linux/arch/powerpc/kernel/hw_breakpoint.c
Jiri Olsa 5aab90ce1e perf/powerpc: Don't call perf_event_disable() from atomic context
The trinity syscall fuzzer triggered following WARN() on powerpc:

  WARNING: CPU: 9 PID: 2998 at arch/powerpc/kernel/hw_breakpoint.c:278
  ...
  NIP [c00000000093aedc] .hw_breakpoint_handler+0x28c/0x2b0
  LR [c00000000093aed8] .hw_breakpoint_handler+0x288/0x2b0
  Call Trace:
  [c0000002f7933580] [c00000000093aed8] .hw_breakpoint_handler+0x288/0x2b0 (unreliable)
  [c0000002f7933630] [c0000000000f671c] .notifier_call_chain+0x7c/0xf0
  [c0000002f79336d0] [c0000000000f6abc] .__atomic_notifier_call_chain+0xbc/0x1c0
  [c0000002f7933780] [c0000000000f6c40] .notify_die+0x70/0xd0
  [c0000002f7933820] [c00000000001a74c] .do_break+0x4c/0x100
  [c0000002f7933920] [c0000000000089fc] handle_dabr_fault+0x14/0x48

Followed by a lockdep warning:

  ===============================
  [ INFO: suspicious RCU usage. ]
  4.8.0-rc5+ #7 Tainted: G        W
  -------------------------------
  ./include/linux/rcupdate.h:556 Illegal context switch in RCU read-side critical section!

  other info that might help us debug this:

  rcu_scheduler_active = 1, debug_locks = 0
  2 locks held by ls/2998:
   #0:  (rcu_read_lock){......}, at: [<c0000000000f6a00>] .__atomic_notifier_call_chain+0x0/0x1c0
   #1:  (rcu_read_lock){......}, at: [<c00000000093ac50>] .hw_breakpoint_handler+0x0/0x2b0

  stack backtrace:
  CPU: 9 PID: 2998 Comm: ls Tainted: G        W       4.8.0-rc5+ #7
  Call Trace:
  [c0000002f7933150] [c00000000094b1f8] .dump_stack+0xe0/0x14c (unreliable)
  [c0000002f79331e0] [c00000000013c468] .lockdep_rcu_suspicious+0x138/0x180
  [c0000002f7933270] [c0000000001005d8] .___might_sleep+0x278/0x2e0
  [c0000002f7933300] [c000000000935584] .mutex_lock_nested+0x64/0x5a0
  [c0000002f7933410] [c00000000023084c] .perf_event_ctx_lock_nested+0x16c/0x380
  [c0000002f7933500] [c000000000230a80] .perf_event_disable+0x20/0x60
  [c0000002f7933580] [c00000000093aeec] .hw_breakpoint_handler+0x29c/0x2b0
  [c0000002f7933630] [c0000000000f671c] .notifier_call_chain+0x7c/0xf0
  [c0000002f79336d0] [c0000000000f6abc] .__atomic_notifier_call_chain+0xbc/0x1c0
  [c0000002f7933780] [c0000000000f6c40] .notify_die+0x70/0xd0
  [c0000002f7933820] [c00000000001a74c] .do_break+0x4c/0x100
  [c0000002f7933920] [c0000000000089fc] handle_dabr_fault+0x14/0x48

While it looks like the first WARN() is probably valid, the other one is
triggered by disabling event via perf_event_disable() from atomic context.

The event is disabled here in case we were not able to emulate
the instruction that hit the breakpoint. By disabling the event
we unschedule the event and make sure it's not scheduled back.

But we can't call perf_event_disable() from atomic context, instead
we need to use the event's pending_disable irq_work method to disable it.

Reported-by: Jan Stancek <jstancek@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Neuling <mikey@neuling.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20161026094824.GA21397@krava
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-28 11:06:25 +02:00

371 lines
9.6 KiB
C

/*
* HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
* using the CPU's debug registers. Derived from
* "arch/x86/kernel/hw_breakpoint.c"
*
* 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.
*
* 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.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
* Copyright 2010 IBM Corporation
* Author: K.Prasad <prasad@linux.vnet.ibm.com>
*
*/
#include <linux/hw_breakpoint.h>
#include <linux/notifier.h>
#include <linux/kprobes.h>
#include <linux/percpu.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/smp.h>
#include <asm/hw_breakpoint.h>
#include <asm/processor.h>
#include <asm/sstep.h>
#include <asm/uaccess.h>
/*
* Stores the breakpoints currently in use on each breakpoint address
* register for every cpu
*/
static DEFINE_PER_CPU(struct perf_event *, bp_per_reg);
/*
* Returns total number of data or instruction breakpoints available.
*/
int hw_breakpoint_slots(int type)
{
if (type == TYPE_DATA)
return HBP_NUM;
return 0; /* no instruction breakpoints available */
}
/*
* Install a perf counter breakpoint.
*
* We seek a free debug address register and use it for this
* breakpoint.
*
* Atomic: we hold the counter->ctx->lock and we only handle variables
* and registers local to this cpu.
*/
int arch_install_hw_breakpoint(struct perf_event *bp)
{
struct arch_hw_breakpoint *info = counter_arch_bp(bp);
struct perf_event **slot = this_cpu_ptr(&bp_per_reg);
*slot = bp;
/*
* Do not install DABR values if the instruction must be single-stepped.
* If so, DABR will be populated in single_step_dabr_instruction().
*/
if (current->thread.last_hit_ubp != bp)
__set_breakpoint(info);
return 0;
}
/*
* Uninstall the breakpoint contained in the given counter.
*
* First we search the debug address register it uses and then we disable
* it.
*
* Atomic: we hold the counter->ctx->lock and we only handle variables
* and registers local to this cpu.
*/
void arch_uninstall_hw_breakpoint(struct perf_event *bp)
{
struct perf_event **slot = this_cpu_ptr(&bp_per_reg);
if (*slot != bp) {
WARN_ONCE(1, "Can't find the breakpoint");
return;
}
*slot = NULL;
hw_breakpoint_disable();
}
/*
* Perform cleanup of arch-specific counters during unregistration
* of the perf-event
*/
void arch_unregister_hw_breakpoint(struct perf_event *bp)
{
/*
* If the breakpoint is unregistered between a hw_breakpoint_handler()
* and the single_step_dabr_instruction(), then cleanup the breakpoint
* restoration variables to prevent dangling pointers.
* FIXME, this should not be using bp->ctx at all! Sayeth peterz.
*/
if (bp->ctx && bp->ctx->task && bp->ctx->task != ((void *)-1L))
bp->ctx->task->thread.last_hit_ubp = NULL;
}
/*
* Check for virtual address in kernel space.
*/
int arch_check_bp_in_kernelspace(struct perf_event *bp)
{
struct arch_hw_breakpoint *info = counter_arch_bp(bp);
return is_kernel_addr(info->address);
}
int arch_bp_generic_fields(int type, int *gen_bp_type)
{
*gen_bp_type = 0;
if (type & HW_BRK_TYPE_READ)
*gen_bp_type |= HW_BREAKPOINT_R;
if (type & HW_BRK_TYPE_WRITE)
*gen_bp_type |= HW_BREAKPOINT_W;
if (*gen_bp_type == 0)
return -EINVAL;
return 0;
}
/*
* Validate the arch-specific HW Breakpoint register settings
*/
int arch_validate_hwbkpt_settings(struct perf_event *bp)
{
int ret = -EINVAL, length_max;
struct arch_hw_breakpoint *info = counter_arch_bp(bp);
if (!bp)
return ret;
info->type = HW_BRK_TYPE_TRANSLATE;
if (bp->attr.bp_type & HW_BREAKPOINT_R)
info->type |= HW_BRK_TYPE_READ;
if (bp->attr.bp_type & HW_BREAKPOINT_W)
info->type |= HW_BRK_TYPE_WRITE;
if (info->type == HW_BRK_TYPE_TRANSLATE)
/* must set alteast read or write */
return ret;
if (!(bp->attr.exclude_user))
info->type |= HW_BRK_TYPE_USER;
if (!(bp->attr.exclude_kernel))
info->type |= HW_BRK_TYPE_KERNEL;
if (!(bp->attr.exclude_hv))
info->type |= HW_BRK_TYPE_HYP;
info->address = bp->attr.bp_addr;
info->len = bp->attr.bp_len;
/*
* Since breakpoint length can be a maximum of HW_BREAKPOINT_LEN(8)
* and breakpoint addresses are aligned to nearest double-word
* HW_BREAKPOINT_ALIGN by rounding off to the lower address, the
* 'symbolsize' should satisfy the check below.
*/
length_max = 8; /* DABR */
if (cpu_has_feature(CPU_FTR_DAWR)) {
length_max = 512 ; /* 64 doublewords */
/* DAWR region can't cross 512 boundary */
if ((bp->attr.bp_addr >> 10) !=
((bp->attr.bp_addr + bp->attr.bp_len - 1) >> 10))
return -EINVAL;
}
if (info->len >
(length_max - (info->address & HW_BREAKPOINT_ALIGN)))
return -EINVAL;
return 0;
}
/*
* Restores the breakpoint on the debug registers.
* Invoke this function if it is known that the execution context is
* about to change to cause loss of MSR_SE settings.
*/
void thread_change_pc(struct task_struct *tsk, struct pt_regs *regs)
{
struct arch_hw_breakpoint *info;
if (likely(!tsk->thread.last_hit_ubp))
return;
info = counter_arch_bp(tsk->thread.last_hit_ubp);
regs->msr &= ~MSR_SE;
__set_breakpoint(info);
tsk->thread.last_hit_ubp = NULL;
}
/*
* Handle debug exception notifications.
*/
int hw_breakpoint_handler(struct die_args *args)
{
int rc = NOTIFY_STOP;
struct perf_event *bp;
struct pt_regs *regs = args->regs;
int stepped = 1;
struct arch_hw_breakpoint *info;
unsigned int instr;
unsigned long dar = regs->dar;
/* Disable breakpoints during exception handling */
hw_breakpoint_disable();
/*
* The counter may be concurrently released but that can only
* occur from a call_rcu() path. We can then safely fetch
* the breakpoint, use its callback, touch its counter
* while we are in an rcu_read_lock() path.
*/
rcu_read_lock();
bp = __this_cpu_read(bp_per_reg);
if (!bp)
goto out;
info = counter_arch_bp(bp);
/*
* Return early after invoking user-callback function without restoring
* DABR if the breakpoint is from ptrace which always operates in
* one-shot mode. The ptrace-ed process will receive the SIGTRAP signal
* generated in do_dabr().
*/
if (bp->overflow_handler == ptrace_triggered) {
perf_bp_event(bp, regs);
rc = NOTIFY_DONE;
goto out;
}
/*
* Verify if dar lies within the address range occupied by the symbol
* being watched to filter extraneous exceptions. If it doesn't,
* we still need to single-step the instruction, but we don't
* generate an event.
*/
info->type &= ~HW_BRK_TYPE_EXTRANEOUS_IRQ;
if (!((bp->attr.bp_addr <= dar) &&
(dar - bp->attr.bp_addr < bp->attr.bp_len)))
info->type |= HW_BRK_TYPE_EXTRANEOUS_IRQ;
/* Do not emulate user-space instructions, instead single-step them */
if (user_mode(regs)) {
current->thread.last_hit_ubp = bp;
regs->msr |= MSR_SE;
goto out;
}
stepped = 0;
instr = 0;
if (!__get_user_inatomic(instr, (unsigned int *) regs->nip))
stepped = emulate_step(regs, instr);
/*
* emulate_step() could not execute it. We've failed in reliably
* handling the hw-breakpoint. Unregister it and throw a warning
* message to let the user know about it.
*/
if (!stepped) {
WARN(1, "Unable to handle hardware breakpoint. Breakpoint at "
"0x%lx will be disabled.", info->address);
perf_event_disable_inatomic(bp);
goto out;
}
/*
* As a policy, the callback is invoked in a 'trigger-after-execute'
* fashion
*/
if (!(info->type & HW_BRK_TYPE_EXTRANEOUS_IRQ))
perf_bp_event(bp, regs);
__set_breakpoint(info);
out:
rcu_read_unlock();
return rc;
}
NOKPROBE_SYMBOL(hw_breakpoint_handler);
/*
* Handle single-step exceptions following a DABR hit.
*/
static int single_step_dabr_instruction(struct die_args *args)
{
struct pt_regs *regs = args->regs;
struct perf_event *bp = NULL;
struct arch_hw_breakpoint *info;
bp = current->thread.last_hit_ubp;
/*
* Check if we are single-stepping as a result of a
* previous HW Breakpoint exception
*/
if (!bp)
return NOTIFY_DONE;
info = counter_arch_bp(bp);
/*
* We shall invoke the user-defined callback function in the single
* stepping handler to confirm to 'trigger-after-execute' semantics
*/
if (!(info->type & HW_BRK_TYPE_EXTRANEOUS_IRQ))
perf_bp_event(bp, regs);
__set_breakpoint(info);
current->thread.last_hit_ubp = NULL;
/*
* If the process was being single-stepped by ptrace, let the
* other single-step actions occur (e.g. generate SIGTRAP).
*/
if (test_thread_flag(TIF_SINGLESTEP))
return NOTIFY_DONE;
return NOTIFY_STOP;
}
NOKPROBE_SYMBOL(single_step_dabr_instruction);
/*
* Handle debug exception notifications.
*/
int hw_breakpoint_exceptions_notify(
struct notifier_block *unused, unsigned long val, void *data)
{
int ret = NOTIFY_DONE;
switch (val) {
case DIE_DABR_MATCH:
ret = hw_breakpoint_handler(data);
break;
case DIE_SSTEP:
ret = single_step_dabr_instruction(data);
break;
}
return ret;
}
NOKPROBE_SYMBOL(hw_breakpoint_exceptions_notify);
/*
* Release the user breakpoints used by ptrace
*/
void flush_ptrace_hw_breakpoint(struct task_struct *tsk)
{
struct thread_struct *t = &tsk->thread;
unregister_hw_breakpoint(t->ptrace_bps[0]);
t->ptrace_bps[0] = NULL;
}
void hw_breakpoint_pmu_read(struct perf_event *bp)
{
/* TODO */
}