Re: [PATCH 02/12] RISC-V: MAXPHYSMEM_2GB doesn't depend on CMODEL_MEDLOW

[Alexandre Ghiti <alex@ghiti.fr>]

Hi Geert,

On 1/14/22 11:12, Geert Uytterhoeven wrote:
> Hi Alex,
>
> On Tue, Jan 11, 2022 at 5:14 PM Alexandre ghiti <alex@ghiti.fr> wrote:
>> On 1/11/22 17:04, Geert Uytterhoeven wrote:
>>> On Fri, Nov 19, 2021 at 5:47 PM Palmer Dabbelt <palmer@rivosinc.com> wrote:
>>>> From: Palmer Dabbelt <palmer@rivosinc.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@vger.kernel.org
>>>> Signed-off-by: Palmer Dabbelt <palmer@rivosinc.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):
>>>
>>>       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.
>>
>> I think you have just encountered what I suspected and mentioned in [1]:
>> we recently moved the kernel to the PAGE_OFFSET address used with
>> MAXPHYSMEM_2GB.
>>
>> I would try to cherry-pick [1] and see if that works better :)
>>
>> Alex
>>
>> [1]
>> https://patchwork.kernel.org/project/linux-riscv/patch/20211206104657.433304-6-alexandre.ghiti@canonical.com/
> Thanks, works fine with just that patch (needed small changes), or with
> the full series.
>
> Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>


Nice to know the full series works too, thank you, I'll add your Tested-by.

Thanks!

Alex


> Gr{oetje,eeting}s,
>
>                          Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-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@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv