On 10/13/2011 02:07 AM, Maneesh Soni wrote:
From: Maneesh Soni<manesoni@xxxxxxxxx>
This patch provides support for kprobes on branch instructions. The branch
instruction at the probed address is actually emulated and not executed
out-of-line like other normal instructions. Instead the delay-slot instruction
is copied and single stepped out of line.
At the time of probe hit, the original branch instruction is evaluated
and the target cp0_epc is computed similar to compute_retrun_epc(). It
is also checked if the delay slot instruction can be skipped, which is
true if there is a NOP in delay slot or branch is taken in case of
branch likely instructions. Once the delay slot instruction is single
stepped the normal execution resume with the cp0_epc updated the earlier
computed cp0_epc as per the branch instructions.
I haven't tested it but...
Signed-off-by: Maneesh Soni<manesoni@xxxxxxxxx>
---
arch/mips/include/asm/kprobes.h | 7 +
arch/mips/kernel/kprobes.c | 341 +++++++++++++++++++++++++++++++++++----
2 files changed, 320 insertions(+), 28 deletions(-)
[...]
+static int evaluate_branch_instruction(struct kprobe *p, struct pt_regs *regs,
+ struct kprobe_ctlblk *kcb)
{
+ union mips_instruction insn = p->opcode;
+ unsigned int dspcontrol;
+ long epc;
+
+ epc = regs->cp0_epc;
+ if (epc& 3)
+ goto unaligned;
+
+ if (p->ainsn.insn->word == 0)
+ kcb->flags |= SKIP_DELAYSLOT;
+ else
+ kcb->flags&= ~SKIP_DELAYSLOT;
+
+ switch (insn.i_format.opcode) {
+ /*
+ * jr and jalr are in r_format format.
+ */
+ case spec_op:
[...]
+ case bgtzl_op:
+ /* rt field assumed to be zero */
+ if ((long)regs->regs[insn.i_format.rs]> 0) {
+ epc = epc + 4 + (insn.i_format.simmediate<< 2);
+ kcb->flags |= SKIP_DELAYSLOT;
+ } else
+ epc += 8;
+ regs->cp0_epc = epc;
+ break;
Where is the handling for:
case cop1_op:
#ifdef CONFIG_CPU_CAVIUM_OCTEON
case lwc2_op: /* This is bbit0 on Octeon */
case ldc2_op: /* This is bbit032 on Octeon */
case swc2_op: /* This is bbit1 on Octeon */
case sdc2_op: /* This is bbit132 on Octeon */
#endif
These are all defined in insn_has_delayslot() but not here.
