[PATCH] ARM: reinsert ARCH_MULTI_V4 Kconfig option

[Russell King - ARM Linux]

On Wed, Apr 09, 2014 at 05:04:48PM +0200, Uwe Kleine-König wrote:
> On Wed, Apr 09, 2014 at 04:54:16PM +0200, Jonas Jensen wrote:
> > On 13 December 2013 12:39, Russell King - ARM Linux
> > <linux at arm.linux.org.uk> wrote:
> > > I see what's causing this: the kuser helpers are using "bx lr" to return
> > > which will be undefined on non-Thumb CPUs.  We generally cope fine with
> > > non-Thumb CPUs, conditionalising where necessary on HWCAP_THUMB or the
> > > T bit in the PSR being set.
> > >
> > > However, it looks like the kuser helpers got missed.  As a check, please
> > > look at arch/arm/kernel/entry-armv.S, find the line with:
> > >
> > >         .macro  usr_ret, reg
> > >
> > > and ensure that the mov pc, \reg case always gets used.  Please report
> > > back.
> > 
> > Uwe and Arnd came up with a solution except it doesn't work when I test it.
> > 
> > The suggested patch is:
> > 
> > diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
> > index 1879e8d..de15bfd 100644
> > --- a/arch/arm/kernel/entry-armv.S
> > +++ b/arch/arm/kernel/entry-armv.S
> > @@ -739,6 +739,18 @@ ENDPROC(__switch_to)
> > 
> >         .macro  usr_ret, reg
> >  #ifdef CONFIG_ARM_THUMB
> > +       /*
> > +        * Having CONFIG_ARM_THUMB isn't a guarantee that the cpu has support
> > +        * for Thumb and so the bx instruction. Use a mov if the address to
> > +        * jump to is 32 bit aligned. (Note that this code is compiled in ARM
> > +        * mode, so this is the right test.)
> > +        */
> > +#if defined(CONFIG_CPU_32v4)
> > +       tst     \reg, #3
> > +       moveq   pc, \reg
> > +       b       .
> > +#endif
> > +
> >         bx      \reg
> >  #else
> >         mov     pc, \reg
> > 
> > 
> > With this applied, and the following two possibilities in next-20140403:
> > 
> > [1] ARCH_MULTI_V4 only (no CONFIG_ARM_THUMB)
> >      CONFIG_CPU_32v4=y
> > 
> > [2] ARCH_MULTI_V4 and ARCH_MULTI_V4T
> >      CONFIG_ARM_THUMB=y
> >      CONFIG_CPU_32v4=y
> >      CONFIG_CPU_32v4T=y
> > 
> > 
> > Booting a kernel with either option [1] or [2] yield the following results:
> > 
> > [1] works
> > [2] hangs after "[    2.730000] Freeing unused kernel memory: 104K
> > (c02f5000 - c030f000)"
> > 
> > 
> > Any help why the moveq doesn't work would be much appreciated.
> doing s/moveq/mov/ does the trick on the machine in question, but this
> is obviously not an option for mainline. But it means that even on this
> non-Thumb capable machine \reg contains an address that is not aligned.
> 
> Where does \reg come from? Is it provided by userspace? If so, is it a
> userspace bug?

Well, in case (2), if usr_ret is returning to a caller running thumb code,
it will lock up - because the register will not be aligned, so moveq will
fail, and the next instruction is a branch which only ever branches to
itself.

-- 
FTTC broadband for 0.8mile line: now at 9.7Mbps down 460kbps up... slowly
improving, and getting towards what was expected from it.