lkml.org 
[lkml]   [2016]   [May]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC PATCH v2 1/3] arch/powerpc : Add detour buffer support for optprobes
On 2016/05/19 08:40PM, Anju T wrote:
> Detour buffer contains instructions to create an in memory pt_regs.
> After the execution of prehandler a call is made for instruction emulation.
> The NIP is decided after the probed instruction is executed. Hence a branch
> instruction is created to the NIP returned by emulate_step().
>
> Instruction slot for detour buffer is allocated from
> the reserved area. For the time being 64KB is reserved
> in memory for this purpose.
>
> Signed-off-by: Anju T <anju@linux.vnet.ibm.com>
> ---
> arch/powerpc/include/asm/kprobes.h | 25 ++++++++
> arch/powerpc/kernel/optprobes_head.S | 108 +++++++++++++++++++++++++++++++++++
> 2 files changed, 133 insertions(+)
> create mode 100644 arch/powerpc/kernel/optprobes_head.S
>
> diff --git a/arch/powerpc/include/asm/kprobes.h b/arch/powerpc/include/asm/kprobes.h
> index 039b583..3e4c998 100644
> --- a/arch/powerpc/include/asm/kprobes.h
> +++ b/arch/powerpc/include/asm/kprobes.h
> @@ -38,7 +38,25 @@ struct pt_regs;
> struct kprobe;
>
> typedef ppc_opcode_t kprobe_opcode_t;
> +
> +extern kprobe_opcode_t optinsn_slot;
> +/* Optinsn template address */
> +extern kprobe_opcode_t optprobe_template_entry[];
> +extern kprobe_opcode_t optprobe_template_call_handler[];
> +extern kprobe_opcode_t optprobe_template_call_emulate[];
> +extern kprobe_opcode_t optprobe_template_ret_branch[];
> +extern kprobe_opcode_t optprobe_template_ret[];
> +extern kprobe_opcode_t optprobe_template_insn[];
> +extern kprobe_opcode_t optprobe_template_op_address1[];
> +extern kprobe_opcode_t optprobe_template_op_address2[];
> +extern kprobe_opcode_t optprobe_template_end[];
> +
> #define MAX_INSN_SIZE 1
> +#define MAX_OPTIMIZED_LENGTH 4
> +#define MAX_OPTINSN_SIZE \
> + ((unsigned long)&optprobe_template_end - \
> + (unsigned long)&optprobe_template_entry)
> +#define RELATIVEJUMP_SIZE 4
>
> #ifdef CONFIG_PPC64
> #if defined(_CALL_ELF) && _CALL_ELF == 2
> @@ -129,5 +147,12 @@ struct kprobe_ctlblk {
> extern int kprobe_exceptions_notify(struct notifier_block *self,
> unsigned long val, void *data);
> extern int kprobe_fault_handler(struct pt_regs *regs, int trapnr);
> +
> +struct arch_optimized_insn {
> + kprobe_opcode_t copied_insn[1];
> + /* detour buffer */
> + kprobe_opcode_t *insn;
> +};
> +
> #endif /* __KERNEL__ */
> #endif /* _ASM_POWERPC_KPROBES_H */
> diff --git a/arch/powerpc/kernel/optprobes_head.S b/arch/powerpc/kernel/optprobes_head.S
> new file mode 100644
> index 0000000..ce32aec
> --- /dev/null
> +++ b/arch/powerpc/kernel/optprobes_head.S
> @@ -0,0 +1,108 @@
> +/*
> + * Code to prepare detour buffer for optprobes in kernel.
> + *
> + * Copyright 2016, Anju T, IBM Corp.
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation; either version
> + * 2 of the License, or (at your option) any later version.
> + */
> +
> +#include <asm/ppc_asm.h>
> +#include <asm/ptrace.h>
> +#include <asm/asm-offsets.h>
> +
> +.global optinsn_slot
> +optinsn_slot:
> + /* Reserve an area to allocate slots for detour buffer */
> + .space 65536
> +.global optprobe_template_entry
> +optprobe_template_entry:
> + stdu r1,-INT_FRAME_SIZE(r1)
> + SAVE_GPR(0,r1)
> + /* Save the previous SP into stack */
> + addi r0,r1,INT_FRAME_SIZE
> + std 0,GPR1(r1)
> + SAVE_2GPRS(2,r1)
> + SAVE_8GPRS(4,r1)
> + SAVE_10GPRS(12,r1)
> + SAVE_10GPRS(22,r1)
> + /* Save SPRS */
> + mfcfar r5
> + std r5,_NIP(r1)

I think this will not work. Looking through the code, I don't see us
saving/restoring CFAR across interrupts.

Also, per the ISA:
"The Come-From Address Register (CFAR) is a 64-bit
register. When an rfebb, rfid, or rfscv instruction is
executed, the register is set to the effective address of
the instruction."

So, it looks like we actually can't save/restore it anyway.

Regardless, NIP is the same as the kprobe'd address, so you should be
able to either use immediate loads or load the address from kprobe
structure.


- Naveen

\
 
 \ /
  Last update: 2016-05-24 13:41    [W:0.074 / U:0.332 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site