CPU stall-warning changes for 3.8
Subsystems (2)
cover CPU stall-warning changes for 3.8
Hello!
This patch series adds updates to RCU's CPU stall warning capability, with patches as follows:
1. For architectures without NMI-based stack-dump capability,
have the current CPU remotely access the stalled CPU's stack to dump it.
2. Include the number of the last grace period started and completed
in RCU CPU stall warning messages.
3. Dump the number of RCU callbacks on the system in order to help
diagnose situations where spurious grace periods were started.
4. Prototype patch to provide RCU CPU stall warning messages for
Tiny RCU.
Thanx, Paul
b/include/linux/sched.h | 2 b/kernel/rcu.h | 7 +++ b/kernel/rcupdate.c | 52 +++++++++++++++++++++++++ b/kernel/rcutiny.c | 6 +- b/kernel/rcutiny_plugin.h | 56 +++++++++++++++++++++++++++ b/kernel/rcutree.c | 94 +++++++++++++++++++++------------------------- b/kernel/rcutree.h | 5 -- b/kernel/sched/core.c | 6 ++ b/lib/Kconfig.debug | 2 9 files changed, 171 insertions(+), 59 deletions(-)
PATCH 1/4 rcu: Print remote CPU's stacks in stall warnings
From: "Paul E. McKenney" <[email protected]>
The RCU CPU stall warnings rely on trigger_all_cpu_backtrace() to do NMI-based dump of the stack traces of all CPUs. Unfortunately, a number of architectures do not implement trigger_all_cpu_backtrace(), in which case RCU falls back to just dumping the stack of the running CPU. This is unhelpful in the case where the running CPU has detected that some other CPU has stalled.
This commit therefore makes the running CPU dump the stacks of the tasks running on the stalled CPUs.
--- include/linux/sched.h | 2 ++ kernel/rcutree.c | 25 ++++++++++++++++++++++++- kernel/sched/core.c | 6 ++++++ 3 files changed, 32 insertions(+), 1 deletions(-)
include/linux/sched.h +2 -0
@@ -109,6 +109,8 @@ extern void update_cpu_load_nohz(void);extern unsigned long get_parent_ip (unsigned long addr); +extern void dump_cpu_task (int cpu); +struct seq_file ;struct cfs_rq ;struct task_group ;
kernel/rcutree.c +24 -1
@@ -873,6 +873,29 @@ static void record_gp_stall_check_time(struct rcu_state *rsp) rsp->jiffies_stall = jiffies +jiffies_till_stall_check (); } +/* +* Dump stacks of all tasks running on stalled CPUs. This is a fallback +* for architectures that do not implement trigger_all_cpu_backtrace(). +* The NMI-triggered stack traces are more accurate because they are +* printed by the target CPU. +*/ +static void rcu_dump_cpu_stacks (struct rcu_state *rsp) +{ +int cpu; +unsigned long flags; +struct rcu_node *rnp; + +rcu_for_each_leaf_node (rsp, rnp) { +raw_spin_lock_irqsave (&rnp->lock, flags); +if (rnp->qsmask !=0 ) { +for (cpu =0 ; cpu <= rnp->grphi - rnp->grplo; cpu++) +if (rnp->qsmask & (1UL << cpu)) +dump_cpu_task (rnp->grplo + cpu); + } +raw_spin_unlock_irqrestore (&rnp->lock, flags); + } +} +static void print_other_cpu_stall (struct rcu_state *rsp) {int cpu;
@@ -929,7 +952,7 @@ static void print_other_cpu_stall(struct rcu_state *rsp)if (ndetected ==0 )printk (KERN_ERR "INFO: Stall ended before state dump start\n" );else if (!trigger_all_cpu_backtrace ()) -dump_stack (); +rcu_dump_cpu_stacks (rsp);/* Complain about tasks blocking the grace period. */
kernel/sched/core.c +6 -0
@@ -8076,3 +8076,9 @@ struct cgroup_subsys cpuacct_subsys = { .base_cftypes = files, };#endif /* CONFIG_CGROUP_CPUACCT */ + +void dump_cpu_task (int cpu) +{ +pr_info ("Task dump for CPU %d:\n" , cpu); +sched_show_task (cpu_curr (cpu)); +}
Signature · 1 line
1.7.8
PATCH 2/4 rcu: Add grace-period information to RCU CPU stall warnings
From: "Paul E. McKenney" <[email protected]>
This commit causes the last grace period started and completed to be printed on RCU CPU stall warning messages in order to aid diagnosis.
--- kernel/rcutree.c | 8 +++++--- 1 files changed, 5 insertions(+), 3 deletions(-)
kernel/rcutree.c +5 -3
@@ -947,8 +947,9 @@ static void print_other_cpu_stall(struct rcu_state *rsp)raw_spin_unlock_irqrestore (&rnp->lock, flags);print_cpu_stall_info_end (); -printk (KERN_CONT "(detected by %d, t=%ld jiffies)\n" , -smp_processor_id (), (long )(jiffies - rsp->gp_start)); +pr_cont ("(detected by %d, t=%ld jiffies, g=%lu, c=%lu)\n" , +smp_processor_id (), (long )(jiffies - rsp->gp_start), + rsp->gpnum, rsp->completed);if (ndetected ==0 )printk (KERN_ERR "INFO: Stall ended before state dump start\n" );else if (!trigger_all_cpu_backtrace ())
@@ -975,7 +976,8 @@ static void print_cpu_stall(struct rcu_state *rsp)print_cpu_stall_info_begin ();print_cpu_stall_info (rsp,smp_processor_id ());print_cpu_stall_info_end (); -printk (KERN_CONT " (t=%lu jiffies)\n" , jiffies - rsp->gp_start); +pr_cont (" (t=%lu jiffies g=%lu c=%lu)\n" , + jiffies - rsp->gp_start, rsp->gpnum, rsp->completed);if (!trigger_all_cpu_backtrace ())dump_stack ();
Signature · 1 line
1.7.8
PATCH 3/4 rcu: Dump number of callbacks in stall warning messages
In reply to Paul E. McKenneyFrom: "Paul E. McKenney" <[email protected]>
In theory, if a grace period manages to get started despite there being no callbacks on any of the CPUs, all CPUs could go into dyntick-idle mode, so that the grace period would never end. This commit updates the RCU CPU stall warning messages to detect this condition by summing up the number of callbacks on all CPUs.
--- kernel/rcutree.c | 15 +++++++++++---- 1 files changed, 11 insertions(+), 4 deletions(-)
kernel/rcutree.c +11 -4
@@ -903,6 +903,7 @@ static void print_other_cpu_stall(struct rcu_state *rsp)unsigned long flags;int ndetected =0 ;struct rcu_node *rnp =rcu_get_root (rsp); +long totqlen =0 ;/* Only let one CPU complain about others per time interval. */
@@ -947,9 +948,11 @@ static void print_other_cpu_stall(struct rcu_state *rsp)raw_spin_unlock_irqrestore (&rnp->lock, flags);print_cpu_stall_info_end (); -pr_cont ("(detected by %d, t=%ld jiffies, g=%lu, c=%lu)\n" , +for_each_possible_cpu (cpu) + totqlen +=per_cpu_ptr (rsp->rda, cpu)->qlen; +pr_cont ("(detected by %d, t=%ld jiffies, g=%lu, c=%lu, q=%lu)\n" ,smp_processor_id (), (long )(jiffies - rsp->gp_start), - rsp->gpnum, rsp->completed); + rsp->gpnum, rsp->completed, totqlen);if (ndetected ==0 )printk (KERN_ERR "INFO: Stall ended before state dump start\n" );else if (!trigger_all_cpu_backtrace ())
@@ -964,8 +967,10 @@ static void print_other_cpu_stall(struct rcu_state *rsp)static void print_cpu_stall (struct rcu_state *rsp) { +int cpu;unsigned long flags;struct rcu_node *rnp =rcu_get_root (rsp); +long totqlen =0 ;/* * OK, time to rat on ourselves...
@@ -976,8 +981,10 @@ static void print_cpu_stall(struct rcu_state *rsp)print_cpu_stall_info_begin ();print_cpu_stall_info (rsp,smp_processor_id ());print_cpu_stall_info_end (); -pr_cont (" (t=%lu jiffies g=%lu c=%lu)\n" , - jiffies - rsp->gp_start, rsp->gpnum, rsp->completed); +for_each_possible_cpu (cpu) + totqlen +=per_cpu_ptr (rsp->rda, cpu)->qlen; +pr_cont (" (t=%lu jiffies g=%lu c=%lu q=%lu)\n" , + jiffies - rsp->gp_start, rsp->gpnum, rsp->completed, totqlen);if (!trigger_all_cpu_backtrace ())dump_stack ();
Signature · 1 line
1.7.8
PATCH 4/4 rcu: Provide RCU CPU stall warnings for tiny RCU
In reply to Paul E. McKenneyFrom: "Paul E. McKenney" <[email protected]>
Tiny RCU has historically omitted RCU CPU stall warnings in order to reduce memory requirements, however, lack of these warnings caused Thomas Gleixner some debugging pain recently. Therefore, this commit adds RCU CPU stall warnings to tiny RCU if RCU_TRACE=y. This keeps the memory footprint small, while still enabling CPU stall warnings in kernels built to enable them.
--- kernel/rcu.h | 7 ++++++ kernel/rcupdate.c | 52 +++++++++++++++++++++++++++++++++++++++++++ kernel/rcutiny.c | 6 +++- kernel/rcutiny_plugin.h | 56 +++++++++++++++++++++++++++++++++++++++++++++++ kernel/rcutree.c | 46 ++------------------------------------ kernel/rcutree.h | 5 ---- lib/Kconfig.debug | 2 +- 7 files changed, 123 insertions(+), 51 deletions(-)
kernel/rcu.h +7 -0
@@ -109,4 +109,11 @@ static inline bool __rcu_reclaim(char *rn, struct rcu_head *head) } } +#if defined(CONFIG_SMP ) || defined(CONFIG_RCU_TRACE ) + +extern int rcu_cpu_stall_suppress; +int rcu_jiffies_till_stall_check (void ); + +#endif /* defined(CONFIG_SMP) || defined(CONFIG_RCU_TRACE) */ +#endif /* __LINUX_RCU_H */
kernel/rcupdate.c +52 -0
61 lines not on this page · raw message
kernel/rcutiny.c +4 -2
@@ -51,10 +51,10 @@ static void __call_rcu(struct rcu_head *head,void (*func)(struct rcu_head *rcu), struct rcu_ctrlblk *rcp); -#include "rcutiny_plugin.h" -static long long rcu_dynticks_nesting =DYNTICK_TASK_EXIT_IDLE ; +#include "rcutiny_plugin.h" +/* Common code for rcu_idle_enter() and rcu_irq_exit(), see kernel/rcutree.c. */ static void rcu_idle_enter_common (long long newval) {
@@ -205,6 +205,7 @@ int rcu_is_cpu_rrupt_from_idle(void)*/ static int rcu_qsctr_help (struct rcu_ctrlblk *rcp) { +reset_cpu_stall_ticks (rcp);if (rcp->rcucblist !=NULL && rcp->donetail != rcp->curtail) { rcp->donetail = rcp->curtail;
@@ -251,6 +252,7 @@ void rcu_bh_qs(int cpu)*/ void rcu_check_callbacks (int cpu,int user) { +check_cpu_stalls ();if (user ||rcu_is_cpu_rrupt_from_idle ())rcu_sched_qs (cpu);else if (!in_softirq ())
kernel/rcutiny_plugin.h +56 -0
78 lines not on this page · raw message
kernel/rcutree.c +3 -43
91 lines not on this page · raw message
kernel/rcutree.h +0 -5
@@ -330,11 +330,6 @@ struct rcu_data {#define RCU_JIFFIES_TILL_FORCE_QS 3/* for rsp->jiffies_force_qs */ -#ifdef CONFIG_PROVE_RCU -#define RCU_STALL_DELAY_DELTA (5 * HZ) -#else -#define RCU_STALL_DELAY_DELTA 0 -#endif #define RCU_STALL_RAT_DELAY 2/* Allow other CPUs time */ /* to take at least one */ /* scheduling clock irq */
lib/Kconfig.debug +1 -1
@@ -970,7 +970,7 @@ config RCU_TORTURE_TEST_RUNNABLE config RCU_CPU_STALL_TIMEOUT int "RCU CPU stall timeout in seconds" - depends on TREE_RCU || TREE_PREEMPT_RCU + depends on TREE_RCU || TREE_PREEMPT_RCU || RCU_TRACE range 3 300 default 60 help
Signature · 1 line
1.7.8