qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Daniel P. Berrange" <berrange@redhat.com>
To: qemu-devel@nongnu.org
Cc: Juan Quintela <quintela@redhat.com>,
	"Dr. David Alan Gilbert" <dgilbert@redhat.com>,
	Gerd Hoffmann <kraxel@redhat.com>,
	Amit Shah <amit.shah@redhat.com>,
	Paolo Bonzini <pbonzini@redhat.com>
Subject: [Qemu-devel] [PATCH FYI 41/46] migration: convert RDMA to use QIOChannel interface
Date: Thu,  3 Sep 2015 16:39:23 +0100	[thread overview]
Message-ID: <1441294768-8712-42-git-send-email-berrange@redhat.com> (raw)
In-Reply-To: <1441294768-8712-1-git-send-email-berrange@redhat.com>

This converts the RDMA code to provide a subclass of
QIOChannel that uses RDMA for the data transport.

The RDMA code would be much better off it it could
be split up in a generic RDMA layer, a QIOChannel
impl based on RMDA, and then the RMDA migration
glue. This is left as a future exercise for the brave.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
---
 migration/rdma.c | 254 ++++++++++++++++++++++++++++++++++---------------------
 1 file changed, 156 insertions(+), 98 deletions(-)

diff --git a/migration/rdma.c b/migration/rdma.c
index ca6b100..4e41c73 100644
--- a/migration/rdma.c
+++ b/migration/rdma.c
@@ -374,14 +374,19 @@ typedef struct RDMAContext {
     GHashTable *blockmap;
 } RDMAContext;
 
-/*
- * Interface to the rest of the migration call stack.
- */
-typedef struct QEMUFileRDMA {
+#define TYPE_QIO_CHANNEL_RDMA "qio-channel-rdma"
+#define QIO_CHANNEL_RDMA(obj)                                     \
+    OBJECT_CHECK(QIOChannelRDMA, (obj), TYPE_QIO_CHANNEL_RDMA)
+
+typedef struct QIOChannelRDMA QIOChannelRDMA;
+
+
+struct QIOChannelRDMA {
+    QIOChannel parent;
     RDMAContext *rdma;
+    QEMUFile *file;
     size_t len;
-    void *file;
-} QEMUFileRDMA;
+};
 
 /*
  * Main structure for IB Send/Recv control messages.
@@ -2519,15 +2524,19 @@ static void *qemu_rdma_data_init(const char *host_port, Error **errp)
  * SEND messages for control only.
  * VM's ram is handled with regular RDMA messages.
  */
-static int qemu_rdma_put_buffer(void *opaque, const uint8_t *buf,
-                                int64_t pos, int size)
-{
-    QEMUFileRDMA *r = opaque;
-    QEMUFile *f = r->file;
-    RDMAContext *rdma = r->rdma;
-    size_t remaining = size;
-    uint8_t * data = (void *) buf;
+static ssize_t qio_channel_rdma_writev(QIOChannel *ioc,
+                                       const struct iovec *iov,
+                                       size_t niov,
+                                       int *fds,
+                                       size_t nfds,
+                                       Error **errp)
+{
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(ioc);
+    QEMUFile *f = rioc->file;
+    RDMAContext *rdma = rioc->rdma;
     int ret;
+    ssize_t done = 0;
+    size_t i;
 
     CHECK_ERROR_STATE();
 
@@ -2541,26 +2550,31 @@ static int qemu_rdma_put_buffer(void *opaque, const uint8_t *buf,
         return ret;
     }
 
-    while (remaining) {
-        RDMAControlHeader head;
+    for (i = 0; i < niov; i++) {
+        size_t remaining = iov[i].iov_len;
+        uint8_t * data = (void *)iov[i].iov_base;
+        while (remaining) {
+            RDMAControlHeader head;
 
-        r->len = MIN(remaining, RDMA_SEND_INCREMENT);
-        remaining -= r->len;
+            rioc->len = MIN(remaining, RDMA_SEND_INCREMENT);
+            remaining -= rioc->len;
 
-        head.len = r->len;
-        head.type = RDMA_CONTROL_QEMU_FILE;
+            head.len = rioc->len;
+            head.type = RDMA_CONTROL_QEMU_FILE;
 
-        ret = qemu_rdma_exchange_send(rdma, &head, data, NULL, NULL, NULL);
+            ret = qemu_rdma_exchange_send(rdma, &head, data, NULL, NULL, NULL);
 
-        if (ret < 0) {
-            rdma->error_state = ret;
-            return ret;
-        }
+            if (ret < 0) {
+                rdma->error_state = ret;
+                return ret;
+            }
 
-        data += r->len;
+            data += rioc->len;
+            done += rioc->len;
+        }
     }
 
-    return size;
+    return done;
 }
 
 static size_t qemu_rdma_fill(RDMAContext *rdma, uint8_t *buf,
@@ -2585,41 +2599,65 @@ static size_t qemu_rdma_fill(RDMAContext *rdma, uint8_t *buf,
  * RDMA links don't use bytestreams, so we have to
  * return bytes to QEMUFile opportunistically.
  */
-static int qemu_rdma_get_buffer(void *opaque, uint8_t *buf,
-                                int64_t pos, int size)
-{
-    QEMUFileRDMA *r = opaque;
-    RDMAContext *rdma = r->rdma;
+static ssize_t qio_channel_rdma_readv(QIOChannel *ioc,
+                                      const struct iovec *iov,
+                                      size_t niov,
+                                      int **fds,
+                                      size_t *nfds,
+                                      Error **errp)
+{
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(ioc);
+    RDMAContext *rdma = rioc->rdma;
     RDMAControlHeader head;
     int ret = 0;
+    ssize_t i;
+    size_t done = 0;
 
     CHECK_ERROR_STATE();
 
-    /*
-     * First, we hold on to the last SEND message we
-     * were given and dish out the bytes until we run
-     * out of bytes.
-     */
-    r->len = qemu_rdma_fill(r->rdma, buf, size, 0);
-    if (r->len) {
-        return r->len;
-    }
+    for (i = 0; i < niov; i++) {
+        size_t want = iov[i].iov_len;
+        uint8_t *data = (void *)iov[i].iov_base;
 
-    /*
-     * Once we run out, we block and wait for another
-     * SEND message to arrive.
-     */
-    ret = qemu_rdma_exchange_recv(rdma, &head, RDMA_CONTROL_QEMU_FILE);
+        /*
+         * First, we hold on to the last SEND message we
+         * were given and dish out the bytes until we run
+         * out of bytes.
+         */
+        ret = qemu_rdma_fill(rioc->rdma, data, want, 0);
+        if (ret > 0) {
+            done += ret;
+            if (ret < want) {
+                break;
+            } else {
+                continue;
+            }
+        }
 
-    if (ret < 0) {
-        rdma->error_state = ret;
-        return ret;
-    }
+        /*
+         * Once we run out, we block and wait for another
+         * SEND message to arrive.
+         */
+        ret = qemu_rdma_exchange_recv(rdma, &head, RDMA_CONTROL_QEMU_FILE);
 
-    /*
-     * SEND was received with new bytes, now try again.
-     */
-    return qemu_rdma_fill(r->rdma, buf, size, 0);
+        if (ret < 0) {
+            rdma->error_state = ret;
+            return ret;
+        }
+
+        /*
+         * SEND was received with new bytes, now try again.
+         */
+        ret = qemu_rdma_fill(rioc->rdma, data, want, 0);
+        if (ret > 0) {
+            done += ret;
+            if (ret < want) {
+                break;
+            }
+        }
+    }
+    rioc->len = done;
+    return rioc->len;
 }
 
 /*
@@ -2646,15 +2684,16 @@ static int qemu_rdma_drain_cq(QEMUFile *f, RDMAContext *rdma)
     return 0;
 }
 
-static int qemu_rdma_close(void *opaque)
+static int qio_channel_rdma_close(QIOChannel *ioc,
+                                  Error **errp)
 {
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(ioc);
     trace_qemu_rdma_close();
-    QEMUFileRDMA *r = opaque;
-    if (r->rdma) {
-        qemu_rdma_cleanup(r->rdma);
-        g_free(r->rdma);
+    if (rioc->rdma) {
+        qemu_rdma_cleanup(rioc->rdma);
+        g_free(rioc->rdma);
+        rioc->rdma = NULL;
     }
-    g_free(r);
     return 0;
 }
 
@@ -2696,8 +2735,8 @@ static size_t qemu_rdma_save_page(QEMUFile *f, void *opaque,
                                   ram_addr_t block_offset, ram_addr_t offset,
                                   size_t size, uint64_t *bytes_sent)
 {
-    QEMUFileRDMA *rfile = opaque;
-    RDMAContext *rdma = rfile->rdma;
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(opaque);
+    RDMAContext *rdma = rioc->rdma;
     int ret;
 
     CHECK_ERROR_STATE();
@@ -2951,8 +2990,8 @@ static int qemu_rdma_registration_handle(QEMUFile *f, void *opaque)
                              };
     RDMAControlHeader blocks = { .type = RDMA_CONTROL_RAM_BLOCKS_RESULT,
                                  .repeat = 1 };
-    QEMUFileRDMA *rfile = opaque;
-    RDMAContext *rdma = rfile->rdma;
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(opaque);
+    RDMAContext *rdma = rioc->rdma;
     RDMALocalBlocks *local = &rdma->local_ram_blocks;
     RDMAControlHeader head;
     RDMARegister *reg, *registers;
@@ -3207,9 +3246,9 @@ out:
  * We've already built our local RAMBlock list, but not yet sent the list to
  * the source.
  */
-static int rdma_block_notification_handle(QEMUFileRDMA *rfile, const char *name)
+static int rdma_block_notification_handle(QIOChannelRDMA *rioc, const char *name)
 {
-    RDMAContext *rdma = rfile->rdma;
+    RDMAContext *rdma = rioc->rdma;
     int curr;
     int found = -1;
 
@@ -3251,8 +3290,8 @@ static int rdma_load_hook(QEMUFile *f, void *opaque, uint64_t flags, void *data)
 static int qemu_rdma_registration_start(QEMUFile *f, void *opaque,
                                         uint64_t flags, void *data)
 {
-    QEMUFileRDMA *rfile = opaque;
-    RDMAContext *rdma = rfile->rdma;
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(opaque);
+    RDMAContext *rdma = rioc->rdma;
 
     CHECK_ERROR_STATE();
 
@@ -3271,8 +3310,8 @@ static int qemu_rdma_registration_stop(QEMUFile *f, void *opaque,
                                        uint64_t flags, void *data)
 {
     Error *local_err = NULL, **errp = &local_err;
-    QEMUFileRDMA *rfile = opaque;
-    RDMAContext *rdma = rfile->rdma;
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(opaque);
+    RDMAContext *rdma = rioc->rdma;
     RDMAControlHeader head = { .len = 0, .repeat = 1 };
     int ret = 0;
 
@@ -3368,55 +3407,74 @@ err:
     return ret;
 }
 
-static int qemu_rdma_get_fd(void *opaque)
-{
-    QEMUFileRDMA *rfile = opaque;
-    RDMAContext *rdma = rfile->rdma;
-
-    return rdma->comp_channel->fd;
-}
-
-static const QEMUFileOps rdma_read_ops = {
-    .get_buffer    = qemu_rdma_get_buffer,
-    .get_fd        = qemu_rdma_get_fd,
-    .close         = qemu_rdma_close,
-};
-
 static const QEMUFileHooks rdma_read_hooks = {
     .hook_ram_load = rdma_load_hook,
 };
 
-static const QEMUFileOps rdma_write_ops = {
-    .put_buffer         = qemu_rdma_put_buffer,
-    .close              = qemu_rdma_close,
-};
-
 static const QEMUFileHooks rdma_write_hooks = {
     .before_ram_iterate = qemu_rdma_registration_start,
     .after_ram_iterate  = qemu_rdma_registration_stop,
     .save_page          = qemu_rdma_save_page,
 };
 
-static void *qemu_fopen_rdma(RDMAContext *rdma, const char *mode)
+
+static void qio_channel_rdma_finalize(Object *obj)
+{
+    QIOChannelRDMA *rioc = QIO_CHANNEL_RDMA(obj);
+    if (rioc->rdma) {
+        qemu_rdma_cleanup(rioc->rdma);
+        g_free(rioc->rdma);
+        rioc->rdma = NULL;
+    }
+}
+
+static void qio_channel_rdma_class_init(ObjectClass *klass,
+                                        void *class_data G_GNUC_UNUSED)
+{
+    QIOChannelClass *ioc_klass = QIO_CHANNEL_CLASS(klass);
+
+    ioc_klass->io_writev = qio_channel_rdma_writev;
+    ioc_klass->io_readv = qio_channel_rdma_readv;
+    //ioc_klass->io_set_blocking = qio_channel_rdma_set_blocking;
+    ioc_klass->io_close = qio_channel_rdma_close;
+    //ioc_klass->io_create_watch = qio_channel_rdma_create_watch;
+}
+
+static const TypeInfo qio_channel_rdma_info = {
+    .parent = TYPE_QIO_CHANNEL,
+    .name = TYPE_QIO_CHANNEL_RDMA,
+    .instance_size = sizeof(QIOChannelRDMA),
+    .instance_finalize = qio_channel_rdma_finalize,
+    .class_init = qio_channel_rdma_class_init,
+};
+
+static void qio_channel_rdma_register_types(void)
+{
+    type_register_static(&qio_channel_rdma_info);
+}
+
+type_init(qio_channel_rdma_register_types);
+
+static QEMUFile *qemu_fopen_rdma(RDMAContext *rdma, const char *mode)
 {
-    QEMUFileRDMA *r;
+    QIOChannelRDMA *rioc;
 
     if (qemu_file_mode_is_not_valid(mode)) {
         return NULL;
     }
 
-    r = g_malloc0(sizeof(QEMUFileRDMA));
-    r->rdma = rdma;
+    rioc = QIO_CHANNEL_RDMA(object_new(TYPE_QIO_CHANNEL_RDMA));
+    rioc->rdma = rdma;
 
     if (mode[0] == 'w') {
-        r->file = qemu_fopen_ops(r, &rdma_write_ops);
-        qemu_file_set_hooks(r->file, &rdma_write_hooks);
+        rioc->file = qemu_fopen_channel_output(QIO_CHANNEL(rioc));
+        qemu_file_set_hooks(rioc->file, &rdma_write_hooks);
     } else {
-        r->file = qemu_fopen_ops(r, &rdma_read_ops);
-        qemu_file_set_hooks(r->file, &rdma_read_hooks);
+        rioc->file = qemu_fopen_channel_input(QIO_CHANNEL(rioc));
+        qemu_file_set_hooks(rioc->file, &rdma_read_hooks);
     }
 
-    return r->file;
+    return rioc->file;
 }
 
 static void rdma_accept_incoming_migration(void *opaque)
-- 
2.4.3

  parent reply	other threads:[~2015-09-03 15:41 UTC|newest]

Thread overview: 57+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-09-03 15:38 [Qemu-devel] [PATCH FYI 00/46] Generic TLS support across VNC/chardev/migration Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 01/46] sockets: add helpers for creating SocketAddress from a socket Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 02/46] sockets: move qapi_copy_SocketAddress into qemu-sockets.c Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 03/46] sockets: allow port to be NULL when listening on IP address Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 04/46] osdep: add qemu_fork() wrapper for safely handling signals Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 05/46] coroutine: move into libqemuutil.a library Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 06/46] io: add abstract QIOChannel classes Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 07/46] io: add helper module for creating watches on FDs Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 08/46] io: pull Buffer code out of VNC module Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 09/46] io: add QIOTask class for async operations Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 10/46] io: add QIOChannelSocket class Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 11/46] io: add QIOChannelFile class Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 12/46] io: add QIOChannelTLS class Daniel P. Berrange
2015-09-07 15:31   ` Dr. David Alan Gilbert
2015-09-07 15:41     ` Daniel P. Berrange
2015-09-07 15:51       ` Dr. David Alan Gilbert
2015-09-07 16:04         ` Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 13/46] io: add QIOChannelWebsock class Daniel P. Berrange
2015-09-07 15:44   ` Dr. David Alan Gilbert
2015-09-07 15:50     ` Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 14/46] io: add QIOChannelCommand class Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 15/46] ui: convert VNC startup code to use SocketAddress Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 16/46] ui: convert VNC server to use QIOChannelSocket Daniel P. Berrange
2015-09-03 15:38 ` [Qemu-devel] [PATCH FYI 17/46] ui: convert VNC server to use QIOChannelTLS Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 18/46] ui: convert VNC server to use QIOChannelWebsock Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 19/46] char: remove fixed length filename allocation Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 20/46] char: convert from GIOChannel to QIOChannel Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 21/46] char: don't assume telnet initialization will not block Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 22/46] char: introduce support for TLS encrypted TCP chardev backend Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 23/46] nbd: convert to use the QAPI SocketAddress object Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 24/46] qemu-nbd: " Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 25/46] sockets: remove use of QemuOpts from header file Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 26/46] sockets: remove use of QemuOpts from socket_listen Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 27/46] sockets: remove use of QemuOpts from socket_connect Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 28/46] sockets: remove use of QemuOpts from socket_dgram Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 29/46] migration: remove use of qemu_bufopen from vmstate tests Daniel P. Berrange
2015-09-07 16:08   ` Dr. David Alan Gilbert
2015-09-07 16:17     ` Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 30/46] migration: remove memory buffer based QEMUFile backend Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 31/46] migration: move definition of struct QEMUFile back into qemu-file.c Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 32/46] migration: split migration hooks out of QEMUFileOps Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 33/46] migration: ensure qemu_fflush() always writes full data amount Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 34/46] migration: introduce qemu_fset_blocking function on QEMUFile Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 35/46] migration: force QEMUFile to blocking mode for outgoing migration Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 36/46] migration: introduce a new QEMUFile impl based on QIOChannel Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 37/46] migration: convert unix socket protocol to use QIOChannel Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 38/46] migration: convert tcp " Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 39/46] migration: convert fd " Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 40/46] migration: convert exec " Daniel P. Berrange
2015-09-03 15:39 ` Daniel P. Berrange [this message]
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 42/46] migration: convert savevm to use QIOChannel for writing to files Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 43/46] migration: delete QEMUFile sockets implementation Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 44/46] migration: delete QEMUFile stdio implementation Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 45/46] migration: support TLS encryption with TCP migration backend Daniel P. Berrange
2015-09-07 16:23   ` Dr. David Alan Gilbert
2015-09-07 16:29     ` Daniel P. Berrange
2015-09-03 15:39 ` [Qemu-devel] [PATCH FYI 46/46] migration: remove support for non-iovec based write handlers Daniel P. Berrange

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1441294768-8712-42-git-send-email-berrange@redhat.com \
    --to=berrange@redhat.com \
    --cc=amit.shah@redhat.com \
    --cc=dgilbert@redhat.com \
    --cc=kraxel@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=quintela@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).