2021-02-26 17:55:38

by Shuah Khan

[permalink] [raw]
Subject: [PATCH v2 0/3] Add lockdep_assert_not_held()

Some kernel functions must not be called holding a specific lock. Doing
so could lead to locking problems. Currently these routines call
lock_is_held() to check for lock hold followed by WARN_ON.

Adding a common lockdep interface will help reduce the duplication of this
logic in the rest of the kernel.

Add lockdep_assert_not_held() to be used in these functions to detect
incorrect calls while holding a lock.

lockdep_assert_not_held() provides the opposite functionality of
lockdep_assert_held() which is used to assert calls that require
holding a specific lock.

The need for lockdep_assert_not_held() came up in a discussion on
ath10k patch. ath10k_drain_tx() and i915_vma_pin_ww() are examples
of functions that can use lockdep_assert_not_held().

Link: https://lore.kernel.org/lkml/[email protected]nuxfoundation.org/
Link: https://lore.kernel.org/linux-wireless/[email protected]/

This patch series adds lockdep_assert_not_held() and uses it in the
second patch in ath10k_drain_tx() function.

Patch 1 incorporates suggestions from Peter Zijlstra on v1 series
to avoid misfires when lockdep_off() is employed.

Patch 2 Johannes Berg's suggestions as it make it easier to read and
maintain the lock states. These are defines and a enum to avoid changes
to lock_is_held_type() and lockdep_is_held() return types.

Patch 2 is a separate patch because it adds defines to lockdep.h and
kernel/locking/lockdep.c now includes lockdep.h - decided make this
a separate patch just in case issues related to header dependencies
pop up. I can combine Patches 1&2 if that is preferred.

Patch 3 uses the new interface in ath10k_drain_tx() function. Added
Kalle Valo's Ack from v1 for this change.

Tested on the mainline from yesterday.

Shuah Khan (3):
lockdep: add lockdep_assert_not_held()
lockdep: add lockdep lock state defines
ath10k: detect conf_mutex held ath10k_drain_tx() calls

drivers/net/wireless/ath/ath10k/mac.c | 2 ++
include/linux/lockdep.h | 18 +++++++++++++++---
kernel/locking/lockdep.c | 6 +++++-
3 files changed, 22 insertions(+), 4 deletions(-)

--
2.27.0


2021-02-26 17:55:49

by Shuah Khan

[permalink] [raw]
Subject: [PATCH v2 2/3] lockdep: add lockdep lock state defines

Adds defines for lock state returns from lock_is_held_type() based on
Johannes Berg's suggestions as it make it easier to read and maintain
the lock states. These are defines and a enum to avoid changes to
lock_is_held_type() and lockdep_is_held() return types.

Link: https://lore.kernel.org/lkml/[email protected]nuxfoundation.org/
Link: https://lore.kernel.org/linux-wireless/[email protected]/
Signed-off-by: Shuah Khan <[email protected]>
---
include/linux/lockdep.h | 11 +++++++++--
kernel/locking/lockdep.c | 3 ++-
2 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
index 93eb5f797fc1..7c2622854152 100644
--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -261,6 +261,11 @@ extern void lock_acquire(struct lockdep_map *lock, unsigned int subclass,

extern void lock_release(struct lockdep_map *lock, unsigned long ip);

+/* lock_is_held_type() returns */
+#define LOCK_STATE_UNKNOWN -1
+#define LOCK_STATE_NOT_HELD 0
+#define LOCK_STATE_HELD 1
+
/*
* Same "read" as for lock_acquire(), except -1 means any.
*/
@@ -295,11 +300,13 @@ extern void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie);
#define lockdep_depth(tsk) (debug_locks ? (tsk)->lockdep_depth : 0)

#define lockdep_assert_held(l) do { \
- WARN_ON(debug_locks && lockdep_is_held(l) == 0); \
+ WARN_ON(debug_locks && \
+ lockdep_is_held(l) == LOCK_STATE_NOT_HELD); \
} while (0)

#define lockdep_assert_not_held(l) do { \
- WARN_ON(debug_locks && lockdep_is_held(l) == 1); \
+ WARN_ON(debug_locks && \
+ lockdep_is_held(l) == LOCK_STATE_HELD); \
} while (0)

#define lockdep_assert_held_write(l) do { \
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index d3299fd85c4a..e3898c888ad2 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -54,6 +54,7 @@
#include <linux/nmi.h>
#include <linux/rcupdate.h>
#include <linux/kprobes.h>
+#include <linux/lockdep.h>

#include <asm/sections.h>

@@ -5475,7 +5476,7 @@ noinstr int lock_is_held_type(const struct lockdep_map *lock, int read)
/* avoid false negative lockdep_assert_not_held()
* and lockdep_assert_held()
*/
- return -1;
+ return LOCK_STATE_UNKNOWN;

raw_local_irq_save(flags);
check_flags(flags);
--
2.27.0

2021-02-26 17:56:37

by Shuah Khan

[permalink] [raw]
Subject: [PATCH v2 3/3] ath10k: detect conf_mutex held ath10k_drain_tx() calls

ath10k_drain_tx() must not be called with conf_mutex held as workers can
use that also. Add call to lockdep_assert_not_held() on conf_mutex to
detect if conf_mutex is held by the caller.

The idea for this patch stemmed from coming across the comment block
above the ath10k_drain_tx() while reviewing the conf_mutex holds during
to debug the conf_mutex lock assert in ath10k_debug_fw_stats_request().

Adding detection to assert on conf_mutex hold will help detect incorrect
usages that could lead to locking problems when async worker routines try
to call this routine.

Link: https://lore.kernel.org/lkml/[email protected]nuxfoundation.org/
Link: https://lore.kernel.org/linux-wireless/[email protected]/
Acked-by: Kalle Valo <[email protected]>
Signed-off-by: Shuah Khan <[email protected]>
---
drivers/net/wireless/ath/ath10k/mac.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index 7d98250380ec..a595ff4f0843 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -4567,6 +4567,8 @@ static void ath10k_mac_op_wake_tx_queue(struct ieee80211_hw *hw,
/* Must not be called with conf_mutex held as workers can use that also. */
void ath10k_drain_tx(struct ath10k *ar)
{
+ lockdep_assert_not_held(&ar->conf_mutex);
+
/* make sure rcu-protected mac80211 tx path itself is drained */
synchronize_net();

--
2.27.0

2021-02-26 18:05:36

by Johannes Berg

[permalink] [raw]
Subject: Re: [PATCH v2 2/3] lockdep: add lockdep lock state defines


> @@ -5475,7 +5476,7 @@ noinstr int lock_is_held_type(const struct lockdep_map *lock, int read)
> /* avoid false negative lockdep_assert_not_held()
> * and lockdep_assert_held()
> */
> - return -1;
> + return LOCK_STATE_UNKNOWN;

I'd argue that then the other two return places here should also be
changed.

johannes

2021-02-26 18:18:02

by Shuah Khan

[permalink] [raw]
Subject: Re: [PATCH v2 2/3] lockdep: add lockdep lock state defines

On 2/26/21 11:03 AM, Johannes Berg wrote:
>
>> @@ -5475,7 +5476,7 @@ noinstr int lock_is_held_type(const struct lockdep_map *lock, int read)
>> /* avoid false negative lockdep_assert_not_held()
>> * and lockdep_assert_held()
>> */
>> - return -1;
>> + return LOCK_STATE_UNKNOWN;
>
> I'd argue that then the other two return places here should also be
> changed.
>

Makes sense.

Since lock_is_held_type() simply returns what __lock_is_held() for the
other cases, __lock_is_held() is the one that needs changes to use these
defines.

thanks,
-- Shuah