Re: drivers/vhost/vhost.c:718:3: error: call to '__compiletime_assert_718' declared with attribute error: BUILD_BUG_ON failed: __alignof__ *vq->avail > VRING_AVAIL_ALIGN_SIZE

[Michael S. Tsirkin]

I keep seeing these errors. Any openrisc/gcc/kbuild people
interested in looking into this?

On Sun, Feb 28, 2016 at 07:15:14AM +0800, kbuild test robot wrote:
> Hi Michael,
> 
> FYI, the error/warning still remains.
> 
> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 
> master
> head:   691429e13dfaf5b0994b07cc166db41bd608ee3d
> commit: 5d9a07b0de512b77bf28d2401e5fe3351f00a240 vhost: relax used address 
> alignment
> date:   1 year, 2 months ago
> config: openrisc-allmodconfig (attached as .config)
> reproduce:
>         wget 
> https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross
>  -O ~/bin/make.cross
>         chmod +x ~/bin/make.cross
>         git checkout 5d9a07b0de512b77bf28d2401e5fe3351f00a240
>         # save the attached .config to linux build tree
>         make.cross ARCH=openrisc 
> 
> All errors (new ones prefixed by >>):
> 
>    drivers/vhost/vhost.c: In function 'vhost_vring_ioctl':
> >> drivers/vhost/vhost.c:718:3: error: call to '__compiletime_assert_718' 
> >> declared with attribute error: BUILD_BUG_ON failed: __alignof__ *vq->avail 
> >> > VRING_AVAIL_ALIGN_SIZE
> 
> vim +/__compiletime_assert_718 +718 drivers/vhost/vhost.c
> 
>    712                            (u64)(unsigned long)a.avail_user_addr != 
> a.avail_user_addr) {
>    713                                r = -EFAULT;
>    714                                break;
>    715                        }
>    716        
>    717                        /* Make sure it's safe to cast pointers to 
> vring types. */
>  > 718                        BUILD_BUG_ON(__alignof__ *vq->avail > 
> VRING_AVAIL_ALIGN_SIZE);
>    719                        BUILD_BUG_ON(__alignof__ *vq->used > 
> VRING_USED_ALIGN_SIZE);
>    720                        if ((a.avail_user_addr & 
> (VRING_AVAIL_ALIGN_SIZE - 1)) ||
>    721                            (a.used_user_addr & (VRING_USED_ALIGN_SIZE 
> - 1)) ||
> 
> ---
> 0-DAY kernel test infrastructure                Open Source Technology Center
> https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

I still think it's a bug in the openrisc compiler used.

struct vring_avail {
        unsigned short flags;
        unsigned short idx;
};

struct vring_avail *x;

int main(int argc, char **argv)
{

        footest();
        if (__alignof__(*x) == 1)
                foo1();
        else if (__alignof__(*x) == 2)
                foo2();
        else if (__alignof__(*x) == 4)
                foo4();
        else if (__alignof__(*x) == 8)
                foo8();
        else
                fooother();
        return 0;
}

when compiled with the openrisc compiler from kernel.org this calls foo4
and it really should call foo2, since the openrisc manual says:

16.1.2
Aggregates and Unions
Aggregates (structures and arrays) and unions assume the alignment of
their most
strictly aligned element.

and alignment for unsigned short is 2.



-- 
MST