linux-kernel - Re: [PATCH 16/17] workqueue: Allow cancel_work_sync() and disable_work() from atomic contexts on BH work items

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: Tue, 20 Feb 2024 15:33:34 +0800
From: Lai Jiangshan <jiangshanlai@...il.com>
To: Tejun Heo <tj@...nel.org>
Cc: torvalds@...ux-foundation.org, linux-kernel@...r.kernel.org, 
	allen.lkml@...il.com, kernel-team@...a.com
Subject: Re: [PATCH 16/17] workqueue: Allow cancel_work_sync() and
 disable_work() from atomic contexts on BH work items

Hello, Tejun

On Sat, Feb 17, 2024 at 2:06 AM Tejun Heo <tj@...nel.org> wrote:

> @@ -4072,7 +4070,32 @@ static bool __flush_work(struct work_struct *work, bool from_cancel)
>         if (!pool)
>                 return false;
>
> -       wait_for_completion(&barr.done);
> +       if ((pool->flags & POOL_BH) && from_cancel) {

pool pointer might be invalid here, please check POOL_BH before
rcu_read_unlock()
or move rcu_read_unlock() here, or use "*work_data_bits(work) & WORK_OFFQ_BH".

> +               /*
> +                * We're flushing a BH work item which is being canceled. It
> +                * must have been executing during start_flush_work() and can't
> +                * currently be queued. If @work is still executing, we know it
> +                * is running in the BH context and thus can be busy-waited.
> +                *
> +                * On RT, prevent a live lock when current preempted soft
> +                * interrupt processing or prevents ksoftirqd from running by
> +                * keeping flipping BH. If the tasklet runs on a different CPU
> +                * then this has no effect other than doing the BH
> +                * disable/enable dance for nothing. This is copied from
> +                * kernel/softirq.c::tasklet_unlock_spin_wait().
> +                */

s/tasklet/BH work/g

Although the comment is copied from kernel/softirq.c, but I can't
envision what the scenario is when the current task
"prevents ksoftirqd from running by keeping flipping BH"
since the @work is still executing or the tasklet is running.


> +               while (!try_wait_for_completion(&barr.done)) {
> +                       if (IS_ENABLED(CONFIG_PREEMPT_RT)) {
> +                               local_bh_disable();
> +                               local_bh_enable();
> +                       } else {
> +                               cpu_relax();
> +                       }
> +               }
> +       } else {
> +               wait_for_completion(&barr.done);
> +       }
> +
>         destroy_work_on_stack(&barr.work);
>         return true;
>  }
> @@ -4090,6 +4113,7 @@ static bool __flush_work(struct work_struct *work, bool from_cancel)
>   */
>  bool flush_work(struct work_struct *work)
>  {
> +       might_sleep();
>         return __flush_work(work, false);
>  }
>  EXPORT_SYMBOL_GPL(flush_work);
> @@ -4179,6 +4203,11 @@ static bool __cancel_work_sync(struct work_struct *work, u32 cflags)
>
>         ret = __cancel_work(work, cflags | WORK_CANCEL_DISABLE);
>
> +       if (*work_data_bits(work) & WORK_OFFQ_BH)
> +               WARN_ON_ONCE(in_hardirq());

When !PREEMPT_RT, this check is sufficient.

But when PREEMP_RT, it should be only in the contexts that allow
local_bh_disable() for synching a BH work, although I'm not sure
what check code is proper.

In PREEMPT_RT, local_bh_disable() is disallowed in not only hardirq
context but also !preemptible() context (I'm not sure about it).

__local_bh_disable_ip() (PREEMP_RT version) doesn't contain
full check except for "WARN_ON_ONCE(in_hardirq())" either.

Since the check is just for debugging, I'm OK with the current check.

Thanks
Lai