mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Peter Zijlstra <peterz@infradead.org>
To: Valentin Schneider <valentin.schneider@arm.com>
Cc: syzbot <syzbot+9362b31a2e0cad8b749d@syzkaller.appspotmail.com>,
	bp@alien8.de, dwmw@amazon.co.uk, hpa@zytor.com,
	linux-kernel@vger.kernel.org, luto@kernel.org, mingo@redhat.com,
	syzkaller-bugs@googlegroups.com, tglx@linutronix.de,
	x86@kernel.org
Subject: Re: [syzbot] WARNING in kthread_is_per_cpu
Date: Tue, 20 Apr 2021 10:51:49 +0200	[thread overview]
Message-ID: <YH6WJc825C4P0FCK@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <874kg2kpwd.mognet@arm.com>

On Mon, Apr 19, 2021 at 08:58:26PM +0100, Valentin Schneider wrote:

> Looks about right, IIUC the key being:
> 
>   p->flags & PF_KTHREAD + p->set_child_tid => the struct kthread is
>   persistent
> 
>   p->flags & PF_KTHREAD => you may or may not have a struct kthread (see
>   kernel/umh.c kernel_thread() uses). PF_KTHREAD isn't even guaranteed to
>   persist (begin_new_exec()), which seems to be what the syzbot hit.

Ack, that's nicely put.

> While we're at it, does free_kthread_struct() want the __to_kthread()
> treatment as well? The other to_kthread() callsites looked like they only
> made sense with a "proper" kthread anyway.

I think free_kthread_struct() is ok, because a task at that point in its
lifetime cannot be also doing exec().

kthread_func() is another 'fun' trainwreck waiting to happen -- luckily
the only caller uses current, still let me go fix it.

kthread_probe_data() relies on PF_WQ_WORKER implying PF_KTHREAD but
otherwise seems very fragile too.

Something like so then?

---
Subject: kthread: Fix PF_KTHREAD vs to_kthread() race
From: Peter Zijlstra <peterz@infradead.org>
Date: Tue Apr 20 10:18:17 CEST 2021

The kthread_is_per_cpu() construct relies on only being called on
PF_KTHREAD tasks (per the WARN in to_kthread). This gives rise to the
following usage pattern:

	if ((p->flags & PF_KTHREAD) && kthread_is_per_cpu(p))

However, as reported by syzcaller, this is broken. The scenario is:

	CPU0				CPU1 (running p)

	(p->flags & PF_KTHREAD) // true

					begin_new_exec()
					  me->flags &= ~(PF_KTHREAD|...);
	kthread_is_per_cpu(p)
	  to_kthread(p)
	    WARN(!(p->flags & PF_KTHREAD) <-- *SPLAT*

Introduce __to_kthread() that omits the WARN and is sure to check both
values.

Use this to remove the problematic pattern for kthread_is_per_cpu()
and fix a number of other kthread_*() functions that have similar
issues but are currently not used in ways that would expose the
problem.

Notably kthread_func() is only ever called on 'current', while
kthread_probe_data() is only used for PF_WQ_WORKER, which implies the
task is from kthread_create*().

Fixes: ac687e6e8c26 ("kthread: Extract KTHREAD_IS_PER_CPU")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 kernel/kthread.c    |   33 +++++++++++++++++++++++++++------
 kernel/sched/core.c |    2 +-
 kernel/sched/fair.c |    2 +-
 3 files changed, 29 insertions(+), 8 deletions(-)

--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -84,6 +84,25 @@ static inline struct kthread *to_kthread
 	return (__force void *)k->set_child_tid;
 }
 
+/*
+ * Variant of to_kthread() that doesn't assume @p is a kthread.
+ *
+ * Per construction; when:
+ *
+ *   (p->flags & PF_KTHREAD) && p->set_child_tid
+ *
+ * the task is both a kthread and struct kthread is persistent. However
+ * PF_KTHREAD on it's own is not, kernel_thread() can exec() (See umh.c and
+ * begin_new_exec()).
+ */
+static inline struct kthread *__to_kthread(struct task_struct *p)
+{
+	void *kthread = (__force void *)p->set_child_tid;
+	if (kthread && !(p->flags & PF_KTHREAD))
+		kthread = NULL;
+	return kthread;
+}
+
 void free_kthread_struct(struct task_struct *k)
 {
 	struct kthread *kthread;
@@ -168,8 +187,9 @@ EXPORT_SYMBOL_GPL(kthread_freezable_shou
  */
 void *kthread_func(struct task_struct *task)
 {
-	if (task->flags & PF_KTHREAD)
-		return to_kthread(task)->threadfn;
+	struct kthread *kthread = __to_kthread(task);
+	if (kthread)
+		return kthread->threadfn;
 	return NULL;
 }
 EXPORT_SYMBOL_GPL(kthread_func);
@@ -199,10 +219,11 @@ EXPORT_SYMBOL_GPL(kthread_data);
  */
 void *kthread_probe_data(struct task_struct *task)
 {
-	struct kthread *kthread = to_kthread(task);
+	struct kthread *kthread = __to_kthread(task);
 	void *data = NULL;
 
-	copy_from_kernel_nofault(&data, &kthread->data, sizeof(data));
+	if (kthread)
+		copy_from_kernel_nofault(&data, &kthread->data, sizeof(data));
 	return data;
 }
 
@@ -514,9 +535,9 @@ void kthread_set_per_cpu(struct task_str
 	set_bit(KTHREAD_IS_PER_CPU, &kthread->flags);
 }
 
-bool kthread_is_per_cpu(struct task_struct *k)
+bool kthread_is_per_cpu(struct task_struct *p)
 {
-	struct kthread *kthread = to_kthread(k);
+	struct kthread *kthread = __to_kthread(p);
 	if (!kthread)
 		return false;
 
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -8505,7 +8505,7 @@ static void balance_push(struct rq *rq)
 	 * histerical raisins.
 	 */
 	if (rq->idle == push_task ||
-	    ((push_task->flags & PF_KTHREAD) && kthread_is_per_cpu(push_task)) ||
+	    kthread_is_per_cpu(push_task) ||
 	    is_migration_disabled(push_task)) {
 
 		/*
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7619,7 +7619,7 @@ int can_migrate_task(struct task_struct
 		return 0;
 
 	/* Disregard pcpu kthreads; they are where they need to be. */
-	if ((p->flags & PF_KTHREAD) && kthread_is_per_cpu(p))
+	if (kthread_is_per_cpu(p))
 		return 0;
 
 	if (!cpumask_test_cpu(env->dst_cpu, p->cpus_ptr)) {

  reply	other threads:[~2021-04-20  8:54 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-19 10:36 syzbot
2021-04-19 11:30 ` Thomas Gleixner
2021-04-19 11:31 ` Valentin Schneider
2021-04-19 18:45   ` Peter Zijlstra
2021-04-19 19:58     ` Valentin Schneider
2021-04-20  8:51       ` Peter Zijlstra [this message]
2021-04-20  9:43         ` Valentin Schneider
2021-04-20 10:11           ` Peter Zijlstra
2021-04-20 10:17             ` Valentin Schneider
2021-04-22  7:36         ` [tip: sched/core] kthread: Fix PF_KTHREAD vs to_kthread() race tip-bot2 for 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=YH6WJc825C4P0FCK@hirez.programming.kicks-ass.net \
    --to=peterz@infradead.org \
    --cc=bp@alien8.de \
    --cc=dwmw@amazon.co.uk \
    --cc=hpa@zytor.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=luto@kernel.org \
    --cc=mingo@redhat.com \
    --cc=syzbot+9362b31a2e0cad8b749d@syzkaller.appspotmail.com \
    --cc=syzkaller-bugs@googlegroups.com \
    --cc=tglx@linutronix.de \
    --cc=valentin.schneider@arm.com \
    --cc=x86@kernel.org \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

Powered by JetHome