[PATCH] tty: always relink the port

[PATCH] tty: always relink the port

[Sudip Mukherjee]

If the serial device is disconnected and reconnected, it re-enumerates
properly but does not link it. fwiw, linking means just saving the port
index, so allow it always as there is no harm in saving the same value
again even if it tries to relink with the same port.

Fixes: fb2b90014d78 ("tty: link tty and port before configuring it as console")
Reported-by: Kenneth R. Crudup <kenny@panix.com>
Signed-off-by: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
---
 drivers/tty/tty_port.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/tty/tty_port.c b/drivers/tty/tty_port.c
index 5023c85ebc6e..044c3cbdcfa4 100644
--- a/drivers/tty/tty_port.c
+++ b/drivers/tty/tty_port.c
@@ -89,8 +89,7 @@ void tty_port_link_device(struct tty_port *port,
 {
 	if (WARN_ON(index >= driver->num))
 		return;
-	if (!driver->ports[index])
-		driver->ports[index] = port;
+	driver->ports[index] = port;
 }
 EXPORT_SYMBOL_GPL(tty_port_link_device);
 
-- 
2.11.0

Re: [PATCH] tty: always relink the port

[Johan Hovold]

On Fri, Dec 27, 2019 at 05:44:34PM +0000, Sudip Mukherjee wrote:
> If the serial device is disconnected and reconnected, it re-enumerates
> properly but does not link it. fwiw, linking means just saving the port
> index, so allow it always as there is no harm in saving the same value
> again even if it tries to relink with the same port.

This is a pretty vague description. Commit fb2b90014d78 ("tty: link tty
and port before configuring it as console") completely broke usb-serial
(and anything else hotpluggable) which obviously depends on being able
to reuse a minor number when a new device is later plugged in after a
disconnect.

Things are crashing left and right due to that stale port-pointer, and I
just had to debug this only to find that this one is sitting in the
tty-linus branch. I know, I know, Christmas and all, but would be nice
to get it into -rc6. :)

> Fixes: fb2b90014d78 ("tty: link tty and port before configuring it as console")

Also note that the offending commit had a stable tag unlike this one.

> Reported-by: Kenneth R. Crudup <kenny@panix.com>
> Signed-off-by: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
> ---
>  drivers/tty/tty_port.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/tty/tty_port.c b/drivers/tty/tty_port.c
> index 5023c85ebc6e..044c3cbdcfa4 100644
> --- a/drivers/tty/tty_port.c
> +++ b/drivers/tty/tty_port.c
> @@ -89,8 +89,7 @@ void tty_port_link_device(struct tty_port *port,
>  {
>  	if (WARN_ON(index >= driver->num))
>  		return;
> -	if (!driver->ports[index])
> -		driver->ports[index] = port;
> +	driver->ports[index] = port;
>  }
>  EXPORT_SYMBOL_GPL(tty_port_link_device);

Johan

Re: [PATCH] tty: always relink the port

[Greg Kroah-Hartman]

On Fri, Jan 10, 2020 at 11:08:17AM +0100, Johan Hovold wrote:
> On Fri, Dec 27, 2019 at 05:44:34PM +0000, Sudip Mukherjee wrote:
> > If the serial device is disconnected and reconnected, it re-enumerates
> > properly but does not link it. fwiw, linking means just saving the port
> > index, so allow it always as there is no harm in saving the same value
> > again even if it tries to relink with the same port.
> 
> This is a pretty vague description. Commit fb2b90014d78 ("tty: link tty
> and port before configuring it as console") completely broke usb-serial
> (and anything else hotpluggable) which obviously depends on being able
> to reuse a minor number when a new device is later plugged in after a
> disconnect.
> 
> Things are crashing left and right due to that stale port-pointer, and I
> just had to debug this only to find that this one is sitting in the
> tty-linus branch. I know, I know, Christmas and all, but would be nice
> to get it into -rc6. :)

Sorry, yes, my fault, will get to Linus either today or tomorrow.

> > Fixes: fb2b90014d78 ("tty: link tty and port before configuring it as console")
> 
> Also note that the offending commit had a stable tag unlike this one.

I'll pick it up properly, I have held off on adding the original to the
stable trees yet.

thanks,

greg k-h