Re: [PATCH] riscv: VMAP_STACK overflow detection thread-safe

From: Deepak Gupta <debug@rivosinc.com>
To: Guo Ren <guoren@kernel.org>
Cc: palmer@dabbelt.com, linux-kernel@vger.kernel.org,
	linux-riscv@lists.infradead.org, paul.walmsley@sifive.com,
	Jisheng Zhang <jszhang@kernel.org>
Subject: Re: [PATCH] riscv: VMAP_STACK overflow detection thread-safe
Date: Thu, 24 Nov 2022 01:07:14 -0800	[thread overview]
Message-ID: <20221124090714.GB1896875@debug.ba.rivosinc.com> (raw)
In-Reply-To: <20221124082612.GC1149630@debug.ba.rivosinc.com>

On Thu, Nov 24, 2022 at 12:26:12AM -0800, Deepak Gupta wrote:
>On Thu, Nov 24, 2022 at 09:32:19AM +0800, Guo Ren wrote:
>>>+.macro asm_per_cpu dst sym tmp
>>>+       REG_L \tmp, TASK_TI_CPU_NUM(tp)
>>>+       slli \tmp, \tmp, 0x3
>>>+       la \dst, __per_cpu_offset
>>>+       add \dst, \dst, \tmp
>>>+       REG_L \tmp, 0(\dst)
>>>+       la \dst, \sym
>>>+       add \dst, \dst, \tmp
>>Another tricky asm code of using percpu, I don't know how the percpu
>>maintenance guy thinks.
>
>On this, if you can point me to someone, I would like to get their feedback.
>While discussing this, one issue that comes in my mind is if the shift 
>of 0x3 is correct or not on 32bit systems. I don't know what's the 
>size of each entry
>in `__per_cpu_offset`. On 64bit it's 8 bytes and thus 3 bit shift.

Yes this would need a fix for 32bit.
`__per_cpu_offset` is an array of `unsigned long` and thus shift should be 2 bit
on 32bit systems.