On Tue, Jan 14, 2014 at 04:20:36PM +0100, Geert Uytterhoeven wrote:
> On Tue, Jan 14, 2014 at 4:15 PM, H. Peter Anvin <hpa@xxxxxxxxx> wrote:
> > On 01/14/2014 04:51 AM, Peter Zijlstra wrote:
> >> On Tue, Jan 14, 2014 at 03:53:31PM +1100, Stephen Rothwell wrote:
> >>> Hi Andrew,
> >>>
> >>> Today's linux-next merge of the akpm-current tree got a conflict in
> >>> kernel/futex.c between commit a52b89ebb6d4 ("futexes: Increase hash table
> >>> size for better performance") from the tip tree and commit 61beee6c76e5
> >>> ("futex: switch to USER_DS for futex test") from the akpm-current tree.
> >>>
> >>> @@@ -2869,10 -2748,13 +2871,13 @@@
> >>> * implementation, the non-functional ones will return
> >>> * -ENOSYS.
> >>> */
> >>> + fs = get_fs();
> >>> + set_fs(USER_DS);
> >>> if (cmpxchg_futex_value_locked(&curval, NULL, 0, 0) == -EFAULT)
> >>> futex_cmpxchg_enabled = 1;
> >>> + set_fs(fs);
> >>>
> >>
> >> This seems terribly broken, the *futex_value*() ops should not need
> >> that; they are supposed to access userspace without any of that.
> >
> > I am *guessing* that m68k is has get_fs() == KERNEL_DS at the point that
> > futex_init() is called. This would seem a bit of a peculiarity to m68k,
> > and as such it would seem like it would be better for it to belong in
> > the m68k-specific code, but since futex_init() is init code and only
> > called once anyway it shouldn't cause any harm...
>
> Yes it does. So when getting the exception on 68030, we notice it's a kernel
> space access error, not a user space access error, and crash.
Is there a good reason m68k works like this? That is, shouldn't we fix
the arch code instead of littering the generic code with weirdness like
this?
--
To unsubscribe from this list: send the line "unsubscribe linux-next" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
