Add qemu 2.4.0
[kvmfornfv.git] / qemu / block / nfs.c
diff --git a/qemu/block/nfs.c b/qemu/block/nfs.c
new file mode 100644 (file)
index 0000000..c026ff6
--- /dev/null
@@ -0,0 +1,516 @@
+/*
+ * QEMU Block driver for native access to files on NFS shares
+ *
+ * Copyright (c) 2014 Peter Lieven <pl@kamp.de>
+ *
+ * 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>
+
+#define QEMU_NFS_MAX_READAHEAD_SIZE 1048576
+
+typedef struct NFSClient {
+    struct nfs_context *context;
+    struct nfsfh *fh;
+    int events;
+    bool has_zero_init;
+    AioContext *aio_context;
+} NFSClient;
+
+typedef struct NFSRPC {
+    int ret;
+    int complete;
+    QEMUIOVector *iov;
+    struct stat *st;
+    Coroutine *co;
+    QEMUBH *bh;
+    NFSClient *client;
+} 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) {
+        aio_set_fd_handler(client->aio_context,
+                           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(),
+        .client         = client,
+    };
+}
+
+static void nfs_co_generic_bh_cb(void *opaque)
+{
+    NFSRPC *task = opaque;
+    task->complete = 1;
+    qemu_bh_delete(task->bh);
+    qemu_coroutine_enter(task->co, NULL);
+}
+
+static void
+nfs_co_generic_cb(int ret, struct nfs_context *nfs, void *data,
+                  void *private_data)
+{
+    NFSRPC *task = private_data;
+    task->ret = ret;
+    if (task->ret > 0 && task->iov) {
+        if (task->ret <= task->iov->size) {
+            qemu_iovec_from_buf(task->iov, 0, data, task->ret);
+        } else {
+            task->ret = -EIO;
+        }
+    }
+    if (task->ret == 0 && task->st) {
+        memcpy(task->st, data, sizeof(struct stat));
+    }
+    if (task->ret < 0) {
+        error_report("NFS Error: %s", nfs_get_error(nfs));
+    }
+    if (task->co) {
+        task->bh = aio_bh_new(task->client->aio_context,
+                              nfs_co_generic_bh_cb, task);
+        qemu_bh_schedule(task->bh);
+    } else {
+        task->complete = 1;
+    }
+}
+
+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.ret < 0) {
+        return task.ret;
+    }
+
+    /* zero pad short reads */
+    if (task.ret < iov->size) {
+        qemu_iovec_memset(iov, task.ret, 0, iov->size - task.ret);
+    }
+
+    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_try_malloc(nb_sectors * BDRV_SECTOR_SIZE);
+    if (nb_sectors && buf == NULL) {
+        return -ENOMEM;
+    }
+
+    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.ret != nb_sectors * BDRV_SECTOR_SIZE) {
+        return task.ret < 0 ? task.ret : -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.ret;
+}
+
+/* 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_detach_aio_context(BlockDriverState *bs)
+{
+    NFSClient *client = bs->opaque;
+
+    aio_set_fd_handler(client->aio_context,
+                       nfs_get_fd(client->context),
+                       NULL, NULL, NULL);
+    client->events = 0;
+}
+
+static void nfs_attach_aio_context(BlockDriverState *bs,
+                                   AioContext *new_context)
+{
+    NFSClient *client = bs->opaque;
+
+    client->aio_context = new_context;
+    nfs_set_events(client);
+}
+
+static void nfs_client_close(NFSClient *client)
+{
+    if (client->context) {
+        if (client->fh) {
+            nfs_close(client->context, client->fh);
+        }
+        aio_set_fd_handler(client->aio_context,
+                           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_client_open(NFSClient *client, const char *filename,
+                               int flags, Error **errp)
+{
+    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;
+    }
+    if (!uri->server) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;
+    }
+    strp = strrchr(uri->path, '/');
+    if (strp == NULL) {
+        error_setg(errp, "Invalid URL specified");
+        goto fail;
+    }
+    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++) {
+        unsigned long long val;
+        if (!qp->p[i].value) {
+            error_setg(errp, "Value for NFS parameter expected: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (parse_uint_full(qp->p[i].value, &val, 0)) {
+            error_setg(errp, "Illegal value for NFS parameter: %s",
+                       qp->p[i].name);
+            goto fail;
+        }
+        if (!strcmp(qp->p[i].name, "uid")) {
+            nfs_set_uid(client->context, val);
+        } else if (!strcmp(qp->p[i].name, "gid")) {
+            nfs_set_gid(client->context, val);
+        } else if (!strcmp(qp->p[i].name, "tcp-syncnt")) {
+            nfs_set_tcp_syncnt(client->context, val);
+#ifdef LIBNFS_FEATURE_READAHEAD
+        } else if (!strcmp(qp->p[i].name, "readahead")) {
+            if (val > QEMU_NFS_MAX_READAHEAD_SIZE) {
+                error_report("NFS Warning: Truncating NFS readahead"
+                             " size to %d", QEMU_NFS_MAX_READAHEAD_SIZE);
+                val = QEMU_NFS_MAX_READAHEAD_SIZE;
+            }
+            nfs_set_readahead(client->context, val);
+#endif
+        } 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,
+                         Error **errp) {
+    NFSClient *client = bs->opaque;
+    int64_t ret;
+    QemuOpts *opts;
+    Error *local_err = NULL;
+
+    client->aio_context = bdrv_get_aio_context(bs);
+
+    opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        ret = -EINVAL;
+        goto out;
+    }
+    ret = nfs_client_open(client, qemu_opt_get(opts, "filename"),
+                          (flags & BDRV_O_RDWR) ? O_RDWR : O_RDONLY,
+                          errp);
+    if (ret < 0) {
+        goto out;
+    }
+    bs->total_sectors = ret;
+    ret = 0;
+out:
+    qemu_opts_del(opts);
+    return ret;
+}
+
+static QemuOptsList nfs_create_opts = {
+    .name = "nfs-create-opts",
+    .head = QTAILQ_HEAD_INITIALIZER(nfs_create_opts.head),
+    .desc = {
+        {
+            .name = BLOCK_OPT_SIZE,
+            .type = QEMU_OPT_SIZE,
+            .help = "Virtual disk size"
+        },
+        { /* end of list */ }
+    }
+};
+
+static int nfs_file_create(const char *url, QemuOpts *opts, Error **errp)
+{
+    int ret = 0;
+    int64_t total_size = 0;
+    NFSClient *client = g_new0(NFSClient, 1);
+
+    client->aio_context = qemu_get_aio_context();
+
+    /* Read out options */
+    total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
+                          BDRV_SECTOR_SIZE);
+
+    ret = nfs_client_open(client, url, O_CREAT, errp);
+    if (ret < 0) {
+        goto out;
+    }
+    ret = nfs_ftruncate(client->context, client->fh, total_size);
+    nfs_client_close(client);
+out:
+    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);
+        aio_poll(client->aio_context, true);
+    }
+
+    return (task.ret < 0 ? task.ret : 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,
+    .create_opts                    = &nfs_create_opts,
+
+    .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,
+
+    .bdrv_detach_aio_context        = nfs_detach_aio_context,
+    .bdrv_attach_aio_context        = nfs_attach_aio_context,
+};
+
+static void nfs_block_init(void)
+{
+    bdrv_register(&bdrv_nfs);
+}
+
+block_init(nfs_block_init);