Add new ASYNC_SPD_WARP13?

Add new ASYNC_SPD_WARP13?

[Grant Edwards]

I'm working on a driver for a serial board that supports 921600 bps,
and it would be awfully nice if there was a way to specify that using
the ASYNC_SPD_xxxxx mechanism.  Is there any reason why this wouldn't
work?


  #define ASYNC_SPD_HI         (1U << ASYNCB_SPD_HI)
  #define ASYNC_SPD_VHI        (1U << ASYNCB_SPD_VHI)
  #define ASYNC_SPD_SHI        (1U << ASYNCB_SPD_SHI)
  #define ASYNC_SPD_CUST       (ASYNC_SPD_HI|ASYNC_SPD_VHI)
  #define ASYNC_SPD_WARP       (ASYNC_SPD_HI|ASYNC_SPD_SHI)
+ #define ASYNC_SPD_WARP13     (ASYNC_SPD_HI|ASYNC_SPD_SHI|ASYNC_SPD_SHI)

  #define ASYNC_SPD_MASK       (ASYNC_SPD_HI|ASYNC_SPD_VHI|ASYNC_SPD_SHI)

  
-- 
Grant Edwards               grant.b.edwards        Yow! I'm wearing PAMPERS!!
                                  at               
                              gmail.com            

Re: Add new ASYNC_SPD_WARP13?

[Alan Cox]

On Wed, 6 Apr 2011 16:48:49 +0000 (UTC)
Grant Edwards <grant.b.edwards@gmail.com> wrote:

> I'm working on a driver for a serial board that supports 921600 bps,
> and it would be awfully nice if there was a way to specify that using
> the ASYNC_SPD_xxxxx mechanism.  Is there any reason why this wouldn't
> work?

The ASYNC_SPD_xxxx mechanism is obsolete. Just set the baud rate to
921600, the kernel has supported arbitary baud rates for some time.

Re: Add new ASYNC_SPD_WARP13?

[Grant Edwards]

On 2011-04-06, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
> On Wed, 6 Apr 2011 16:48:49 +0000 (UTC)
> Grant Edwards <grant.b.edwards@gmail.com> wrote:
>
>> I'm working on a driver for a serial board that supports 921600 bps,
>> and it would be awfully nice if there was a way to specify that using
>> the ASYNC_SPD_xxxxx mechanism.  Is there any reason why this wouldn't
>> work?
>
> The ASYNC_SPD_xxxx mechanism is obsolete.

I know -- unfortunately, so are many customer applications.

> Just set the baud rate to 921600, the kernel has supported arbitary
> baud rates for some time.

Yep, but there are customers with existing, fielded, applications that
depend on the ASYNC_SPD_xxxx mechanism to map 38400 to <whatever>
[where <whatever> is determined by the SPD_xxxx flags controlled by
setserial].

In the past, we've provided a customized version of setserial that
supported a SPD_WARP13 flag setting for 921600, but it only worked
with a particular driver.  I'd like to ditch that and in future use
the standard setserial and standard SPD_xxxx flags.  But, there is no
standard way to set flags for 921600.

I could tell them to go the SPD_CUST route to get 921600, but I was
hoping to provide an option that would support 921600 the same way
that 57600, 230400, and 460800 are supported.

-- 
Grant Edwards               grant.b.edwards        Yow! I'm dressing up in
                                  at               an ill-fitting IVY-LEAGUE
                              gmail.com            SUIT!!  Too late...

Re: Add new ASYNC_SPD_WARP13?

[Grant Edwards]

On 2011-04-06, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:

> The ASYNC_SPD_xxxx mechanism is obsolete. Just set the baud rate to
> 921600, the kernel has supported arbitary baud rates for some time.

This raises a question I've been wondering about for some time.  While
the kernel and the serial driver API support arbitrary baud rates,
termios doesn't AFAICT.

How does one use the standard POSIX serial port API to obtain an
abritrary baud rate?

The only way I can find to get arbitrary baud rates from a user
perspective is to use the ASYNC_SPD_xxxx mechanism.

-- 
Grant Edwards               grant.b.edwards        Yow! Catsup and Mustard all
                                  at               over the place!  It's the
                              gmail.com            Human Hamburger!

Re: Add new ASYNC_SPD_WARP13?

[Alan Cox]

> I could tell them to go the SPD_CUST route to get 921600, but I was
> hoping to provide an option that would support 921600 the same way
> that 57600, 230400, and 460800 are supported.

And no doubt in time other speeds as well.

ASYNC_SPD_xxx is legacy, it remains simply because we provided it as a
historic interface to some drivers and we don't want to break stuff

So please use SPD_CUST if you must do this.

Alan

Re: Add new ASYNC_SPD_WARP13?

[Grant Edwards]

On 2011-04-06, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
>> I could tell them to go the SPD_CUST route to get 921600, but I was
>> hoping to provide an option that would support 921600 the same way
>> that 57600, 230400, and 460800 are supported.
>
> And no doubt in time other speeds as well.
>
> ASYNC_SPD_xxx is legacy, it remains simply because we provided it as a
> historic interface to some drivers and we don't want to break stuff
>
> So please use SPD_CUST if you must do this.

I don't understand: what's the non-legacy alternative to using SPD_CUST?

-- 
Grant Edwards               grant.b.edwards        Yow! I'm having a
                                  at               tax-deductible experience!
                              gmail.com            I need an energy crunch!!

Re: Add new ASYNC_SPD_WARP13?

[Alan Cox]

> This raises a question I've been wondering about for some time.  While
> the kernel and the serial driver API support arbitrary baud rates,
> termios doesn't AFAICT.
> 
> How does one use the standard POSIX serial port API to obtain an
> abritrary baud rate?

The POSIX API is completely broken in this respect. So the kernel
provides its own interface.

> The only way I can find to get arbitrary baud rates from a user
> perspective is to use the ASYNC_SPD_xxxx mechanism.

Which is also not POSIX


Simple example app:
 
http://www.kerneltrap.com/mailarchive/linux-usb/2008/10/2/3484644

Re: Add new ASYNC_SPD_WARP13?

[Grant Edwards]

On 2011-04-06, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
>> This raises a question I've been wondering about for some time.  While
>> the kernel and the serial driver API support arbitrary baud rates,
>> termios doesn't AFAICT.
>> 
>> How does one use the standard POSIX serial port API to obtain an
>> abritrary baud rate?
>
> The POSIX API is completely broken in this respect. So the kernel
> provides its own interface.
>
>> The only way I can find to get arbitrary baud rates from a user
>> perspective is to use the ASYNC_SPD_xxxx mechanism.
>
> Which is also not POSIX

Right.

> Simple example app:
>  
> http://www.kerneltrap.com/mailarchive/linux-usb/2008/10/2/3484644

Brilliant!

I've been looking for that (off and on) for months.  I knew that the
kernel/driver API supported arbitrary baud rates and that termios
didn't.  So I figured there must be an alternative to SPD_CUST, but
hadn't been able to figure out what it was.

I had found the BSD IOSSIOSPEED ioctl, but hadn't stumbled across the
Linux equivalent.

-- 
Grant Edwards               grant.b.edwards        Yow! Look DEEP into the
                                  at               OPENINGS!!  Do you see any
                              gmail.com            ELVES or EDSELS ... or a
                                                   HIGHBALL?? ...