Re: [Qemu-devel] [PATCH 06/10] r2d: Flash memory creation is confused about size, mark FIXME

[Markus Armbruster]

Philippe Mathieu-Daudé <address@hidden> writes:

> On 2/19/19 4:45 PM, Markus Armbruster wrote:
>> Peter Maydell <address@hidden> writes:
>> 
>>> On Mon, 18 Feb 2019 at 13:07, Markus Armbruster <address@hidden> wrote:
>>>>
>>>> pflash_cfi02_register() takes a size in bytes, a block size in bytes
>>>> and a number of blocks.  r2d_init() passes FLASH_SIZE, 16 * KiB,
>>>> FLASH_SIZE >> 16.  Does not compute: size doesn't match block size *
>>>> number of blocks.  The latter happens to win.  I tried to find
>>>> documentation on the physcial hardware, no luck.
>
> "physical"

Thanks, will fix.

>>>>
>>>> For now, adjust the byte size passed to match the actual size created,
>>>> and add a FIXME comment.
>>>
>>> I'm pretty sure that FLASH_SIZE here is supposed to be a
>>> byte count of the size of the pflash. That matches what
>>> Linux has in arch/sh/boards/mach-r2d/setup.c where it
>>> sets up the flash_resource struct.
>> 
>> Okay, that's some evidence for size 0x02000000 (32MiB).
>> 
>> However, we've created size (16 * KiB) * (FLASH_SIZE >> 16), i.e. 8MiB,
>> since at least commit 368a354f02b (v1.3.0), possibly since forever.
>> 
>>> The r2dplus board is also I think known as RTS7751R2D. That
>>> takes us to https://elinux.org/RTS7751R2D_Handling_Manual
>>> (sadly the link to the "hardware manual" is broken).
>> 
>> Quote section Flash ROM Mapping:
>> 
>>     Currently, MTD device mapping on Flash ROM is set as below.
>>     0x00000000-0x00020000    "bootloader"    
>>     0x00020000-0x00320000    "mtdblock1"     XIP kernel
>>     0x00320000-0x00520000    "mtdblock2"     
>>     0x00520000-0x01000000    "mtdblock3"     
>> 
>> Suggests a size of 0x01000000 (16MiB).  Now we have three candidates.
>> 
>> Pick one, any one, and I'll adjust my patch.  All I really care about is
>> getting argument @size consistent with arguments @sector_len and
>> @nb_blocs, so I can ditch @nb_blocs in PATCH 09.
>> 
>>> No idea what the block size would be.
>> 
>> As long as that's the case, inertia wins by default.
>
> There is also a paper [*]:
>
>   The Renesas Technology R0P751RLC001RL (R2DPLUS) board was used
>   as our target device.
>   This board is often used to evaluate software for CE devices.
>   The specification is shown below.
>     CPU: SH7751R(SH4) 240Mhz
>     RAM: 64Mbyte
>     Compact flash: 512Mbyte
>     Flash ROM: 64Mbyte (32Mbyte available for root file system)

Candidate #4: 64MiB.  Bring 'em on!

>
> Let's use at least 16MB to be able to run the elinux cited kernel.
>
> [*] https://www.kernel.org/doc/ols/2008/ols2008v2-pages-125-134.pdf

That's a vote for changing the status quo (8 MiB).

Perhaps Magnus, who maintains the machine, can pick a new value for us.