Re: [PATCH 0/2] arm: Add collie and sx functional tests

["Philippe Mathieu-Daudé" <philmd@linaro.org>]

Cc'ing Jan.

On 22/10/24 12:04, Guenter Roeck wrote:
> On 10/21/24 21:09, Philippe Mathieu-Daudé wrote:
>> Hi Guenter,
>>
>> On 21/10/24 11:02, Guenter Roeck wrote:
>>
>>> Unrelated to this, but I found that the sd emulation in 9.1 is also 
>>> broken
>>> for loongarch and sifive_u, and partially for ast2600-evb (it has two
>>> controllers, with one of them no longer working). That is too much 
>>> for me
>>> to track down quickly, so this is just a heads-up.
>>
>> Please Cc me with reproducer or assign Gitlab issues to me,
>> I'll have a look.
>>
> 
> If it wasn't funny, it would be sad.
> 
> hw/sd/sd.c:sd_reset() calls sd_bootpart_offset() t determine the boot 
> partition
> offset. That function needs to have sd->ext_csd[EXT_CSD_PART_CONFIG] 
> configured.
> However, the value is only set later in sd_reset() with the call to 
> sc->set_csd().
> One of the parameters of that function is the previously calculated size.
> 
> So in other words there is a circular dependency. The call to 
> sd_bootpart_offset()
> returns 0 because sd->ext_csd[EXT_CSD_PART_CONFIG] isn't set, then
> the call to sc->set_csd() sets it ... too late. Subsequent calls to
> sd_bootpart_offset() will then return the expected offset.
> 
> I have no idea how this is supposed to work, and I don't think it works
> as implemented. I'll revert commit e32ac563b83 in my local copy of qemu.

Jan, can you have a look at this bug report please? Otherwise I'll
have a look during soft freeze.

Regards,

Phil.