DT vs ARM static mappings

Rob Herring robherring2 at gmail.com
Tue Sep 20 08:58:58 EDT 2011


Pawel,

On 09/20/2011 06:51 AM, Pawel Moll wrote:
> Hi All,
> 
> Apologies for the length of the text below, but the problem is
> complex... (or at least it seems to be)
> 
> While working on DT support for ARM Versatile Express I faced an
> interesting problem... Some background first, to set a common ground and
> test my understanding of the problem ;-)
> 
> ARM machine description contains a "map_io" method, which is used to
> create static memory mappings (using iotable_init() function) for things
> like peripherals or SRAMs. At least that's the theory, because most of
> the platforms are doing much more stuff there, like clocking/GPIOs/UARTs
> initialization, hardware probing etc.
> 
> Now the Versatile Express platform: it consists of a motherboard with a
> set of peripherals and a processor daughterboard (core tile), both
> connected via a static memory bus, which is mapped into processor's
> physical address space. The motherboard can be shared between different
> types of the tiles (eg. A9, A5, A15 etc.). The present one is probed byt
> he motherboard firmware and exposed in "system registers".
> 
> Everything is fine so far. The interesting part starts here:
> 
> http://infocenter.arm.com/help/topic/com.arm.doc.dui0447e/CACIHGFE.html
> 
> In brief, depending on the configuration, the system can have one of
> two, totally different, memory maps (please, do spare me the "but why
> did they do that" comments - not my idea nor decision, I just have to
> live with this ;-), depending on the core tile being used.
> 
> In result, the static mapping as defined currently in
> arch/arm/mach-vexpress/v2m.c for A9 variant:
> 
> #define V2M_PA_CS7      0x10000000
> 
> static struct map_desc v2m_io_desc[] __initdata = {
>         {
>                 .virtual        = __MMIO_P2V(V2M_PA_CS7),
>                 .pfn            = __phys_to_pfn(V2M_PA_CS7),
>                 .length         = SZ_128K,
>                 .type           = MT_DEVICE,
>         },
> };
> 
> is no longer valid for A5/A15. It would rather look like this:
> 
> #define V2M_PA_CS3      0x1c000000
> 
> static struct map_desc v2m_io_desc[] __initdata = {
>         {
>                 .virtual        = __MMIO_P2V(V2M_PA_CS3),
>                 .pfn            = __phys_to_pfn(V2M_PA_CS3),
>                 .length         = SZ_2M,
>                 .type           = MT_DEVICE,
>         },
> };
> 
> Not only the peripherals base address is changed but also "internal"
> alignment, thus offsets to peripherals. Some of them are not being
> ioremap()ed, but directly used via the static mapping and MMIO_P2V macro
> (like "readl(MMIO_P2V(V2M_SYS_PROCID0))" in v2m_populate_ct_desc(void)
> function). For example, these two:
> 
> #define V2M_SYSREGS             (V2M_PA_CS7 + 0x00000000)
> #define V2M_SYSCTL              (V2M_PA_CS7 + 0x00001000)
> 
> would have to become:
> 
> #define V2M_SYSREGS             (V2M_PA_CS3 + 0x00010000)
> #define V2M_SYSCTL              (V2M_PA_CS3 + 0x00020000)
> 
> My current DTS for the original memory map looks like that (fragments):
> 
> {
>         motherboard {
>                 ranges = <7 0 0x10000000 0x00020000>;
>                 #address-cells = <2>; // SMB chipselect number and offset
>                 #size-cells = <1>;
> 	
>                 peripherals at 7 {
> 			ranges = <0 7 0 0x20000>;
>                         #address-cells = <1>;
>                         #size-cells = <1>;
> 
>                         sysreg at 00000 {
>                                 reg = <0x00000 0x1000>;
>                         };      
>                         
>                         sysctl at 01000 {
>                                 reg = <0x01000 0x1000>;
>                         };
> 		}
> 	}		
> }
> 		
> DTS for the second memory map would just have the addresses changed.
> Unfortunately, because of the static mappings being hardcoded in the
> board support file, one kernel won't Just Work (TM) with different DTBs.
> 
> Of course the simplest solution would be to define two different
> compatible values, eg. "arm,vexpress-legacy" would execute the current
> map_io implementation, while "arm,vexpress-rs1" would use different one,
> setting up the other map_desc (the MMIO_P2V macro must die of course,
> replaced with a runtime-defined virtual base address for the
> peripherals).
> 
> If you believe that's what I should do, say it and stop reading :-)
> 

Yes. Different tiles are fundamentally different boards, so they should
have different DTs. Using includes should help minimize duplication though.

Think about it this way. How would you solve this without DT? You would
have a bunch of duplicated data in the kernel for the different configs.
So you're not any worse off in this regard and still have the other
advantages of DT.

> To my mind it looked like the whole mechanism was not flexible enough,
> so I wanted to explore other options...
> 
> The obvious one was to describe the required static mapping in the DTS.
> I don't like this idea, though. It can hardly be called "hardware
> description". Besides, what node would carry such data? "chosen"?
> Hardly...
> 
> Would it contain a "regs" property with the physical address and
> "virtual-reg" with the virtual one? Again, doesn't sound right to me
> (especially the virtual bit, however the virtual address could be common
> between different variants and be defined in the board support code, not
> the DTS).
> 
> I have considered a reference (phandle or an alias?) to the node to be
> mapped ("peripherals" in my case), but where to define this reference?
> Any ideas?

In "chosen" like the kernel command line would be the place, but I don't
think that is the right approach. Chosen is really for things that
change frequently and this doesn't really fall in that category.

> 
> There is an additional problem here... The "map_io" is executed before
> the tree is un-flattened, so:
> 
> 1. One can't simply use "of_find_matching_node()" (as in the latest l2x0
> patches) to find the interesting nodes - the only way of going through
> the tree is raw of_scan_flat_dt() function. Therefore any conditions
> more complex then string comparison with the (full) node name are
> problematic.
> 
> 2. The tree mappings (ranges) are not resolved yet, so one can't simply
> get the effective address of a node. Only "raw" properties are
> available, so all one can get scanning for "peripherals at 7" node is "0 7
> 0 0x20000" array, instead of the "0x10000000 0x00020000" that is really
> important.
> 

If you add a compatible field to "motherboard" node, then you can read
the ranges.

> Initially I wanted to find the mentioned devices and create individual
> mappings for them, so the MMIO_P2V would be still valid (if slightly
> "abused"), but I failed due to the problems mentioned above. And I can't
> delay this operation till the tree is un-flattened, as the core tile
> must be probed (via sysreg) in map_io (tile's specific code must be able
> to create its own mappings):

Do you really need MMIO_P2V? If you have fixed virtual addresses in the
kernel and can pull the phys addresses from DT to populate the iotable,
is that sufficient?

> 
> static void __init v2m_map_io(void)
> {       
>         iotable_init(v2m_io_desc, ARRAY_SIZE(v2m_io_desc)); 
>         v2m_populate_ct_desc();
>         ct_desc->map_io();
> }
> 
> Any comments, ideas and suggestions how to tackle this situation are
> more than welcome.
> 
Generally, the trend is to get rid of static mappings as much as
possible. Doing that first might simplify things.

Rob

> Cheers!
> 
> Paweł
> 
> 
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel




More information about the linux-arm-kernel mailing list