[PATCH] riscv: Fix kernel crash due to PR_SET_TAGGED_ADDR_CTRL

[Alexandre Ghiti]

On 05/05/2025 18:07, Nam Cao wrote:
> Hi Alex,
>
> On Mon, May 05, 2025 at 06:02:26PM +0200, Alexandre Ghiti wrote:
>> On 04/05/2025 12:19, Nam Cao wrote:
>>> When userspace does PR_SET_TAGGED_ADDR_CTRL, but Supm extension is not
>>> available, the kernel crashes:
>>>
>>> Oops - illegal instruction [#1]
>>>       [snip]
>>> epc : set_tagged_addr_ctrl+0x112/0x15a
>>>    ra : set_tagged_addr_ctrl+0x74/0x15a
>>> epc : ffffffff80011ace ra : ffffffff80011a30 sp : ffffffc60039be10
>>>       [snip]
>>> status: 0000000200000120 badaddr: 0000000010a79073 cause: 0000000000000002
>>>       set_tagged_addr_ctrl+0x112/0x15a
>>>       __riscv_sys_prctl+0x352/0x73c
>>>       do_trap_ecall_u+0x17c/0x20c
>>>       andle_exception+0x150/0x15c
>>
>> It seems like the csr write is triggering this illegal instruction, can you
>> confirm it is?
> Yes, it is the "csr_write(CSR_ENVCFG, envcfg);" in envcfg_update_bits().
>
>> If so, I can't find in the specification that an implementation should do
>> that when writing envcfg and I can't reproduce it on qemu. Where did you
>> see this oops?
> I can't find it in the spec either. I think it is up to the implementation.


The reserved fields of senvcfg are WPRI and contrary to WLRL, it does 
not explicitly "permit" to raise an illegal instruction so I'd say it is 
not up to the implementation, I'll ask around.

Thanks,

Alex


>
> I got this crash on the MangoPI board:
> https://mangopi.org/mqpro
>
> Best regards,
> Nam
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv