Re: [PATCH 1/3] x86: Use pt_regs pointer in do_device_not_available()

[Ingo Molnar <mingo@elte.hu>]

* Tejun Heo <tj@kernel.org> wrote:

> Hello, Brian.
> 
> Brian Gerst wrote:
> > The generic exception handler (error_code) passes in the pt_regs
> > pointer and the error code (unused in this case).  The commit
> > "x86: fix math_emu register frame access" changed this to pass by
> > value, which doesn't work correctly with stack protector enabled.
> > Change it back to use the pt_regs pointer.
> > 
> > Signed-off-by: Brian Gerst <brgerst@gmail.com>
> > ---
> >  arch/x86/include/asm/traps.h |    2 +-
> >  arch/x86/kernel/traps.c      |    9 +++++----
> >  2 files changed, 6 insertions(+), 5 deletions(-)
> > 
> > diff --git a/arch/x86/include/asm/traps.h b/arch/x86/include/asm/traps.h
> > index cf3bb05..0d53425 100644
> > --- a/arch/x86/include/asm/traps.h
> > +++ b/arch/x86/include/asm/traps.h
> > @@ -41,7 +41,7 @@ dotraplinkage void do_int3(struct pt_regs *, long);
> >  dotraplinkage void do_overflow(struct pt_regs *, long);
> >  dotraplinkage void do_bounds(struct pt_regs *, long);
> >  dotraplinkage void do_invalid_op(struct pt_regs *, long);
> > -dotraplinkage void do_device_not_available(struct pt_regs);
> > +dotraplinkage void do_device_not_available(struct pt_regs *, long);
> >  dotraplinkage void do_coprocessor_segment_overrun(struct pt_regs *, long);
> >  dotraplinkage void do_invalid_TSS(struct pt_regs *, long);
> >  dotraplinkage void do_segment_not_present(struct pt_regs *, long);
> > diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
> > index 3b7b2e1..71a8f87 100644
> > --- a/arch/x86/kernel/traps.c
> > +++ b/arch/x86/kernel/traps.c
> > @@ -905,19 +905,20 @@ void math_emulate(struct math_emu_info *info)
> >  }
> >  #endif /* CONFIG_MATH_EMULATION */
> >  
> > -dotraplinkage void __kprobes do_device_not_available(struct pt_regs regs)
> > +dotraplinkage void __kprobes
> > +do_device_not_available(struct pt_regs *regs, long error_code)
> 
> What do you think about just taking pt_regs and accessing
> regs->orig_eax instead of having separate call convention for pt_regs
> ones and trap ones?  Too much work without enough benefit?

Looks worthwile to me. [ Cleanups rarely have clear benefits of their own,
it's the sheer mass of them that makes a difference in the end. Like the
many snowflakes that can bend a tree ;-) ]

There's one small namespace complication here: it's pt_regs->orig_eax on
32-bit and pt_regs->orig_rax on 64-bit.

So i'd suggest the introduction of an anonymous union "error_code" field
which overlays orig_eax and that can be used just fine from unified code too.
That would also be more readable.

Or we could rename orig_eax/orig_rax to error_code. (it might be a bit
confusing in the syscall entry context though.)

	Ingo