Re: [PATCH v2 00/48] ARM: PXA multiplatform support

[Guenter Roeck <linux@roeck-us.net>]

On 4/23/22 12:55, Arnd Bergmann wrote:
> On Sat, Apr 23, 2022 at 1:41 AM Guenter Roeck <linux@roeck-us.net> wrote:
>>
>> On Sat, Apr 23, 2022 at 12:04:31AM +0200, Arnd Bergmann wrote:
>>> On Fri, Apr 22, 2022 at 10:55 PM Guenter Roeck <linux@roeck-us.net> wrote:
>>>> On 4/22/22 12:16, Arnd Bergmann wrote:
>>>>> On Fri, Apr 22, 2022 at 7:05 PM Guenter Roeck <linux@roeck-us.net> wrote:
>>>>>
>>>>> Which machine did you hit this on? Is this on hardware or in qemu?
>>>>>
>>>> qemu, as always. borzoi, spitz, terrier, tosa, z2, and sx1 fail.
>>>> Also, I just noticed that the failure is not always the same.
>>>> z2 fails to boot from initrd, and sx1 fails to boot completely.
>>>
>>> That's a lot of machines failing, I hope at least we got the same bugs more
>>> than once here.
>>>
>>> For the I/O space, I found now that PXA was not using the standard
>>> virtual I/O address yet, but instead used a NULL-based offset.
>>>
>>> I'm not entirely happy with this patch, but this is an outline of what
>>> I think we need to fix that: https://pastebin.com/3nVgQsEw
>>> This one is probably incomplete, at least it breaks sa1100 for now,
>>> and it adds a bogus CONFIG_PCI dependency. I'm also not sure
>>> in what way the last patch in the series triggers it, rather than the
>>> one that removed mach/io.h.
>>>
>>> I had sx1 booting in qemu at least, with the omap1 multiplatform series only.
>>> If you have a custom config for this one, make sure you get the right
>>> DEBUG_LL address.
>>>
>>>> I'll do another round of bisects.
>>>
>>
>> Here is the bisect for the sx1 boot failure.
> 
> Odd, I can't reproduce this at all. Do you get any console output at
> all for this?
> 
> Is this the plain omap1_defconfig, or something else?
> 

No, it is my own sx1 specific configuration.

https://github.com/groeck/linux-build-test/blob/master/rootfs/arm/qemu_sx1_defconfig

I don't recall where I got it from but ...

> One thing I keep having to apply myself is this snippet:
> 
> diff --git a/arch/arm/mm/proc-arm925.S b/arch/arm/mm/proc-arm925.S
> index 0bfad62ea858..87c695703580 100644
> --- a/arch/arm/mm/proc-arm925.S
> +++ b/arch/arm/mm/proc-arm925.S
> @@ -441,7 +441,6 @@ __arm925_setup:
> 
>   #ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
>          mov     r0, #4                          @ disable write-back
> on caches explicitly
> -       mcr     p15, 7, r0, c15, c0, 0
>   #endif

it does not have CONFIG_CPU_DCACHE_WRITETHROUGH enabled.

Guenter

> 
>          adr     r5, arm925_crval
> 
> I don't remember what the story is behind this, but I can't actually manage
> to boot omap1_defconfig on qemu with the instruction intact.
> 
>         Arnd