[Qemu-devel] [PATCH 10/10] scsi: add persistent reservation manager using qemu-pr-helper

[Paolo Bonzini <pbonzini@redhat.com>]

This adds a concrete subclass of pr-manager that talks to qemu-pr-helper.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 scsi/Makefile.objs       |   2 +-
 scsi/pr-manager-helper.c | 288 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 289 insertions(+), 1 deletion(-)
 create mode 100644 scsi/pr-manager-helper.c

diff --git a/scsi/Makefile.objs b/scsi/Makefile.objs
index 5496d2ae6a..4d25e476cf 100644
--- a/scsi/Makefile.objs
+++ b/scsi/Makefile.objs
@@ -1,3 +1,3 @@
 block-obj-y += utils.o
 
-block-obj-$(CONFIG_LINUX) += pr-manager.o
+block-obj-$(CONFIG_LINUX) += pr-manager.o pr-manager-helper.o
diff --git a/scsi/pr-manager-helper.c b/scsi/pr-manager-helper.c
new file mode 100644
index 0000000000..c9d9606696
--- /dev/null
+++ b/scsi/pr-manager-helper.c
@@ -0,0 +1,288 @@
+/*
+ * Persistent reservation manager that talks to qemu-mpath-helper
+ *
+ * Copyright (c) 2017 Red Hat, Inc.
+ *
+ * Author: Paolo Bonzini <pbonzini@redhat.com>
+ *
+ * This code is licensed under the LGPL.
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qapi/error.h"
+#include "scsi/constants.h"
+#include "scsi/pr-manager.h"
+#include "scsi/utils.h"
+#include "io/channel.h"
+#include "io/channel-socket.h"
+#include "pr-helper.h"
+
+#include <scsi/sg.h>
+
+#define PR_MAX_RECONNECT_ATTEMPTS 5
+
+#define TYPE_PR_MANAGER_HELPER "pr-manager-helper"
+
+#define PR_MANAGER_HELPER(obj) \
+     INTERFACE_CHECK(PRManagerHelper, (obj), \
+                     TYPE_PR_MANAGER_HELPER)
+
+typedef struct PRManagerHelper {
+    /* <private> */
+    PRManager parent;
+
+    char *path;
+
+    QemuMutex lock;
+    QIOChannel *ioc;
+} PRManagerHelper;
+
+/* Called with lock held.  */
+static int pr_manager_helper_read(PRManagerHelper *pr_mgr,
+                                  void *buf, int sz, Error **errp)
+{
+    ssize_t r = qio_channel_read_all(pr_mgr->ioc, buf, sz, errp);
+
+    if (r < 0) {
+        object_unref(OBJECT(pr_mgr->ioc));
+        pr_mgr->ioc = NULL;
+        return r;
+    }
+
+    return r < 0 ? r : 0;
+}
+
+/* Called with lock held.  */
+static int pr_manager_helper_write(PRManagerHelper *pr_mgr,
+                                   int fd,
+                                   const void *buf, int sz, Error **errp)
+{
+    size_t nfds = (fd != -1);
+    while (sz > 0) {
+        struct iovec iov;
+        ssize_t n_written;
+
+        iov.iov_base = (void *)buf;
+        iov.iov_len = sz;
+        n_written = qio_channel_writev_full(QIO_CHANNEL(pr_mgr->ioc), &iov, 1,
+                                            nfds ? &fd : NULL, nfds, errp);
+
+        if (n_written <= 0) {
+            assert(n_written != QIO_CHANNEL_ERR_BLOCK);
+            object_unref(OBJECT(pr_mgr->ioc));
+            pr_mgr->ioc = NULL;
+            return n_written;
+        }
+
+        nfds = 0;
+        buf += n_written;
+        sz -= n_written;
+    }
+
+    return 0;
+}
+
+/* Called with lock held.  */
+static int pr_manager_helper_initialize(PRManagerHelper *pr_mgr,
+                                        Error **errp)
+{
+    uint32_t flags;
+
+    SocketAddress saddr = {
+        .type = SOCKET_ADDRESS_TYPE_UNIX,
+        .u.q_unix.path = g_strdup(pr_mgr->path)
+    };
+    QIOChannelSocket *sioc = qio_channel_socket_new();
+    Error *local_err = NULL;
+
+    qio_channel_set_name(QIO_CHANNEL(sioc), "pr-manager-helper");
+    qio_channel_socket_connect_sync(sioc,
+                                    &saddr,
+                                    &local_err);
+    if (local_err) {
+        object_unref(OBJECT(sioc));
+        error_propagate(errp, local_err);
+        return -ENOTCONN;
+    }
+
+    qio_channel_set_delay(QIO_CHANNEL(sioc), false);
+    pr_mgr->ioc = QIO_CHANNEL(sioc);
+
+    /* A simple feature negotation protocol, even though there is
+     * no optional feature right now.
+     */
+    if (pr_manager_helper_read(pr_mgr, &flags, sizeof(flags), errp) < 0) {
+        return -EINVAL;
+    }
+
+    flags = 0;
+    if (pr_manager_helper_write(pr_mgr, -1, &flags, sizeof(flags), errp) < 0) {
+        return -EINVAL;
+    }
+
+    return 0;
+}
+
+static int pr_manager_helper_run(PRManager *p,
+                                 int fd, struct sg_io_hdr *io_hdr)
+{
+    PRManagerHelper *pr_mgr = PR_MANAGER_HELPER(p);
+
+    uint32_t len;
+    PRHelperResponse resp;
+    int ret;
+    int expected_dir;
+    int attempts;
+    uint8_t cdb[PR_HELPER_CDB_SIZE] = { 0 };
+
+    if (!io_hdr->cmd_len || io_hdr->cmd_len > PR_HELPER_CDB_SIZE) {
+        return -EINVAL;
+    }
+
+    memcpy(cdb, io_hdr->cmdp, io_hdr->cmd_len);
+    assert(cdb[0] == PERSISTENT_RESERVE_OUT || cdb[0] == PERSISTENT_RESERVE_IN);
+    expected_dir =
+        (cdb[0] == PERSISTENT_RESERVE_OUT ? SG_DXFER_TO_DEV : SG_DXFER_FROM_DEV);
+    if (io_hdr->dxfer_direction != expected_dir) {
+        return -EINVAL;
+    }
+
+    len = scsi_cdb_xfer(cdb);
+    if (io_hdr->dxfer_len < len || len > PR_HELPER_DATA_SIZE) {
+        return -EINVAL;
+    }
+
+    ret = 0;
+    qemu_mutex_lock(&pr_mgr->lock);
+
+    /* Try to reconnect while sending the CDB.  */
+    for (attempts = 0; attempts < PR_MAX_RECONNECT_ATTEMPTS; attempts++) {
+        if (!pr_mgr->ioc) {
+            if (pr_manager_helper_initialize(pr_mgr, NULL) < 0) {
+                qemu_mutex_unlock(&pr_mgr->lock);
+                g_usleep(G_USEC_PER_SEC);
+                qemu_mutex_lock(&pr_mgr->lock);
+            }
+        }
+
+        if (pr_mgr->ioc) {
+            if (pr_manager_helper_write(pr_mgr, fd, cdb,
+                                        ARRAY_SIZE(cdb), NULL) >= 0) {
+                break;
+            }
+        }
+    }
+    if (attempts == PR_MAX_RECONNECT_ATTEMPTS) {
+        ret = -EINVAL;
+        goto out;
+    }
+
+    /* After sending the CDB, any communications failure causes the
+     * command to fail.  The failure is transient, retrying the command
+     * will invoke pr_manager_helper_initialize again.
+     */
+    if (expected_dir == SG_DXFER_TO_DEV) {
+        if (pr_manager_helper_write(pr_mgr, -1, io_hdr->dxferp,
+                                    len, NULL) < 0) {
+            ret = -EINVAL;
+            goto out;
+        }
+    }
+    if (pr_manager_helper_read(pr_mgr, &resp, sizeof(resp), NULL) < 0) {
+        ret = -EINVAL;
+        goto out;
+    }
+    if (expected_dir == SG_DXFER_FROM_DEV && resp.result == 0) {
+        if (pr_manager_helper_read(pr_mgr, io_hdr->dxferp, len, NULL) < 0) {
+            ret = -EINVAL;
+            goto out;
+        }
+    }
+
+    io_hdr->status = resp.result;
+    if (resp.result == CHECK_CONDITION) {
+        io_hdr->driver_status = SG_ERR_DRIVER_SENSE;
+        io_hdr->sb_len_wr = MIN(io_hdr->mx_sb_len, PR_HELPER_SENSE_SIZE);
+        memcpy(io_hdr->sbp, resp.sense, io_hdr->sb_len_wr);
+    }
+
+out:
+    if (ret < 0) {
+        int sense_len = scsi_build_sense(io_hdr->sbp,
+                                         SENSE_CODE(LUN_COMM_FAILURE));
+        io_hdr->driver_status = SG_ERR_DRIVER_SENSE;
+        io_hdr->sb_len_wr = MIN(io_hdr->mx_sb_len, sense_len);
+        io_hdr->status = CHECK_CONDITION;
+    }
+    qemu_mutex_unlock(&pr_mgr->lock);
+    return ret;
+}
+
+static void pr_manager_helper_complete(UserCreatable *uc, Error **errp)
+{
+    PRManagerHelper *pr_mgr = PR_MANAGER_HELPER(uc);
+
+    qemu_mutex_lock(&pr_mgr->lock);
+    pr_manager_helper_initialize(pr_mgr, errp);
+    qemu_mutex_unlock(&pr_mgr->lock);
+}
+
+static char *get_path(Object *obj, Error **errp)
+{
+    PRManagerHelper *pr_mgr = PR_MANAGER_HELPER(obj);
+
+    return g_strdup(pr_mgr->path);
+}
+
+static void set_path(Object *obj, const char *str, Error **errp)
+{
+    PRManagerHelper *pr_mgr = PR_MANAGER_HELPER(obj);
+
+    g_free(pr_mgr->path);
+    pr_mgr->path = g_strdup(str);
+}
+
+static void pr_manager_helper_instance_finalize(Object *obj)
+{
+    PRManagerHelper *pr_mgr = PR_MANAGER_HELPER(obj);
+
+    g_free(pr_mgr->path);
+    object_unref(OBJECT(pr_mgr->ioc));
+    qemu_mutex_destroy(&pr_mgr->lock);
+}
+
+static void pr_manager_helper_instance_init(Object *obj)
+{
+    PRManagerHelper *pr_mgr = PR_MANAGER_HELPER(obj);
+
+    qemu_mutex_init(&pr_mgr->lock);
+}
+
+static void pr_manager_helper_class_init(ObjectClass *klass,
+                                         void *class_data G_GNUC_UNUSED)
+{
+    PRManagerClass *prmgr_klass = PR_MANAGER_CLASS(klass);
+    UserCreatableClass *uc_klass = USER_CREATABLE_CLASS(klass);
+
+    object_class_property_add_str(klass, "path", get_path, set_path,
+                                  &error_abort);
+    uc_klass->complete = pr_manager_helper_complete;
+    prmgr_klass->run = pr_manager_helper_run;
+}
+
+static const TypeInfo pr_manager_helper_info = {
+    .parent = TYPE_PR_MANAGER,
+    .name = TYPE_PR_MANAGER_HELPER,
+    .instance_size = sizeof(PRManagerHelper),
+    .instance_init = pr_manager_helper_instance_init,
+    .instance_finalize = pr_manager_helper_instance_finalize,
+    .class_init = pr_manager_helper_class_init,
+};
+
+static void pr_manager_helper_register_types(void)
+{
+    type_register_static(&pr_manager_helper_info);
+}
+
+type_init(pr_manager_helper_register_types);
-- 
2.13.5