LKML Archive on lore.kernel.org
 help / color / Atom feed
From: Alessio Balsini <balsini@android.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: mingo@kernel.org, juri.lelli@redhat.com,
	linux-kernel@vger.kernel.org, dietmar.eggemann@arm.com,
	luca.abeni@santannapisa.it, bristot@redhat.com,
	dvyukov@google.com, tglx@linutronix.de, vpillai@digitalocean.com,
	rostedt@goodmis.org, kernel-team@android.com
Subject: Re: [RFC][PATCH 01/13] sched/deadline: Impose global limits on sched_attr::sched_period
Date: Sat, 31 Aug 2019 15:41:17 +0100
Message-ID: <20190831144117.GA133727@google.com> (raw)
In-Reply-To: <20190822165125.GW2369@hirez.programming.kicks-ass.net>

Right!

Verified that sysctl_sched_dl_period_max and sysctl_sched_dl_period_min values
are now always consistent.

I spent some time in trying to figure out if not having any mutex in
__checkparam_dl() is safe. There can surely happen that "max < min", e.g.:

          |              |               periods
User1     | User2        | checkparam_dl()  | sysctl_sched_dl_*
----------|--------------|------------------|-------------------
          |              |                  | [x, x]
p_min = 5 |              |                  |
          |              |                  | [5, x]
p_max = 5 |              |                  |
          |              |                  | [5, 5]
          | setattr(p=8) |                  |
          |              | p = 8            |
          |              | [x, 5]           |
p_max = 9 |              |                  |
          |              |                  | [5, 9]
p_min = 6 |              |                  |
          |              |                  | [6, 9]
          |              | [6, 5]           |
----------|--------------|------------------|-------------------

Sharing my thoughts, a "BUG_ON(max < min)" in __checkparam_dl() is then a
guaranteed source of explosions, but the good news is that "if (period < min ||
period > max" in __checkparam_dl() surely fails if "max < min".  Also the fact
that, when we are writing the new sysctl_sched_dl_* values, only one is
actually changed at a time, that surely helps to preserve the consistency.

But is that enough?

Well, I couldn't find any counterexample to make __checkparam_dl() pass with
wrong parameters. And the tests I made are happy.

On Thu, Aug 22, 2019 at 06:51:25PM +0200, Peter Zijlstra wrote:
>  include/linux/sched/sysctl.h |    7 +++++
>  kernel/sched/deadline.c      |   58 +++++++++++++++++++++++++++++++++++++++++--
>  kernel/sysctl.c              |   14 ++++++++++
>  3 files changed, 77 insertions(+), 2 deletions(-)
> 
> --- a/kernel/sched/deadline.c
> +++ b/kernel/sched/deadline.c
> +int sched_dl_period_handler(struct ctl_table *table, int write,
> +			    void __user *buffer, size_t *lenp,
> +			    loff_t *ppos)
> +{
> +		if (min > 1ULL << DL_SCALE && max > min) {

s/>/>=/

> +			WRITE_ONCE(sysctl_sched_dl_period_max, new_max);
> +			WRITE_ONCE(sysctl_sched_dl_period_min, new_min);

Besides the inline comment above, this is my ack to your patch.

Otherwise, here follows a slightly more convoluted version of your patch with a
couple of changes to sched_dl_period_handler(), summarized as:
 - handle new_table only if writing;
 - directly compare the us min and max (saves one multiplication);
 - atomic-writes only the sysctl_sched_dl_period_XXX which changed.

M2c.

Thanks!
-Alessio

---
From cb4481233b57e42ff9dd315811f7919168a28162 Mon Sep 17 00:00:00 2001
From: Peter Zijlstra <peterz@infradead.org>
Date: Thu, 22 Aug 2019 18:51:25 +0200
Subject: [PATCH] sched/deadline: Impose global limits on
 sched_attr::sched_period

There are two DoS scenarios with SCHED_DEADLINE related to
sched_attr::sched_period:

 - since access-control only looks at utilization and density, a very
   large period can allow a very large runtime, which in turn can
   incur a very large latency to lower priority tasks.

 - for very short periods we can end up spending more time programming
   the hardware timer than actually running the task.

Mitigate these by imposing limits on the period.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Alessio Balsini <balsini@android.com>
Cc: rostedt@goodmis.org
Cc: mingo@kernel.org
Cc: luca.abeni@santannapisa.it
Cc: bristot@redhat.com
Cc: vpillai@digitalocean.com
Cc: kernel-team@android.com
Cc: juri.lelli@redhat.com
Cc: dietmar.eggemann@arm.com
Cc: dvyukov@google.com
Cc: tglx@linutronix.de
---
 include/linux/sched/sysctl.h |  7 ++++
 kernel/sched/deadline.c      | 68 ++++++++++++++++++++++++++++++++++--
 kernel/sysctl.c              | 14 ++++++++
 3 files changed, 87 insertions(+), 2 deletions(-)

diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index d4f6215ee03f7..7c8ef07e52133 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -56,6 +56,13 @@ int sched_proc_update_handler(struct ctl_table *table, int write,
 extern unsigned int sysctl_sched_rt_period;
 extern int sysctl_sched_rt_runtime;
 
+extern unsigned int sysctl_sched_dl_period_max;
+extern unsigned int sysctl_sched_dl_period_min;
+
+extern int sched_dl_period_handler(struct ctl_table *table, int write,
+		void __user *buffer, size_t *lenp,
+		loff_t *ppos);
+
 #ifdef CONFIG_UCLAMP_TASK
 extern unsigned int sysctl_sched_uclamp_util_min;
 extern unsigned int sysctl_sched_uclamp_util_max;
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 0b9cbfb2b1d4f..c4a6107e055c7 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -2640,6 +2640,59 @@ void __getparam_dl(struct task_struct *p, struct sched_attr *attr)
 	attr->sched_flags = dl_se->flags;
 }
 
+/*
+ * Default limits for DL period: on the top end we guard against small util
+ * tasks still getting ridiculous long effective runtimes, on the bottom end we
+ * guard against timer DoS.
+ */
+unsigned int sysctl_sched_dl_period_max = 1 << 22; /* ~4.2 seconds */
+unsigned int sysctl_sched_dl_period_min = 100;     /* 100 us */
+
+int sched_dl_period_handler(struct ctl_table *table, int write,
+			    void __user *buffer, size_t *lenp,
+			    loff_t *ppos)
+{
+	static DEFINE_MUTEX(mutex);
+	int ret;
+
+	mutex_lock(&mutex);
+	if (write) {
+		/*
+		 * Use a temporary data structure to store the value read from
+		 * userspace. The sysctl_sched_dl_period_{max,min} value will
+		 * be updated only if the data is consistent.
+		 */
+		struct ctl_table new_table = *table;
+		unsigned int max, min;
+
+		if (new_table.data == &sysctl_sched_dl_period_max) {
+			new_table.data = &max;
+			min = sysctl_sched_dl_period_min;
+		} else {
+			new_table.data = &min;
+			max = sysctl_sched_dl_period_max;
+		}
+
+		ret = proc_douintvec(&new_table, write, buffer, lenp, ppos);
+		if (!ret) {
+			if (min > max ||
+			    (u64)min * NSEC_PER_USEC < 1ULL << DL_SCALE) {
+				ret = -EINVAL;
+			} else {
+				unsigned int *src = new_table.data;
+				unsigned int *dst = table->data;
+
+				WRITE_ONCE(*dst, *src);
+			}
+		}
+	} else {
+		ret = proc_douintvec(table, write, buffer, lenp, ppos);
+	}
+	mutex_unlock(&mutex);
+
+	return ret;
+}
+
 /*
  * This function validates the new parameters of a -deadline task.
  * We ask for the deadline not being zero, and greater or equal
@@ -2652,6 +2705,8 @@ void __getparam_dl(struct task_struct *p, struct sched_attr *attr)
  */
 bool __checkparam_dl(const struct sched_attr *attr)
 {
+	u64 period, max, min;
+
 	/* special dl tasks don't actually use any parameter */
 	if (attr->sched_flags & SCHED_FLAG_SUGOV)
 		return true;
@@ -2675,12 +2730,21 @@ bool __checkparam_dl(const struct sched_attr *attr)
 	    attr->sched_period & (1ULL << 63))
 		return false;
 
+	period = attr->sched_period;
+	if (!period)
+		period = attr->sched_deadline;
+
 	/* runtime <= deadline <= period (if period != 0) */
-	if ((attr->sched_period != 0 &&
-	     attr->sched_period < attr->sched_deadline) ||
+	if (period < attr->sched_deadline ||
 	    attr->sched_deadline < attr->sched_runtime)
 		return false;
 
+	max = (u64)READ_ONCE(sysctl_sched_dl_period_max) * NSEC_PER_USEC;
+	min = (u64)READ_ONCE(sysctl_sched_dl_period_min) * NSEC_PER_USEC;
+
+	if (period < min || period > max)
+		return false;
+
 	return true;
 }
 
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 078950d9605ba..0d07e4707e9d2 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -442,6 +442,20 @@ static struct ctl_table kern_table[] = {
 		.mode		= 0644,
 		.proc_handler	= sched_rt_handler,
 	},
+	{
+		.procname	= "sched_deadline_period_max_us",
+		.data		= &sysctl_sched_dl_period_max,
+		.maxlen		= sizeof(unsigned int),
+		.mode		= 0644,
+		.proc_handler	= sched_dl_period_handler,
+	},
+	{
+		.procname	= "sched_deadline_period_min_us",
+		.data		= &sysctl_sched_dl_period_min,
+		.maxlen		= sizeof(unsigned int),
+		.mode		= 0644,
+		.proc_handler	= sched_dl_period_handler,
+	},
 	{
 		.procname	= "sched_rr_timeslice_ms",
 		.data		= &sysctl_sched_rr_timeslice,
-- 
2.23.0.187.g17f5b7556c-goog



  reply index

Thread overview: 63+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-07-26 14:54 [RFC][PATCH 00/13] SCHED_DEADLINE server infrastructure Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 01/13] sched/deadline: Impose global limits on sched_attr::sched_period Peter Zijlstra
2019-07-29  8:57   ` Juri Lelli
2019-07-29 11:45     ` Daniel Bristot de Oliveira
2019-08-02 17:21   ` Alessio Balsini
2019-08-05 11:53     ` Peter Zijlstra
2019-08-22 12:29       ` Alessio Balsini
2019-08-22 16:51         ` Peter Zijlstra
2019-08-31 14:41           ` Alessio Balsini [this message]
2019-09-02  9:16             ` Peter Zijlstra
2019-09-02 12:31               ` Peter Zijlstra
2019-09-04 10:16               ` Steven Rostedt
2019-09-04 11:30                 ` Peter Zijlstra
2019-09-04 13:24                   ` Joel Fernandes
2019-09-04 14:11                     ` Will Deacon
2019-09-04 14:35                       ` Joel Fernandes
2019-09-04 15:52                     ` Peter Zijlstra
2019-10-23 17:17       ` [PATCH 4.4 4.9 4.14] loop: Add LOOP_SET_DIRECT_IO to compat ioctl Alessio Balsini
2019-10-23 17:22         ` Alessio Balsini
2019-10-25  0:17         ` Sasha Levin
2020-05-20 18:38   ` [RFC][PATCH 01/13] sched/deadline: Impose global limits on sched_attr::sched_period Juri Lelli
2020-05-21 13:45     ` Daniel Bristot de Oliveira
2020-06-16 12:21   ` [tip: sched/core] " tip-bot2 for Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 02/13] stop_machine: Fix stop_cpus_in_progress ordering Peter Zijlstra
2019-07-30 13:16   ` Phil Auld
2019-07-30 13:22   ` Steven Rostedt
2019-07-26 14:54 ` [RFC][PATCH 03/13] sched: Fix kerneldoc comment for ia64_set_curr_task Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 04/13] sched/{rt,deadline}: Fix set_next_task vs pick_next_task Peter Zijlstra
2019-07-29  9:25   ` Juri Lelli
2019-07-29 11:15     ` Peter Zijlstra
2019-07-29 11:27       ` Juri Lelli
2019-07-29 13:04         ` Peter Zijlstra
2019-07-29 13:17           ` Juri Lelli
2019-07-29 14:40             ` Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 05/13] sched: Add task_struct pointer to sched_class::set_curr_task Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 06/13] sched/fair: Export newidle_balance() Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 07/13] sched: Allow put_prev_task() to drop rq->lock Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 08/13] sched: Rework pick_next_task() slow-path Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 09/13] sched: Unify runtime accounting across classes Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 10/13] sched/deadline: Collect sched_dl_entity initialization Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 11/13] sched/deadline: Move bandwidth accounting into {en,de}queue_dl_entity Peter Zijlstra
2019-07-26 14:54 ` [RFC][PATCH 12/13] sched/deadline: Introduce deadline servers Peter Zijlstra
2019-08-07 16:31   ` Dietmar Eggemann
2019-08-08  6:52     ` Juri Lelli
2019-08-08  7:52       ` Dietmar Eggemann
2019-08-08  7:56     ` Peter Zijlstra
2019-08-08  8:11       ` Dietmar Eggemann
2019-08-08  8:46         ` Juri Lelli
2019-08-08  8:57           ` Dietmar Eggemann
2019-08-08  9:27             ` Juri Lelli
2019-08-08  9:45               ` Juri Lelli
2019-08-30 11:24                 ` Peter Zijlstra
2019-09-06  9:36                   ` Juri Lelli
2019-08-08 10:31           ` Peter Zijlstra
2019-08-09  7:13             ` Juri Lelli
2019-08-08  6:59   ` Juri Lelli
2019-08-09  9:17   ` Dietmar Eggemann
2019-08-09 12:16     ` Juri Lelli
2019-07-26 14:54 ` [RFC][PATCH 13/13] sched/fair: Add trivial fair server Peter Zijlstra
2019-07-26 20:01 ` [RFC][PATCH 00/13] SCHED_DEADLINE server infrastructure luca abeni
2019-09-03 14:27 ` Alessio Balsini
2019-09-04 10:50   ` Juri Lelli
2019-09-04 11:32     ` Peter Zijlstra

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=20190831144117.GA133727@google.com \
    --to=balsini@android.com \
    --cc=bristot@redhat.com \
    --cc=dietmar.eggemann@arm.com \
    --cc=dvyukov@google.com \
    --cc=juri.lelli@redhat.com \
    --cc=kernel-team@android.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=luca.abeni@santannapisa.it \
    --cc=mingo@kernel.org \
    --cc=peterz@infradead.org \
    --cc=rostedt@goodmis.org \
    --cc=tglx@linutronix.de \
    --cc=vpillai@digitalocean.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

LKML Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/lkml/0 lkml/git/0.git
	git clone --mirror https://lore.kernel.org/lkml/1 lkml/git/1.git
	git clone --mirror https://lore.kernel.org/lkml/2 lkml/git/2.git
	git clone --mirror https://lore.kernel.org/lkml/3 lkml/git/3.git
	git clone --mirror https://lore.kernel.org/lkml/4 lkml/git/4.git
	git clone --mirror https://lore.kernel.org/lkml/5 lkml/git/5.git
	git clone --mirror https://lore.kernel.org/lkml/6 lkml/git/6.git
	git clone --mirror https://lore.kernel.org/lkml/7 lkml/git/7.git
	git clone --mirror https://lore.kernel.org/lkml/8 lkml/git/8.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 lkml lkml/ https://lore.kernel.org/lkml \
		linux-kernel@vger.kernel.org
	public-inbox-index lkml

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-kernel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git