Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [RFC PATCH 3/5] arm64: aarch64-insn: Add barrier encodings
To:     Mark Rutland <mark.rutland@arm.com>
Cc:     linux-kernel@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        catalin.marinas@arm.com, will@kernel.org, broonie@kernel.org
References: <20210120171745.1657762-1-jthierry@redhat.com>
 <20210120171745.1657762-4-jthierry@redhat.com>
 <20210202111512.GC59049@C02TD0UTHF1T.local>
From:   Julien Thierry <jthierry@redhat.com>
Message-ID: <a2466e6b-5b9b-6db6-496e-f8353e097d7e@redhat.com>
Date:   Wed, 3 Feb 2021 09:47:14 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.11.0
MIME-Version: 1.0
In-Reply-To: <20210202111512.GC59049@C02TD0UTHF1T.local>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 2/2/21 12:15 PM, Mark Rutland wrote:
> On Wed, Jan 20, 2021 at 06:17:43PM +0100, Julien Thierry wrote:
>> Create necessary functions to encode/decode aarch64 data/instruction
>> barriers.
>>
>> Signed-off-by: Julien Thierry <jthierry@redhat.com>
>> ---
>>   arch/arm64/include/asm/aarch64-insn.h | 9 +++++++++
>>   1 file changed, 9 insertions(+)
>>
>> diff --git a/arch/arm64/include/asm/aarch64-insn.h b/arch/arm64/include/asm/aarch64-insn.h
>> index 200bee726172..d0fee47bbe6e 100644
>> --- a/arch/arm64/include/asm/aarch64-insn.h
>> +++ b/arch/arm64/include/asm/aarch64-insn.h
>> @@ -379,6 +379,9 @@ __AARCH64_INSN_FUNCS(eret_auth,	0xFFFFFBFF, 0xD69F0BFF)
>>   __AARCH64_INSN_FUNCS(mrs,	0xFFF00000, 0xD5300000)
>>   __AARCH64_INSN_FUNCS(msr_imm,	0xFFF8F01F, 0xD500401F)
>>   __AARCH64_INSN_FUNCS(msr_reg,	0xFFF00000, 0xD5100000)
>> +__AARCH64_INSN_FUNCS(dmb,	0xFFFFF0FF, 0xD50330BF)
>> +__AARCH64_INSN_FUNCS(dsb,	0xFFFFF0FF, 0xD503309F)
>> +__AARCH64_INSN_FUNCS(isb,	0xFFFFF0FF, 0xD50330DF)
> 
> These match the encodings in ARM DDI 0487G.a, with a couple of caveats
> for DSB.
> 
> Per section C6.2.82 on page C6-1000, when CRm != 0x00, the instruction
> isn't considered a DSB. I believe per the "barriers" decode table on
> page C4-289 that here "0x00" is actually a binary string and 'x' is a
> "don't care" -- I've raised a ticket to get the documentation clarified.
> I suspect we need to write a function to handle that.
> 

Ah, I did miss that part. Thanks for pointing it out (and for clarifying 
it's probably not hexa, but the binary string makes sense since it's a 4 
bits field)

> There's also a secondary encoding for DSB with FEAT_XS, which we don't
> currently use but might want to add.
> 

Ah, yes, had to pick up a newer version of the Arm ARM! I'll add it.

>>   #undef	__AARCH64_INSN_FUNCS
>>   
>> @@ -390,6 +393,12 @@ static inline bool aarch64_insn_is_adr_adrp(u32 insn)
>>   	return aarch64_insn_is_adr(insn) || aarch64_insn_is_adrp(insn);
>>   }
>>   
>> +static inline bool aarch64_insn_is_barrier(u32 insn)
>> +{
>> +	return aarch64_insn_is_dmb(insn) || aarch64_insn_is_dsb(insn) ||
>> +	       aarch64_insn_is_isb(insn);
>> +}
> 
> I assume this is meant to match the barriers instruction class, as per
> the table on page C4-289? That also contains CLREX, SB, SSBB, and PSSBB,
> and it might be worth adding them at the same time.
> 

Yes, I have to admit I only added the ones that objtool saw and 
complained about "unreachable instruction" (mostly barriers after 
ret/eret). I'll add them as well.

Thanks,

-- 
Julien Thierry