lkml.org 
[lkml]   [2018]   [Dec]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    Date
    From
    SubjectRe: rcu_preempt caused oom
    On Wed, Dec 12, 2018 at 01:37:40AM +0000, He, Bo wrote:
    > We reproduced the issue panic in hung_task with the patch "Improve diagnostics for failed RCU grace-period start", but unfortunately maybe it's due to the loglevel, the show_rcu_gp_kthreads doesn't print any logs, we will improve the build and run the test to double check.

    Well, at least the diagnostics didn't prevent the problem from happening. ;-)

    Thanx, Paul

    > -----Original Message-----
    > From: Paul E. McKenney <paulmck@linux.ibm.com>
    > Sent: Tuesday, December 11, 2018 12:47 PM
    > To: He, Bo <bo.he@intel.com>
    > Cc: Steven Rostedt <rostedt@goodmis.org>; linux-kernel@vger.kernel.org; josh@joshtriplett.org; mathieu.desnoyers@efficios.com; jiangshanlai@gmail.com; Zhang, Jun <jun.zhang@intel.com>; Xiao, Jin <jin.xiao@intel.com>; Zhang, Yanmin <yanmin.zhang@intel.com>; Bai, Jie A <jie.a.bai@intel.com>
    > Subject: Re: rcu_preempt caused oom
    >
    > On Mon, Dec 10, 2018 at 04:38:38PM -0800, Paul E. McKenney wrote:
    > > On Mon, Dec 10, 2018 at 06:56:18AM +0000, He, Bo wrote:
    > > > Hi,
    > > > We have start the test with the CONFIG_PROVE_RCU=y, and also add one 2s to detect the preempt rcu hang, hope we can get more useful logs tomorrow.
    > > > I also enclosed the config and the debug patches for you review.
    > >
    > > I instead suggest the (lightly tested) debug patch shown below, which
    > > tracks wakeups of RCU's grace-period kthreads and dumps them out if a
    > > given requested grace period fails to start. Again, it is necessary
    > > to build with CONFIG_PROVE_RCU=y, that is, with CONFIG_PROVE_LOCKING=y.
    >
    > Right. This time without commenting out the wakeup as a test of the diagnostic. :-/
    >
    > Please use the patch below instead of the one that I sent in my previous email.
    >
    > Thanx, Paul
    >
    > ------------------------------------------------------------------------
    >
    > commit adfc7dff659495a3433d5084256be59eee0ac6df
    > Author: Paul E. McKenney <paulmck@linux.ibm.com>
    > Date: Mon Dec 10 16:33:59 2018 -0800
    >
    > rcu: Improve diagnostics for failed RCU grace-period start
    >
    > Backported from v4.21/v5.0
    >
    > If a grace period fails to start (for example, because you commented
    > out the last two lines of rcu_accelerate_cbs_unlocked()), rcu_core()
    > will invoke rcu_check_gp_start_stall(), which will notice and complain.
    > However, this complaint is lacking crucial debugging information such
    > as when the last wakeup executed and what the value of ->gp_seq was at
    > that time. This commit therefore removes the current pr_alert() from
    > rcu_check_gp_start_stall(), instead invoking show_rcu_gp_kthreads(),
    > which has been updated to print the needed information, which is collected
    > by rcu_gp_kthread_wake().
    >
    > Signed-off-by: Paul E. McKenney <paulmck@linux.ibm.com>
    >
    > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index 0b760c1369f7..4bcd8753e293 100644
    > --- a/kernel/rcu/tree.c
    > +++ b/kernel/rcu/tree.c
    > @@ -626,25 +626,57 @@ void rcu_sched_force_quiescent_state(void)
    > }
    > EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);
    >
    > +/*
    > + * Convert a ->gp_state value to a character string.
    > + */
    > +static const char *gp_state_getname(short gs) {
    > + if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names))
    > + return "???";
    > + return gp_state_names[gs];
    > +}
    > +
    > +/*
    > + * Return the root node of the specified rcu_state structure.
    > + */
    > +static struct rcu_node *rcu_get_root(struct rcu_state *rsp) {
    > + return &rsp->node[0];
    > +}
    > +
    > /*
    > * Show the state of the grace-period kthreads.
    > */
    > void show_rcu_gp_kthreads(void)
    > {
    > int cpu;
    > + unsigned long j;
    > + unsigned long ja;
    > + unsigned long jr;
    > + unsigned long jw;
    > struct rcu_data *rdp;
    > struct rcu_node *rnp;
    > struct rcu_state *rsp;
    >
    > + j = jiffies;
    > for_each_rcu_flavor(rsp) {
    > - pr_info("%s: wait state: %d ->state: %#lx\n",
    > - rsp->name, rsp->gp_state, rsp->gp_kthread->state);
    > + ja = j - READ_ONCE(rsp->gp_activity);
    > + jr = j - READ_ONCE(rsp->gp_req_activity);
    > + jw = j - READ_ONCE(rsp->gp_wake_time);
    > + pr_info("%s: wait state: %s(%d) ->state: %#lx delta ->gp_activity %lu ->gp_req_activity %lu ->gp_wake_time %lu ->gp_wake_seq %ld ->gp_seq %ld ->gp_seq_needed %ld ->gp_flags %#x\n",
    > + rsp->name, gp_state_getname(rsp->gp_state),
    > + rsp->gp_state,
    > + rsp->gp_kthread ? rsp->gp_kthread->state : 0x1ffffL,
    > + ja, jr, jw, (long)READ_ONCE(rsp->gp_wake_seq),
    > + (long)READ_ONCE(rsp->gp_seq),
    > + (long)READ_ONCE(rcu_get_root(rsp)->gp_seq_needed),
    > + READ_ONCE(rsp->gp_flags));
    > rcu_for_each_node_breadth_first(rsp, rnp) {
    > if (ULONG_CMP_GE(rsp->gp_seq, rnp->gp_seq_needed))
    > continue;
    > - pr_info("\trcu_node %d:%d ->gp_seq %lu ->gp_seq_needed %lu\n",
    > - rnp->grplo, rnp->grphi, rnp->gp_seq,
    > - rnp->gp_seq_needed);
    > + pr_info("\trcu_node %d:%d ->gp_seq %ld ->gp_seq_needed %ld\n",
    > + rnp->grplo, rnp->grphi, (long)rnp->gp_seq,
    > + (long)rnp->gp_seq_needed);
    > if (!rcu_is_leaf_node(rnp))
    > continue;
    > for_each_leaf_node_possible_cpu(rnp, cpu) { @@ -653,8 +685,8 @@ void show_rcu_gp_kthreads(void)
    > ULONG_CMP_GE(rsp->gp_seq,
    > rdp->gp_seq_needed))
    > continue;
    > - pr_info("\tcpu %d ->gp_seq_needed %lu\n",
    > - cpu, rdp->gp_seq_needed);
    > + pr_info("\tcpu %d ->gp_seq_needed %ld\n",
    > + cpu, (long)rdp->gp_seq_needed);
    > }
    > }
    > /* sched_show_task(rsp->gp_kthread); */ @@ -690,14 +722,6 @@ void rcutorture_get_gp_data(enum rcutorture_type test_type, int *flags, } EXPORT_SYMBOL_GPL(rcutorture_get_gp_data);
    >
    > -/*
    > - * Return the root node of the specified rcu_state structure.
    > - */
    > -static struct rcu_node *rcu_get_root(struct rcu_state *rsp) -{
    > - return &rsp->node[0];
    > -}
    > -
    > /*
    > * Enter an RCU extended quiescent state, which can be either the
    > * idle loop or adaptive-tickless usermode execution.
    > @@ -1285,16 +1309,6 @@ static void record_gp_stall_check_time(struct rcu_state *rsp)
    > rsp->n_force_qs_gpstart = READ_ONCE(rsp->n_force_qs); }
    >
    > -/*
    > - * Convert a ->gp_state value to a character string.
    > - */
    > -static const char *gp_state_getname(short gs) -{
    > - if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names))
    > - return "???";
    > - return gp_state_names[gs];
    > -}
    > -
    > /*
    > * Complain about starvation of grace-period kthread.
    > */
    > @@ -1693,7 +1707,8 @@ static bool rcu_future_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
    > * Don't do a self-awaken, and don't bother awakening when there is
    > * nothing for the grace-period kthread to do (as in several CPUs
    > * raced to awaken, and we lost), and finally don't try to awaken
    > - * a kthread that has not yet been created.
    > + * a kthread that has not yet been created. If all those checks are
    > + * passed, track some debug information and awaken.
    > */
    > static void rcu_gp_kthread_wake(struct rcu_state *rsp) { @@ -1701,6 +1716,8 @@ static void rcu_gp_kthread_wake(struct rcu_state *rsp)
    > !READ_ONCE(rsp->gp_flags) ||
    > !rsp->gp_kthread)
    > return;
    > + WRITE_ONCE(rsp->gp_wake_time, jiffies);
    > + WRITE_ONCE(rsp->gp_wake_seq, READ_ONCE(rsp->gp_seq));
    > swake_up_one(&rsp->gp_wq);
    > }
    >
    > @@ -2802,16 +2819,11 @@ rcu_check_gp_start_stall(struct rcu_state *rsp, struct rcu_node *rnp,
    > raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
    > return;
    > }
    > - pr_alert("%s: g%ld->%ld gar:%lu ga:%lu f%#x gs:%d %s->state:%#lx\n",
    > - __func__, (long)READ_ONCE(rsp->gp_seq),
    > - (long)READ_ONCE(rnp_root->gp_seq_needed),
    > - j - rsp->gp_req_activity, j - rsp->gp_activity,
    > - rsp->gp_flags, rsp->gp_state, rsp->name,
    > - rsp->gp_kthread ? rsp->gp_kthread->state : 0x1ffffL);
    > WARN_ON(1);
    > if (rnp_root != rnp)
    > raw_spin_unlock_rcu_node(rnp_root);
    > raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
    > + show_rcu_gp_kthreads();
    > }
    >
    > /*
    > diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index 4e74df768c57..0e051d9b5f1a 100644
    > --- a/kernel/rcu/tree.h
    > +++ b/kernel/rcu/tree.h
    > @@ -327,6 +327,8 @@ struct rcu_state {
    > struct swait_queue_head gp_wq; /* Where GP task waits. */
    > short gp_flags; /* Commands for GP task. */
    > short gp_state; /* GP kthread sleep state. */
    > + unsigned long gp_wake_time; /* Last GP kthread wake. */
    > + unsigned long gp_wake_seq; /* ->gp_seq at ^^^. */
    >
    > /* End of fields guarded by root rcu_node's lock. */
    >
    >


    \
     
     \ /
      Last update: 2018-12-12 03:25    [W:3.952 / U:0.240 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site