On Fri, Feb 24, 2023 at 07:32:22PM -0800, Randy Dunlap wrote:

Hi--

On 2/24/23 19:27, Joel Fernandes (Google) wrote:

quoted

On many systems, a great deal of boot happens after the kernel thinks the boot
has completed. It is difficult to determine if the system has really booted
from the kernel side. Some features like lazy-RCU can risk slowing down boot
time if, say, a callback has been added that the boot synchronously depends on.

Further, it is better to boot systems which pass 'rcu_normal_after_boot' to
stay expedited for as long as the system is still booting.

For these reasons, this commit adds a config option
'CONFIG_RCU_BOOT_END_DELAY' and a boot parameter rcupdate.boot_end_delay.

By default, this value is 20s. A system designer can choose to specify a value
here to keep RCU from marking boot completion.  The boot sequence will not be
marked ended until at least boot_end_delay milliseconds have passed.

Signed-off-by: Joel Fernandes (Google) <redacted>
---
 .../admin-guide/kernel-parameters.txt         |  4 +++
 kernel/rcu/Kconfig                            | 12 +++++++++
 kernel/rcu/update.c                           | 25 +++++++++++++++++--
 3 files changed, 39 insertions(+), 2 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 2429b5e3184b..0943139fdf01 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt

@@ -5085,6 +5085,10 @@
 	rcutorture.verbose= [KNL]
 			Enable additional printk() statements.
 
+	rcupdate.boot_end_delay= [KNL]

Tell units:

quoted

+			Minimum time that must elapse before the boot

Fixed.

+			Minimum time in milliseconds that must elapse before the boot

quoted

+			sequence can be marked as completed.
+
 	rcupdate.rcu_cpu_stall_ftrace_dump= [KNL]
 			Dump ftrace buffer after reporting RCU CPU
 			stall warning.

diff --git a/kernel/rcu/Kconfig b/kernel/rcu/Kconfig
index 9071182b1284..1033a38bddad 100644
--- a/kernel/rcu/Kconfig
+++ b/kernel/rcu/Kconfig

@@ -217,6 +217,18 @@ config RCU_BOOST_DELAY
 
 	  Accept the default if unsure.
 
+config RCU_BOOT_END_DELAY
+	int "Minimum delay before RCU considers boot has completed"
+	range 0 120000
+	default 20000
+	help
+	  This option specifies the minmum time since boot before which

tpyo:	                            minimum

quoted

+	  RCU believes the system is booted. The actual delay can be
+	  higher than this if the kernel takes a long time to initialize
+	  but it will never be smaller than this.

Fixed.

+	  Specified in milliseconds.

quoted

+
+	  Accept the default if unsure.
+
 config RCU_EXP_KTHREAD
 	bool "Perform RCU expedited work in a real-time kthread"
 	depends on RCU_BOOST && RCU_EXPERT

diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
index 19bf6fa3ee6a..5b73341d9b89 100644
--- a/kernel/rcu/update.c
+++ b/kernel/rcu/update.c

@@ -62,6 +62,10 @@ module_param(rcu_normal_after_boot, int, 0444);
 #endif
 #endif /* #ifndef CONFIG_TINY_RCU */
 
+/* Minimum time until RCU considers boot as completed. */
+static int boot_end_delay = CONFIG_RCU_BOOT_END_DELAY;
+module_param(boot_end_delay, int, 0444);
+
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 /**
  * rcu_read_lock_held_common() - might we be in RCU-sched read-side critical section?

@@ -225,12 +229,29 @@ void rcu_unexpedite_gp(void)
 EXPORT_SYMBOL_GPL(rcu_unexpedite_gp);
 
 static bool rcu_boot_ended __read_mostly;
-
 /*
- * Inform RCU of the end of the in-kernel boot sequence.
+ * Inform RCU of the end of the in-kernel boot sequence. The boot sequence will
+ * not be marked ended until at least boot_end_delay milliseconds have passed.
  */
+void rcu_end_inkernel_boot(void);
+static void boot_rcu_work_fn(struct work_struct *work)
+{
+	rcu_end_inkernel_boot();
+}
+static DECLARE_DELAYED_WORK(boot_rcu_work, boot_rcu_work_fn);
+
 void rcu_end_inkernel_boot(void)
 {
+	if (boot_end_delay) {
+		u64 boot_ms = ktime_get_boot_fast_ns() / 1000000UL;

Is that division OK on 32-bit?  Might have to use a helper macro. (I dunno.)

Ah, maybe. I am not sure if the compiler generates the right stubs. At least
in userspace it does and there's no problem on 32-bit system. I will research
it more, perhaps by building a 32-bit kernel and seeing what the compiler
does.

Will re-spin later after any more feedback.

thanks,

 - Joel

quoted

+
+		if (boot_ms < boot_end_delay) {
+			schedule_delayed_work(&boot_rcu_work,
+					boot_end_delay - boot_ms);
+			return;
+		}
+	}
+
 	rcu_unexpedite_gp();
 	rcu_async_relax();
 	if (rcu_normal_after_boot)

-- 
~Randy