Skip to contentSkip to conversation

block: add native support for NFS

PATCH Superseded Peter Lieven 4 participants 23 messages Jan 29, 2014 – Jan 31, 2014 Block layer

v7 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v7 cover block: add native support for NFS

This adds v6 of the NFS protocol driver + qemu-iotest adjustments.

v6->v7:
- removed unsused and partially non-existing libnfs header file includes [Stefan]
- adjusted usage of qemu_opts_create_nofail [Stefan]
- removed patch to fix output of qemu-iotest 067
- only blacklist qemu-iotest 020 for NFS do not change proto from generic to file [Fam]

v5->v6:
- use internal qemu function to parse the NFS url [Kevin]
- zero pad short reads [Kevin, Paolo]
- added qemu-iotests patches for basic nfs protocol support

v4->v5:
- disussed with Ronnie and decided to move URL + Paramter parsing to LibNFS. This allows for URL parameter processing directly in LibNFS without altering the qemu NFS block driver. This bumps the version requirement for LibNFS to 1.9.0 though.
- added a pointer to the LibNFS readme where additional information about ROOT privilidge requirements can be found as this raised a few concerns.
- removed a trailing dot in an error statement [Fam].

v3->v4:
- finally added full implementation of bdrv_get_allocated_file_size [Stefan]
- removed trailing \n from error statements [Stefan]

v2->v3:
- rebased the stefanha/block
- use pkg_config to check for libnfs (ignoring cflags which are broken in 1.8.0) [Stefan]
- fixed NFSClient declaration [Stefan]
- renamed Task variables to task [Stefan]
- renamed NFSTask to NFSRPC [Ronnie]
- do not update bs->total_sectors in nfs_co_writev [Stefan]
- return -ENOMEM on all async call failures [Stefan,Ronnie]
- fully implement ftruncate
- use util/uri.c for URL parsing [Stefan]
- reworked nfs_file_open_common to nfs_client_open which works on NFSClient [Stefan]
- added a comment ot the connect message that libnfs support NFSv3 only at the moment.
- DID NOT add full implementation of bdrv_get_allocated_file_size because we are not in a coroutine context and I cannot do an async call here.
I could do a sync call if there would be a guarantee that no requests are in flight. [Stefan]

v1->v2:
- fixed block/Makefile.objs [Ronnie]
- do not always register a read handler [Ronnie]
- add support for reading beyond EOF [Fam]
- fixed struct and paramter naming [Fam]
- fixed overlong lines and whitespace errors [Fam]
- return return status from libnfs whereever possible [Fam]
- added comment why we set allocated_file_size to -ENOTSUP after write [Fam]
- avoid segfault when parsing filname [Fam]
- remove unused close_bh from NFSClient [Fam]
- avoid dividing and mutliplying total_size by BDRV_SECTOR_SIZE in nfs_file_create [Fam]

Peter Lieven (5):
block: add native support for NFS
qemu-iotests: change _supported_proto to file for various tests
qemu-iotests: enable support for NFS protocol
qemu-iotests: enable test 016 and 025 to work with NFS protocol
qemu-iotests: blacklist test 020 for NFS protocol

 MAINTAINERS                  |    5 +
 block/Makefile.objs          |    1 +
 block/nfs.c                  |  440 ++++++++++++++++++++++++++++++++++++++++++
 configure                    |   26 +++
 qapi-schema.json             |    1 +
 tests/qemu-iotests/013       |    2 +-
 tests/qemu-iotests/014       |    2 +-
 tests/qemu-iotests/016       |    2 +-
 tests/qemu-iotests/018       |    2 +-
 tests/qemu-iotests/019       |    2 +-
 tests/qemu-iotests/020       |    5 +
 tests/qemu-iotests/023       |    2 +-
 tests/qemu-iotests/024       |    2 +-
 tests/qemu-iotests/025       |    2 +-
 tests/qemu-iotests/026       |    2 +-
 tests/qemu-iotests/028       |    2 +-
 tests/qemu-iotests/031       |    2 +-
 tests/qemu-iotests/034       |    2 +-
 tests/qemu-iotests/036       |    2 +-
 tests/qemu-iotests/037       |    2 +-
 tests/qemu-iotests/038       |    2 +-
 tests/qemu-iotests/039       |    2 +-
 tests/qemu-iotests/043       |    2 +-
 tests/qemu-iotests/046       |    2 +-
 tests/qemu-iotests/052       |    2 +-
 tests/qemu-iotests/054       |    2 +-
 tests/qemu-iotests/059       |    2 +-
 tests/qemu-iotests/060       |    2 +-
 tests/qemu-iotests/061       |    2 +-
 tests/qemu-iotests/063       |    2 +-
 tests/qemu-iotests/069       |    2 +-
 tests/qemu-iotests/common    |   22 ++-
 tests/qemu-iotests/common.rc |    3 +
 33 files changed, 525 insertions(+), 28 deletions(-)
 create mode 100644 block/nfs.c
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH v7 1/5 block: add native support for NFS

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---
 MAINTAINERS         |    5 +
 block/Makefile.objs |    1 +
 block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
 configure           |   26 +++
 qapi-schema.json    |    1 +
 5 files changed, 473 insertions(+)
 create mode 100644 block/nfs.c
MAINTAINERS +5 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>
 S: Supported
 F: block/iscsi.c
 
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+
 SSH
 M: Richard W.M. Jones <[email protected]>
 S: Supported
block/Makefile.objs +1 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
 ifeq ($(CONFIG_POSIX),y)
 block-obj-y += nbd.o nbd-client.o sheepdog.o
 block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
 block-obj-$(CONFIG_CURL) += curl.o
 block-obj-$(CONFIG_RBD) += rbd.o
 block-obj-$(CONFIG_GLUSTERFS) += gluster.o
block/nfs.c new file +440 -0

440 lines not on this page · raw message

configure +26 -0
index b472694..9ca0c63 100755
@@ -251,6 +251,7 @@ vss_win32_sdk=""
 win_sdk="no"
 want_tools="yes"
 libiscsi=""
+libnfs=""
 coroutine=""
 coroutine_pool=""
 seccomp=""
@@ -840,6 +841,10 @@ for opt do
   ;;
   --enable-libiscsi) libiscsi="yes"
   ;;
+  --disable-libnfs) libnfs="no"
+  ;;
+  --enable-libnfs) libnfs="yes"
+  ;;
   --enable-profiler) profiler="yes"
   ;;
   --disable-cocoa) cocoa="no"
@@ -1229,6 +1234,8 @@ Advanced options (experts only):
   --enable-rbd             enable building the rados block device (rbd)
   --disable-libiscsi       disable iscsi support
   --enable-libiscsi        enable iscsi support
+  --disable-libnfs         disable nfs support
+  --enable-libnfs          enable nfs support
   --disable-smartcard-nss  disable smartcard nss support
   --enable-smartcard-nss   enable smartcard nss support
   --disable-libusb         disable libusb (for usb passthrough)
@@ -3600,6 +3607,20 @@ elif test "$debug" = "no" ; then
   CFLAGS="-O2 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 $CFLAGS"
 fi
 
+##########################################
+# Do we have libnfs
+if test "$libnfs" != "no" ; then
+  if $pkg_config --atleast-version=1.8.91 libnfs; then
+    libnfs="yes"
+    libnfs_libs=$($pkg_config --libs libnfs)
+    LIBS="$LIBS $libnfs_libs"
+  else
+    if test "$libnfs" = "yes" ; then
+      feature_not_found "libnfs"
+    fi
+    libnfs="no"
+  fi
+fi
 
 # Disable zero malloc errors for official releases unless explicitly told to
 # enable/disable
@@ -3829,6 +3850,7 @@ echo "libiscsi support  $libiscsi (1.4.0)"
 else
 echo "libiscsi support  $libiscsi"
 fi
+echo "libnfs support    $libnfs"
 echo "build guest agent $guest_agent"
 echo "QGA VSS support   $guest_agent_with_vss"
 echo "seccomp support   $seccomp"
@@ -4165,6 +4187,10 @@ if test "$libiscsi" = "yes" ; then
   fi
 fi
 
+if test "$libnfs" = "yes" ; then
+  echo "CONFIG_LIBNFS=y" >> $config_host_mak
+fi
+
 if test "$seccomp" = "yes"; then
   echo "CONFIG_SECCOMP=y" >> $config_host_mak
 fi
qapi-schema.json +1 -0
index 05ced9d..7cfb5e5 100644
@@ -4371,6 +4371,7 @@
 # TODO gluster: Wait for structured options
 # TODO iscsi: Wait for structured options
 # TODO nbd: Should take InetSocketAddress for 'host'?
+# TODO nfs: Wait for structured options
 # TODO rbd: Wait for structured options
 # TODO sheepdog: Wait for structured options
 # TODO ssh: Should take InetSocketAddress for 'host'?
Signature · 1 line
1.7.9.5
#2 lore raw
Quoted 170 lines · Peter Lieven

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---
 MAINTAINERS         |    5 +
 block/Makefile.objs |    1 +
 block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
 configure           |   26 +++
 qapi-schema.json    |    1 +
 5 files changed, 473 insertions(+)
 create mode 100644 block/nfs.c
MAINTAINERS +5 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>
 S: Supported
 F: block/iscsi.c
 
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+
 SSH
 M: Richard W.M. Jones <[email protected]>
 S: Supported
block/Makefile.objs +1 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
 ifeq ($(CONFIG_POSIX),y)
 block-obj-y += nbd.o nbd-client.o sheepdog.o
 block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
 block-obj-$(CONFIG_CURL) += curl.o
 block-obj-$(CONFIG_RBD) += rbd.o
 block-obj-$(CONFIG_GLUSTERFS) += gluster.o
block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 138 lines
block/nfs.c new file +138 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{

block/nfs.c line 254

Quoted 10 lines
block/nfs.c new file +10 -0
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

block/nfs.c line 259

Quoted 5 lines
block/nfs.c new file +5 -0
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

goto out ?

Quoted 75 lines
block/nfs.c new file +75 -0
+    }
+    file = g_strdup(strp);
+    *strp = 0;
+
+    client->context = nfs_init_context();
+    if (client->context == NULL) {
+        error_setg(errp, "Failed to init NFS context");
+        goto fail;
+    }
+
+    qp = query_params_parse(uri->query);
+    for (i = 0; i < qp->n; i++) {
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strncmp(qp->p[i].name, "uid", 3)) {
+            nfs_set_uid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "gid", 3)) {
+            nfs_set_gid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "tcp-syncnt", 10)) {
+            nfs_set_tcp_syncnt(client->context, atoi(qp->p[i].value));
+        } else {
+            error_setg(errp, "Unknown NFS parameter name: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+    }
+
+    ret = nfs_mount(client->context, uri->server, uri->path);
+    if (ret < 0) {
+        error_setg(errp, "Failed to mount nfs share: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    if (flags & O_CREAT) {
+        ret = nfs_creat(client->context, file, 0600, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to create file: %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    } else {
+        ret = nfs_open(client->context, file, flags, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to open file : %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    }
+
+    ret = nfs_fstat(client->context, client->fh, &st);
+    if (ret < 0) {
+        error_setg(errp, "Failed to fstat file: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    ret = DIV_ROUND_UP(st.st_size, BDRV_SECTOR_SIZE);
+    client->has_zero_init = S_ISREG(st.st_mode);
+    goto out;
+fail:
+    nfs_client_close(client);
+out:
+    if (qp) {
+        query_params_free(qp);
+    }
+    uri_free(uri);
+    g_free(file);
+    return ret;
+}
+
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c line 344

Quoted 10 lines
block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 24 lines
block/nfs.c new file +24 -0
+        error_free(local_err);
+        return -EINVAL;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        return ret;
+    }
+    bs->total_sectors = ret;
+    return 0;
+}
+
+static int nfs_file_create(const char *url, QEMUOptionParameter *options,
+                         Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_malloc0(sizeof(NFSClient));
+
+    /* Read out options */
+    while (options && options->name) {
+        if (!strcmp(options->name, "size")) {
+            total_size = options->value.n;

block/nfs.c line 378

Quoted 10 lines
block/nfs.c new file +10 -0
+        }
+        options++;
+    }
+
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+out:

Should the out: label be after nfs_client_close(client); since the code will jump to it on nfs_client_open failure ?

Quoted 150 lines
block/nfs.c new file +62 -0
+    nfs_client_close(client);
+    g_free(client);
+    return ret;
+}
+
+static int nfs_has_zero_init(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    return client->has_zero_init;
+}
+
+static int64_t nfs_get_allocated_file_size(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task = {0};
+    struct stat st;
+
+    task.st = &st;
+    if (nfs_fstat_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_aio_wait();
+    }
+
+    return (task.status < 0 ? task.status : st.st_blocks * st.st_blksize);
+}
+
+static int nfs_file_truncate(BlockDriverState *bs, int64_t offset)
+{
+    NFSClient *client = bs->opaque;
+    return nfs_ftruncate(client->context, client->fh, offset);
+}
+
+static BlockDriver bdrv_nfs = {
+    .format_name     = "nfs",
+    .protocol_name   = "nfs",
+
+    .instance_size   = sizeof(NFSClient),
+    .bdrv_needs_filename = true,
+    .bdrv_has_zero_init = nfs_has_zero_init,
+    .bdrv_get_allocated_file_size = nfs_get_allocated_file_size,
+    .bdrv_truncate = nfs_file_truncate,
+
+    .bdrv_file_open  = nfs_file_open,
+    .bdrv_close      = nfs_file_close,
+    .bdrv_create     = nfs_file_create,
+
+    .bdrv_co_readv         = nfs_co_readv,
+    .bdrv_co_writev        = nfs_co_writev,
+    .bdrv_co_flush_to_disk = nfs_co_flush,
+};
+
+static void nfs_block_init(void)
+{
+    bdrv_register(&bdrv_nfs);
+}
+
+block_init(nfs_block_init);
configure +26 -0
index b472694..9ca0c63 100755
@@ -251,6 +251,7 @@ vss_win32_sdk=""
 win_sdk="no"
 want_tools="yes"
 libiscsi=""
+libnfs=""
 coroutine=""
 coroutine_pool=""
 seccomp=""
@@ -840,6 +841,10 @@ for opt do
   ;;
   --enable-libiscsi) libiscsi="yes"
   ;;
+  --disable-libnfs) libnfs="no"
+  ;;
+  --enable-libnfs) libnfs="yes"
+  ;;
   --enable-profiler) profiler="yes"
   ;;
   --disable-cocoa) cocoa="no"
@@ -1229,6 +1234,8 @@ Advanced options (experts only):
   --enable-rbd             enable building the rados block device (rbd)
   --disable-libiscsi       disable iscsi support
   --enable-libiscsi        enable iscsi support
+  --disable-libnfs         disable nfs support
+  --enable-libnfs          enable nfs support
   --disable-smartcard-nss  disable smartcard nss support
   --enable-smartcard-nss   enable smartcard nss support
   --disable-libusb         disable libusb (for usb passthrough)
@@ -3600,6 +3607,20 @@ elif test "$debug" = "no" ; then
   CFLAGS="-O2 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 $CFLAGS"
 fi
 
+##########################################
+# Do we have libnfs
+if test "$libnfs" != "no" ; then
+  if $pkg_config --atleast-version=1.8.91 libnfs; then
+    libnfs="yes"
+    libnfs_libs=$($pkg_config --libs libnfs)
+    LIBS="$LIBS $libnfs_libs"
+  else
+    if test "$libnfs" = "yes" ; then
+      feature_not_found "libnfs"
+    fi
+    libnfs="no"
+  fi
+fi
 
 # Disable zero malloc errors for official releases unless explicitly told to
 # enable/disable
@@ -3829,6 +3850,7 @@ echo "libiscsi support  $libiscsi (1.4.0)"
 else
 echo "libiscsi support  $libiscsi"
 fi
+echo "libnfs support    $libnfs"
 echo "build guest agent $guest_agent"
 echo "QGA VSS support   $guest_agent_with_vss"
 echo "seccomp support   $seccomp"
@@ -4165,6 +4187,10 @@ if test "$libiscsi" = "yes" ; then
   fi
 fi
 
+if test "$libnfs" = "yes" ; then
+  echo "CONFIG_LIBNFS=y" >> $config_host_mak
+fi
+
 if test "$seccomp" = "yes"; then
   echo "CONFIG_SECCOMP=y" >> $config_host_mak
 fi
qapi-schema.json +1 -0
index 05ced9d..7cfb5e5 100644
@@ -4371,6 +4371,7 @@
 # TODO gluster: Wait for structured options
 # TODO iscsi: Wait for structured options
 # TODO nbd: Should take InetSocketAddress for 'host'?
+# TODO nfs: Wait for structured options
 # TODO rbd: Wait for structured options
 # TODO sheepdog: Wait for structured options
 # TODO ssh: Should take InetSocketAddress for 'host'?
Signature · 1 line
1.7.9.5
#3 lore raw
Quoted 454 lines · Benoît Canet
Quoted 170 lines · Peter Lieven

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---
  MAINTAINERS         |    5 +
  block/Makefile.objs |    1 +
  block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
  configure           |   26 +++
  qapi-schema.json    |    1 +
  5 files changed, 473 insertions(+)

create mode 100644 block/nfs.c

MAINTAINERS +5 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>
  S: Supported
  F: block/iscsi.c
 
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+
  SSH
  M: Richard W.M. Jones <[email protected]>
  S: Supported
block/Makefile.objs +1 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
  ifeq ($(CONFIG_POSIX),y)
  block-obj-y += nbd.o nbd-client.o sheepdog.o
  block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
  block-obj-$(CONFIG_CURL) += curl.o
  block-obj-$(CONFIG_RBD) += rbd.o
  block-obj-$(CONFIG_GLUSTERFS) += gluster.o
block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 138 lines
block/nfs.c new file +138 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{

block/nfs.c line 254

Quoted 10 lines
block/nfs.c new file +10 -0
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

block/nfs.c line 259

Quoted 5 lines
block/nfs.c new file +5 -0
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

goto out ?

Quoted 75 lines
block/nfs.c new file +75 -0
+    }
+    file = g_strdup(strp);
+    *strp = 0;
+
+    client->context = nfs_init_context();
+    if (client->context == NULL) {
+        error_setg(errp, "Failed to init NFS context");
+        goto fail;
+    }
+
+    qp = query_params_parse(uri->query);
+    for (i = 0; i < qp->n; i++) {
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strncmp(qp->p[i].name, "uid", 3)) {
+            nfs_set_uid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "gid", 3)) {
+            nfs_set_gid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "tcp-syncnt", 10)) {
+            nfs_set_tcp_syncnt(client->context, atoi(qp->p[i].value));
+        } else {
+            error_setg(errp, "Unknown NFS parameter name: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+    }
+
+    ret = nfs_mount(client->context, uri->server, uri->path);
+    if (ret < 0) {
+        error_setg(errp, "Failed to mount nfs share: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    if (flags & O_CREAT) {
+        ret = nfs_creat(client->context, file, 0600, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to create file: %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    } else {
+        ret = nfs_open(client->context, file, flags, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to open file : %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    }
+
+    ret = nfs_fstat(client->context, client->fh, &st);
+    if (ret < 0) {
+        error_setg(errp, "Failed to fstat file: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    ret = DIV_ROUND_UP(st.st_size, BDRV_SECTOR_SIZE);
+    client->has_zero_init = S_ISREG(st.st_mode);
+    goto out;
+fail:
+    nfs_client_close(client);
+out:
+    if (qp) {
+        query_params_free(qp);
+    }
+    uri_free(uri);
+    g_free(file);
+    return ret;
+}
+
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c line 344

Quoted 10 lines
block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 28 lines
block/nfs.c new file +28 -0
+        error_free(local_err);
+        return -EINVAL;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        return ret;
+    }
+    bs->total_sectors = ret;
+    return 0;
+}
+
+static int nfs_file_create(const char *url, QEMUOptionParameter *options,
+                         Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_malloc0(sizeof(NFSClient));
+
+    /* Read out options */
+    while (options && options->name) {
+        if (!strcmp(options->name, "size")) {
+            total_size = options->value.n;
+        }
+        options++;
+    }
+
Quoted 8 lines

block/nfs.c line 378

Quoted 6 lines
block/nfs.c new file +6 -0
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+out:

Should the out: label be after nfs_client_close(client); since the code will jump to it on nfs_client_open failure ?

You are right, but it doesn't hurt since you can call nfs_client_close multiple times.

Peter

Quoted 151 lines
Quoted 151 lines
block/nfs.c new file +62 -0
+    nfs_client_close(client);
+    g_free(client);
+    return ret;
+}
+
+static int nfs_has_zero_init(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    return client->has_zero_init;
+}
+
+static int64_t nfs_get_allocated_file_size(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task = {0};
+    struct stat st;
+
+    task.st = &st;
+    if (nfs_fstat_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_aio_wait();
+    }
+
+    return (task.status < 0 ? task.status : st.st_blocks * st.st_blksize);
+}
+
+static int nfs_file_truncate(BlockDriverState *bs, int64_t offset)
+{
+    NFSClient *client = bs->opaque;
+    return nfs_ftruncate(client->context, client->fh, offset);
+}
+
+static BlockDriver bdrv_nfs = {
+    .format_name     = "nfs",
+    .protocol_name   = "nfs",
+
+    .instance_size   = sizeof(NFSClient),
+    .bdrv_needs_filename = true,
+    .bdrv_has_zero_init = nfs_has_zero_init,
+    .bdrv_get_allocated_file_size = nfs_get_allocated_file_size,
+    .bdrv_truncate = nfs_file_truncate,
+
+    .bdrv_file_open  = nfs_file_open,
+    .bdrv_close      = nfs_file_close,
+    .bdrv_create     = nfs_file_create,
+
+    .bdrv_co_readv         = nfs_co_readv,
+    .bdrv_co_writev        = nfs_co_writev,
+    .bdrv_co_flush_to_disk = nfs_co_flush,
+};
+
+static void nfs_block_init(void)
+{
+    bdrv_register(&bdrv_nfs);
+}
+
+block_init(nfs_block_init);
configure +26 -0
index b472694..9ca0c63 100755
@@ -251,6 +251,7 @@ vss_win32_sdk=""
  win_sdk="no"
  want_tools="yes"
  libiscsi=""
+libnfs=""
  coroutine=""
  coroutine_pool=""
  seccomp=""
@@ -840,6 +841,10 @@ for opt do
    ;;
    --enable-libiscsi) libiscsi="yes"
    ;;
+  --disable-libnfs) libnfs="no"
+  ;;
+  --enable-libnfs) libnfs="yes"
+  ;;
    --enable-profiler) profiler="yes"
    ;;
    --disable-cocoa) cocoa="no"
@@ -1229,6 +1234,8 @@ Advanced options (experts only):
    --enable-rbd             enable building the rados block device (rbd)
    --disable-libiscsi       disable iscsi support
    --enable-libiscsi        enable iscsi support
+  --disable-libnfs         disable nfs support
+  --enable-libnfs          enable nfs support
    --disable-smartcard-nss  disable smartcard nss support
    --enable-smartcard-nss   enable smartcard nss support
    --disable-libusb         disable libusb (for usb passthrough)
@@ -3600,6 +3607,20 @@ elif test "$debug" = "no" ; then
    CFLAGS="-O2 -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 $CFLAGS"
  fi
 
+##########################################
+# Do we have libnfs
+if test "$libnfs" != "no" ; then
+  if $pkg_config --atleast-version=1.8.91 libnfs; then
+    libnfs="yes"
+    libnfs_libs=$($pkg_config --libs libnfs)
+    LIBS="$LIBS $libnfs_libs"
+  else
+    if test "$libnfs" = "yes" ; then
+      feature_not_found "libnfs"
+    fi
+    libnfs="no"
+  fi
+fi
 
  # Disable zero malloc errors for official releases unless explicitly told to
  # enable/disable
@@ -3829,6 +3850,7 @@ echo "libiscsi support  $libiscsi (1.4.0)"
  else
  echo "libiscsi support  $libiscsi"
  fi
+echo "libnfs support    $libnfs"
  echo "build guest agent $guest_agent"
  echo "QGA VSS support   $guest_agent_with_vss"
  echo "seccomp support   $seccomp"
@@ -4165,6 +4187,10 @@ if test "$libiscsi" = "yes" ; then
    fi
  fi
 
+if test "$libnfs" = "yes" ; then
+  echo "CONFIG_LIBNFS=y" >> $config_host_mak
+fi
+
  if test "$seccomp" = "yes"; then
    echo "CONFIG_SECCOMP=y" >> $config_host_mak
  fi
qapi-schema.json +1 -0
index 05ced9d..7cfb5e5 100644
@@ -4371,6 +4371,7 @@
  # TODO gluster: Wait for structured options
  # TODO iscsi: Wait for structured options
  # TODO nbd: Should take InetSocketAddress for 'host'?
+# TODO nfs: Wait for structured options
  # TODO rbd: Wait for structured options
  # TODO sheepdog: Wait for structured options
  # TODO ssh: Should take InetSocketAddress for 'host'?
Signature · 1 line
1.7.9.5
Signature · 16 lines
Mit freundlichen Grüßen

Peter Lieven

...........................................................

  KAMP Netzwerkdienste GmbH
  Vestische Str. 89-91 | 46117 Oberhausen
  Tel: +49 (0) 208.89 402-50 | Fax: +49 (0) 208.89 402-40
  [email protected] | http://www.kamp.de

  Geschäftsführer: Heiner Lante | Michael Lante
  Amtsgericht Duisburg | HRB Nr. 12154
  USt-Id-Nr.: DE 120607556

...........................................................
#4 lore raw
Quoted 462 lines · Peter Lieven
Quoted 453 lines · Benoît Canet
Quoted 169 lines · Peter Lieven

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---
 MAINTAINERS         |    5 +
 block/Makefile.objs |    1 +
 block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
 configure           |   26 +++
 qapi-schema.json    |    1 +
 5 files changed, 473 insertions(+)
 create mode 100644 block/nfs.c
MAINTAINERS +5 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>
 S: Supported
 F: block/iscsi.c
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+
 SSH
 M: Richard W.M. Jones <[email protected]>
 S: Supported
block/Makefile.objs +1 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
 ifeq ($(CONFIG_POSIX),y)
 block-obj-y += nbd.o nbd-client.o sheepdog.o
 block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
 block-obj-$(CONFIG_CURL) += curl.o
 block-obj-$(CONFIG_RBD) += rbd.o
 block-obj-$(CONFIG_GLUSTERFS) += gluster.o
block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 138 lines
block/nfs.c new file +138 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{

block/nfs.c line 254

Quoted 10 lines
block/nfs.c new file +10 -0
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

block/nfs.c line 259

Quoted 5 lines
block/nfs.c new file +5 -0
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

goto out ?

Quoted 75 lines
block/nfs.c new file +75 -0
+    }
+    file = g_strdup(strp);
+    *strp = 0;
+
+    client->context = nfs_init_context();
+    if (client->context == NULL) {
+        error_setg(errp, "Failed to init NFS context");
+        goto fail;
+    }
+
+    qp = query_params_parse(uri->query);
+    for (i = 0; i < qp->n; i++) {
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strncmp(qp->p[i].name, "uid", 3)) {
+            nfs_set_uid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "gid", 3)) {
+            nfs_set_gid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "tcp-syncnt", 10)) {
+            nfs_set_tcp_syncnt(client->context, atoi(qp->p[i].value));
+        } else {
+            error_setg(errp, "Unknown NFS parameter name: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+    }
+
+    ret = nfs_mount(client->context, uri->server, uri->path);
+    if (ret < 0) {
+        error_setg(errp, "Failed to mount nfs share: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    if (flags & O_CREAT) {
+        ret = nfs_creat(client->context, file, 0600, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to create file: %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    } else {
+        ret = nfs_open(client->context, file, flags, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to open file : %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    }
+
+    ret = nfs_fstat(client->context, client->fh, &st);
+    if (ret < 0) {
+        error_setg(errp, "Failed to fstat file: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    ret = DIV_ROUND_UP(st.st_size, BDRV_SECTOR_SIZE);
+    client->has_zero_init = S_ISREG(st.st_mode);
+    goto out;
+fail:
+    nfs_client_close(client);
+out:
+    if (qp) {
+        query_params_free(qp);
+    }
+    uri_free(uri);
+    g_free(file);
+    return ret;
+}
+
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c line 344

Quoted 10 lines
block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 28 lines
block/nfs.c new file +28 -0
+        error_free(local_err);
+        return -EINVAL;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        return ret;
+    }
+    bs->total_sectors = ret;
+    return 0;
+}
+
+static int nfs_file_create(const char *url, QEMUOptionParameter *options,
+                         Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_malloc0(sizeof(NFSClient));
+
+    /* Read out options */
+    while (options && options->name) {
+        if (!strcmp(options->name, "size")) {
+            total_size = options->value.n;
+        }
+        options++;
+    }
+
Quoted 8 lines

block/nfs.c line 378

Quoted 6 lines
block/nfs.c new file +6 -0
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+out:

Should the out: label be after nfs_client_close(client); since the code will jump to it on nfs_client_open failure ?

Quoted 2 lines

You are right, but it doesn't hurt since you can call nfs_client_close multiple times.

Will you send another revision addressing the other comments?

Stefan

#5 lore raw

Am 30.01.2014 um 10:05 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 463 lines · Peter Lieven
Quoted 462 lines · Benoît Canet
Quoted 461 lines · Peter Lieven
Quoted 169 lines

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---

MAINTAINERS         |    5 +
block/Makefile.objs |    1 +
block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
configure           |   26 +++
qapi-schema.json    |    1 +
5 files changed, 473 insertions(+)

create mode 100644 block/nfs.c

MAINTAINERS +0 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>

S: Supported
F: block/iscsi.c

MAINTAINERS +5 -0
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+

SSH
M: Richard W.M. Jones <[email protected]>
S: Supported

block/Makefile.objs +0 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
 ifeq ($(CONFIG_POSIX),y)

block-obj-y += nbd.o nbd-client.o sheepdog.o
block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
block-obj-$(CONFIG_CURL) += curl.o
block-obj-$(CONFIG_RBD) += rbd.o
block-obj-$(CONFIG_GLUSTERFS) += gluster.o

block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 138 lines
block/nfs.c new file +138 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{

block/nfs.c line 254

Quoted 10 lines
block/nfs.c new file +10 -0
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

block/nfs.c line 259

Quoted 5 lines
block/nfs.c new file +5 -0
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

goto out ?

Quoted 75 lines
block/nfs.c new file +75 -0
+    }
+    file = g_strdup(strp);
+    *strp = 0;
+
+    client->context = nfs_init_context();
+    if (client->context == NULL) {
+        error_setg(errp, "Failed to init NFS context");
+        goto fail;
+    }
+
+    qp = query_params_parse(uri->query);
+    for (i = 0; i < qp->n; i++) {
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strncmp(qp->p[i].name, "uid", 3)) {
+            nfs_set_uid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "gid", 3)) {
+            nfs_set_gid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "tcp-syncnt", 10)) {
+            nfs_set_tcp_syncnt(client->context, atoi(qp->p[i].value));
+        } else {
+            error_setg(errp, "Unknown NFS parameter name: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+    }
+
+    ret = nfs_mount(client->context, uri->server, uri->path);
+    if (ret < 0) {
+        error_setg(errp, "Failed to mount nfs share: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    if (flags & O_CREAT) {
+        ret = nfs_creat(client->context, file, 0600, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to create file: %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    } else {
+        ret = nfs_open(client->context, file, flags, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to open file : %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    }
+
+    ret = nfs_fstat(client->context, client->fh, &st);
+    if (ret < 0) {
+        error_setg(errp, "Failed to fstat file: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    ret = DIV_ROUND_UP(st.st_size, BDRV_SECTOR_SIZE);
+    client->has_zero_init = S_ISREG(st.st_mode);
+    goto out;
+fail:
+    nfs_client_close(client);
+out:
+    if (qp) {
+        query_params_free(qp);
+    }
+    uri_free(uri);
+    g_free(file);
+    return ret;
+}
+
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c line 344

Quoted 10 lines
block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 24 lines
block/nfs.c new file +24 -0
+        error_free(local_err);
+        return -EINVAL;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        return ret;
+    }
+    bs->total_sectors = ret;
+    return 0;
+}
+
+static int nfs_file_create(const char *url, QEMUOptionParameter *options,
+                         Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_malloc0(sizeof(NFSClient));
+
+    /* Read out options */
+    while (options && options->name) {
+        if (!strcmp(options->name, "size")) {
+            total_size = options->value.n;

block/nfs.c line 378

Quoted 10 lines
block/nfs.c new file +10 -0
+        }
+        options++;
+    }
+
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+out:

Should the out: label be after nfs_client_close(client); since the code will jump to it on nfs_client_open failure ?

Quoted 4 lines
Quoted 2 lines

You are right, but it doesn't hurt since you can call nfs_client_close multiple times.

Will you send another revision addressing the other comments?

Sorry, which other comments are you referring to?

Have you been able to test the patch without the bad commit?

BR,
Peter

#6 lore raw
Quoted 466 lines · Peter Lieven

Am 30.01.2014 um 10:05 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 463 lines · Peter Lieven
Quoted 462 lines · Benoît Canet
Quoted 461 lines · Peter Lieven
Quoted 169 lines

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---

MAINTAINERS         |    5 +
block/Makefile.objs |    1 +
block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
configure           |   26 +++
qapi-schema.json    |    1 +
5 files changed, 473 insertions(+)

create mode 100644 block/nfs.c

MAINTAINERS +0 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>

S: Supported
F: block/iscsi.c

MAINTAINERS +5 -0
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+

SSH
M: Richard W.M. Jones <[email protected]>
S: Supported

block/Makefile.objs +0 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
 ifeq ($(CONFIG_POSIX),y)

block-obj-y += nbd.o nbd-client.o sheepdog.o
block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
block-obj-$(CONFIG_CURL) += curl.o
block-obj-$(CONFIG_RBD) += rbd.o
block-obj-$(CONFIG_GLUSTERFS) += gluster.o

block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 138 lines
block/nfs.c new file +138 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{

block/nfs.c line 254

Quoted 10 lines
block/nfs.c new file +10 -0
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

block/nfs.c line 259

Quoted 5 lines
block/nfs.c new file +5 -0
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

goto out ?

Quoted 75 lines
block/nfs.c new file +75 -0
+    }
+    file = g_strdup(strp);
+    *strp = 0;
+
+    client->context = nfs_init_context();
+    if (client->context == NULL) {
+        error_setg(errp, "Failed to init NFS context");
+        goto fail;
+    }
+
+    qp = query_params_parse(uri->query);
+    for (i = 0; i < qp->n; i++) {
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strncmp(qp->p[i].name, "uid", 3)) {
+            nfs_set_uid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "gid", 3)) {
+            nfs_set_gid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "tcp-syncnt", 10)) {
+            nfs_set_tcp_syncnt(client->context, atoi(qp->p[i].value));
+        } else {
+            error_setg(errp, "Unknown NFS parameter name: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+    }
+
+    ret = nfs_mount(client->context, uri->server, uri->path);
+    if (ret < 0) {
+        error_setg(errp, "Failed to mount nfs share: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    if (flags & O_CREAT) {
+        ret = nfs_creat(client->context, file, 0600, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to create file: %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    } else {
+        ret = nfs_open(client->context, file, flags, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to open file : %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    }
+
+    ret = nfs_fstat(client->context, client->fh, &st);
+    if (ret < 0) {
+        error_setg(errp, "Failed to fstat file: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    ret = DIV_ROUND_UP(st.st_size, BDRV_SECTOR_SIZE);
+    client->has_zero_init = S_ISREG(st.st_mode);
+    goto out;
+fail:
+    nfs_client_close(client);
+out:
+    if (qp) {
+        query_params_free(qp);
+    }
+    uri_free(uri);
+    g_free(file);
+    return ret;
+}
+
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c line 344

Quoted 10 lines
block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 24 lines
block/nfs.c new file +24 -0
+        error_free(local_err);
+        return -EINVAL;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        return ret;
+    }
+    bs->total_sectors = ret;
+    return 0;
+}
+
+static int nfs_file_create(const char *url, QEMUOptionParameter *options,
+                         Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_malloc0(sizeof(NFSClient));
+
+    /* Read out options */
+    while (options && options->name) {
+        if (!strcmp(options->name, "size")) {
+            total_size = options->value.n;

block/nfs.c line 378

Quoted 10 lines
block/nfs.c new file +10 -0
+        }
+        options++;
+    }
+
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+out:

Should the out: label be after nfs_client_close(client); since the code will jump to it on nfs_client_open failure ?

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

You are right, but it doesn't hurt since you can call nfs_client_close multiple times.

Will you send another revision addressing the other comments?

Sorry, which other comments are you referring to?

I replied to Benoit's email hilighting his questions which were not addressed.

Quoted 1 line

Have you been able to test the patch without the bad commit?

Waiting for a final version to test.

Stefan

#7 lore raw

Am 30.01.2014 um 15:23 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 470 lines · Peter Lieven
Quoted 469 lines

Am 30.01.2014 um 10:05 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 466 lines · Peter Lieven
Quoted 464 lines · Benoît Canet
Quoted 453 lines · Peter Lieven
Quoted 169 lines

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---

MAINTAINERS         |    5 +
block/Makefile.objs |    1 +
block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
configure           |   26 +++
qapi-schema.json    |    1 +
5 files changed, 473 insertions(+)

create mode 100644 block/nfs.c

MAINTAINERS +0 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>

S: Supported
F: block/iscsi.c

MAINTAINERS +5 -0
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+

SSH
M: Richard W.M. Jones <[email protected]>
S: Supported

block/Makefile.objs +0 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
 ifeq ($(CONFIG_POSIX),y)

block-obj-y += nbd.o nbd-client.o sheepdog.o
block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
block-obj-$(CONFIG_CURL) += curl.o
block-obj-$(CONFIG_RBD) += rbd.o
block-obj-$(CONFIG_GLUSTERFS) += gluster.o

block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 138 lines
block/nfs.c new file +138 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{

block/nfs.c line 254

Quoted 10 lines
block/nfs.c new file +10 -0
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

block/nfs.c line 259

Quoted 5 lines
block/nfs.c new file +5 -0
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

goto out ?

Quoted 75 lines
block/nfs.c new file +75 -0
+    }
+    file = g_strdup(strp);
+    *strp = 0;
+
+    client->context = nfs_init_context();
+    if (client->context == NULL) {
+        error_setg(errp, "Failed to init NFS context");
+        goto fail;
+    }
+
+    qp = query_params_parse(uri->query);
+    for (i = 0; i < qp->n; i++) {
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strncmp(qp->p[i].name, "uid", 3)) {
+            nfs_set_uid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "gid", 3)) {
+            nfs_set_gid(client->context, atoi(qp->p[i].value));
+        } else if (!strncmp(qp->p[i].name, "tcp-syncnt", 10)) {
+            nfs_set_tcp_syncnt(client->context, atoi(qp->p[i].value));
+        } else {
+            error_setg(errp, "Unknown NFS parameter name: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+    }
+
+    ret = nfs_mount(client->context, uri->server, uri->path);
+    if (ret < 0) {
+        error_setg(errp, "Failed to mount nfs share: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    if (flags & O_CREAT) {
+        ret = nfs_creat(client->context, file, 0600, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to create file: %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    } else {
+        ret = nfs_open(client->context, file, flags, &client->fh);
+        if (ret < 0) {
+            error_setg(errp, "Failed to open file : %s",
+                       nfs_get_error(client->context));
+            goto fail;
+        }
+    }
+
+    ret = nfs_fstat(client->context, client->fh, &st);
+    if (ret < 0) {
+        error_setg(errp, "Failed to fstat file: %s",
+                   nfs_get_error(client->context));
+        goto fail;
+    }
+
+    ret = DIV_ROUND_UP(st.st_size, BDRV_SECTOR_SIZE);
+    client->has_zero_init = S_ISREG(st.st_mode);
+    goto out;
+fail:
+    nfs_client_close(client);
+out:
+    if (qp) {
+        query_params_free(qp);
+    }
+    uri_free(uri);
+    g_free(file);
+    return ret;
+}
+
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c line 344

Quoted 10 lines
block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 28 lines
block/nfs.c new file +28 -0
+        error_free(local_err);
+        return -EINVAL;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        return ret;
+    }
+    bs->total_sectors = ret;
+    return 0;
+}
+
+static int nfs_file_create(const char *url, QEMUOptionParameter *options,
+                         Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_malloc0(sizeof(NFSClient));
+
+    /* Read out options */
+    while (options && options->name) {
+        if (!strcmp(options->name, "size")) {
+            total_size = options->value.n;
+        }
+        options++;
+    }
+
Quoted 8 lines

block/nfs.c line 378

Quoted 6 lines
block/nfs.c new file +6 -0
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+out:

Should the out: label be after nfs_client_close(client); since the code will jump to it on nfs_client_open failure ?

You are right, but it doesn't hurt since you can call nfs_client_close multiple times.

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

Will you send another revision addressing the other comments?

Sorry, which other comments are you referring to?

I replied to Benoit's email hilighting his questions which were not addressed.

Sorry, I missed the other comments.

Quoted 4 lines
Quoted 2 lines

Have you been able to test the patch without the bad commit?

Waiting for a final version to test.

I cannot do that, unfortunately.

Ronnie, can you revert the bad commit and leave it for the 2.0 development. It would be good if you bump the version to 1.9.2 so I can check for that in the configure script

Peter

#8 lore raw
In reply to Benoît Canet
Quoted 1 line · Benoît Canet

Le Wednesday 29 Jan 2014 à 09:50:21 (+0100), Peter Lieven a écrit :

Hi Peter,
If I read your reply to Benoit correctly, you only addressed the questions about nfs_client_close(). Here are Benoits other comments and my thoughts on them:

Quoted 10 lines

block/nfs.c line 106

Quoted 9 lines
block/nfs.c new file +9 -0
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Maybe the argument name isn't ideal. Something like 'ret' or 'result' would be more commonplace.

Not critical but would be nice to change.

Quoted 5 lines
Quoted 5 lines
block/nfs.c new file +5 -0
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
Quoted 8 lines

block/nfs.c line 344

Quoted 6 lines
block/nfs.c new file +6 -0
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Yes, I think you are right. The Error should be propagated to the caller. It's not clear to me whether we can ever get an error from qemu_opts_absorb_qdict() in this call site though.

#9 lore raw

Am 30.01.2014 um 15:22 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 19 lines · Benoît Canet
Quoted 1 line · Peter Lieven

Hi Peter,
If I read your reply to Benoit correctly, you only addressed the questions about nfs_client_close(). Here are Benoits other comments and my thoughts on them:

Quoted 10 lines

block/nfs.c line 106

Quoted 9 lines
block/nfs.c new file +9 -0
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 2 lines

Maybe the argument name isn't ideal. Something like 'ret' or 'result' would be more commonplace.

I took the "status" from libnfs examples. I can change it in the qemu code if you like.

Quoted 17 lines

Not critical but would be nice to change.

Quoted 5 lines
Quoted 5 lines
block/nfs.c new file +5 -0
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
Quoted 8 lines

block/nfs.c line 344

Quoted 6 lines
block/nfs.c new file +6 -0
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 3 lines

Yes, I think you are right. The Error should be propagated to the caller. It's not clear to me whether we can ever get an error from qemu_opts_absorb_qdict() in this call site though.

Is there any action I should take here? If yes, can you advise what to do please.

Peter

#10 lore raw
Quoted 19 lines · Peter Lieven

Am 30.01.2014 um 15:22 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 16 lines · Benoît Canet
Quoted 14 lines · Peter Lieven

block/nfs.c line 344

Quoted 11 lines
block/nfs.c new file +11 -0
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 6 lines
Quoted 3 lines

Yes, I think you are right. The Error should be propagated to the caller. It's not clear to me whether we can ever get an error from qemu_opts_absorb_qdict() in this call site though.

Is there any action I should take here? If yes, can you advise what to do please.

The issue is that nfs_file_open() takes an Error **errp argument. This means the function should report detailed errors using the Error object.

The patch prints and then discards the local_error instead of propagating it to the caller's errp.

We should just propagate the error instead of printing it:

if (error_is_set(&local_err)) {
    error_propagate(errp, local_err);
    goto ...;
}

Stefan

#11 lore raw
Quoted 27 lines · Stefan Hajnoczi
Quoted 17 lines · Peter Lieven

Am 30.01.2014 um 15:22 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 15 lines · Benoît Canet
Quoted 14 lines · Peter Lieven

block/nfs.c line 344

Quoted 11 lines
block/nfs.c new file +11 -0
+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code. Maybe I am missing the point.

Quoted 5 lines
Quoted 3 lines

Yes, I think you are right. The Error should be propagated to the caller. It's not clear to me whether we can ever get an error from qemu_opts_absorb_qdict() in this call site though.

Is there any action I should take here? If yes, can you advise what to do please.

The issue is that nfs_file_open() takes an Error **errp argument. This means the function should report detailed errors using the Error object.

Quoted 7 lines

The patch prints and then discards the local_error instead of propagating it to the caller's errp.

We should just propagate the error instead of printing it:

if (error_is_set(&local_err)) {
     error_propagate(errp, local_err);
     goto ...;

Ok, you are just referring to this part in nfs_file_open:

    if (error_is_set(&local_err)) {
        qerror_report_err(local_err);
        error_free(local_err);
        return -EINVAL;
    }

which I would change to:

    if (error_is_set(&local_err)) {
        error_propagate(errp, local_err);
        return -EINVAL;
    }

The use of error_setg in nfs_client_open is ok?

Peter

#12 lore raw
Quoted 55 lines · Peter Lieven
Quoted 37 lines · Stefan Hajnoczi
Quoted 24 lines · Peter Lieven

Am 30.01.2014 um 15:22 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 21 lines · Benoīt Canet
Quoted 19 lines · Peter Lieven

block/nfs.c line 344

Quoted 13 lines

+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

block/nfs.c new file +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code.
Maybe I am missing the point.

Quoted 6 lines
Quoted 3 lines

Yes, I think you are right. The Error should be propagated to the caller. It's not clear to me whether we can ever get an error from qemu_opts_absorb_qdict() in this call site though.

Is there any action I should take here? If yes, can you advise what to do please.

The issue is that nfs_file_open() takes an Error **errp argument. This means the function should report detailed errors using the Error object.

Quoted 7 lines

The patch prints and then discards the local_error instead of propagating it to the caller's errp.

We should just propagate the error instead of printing it:

if (error_is_set(&local_err)) {
     error_propagate(errp, local_err);
     goto ...;

Ok, you are just referring to this part in nfs_file_open:

    if (error_is_set(&local_err)) {
        qerror_report_err(local_err);
        error_free(local_err);
        return -EINVAL;
    }
Quoted 6 lines

which I would change to:

    if (error_is_set(&local_err)) {
        error_propagate(errp, local_err);
        return -EINVAL;
    }

Yes.

Quoted 1 line

The use of error_setg in nfs_client_open is ok?

Yes, it's fine.

The Error API is not 100% obvious when you first see it, but once you learn the conventions it's pretty usable:

Functions take an Error **errp argument. This argument is optional, so a caller that doesn't care about detailed error messages may pass NULL. This has implications...

error_setg(errp, fmt, ...) handles errp == NULL internally so you can call it unconditionally.

The tricky thing is that error_is_set() only works if errp is non-NULL (otherwise error_setg() skips creating an Error object). So it means you cannot rely on your errp argument and often functions will declare Error *local_err = NULL, so they can pass &local_err to child functions. Finally, error_propagate(errp, local_err) is used to pass out the Error object to our caller.

Hope this summary makes the Error API clear.

Stefan

#13 lore raw
Quoted 75 lines · Stefan Hajnoczi
Quoted 46 lines · Peter Lieven
Quoted 29 lines · Stefan Hajnoczi
Quoted 19 lines · Peter Lieven

Am 30.01.2014 um 15:22 schrieb Stefan Hajnoczi <[email protected]>:

Quoted 17 lines · Benoīt Canet
Quoted 16 lines · Peter Lieven

Review

Quoted 12 lines

+static int nfs_file_open(BlockDriverState *bs, QDict *options, int flags,

(quoted diff) +10 -0
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (error_is_set(&local_err)) {
+        qerror_report_err(local_err);

I have seen more usage of error_propagate(errp, local_err); in QEMU code.
Maybe I am missing the point.

Quoted 5 lines
Quoted 3 lines

Yes, I think you are right. The Error should be propagated to the caller. It's not clear to me whether we can ever get an error from qemu_opts_absorb_qdict() in this call site though.

Is there any action I should take here? If yes, can you advise what to do please.

The issue is that nfs_file_open() takes an Error **errp argument. This means the function should report detailed errors using the Error object.

Quoted 7 lines

The patch prints and then discards the local_error instead of propagating it to the caller's errp.

We should just propagate the error instead of printing it:

if (error_is_set(&local_err)) {
      error_propagate(errp, local_err);
      goto ...;

Ok, you are just referring to this part in nfs_file_open:

     if (error_is_set(&local_err)) {
         qerror_report_err(local_err);
         error_free(local_err);
         return -EINVAL;
     }
Quoted 6 lines

which I would change to:

     if (error_is_set(&local_err)) {
         error_propagate(errp, local_err);
         return -EINVAL;
     }

Yes.

Quoted 1 line

The use of error_setg in nfs_client_open is ok?

Yes, it's fine.

The Error API is not 100% obvious when you first see it, but once you learn the conventions it's pretty usable:

Functions take an Error **errp argument. This argument is optional, so a caller that doesn't care about detailed error messages may pass NULL. This has implications...

error_setg(errp, fmt, ...) handles errp == NULL internally so you can call it unconditionally.

The tricky thing is that error_is_set() only works if errp is non-NULL (otherwise error_setg() skips creating an Error object). So it means you cannot rely on your errp argument and often functions will declare Error *local_err = NULL, so they can pass &local_err to child functions. Finally, error_propagate(errp, local_err) is used to pass out the Error object to our caller.

Quoted 3 lines

Hope this summary makes the Error API clear.

Stefan

Thanks for the explanation. I will sent you an updated series shortly.

Peter

Signature · 16 lines
Mit freundlichen Grüßen

Peter Lieven

...........................................................

  KAMP Netzwerkdienste GmbH
  Vestische Str. 89-91 | 46117 Oberhausen
  Tel: +49 (0) 208.89 402-50 | Fax: +49 (0) 208.89 402-40
  [email protected] | http://www.kamp.de

  Geschäftsführer: Heiner Lante | Michael Lante
  Amtsgericht Duisburg | HRB Nr. 12154
  USt-Id-Nr.: DE 120607556

...........................................................
#14 lore raw
In reply to Benoît Canet
Quoted 324 lines · Benoît Canet
Quoted 170 lines · Peter Lieven

This patch adds native support for accessing images on NFS shares without the requirement to actually mount the entire NFS share on the host.

NFS Images can simply be specified by an url of the form:
nfs://<host>/<export>/<filename>[?param=value[&param2=value2[&...]]]

For example:
qemu-img create -f qcow2 nfs://10.0.0.1/qemu-images/test.qcow2

You need LibNFS from Ronnie Sahlberg available at:
git://github.com/sahlberg/libnfs.git
for this to work.

During configure it is automatically probed for libnfs and support is enabled on-the-fly. You can forbid or enforce libnfs support with --disable-libnfs or --enable-libnfs respectively.

Due to NFS restrictions you might need to execute your binaries as root, allow them to open priviledged ports (<1024) or specify insecure option on the NFS server.

For additional information on ROOT vs. non-ROOT operation and URL format + parameters see:
https://raw.github.com/sahlberg/libnfs/master/README

Supported by qemu are the uid, gid and tcp-syncnt URL parameters.

LibNFS currently support NFS version 3 only.

---
  MAINTAINERS         |    5 +
  block/Makefile.objs |    1 +
  block/nfs.c         |  440 +++++++++++++++++++++++++++++++++++++++++++++++++++
  configure           |   26 +++
  qapi-schema.json    |    1 +
  5 files changed, 473 insertions(+)

create mode 100644 block/nfs.c

MAINTAINERS +5 -0
index fb53242..f8411f9 100644
@@ -936,6 +936,11 @@ M: Peter Lieven <[email protected]>
  S: Supported
  F: block/iscsi.c
 
+NFS
+M: Peter Lieven <[email protected]>
+S: Maintained
+F: block/nfs.c
+
  SSH
  M: Richard W.M. Jones <[email protected]>
  S: Supported
block/Makefile.objs +1 -0
index 4e8c91e..e254a21 100644
@@ -12,6 +12,7 @@ block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
  ifeq ($(CONFIG_POSIX),y)
  block-obj-y += nbd.o nbd-client.o sheepdog.o
  block-obj-$(CONFIG_LIBISCSI) += iscsi.o
+block-obj-$(CONFIG_LIBNFS) += nfs.o
  block-obj-$(CONFIG_CURL) += curl.o
  block-obj-$(CONFIG_RBD) += rbd.o
  block-obj-$(CONFIG_GLUSTERFS) += gluster.o
block/nfs.c new file +96 -0
new file mode 100644
index 0000000..2bbf7a2
@@ -0,0 +1,440 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <[email protected]>
+ *
+ * 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 "config-host.h"
+
+#include <poll.h>
+#include "qemu-common.h"
+#include "qemu/config-file.h"
+#include "qemu/error-report.h"
+#include "block/block_int.h"
+#include "trace.h"
+#include "qemu/iov.h"
+#include "qemu/uri.h"
+#include "sysemu/sysemu.h"
+#include <nfsc/libnfs.h>
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int status;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+} NFSRPC;
+
+static void nfs_process_read(void *arg);
+static void nfs_process_write(void *arg);
+
+static void nfs_set_events(NFSClient *client)
+{
+    int ev = nfs_which_events(client->context);
+    if (ev != client->events) {
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context),
+                      (ev & POLLIN) ? nfs_process_read : NULL,
+                      (ev & POLLOUT) ? nfs_process_write : NULL,
+                      client);
+
+    }
+    client->events = ev;
+}
+
+static void nfs_process_read(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLIN);
+    nfs_set_events(client);
+}
+
+static void nfs_process_write(void *arg)
+{
+    NFSClient *client = arg;
+    nfs_service(client->context, POLLOUT);
+    nfs_set_events(client);
+}
+
+static void nfs_co_init_task(NFSClient *client, NFSRPC *task)
+{
+    *task = (NFSRPC) {
+        .co         = qemu_coroutine_self(),
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}

block/nfs.c line 106

Quoted 10 lines
block/nfs.c new file +10 -0
+
+static void
+nfs_co_generic_cb(int status, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->complete = 1;
+    task->status = status;
+    if (task->status > 0 && task->iov) {
+        if (task->status <= task->iov->size) {

It feel very odd to compare something named status with something named size.

Quoted 141 lines
block/nfs.c new file +141 -0
+            qemu_iovec_from_buf(task->iov, 0, data, task->status);
+        } else {
+            task->status = -EIO;
+        }
+    }
+    if (task->status == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->co) {
+        task->bh = qemu_bh_new(nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    }
+}
+
+static int coroutine_fn nfs_co_readv(BlockDriverState *bs,
+                                     int64_t sector_num, int nb_sectors,
+                                     QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+    task.iov = iov;
+
+    if (nfs_pread_async(client->context, client->fh,
+                        sector_num * BDRV_SECTOR_SIZE,
+                        nb_sectors * BDRV_SECTOR_SIZE,
+                        nfs_co_generic_cb, &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    if (task.status < 0) {
+        return task.status;
+    }
+
+    /* zero pad short reads */
+    if (task.status < iov->size) {
+        qemu_iovec_memset(iov, task.status, 0, iov->size - task.status);
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_writev(BlockDriverState *bs,
+                                        int64_t sector_num, int nb_sectors,
+                                        QEMUIOVector *iov)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+    char *buf = NULL;
+
+    nfs_co_init_task(client, &task);
+
+    buf = g_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    qemu_iovec_to_buf(iov, 0, buf, nb_sectors * BDRV_SECTOR_SIZE);
+
+    if (nfs_pwrite_async(client->context, client->fh,
+                         sector_num * BDRV_SECTOR_SIZE,
+                         nb_sectors * BDRV_SECTOR_SIZE,
+                         buf, nfs_co_generic_cb, &task) != 0) {
+        g_free(buf);
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    g_free(buf);
+
+    if (task.status != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.status < 0 ? task.status : -EIO;
+    }
+
+    return 0;
+}
+
+static int coroutine_fn nfs_co_flush(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    NFSRPC task;
+
+    nfs_co_init_task(client, &task);
+
+    if (nfs_fsync_async(client->context, client->fh, nfs_co_generic_cb,
+                        &task) != 0) {
+        return -ENOMEM;
+    }
+
+    while (!task.complete) {
+        nfs_set_events(client);
+        qemu_coroutine_yield();
+    }
+
+    return task.status;
+}
+
+/* TODO Convert to fine grained options */
+static QemuOptsList runtime_opts = {
+    .name = "nfs",
+    .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+    .desc = {
+        {
+            .name = "filename",
+            .type = QEMU_OPT_STRING,
+            .help = "URL to the NFS file",
+        },
+        { /* end of list */ }
+    },
+};
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        qemu_aio_set_fd_handler(nfs_get_fd(client->context), NULL, NULL, NULL);
+        nfs_destroy_context(client->context);
+    }
+    memset(client, 0, sizeof(NFSClient));
+}
+
+static void nfs_file_close(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+    nfs_client_close(client);
+}
+
+static int64_t nfs_.cilclient_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{
+    int ret = -EINVAL, i;
+    struct stat st;
+    URI *uri;
Quoted 8 lines

block/nfs.c line 254

Quoted 7 lines
block/nfs.c new file +7 -0
+    QueryParams *qp = NULL;
+    char *file = NULL, *strp = NULL;
+
+    uri = uri_parse(filename);
+    if (!uri) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;

Should this be goto out since we don't have done nfs_init_context yet ?

it doesn't matter nfs_client_close can be called with client->context == NULL.

Sorry I entirely missed your other comments.

Peter

#15 lore raw

PATCH v7 2/5 qemu-iotests: change _supported_proto to file for various tests

In reply to Peter Lieven

all these tests do anything of the following and thus fail with any protocol other than file:
- the tests use rm, cp or mv shell commands which only work on file
- the tests use qcow2.py
- the images construct new filenames (e.g. backing file names) and the logic is broken for anything else than file

---
 tests/qemu-iotests/013 |    2 +-
 tests/qemu-iotests/014 |    2 +-
 tests/qemu-iotests/018 |    2 +-
 tests/qemu-iotests/019 |    2 +-
 tests/qemu-iotests/023 |    2 +-
 tests/qemu-iotests/024 |    2 +-
 tests/qemu-iotests/026 |    2 +-
 tests/qemu-iotests/028 |    2 +-
 tests/qemu-iotests/031 |    2 +-
 tests/qemu-iotests/034 |    2 +-
 tests/qemu-iotests/036 |    2 +-
 tests/qemu-iotests/037 |    2 +-
 tests/qemu-iotests/038 |    2 +-
 tests/qemu-iotests/039 |    2 +-
 tests/qemu-iotests/043 |    2 +-
 tests/qemu-iotests/046 |    2 +-
 tests/qemu-iotests/052 |    2 +-
 tests/qemu-iotests/054 |    2 +-
 tests/qemu-iotests/059 |    2 +-
 tests/qemu-iotests/060 |    2 +-
 tests/qemu-iotests/061 |    2 +-
 tests/qemu-iotests/063 |    2 +-
 tests/qemu-iotests/069 |    2 +-
 23 files changed, 23 insertions(+), 23 deletions(-)
tests/qemu-iotests/013 +1 -1
index 389f4b8..ea3cab9 100755
@@ -41,7 +41,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # much of this could be generic for any format supporting compression.
 _supported_fmt qcow qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 TEST_OFFSETS="0 4294967296"
tests/qemu-iotests/014 +1 -1
index 0edeb4b..b23c2db 100755
@@ -43,7 +43,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # much of this could be generic for any format supporting snapshots
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 TEST_OFFSETS="0 4294967296"
tests/qemu-iotests/018 +1 -1
index 6f7f054..d8a7d43 100755
@@ -41,7 +41,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # Any format supporting backing files
 _supported_fmt qcow qcow2 vmdk qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" "subformat=twoGbMaxExtentFlat"
 
tests/qemu-iotests/019 +1 -1
index b43e70f..e67445c 100755
@@ -45,7 +45,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # Any format supporting backing files
 _supported_fmt qcow qcow2 vmdk qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" \
                      "subformat=twoGbMaxExtentFlat" \
tests/qemu-iotests/023 +1 -1
index 090ed23..9ad06b9 100755
@@ -41,7 +41,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # much of this could be generic for any format supporting compression.
 _supported_fmt qcow qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 TEST_OFFSETS="0 4294967296"
tests/qemu-iotests/024 +1 -1
index be974f0..9bf99e1 100755
@@ -43,7 +43,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # Currently only qcow2 and qed support rebasing
 _supported_fmt qcow2 qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 CLUSTER_SIZE=65536
tests/qemu-iotests/026 +1 -1
index c9c5f83..df2884b 100755
@@ -42,7 +42,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # Currently only qcow2 supports rebasing
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _default_cache_mode "writethrough"
 _supported_cache_modes "writethrough" "none"
tests/qemu-iotests/028 +1 -1
index 93a9fa6..a99e4fa 100755
@@ -45,7 +45,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 # Any format supporting backing files except vmdk and qcow which do not support
 # smaller backing files.
 _supported_fmt qcow2 qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 # Choose a size that is not necessarily a cluster size multiple for image
tests/qemu-iotests/031 +1 -1
index c9070b0..1d920ea 100755
@@ -41,7 +41,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # This tests qcow2-specific low-level functionality
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 CLUSTER_SIZE=65536
tests/qemu-iotests/034 +1 -1
index 7349789..69c7858 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt qcow qcow2 vmdk qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" \
                      "subformat=twoGbMaxExtentFlat" \
tests/qemu-iotests/036 +1 -1
index e049a64..03b6aa9 100755
@@ -44,7 +44,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # This tests qcow2-specific low-level functionality
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 # Only qcow2v3 and later supports feature bits
tests/qemu-iotests/037 +1 -1
index e444349..9171d8c 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt qcow qcow2 vmdk qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" \
                      "subformat=twoGbMaxExtentFlat" \
tests/qemu-iotests/038 +1 -1
index 7bb7906..cfaf00a 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt qcow2 qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 CLUSTER_SIZE=2M
tests/qemu-iotests/039 +1 -1
index 6abf472..9b355c0 100755
@@ -42,7 +42,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _default_cache_mode "writethrough"
 _supported_cache_modes "writethrough"
tests/qemu-iotests/043 +1 -1
index d7f1231..b316b97 100755
@@ -41,7 +41,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # Any format supporting backing files
 _supported_fmt qcow2 qed
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 
tests/qemu-iotests/046 +1 -1
index 3f17ceb..2d44bbb 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 CLUSTER_SIZE=64k
tests/qemu-iotests/052 +1 -1
index 4d4e411..6bdae92 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt generic
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _default_cache_mode "writethrough"
 _supported_cache_modes "writethrough"
tests/qemu-iotests/054 +1 -1
index 5a0d1b1..c8b7082 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 echo
tests/qemu-iotests/059 +1 -1
index 2d604d3..ca5aa16 100755
@@ -40,7 +40,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # This tests vmdk-specific low-level functionality
 _supported_fmt vmdk
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" \
                      "subformat=twoGbMaxExtentFlat" \
tests/qemu-iotests/060 +1 -1
index bbb1909..af8ed9f 100755
@@ -40,7 +40,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # This tests qocw2-specific low-level functionality
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 rt_offset=65536  # 0x10000 (XXX: just an assumption)
tests/qemu-iotests/061 +1 -1
index e42f9bd..d3a6b38 100755
@@ -40,7 +40,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 
 # This tests qocw2-specific low-level functionality
 _supported_fmt qcow2
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 
 echo
tests/qemu-iotests/063 +1 -1
index 77503a2..a47493a 100755
@@ -42,7 +42,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.pattern
 
 _supported_fmt qcow qcow2 vmdk qed raw
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" \
                      "subformat=twoGbMaxExtentFlat" \
tests/qemu-iotests/069 +1 -1
index 50347d9..e661598 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt cow qed qcow qcow2 vmdk
-_supported_proto generic
+_supported_proto file
 _supported_os Linux
 _unsupported_imgopts "subformat=monolithicFlat" "subformat=twoGbMaxExtentFlat"
 
Signature · 1 line
1.7.9.5
#16 lore raw

PATCH v7 3/5 qemu-iotests: enable support for NFS protocol

In reply to Peter Lieven
---
 tests/qemu-iotests/common    |   22 +++++++++++++++++++---
 tests/qemu-iotests/common.rc |    3 +++
 2 files changed, 22 insertions(+), 3 deletions(-)
tests/qemu-iotests/common +19 -3
index 8b4e22c..5795358 100644
@@ -144,10 +144,12 @@ check options
     -vpc                test vpc
     -vhdx               test vhdx
     -vmdk               test vmdk
+    -file               test file (default)
     -rbd                test rbd
     -sheepdog           test sheepdog
     -nbd                test nbd
     -ssh                test ssh
+    -nfs                test nfs
     -xdiff              graphical mode diff
     -nocache            use O_DIRECT on backing file
     -misalign           misalign memory allocations
@@ -211,22 +213,36 @@ testlist options
             xpand=false
             ;;
 
+        -file)
+            IMGPROTO=file
+            xpand=false
+            ;;
+
         -rbd)
             IMGPROTO=rbd
             xpand=false
             ;;
+
         -sheepdog)
             IMGPROTO=sheepdog
             xpand=false
             ;;
+
         -nbd)
             IMGPROTO=nbd
             xpand=false
             ;;
+
         -ssh)
             IMGPROTO=ssh
             xpand=false
             ;;
+
+        -nfs)
+            IMGPROTO=nfs
+            xpand=false
+            ;;
+
         -nocache)
             CACHEMODE="none"
             CACHEMODE_IS_DEFAULT=false
@@ -238,10 +254,10 @@ testlist options
             xpand=false
             ;;
 
-    -valgrind)
-        valgrind=true
+        -valgrind)
+            valgrind=true
             xpand=false
-        ;;
+            ;;
 
         -g)        # -g group ... pick from group file
             group=true
tests/qemu-iotests/common.rc +3 -0
index 0f68156..71e9a74 100644
@@ -61,6 +61,9 @@ elif [ "$IMGPROTO" = "nbd" ]; then
 elif [ "$IMGPROTO" = "ssh" ]; then
     TEST_IMG_FILE=$TEST_DIR/t.$IMGFMT
     TEST_IMG="ssh://127.0.0.1$TEST_IMG_FILE"
+elif [ "$IMGPROTO" = "nfs" ]; then
+    TEST_DIR="nfs://127.0.0.1/$TEST_DIR"
+    TEST_IMG=$TEST_DIR/t.$IMGFMT
 else
     TEST_IMG=$IMGPROTO:$TEST_DIR/t.$IMGFMT
 fi
Signature · 1 line
1.7.9.5
#17 lore raw

PATCH v7 4/5 qemu-iotests: enable test 016 and 025 to work with NFS protocol

In reply to Peter Lieven
---
 tests/qemu-iotests/016 |    2 +-
 tests/qemu-iotests/025 |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
tests/qemu-iotests/016 +1 -1
index b87a32b..7ea9e94 100755
@@ -39,7 +39,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.filter
 
 _supported_fmt raw
-_supported_proto file sheepdog
+_supported_proto file sheepdog nfs
 _supported_os Linux
 
 
tests/qemu-iotests/025 +1 -1
index 9426c93..a5f45b4 100755
@@ -40,7 +40,7 @@ trap "_cleanup; exit \$status" 0 1 2 3 15
 . ./common.pattern
 
 _supported_fmt raw qcow2 qed
-_supported_proto file sheepdog rbd
+_supported_proto file sheepdog rbd nfs
 _supported_os Linux
 
 echo "=== Creating image"
Signature · 1 line
1.7.9.5
#18 lore raw

PATCH v7 5/5 qemu-iotests: blacklist test 020 for NFS protocol

In reply to Peter Lieven

reopening is currently not supported.

---
 tests/qemu-iotests/020 |    5 +++++
 1 file changed, 5 insertions(+)
tests/qemu-iotests/020 +5 -0
index 73a0429..2f258dc 100755
@@ -49,6 +49,11 @@ _unsupported_imgopts "subformat=monolithicFlat" \
                      "subformat=twoGbMaxExtentFlat" \
                      "subformat=twoGbMaxExtentSparse"
 
+# NFS does not support bdrv_reopen_prepare thus qemu-img commit fails.
+if [ "$IMGPROTO" = "nfs" ]; then
+    _notrun "image protocol $IMGPROTO does not support bdrv_commit"
+fi
+
 TEST_OFFSETS="0 4294967296"
 
 _make_test_img 6G
Signature · 1 line
1.7.9.5
#19 lore raw
In reply to Peter Lieven
Quoted 7 lines · Peter Lieven

This adds v6 of the NFS protocol driver + qemu-iotest adjustments.

v6->v7:
- removed unsused and partially non-existing libnfs header file includes [Stefan]
- adjusted usage of qemu_opts_create_nofail [Stefan]
- removed patch to fix output of qemu-iotest 067
- only blacklist qemu-iotest 020 for NFS do not change proto from generic to file [Fam]

It compiles now but I can't create an image on my /home NFS export:

$ ./qemu-img create -f qcow2 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000' 8G
Formatting 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000', fmt=qcow2 size=8589934592 encryption=off cluster_size=65536 lazy_refcounts=off 
qemu-img: nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000: Failed to create file: nfs_creat_async failed

I added printfs to the code, block/nfs.c is mounting /home/stefanha/qemu and tries to create the file /test-nfs.img.

I'm using the Linux kernel NFS server. Unfortunately I don't have more time to debug this but I'd like to test this feature successfully before merging it into QEMU. Any ideas why libnfs refuses to create the file? (It's failing in nfs_lookuppath_async() with "Path is empty".)

Ronnie: libnfs clobbers the nfs->rpc error string so you cannot see the detailed error message:

In this case nfs_creat() clobbers the error with "Failed to create file:
nfs_creat_async failed". In fact the message was "Out of memory: failed to start parsing the path components" from nfs_creat_async(), but that's clobbering the true error too. It really should be "Path is empty" from nfs_lookuppath_async().

It would be nice to preserve the full error string, otherwise the user cannot make sense of the problem.

Stefan

#20 lore raw
Quoted 31 lines · Stefan Hajnoczi
Quoted 7 lines · Peter Lieven

This adds v6 of the NFS protocol driver + qemu-iotest adjustments.

v6->v7:
- removed unsused and partially non-existing libnfs header file includes [Stefan]
- adjusted usage of qemu_opts_create_nofail [Stefan]
- removed patch to fix output of qemu-iotest 067
- only blacklist qemu-iotest 020 for NFS do not change proto from generic to file [Fam]

It compiles now but I can't create an image on my /home NFS export:

$ ./qemu-img create -f qcow2 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000' 8G
Formatting 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000', fmt=qcow2 size=8589934592 encryption=off cluster_size=65536 lazy_refcounts=off
qemu-img: nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000: Failed to create file: nfs_creat_async failed

I added printfs to the code, block/nfs.c is mounting /home/stefanha/qemu and tries to create the file /test-nfs.img.

I'm using the Linux kernel NFS server. Unfortunately I don't have more time to debug this but I'd like to test this feature successfully before merging it into QEMU. Any ideas why libnfs refuses to create the file? (It's failing in nfs_lookuppath_async() with "Path is empty".)

Ronnie: libnfs clobbers the nfs->rpc error string so you cannot see the detailed error message:

In this case nfs_creat() clobbers the error with "Failed to create file:
nfs_creat_async failed". In fact the message was "Out of memory: failed to start parsing the path components" from nfs_creat_async(), but that's clobbering the true error too. It really should be "Path is empty" from nfs_lookuppath_async().

Quoted 4 lines

It would be nice to preserve the full error string, otherwise the user cannot make sense of the problem.

Stefan

I do not understand why your path is empty...

Running almost the same test:

~/git/qemu$ ./qemu-img create -f qcow2 'nfs://127.0.0.1/home/lieven/git/qemu/test-nfs.img?uid=1000&gid=1000' 8G
Formatting 'nfs://127.0.0.1/home/lieven/git/qemu/test-nfs.img?uid=1000&gid=1000', fmt=qcow2 size=8589934592 encryption=off cluster_size=65536 lazy_refcounts=off
qemu-img: nfs://127.0.0.1/home/lieven/git/qemu/test-nfs.img?uid=1000&gid=1000: Failed to mount nfs share: mount/mnt call failed with "RPC error: Mount failed with error MNT3ERR_ACCES(13) Permission denied(13)"

I have to run as root or specify 'insecure' option in /etc/exports

~/git/qemu$ sudo ./qemu-img create -f qcow2 'nfs://127.0.0.1/home/lieven/git/qemu/test-nfs.img?uid=1000&gid=1000' 8G
Formatting 'nfs://127.0.0.1/home/lieven/git/qemu/test-nfs.img?uid=1000&gid=1000', fmt=qcow2 size=8589934592 encryption=off cluster_size=65536 lazy_refcounts=off

The last succeeds without problems.

Peter

#21 lore raw
In reply to Stefan Hajnoczi
Quoted 22 lines · Stefan Hajnoczi
Quoted 7 lines · Peter Lieven

This adds v6 of the NFS protocol driver + qemu-iotest adjustments.

v6->v7:
- removed unsused and partially non-existing libnfs header file includes [Stefan]
- adjusted usage of qemu_opts_create_nofail [Stefan]
- removed patch to fix output of qemu-iotest 067
- only blacklist qemu-iotest 020 for NFS do not change proto from generic to file [Fam]

It compiles now but I can't create an image on my /home NFS export:

$ ./qemu-img create -f qcow2 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000' 8G
Formatting 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000', fmt=qcow2 size=8589934592 encryption=off cluster_size=65536 lazy_refcounts=off
qemu-img: nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000: Failed to create file: nfs_creat_async failed

I added printfs to the code, block/nfs.c is mounting /home/stefanha/qemu and tries to create the file /test-nfs.img.

I'm using the Linux kernel NFS server. Unfortunately I don't have more time to debug this but I'd like to test this feature successfully before merging it into QEMU. Any ideas why libnfs refuses to create the file? (It's failing in nfs_lookuppath_async() with "Path is empty".)

Quoted 8 lines

Ronnie: libnfs clobbers the nfs->rpc error string so you cannot see the detailed error message:

In this case nfs_creat() clobbers the error with "Failed to create file:
nfs_creat_async failed". In fact the message was "Out of memory: failed to start parsing the path components" from nfs_creat_async(), but that's clobbering the true error too. It really should be "Path is empty" from nfs_lookuppath_async().

It seems the version in Ronnies repository got broken somewhere after my commit d43a8953f5. Can you try with:

https://github.com/plieven/libnfs

Thanks,
Peter

#22 lore raw
Quoted 31 lines · Peter Lieven
Quoted 22 lines · Stefan Hajnoczi
Quoted 7 lines · Peter Lieven

This adds v6 of the NFS protocol driver + qemu-iotest adjustments.

v6->v7:
- removed unsused and partially non-existing libnfs header file includes [Stefan]
- adjusted usage of qemu_opts_create_nofail [Stefan]
- removed patch to fix output of qemu-iotest 067
- only blacklist qemu-iotest 020 for NFS do not change proto from generic to file [Fam]

It compiles now but I can't create an image on my /home NFS export:

$ ./qemu-img create -f qcow2 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000' 8G
Formatting 'nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000', fmt=qcow2 size=8589934592 encryption=off cluster_size=65536 lazy_refcounts=off
qemu-img: nfs://127.0.0.1/home/stefanha/qemu/test-nfs.img?uid=1000&gid=1000: Failed to create file: nfs_creat_async failed

I added printfs to the code, block/nfs.c is mounting /home/stefanha/qemu and tries to create the file /test-nfs.img.

I'm using the Linux kernel NFS server. Unfortunately I don't have more time to debug this but I'd like to test this feature successfully before merging it into QEMU. Any ideas why libnfs refuses to create the file? (It's failing in nfs_lookuppath_async() with "Path is empty".)

Quoted 8 lines

Ronnie: libnfs clobbers the nfs->rpc error string so you cannot see the detailed error message:

In this case nfs_creat() clobbers the error with "Failed to create file:
nfs_creat_async failed". In fact the message was "Out of memory: failed to start parsing the path components" from nfs_creat_async(), but that's clobbering the true error too. It really should be "Path is empty" from nfs_lookuppath_async().

Quoted 4 lines

It seems the version in Ronnies repository got broken somewhere after my commit d43a8953f5. Can you try with:

https://github.com/plieven/libnfs

Ronnie, commit fb6f3152 "Add support for chdir and getcwd" is bad.

Peter