<div dir="ltr"><div dir="ltr">On Fri, Dec 27, 2019 at 1:58 AM Christian Mauderer <<a href="mailto:list@c-mauderer.de">list@c-mauderer.de</a>> wrote:<br></div><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hello Niteesh,<br>
<br>
sorry for not answering earlier. During this time of the year you have<br>
to expect some delays on the mailing list due to public holydays and<br>
vacations.</blockquote><div>That's okay, I understand. </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> <br></blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
On 25/12/2019 10:50, Niteesh wrote:<br>
> Just to make sure I am going in the right track.<br>
> I moved the uart register definitions to bsp/usart.h into a struct of<br>
> uint32_t called usart0_regs<br>
> here is git diff of usart.c after changing it to the latest console<br>
> interface.<br>
<br>
Do you have a plan how you want to test these changes?<br>
<br>
The direction looks OK. Some notes below.<br>
<br>
By the way: Maybe it would be a better idea to just remove it completely<br>
and use the bsps/arm/shared/serial/arm-pl011.c driver instead? That<br>
reduces the ammount of code and therefore ammount of bugs we have in<br>
this BSP. </blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> <br></blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
> <br>
> diff --git a/bsps/arm/raspberrypi/console/usart.c<br>
> b/bsps/arm/raspberrypi/console/usart.c<br>
> index 25fb523621..b12f375a1c 100644<br>
> --- a/bsps/arm/raspberrypi/console/usart.c<br>
> +++ b/bsps/arm/raspberrypi/console/usart.c<br>
> @@ -47,6 +47,12 @@ static uint32_t usart_get_baud(const console_tbl *ct)<br>
>  }<br>
>  #endif<br>
>  <br>
> +typedef struct {<br>
> +  rtems_termios_device_context base;<br>
> +  const char *device_name;<br>
> +  volatile usart0_regs *regs;<br>
> +}uart0_context;<br>
<br>
Why uart0_context and not usart_context? All other names in this file<br>
are called usart_...<br>
<br></blockquote><div>Sorry, for the inconsistent naming, should I rename it as pl011_context since we will be adding</div><div>mini uart for rpi3, IMHO it would be better.</div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
> +<br>
>  static void usart_set_baud(int minor, int baud)<br>
>  {<br>
>   /*<br>
> @@ -55,10 +61,17 @@ static void usart_set_baud(int minor, int baud)<br>
>   return;<br>
>  }<br>
>  <br>
> -static void usart_initialize(int minor)<br>
> +static volatile usart0_regs<br>
> *rpi_uart_get_regs(rtems_termios_device_context *base)<br>
>  {<br>
> -  unsigned int gpio_reg;<br>
> +  uart0_context *ctx;<br>
> +<br>
> +  ctx = (usart0_regs *) base;<br>
> +  return ctx->regs;<br>
> +}<br>
>  <br>
> +static void usart_initialize(rtems_termios_device_context *base)<br>
> +{<br>
> +  unsigned int gpio_reg;<br>
>    /*<br>
>    ** Program GPIO pins for UART 0<br>
>    */<br>
> @@ -75,67 +88,81 @@ static void usart_initialize(int minor)<br>
>    usart_delay(150);<br>
>    BCM2835_REG(BCM2835_GPIO_GPPUDCLK0) = 0;<br>
>  <br>
> +  volatile uint32_t *uart_regs = rpi_uart_get_regs(base);<br>
> +<br>
>    /*<br>
>    ** Init the PL011 UART<br>
>    */<br>
> -  BCM2835_REG(BCM2835_UART0_CR)   = 0;<br>
> -  BCM2835_REG(BCM2835_UART0_ICR)  = 0x7FF;<br>
> -  BCM2835_REG(BCM2835_UART0_IMSC) = 0;<br>
> -  BCM2835_REG(BCM2835_UART0_IBRD) = 1;<br>
> -  BCM2835_REG(BCM2835_UART0_FBRD) = 40;<br>
> -  BCM2835_REG(BCM2835_UART0_LCRH) = 0x70;<br>
> -  BCM2835_REG(BCM2835_UART0_RSRECR) =  0;<br>
> -<br>
> -  BCM2835_REG(BCM2835_UART0_CR)   = 0x301;<br>
> -<br>
> -  BCM2835_REG(BCM2835_UART0_IMSC) = BCM2835_UART0_IMSC_RX;<br>
> -<br>
> -  usart_set_baud(minor, 115000);<br>
> +  uart_regs->cr  = 0;<br>
> +  uart_regs->icr = 0x7ff;<br>
> +  uart_regs->imsc = 0;<br>
> +  uart_regs->ibrd = 1;<br>
> +  uart_regs->fbrd= 40;<br>
> +  uart_regs->lcrh= 0x70;<br>
> +  uart_regs->rsrecr= 0;<br>
> +  uart_regs->cr = 0x301;<br>
> +  uart_regs->imsc = BCM2835_UART0_IMSC_RX;<br>
> +  // usart_set_baud(minor, 115000);<br>
<br>
Why is this line commented now?<br></blockquote><div>It actually does nothing. The function body was empty. The current baud rate is set directly</div><div>in the initialization function. I was planning to update it once, I was finished with the interface. </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
>  }<br>
>  <br>
> -static int usart_first_open(int major, int minor, void *arg)<br>
> +static bool usart_first_open(<br>
> +  rtems_termios_tty *tty,<br>
> +  rtems_termios_device_context *base,<br>
> +  struct termios *term,<br>
> +  rtems_libio_open_close_args_t *args<br>
> +)<br>
>  {<br>
> -  rtems_libio_open_close_args_t *oc = (rtems_libio_open_close_args_t *)<br>
> arg;<br>
> -  struct rtems_termios_tty *tty = (struct rtems_termios_tty *)<br>
> oc->iop->data1;<br>
> -  const console_tbl *ct = Console_Port_Tbl [minor];<br>
> -  console_data *cd = &Console_Port_Data [minor];<br>
> +  rtems_status_code sc;<br>
> +  uart0_context *ctx;<br>
> +  bool ok;<br>
>  <br>
> -  cd->termios_data = tty;<br>
> -  rtems_termios_set_initial_baud(tty, ct->ulClock);<br>
> +  ctx = (uart0_context *) base;<br>
>  <br>
> -  return 0;<br>
> +  usart_initialize(base);<br>
> +<br>
> +  sc = rtems_termios_set_initial_baud(tty,  USART0_DEFAULT_BAUD);<br>
> +  if ( sc != RTEMS_SUCCESSFUL ){<br>
> +    printk("Error setting the baud for termios\n");<br>
> +    return false;<br>
> +  }<br>
<br>
There is a return missing here. Did you compile the code? The compiler<br>
should give you a warning about that. </blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
>  }<br>
>  <br>
> -static int usart_last_close(int major, int minor, void *arg)<br>
> +static int usart_last_close(<br>
<br>
The first_open returns a bool but last_close returns still an int? Is<br>
this correct? I don't have the interface memorized.<br></blockquote><div>last_close return type is void. </div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
> +  rtems_termios_tty *tty,<br>
> +  rtems_termios_device_context *base,<br>
> +  rtems_termios_open_close_args_t *arg)<br>
>  {<br>
>    return 0;<br>
>  }<br>
>  <br>
> -static int usart_read_polled(int minor)<br>
> +static int usart_read_polled(rtems_termios_device_context *base)<br>
>  {<br>
> -  if (minor == 0) {<br>
> -    if (((BCM2835_REG(BCM2835_UART0_FR)) & BCM2835_UART0_FR_RXFE) == 0) {<br>
> -       return((BCM2835_REG(BCM2835_UART0_DR)) & 0xFF );<br>
> -    } else {<br>
> -      return -1;<br>
> -    }<br>
> -  } else {<br>
> -    printk("Unknown console minor number: %d\n", minor);<br>
> -    return -1;<br>
> +  volatile usart0_regs *regs;<br>
> +<br>
> +  regs = rpi_uart_get_regs(base);<br>
<br>
Just noted that here: Why rpi_uart_get_regs and not usart_get_regs?<br>
Please use a consitent naming scheme.<br>
<br>
> +<br>
> +  if ((regs->fr & BCM2835_UART0_FR_RXFE) == 0) {<br>
> +    return (regs->dr & 0xFF);<br>
>    }<br>
> +<br>
> +  return -1;<br>
>  }<br>
>  <br>
> -static void usart_write_polled(int minor, char c)<br>
> +static void usart_write_polled(rtems_termios_device_context *base, char c)<br>
>  {<br>
> -   while (1) {<br>
> -     if ((BCM2835_REG(BCM2835_UART0_FR) & BCM2835_UART0_FR_TXFF) == 0)<br>
> -       break;<br>
> -   }<br>
> -   BCM2835_REG(BCM2835_UART0_DR) = c;<br>
> +  volatile usart0_regs *regs;<br>
> +<br>
> +  regs = rpi_uart_get_regs(base);<br>
> +<br>
> +  while (1) {<br>
> +    if (((regs->fr) & BCM2835_UART0_FR_TXFF) == 0)<br>
> +      break;<br>
> +  }<br>
> +  regs->dr = c;<br>
>  }<br>
>  <br>
>  static ssize_t usart_write_support_polled(<br>
> -  int minor,<br>
> +  rtems_termios_device_context *base,<br>
>    const char *s,<br>
>    size_t n<br>
>  )<br>
> @@ -143,7 +170,7 @@ static ssize_t usart_write_support_polled(<br>
>    ssize_t i = 0;<br>
>  <br>
>    for (i = 0; i < n; ++i) {<br>
> -    usart_write_polled(minor, s [i]);<br>
> +    usart_write_polled(base, s[i]);<br>
>    }<br>
>  <br>
>    return n;<br>
> @@ -154,14 +181,11 @@ static int usart_set_attributes(int minor, const<br>
> struct termios *term)<br>
>    return -1;<br>
>  }<br>
>  <br>
> -const console_fns bcm2835_usart_fns = {<br>
> -  .deviceProbe = libchip_serial_default_probe,<br>
> -  .deviceFirstOpen = usart_first_open,<br>
> -  .deviceLastClose = usart_last_close,<br>
> -  .deviceRead = usart_read_polled,<br>
> -  .deviceWrite = usart_write_support_polled,<br>
> -  .deviceInitialize = usart_initialize,<br>
> -  .deviceWritePolled = usart_write_polled,<br>
> -  .deviceSetAttributes = usart_set_attributes,<br>
> -  .deviceOutputUsesInterrupts = false<br>
> -};<br>
> +const rtems_termios_device_handler bcm2835_uart0_handler_polled = {<br>
> +  .first_open = usart_first_open,<br>
> +  .last_close = usart_last_close,<br>
> +  .poll_read = usart_read_polled,<br>
> +  .set_attributes = usart_set_attributes,<br>
> +  .write = usart_write_support_polled,<br>
> +  .mode = TERMIOS_POLLED<br>
> +}<br>
> \ No newline at end of file<br>
> <br>
> On Wed, Dec 25, 2019 at 12:36 AM Joel Sherrill <<a href="mailto:joel@rtems.org" target="_blank">joel@rtems.org</a><br>
> <mailto:<a href="mailto:joel@rtems.org" target="_blank">joel@rtems.org</a>>> wrote:<br>
> <br>
> <br>
> <br>
>     On Tue, Dec 24, 2019, 12:19 PM Niteesh <<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>     <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>> wrote:<br>
> <br>
>         And also the register definitions are in raspberrpi.h file<br>
>         should I move them to usart.h.<br>
> <br>
> <br>
>     Sounds right if you mean bsp/usart.h<br>
> <br>
>         I have a doubt we have a register field in device_context<br>
>         typedef struct { <br>
> <br>
>             rtems_termios_device_context base; <br>
> <br>
>             const char *device_name; <br>
> <br>
>             volatile some_chip_registers *regs;<br>
> <br>
>         } my_driver_context;<br>
> <br>
>         How does the reg field point to the correct memory location? for<br>
>         instance in IMX BSP,<br>
>         there is a struct with register field's but none of the define a<br>
>         memory location?<br>
> <br>
> <br>
>     Make sure the structure has volatiles and proper alignment. :)<br>
> <br>
> <br>
>         On Tue, Dec 24, 2019 at 11:37 PM Niteesh <<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>         <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>> wrote:<br>
> <br>
>             How to handle different serial devices? In other BSPs the<br>
>             uart devices are the same, so<br>
>             they were able to put it under a single array? But here we<br>
>             have 2 uarts and a FB?<br>
> <br>
> <br>
>             On Tue, Dec 24, 2019 at 8:18 PM Christian Mauderer<br>
>             <<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>> wrote:<br>
> <br>
>                 On 24/12/2019 12:06, Niteesh wrote:<br>
>                 > The current raspi console section is like this:<br>
>                 > The bsp_console_select in console_select.c is<br>
>                 responsible for selecting<br>
>                 > between uart and the framebuffer. It does so<br>
>                 > by setting the Console_port_minor.<br>
>                 > The console_config is responsible for output_char<br>
>                 function.<br>
>                 > And other files are driver code.<br>
>                 > If rewriting, this would be my approach,<br>
>                 > Rewrite the bsp_console_select to set some kind of a<br>
>                 variable like in<br>
>                 > IMX, then in console_initialize function<br>
>                 > link the right driver to /dev/console.<br>
>                 > Replace the console_tbl with the device_context and<br>
>                 console_fns with<br>
>                 > termios_device_handlers and<br>
>                 > finally add in the console_initialization function.<br>
> <br>
>                 I agree that this would be a clean solution. So if you<br>
>                 want you can do<br>
>                 that. But there might is a hurdle: As far as I<br>
>                 understood you you only<br>
>                 have a Pi3? So you might have a hard time testing the<br>
>                 changes. Maybe the<br>
>                 simulator could work.<br>
> <br>
>                 Another possibility could be to set the<br>
>                 "Console_port_minor" to<br>
>                 something unused (for example -1). In that case you can<br>
>                 define another<br>
>                 /dev/console.<br>
> <br>
>                 Best regards and merry Christmas (in case you celebrate)<br>
> <br>
>                 Christian<br>
> <br>
>                 ><br>
>                 > On Tue, Dec 24, 2019 at 2:13 PM Niteesh<br>
>                 <<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>><br>
>                 > <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>>><br>
>                 wrote:<br>
>                 ><br>
>                 >     Thank you so much, for such a detailed answer. Now<br>
>                 things make<br>
>                 >     really good sense to me, <br>
>                 >     going through the code now is just a breeze. But I<br>
>                 still have one<br>
>                 >     question<br>
>                 >     for the newer driver interface is<br>
>                 console_initialize the function<br>
>                 >     which RTEMS calls while initializing<br>
>                 >     the console? Which means I can't mess with the<br>
>                 name right? It is<br>
>                 >     similar to the main function, right?<br>
>                 ><br>
>                 >     The current driver is a legacy one, how do you<br>
>                 want me to proceed,<br>
>                 >     shall I rewrite the legacy to a<br>
>                 >     the new one, this is will be a great<br>
>                 learning experience for me also<br>
>                 >     and we also get the BSP updated to the latest<br>
>                 interface.<br>
>                 ><br>
>                 ><br>
>                 >     On Tue, Dec 24, 2019 at 3:20 AM Christian Mauderer<br>
>                 >     <<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>>><br>
>                 wrote:<br>
>                 ><br>
>                 >         Hello Niteesh,<br>
>                 ><br>
>                 >         quite a lot of questions. I'll try to answer<br>
>                 them. Note that it<br>
>                 >         has been<br>
>                 >         some time since I had a detailed look at that<br>
>                 code so if something I<br>
>                 >         tell seems odd please don't hesitate to<br>
>                 question it.<br>
>                 ><br>
>                 >         Please note that in RTEMS their are more or<br>
>                 less two "levels" of<br>
>                 >         support<br>
>                 >         for a serial console:<br>
>                 ><br>
>                 >         1. A very basic polled system console (also<br>
>                 known as<br>
>                 >         "debug-console" in<br>
>                 >         some BSPs). This one is used for printk and<br>
>                 should work in basically<br>
>                 >         every case. It is used for critical system<br>
>                 messages like<br>
>                 >         printing the<br>
>                 >         exception frame. For that a BSP has to provide a<br>
>                 >         "BSP_output_char" function.<br>
>                 ><br>
>                 >         2. A full featured UART driver integrated into<br>
>                 Termios. That one<br>
>                 >         will be<br>
>                 >         used for all normal I/O on the UARTs.<br>
>                 ><br>
>                 >         As far as I know the "console_tbl<br>
>                 Console_Configuration_Ports"<br>
>                 >         belongs<br>
>                 >         to a table based legacy interface. It is<br>
>                 handled in the file<br>
>                 >         bsps/shared/dev/serial/legacy-console.c. I'm<br>
>                 not sure whether it is<br>
>                 >         documented in the BSP guide because it<br>
>                 shouldn't be used for new<br>
>                 >         BSPs.<br>
>                 >         Same is true for the "major" and "minor"<br>
>                 stuff: It's not really<br>
>                 >         used for<br>
>                 >         new drivers.<br>
>                 ><br>
>                 >         Newer drivers use the initialization that is<br>
>                 described in the manual<br>
>                 >         that you have already found. Basically they use<br>
>                 >         "rtems_termios_device_install" to register a<br>
>                 new UART as<br>
>                 >         "/dev/ttySomething". Some recent (ARM) BSPs<br>
>                 that do that are the<br>
>                 >         imx or<br>
>                 >         the atsam.<br>
>                 ><br>
>                 >         The console that is used for stdin, stdout and<br>
>                 stderr (printf,<br>
>                 >         scanf,<br>
>                 >         ...) is the one called "/dev/console" (defined in<br>
>                 >         CONSOLE_DEVICE_NAME).<br>
>                 >         For the legacy table based interface it's the<br>
>                 one with the index of<br>
>                 >         "Console_Port_Minor".<br>
>                 ><br>
>                 ><br>
>                 >         If you want to access any UART other than the<br>
>                 one for stdin and<br>
>                 >         stdout<br>
>                 >         you do that the same way like on Linux: Just<br>
>                 use the "open"<br>
>                 >         function on<br>
>                 >         the "/dev/ttySomething" and use "read",<br>
>                 "write" and simmilar or use<br>
>                 >         "fopen" together with "fread", "fwrite",<br>
>                 "fprintf", ...<br>
>                 ><br>
>                 ><br>
>                 >         "printf" (and family) is a function belonging<br>
>                 to the C library.<br>
>                 >         In our<br>
>                 >         case that's newlib. It will format your<br>
>                 message and after some other<br>
>                 >         preprocessing will call the "write" function<br>
>                 of the file that is<br>
>                 >         opened<br>
>                 >         as stdout (which is "/dev/console" in the<br>
>                 default case).<br>
>                 ><br>
>                 ><br>
>                 >         I hope that I helped you with that<br>
>                 explanation. Please feel free<br>
>                 >         to ask<br>
>                 >         anything if it isn't clear.<br>
>                 ><br>
>                 >         Best regards<br>
>                 ><br>
>                 >         Christian<br>
>                 ><br>
>                 >         On 23/12/2019 19:50, Niteesh wrote:<br>
>                 >         > And finally, how does printf work? It is a<br>
>                 macro? In that<br>
>                 >         case, how does<br>
>                 >         > any write to<br>
>                 >         > a console work?<br>
>                 >         ><br>
>                 >         > On Tue, Dec 24, 2019 at 12:18 AM Niteesh<br>
>                 <<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>><br>
>                 >         <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>><br>
>                 >         > <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>>>> wrote:<br>
>                 >         ><br>
>                 >         >     Is the correct port minor number set<br>
>                 during the<br>
>                 >         initialization? What<br>
>                 >         >     is the application want's to<br>
>                 >         >     access some other port?<br>
>                 >         ><br>
>                 >         >     On Tue, Dec 24, 2019 at 12:16 AM Niteesh<br>
>                 >         <<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>><br>
>                 >         >     <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>>>> wrote:<br>
>                 >         ><br>
>                 >         >         I would like to clarify my doubts<br>
>                 regarding the<br>
>                 >         console driver.<br>
>                 >         >         I went through the documentation<br>
>                 >         >         for the console<br>
>                 >         >       <br>
>                 >       <br>
>                   driver <a href="https://docs.rtems.org/branches/master/bsp-howto/console.html#introduction" rel="noreferrer" target="_blank">https://docs.rtems.org/branches/master/bsp-howto/console.html#introduction</a>.<br>
>                 >         >         But it is quite different from how<br>
>                 some BSPs initialize.<br>
>                 >         >         Correct me if I am wrong<br>
>                 >         >         The console_tbl contains the various<br>
>                 entries of serial<br>
>                 >         ports.<br>
>                 >         >         The console_fns is a struct of<br>
>                 function pointers,<br>
>                 >         which point to<br>
>                 >         >         the BSP uart functions.<br>
>                 >         >         The BSP_output_char_function_type is<br>
>                 what will be<br>
>                 >         called for<br>
>                 >         >         printing a char on to the console.<br>
>                 >         >         How does RTEMS initialize the uart?<br>
>                 It's seems not to<br>
>                 >         be same<br>
>                 >         >         for all BSPs.<br>
>                 >         >         The doc says that the driver's<br>
>                 initialization function<br>
>                 >         is called<br>
>                 >         >         once during the rtems initialization<br>
>                 process.<br>
>                 >         >         The console init function install<br>
>                 the serial driver using<br>
>                 >         >         rtems_termios_device_install but<br>
>                 there seems to be<br>
>                 >         >         no such function in the raspberry<br>
>                 pi? But there is a<br>
>                 >         entry in<br>
>                 >         >         console_fns for init function, but<br>
>                 then how does it<br>
>                 >         >         gets called?<br>
>                 >         >         And for BSP's with multiple<br>
>                 serial's, the output function<br>
>                 >         >         chooses the right serial using<br>
>                 console_port_minor,<br>
>                 >         >         Is it during initialization?<br>
>                 >         >         What is the need for get and set<br>
>                 register functions? <br>
>                 >         ><br>
>                 >         >         On Mon, Dec 23, 2019 at 1:04 AM<br>
>                 Christian Mauderer<br>
>                 >         >         <<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>><br>
>                 >         <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>>>> wrote:<br>
>                 >         ><br>
>                 >         >             On 22/12/2019 19:45, Joel<br>
>                 Sherrill wrote:<br>
>                 >         >             ><br>
>                 >         >             ><br>
>                 >         >             > On Sun, Dec 22, 2019, 12:29 PM<br>
>                 Niteesh<br>
>                 >         <<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>><br>
>                 >         >             <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>>><br>
>                 >         >             > <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>><br>
>                 >         <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>> <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>><br>
>                 >         <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a><br>
>                 <mailto:<a href="mailto:gsnb.gn@gmail.com" target="_blank">gsnb.gn@gmail.com</a>>>>>> wrote:<br>
>                 >         >             ><br>
>                 >         >             >     On Sun, Dec 22, 2019 at<br>
>                 8:44 PM Christian<br>
>                 >         Mauderer<br>
>                 >         >             >     <<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>><br>
>                 >         <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>><br>
>                 >         <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>>><br>
>                 >         >             <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>><br>
>                 >         <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>> <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>><br>
>                 >         <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a><br>
>                 <mailto:<a href="mailto:list@c-mauderer.de" target="_blank">list@c-mauderer.de</a>>>>>> wrote:<br>
>                 >         >             ><br>
>                 >         >             >         Hello Niteesh,<br>
>                 >         >             ><br>
>                 >         >             >         thanks for doing that<br>
>                 work.<br>
>                 >         >             ><br>
>                 >         >             >         On 22/12/2019 12:10,<br>
>                 Niteesh wrote:<br>
>                 >         >             >         > The rpi1 and rpi2<br>
>                 use the PL011 UART,<br>
>                 >         whereas,<br>
>                 >         >             with RPI's<br>
>                 >         >             >         equipped with<br>
>                 >         >             >         > wireless/Bluetooth<br>
>                 module, the PL011 is<br>
>                 >         >             connected to the Bluetooth<br>
>                 >         >             >         > module, and the mini<br>
>                 UART is used as<br>
>                 >         the primary<br>
>                 >         >             UART.<br>
>                 >         >             ><br>
>                 >         >             >         In my opinion it would<br>
>                 be great if you<br>
>                 >         could use<br>
>                 >         >             the FDT to<br>
>                 >         >             >         distinguish<br>
>                 >         >             >         between the boards.<br>
>                 That should allow to add<br>
>                 >         >             raspberry 3 (and<br>
>                 >         >             >         maybe 4)<br>
>                 >         >             >         support without adding<br>
>                 another BSP. More<br>
>                 >         BSPs mean<br>
>                 >         >             a bigger<br>
>                 >         >             >         maintenance<br>
>                 >         >             >         effort for the RTEMS<br>
>                 community.<br>
>                 >         >             ><br>
>                 >         >             >     Learning more about FDT is<br>
>                 on my list for a long<br>
>                 >         >             time.  I would love<br>
>                 >         >             >     to work on that<br>
>                 >         >             >     but I have almost no exp<br>
>                 with FDT's.<br>
>                 >         >             >     But another thing could<br>
>                 also be done, in<br>
>                 >         >             >   <br>
>                  raspberrypi/start/bspstart.c we get the<br>
>                 >         revision and<br>
>                 >         >             >     model of the board using<br>
>                 the mailbox. Every<br>
>                 >         board has<br>
>                 >         >             a unique id,<br>
>                 >         >             >     which we could use to<br>
>                 initialize<br>
>                 >         >             >     the BSP. But using FDT<br>
>                 seems to be a more<br>
>                 >         elegant<br>
>                 >         >             option, it is a<br>
>                 >         >             >     lot of work I think, but<br>
>                 we could take<br>
>                 >         >             >     help from libbsd and linux<br>
>                 I suppose. What<br>
>                 >         do you think?<br>
>                 >         >             ><br>
>                 >         >             ><br>
>                 >         >             > I think there are almost<br>
>                 always two steps to a<br>
>                 >         project<br>
>                 >         >             like this: get it<br>
>                 >         >             > to work and make it nice. :)<br>
>                 >         >             ><br>
>                 >         >             > If you fix the startup code to<br>
>                 read the board<br>
>                 >         revision and<br>
>                 >         >             memory size,<br>
>                 >         >             > you can get a working BSP that<br>
>                 dynamically<br>
>                 >         adapts to the<br>
>                 >         >             models and<br>
>                 >         >             > memory variations with minimal<br>
>                 modifications. If<br>
>                 >         you want<br>
>                 >         >             to then<br>
>                 >         >             > convert the BSP to FDT, it<br>
>                 will be a LOT easier<br>
>                 >         to debug<br>
>                 >         >             with a working BSP.<br>
>                 >         >             ><br>
>                 >         >             > Plus you may be able to<br>
>                 identify every variation<br>
>                 >         point<br>
>                 >         >             based on just the<br>
>                 >         >             > model info. Then FDT is just a<br>
>                 matter of<br>
>                 >         switching the<br>
>                 >         >             source of<br>
>                 >         >             > some/all of the info.<br>
>                 >         >             ><br>
>                 >         >             > That would be my work plan anyway.<br>
>                 >         ><br>
>                 >         >             I agree with Joel that a secure<br>
>                 development basis<br>
>                 >         (also<br>
>                 >         >             known as "hack")<br>
>                 >         >             as a first step is a good idea.<br>
>                 You maybe even<br>
>                 >         just make the<br>
>                 >         >             mini UART<br>
>                 >         >             the default driver while you are<br>
>                 developing. Then<br>
>                 >         you can be<br>
>                 >         >             sure that<br>
>                 >         >             you have the right driver.<br>
>                 >         ><br>
>                 >         >             As soon as that works you can<br>
>                 either change to the<br>
>                 >         revision<br>
>                 >         >             method or<br>
>                 >         >             (better) to the FDT one and<br>
>                 after that the patches<br>
>                 >         can be<br>
>                 >         >             merged. Using<br>
>                 >         >             the FDT isn't that complicated.<br>
>                 Basically you<br>
>                 >         search for a<br>
>                 >         >             node based on<br>
>                 >         >             different parameters. For an<br>
>                 example you can take<br>
>                 >         a look at<br>
>                 >         >             the imx BSP.<br>
>                 >         >             In imx_uart_probe<br>
>                 >         (bsps/arm/imx/console/console-config.c) a<br>
>                 >         >             fdt node is<br>
>                 >         >             searched and based on that a<br>
>                 UART driver is used.<br>
>                 >         But again:<br>
>                 >         >             Follow<br>
>                 >         >             Joels suggestion to start simple<br>
>                 and secure.<br>
>                 >         ><br>
>                 >         >             ><br>
>                 >         >             >         ><br>
>                 >         >             >         ><br>
>                 >         >             >       <br>
>                 >         >           <br>
>                 >       <br>
>                    <a href="https://www.raspberrypi.org/app/uploads/2012/02/BCM2835-ARM-Peripherals.pdf" rel="noreferrer" target="_blank">https://www.raspberrypi.org/app/uploads/2012/02/BCM2835-ARM-Peripherals.pdf</a><br>
>                 >         >             >         > But from the above<br>
>                 doc (PAGE 10), the<br>
>                 >         mini uart<br>
>                 >         >             has 16550 like<br>
>                 >         >             >         registers<br>
>                 >         >             >         > and RTEMS already<br>
>                 has the driver for it<br>
>                 >         >             >         ><br>
>                 bsps/shared/dev/serial/ns16550.c. But<br>
>                 >         I am not<br>
>                 >         >             sure how<br>
>                 >         >             >         compatible they<br>
>                 >         >             >         > are? Should a new<br>
>                 driver be<br>
>                 >         implemented from<br>
>                 >         >             scratch or use<br>
>                 >         >             >         ns16550 if<br>
>                 >         >             >         > possible?<br>
>                 >         >             ><br>
>                 >         >             >         In general it's better<br>
>                 to re-use<br>
>                 >         existing code.<br>
>                 >         >             That has multiple<br>
>                 >         >             >         advantages:<br>
>                 >         >             ><br>
>                 >         >             >         - It reduces the<br>
>                 maintenance effort.<br>
>                 >         Fewer code<br>
>                 >         >             means fewer work.<br>
>                 >         >             >         - If you have multiple<br>
>                 driver for the<br>
>                 >         same or<br>
>                 >         >             similar hardware<br>
>                 >         >             >         it can<br>
>                 >         >             >         happen that a bug is<br>
>                 fixed in one but<br>
>                 >         not the other.<br>
>                 >         >             >         - It's simpler to find<br>
>                 a hardware to<br>
>                 >         test changes.<br>
>                 >         >             >         - The driver becomes<br>
>                 more universal with<br>
>                 >         every new<br>
>                 >         >             supported<br>
>                 >         >             >         hardware.<br>
>                 >         >             >         That increases the<br>
>                 chance that it fits<br>
>                 >         the next<br>
>                 >         >             new hardware.<br>
>                 >         >             ><br>
>                 >         >             >         I'm sure there are<br>
>                 some more if you ask<br>
>                 >         someone else.<br>
>                 >         >             ><br>
>                 >         >             >     I do understand the<br>
>                 issues, I just spent<br>
>                 >         some time<br>
>                 >         >             reading the<br>
>                 >         >             >     driver code.<br>
>                 >         >             >     I think we could most<br>
>                 probably use it. I<br>
>                 >         will take a<br>
>                 >         >             closer look and<br>
>                 >         >             >     will update.<br>
>                 >         >             ><br>
>                 >         ><br>
>                 >         >             Great.<br>
>                 >         ><br>
>                 >         >             >          <br>
>                 >         >             ><br>
>                 >         >             >         ><br>
>                 >         >             >         > Also, the core clock<br>
>                 on which the<br>
>                 >         PL011 is based<br>
>                 >         >             on is changed<br>
>                 >         >             >         in rpi3.<br>
>                 >         >             >         > Rpi1 and 2 use<br>
>                 250Mhz as the default<br>
>                 >         clock but<br>
>                 >         >             it was changed<br>
>                 >         >             >         to 400Mhz<br>
>                 >         >             >         > in Rpi3 and newer<br>
>                 >         >             ><br>
>                 >         >             >         Again: Would be great<br>
>                 if that could be<br>
>                 >         adapted<br>
>                 >         >             based on FDT or by<br>
>                 >         >             >         reading the right<br>
>                 registers.<br>
>                 >         >             ><br>
>                 >         >             >         ><br>
>                 >         >             >         > Few differences<br>
>                 between PL011 and Mini<br>
>                 >         uart<br>
>                 >         >             >         > The mini UART has<br>
>                 smaller FIFOs.<br>
>                 >         Combined with<br>
>                 >         >             the lack of<br>
>                 >         >             >         flow control,<br>
>                 >         >             >         > this makes it more<br>
>                 prone to losing<br>
>                 >         characters at<br>
>                 >         >             higher baud<br>
>                 >         >             >         rates. It<br>
>                 >         >             >         > is also generally<br>
>                 less capable than<br>
>                 >         the PL011,<br>
>                 >         >             mainly due to<br>
>                 >         >             >         its baud<br>
>                 >         >             >         > rate link to the VPU<br>
>                 clock speed.<br>
>                 >         >             ><br>
>                 >         >             >         That shouldn't really<br>
>                 be a problem for<br>
>                 >         the system<br>
>                 >         >             console.<br>
>                 >         >             ><br>
>                 >         >             >         ><br>
>                 >         >             >         > The particular<br>
>                 deficiencies of the<br>
>                 >         mini UART<br>
>                 >         >             compared to the<br>
>                 >         >             >         PL011 are :<br>
>                 >         >             >         ><br>
>                 >         >             >         > No break detection<br>
>                 >         >             >         > No framing errors<br>
>                 detection<br>
>                 >         >             >         > No parity bit<br>
>                 >         >             >         > No receive timeout<br>
>                 interrupt<br>
>                 >         >             >         > No DCD, DSR, DTR or<br>
>                 RI signals<br>
>                 >         >             >         ><br>
>                 >         >             ><br>
>                 >         >             >   <br>
>                  _______________________________________________<br>
>                 >         >             >     devel mailing list<br>
>                 >         >             >     <a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>> <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>>><br>
>                 >         <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>> <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>>>><br>
>                 >         >             <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>> <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>>><br>
>                 >         <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>> <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a><br>
>                 <mailto:<a href="mailto:devel@rtems.org" target="_blank">devel@rtems.org</a>>>>><br>
>                 >         >             >   <br>
>                  <a href="http://lists.rtems.org/mailman/listinfo/devel" rel="noreferrer" target="_blank">http://lists.rtems.org/mailman/listinfo/devel</a><br>
>                 >         >             ><br>
>                 >         ><br>
>                 ><br>
> <br>
</blockquote></div></div>