From: Matthias Schiffer <matthias.schiffer@ew.tq-group.com>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Jiri Slaby <jirislaby@kernel.org>, Lukas Wunner <lukas@wunner.de>
Cc: "Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>,
linux-serial@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: 8250_omap: incorrect RS485 RTS on close during transmission
Date: Tue, 22 Nov 2022 09:52:40 +0100 [thread overview]
Message-ID: <bce6407bc42eeec8f459f09bc1326761327c98ec.camel@ew.tq-group.com> (raw)
In-Reply-To: <f7cf39a9c152031a09608a68115e32988617f626.camel@ew.tq-group.com>
On Fri, 2022-11-18 at 12:57 +0100, Matthias Schiffer wrote:
> Hi Lukas,
>
> I've discovered a new issue with EM485 in 8250_omap (and probably other
> serial drivers as well?): When closing the TTY device while there is
> still an ongoing transmission, the RTS pin will get stuck active until
> the TTY is opened again. This can be easily reproduced by running `cat
> /dev/zero > /dev/ttySx` and then stopping it using Ctrl-C.
>
> The issue exists in current mainline (6.1-rc5+), and applying "serial:
> 8250: 8250_omap: Avoid RS485 RTS glitch on ->set_termios()" from next
> doesn't solve it. I think it should work with "serial: 8250: 8250_omap:
> Support native RS485", but we also need RS485 on AM57xx CPUs, which
> don't have native RS485.
>
> I intend to look into this myself next week, but as you've worked on
> this code a lot lately, maybe you already have an idea how to fix it?
>
> Regards,
> Matthias
>
Okay, I've narrowed down the issue to uart_shutdown() ->
uart_port_shutdown() - here, ops->shutdown() is called, which clears
the FIFOs (in the case of the default 8250_port and 8250_omap)
implementations, but doesn't wait for the clear to finish, and doesn't
call ops->stop_tx().
I believe a similar problem can occur in uart_suspend_port(). Here,
ops->stop_tx() is called before ops->shutdown(), but stop_tx() is a no-
op when the TX FIFO is not empty (again, for the 8250 case - I haven't
really looked at other drivers).
Given that EM485 is handled in 8250_port by __stop_tx(), which is
called both internally in the 8250 driver and from the serial core
through stop_tx(), it is not clear to me what the correct layer to fix
this is. Two ideas:
(1) Have ops->shutdown() wait until the TX FIFO is empty. Call
ops->stop_tx() after ops->shutdown() in serial core. For this to work,
calling stop_tx() after shutdown() must be allowed in all drivers; I'm
not sure if this is currently the case, and if it is desirable at all.
(2) Do the whole handling in the driver's ops->shutdown(), do not
further involve the serial core.
So far, I've implemented (1) locally for 8250_omap, and together with
"serial: 8250: 8250_omap: Avoid RS485 RTS glitch on ->set_termios()" it
fixes the issue on my test devices, but I'm not too happy about the
"shutdown then tx_stop" order of calls. So maybe (2) is the better
solution after all?
prev parent reply other threads:[~2022-11-22 8:52 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-18 11:57 Matthias Schiffer
2022-11-22 8:52 ` Matthias Schiffer [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=bce6407bc42eeec8f459f09bc1326761327c98ec.camel@ew.tq-group.com \
--to=matthias.schiffer@ew.tq-group.com \
--cc=gregkh@linuxfoundation.org \
--cc=ilpo.jarvinen@linux.intel.com \
--cc=jirislaby@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-serial@vger.kernel.org \
--cc=lukas@wunner.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®