0001-got-a-first-xensocket-functioality-that-compiles-but.patch

From 97b4f22565ce0aed81c633295c30a591051e37ee Mon Sep 17 00:00:00 2001
From: Haris Rotsos <[email protected]>
Date: Tue, 14 May 2019 21:51:34 +0000
Subject: [PATCH 1/2] got a first xensocket functioality that compiles but
 still needs work to complete

---
 Makefile            |    9 +-
 include/xensocket.h |  102 +++++
 stub.mk             |    1 +
 xensocket.c         | 1030 +++++++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 1138 insertions(+), 4 deletions(-)
 create mode 100644 include/xensocket.h
 create mode 100644 xensocket.c

diff --git a/Makefile b/Makefile
index 360af84..25d5bfe 100644
--- a/Makefile
+++ b/Makefile
@@ -64,8 +64,8 @@ include minios.mk
 # CFLAGS += -DBLKTEST_WRITE
 
 # Define some default flags for linking.
-LDLIBS := 
-APP_LDLIBS := 
+LDLIBS :=
+APP_LDLIBS :=
 LDARCHLIB := -L$(OBJ_DIR)/$(TARGET_ARCH_DIR) -l$(ARCH_LIB_NAME)
 LDFLAGS_FINAL := -T $(TARGET_ARCH_DIR)/minios-$(XEN_TARGET_ARCH).lds
 
@@ -93,6 +93,7 @@ src-y += kernel.c
 src-y += lock.c
 src-y += main.c
 src-y += mm.c
+src-y += xensocket.c
 src-$(CONFIG_NETFRONT) += netfront.c
 src-$(CONFIG_PCIFRONT) += pcifront.c
 src-y += sched.c
@@ -179,7 +180,7 @@ $(OBJ_DIR)/$(TARGET)_app.o: $(APP_OBJS) app.lds
 	$(LD) -r -d $(LDFLAGS) -\( $^ -\) $(APP_LDLIBS) --undefined main -o $@
 
 ifneq ($(APP_OBJS),)
-APP_O=$(OBJ_DIR)/$(TARGET)_app.o 
+APP_O=$(OBJ_DIR)/$(TARGET)_app.o
 endif
 
 $(OBJ_DIR)/$(TARGET): links include/list.h $(OBJS) $(APP_O) arch_lib
@@ -212,7 +213,7 @@ endef
 cscope:
 	$(all_sources) > cscope.files
 	cscope -k -b -q
-    
+
 .PHONY: tags
 tags:
 	$(all_sources) | xargs ctags
diff --git a/include/xensocket.h b/include/xensocket.h
new file mode 100644
index 0000000..3e97fcf
--- /dev/null
+++ b/include/xensocket.h
@@ -0,0 +1,102 @@
+/* xensocket.h
+ *
+ * Header file for shared-memory sockets transport for communications
+ * between two domains on the same machine, under the Xen hypervisor.
+ *
+ * Authors: Xiaolan (Catherine) Zhang <[email protected]>
+ *          Suzanne McIntosh <[email protected]>
+ *          John Griffin
+ *
+ * History:
+ *          Suzanne McIntosh    13-Aug-07     Initial open source version
+ *
+ * Copyright (c) 2007, IBM Corporation
+ *
+ */
+
+#ifndef __XENSOCKET_H__
+#define __XENSOCKET_H__
+
+struct descriptor_page;
+struct xen_sock;
+
+static void
+initialize_xen_sock (struct xen_sock *sock);
+
+static int
+xen_bind (struct xen_sock *sock, uint16_t);
+
+static int
+server_allocate_descriptor_page (struct xen_sock *x);
+
+static int
+server_allocate_event_channel (struct xen_sock *x);
+
+static int
+server_allocate_buffer_pages (struct xen_sock *x);
+
+static int
+xen_connect (struct xen_sock *, uint16_t, int);
+
+static int
+client_map_descriptor_page (struct xen_sock *x);
+
+static int
+client_bind_event_channel (struct xen_sock *x);
+
+static int
+client_map_buffer_pages (struct xen_sock *x);
+
+static int
+xen_sendmsg (struct xen_sock *sock, void *data, size_t len);
+
+static inline int
+is_writeable (struct descriptor_page *d);
+
+//static long
+//send_data_wait (struct sock *sk, long timeo);
+
+static int
+xen_recvmsg (struct xen_sock *, void *, size_t);
+
+static inline int
+is_readable (struct descriptor_page *d);
+
+static long
+receive_data_wait (struct xen_sock *, long);
+
+/*static irqreturn_t
+server_interrupt (int irq, void *dev_id, struct pt_regs *regs); */
+
+//static int
+// local_memcpy_toiovecend (struct iovec *iov, unsigned char *kdata, int offset, int len);
+
+static int
+xen_release (struct xen_sock *sock);
+
+static int
+xen_shutdown (struct xen_sock *sock, int how);
+
+static void
+server_unallocate_buffer_pages (struct xen_sock *x);
+
+static void
+server_unallocate_descriptor_page (struct xen_sock *x);
+
+static void
+client_unmap_buffer_pages (struct xen_sock *x);
+
+static void
+client_unmap_descriptor_page (struct xen_sock *x);
+
+/* static int __init
+xensocket_init (void);
+
+static void __exit
+xensocket_exit (void); */
+
+
+
+#define xen_sk(__sk) ((struct xen_sock *)__sk)
+
+#endif /* __XENSOCKET_H__ */
diff --git a/stub.mk b/stub.mk
index 240aa95..373fa46 100644
--- a/stub.mk
+++ b/stub.mk
@@ -177,6 +177,7 @@ MINIOS_OBJS0-y		:=	\
 	string.o			\
 	sys.o				\
 	xencons_ring.o		\
+	xensocket.o			\
 	xmalloc.o
 MINIOS_OBJS0-$(CONFIG_XENBUS)		+= xenbus.o
 MINIOS_OBJS0-$(CONFIG_XENBUS)		+= xs.o
diff --git a/xensocket.c b/xensocket.c
new file mode 100644
index 0000000..f4ce035
--- /dev/null
+++ b/xensocket.c
@@ -0,0 +1,1030 @@
+/* xensocket.c
+ *
+ * XVMSocket module for a shared-memory sockets transport for communications
+ * between two domains on the same machine, under the Xen hypervisor.
+ *
+ * Authors: Xiaolan (Catherine) Zhang <[email protected]>
+ *          Suzanne McIntosh <[email protected]>
+ *          John Griffin
+ *
+ * History:
+ *          Suzanne McIntosh    13-Aug-07     Initial open source version
+ *
+ * Copyright (c) 2007, IBM Corporation
+ *
+ */
+
+// #include <linux/module.h>
+// #include <linux/kernel.h>
+
+//#include <net/sock.h>
+//#include <net/tcp_states.h>
+
+//#include <xen/driver_util.h>
+#include <stdint.h>
+#include <mini-os/os.h>
+#include <mini-os/errno.h>
+#include <mini-os/types.h>
+#include <mini-os/xmalloc.h>
+#include <mini-os/gnttab.h>
+#include <mini-os/events.h>
+
+#include <mini-os/xensocket.h>
+
+#define DPRINTK( x, args... ) printk("%s: line %d: " x, \
+        __FUNCTION__ , __LINE__ , ## args );
+
+#define DEBUG
+#ifdef DEBUG
+#define TRACE_ENTRY printk("Entering %s\n", __func__)
+#define TRACE_EXIT  printk("Exiting %s\n", __func__)
+#else
+#define TRACE_ENTRY do {} while (0)
+#define TRACE_EXIT  do {} while (0)
+#endif
+#define TRACE_ERROR printk("Exiting (ERROR) %s\n", __func__)
+
+/* ++++++++++++++++++++++++++++++++++++++++++++ */
+#define CM_SET_GREF  0x01
+#define CM_GET_GREF	0x02
+#define CM_FREE_NODE	0x03
+typedef struct{
+    uint16_t   remote_domid;
+    int         shared_page_gref;
+}SHARE_PAGE_GREF;
+/* +++++++++++++++++++++++++++++++++++++++++++ */
+
+/************************************************************************
+ * Data structures for internal recordkeeping and shared memory.
+ ************************************************************************/
+
+struct descriptor_page {
+    uint32_t        server_evtchn_port;
+    int             buffer_order; /* num_pages = (1 << buffer_order) */
+    int             buffer_first_gref;
+    unsigned int    send_offset;
+    unsigned int    recv_offset;
+    unsigned int    total_bytes_sent;
+    unsigned int    total_bytes_received;
+    unsigned int    sender_is_blocking;
+    atomic_t        avail_bytes;
+    atomic_t        sender_has_shutdown;
+    atomic_t        force_sender_shutdown;
+};
+
+static inline void atomic_set(atomic_t *v, int i) {v-> counter = i;}
+static inline int atomic_read(atomic_t *v) {return v-> counter;}
+static inline void atomic_add(int i, atomic_t *v) {v->counter += i;}
+static inline void atomic_sub(int i, atomic_t *v) {v->counter -= i;}
+
+static inline unsigned int min(unsigned int a, unsigned int b) {return a<b?a:b;}
+
+    static void
+initialize_descriptor_page (struct descriptor_page *d)
+{
+    d->server_evtchn_port = -1;
+    d->buffer_order = -1;
+    d->buffer_first_gref = -ENOSPC;
+    d->send_offset = 0;
+    d->recv_offset = 0;
+    d->total_bytes_sent = 0;
+    d->total_bytes_received = 0;
+    d->sender_is_blocking = 0;
+    atomic_set(&d->avail_bytes, 0);
+    atomic_set(&d->sender_has_shutdown, 0);
+    atomic_set(&d->force_sender_shutdown, 0);
+}
+
+/* struct xen_sock:
+ *
+ * @sk: this must be the first element in the structure.
+ */
+struct xen_sock {
+    //struct sock             sk;
+    unsigned char           is_server, is_client;
+    domid_t                 otherend_id;
+    struct descriptor_page *descriptor_addr;    /* server and client */
+    int                     descriptor_gref;    /* server only */
+    //  struct vm_struct       *descriptor_area;    /* client only */
+    grant_handle_t          descriptor_handle;  /* client only */
+    unsigned int            evtchn_local_port;
+    unsigned int            irq;
+    unsigned long           buffer_addr;    /* server and client */
+    int                    *buffer_grefs;   /* server */
+    grant_handle_t         *buffer_handles; /* client */
+    int                     buffer_order;
+};
+
+static void
+initialize_xen_sock (struct xen_sock *x) {
+    x->is_server = 0;
+    x->is_client = 0;
+    x->otherend_id = -1;
+    x->descriptor_addr = NULL;
+    x->descriptor_gref = -ENOSPC;
+    //x->descriptor_area = NULL;
+    x->descriptor_handle = -1;
+    x->evtchn_local_port = -1;
+    x->irq = -1;
+    x->buffer_addr = 0;
+    x->buffer_handles = NULL;
+    x->buffer_order = -1;
+}
+
+static void server_interrupt (evtchn_port_t, struct pt_regs *, void *);
+static void client_interrupt (evtchn_port_t, struct pt_regs *, void *);
+
+/* static struct proto xen_proto = {
+   .name           = "XEN",
+   .owner          = THIS_MODULE,
+   .obj_size       = sizeof(struct xen_sock),
+   }; */
+
+/* static const struct proto_ops xen_stream_ops = {
+   .family         = AF_XEN,
+   .owner          = THIS_MODULE,
+   .release        = xen_release,
+   .bind           = xen_bind,
+   .connect        = xen_connect,
+   .socketpair     = sock_no_socketpair,
+   .accept         = sock_no_accept,
+   .getname        = sock_no_getname,
+   .poll           = sock_no_poll,
+   .ioctl          = sock_no_ioctl,
+   .listen         = sock_no_listen,
+   .shutdown       = xen_shutdown,
+   .getsockopt     = sock_no_getsockopt,
+   .setsockopt     = sock_no_setsockopt,
+   .sendmsg        = xen_sendmsg,
+   .recvmsg        = xen_recvmsg,
+   .mmap           = sock_no_mmap,
+   .sendpage       = sock_no_sendpage,
+   }; */
+
+/* static struct net_proto_family xen_family_ops = {
+   .family         = AF_XEN,
+   .create         = xen_create,
+   .owner          = THIS_MODULE,
+   }; */
+
+
+/* TODO make this simpler */
+static int
+xen_shutdown (struct xen_sock *x, int how) {
+    //struct sock *sk = sock->sk;
+    //struct xen_sock *x;
+    struct descriptor_page *d;
+    SHARE_PAGE_GREF hypercall_arg;
+    //x = xen_sk(sk);
+    d = x->descriptor_addr;
+
+    if (x->is_server) {
+        /* +++++++++++++++++++++++++++++++++++++++++++ */
+        hypercall_arg.remote_domid = x->otherend_id;
+        hypercall_arg.shared_page_gref = -1;
+        if ( _hypercall2(long, myhpcall_gref_handler, CM_FREE_NODE, &hypercall_arg) )
+        {
+            DPRINTK("ERR: free node failed.\n");
+        }
+        /* +++++++++++++++++++++++++++++++++++++++++++ */
+        atomic_set(&d->force_sender_shutdown, 1);
+    }
+
+    return xen_release(x);
+}
+
+/************************************************************************
+ * Server-side connection setup functions.
+ ************************************************************************/
+
+/* In our nonstandard use of the bind function, the return value is the
+ * grant table entry of the descriptor page.
+ */
+static int
+xen_bind (struct xen_sock *x, uint16_t remote_domid) {
+    int    rc = -EINVAL;
+    SHARE_PAGE_GREF hypercall_arg;
+    TRACE_ENTRY;
+
+    /* Ensure that bind() is only called once for this socket.
+    */
+    if (x->is_server) {
+        DPRINTK("error: cannot call bind() more than once on a socket\n");
+        goto err;
+    }
+    if (x->is_client) {
+        DPRINTK("error: cannot call both bind() and connect() on the same socket\n");
+        goto err;
+    }
+    x->is_server = 1;
+    x->otherend_id = remote_domid;
+    if ((rc = server_allocate_descriptor_page(x)) != 0) {
+        goto err;
+    }
+
+    if ((rc = server_allocate_event_channel(x)) != 0) {
+        goto err;
+    }
+
+    if ((rc = server_allocate_buffer_pages(x)) != 0) {
+        goto err;
+    }
+
+    /* A successful function exit returns the grant table reference. */
+    hypercall_arg.remote_domid = x->otherend_id;
+    hypercall_arg.shared_page_gref = x->descriptor_gref;
+    if ( _hypercall2(long, myhpcall_gref_handler, CM_SET_GREF, &hypercall_arg) )
+    {
+        DPRINTK("ERR: set gref failed.\n");
+    }
+    TRACE_EXIT;
+    return x->descriptor_gref;
+
+err:
+    TRACE_ERROR;
+    return rc;
+}
+
+static int
+server_allocate_descriptor_page (struct xen_sock *x) {
+    TRACE_ENTRY;
+
+    if (x->descriptor_addr) {
+        DPRINTK("error: already allocated server descriptor page\n");
+        goto err;
+    }
+
+    if (!(x->descriptor_addr =
+                (struct descriptor_page *)alloc_page())) {
+        DPRINTK("error: cannot allocate free page\n");
+        goto err_unalloc;
+    }
+
+    initialize_descriptor_page(x->descriptor_addr);
+    x->descriptor_gref = gnttab_grant_access(x->otherend_id,
+            virt_to_mfn(x->descriptor_addr), 0);
+    if (x->descriptor_gref == -ENOSPC) {
+        DPRINTK("error: cannot share descriptor page %p\n", x->descriptor_addr);
+        goto err_unalloc;
+    }
+
+    TRACE_EXIT;
+    return 0;
+
+err_unalloc:
+    server_unallocate_descriptor_page(x);
+
+err:
+    TRACE_ERROR;
+    return -ENOMEM;
+}
+
+static int
+server_allocate_event_channel (struct xen_sock *x) {
+    evtchn_port_t port;
+    int         rc;
+
+    TRACE_ENTRY;
+    rc = evtchn_alloc_unbound(x->otherend_id, server_interrupt,
+            x, &port);
+    if ( rc != 0) {
+        DPRINTK("Unable to allocate event channel\n");
+        goto err;
+
+    }
+
+    x->evtchn_local_port = port;
+    x->descriptor_addr->server_evtchn_port = x->evtchn_local_port;
+
+    TRACE_EXIT;
+    return 0;
+
+err:
+    TRACE_ERROR;
+    return rc;
+}
+
+static int
+server_allocate_buffer_pages (struct xen_sock *x) {
+    struct descriptor_page *d = x->descriptor_addr;
+    int    buffer_num_pages;
+    int    i;
+
+    TRACE_ENTRY;
+
+    if (!d) {
+        /* must call server_allocate_descriptor_page first */
+        DPRINTK("error: descriptor page not yet allocated\n");
+        goto err;
+    }
+
+    if (x->buffer_addr) {
+        DPRINTK("error: already allocated server buffer pages\n");
+        goto err;
+    }
+
+    x->buffer_order = 5;  //32 pages    /* you can change this as desired */
+    buffer_num_pages = (1 << x->buffer_order);
+
+    if (!(x->buffer_addr = alloc_pages(x->buffer_order))) {
+        DPRINTK("error: cannot allocate %d pages\n", buffer_num_pages);
+        goto err;
+    }
+
+    if (!(x->buffer_grefs = malloc(buffer_num_pages * sizeof(int)))) {
+        DPRINTK("error: unexpected memory allocation failure\n");
+        goto err_unallocate;
+    }
+    else {
+        /* Success, so first invalidate all the entries */
+        for (i = 0; i < buffer_num_pages; i++) {
+            x->buffer_grefs[i] = -ENOSPC;
+        }
+    }
+
+    printk("x->buffer_addr = %lx  PAGE_SIZE = %li  buffer_num_pages = %d\n", \
+            x->buffer_addr, PAGE_SIZE, buffer_num_pages);
+    for (i = 0; i < buffer_num_pages; i++) {
+        x->buffer_grefs[i] =
+            gnttab_grant_access(x->otherend_id,
+                    virt_to_mfn(x->buffer_addr + i * PAGE_SIZE), 0);
+        if (x->buffer_grefs[i] == -ENOSPC) {
+            DPRINTK("error: cannot share buffer page #%d\n", i);
+            goto err_unallocate;
+        }
+    }
+
+    /* In this scheme, we initially use each page to hold
+     * the grant table reference for the next page.  The client maps
+     * the next page by reading the gref from the current page.
+     */
+
+    d->buffer_first_gref = x->buffer_grefs[0];
+    for (i = 1; i < buffer_num_pages; i++) {
+        int *next_gref = (int *)(x->buffer_addr + (i-1) * PAGE_SIZE);
+        *next_gref = x->buffer_grefs[i];
+    }
+
+    d->buffer_order = x->buffer_order;
+    atomic_set(&d->avail_bytes, (1 << d->buffer_order) * PAGE_SIZE);
+
+    TRACE_EXIT;
+    return 0;
+
+err_unallocate:
+    server_unallocate_buffer_pages(x);
+
+err:
+    TRACE_ERROR;
+    return -ENOMEM;
+}
+
+/************************************************************************
+ * Client-side connection setup functions.
+ ************************************************************************/
+
+static int
+xen_connect (struct xen_sock *x, uint16_t remote_domid, int shared_page_gref) {
+    int    rc = -EINVAL;
+    SHARE_PAGE_GREF hypercall_arg;
+    TRACE_ENTRY;
+    /* Ensure that connect() is only called once for this socket.
+    */
+
+    if (x->is_client) {
+        DPRINTK("error: cannot call connect() more than once on a socket\n");
+        goto err;
+    }
+    if (x->is_server) {
+        DPRINTK("error: cannot call both bind() and connect() on the same socket\n");
+        goto err;
+    }
+    x->is_client = 1;
+
+    x->otherend_id = remote_domid;
+    x->descriptor_gref = shared_page_gref;
+
+    /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
+    if( x->descriptor_gref <=0 )
+    {
+        DPRINTK("####get gref by hypercall.\n");
+        hypercall_arg.remote_domid = remote_domid;
+        hypercall_arg.shared_page_gref = -1;
+        if ( _hypercall2(long, myhpcall_gref_handler, CM_GET_GREF, &hypercall_arg) )
+        {
+            DPRINTK("ERR: get gref failed.\n");
+            goto err;
+        }
+        x->descriptor_gref = hypercall_arg.shared_page_gref;
+        DPRINTK("shared_page_gref = %d.\n", x->descriptor_gref);
+    }
+    /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
+
+    if ((rc = client_map_descriptor_page(x)) != 0) {
+        goto err;
+    }
+
+    if ((rc = client_bind_event_channel(x)) != 0) {
+        goto err_unmap_descriptor;
+    }
+
+    if ((rc = client_map_buffer_pages(x)) != 0) {
+        goto err_unmap_buffer;
+    }
+
+    TRACE_EXIT;
+    return 0;
+
+err_unmap_buffer:
+    client_unmap_buffer_pages(x);
+
+err_unmap_descriptor:
+    client_unmap_descriptor_page(x);
+    notify_remote_via_evtchn(x->evtchn_local_port);
+
+err:
+    return rc;
+}
+
+static int
+client_map_descriptor_page (struct xen_sock *x) {
+    struct gnttab_map_grant_ref op;
+    int    rc = -ENOMEM;
+
+    TRACE_ENTRY;
+
+    if (x->descriptor_addr) {
+        DPRINTK("error: already allocated client descriptor page\n");
+        goto err;
+    }
+
+    if ((x->descriptor_addr = (struct descriptor_page *)alloc_page()) == NULL) {
+        DPRINTK("error: cannot allocate memory for descriptor page\n");
+        goto err;
+    }
+
+    memset(&op, 0, sizeof(op));
+    op.host_addr = (uint64_t)x->descriptor_addr;
+    op.flags = GNTMAP_host_map;
+    op.ref = x->descriptor_gref;
+    op.dom = x->otherend_id;
+
+    rc = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, &op, 1);
+    if (rc == -ENOSYS) {
+        goto err_unmap;
+    }
+
+    if (op.status) {
+        DPRINTK("error: grant table mapping operation failed\n");
+        goto err_unmap;
+    }
+
+    x->descriptor_handle = op.handle;
+
+    TRACE_EXIT;
+    return 0;
+
+err_unmap:
+    client_unmap_descriptor_page(x);
+
+err:
+    TRACE_ERROR;
+    return rc;
+}
+
+static int
+client_bind_event_channel (struct xen_sock *x) {
+    int         rc;
+    evtchn_port_t port;
+
+    TRACE_ENTRY;
+
+    /* Start by binding this end of the event channel to the other
+     * end of the event channel. */
+    if((rc=evtchn_bind_interdomain(x->otherend_id,
+                    x->descriptor_addr->server_evtchn_port,
+                    client_interrupt, x, &port)) != 0) {
+        DPRINTK("Unable to bind to sender's event channel\n");
+        goto err;
+    }
+
+    x->evtchn_local_port = port;
+
+    DPRINTK("Other port is %d\n", x->descriptor_addr->server_evtchn_port);
+    DPRINTK("My port is %d\n", port);
+
+    /* Next bind this end of the event channel to our local callback
+     * function. */
+    x->irq = port;
+
+    TRACE_EXIT;
+    return 0;
+
+err:
+    TRACE_ERROR;
+    return rc;
+}
+
+static int
+client_map_buffer_pages (struct xen_sock *x) {
+    struct descriptor_page *d = x->descriptor_addr;
+    int    buffer_num_pages;
+    int    *grefp;
+    int    i;
+    struct gnttab_map_grant_ref op;
+    int    rc = -ENOMEM;
+
+    TRACE_ENTRY;
+
+    if (!d) {
+        /* must call client_map_descriptor_page first */
+        DPRINTK("error: descriptor page not yet mapped\n");
+        goto err;
+    }
+
+    if (d->buffer_order == -1) {
+        DPRINTK("error: server has not yet allocated buffer pages\n");
+        goto err;
+    }
+
+    x->buffer_order = d->buffer_order;
+    buffer_num_pages = (1 << x->buffer_order);
+
+    if (!(x->buffer_handles = malloc(buffer_num_pages * sizeof(grant_handle_t)))) {
+        DPRINTK("error: unexpected memory allocation failure\n");
+        goto err;
+    }
+    else {
+        for (i = 0; i < buffer_num_pages; i++) {
+            x->buffer_handles[i] = -1;
+        }
+    }
+
+    // this might be rather incorrect.
+    if (!(x->buffer_addr = (unsigned long)alloc_pages(buffer_num_pages))) {
+        DPRINTK("error: cannot allocate %d buffer pages\n", buffer_num_pages);
+        goto err_unmap;
+    }
+
+    grefp = &d->buffer_first_gref;
+    for (i = 0; i < buffer_num_pages; i++) {
+        memset(&op, 0, sizeof(op));
+        op.host_addr = x->buffer_addr + i * PAGE_SIZE;
+        op.flags = GNTMAP_host_map;
+        op.ref = *grefp;
+        op.dom = x->otherend_id;
+
+        rc = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, &op, 1);
+        if (rc == -ENOSYS) {
+            goto err_unmap;
+        }
+
+        if (op.status) {
+            DPRINTK("error: grant table mapping failed\n");
+            goto err_unmap;
+        }
+
+        x->buffer_handles[i] = op.handle;
+        grefp = (int *)(x->buffer_addr + i * PAGE_SIZE);
+    }
+
+    TRACE_EXIT;
+    return 0;
+
+err_unmap:
+    client_unmap_buffer_pages(x);
+
+err:
+    TRACE_ERROR;
+    return rc;
+}
+
+/************************************************************************
+ * Data transmission functions (client-only in a one-way communication
+ * channel).
+ ************************************************************************/
+
+static int
+xen_sendmsg (struct xen_sock *x, void *data, size_t len) {
+    // int                     rc = -EINVAL;
+    struct descriptor_page *d = x->descriptor_addr;
+    unsigned int            max_offset = (1 << x->buffer_order) * PAGE_SIZE;
+    // long                    timeo;
+    unsigned int            copied = 0;
+
+    TRACE_ENTRY;
+
+    while (copied < len) {
+        unsigned int send_offset = d->send_offset;
+        unsigned int avail_bytes = atomic_read(&d->avail_bytes);
+        unsigned int bytes;
+
+        if (atomic_read(&d->force_sender_shutdown) != 0) {
+            xen_release(x);
+            goto err;
+        }
+
+        /* Determine the maximum amount that can be written */
+        bytes = len - copied;
+        bytes = min(bytes, avail_bytes);
+
+        /* Block if no space is available */
+        // TODO How can I block the write?
+        /*    if (bytes == 0) {
+              timeo = send_data_wait(sk, timeo);
+              if (signal_pending(current)) {
+              rc = sock_intr_errno(timeo);
+              goto err;
+              }
+              continue;
+              }*/
+
+        if ((send_offset + bytes) > max_offset) {
+            /* wrap around, need to copy twice */
+            unsigned int bytes_segment1 = max_offset - send_offset;
+            unsigned int bytes_segment2 = bytes - bytes_segment1;
+
+            /* TODO this may be wrong */
+            memcpy((unsigned char *)(x->buffer_addr + send_offset),
+                    (data + copied), bytes_segment1);
+            memcpy((unsigned char *)(x->buffer_addr),
+                    (data + copied + bytes_segment1), bytes_segment2);
+        }
+        else {
+            /* no need to wrap around */
+            memcpy((unsigned char *)(x->buffer_addr + send_offset),
+                    (data + copied), bytes);
+        }
+
+        /* Update values */
+        copied += bytes;
+        d->send_offset = (send_offset + bytes) % max_offset;
+        d->total_bytes_sent += bytes;
+        atomic_sub(bytes, &d->avail_bytes);
+    }
+
+    notify_remote_via_evtchn(x->evtchn_local_port);
+
+    TRACE_EXIT;
+    return copied;
+
+err:
+    TRACE_ERROR;
+    return copied;
+}
+
+static inline int
+is_writeable (struct descriptor_page *d) {
+    unsigned int avail_bytes = atomic_read(&d->avail_bytes);
+    if (avail_bytes > 0)
+        return 1;
+
+    return 0;
+}
+
+/* static long
+   send_data_wait (struct xen_sock *x, long timeo) {
+   struct descriptor_page *d = x->descriptor_addr;
+   DEFINE_WAIT(wait);
+
+   TRACE_ENTRY;
+
+   d->sender_is_blocking = 1;
+   notify_remote_via_evtchn(x->evtchn_local_port);
+
+   for (;;) {
+   prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
+
+   if (is_writeable(d)
+   || !skb_queue_empty(&sk->sk_receive_queue)
+   || sk->sk_err
+   || (sk->sk_shutdown & RCV_SHUTDOWN)
+   || signal_pending(current)
+   || !timeo
+   || atomic_read(&d->force_sender_shutdown)) {
+   break;
+   }
+
+   timeo = schedule_timeout(timeo);
+   }
+
+   d->sender_is_blocking = 0;
+
+   finish_wait(sk->sk_sleep, &wait);
+
+   TRACE_EXIT;
+   return timeo;
+   }*/
+
+static void
+client_interrupt (evtchn_port_t irq, struct pt_regs *regs, void *dev_id) {
+    // struct xen_sock *x = dev_id;
+    TRACE_ENTRY;
+    // TODO FIXME what happens when the interrupt fires?
+    /*if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) {
+      wake_up_interruptible(sk->sk_sleep);
+      } */
+    TRACE_EXIT;
+}
+
+/************************************************************************
+ * Data reception functions (server-only in a one-way communication
+ * channel, but common to both in a two-way channel).
+ ***********************************************************************/
+
+static int
+xen_recvmsg (struct xen_sock *x, void *data, size_t size) {
+    struct descriptor_page *d = x->descriptor_addr;
+    unsigned int            max_offset = (1 << x->buffer_order) * PAGE_SIZE;
+    int                     copied = 0;
+    int                     target;
+    long                      timeo;
+
+    TRACE_ENTRY;
+
+    while (copied < size) {
+        unsigned int recv_offset = d->recv_offset;
+        unsigned int bytes;
+        unsigned int avail_bytes = max_offset - atomic_read(&d->avail_bytes);  /* bytes available for read */
+
+        /* Determine the maximum amount that can be read */
+        bytes = min(size - copied, avail_bytes);
+
+        if (atomic_read(&d->sender_has_shutdown) != 0) {
+            if (avail_bytes == 0) {
+                copied = 0;
+                break;
+            }
+        }
+
+        /* Block if the buffer is empty */
+        if (bytes == 0) {
+            if (copied > target) {
+                break;
+            }
+
+            timeo = receive_data_wait(x, timeo);
+            /*      if (signal_pending(current)) {
+                    rc = sock_intr_errno(timeo);
+                    DPRINTK("error: signal\n");
+                    goto err;
+                    }
+                    continue; */
+        }
+
+        /* Perform the read */
+        if ((recv_offset + bytes) > max_offset) {
+            /* wrap around, need to perform the read twice */
+            unsigned int bytes_segment1 = max_offset - recv_offset;
+            unsigned int bytes_segment2 = bytes - bytes_segment1;
+            memcpy(data+copied, (unsigned char *)(x->buffer_addr + recv_offset),
+                    bytes_segment1);
+            memcpy(data+copied+bytes_segment1, (unsigned char *)(x->buffer_addr),
+                    bytes_segment2);
+        }
+        else {
+            /* no wrap around, proceed with one copy */
+            memcpy(data+copied, (unsigned char *)(x->buffer_addr + recv_offset),
+                    bytes);
+        }
+
+        /* Update values */
+        copied += bytes;
+        d->recv_offset = (recv_offset + bytes) % max_offset;
+        d->total_bytes_received += bytes;
+        atomic_add(bytes, &d->avail_bytes);
+        if (d->sender_is_blocking) {
+            notify_remote_via_evtchn(x->evtchn_local_port);
+        }
+    }
+
+    TRACE_EXIT;
+    return copied;
+}
+
+static inline int
+is_readable (struct descriptor_page *d) {
+    unsigned int max_offset = (1 << d->buffer_order) * PAGE_SIZE;
+    unsigned int avail_bytes = max_offset - atomic_read(&d->avail_bytes);
+    if (avail_bytes > 0)
+        return 1;
+
+    return 0;
+}
+
+static long
+receive_data_wait (struct xen_sock *x, long timeo) {
+    struct descriptor_page *d = x->descriptor_addr;
+//    DEFINE_WAIT(wait);
+
+    TRACE_ENTRY;
+
+    for (;;) {
+/*        prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
+        if (is_readable(d)
+                || (atomic_read(&d->sender_has_shutdown) != 0)
+                || !skb_queue_empty(&sk->sk_receive_queue)
+                || sk->sk_err
+                || (sk->sk_shutdown & RCV_SHUTDOWN)
+                || signal_pending(current)
+                || !timeo) {
+            break;
+        }
+
+        timeo = schedule_timeout(timeo); */
+        break;
+    }
+
+//    finish_wait(sk->sk_sleep, &wait);
+
+    TRACE_EXIT;
+    return timeo;
+}
+
+static void
+server_interrupt (evtchn_port_t irq, struct pt_regs *regs, void *dev_id) {
+//    struct xen_sock *x = dev_id;
+
+    TRACE_ENTRY;
+
+/*    if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) {
+        wake_up_interruptible(sk->sk_sleep);
+    } */
+
+    TRACE_EXIT;
+}
+
+// static int
+// local_memcpy_toiovecend (struct iovec *iov, unsigned char *kdata, int offset, int len) {
+//     int err = -EFAULT;
+//
+//     /* Skip over the finished iovecs */
+//     while (offset >= iov->iov_len) {
+//         offset -= iov->iov_len;
+//         iov++;
+//     }
+//
+//     while (len > 0) {
+//         u8 *base = iov->iov_base + offset;
+//         int copy = min((unsigned int)len, iov->iov_len - offset);
+//
+//         offset = 0;
+//         if (copy_to_user(base, kdata, copy)) {
+//             goto out;
+//         }
+//         kdata += copy;
+//         len -= copy;
+//         iov++;
+//     }
+//     err = 0;
+//
+// out:
+//     return err;
+// } */
+
+/************************************************************************
+ * Connection teardown functions (common to both server and client).
+ ************************************************************************/
+
+static int
+xen_release (struct xen_sock *x) {
+    //struct sock            *sk = sock->sk;
+    //struct xen_sock        *x;
+    struct descriptor_page *d;
+
+    TRACE_ENTRY;
+    //if (!sk) {
+    //  return 0;
+    //}
+
+    //sock->sk = NULL;
+    //x = xen_sk(sk);
+    d = x->descriptor_addr;
+
+    // if map didn't succeed, gracefully exit
+    if (x->descriptor_handle == -1)
+        goto out;
+
+    if (x->is_server) {
+        while (atomic_read(&d->sender_has_shutdown) == 0 ) {
+        }
+
+        server_unallocate_buffer_pages(x);
+        server_unallocate_descriptor_page(x);
+    }
+
+    if (x->is_client) {
+        if ((atomic_read(&d->sender_has_shutdown)) == 0) {
+            client_unmap_buffer_pages(x);
+            client_unmap_descriptor_page(x);
+            notify_remote_via_evtchn(x->evtchn_local_port);
+        }
+        else {
+            printk("    xen_release: SENDER ALREADY SHUT DOWN!\n");
+        }
+    }
+
+out:
+    TRACE_EXIT;
+    return 0;
+}
+
+static void
+server_unallocate_buffer_pages (struct xen_sock *x) {
+    if (x->buffer_grefs) {
+        int buffer_num_pages = (1 << x->buffer_order);
+        int i;
+
+        for (i = 0; i < buffer_num_pages; i++) {
+            if (x->buffer_grefs[i] == -ENOSPC) {
+                break;
+            }
+
+            gnttab_end_access(x->buffer_grefs[i]);
+            x->buffer_grefs[i] = -ENOSPC;
+        }
+
+        free(x->buffer_grefs);
+        x->buffer_grefs = NULL;
+    }
+
+    if (x->buffer_addr) {
+        struct descriptor_page *d = x->descriptor_addr;
+
+        free_pages((void *)x->buffer_addr, x->buffer_order);
+        x->buffer_addr = 0;
+        x->buffer_order = -1;
+        if (d) {
+            d->buffer_order = -1;
+        }
+    }
+}
+
+static void
+server_unallocate_descriptor_page (struct xen_sock *x) {
+    if (x->descriptor_gref != -ENOSPC) {
+        gnttab_end_access(x->descriptor_gref);
+        x->descriptor_gref = -ENOSPC;
+    }
+    if (x->descriptor_addr) {
+        free_page((void *)(x->descriptor_addr));
+        x->descriptor_addr = NULL;
+    }
+}
+
+static void
+client_unmap_buffer_pages (struct xen_sock *x) {
+
+    if (x->buffer_handles) {
+        struct descriptor_page *d = x->descriptor_addr;
+        int                     buffer_order = d->buffer_order;
+        int                     buffer_num_pages = (1 << buffer_order);
+        int                     i;
+        struct                  gnttab_unmap_grant_ref op;
+        int                     rc = 0;
+
+        for (i = 0; i < buffer_num_pages; i++) {
+            if (x->buffer_handles[i] == -1) {
+                break;
+            }
+
+            memset(&op, 0, sizeof(op));
+            op.host_addr = x->buffer_addr + i * PAGE_SIZE;
+            op.handle = x->buffer_handles[i];
+            op.dev_bus_addr = 0;
+
+            rc = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, &op, 1);
+            if (rc == -ENOSYS) {
+                printk("Failure to unmap grant reference \n");
+            }
+        }
+
+        free(x->buffer_handles);
+        x->buffer_handles = NULL;
+    }
+}
+
+static void
+client_unmap_descriptor_page (struct xen_sock *x) {
+    struct descriptor_page *d;
+    int                     rc = 0;
+
+    d = x->descriptor_addr;
+
+    if (x->descriptor_handle != -1) {
+        struct gnttab_unmap_grant_ref op;
+
+        memset(&op, 0, sizeof(op));
+        op.host_addr = (unsigned long)x->descriptor_addr;
+        op.handle = x->descriptor_handle;
+        op.dev_bus_addr = 0;
+
+        atomic_set(&d->sender_has_shutdown, 1);
+        rc = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, &op, 1);
+        if (rc == -ENOSYS) {
+            printk("Failure to unmap grant reference for descriptor page\n");
+        }
+
+        x->descriptor_handle = -1;
+    }
+}
-- 
2.7.4