The current RCU CPU stall-warning code can say "INFO: Stall ended before state dump start" but then panic the system anyway. This commit therefore avoids panicking the system if the stall has already ended.
However, the sched_ext scheduler will still be ejected, as it might have been the cause of the almost stall. This can easily be changed if needed. Signed-off-by: Paul E. McKenney <[email protected]> --- kernel/rcu/tree_exp.h | 2 +- kernel/rcu/tree_stall.h | 12 +++++++++--- 2 files changed, 10 insertions(+), 4 deletions(-) diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h index 81d42b6e2e31..01bdbd7e28cc 100644 --- a/kernel/rcu/tree_exp.h +++ b/kernel/rcu/tree_exp.h @@ -678,7 +678,7 @@ static void synchronize_rcu_expedited_wait(void) nbcon_cpu_emergency_exit(); - panic_on_rcu_stall(&rcu_exp_stall_cpumask); + panic_on_rcu_stall(&rcu_exp_stall_cpumask, NULL, 0); } } diff --git a/kernel/rcu/tree_stall.h b/kernel/rcu/tree_stall.h index 091e7850ab6e..93ba31a619b6 100644 --- a/kernel/rcu/tree_stall.h +++ b/kernel/rcu/tree_stall.h @@ -159,7 +159,8 @@ static int __init check_cpu_stall_init(void) early_initcall(check_cpu_stall_init); /* If so specified via sysctl, panic, yielding cleaner stall-warning output. */ -static void panic_on_rcu_stall(const struct cpumask *stalled_mask) +static void panic_on_rcu_stall(const struct cpumask *stalled_mask, + unsigned long *gsp, unsigned long gp_seq) { static int cpu_stall; @@ -170,6 +171,11 @@ static void panic_on_rcu_stall(const struct cpumask *stalled_mask) if (scx_rcu_cpu_stall(stalled_mask)) return; + if (gsp && rcu_seq_current(gsp) != gp_seq) { + pr_err("INFO: Stall ended before panic check.\n"); + return; + } + if (++cpu_stall < sysctl_max_rcu_stall_to_panic) return; @@ -703,7 +709,7 @@ static void print_other_cpu_stall(unsigned long gp_seq, unsigned long gps) nbcon_cpu_emergency_exit(); - panic_on_rcu_stall(&rcu_stall_cpumask); + panic_on_rcu_stall(&rcu_stall_cpumask, &rcu_state.gp_seq, gp_seq); rcu_force_quiescent_state(); /* Kick them all. */ } @@ -758,7 +764,7 @@ static void print_cpu_stall(unsigned long gp_seq, unsigned long gps) cpumask_clear(&rcu_stall_cpumask); cpumask_set_cpu(smp_processor_id(), &rcu_stall_cpumask); - panic_on_rcu_stall(&rcu_stall_cpumask); + panic_on_rcu_stall(&rcu_stall_cpumask, &rcu_state.gp_seq, gp_seq); /* * Attempt to revive the RCU machinery by forcing a context switch. -- 2.40.1
