Re: [PATCH 00/13] srcu: Round out atomic SRCU support

[Zqiang]

> 
> On Thu, Sep 10, 2026 at 8:16 PM Zqiang <qiang.zhang@xxxxxxxxx> wrote:
> 
> > 
> > From: Kunwu Chan <kunwu.chan@xxxxxxxxx>
> > 
> >  This series rounds out atomic SRCU support with a reader-free fastpath,
> >  some atomic-context fixes, and additional test coverage.
> > 
> >  The core change is a reader-free fastpath for synchronize_srcu_atomic().
> >  Atomic SRCU's read-side critical sections disable preemption, so when
> >  the summed lock and unlock counts match for both ranks,
> >  synchronize_srcu_atomic() can end the grace period immediately, without
> >  the index flip or the srcu_advance_state() scans. Correctness depends
> >  on the grace-period anchor written by srcu_gp_start() preceding the
> >  lock-count scan; two LKMM litmus tests (patches 1-2) verify that
> >  ordering.
> > 
> >  For synchronize_srcu_atomic(), the following items maybe also
> >  need to be added:
> > 
> Hi Zqiang
> Sent as: https://lore.kernel.org/rcu/20260911093515.2802551-1-kunwu.chan@xxxxxxxxx/
> 
> Did a small late_initcall that enters an atomic read-side critical section and
> calls synchronize_srcu_atomic(). srcu_lock_sync() catches it:
>  WARNING: possible recursive locking detected
>  swapper/0/1 is trying to acquire lock: (&ss) at:
> synchronize_srcu_atomic+0x28/0x380
>  but task is already holding lock: (&ss) at:
> srcu_lockdep_test_init+0xd0/0x170
>  *** DEADLOCK ***
> Rcutorture SRCU-N/P also pass.

Hi, kunwu

You can add srcu_atomic lockdep test to rcutorture to verify
your change, and I also saw that Paul suggested the same thing :)

For example, as shown below (I am not test or verify it):

diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
index d795e07a12ad..d4f7399acbd5 100644
--- a/kernel/rcu/rcutorture.c
+++ b/kernel/rcu/rcutorture.c
@@ -4662,6 +4662,17 @@ static DECLARE_RWSEM(rwsem7);
 static DECLARE_RWSEM(rwsem8);
 static DECLARE_RWSEM(rwsem9);

+static DEFINE_RAW_SPINLOCK(spin0);
+static DEFINE_RAW_SPINLOCK(spin1);
+static DEFINE_RAW_SPINLOCK(spin2);
+static DEFINE_RAW_SPINLOCK(spin3);
+static DEFINE_RAW_SPINLOCK(spin4);
+static DEFINE_RAW_SPINLOCK(spin5);
+static DEFINE_RAW_SPINLOCK(spin6);
+static DEFINE_RAW_SPINLOCK(spin7);
+static DEFINE_RAW_SPINLOCK(spin8);
+static DEFINE_RAW_SPINLOCK(spin9);
+
 DEFINE_STATIC_SRCU(srcu0);
 DEFINE_STATIC_SRCU(srcu1);
 DEFINE_STATIC_SRCU(srcu2);
@@ -4673,6 +4684,17 @@ DEFINE_STATIC_SRCU(srcu7);
 DEFINE_STATIC_SRCU(srcu8);
 DEFINE_STATIC_SRCU(srcu9);

+DEFINE_STATIC_SRCU_ATOMIC(srcu0_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu1_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu2_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu3_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu4_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu5_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu6_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu7_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu8_atomic);
+DEFINE_STATIC_SRCU_ATOMIC(srcu9_atomic);
+
 static int srcu_lockdep_next(const char *f, const char *fl, const char *fs, const char *fu, int i,
                             int cyclelen, int deadlock)
 {
@@ -4702,6 +4724,13 @@ static void rcu_torture_init_srcu_lockdep(void)
                                          &rwsem5, &rwsem6, &rwsem7, &rwsem8, &rwsem9 };
        struct srcu_struct *srcus[] = { &srcu0, &srcu1, &srcu2, &srcu3, &srcu4,
                                        &srcu5, &srcu6, &srcu7, &srcu8, &srcu9 };
+
+       raw_spinlock_t *spins[] = { &spin0, &spin1, &spin2, &spin3, &spin4, &spin5,
+                                       &spin6, &spin7, &spin8, &spin9};
+       struct srcu_struct *srcus_atomic[] = {&srcu0_atomic, &srcu1_atomic, &srcu2_atomic,
+                                               &srcu3_atomic, &srcu4_atomic, &srcu5_atomic,
+                                               &srcu6_atomic, &srcu7_atomic, &srcu8_atomic,
+                                               &srcu9_atomic};
        int testtype;

        if (!test_srcu_lockdep)
@@ -4812,11 +4841,48 @@ static void rcu_torture_init_srcu_lockdep(void)
        }
 #endif // #ifdef CONFIG_TASKS_TRACE_RCU

+        if (testtype == 4) {
+                pr_info("%s: test_srcu_lockdep = %05d: SRCU_ATOMIC %d-way %sdeadlock.\n",
+                        __func__, test_srcu_lockdep, cyclelen, deadlock ? "" : "non-");
+                if (deadlock && cyclelen == 1)
+                        pr_info("%s: Expect hang.\n", __func__);
+                for (i = 0; i < cyclelen; i++) {
+                        j = srcu_lockdep_next(__func__, "srcu_read_lock_atomic", "synchronize_srcu_atomic",
+                                              "srcu_read_unlock_atomic", i, cyclelen, deadlock);
+                        idx = srcu_read_lock_atomic(srcus_atomic[i]);
+                        if (j >= 0)
+                                synchronize_srcu_atomic(srcus_atomic[j]);
+                        srcu_read_unlock_atomic(srcus_atomic[i], idx);
+                }
+                return;
+        }
+
+        if (testtype == 1) {
+                pr_info("%s: test_srcu_lockdep = %05d: SRCU_ATOMIC/raw_spinlock %d-way %sdeadlock.\n",
+                        __func__, test_srcu_lockdep, cyclelen, deadlock ? "" : "non-");
+                for (i = 0; i < cyclelen; i++) {
+                        pr_info("%s: srcu_read_lock_atomic(%d), raw_spin_lock(%d), raw_spin_unlock(%d), srcu_read_unlock_atomic(%d)\n",
+                                __func__, i, i, i, i);
+                        idx = srcu_read_lock_atomic(srcus_atomic[i]);
+                        raw_spin_lock(spins[i]);
+                        raw_spin_unlock(spins[i]);
+                        srcu_read_unlock_atomic(srcus_atomic[i], idx);
+
+                        j = srcu_lockdep_next(__func__, "raw_spin_lock", "synchronize_srcu_atomic",
+                                              "raw_spin_unlock", i, cyclelen, deadlock);
+                        raw_spin_lock(spins[i]);
+                        if (j >= 0)
+                                synchronize_srcu_atomic(srcus_atomic[j]);
+                        raw_spin_unlock(spins[i]);
+                }
+                return;
+        }
+
 err_out:
        pr_info("%s: test_srcu_lockdep = %05d does nothing.\n", __func__, test_srcu_lockdep);
        pr_info("%s: test_srcu_lockdep = DNNL.\n", __func__);
        pr_info("%s: D: Deadlock if nonzero.\n", __func__);
-       pr_info("%s: NN: Test number, 0=SRCU, 1=SRCU/mutex, 2=SRCU/rwsem, 3=SRCU/Tasks Trace RCU.\n", __func__);
+       pr_info("%s: NN: Test number, 0=SRCU, 1=SRCU/mutex, 2=SRCU/rwsem, 3=SRCU/Tasks Trace RCU, 4=SRCU_ATOMIC, 5=SRCU_ATOMIC/raw_spinlock.\n", __func__);
        pr_info("%s: L: Cycle length.\n", __func__);
        if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU))
                pr_info("%s: NN=3 disallowed because kernel is built with CONFIG_TASKS_TRACE_RCU=n\n", __func__);


Thanks
Zqiang





> 
> Thanks, Kunwu
> 
> > 
> > srcu_lock_sync(&ssp->dep_map);
> > 
> >  RCU_LOCKDEP_WARN(lockdep_is_held(ssp),
> >  "Illegal synchronize_srcu() in same-type SRCU read-side critical section");
> > 
> >  if (rcu_scheduler_active == RCU_SCHEDULER_INACTIVE)
> >  return;
> > 
> >  Thanks
> >  Zqiang
> > 
> >  The series is:
> > 
> >  1-2 LKMM litmus tests for the reader-free fastpath ordering.
> >  3 srcutree: add the reader-free fastpath.
> >  4 rcutorture: add --do-atomic-srcu to torture.sh.
> >  5-7 srcutree: honor and propagate is_atomic so atomic SRCU never
> >  transitions to big.
> >  8 srcutree: forbid srcu_expedite_current() on atomic SRCU.
> >  9 rcutorture: disable srcu_expedite_current() for atomic SRCU.
> >  10 srcutree: skip callback scheduling for atomic SRCU grace periods.
> >  11 srcutree: drop the srcu_barrier() sleep for atomic SRCU.
> >  12 srcutree: remove leftover debug pr_alert()s.
> >  13 srcu: restrict the non_block annotation to task context, fixing a
> >  KCSAN data race.
> > 
> >  The KCSAN race in patch 13 is a data race on the interrupted task's
> >  ->non_block_count when a timer callback runs atomic-SRCU readers in
> >  inline softirq context.
> > 
> >  Tested with herd7 7.58 (litmus tests), torture.sh --do-atomic-srcu, and
> >  torture.sh --do-atomic-srcu --do-kcsan (KCSAN+PROVE_LOCKING).
> > 
> >  Kunwu Chan (13):
> >  litmus: Add SRCU fastpath anchor-before-scan test
> >  litmus: Add SRCU fastpath scan-before-anchor test
> >  srcutree: Add reader-free fastpath to synchronize_srcu_atomic()
> >  rcutorture: Add atomic-SRCU support to torture.sh
> >  srcutree: Honor is_atomic in check_init_srcu_struct()
> >  srcutree: Make init_srcu_struct_atomic() prevent transition to big
> >  srcutree: Don't transition atomic SRCU to big in srcu_gp_end()
> >  srcutree: Forbid srcu_expedite_current() on atomic SRCU
> >  rcutorture: Disable srcu_expedite_current() for atomic SRCU
> >  srcutree: Skip callback scheduling for atomic SRCU grace periods
> >  srcutree: Remove srcu_barrier() sleep for atomic SRCU
> >  srcutree: Remove debug pr_alert()s
> >  srcu: Restrict atomic-SRCU non_block annotation to task context
> > 
> >  include/linux/srcu.h | 10 +-
> >  kernel/rcu/rcutorture.c | 1 +
> >  kernel/rcu/srcutree.c | 95 ++++++++++++++-----
> >  tools/memory-model/litmus-tests/README | 16 ++++
> >  .../SRCU-fastpath-anchor-before-scan.litmus | 56 +++++++++++
> >  .../SRCU-fastpath-scan-before-anchor.litmus | 53 +++++++++++
> >  .../selftests/rcutorture/bin/torture.sh | 24 +++++
> >  7 files changed, 229 insertions(+), 26 deletions(-)
> >  create mode 100644 tools/memory-model/litmus-tests/SRCU-fastpath-anchor-before-scan.litmus
> >  create mode 100644 tools/memory-model/litmus-tests/SRCU-fastpath-scan-before-anchor.litmus
> > 
> >  base-commit: ef15070f96c5f07de3a9b593c36f6e83bc9e3d58
> >  --
> >  2.43.0
> >
>