[PATCH 11/13] arm64: uaccess: remove set_fs()

Mark Rutland mark.rutland at arm.com
Mon Sep 28 11:29:58 EDT 2020


On Mon, Sep 28, 2020 at 10:02:28AM +0100, Mark Rutland wrote:
> On Mon, Sep 28, 2020 at 09:04:43AM +0200, Christoph Hellwig wrote:
> > > +static inline void init_hw_uaccess_state(void)
> > >  {
> > > +	asm(ALTERNATIVE("nop", SET_PSTATE_PAN(1), ARM64_HAS_PAN));
> > > +	asm(ALTERNATIVE("nop", SET_PSTATE_UAO(0), ARM64_HAS_UAO));
> > >  }
> > 
> > The way I read this code it is about per-task state, maybe that
> > could be reflected in the name?  init_task_hw_access_state or
> > init_task_uaccess_flags?
> 
> My intent writing this was that this resets the HW into the expected
> ambient state common to all tasks, so I don't think about this as a
> per-task thing (but I agree the transient PAN toggling for atomics means
> PAN is per-task state).
> 
> I'd prefer not to put 'task' in the name as to me that implies messing
> with something like task_struct or other SW state, but if you have
> strong feeling about this, I think something like
> init_task_hw_access_state() would be fine.

Luckily this disappeared entirely as part of removing the vestigal UAO
support, so no need to debate the name. ;)

Thanks,
Mark.



More information about the linux-arm-kernel mailing list