Re: [PATCH] thread_info: use unsigned long for flags

From: Mark Rutland
Date: Fri Sep 23 2016 - 19:48:47 EST


[Adding Ingo, so this doesn't get lost -- please see the end of the mail]

On Fri, Sep 23, 2016 at 11:08:46AM -0700, Andy Lutomirski wrote:
> On Fri, Sep 23, 2016 at 10:24 AM, Mark Rutland <mark.rutland@xxxxxxx> wrote:
> > The generic THREAD_INFO_IN_TASK definition of thread_info::flags is a
> > u32, matching x86 prior to the introduction of THREAD_INFO_IN_TASK.
> >
> > However, common helpers like test_ti_thread_flag() implicitly assume
> > that thread_info::flags has at least the size and alignment of unsigned
> > long, and relying on padding and alignment provided by other elements of
> > task_struct is somewhat fragile. Additionally, some architectures use
> > more that 32 bits for thread_info::flags, and others may need to in
> > future.
> >
> > With THREAD_INFO_IN_TASK, task struct follows thread_info with a long
> > field, and thus we no longer save any space as we did back in commit
> > affa219b60a11b32 ("x86: change thread_info's flag field back to 32
> > bits").
> >
> > Given all this, it makes more sense for the generic thread_info::flags
> > to be an unsigned long. Make it so.
>
> I have only one problem with this, and it's a general objection that's
> mostly off topic: why the [expletive] do the arch-independent bitfield
> helpers think in units of variable size? It's *absurd*, especially on
> big-endian architectures.
>
> Now that that's out of my system, I think this patch is fine.
> Big-endian arches that opt in will have to deal with it somehow, but I
> don't see why making it 'unsigned long' is worse than anything else.
> x86 is fine with this change.

FWIW, given <linux/thread_info.h> contains/uses the helpers mentioned above, BE
arches *must* use unsigned long (or something of the same size) today, or they
wouldn't work. In v4.8-rc7 that is the case:

$ ls -l arch | wc -l
33

# Note the above includes Kconfig, so there are 32 to consider

$ git grep -W 'struct thread_info {' -- arch/*/include | \
grep flags | grep 'unsigned long\s\+flags' | wc -l
29

$ git grep -W 'struct thread_info {' -- arch/*/include | \
grep flags | grep 'int\s\+flags'
arch/alpha/include/asm/thread_info.h- unsigned int flags; /* low level flags */

$ git grep -W 'struct thread_info {' -- arch/*/include | \
grep flags | grep 'u32\s\+flags'
arch/ia64/include/asm/thread_info.h- __u32 flags; /* thread_info flags (see TIF_*) */
arch/x86/include/asm/thread_info.h- __u32 flags; /* low level flags */

> Acked-by: Andy Lutomirski <luto@xxxxxxxxxx>
>
> Ingo, can you apply this for 4.9 so that we can make this change
> before other arches might start depending on the field being u32?
>
> --Andy

Thanks,
Mark.