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: <1a2d7583-ac56-52ea-2ef7-887958966a07@kernel.org>
Date:   Fri, 26 May 2023 08:24:12 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
Subject: Re: [PATCH v2] x86/lib: Do not use local symbols with
 SYM_CODE_START_LOCAL()
To:     Nadav Amit <nadav.amit@gmail.com>,
        Dave Hansen <dave.hansen@intel.com>,
        Borislav Petkov <bp@alien8.de>
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        X86 ML <x86@kernel.org>, LKML <linux-kernel@vger.kernel.org>
References: <20230525184244.2311-1-namit@vmware.com>
 <38e24fd4-9213-229d-9919-7ae3bfb113bb@intel.com>
 <24E47178-C177-425F-A8EF-CFFAE22597D4@gmail.com>
Content-Language: en-US
From:   Jiri Slaby <jirislaby@kernel.org>
In-Reply-To: <24E47178-C177-425F-A8EF-CFFAE22597D4@gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 25. 05. 23, 21:39, Nadav Amit wrote:
> 
>> On May 25, 2023, at 12:05 PM, Dave Hansen <dave.hansen@intel.com> wrote:
>>
>> On 5/25/23 11:42, Nadav Amit wrote:
>>> From: Nadav Amit <namit@vmware.com>
>>>
>>> When SYM_CODE_START_LOCAL() is used, the symbols are local but need to
>>> be preserved in the object. However, using the ".L" label prefix does
>>> not retain the symbol in the object.
>>>
>>> It is beneficial to be able to map instruction pointers back to symbols,
>>> for instance for profiling. Otherwise, there are code addresses that do
>>> not map back to any symbol.  Consequently, the ".L" label prefix should
>>> not be used when SYM_CODE_START_LOCAL() is used.
>>>
>>> Few symbols, such as .Lbad_put_user_clac and currently have both the
>>> SYM_CODE_START_LOCAL() invocation and the ".L" prefix.  This commit
>>> removes the ".L" prefix from these symbols.
>>>
>>> No functional change, other then emitting these symbols into the object,
>>> is intended.
>>
>> Nadav, could you perhaps do a bit of research on how this situation came
>> to be?  Was it an accident or on purpose that these symbols came to be
>> .L?  Then, could you CC the folks who made this change and ask them
>> directly if they intended to induce the effects that you find undesirable?
> 
> Fair enough. I actually thought it is an oversight, but it now seems
> intentional (although I am not sure I understand/agree with the reason).
> 
> So apparently, for one of the symbols from my v1 (which was already
> removed), I see that Borislav Petkov suggested to prepend the “.L” in
> order to for them not to be visible [1].
> 
> The reason that is given for not making the functions visible is that
> these are "functions with very local names”.
> 
> I do not think in this tradeoff not exposing local names worth
> preventing profilers (and their users) from understanding where a
> sample/trace is was taken. If for instance you look at a branch
> trace (e.g., using Intel PT) you want to see the symbol to which a
> branch goes to.
> 
> Borislav, Jiri - do you agree?

Ah, if it makes tools' output harder to follow, I would indeed switch 
back to emitting these symbols, i.e. remove the .L prefix from them.

That said:

Acked-by: Jiri Slaby <jirislaby@kernel.org>

thanks,
-- 
js
suse labs