1
linux/kernel
Paul E. McKenney d7c0651390 rcu: fix rcu_try_flip_waitack_needed() to prevent grace-period stall
The comment was correct -- need to make the code match the comment.
Without this patch, if a CPU goes dynticks idle (and stays there forever)
in just the right phase of preemptible-RCU grace-period processing,
grace periods stall.  The offending sequence of events (courtesy
of Promela/spin, at least after I got the liveness criterion coded
correctly...) is as follows:

o	CPU 0 is in dynticks-idle mode.  Its dynticks_progress_counter
	is (say) 10.

o	CPU 0 takes an interrupt, so rcu_irq_enter() increments CPU 0's
	dynticks_progress_counter to 11.

o	CPU 1 is doing RCU grace-period processing in rcu_try_flip_idle(),
	sees rcu_pending(), so invokes dyntick_save_progress_counter(),
	which in turn takes a snapshot of CPU 0's dynticks_progress_counter
	into CPU 0's rcu_dyntick_snapshot -- now set to 11.  CPU 1 then
	updates the RCU grace-period state to rcu_try_flip_waitack().

o	CPU 0 returns from its interrupt, so rcu_irq_exit() increments
	CPU 0's dynticks_progress_counter to 12.

o	CPU 1 later invokes rcu_try_flip_waitack(), which notices that
	CPU 0 has not yet responded, and hence in turn invokes
	rcu_try_flip_waitack_needed().  This function examines the
	state of CPU 0's dynticks_progress_counter and rcu_dyntick_snapshot
	variables, which it copies to curr (== 12) and snap (== 11),
	respectively.

	Because curr!=snap, the first condition fails.

	Because curr-snap is only 1 and snap is odd, the second
	condition fails.

	rcu_try_flip_waitack_needed() therefore incorrectly concludes
	that it must wait for CPU 0 to explicitly acknowledge the
	counter flip.

o	CPU 0 remains forever in dynticks-idle mode, never taking
	any more hardware interrupts or any NMIs, and never running
	any more tasks.  (Of course, -something- will usually eventually
	happen, which might be why we haven't seen this one in the
	wild.  Still should be fixed!)

Therefore the grace period never ends.  Fix is to make the code match
the comment, as shown below.  With this fix, the above scenario
would be satisfied with curr being even, and allow the grace period
to proceed.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Josh Triplett <josh@kernel.org>
Cc: Dipankar Sarma <dipankar@in.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-05-19 10:03:38 +02:00
..
irq genirq: reenable a nobody cared disabled irq when a new driver arrives 2008-05-02 13:40:34 +02:00
power Merge branches 'release', 'acpica', 'bugzilla-10224', 'bugzilla-9772', 'bugzilla-9916', 'ec', 'eeepc', 'idle', 'misc', 'pm-legacy', 'sysfs-links-2.6.26', 'thermal', 'thinkpad' and 'video' into release 2008-04-30 13:58:00 -04:00
time clocksource: allow read access to available/current_clocksource 2008-05-03 18:11:48 +02:00
.gitignore
acct.c
audit_tree.c
audit.c
audit.h
auditfilter.c Merge branch 'audit.b50' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current 2008-04-29 11:41:22 -07:00
auditsc.c
backtracetest.c
bounds.c Add kbuild.h that contains common definitions for kbuild users 2008-04-29 08:06:29 -07:00
capability.c
cgroup_debug.c CGroup API files: move "releasable" to cgroup_debug subsystem 2008-04-29 08:06:09 -07:00
cgroup.c mm: bdi: add separate writeback accounting capability 2008-04-30 08:29:50 -07:00
compat.c ntp: support for TAI 2008-05-01 08:03:59 -07:00
configs.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
cpu.c kernel: replace remaining __FUNCTION__ occurrences 2008-04-30 08:29:54 -07:00
cpuset.c Fix cpuset sched_relax_domain_level control file 2008-05-08 10:46:56 -07:00
delayacct.c
dma.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
exec_domain.c
exit.c [PATCH] split linux/file.h 2008-05-01 13:08:16 -04:00
extable.c
fork.c [PATCH] split linux/file.h 2008-05-01 13:08:16 -04:00
futex_compat.c
futex.c Removal of FUTEX_FD 2008-05-05 08:18:45 -07:00
hrtimer.c hrtimer: remove duplicate helper function 2008-05-03 18:11:48 +02:00
itimer.c
kallsyms.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
Kconfig.hz
Kconfig.preempt
kexec.c kexec: make extended crashkernel= syntax less confusing 2008-05-01 08:04:00 -07:00
kfifo.c
kgdb.c lib: create common ascii hex array 2008-05-14 19:11:14 -07:00
kmod.c [PATCH] split linux/file.h 2008-05-01 13:08:16 -04:00
kprobes.c
ksysfs.c
kthread.c Deprecate find_task_by_pid() 2008-04-30 08:29:48 -07:00
latencytop.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
lockdep_internals.h
lockdep_proc.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
lockdep.c
Makefile sched: add optional support for CONFIG_HAVE_UNSTABLE_SCHED_CLOCK 2008-05-05 23:56:18 +02:00
marker.c make marker_debug static 2008-04-30 08:29:49 -07:00
module.c module: don't ignore vermagic string if module doesn't have modversions 2008-05-09 07:45:18 -07:00
mutex-debug.c
mutex-debug.h
mutex.c
mutex.h
notifier.c ipc: re-enable msgmni automatic recomputing msgmni if set to negative 2008-04-29 08:06:13 -07:00
ns_cgroup.c
nsproxy.c ipc: sysvsem: refuse clone(CLONE_SYSVSEM|CLONE_NEWIPC) 2008-04-29 08:06:14 -07:00
panic.c
params.c
pid_namespace.c pidns: make pid->level and pid_ns->level unsigned 2008-04-30 08:29:49 -07:00
pid.c rcu: split list.h and move rcu-protected lists into rculist.h 2008-05-19 10:01:37 +02:00
pm_qos_params.c
posix-cpu-timers.c remove div_long_long_rem 2008-05-01 08:03:58 -07:00
posix-timers.c signals: join send_sigqueue() with send_group_sigqueue() 2008-04-30 08:29:36 -07:00
printk.c printk: don't read beyond string arguments' terminating zero 2008-04-30 08:29:52 -07:00
profile.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
ptrace.c make generic sys_ptrace unconditional 2008-05-01 10:21:54 -07:00
rcuclassic.c rcu: add memory barriers and comments to rcu_check_callbacks() 2008-05-19 10:01:36 +02:00
rcupdate.c rcu: add rcu_barrier_sched() and rcu_barrier_bh() 2008-05-19 10:01:36 +02:00
rcupreempt_trace.c
rcupreempt.c rcu: fix rcu_try_flip_waitack_needed() to prevent grace-period stall 2008-05-19 10:03:38 +02:00
rcutorture.c rcu: add call_rcu_sched() and friends to rcutorture 2008-05-19 10:01:37 +02:00
relay.c Revert "relay: fix splice problem" 2008-05-08 14:06:19 +02:00
res_counter.c memcgroup: add the max_usage member on the res_counter 2008-04-29 08:06:10 -07:00
resource.c kernel: use non-racy method for proc entries creation 2008-04-29 08:06:22 -07:00
rtmutex_common.h
rtmutex-debug.c
rtmutex-debug.h
rtmutex-tester.c
rtmutex.c
rtmutex.h
rwsem.c
sched_clock.c sched: add optional support for CONFIG_HAVE_UNSTABLE_SCHED_CLOCK 2008-05-05 23:56:18 +02:00
sched_debug.c sched: add optional support for CONFIG_HAVE_UNSTABLE_SCHED_CLOCK 2008-05-05 23:56:18 +02:00
sched_fair.c sched: fix weight calculations 2008-05-08 17:00:42 +02:00
sched_features.h
sched_idletask.c sched: make rt_sched_class, idle_sched_class static 2008-05-05 23:56:17 +02:00
sched_rt.c sched: fix RT task-wakeup logic 2008-05-05 23:56:18 +02:00
sched_stats.h
sched.c cgroups: fix compile warning 2008-05-14 19:11:14 -07:00
seccomp.c
semaphore.c Revert "semaphore: fix" 2008-05-10 20:43:22 -07:00
signal.c signals: add set_restore_sigmask 2008-04-30 08:29:37 -07:00
softirq.c Fix cpu hotplug problem in softirq code 2008-05-01 08:03:58 -07:00
softlockup.c
spinlock.c
srcu.c
stacktrace.c
stop_machine.c
sys_ni.c
sys.c pids: sys_getpgid: fix unsafe *pid usage, s/tasklist/rcu/ 2008-04-30 08:29:49 -07:00
sysctl_check.c
sysctl.c sysctl: add the ->permissions callback on the ctl_table_root 2008-04-29 08:06:23 -07:00
taskstats.c Use find_task_by_vpid in taskstats 2008-04-30 08:29:48 -07:00
test_kprobes.c
time.c Make constants in kernel/timeconst.h fixed 64 bits 2008-05-02 16:18:42 -07:00
timeconst.pl Make constants in kernel/timeconst.h fixed 64 bits 2008-05-02 16:18:42 -07:00
timer.c debugobjects: add timer specific object debugging code 2008-04-30 08:29:53 -07:00
tsacct.c
uid16.c
user_namespace.c
user.c alloc_uid: cleanup 2008-04-30 08:29:53 -07:00
utsname_sysctl.c
utsname.c
wait.c
workqueue.c workqueue: remove redundant function invocation 2008-05-01 08:04:02 -07:00