Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
Date:   Mon, 20 Nov 2023 13:29:41 +0100
From:   Corentin LABBE <clabbe@baylibre.com>
To:     Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Cc:     gregkh@linuxfoundation.org, johan@kernel.org,
        linux-kernel@vger.kernel.org, linux-usb@vger.kernel.org,
        david@ixit.cz
Subject: Re: [PATCH v6 0/2] usb: serial: add support for CH348
Message-ID: <ZVtRNZmCMImCT9sN@Red>
References: <20230628133834.1527941-1-clabbe@baylibre.com>
 <2595072.9XhBIDAVAK@archbook>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <2595072.9XhBIDAVAK@archbook>
Precedence: bulk

Le Wed, Aug 30, 2023 at 07:43:03PM +0200, Nicolas Frattaroli a ?crit :
> On Mittwoch, 28. Juni 2023 15:38:32 CEST Corentin Labbe wrote:
> > Hello
> > 
> > The CH348 is an octo serial to USB adapter.
> > The following patch adds a driver for supporting it.
> > Since there is no public datasheet, unfortunatly it remains some magic values.
> > 
> > It was tested with a large range of baud from 1200 to 1500000 and used with
> > success in one of our kernel CI testlab.
> > 
> > Regards
> > 
> > [...]
> 
> Hello,
> 
> thank you for your work on this. I recently made myself a CH348
> board and used this patchset with a small test application[1]
> to see how it performs. Specifically, I ran this on an RK3566
> single board computer, connecting one serial adapter to the
> other, with the test as follows:
> 
>  ./serialtest /dev/ttyUSB0 9600 # UART0 of 1st CH348 board
>  ./serialtest /dev/ttyUSB8 9600 # UART0 of 2nd CH348 board
> 
> One problem I've noticed is that writes to the tty fd never
> seem to block. On two CH340 adapters I have, they do seem to
> block, whereas here, you can see from the statistics at the
> end that magnitudes more bytes were written than read, with
> seemingly most of them being discarded. From my reading of
> the termios parameters I set, this shouldn't be the case,
> right?
> 
> You can see from the error percentage that it gets less
> bad as you increase the serial baudrate; I've tested up
> to 6 mbaud like this. I assume that's because less written
> bytes get discarded.
> 
> Any ideas on whether I'm relying on weird driver behaviour
> with the blocking here or if this driver actually has a
> defect whereby it never signals to userspace that less
> bytes were written than have been submitted?
> 
> Kind regards,
> Nicolas Frattaroli
> 
> [1]: https://github.com/CounterPillow/serialtest
> 

Hello

Sorry for the very long delay of the answer.
I have reproduced the problem on my board.
My reproducer is https://github.com/montjoie/lava-tests/blob/master/test2a2.py

This problem seems to be here since the v1 of my patchset.
The vendor driver seems to work so it is not an hardware problem.

I have no clue at the moment, it is hard to diff with vendor driver since it create tty directly and do not use usbserial.

Regards