linux-kernel - Re: [PATCH 1/1, v6] cgroup/freezer: add per freezer duty ratio control

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 10 Feb 2011 11:15:22 +0200
From:	"Kirill A. Shutemov" <kirill@...temov.name>
To:	Matt Helsley <matthltc@...ibm.com>
Cc:	jacob.jun.pan@...ux.intel.com, LKML <linux-kernel@...r.kernel.org>,
	Arjan van de Ven <arjan@...ux.intel.com>,
	container cgroup <containers@...ts.linux-foundation.org>,
	Li Zefan <lizf@...fujitsu.com>,
	Paul Menage <menage@...gle.com>, akpm@...ux-foundation.org,
	rdunlap@...otime.net, Cedric Le Goater <clg@...t.ibm.com>,
	Wysocki Rafael Wysocki <rjw@...k.pl>
Subject: Re: [PATCH 1/1, v6] cgroup/freezer: add per freezer duty ratio
 control

On Wed, Feb 09, 2011 at 07:04:42PM -0800, Matt Helsley wrote:
> > +{
> > +	struct cgroup *cgroup = (struct cgroup *)data;
> > +	struct freezer *freezer = cgroup_freezer(cgroup);
> > +
> > +	do {
> > +		if (freezer->duty.ratio < 100 && freezer->duty.ratio > 0 &&
> > +			freezer->duty.period_pct_ms) {
> > +			if (try_to_freeze_cgroup(cgroup, freezer))
> > +				pr_info("cannot freeze\n");
> > +			msleep(freezer->duty.period_pct_ms *
> > +				freezer->duty.ratio);
> > +			unfreeze_cgroup(cgroup, freezer);
> > +			msleep(freezer->duty.period_pct_ms *
> > +				(100 - freezer->duty.ratio));
> > +		} else {
> > +			sleep_on(&freezer_wait);
> > +			pr_debug("freezer thread wake up\n");
> > +		}
> > +	} while (!kthread_should_stop());
> > +	return 0;
> > +}
> 
> Seems to me you could avoid the thread-per-cgroup overhead and the
> sleep-loop code by using one timer-per-cgroup. When the timer expires
> you freeze/thaw the cgroup associated with the timer, setup the next
> wakeup timer, and use only one kernel thread to do it all. If you
> use workqueues you might even avoid the single kernel thread.
> 
> Seems to me like that'd be a good fit for embedded devices.

I proposed to use delayed workqueues (schedule_delayed_work()).

> > +#define FREEZER_KH_PREFIX  "freezer_"
> > +static int freezer_write_param(struct cgroup *cgroup, struct cftype *cft,
> > +		u64 val)
> > +{
> > +	struct freezer *freezer;
> > +	char thread_name[32];
> > +	int ret = 0;
> > +
> > +	freezer = cgroup_freezer(cgroup);
> > +
> > +	if (!cgroup_lock_live_group(cgroup))
> > +		return -ENODEV;
> > +
> > +	switch (cft->private) {
> > +	case FREEZER_DUTY_RATIO:
> > +		if (val >= 100 || val < 0) {
> > +			ret = -EINVAL;
> > +			goto exit;
> > +		}
> > +		freezer->duty.ratio = val;
> 
> Why can't val == 100? At that point it's always THAWED and no kernel
> thread is necessary (just like at 0 it's always FROZEN and no kernel
> thread is necessary).

val == 100 is interface abuse, I think. I just turn off the feature, if
you want.
> >  static struct cftype files[] = {
> >  	{
> >  		.name = "state",
> >  		.read_seq_string = freezer_read,
> >  		.write_string = freezer_write,
> 
> It's not clear what should happen when userspace writes the state
> file after writing a duty_ratio_pct.

It should return -EBUSY, I think.

> >  	},
> > +	{
> > +		.name = "duty_ratio_pct",
> > +		.private = FREEZER_DUTY_RATIO,
> > +		.read_u64 = freezer_read_duty_ratio,
> > +		.write_u64 = freezer_write_param,
> > +	},
> 
> nit: Why use a u64 for a value that can only be 0-100? (or perhaps
> 0-1000 if you wanted sub-1% granularity...)

.read_u64/.write_64 is a standard cgroup's interface.

-- 
 Kirill A. Shutemov
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/