Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <c5bdf7af-112e-e92c-7df0-8bc8a652bcd8@huawei.com>
Date:   Wed, 4 Jan 2023 16:34:26 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101
 Thunderbird/102.4.2
Subject: Re: [PATCH v5 1/9] riscv/kprobe: Prepare the skeleton to implement
 RISCV OPTPROBES feature
To:     =?UTF-8?B?QmrDtnJuIFTDtnBlbA==?= <bjorn@kernel.org>,
        Chen Guokai <chenguokai17@mails.ucas.ac.cn>,
        <paul.walmsley@sifive.com>, <palmer@dabbelt.com>,
        <aou@eecs.berkeley.edu>, <rostedt@goodmis.org>, <mingo@redhat.com>,
        <sfr@canb.auug.org.au>
CC:     <linux-riscv@lists.infradead.org>, <linux-kernel@vger.kernel.org>
References: <20221224114315.850130-1-chenguokai17@mails.ucas.ac.cn>
 <20221224114315.850130-2-chenguokai17@mails.ucas.ac.cn>
 <87wn64vmoq.fsf@all.your.base.are.belong.to.us>
From:   "liaochang (A)" <liaochang1@huawei.com>
In-Reply-To: <87wn64vmoq.fsf@all.your.base.are.belong.to.us>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8bit
Precedence: bulk


在 2023/1/3 2:03, Björn Töpel 写道:
> Chen Guokai <chenguokai17@mails.ucas.ac.cn> writes:
> 
>> From: Liao Chang <liaochang1@huawei.com>
>>
>> Prepare skeleton to implement optimized kprobe on RISCV, it is consist
>> of Makfile, Kconfig and some architecture specific files: kprobe.h and
>> opt.c opt.c include some macro, type definition and functions required
>> by kprobe framework, opt_trampoline.S provide a piece of assembly code
>> template used to construct the detour buffer as the target of long jump
>> instruction(s) for each optimzed kprobe.
> 
> This is pretty much just reiterating what diff-stat says. Please try to
> explain why a certain change is done, instead of what. What is already
> in the patch.

Thanks for your suggestion, i will explain further in next revision.

> 
>> Since the jump range of PC-relative instruction JAL is +/-2M, that is
>> too small to reach the detour buffer, hence the foudamental idea to
>> address OPTPROBES on RISCV is replace 'EBREAK' with 'AUIPC/JALR'. which
>> means it needs to clobber one more instruction beside the kprobe
>> instruction, furthermore, RISCV supports hybird RVI and RVC in single
>> kernel binary, so in theory a pair of 'AUIPC/JALR' is about to clobber
>> 10 bytes(3 RVC and 1 RVI, 2 bytes is padding for alignment) at worst
>> case. The second hardsome problem is looking for one integer register as
>> the destination of 'AUIPC/JALR' without any side-effect.
> 
> There are a number of spelling errors, please use a spellchecker and if
> you reference a file (e.g. Makefile), make sure it is correctly spelled
> out.
> 
> The comments above applies to all the commit messages of this series.

Thanks for reviewing, i will correct these spelling errors.

> 
> 
> Björn

-- 
BR,
Liao, Chang