From: Josh Triplett <josh@kernel.org>
To: paulmck@linux.vnet.ibm.com
Cc: linux-kernel@vger.kernel.org, mingo@elte.hu, tglx@linuxtronix.de,
dipankar@in.ibm.com, tytso@us.ibm.com, dvhltc@us.ibm.com,
oleg@tv-sign.ru, twoerner.k@gmail.com, billh@gnuppy.monkey.org,
nielsen.esben@googlemail.com, corbet@lwn.net
Subject: Re: [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture
Date: Thu, 25 Jan 2007 00:47:04 -0800 [thread overview]
Message-ID: <45B86E88.6020209@kernel.org> (raw)
In-Reply-To: <20070125022338.GB22540@linux.vnet.ibm.com>
Paul E. McKenney wrote:
> This patch adds an optional preemption kernel thread to the rcutorture
> tests. This thread sets itself to a low RT priority and chews up
> CPU in 10-second bursts, verifying that grace periods progress during
> this 10-second interval. This has thus far passed about 30 hours of
> RCU torture testing on a 4-CPU (a pair of 2-CPU dies) 64-bit Xeon
> system.
>
> I am experimenting with more-vicious tests, but extra viciousness thus
> far comes at the expense of grotesque code.
Overall, the new feature seems like a good idea, and it should exercise the
new RCU boosting code. Some comments below.
One major item: this new test feature really needs a new module parameter to
enable or disable it.
> diff -urpNa -X dontdiff linux-2.6.20-rc4-rt1/kernel/rcutorture.c linux-2.6.20-rc4-rt1-rcubtorture/kernel/rcutorture.c
> --- linux-2.6.20-rc4-rt1/kernel/rcutorture.c 2007-01-09 10:59:54.000000000 -0800
> +++ linux-2.6.20-rc4-rt1-rcubtorture/kernel/rcutorture.c 2007-01-23 11:27:49.000000000 -0800
> @@ -194,6 +194,8 @@ struct rcu_torture_ops {
> int (*completed)(void);
> void (*deferredfree)(struct rcu_torture *p);
> void (*sync)(void);
> + void (*preemptstart)(void);
> + void (*preemptend)(void);
> int (*stats)(char *page);
> char *name;
> };
> @@ -258,6 +260,71 @@ static void rcu_torture_deferred_free(st
> call_rcu(&p->rtort_rcu, rcu_torture_cb);
> }
>
> +#ifndef CONFIG_PREEMPT_RCU_BOOST
> +static void rcu_preempt_start(void) { }
> +static void rcu_preempt_end(void) { }
> +static int rcu_preempt_stats(char *page) { return 0; }
> +#else /* #ifndef CONFIG_PREEMPT_RCU_BOOST */
> +
> +static struct task_struct *rcu_preeempt_task;
> +static long rcu_torture_preempt_errors = 0;
Might as well make this an unsigned long; negative values wouldn't make sense.
> +static int rcu_torture_preempt(void *arg)
> +{
> + int completedstart;
> + time_t gcstart;
> + struct sched_param sp;
> +
> + sp.sched_priority = MAX_RT_PRIO - 1;
> + sched_setscheduler(current, SCHED_RR, &sp);
> + current->flags |= PF_NOFREEZE;
> +
> + do {
> + completedstart = rcu_torture_completed();
> + gcstart = xtime.tv_sec;
> + while ((xtime.tv_sec - gcstart < 10) &&
> + (rcu_torture_completed() == completedstart))
> + cond_resched();
> + if (rcu_torture_completed() == completedstart)
> + rcu_torture_preempt_errors++;
> + schedule_timeout_interruptible(shuffle_interval * HZ);
Why call schedule_timeout_interruptible here without actually handling
interruptions? So that you can send it a signal to cause the shuffle early?
> + } while (!kthread_should_stop());
> + return NULL;
> +}
> +
> +static void rcu_preempt_start(void)
> +{
> + rcu_preeempt_task = kthread_run(rcu_torture_preempt, NULL,
> + "rcu_torture_preempt");
> + if (IS_ERR(rcu_preeempt_task)) {
> + VERBOSE_PRINTK_ERRSTRING("Failed to create preempter");
This ought to include the errno value, PTR_ERR(rcu_preempt_task).
> + rcu_preeempt_task = NULL;
> + }
> +}
> +
> +static void rcu_preempt_end(void)
> +{
> + if (rcu_preeempt_task != NULL) {
if (rcu_preempt_task) would work just as well here.
> + VERBOSE_PRINTK_STRING("Stopping rcu_preempt task");
> + kthread_stop(rcu_preeempt_task);
> + }
> + rcu_preeempt_task = NULL;
> +}
> +
> +static int rcu_preempt_stats(char *page) {
> + int cnt = 0;
> +
> + cnt += sprintf(&page[cnt],
> + "Preemption stalls: %ld\n", rcu_torture_preempt_errors);
> + return (cnt);
> +}
How about just:
return sprintf(page, ...);
?
Also, if you decide to make rcu_torture_preempt_errors an unsigned long as
suggested above, this should use %lu.
> +#endif /* #else #ifndef CONFIG_PREEMPT_RCU_BOOST */
> +
> +static void rcu_preemptstart(void)
> +{
> +
> +}
> +
This looks like a bit of stray code.
> static struct rcu_torture_ops rcu_ops = {
> .init = NULL,
> .cleanup = NULL,
> @@ -267,7 +334,9 @@ static struct rcu_torture_ops rcu_ops =
> .completed = rcu_torture_completed,
> .deferredfree = rcu_torture_deferred_free,
> .sync = synchronize_rcu,
> - .stats = NULL,
> + .preemptstart = rcu_preempt_start,
> + .preemptend = rcu_preempt_end,
> + .stats = rcu_preempt_stats,
> .name = "rcu"
> };
>
> @@ -306,6 +375,8 @@ static struct rcu_torture_ops rcu_sync_o
> .completed = rcu_torture_completed,
> .deferredfree = rcu_sync_torture_deferred_free,
> .sync = synchronize_rcu,
> + .preemptstart = NULL,
> + .preemptend = NULL,
> .stats = NULL,
> .name = "rcu_sync"
> };
Much like other common structures such as struct file_operations, no need to
explicitly specify members as NULL here; any member you don't specify will get
a NULL value. That avoids the need to update every use of this structure
whenever you add a new member used by only some of them.
> @@ -370,6 +441,8 @@ static struct rcu_torture_ops rcu_bh_ops
> .completed = rcu_bh_torture_completed,
> .deferredfree = rcu_bh_torture_deferred_free,
> .sync = rcu_bh_torture_synchronize,
> + .preemptstart = NULL,
> + .preemptend = NULL,
> .stats = NULL,
> .name = "rcu_bh"
> };
Likewise.
> @@ -383,6 +456,8 @@ static struct rcu_torture_ops rcu_bh_syn
> .completed = rcu_bh_torture_completed,
> .deferredfree = rcu_sync_torture_deferred_free,
> .sync = rcu_bh_torture_synchronize,
> + .preemptstart = NULL,
> + .preemptend = NULL,
> .stats = NULL,
> .name = "rcu_bh_sync"
> };
Likewise.
> @@ -464,6 +539,8 @@ static struct rcu_torture_ops srcu_ops =
> .completed = srcu_torture_completed,
> .deferredfree = rcu_sync_torture_deferred_free,
> .sync = srcu_torture_synchronize,
> + .preemptstart = NULL,
> + .preemptend = NULL,
> .stats = srcu_torture_stats,
> .name = "srcu"
> };
Likewise.
> @@ -502,6 +579,8 @@ static struct rcu_torture_ops sched_ops
> .completed = sched_torture_completed,
> .deferredfree = rcu_sync_torture_deferred_free,
> .sync = sched_torture_synchronize,
> + .preemptstart = NULL,
> + .preemptend = NULL,
> .stats = NULL,
> .name = "sched"
Likewise.
> @@ -856,6 +935,8 @@ rcu_torture_cleanup(void)
> kthread_stop(stats_task);
> }
> stats_task = NULL;
> + if (cur_ops->preemptend != NULL)
if (cur_ops->preemptend) would work as well.
> @@ -997,6 +1078,8 @@ rcu_torture_init(void)
> goto unwind;
> }
> }
> + if (cur_ops->preemptstart != NULL)
Likewise.
- Josh Triplett
next prev parent reply other threads:[~2007-01-25 8:47 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-01-25 2:11 [RFC PATCH -rt 0/2] RCU priority boosting that survives semi-vicious testing Paul E. McKenney
2007-01-25 2:14 ` [RFC PATCH -rt 1/2] " Paul E. McKenney
2007-01-25 2:23 ` [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture Paul E. McKenney
2007-01-25 8:47 ` Josh Triplett [this message]
2007-01-25 18:01 ` Paul E. McKenney
2007-01-25 19:06 ` Josh Triplett
2007-01-26 1:52 ` Paul E. McKenney
2007-01-26 6:29 ` Josh Triplett
2007-01-29 2:11 ` Paul E. McKenney
2007-01-29 6:05 ` Josh Triplett
2007-01-25 9:29 ` [RFC PATCH -rt 1/2] RCU priority boosting that survives semi-vicious testing Josh Triplett
2007-01-25 19:58 ` Paul E. McKenney
2007-01-26 1:47 ` Paul E. McKenney
2007-01-26 6:13 ` Josh Triplett
-- strict thread matches above, loose matches on Subject: below --
2007-02-01 1:21 [RFC PATCH -rt 0/2] RCU priority boosting that survives vicious testing Paul E. McKenney
2007-02-01 1:26 ` [RFC PATCH -rt 2/2] RCU priority boosting additions to rcutorture Paul E. McKenney
2007-02-01 2:12 ` Nigel Cunningham
2007-02-01 2:31 ` Paul E. McKenney
2007-02-01 2:42 ` Nigel Cunningham
2007-02-01 5:46 ` Paul E. McKenney
2007-02-01 22:13 ` Nigel Cunningham
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=45B86E88.6020209@kernel.org \
--to=josh@kernel.org \
--cc=billh@gnuppy.monkey.org \
--cc=corbet@lwn.net \
--cc=dipankar@in.ibm.com \
--cc=dvhltc@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=nielsen.esben@googlemail.com \
--cc=oleg@tv-sign.ru \
--cc=paulmck@linux.vnet.ibm.com \
--cc=tglx@linuxtronix.de \
--cc=twoerner.k@gmail.com \
--cc=tytso@us.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.