Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Sender: Guenter Roeck <groeck7@gmail.com>
To:     Sam Ravnborg <sam@ravnborg.org>
Cc:     "David S . Miller" <davem@davemloft.net>,
        sparclinux@vger.kernel.org, linux-kernel@vger.kernel.org,
        Arnd Bergmann <arnd@kernel.org>
References: <20210913163712.922188-1-linux@roeck-us.net>
 <YT+SPIAl0IdWOAn/@ravnborg.org>
From:   Guenter Roeck <linux@roeck-us.net>
Subject: Re: [PATCH] sparc: mdesc: Fix compile error seen with gcc 11.x
Message-ID: <d0a4b46a-2f0e-f6a2-1342-777e738d9525@roeck-us.net>
Date:   Mon, 13 Sep 2021 11:52:52 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.13.0
MIME-Version: 1.0
In-Reply-To: <YT+SPIAl0IdWOAn/@ravnborg.org>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 9/13/21 11:02 AM, Sam Ravnborg wrote:
> Hi Guenter,
> 
> On Mon, Sep 13, 2021 at 09:37:12AM -0700, Guenter Roeck wrote:
>> sparc64 images fail to compile with gcc 11.x, reporting the following
>> errors.
>>
>> arch/sparc/kernel/mdesc.c:647:22: error:
>> 	'strcmp' reading 1 or more bytes from a region of size 0
>> arch/sparc/kernel/mdesc.c:692:22: error:
>> 	'strcmp' reading 1 or more bytes from a region of size 0
>> arch/sparc/kernel/mdesc.c:719:21:
>> 	error: 'strcmp' reading 1 or more bytes from a region of size 0
>>
>> The underlying problem is that node_block() returns a pointer beyond
>> the end of struct mdesc_hdr. gcc 11.x detects that and reports the error.
>> Adding an additional zero-length field to struct mdesc_hdr and pointing
>> to that field fixes the problem.
>>
>> Cc: Arnd Bergmann <arnd@kernel.org>
>> Signed-off-by: Guenter Roeck <linux@roeck-us.net>
>> ---
>> My apologies if a similar patch was submitted already; I was unable to find it.
>> I did find the following patch:
>>      https://git.busybox.net/buildroot/commit/?id=6e1106b4a9aee25d1556310d5cd1cb6dde2e6e3f
>> but I failed to find it in patchwork or on lore.kernel.org, and it
>> seems to be more expensive than the solution suggested here.
>>
>>   arch/sparc/kernel/mdesc.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/sparc/kernel/mdesc.c b/arch/sparc/kernel/mdesc.c
>> index 8e645ddac58e..c67bdcc23727 100644
>> --- a/arch/sparc/kernel/mdesc.c
>> +++ b/arch/sparc/kernel/mdesc.c
>> @@ -39,6 +39,7 @@ struct mdesc_hdr {
>>   	u32	node_sz; /* node block size */
>>   	u32	name_sz; /* name block size */
>>   	u32	data_sz; /* data block size */
>> +	char	data[0];
>>   } __attribute__((aligned(16)));
> 
> I do not think this will works.
> See following comment:
>   * mdesc_hdr and mdesc_elem describe the layout of the data structure
>   * we get from the Hypervisor.
> 
> With the above change you increased the size from 16 to 32 bytes,
> and any code using sizeof(struct mdesc_hdr) will now point too far in
> memory for the second and subsequent entries.
> 
> I did not take any closer look, but this was from a quick analysis.
> 

Sorry, I didn't realize that a field of size 0 increases the structure size
on sparc. I had checked the size of the old and the new structure with gcc
on x86_64 and didn't see a field size increase.

Guenter

---
Test code I had used:

#include <stddef.h>
#include <stdio.h>

typedef unsigned int u32;

struct mdesc_hdr {
         u32     version; /* Transport version */
         u32     node_sz; /* node block size */
         u32     name_sz; /* name block size */
         u32     data_sz; /* data block size */
} __attribute__((aligned(16)));

struct mdesc_hdr2 {
         u32     version; /* Transport version */
         u32     node_sz; /* node block size */
         u32     name_sz; /* name block size */
         u32     data_sz; /* data block size */
         char    data[0];
} __attribute__((aligned(16)));

int main()
{
	printf("%ld %ld\n", sizeof(struct mdesc_hdr), sizeof(struct mdesc_hdr2));

	return 0;
}