[PATCH 0/2] fixes for empty tx buffer breakage

From: Peter Hurley
Date: Sun Jul 06 2014 - 11:30:36 EST


Cc: Seth Bollinger <sethb@xxxxxxxx>
Cc: Aaro Koskinen <aaro.koskinen@xxxxxx>
Cc: Sam Ravnborg <sam@xxxxxxxxxxxx>
Cc: "David S. Miller" <davem@xxxxxxxxxxxxx>
Cc: Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx>

Greg,

I completed the audit of serial drivers after reports that
several Sun serial drivers were broken by
commit 717f3bbab3c7628736ef738fdbf3d9a28578c26c,
'serial_core: Fix conditional start_tx on ring buffer not empty'.

I apologize for not submitting this sooner. The delay was due to
an ongoing analysis of serial flow control prompted by Sam Ravnborg's
question:

On 06/10/2014 03:24 PM, Sam Ravnborg wrote:
> I also noticed the typical pattern is:
>
> if (uart_circ_empty(xmit) || uart_tx_stopped(port))
>
> Should you use this pattern also in sunsab.c?

Unfortunately, that analysis revealed that tx flow control is
largely SMP-unsafe, and it's fairly easy to corrupt the hardware
state wrt. the tty flow control state.

I'm still working on the solutions to that; they're too
extensive to submit for 3.16 anyway.

Regards,

Peter Hurley (2):
serial: Test for no tx data on tx restart
serial: arc_uart: Use uart_circ_empty() for open-coded comparison

drivers/tty/serial/arc_uart.c | 2 +-
drivers/tty/serial/imx.c | 3 +++
drivers/tty/serial/ip22zilog.c | 2 ++
drivers/tty/serial/m32r_sio.c | 8 +++++---
drivers/tty/serial/pmac_zilog.c | 3 +++
drivers/tty/serial/sunsab.c | 3 +++
drivers/tty/serial/sunzilog.c | 2 ++
7 files changed, 19 insertions(+), 4 deletions(-)

--
2.0.0

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/