[PATCH] IPMI: Fix some RCU problems

Fix some RCU problem pointed out by Paul McKenney of IBM.  These are:

The wholesale move of the command receivers list into a new list was not
safe because the list will point to the new tail during a traversal, so the
traversal will never end on a reader if this happens during a read.

Memory barriers were needed to handle proper ordering of the setting of the
IPMI interface as valid.  Readers might not see proper ordering of data
otherwise.

In ipmi_smi_watcher_register(), the use of the _rcu suffix on the list is
unnecessary.

This require the list_splice_init_rcu() patch previously posted.

Signed-off-by: Corey Minyard <minyard@acm.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/drivers/char/ipmi/ipmi_msghandler.c b/drivers/char/ipmi/ipmi_msghandler.c
index 53582b53da95c449a9cb730f1b93734793139c30..230064ede08d9ade7c9fa26f3ce5e7f76127ccd6 100644 (file)
--- a/drivers/char/ipmi/ipmi_msghandler.c
+++ b/drivers/char/ipmi/ipmi_msghandler.c
@@ -406,13 +406,14 @@ static void clean_up_interface_data(ipmi_smi_t intf)
        free_smi_msg_list(&intf->waiting_msgs);
        free_recv_msg_list(&intf->waiting_events);
 
-       /* Wholesale remove all the entries from the list in the
-        * interface and wait for RCU to know that none are in use. */
+       /*
+        * Wholesale remove all the entries from the list in the
+        * interface and wait for RCU to know that none are in use.
+        */
        mutex_lock(&intf->cmd_rcvrs_mutex);
-       list_add_rcu(&list, &intf->cmd_rcvrs);
-       list_del_rcu(&intf->cmd_rcvrs);
+       INIT_LIST_HEAD(&list);
+       list_splice_init_rcu(&intf->cmd_rcvrs, &list, synchronize_rcu);
        mutex_unlock(&intf->cmd_rcvrs_mutex);
-       synchronize_rcu();
 
        list_for_each_entry_safe(rcvr, rcvr2, &list, link)
                kfree(rcvr);
@@ -451,7 +452,7 @@ int ipmi_smi_watcher_register(struct ipmi_smi_watcher *watcher)
        mutex_lock(&ipmi_interfaces_mutex);
 
        /* Build a list of things to deliver. */
-       list_for_each_entry_rcu(intf, &ipmi_interfaces, link) {
+       list_for_each_entry(intf, &ipmi_interfaces, link) {
                if (intf->intf_num == -1)
                        continue;
                e = kmalloc(sizeof(*e), GFP_KERNEL);
@@ -2760,9 +2761,15 @@ int ipmi_register_smi(struct ipmi_smi_handlers *handlers,
                synchronize_rcu();
                kref_put(&intf->refcount, intf_free);
        } else {
-               /* After this point the interface is legal to use. */
+               /*
+                * Keep memory order straight for RCU readers.  Make
+                * sure everything else is committed to memory before
+                * setting intf_num to mark the interface valid.
+                */
+               smp_wmb();
                intf->intf_num = i;
                mutex_unlock(&ipmi_interfaces_mutex);
+               /* After this point the interface is legal to use. */
                call_smi_watchers(i, intf->si_dev);
                mutex_unlock(&smi_watchers_mutex);
        }
@@ -3923,6 +3930,14 @@ static void send_panic_events(char *str)
                        /* Interface was not ready yet. */
                        continue;
 
+               /*
+                * intf_num is used as an marker to tell if the
+                * interface is valid.  Thus we need a read barrier to
+                * make sure data fetched before checking intf_num
+                * won't be used.
+                */
+               smp_rmb();
+
                /* First job here is to figure out where to send the
                   OEM events.  There's no way in IPMI to send OEM
                   events using an event send command, so we have to