This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / drivers / xen / xenbus / xenbus_dev.c
diff --git a/drivers/xen/xenbus/xenbus_dev.c b/drivers/xen/xenbus/xenbus_dev.c
new file mode 100644 (file)
index 0000000..02d3b46
--- /dev/null
@@ -0,0 +1,359 @@
+/*
+ * xenbus_dev.c
+ * 
+ * Driver giving user-space access to the kernel's xenbus connection
+ * to xenstore.
+ * 
+ * Copyright (c) 2005, Christian Limpach
+ * Copyright (c) 2005, Rusty Russell, IBM Corporation
+ * 
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version 2
+ * as published by the Free Software Foundation; or, when distributed
+ * separately from the Linux kernel or incorporated into other
+ * software packages, subject to the following license:
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this source file (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 <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/uio.h>
+#include <linux/notifier.h>
+#include <linux/wait.h>
+#include <linux/fs.h>
+#include <linux/poll.h>
+
+#include "xenbus_comms.h"
+
+#include <asm/uaccess.h>
+#include <asm/hypervisor.h>
+#include <xen/xenbus.h>
+#include <xen/xen_proc.h>
+#include <asm/hypervisor.h>
+
+struct xenbus_dev_transaction {
+       struct list_head list;
+       struct xenbus_transaction handle;
+};
+
+struct xenbus_dev_data {
+       /* In-progress transaction. */
+       struct list_head transactions;
+
+       /* Active watches. */
+       struct list_head watches;
+
+       /* Partial request. */
+       unsigned int len;
+       union {
+               struct xsd_sockmsg msg;
+               char buffer[PAGE_SIZE];
+       } u;
+
+       /* Response queue. */
+#define MASK_READ_IDX(idx) ((idx)&(PAGE_SIZE-1))
+       char read_buffer[PAGE_SIZE];
+       unsigned int read_cons, read_prod;
+       wait_queue_head_t read_waitq;
+
+       struct mutex reply_mutex;
+};
+
+static struct proc_dir_entry *xenbus_dev_intf;
+
+static ssize_t xenbus_dev_read(struct file *filp,
+                              char __user *ubuf,
+                              size_t len, loff_t *ppos)
+{
+       struct xenbus_dev_data *u = filp->private_data;
+       int i;
+
+       if (wait_event_interruptible(u->read_waitq,
+                                    u->read_prod != u->read_cons))
+               return -EINTR;
+
+       for (i = 0; i < len; i++) {
+               if (u->read_cons == u->read_prod)
+                       break;
+               put_user(u->read_buffer[MASK_READ_IDX(u->read_cons)], ubuf+i);
+               u->read_cons++;
+       }
+
+       return i;
+}
+
+static void queue_reply(struct xenbus_dev_data *u,
+                       char *data, unsigned int len)
+{
+       int i;
+
+       mutex_lock(&u->reply_mutex);
+
+       for (i = 0; i < len; i++, u->read_prod++)
+               u->read_buffer[MASK_READ_IDX(u->read_prod)] = data[i];
+
+       BUG_ON((u->read_prod - u->read_cons) > sizeof(u->read_buffer));
+
+       mutex_unlock(&u->reply_mutex);
+
+       wake_up(&u->read_waitq);
+}
+
+struct watch_adapter
+{
+       struct list_head list;
+       struct xenbus_watch watch;
+       struct xenbus_dev_data *dev_data;
+       char *token;
+};
+
+static void free_watch_adapter (struct watch_adapter *watch)
+{
+       kfree(watch->watch.node);
+       kfree(watch->token);
+       kfree(watch);
+}
+
+static void watch_fired(struct xenbus_watch *watch,
+                       const char **vec,
+                       unsigned int len)
+{
+       struct watch_adapter *adap =
+            container_of(watch, struct watch_adapter, watch);
+       struct xsd_sockmsg hdr;
+       const char *path, *token;
+       int path_len, tok_len, body_len;
+
+       path = vec[XS_WATCH_PATH];
+       token = adap->token;
+
+       path_len = strlen(path) + 1;
+       tok_len = strlen(token) + 1;
+       body_len = path_len + tok_len;
+
+       hdr.type = XS_WATCH_EVENT;
+       hdr.len = body_len;
+       
+       queue_reply(adap->dev_data, (char *)&hdr, sizeof(hdr));
+       queue_reply(adap->dev_data, (char *)path, path_len);
+       queue_reply(adap->dev_data, (char *)token, tok_len);
+}
+
+static LIST_HEAD(watch_list);
+
+static ssize_t xenbus_dev_write(struct file *filp,
+                               const char __user *ubuf,
+                               size_t len, loff_t *ppos)
+{
+       struct xenbus_dev_data *u = filp->private_data;
+       struct xenbus_dev_transaction *trans = NULL;
+       uint32_t msg_type;
+       void *reply;
+       char *path, *token;
+       struct watch_adapter *watch, *tmp_watch;
+       int err;
+
+       if ((len + u->len) > sizeof(u->u.buffer))
+               return -EINVAL;
+
+       if (copy_from_user(u->u.buffer + u->len, ubuf, len) != 0)
+               return -EFAULT;
+
+       u->len += len;
+       if (u->len < (sizeof(u->u.msg) + u->u.msg.len))
+               return len;
+
+       msg_type = u->u.msg.type;
+
+       switch (msg_type) {
+       case XS_TRANSACTION_START:
+       case XS_TRANSACTION_END:
+       case XS_DIRECTORY:
+       case XS_READ:
+       case XS_GET_PERMS:
+       case XS_RELEASE:
+       case XS_GET_DOMAIN_PATH:
+       case XS_WRITE:
+       case XS_MKDIR:
+       case XS_RM:
+       case XS_SET_PERMS:
+               if (msg_type == XS_TRANSACTION_START) {
+                       trans = kmalloc(sizeof(*trans), GFP_KERNEL);
+                       if (!trans)
+                               return -ENOMEM;
+               }
+
+               reply = xenbus_dev_request_and_reply(&u->u.msg);
+               if (IS_ERR(reply)) {
+                       kfree(trans);
+                       return PTR_ERR(reply);
+               }
+
+               if (msg_type == XS_TRANSACTION_START) {
+                       trans->handle.id = simple_strtoul(reply, NULL, 0);
+                       list_add(&trans->list, &u->transactions);
+               } else if (msg_type == XS_TRANSACTION_END) {
+                       list_for_each_entry(trans, &u->transactions, list)
+                               if (trans->handle.id == u->u.msg.tx_id)
+                                       break;
+                       BUG_ON(&trans->list == &u->transactions);
+                       list_del(&trans->list);
+                       kfree(trans);
+               }
+               queue_reply(u, (char *)&u->u.msg, sizeof(u->u.msg));
+               queue_reply(u, (char *)reply, u->u.msg.len);
+               kfree(reply);
+               break;
+
+       case XS_WATCH:
+       case XS_UNWATCH:
+               path = u->u.buffer + sizeof(u->u.msg);
+               token = memchr(path, 0, u->u.msg.len);
+               if (token == NULL)
+                       return -EILSEQ;
+               token++;
+
+               if (msg_type == XS_WATCH) {
+                       static const char * XS_WATCH_RESP = "OK";
+                       struct xsd_sockmsg hdr;
+
+                       watch = kmalloc(sizeof(*watch), GFP_KERNEL);
+                       watch->watch.node = kmalloc(strlen(path)+1,
+                                                    GFP_KERNEL);
+                       strcpy((char *)watch->watch.node, path);
+                       watch->watch.callback = watch_fired;
+                       watch->token = kmalloc(strlen(token)+1, GFP_KERNEL);
+                       strcpy(watch->token, token);
+                       watch->dev_data = u;
+
+                       err = register_xenbus_watch(&watch->watch);
+                       if (err) {
+                               free_watch_adapter(watch);
+                               return err;
+                       }
+                       
+                       list_add(&watch->list, &u->watches);
+
+                       hdr.type = XS_WATCH;
+                       hdr.len = strlen(XS_WATCH_RESP) + 1;
+                       queue_reply(u, (char *)&hdr, sizeof(hdr));
+                       queue_reply(u, (char *)XS_WATCH_RESP, hdr.len);
+               } else {
+                       list_for_each_entry_safe(watch, tmp_watch,
+                                                 &u->watches, list) {
+                               if (!strcmp(watch->token, token) &&
+                                   !strcmp(watch->watch.node, path))
+                                       break;
+                               {
+                                       unregister_xenbus_watch(&watch->watch);
+                                       list_del(&watch->list);
+                                       free_watch_adapter(watch);
+                                       break;
+                               }
+                       }
+               }
+
+               break;
+
+       default:
+               return -EINVAL;
+       }
+
+       u->len = 0;
+       return len;
+}
+
+static int xenbus_dev_open(struct inode *inode, struct file *filp)
+{
+       struct xenbus_dev_data *u;
+
+       if (xen_start_info->store_evtchn == 0)
+               return -ENOENT;
+
+       nonseekable_open(inode, filp);
+
+       u = kzalloc(sizeof(*u), GFP_KERNEL);
+       if (u == NULL)
+               return -ENOMEM;
+
+       INIT_LIST_HEAD(&u->transactions);
+       INIT_LIST_HEAD(&u->watches);
+       init_waitqueue_head(&u->read_waitq);
+
+       mutex_init(&u->reply_mutex);
+
+       filp->private_data = u;
+
+       return 0;
+}
+
+static int xenbus_dev_release(struct inode *inode, struct file *filp)
+{
+       struct xenbus_dev_data *u = filp->private_data;
+       struct xenbus_dev_transaction *trans, *tmp;
+       struct watch_adapter *watch, *tmp_watch;
+
+       list_for_each_entry_safe(trans, tmp, &u->transactions, list) {
+               xenbus_transaction_end(trans->handle, 1);
+               list_del(&trans->list);
+               kfree(trans);
+       }
+
+       list_for_each_entry_safe(watch, tmp_watch, &u->watches, list) {
+               unregister_xenbus_watch(&watch->watch);
+               list_del(&watch->list);
+               free_watch_adapter(watch);
+       }
+
+       kfree(u);
+
+       return 0;
+}
+
+static unsigned int xenbus_dev_poll(struct file *file, poll_table *wait)
+{
+       struct xenbus_dev_data *u = file->private_data;
+
+       poll_wait(file, &u->read_waitq, wait);
+       if (u->read_cons != u->read_prod)
+               return POLLIN | POLLRDNORM;
+       return 0;
+}
+
+static struct file_operations xenbus_dev_file_ops = {
+       .read = xenbus_dev_read,
+       .write = xenbus_dev_write,
+       .open = xenbus_dev_open,
+       .release = xenbus_dev_release,
+       .poll = xenbus_dev_poll,
+};
+
+static int __init
+xenbus_dev_init(void)
+{
+       xenbus_dev_intf = create_xen_proc_entry("xenbus", 0400);
+       if (xenbus_dev_intf)
+               xenbus_dev_intf->proc_fops = &xenbus_dev_file_ops;
+
+       return 0;
+}
+
+__initcall(xenbus_dev_init);