mbox series

[v4,0/2] serial: sifive: Convert sifive console to nbcon

Message ID 20250405043833.397020-1-ryotkkr98@gmail.com
Headers show
Series serial: sifive: Convert sifive console to nbcon | expand

Message

Ryo Takakura April 5, 2025, 4:38 a.m. UTC 
Hi!

This is v4 of series converting sifive console to nbcon.

The first patch fixes the issue which was pointed out by John [0] 
that the driver has been accessing SIFIVE_SERIAL_IE_OFFS register 
on its ->startup() and ->shutdown() without port lock synchronization 
against ->write().

The fix on the first patch still applies to the second patch which 
converts the console to nbcon as ->write_thread() holds port lock
and ->write_atomic() checks for the console ownership.

Sincerely,
Ryo Takakura

[0] https://lore.kernel.org/lkml/84sen2fo4b.fsf@jogness.linutronix.de/

---

Changes since v1:
[1] https://lore.kernel.org/lkml/20250323060603.388621-1-ryotkkr98@gmail.com/

- Thank you John for the feedback!
- Add a patch for synchronizing startup()/shutdown() vs write(). 
- Add <Reviewed-by> by John.

Changes since v2:
[2] https://lore.kernel.org/all/20250330003058.386447-1-ryotkkr98@gmail.com/ 

- Add Cc stable for the first patch.

Changes since v3:
[3] https://lore.kernel.org/all/20250330110957.392460-1-ryotkkr98@gmail.com/

- Avoid using return statement on return-void function. Thanks Sebastian 
  for pointing out!

---

Ryo Takakura (2):
  serial: sifive: lock port in startup()/shutdown() callbacks
  serial: sifive: Switch to nbcon console

 drivers/tty/serial/sifive.c | 93 +++++++++++++++++++++++++++++++------
 1 file changed, 80 insertions(+), 13 deletions(-)

Comments

Greg KH April 5, 2025, 7:35 a.m. UTC | #1 
On Sat, Apr 05, 2025 at 01:43:38PM +0900, Ryo Takakura wrote:
> startup()/shutdown() callbacks access SIFIVE_SERIAL_IE_OFFS.
> The register is also accessed from write() callback.
> 
> If console were printing and startup()/shutdown() callback
> gets called, its access to the register could be overwritten.
> 
> Add port->lock to startup()/shutdown() callbacks to make sure
> their access to SIFIVE_SERIAL_IE_OFFS is synchronized against
> write() callback.
> 
> Signed-off-by: Ryo Takakura <ryotkkr98@gmail.com>
> Cc: stable@vger.kernel.org

What commit id does this fix?

Why does patch 1/2 need to go to stable, but patch 2/2 does not?  Please
do not mix changes like this in the same series, otherwise we have to
split them up manually when we apply them to the different branches,
right?

thanks,

greg k-h