On 10/2/21 10:21 am, Joel Sherrill wrote:
> On Tue, Feb 9, 2021 at 5:08 PM Chris Johns <chr...@rtems.org
> <mailto:chr...@rtems.org>> wrote:
> 
>     On 9/2/21 11:42 pm, Jan Sommer wrote:
>     > From: Kinsey Moore <kinsey.mo...@oarcorp.com
>     <mailto:kinsey.mo...@oarcorp.com>>
>     >
>     > The zynq-uart set_attributes implementation was configured to always
>     > return false which causes spconsole01 to fail. This restores the
>     > disabled implementation which sets the baud rate registers
>     > appropriately and allows spconsole01 to pass. This also expands the
>     > set_attributes functionality to allow setting of the stop bits,
>     > character width, and parity.
> 
>     Hmmm I missed this change.
> 
>     Sebastian changed the Zynq driver to return an error. Returning an error 
> is
>     correct because the attributes are not being set however we have a couple 
> of
>     design approaches clashing.
> 
>     The issue pulls right back to the Xilinx SystemZ design dialogue box. The
>     hardware designer has the ability to set the serial port defaults. These
>     settings form part of the ps7init data the FSBL paints into the hardware 
> very
>     early in the boot process. Also in this data are parallel port pin 
> settings, AXI
>     clocks, memory and more so lots of important stuff. The Xilinx bootloader 
> and
>     ones I have created (I do not know about uboot) do not touch the UART 
> hardware
>     on purpose so the system designer has control. Then in 2017 Sebastian 
> pushed a
>     change to force the baudrate to a BSP specific default.
> 
>     This is normally all OK because all Zynq designs I know of use the 
> default of
>     115200,8,n,1 for the configuration. A problem arises when these defaults 
> are
>     varied.
> 
>     Maybe something explaining this in the user manual ...
> 
>     https://docs.rtems.org/branches/master/user/bsps/bsps-arm.html#xilinx-zynq
>     
> <https://docs.rtems.org/branches/master/user/bsps/bsps-arm.html#xilinx-zynq>
> 
>     so we have provided information on what to change?
> 
> 
> I wonder if there should be a configuration parameter for the default 
> settings.
> At least "don't override the defaults"

I am not sure this is worth the effort. Personally I cannot see any reason
anyone would move away from 115200. All designs I have see use the eval board's
USB serial interface so 115200 is easy to support.

I think a small fragment of user doco will work.

>     The other approach is not to touch the defaults in hardware and allow a 
> get
>     attributes to discover them? Does termios drivers have a low level get? I 
> can
>     see a set but no get in the zynq uart driver.
> 
> No. This reflects termios from the POSIX perspective. 


Ah well that rules that option out.

Chris
_______________________________________________
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel

Reply via email to