[SCSI] cxgb3i: convert cdev->l2opt to use rcu to prevent NULL dereference

This oops was reported recently:
d:mon> e
cpu 0xd: Vector: 300 (Data Access) at [c0000000fd4c7120]
    pc: d00000000076f194: .t3_l2t_get+0x44/0x524 [cxgb3]
    lr: d000000000b02108: .init_act_open+0x150/0x3d4 [cxgb3i]
    sp: c0000000fd4c73a0
   msr: 8000000000009032
   dar: 0
 dsisr: 40000000
  current = 0xc0000000fd640d40
  paca    = 0xc00000000054ff80
    pid   = 5085, comm = iscsid
d:mon> t
[c0000000fd4c7450] d000000000b02108 .init_act_open+0x150/0x3d4 [cxgb3i]
[c0000000fd4c7500] d000000000e45378 .cxgbi_ep_connect+0x784/0x8e8 [libcxgbi]
[c0000000fd4c7650] d000000000db33f0 .iscsi_if_rx+0x71c/0xb18
[scsi_transport_iscsi2]
[c0000000fd4c7740] c000000000370c9c .netlink_data_ready+0x40/0xa4
[c0000000fd4c77c0] c00000000036f010 .netlink_sendskb+0x4c/0x9c
[c0000000fd4c7850] c000000000370c18 .netlink_sendmsg+0x358/0x39c
[c0000000fd4c7950] c00000000033be24 .sock_sendmsg+0x114/0x1b8
[c0000000fd4c7b50] c00000000033d208 .sys_sendmsg+0x218/0x2ac
[c0000000fd4c7d70] c00000000033f55c .sys_socketcall+0x228/0x27c
[c0000000fd4c7e30] c0000000000086a4 syscall_exit+0x0/0x40
--- Exception: c01 (System Call) at 00000080da560cfc

The root cause was an EEH error, which sent us down the offload_close path in
the cxgb3 driver, which in turn sets cdev->l2opt to NULL, without regard for
upper layer driver (like the cxgbi drivers) which might have execution contexts
in the middle of its use. The result is the oops above, when t3_l2t_get attempts
to dereference L2DATA(cdev)->nentries in arp_hash right after the EEH error handler sets it to NULL.

The fix is to prevent the setting of the NULL pointer until after there are no
further users of it.  The t3cdev->l2opt pointer is now converted to be an rcu
pointer and the L2DATA macro is now called under the protection of the
rcu_read_lock().  When the EEH error path:
t3_adapter_error->offload_close->cxgb3_offload_deactivate
Is exectured, setting of that l2opt pointer to NULL, is now gated on an rcu
quiescence point, preventing, allowing L2DATA callers to safely check for a NULL
pointer without concern that the underlying data will be freeded before the
pointer is dereferenced.

This has been tested by the reporter and shown to fix the reproted oops

[nhorman: fix up unitinialised variable reported by Dan Carpenter]
Signed-off-by: Neil Horman <[email protected]>
Reviewed-by: Karen Xie <[email protected]>
Cc: [email protected]
Signed-off-by: James Bottomley <[email protected]>

Author: nhorman

drivers/infiniband/hw/cxgb3/iwch_cm.c

@@ -287,7 +287,7 @@ void __free_ep(struct kref *kref)
 	if (test_bit(RELEASE_RESOURCES, &ep->com.flags)) {
 		cxgb3_remove_tid(ep->com.tdev, (void *)ep, ep->hwtid);
 		dst_release(ep->dst);
-		l2t_release(L2DATA(ep->com.tdev), ep->l2t);
+		l2t_release(ep->com.tdev, ep->l2t);
 	}
 	kfree(ep);
 }
@@ -1178,7 +1178,7 @@ static int act_open_rpl(struct t3cdev *tdev, struct sk_buff *skb, void *ctx)
 		release_tid(ep->com.tdev, GET_TID(rpl), NULL);
 	cxgb3_free_atid(ep->com.tdev, ep->atid);
 	dst_release(ep->dst);
-	l2t_release(L2DATA(ep->com.tdev), ep->l2t);
+	l2t_release(ep->com.tdev, ep->l2t);
 	put_ep(&ep->com);
 	return CPL_RET_BUF_DONE;
 }
@@ -1377,7 +1377,7 @@ static int pass_accept_req(struct t3cdev *tdev, struct sk_buff *skb, void *ctx)
 	if (!child_ep) {
 		printk(KERN_ERR MOD "%s - failed to allocate ep entry!\n",
 		       __func__);
-		l2t_release(L2DATA(tdev), l2t);
+		l2t_release(tdev, l2t);
 		dst_release(dst);
 		goto reject;
 	}
@@ -1956,7 +1956,7 @@ int iwch_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	if (!err)
 		goto out;
 
-	l2t_release(L2DATA(h->rdev.t3cdev_p), ep->l2t);
+	l2t_release(h->rdev.t3cdev_p, ep->l2t);
 fail4:
 	dst_release(ep->dst);
 fail3:
@@ -2127,7 +2127,7 @@ int iwch_ep_redirect(void *ctx, struct dst_entry *old, struct dst_entry *new,
 	PDBG("%s ep %p redirect to dst %p l2t %p\n", __func__, ep, new,
 	     l2t);
 	dst_hold(new);
-	l2t_release(L2DATA(ep->com.tdev), ep->l2t);
+	l2t_release(ep->com.tdev, ep->l2t);
 	ep->l2t = l2t;
 	dst_release(old);
 	ep->dst = new;

drivers/net/cxgb3/cxgb3_offload.c

@@ -1146,12 +1146,14 @@ static void cxgb_redirect(struct dst_entry *old, struct dst_entry *new)
 		if (te && te->ctx && te->client && te->client->redirect) {
 			update_tcb = te->client->redirect(te->ctx, old, new, e);
 			if (update_tcb) {
+				rcu_read_lock();
 				l2t_hold(L2DATA(tdev), e);
+				rcu_read_unlock();
 				set_l2t_ix(tdev, tid, e);
 			}
 		}
 	}
-	l2t_release(L2DATA(tdev), e);
+	l2t_release(tdev, e);
 }
 
 /*
@@ -1264,7 +1266,7 @@ int cxgb3_offload_activate(struct adapter *adapter)
 		goto out_free;
 
 	err = -ENOMEM;
-	L2DATA(dev) = t3_init_l2t(l2t_capacity);
+	RCU_INIT_POINTER(dev->l2opt, t3_init_l2t(l2t_capacity));
 	if (!L2DATA(dev))
 		goto out_free;
 
@@ -1298,25 +1300,36 @@ int cxgb3_offload_activate(struct adapter *adapter)
 
 out_free_l2t:
 	t3_free_l2t(L2DATA(dev));
-	L2DATA(dev) = NULL;
+	rcu_assign_pointer(dev->l2opt, NULL);
 out_free:
 	kfree(t);
 	return err;
 }
 
+static void clean_l2_data(struct rcu_head *head)
+{
+	struct l2t_data *d = container_of(head, struct l2t_data, rcu_head);
+	t3_free_l2t(d);
+}
+
+
 void cxgb3_offload_deactivate(struct adapter *adapter)
 {
 	struct t3cdev *tdev = &adapter->tdev;
 	struct t3c_data *t = T3C_DATA(tdev);
+	struct l2t_data *d;
 
 	remove_adapter(adapter);
 	if (list_empty(&adapter_list))
 		unregister_netevent_notifier(&nb);
 
 	free_tid_maps(&t->tid_maps);
 	T3C_DATA(tdev) = NULL;
-	t3_free_l2t(L2DATA(tdev));
-	L2DATA(tdev) = NULL;
+	rcu_read_lock();
+	d = L2DATA(tdev);
+	rcu_read_unlock();
+	rcu_assign_pointer(tdev->l2opt, NULL);
+	call_rcu(&d->rcu_head, clean_l2_data);
 	if (t->nofail_skb)
 		kfree_skb(t->nofail_skb);
 	kfree(t);

drivers/net/cxgb3/l2t.c

@@ -300,14 +300,21 @@ static inline void reuse_entry(struct l2t_entry *e, struct neighbour *neigh)
 struct l2t_entry *t3_l2t_get(struct t3cdev *cdev, struct neighbour *neigh,
 			     struct net_device *dev)
 {
-	struct l2t_entry *e;
-	struct l2t_data *d = L2DATA(cdev);
+	struct l2t_entry *e = NULL;
+	struct l2t_data *d;
+	int hash;
 	u32 addr = *(u32 *) neigh->primary_key;
 	int ifidx = neigh->dev->ifindex;
-	int hash = arp_hash(addr, ifidx, d);
 	struct port_info *p = netdev_priv(dev);
 	int smt_idx = p->port_id;
 
+	rcu_read_lock();
+	d = L2DATA(cdev);
+	if (!d)
+		goto done_rcu;
+
+	hash = arp_hash(addr, ifidx, d);
+
 	write_lock_bh(&d->lock);
 	for (e = d->l2tab[hash].first; e; e = e->next)
 		if (e->addr == addr && e->ifindex == ifidx &&
@@ -338,6 +345,8 @@ struct l2t_entry *t3_l2t_get(struct t3cdev *cdev, struct neighbour *neigh,
 	}
 done:
 	write_unlock_bh(&d->lock);
+done_rcu:
+	rcu_read_unlock();
 	return e;
 }
 

drivers/net/cxgb3/l2t.h

@@ -76,6 +76,7 @@ struct l2t_data {
 	atomic_t nfree;		/* number of free entries */
 	rwlock_t lock;
 	struct l2t_entry l2tab[0];
+	struct rcu_head rcu_head;	/* to handle rcu cleanup */
 };
 
 typedef void (*arp_failure_handler_func)(struct t3cdev * dev,
@@ -99,7 +100,7 @@ static inline void set_arp_failure_handler(struct sk_buff *skb,
 /*
  * Getting to the L2 data from an offload device.
  */
-#define L2DATA(dev) ((dev)->l2opt)
+#define L2DATA(cdev) (rcu_dereference((cdev)->l2opt))
 
 #define W_TCB_L2T_IX    0
 #define S_TCB_L2T_IX    7
@@ -126,15 +127,22 @@ static inline int l2t_send(struct t3cdev *dev, struct sk_buff *skb,
 	return t3_l2t_send_slow(dev, skb, e);
 }
 
-static inline void l2t_release(struct l2t_data *d, struct l2t_entry *e)
+static inline void l2t_release(struct t3cdev *t, struct l2t_entry *e)
 {
-	if (atomic_dec_and_test(&e->refcnt))
+	struct l2t_data *d;
+
+	rcu_read_lock();
+	d = L2DATA(t);
+
+	if (atomic_dec_and_test(&e->refcnt) && d)
 		t3_l2e_free(d, e);
+
+	rcu_read_unlock();
 }
 
 static inline void l2t_hold(struct l2t_data *d, struct l2t_entry *e)
 {
-	if (atomic_add_return(1, &e->refcnt) == 1)	/* 0 -> 1 transition */
+	if (d && atomic_add_return(1, &e->refcnt) == 1)	/* 0 -> 1 transition */
 		atomic_dec(&d->nfree);
 }
 

drivers/scsi/cxgbi/cxgb3i/cxgb3i.c

@@ -913,7 +913,7 @@ static void l2t_put(struct cxgbi_sock *csk)
 	struct t3cdev *t3dev = (struct t3cdev *)csk->cdev->lldev;
 
 	if (csk->l2t) {
-		l2t_release(L2DATA(t3dev), csk->l2t);
+		l2t_release(t3dev, csk->l2t);
 		csk->l2t = NULL;
 		cxgbi_sock_put(csk);
 	}