[PATCH 1/1] Fix hang on IBM Token Ring PCMCIA card ejection

[PATCH 1/1] Fix hang on IBM Token Ring PCMCIA card ejection

[Paul Walmsley]

From: Paul Walmsley <paul@booyaka.com>

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.  It seems that ibmtr.c:tok_interrupt() doesn't
expect to be called before tok_open() is called, but tok_interrupt()
gets called anyway when the card is ejected.  So, set an already-extant
flag which causes tok_interrupt() to bail out early upon card ejection.
Tested by inserting and removing the PCMCIA card several times.

Applies against 2.6.21-rc5.


Signed-off-by: Paul Walmsley <paul@booyaka.com>

---
  dev/drivers/net/pcmcia/ibmtr_cs.c |   12 ++++++++----
  1 file changed, 8 insertions(+), 4 deletions(-)

Index: linux/dev/drivers/net/pcmcia/ibmtr_cs.c
===================================================================
--- linux.orig/dev/drivers/net/pcmcia/ibmtr_cs.c
+++ linux/dev/drivers/net/pcmcia/ibmtr_cs.c
@@ -189,16 +189,20 @@ static void ibmtr_detach(struct pcmcia_d
  {
      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);

Re: [PATCH 1/1] Fix hang on IBM Token Ring PCMCIA card ejection

[Jeff Garzik]

Paul Walmsley wrote:
> From: Paul Walmsley <paul@booyaka.com>
> 
> 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.  It seems that ibmtr.c:tok_interrupt() doesn't
> expect to be called before tok_open() is called, but tok_interrupt()
> gets called anyway when the card is ejected.  So, set an already-extant
> flag which causes tok_interrupt() to bail out early upon card ejection.
> Tested by inserting and removing the PCMCIA card several times.
> 
> Applies against 2.6.21-rc5.
> 
> 
> Signed-off-by: Paul Walmsley <paul@booyaka.com>

ACK but does not apply

Re: [PATCH 1/1] Fix hang on IBM Token Ring PCMCIA card ejection

[Paul Walmsley]

On Tue, 8 May 2007, Jeff Garzik wrote:

> ACK but does not apply

Thanks.  Some mailer trouble I guess - will resend via quilt.

- Paul