[PATCH v7 0/1] cu/nocb: Delete local variable 'need_rcu_nocb_mask' in rcu_init_nohz()

[PATCH v7 0/1] cu/nocb: Delete local variable 'need_rcu_nocb_mask' in rcu_init_nohz()

[Zhen Lei]

v6 --> v7:
1. Replace "#if defined(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL)" with
   "IS_ENABLED(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL)".
2. Drop patch 2, Keep the semantics of "CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y"
   unchanged. So the test result of case 7) is different in v6 and v7.

Test cases and test results as below:
 ----------------------------------------------------------------------
| CONFIG_NO_HZ_FULL  | CONFIG_RCU_NOCB_CPU_DEFAULT_ALL |    cmdline    |
 ----------------------------------------------------------------------
1)         N         |                 N               |     none      |
2)         N         |                 N               | rcu_nocbs=1-2 |
3)         Y         |                 N               | nohz_full=2-3 |
4)         Y         |                 N               | rcu_nocbs=1 nohz_full=3 |
5)         Y         |                 Y               |      none     |
6)         Y         |                 Y               | rcu_nocbs=1-2 |
7)         Y         |                 Y               | nohz_full=2-3 |
8)         Y         |                 Y               | rcu_nocbs=1 nohz_full=3 |
 ----------------------------------------------------------------------
 
dmesg | grep "Offload RCU callbacks"
1) none
2) Offload RCU callbacks from CPUs: 1-2.
3) Offload RCU callbacks from CPUs: 2-3.
4) Offload RCU callbacks from CPUs: 1,3.
5) Offload RCU callbacks from CPUs: 0-3.
6) Offload RCU callbacks from CPUs: 1-2.
7) Offload RCU callbacks from CPUs: 2-3.
8) Offload RCU callbacks from CPUs: 1,3.


v5 --> v6:
1. Frist, update patch 1/2, so that, CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y takes
   effect only when there are no 'rcu_nocbs=' and 'nohz_full='. Retain the same
   semantics as the original.
2. Then, add patch 2/2, so that, CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y takes effect
   as long as there is no "rcu_nocbs=" boot parameter. This is suggested by
   Frederic Weisbecker.

Test cases and test results as below:
 ----------------------------------------------------------------------
| CONFIG_NO_HZ_FULL  | CONFIG_RCU_NOCB_CPU_DEFAULT_ALL |    cmdline    |
 ----------------------------------------------------------------------
1)         N         |                 N               |     none      |
2)         N         |                 N               | rcu_nocbs=1-2 |
3)         Y         |                 N               | nohz_full=2-3 |
4)         Y         |                 N               | rcu_nocbs=1 nohz_full=3 |
5)         Y         |                 Y               |      none     |
6)         Y         |                 Y               | rcu_nocbs=1-2 |
7)         Y         |                 Y               | nohz_full=2-3 |
8)         Y         |                 Y               | rcu_nocbs=1 nohz_full=3 |
 ----------------------------------------------------------------------

dmesg | grep "Offload RCU callbacks"
1) none
2) Offload RCU callbacks from CPUs: 1-2.
3) Offload RCU callbacks from CPUs: 2-3.
4) Offload RCU callbacks from CPUs: 1,3.
5) Offload RCU callbacks from CPUs: 0-3.
6) Offload RCU callbacks from CPUs: 1-2.
7) Offload RCU callbacks from CPUs: 0-3.
8) Offload RCU callbacks from CPUs: 1,3.


v4 --> v5:
1. Cancel eliminate rcu_state.nocb_is_setup. It is useful for case 1) below.

Test cases and test results as below:
 ----------------------------------------------------------------------
| CONFIG_NO_HZ_FULL  | CONFIG_RCU_NOCB_CPU_DEFAULT_ALL |    cmdline    |
 ----------------------------------------------------------------------
1)         N         |                 N               |     none      |
2)         N         |                 N               | rcu_nocbs=1-2 |
3)         Y         |                 N               | nohz_full=1-2 |
4)         Y         |                 Y               |     none      |
 ----------------------------------------------------------------------
 
1)
[ 1018.082417] rcu-torture:--- End of test: SUCCESS (default)
root@genericarmv8:~# dmesg | grep "Offload RCU callbacks"

root@genericarmv8:~# zcat /proc/config.gz | grep NOCB
CONFIG_RCU_NOCB_CPU=y
# CONFIG_RCU_NOCB_CPU_DEFAULT_ALL is not set
root@genericarmv8:~# zcat /proc/config.gz | grep NO_HZ_FULL
# CONFIG_NO_HZ_FULL is not set
root@genericarmv8:~# cat /proc/cmdline
console=ttyS0 earlyprintk=serial root=/dev/sda rw

2) 3) 4)
The same to v4.


v3 --> v4:
Fix a build warning.
-       struct cpumask *cpumask = NULL;
+       const struct cpumask *cpumask = NULL;

v2 --> v3:
1. Eliminate rcu_state.nocb_is_setup
2. Update the code based on the above patch and commit b37a667c6242
   ("rcu/nocb: Add an option to offload all CPUs on boot").

Test cases and test results as below:

 ----------------------------------------------------------------------
| CONFIG_NO_HZ_FULL  | CONFIG_RCU_NOCB_CPU_DEFAULT_ALL |    cmdline    |
 ----------------------------------------------------------------------
1)         N         |                 N               |     none      |
2)         N         |                 N               | rcu_nocbs=1-2 |
3)         Y         |                 N               | nohz_full=1-2 |
4)         Y         |                 Y               |     none      |
 ----------------------------------------------------------------------

1)
[ 1085.691770] rcu-torture:--- End of test: SUCCESS: (default)
root@genericarmv8:~# dmesg | grep "Offload RCU callbacks"
[    0.000000] rcu:     Offload RCU callbacks from CPUs: (none).
root@genericarmv8:~# zcat /proc/config.gz | grep NOCB
CONFIG_RCU_NOCB_CPU=y
# CONFIG_RCU_NOCB_CPU_DEFAULT_ALL is not set
root@genericarmv8:~# zcat /proc/config.gz | grep NO_HZ_FULL
# CONFIG_NO_HZ_FULL is not set
root@genericarmv8:~# cat /proc/cmdline
console=ttyAMA0 root=/dev/vda2 crashkernel=256M@ee000000

2)
[ 1022.532198] rcu-torture:--- End of test: SUCCESS: (default)
root@genericarmv8:~# dmesg | grep "Offload RCU callbacks"
[    0.000000] rcu:     Offload RCU callbacks from CPUs: 1-2.
root@genericarmv8:~# zcat /proc/config.gz | grep NOCB
CONFIG_RCU_NOCB_CPU=y
# CONFIG_RCU_NOCB_CPU_DEFAULT_ALL is not set
root@genericarmv8:~# zcat /proc/config.gz | grep NO_HZ_FULL
# CONFIG_NO_HZ_FULL is not set
root@genericarmv8:~# cat /proc/cmdline
console=ttyAMA0 root=/dev/vda2 crashkernel=256M@ee000000 rcu_nocbs=1-2

3)
[ 1011.714712] rcu-torture:--- End of test: SUCCESS: (default)
root@genericarmv8:~# dmesg | grep "Offload RCU callbacks"
[    0.000000] rcu:     Offload RCU callbacks from CPUs: 1-2.
root@genericarmv8:~# zcat /proc/config.gz | grep NOCB
CONFIG_RCU_NOCB_CPU=y
# CONFIG_RCU_NOCB_CPU_DEFAULT_ALL is not set
root@genericarmv8:~# zcat /proc/config.gz | grep NO_HZ_FULL
CONFIG_NO_HZ_FULL=y
root@genericarmv8:~# cat /proc/cmdline
console=ttyAMA0 root=/dev/vda2 crashkernel=256M@ee000000 nohz_full=1-2

4)
[ 1031.047094] rcu-torture:--- End of test: SUCCESS: (default)
root@genericarmv8:~# dmesg | grep "Offload RCU callbacks"
[    0.000000] rcu:     Offload RCU callbacks from CPUs: 0-3.
root@genericarmv8:~# zcat /proc/config.gz | grep NOCB
CONFIG_RCU_NOCB_CPU=y
CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y
root@genericarmv8:~# zcat /proc/config.gz | grep NO_HZ_FULL
CONFIG_NO_HZ_FULL=y
root@genericarmv8:~# cat /proc/cmdline
console=ttyAMA0 root=/dev/vda2 crashkernel=256M@ee000000


v1 --> v2:
Update commit message.

Zhen Lei (1):
  rcu: Simplify rcu_init_nohz() cpumask handling

 kernel/rcu/tree_nocb.h | 34 +++++++++++-----------------------
 1 file changed, 11 insertions(+), 23 deletions(-)

-- 
2.25.1

[PATCH v7 1/1] rcu: Simplify rcu_init_nohz() cpumask handling

[Zhen Lei]

In kernels built with either CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y or
CONFIG_NO_HZ_FULL=y, additional CPUs must be added to rcu_nocb_mask.
Except that kernels booted without the rcu_nocbs= will not have
allocated rcu_nocb_mask.  And the current rcu_init_nohz() function uses
its need_rcu_nocb_mask and offload_all local variables to track the
rcu_nocb and nohz_full state.

But there is a much simpler approach, namely creating a cpumask pointer
to track the default and then using cpumask_available() to check the
rcu_nocb_mask state.  This commit takes this approach, thereby simplifying
and shortening the rcu_init_nohz() function.

Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Acked-by: Frederic Weisbecker <frederic@kernel.org>
---
 kernel/rcu/tree_nocb.h | 34 +++++++++++-----------------------
 1 file changed, 11 insertions(+), 23 deletions(-)

diff --git a/kernel/rcu/tree_nocb.h b/kernel/rcu/tree_nocb.h
index 0a5f0ef41484518..ce526cc2791ca6c 100644
--- a/kernel/rcu/tree_nocb.h
+++ b/kernel/rcu/tree_nocb.h
@@ -1210,45 +1210,33 @@ EXPORT_SYMBOL_GPL(rcu_nocb_cpu_offload);
 void __init rcu_init_nohz(void)
 {
 	int cpu;
-	bool need_rcu_nocb_mask = false;
-	bool offload_all = false;
 	struct rcu_data *rdp;
-
-#if defined(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL)
-	if (!rcu_state.nocb_is_setup) {
-		need_rcu_nocb_mask = true;
-		offload_all = true;
-	}
-#endif /* #if defined(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL) */
+	const struct cpumask *cpumask = NULL;
 
 #if defined(CONFIG_NO_HZ_FULL)
-	if (tick_nohz_full_running && !cpumask_empty(tick_nohz_full_mask)) {
-		need_rcu_nocb_mask = true;
-		offload_all = false; /* NO_HZ_FULL has its own mask. */
-	}
-#endif /* #if defined(CONFIG_NO_HZ_FULL) */
+	if (tick_nohz_full_running && !cpumask_empty(tick_nohz_full_mask))
+		cpumask = tick_nohz_full_mask;
+#endif
+
+	if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL) &&
+	    !rcu_state.nocb_is_setup && !cpumask)
+		cpumask = cpu_possible_mask;
 
-	if (need_rcu_nocb_mask) {
+	if (cpumask) {
 		if (!cpumask_available(rcu_nocb_mask)) {
 			if (!zalloc_cpumask_var(&rcu_nocb_mask, GFP_KERNEL)) {
 				pr_info("rcu_nocb_mask allocation failed, callback offloading disabled.\n");
 				return;
 			}
 		}
+
+		cpumask_or(rcu_nocb_mask, rcu_nocb_mask, cpumask);
 		rcu_state.nocb_is_setup = true;
 	}
 
 	if (!rcu_state.nocb_is_setup)
 		return;
 
-#if defined(CONFIG_NO_HZ_FULL)
-	if (tick_nohz_full_running)
-		cpumask_or(rcu_nocb_mask, rcu_nocb_mask, tick_nohz_full_mask);
-#endif /* #if defined(CONFIG_NO_HZ_FULL) */
-
-	if (offload_all)
-		cpumask_setall(rcu_nocb_mask);
-
 	if (!cpumask_subset(rcu_nocb_mask, cpu_possible_mask)) {
 		pr_info("\tNote: kernel parameter 'rcu_nocbs=', 'nohz_full', or 'isolcpus=' contains nonexistent CPUs.\n");
 		cpumask_and(rcu_nocb_mask, cpu_possible_mask,
-- 
2.25.1

Re: [PATCH v7 1/1] rcu: Simplify rcu_init_nohz() cpumask handling

[Frederic Weisbecker]

On Tue, Sep 13, 2022 at 11:00:36AM +0800, Zhen Lei wrote:
> In kernels built with either CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y or
> CONFIG_NO_HZ_FULL=y, additional CPUs must be added to rcu_nocb_mask.
> Except that kernels booted without the rcu_nocbs= will not have
> allocated rcu_nocb_mask.  And the current rcu_init_nohz() function uses
> its need_rcu_nocb_mask and offload_all local variables to track the
> rcu_nocb and nohz_full state.
> 
> But there is a much simpler approach, namely creating a cpumask pointer
> to track the default and then using cpumask_available() to check the
> rcu_nocb_mask state.  This commit takes this approach, thereby simplifying
> and shortening the rcu_init_nohz() function.
> 
> Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
> Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> Acked-by: Frederic Weisbecker <frederic@kernel.org>

Looks good, thanks!

Re: [PATCH v7 1/1] rcu: Simplify rcu_init_nohz() cpumask handling

[Paul E. McKenney]

On Wed, Sep 14, 2022 at 12:46:34PM +0200, Frederic Weisbecker wrote:
> On Tue, Sep 13, 2022 at 11:00:36AM +0800, Zhen Lei wrote:
> > In kernels built with either CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y or
> > CONFIG_NO_HZ_FULL=y, additional CPUs must be added to rcu_nocb_mask.
> > Except that kernels booted without the rcu_nocbs= will not have
> > allocated rcu_nocb_mask.  And the current rcu_init_nohz() function uses
> > its need_rcu_nocb_mask and offload_all local variables to track the
> > rcu_nocb and nohz_full state.
> > 
> > But there is a much simpler approach, namely creating a cpumask pointer
> > to track the default and then using cpumask_available() to check the
> > rcu_nocb_mask state.  This commit takes this approach, thereby simplifying
> > and shortening the rcu_init_nohz() function.
> > 
> > Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
> > Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > Acked-by: Frederic Weisbecker <frederic@kernel.org>
> 
> Looks good, thanks!

Queued and pushed, thank you all!

							Thanx, Paul

Re: [PATCH v7 1/1] rcu: Simplify rcu_init_nohz() cpumask handling

[Paul E. McKenney]

On Wed, Sep 14, 2022 at 04:30:06AM -0700, Paul E. McKenney wrote:
> On Wed, Sep 14, 2022 at 12:46:34PM +0200, Frederic Weisbecker wrote:
> > On Tue, Sep 13, 2022 at 11:00:36AM +0800, Zhen Lei wrote:
> > > In kernels built with either CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y or
> > > CONFIG_NO_HZ_FULL=y, additional CPUs must be added to rcu_nocb_mask.
> > > Except that kernels booted without the rcu_nocbs= will not have
> > > allocated rcu_nocb_mask.  And the current rcu_init_nohz() function uses
> > > its need_rcu_nocb_mask and offload_all local variables to track the
> > > rcu_nocb and nohz_full state.
> > > 
> > > But there is a much simpler approach, namely creating a cpumask pointer
> > > to track the default and then using cpumask_available() to check the
> > > rcu_nocb_mask state.  This commit takes this approach, thereby simplifying
> > > and shortening the rcu_init_nohz() function.
> > > 
> > > Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
> > > Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > > Acked-by: Frederic Weisbecker <frederic@kernel.org>
> > 
> > Looks good, thanks!
> 
> Queued and pushed, thank you all!

And this time, actually queued the most recent version.  Apologies for
my confusion.

							Thanx, Paul