2022-05-28 08:14:53

by John Ogness

Subject: [PATCH] Revert "printk: wake up all waiters"

This reverts commit 938ba4084abcf6fdd21d9078513c52f8fb9b00d0.

The wait queue @log_wait never has exclusive waiters, so there
is no need to use wake_up_interruptible_all(). Using
wake_up_interruptible() was the correct function to wake all
waiters.

Since there are no exclusive waiters, erroneously changing
wake_up_interruptible() to wake_up_interruptible_all() did not
result in any behavior change. However, using
wake_up_interruptible_all() on a wait queue without exclusive
waiters is fundamentally wrong.

Go back to using wake_up_interruptible() to wake all waiters.

Signed-off-by: John Ogness <[email protected]>
---
kernel/printk/printk.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index a3e1035929b0..ea3dd55709e7 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -3904,7 +3904,7 @@ static void wake_up_klogd_work_func(struct irq_work *irq_work)
}

if (pending & PRINTK_PENDING_WAKEUP)
- wake_up_interruptible_all(&log_wait);
+ wake_up_interruptible(&log_wait);
}

static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) =
--
2.30.2

2022-05-28 19:45:54

by Steven Rostedt

Subject: Re: [PATCH] Revert "printk: wake up all waiters"

On Thu, 26 May 2022 22:36:56 +0206
John Ogness <[email protected]> wrote:

> This reverts commit 938ba4084abcf6fdd21d9078513c52f8fb9b00d0.
>
> The wait queue @log_wait never has exclusive waiters, so there
> is no need to use wake_up_interruptible_all(). Using
> wake_up_interruptible() was the correct function to wake all
> waiters.
>
> Since there are no exclusive waiters, erroneously changing
> wake_up_interruptible() to wake_up_interruptible_all() did not
> result in any behavior change. However, using
> wake_up_interruptible_all() on a wait queue without exclusive
> waiters is fundamentally wrong.
>
> Go back to using wake_up_interruptible() to wake all waiters.

(Just to show that I have been paying attention ;-)

Acked-by: Steven Rostedt (Google) <[email protected]>

-- Steve

>
> Signed-off-by: John Ogness <[email protected]>
> ---
> kernel/printk/printk.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index a3e1035929b0..ea3dd55709e7 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -3904,7 +3904,7 @@ static void wake_up_klogd_work_func(struct irq_work *irq_work)
> }
>
> if (pending & PRINTK_PENDING_WAKEUP)
> - wake_up_interruptible_all(&log_wait);
> + wake_up_interruptible(&log_wait);
> }
>
> static DEFINE_PER_CPU(struct irq_work, wake_up_klogd_work) =

2022-05-28 20:04:33

by Petr Mladek

Subject: Re: [PATCH] Revert "printk: wake up all waiters"

On Thu 2022-05-26 22:36:56, John Ogness wrote:
> This reverts commit 938ba4084abcf6fdd21d9078513c52f8fb9b00d0.
>
> The wait queue @log_wait never has exclusive waiters, so there
> is no need to use wake_up_interruptible_all(). Using
> wake_up_interruptible() was the correct function to wake all
> waiters.
>
> Since there are no exclusive waiters, erroneously changing
> wake_up_interruptible() to wake_up_interruptible_all() did not
> result in any behavior change. However, using
> wake_up_interruptible_all() on a wait queue without exclusive
> waiters is fundamentally wrong.
>
> Go back to using wake_up_interruptible() to wake all waiters.
>
> Signed-off-by: John Ogness <[email protected]>

The patch has been committed into printk/linux.git, branch
rework/printk.

I am going to create the pull request the following week
after it gets at least one spin in linux-next.

It is trivial patch so that it might be sent directly.
But it does not affect the functionality and is is not
urgent either.

Best Regards,
Petr