[PATCH 05/10] VMCI: event handling implementation.

From: George Zhang
Date: Mon Oct 15 2012 - 20:38:39 EST


VMCI event code that manages event handlers and handles callbacks when specific events fire.


Signed-off-by: George Zhang <georgezhang@xxxxxxxxxx>
---
drivers/misc/vmw_vmci/vmci_event.c | 415 ++++++++++++++++++++++++++++++++++++
drivers/misc/vmw_vmci/vmci_event.h | 25 ++
2 files changed, 440 insertions(+), 0 deletions(-)
create mode 100644 drivers/misc/vmw_vmci/vmci_event.c
create mode 100644 drivers/misc/vmw_vmci/vmci_event.h

diff --git a/drivers/misc/vmw_vmci/vmci_event.c b/drivers/misc/vmw_vmci/vmci_event.c
new file mode 100644
index 0000000..977550d
--- /dev/null
+++ b/drivers/misc/vmw_vmci/vmci_event.c
@@ -0,0 +1,415 @@
+/*
+ * VMware VMCI Driver
+ *
+ * Copyright (C) 2012 VMware, Inc. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation version 2 and no later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ * for more details.
+ */
+
+#include <linux/vmw_vmci_defs.h>
+#include <linux/vmw_vmci_api.h>
+#include <linux/completion.h>
+#include <linux/kref.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+
+#include "vmci_driver.h"
+#include "vmci_event.h"
+
+#define EVENT_MAGIC 0xEABE0000
+#define VMCI_EVENT_MAX_ATTEMPTS 10
+
+struct vmci_subscription {
+ u32 id;
+ u32 event;
+ struct kref kref;
+ struct completion done; /* unregistered, ready to be freed */
+ vmci_event_cb callback;
+ void *callback_data;
+ struct list_head node; /* on one of subscriber lists */
+ bool run_delayed;
+};
+
+static struct list_head subscriber_array[VMCI_EVENT_MAX];
+static DEFINE_MUTEX(subscriber_mutex);
+
+struct delayed_event_info {
+ struct work_struct work;
+ struct vmci_subscription *sub;
+ u8 event_payload[sizeof(struct vmci_event_data_max)];
+};
+
+struct event_ref {
+ struct vmci_subscription *sub;
+ struct list_head list_item;
+};
+
+int __init vmci_event_init(void)
+{
+ int i;
+
+ for (i = 0; i < VMCI_EVENT_MAX; i++)
+ INIT_LIST_HEAD(&subscriber_array[i]);
+
+ return VMCI_SUCCESS;
+}
+
+void vmci_event_exit(void)
+{
+ int e;
+
+ /* We free all memory at exit. */
+ for (e = 0; e < VMCI_EVENT_MAX; e++) {
+ struct vmci_subscription *cur, *p2;
+ list_for_each_entry_safe(cur, p2, &subscriber_array[e], node) {
+
+ /*
+ * We should never get here because all events
+ * should have been unregistered before we try
+ * to unload the driver module. Also, delayed
+ * callbacks could still be firing so this
+ * cleanup would not be safe. Still it is
+ * better to free the memory than not ... so
+ * we leave this code in just in case....
+ */
+ pr_warn("Unexpected free events occuring.");
+ kfree(cur);
+ }
+ }
+
+}
+
+/*
+ * Gets a reference to the given VMCISubscription.
+ */
+static struct vmci_subscription *event_get(struct vmci_subscription *entry)
+{
+ kref_get(&entry->kref);
+
+ return entry;
+}
+
+static void event_signal_destroy(struct kref *kref)
+{
+ struct vmci_subscription *entry =
+ container_of(kref, struct vmci_subscription, kref);
+
+ complete(&entry->done);
+}
+
+/*
+ * Releases the given VMCISubscription.
+ * Fires the destroy event if the reference count has gone to zero.
+ */
+static void event_release(struct vmci_subscription *entry)
+{
+ kref_put(&entry->kref, event_signal_destroy);
+}
+
+/*
+ * Find entry. Assumes lock is held.
+ */
+static struct vmci_subscription *event_find(u32 sub_id)
+{
+ int e;
+
+ for (e = 0; e < VMCI_EVENT_MAX; e++) {
+ struct vmci_subscription *cur;
+ list_for_each_entry(cur, &subscriber_array[e], node) {
+ if (cur->id == sub_id)
+ return cur;
+ }
+ }
+ return NULL;
+}
+
+/*
+ * Calls the specified callback in a delayed context.
+ */
+static void event_delayed_dispatch(struct work_struct *work)
+{
+ struct delayed_event_info *event_info =
+ container_of(work, struct delayed_event_info, work);
+ struct vmci_subscription *sub = event_info->sub;
+ struct vmci_event_data *ed;
+
+ BUG_ON(!sub);
+
+ ed = (struct vmci_event_data *)event_info->event_payload;
+
+ sub->callback(sub->id, ed, sub->callback_data);
+ event_release(sub);
+
+ kfree(event_info);
+}
+
+/*
+ * Actually delivers the events to the subscribers.
+ * The callback function for each subscriber is invoked.
+ */
+static int event_deliver(struct vmci_event_msg *event_msg)
+{
+ int err = VMCI_SUCCESS;
+ struct vmci_subscription *cur;
+ struct list_head *subscriber_list;
+ struct list_head no_delay_list;
+ struct vmci_event_data *ed;
+ struct event_ref *event_ref, *p2;
+
+ ASSERT(event_msg);
+
+ INIT_LIST_HEAD(&no_delay_list);
+
+ rcu_read_lock();
+ subscriber_list = &subscriber_array[event_msg->event_data.event];
+ list_for_each_entry_rcu(cur, subscriber_list, node) {
+ ASSERT(cur && cur->event == event_msg->event_data.event);
+
+ if (cur->run_delayed) {
+ struct delayed_event_info *event_info;
+
+ event_info = kzalloc(sizeof(*event_info), GFP_ATOMIC);
+ if (!event_info) {
+ err = VMCI_ERROR_NO_MEM;
+ goto out;
+ }
+
+ INIT_WORK(&event_info->work, event_delayed_dispatch);
+ event_info->sub = event_get(cur);
+ memcpy(event_info->event_payload,
+ VMCI_DG_PAYLOAD(event_msg),
+ (size_t) event_msg->hdr.payload_size);
+
+ schedule_work(&event_info->work);
+
+ } else {
+ /*
+ * To avoid possible lock rank voilation when holding
+ * subscriber_lock, we construct a local list of
+ * subscribers and release subscriber_lock before
+ * invokes the callbacks. This is similar to delayed
+ * callbacks, but callbacks is invoked right away here.
+ */
+ event_ref = kmalloc(sizeof(*event_ref), GFP_ATOMIC);
+ if (!event_ref) {
+ err = VMCI_ERROR_NO_MEM;
+ goto out;
+ }
+
+ event_ref->sub = event_get(cur);
+ INIT_LIST_HEAD(&event_ref->list_item);
+ list_add(&event_ref->list_item, &no_delay_list);
+ }
+ }
+
+ out:
+ rcu_read_unlock();
+
+ list_for_each_entry_safe(event_ref, p2, &no_delay_list, list_item) {
+ u8 event_payload[sizeof(struct vmci_event_data_max)] = { 0 };
+
+ /*
+ * We set event data before each callback to ensure
+ * isolation.
+ */
+ memcpy(event_payload, VMCI_DG_PAYLOAD(event_msg),
+ (size_t) event_msg->hdr.payload_size);
+ ed = (struct vmci_event_data *)event_payload;
+ cur = event_ref->sub;
+ cur->callback(cur->id, ed, cur->callback_data);
+ event_release(cur);
+
+ kfree(event_ref);
+ }
+
+ return err;
+}
+
+/*
+ * Dispatcher for the VMCI_EVENT_RECEIVE datagrams. Calls all
+ * subscribers for given event.
+ */
+int vmci_event_dispatch(struct vmci_datagram *msg)
+{
+ struct vmci_event_msg *event_msg = (struct vmci_event_msg *)msg;
+
+ ASSERT(msg &&
+ msg->src.context == VMCI_HYPERVISOR_CONTEXT_ID &&
+ msg->dst.resource == VMCI_EVENT_HANDLER);
+
+ if (msg->payload_size < sizeof(u32) ||
+ msg->payload_size > sizeof(struct vmci_event_data_max))
+ return VMCI_ERROR_INVALID_ARGS;
+
+ if (!VMCI_EVENT_VALID(event_msg->event_data.event))
+ return VMCI_ERROR_EVENT_UNKNOWN;
+
+ event_deliver(event_msg);
+ return VMCI_SUCCESS;
+}
+
+/*
+ * Initialize and add subscription to subscriber list.
+ */
+static int event_register_subscription(struct vmci_subscription *sub,
+ u32 event,
+ u32 flags,
+ vmci_event_cb callback,
+ void *callback_data)
+{
+ static u32 subscription_id;
+ int attempts;
+ int result;
+ bool success = false;
+
+ ASSERT(sub);
+
+ if (!VMCI_EVENT_VALID(event) || callback == NULL) {
+ pr_devel("Failed to subscribe to event (type=%d) "
+ "(callback=%p) (data=%p).", event,
+ callback, callback_data);
+ return VMCI_ERROR_INVALID_ARGS;
+ }
+
+ sub->run_delayed = flags & VMCI_FLAG_EVENT_DELAYED_CB;
+ sub->event = event;
+ sub->callback = callback;
+ sub->callback_data = callback_data;
+ kref_init(&sub->kref);
+ init_completion(&sub->done);
+ INIT_LIST_HEAD(&sub->node);
+
+ mutex_lock(&subscriber_mutex);
+
+ /* Creation of a new event is always allowed. */
+ for (attempts = 0; attempts < VMCI_EVENT_MAX_ATTEMPTS; attempts++) {
+ struct vmci_subscription *existing_sub = NULL;
+
+ /*
+ * We try to get an id a couple of time before
+ * claiming we are out of resources.
+ */
+ sub->id = ++subscription_id;
+
+ /* Test for duplicate id. */
+ existing_sub = event_find(sub->id);
+ if (!existing_sub) {
+ success = true;
+ break;
+ }
+ }
+
+ if (success) {
+ list_add_rcu(&sub->node, &subscriber_array[event]);
+ result = VMCI_SUCCESS;
+ } else {
+ result = VMCI_ERROR_NO_RESOURCES;
+ }
+
+ mutex_unlock(&subscriber_mutex);
+ return result;
+}
+
+/*
+ * Remove subscription from subscriber list.
+ */
+static struct vmci_subscription *event_unregister_subscription(u32 sub_id)
+{
+ struct vmci_subscription *s;
+
+ mutex_lock(&subscriber_mutex);
+ s = event_find(sub_id);
+ if (s)
+ list_del_rcu(&s->node);
+ mutex_unlock(&subscriber_mutex);
+
+ if (s) {
+ synchronize_rcu();
+ event_release(s);
+ wait_for_completion(&s->done);
+ }
+
+ return s;
+}
+
+/*
+ * vmci_event_subscribe() - Subscribe to a given event.
+ * @event: The event to subscribe to.
+ * @flags: Event flags. VMCI_FLAG_EVENT_*
+ * @callback: The callback to invoke upon the event.
+ * @callback_data: Data to pass to the callback.
+ * @subscription_id: ID used to track subscription. Used with
+ * vmci_event_unscribe()
+ *
+ * Subscribes to the provided event. The callback specified can be fired
+ * in different contexts depending on what flag is specified while
+ * registering. If flags contains VMCI_FLAG_EVENT_NONE then the
+ * callback is fired with the subscriber lock held (and BH context
+ * on the guest). If flags contain VMCI_FLAG_EVENT_DELAYED_CB then
+ * the callback is fired with no locks held in thread context.
+ * This is useful because other VMCIEvent functions can be called,
+ * but it also increases the chances that an event will be dropped.
+ */
+int vmci_event_subscribe(u32 event,
+ u32 flags,
+ vmci_event_cb callback,
+ void *callback_data,
+ u32 *subscription_id)
+{
+ int retval;
+ struct vmci_subscription *s = NULL;
+
+ if (subscription_id == NULL) {
+ pr_devel("Invalid subscription (NULL).");
+ return VMCI_ERROR_INVALID_ARGS;
+ }
+
+ s = kmalloc(sizeof(*s), GFP_KERNEL);
+ if (s == NULL)
+ return VMCI_ERROR_NO_MEM;
+
+ retval = event_register_subscription(s, event, flags,
+ callback, callback_data);
+ if (retval < VMCI_SUCCESS) {
+ kfree(s);
+ return retval;
+ }
+
+ *subscription_id = s->id;
+ return retval;
+}
+EXPORT_SYMBOL(vmci_event_subscribe);
+
+/*
+ * vmci_event_unsubscribe() - Unsubscribe to an event.
+ * @sub_id: A subscription ID ad provided by vmci_event_subscribe()
+ *
+ * Unsubscribe to given event. Removes it from list and frees it.
+ * Will return callback_data if requested by caller.
+ */
+int vmci_event_unsubscribe(u32 sub_id)
+{
+ struct vmci_subscription *s;
+
+ /*
+ * Return subscription. At this point we know noone else is accessing
+ * the subscription so we can free it.
+ */
+ s = event_unregister_subscription(sub_id);
+ if (s == NULL)
+ return VMCI_ERROR_NOT_FOUND;
+
+ kfree(s);
+
+ return VMCI_SUCCESS;
+}
+EXPORT_SYMBOL(vmci_event_unsubscribe);
diff --git a/drivers/misc/vmw_vmci/vmci_event.h b/drivers/misc/vmw_vmci/vmci_event.h
new file mode 100644
index 0000000..7df9b1c
--- /dev/null
+++ b/drivers/misc/vmw_vmci/vmci_event.h
@@ -0,0 +1,25 @@
+/*
+ * VMware VMCI Driver
+ *
+ * Copyright (C) 2012 VMware, Inc. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation version 2 and no later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ * for more details.
+ */
+
+#ifndef __VMCI_EVENT_H__
+#define __VMCI_EVENT_H__
+
+#include <linux/vmw_vmci_api.h>
+
+int vmci_event_init(void);
+void vmci_event_exit(void);
+int vmci_event_dispatch(struct vmci_datagram *msg);
+
+#endif /*__VMCI_EVENT_H__ */

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/