Received-SPF: pass (google.com: domain of linux-kernel+bounces-183029-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Message-ID: <f7744e88-8123-4ab9-8542-209b10b09321@schmorgal.com>
Date: Sat, 18 May 2024 22:51:54 -0700
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH v2] serial: core: only stop transmit when HW fifo is empty
From: Doug Brown <doug@schmorgal.com>
To: Jonas Gorski <jonas.gorski@gmail.com>, linux-kernel@vger.kernel.org,
 linux-serial@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
 Jiri Slaby <jirislaby@kernel.org>,
 =?UTF-8?Q?Ilpo_J=C3=A4rvinen?= <ilpo.jarvinen@linux.intel.com>,
 Florian Fainelli <florian.fainelli@broadcom.com>, stable@vger.kernel.org
References: <20240303150807.68117-1-jonas.gorski@gmail.com>
 <77b71bd9-42be-40e8-8b96-196e214c8afb@schmorgal.com>
Content-Language: en-US
In-Reply-To: <77b71bd9-42be-40e8-8b96-196e214c8afb@schmorgal.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

Hi again,

On 5/16/2024 9:22 PM, Doug Brown wrote:

> I'm hoping there is some kind of simple fix that can be made to the pxa
> driver to work around it with this new behavior. Can anyone think of a
> reason that this driver would not like this change? It seems
> counterintuitive to me -- the patch makes perfect sense.

After further experimentation, I've come to the conclusion that this is
a bug in the pxa uart driver, and this patch simply exposed the bug.
I'll submit a patch to fix the issue in the pxa driver.

If anyone's interested in the details: basically, the pxa driver in its
current state doesn't work correctly if it receives a TX interrupt when
the circular buffer is empty. It handles it, but then gets stuck waiting
for the next TX IRQ that will never happen because no characters were
transmitted. The way stop_tx() was previously being called before the
transmitter was empty, it prevented that situation from happening
because toggling the TX interrupt enable flag off (with stop_tx) and
back on (with the next start_tx) causes a new TX interrupt to fire and
kickstarts the transmit process again.

The 8250 driver, for example, isn't affected by this problem because it
effectively does stop_tx() on its own if it detects an empty circular
buffer in the TX interrupt handler. Adding similar logic to the pxa
driver fixes it.

Doug