From: Zhouyi Zhou <[email protected]>
Hi Paul,
During my studying of RCU, I did a grep in the kernel source tree.
I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
should be CONFIG_TASKS_TRACE_RCU instead.
Without memory fencing, the idle/userspace task inspection may not
be so accurate.
Thanks for your constant encouragement for my studying.
Best Wishes
Zhouyi
Signed-off-by: Zhouyi Zhou <[email protected]>
---
include/linux/rcupdate.h | 2 +-
kernel/rcu/tree_plugin.h | 8 ++++----
2 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index d9680b798b21..955c82b4737c 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -167,7 +167,7 @@ void synchronize_rcu_tasks(void);
# define synchronize_rcu_tasks synchronize_rcu
# endif
-# ifdef CONFIG_TASKS_RCU_TRACE
+# ifdef CONFIG_TASKS_TRACE_RCU
# define rcu_tasks_trace_qs(t) \
do { \
if (!likely(READ_ONCE((t)->trc_reader_checked)) && \
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index de1dc3bb7f70..6ce104242b23 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -2982,17 +2982,17 @@ static void noinstr rcu_dynticks_task_exit(void)
/* Turn on heavyweight RCU tasks trace readers on idle/user entry. */
static void rcu_dynticks_task_trace_enter(void)
{
-#ifdef CONFIG_TASKS_RCU_TRACE
+#ifdef CONFIG_TASKS_TRACE_RCU
if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
current->trc_reader_special.b.need_mb = true;
-#endif /* #ifdef CONFIG_TASKS_RCU_TRACE */
+#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
}
/* Turn off heavyweight RCU tasks trace readers on idle/user exit. */
static void rcu_dynticks_task_trace_exit(void)
{
-#ifdef CONFIG_TASKS_RCU_TRACE
+#ifdef CONFIG_TASKS_TRACE_RCU
if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
current->trc_reader_special.b.need_mb = false;
-#endif /* #ifdef CONFIG_TASKS_RCU_TRACE */
+#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
}
--
2.25.1
On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> From: Zhouyi Zhou <[email protected]>
>
> Hi Paul,
>
> During my studying of RCU, I did a grep in the kernel source tree.
> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> should be CONFIG_TASKS_TRACE_RCU instead.
>
> Without memory fencing, the idle/userspace task inspection may not
> be so accurate.
>
> Thanks for your constant encouragement for my studying.
>
> Best Wishes
> Zhouyi
>
> Signed-off-by: Zhouyi Zhou <[email protected]>
Good eyes, and those could cause real bugs, so thank you!
Could you please check the wordsmithed version below?
Thanx, Paul
------------------------------------------------------------------------
commit fdcf5524b64f2cc8e6201447644079d9f8d4c821
Author: Zhouyi Zhou <[email protected]>
Date: Tue Jul 13 08:56:45 2021 +0800
RCU: Fix macro name CONFIG_TASKS_RCU_TRACE
This commit fixes several typos where CONFIG_TASKS_RCU_TRACE should
instead be CONFIG_TASKS_TRACE_RCU. Among other things, these typos
could cause CONFIG_TASKS_TRACE_RCU_READ_MB=y kernels to suffer from
memory-ordering bugs that could result in false-positive quiescent
states and too-short grace periods.
Signed-off-by: Zhouyi Zhou <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index cfeb43bfc719..434d12fe2d4f 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -167,7 +167,7 @@ void synchronize_rcu_tasks(void);
# define synchronize_rcu_tasks synchronize_rcu
# endif
-# ifdef CONFIG_TASKS_RCU_TRACE
+# ifdef CONFIG_TASKS_TRACE_RCU
# define rcu_tasks_trace_qs(t) \
do { \
if (!likely(READ_ONCE((t)->trc_reader_checked)) && \
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 27b74352cccf..a8e3acead6f6 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -1498,17 +1498,17 @@ static void noinstr rcu_dynticks_task_exit(void)
/* Turn on heavyweight RCU tasks trace readers on idle/user entry. */
static void rcu_dynticks_task_trace_enter(void)
{
-#ifdef CONFIG_TASKS_RCU_TRACE
+#ifdef CONFIG_TASKS_TRACE_RCU
if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
current->trc_reader_special.b.need_mb = true;
-#endif /* #ifdef CONFIG_TASKS_RCU_TRACE */
+#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
}
/* Turn off heavyweight RCU tasks trace readers on idle/user exit. */
static void rcu_dynticks_task_trace_exit(void)
{
-#ifdef CONFIG_TASKS_RCU_TRACE
+#ifdef CONFIG_TASKS_TRACE_RCU
if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
current->trc_reader_special.b.need_mb = false;
-#endif /* #ifdef CONFIG_TASKS_RCU_TRACE */
+#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
}
Thank you very much,
I went through the wordsmithed version, it looks exquisite!
Thank you for your encouragement.
Cheers
Zhouyi
On Tue, Jul 13, 2021 at 12:16 PM Paul E. McKenney <[email protected]> wrote:
>
> On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > From: Zhouyi Zhou <[email protected]>
> >
> > Hi Paul,
> >
> > During my studying of RCU, I did a grep in the kernel source tree.
> > I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > should be CONFIG_TASKS_TRACE_RCU instead.
> >
> > Without memory fencing, the idle/userspace task inspection may not
> > be so accurate.
> >
> > Thanks for your constant encouragement for my studying.
> >
> > Best Wishes
> > Zhouyi
> >
> > Signed-off-by: Zhouyi Zhou <[email protected]>
>
> Good eyes, and those could cause real bugs, so thank you!
>
> Could you please check the wordsmithed version below?
>
> Thanx, Paul
>
> ------------------------------------------------------------------------
>
> commit fdcf5524b64f2cc8e6201447644079d9f8d4c821
> Author: Zhouyi Zhou <[email protected]>
> Date: Tue Jul 13 08:56:45 2021 +0800
>
> RCU: Fix macro name CONFIG_TASKS_RCU_TRACE
>
> This commit fixes several typos where CONFIG_TASKS_RCU_TRACE should
> instead be CONFIG_TASKS_TRACE_RCU. Among other things, these typos
> could cause CONFIG_TASKS_TRACE_RCU_READ_MB=y kernels to suffer from
> memory-ordering bugs that could result in false-positive quiescent
> states and too-short grace periods.
>
> Signed-off-by: Zhouyi Zhou <[email protected]>
> Signed-off-by: Paul E. McKenney <[email protected]>
>
> diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> index cfeb43bfc719..434d12fe2d4f 100644
> --- a/include/linux/rcupdate.h
> +++ b/include/linux/rcupdate.h
> @@ -167,7 +167,7 @@ void synchronize_rcu_tasks(void);
> # define synchronize_rcu_tasks synchronize_rcu
> # endif
>
> -# ifdef CONFIG_TASKS_RCU_TRACE
> +# ifdef CONFIG_TASKS_TRACE_RCU
> # define rcu_tasks_trace_qs(t) \
> do { \
> if (!likely(READ_ONCE((t)->trc_reader_checked)) && \
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index 27b74352cccf..a8e3acead6f6 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -1498,17 +1498,17 @@ static void noinstr rcu_dynticks_task_exit(void)
> /* Turn on heavyweight RCU tasks trace readers on idle/user entry. */
> static void rcu_dynticks_task_trace_enter(void)
> {
> -#ifdef CONFIG_TASKS_RCU_TRACE
> +#ifdef CONFIG_TASKS_TRACE_RCU
> if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
> current->trc_reader_special.b.need_mb = true;
> -#endif /* #ifdef CONFIG_TASKS_RCU_TRACE */
> +#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
> }
>
> /* Turn off heavyweight RCU tasks trace readers on idle/user exit. */
> static void rcu_dynticks_task_trace_exit(void)
> {
> -#ifdef CONFIG_TASKS_RCU_TRACE
> +#ifdef CONFIG_TASKS_TRACE_RCU
> if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB))
> current->trc_reader_special.b.need_mb = false;
> -#endif /* #ifdef CONFIG_TASKS_RCU_TRACE */
> +#endif /* #ifdef CONFIG_TASKS_TRACE_RCU */
> }
----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
>> From: Zhouyi Zhou <[email protected]>
>>
>> Hi Paul,
>>
>> During my studying of RCU, I did a grep in the kernel source tree.
>> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
>> should be CONFIG_TASKS_TRACE_RCU instead.
>>
>> Without memory fencing, the idle/userspace task inspection may not
>> be so accurate.
>>
>> Thanks for your constant encouragement for my studying.
>>
>> Best Wishes
>> Zhouyi
>>
>> Signed-off-by: Zhouyi Zhou <[email protected]>
>
> Good eyes, and those could cause real bugs, so thank you!
Hi Paul,
This makes me wonder: what is missing testing-wise in rcutorture to
catch those issues with testing before they reach mainline ?
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
>
> > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> >> From: Zhouyi Zhou <[email protected]>
> >>
> >> Hi Paul,
> >>
> >> During my studying of RCU, I did a grep in the kernel source tree.
> >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> >> should be CONFIG_TASKS_TRACE_RCU instead.
> >>
> >> Without memory fencing, the idle/userspace task inspection may not
> >> be so accurate.
> >>
> >> Thanks for your constant encouragement for my studying.
> >>
> >> Best Wishes
> >> Zhouyi
> >>
> >> Signed-off-by: Zhouyi Zhou <[email protected]>
> >
> > Good eyes, and those could cause real bugs, so thank you!
>
> Hi Paul,
>
> This makes me wonder: what is missing testing-wise in rcutorture to
> catch those issues with testing before they reach mainline ?
My guess: Running on weakly ordered architectures. ;-)
Thanx, Paul
On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> >
> > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > >> From: Zhouyi Zhou <[email protected]>
> > >>
> > >> Hi Paul,
> > >>
> > >> During my studying of RCU, I did a grep in the kernel source tree.
> > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > >>
> > >> Without memory fencing, the idle/userspace task inspection may not
> > >> be so accurate.
> > >>
> > >> Thanks for your constant encouragement for my studying.
> > >>
> > >> Best Wishes
> > >> Zhouyi
> > >>
> > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > >
> > > Good eyes, and those could cause real bugs, so thank you!
> >
> > Hi Paul,
> >
> > This makes me wonder: what is missing testing-wise in rcutorture to
> > catch those issues with testing before they reach mainline ?
>
> My guess: Running on weakly ordered architectures. ;-)
And another guess: A tool that identifies use of Kconfig options
that are not defined in any Kconfig* file.
Thanx, Paul
On Tue, Jul 13, 2021 at 11:19 PM Paul E. McKenney <[email protected]> wrote:
>
> On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> > On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> > >
> > > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > > >> From: Zhouyi Zhou <[email protected]>
> > > >>
> > > >> Hi Paul,
> > > >>
> > > >> During my studying of RCU, I did a grep in the kernel source tree.
> > > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > > >>
> > > >> Without memory fencing, the idle/userspace task inspection may not
> > > >> be so accurate.
> > > >>
> > > >> Thanks for your constant encouragement for my studying.
> > > >>
> > > >> Best Wishes
> > > >> Zhouyi
> > > >>
> > > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > > >
> > > > Good eyes, and those could cause real bugs, so thank you!
> > >
> > > Hi Paul,
> > >
> > > This makes me wonder: what is missing testing-wise in rcutorture to
> > > catch those issues with testing before they reach mainline ?
> >
> > My guess: Running on weakly ordered architectures. ;-)
>
> And another guess: A tool that identifies use of Kconfig options
> that are not defined in any Kconfig* file.
Based on Paul's second guess ;-), I did a small research, and I think
the best answer is to modify scripts/checkpatch.pl. We modify checkpatch.pl
to identify use of Kconfig options that are not defined in any Kconfig* file.
As I am a C/C++ programmer, I would be glad to take some time to learn
perl (checkpatch is implented in perl) first if no other volunteer is
about to do it ;-)
>
> Thanx, Paul
Thanx
Zhouyi
On Wed, Jul 14, 2021 at 12:44:36PM +0800, Zhouyi Zhou wrote:
> On Tue, Jul 13, 2021 at 11:19 PM Paul E. McKenney <[email protected]> wrote:
> >
> > On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> > > On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > > > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> > > >
> > > > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > > > >> From: Zhouyi Zhou <[email protected]>
> > > > >>
> > > > >> Hi Paul,
> > > > >>
> > > > >> During my studying of RCU, I did a grep in the kernel source tree.
> > > > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > > > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > > > >>
> > > > >> Without memory fencing, the idle/userspace task inspection may not
> > > > >> be so accurate.
> > > > >>
> > > > >> Thanks for your constant encouragement for my studying.
> > > > >>
> > > > >> Best Wishes
> > > > >> Zhouyi
> > > > >>
> > > > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > > > >
> > > > > Good eyes, and those could cause real bugs, so thank you!
> > > >
> > > > Hi Paul,
> > > >
> > > > This makes me wonder: what is missing testing-wise in rcutorture to
> > > > catch those issues with testing before they reach mainline ?
> > >
> > > My guess: Running on weakly ordered architectures. ;-)
> >
> > And another guess: A tool that identifies use of Kconfig options
> > that are not defined in any Kconfig* file.
> Based on Paul's second guess ;-), I did a small research, and I think
> the best answer is to modify scripts/checkpatch.pl. We modify checkpatch.pl
> to identify use of Kconfig options that are not defined in any Kconfig* file.
>
> As I am a C/C++ programmer, I would be glad to take some time to learn
> perl (checkpatch is implented in perl) first if no other volunteer is
> about to do it ;-)
I haven't heard anyone else volunteer. ;-)
Others might have opinions on where best to implement these checks,
but I must confess that I have not given it much thought.
Thanx, Paul
On Thu, Jul 15, 2021 at 11:51 AM Paul E. McKenney <[email protected]> wrote:
>
> On Wed, Jul 14, 2021 at 12:44:36PM +0800, Zhouyi Zhou wrote:
> > On Tue, Jul 13, 2021 at 11:19 PM Paul E. McKenney <[email protected]> wrote:
> > >
> > > On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> > > > On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > > > > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> > > > >
> > > > > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > > > > >> From: Zhouyi Zhou <[email protected]>
> > > > > >>
> > > > > >> Hi Paul,
> > > > > >>
> > > > > >> During my studying of RCU, I did a grep in the kernel source tree.
> > > > > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > > > > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > > > > >>
> > > > > >> Without memory fencing, the idle/userspace task inspection may not
> > > > > >> be so accurate.
> > > > > >>
> > > > > >> Thanks for your constant encouragement for my studying.
> > > > > >>
> > > > > >> Best Wishes
> > > > > >> Zhouyi
> > > > > >>
> > > > > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > > > > >
> > > > > > Good eyes, and those could cause real bugs, so thank you!
> > > > >
> > > > > Hi Paul,
> > > > >
> > > > > This makes me wonder: what is missing testing-wise in rcutorture to
> > > > > catch those issues with testing before they reach mainline ?
> > > >
> > > > My guess: Running on weakly ordered architectures. ;-)
> > >
> > > And another guess: A tool that identifies use of Kconfig options
> > > that are not defined in any Kconfig* file.
> > Based on Paul's second guess ;-), I did a small research, and I think
> > the best answer is to modify scripts/checkpatch.pl. We modify checkpatch.pl
> > to identify use of Kconfig options that are not defined in any Kconfig* file.
> >
> > As I am a C/C++ programmer, I would be glad to take some time to learn
> > perl (checkpatch is implented in perl) first if no other volunteer is
> > about to do it ;-)
>
> I haven't heard anyone else volunteer. ;-)
>
> Others might have opinions on where best to implement these checks,
> but I must confess that I have not given it much thought.
I recklessly cc the maintainers of checkpatch.pl without your
permission to see others' opion,
and I begin to study perl at the same time, after all, learning
something is always good ;-)
>
> Thanx, Paul
Best Wishes
Zhouyi
On Thu, Jul 15, 2021 at 04:45:04PM +0800, Zhouyi Zhou wrote:
> On Thu, Jul 15, 2021 at 11:51 AM Paul E. McKenney <[email protected]> wrote:
> >
> > On Wed, Jul 14, 2021 at 12:44:36PM +0800, Zhouyi Zhou wrote:
> > > On Tue, Jul 13, 2021 at 11:19 PM Paul E. McKenney <[email protected]> wrote:
> > > >
> > > > On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> > > > > On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > > > > > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> > > > > >
> > > > > > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > > > > > >> From: Zhouyi Zhou <[email protected]>
> > > > > > >>
> > > > > > >> Hi Paul,
> > > > > > >>
> > > > > > >> During my studying of RCU, I did a grep in the kernel source tree.
> > > > > > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > > > > > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > > > > > >>
> > > > > > >> Without memory fencing, the idle/userspace task inspection may not
> > > > > > >> be so accurate.
> > > > > > >>
> > > > > > >> Thanks for your constant encouragement for my studying.
> > > > > > >>
> > > > > > >> Best Wishes
> > > > > > >> Zhouyi
> > > > > > >>
> > > > > > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > > > > > >
> > > > > > > Good eyes, and those could cause real bugs, so thank you!
> > > > > >
> > > > > > Hi Paul,
> > > > > >
> > > > > > This makes me wonder: what is missing testing-wise in rcutorture to
> > > > > > catch those issues with testing before they reach mainline ?
> > > > >
> > > > > My guess: Running on weakly ordered architectures. ;-)
> > > >
> > > > And another guess: A tool that identifies use of Kconfig options
> > > > that are not defined in any Kconfig* file.
> > > Based on Paul's second guess ;-), I did a small research, and I think
> > > the best answer is to modify scripts/checkpatch.pl. We modify checkpatch.pl
> > > to identify use of Kconfig options that are not defined in any Kconfig* file.
> > >
> > > As I am a C/C++ programmer, I would be glad to take some time to learn
> > > perl (checkpatch is implented in perl) first if no other volunteer is
> > > about to do it ;-)
> >
> > I haven't heard anyone else volunteer. ;-)
> >
> > Others might have opinions on where best to implement these checks,
> > but I must confess that I have not given it much thought.
> I recklessly cc the maintainers of checkpatch.pl without your
> permission to see others' opion,
> and I begin to study perl at the same time, after all, learning
> something is always good ;-)
Works for me!
Thanx, Paul
Hi Paul
During the research, I found a already existing tool to detect
undefined Kconfig macro:
scripts/checkkconfigsymbols.py. It is marvellous!
By invoking ./scripts/checkkconfigsymbols.py > /tmp/log, I found
following possibly undefined Kconfig macros
which may need our attention:
PREEMPT_LOCK
Referencing files: include/linux/lockdep_types.h
PREEMT_DYNAMIC
Referencing files: kernel/entry/common.c
TREE_PREEMPT_RCU
Referencing files: arch/sh/configs/sdk7786_defconfig
RCU_CPU_STALL_INFO
Referencing files: arch/xtensa/configs/nommu_kc705_defconfig
RCU_NOCB_CPU_ALL
Referencing files:
Documentation/RCU/Design/Memory-Ordering/Tree-RCU-Memory-Ordering.rst
RCU_TORTURE_TESTS
Referencing files: kernel/rcu/rcutorture.c
and finally the macro which drive me to do this research
TASKS_RCU_TRACE
Referencing files: include/linux/rcupdate.h, kernel/rcu/tree_plugin.h
On Fri, Jul 16, 2021 at 2:09 AM Paul E. McKenney <[email protected]> wrote:
>
> On Thu, Jul 15, 2021 at 04:45:04PM +0800, Zhouyi Zhou wrote:
> > On Thu, Jul 15, 2021 at 11:51 AM Paul E. McKenney <[email protected]> wrote:
> > >
> > > On Wed, Jul 14, 2021 at 12:44:36PM +0800, Zhouyi Zhou wrote:
> > > > On Tue, Jul 13, 2021 at 11:19 PM Paul E. McKenney <[email protected]> wrote:
> > > > >
> > > > > On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> > > > > > On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > > > > > > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> > > > > > >
> > > > > > > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > > > > > > >> From: Zhouyi Zhou <[email protected]>
> > > > > > > >>
> > > > > > > >> Hi Paul,
> > > > > > > >>
> > > > > > > >> During my studying of RCU, I did a grep in the kernel source tree.
> > > > > > > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > > > > > > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > > > > > > >>
> > > > > > > >> Without memory fencing, the idle/userspace task inspection may not
> > > > > > > >> be so accurate.
> > > > > > > >>
> > > > > > > >> Thanks for your constant encouragement for my studying.
> > > > > > > >>
> > > > > > > >> Best Wishes
> > > > > > > >> Zhouyi
> > > > > > > >>
> > > > > > > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > > > > > > >
> > > > > > > > Good eyes, and those could cause real bugs, so thank you!
> > > > > > >
> > > > > > > Hi Paul,
> > > > > > >
> > > > > > > This makes me wonder: what is missing testing-wise in rcutorture to
> > > > > > > catch those issues with testing before they reach mainline ?
> > > > > >
> > > > > > My guess: Running on weakly ordered architectures. ;-)
> > > > >
> > > > > And another guess: A tool that identifies use of Kconfig options
> > > > > that are not defined in any Kconfig* file.
> > > > Based on Paul's second guess ;-), I did a small research, and I think
> > > > the best answer is to modify scripts/checkpatch.pl. We modify checkpatch.pl
> > > > to identify use of Kconfig options that are not defined in any Kconfig* file.
> > > >
> > > > As I am a C/C++ programmer, I would be glad to take some time to learn
> > > > perl (checkpatch is implented in perl) first if no other volunteer is
> > > > about to do it ;-)
> > >
> > > I haven't heard anyone else volunteer. ;-)
> > >
> > > Others might have opinions on where best to implement these checks,
> > > but I must confess that I have not given it much thought.
> > I recklessly cc the maintainers of checkpatch.pl without your
> > permission to see others' opion,
> > and I begin to study perl at the same time, after all, learning
> > something is always good ;-)
>
> Works for me!
>
> Thanx, Paul
Best Wishes
Zhouyi
On Sun, Jul 18, 2021 at 06:03:34AM +0800, Zhouyi Zhou wrote:
> Hi Paul
> During the research, I found a already existing tool to detect
> undefined Kconfig macro:
> scripts/checkkconfigsymbols.py. It is marvellous!
Nice! Maybe I should add this to torture.sh.
> By invoking ./scripts/checkkconfigsymbols.py > /tmp/log, I found
> following possibly undefined Kconfig macros
> which may need our attention:
>
> PREEMPT_LOCK
> Referencing files: include/linux/lockdep_types.h
Not sure about this one. It might be in anticipation of -rt functionality.
Or another typo.
> PREEMT_DYNAMIC
> Referencing files: kernel/entry/common.c
This needs to be PREEMPT_DYNAMIC. Please CC Frederic Weisbecker and
myself if you send a patch.
> TREE_PREEMPT_RCU
> Referencing files: arch/sh/configs/sdk7786_defconfig
This would have been correct back in the day. It should now be
CONFIG_PREEMPT_RCU. Except that the CONFIG_PREEMPT=y in that same
file implies CONFIG_PREEMPT_RCU=y, so best to simply delete the
CONFIG_TREE_PREEMPT_RCU=y line.
> RCU_CPU_STALL_INFO
> Referencing files: arch/xtensa/configs/nommu_kc705_defconfig
You now get RCU_CPU_STALL_INFO whether you want it or not, so this
line should be deleted.
> RCU_NOCB_CPU_ALL
> Referencing files:
> Documentation/RCU/Design/Memory-Ordering/Tree-RCU-Memory-Ordering.rst
This is an old snapshot of the code. One approach would be to
update this from the real rcu_prepare_for_idle() function in
kernel/rcu/tree_plugin.h. The line numbers in the following paragraph
would need to be updated, but the figure is unaffected.
> RCU_TORTURE_TESTS
> Referencing files: kernel/rcu/rcutorture.c
The final "S" needs to be dropped.
> and finally the macro which drive me to do this research
>
> TASKS_RCU_TRACE
> Referencing files: include/linux/rcupdate.h, kernel/rcu/tree_plugin.h
The fix for this one is of course already queued.
Please CC me if you decide to create patches. Otherwise, let me know,
and I can produce fixes.
Thanx, Paul
> On Fri, Jul 16, 2021 at 2:09 AM Paul E. McKenney <[email protected]> wrote:
> >
> > On Thu, Jul 15, 2021 at 04:45:04PM +0800, Zhouyi Zhou wrote:
> > > On Thu, Jul 15, 2021 at 11:51 AM Paul E. McKenney <[email protected]> wrote:
> > > >
> > > > On Wed, Jul 14, 2021 at 12:44:36PM +0800, Zhouyi Zhou wrote:
> > > > > On Tue, Jul 13, 2021 at 11:19 PM Paul E. McKenney <[email protected]> wrote:
> > > > > >
> > > > > > On Tue, Jul 13, 2021 at 06:18:12AM -0700, Paul E. McKenney wrote:
> > > > > > > On Tue, Jul 13, 2021 at 09:09:04AM -0400, Mathieu Desnoyers wrote:
> > > > > > > > ----- On Jul 13, 2021, at 12:16 AM, paulmck [email protected] wrote:
> > > > > > > >
> > > > > > > > > On Tue, Jul 13, 2021 at 08:56:45AM +0800, [email protected] wrote:
> > > > > > > > >> From: Zhouyi Zhou <[email protected]>
> > > > > > > > >>
> > > > > > > > >> Hi Paul,
> > > > > > > > >>
> > > > > > > > >> During my studying of RCU, I did a grep in the kernel source tree.
> > > > > > > > >> I found there are 3 places where the macro name CONFIG_TASKS_RCU_TRACE
> > > > > > > > >> should be CONFIG_TASKS_TRACE_RCU instead.
> > > > > > > > >>
> > > > > > > > >> Without memory fencing, the idle/userspace task inspection may not
> > > > > > > > >> be so accurate.
> > > > > > > > >>
> > > > > > > > >> Thanks for your constant encouragement for my studying.
> > > > > > > > >>
> > > > > > > > >> Best Wishes
> > > > > > > > >> Zhouyi
> > > > > > > > >>
> > > > > > > > >> Signed-off-by: Zhouyi Zhou <[email protected]>
> > > > > > > > >
> > > > > > > > > Good eyes, and those could cause real bugs, so thank you!
> > > > > > > >
> > > > > > > > Hi Paul,
> > > > > > > >
> > > > > > > > This makes me wonder: what is missing testing-wise in rcutorture to
> > > > > > > > catch those issues with testing before they reach mainline ?
> > > > > > >
> > > > > > > My guess: Running on weakly ordered architectures. ;-)
> > > > > >
> > > > > > And another guess: A tool that identifies use of Kconfig options
> > > > > > that are not defined in any Kconfig* file.
> > > > > Based on Paul's second guess ;-), I did a small research, and I think
> > > > > the best answer is to modify scripts/checkpatch.pl. We modify checkpatch.pl
> > > > > to identify use of Kconfig options that are not defined in any Kconfig* file.
> > > > >
> > > > > As I am a C/C++ programmer, I would be glad to take some time to learn
> > > > > perl (checkpatch is implented in perl) first if no other volunteer is
> > > > > about to do it ;-)
> > > >
> > > > I haven't heard anyone else volunteer. ;-)
> > > >
> > > > Others might have opinions on where best to implement these checks,
> > > > but I must confess that I have not given it much thought.
> > > I recklessly cc the maintainers of checkpatch.pl without your
> > > permission to see others' opion,
> > > and I begin to study perl at the same time, after all, learning
> > > something is always good ;-)
> >
> > Works for me!
> >
> > Thanx, Paul
> Best Wishes
> Zhouyi
On 7/18/21 2:08 PM, Paul E. McKenney wrote:
> On Sun, Jul 18, 2021 at 06:03:34AM +0800, Zhouyi Zhou wrote:
>> Hi Paul
>> During the research, I found a already existing tool to detect
>> undefined Kconfig macro:
>> scripts/checkkconfigsymbols.py. It is marvellous!
>
> Nice! Maybe I should add this to torture.sh.
>
Paul, I believe that subsystems should take care of themselves,
so you can do that for RCU, e.g., but at the same time, I think that
some CI should be running that script (and other relevant scripts)
on the entire kernel tree and reporting problems that are found.
--
~Randy
On Mon, Jul 19, 2021 at 06:39:17PM -0700, Randy Dunlap wrote:
> On 7/18/21 2:08 PM, Paul E. McKenney wrote:
> > On Sun, Jul 18, 2021 at 06:03:34AM +0800, Zhouyi Zhou wrote:
> >> Hi Paul
> >> During the research, I found a already existing tool to detect
> >> undefined Kconfig macro:
> >> scripts/checkkconfigsymbols.py. It is marvellous!
> >
> > Nice! Maybe I should add this to torture.sh.
>
> Paul, I believe that subsystems should take care of themselves,
> so you can do that for RCU, e.g., but at the same time, I think that
> some CI should be running that script (and other relevant scripts)
> on the entire kernel tree and reporting problems that are found.
Even better! ;-)
Thanx, Paul