Re: [tip:x86/urgent] x86, mce: Fix RCU lockdep from mce_poll()

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Thu, Mar 31, 2011 at 12:03:48PM +0200, Zdenek Kabelac wrote:
> However on my machine - I do not a see a difference ?
> 
> I've this patch applied - and it still gives me the same error.
> (using 6aba74f2791287ec407e0f92487a725a25908067 and this patch)
> 
> Here are my RCU config options:
> 
> # grep RCU .config
> # RCU Subsystem
> CONFIG_TREE_PREEMPT_RCU=y
> CONFIG_PREEMPT_RCU=y
> CONFIG_RCU_TRACE=y
> CONFIG_RCU_FANOUT=64
> # CONFIG_RCU_FANOUT_EXACT is not set
> CONFIG_TREE_RCU_TRACE=y
> CONFIG_PROVE_RCU=y
> # CONFIG_PROVE_RCU_REPEATEDLY is not set
> # CONFIG_SPARSE_RCU_POINTER is not set
> # CONFIG_RCU_TORTURE_TEST is not set
> CONFIG_RCU_CPU_STALL_DETECTOR=y
> CONFIG_RCU_CPU_STALL_TIMEOUT=60
> CONFIG_RCU_CPU_STALL_DETECTOR_RUNNABLE=y
> CONFIG_RCU_CPU_STALL_VERBOSE=y

What idiot created that patch, anyway???  ;-)

Here is a patch that is much more likely to do something useful.
Could you please test it?  It is on top of the patch you just tested.

							Thanx, Paul

------------------------------------------------------------------------

rcu: create new rcu_access_index() and use in mce

The MCE subsystem needs to sample an RCU-protected index outside of
any protection for that index.  If this was a pointer, we would use
rcu_access_pointer(), but there is no corresponding rcu_access_index().
This commit therefore creates an rcu_access_index() and applies it
to MCE.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index a2d664f..3385ea2 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -1626,7 +1626,7 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
 	poll_wait(file, &mce_wait, wait);
-	if (rcu_dereference_index_check(mcelog.next, rcu_read_lock_sched_held()))
+	if (rcu_access_index(mcelog.next))
 		return POLLIN | POLLRDNORM;
 	if (!mce_apei_read_done && apei_check_mce())
 		return POLLIN | POLLRDNORM;
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index af56148..ff422d2 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -339,6 +339,12 @@ extern int rcu_my_thread_group_empty(void);
 		((typeof(*p) __force __kernel *)(p)); \
 	})
 
+#define __rcu_access_index(p, space) \
+	({ \
+		typeof(p) _________p1 = ACCESS_ONCE(p); \
+		rcu_dereference_sparse(p, space); \
+		(_________p1); \
+	})
 #define __rcu_dereference_index_check(p, c) \
 	({ \
 		typeof(p) _________p1 = ACCESS_ONCE(p); \
@@ -429,6 +435,20 @@ extern int rcu_my_thread_group_empty(void);
 #define rcu_dereference_raw(p) rcu_dereference_check(p, 1) /*@@@ needed? @@@*/
 
 /**
+ * rcu_access_index() - fetch RCU index with no dereferencing
+ * @p: The index to read
+ *
+ * Return the value of the specified RCU-protected index, but omit the
+ * smp_read_barrier_depends() and keep the ACCESS_ONCE().  This is useful
+ * when the value of this index is accessed, but the index is not
+ * dereferenced, for example, when testing an RCU-protected index against
+ * -1.  Although rcu_access_index() may also be used in cases where
+ * update-side locks prevent the value of the index from changing, you
+ * should instead use rcu_dereference_index_protected() for this use case.
+ */
+#define rcu_access_index(p) __rcu_access_index((p), __rcu)
+
+/**
  * rcu_dereference_index_check() - rcu_dereference for indices with debug checking
  * @p: The pointer to read, prior to dereferencing
  * @c: The conditions under which the dereference will take place