From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1754408AbcFPLWM (ORCPT <rfc822;w@1wt.eu>);
	Thu, 16 Jun 2016 07:22:12 -0400
Received: from szxga04-in.huawei.com ([119.145.14.52]:32297 "EHLO
	szxga04-in.huawei.com" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org
	with ESMTP id S1754155AbcFPLWF (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Thu, 16 Jun 2016 07:22:05 -0400
Subject: Re: [PATCH 21/23] arm64: ilp32: introduce ilp32-specific handlers for
 sigframe and ucontext
To: Yury Norov <ynorov@caviumnetworks.com>
References: <1464048292-30136-1-git-send-email-ynorov@caviumnetworks.com>
 <1464048292-30136-22-git-send-email-ynorov@caviumnetworks.com>
 <5752BCC8.7080205@huawei.com> <20160612174430.GA12067@yury-N73SV>
CC: <arnd@arndb.de>, <catalin.marinas@arm.com>,
        <linux-arm-kernel@lists.infradead.org>, <linux-kernel@vger.kernel.org>,
        <linux-doc@vger.kernel.org>, <linux-arch@vger.kernel.org>,
        <linux-s390@vger.kernel.org>, <libc-alpha@sourceware.org>,
        <schwidefsky@de.ibm.com>, <heiko.carstens@de.ibm.com>,
        <pinskia@gmail.com>, <broonie@kernel.org>, <joseph@codesourcery.com>,
        <christoph.muellner@theobroma-systems.com>, <szabolcs.nagy@arm.com>,
        <klimov.linux@gmail.com>, <Nathan_Lynch@mentor.com>, <agraf@suse.de>,
        <Prasun.Kapoor@caviumnetworks.com>, <kilobyte@angband.pl>,
        <geert@linux-m68k.org>, <philipp.tomsich@theobroma-systems.com>,
        "Andrew Pinski" <apinski@cavium.com>,
        Andrew Pinski <Andrew.Pinski@caviumnetworks.com>,
        Hanjun Guo <guohanjun@huawei.com>,
        "Zhangjian (Bamvor)" <bamvor.zhangjian@huawei.com>
From: "Zhangjian (Bamvor)" <bamvor.zhangjian@huawei.com>
Message-ID: <57628BB2.7080100@huawei.com>
Date: Thu, 16 Jun 2016 19:21:22 +0800
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101
 Thunderbird/38.1.0
MIME-Version: 1.0
In-Reply-To: <20160612174430.GA12067@yury-N73SV>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
X-Originating-IP: [10.111.72.170]
X-CFilter-Loop: Reflected
X-Mirapoint-Virus-RAPID-Raw: score=unknown(0),
	refid=str=0001.0A090201.57628BC3.0078,ss=1,re=0.000,recu=0.000,reip=0.000,cl=1,cld=1,fgs=0,
	ip=0.0.0.0,
	so=2014-11-16 11:51:01,
	dmn=2013-03-21 17:37:32
X-Mirapoint-Loop-Id: 17951ba79ac8b4969aa89c8fe62dfbac
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

Hi,

On 2016/6/13 1:44, Yury Norov wrote:
> Hi Bamvor,
>
> Sorry, I missed this patch.
>
> On Sat, Jun 04, 2016 at 07:34:32PM +0800, Zhangjian (Bamvor) wrote:
>> Hi,
>>
>> I found an issue of unwind with the following code. The correct backtrace
>> should be:
>> (gdb) where
>> #0 0x004004d0 in my_sig (sig=11) at test_force3.c:16
>> #1 <signal handler called>
>> #2 func2 (num=0) at test_force3.c:22
>> #3 0x00400540 in func1 (num=1) at test_force3.c:28
>> #4 0x00400574 in main (argc=1, argv=0xffd7bc04) at test_force3.c:33
>>
>> Without my patch, the backtrace is:
>> (gdb) where
>> #0 0x00400490 in my_sig (sig=11) at test_force3.c:16
>> #1 <signal handler called>
>> #2 0x004004e4 in main (argc=1, argv=0xffe6f8f4) at test_force3.c:33
>>
>> With my patch which fix the wrong frame pointer(setup_return calculate the offset
>> of fp through ilp32_sigframe instead of sigfreame), the backtrace is:
>> (gdb) where
>> #0 0x00400490 in my_sig (sig=11) at test_force3.c:16
>> #1 <signal handler called>
>> #2 func1 () at test_force3.c:28
>> #3 0x004004e4 in main (argc=1, argv=0xffe6f8f4) at test_force3.c:33
>>
>> I am not sure there is still some issue in kernel. But it seem that the gdb of ilp32
>> does not work correctly when unwind without framepointer.
>>
>> The test code is:
>>
>>  From 7e364a765097f57aed2d73f94c1688c2e7343e79 Mon Sep 17 00:00:00 2001
>> From: Bamvor Jian Zhang <bamvor.zhangjian@huawei.com>
>> Date: Sat, 4 Jun 2016 14:30:05 +0800
>> Subject: [PATCH] arm64: ilp32: fix for wrong fp offset when calculate the
>>   new fp
>>
>> ILP32 define its own sigframe(ilp32_sigframe) because of the
>> difference uc_context. setup_return do not use ilp32 specific
>> sigframe to calculate the new offset of fp which lead to wrong
>> fp in signal handler. At this circumstance, gdb backtrace will miss
>> one item:
>> (gdb) where
>>
>> It should be:
>> (gdb) where
>>
>> The test code is as follows:
>>
>> void my_sig(int sig)
>> {
>>          printf("sig=%d\n", sig);
>>          *(int *)0 = 0x0;
>> }
>>
>> void func2(int num)
>> {
>>          printf("%s: %d\n", __FUNCTION__, num);
>>          *(int *)0 = 0x0;
>>          func2(num-1);
>> }
>>
>> void func1(int num)
>> {
>>          printf("%s\n", __FUNCTION__);
>>          func2(num - 1);
>> }
>>
>> int main(int argc, char **argv)
>> {
>>          signal(11, my_sig);
>>          func1(argc);
>>          return 0;
>> }
>>
>> This patch fix this by passing the correct offset of fp to
>> setup_return.
>> Test pass on both ILP32 and LP64 in aarch64 EE.
>>
>> Signed-off-by: Bamvor Jian Zhang <bamvor.zhangjian@huawei.com>
>> ---
>>   arch/arm64/include/asm/signal_common.h | 3 ++-
>>   arch/arm64/kernel/signal.c             | 9 +++++----
>>   arch/arm64/kernel/signal_ilp32.c       | 4 ++--
>>   3 files changed, 9 insertions(+), 7 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/signal_common.h b/arch/arm64/include/asm/signal_common.h
>> index de93c71..a5d7b63 100644
>> --- a/arch/arm64/include/asm/signal_common.h
>> +++ b/arch/arm64/include/asm/signal_common.h
>> @@ -29,6 +29,7 @@ int setup_sigcontex(struct sigcontext __user *uc_mcontext,
>>   		    struct pt_regs *regs);
>>   int restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sf);
>>   void setup_return(struct pt_regs *regs, struct k_sigaction *ka,
>> -			void __user *frame, off_t sigframe_off, int usig);
>> +			void __user *frame, off_t sigframe_off, off_t fp_off,
>> +			int usig);
>>
>>   #endif /* __ASM_SIGNAL_COMMON_H */
>> diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c
>> index 038bebe..e66a6e9 100644
>> --- a/arch/arm64/kernel/signal.c
>> +++ b/arch/arm64/kernel/signal.c
>> @@ -256,14 +256,14 @@ static struct rt_sigframe __user *get_sigframe(struct ksignal *ksig,
>>   }
>>
>>   void setup_return(struct pt_regs *regs, struct k_sigaction *ka,
>> -			 void __user *frame, off_t sigframe_off, int usig)
>> +			 void __user *frame, off_t sigframe_off, off_t fp_off,
>> +			 int usig)
>>   {
>>   	__sigrestore_t sigtramp;
>>
>>   	regs->regs[0] = usig;
>>   	regs->sp = (unsigned long)frame;
>> -	regs->regs[29] = regs->sp + sigframe_off +
>> -		offsetof(struct sigframe, fp);
>> +	regs->regs[29] = regs->sp + sigframe_off + fp_off;
>
> I think you are right here. The only nitpick is what for we send 2
> offsets just to add one to another inside setup_return()?
> We can do like this:
>
>          void setup_return(struct pt_regs *regs, struct k_sigaction *ka,
>          			 void __user *frame, off_t fp_off, int usig)
>          {
>          	__sigrestore_t sigtramp;
>
>          	regs->regs[0] = usig;
>          	regs->sp = (unsigned long)frame;
>          	regs->regs[29] = regs->sp + fp_off;
>                  [...]
>          }
>
> Where fp_off calculation is done by caller.
>
> 	setup_return(regs, &ksig->ka, frame,
> 		offsetof(struct rt_sigframe, sig) + offsetof(struct sigframe, fp),
>                  usig);
>
> For me it's more clear to understand what happens with this approach.
> I don't think struct rt_sigframe will grow, but we can even introduce
> some helper for it:
>          #define RT_SIGFRAME_FP_POS (offsetof(struct rt_sigframe, sig) + offsetof(struct sigframe, fp))
>
> If no objections, I'll apply your patch with my fix in next series.
Sure. Thanks.

Regards

Bamvor
>
>>   	regs->pc = (unsigned long)ka->sa.sa_handler;
>>
>>   	if (ka->sa.sa_flags & SA_RESTORER)
>> @@ -294,7 +294,8 @@ static int setup_rt_frame(int usig, struct ksignal *ksig, sigset_t *set,
>>   	err |= setup_sigframe(&frame->sig, regs, set);
>>   	if (err == 0) {
>>   		setup_return(regs, &ksig->ka, frame,
>> -			offsetof(struct rt_sigframe, sig), usig);
>> +			offsetof(struct rt_sigframe, sig),
>> +			offsetof(struct sigframe, fp), usig);
>>   		if (ksig->ka.sa.sa_flags & SA_SIGINFO) {
>>   			err |= copy_siginfo_to_user(&frame->info, &ksig->info);
>>   			regs->regs[1] = (unsigned long)&frame->info;
>> diff --git a/arch/arm64/kernel/signal_ilp32.c b/arch/arm64/kernel/signal_ilp32.c
>> index a8ea73e..9030f14 100644
>> --- a/arch/arm64/kernel/signal_ilp32.c
>> +++ b/arch/arm64/kernel/signal_ilp32.c
>> @@ -147,7 +147,6 @@ static struct ilp32_rt_sigframe __user *ilp32_get_sigframe(struct ksignal *ksig,
>>   	struct ilp32_rt_sigframe __user *frame;
>>
>>   	sp = sp_top = sigsp(regs->sp, ksig);
>> -
>>   	sp = (sp - sizeof(struct ilp32_rt_sigframe)) & ~15;
>>   	frame = (struct ilp32_rt_sigframe __user *)sp;
>>
>> @@ -183,7 +182,8 @@ int ilp32_setup_rt_frame(int usig, struct ksignal *ksig,
>>   	err |= setup_ilp32_sigframe(&frame->sig, regs, set);
>>   	if (err == 0) {
>>   		setup_return(regs, &ksig->ka, frame,
>> -			     offsetof(struct ilp32_rt_sigframe, sig), usig);
>> +			     offsetof(struct ilp32_rt_sigframe, sig),
>> +			     offsetof(struct ilp32_sigframe, fp), usig);
>>   		regs->regs[1] = (unsigned long)&frame->info;
>>   		regs->regs[2] = (unsigned long)&frame->sig.uc;
>>   	}
>> --
>> 1.8.4.5
>>
>> Regards
>>
>> Bamvor
>>
>>