Message-ID: <5487118F.4010606@hitachi.com>
Date: Wed, 10 Dec 2014 00:13:19 +0900
From: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Organization: Hitachi, Ltd., Japan
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:13.0) Gecko/20120614 Thunderbird/13.0.1
MIME-Version: 1.0
To: "Jon Medhurst (Tixy)" <tixy@linaro.org>
Cc: Wang Nan <wangnan0@huawei.com>, lizefan@huawei.com, linux@arm.linux.org.uk,
        linux-kernel@vger.kernel.org, linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH v14 7/7] ARM: kprobes: enable OPTPROBES for ARM 32
References: <1418020040-68977-1-git-send-email-wangnan0@huawei.com>  <1418020131-69375-1-git-send-email-wangnan0@huawei.com>  <1418036666.3647.33.camel@linaro.org> <5485886E.2060303@huawei.com>  <1418039451.3647.48.camel@linaro.org> <5486CB9F.6030804@hitachi.com> <1418121037.3641.22.camel@linaro.org>
In-Reply-To: <1418121037.3641.22.camel@linaro.org>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org

(2014/12/09 19:30), Jon Medhurst (Tixy) wrote:
> On Tue, 2014-12-09 at 19:14 +0900, Masami Hiramatsu wrote:
>> (2014/12/08 20:50), Jon Medhurst (Tixy) wrote:> arch_optimize_kprobes is calling __arch_optimize_kprobes, which is
>>> iterating over a list of probes and removing each one in turn, if this
>>> is happening on multiple cpu's simultaneously, it's not clear to me that
>>> such an operation is safe. list_del_init calls __list_del which does
>>>
>>> 	next->prev = prev;
>>> 	prev->next = next;
>>>
>>> so what happens if another cpu is at the same time updating any of those
>>> list entries? Without even fully analysing the code I can see that with
>>> the fact that the list handling helpers have no memory barriers, that
>>> the above two lines could be seen to execute in the reverse order, e.g.
>>>
>>> 	prev->next = next;
>>> 	next->prev = prev;
>>>
>>> so another CPU could find and delete next before this one has finished
>>> doing so. Would the list end up in a consistent state where no loops
>>> develop and no probes are missed? I don't know the answer and a full
>>> analysis would be complicated, but my gut feeling is that if a cpu can
>>> observe the links in the list in an inconsistent state then only bad
>>> things can result.
>>
>> Just a comment, arch_optimize_kprobes() are only called under
>> kprobe_mutex held. No concurrent update happens :)
> 
> Except in the case of the code I was commenting on which was using
> stop_machine to make all cpu's simultaneously do the work of
> arch_optimize_kprobes :-)

Ah, right! stop_machine with cpu_online_mask cause that problem.

Thanks,

-- 
Masami HIRAMATSU
Software Platform Research Dept. Linux Technology Research Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@hitachi.com


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/