Re: [Qemu-devel] [PATCH] linux-user: Implement renameat2 when defined

[Palmer Dabbelt]

On Thu, 21 Dec 2017 06:01:25 PST (-0800), address@hidden wrote:
> On 20 December 2017 at 00:29, Palmer Dabbelt <address@hidden> wrote:
> > +#if defined(TARGET_NR_renameat2) && defined(__NR_renameat2)
> > +    case TARGET_NR_renameat2:
> > +        {
> > +            void *p2;
> > +            p  = lock_user_string(arg2);
> > +            p2 = lock_user_string(arg4);
> > +            if (!p || !p2)
> > +                ret = -TARGET_EFAULT;
> > +            else
> > +                ret = get_errno(syscall(__NR_renameat2, arg1, p, arg3, p2, 
> > arg5));
> > +            unlock_user(p2, arg4, 0);
> > +            unlock_user(p, arg2, 0);
> > +        }
> > +        break;
> > +#endif
>
> Should we have code to handle using plain renameat for flags==0
> calls? renameat2() only arrived in 3.15 kernels, so as it stands
> this patch will work on a smaller set of hosts than it might.

That sound sane: there isn't even a glibc wrapper for renameat2, so I assume 
most calls will be with flags==0.  Do you want me to re-spin the patch, or can 
you take it from here?