On 05.02.2026 00:37, Stefano Stabellini wrote:
> --- a/xen/drivers/char/console.c
> +++ b/xen/drivers/char/console.c
> @@ -541,6 +541,12 @@ void console_put_domain(struct domain *d)
> rcu_unlock_domain(d);
> }
>
> +static bool is_focus_domain(const struct domain *d)
> +{
> + ASSERT(rspin_is_locked(&console_lock));
> + return d != NULL && d->domain_id == console_rx - 1;
> +}
Just as a remark: You using only nrspin_*() further down and in patch 3,
using rspin_is_locked() here is slightly irritating. It looks to be
technically correct, but maybe a comment is warranted nevertheless?
> @@ -559,6 +565,7 @@ static void console_switch_input(void)
> {
> nrspin_lock_irqsave(&console_lock, flags);
> ACCESS_ONCE(console_rx) = 0;
> + serial_rx_cons = serial_rx_prod;
I think you want to have the same comment here as you put ...
> @@ -581,6 +588,8 @@ static void console_switch_input(void)
>
> nrspin_lock_irqsave(&console_lock, flags);
> ACCESS_ONCE(console_rx) = next_rx;
> + /* Don't let the next dom read the previous dom's unread data. */
> + serial_rx_cons = serial_rx_prod;
... here.
> @@ -610,7 +619,7 @@ static void __serial_rx(char c)
> unsigned long flags;
>
> /*
> - * Deliver input to the hardware domain buffer, unless it is
> + * Deliver input to the focus domain buffer, unless it is
> * already full.
> */
> nrspin_lock_irqsave(&console_lock, flags);
The conditional ahead of this comment isn't changed here yet, so changing the
comment is premature. (I'm pretty sure I said so before.)
Jan
