mce: fix RCU lockdep from mce_log()

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

Thanks,
Davidlohr

From: Davidlohr Bueso <[email protected]>

Based on the following message:

===================================================
[ INFO: suspicious rcu_dereference_check() usage. ]
---------------------------------------------------
arch/x86/kernel/cpu/mcheck/mce.c:1628 invoked rcu_dereference_check() without protection!

other info that might help us debug this:

rcu_scheduler_active = 1, debug_locks = 1
no locks held by mcelog/2350.

stack backtrace:
Pid: 2350, comm: mcelog Tainted: G        W   2.6.37-rc1+ #7
Call Trace:
 [<ffffffff8108e6d4>] lockdep_rcu_dereference+0xa4/0xc0
 [<ffffffff810189e9>] mce_poll+0xa9/0xd0
 [<ffffffff81160585>] do_sys_poll+0x275/0x550
 [<ffffffff8115f0e0>] ? __pollwait+0x0/0xf0
 [<ffffffff8115f1d0>] ? pollwake+0x0/0x60
 [<ffffffff8115f1d0>] ? pollwake+0x0/0x60
 [<ffffffff8130431c>] ? rcu_read_lock_held+0x2c/0x30
 [<ffffffff8130549a>] ? radix_tree_lookup_element+0xda/0x100
 [<ffffffff81121f08>] ? __do_fault+0x128/0x470
 [<ffffffff81100bdb>] ? filemap_fault+0xdb/0x4e0
 [<ffffffff810ffe75>] ? unlock_page+0x25/0x30
 [<ffffffff81069ddf>] ? sigprocmask+0x3f/0x100
 [<ffffffff8183672b>] ? _raw_spin_unlock_irq+0x2b/0x60
 [<ffffffff8108fefd>] ? trace_hardirqs_on_caller+0x13d/0x180
 [<ffffffff8108ff4d>] ? trace_hardirqs_on+0xd/0x10
 [<ffffffff8183672b>] ? _raw_spin_unlock_irq+0x2b/0x60
 [<ffffffff811608a7>] sys_ppoll+0x47/0x190
 [<ffffffff8108fefd>] ? trace_hardirqs_on_caller+0x13d/0x180
 [<ffffffff81835d39>] ? trace_hardirqs_on_thunk+0x3a/0x3f
 [<ffffffff810030eb>] system_call_fastpath+0x16/0x1b

At this point the arch/x86/kernel/cpu/mcheck/mce.c:1628 invoked rcu_dereference_check() without protection!

other info that might help us debug this:

rcu_scheduler_active = 1, debug_locks = 1
no locks held by mcelog/2350.

stack backtrace:
Pid: 2350, comm: mcelog Tainted: G        W   2.6.37-rc1+ #7
Call Trace:
 [<ffffffff8108e6d4>] lockdep_rcu_dereference+0xa4/0xc0
 [<ffffffff810189e9>] mce_poll+0xa9/0xd0
 [<ffffffff81160585>] do_sys_poll+0x275/0x550
 [<ffffffff8115f0e0>] ? __pollwait+0x0/0xf0
 [<ffffffff8115f1d0>] ? pollwake+0x0/0x60
 [<ffffffff8115f1d0>] ? pollwake+0x0/0x60
 [<ffffffff8130431c>] ? rcu_read_lock_held+0x2c/0x30
 [<ffffffff8130549a>] ? radix_tree_lookup_element+0xda/0x100
 [<ffffffff81121f08>] ? __do_fault+0x128/0x470
 [<ffffffff81100bdb>] ? filemap_fault+0xdb/0x4e0
 [<ffffffff810ffe75>] ? unlock_page+0x25/0x30
 [<ffffffff81069ddf>] ? sigprocmask+0x3f/0x100
 [<ffffffff8183672b>] ? _raw_spin_unlock_irq+0x2b/0x60
 [<ffffffff8108fefd>] ? trace_hardirqs_on_caller+0x13d/0x180
 [<ffffffff8108ff4d>] ? trace_hardirqs_on+0xd/0x10
 [<ffffffff8183672b>] ? _raw_spin_unlock_irq+0x2b/0x60
 [<ffffffff811608a7>] sys_ppoll+0x47/0x190
 [<ffffffff8108fefd>] ? trace_hardirqs_on_caller+0x13d/0x180
 [<ffffffff81835d39>] ? trace_hardirqs_on_thunk+0x3a/0x3f
 [<ffffffff810030eb>] system_call_fastpath+0x16/0x1b

At this point the lockdep_is_held(&mce_read_mutex) call is failing.
So check if the mce_read_mutex is held before derefencing instead of using rcu_dereference_check_mce()

Signed-off-by Davidlohr Bueso <[email protected]>

---
 arch/x86/kernel/cpu/mcheck/mce.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

arch/x86/kernel/cpu/mcheck/mce.c +6 -2

index 7a35b72..6f95b2c 100644
@@ -1625,8 +1625,12 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
 	poll_wait(file, &mce_wait, wait);
-	if (rcu_dereference_check_mce(mcelog.next))
-		return POLLIN | POLLRDNORM;
+
+	if (mutex_is_locked(&mce_read_mutex)) {
+		if (rcu_dereference_index_check(mcelog.next, 
+						rcu_read_lock_sched_held()))
+			return POLLIN | POLLRDNORM;
+	}
 	if (!mce_apei_read_done && apei_check_mce())
 		return POLLIN | POLLRDNORM;
 	return 0;

Signature · 9 lines

1.7.1



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

Quoted 4 lines · Davidlohr Bueso

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

The patch is certainly not correct. The variable needs to be read independently of the mutex.

Quoted 8 lines · Andi Kleen

Quoted 4 lines · Davidlohr Bueso

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

The patch is certainly not correct. The variable needs to be read independently of the mutex.

This code is simply checking the value of the pointer, and therefore need not protect any actual dereferences. So why not replace the rcu_dereference_check_mce() with rcu_access_pointer()? If this is OK, please see the patch below.

BTW, assigning the value returned by rcu_access_pointer() into a variable often indicates a bug. ;-)

Thanx, Paul

Signed-off-by Paul E. McKenney <[email protected]>

arch/x86/kernel/cpu/mcheck/mce.c +0 -0

index 7a35b72..4d29d50 100644

Quoted 27 lines · Paul E. McKenney

Quoted 8 lines · Andi Kleen

Quoted 4 lines · Davidlohr Bueso

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

The patch is certainly not correct. The variable needs to be read independently of the mutex.

This code is simply checking the value of the pointer, and therefore need not protect any actual dereferences. So why not replace the rcu_dereference_check_mce() with rcu_access_pointer()? If this is OK, please see the patch below.

BTW, assigning the value returned by rcu_access_pointer() into a variable often indicates a bug. ;-)

Thanx, Paul

Signed-off-by Paul E. McKenney <[email protected]>

arch/x86/kernel/cpu/mcheck/mce.c +0 -0

index 7a35b72..4d29d50 100644

arch/x86/kernel/cpu/mcheck/mce.c out: · line 1628

Quoted 5 lines

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

@@ -1625,7 +1625,7 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
 	poll_wait(file, &mce_wait, wait);
-	if (rcu_dereference_check_mce(mcelog.next))
+	if (rcu_access_pointer(mcelog.next))

this doesn't compile (mcelog.next is an index):

arch/x86/kernel/cpu/mcheck/mce.c: In function ‘mce_poll’:
arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘_________p1’ arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’ arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: cast to pointer from integer of different size arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’
make[4]: *** [arch/x86/kernel/cpu/mcheck/mce.o] Error 1

Since the mutex is independent, what about this patch?

Signed-off-by Davidlohr Bueso <[email protected]>

---
 arch/x86/kernel/cpu/mcheck/mce.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

index 7a35b72..cc1c673 100644
@@ -1625,7 +1625,7 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
 	poll_wait(file, &mce_wait, wait);
-	if (rcu_dereference_check_mce(mcelog.next))
+	if (rcu_dereference_index_check(mcelog.next,
 rcu_read_lock_sched_held()))
 		return POLLIN | POLLRDNORM;
 	if (!mce_apei_read_done && apei_check_mce())

return POLLIN | POLLRDNORM;

Signature · 1 line

1.7.1

Quoted 54 lines · Davidlohr Bueso

Quoted 27 lines · Paul E. McKenney

Quoted 8 lines · Andi Kleen

Quoted 4 lines · Davidlohr Bueso

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

The patch is certainly not correct. The variable needs to be read independently of the mutex.

This code is simply checking the value of the pointer, and therefore need not protect any actual dereferences. So why not replace the rcu_dereference_check_mce() with rcu_access_pointer()? If this is OK, please see the patch below.

BTW, assigning the value returned by rcu_access_pointer() into a variable often indicates a bug. ;-)

Thanx, Paul

Signed-off-by Paul E. McKenney <[email protected]>

arch/x86/kernel/cpu/mcheck/mce.c +0 -0

index 7a35b72..4d29d50 100644

arch/x86/kernel/cpu/mcheck/mce.c out: · line 1628

Quoted 5 lines

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

@@ -1625,7 +1625,7 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
 	poll_wait(file, &mce_wait, wait);
-	if (rcu_dereference_check_mce(mcelog.next))
+	if (rcu_access_pointer(mcelog.next))

this doesn't compile (mcelog.next is an index):

arch/x86/kernel/cpu/mcheck/mce.c: In function ‘mce_poll’:
arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘_________p1’ arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’ arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: cast to pointer from integer of different size arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’
make[4]: *** [arch/x86/kernel/cpu/mcheck/mce.o] Error 1

Quoted 1 line

Since the mutex is independent, what about this patch?

Looks good to me!

Acked-by Paul E. McKenney <[email protected]>

Quoted 54 lines · Paul E. McKenney

Quoted 27 lines · Andi Kleen

Quoted 8 lines · Davidlohr Bueso

Quoted 4 lines

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

The patch is certainly not correct. The variable needs to be read independently of the mutex.

This code is simply checking the value of the pointer, and therefore need not protect any actual dereferences.  So why not replace the rcu_dereference_check_mce() with rcu_access_pointer()?  If this is OK, please see the patch below.

BTW, assigning the value returned by rcu_access_pointer() into a variable often indicates a bug.  ;-)

                                                    Thanx, Paul

Signed-off-by Paul E. McKenney <[email protected]>

arch/x86/kernel/cpu/mcheck/mce.c +0 -0

index 7a35b72..4d29d50 100644

arch/x86/kernel/cpu/mcheck/mce.c out: · line 1628

Quoted 5 lines

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

@@ -1625,7 +1625,7 @@ out:
  static unsigned int mce_poll(struct file *file, poll_table *wait)
  {
     poll_wait(file, &mce_wait, wait);
-   if (rcu_dereference_check_mce(mcelog.next))
+   if (rcu_access_pointer(mcelog.next))

this doesn't compile (mcelog.next is an index):

arch/x86/kernel/cpu/mcheck/mce.c: In function ‘mce_poll’:
arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘_________p1’ arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’ arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: cast to pointer from integer of different size arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’
make[4]: *** [arch/x86/kernel/cpu/mcheck/mce.o] Error 1

Quoted 1 line

Since the mutex is independent, what about this patch?

Looks good to me!

Acked-by Paul E. McKenney <[email protected]>

Quoted 6 lines

Signed-off-by Davidlohr Bueso <[email protected]>

---
 arch/x86/kernel/cpu/mcheck/mce.c |    2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)

Quoted 14 lines

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

index 7a35b72..cc1c673 100644
@@ -1625,7 +1625,7 @@ out:
  static unsigned int mce_poll(struct file *file, poll_table *wait)
  {
       poll_wait(file, &mce_wait, wait);
-     if (rcu_dereference_check_mce(mcelog.next))
+     if (rcu_dereference_index_check(mcelog.next,
 rcu_read_lock_sched_held()))
               return POLLIN | POLLRDNORM;
       if (!mce_apei_read_done && apei_check_mce())

Quoted 1 line

              return POLLIN | POLLRDNORM;

2010/11/8 Paul E. McKenney <[email protected]>:

Quoted 79 lines · Davidlohr Bueso

Quoted 54 lines · Paul E. McKenney

Quoted 27 lines · Andi Kleen

Quoted 8 lines · Davidlohr Bueso

Quoted 4 lines

Hi,

Please review this patch, I am not very familiar with MCE/RCU so I'm not sure that this is the correct fix (otherwise consider it a bug report :)). This does "fix" the message though and I can use MCE normally.

The patch is certainly not correct. The variable needs to be read independently of the mutex.

This code is simply checking the value of the pointer, and therefore need not protect any actual dereferences. So why not replace the rcu_dereference_check_mce() with rcu_access_pointer()? If this is OK, please see the patch below.

BTW, assigning the value returned by rcu_access_pointer() into a variable often indicates a bug. ;-)

Thanx, Paul

Signed-off-by Paul E. McKenney <[email protected]>

arch/x86/kernel/cpu/mcheck/mce.c +0 -0

index 7a35b72..4d29d50 100644

arch/x86/kernel/cpu/mcheck/mce.c out: · line 1628

Quoted 5 lines

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

@@ -1625,7 +1625,7 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
    poll_wait(file, &mce_wait, wait);
-   if (rcu_dereference_check_mce(mcelog.next))
+   if (rcu_access_pointer(mcelog.next))

this doesn't compile (mcelog.next is an index):

arch/x86/kernel/cpu/mcheck/mce.c: In function ‘mce_poll’:
arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘_________p1’ arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’ arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: cast to pointer from integer of different size arch/x86/kernel/cpu/mcheck/mce.c:1628: error: invalid type argument of ‘unary *’ (have ‘unsigned int’) arch/x86/kernel/cpu/mcheck/mce.c:1628: warning: type defaults to ‘int’ in declaration of ‘type name’
make[4]: *** [arch/x86/kernel/cpu/mcheck/mce.o] Error 1

Quoted 1 line

Since the mutex is independent, what about this patch?

Looks good to me!

Acked-by Paul E. McKenney <[email protected]>

Quoted 18 lines

Signed-off-by Davidlohr Bueso <[email protected]>

---
 arch/x86/kernel/cpu/mcheck/mce.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

index 7a35b72..cc1c673 100644
@@ -1625,7 +1625,7 @@ out:
 static unsigned int mce_poll(struct file *file, poll_table *wait)
 {
      poll_wait(file, &mce_wait, wait);
-     if (rcu_dereference_check_mce(mcelog.next))
+     if (rcu_dereference_index_check(mcelog.next,
 rcu_read_lock_sched_held()))

Quoted 3 lines

Quoted 3 lines

arch/x86/kernel/cpu/mcheck/mce.c +0 -0

              return POLLIN | POLLRDNORM;
      if (!mce_apei_read_done && apei_check_mce())

              return POLLIN | POLLRDNORM;

Any chance to have this ever fixed upstream ?
(still happens with today's vanialla build)

I'm still quite interested in getting this fixed, I run into it several times a day.

Commit-ID:  706453b0d1fabf4dccbcffa29356d0dd6ab9afb9
Gitweb:     http://git.kernel.org/tip/706453b0d1fabf4dccbcffa29356d0dd6ab9afb9
Author:     Davidlohr Bueso <[email protected]>
AuthorDate: Wed, 30 Mar 2011 23:13:24 -0300
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 31 Mar 2011 11:06:08 +0200

x86, mce: Fix RCU lockdep from mce_poll()

Based on the following message:

 ===================================================
 [ INFO: suspicious rcu_dereference_check() usage. ]
 ---------------------------------------------------
 arch/x86/kernel/cpu/mcheck/mce.c:1628 invoked rcu_dereference_check() without protection!

 other info that might help us debug this:

 rcu_scheduler_active = 1, debug_locks = 1
 no locks held by mcelog/2350.

 stack backtrace:
 Pid: 2350, comm: mcelog Tainted: G        W   2.6.37-rc1+ #7
 Call Trace:
  [<ffffffff8108e6d4>] lockdep_rcu_dereference+0xa4/0xc0
  [<ffffffff810189e9>] mce_poll+0xa9/0xd0
  [<ffffffff81160585>] do_sys_poll+0x275/0x550
  [<ffffffff8115f0e0>] ? __pollwait+0x0/0xf0
  [<ffffffff8115f1d0>] ? pollwake+0x0/0x60
  [<ffffffff8115f1d0>] ? pollwake+0x0/0x60
  [<ffffffff8130431c>] ? rcu_read_lock_held+0x2c/0x30
  [<ffffffff8130549a>] ? radix_tree_lookup_element+0xda/0x100
  [<ffffffff81121f08>] ? __do_fault+0x128/0x470
  [<ffffffff81100bdb>] ? filemap_fault+0xdb/0x4e0
  [<ffffffff810ffe75>] ? unlock_page+0x25/0x30
  [<ffffffff81069ddf>] ? sigprocmask+0x3f/0x100
  [<ffffffff8183672b>] ? _raw_spin_unlock_irq+0x2b/0x60
  [<ffffffff8108fefd>] ? trace_hardirqs_on_caller+0x13d/0x180
  [<ffffffff8108ff4d>] ? trace_hardirqs_on+0xd/0x10
  [<ffffffff8183672b>] ? _raw_spin_unlock_irq+0x2b/0x60
  [<ffffffff811608a7>] sys_ppoll+0x47/0x190
  [<ffffffff8108fefd>] ? trace_hardirqs_on_caller+0x13d/0x180
  [<ffffffff81835d39>] ? trace_hardirqs_on_thunk+0x3a/0x3f
  [<ffffffff810030eb>] system_call_fastpath+0x16/0x1b

This code is simply checking the value of the pointer, and therefore need not protect any actual dereferences. Replace rcu_dereference_check_mce() with rcu_dereference_index_check().

Signed-off-by Davidlohr Bueso <[email protected]>

Acked-by Paul E. McKenney <[email protected]>

Cc Zdenek Kabelac <[email protected]>

Cc Andi Kleen <[email protected]>

LKML-Reference: <1301537604.2140.21.camel@offworld>

Signed-off-by Ingo Molnar <[email protected]>

---
 arch/x86/kernel/cpu/mcheck/mce.c |    2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

index 5a05ef6..a2d664f 100644
@@ -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_check_mce(mcelog.next))
+       if (rcu_dereference_index_check(mcelog.next, rcu_read_lock_sched_held()))
                return POLLIN | POLLRDNORM;
        if (!mce_apei_read_done && apei_check_mce())
                return POLLIN | POLLRDNORM;

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

Quoted 7 lines

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:

Quoted 16 lines

# 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 <[email protected]>

arch/x86/kernel/cpu/mcheck/mce.c +1 -1

index a2d664f..3385ea2 100644
@@ -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;

include/linux/rcupdate.h +13 -0

index af56148..ff422d2 100644
@@ -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

2011/3/31 Paul E. McKenney <[email protected]>:

Quoted 1 line · Zdenek Kabelac

Quoted 5 lines

+/**
 * 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

Works for me.

Tested-by Zdenek Kabelac <[email protected]>