Fix hang on IBM Token Ring PCMCIA card ejection
authorPaul Walmsley <paul@booyaka.com>
Wed, 9 May 2007 16:47:16 +0000 (10:47 -0600)
committerJeff Garzik <jeff@garzik.org>
Wed, 9 May 2007 23:25:15 +0000 (19:25 -0400)
Ejecting a PCMCIA IBM Token Ring card that has not had its dev->open()
called will reliably trigger an uninitialized spinlock oops when
spinlock debugging is enabled. The system then hangs, occasionally
softlockup oopsing.  Apparently ibmtr.c:tok_interrupt() doesn't expect
to be called before tok_open(), but tok_interrupt() gets called anyway
when the card is ejected.  So, set an already-existing flag which
causes tok_interrupt() to bail out early upon card ejection. Tested by
inserting and removing the PCMCIA card several times.

Signed-off-by: Paul Walmsley <paul@booyaka.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
drivers/net/pcmcia/ibmtr_cs.c

index 1060154ae750279138d8bd9414a3049beba5a369..4ecb8ca5a992ec318cfd9c58cbc21f116d56287c 100644 (file)
@@ -189,16 +189,20 @@ static void ibmtr_detach(struct pcmcia_device *link)
 {
     struct ibmtr_dev_t *info = link->priv;
     struct net_device *dev = info->dev;
+     struct tok_info *ti = netdev_priv(dev);
 
     DEBUG(0, "ibmtr_detach(0x%p)\n", link);
+    
+    /* 
+     * When the card removal interrupt hits tok_interrupt(), 
+     * bail out early, so we don't crash the machine 
+     */
+    ti->sram_phys |= 1;
 
     if (link->dev_node)
        unregister_netdev(dev);
-
-    {
-       struct tok_info *ti = netdev_priv(dev);
-       del_timer_sync(&(ti->tr_timer));
-    }
+    
+    del_timer_sync(&(ti->tr_timer));
 
     ibmtr_release(link);