Re: [PATCH] serial: stm32: optimize spin lock usage

From: Greg KH <gregkh@linuxfoundation.org>
To: dillon min <dillon.minfei@gmail.com>
Cc: jirislaby@kernel.org, Maxime Coquelin <mcoquelin.stm32@gmail.com>,
	Alexandre TORGUE <alexandre.torgue@foss.st.com>,
	linux-serial@vger.kernel.org,
	linux-stm32@st-md-mailman.stormreply.com,
	Linux ARM <linux-arm-kernel@lists.infradead.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] serial: stm32: optimize spin lock usage
Date: Mon, 12 Apr 2021 10:25:39 +0200	[thread overview]
Message-ID: <YHQEA9jn5uXQCtrN@kroah.com> (raw)
In-Reply-To: <CAL9mu0Lt-3_O7V5HLxd5Hbt9afx9ryBUzWqmsc+2n3SP7JS6ig@mail.gmail.com>

On Mon, Apr 12, 2021 at 02:50:20PM +0800, dillon min wrote:
> Hi Greg，
> 
> Thanks for the quick response, please ignore the last private mail.
> 
> On Mon, Apr 12, 2021 at 1:52 PM Greg KH <gregkh@linuxfoundation.org> wrote:
> >
> > On Mon, Apr 12, 2021 at 12:34:21PM +0800, dillon.minfei@gmail.com wrote:
> > > From: dillon min <dillon.minfei@gmail.com>
> > >
> > > To avoid potential deadlock in spin_lock usage, change to use
> > > spin_lock_irqsave(), spin_unlock_irqrestore() in process(thread_fn) context.
> > > spin_lock(), spin_unlock() under handler context.
> > >
> > > remove unused local_irq_save/restore call.
> > >
> > > Signed-off-by: dillon min <dillon.minfei@gmail.com>
> > > ---
> > > Was verified on stm32f469-disco board. need more test on stm32mp platform.
> > >
> > >  drivers/tty/serial/stm32-usart.c | 27 +++++++++++++++++----------
> > >  1 file changed, 17 insertions(+), 10 deletions(-)
> > >
> > > diff --git a/drivers/tty/serial/stm32-usart.c b/drivers/tty/serial/stm32-usart.c
> > > index b3675cf25a69..c4c859b34367 100644
> > > --- a/drivers/tty/serial/stm32-usart.c
> > > +++ b/drivers/tty/serial/stm32-usart.c
> > > @@ -214,7 +214,7 @@ static void stm32_usart_receive_chars(struct uart_port *port, bool threaded)
> > >       struct tty_port *tport = &port->state->port;
> > >       struct stm32_port *stm32_port = to_stm32_port(port);
> > >       const struct stm32_usart_offsets *ofs = &stm32_port->info->ofs;
> > > -     unsigned long c;
> > > +     unsigned long c, flags;
> > >       u32 sr;
> > >       char flag;
> > >
> > > @@ -276,9 +276,17 @@ static void stm32_usart_receive_chars(struct uart_port *port, bool threaded)
> > >               uart_insert_char(port, sr, USART_SR_ORE, c, flag);
> > >       }
> > >
> > > -     spin_unlock(&port->lock);
> > > +     if (threaded)
> > > +             spin_unlock_irqrestore(&port->lock, flags);
> > > +     else
> > > +             spin_unlock(&port->lock);
> >
> > You shouldn't have to check for this, see the other patches on the list
> > recently that fixed this up to not be an issue for irq handlers.
> Can you help to give more hints, or the commit id of the patch which
> fixed this. thanks.
> 
> I'm still confused with this.
> 
> The stm32_usart_threaded_interrupt() is a kthread context, once
> port->lock holds by this function, another serial interrupts raised,
> such as USART_SR_TXE,stm32_usart_interrupt() can't get the lock,
> there will be a deadlock. isn't it?
> 
>  So, shouldn't I use spin_lock{_irqsave} according to the caller's context ?

Please see 81e2073c175b ("genirq: Disable interrupts for force threaded
handlers") for when threaded irq handlers have irqs disabled, isn't that
the case you are trying to "protect" from here?

Why is the "threaded" flag used at all?  The driver should not care.

Also see 9baedb7baeda ("serial: imx: drop workaround for forced irq
threading") in linux-next for an example of how this was fixed up in a
serial driver.

does that help?

thanks,

greg k-h