On Wed, Jan 30, 2019 at 12:48:42PM +0000, Li,Rongqing wrote:
>
>
> > -----邮件原件-----
> > 发件人: linux-kernel-ow...@vger.kernel.org
> > [mailto:linux-kernel-ow...@vger.kernel.org] 代表 Greg KH
> > 发送时间: 2019年1月30日 18:19
> > 收件人: Li,Rongqing <lirongq...@baidu.com>
> > 抄送: jsl...@suse.com; linux-kernel@vger.kernel.org; gko...@codeaurora.org
> > 主题: Re: [PATCH][v4] tty: fix race between flush_to_ldisc and tty_open
> >
> > On Fri, Jan 18, 2019 at 05:27:17PM +0800, Li RongQing wrote:
> > > There still is a race window after the commit b027e2298bd588
> > > ("tty: fix data race between tty_init_dev and flush of buf"), and we
> > > encountered this crash issue if receive_buf call comes before tty
> > > initialization completes in n_tty_open and
> > > tty->driver_data may be NULL.
> > >
> > > CPU0 CPU1
> > > ---- ----
> > > n_tty_open
> > > tty_init_dev
> > > tty_ldisc_unlock
> > > schedule flush_to_ldisc
> > > receive_buf
> > > tty_port_default_receive_buf
> > > tty_ldisc_receive_buf
> > > n_tty_receive_buf_common
> > > __receive_buf
> > > uart_flush_chars
> > > uart_start
> > > /*tty->driver_data is NULL*/
> > > tty->ops->open
> > > /*init tty->driver_data*/
> > >
> > > it can be fixed by extending ldisc semaphore lock in tty_init_dev to
> > > driver_data initialized completely after tty->ops->open(), but this
> > > will lead to put lock on one function and unlock in some other
> > > function, and hard to maintain, so fix this race only by checking
> > > tty->driver_data when receiving, and return if tty->driver_data
> > > is NULL
> > >
> > > Signed-off-by: Wang Li <wangl...@baidu.com>
> > > Signed-off-by: Zhang Yu <zhangy...@baidu.com>
> > > Signed-off-by: Li RongQing <lirongq...@baidu.com>
> > > ---
> > > V4: add version information
> > > V3: not used ldisc semaphore lock, only checking tty->driver_data with
> > > NULL
> > > V2: fix building error by EXPORT_SYMBOL tty_ldisc_unlock
> > > V1: extend ldisc lock to protect that tty->driver_data is inited
> > >
> > > drivers/tty/tty_port.c | 3 +++
> > > 1 file changed, 3 insertions(+)
> > >
> > > diff --git a/drivers/tty/tty_port.c b/drivers/tty/tty_port.c index
> > > 044c3cbdcfa4..86d0bec38322 100644
> > > --- a/drivers/tty/tty_port.c
> > > +++ b/drivers/tty/tty_port.c
> > > @@ -31,6 +31,9 @@ static int tty_port_default_receive_buf(struct tty_port
> > *port,
> > > if (!tty)
> > > return 0;
> > >
> > > + if (!tty->driver_data)
> > > + return 0;
> > > +
> >
> > How is this working? What is setting driver_data to NULL to "stop" this
> > race?
> >
>
>
> if tty->driver_data is NULL and return, tty_port_default_receive_buf will
> not step to
> uart_start which access tty->driver_data and trigger panic before tty_open,
> so it can
> fix the system panic
>
> > There's no requirement that a tty driver set this field to NULL when it is
> > "done"
> > with the tty device, so I think you are just getting lucky in that your
> > specific
> > driver happens to be doing this.
> >
>
> when tty_open is running, tty is allocated by kzalloc in tty_init_dev which
> called
> by tty_open_by_driver, tty is inited to 0
>
> > What driver are you testing this against?
> >
>
> 8250
Ok, as this is specific to the uart core, how about this patch instead:
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 5c01bb6d1c24..b56a6250df3f 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -130,6 +130,9 @@ static void uart_start(struct tty_struct *tty)
struct uart_port *port;
unsigned long flags;
+ if (!state)
+ return;
+
port = uart_port_lock(state, flags);
__uart_start(tty);
uart_port_unlock(port, flags);
