inst->handles is traversed using list_for_each_entry_rcu
outside an RCU read-side critical section but under the protection
of knav_dev_lock.
Hence, add corresponding lockdep expression to silence false-positive
lockdep warnings, and harden RCU lists.
Add macro for the corresponding lockdep expression.
Signed-off-by: Amol Grover <[email protected]>
---
v2:
- Remove rcu_read_lock_held() from lockdep expression since it is
implicitly checked for.
drivers/soc/ti/knav_qmss_queue.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/soc/ti/knav_qmss_queue.c b/drivers/soc/ti/knav_qmss_queue.c
index 1ccc9064e1eb..37f3db6c041c 100644
--- a/drivers/soc/ti/knav_qmss_queue.c
+++ b/drivers/soc/ti/knav_qmss_queue.c
@@ -25,6 +25,8 @@
static struct knav_device *kdev;
static DEFINE_MUTEX(knav_dev_lock);
+#define knav_dev_lock_held() \
+ lockdep_is_held(&knav_dev_lock)
/* Queue manager register indices in DTS */
#define KNAV_QUEUE_PEEK_REG_INDEX 0
@@ -52,8 +54,9 @@ static DEFINE_MUTEX(knav_dev_lock);
#define knav_queue_idx_to_inst(kdev, idx) \
(kdev->instances + (idx << kdev->inst_shift))
-#define for_each_handle_rcu(qh, inst) \
- list_for_each_entry_rcu(qh, &inst->handles, list)
+#define for_each_handle_rcu(qh, inst) \
+ list_for_each_entry_rcu(qh, &inst->handles, list, \
+ knav_dev_lock_held())
#define for_each_instance(idx, inst, kdev) \
for (idx = 0, inst = kdev->instances; \
--
2.24.1
On 1/17/20 5:30 AM, Amol Grover wrote:
> inst->handles is traversed using list_for_each_entry_rcu
> outside an RCU read-side critical section but under the protection
> of knav_dev_lock.
>
> Hence, add corresponding lockdep expression to silence false-positive
> lockdep warnings, and harden RCU lists.
>
> Add macro for the corresponding lockdep expression.
>
> Signed-off-by: Amol Grover <[email protected]>
> ---
> v2:
> - Remove rcu_read_lock_held() from lockdep expression since it is
> implicitly checked for.
>
Looks fine to me.
Hi Olof, Arnd,
Can you please pick this one and apply to your driver-soc branch ?
I already sent out pull request and hence the request.
regards,
Santosh
On Fri, Jan 17, 2020 at 10:24:04AM -0800, [email protected] wrote:
> On 1/17/20 5:30 AM, Amol Grover wrote:
> > inst->handles is traversed using list_for_each_entry_rcu
> > outside an RCU read-side critical section but under the protection
> > of knav_dev_lock.
> >
> > Hence, add corresponding lockdep expression to silence false-positive
> > lockdep warnings, and harden RCU lists.
> >
> > Add macro for the corresponding lockdep expression.
> >
> > Signed-off-by: Amol Grover <[email protected]>
> > ---
> > v2:
> > - Remove rcu_read_lock_held() from lockdep expression since it is
> > implicitly checked for.
> >
> Looks fine to me.
>
> Hi Olof, Arnd,
> Can you please pick this one and apply to your driver-soc branch ?
> I already sent out pull request and hence the request.
Hi,
Can you please email the whole patch with sign-off to
[email protected]? Otherwise it won't end up in patchwork, which is how we track
patches and pull requests these days.
Thanks,
-Olof
On Fri, Jan 17, 2020 at 11:25:55AM -0800, Olof Johansson wrote:
> On Fri, Jan 17, 2020 at 10:24:04AM -0800, [email protected] wrote:
> > On 1/17/20 5:30 AM, Amol Grover wrote:
> > > inst->handles is traversed using list_for_each_entry_rcu
> > > outside an RCU read-side critical section but under the protection
> > > of knav_dev_lock.
> > >
> > > Hence, add corresponding lockdep expression to silence false-positive
> > > lockdep warnings, and harden RCU lists.
> > >
> > > Add macro for the corresponding lockdep expression.
> > >
> > > Signed-off-by: Amol Grover <[email protected]>
> > > ---
> > > v2:
> > > - Remove rcu_read_lock_held() from lockdep expression since it is
> > > implicitly checked for.
> > >
> > Looks fine to me.
> >
> > Hi Olof, Arnd,
> > Can you please pick this one and apply to your driver-soc branch ?
> > I already sent out pull request and hence the request.
>
> Hi,
>
> Can you please email the whole patch with sign-off to
> [email protected]? Otherwise it won't end up in patchwork, which is how we track
> patches and pull requests these days.
>
Thank you Santosh and Olof. I'm resending the patch to [email protected]
as well.
Thanks
Amol
>
> Thanks,
>
> -Olof