Date:	Wed, 14 Jan 2015 10:39:04 -0800
From:	Felipe Tonello <eu@...ipetonello.com>
To:	One Thousand Gnomes <gnomes@...rguk.ukuu.org.uk>
Cc:	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Jiri Slaby <jslaby@...e.cz>, linux-kernel@...r.kernel.org,
	linux-serial@...r.kernel.org
Subject: Re: [PATCH] char: Added support for u-blox 6 i2c gps driver

Hi Alan,

On Wed, Jan 14, 2015 at 7:48 AM, One Thousand Gnomes
<gnomes@...rguk.ukuu.org.uk> wrote:
> On Tue, 13 Jan 2015 17:16:42 -0800
> "Felipe F. Tonello" <eu@...ipetonello.com> wrote:
>
>> This driver will basically translate serial communication to i2c communication
>> between the user-space and the GPS module.
>>
>> It creates a /dev/ttyS device node.
>
> It shouldn't. It should use its own name. (ttyubl or something .. )

Ok.

>
>
>> There are specific tty termios flags in order to the tty line discipliner not
>> to change the date it is pushed to user space. That is necessary so the NMEA
>> data is not corrupted.
>> * iflags: added IGNCR: Ignore carriage return on input.
>> * oflags: removed ONLCR: (XSI) Map NL to CR-NL on output.
>
> The user space should probably set these but if its a single purpose
> device then it's not unreasonable to just set defaults I guess.

Yes, that was my idea.

>
>> +static DECLARE_DELAYED_WORK(ublox_gps_wq, ublox_gps_read_worker);
>> +
>> +static void ublox_gps_read_worker(struct work_struct *private)
>> +{
>> +     s32 gps_buf_size, buf_size = 0;
>> +     u8 *buf;
>> +
>> +     if (!ublox_gps_is_open)
>> +             return;
>> +
>> +     /* check if driver was removed */
>> +     if (!ublox_gps_i2c_client)
>> +             return;
>
>
> Greg has pointed out your locking is nonsensical - you want to be
> refcounting. In the tty side case it's done for you for the most part,
> but for the i2c side it's your problem and you need to ensure that you
> hold suitable references between tty open and close, and on the close you
> kill and wait for any work queues.
>
>
>> +static int ublox_gps_serial_open(struct tty_struct *tty, struct file *filp)
>> +{
>> +     if (ublox_gps_is_open)
>> +             return -EBUSY;
>> +
>> +     ublox_gps_filp = filp;
>> +     ublox_gps_tty_port = tty->port;
>> +     ublox_gps_tty_port->low_latency = true; /* make sure we push data immediately */
>> +     ublox_gps_is_open = true;
>
> tty semantics for exclusive are standardised in POSIX and this is not the
> expected behaviour. See the TIOCEXCL ioctl.
>
>> +static void ublox_gps_serial_close(struct tty_struct *tty, struct file *filp)
>> +{
>> +     if (!ublox_gps_is_open)
>> +             return;
>> +
>> +     /* avoid stop when the denied (in open) file structure closes itself */
>> +     if (ublox_gps_filp != filp)
>> +             return;
>
> That should never be possible. If it can happen you've got a locking bug
> that needs fixing instead
>
>> +static int ublox_gps_serial_write(struct tty_struct *tty, const unsigned char *buf,
>> +     int count)
>> +{
>> +     if (!ublox_gps_is_open)
>> +             return 0;
>> +
>> +     /* check if driver was removed */
>> +     if (!ublox_gps_i2c_client)
>> +             return 0;
>> +
>> +     /* we don't write back to the GPS so just return same value here */
>> +     return count;
>
> Just omit the serial_write method, or return -EIO; don't pretend it
> worked.

Ok. I had to write the write because the tty layer was calling
write_room without testing if the pointer was NULL. That was causing a
segmentation fault. And If you implement the write_room you have to
implement the write too.

Maybe more recent versions fixed this bug.

>
>> +     ublox_gps_tty_driver->driver_name = "ublox_gps";
>> +     ublox_gps_tty_driver->name = "ttyS";
>> +     ublox_gps_tty_driver->major = UBLOX_GPS_MAJOR;
>
> Not ttyS and use dynamic numbering
>
>> +     ublox_gps_tty_driver->minor_start = 0;
>> +     ublox_gps_tty_driver->type = TTY_DRIVER_TYPE_SERIAL;
>> +     ublox_gps_tty_driver->subtype = SERIAL_TYPE_NORMAL;
>> +     ublox_gps_tty_driver->flags = TTY_DRIVER_REAL_RAW;
>> +     ublox_gps_tty_driver->init_termios = tty_std_termios;
>> +     ublox_gps_tty_driver->init_termios.c_iflag = IGNCR | IXON;
>> +     ublox_gps_tty_driver->init_termios.c_oflag = OPOST;
>> +     ublox_gps_tty_driver->init_termios.c_cflag = B9600 | CS8 | CREAD |
>> +             HUPCL | CLOCAL;
>> +     ublox_gps_tty_driver->init_termios.c_ispeed = 9600;
>> +     ublox_gps_tty_driver->init_termios.c_ospeed = 9600;
>> +     tty_set_operations(ublox_gps_tty_driver, &ublox_gps_serial_ops);
>> +     result = tty_register_driver(ublox_gps_tty_driver);
>> +     if (result) {
>> +             dev_err(&ublox_gps_i2c_client->dev, KBUILD_MODNAME ": %s - tty_register_driver failed\n",
>> +                     __func__);
>> +             goto err;
>> +     }
>> +
>> +     ublox_gps_i2c_client = client;
>> +     ublox_gps_filp = NULL;
>> +     ublox_gps_tty_port = NULL;
>> +     ublox_gps_is_open = false;
>
> There are some other i2c based tty drivers in the kernel - notably those
> in drivers/tty/serial that use the uart layer to deal with most of the
> awkward locking cases.
>
> You can do it by hand but it's fairly hairy (see
> drivers/mmc/card/sdio_uart.c, so it might be simplest to tweak the driver
> to use the uart layer. You don't really gain much from it for your driver
> except easier locking - but the locking is rather handy.
>
> Alan

Ok.

The thing is that: I wrote this driver to work with only one gps
module, because that's my configuration here. I cannot really test
multiple i2c gps at the same time. If you guys really want a driver
that works for multiple gps drivers, I cannot test it.

So my question is, can I leave the driver working for one gps only and
then improve what needs to be improved?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists