Message-ID: <57708ADC.3050002@huawei.com>
Date:	Mon, 27 Jun 2016 10:09:32 +0800
From:	zhouchengming <zhouchengming1@...wei.com>
To:	Bamvor Zhang <bamv2005@...il.com>
CC:	Yury Norov <ynorov@...iumnetworks.com>,
	Arnd Bergmann <arnd@...db.de>,
	Catalin Marinas <catalin.marinas@....com>,
	<linux-arm-kernel@...ts.infradead.org>,
	<linux-kernel@...r.kernel.org>, <linux-doc@...r.kernel.org>,
	<linux-arch@...r.kernel.org>, <linux-s390@...r.kernel.org>,
	<libc-alpha@...rceware.org>, <kilobyte@...band.pl>,
	<pinskia@...il.com>, <szabolcs.nagy@....com>,
	<Nathan_Lynch@...tor.com>, <heiko.carstens@...ibm.com>,
	Alexander Graf <agraf@...e.de>, <geert@...ux-m68k.org>,
	<Prasun.Kapoor@...iumnetworks.com>, <klimov.linux@...il.com>,
	<broonie@...nel.org>, <schwidefsky@...ibm.com>,
	Bamvor Jian Zhang <bamvor.zhangjian@...wei.com>,
	<philipp.tomsich@...obroma-systems.com>, <joseph@...esourcery.com>,
	<christoph.muellner@...obroma-systems.com>, <guohanjun@...wei.com>
Subject: Re: [PATCH 17/23] arm64: ptrace: handle ptrace_request differently
 for aarch32 and ilp32

On 2016/6/25 22:15, Bamvor Zhang wrote:
> Hi, Chengming
>
> On Sat, Jun 25, 2016 at 5:36 PM, zhouchengming
> <zhouchengming1@...wei.com>  wrote:
>> On 2016/6/9 1:00, Yury Norov wrote:
>>>
>>> On Wed, Jun 08, 2016 at 09:34:09AM +0800, zhouchengming wrote:
>>>>
>>>> On 2016/5/24 8:04, Yury Norov wrote:
>>>>>
>>>>> Here new aarch32 ptrace syscall handler is introsuced to avoid run-time
>>>>> detection of the task type.
>>>>>
>>>>> Signed-off-by: Yury Norov<ynorov@...iumnetworks.com>
>>>
>>>
>>> [...]
>>>
>>>> Hello, I found ilp32 will use sys_ptrace, not compat_sys_ptrace. So I
>>>> write
>>>> a little patch to see if can solve the problem correctly.
>>>>
>>>> Thanks.
>>>>
>>>>   From f6156236df578bb05c4a17e7f9776ceaf8f7afe6 Mon Sep 17 00:00:00 2001
>>>> From: Zhou Chengming<zhouchengming1@...wei.com>
>>>> Date: Wed, 8 Jun 2016 09:46:23 +0800
>>>> Subject: [PATCH] ilp32: use compat_sys_ptrace instead of sys_ptrace
>>>>
>>>> When we analyze a testcase of ptrace that failed on ilp32, we found
>>>> the syscall that the ilp32 uses is sys_ptrace, not compat_sys_ptrace.
>>>> Because in include/uapi/asm-generic/unistd.h it's defined like:
>>>> __SYSCALL(__NR_ptrace, sys_ptrace)
>>>> So we change it to __SC_COMP(__NR_ptrace, sys_ptrace, compat_sys_ptrace),
>>>> let compat tasks use the compat_sys_ptrace.
>>>>
>>>> Signed-off-by: Zhou Chengming<zhouchengming1@...wei.com>
>>>> ---
>>>>    include/uapi/asm-generic/unistd.h |    2 +-
>>>>    1 files changed, 1 insertions(+), 1 deletions(-)
>>>>
>>>> diff --git a/include/uapi/asm-generic/unistd.h
>>>> b/include/uapi/asm-generic/unistd.h
>>>> index 2862d2e..50ee770 100644
>>>> --- a/include/uapi/asm-generic/unistd.h
>>>> +++ b/include/uapi/asm-generic/unistd.h
>>>> @@ -364,7 +364,7 @@ __SC_WRAP(__NR_syslog, sys_syslog)
>>>>
>>>>    /* kernel/ptrace.c */
>>>>    #define __NR_ptrace 117
>>>> -__SYSCALL(__NR_ptrace, sys_ptrace)
>>>> +__SC_COMP(__NR_ptrace, sys_ptrace, compat_sys_ptrace)
>>>>
>>>>    /* kernel/sched/core.c */
>>>>    #define __NR_sched_setparam 118
>>>> --
>>>> 1.7.7
>>>>
>>>
>>> Hi Zhou,
>>>
>>> Thank you for the catch.
>>>
>>> Could you also show the test that is failed for you. It should
>>> probably be sent to LTP maillist.
>>>
>>> I'm not sure your fix correct as it affects other architectures that
>>> use standard unistd.h. I think it's better to redirect the syscall in
>>> arch/arm64/kernel/sys_ilp32.c with corresponding definition.
>>>
>>> Yury
>>>
>>> .
>>>
>>
>> Sorry, I missed this mail. Thanks for your reply. :)
>> I attach the testcase file of ptrace that failed on ilp32.
>> I also think it's better to redirect the syscall in ilp32, so I changed
>> the patch.
>
> Thanks for your patch. But Yury has already sent an new series this week
> which define ptrace to compat one.
>
> It seems that Yury do not take GET/SETSIGMASK into account. You
> could share your test case and patches at this point.
>
> Best wishes
>
> Bamvor

Ok, I get it. So the new series can handle ptrace correctly. :)
But as for the GET/SETSIGMASK request, both ilp32 and aarch32 are wrong, 
because they are handled in ptrace_request(like aarch64). But I don't 
have a good way to correct it in all architectures.
The architectures that use compat_ptrace_request are:
arch/arm64, arch/mips, arch/parisc, arch/powerpc, arch/s390, arch/sparc,
arch/tile, arch/x86.
We have to use two architecture dependent functions in 
compat_ptrace_request to fix the GET/SETSIGMASK request. (as for 
arch/arm64, the two functions are: put_sigset_t and get_sigset_t)
So we maybe have to prepare these functions for all these architectures.
But I don't have much time to work on it. So I just fix it for ilp32 of 
arch/arm64 (Ah, easier way for me). I will put my patch of this fix 
under Yury's new series.

Thanks!

>
> [1] https://www.mail-archive.com/linux-doc@vger.kernel.org/msg03811.html
>>
>>
>>  From 7e692ba1adf02c2a2f125836f5222f455c9ffe56 Mon Sep 17 00:00:00 2001
>> From: Zhou Chengming<zhouchengming1@...wei.com>
>> Date: Sat, 25 Jun 2016 18:02:51 +0800
>> Subject: [PATCH] ilp32 should use compat_sys_ptrace
>>
>> The file include/uapi/asm-generic/unistd.h defines this:
>> __SYSCALL(__NR_ptrace, sys_ptrace)
>> It may cause some ptrace tests failed on ilp32. So we redirect the ptrace
>> syscall in arch/arm64/kernel/sys_ilp32.c with corresponding definition.
>>
>> Signed-off-by: Zhou Chengming<zhouchengming1@...wei.com>
>> ---
>>   arch/arm64/kernel/sys_ilp32.c |    3 +++
>>   1 files changed, 3 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/arm64/kernel/sys_ilp32.c b/arch/arm64/kernel/sys_ilp32.c
>> index d85fe94..06d5e1b 100644
>> --- a/arch/arm64/kernel/sys_ilp32.c
>> +++ b/arch/arm64/kernel/sys_ilp32.c
>> @@ -46,6 +46,9 @@
>>   asmlinkage long ilp32_sys_rt_sigreturn_wrapper(void);
>>   #define compat_sys_rt_sigreturn        ilp32_sys_rt_sigreturn_wrapper
>>
>> +/* ilp32 should use compat_sys_ptrace */
>> +#define sys_ptrace                    compat_sys_ptrace
>> +
>>   #include<asm/syscall.h>
>>
>>   #undef __SYSCALL
>> --
>> 1.7.7
>>
>>
>>
>
>
>

Powered by blists - more mailing lists