[RFC v2 1/1] RTL8712 alignment bug in 3.6 and up on ARMV5

[Josh Coombs]

I've been banging on my test setup for the past week.  I haven't
tracked down the cause yet, but something in my prior build system
results in this problem being masked with some builds.  Doing pure
vanilla builds from git I'm able to reproduce with 3.6 and the 3.7-rc
series reliably.  Support for my GFN doesn't exist officially prior to
3.6 and I didn't want to introduce new variables by back porting
support and trying to test with that patch set so I was unable to
isolate a start point for the failure I found.  I can reproduce the
failure with 3.7-rc6 currently.

Currently I'm running stable on 3.7-rc6 with the below patch and my
RealTek 8712 USB NIC.

I'm open to any and all suggestions on how to test further.

Signed-off-by: Joshua Coombs <josh.coombs at gmail.com>
--
diff -ruN a/drivers/staging/rtl8712/rtl871x_sta_mgt.c
b/drivers/staging/rtl8712/rtl871x_sta_mgt.c
--- a/drivers/staging/rtl8712/rtl871x_sta_mgt.c 2012-11-17
16:21:23.000000000 -0500
+++ b/drivers/staging/rtl8712/rtl871x_sta_mgt.c 2012-11-18
14:58:27.000000000 -0500
@@ -52,11 +52,11 @@
  s32 i;

  pstapriv->pallocated_stainfo_buf = _malloc(sizeof(struct sta_info) *
-   NUM_STA + 4);
+   NUM_STA + 8);
  if (pstapriv->pallocated_stainfo_buf == NULL)
  return _FAIL;
- pstapriv->pstainfo_buf = pstapriv->pallocated_stainfo_buf + 4 -
- ((addr_t)(pstapriv->pallocated_stainfo_buf) & 3);
+ pstapriv->pstainfo_buf = pstapriv->pallocated_stainfo_buf + 8 -
+ ((addr_t)(pstapriv->pallocated_stainfo_buf) & 7);
  _init_queue(&pstapriv->free_sta_queue);
  spin_lock_init(&pstapriv->sta_hash_lock);
  pstapriv->asoc_sta_count = 0;