Hi,
Until now, x86 and superh32 had to always build hw_breakpoint and perf.
And powerpc and arm were always building hw_breakpoints when perf
was enabled.
After this patchset, one can simply disable hw_breakpoints if he
selected CONFIG_EXPERT. Afterward the user is free to select perf
or not. Except on SH unfortunately because that would require a
bit of change in ptrace to support singlestep failures.
Frederic Weisbecker (6):
hw_breakpoints: Split hardware breakpoints config
hw_breakpoints: Migrate breakpoint conditional build under new config
x86: Allow the user not to build hw_breakpoints
hw_breakpoints: Breakpoints arch ability don't need perf events
hw_breakpoints: Only force perf events if breakpoints are selected
hw_breakpoints: Drop remaining misplaced dependency on perf
arch/Kconfig | 1 -
arch/arm/Kconfig | 2 +-
arch/arm/include/asm/hw_breakpoint.h | 4 +-
arch/arm/include/asm/processor.h | 2 +-
arch/arm/kernel/Makefile | 2 +-
arch/arm/kernel/entry-header.S | 2 +-
arch/arm/kernel/ptrace.c | 4 +-
arch/powerpc/Kconfig | 2 +-
arch/powerpc/include/asm/cputable.h | 4 +-
arch/powerpc/include/asm/hw_breakpoint.h | 6 ++--
arch/powerpc/include/asm/processor.h | 4 +-
arch/powerpc/kernel/Makefile | 2 +-
arch/powerpc/kernel/process.c | 18 ++++++++--------
arch/powerpc/kernel/ptrace.c | 12 +++++-----
arch/powerpc/lib/Makefile | 2 +-
arch/sh/Kconfig | 2 +-
arch/sh/kernel/Makefile | 2 +-
arch/sh/kernel/cpu/sh4a/Makefile | 2 +-
arch/x86/Kconfig | 3 +-
arch/x86/include/asm/debugreg.h | 33 ++++++++++++++++++++++++++++-
arch/x86/kernel/Makefile | 3 +-
arch/x86/kernel/process.c | 1 +
arch/x86/kernel/ptrace.c | 17 +++++++++++++++
include/linux/hw_breakpoint.h | 6 ++--
include/linux/perf_event.h | 4 +-
init/Kconfig | 11 ++++++++++
kernel/Makefile | 2 +-
kernel/perf_event.c | 4 +-
samples/Kconfig | 2 +-
29 files changed, 108 insertions(+), 51 deletions(-)
--
1.7.3.2
Previously, arch were forced to always build perf events if they
supported hw_breakpoints.
Now that the user can choose not to build hw_breakpoints, let
only force perf events if hw_breakpoints are selected.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Prasad <[email protected]>
Cc: Paul Mundt <[email protected]>
---
arch/sh/Kconfig | 1 -
arch/x86/Kconfig | 1 -
init/Kconfig | 1 +
3 files changed, 1 insertions(+), 2 deletions(-)
diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index 0d4d124..d59e6c2 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -46,7 +46,6 @@ config SUPERH32
select HAVE_HW_BREAKPOINT
select HW_BREAKPOINT
select HAVE_MIXED_BREAKPOINTS_REGS
- select PERF_EVENTS
select ARCH_HIBERNATION_POSSIBLE if MMU
select SPARSE_IRQ
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index fb28dd9..5317f42 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -55,7 +55,6 @@ config X86
select HAVE_KERNEL_LZO
select HAVE_HW_BREAKPOINT
select HAVE_MIXED_BREAKPOINTS_REGS
- select PERF_EVENTS
select HAVE_PERF_EVENTS_NMI
select ANON_INODES
select HAVE_ARCH_KMEMCHECK
diff --git a/init/Kconfig b/init/Kconfig
index 51f587f..797c98c 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -927,6 +927,7 @@ menuconfig EXPERT
config HW_BREAKPOINT
bool "Hardware breakpoints" if EXPERT
depends on HAVE_HW_BREAKPOINT
+ select PERF_EVENTS
default y
help
Hardware breakpoints are a feature implemented by most CPUs
--
1.7.3.2
Powerpc and Arm select breakpoint support ability only if
Perf is built. This is not necessary anymore now that we
enable perf once breakpoints support is selected.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Prasad <[email protected]>
Cc: Paul Mundt <[email protected]>
---
arch/arm/Kconfig | 2 +-
arch/powerpc/Kconfig | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index 377a7a5..1d3a0b4 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -24,7 +24,7 @@ config ARM
select HAVE_PERF_EVENTS
select PERF_USE_VMALLOC
select HAVE_REGS_AND_STACK_ACCESS_API
- select HAVE_HW_BREAKPOINT if (PERF_EVENTS && (CPU_V6 || CPU_V6K || CPU_V7))
+ select HAVE_HW_BREAKPOINT if (CPU_V6 || CPU_V6K || CPU_V7)
select HAVE_C_RECORDMCOUNT
select HAVE_GENERIC_HARDIRQS
select HAVE_SPARSE_IRQ
diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 8f4d50b..e944eb8 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -134,7 +134,7 @@ config PPC
select HAVE_IRQ_WORK
select HAVE_PERF_EVENTS
select HAVE_REGS_AND_STACK_ACCESS_API
- select HAVE_HW_BREAKPOINT if PERF_EVENTS && PPC_BOOK3S_64
+ select HAVE_HW_BREAKPOINT if PPC_BOOK3S_64
select HAVE_GENERIC_HARDIRQS
select HAVE_SPARSE_IRQ
select IRQ_PER_CPU
--
1.7.3.2
Migrate conditional hw_breakpoint code compilation under
the new config to prepare for letting the user chose whether
or not to build this feature
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Prasad <[email protected]>
Cc: Paul Mundt <[email protected]>
---
arch/arm/include/asm/hw_breakpoint.h | 4 ++--
arch/arm/include/asm/processor.h | 2 +-
arch/arm/kernel/Makefile | 2 +-
arch/arm/kernel/entry-header.S | 2 +-
arch/arm/kernel/ptrace.c | 4 ++--
arch/powerpc/include/asm/cputable.h | 4 ++--
arch/powerpc/include/asm/hw_breakpoint.h | 6 +++---
arch/powerpc/include/asm/processor.h | 4 ++--
arch/powerpc/kernel/Makefile | 2 +-
arch/powerpc/kernel/process.c | 18 +++++++++---------
arch/powerpc/kernel/ptrace.c | 12 ++++++------
arch/powerpc/lib/Makefile | 2 +-
arch/sh/kernel/Makefile | 2 +-
arch/sh/kernel/cpu/sh4a/Makefile | 2 +-
include/linux/hw_breakpoint.h | 6 +++---
include/linux/perf_event.h | 4 ++--
kernel/Makefile | 2 +-
kernel/perf_event.c | 4 ++--
samples/Kconfig | 2 +-
19 files changed, 42 insertions(+), 42 deletions(-)
diff --git a/arch/arm/include/asm/hw_breakpoint.h b/arch/arm/include/asm/hw_breakpoint.h
index f389b27..fc6ba18 100644
--- a/arch/arm/include/asm/hw_breakpoint.h
+++ b/arch/arm/include/asm/hw_breakpoint.h
@@ -5,7 +5,7 @@
struct task_struct;
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
struct arch_hw_breakpoint_ctrl {
u32 __reserved : 9,
@@ -128,6 +128,6 @@ int hw_breakpoint_slots(int type);
#else
static inline void clear_ptrace_hw_breakpoint(struct task_struct *tsk) {}
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
#endif /* __KERNEL__ */
#endif /* _ARM_HW_BREAKPOINT_H */
diff --git a/arch/arm/include/asm/processor.h b/arch/arm/include/asm/processor.h
index b2d9df5..b86d135 100644
--- a/arch/arm/include/asm/processor.h
+++ b/arch/arm/include/asm/processor.h
@@ -30,7 +30,7 @@
#endif
struct debug_info {
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
struct perf_event *hbp[ARM_MAX_HBP_SLOTS];
#endif
};
diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile
index 8d95446..e6c4b04 100644
--- a/arch/arm/kernel/Makefile
+++ b/arch/arm/kernel/Makefile
@@ -47,7 +47,7 @@ obj-$(CONFIG_HAVE_TCM) += tcm.o
obj-$(CONFIG_CRASH_DUMP) += crash_dump.o
obj-$(CONFIG_SWP_EMULATE) += swp_emulate.o
CFLAGS_swp_emulate.o := -Wa,-march=armv7-a
-obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
+obj-$(CONFIG_HW_BREAKPOINT) += hw_breakpoint.o
obj-$(CONFIG_CRUNCH) += crunch.o crunch-bits.o
AFLAGS_crunch-bits.o := -Wa,-mcpu=ep9312
diff --git a/arch/arm/kernel/entry-header.S b/arch/arm/kernel/entry-header.S
index 051166c..fbc7cc9 100644
--- a/arch/arm/kernel/entry-header.S
+++ b/arch/arm/kernel/entry-header.S
@@ -171,7 +171,7 @@
@ we can access the debug registers safely.
@
.macro debug_entry, fsr
-#if defined(CONFIG_HAVE_HW_BREAKPOINT) && defined(CONFIG_PREEMPT)
+#if defined(CONFIG_HW_BREAKPOINT) && defined(CONFIG_PREEMPT)
ldr r4, =0x40f @ mask out fsr.fs
and r5, r4, \fsr
cmp r5, #2 @ debug exception
diff --git a/arch/arm/kernel/ptrace.c b/arch/arm/kernel/ptrace.c
index 2bf27f3..a3f8a0b 100644
--- a/arch/arm/kernel/ptrace.c
+++ b/arch/arm/kernel/ptrace.c
@@ -468,7 +468,7 @@ static int ptrace_setvfpregs(struct task_struct *tsk, void __user *data)
}
#endif
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
/*
* Convert a virtual register number into an index for a thread_info
* breakpoint array. Breakpoints are identified using positive numbers
@@ -765,7 +765,7 @@ long arch_ptrace(struct task_struct *child, long request,
break;
#endif
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
case PTRACE_GETHBPREGS:
ret = ptrace_gethbpregs(child, addr,
(unsigned long __user *)data);
diff --git a/arch/powerpc/include/asm/cputable.h b/arch/powerpc/include/asm/cputable.h
index 1833d1a..36b5568 100644
--- a/arch/powerpc/include/asm/cputable.h
+++ b/arch/powerpc/include/asm/cputable.h
@@ -538,9 +538,9 @@ static inline int cpu_has_feature(unsigned long feature)
& feature);
}
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
#define HBP_NUM 1
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
#endif /* !__ASSEMBLY__ */
diff --git a/arch/powerpc/include/asm/hw_breakpoint.h b/arch/powerpc/include/asm/hw_breakpoint.h
index 1c33ec1..6ecd3b6 100644
--- a/arch/powerpc/include/asm/hw_breakpoint.h
+++ b/arch/powerpc/include/asm/hw_breakpoint.h
@@ -24,7 +24,7 @@
#define _PPC_BOOK3S_64_HW_BREAKPOINT_H
#ifdef __KERNEL__
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
struct arch_hw_breakpoint {
bool extraneous_interrupt;
@@ -65,10 +65,10 @@ static inline void hw_breakpoint_disable(void)
}
extern void thread_change_pc(struct task_struct *tsk, struct pt_regs *regs);
-#else /* CONFIG_HAVE_HW_BREAKPOINT */
+#else /* CONFIG_HW_BREAKPOINT */
static inline void hw_breakpoint_disable(void) { }
static inline void thread_change_pc(struct task_struct *tsk,
struct pt_regs *regs) { }
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
#endif /* __KERNEL__ */
#endif /* _PPC_BOOK3S_64_HW_BREAKPOINT_H */
diff --git a/arch/powerpc/include/asm/processor.h b/arch/powerpc/include/asm/processor.h
index de1967a..3fe688d 100644
--- a/arch/powerpc/include/asm/processor.h
+++ b/arch/powerpc/include/asm/processor.h
@@ -207,14 +207,14 @@ struct thread_struct {
#ifdef CONFIG_PPC64
unsigned long start_tb; /* Start purr when proc switched in */
unsigned long accum_tb; /* Total accumilated purr for process */
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
struct perf_event *ptrace_bps[HBP_NUM];
/*
* Helps identify source of single-step exception and subsequent
* hw-breakpoint enablement
*/
struct perf_event *last_hit_ubp;
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
#endif
unsigned long dabr; /* Data address breakpoint register */
#ifdef CONFIG_ALTIVEC
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index 3bb2a3e..5df8585 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -36,7 +36,7 @@ obj-y := cputable.o ptrace.o syscalls.o \
obj-$(CONFIG_PPC64) += setup_64.o sys_ppc32.o \
signal_64.o ptrace32.o \
paca.o nvram_64.o firmware.o
-obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
+obj-$(CONFIG_HW_BREAKPOINT) += hw_breakpoint.o
obj-$(CONFIG_PPC_BOOK3S_64) += cpu_setup_ppc970.o cpu_setup_pa6t.o
obj64-$(CONFIG_RELOCATABLE) += reloc_64.o
obj-$(CONFIG_PPC_BOOK3E_64) += exceptions-64e.o idle_book3e.o
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index f74f355..3faf61a 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -353,7 +353,7 @@ static void switch_booke_debug_regs(struct thread_struct *new_thread)
prime_debug_regs(new_thread);
}
#else /* !CONFIG_PPC_ADV_DEBUG_REGS */
-#ifndef CONFIG_HAVE_HW_BREAKPOINT
+#ifndef CONFIG_HW_BREAKPOINT
static void set_debug_reg_defaults(struct thread_struct *thread)
{
if (thread->dabr) {
@@ -361,7 +361,7 @@ static void set_debug_reg_defaults(struct thread_struct *thread)
set_dabr(0);
}
}
-#endif /* !CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* !CONFIG_HW_BREAKPOINT */
#endif /* CONFIG_PPC_ADV_DEBUG_REGS */
int set_dabr(unsigned long dabr)
@@ -469,10 +469,10 @@ struct task_struct *__switch_to(struct task_struct *prev,
* For PPC_BOOK3S_64, we use the hw-breakpoint interfaces that would
* schedule DABR
*/
-#ifndef CONFIG_HAVE_HW_BREAKPOINT
+#ifndef CONFIG_HW_BREAKPOINT
if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr))
set_dabr(new->thread.dabr);
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
#endif
@@ -672,11 +672,11 @@ void flush_thread(void)
{
discard_lazy_cpu_state();
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
flush_ptrace_hw_breakpoint(current);
-#else /* CONFIG_HAVE_HW_BREAKPOINT */
+#else /* CONFIG_HW_BREAKPOINT */
set_debug_reg_defaults(¤t->thread);
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
}
void
@@ -694,9 +694,9 @@ void prepare_to_copy(struct task_struct *tsk)
flush_altivec_to_thread(current);
flush_vsx_to_thread(current);
flush_spe_to_thread(current);
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
flush_ptrace_hw_breakpoint(tsk);
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
}
/*
diff --git a/arch/powerpc/kernel/ptrace.c b/arch/powerpc/kernel/ptrace.c
index 55613e3..6a3f54e 100644
--- a/arch/powerpc/kernel/ptrace.c
+++ b/arch/powerpc/kernel/ptrace.c
@@ -877,7 +877,7 @@ void user_disable_single_step(struct task_struct *task)
clear_tsk_thread_flag(task, TIF_SINGLESTEP);
}
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
void ptrace_triggered(struct perf_event *bp, int nmi,
struct perf_sample_data *data, struct pt_regs *regs)
{
@@ -893,17 +893,17 @@ void ptrace_triggered(struct perf_event *bp, int nmi,
attr.disabled = true;
modify_user_hw_breakpoint(bp, &attr);
}
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
unsigned long data)
{
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
int ret;
struct thread_struct *thread = &(task->thread);
struct perf_event *bp;
struct perf_event_attr attr;
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
/* For ppc64 we support one DABR and no IABR's at the moment (ppc64).
* For embedded processors we support one DAC and no IAC's at the
@@ -932,7 +932,7 @@ int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
/* Ensure breakpoint translation bit is set */
if (data && !(data & DABR_TRANSLATION))
return -EIO;
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
bp = thread->ptrace_bps[0];
if ((!data) || !(data & (DABR_DATA_WRITE | DABR_DATA_READ))) {
if (bp) {
@@ -968,7 +968,7 @@ int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
return PTR_ERR(bp);
}
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
/* Move contents to the DABR register */
task->thread.dabr = data;
diff --git a/arch/powerpc/lib/Makefile b/arch/powerpc/lib/Makefile
index 166a6a0..515d044 100644
--- a/arch/powerpc/lib/Makefile
+++ b/arch/powerpc/lib/Makefile
@@ -19,7 +19,7 @@ obj-$(CONFIG_PPC64) += copypage_64.o copyuser_64.o \
checksum_wrappers_64.o hweight_64.o
obj-$(CONFIG_XMON) += sstep.o ldstfp.o
obj-$(CONFIG_KPROBES) += sstep.o ldstfp.o
-obj-$(CONFIG_HAVE_HW_BREAKPOINT) += sstep.o ldstfp.o
+obj-$(CONFIG_HW_BREAKPOINT) += sstep.o ldstfp.o
ifeq ($(CONFIG_PPC64),y)
obj-$(CONFIG_SMP) += locks.o
diff --git a/arch/sh/kernel/Makefile b/arch/sh/kernel/Makefile
index 77f7ae1..9d5075c 100644
--- a/arch/sh/kernel/Makefile
+++ b/arch/sh/kernel/Makefile
@@ -45,7 +45,7 @@ obj-$(CONFIG_HIBERNATION) += swsusp.o
obj-$(CONFIG_DWARF_UNWINDER) += dwarf.o
obj-$(CONFIG_PERF_EVENTS) += perf_event.o perf_callchain.o
-obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
+obj-$(CONFIG_HW_BREAKPOINT) += hw_breakpoint.o
obj-$(CONFIG_GENERIC_CLOCKEVENTS_BROADCAST) += localtimer.o
ccflags-y := -Werror
diff --git a/arch/sh/kernel/cpu/sh4a/Makefile b/arch/sh/kernel/cpu/sh4a/Makefile
index cc122b1..5bec639 100644
--- a/arch/sh/kernel/cpu/sh4a/Makefile
+++ b/arch/sh/kernel/cpu/sh4a/Makefile
@@ -46,4 +46,4 @@ obj-y += $(clock-y)
obj-$(CONFIG_SMP) += $(smp-y)
obj-$(CONFIG_GENERIC_GPIO) += $(pinmux-y)
obj-$(CONFIG_PERF_EVENTS) += perf_event.o
-obj-$(CONFIG_HAVE_HW_BREAKPOINT) += ubc.o
+obj-$(CONFIG_HW_BREAKPOINT) += ubc.o
diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h
index d1e55fe..9aad682 100644
--- a/include/linux/hw_breakpoint.h
+++ b/include/linux/hw_breakpoint.h
@@ -31,7 +31,7 @@ enum bp_type_idx {
#include <linux/perf_event.h>
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
extern int __init init_hw_breakpoint(void);
@@ -108,7 +108,7 @@ static inline struct arch_hw_breakpoint *counter_arch_bp(struct perf_event *bp)
return &bp->hw.info;
}
-#else /* !CONFIG_HAVE_HW_BREAKPOINT */
+#else /* !CONFIG_HW_BREAKPOINT */
static inline int __init init_hw_breakpoint(void) { return 0; }
@@ -144,7 +144,7 @@ static inline struct arch_hw_breakpoint *counter_arch_bp(struct perf_event *bp)
return NULL;
}
-#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+#endif /* CONFIG_HW_BREAKPOINT */
#endif /* __KERNEL__ */
#endif /* _LINUX_HW_BREAKPOINT_H */
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index ee9f1e7..0f0f9bb 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -489,7 +489,7 @@ struct perf_guest_info_callbacks {
unsigned long (*get_guest_ip) (void);
};
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
#include <asm/hw_breakpoint.h>
#endif
@@ -554,7 +554,7 @@ struct hw_perf_event {
struct { /* software */
struct hrtimer hrtimer;
};
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
struct { /* breakpoint */
struct arch_hw_breakpoint info;
struct list_head bp_list;
diff --git a/kernel/Makefile b/kernel/Makefile
index 85cbfb3..c2739c5 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -104,7 +104,7 @@ obj-$(CONFIG_TRACEPOINTS) += trace/
obj-$(CONFIG_SMP) += sched_cpupri.o
obj-$(CONFIG_IRQ_WORK) += irq_work.o
obj-$(CONFIG_PERF_EVENTS) += perf_event.o
-obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
+obj-$(CONFIG_HW_BREAKPOINT) += hw_breakpoint.o
obj-$(CONFIG_USER_RETURN_NOTIFIER) += user-return-notifier.o
obj-$(CONFIG_PADATA) += padata.o
obj-$(CONFIG_CRASH_DUMP) += crash_dump.o
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 8e81a98..cc89f16 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -5616,7 +5616,7 @@ static void perf_event_free_filter(struct perf_event *event)
#endif /* CONFIG_EVENT_TRACING */
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
void perf_bp_event(struct perf_event *bp, void *data)
{
struct perf_sample_data sample;
@@ -6207,7 +6207,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
if (task) {
event->attach_state = PERF_ATTACH_TASK;
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#ifdef CONFIG_HW_BREAKPOINT
/*
* hw_breakpoint is a bit difficult here..
*/
diff --git a/samples/Kconfig b/samples/Kconfig
index 41063e7..d1e41e9 100644
--- a/samples/Kconfig
+++ b/samples/Kconfig
@@ -40,7 +40,7 @@ config SAMPLE_KRETPROBES
config SAMPLE_HW_BREAKPOINT
tristate "Build kernel hardware breakpoint examples -- loadable module only"
- depends on HAVE_HW_BREAKPOINT && m
+ depends on HW_BREAKPOINT && m
help
This builds kernel hardware breakpoint example modules.
--
1.7.3.2
So that hw_breakpoints and perf can be not built on
specific embedded systems.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Jason Wessel <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Thomas Gleixner <[email protected]>
---
arch/x86/Kconfig | 3 +--
arch/x86/include/asm/debugreg.h | 33 +++++++++++++++++++++++++++++++--
arch/x86/kernel/Makefile | 3 ++-
arch/x86/kernel/process.c | 1 +
arch/x86/kernel/ptrace.c | 17 +++++++++++++++++
5 files changed, 52 insertions(+), 5 deletions(-)
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 8b49bff..fb28dd9 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -41,7 +41,7 @@ config X86
select HAVE_FTRACE_NMI_ENTER if DYNAMIC_FTRACE
select HAVE_SYSCALL_TRACEPOINTS
select HAVE_KVM
- select HAVE_ARCH_KGDB
+ select HAVE_ARCH_KGDB if HW_BREAKPOINT
select HAVE_ARCH_TRACEHOOK
select HAVE_GENERIC_DMA_COHERENT if X86_32
select HAVE_EFFICIENT_UNALIGNED_ACCESS
@@ -54,7 +54,6 @@ config X86
select HAVE_KERNEL_XZ
select HAVE_KERNEL_LZO
select HAVE_HW_BREAKPOINT
- select HW_BREAKPOINT
select HAVE_MIXED_BREAKPOINTS_REGS
select PERF_EVENTS
select HAVE_PERF_EVENTS_NMI
diff --git a/arch/x86/include/asm/debugreg.h b/arch/x86/include/asm/debugreg.h
index 078ad0c..c40d6d4 100644
--- a/arch/x86/include/asm/debugreg.h
+++ b/arch/x86/include/asm/debugreg.h
@@ -78,8 +78,6 @@
*/
#ifdef __KERNEL__
-DECLARE_PER_CPU(unsigned long, cpu_dr7);
-
static inline void hw_breakpoint_disable(void)
{
/* Zero the control register for HW Breakpoint */
@@ -92,6 +90,10 @@ static inline void hw_breakpoint_disable(void)
set_debugreg(0UL, 3);
}
+#ifdef CONFIG_HW_BREAKPOINT
+
+DECLARE_PER_CPU(unsigned long, cpu_dr7);
+
static inline int hw_breakpoint_active(void)
{
return __this_cpu_read(cpu_dr7) & DR_GLOBAL_ENABLE_MASK;
@@ -100,6 +102,33 @@ static inline int hw_breakpoint_active(void)
extern void aout_dump_debugregs(struct user *dump);
extern void hw_breakpoint_restore(void);
+#else
+static inline int hw_breakpoint_active(void)
+{
+ return 0;
+}
+
+static inline void hw_breakpoint_restore(void)
+{
+ set_debugreg(0UL, 0);
+ set_debugreg(0UL, 1);
+ set_debugreg(0UL, 2);
+ set_debugreg(0UL, 3);
+ set_debugreg(current->thread.debugreg6, 6);
+ set_debugreg(0UL, 7);
+}
+
+static inline void aout_dump_debugregs(struct user *dump)
+{
+ int i;
+
+ for (i = 0; i < 6; i++)
+ dump->u_debugreg[i] = 0;
+
+ dump->u_debugreg[6] = current->thread.debugreg6;
+ dump->u_debugreg[7] = 0;
+}
+#endif /* CONFIG_HW_BREAKPOINT */
#endif /* __KERNEL__ */
diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 7338ef2..270e439 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -42,7 +42,7 @@ obj-$(CONFIG_X86_64) += sys_x86_64.o x8664_ksyms_64.o
obj-$(CONFIG_X86_64) += syscall_64.o vsyscall_64.o
obj-y += bootflag.o e820.o
obj-y += pci-dma.o quirks.o topology.o kdebugfs.o
-obj-y += alternative.o i8253.o pci-nommu.o hw_breakpoint.o
+obj-y += alternative.o i8253.o pci-nommu.o
obj-y += tsc.o io_delay.o rtc.o
obj-y += pci-iommu_table.o
obj-y += resource.o
@@ -51,6 +51,7 @@ obj-y += trampoline.o trampoline_$(BITS).o
obj-y += process.o
obj-y += i387.o xsave.o
obj-y += ptrace.o
+obj-$(CONFIG_HW_BREAKPOINT) += hw_breakpoint.o
obj-$(CONFIG_X86_32) += tls.o
obj-$(CONFIG_IA32_EMULATION) += tls.o
obj-y += step.o
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index d46cbe4..d1adbd1 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -22,6 +22,7 @@
#include <asm/uaccess.h>
#include <asm/i387.h>
#include <asm/debugreg.h>
+#include <asm/kdebug.h>
struct kmem_cache *task_xstate_cachep;
EXPORT_SYMBOL_GPL(task_xstate_cachep);
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 45892dc..a1e1d6d 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -528,6 +528,7 @@ static int genregs_set(struct task_struct *target,
return ret;
}
+#ifdef CONFIG_HW_BREAKPOINT
static void ptrace_triggered(struct perf_event *bp, int nmi,
struct perf_sample_data *data,
struct pt_regs *regs)
@@ -761,6 +762,22 @@ ret_path:
return rc;
}
+#else /* !CONFIG_HW_BREAKPOINT */
+
+static inline unsigned long
+ptrace_get_debugreg(struct task_struct *tsk, int n)
+{
+ return -ENOSYS;
+}
+
+static inline
+int ptrace_set_debugreg(struct task_struct *tsk, int n, unsigned long val)
+{
+ return -ENOSYS;
+}
+
+#endif /* CONFIG_HW_BREAKPOINT */
+
/*
* These access the current or another (stopped) task's io permission
* bitmap for debugging or core dump.
--
1.7.3.2
The breakpoint support ability in an arch is not related
to the fact perf events is built or not. HAVE_HW_BREAKPOINT
only shows an ability so this dependency makes no sense
anymore. Archs that select HAVE_HW_BREAKPOINT already
ensure that perf event is built.
Remove that dependency.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Prasad <[email protected]>
Cc: Paul Mundt <[email protected]>
---
arch/Kconfig | 1 -
1 files changed, 0 insertions(+), 1 deletions(-)
diff --git a/arch/Kconfig b/arch/Kconfig
index f78c2be..ce4be89 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -149,7 +149,6 @@ config HAVE_DEFAULT_NO_SPIN_MUTEXES
config HAVE_HW_BREAKPOINT
bool
- depends on PERF_EVENTS
config HAVE_MIXED_BREAKPOINTS_REGS
bool
--
1.7.3.2
The hardware breakpoint config is only made of an ability. An
arch that support this feature selects HAVE_HW_BREAKPOINT. If so,
the feature is definetly built-in, the user can't decide to turn
it off. As hw_breakpoints depend on perf, it also makes perf
a mandatory feature. The whole is quite a piece of code and
may not be desired on some embedded systems.
In order to prepare to make this optable by the user, split the
config into the more traditional couple (ability, user choice) by
providing a new HW_BREAKPOINT config. It is default on and depends
on CONFIG_EXPERT because breakpoint ptrace requests are part of the
usual user ABI. The user must know what he's doing before turning
that off.
For now, only the archs that already implemented a conditional
HAVE_HW_BREAKPOINT can turn off HW_BREAKPOINT. x86 and sh have it
always selected because they need more background work to support
this new modularity.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Prasad <[email protected]>
Cc: Paul Mundt <[email protected]>
---
arch/sh/Kconfig | 1 +
arch/x86/Kconfig | 1 +
init/Kconfig | 10 ++++++++++
3 files changed, 12 insertions(+), 0 deletions(-)
diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index 4b89da2..0d4d124 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -44,6 +44,7 @@ config SUPERH32
select HAVE_FUNCTION_GRAPH_TRACER
select HAVE_ARCH_KGDB
select HAVE_HW_BREAKPOINT
+ select HW_BREAKPOINT
select HAVE_MIXED_BREAKPOINTS_REGS
select PERF_EVENTS
select ARCH_HIBERNATION_POSSIBLE if MMU
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index cc6c53a..8b49bff 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -54,6 +54,7 @@ config X86
select HAVE_KERNEL_XZ
select HAVE_KERNEL_LZO
select HAVE_HW_BREAKPOINT
+ select HW_BREAKPOINT
select HAVE_MIXED_BREAKPOINTS_REGS
select PERF_EVENTS
select HAVE_PERF_EVENTS_NMI
diff --git a/init/Kconfig b/init/Kconfig
index 56240e7..51f587f 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -924,6 +924,16 @@ menuconfig EXPERT
environments which can tolerate a "non-standard" kernel.
Only use this if you really know what you are doing.
+config HW_BREAKPOINT
+ bool "Hardware breakpoints" if EXPERT
+ depends on HAVE_HW_BREAKPOINT
+ default y
+ help
+ Hardware breakpoints are a feature implemented by most CPUs
+ to trigger an event when an instruction or data fetch
+ matches a given pattern. This is typically used by ptrace
+ and perf events.
+
config EMBEDDED
bool "Embedded system"
select EXPERT
--
1.7.3.2
On Wed, Apr 27, 2011 at 06:59:56PM +0200, Frederic Weisbecker wrote:
> Hi,
>
> Until now, x86 and superh32 had to always build hw_breakpoint and perf.
> And powerpc and arm were always building hw_breakpoints when perf
> was enabled.
>
> After this patchset, one can simply disable hw_breakpoints if he
> selected CONFIG_EXPERT. Afterward the user is free to select perf
> or not. Except on SH unfortunately because that would require a
> bit of change in ptrace to support singlestep failures.
>
> Frederic Weisbecker (6):
> hw_breakpoints: Split hardware breakpoints config
> hw_breakpoints: Migrate breakpoint conditional build under new config
> x86: Allow the user not to build hw_breakpoints
> hw_breakpoints: Breakpoints arch ability don't need perf events
> hw_breakpoints: Only force perf events if breakpoints are selected
> hw_breakpoints: Drop remaining misplaced dependency on perf
Ah, and every !x86 archs touched by this patchset have been compile tested only.
On 04/27/2011 09:59 AM, Frederic Weisbecker wrote:
> So that hw_breakpoints and perf can be not built on
> specific embedded systems.
How big of a win, if any, is this, really?
-hpa
--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.
On Wed, Apr 27, 2011 at 10:38:50AM -0700, H. Peter Anvin wrote:
> On 04/27/2011 09:59 AM, Frederic Weisbecker wrote:
> > So that hw_breakpoints and perf can be not built on
> > specific embedded systems.
>
> How big of a win, if any, is this, really?
>
> -hpa
That's pretty significant, because if you can disable breakpoints,
then you can disable perf. Together they make quite a chunk.
breakpoints + perf:
text data bss dec hex filename
1567422 348628 511680 2427730 250b52 vmlinux
none:
text data bss dec hex filename
1478031 333220 508256 2319507 236493 vmlinux
On 4/27/2011 11:26 AM, Frederic Weisbecker wrote:
> On Wed, Apr 27, 2011 at 10:38:50AM -0700, H. Peter Anvin wrote:
>>
>> How big of a win, if any, is this, really?
>
> That's pretty significant, because if you can disable breakpoints,
> then you can disable perf. Together they make quite a chunk.
Another advantage is the ability to disable these features for kernel
bringup on new targets and kernel versions. Recently in v2.6.38
hw_breakpoints were broken in ARM. A change like this would have saved
me a hack to the Kconfig at the least.
Mike
--
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum
On Wed, Apr 27, 2011 at 11:54:17AM -0700, H. Peter Anvin wrote:
> Why do you have to be able to disable breakpoints to disable perf? That seems seriously backwards, especially since we had breakpoints long before perf...
That started when we implemented breakpoints as counters. Then we realized that
ptrace had its own scheduling that was somehow duplicating what perf was doing.
So we have finally unified that under perf. The good point is that archs don't need
to care much about ptrace breakpoints tracking, just the interface.
But yeah the bad point is that dependency.
On Wed, 2011-04-27 at 17:59 +0100, Frederic Weisbecker wrote:
> Hi,
Hi Frederic,
> Until now, x86 and superh32 had to always build hw_breakpoint and perf.
> And powerpc and arm were always building hw_breakpoints when perf
> was enabled.
>
> After this patchset, one can simply disable hw_breakpoints if he
> selected CONFIG_EXPERT. Afterward the user is free to select perf
> or not. Except on SH unfortunately because that would require a
> bit of change in ptrace to support singlestep failures.
For the ARM bits:
Acked-by: Will Deacon <[email protected]>
Cheers,
Will
On 04/27/2011 12:50 PM, Frederic Weisbecker wrote:
> On Wed, Apr 27, 2011 at 11:54:17AM -0700, H. Peter Anvin wrote:
>> Why do you have to be able to disable breakpoints to disable perf? That seems seriously backwards, especially since we had breakpoints long before perf...
>
> That started when we implemented breakpoints as counters. Then we realized that
> ptrace had its own scheduling that was somehow duplicating what perf was doing.
> So we have finally unified that under perf. The good point is that archs don't need
> to care much about ptrace breakpoints tracking, just the interface.
>
> But yeah the bad point is that dependency.
It really is very bad... without breakpoints, you lose almost all
debugging support.
-hpa
--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.
On Tue, May 03, 2011 at 08:35:53AM -0700, H. Peter Anvin wrote:
> On 04/27/2011 12:50 PM, Frederic Weisbecker wrote:
> > On Wed, Apr 27, 2011 at 11:54:17AM -0700, H. Peter Anvin wrote:
> >> Why do you have to be able to disable breakpoints to disable perf? That seems seriously backwards, especially since we had breakpoints long before perf...
> >
> > That started when we implemented breakpoints as counters. Then we realized that
> > ptrace had its own scheduling that was somehow duplicating what perf was doing.
> > So we have finally unified that under perf. The good point is that archs don't need
> > to care much about ptrace breakpoints tracking, just the interface.
> >
> > But yeah the bad point is that dependency.
>
> It really is very bad... without breakpoints, you lose almost all
> debugging support.
Right, so it should be fine for embedded environment to disable breakpoints.
It depends on CONFIG_EXPERT now.
On 05/03/2011 04:12 PM, Frederic Weisbecker wrote:
>>
>> It really is very bad... without breakpoints, you lose almost all
>> debugging support.
>
> Right, so it should be fine for embedded environment to disable breakpoints.
> It depends on CONFIG_EXPERT now.
Uh... even embedded environments need to be able to debug.
-hpa
On Tue, May 03, 2011 at 04:40:14PM -0700, H. Peter Anvin wrote:
> On 05/03/2011 04:12 PM, Frederic Weisbecker wrote:
> >>
> >> It really is very bad... without breakpoints, you lose almost all
> >> debugging support.
> >
> > Right, so it should be fine for embedded environment to disable breakpoints.
> > It depends on CONFIG_EXPERT now.
>
> Uh... even embedded environments need to be able to debug.
For development yeah, but is it needed for production evironments?
May be sometimes. But even though, the main functionalities of
ptrace are still available since instruction breakpoints are
implemented through int3 and not debugreg in ptrace. Only data
breakpoints would be unusable, but I believe they are a minor use compared
to instruction breakpoints. I rarely make use of them for debugging
personally.
That's really deemed for specific cases where people know what
they are doing.
On 05/03/2011 04:54 PM, Frederic Weisbecker wrote:
> On Tue, May 03, 2011 at 04:40:14PM -0700, H. Peter Anvin wrote:
>> On 05/03/2011 04:12 PM, Frederic Weisbecker wrote:
>>>>
>>>> It really is very bad... without breakpoints, you lose almost all
>>>> debugging support.
>>>
>>> Right, so it should be fine for embedded environment to disable breakpoints.
>>> It depends on CONFIG_EXPERT now.
>>
>> Uh... even embedded environments need to be able to debug.
>
> For development yeah, but is it needed for production evironments?
>
Most of the time people don't want to debug something other than they're
going to ship.
-hpa
On Tue, May 03, 2011 at 04:56:03PM -0700, H. Peter Anvin wrote:
> On 05/03/2011 04:54 PM, Frederic Weisbecker wrote:
> > On Tue, May 03, 2011 at 04:40:14PM -0700, H. Peter Anvin wrote:
> >> On 05/03/2011 04:12 PM, Frederic Weisbecker wrote:
> >>>>
> >>>> It really is very bad... without breakpoints, you lose almost all
> >>>> debugging support.
> >>>
> >>> Right, so it should be fine for embedded environment to disable breakpoints.
> >>> It depends on CONFIG_EXPERT now.
> >>
> >> Uh... even embedded environments need to be able to debug.
> >
> > For development yeah, but is it needed for production evironments?
> >
>
> Most of the time people don't want to debug something other than they're
> going to ship.
Well, it's like systems that don't care about having BUG(), sysctl, futexes, etc...
support in the kernel, or whatever standard feature that makes sense most of the time
but sometimes you simply don't need it.
Also consider that as a first step to be able to build a !CONFIG_PERF_EVENTS
kernel. If people show strong interest in beeing able to build a kernel
with breakpoints but without perf events, then I'll invest more efforts in
breaking the dependency.