alistair23-linux/arch/arm/probes/kprobes/actions-arm.c

345 lines
11 KiB
C
Raw Normal View History

ARM kprobes: instruction single-stepping support This is the code implementing instruction single-stepping for kprobes on ARM. To get around the limitation of no Next-PC and no hardware single- stepping, all kprobe'd instructions are split into three camps: simulation, emulation, and rejected. "Simulated" instructions are those instructions which behavior is reproduced by straight C code. "Emulated" instructions are ones that are copied, slightly altered and executed directly in the instruction slot to reproduce their behavior. "Rejected" instructions are ones that could be simulated, but work hasn't been put into simulating them. These instructions should be very rare, if not unencountered, in the kernel. If ever needed, code could be added to simulate them. One might wonder why this and the ptrace singlestep facility are not sharing some code. Both approaches are fundamentally different because the ptrace code regains control after the stepped instruction by installing a breakpoint after the instruction itself, and possibly at the location where the instruction might be branching to, instead of simulating or emulating the target instruction. The ptrace approach isn't suitable for kprobes because the breakpoints would have to be moved back, and the icache flushed, everytime the probe is hit to let normal code execution resume, which would have a significant performance impact. It is also racy on SMP since another CPU could, with the right timing, sail through the probe point without being caught. Because ptrace single-stepping always result in a different process to be scheduled, the concern for performance is much less significant. On the other hand, the kprobes approach isn't (currently) suitable for ptrace because it has no provision for proper user space memory protection and translation, and even if that was implemented, the gain wouldn't be worth the added complexity in the ptrace path compared to the current approach. So, until kprobes does support user space, both kprobes and ptrace are best kept independent and separate. Signed-off-by: Quentin Barnes <qbarnes@gmail.com> Signed-off-by: Abhishek Sagar <sagar.abhishek@gmail.com> Signed-off-by: Nicolas Pitre <nico@marvell.com>
2007-06-11 16:20:10 -06:00
/*
* arch/arm/probes/kprobes/actions-arm.c
ARM kprobes: instruction single-stepping support This is the code implementing instruction single-stepping for kprobes on ARM. To get around the limitation of no Next-PC and no hardware single- stepping, all kprobe'd instructions are split into three camps: simulation, emulation, and rejected. "Simulated" instructions are those instructions which behavior is reproduced by straight C code. "Emulated" instructions are ones that are copied, slightly altered and executed directly in the instruction slot to reproduce their behavior. "Rejected" instructions are ones that could be simulated, but work hasn't been put into simulating them. These instructions should be very rare, if not unencountered, in the kernel. If ever needed, code could be added to simulate them. One might wonder why this and the ptrace singlestep facility are not sharing some code. Both approaches are fundamentally different because the ptrace code regains control after the stepped instruction by installing a breakpoint after the instruction itself, and possibly at the location where the instruction might be branching to, instead of simulating or emulating the target instruction. The ptrace approach isn't suitable for kprobes because the breakpoints would have to be moved back, and the icache flushed, everytime the probe is hit to let normal code execution resume, which would have a significant performance impact. It is also racy on SMP since another CPU could, with the right timing, sail through the probe point without being caught. Because ptrace single-stepping always result in a different process to be scheduled, the concern for performance is much less significant. On the other hand, the kprobes approach isn't (currently) suitable for ptrace because it has no provision for proper user space memory protection and translation, and even if that was implemented, the gain wouldn't be worth the added complexity in the ptrace path compared to the current approach. So, until kprobes does support user space, both kprobes and ptrace are best kept independent and separate. Signed-off-by: Quentin Barnes <qbarnes@gmail.com> Signed-off-by: Abhishek Sagar <sagar.abhishek@gmail.com> Signed-off-by: Nicolas Pitre <nico@marvell.com>
2007-06-11 16:20:10 -06:00
*
* Copyright (C) 2006, 2007 Motorola Inc.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*/
/*
* We do not have hardware single-stepping on ARM, This
* effort is further complicated by the ARM not having a
* "next PC" register. Instructions that change the PC
* can't be safely single-stepped in a MP environment, so
* we have a lot of work to do:
*
* In the prepare phase:
* *) If it is an instruction that does anything
* with the CPU mode, we reject it for a kprobe.
* (This is out of laziness rather than need. The
* instructions could be simulated.)
*
* *) Otherwise, decode the instruction rewriting its
* registers to take fixed, ordered registers and
* setting a handler for it to run the instruction.
*
* In the execution phase by an instruction's handler:
*
* *) If the PC is written to by the instruction, the
* instruction must be fully simulated in software.
*
* *) Otherwise, a modified form of the instruction is
* directly executed. Its handler calls the
* instruction in insn[0]. In insn[1] is a
* "mov pc, lr" to return.
*
* Before calling, load up the reordered registers
* from the original instruction's registers. If one
* of the original input registers is the PC, compute
* and adjust the appropriate input register.
*
* After call completes, copy the output registers to
* the original instruction's original registers.
*
* We don't use a real breakpoint instruction since that
* would have us in the kernel go from SVC mode to SVC
* mode losing the link register. Instead we use an
* undefined instruction. To simplify processing, the
* undefined instruction used for kprobes must be reserved
* exclusively for kprobes use.
*
* TODO: ifdef out some instruction decoding based on architecture.
*/
#include <linux/kernel.h>
#include <linux/kprobes.h>
#include <linux/ptrace.h>
ARM kprobes: instruction single-stepping support This is the code implementing instruction single-stepping for kprobes on ARM. To get around the limitation of no Next-PC and no hardware single- stepping, all kprobe'd instructions are split into three camps: simulation, emulation, and rejected. "Simulated" instructions are those instructions which behavior is reproduced by straight C code. "Emulated" instructions are ones that are copied, slightly altered and executed directly in the instruction slot to reproduce their behavior. "Rejected" instructions are ones that could be simulated, but work hasn't been put into simulating them. These instructions should be very rare, if not unencountered, in the kernel. If ever needed, code could be added to simulate them. One might wonder why this and the ptrace singlestep facility are not sharing some code. Both approaches are fundamentally different because the ptrace code regains control after the stepped instruction by installing a breakpoint after the instruction itself, and possibly at the location where the instruction might be branching to, instead of simulating or emulating the target instruction. The ptrace approach isn't suitable for kprobes because the breakpoints would have to be moved back, and the icache flushed, everytime the probe is hit to let normal code execution resume, which would have a significant performance impact. It is also racy on SMP since another CPU could, with the right timing, sail through the probe point without being caught. Because ptrace single-stepping always result in a different process to be scheduled, the concern for performance is much less significant. On the other hand, the kprobes approach isn't (currently) suitable for ptrace because it has no provision for proper user space memory protection and translation, and even if that was implemented, the gain wouldn't be worth the added complexity in the ptrace path compared to the current approach. So, until kprobes does support user space, both kprobes and ptrace are best kept independent and separate. Signed-off-by: Quentin Barnes <qbarnes@gmail.com> Signed-off-by: Abhishek Sagar <sagar.abhishek@gmail.com> Signed-off-by: Nicolas Pitre <nico@marvell.com>
2007-06-11 16:20:10 -06:00
#include "../decode-arm.h"
#include "core.h"
#include "checkers.h"
ARM kprobes: instruction single-stepping support This is the code implementing instruction single-stepping for kprobes on ARM. To get around the limitation of no Next-PC and no hardware single- stepping, all kprobe'd instructions are split into three camps: simulation, emulation, and rejected. "Simulated" instructions are those instructions which behavior is reproduced by straight C code. "Emulated" instructions are ones that are copied, slightly altered and executed directly in the instruction slot to reproduce their behavior. "Rejected" instructions are ones that could be simulated, but work hasn't been put into simulating them. These instructions should be very rare, if not unencountered, in the kernel. If ever needed, code could be added to simulate them. One might wonder why this and the ptrace singlestep facility are not sharing some code. Both approaches are fundamentally different because the ptrace code regains control after the stepped instruction by installing a breakpoint after the instruction itself, and possibly at the location where the instruction might be branching to, instead of simulating or emulating the target instruction. The ptrace approach isn't suitable for kprobes because the breakpoints would have to be moved back, and the icache flushed, everytime the probe is hit to let normal code execution resume, which would have a significant performance impact. It is also racy on SMP since another CPU could, with the right timing, sail through the probe point without being caught. Because ptrace single-stepping always result in a different process to be scheduled, the concern for performance is much less significant. On the other hand, the kprobes approach isn't (currently) suitable for ptrace because it has no provision for proper user space memory protection and translation, and even if that was implemented, the gain wouldn't be worth the added complexity in the ptrace path compared to the current approach. So, until kprobes does support user space, both kprobes and ptrace are best kept independent and separate. Signed-off-by: Quentin Barnes <qbarnes@gmail.com> Signed-off-by: Abhishek Sagar <sagar.abhishek@gmail.com> Signed-off-by: Nicolas Pitre <nico@marvell.com>
2007-06-11 16:20:10 -06:00
#if __LINUX_ARM_ARCH__ >= 6
#define BLX(reg) "blx "reg" \n\t"
#else
#define BLX(reg) "mov lr, pc \n\t" \
"mov pc, "reg" \n\t"
#endif
static void __kprobes
emulate_ldrdstrd(probes_opcode_t insn,
struct arch_probes_insn *asi, struct pt_regs *regs)
{
unsigned long pc = regs->ARM_pc + 4;
int rt = (insn >> 12) & 0xf;
int rn = (insn >> 16) & 0xf;
int rm = insn & 0xf;
register unsigned long rtv asm("r0") = regs->uregs[rt];
register unsigned long rt2v asm("r1") = regs->uregs[rt+1];
register unsigned long rnv asm("r2") = (rn == 15) ? pc
: regs->uregs[rn];
register unsigned long rmv asm("r3") = regs->uregs[rm];
__asm__ __volatile__ (
BLX("%[fn]")
: "=r" (rtv), "=r" (rt2v), "=r" (rnv)
: "0" (rtv), "1" (rt2v), "2" (rnv), "r" (rmv),
[fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
regs->uregs[rt] = rtv;
regs->uregs[rt+1] = rt2v;
if (is_writeback(insn))
regs->uregs[rn] = rnv;
}
static void __kprobes
emulate_ldr(probes_opcode_t insn,
struct arch_probes_insn *asi, struct pt_regs *regs)
{
unsigned long pc = regs->ARM_pc + 4;
int rt = (insn >> 12) & 0xf;
int rn = (insn >> 16) & 0xf;
int rm = insn & 0xf;
register unsigned long rtv asm("r0");
register unsigned long rnv asm("r2") = (rn == 15) ? pc
: regs->uregs[rn];
register unsigned long rmv asm("r3") = regs->uregs[rm];
__asm__ __volatile__ (
BLX("%[fn]")
: "=r" (rtv), "=r" (rnv)
: "1" (rnv), "r" (rmv), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
if (rt == 15)
load_write_pc(rtv, regs);
else
regs->uregs[rt] = rtv;
if (is_writeback(insn))
regs->uregs[rn] = rnv;
}
static void __kprobes
emulate_str(probes_opcode_t insn,
struct arch_probes_insn *asi, struct pt_regs *regs)
{
unsigned long rtpc = regs->ARM_pc - 4 + str_pc_offset;
unsigned long rnpc = regs->ARM_pc + 4;
int rt = (insn >> 12) & 0xf;
int rn = (insn >> 16) & 0xf;
int rm = insn & 0xf;
register unsigned long rtv asm("r0") = (rt == 15) ? rtpc
: regs->uregs[rt];
register unsigned long rnv asm("r2") = (rn == 15) ? rnpc
: regs->uregs[rn];
register unsigned long rmv asm("r3") = regs->uregs[rm];
__asm__ __volatile__ (
BLX("%[fn]")
: "=r" (rnv)
: "r" (rtv), "0" (rnv), "r" (rmv), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
if (is_writeback(insn))
regs->uregs[rn] = rnv;
}
static void __kprobes
emulate_rd12rn16rm0rs8_rwflags(probes_opcode_t insn,
struct arch_probes_insn *asi, struct pt_regs *regs)
{
unsigned long pc = regs->ARM_pc + 4;
int rd = (insn >> 12) & 0xf;
int rn = (insn >> 16) & 0xf;
int rm = insn & 0xf;
int rs = (insn >> 8) & 0xf;
register unsigned long rdv asm("r0") = regs->uregs[rd];
register unsigned long rnv asm("r2") = (rn == 15) ? pc
: regs->uregs[rn];
register unsigned long rmv asm("r3") = (rm == 15) ? pc
: regs->uregs[rm];
register unsigned long rsv asm("r1") = regs->uregs[rs];
unsigned long cpsr = regs->ARM_cpsr;
__asm__ __volatile__ (
"msr cpsr_fs, %[cpsr] \n\t"
BLX("%[fn]")
"mrs %[cpsr], cpsr \n\t"
: "=r" (rdv), [cpsr] "=r" (cpsr)
: "0" (rdv), "r" (rnv), "r" (rmv), "r" (rsv),
"1" (cpsr), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
if (rd == 15)
alu_write_pc(rdv, regs);
else
regs->uregs[rd] = rdv;
regs->ARM_cpsr = (regs->ARM_cpsr & ~APSR_MASK) | (cpsr & APSR_MASK);
}
static void __kprobes
emulate_rd12rn16rm0_rwflags_nopc(probes_opcode_t insn,
struct arch_probes_insn *asi, struct pt_regs *regs)
{
int rd = (insn >> 12) & 0xf;
int rn = (insn >> 16) & 0xf;
int rm = insn & 0xf;
register unsigned long rdv asm("r0") = regs->uregs[rd];
register unsigned long rnv asm("r2") = regs->uregs[rn];
register unsigned long rmv asm("r3") = regs->uregs[rm];
unsigned long cpsr = regs->ARM_cpsr;
__asm__ __volatile__ (
"msr cpsr_fs, %[cpsr] \n\t"
BLX("%[fn]")
"mrs %[cpsr], cpsr \n\t"
: "=r" (rdv), [cpsr] "=r" (cpsr)
: "0" (rdv), "r" (rnv), "r" (rmv),
"1" (cpsr), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
regs->uregs[rd] = rdv;
regs->ARM_cpsr = (regs->ARM_cpsr & ~APSR_MASK) | (cpsr & APSR_MASK);
}
static void __kprobes
emulate_rd16rn12rm0rs8_rwflags_nopc(probes_opcode_t insn,
struct arch_probes_insn *asi,
struct pt_regs *regs)
{
int rd = (insn >> 16) & 0xf;
int rn = (insn >> 12) & 0xf;
int rm = insn & 0xf;
int rs = (insn >> 8) & 0xf;
register unsigned long rdv asm("r2") = regs->uregs[rd];
register unsigned long rnv asm("r0") = regs->uregs[rn];
register unsigned long rmv asm("r3") = regs->uregs[rm];
register unsigned long rsv asm("r1") = regs->uregs[rs];
unsigned long cpsr = regs->ARM_cpsr;
__asm__ __volatile__ (
"msr cpsr_fs, %[cpsr] \n\t"
BLX("%[fn]")
"mrs %[cpsr], cpsr \n\t"
: "=r" (rdv), [cpsr] "=r" (cpsr)
: "0" (rdv), "r" (rnv), "r" (rmv), "r" (rsv),
"1" (cpsr), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
regs->uregs[rd] = rdv;
regs->ARM_cpsr = (regs->ARM_cpsr & ~APSR_MASK) | (cpsr & APSR_MASK);
}
static void __kprobes
emulate_rd12rm0_noflags_nopc(probes_opcode_t insn,
struct arch_probes_insn *asi, struct pt_regs *regs)
{
int rd = (insn >> 12) & 0xf;
int rm = insn & 0xf;
register unsigned long rdv asm("r0") = regs->uregs[rd];
register unsigned long rmv asm("r3") = regs->uregs[rm];
__asm__ __volatile__ (
BLX("%[fn]")
: "=r" (rdv)
: "0" (rdv), "r" (rmv), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
regs->uregs[rd] = rdv;
}
static void __kprobes
emulate_rdlo12rdhi16rn0rm8_rwflags_nopc(probes_opcode_t insn,
struct arch_probes_insn *asi,
struct pt_regs *regs)
{
int rdlo = (insn >> 12) & 0xf;
int rdhi = (insn >> 16) & 0xf;
int rn = insn & 0xf;
int rm = (insn >> 8) & 0xf;
register unsigned long rdlov asm("r0") = regs->uregs[rdlo];
register unsigned long rdhiv asm("r2") = regs->uregs[rdhi];
register unsigned long rnv asm("r3") = regs->uregs[rn];
register unsigned long rmv asm("r1") = regs->uregs[rm];
unsigned long cpsr = regs->ARM_cpsr;
__asm__ __volatile__ (
"msr cpsr_fs, %[cpsr] \n\t"
BLX("%[fn]")
"mrs %[cpsr], cpsr \n\t"
: "=r" (rdlov), "=r" (rdhiv), [cpsr] "=r" (cpsr)
: "0" (rdlov), "1" (rdhiv), "r" (rnv), "r" (rmv),
"2" (cpsr), [fn] "r" (asi->insn_fn)
: "lr", "memory", "cc"
);
regs->uregs[rdlo] = rdlov;
regs->uregs[rdhi] = rdhiv;
regs->ARM_cpsr = (regs->ARM_cpsr & ~APSR_MASK) | (cpsr & APSR_MASK);
}
const union decode_action kprobes_arm_actions[NUM_PROBES_ARM_ACTIONS] = {
[PROBES_PRELOAD_IMM] = {.handler = probes_simulate_nop},
[PROBES_PRELOAD_REG] = {.handler = probes_simulate_nop},
[PROBES_BRANCH_IMM] = {.handler = simulate_blx1},
[PROBES_MRS] = {.handler = simulate_mrs},
[PROBES_BRANCH_REG] = {.handler = simulate_blx2bx},
[PROBES_CLZ] = {.handler = emulate_rd12rm0_noflags_nopc},
[PROBES_SATURATING_ARITHMETIC] = {
.handler = emulate_rd12rn16rm0_rwflags_nopc},
[PROBES_MUL1] = {.handler = emulate_rdlo12rdhi16rn0rm8_rwflags_nopc},
[PROBES_MUL2] = {.handler = emulate_rd16rn12rm0rs8_rwflags_nopc},
[PROBES_SWP] = {.handler = emulate_rd12rn16rm0_rwflags_nopc},
[PROBES_LDRSTRD] = {.handler = emulate_ldrdstrd},
[PROBES_LOAD_EXTRA] = {.handler = emulate_ldr},
[PROBES_LOAD] = {.handler = emulate_ldr},
[PROBES_STORE_EXTRA] = {.handler = emulate_str},
[PROBES_STORE] = {.handler = emulate_str},
[PROBES_MOV_IP_SP] = {.handler = simulate_mov_ipsp},
[PROBES_DATA_PROCESSING_REG] = {
.handler = emulate_rd12rn16rm0rs8_rwflags},
[PROBES_DATA_PROCESSING_IMM] = {
.handler = emulate_rd12rn16rm0rs8_rwflags},
[PROBES_MOV_HALFWORD] = {.handler = emulate_rd12rm0_noflags_nopc},
[PROBES_SEV] = {.handler = probes_emulate_none},
[PROBES_WFE] = {.handler = probes_simulate_nop},
[PROBES_SATURATE] = {.handler = emulate_rd12rn16rm0_rwflags_nopc},
[PROBES_REV] = {.handler = emulate_rd12rm0_noflags_nopc},
[PROBES_MMI] = {.handler = emulate_rd12rn16rm0_rwflags_nopc},
[PROBES_PACK] = {.handler = emulate_rd12rn16rm0_rwflags_nopc},
[PROBES_EXTEND] = {.handler = emulate_rd12rm0_noflags_nopc},
[PROBES_EXTEND_ADD] = {.handler = emulate_rd12rn16rm0_rwflags_nopc},
[PROBES_MUL_ADD_LONG] = {
.handler = emulate_rdlo12rdhi16rn0rm8_rwflags_nopc},
[PROBES_MUL_ADD] = {.handler = emulate_rd16rn12rm0rs8_rwflags_nopc},
[PROBES_BITFIELD] = {.handler = emulate_rd12rm0_noflags_nopc},
[PROBES_BRANCH] = {.handler = simulate_bbl},
[PROBES_LDMSTM] = {.decoder = kprobe_decode_ldmstm}
};
const struct decode_checker *kprobes_arm_checkers[] = {arm_stack_checker, arm_regs_checker, NULL};