Hi Rob,
On Mon, Oct 3, 2016 at 10:30 PM, Rob Herring <robh@xxxxxxxxxx> wrote:
> On Mon, Oct 3, 2016 at 2:42 PM, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
>> On Mon, Oct 3, 2016 at 8:25 PM, Rob Herring <robh@xxxxxxxxxx> wrote:
>>> On Thu, Sep 29, 2016 at 10:31 AM, Geert Uytterhoeven
>>> <geert+renesas@xxxxxxxxx> wrote:
>>>> The port->console flag is always false, as uart_console() is called
>>>> before the serial console has been registered.
>>>
>>> I'm not seeing how that is. Everything uart_console() depends on
>>> (port->cons, port->cons->index, and port->line) should be set already.
>>> Maybe you pass in -1 for index which gets changed to 0 is the only
>>> thing I can see. Is doing that valid if you have multiple ports?
>>
>> Isn't .index always initialized to -1, and set to the actual line number when
>> the console is registered? That's how I remember it from when
>> register_console() was introduced (in 2.1)...
>
> Okay, the index is also set and gets marked enabled when specified on
> the command line (or DT).
I'm using DT and stdout-path. Where is the index set?
>>>> Hence for a serial port used as the console, uart_tty_port_shutdown()
>>>> will still be called when userspace closes the port, powering it down.
>>>> This will lead to a system lock up when the serial console driver writes
>>>> to the serial port's registers.
>>>>
>>>> To fix this, move the setting of port->console after the call to
>>>> uart_configure_port(), which registers the serial console.
>>>>
>>>> Fixes: 761ed4a94582ab29 ("tty: serial_core: convert uart_close to use tty_port_close")
>>>> Reported-by: Niklas Söderlund <niklas.soderlund+renesas@xxxxxxxxxxxx>
>>>> Reported-by: Takeshi Kihara <takeshi.kihara.df@xxxxxxxxxxx>
>>>> Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
>>>> ---
>>>> RFC because of the comment "If this port is a console, then the spinlock
>>>> is already initialised", and the pre-existing code calling
>>>> uart_console() before uart_configure_port().
>>>
>>> So the spinlock is initialized twice which is probably harmless? If
>>> the spinlock was initialized by the console, then the index would
>>> probably not be -1.
>>
>> Shouldn't a double initialization cause a warning, with DEBUG_SPINLOCK=y
>> (which I have enabled)?
>
> I though only missing init caused warnings.
>
> Anyway, we can't really re-order things as uart_configure_port uses
> the spinlock. This has existed this way for some time and is all an
> independent of my change and your fix. So for your fix:
>
> Acked-by: Rob Herring <robh@xxxxxxxxxx>
Thanks!
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
