Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
authorLinus Torvalds <torvalds@woody.linux-foundation.org>
Fri, 22 Jun 2007 18:10:05 +0000 (11:10 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Fri, 22 Jun 2007 18:10:05 +0000 (11:10 -0700)
* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [AF_RXRPC]: Return the number of bytes buffered in rxrpc_send_data()
  [IPVS]: Fix state variable on failure to start ipvs threads
  [XFRM]: Fix MTU calculation for non-ESP SAs

net/ipv4/ipvs/ip_vs_sync.c
net/rxrpc/ar-output.c
net/xfrm/xfrm_state.c

index 7ea2d981a9328490df6dc684b6f1768fe0304f3d..356f067484e393634176357f193b9565cfdb2814 100644 (file)
@@ -67,6 +67,11 @@ struct ip_vs_sync_conn_options {
        struct ip_vs_seq        out_seq;        /* outgoing seq. struct */
 };
 
+struct ip_vs_sync_thread_data {
+       struct completion *startup;
+       int state;
+};
+
 #define IP_VS_SYNC_CONN_TIMEOUT (3*60*HZ)
 #define SIMPLE_CONN_SIZE  (sizeof(struct ip_vs_sync_conn))
 #define FULL_CONN_SIZE  \
@@ -751,6 +756,7 @@ static int sync_thread(void *startup)
        mm_segment_t oldmm;
        int state;
        const char *name;
+       struct ip_vs_sync_thread_data *tinfo = startup;
 
        /* increase the module use count */
        ip_vs_use_count_inc();
@@ -789,7 +795,14 @@ static int sync_thread(void *startup)
        add_wait_queue(&sync_wait, &wait);
 
        set_sync_pid(state, current->pid);
-       complete((struct completion *)startup);
+       complete(tinfo->startup);
+
+       /*
+        * once we call the completion queue above, we should
+        * null out that reference, since its allocated on the
+        * stack of the creating kernel thread
+        */
+       tinfo->startup = NULL;
 
        /* processing master/backup loop here */
        if (state == IP_VS_STATE_MASTER)
@@ -801,6 +814,14 @@ static int sync_thread(void *startup)
        remove_wait_queue(&sync_wait, &wait);
 
        /* thread exits */
+
+       /*
+        * If we weren't explicitly stopped, then we
+        * exited in error, and should undo our state
+        */
+       if ((!stop_master_sync) && (!stop_backup_sync))
+               ip_vs_sync_state -= tinfo->state;
+
        set_sync_pid(state, 0);
        IP_VS_INFO("sync thread stopped!\n");
 
@@ -812,6 +833,11 @@ static int sync_thread(void *startup)
        set_stop_sync(state, 0);
        wake_up(&stop_sync_wait);
 
+       /*
+        * we need to free the structure that was allocated
+        * for us in start_sync_thread
+        */
+       kfree(tinfo);
        return 0;
 }
 
@@ -838,11 +864,19 @@ int start_sync_thread(int state, char *mcast_ifn, __u8 syncid)
 {
        DECLARE_COMPLETION_ONSTACK(startup);
        pid_t pid;
+       struct ip_vs_sync_thread_data *tinfo;
 
        if ((state == IP_VS_STATE_MASTER && sync_master_pid) ||
            (state == IP_VS_STATE_BACKUP && sync_backup_pid))
                return -EEXIST;
 
+       /*
+        * Note that tinfo will be freed in sync_thread on exit
+        */
+       tinfo = kmalloc(sizeof(struct ip_vs_sync_thread_data), GFP_KERNEL);
+       if (!tinfo)
+               return -ENOMEM;
+
        IP_VS_DBG(7, "%s: pid %d\n", __FUNCTION__, current->pid);
        IP_VS_DBG(7, "Each ip_vs_sync_conn entry need %Zd bytes\n",
                  sizeof(struct ip_vs_sync_conn));
@@ -858,8 +892,11 @@ int start_sync_thread(int state, char *mcast_ifn, __u8 syncid)
                ip_vs_backup_syncid = syncid;
        }
 
+       tinfo->state = state;
+       tinfo->startup = &startup;
+
   repeat:
-       if ((pid = kernel_thread(fork_sync_thread, &startup, 0)) < 0) {
+       if ((pid = kernel_thread(fork_sync_thread, tinfo, 0)) < 0) {
                IP_VS_ERR("could not create fork_sync_thread due to %d... "
                          "retrying.\n", pid);
                msleep_interruptible(1000);
index 591c4422205e4bb9e76d3b7c71ca1d1c13fc8795..cc9102c5b588319e09b9769302bd5704e9bfb8e3 100644 (file)
@@ -640,6 +640,7 @@ static int rxrpc_send_data(struct kiocb *iocb,
                        goto efault;
                sp->remain -= copy;
                skb->mark += copy;
+               copied += copy;
 
                len -= copy;
                segment -= copy;
@@ -709,6 +710,8 @@ static int rxrpc_send_data(struct kiocb *iocb,
 
        } while (segment > 0);
 
+success:
+       ret = copied;
 out:
        call->tx_pending = skb;
        _leave(" = %d", ret);
@@ -725,7 +728,7 @@ call_aborted:
 
 maybe_error:
        if (copied)
-               ret = copied;
+               goto success;
        goto out;
 
 efault:
index 85f3f43a6cca402a3339aa4ef8317d7b41eaaa00..dfacb9c2a6e3861837b1a25b2d919bcc19b88d5a 100644 (file)
@@ -1729,7 +1729,7 @@ int xfrm_state_mtu(struct xfrm_state *x, int mtu)
            x->type && x->type->get_mtu)
                res = x->type->get_mtu(x, mtu);
        else
-               res = mtu;
+               res = mtu - x->props.header_len;
        spin_unlock_bh(&x->lock);
        return res;
 }