Merge tag 'rcu.release.v6.17' of git://git.kernel.org/pub/scm/linux/kernel/git/rcu/linux

Pull RCU updates from Neeraj Upadhyay:
 "Expedited grace period updates:

   - Protect against early RCU exp quiescent state reporting during exp
     grace period initialization

   - Remove superfluous barrier in task unblock path

   - Remove the CPU online quiescent state report optimization, which is
     error prone for certain scenarios

   - Add warning for unexpected pending requested expedited quiescent
     state on dying CPU

  Core:

   - Robustify rcu_is_cpu_rrupt_from_idle() by using more accurate
     indicators of the actual context tracking state of a CPU

   - Handle ->defer_qs_iw_pending field data race

   - Enable rcu_normal_wake_from_gp by default on systems with <= 16
     CPUs

   - Fix lockup in rcu_read_unlock() due to recursive irq_exit() calls

   - Refactor expedited handling condition in rcu_read_unlock_special()

   - Documentation updates for hotplug and GP init scan ordering,
     separation of rcu_state and rnp's gp_seq states, quiescent state
     reporting for offline CPUs

  torture-scripts:

   - Cleanup and improve scripts : remove superfluous warnings for
     disabled tests; better handling of kvm.sh --kconfig arg; suppress
     some confusing diagnostics; tolerate bad kvm.sh args; add new
     diagnostic for build output; fail allmodconfig testing on warnings

   - Include RCU_TORTURE_TEST_CHK_RDR_STATE config for KCSAN kernels

   - Disable default RCU-tasks and clocksource-wdog testing on arm64

   - Add EXPERT Kconfig option for arm64 KCSAN runs

   - Remove SRCU-lite testing

  rcutorture:

   - Start torture writer threads creation after reader threads to
     handle race in SRCU-P scenario

   - Add SRCU down_read()/up_read() test

   - Add diagnostics for delayed SRCU up_read(), unmatched up_read(),
     print number of up/down readers and the number of such readers
     which migrated to other CPU

   - Ignore certain unsupported configurations for trivial RCU test

   - Fix splats in RT kernels due to inaccurate checks for BH-disabled
     context

   - Enable checks and logs to capture intentionally exercised
     unexpected scenarios (too short readers) for BUSTED test

   - Remove SRCU-lite testing

  srcu:

   - Expedite SRCU-fast grace periods

   - Remove SRCU-lite implementation

   - Add guards for SRCU-fast readers

  rcu nocb:

   - Dump NOCB group leader state on stall detection

   - Robustify nocb_cb_kthread pointer accesses

   - Fix delayed execution of hurry callbacks when LAZY_RCU is enabled

  refscale:

   - Fix multiplication overflow in "loops" and "nreaders" calculations"

* tag 'rcu.release.v6.17' of git://git.kernel.org/pub/scm/linux/kernel/git/rcu/linux: (49 commits)
  rcu: Document concurrent quiescent state reporting for offline CPUs
  rcu: Document separation of rcu_state and rnp's gp_seq
  rcu: Document GP init vs hotplug-scan ordering requirements
  srcu: Add guards for SRCU-fast readers
  rcu: Fix delayed execution of hurry callbacks
  rcu: Refactor expedited handling check in rcu_read_unlock_special()
  checkpatch: Remove SRCU-lite deprecation
  srcu: Remove SRCU-lite implementation
  srcu: Expedite SRCU-fast grace periods
  rcutorture: Remove support for SRCU-lite
  rcutorture: Remove SRCU-lite scenarios
  torture: Remove support for SRCU-lite
  torture: Make torture.sh --allmodconfig testing fail on warnings
  torture: Add "ERROR" diagnostic for testing kernel-build output
  torture: Make torture.sh tolerate runs having bad kvm.sh arguments
  torture: Add textid.txt file to --do-allmodconfig and --do-rcu-rust runs
  torture: Extract testid.txt generation to separate script
  torture: Suppress "find" diagnostics from torture.sh --do-none run
  torture: Provide EXPERT Kconfig option for arm64 KCSAN torture.sh runs
  rcu: Fix rcu_read_unlock() deadloop due to IRQ work
  ...

1 files changed, 33 insertions, 0 deletions

diff --git a/Documentation/RCU/Design/Data-Structures/Data-Structures.rst b/Documentation/RCU/Design/Data-Structures/Data-Structures.rst
index 04e16775c752..1b0aad184dd7 100644
--- a/Documentation/RCU/Design/Data-Structures/Data-Structures.rst
+++ b/Documentation/RCU/Design/Data-Structures/Data-Structures.rst
@@ -286,6 +286,39 @@ in order to detect the beginnings and ends of grace periods in a
 distributed fashion. The values flow from ``rcu_state`` to ``rcu_node``
 (down the tree from the root to the leaves) to ``rcu_data``.
 
++-----------------------------------------------------------------------+
+| **Quick Quiz**:                                                       |
++-----------------------------------------------------------------------+
+| Given that the root rcu_node structure has a gp_seq field,            |
+| why does RCU maintain a separate gp_seq in the rcu_state structure?   |
+| Why not just use the root rcu_node's gp_seq as the official record    |
+| and update it directly when starting a new grace period?              |
++-----------------------------------------------------------------------+
+| **Answer**:                                                           |
++-----------------------------------------------------------------------+
+| On single-node RCU trees (where the root node is also a leaf),        |
+| updating the root node's gp_seq immediately would create unnecessary  |
+| lock contention. Here's why:                                          |
+|                                                                       |
+| If we did rcu_seq_start() directly on the root node's gp_seq:         |
+|                                                                       |
+| 1. All CPUs would immediately see their node's gp_seq from their rdp's|
+|    gp_seq, in rcu_pending(). They would all then invoke the RCU-core. |
+| 2. Which calls note_gp_changes() and try to acquire the node lock.    |
+| 3. But rnp->qsmask isn't initialized yet (happens later in            |
+|    rcu_gp_init())                                                     |
+| 4. So each CPU would acquire the lock, find it can't determine if it  |
+|    needs to report quiescent state (no qsmask), update rdp->gp_seq,   |
+|    and release the lock.                                              |
+| 5. Result: Lots of lock acquisitions with no grace period progress    |
+|                                                                       |
+| By having a separate rcu_state.gp_seq, we can increment the official  |
+| grace period counter without immediately affecting what CPUs see in   |
+| their nodes. The hierarchical propagation in rcu_gp_init() then       |
+| updates the root node's gp_seq and qsmask together under the same lock|
+| acquisition, avoiding this useless contention.                        |
++-----------------------------------------------------------------------+
+
 Miscellaneous
 '''''''''''''