
+    off_t cur = 0;
+    int saved_errno = errno;
+
+    if ( flush && (fsync(fd) < 0) )
+    {
+        /*PERROR("Failed to flush file: %s", strerror(errno));*/
+        goto out;
+    }
+
+    /*
+     * Calculate last page boundry of amount written so far
+     * unless we are flushing in which case entire cache
+     * is discarded.
+     */
+    if ( !flush )
+    {
+        if ( ( cur = lseek(fd, 0, SEEK_CUR)) == (off_t)-1 )
+            cur = 0;
+        cur &= ~(PAGE_SIZE - 1);
+    }
+
+    /* Discard from the buffer cache. */
+    if ( posix_fadvise(fd, 0, cur, POSIX_FADV_DONTNEED) < 0 )
+    {
+        /*PERROR("Failed to discard cache: %s", strerror(errno));*/
+        goto out;
+    }
+
+ out:
+    errno = saved_errno;
+}
+
+void *xc_memalign(xc_interface *xch, size_t alignment, size_t size)
+{
+    return valloc(size);
+}
-- 
1.9.1

[PATCH v2 4/5] libxc: minios: introduce abstraction for files[]

[Wei Liu]

We are going to want to reuse this code for NetBSD rump kernels, where
there is no gntmap device and we just want to call the MiniOS gntmap
code directly.

As part of this we want to abstract away the use of files[] inside the
actual functions.  Do this with a #define whose definition we are
going to make conditional in just a moment.

No functional change in this patch.

Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
Changes in v2:
1. Fix a bug spotted by Jürgen.
---
 tools/libxc/xc_minios_common.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/tools/libxc/xc_minios_common.c b/tools/libxc/xc_minios_common.c
index 8b08cde..c38b5f2 100644
--- a/tools/libxc/xc_minios_common.c
+++ b/tools/libxc/xc_minios_common.c
@@ -208,12 +208,14 @@ struct xc_osdep_ops xc_privcmd_ops = {
     },
 };
 
+#define GNTMAP(h) (files[(int)(h)].gntmap)
+
 static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
 {
     int fd = alloc_fd(FTYPE_GNTMAP);
     if ( fd == -1 )
         return XC_OSDEP_OPEN_ERROR;
-    gntmap_init(&files[fd].gntmap);
+    gntmap_init(&GNTMAP(fd));
     return (xc_osdep_handle)fd;
 }
 
@@ -225,7 +227,7 @@ static int minios_gnttab_close(xc_gnttab *xcg, xc_osdep_handle h)
 
 void minios_gnttab_close_fd(int fd)
 {
-    gntmap_fini(&files[fd].gntmap);
+    gntmap_fini(&GNTMAP(fd));
     files[fd].type = FTYPE_NONE;
 }
 
@@ -235,7 +237,6 @@ static void *minios_gnttab_grant_map(xc_gnttab *xcg, xc_osdep_handle h,
                                      uint32_t notify_offset,
                                      evtchn_port_t notify_port)
 {
-    int fd = (int)h;
     int stride = 1;
     if (flags & XC_GRANT_MAP_SINGLE_DOMAIN)
         stride = 0;
@@ -243,7 +244,7 @@ static void *minios_gnttab_grant_map(xc_gnttab *xcg, xc_osdep_handle h,
         errno = ENOSYS;
         return NULL;
     }
-    return gntmap_map_grant_refs(&files[fd].gntmap,
+    return gntmap_map_grant_refs(&GNTMAP(fd),
                                  count, domids, stride,
                                  refs, prot & PROT_WRITE);
 }
@@ -252,9 +253,8 @@ static int minios_gnttab_munmap(xc_gnttab *xcg, xc_osdep_handle h,
                                 void *start_address,
                                 uint32_t count)
 {
-    int fd = (int)h;
     int ret;
-    ret = gntmap_munmap(&files[fd].gntmap,
+    ret = gntmap_munmap(&GNTMAP(fd),
                         (unsigned long) start_address,
                         count);
     if (ret < 0) {
@@ -267,9 +267,8 @@ static int minios_gnttab_munmap(xc_gnttab *xcg, xc_osdep_handle h,
 static int minios_gnttab_set_max_grants(xc_gnttab *xcg, xc_osdep_handle h,
                              uint32_t count)
 {
-    int fd = (int)h;
     int ret;
-    ret = gntmap_set_max_grants(&files[fd].gntmap,
+    ret = gntmap_set_max_grants(&GNTMAP(fd),
                                 count);
     if (ret < 0) {
         errno = -ret;
-- 
1.9.1


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

[PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Wei Liu]

This allows programs which use the bulk of libxc to link.  We use
/dev/xenevt for event channels, the raw minios functions for privcmd
and gnttab, and the netbsd versions of discard_file_cache and
xc_memalign.

Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
---
 tools/libxc/Makefile             |  2 ++
 tools/libxc/xc_minios_common.c   | 34 ++++++++++++++++++++++-
 tools/libxc/xc_netbsd_rumpkern.c | 59 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 94 insertions(+), 1 deletion(-)
 create mode 100644 tools/libxc/xc_netbsd_rumpkern.c

diff --git a/tools/libxc/Makefile b/tools/libxc/Makefile
index 8bf7344..c8a5442 100644
--- a/tools/libxc/Makefile
+++ b/tools/libxc/Makefile
@@ -48,6 +48,8 @@ CTRL_SRCS-$(CONFIG_FreeBSD) += xc_freebsd.c xc_freebsd_osdep.c
 CTRL_SRCS-$(CONFIG_SunOS) += xc_solaris.c
 CTRL_SRCS-$(CONFIG_NetBSD) += xc_netbsd.c xc_netbsd_evtchn.c
 CTRL_SRCS-$(CONFIG_MiniOS) += xc_minios.c xc_minios_common.c
+CTRL_SRCS-$(CONFIG_NetBSDRump) += xc_netbsd_rumpkern.c xc_netbsd_evtchn.c
+CTRL_SRCS-$(CONFIG_NetBSDRump) += xc_minios_common.c
 
 GUEST_SRCS-y :=
 GUEST_SRCS-y += xg_private.c xc_suspend.c
diff --git a/tools/libxc/xc_minios_common.c b/tools/libxc/xc_minios_common.c
index c38b5f2..fd201fd 100644
--- a/tools/libxc/xc_minios_common.c
+++ b/tools/libxc/xc_minios_common.c
@@ -43,7 +43,17 @@
 
 #ifdef __RUMPRUN__
 # define map_frames_ex minios_map_frames_ex
-#endif /* __RUMPRUN__ */
+
+static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
+{
+    return 1;
+}
+static int minios_privcmd_close(xc_interface *xch, xc_osdep_handle h)
+{
+    return 0;
+}
+
+#else /* !__RUMPRUN__ */
 
 void minios_interface_close_fd(int fd);
 void minios_gnttab_close_fd(int fd);
@@ -69,6 +79,8 @@ void minios_interface_close_fd(int fd)
     files[fd].type = FTYPE_NONE;
 }
 
+#endif /* !__RUMPRUN__ */
+
 static void *minios_privcmd_alloc_hypercall_buffer(xc_interface *xch,
 						   xc_osdep_handle h,
 						   int npages)
@@ -208,6 +220,24 @@ struct xc_osdep_ops xc_privcmd_ops = {
     },
 };
 
+#ifdef __RUMPRUN__
+
+static struct gntmap static_gntmap;
+
+#define GNTMAP(h) static_gntmap
+
+static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
+{
+    return 1;
+}
+
+static int minios_gnttab_close(xc_gnttab *xcg, xc_osdep_handle h)
+{
+    return 0;
+}
+
+#else /* !__RUMPRUN__ */
+
 #define GNTMAP(h) (files[(int)(h)].gntmap)
 
 static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
@@ -231,6 +261,8 @@ void minios_gnttab_close_fd(int fd)
     files[fd].type = FTYPE_NONE;
 }
 
+#endif /* !__RUMPRUN__ */
+
 static void *minios_gnttab_grant_map(xc_gnttab *xcg, xc_osdep_handle h,
                                      uint32_t count, int flags, int prot,
                                      uint32_t *domids, uint32_t *refs,
diff --git a/tools/libxc/xc_netbsd_rumpkern.c b/tools/libxc/xc_netbsd_rumpkern.c
new file mode 100644
index 0000000..882374a
--- /dev/null
+++ b/tools/libxc/xc_netbsd_rumpkern.c
@@ -0,0 +1,59 @@
+/******************************************************************************
+ *
+ * Copyright 2013 Citrix.
+ * Use is subject to license terms.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation;
+ * version 2.1 of the License.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+#include "xc_private.h"
+
+#include <xen/sys/evtchn.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <malloc.h>
+#include <sys/mman.h>
+
+static struct xc_osdep_ops *rumpxen_osdep_init(xc_interface *xch,
+                                               enum xc_osdep_type type)
+{
+    switch ( type )
+    {
+    case XC_OSDEP_PRIVCMD:
+        return &xc_privcmd_ops;
+    case XC_OSDEP_EVTCHN:
+        return &xc_evtchn_ops;
+    case XC_OSDEP_GNTTAB:
+        return &xc_gnttab_ops;
+    default:
+        return NULL;
+    }
+}
+
+xc_osdep_info_t xc_osdep_info = {
+    .name = "Rump kernel OS interface",
+    .init = &rumpxen_osdep_init,
+    .fake = 0,
+};
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
-- 
1.9.1

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Samuel Thibault]

Wei Liu, le Tue 03 Mar 2015 15:57:21 +0000, a écrit :
> This allows programs which use the bulk of libxc to link.  We use
> /dev/xenevt for event channels, the raw minios functions for privcmd
> and gnttab, and the netbsd versions of discard_file_cache and
> xc_memalign.
> 
> Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
> Signed-off-by: Wei Liu <wei.liu2@citrix.com>

Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

> +static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
> +{
> +    return 1;
> +}
> +static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
> +{
> +    return 1;
> +}

Perhaps return a different fd number?

Samuel

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Wei Liu]

On Tue, Mar 03, 2015 at 05:05:29PM +0100, Samuel Thibault wrote:
> Wei Liu, le Tue 03 Mar 2015 15:57:21 +0000, a écrit :
> > This allows programs which use the bulk of libxc to link.  We use
> > /dev/xenevt for event channels, the raw minios functions for privcmd
> > and gnttab, and the netbsd versions of discard_file_cache and
> > xc_memalign.
> > 
> > Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
> > Signed-off-by: Wei Liu <wei.liu2@citrix.com>
> 
> Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
> 
> > +static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
> > +{
> > +    return 1;
> > +}
> > +static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
> > +{
> > +    return 1;
> > +}
> 
> Perhaps return a different fd number?
> 

After thinking a bit more I think we should use alloc_fd(FTYPE_*) for
all file descriptor allocation.

Wei.

> Samuel

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Wei Liu]

On Tue, Mar 03, 2015 at 04:41:23PM +0000, Wei Liu wrote:
> On Tue, Mar 03, 2015 at 05:05:29PM +0100, Samuel Thibault wrote:
> > Wei Liu, le Tue 03 Mar 2015 15:57:21 +0000, a écrit :
> > > This allows programs which use the bulk of libxc to link.  We use
> > > /dev/xenevt for event channels, the raw minios functions for privcmd
> > > and gnttab, and the netbsd versions of discard_file_cache and
> > > xc_memalign.
> > > 
> > > Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
> > > Signed-off-by: Wei Liu <wei.liu2@citrix.com>
> > 
> > Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
> > 
> > > +static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
> > > +{
> > > +    return 1;
> > > +}
> > > +static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
> > > +{
> > > +    return 1;
> > > +}
> > 
> > Perhaps return a different fd number?
> > 
> 
> After thinking a bit more I think we should use alloc_fd(FTYPE_*) for
> all file descriptor allocation.
> 

(Sorry for speaking to myself)

Just tried that, failed.

Rump kernel doesn't use mini-os's fd management facility so there is no
alloc_fd etc. On the other hand this makes me think number 1 is as good
as any other number we can use.

Wei.

> Wei.
> 
> > Samuel

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Samuel Thibault]

Wei Liu, le Tue 03 Mar 2015 16:41:23 +0000, a écrit :
> On Tue, Mar 03, 2015 at 05:05:29PM +0100, Samuel Thibault wrote:
> > Wei Liu, le Tue 03 Mar 2015 15:57:21 +0000, a écrit :
> > > This allows programs which use the bulk of libxc to link.  We use
> > > /dev/xenevt for event channels, the raw minios functions for privcmd
> > > and gnttab, and the netbsd versions of discard_file_cache and
> > > xc_memalign.
> > > 
> > > Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
> > > Signed-off-by: Wei Liu <wei.liu2@citrix.com>
> > 
> > Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
> > 
> > > +static xc_osdep_handle minios_privcmd_open(xc_interface *xch)
> > > +{
> > > +    return 1;
> > > +}
> > > +static xc_osdep_handle minios_gnttab_open(xc_gnttab *xcg)
> > > +{
> > > +    return 1;
> > > +}
> > 
> > Perhaps return a different fd number?
> > 
> 
> After thinking a bit more I think we should use alloc_fd(FTYPE_*) for
> all file descriptor allocation.

That would make sense, yes.

Samuel

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Samuel Thibault]

Wei Liu, le Tue 03 Mar 2015 16:58:07 +0000, a écrit :
> Rump kernel doesn't use mini-os's fd management facility so there is no
> alloc_fd etc. On the other hand this makes me think number 1 is as good
> as any other number we can use.

Does Rump have not fd at all?  It'd probably still be good to have
different, even if hardcoded, fd numbers, for potentially easier
debugging.

Samuel

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Ian Campbell]

On Tue, 2015-03-03 at 17:59 +0100, Samuel Thibault wrote:
> Wei Liu, le Tue 03 Mar 2015 16:58:07 +0000, a écrit :
> > Rump kernel doesn't use mini-os's fd management facility so there is no
> > alloc_fd etc. On the other hand this makes me think number 1 is as good
> > as any other number we can use.
> 
> Does Rump have not fd at all?  It'd probably still be good to have
> different, even if hardcoded, fd numbers, for potentially easier
> debugging.

In particular 1 is normally stdout, which could be confusing.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH v2 5/5] libxc: rumpxen: provide xc_osdep_info

[Wei Liu]

On Tue, Mar 03, 2015 at 05:59:39PM +0100, Samuel Thibault wrote:
> Wei Liu, le Tue 03 Mar 2015 16:58:07 +0000, a écrit :
> > Rump kernel doesn't use mini-os's fd management facility so there is no
> > alloc_fd etc. On the other hand this makes me think number 1 is as good
> > as any other number we can use.
> 
> Does Rump have not fd at all?  It'd probably still be good to have

AIUI it's using NetBSD's facility or it just doesn't care.

The mini-os fork (!) in rump kernel doesn't have alloc_fd etc.

> different, even if hardcoded, fd numbers, for potentially easier
> debugging.

Fair enough. I'm going to use ... 0xff (and probably comes with a
#define in code).

Wei.

> 
> Samuel

Re: [PATCH v2 1/5] NetBSDRump: provide evtchn.h

[Ian Campbell]

On Tue, 2015-03-03 at 15:57 +0000, Wei Liu wrote:
> Xen's build system has a target for rump kernel called NetBSDRump. We
> want to build libxc against rump kernel, so we need to copy NetBSD's
> evtchn.h to NetBSDRump. This copy is not very likely to diverge from
> NetBSD's copy, but we don't preclude such possibility.
> 
> The actual implementation of evtch code will be factored out from
> libxc's NetBSD osdep code.
> 
> Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
> Signed-off-by: Wei Liu <wei.liu2@citrix.com>

Acked-by: Ian Campbell <ian.campbell@citrix.com>

Re: [PATCH v2 2/5] libxc: split off xc_minios_common.c

[Ian Campbell]

On Tue, 2015-03-03 at 15:57 +0000, Wei Liu wrote:
> Rump kernel is going to use some but not all of the machinery previously
> in xc_minios.c.  Split the privcmd and gnttab code into its own file.
> This part is pure code motion.
> 
> But we also have to:
> 
>  - Alter the Makefile to build and link xc_minios_common.c too.
> 
>  - Rename some of the minios_*_ops symbols to have proper namespaceing
>    and make them have external linkage, so that the init code (which
>    remains in xc_minios.c) can reference them.
> 
>  - Call these *_ops symbols xc_*_ops so that we can mix and match in
>    the future.  This does not impede the existing mechanisms for
>    run-time overriding.  (But leave a comment next to the new
>    declarations in xc_private.h saying not to use these.)
> 
>  - Change map_frames_ex to minios_map_frames_ex if compiling on rump
>    kernel.
> 
> Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
> Signed-off-by: Wei Liu <wei.liu2@citrix.com>
> Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

Acked-by: Ian Campbell <ian.campbell@citrix.com>

Re: [PATCH v2 3/5] libxc: split off xc_netbsd_evtchn.c

[Ian Campbell]

On Tue, 2015-03-03 at 15:57 +0000, Wei Liu wrote:
> iff --git a/tools/libxc/xc_netbsd_evtchn.c
> b/tools/libxc/xc_netbsd_evtchn.c
> new file mode 100644
> index 0000000..b5c2491
> --- /dev/null
> +++ b/tools/libxc/xc_netbsd_evtchn.c
> @@ -0,0 +1,196 @@
> +/******************************************************************************
> + *
> + * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
> + * Use is subject to license terms.
> + *
> + * xc_gnttab functions:
> + * Copyright (c) 2007-2008, D G Murray <Derek.Murray@cl.cam.ac.uk>

Is this the same cut-and-paste-o as last time or was there a second one
I didn't spot?

Ian.

Re: [PATCH v2 3/5] libxc: split off xc_netbsd_evtchn.c

[Wei Liu]

On Thu, Mar 05, 2015 at 05:40:44PM +0000, Ian Campbell wrote:
> On Tue, 2015-03-03 at 15:57 +0000, Wei Liu wrote:
> > iff --git a/tools/libxc/xc_netbsd_evtchn.c
> > b/tools/libxc/xc_netbsd_evtchn.c
> > new file mode 100644
> > index 0000000..b5c2491
> > --- /dev/null
> > +++ b/tools/libxc/xc_netbsd_evtchn.c
> > @@ -0,0 +1,196 @@
> > +/******************************************************************************
> > + *
> > + * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
> > + * Use is subject to license terms.
> > + *
> > + * xc_gnttab functions:
> > + * Copyright (c) 2007-2008, D G Murray <Derek.Murray@cl.cam.ac.uk>
> 
> Is this the same cut-and-paste-o as last time or was there a second one
> I didn't spot?
> 

This is a different patch. I fixed the one you mentioned last time.

I will fix this as well.

Wei.

> Ian.
>