serial: stm32: fix threaded interrupt handling
authorJohan Hovold <johan@kernel.org>
Fri, 16 Apr 2021 14:05:56 +0000 (16:05 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 22 Apr 2021 10:04:26 +0000 (12:04 +0200)
commite359b4411c2836cf87c8776682d1b594635570de
tree6157b722187b53ec336ea6607957be90f5ba6061
parent75f4e830fa9c47637054a3b7201765f2a314bda2
serial: stm32: fix threaded interrupt handling

When DMA is enabled the receive handler runs in a threaded handler, but
the primary handler up until very recently neither disabled interrupts
in the device or used IRQF_ONESHOT. This would lead to a deadlock if an
interrupt comes in while the threaded receive handler is running under
the port lock.

Commit ad7676812437 ("serial: stm32: fix a deadlock condition with
wakeup event") claimed to fix an unrelated deadlock, but unfortunately
also disabled interrupts in the threaded handler. While this prevents
the deadlock mentioned in the previous paragraph it also defeats the
purpose of using a threaded handler in the first place.

Fix this by making the interrupt one-shot and not disabling interrupts
in the threaded handler.

Note that (receive) DMA must not be used for a console port as the
threaded handler could be interrupted while holding the port lock,
something which could lead to a deadlock in case an interrupt handler
ends up calling printk.

Fixes: ad7676812437 ("serial: stm32: fix a deadlock condition with wakeup event")
Fixes: 3489187204eb ("serial: stm32: adding dma support")
Cc: stable@vger.kernel.org # 4.9
Cc: Alexandre TORGUE <alexandre.torgue@st.com>
Cc: Gerald Baeza <gerald.baeza@st.com>
Reviewed-by: Valentin Caron<valentin.caron@foss.st.com>
Signed-off-by: Johan Hovold <johan@kernel.org>
Link: https://lore.kernel.org/r/20210416140557.25177-3-johan@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/tty/serial/stm32-usart.c