Date:   Fri, 14 Jan 2022 09:41:51 +0000
From:   <Conor.Dooley@...rochip.com>
To:     <alex@...ti.fr>, <geert@...ux-m68k.org>, <palmer@...osinc.com>
CC:     <linux-riscv@...ts.infradead.org>, <paul.walmsley@...ive.com>,
        <palmer@...belt.com>, <aou@...s.berkeley.edu>,
        <heinrich.schuchardt@...onical.com>, <bin.meng@...driver.com>,
        <sagar.kadam@...ive.com>, <damien.lemoal@....com>,
        <axboe@...nel.dk>, <linux-kernel@...r.kernel.org>,
        <stable@...r.kernel.org>
Subject: Re: [PATCH 02/12] RISC-V: MAXPHYSMEM_2GB doesn't depend on
 CMODEL_MEDLOW

On 14/01/2022 09:09, Alexandre ghiti wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know 
> the content is safe
> 
> Hi Conor,
> 
> On 1/14/22 09:40, Conor.Dooley@...rochip.com wrote:
>> On 11/01/2022 16:04, Geert Uytterhoeven wrote:
>>> EXTERNAL EMAIL: Do not click links or open attachments unless you 
>>> know the content is safe
>>>
>>> Hi Palmer,
>>>
>>> On Fri, Nov 19, 2021 at 5:47 PM Palmer Dabbelt <palmer@...osinc.com> 
>>> wrote:
>>>> From: Palmer Dabbelt <palmer@...osinc.com>
>>>>
>>>> For non-relocatable kernels we need to be able to link the kernel at
>>>> approximately PAGE_OFFSET, thus requiring medany (as medlow requires 
>>>> the
>>>> code to be linked within 2GiB of 0).  The inverse doesn't apply, 
>>>> though:
>>>> since medany code can be linked anywhere it's fine to link it close to
>>>> 0, so we can support the smaller memory config.
>>>>
>>>> Fixes: de5f4b8f634b ("RISC-V: Define MAXPHYSMEM_1GB only for RV32")
>>>> Cc: stable@...r.kernel.org
>>>> Signed-off-by: Palmer Dabbelt <palmer@...osinc.com>
>>> Thanks for your patch, which is now commit 9f36b96bc70f9707 ("RISC-V:
>>> MAXPHYSMEM_2GB doesn't depend on CMODEL_MEDLOW").
>>>
>>>> I found this when going through the savedefconfig diffs for the K210
>>>> defconfigs.  I'm not entirely sure they're doing the right thing here
>>>> (they should probably be setting CMODEL_LOW to take advantage of the
>>>> better code generation), but I don't have any way to test those
>>>> platforms so I don't want to change too much.
>>> I can confirm MAXPHYSMEM_2GB works on K210 with CMODEL_MEDANY.
>>>
>>> As the Icicle has 1760 MiB of RAM, I gave it a try with MAXPHYSMEM_2GB
>>> (and CMODEL_MEDANY), too.  Unfortunately it crashes very early
>>> (needs earlycon to see):
>> Given you said 1760 MiB I assume you're not running the device tree
>> currently in the kernel?
>> But the defconfig is /arch/riscv/configs/defconfig?
>>
>> I tested it w/ my newer version of the dts, using both 1760 & 736 MiB
>> (ddrc_cache_lo only) w/ MAXPHYSMEM_2GB.
>> Enabling MAXPHYSMEM_2GB with either CMODEL_MEDANY or CMODEL_MEDLOW
>> lead to the same boot failure as you got.
> 
> 
> Any chance you can give a try to [1] so that I can extract it from my
> sv48 patchset and propose it to fixes if it works?
Applied, tested with 1760 & 736 MiB - booted fine. :)
> 
> Thanks,
> 
> Alex
> 
> https://patchwork.kernel.org/project/linux-riscv/patch/20211206104657.433304-6-alexandre.ghiti@canonical.com/ 
> 
> 
> 
>>>       OF: fdt: Ignoring memory range 0x80000000 - 0x80200000
>>>       Machine model: Microchip PolarFire-SoC Icicle Kit
>>>       printk: debug: ignoring loglevel setting.
>>>       earlycon: ns16550a0 at MMIO32 0x0000000020100000 (options 
>>> '115200n8')
>>>       printk: bootconsole [ns16550a0] enabled
>>>       printk: debug: skip boot console de-registration.
>>>       efi: UEFI not found.
>>>       Unable to handle kernel paging request at virtual address 
>>> ffffffff87e00001
>>>       Oops [#1]
>>>       Modules linked in:
>>>       CPU: 0 PID: 0 Comm: swapper Not tainted 
>>> 5.16.0-08771-g85515233477d #56
>>>       Hardware name: Microchip PolarFire-SoC Icicle Kit (DT)
>>>       epc : fdt_check_header+0x14/0x208
>>>        ra : early_init_dt_verify+0x16/0x94
>>>       epc : ffffffff802ddacc ra : ffffffff8082415a sp : ffffffff81203ee0
>>>        gp : ffffffff812ec3a8 tp : ffffffff8120cd80 t0 : 0000000000000005
>>>        t1 : 0000001040000000 t2 : ffffffff80000000 s0 : ffffffff81203f00
>>>        s1 : ffffffff87e00000 a0 : ffffffff87e00000 a1 : 000000040ffffce7
>>>        a2 : 00000000000000e7 a3 : ffffffff8080394c a4 : 0000000000000000
>>>        a5 : 0000000000000000 a6 : 0000000000000000 a7 : 0000000000000000
>>>        s2 : ffffffff81203f98 s3 : 8000000a00006800 s4 : fffffffffffffff3
>>>        s5 : 0000000000000000 s6 : 0000000000000001 s7 : 0000000000000000
>>>        s8 : 0000000020236c20 s9 : 0000000000000000 s10: 0000000000000000
>>>        s11: 0000000000000000 t3 : 0000000000000018 t4 : 00ff000000000000
>>>        t5 : 0000000000000000 t6 : 0000000000000010
>>>       status: 0000000200000100 badaddr: ffffffff87e00001 cause: 
>>> 000000000000000d
>>>       [<ffffffff802ddacc>] fdt_check_header+0x14/0x208
>>>       [<ffffffff8082415a>] early_init_dt_verify+0x16/0x94
>>>       [<ffffffff80802dee>] setup_arch+0xec/0x4ec
>>>       [<ffffffff80800700>] start_kernel+0x88/0x6d6
>>>       random: get_random_bytes called from
>>> print_oops_end_marker+0x22/0x44 with crng_init=0
>>>       ---[ end trace 903df1a0ade0b876 ]---
>>>       Kernel panic - not syncing: Attempted to kill the idle task!
>>>       ---[ end Kernel panic - not syncing: Attempted to kill the idle 
>>> task! ]---
>>>
>>> So the FDT is at 0xffffffff87e00000, i.e. at 0x7e00000 from the start
>>> of virtual memory (CONFIG_PAGE_OFFSET=0xffffffff80000000), and thus
>>> within the 2 GiB range.
>>>
>>>> --- a/arch/riscv/Kconfig
>>>> +++ b/arch/riscv/Kconfig
>>>> @@ -280,7 +280,7 @@ choice
>>>>                   depends on 32BIT
>>>>                   bool "1GiB"
>>>>           config MAXPHYSMEM_2GB
>>>> -               depends on 64BIT && CMODEL_MEDLOW
>>>> +               depends on 64BIT
>>>>                   bool "2GiB"
>>>>           config MAXPHYSMEM_128GB
>>>>                   depends on 64BIT && CMODEL_MEDANY
>>> Gr{oetje,eeting}s,
>>>
>>>                           Geert
>>>
>>> -- 
>>> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- 
>>> geert@...ux-m68k.org
>>>
>>> In personal conversations with technical people, I call myself a 
>>> hacker. But
>>> when I'm talking to journalists I just say "programmer" or something 
>>> like that.
>>>                                   -- Linus Torvalds
>>>
>>> _______________________________________________
>>> linux-riscv mailing list
>>> linux-riscv@...ts.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/linux-riscv
>>>
>> _______________________________________________
>> linux-riscv mailing list
>> linux-riscv@...ts.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-riscv

Powered by blists - more mailing lists