[6-freebsd-12 PATCH 1/2] rtemsbsd/bus: Add PCI support to the nexus bus

[Sebastian Huber]

On 16/02/2021 03:02, chrisj at rtems.org wrote:
> +/*
> + * Values for the bus space tag, not to be used directly by MI code.
> + */
> +#define	BSP_BUS_SPACE_IO	0	/* space is i/o space */
> +#define	BSP_BUS_SPACE_MEM	1	/* space is mem space */
> +
> +/*
> + * BSP PCI Support
> + *
> + * The RTEMS Nexus bus support can optionaly support PCI spaces that
> + * mapped to BSP speciic address spaces. Add the following define to
> + * the BSP header file to enable this support:
> + *
> + *  #define BSP_HAS_PCI
> + *
> + * If enabled a BSP must the following IO region calls:
> + *
> + * inb  : read 8 bits
> + * outb : write 8 bits
> + * inw  : read 16 bits
> + * outw : write 16 bits
> + * inl  : read 32 bits
> + * outl : write 32 bits
> + *
> + * The BSP needs to provide the DRAM address space offset
> + * PCI_DRAM_OFFSET. This is the base address of the DRAM as seen by a
> + * PCI master.
Why is it not possible to account for this offset in the 
bus_space_handle_t bsh? I thought the purpose of the bus_space_tag_t bst 
was to select different instructions to access the memory space. The 
generic file is for systems with memory mapped access only.
> + *
> + * i386 BSPs have a special bus.h file and do not use this file.
> + */
> +#include <bsp.h>
> +
>   /*
>    * Bus address alignment.
>    */
> @@ -144,6 +180,7 @@
>   /*
>    * Bus access.
>    */
> +#define BUS_SPACE_INVALID_DATA	(~0)
>   #define BUS_SPACE_UNRESTRICTED	(~0U)
>   
>   /*
> @@ -228,29 +265,52 @@ bus_space_barrier(bus_space_tag_t bst __unused, bus_space_handle_t bsh, bus_size
>    * data is returned.
>    */
>   static __inline uint8_t
> -bus_space_read_1(bus_space_tag_t bst __unused, bus_space_handle_t bsh, bus_size_t ofs)
> +bus_space_read_1(bus_space_tag_t bst, bus_space_handle_t bsh, bus_size_t ofs)
>   {
> +	if (bst == BSP_BUS_SPACE_IO) {
> +#ifdef BSP_HAS_PCI
> +		return inb(bsh + ofs);
> +#else
> +		return BUS_SPACE_INVALID_DATA;
> +#endif
> +	}
This adds a conditional to all memory mapped generic architectures. From 
my point of view this is a potential code size and performance 
regression. It also includes <bsp.h> and thus <rtems.h> in a file which 
is included all over the place.
>   	uint8_t __volatile *bsp = (uint8_t __volatile *)(bsh + ofs);
>   	return (*bsp);
>   }

-- 
embedded brains GmbH
Herr Sebastian HUBER
Dornierstr. 4
82178 Puchheim
Germany
email: sebastian.huber at embedded-brains.de
phone: +49-89-18 94 741 - 16
fax:   +49-89-18 94 741 - 08

Registergericht: Amtsgericht München
Registernummer: HRB 157899
Vertretungsberechtigte Geschäftsführer: Peter Rasmussen, Thomas Dörfler
Unsere Datenschutzerklärung finden Sie hier:
https://embedded-brains.de/datenschutzerklaerung/