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;
Date:   Tue, 15 Feb 2022 19:06:48 +0530
From:   "Naveen N. Rao" <naveen.n.rao@linux.vnet.ibm.com>
Subject: Re: [PATCH v2 09/13] powerpc/ftrace: Implement
 CONFIG_DYNAMIC_FTRACE_WITH_ARGS
To:     Christophe Leroy <christophe.leroy@csgroup.eu>,
        Jiri Kosina <jikos@kernel.org>,
        Joe Lawrence <joe.lawrence@redhat.com>,
        Josh Poimboeuf <jpoimboe@redhat.com>,
        Miroslav Benes <mbenes@suse.cz>,
        Ingo Molnar <mingo@redhat.com>,
        Michael Ellerman <mpe@ellerman.id.au>,
        Petr Mladek <pmladek@suse.com>,
        Steven Rostedt <rostedt@goodmis.org>
Cc:     "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "linuxppc-dev@lists.ozlabs.org" <linuxppc-dev@lists.ozlabs.org>,
        "live-patching@vger.kernel.org" <live-patching@vger.kernel.org>
References: <cover.1640017960.git.christophe.leroy@csgroup.eu>
        <5831f711a778fcd6eb51eb5898f1faae4378b35b.1640017960.git.christophe.leroy@csgroup.eu>
        <1644852011.qg7ud9elo2.naveen@linux.ibm.com>
        <1b28f52a-f8b7-6b5c-e726-feac4123517d@csgroup.eu>
        <875ypgo0f3.fsf@mpe.ellerman.id.au>
In-Reply-To: <875ypgo0f3.fsf@mpe.ellerman.id.au>
User-Agent: astroid/4d6b06ad (https://github.com/astroidmail/astroid)
Message-Id: <1644930705.g64na2kgvd.naveen@linux.ibm.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Precedence: bulk

Michael Ellerman wrote:
> Christophe Leroy <christophe.leroy@csgroup.eu> writes:
>> Le 14/02/2022 =C3=A0 16:25, Naveen N. Rao a =C3=A9crit=C2=A0:
>>> Christophe Leroy wrote:
>>>> Implement CONFIG_DYNAMIC_FTRACE_WITH_ARGS. It accelerates the call
>>>> of livepatching.
>>>>
>>>> Also note that powerpc being the last one to convert to
>>>> CONFIG_DYNAMIC_FTRACE_WITH_ARGS, it will now be possible to remove
>>>> klp_arch_set_pc() on all architectures.
>>>>
>>>> Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
>>>> ---
>>>> =C2=A0arch/powerpc/Kconfig=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 |=C2=A0 1 +
>>>> =C2=A0arch/powerpc/include/asm/ftrace.h=C2=A0=C2=A0=C2=A0 | 17 +++++++=
++++++++++
>>>> =C2=A0arch/powerpc/include/asm/livepatch.h |=C2=A0 4 +---
>>>> =C2=A03 files changed, 19 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
>>>> index cdac2115eb00..e2b1792b2aae 100644
>>>> --- a/arch/powerpc/Kconfig
>>>> +++ b/arch/powerpc/Kconfig
>>>> @@ -210,6 +210,7 @@ config PPC
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 select HAVE_DEBUG_KMEMLEAK
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 select HAVE_DEBUG_STACKOVERFLOW
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 select HAVE_DYNAMIC_FTRACE
>>>> +=C2=A0=C2=A0=C2=A0 select HAVE_DYNAMIC_FTRACE_WITH_ARGS=C2=A0=C2=A0=
=C2=A0 if MPROFILE_KERNEL || PPC32
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 select HAVE_DYNAMIC_FTRACE_WITH_REGS=C2=A0=C2=
=A0=C2=A0 if MPROFILE_KERNEL || PPC32
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 select HAVE_EBPF_JIT
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 select HAVE_EFFICIENT_UNALIGNED_ACCESS=C2=A0=
=C2=A0=C2=A0 if !(CPU_LITTLE_ENDIAN=20
>>>> && POWER7_CPU)
>>>> diff --git a/arch/powerpc/include/asm/ftrace.h=20
>>>> b/arch/powerpc/include/asm/ftrace.h
>>>> index b3f6184f77ea..45c3d6f11daa 100644
>>>> --- a/arch/powerpc/include/asm/ftrace.h
>>>> +++ b/arch/powerpc/include/asm/ftrace.h
>>>> @@ -22,6 +22,23 @@ static inline unsigned long=20
>>>> ftrace_call_adjust(unsigned long addr)
>>>> =C2=A0struct dyn_arch_ftrace {
>>>> =C2=A0=C2=A0=C2=A0=C2=A0 struct module *mod;
>>>> =C2=A0};
>>>> +
>>>> +#ifdef CONFIG_DYNAMIC_FTRACE_WITH_ARGS
>>>> +struct ftrace_regs {
>>>> +=C2=A0=C2=A0=C2=A0 struct pt_regs regs;
>>>> +};
>>>> +
>>>> +static __always_inline struct pt_regs *arch_ftrace_get_regs(struct=20
>>>> ftrace_regs *fregs)
>>>> +{
>>>> +=C2=A0=C2=A0=C2=A0 return &fregs->regs;
>>>> +}
>>>=20
>>> I think this is wrong. We need to differentiate between ftrace_caller()=
=20
>>> and ftrace_regs_caller() here, and only return pt_regs if coming in=20
>>> through ftrace_regs_caller() (i.e., FL_SAVE_REGS is set).
>>
>> Not sure I follow you.
>>
>> This is based on 5740a7c71ab6 ("s390/ftrace: add=20
>> HAVE_DYNAMIC_FTRACE_WITH_ARGS support")
>>
>> It's all the point of HAVE_DYNAMIC_FTRACE_WITH_ARGS, have the regs also=
=20
>> with ftrace_caller().
>>
>> Sure you only have the params, but that's the same on s390, so what did=
=20
>> I miss ?

It looks like s390 is special since it apparently saves all registers=20
even for ftrace_caller:=20
https://lore.kernel.org/all/YbipdU5X4HNDWIni@osiris/

As I understand it, the reason ftrace_get_regs() was introduced was to=20
be able to only return the pt_regs, if _all_ registers were saved into=20
it, which we don't do when coming in through ftrace_caller(). See the=20
x86 implementation (commit 02a474ca266a47 ("ftrace/x86: Allow for=20
arguments to be passed in to ftrace_regs by default"), which returns=20
pt_regs conditionally.

>=20
> I already have this series in next, I can pull it out, but I'd rather
> not.

Yeah, I'm sorry about the late review on this one.

>=20
> I'll leave it in for now, hopefully you two can agree overnight my time
> whether this is a big problem or something we can fix with a fixup
> patch.

I think changes to this particular patch can be added as an incremental=20
patch. If anything, pt_regs won't have all valid registers, but no one=20
should depend on it without also setting FL_SAVE_REGS anyway.

I was concerned about patch 8 though, where we are missing saving r1=20
into pt_regs. That gets used in patch 11, and will be used during=20
unwinding when the function_graph tracer is active. But, this should=20
still just result in us being unable to unwind the stack, so I think=20
that can also be an incremental patch.


Thanks,
Naveen