From: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
To: Li Bin <huawei.libin@huawei.com>,
AKASHI Takahiro <takahiro.akashi@linaro.org>,
rostedt@goodmis.org, mingo@kernel.org, jpoimboe@redhat.com,
sjenning@redhat.com, jkosina@suse.cz, vojtech@suse.cz,
catalin.marinas@arm.com, will.deacon@arm.com
Cc: live-patching@vger.kernel.org,
linux-arm-kernel@lists.infradead.org,
linux-kernel@vger.kernel.org, lizefan@huawei.com,
felix.yang@huawei.com, guohanjun@huawei.com, xiexiuqi@huawei.com,
David Long <dave.long@linaro.org>
Subject: Re: Re: [RFC PATCH 0/5] livepatch: add support on arm64
Date: Wed, 03 Jun 2015 06:04:48 +0900 [thread overview]
Message-ID: <556E1A70.5010500@hitachi.com> (raw)
In-Reply-To: <556D8CBD.8060105@huawei.com>
On 2015/06/02 20:00, Li Bin wrote:
> On 2015/6/2 10:15, AKASHI Takahiro wrote:
>> On 05/30/2015 09:01 AM, Masami Hiramatsu wrote:
>>> On 2015/05/28 14:51, Li Bin wrote:
>>>> This patchset propose a method for gcc -mfentry feature(profile
>>>> before prologue) implementation for arm64, and propose the livepatch
>>>> implementation for arm64 based on this feature.
>>>> The gcc implementation about this feature will be post to the gcc
>>>> community soon.
>>>>
>>>> With this -mfentry feature, the entry of each function like:
>>>>
>>>> foo:
>>>> mov x9, x30
>>>> bl __fentry__
>>>> mov x30, x9
>>>> [prologue]
>>>> ...
>>>>
>>>> The x9 is a callee corruptible register, and the __fentry__ function
>>>> is responsible to protect all registers, so it can be used to protect
>>>> the x30. And the added two instructions which is register mov operation
>>>> have ralatively small impact on performance.
>>>
>>> Hm, this implementation looks good to me :)
>>> This also enables us to KPROBES_ON_FTRACE too.
>>
>> Even if x9 is a callee-saved register, there is no way to restore its original
>> value in setting up a pt_regs in ftrace_reg_caller.
Good point :)
>
> Hi, Takahiro AKASHI
>
> Firstly, x9 is not a callee-saved but a caller-saved register(or being called
> corruptible register).
> Secondly, I think x9 is already protected properly, please reference the patch:
> [PATCH 1/5] livepatch: ftrace: arm64: Add support for DYNAMIC_FTRACE_WITH_REGS
> [PATCH 3/5] livepatch: ftrace: arm64: Add support for -mfentry on arm64
I guess he concern about the x9 value at the function entrance is lost. For
example, regs->x9 at the handler of ftrace_regs_call is always same as flags
(if I correctly understand).
If it is right, it should be documented in the commit log and
Documentation/trace/ftrace.txt. However, that is practically no problem,
since;
- x9 is caller saved register, so functions MUST not depend on its value.
(this means ftrace handlers also should not expect any meaningful value
in regs->x9)
- Even if a function is wrongly coded and access x9, it is always same as
caller address (link register). easy to debug :)
So, finally, I think it's OK to use x9 for this purpose.
Thank you,
>
>> It's not the right thing for KPROBES_ON_FTRACE, is it?
>>
>> Saving Link register in stack is not a big deal since the overhead of ftrace
>> is much bigger.
>
> Performance overhead is only one aspect of the problem, and more importantly,
> even worse is that it would break the arm64 ABI rules.
>
> Thanks,
> Li Bin
>>
>> -Takahiro AKASHI
>>
>>
>>> Thanks,
>>>
>>>>
>>>> This patchset has been tested on arm64 platform.
>>>>
>>>> Li Bin (4):
>>>> livepatch: ftrace: arm64: Add support for DYNAMIC_FTRACE_WITH_REGS
>>>> livepatch: ftrace: add ftrace_function_stub_ip function
>>>> livepatch: ftrace: arm64: Add support for -mfentry on arm64
>>>> livepatch: arm64: add support for livepatch on arm64
>>>>
>>>> Xie XiuQi (1):
>>>> livepatch: arm64: support relocation in a module
>>>>
>>>> arch/arm64/Kconfig | 5 +
>>>> arch/arm64/include/asm/ftrace.h | 9 +
>>>> arch/arm64/include/asm/livepatch.h | 45 +++++
>>>> arch/arm64/kernel/Makefile | 1 +
>>>> arch/arm64/kernel/arm64ksyms.c | 4 +
>>>> arch/arm64/kernel/entry-ftrace.S | 154 +++++++++++++++-
>>>> arch/arm64/kernel/ftrace.c | 28 +++-
>>>> arch/arm64/kernel/livepatch.c | 41 ++++
>>>> arch/arm64/kernel/module.c | 355 ++++++++++++++++++------------------
>>>> include/linux/ftrace.h | 1 +
>>>> kernel/livepatch/core.c | 17 ++-
>>>> kernel/trace/ftrace.c | 32 ++++
>>>> scripts/recordmcount.pl | 2 +-
>>>> 13 files changed, 508 insertions(+), 186 deletions(-)
>>>> create mode 100644 arch/arm64/include/asm/livepatch.h
>>>> create mode 100644 arch/arm64/kernel/livepatch.c
>>>>
--
Masami HIRAMATSU
Linux Technology Research Center, System Productivity Research Dept.
Center for Technology Innovation - Systems Engineering
Hitachi, Ltd., Research & Development Group
E-mail: masami.hiramatsu.pt@hitachi.com
prev parent reply other threads:[~2015-06-02 21:05 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-05-28 5:51 [RFC PATCH 0/5] livepatch: add support on arm64 Li Bin
2015-05-28 5:51 ` [RFC PATCH 1/5] livepatch: ftrace: arm64: Add support for DYNAMIC_FTRACE_WITH_REGS Li Bin
2015-05-29 7:14 ` Paul Bolle
2015-05-29 8:01 ` Li Bin
2015-05-28 5:51 ` [RFC PATCH 2/5] livepatch: ftrace: add ftrace_function_stub_ip function Li Bin
2015-05-28 5:51 ` [RFC PATCH 3/5] livepatch: ftrace: arm64: Add support for -mfentry on arm64 Li Bin
2015-05-28 5:51 ` [RFC PATCH 4/5] livepatch: arm64: add support for livepatch " Li Bin
2015-05-28 5:51 ` [RFC PATCH 5/5] livepatch: arm64: support relocation in a module Li Bin
2015-05-29 11:52 ` [RFC PATCH 0/5] livepatch: add support on arm64 Jiri Kosina
2015-05-30 0:01 ` Masami Hiramatsu
2015-06-02 2:15 ` AKASHI Takahiro
2015-06-02 11:00 ` Li Bin
2015-06-02 21:04 ` Masami Hiramatsu [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=556E1A70.5010500@hitachi.com \
--to=masami.hiramatsu.pt@hitachi.com \
--cc=catalin.marinas@arm.com \
--cc=dave.long@linaro.org \
--cc=felix.yang@huawei.com \
--cc=guohanjun@huawei.com \
--cc=huawei.libin@huawei.com \
--cc=jkosina@suse.cz \
--cc=jpoimboe@redhat.com \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=live-patching@vger.kernel.org \
--cc=lizefan@huawei.com \
--cc=mingo@kernel.org \
--cc=rostedt@goodmis.org \
--cc=sjenning@redhat.com \
--cc=takahiro.akashi@linaro.org \
--cc=vojtech@suse.cz \
--cc=will.deacon@arm.com \
--cc=xiexiuqi@huawei.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).