[RFC Patch 1/5] Allow arch-specific cleanup before breakpoint unregistration

[RFC Patch 1/5] Allow arch-specific cleanup before breakpoint unregistration

[K.Prasad]

Certain architectures (such as PowerPC Book III S) have a need to cleanup
data-structures before the breakpoint is unregistered. This patch introduces
an arch-specific hook in release_bp_slot() along with a weak definition in
the form of a stub funciton.

Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
---
 kernel/hw_breakpoint.c |   12 ++++++++++++
 1 file changed, 12 insertions(+)

Index: linux-2.6.ppc64_test/kernel/hw_breakpoint.c
===================================================================
--- linux-2.6.ppc64_test.orig/kernel/hw_breakpoint.c
+++ linux-2.6.ppc64_test/kernel/hw_breakpoint.c
@@ -242,6 +242,17 @@ toggle_bp_slot(struct perf_event *bp, bo
 }
 
 /*
+ * Function to perform processor-specific cleanup during unregistration
+ */
+__weak void arch_unregister_hw_breakpoint(struct perf_event *bp)
+{
+	/*
+	 * A weak stub function here for those archs that don't define
+	 * it inside arch/.../kernel/hw_breakpoint.c
+	 */
+}
+
+/*
  * Contraints to check before allowing this new breakpoint counter:
  *
  *  == Non-pinned counter == (Considered as pinned for now)
@@ -339,6 +350,7 @@ void release_bp_slot(struct perf_event *
 {
 	mutex_lock(&nr_bp_mutex);
 
+	arch_unregister_hw_breakpoint(bp);
 	__release_bp_slot(bp);
 
 	mutex_unlock(&nr_bp_mutex);

[RFC Patch 2/5] PPC64-HWBKPT: Implement hw-breakpoints for PowerPC Book III S

[K.Prasad]

Implement perf-events based hw-breakpoint interfaces for PowerPC Book III S
processors. These interfaces help arbitrate requests from various users and
schedules them as appropriate.

Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
---
 arch/powerpc/Kconfig                     |    1 
 arch/powerpc/include/asm/cputable.h      |    4 
 arch/powerpc/include/asm/hw_breakpoint.h |   49 ++++
 arch/powerpc/include/asm/processor.h     |    8 
 arch/powerpc/kernel/Makefile             |    1 
 arch/powerpc/kernel/hw_breakpoint.c      |  365 +++++++++++++++++++++++++++++++
 arch/powerpc/kernel/machine_kexec_64.c   |    3 
 arch/powerpc/kernel/process.c            |    6 
 arch/powerpc/kernel/ptrace.c             |   64 +++++
 arch/powerpc/lib/Makefile                |    1 
 include/linux/hw_breakpoint.h            |    1 
 11 files changed, 503 insertions(+)

Index: linux-2.6.ppc64_test/arch/powerpc/include/asm/hw_breakpoint.h
===================================================================
--- /dev/null
+++ linux-2.6.ppc64_test/arch/powerpc/include/asm/hw_breakpoint.h
@@ -0,0 +1,49 @@
+#ifndef _PPC_BOOK3S_64_HW_BREAKPOINT_H
+#define _PPC_BOOK3S_64_HW_BREAKPOINT_H
+
+#ifdef	__KERNEL__
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+
+struct arch_hw_breakpoint {
+	u8		len; /* length of the target data symbol */
+	int		type;
+	unsigned long	address;
+};
+
+#include <linux/kdebug.h>
+#include <asm/reg.h>
+#include <asm/system.h>
+
+static inline int hw_breakpoint_slots(int type)
+{
+	return HBP_NUM;
+}
+struct perf_event;
+struct pmu;
+struct perf_sample_data;
+
+#define HW_BREAKPOINT_ALIGN 0x7
+/* Maximum permissible length of any HW Breakpoint */
+#define HW_BREAKPOINT_LEN 0x8
+
+extern int arch_bp_generic_fields(int type, int *gen_bp_type);
+extern int arch_check_bp_in_kernelspace(struct perf_event *bp);
+extern int arch_validate_hwbkpt_settings(struct perf_event *bp);
+extern int hw_breakpoint_exceptions_notify(struct notifier_block *unused,
+						unsigned long val, void *data);
+int arch_install_hw_breakpoint(struct perf_event *bp);
+void arch_uninstall_hw_breakpoint(struct perf_event *bp);
+void hw_breakpoint_pmu_read(struct perf_event *bp);
+extern void flush_ptrace_hw_breakpoint(struct task_struct *tsk);
+
+extern struct pmu perf_ops_bp;
+extern void ptrace_triggered(struct perf_event *bp, int nmi,
+			struct perf_sample_data *data, struct pt_regs *regs);
+static inline void hw_breakpoint_disable(void)
+{
+	set_dabr(0);
+}
+
+#endif	/* CONFIG_HAVE_HW_BREAKPOINT */
+#endif	/* __KERNEL__ */
+#endif	/* _PPC_BOOK3S_64_HW_BREAKPOINT_H */
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/hw_breakpoint.c
===================================================================
--- /dev/null
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/hw_breakpoint.c
@@ -0,0 +1,365 @@
+/*
+ * 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/module.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/smp.h>
+
+#include <asm/hw_breakpoint.h>
+#include <asm/processor.h>
+#include <asm/sstep.h>
+
+/*
+ * Store the 'bp' that caused the hw-breakpoint exception just before we
+ * single-step. Use it to distinguish a single-step exception (due to a
+ * previous hw-breakpoint exception) from a normal one
+ */
+static DEFINE_PER_CPU(struct perf_event *, last_hit_bp);
+
+/*
+ * Stores the breakpoints currently in use on each breakpoint address
+ * register for every cpu
+ */
+static DEFINE_PER_CPU(struct perf_event *, bp_per_reg);
+
+/*
+ * 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 = &__get_cpu_var(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_dabr(info->address | info->type | DABR_TRANSLATION);
+
+	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 = &__get_cpu_var(bp_per_reg);
+
+	if (*slot != bp) {
+		WARN_ONCE(1, "Can't find the breakpoint");
+		return;
+	}
+
+	*slot = NULL;
+	set_dabr(0);
+}
+
+/*
+ * Perform cleanup of arch-specific counters during unregistration
+ * of the perf-event
+ */
+void arch_unregister_hw_breakpoint(struct perf_event *bp)
+{
+	int cpu = get_cpu();
+
+	/*
+	 * 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.
+	 */
+	if (per_cpu(last_hit_bp, cpu) == bp) {
+			per_cpu(last_hit_bp, cpu) = NULL;
+		return;
+	}
+	if (bp->ctx->task)
+		bp->ctx->task->thread.last_hit_ubp = NULL;
+
+	put_cpu();
+}
+
+/*
+ * 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)
+{
+	switch (type) {
+	case DABR_DATA_READ:
+		*gen_bp_type = HW_BREAKPOINT_R;
+		break;
+	case DABR_DATA_WRITE:
+		*gen_bp_type = HW_BREAKPOINT_W;
+		break;
+	case (DABR_DATA_WRITE | DABR_DATA_READ):
+		*gen_bp_type = (HW_BREAKPOINT_W | HW_BREAKPOINT_R);
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+/*
+ * Validate the arch-specific HW Breakpoint register settings
+ */
+int arch_validate_hwbkpt_settings(struct perf_event *bp)
+{
+	int ret = -EINVAL;
+	struct arch_hw_breakpoint *info = counter_arch_bp(bp);
+
+	if (!bp)
+		return ret;
+
+	switch (bp->attr.bp_type) {
+	case HW_BREAKPOINT_R:
+		info->type = DABR_DATA_READ;
+		break;
+	case HW_BREAKPOINT_W:
+		info->type = DABR_DATA_WRITE;
+		break;
+	case HW_BREAKPOINT_R | HW_BREAKPOINT_W:
+		info->type = (DABR_DATA_READ | DABR_DATA_WRITE);
+		break;
+	default:
+		return ret;
+	}
+
+	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.
+	 */
+	if (info->len >
+	    (HW_BREAKPOINT_LEN - (info->address & HW_BREAKPOINT_ALIGN)))
+		return -EINVAL;
+	return 0;
+}
+
+/*
+ * Handle debug exception notifications.
+ */
+int __kprobes hw_breakpoint_handler(struct die_args *args)
+{
+	bool is_kernel, is_ptrace_bp = false;
+	int rc = NOTIFY_STOP;
+	struct perf_event *bp;
+	struct pt_regs *regs = args->regs;
+	unsigned long dar = regs->dar;
+	int stepped = 1;
+	struct arch_hw_breakpoint *info;
+
+	/* Disable breakpoints during exception handling */
+	set_dabr(0);
+	/*
+	 * 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 = __get_cpu_var(bp_per_reg);
+	if (!bp)
+		goto out;
+	info = counter_arch_bp(bp);
+	is_kernel = is_kernel_addr(bp->attr.bp_addr);
+	is_ptrace_bp = (bp->overflow_handler == ptrace_triggered) ?
+			true : false;
+
+	/*
+	 * Verify if dar lies within the address range occupied by the symbol
+	 * being watched to filter extraneous exceptions.
+	 */
+	if (!((bp->attr.bp_addr <= dar) &&
+	    (dar <= (bp->attr.bp_addr + bp->attr.bp_len))) &&
+	    (!is_ptrace_bp))
+		/*
+		 * This exception is triggered not because of a memory access on
+		 * the monitored variable but in the double-word address range
+		 * in which it is contained. We will consume this exception,
+		 * considering it as 'noise'.
+		 */
+		goto restore_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 (is_ptrace_bp) {
+		perf_bp_event(bp, regs);
+		rc = NOTIFY_DONE;
+		goto out;
+	}
+
+	/*
+	 * Do not emulate user-space instructions from kernel-space,
+	 * instead single-step them.
+	 */
+	if (!is_kernel) {
+		bp->ctx->task->thread.last_hit_ubp = bp;
+		regs->msr |= MSR_SE;
+		goto out;
+	}
+
+	stepped = emulate_step(regs, regs->nip);
+	/* emulate_step() could not execute it, single-step them */
+	if (stepped == 0) {
+		regs->msr |= MSR_SE;
+		__get_cpu_var(last_hit_bp) = bp;
+		goto out;
+	}
+	/*
+	 * As a policy, the callback is invoked in a 'trigger-after-execute'
+	 * fashion
+	 */
+	perf_bp_event(bp, regs);
+
+restore_bp:
+	set_dabr(info->address | info->type | DABR_TRANSLATION);
+out:
+	rcu_read_unlock();
+	return rc;
+}
+
+/*
+ * Handle single-step exceptions following a DABR hit.
+ */
+int __kprobes single_step_dabr_instruction(struct die_args *args)
+{
+	struct pt_regs *regs = args->regs;
+	struct perf_event *bp = NULL, *kernel_bp, *user_bp;
+	struct arch_hw_breakpoint *bp_info;
+
+	/*
+	 * Identify the cause of single-stepping and find the corresponding
+	 * breakpoint structure
+	 */
+	user_bp = current->thread.last_hit_ubp;
+	kernel_bp = __get_cpu_var(last_hit_bp);
+	if (user_bp) {
+		bp = user_bp;
+		current->thread.last_hit_ubp = NULL;
+	} else if (kernel_bp) {
+		bp = kernel_bp;
+		__get_cpu_var(last_hit_bp) = NULL;
+	}
+
+	/*
+	 * Check if we are single-stepping as a result of a
+	 * previous HW Breakpoint exception
+	 */
+	if (!bp)
+		return NOTIFY_DONE;
+
+	bp_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
+	 */
+	perf_bp_event(bp, regs);
+
+	/*
+	 * Do not disable MSR_SE if the process was already in
+	 * single-stepping mode. We cannot reliable detect single-step mode
+	 * for kernel-space breakpoints, so this cannot work along with other
+	 * debuggers (like KGDB, xmon) which may be single-stepping kernel code.
+	 */
+	if (!(user_bp && test_thread_flag(TIF_SINGLESTEP)))
+		regs->msr &= ~MSR_SE;
+
+	set_dabr(bp_info->address | bp_info->type | DABR_TRANSLATION);
+	return NOTIFY_STOP;
+}
+
+/*
+ * Handle debug exception notifications.
+ */
+int __kprobes 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;
+}
+
+/*
+ * 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 */
+}
+
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/Makefile
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/Makefile
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/Makefile
@@ -34,6 +34,7 @@ obj-y				+= vdso32/
 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-$(CONFIG_PPC_BOOK3S_64)	+= cpu_setup_ppc970.o cpu_setup_pa6t.o
 obj64-$(CONFIG_RELOCATABLE)	+= reloc_64.o
 obj-$(CONFIG_PPC_BOOK3E_64)	+= exceptions-64e.o
Index: linux-2.6.ppc64_test/arch/powerpc/include/asm/processor.h
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/include/asm/processor.h
+++ linux-2.6.ppc64_test/arch/powerpc/include/asm/processor.h
@@ -209,6 +209,14 @@ 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
+	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
 	unsigned long	dabr;		/* Data address breakpoint register */
 #ifdef CONFIG_ALTIVEC
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/ptrace.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/ptrace.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/ptrace.c
@@ -32,6 +32,8 @@
 #ifdef CONFIG_PPC32
 #include <linux/module.h>
 #endif
+#include <linux/hw_breakpoint.h>
+#include <linux/perf_event.h>
 
 #include <asm/uaccess.h>
 #include <asm/page.h>
@@ -866,9 +868,34 @@ void user_disable_single_step(struct tas
 	clear_tsk_thread_flag(task, TIF_SINGLESTEP);
 }
 
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+void ptrace_triggered(struct perf_event *bp, int nmi,
+		      struct perf_sample_data *data, struct pt_regs *regs)
+{
+	struct perf_event_attr attr;
+
+	/*
+	 * Disable the breakpoint request here since ptrace has defined a
+	 * one-shot behaviour for breakpoint exceptions in PPC64.
+	 * The SIGTRAP signal is generated automatically for us in do_dabr().
+	 * We don't have to do anything about that here
+	 */
+	attr = bp->attr;
+	attr.disabled = true;
+	modify_user_hw_breakpoint(bp, &attr);
+}
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+
 int ptrace_set_debugreg(struct task_struct *task, unsigned long addr,
 			       unsigned long data)
 {
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	int ret;
+	struct thread_struct *thread = &(task->thread);
+	struct perf_event *bp;
+	struct perf_event_attr attr;
+#endif /* CONFIG_HAVE_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
 	 *  moment.
@@ -896,6 +923,43 @@ int ptrace_set_debugreg(struct task_stru
 	/* Ensure breakpoint translation bit is set */
 	if (data && !(data & DABR_TRANSLATION))
 		return -EIO;
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	bp = thread->ptrace_bps[0];
+	if ((!data) || !(data & (DABR_DATA_WRITE | DABR_DATA_READ))) {
+		if (bp) {
+			unregister_hw_breakpoint(bp);
+			thread->ptrace_bps[0] = NULL;
+		}
+		return 0;
+	}
+	if (bp) {
+		attr = bp->attr;
+		attr.bp_addr = data & ~HW_BREAKPOINT_ALIGN;
+		arch_bp_generic_fields(data &
+					(DABR_DATA_WRITE | DABR_DATA_READ),
+							&attr.bp_type);
+		ret =  modify_user_hw_breakpoint(bp, &attr);
+		if (ret)
+			return ret;
+		thread->ptrace_bps[0] = bp;
+		thread->dabr = data;
+		return 0;
+	}
+
+	/* Create a new breakpoint request if one doesn't exist already */
+	hw_breakpoint_init(&attr);
+	attr.bp_addr = data & ~HW_BREAKPOINT_ALIGN;
+	arch_bp_generic_fields(data & (DABR_DATA_WRITE | DABR_DATA_READ),
+								&attr.bp_type);
+
+	thread->ptrace_bps[0] = bp = register_user_hw_breakpoint(&attr,
+							ptrace_triggered, task);
+	if (IS_ERR(bp)) {
+		thread->ptrace_bps[0] = NULL;
+		return PTR_ERR(bp);
+	}
+
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 
 	/* Move contents to the DABR register */
 	task->thread.dabr = data;
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/process.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/process.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/process.c
@@ -462,8 +462,14 @@ struct task_struct *__switch_to(struct t
 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
 	switch_booke_debug_regs(&new->thread);
 #else
+/*
+ * For PPC_BOOK3S_64, we use the hw-breakpoint interfaces that would
+ * schedule DABR
+ */
+#ifndef CONFIG_HAVE_HW_BREAKPOINT
 	if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr))
 		set_dabr(new->thread.dabr);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 #endif
 
 
Index: linux-2.6.ppc64_test/arch/powerpc/include/asm/cputable.h
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/include/asm/cputable.h
+++ linux-2.6.ppc64_test/arch/powerpc/include/asm/cputable.h
@@ -516,6 +516,10 @@ static inline int cpu_has_feature(unsign
 		& feature);
 }
 
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+#define HBP_NUM 1
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
+
 #endif /* !__ASSEMBLY__ */
 
 #endif /* __KERNEL__ */
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/machine_kexec_64.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/machine_kexec_64.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/machine_kexec_64.c
@@ -15,6 +15,7 @@
 #include <linux/thread_info.h>
 #include <linux/init_task.h>
 #include <linux/errno.h>
+#include <linux/hw_breakpoint.h>
 
 #include <asm/page.h>
 #include <asm/current.h>
@@ -165,6 +166,7 @@ static void kexec_smp_down(void *arg)
 	while(kexec_all_irq_disabled == 0)
 		cpu_relax();
 	mb(); /* make sure all irqs are disabled before this */
+	hw_breakpoint_disable();
 	/*
 	 * Now every CPU has IRQs off, we can clear out any pending
 	 * IPIs and be sure that no more will come in after this.
@@ -180,6 +182,7 @@ static void kexec_prepare_cpus_wait(int 
 {
 	int my_cpu, i, notified=-1;
 
+	hw_breakpoint_disable();
 	my_cpu = get_cpu();
 	/* Make sure each CPU has atleast made it to the state we need */
 	for (i=0; i < NR_CPUS; i++) {
Index: linux-2.6.ppc64_test/arch/powerpc/lib/Makefile
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/lib/Makefile
+++ linux-2.6.ppc64_test/arch/powerpc/lib/Makefile
@@ -20,6 +20,7 @@ obj-$(CONFIG_PPC64)	+= copypage_64.o cop
 			   memcpy_64.o usercopy_64.o mem_64.o string.o
 obj-$(CONFIG_XMON)	+= sstep.o
 obj-$(CONFIG_KPROBES)	+= sstep.o
+obj-$(CONFIG_HAVE_HW_BREAKPOINT)	+= sstep.o
 
 ifeq ($(CONFIG_PPC64),y)
 obj-$(CONFIG_SMP)	+= locks.o
Index: linux-2.6.ppc64_test/include/linux/hw_breakpoint.h
===================================================================
--- linux-2.6.ppc64_test.orig/include/linux/hw_breakpoint.h
+++ linux-2.6.ppc64_test/include/linux/hw_breakpoint.h
@@ -139,6 +139,7 @@ static inline struct arch_hw_breakpoint 
 {
 	return NULL;
 }
+static inline void hw_breakpoint_disable(void) { }
 
 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
 #endif /* __KERNEL__ */
Index: linux-2.6.ppc64_test/arch/powerpc/Kconfig
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/Kconfig
+++ linux-2.6.ppc64_test/arch/powerpc/Kconfig
@@ -141,6 +141,7 @@ config PPC
 	select GENERIC_ATOMIC64 if PPC32
 	select HAVE_PERF_EVENTS
 	select HAVE_REGS_AND_STACK_ACCESS_API
+	select HAVE_HW_BREAKPOINT if PERF_EVENTS && PPC_BOOK3S_64
 
 config EARLY_PRINTK
 	bool

Re: [RFC Patch 2/5] PPC64-HWBKPT: Implement hw-breakpoints for PowerPC Book III S

[K.Prasad]

On Mon, May 24, 2010 at 09:32:25AM +0530, K.Prasad wrote:
> Implement perf-events based hw-breakpoint interfaces for PowerPC Book III S
> processors. These interfaces help arbitrate requests from various users and
> schedules them as appropriate.
> 
<snipped>
> +/*
> + * Handle debug exception notifications.
> + */
> +int __kprobes hw_breakpoint_handler(struct die_args *args)
> +{
> +	bool is_kernel, is_ptrace_bp = false;
> +	int rc = NOTIFY_STOP;
> +	struct perf_event *bp;
> +	struct pt_regs *regs = args->regs;
> +	unsigned long dar = regs->dar;
> +	int stepped = 1;
> +	struct arch_hw_breakpoint *info;
> +
> +	/* Disable breakpoints during exception handling */
> +	set_dabr(0);
> +	/*
> +	 * 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 = __get_cpu_var(bp_per_reg);
> +	if (!bp)
> +		goto out;
> +	info = counter_arch_bp(bp);
> +	is_kernel = is_kernel_addr(bp->attr.bp_addr);
> +	is_ptrace_bp = (bp->overflow_handler == ptrace_triggered) ?
> +			true : false;
> +
> +	/*
> +	 * Verify if dar lies within the address range occupied by the symbol
> +	 * being watched to filter extraneous exceptions.
> +	 */
> +	if (!((bp->attr.bp_addr <= dar) &&
> +	    (dar <= (bp->attr.bp_addr + bp->attr.bp_len))) &&
> +	    (!is_ptrace_bp))
> +		/*
> +		 * This exception is triggered not because of a memory access on
> +		 * the monitored variable but in the double-word address range
> +		 * in which it is contained. We will consume this exception,
> +		 * considering it as 'noise'.
> +		 */
> +		goto restore_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 (is_ptrace_bp) {
> +		perf_bp_event(bp, regs);
> +		rc = NOTIFY_DONE;
> +		goto out;
> +	}
> +
> +	/*
> +	 * Do not emulate user-space instructions from kernel-space,
> +	 * instead single-step them.
> +	 */
> +	if (!is_kernel) {
> +		bp->ctx->task->thread.last_hit_ubp = bp;
> +		regs->msr |= MSR_SE;
> +		goto out;
> +	}
> +
> +	stepped = emulate_step(regs, regs->nip);
> +	/* emulate_step() could not execute it, single-step them */
> +	if (stepped == 0) {

As I was responding to one of the previous mails, I realised that I
had not made changes here as Paul Mackerras had suggested
(reference linuxppc-dev message-id:
20100520131003.GB29903@brick.ozlabs.ibm.com) i.e. uninstall breakpoint
if single-stepping failed.

I'll quickly send out a revised patch as a reply to this one. Regrets
for the confusion caused.

Thanks,
K.Prasad

> +		regs->msr |= MSR_SE;
> +		__get_cpu_var(last_hit_bp) = bp;
> +		goto out;
> +	}
> +	/*
> +	 * As a policy, the callback is invoked in a 'trigger-after-execute'
> +	 * fashion
> +	 */
> +	perf_bp_event(bp, regs);
> +
> +restore_bp:
> +	set_dabr(info->address | info->type | DABR_TRANSLATION);
> +out:
> +	rcu_read_unlock();
> +	return rc;
> +}

[RFC Patch 3/5] PPC64-HWBKPT: Handle concurrent alignment interrupts

[K.Prasad]

An alignment interrupt may intervene between a DSI/hw-breakpoint exception
and the single-step exception. Enable the alignment interrupt (through
modifications to emulate_single_step()) to notify the single-step exception
handler for proper restoration of hw-breakpoints.

Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
---
 arch/powerpc/kernel/traps.c |    7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

Index: linux-2.6.ppc64_test/arch/powerpc/kernel/traps.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/traps.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/traps.c
@@ -602,7 +602,7 @@ void RunModeException(struct pt_regs *re
 
 void __kprobes single_step_exception(struct pt_regs *regs)
 {
-	regs->msr &= ~(MSR_SE | MSR_BE);  /* Turn off 'trace' bits */
+	clear_single_step(regs);
 
 	if (notify_die(DIE_SSTEP, "single_step", regs, 5,
 					5, SIGTRAP) == NOTIFY_STOP)
@@ -621,10 +621,7 @@ void __kprobes single_step_exception(str
  */
 static void emulate_single_step(struct pt_regs *regs)
 {
-	if (single_stepping(regs)) {
-		clear_single_step(regs);
-		_exception(SIGTRAP, regs, TRAP_TRACE, 0);
-	}
+	single_step_exception(regs);
 }
 
 static inline int __parse_fpscr(unsigned long fpscr)

[RFC Patch 4/5] PPC64-HWBKPT: Enable hw-breakpoints while handling intervening signals

[K.Prasad]

A signal delivered between a hw_breakpoint_handler() and the
single_step_dabr_instruction() will not have the breakpoint active during
signal handling (since breakpoint will not be restored through single-stepping
due to absence of MSR_SE bit on the signal frame). Enable breakpoints before
signal delivery and clear them during sigreturn() syscall.

Limitation: Nested hw-breakpoint exceptions (where second exception is raised
inside signal context) will cause a 'double-hit' i.e. the first breakpoint
exception will be taken twice.

Restore hw-breakpoints if the user-context is altered in the signal handler
(causing loss of MSR_SE).
Side-effect: 'Double-hit' of breakpoint if the instruction pointer is
unaltered in the new context.

Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/hw_breakpoint.h |    3 +++
 arch/powerpc/kernel/hw_breakpoint.c      |   28 ++++++++++++++++++++++++++++
 arch/powerpc/kernel/signal.c             |    8 ++++++++
 arch/powerpc/kernel/signal_32.c          |   10 ++++++++++
 arch/powerpc/kernel/signal_64.c          |    7 +++++++
 5 files changed, 56 insertions(+)

Index: linux-2.6.ppc64_test/arch/powerpc/include/asm/hw_breakpoint.h
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/include/asm/hw_breakpoint.h
+++ linux-2.6.ppc64_test/arch/powerpc/include/asm/hw_breakpoint.h
@@ -43,6 +43,9 @@ static inline void hw_breakpoint_disable
 {
 	set_dabr(0);
 }
+extern void sighandler_install_bp(struct task_struct *tsk);
+extern void sigreturn_uninstall_bp(struct task_struct *tsk);
+extern void thread_change_pc(struct task_struct *tsk, unsigned long msr);
 
 #endif	/* CONFIG_HAVE_HW_BREAKPOINT */
 #endif	/* __KERNEL__ */
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/hw_breakpoint.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/hw_breakpoint.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/hw_breakpoint.c
@@ -188,6 +188,34 @@ int arch_validate_hwbkpt_settings(struct
 	return 0;
 }
 
+void sighandler_install_bp(struct task_struct *tsk)
+{
+	struct arch_hw_breakpoint *info;
+
+	if (likely(!tsk->thread.last_hit_ubp))
+		return;
+
+	info = counter_arch_bp(tsk->thread.last_hit_ubp);
+	set_dabr(info->address | info->type | DABR_TRANSLATION);
+}
+
+void sigreturn_uninstall_bp(struct task_struct *tsk)
+{
+	if (unlikely(tsk->thread.last_hit_ubp))
+		set_dabr(0);
+}
+
+void thread_change_pc(struct task_struct *tsk, unsigned long new_msr)
+{
+	/*
+	 * Do not bother to restore breakpoints if single-stepping is not
+	 * cleared. single_step_dabr_instruction() will handle it if MSR_SE
+	 * is set.
+	 */
+	if (!(new_msr & MSR_SE))
+		sighandler_install_bp(tsk);
+}
+
 /*
  * Handle debug exception notifications.
  */
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/signal.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/signal.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/signal.c
@@ -11,6 +11,7 @@
 
 #include <linux/tracehook.h>
 #include <linux/signal.h>
+#include <asm/hw_breakpoint.h>
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
 
@@ -149,6 +150,13 @@ static int do_signal_pending(sigset_t *o
 	if (current->thread.dabr)
 		set_dabr(current->thread.dabr);
 #endif
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	/*
+	 * Re-enable the breakpoints (if it was previously cleared in
+	 * hw_breakpoint_handler()) for the signal stack.
+	 */
+	sighandler_install_bp(current);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 
 	if (is32) {
         	if (ka.sa.sa_flags & SA_SIGINFO)
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/signal_64.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/signal_64.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/signal_64.c
@@ -33,6 +33,7 @@
 #include <asm/cacheflush.h>
 #include <asm/syscalls.h>
 #include <asm/vdso.h>
+#include <asm/hw_breakpoint.h>
 
 #include "signal.h"
 
@@ -312,6 +313,9 @@ int sys_swapcontext(struct ucontext __us
 		    || __copy_to_user(&old_ctx->uc_sigmask,
 				      &current->blocked, sizeof(sigset_t)))
 			return -EFAULT;
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+		thread_change_pc(current, new_msr);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 	}
 	if (new_ctx == NULL)
 		return 0;
@@ -364,6 +368,9 @@ int sys_rt_sigreturn(unsigned long r3, u
 	if (__copy_from_user(&set, &uc->uc_sigmask, sizeof(set)))
 		goto badframe;
 	restore_sigmask(&set);
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	sigreturn_uninstall_bp(current);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 	if (restore_sigcontext(regs, NULL, 1, &uc->uc_mcontext))
 		goto badframe;
 
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/signal_32.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/signal_32.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/signal_32.c
@@ -42,6 +42,7 @@
 #include <asm/syscalls.h>
 #include <asm/sigcontext.h>
 #include <asm/vdso.h>
+#include <asm/hw_breakpoint.h>
 #ifdef CONFIG_PPC64
 #include "ppc32.h"
 #include <asm/unistd.h>
@@ -996,6 +997,9 @@ long sys_swapcontext(struct ucontext __u
 		    || put_sigset_t(&old_ctx->uc_sigmask, &current->blocked)
 		    || __put_user(to_user_ptr(mctx), &old_ctx->uc_regs))
 			return -EFAULT;
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+		thread_change_pc(current, new_msr);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 	}
 	if (new_ctx == NULL)
 		return 0;
@@ -1034,6 +1038,9 @@ long sys_rt_sigreturn(int r3, int r4, in
 		(regs->gpr[1] + __SIGNAL_FRAMESIZE + 16);
 	if (!access_ok(VERIFY_READ, rt_sf, sizeof(*rt_sf)))
 		goto bad;
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	sigreturn_uninstall_bp(current);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 	if (do_setcontext(&rt_sf->uc, regs, 1))
 		goto bad;
 
@@ -1279,6 +1286,9 @@ long sys_sigreturn(int r3, int r4, int r
 #endif
 	restore_sigmask(&set);
 
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+	sigreturn_uninstall_bp(current);
+#endif /* CONFIG_HAVE_HW_BREAKPOINT */
 	sr = (struct mcontext __user *)from_user_ptr(sigctx.regs);
 	addr = sr;
 	if (!access_ok(VERIFY_READ, sr, sizeof(*sr))

[RFC Patch 5/5] PPC64-HWBKPT: Enable proper distinction of per-task and per-cpu breakpoints

[K.Prasad]

Per-task and per-cpu breakpoints have to be unambiguously identified for
proper restoration of hw-breakpoints. The notion of pure user- and kernel-
space breakpoints is antiquated. Store 'pid' of the process against which
the perf-counter was requested to help proper distinction.

This helps seamless handling of kernel-space breakpoints within the context
of a user-space process and breakpoints for kernel-threads.

Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
---
 arch/powerpc/kernel/hw_breakpoint.c |   24 +++++++++++++++++-------
 include/linux/perf_event.h          |    1 +
 kernel/perf_event.c                 |    9 ++++++---
 3 files changed, 24 insertions(+), 10 deletions(-)

Index: linux-2.6.ppc64_test/include/linux/perf_event.h
===================================================================
--- linux-2.6.ppc64_test.orig/include/linux/perf_event.h
+++ linux-2.6.ppc64_test/include/linux/perf_event.h
@@ -698,6 +698,7 @@ struct perf_event {
 
 	int				oncpu;
 	int				cpu;
+	pid_t				pid;
 
 	struct list_head		owner_entry;
 	struct task_struct		*owner;
Index: linux-2.6.ppc64_test/kernel/perf_event.c
===================================================================
--- linux-2.6.ppc64_test.orig/kernel/perf_event.c
+++ linux-2.6.ppc64_test/kernel/perf_event.c
@@ -4684,6 +4684,7 @@ static const struct pmu *sw_perf_event_i
 static struct perf_event *
 perf_event_alloc(struct perf_event_attr *attr,
 		   int cpu,
+		   pid_t pid,
 		   struct perf_event_context *ctx,
 		   struct perf_event *group_leader,
 		   struct perf_event *parent_event,
@@ -4717,6 +4718,7 @@ perf_event_alloc(struct perf_event_attr 
 	mutex_init(&event->mmap_mutex);
 
 	event->cpu		= cpu;
+	event->pid		= pid;
 	event->attr		= *attr;
 	event->group_leader	= group_leader;
 	event->pmu		= NULL;
@@ -5015,7 +5017,7 @@ SYSCALL_DEFINE5(perf_event_open,
 			goto err_put_context;
 	}
 
-	event = perf_event_alloc(&attr, cpu, ctx, group_leader,
+	event = perf_event_alloc(&attr, cpu, pid, ctx, group_leader,
 				     NULL, NULL, GFP_KERNEL);
 	err = PTR_ERR(event);
 	if (IS_ERR(event))
@@ -5090,7 +5092,7 @@ perf_event_create_kernel_counter(struct 
 		goto err_exit;
 	}
 
-	event = perf_event_alloc(attr, cpu, ctx, NULL,
+	event = perf_event_alloc(attr, cpu, pid, ctx, NULL,
 				 NULL, overflow_handler, GFP_KERNEL);
 	if (IS_ERR(event)) {
 		err = PTR_ERR(event);
@@ -5142,7 +5144,8 @@ inherit_event(struct perf_event *parent_
 		parent_event = parent_event->parent;
 
 	child_event = perf_event_alloc(&parent_event->attr,
-					   parent_event->cpu, child_ctx,
+					   parent_event->cpu, child->pid,
+					   child_ctx,
 					   group_leader, parent_event,
 					   NULL, GFP_KERNEL);
 	if (IS_ERR(child_event))
Index: linux-2.6.ppc64_test/arch/powerpc/kernel/hw_breakpoint.c
===================================================================
--- linux-2.6.ppc64_test.orig/arch/powerpc/kernel/hw_breakpoint.c
+++ linux-2.6.ppc64_test/arch/powerpc/kernel/hw_breakpoint.c
@@ -221,7 +221,7 @@ void thread_change_pc(struct task_struct
  */
 int __kprobes hw_breakpoint_handler(struct die_args *args)
 {
-	bool is_kernel, is_ptrace_bp = false;
+	bool is_kernel, is_taskbound_bp, is_ptrace_bp = false;
 	int rc = NOTIFY_STOP;
 	struct perf_event *bp;
 	struct pt_regs *regs = args->regs;
@@ -246,6 +246,7 @@ int __kprobes hw_breakpoint_handler(stru
 	is_kernel = is_kernel_addr(bp->attr.bp_addr);
 	is_ptrace_bp = (bp->overflow_handler == ptrace_triggered) ?
 			true : false;
+	is_taskbound_bp = (bp->pid > 0) ? true : false;
 
 	/*
 	 * Verify if dar lies within the address range occupied by the symbol
@@ -288,7 +289,14 @@ int __kprobes hw_breakpoint_handler(stru
 	/* emulate_step() could not execute it, single-step them */
 	if (stepped == 0) {
 		regs->msr |= MSR_SE;
-		__get_cpu_var(last_hit_bp) = bp;
+		/*
+		 * Kernel-space addresses can also be bound to a task. If so,
+		 * store the breakpoint in its 'thread_struct'
+		 */
+		if (is_taskbound_bp)
+			bp->ctx->task->thread.last_hit_ubp = bp;
+		else
+			__get_cpu_var(last_hit_bp) = bp;
 		goto out;
 	}
 	/*
@@ -310,17 +318,17 @@ out:
 int __kprobes single_step_dabr_instruction(struct die_args *args)
 {
 	struct pt_regs *regs = args->regs;
-	struct perf_event *bp = NULL, *kernel_bp, *user_bp;
+	struct perf_event *bp = NULL, *kernel_bp, *per_task_bp;
 	struct arch_hw_breakpoint *bp_info;
 
 	/*
 	 * Identify the cause of single-stepping and find the corresponding
 	 * breakpoint structure
 	 */
-	user_bp = current->thread.last_hit_ubp;
+	per_task_bp = current->thread.last_hit_ubp;
 	kernel_bp = __get_cpu_var(last_hit_bp);
-	if (user_bp) {
-		bp = user_bp;
+	if (per_task_bp) {
+		bp = per_task_bp;
 		current->thread.last_hit_ubp = NULL;
 	} else if (kernel_bp) {
 		bp = kernel_bp;
@@ -348,7 +356,9 @@ int __kprobes single_step_dabr_instructi
 	 * for kernel-space breakpoints, so this cannot work along with other
 	 * debuggers (like KGDB, xmon) which may be single-stepping kernel code.
 	 */
-	if (!(user_bp && test_thread_flag(TIF_SINGLESTEP)))
+	if (!(per_task_bp &&
+	     (!is_kernel_addr(bp->attr.bp_addr)) &&
+	     test_thread_flag(TIF_SINGLESTEP)))
 		regs->msr &= ~MSR_SE;
 
 	set_dabr(bp_info->address | bp_info->type | DABR_TRANSLATION);