Re: Anyone working on ftrace function graph support on ARM?

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Tue, Mar 24, 2009 at 11:14:39PM +0100, Ingo Molnar wrote:
> So the rule is: mcount must not destroy _any_ register state. 
> (beyond flags)

As has already been discussed on the ARM lists, we have a problem with
using mcount along with EABI, or OABI without frame pointers.  We can
not avoid destroying 'lr' - the return address for the function.

OABI works around this by assuming that the parent function saved a
stack frame, which means it's at a known offset from the frame pointer
(provided frame pointers are enabled.)  mcount reloads 'lr' before
returning:

ENTRY(mcount)
        stmdb sp!, {r0-r3, lr}
        mov r0, lr
        sub r0, r0, #MCOUNT_INSN_SIZE

        .globl mcount_call
mcount_call:
        bl ftrace_stub
        ldr lr, [fp, #-4]                       @ restore lr
        ldmia sp!, {r0-r3, pc}

So I think if Tim is expecting to be able to use this feature on EABI
kernels, he's going to struggle.