[net-next-2.6.git] / 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 (C) 2007 Alan Stern
 * Copyright (C) IBM Corporation, 2009
 * Copyright (C) 2009, Frederic Weisbecker <fweisbec@gmail.com>
 *
 * Thanks to Ingo Molnar for his many suggestions.
 *
 * Authors: Alan Stern <stern@rowland.harvard.edu>
 *          K.Prasad <prasad@linux.vnet.ibm.com>
 *          Frederic Weisbecker <fweisbec@gmail.com>
 */

/*
 * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
 * using the CPU's debug registers.
 * This file contains the arch-independent routines.
 */

#include <linux/irqflags.h>
#include <linux/kallsyms.h>
#include <linux/notifier.h>
#include <linux/kprobes.h>
#include <linux/kdebug.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/percpu.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/cpu.h>
#include <linux/smp.h>

#include <linux/hw_breakpoint.h>

enum bp_type_idx {
	TYPE_INST 	= 0,
#ifdef CONFIG_HAVE_MIXED_BREAKPOINTS_REGS
	TYPE_DATA	= 0,
#else
	TYPE_DATA	= 1,
#endif
	TYPE_MAX
};

/*
 * Constraints data
 */

/* Number of pinned cpu breakpoints in a cpu */
static DEFINE_PER_CPU(unsigned int, nr_cpu_bp_pinned[TYPE_MAX]);

/* Number of pinned task breakpoints in a cpu */
static DEFINE_PER_CPU(unsigned int, nr_task_bp_pinned[TYPE_MAX][HBP_NUM]);

/* Number of non-pinned cpu/task breakpoints in a cpu */
static DEFINE_PER_CPU(unsigned int, nr_bp_flexible[TYPE_MAX]);

/* Gather the number of total pinned and un-pinned bp in a cpuset */
struct bp_busy_slots {
	unsigned int pinned;
	unsigned int flexible;
};

/* Serialize accesses to the above constraints */
static DEFINE_MUTEX(nr_bp_mutex);

static inline enum bp_type_idx find_slot_idx(struct perf_event *bp)
{
	if (bp->attr.bp_type & HW_BREAKPOINT_RW)
		return TYPE_DATA;

	return TYPE_INST;
}

/*
 * Report the maximum number of pinned breakpoints a task
 * have in this cpu
 */
static unsigned int max_task_bp_pinned(int cpu, enum bp_type_idx type)
{
	int i;
	unsigned int *tsk_pinned = per_cpu(nr_task_bp_pinned[type], cpu);

	for (i = HBP_NUM -1; i >= 0; i--) {
		if (tsk_pinned[i] > 0)
			return i + 1;
	}

	return 0;
}

static int task_bp_pinned(struct task_struct *tsk, enum bp_type_idx type)
{
	struct perf_event_context *ctx = tsk->perf_event_ctxp;
	struct list_head *list;
	struct perf_event *bp;
	unsigned long flags;
	int count = 0;

	if (WARN_ONCE(!ctx, "No perf context for this task"))
		return 0;

	list = &ctx->event_list;

	raw_spin_lock_irqsave(&ctx->lock, flags);

	/*
	 * The current breakpoint counter is not included in the list
	 * at the open() callback time
	 */
	list_for_each_entry(bp, list, event_entry) {
		if (bp->attr.type == PERF_TYPE_BREAKPOINT)
			if (find_slot_idx(bp) == type)
				count++;
	}

	raw_spin_unlock_irqrestore(&ctx->lock, flags);

	return count;
}

/*
 * Report the number of pinned/un-pinned breakpoints we have in
 * a given cpu (cpu > -1) or in all of them (cpu = -1).
 */
static void
fetch_bp_busy_slots(struct bp_busy_slots *slots, struct perf_event *bp,
		    enum bp_type_idx type)
{
	int cpu = bp->cpu;
	struct task_struct *tsk = bp->ctx->task;

	if (cpu >= 0) {
		slots->pinned = per_cpu(nr_cpu_bp_pinned[type], cpu);
		if (!tsk)
			slots->pinned += max_task_bp_pinned(cpu, type);
		else
			slots->pinned += task_bp_pinned(tsk, type);
		slots->flexible = per_cpu(nr_bp_flexible[type], cpu);

		return;
	}

	for_each_online_cpu(cpu) {
		unsigned int nr;

		nr = per_cpu(nr_cpu_bp_pinned[type], cpu);
		if (!tsk)
			nr += max_task_bp_pinned(cpu, type);
		else
			nr += task_bp_pinned(tsk, type);

		if (nr > slots->pinned)
			slots->pinned = nr;

		nr = per_cpu(nr_bp_flexible[type], cpu);

		if (nr > slots->flexible)
			slots->flexible = nr;
	}
}

/*
 * Add a pinned breakpoint for the given task in our constraint table
 */
static void toggle_bp_task_slot(struct task_struct *tsk, int cpu, bool enable,
				enum bp_type_idx type)
{
	unsigned int *tsk_pinned;
	int count = 0;

	count = task_bp_pinned(tsk, type);

	tsk_pinned = per_cpu(nr_task_bp_pinned[type], cpu);
	if (enable) {
		tsk_pinned[count]++;
		if (count > 0)
			tsk_pinned[count-1]--;
	} else {
		tsk_pinned[count]--;
		if (count > 0)
			tsk_pinned[count-1]++;
	}
}

/*
 * Add/remove the given breakpoint in our constraint table
 */
static void
toggle_bp_slot(struct perf_event *bp, bool enable, enum bp_type_idx type)
{
	int cpu = bp->cpu;
	struct task_struct *tsk = bp->ctx->task;

	/* Pinned counter task profiling */
	if (tsk) {
		if (cpu >= 0) {
			toggle_bp_task_slot(tsk, cpu, enable, type);
			return;
		}

		for_each_online_cpu(cpu)
			toggle_bp_task_slot(tsk, cpu, enable, type);
		return;
	}

	/* Pinned counter cpu profiling */
	if (enable)
		per_cpu(nr_cpu_bp_pinned[type], bp->cpu)++;
	else
		per_cpu(nr_cpu_bp_pinned[type], bp->cpu)--;
}

/*
 * Contraints to check before allowing this new breakpoint counter:
 *
 *  == Non-pinned counter == (Considered as pinned for now)
 *
 *   - If attached to a single cpu, check:
 *
 *       (per_cpu(nr_bp_flexible, cpu) || (per_cpu(nr_cpu_bp_pinned, cpu)
 *           + max(per_cpu(nr_task_bp_pinned, cpu)))) < HBP_NUM
 *
 *       -> If there are already non-pinned counters in this cpu, it means
 *          there is already a free slot for them.
 *          Otherwise, we check that the maximum number of per task
 *          breakpoints (for this cpu) plus the number of per cpu breakpoint
 *          (for this cpu) doesn't cover every registers.
 *
 *   - If attached to every cpus, check:
 *
 *       (per_cpu(nr_bp_flexible, *) || (max(per_cpu(nr_cpu_bp_pinned, *))
 *           + max(per_cpu(nr_task_bp_pinned, *)))) < HBP_NUM
 *
 *       -> This is roughly the same, except we check the number of per cpu
 *          bp for every cpu and we keep the max one. Same for the per tasks
 *          breakpoints.
 *
 *
 * == Pinned counter ==
 *
 *   - If attached to a single cpu, check:
 *
 *       ((per_cpu(nr_bp_flexible, cpu) > 1) + per_cpu(nr_cpu_bp_pinned, cpu)
 *            + max(per_cpu(nr_task_bp_pinned, cpu))) < HBP_NUM
 *
 *       -> Same checks as before. But now the nr_bp_flexible, if any, must keep
 *          one register at least (or they will never be fed).
 *
 *   - If attached to every cpus, check:
 *
 *       ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *))
 *            + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM
 */
static int __reserve_bp_slot(struct perf_event *bp)
{
	struct bp_busy_slots slots = {0};
	enum bp_type_idx type;

	/* Basic checks */
	if (bp->attr.bp_type == HW_BREAKPOINT_EMPTY ||
	    bp->attr.bp_type == HW_BREAKPOINT_INVALID)
		return -EINVAL;

	type = find_slot_idx(bp);
	fetch_bp_busy_slots(&slots, bp, type);

	/* Flexible counters need to keep at least one slot */
	if (slots.pinned + (!!slots.flexible) == HBP_NUM)
		return -ENOSPC;

	toggle_bp_slot(bp, true, type);

	return 0;
}

int reserve_bp_slot(struct perf_event *bp)
{
	int ret;

	mutex_lock(&nr_bp_mutex);

	ret = __reserve_bp_slot(bp);

	mutex_unlock(&nr_bp_mutex);

	return ret;
}

static void __release_bp_slot(struct perf_event *bp)
{
	enum bp_type_idx type;

	type = find_slot_idx(bp);
	toggle_bp_slot(bp, false, type);
}

void release_bp_slot(struct perf_event *bp)
{
	mutex_lock(&nr_bp_mutex);

	__release_bp_slot(bp);

	mutex_unlock(&nr_bp_mutex);
}

/*
 * Allow the kernel debugger to reserve breakpoint slots without
 * taking a lock using the dbg_* variant of for the reserve and
 * release breakpoint slots.
 */
int dbg_reserve_bp_slot(struct perf_event *bp)
{
	if (mutex_is_locked(&nr_bp_mutex))
		return -1;

	return __reserve_bp_slot(bp);
}

int dbg_release_bp_slot(struct perf_event *bp)
{
	if (mutex_is_locked(&nr_bp_mutex))
		return -1;

	__release_bp_slot(bp);

	return 0;
}

static int validate_hw_breakpoint(struct perf_event *bp)
{
	int ret;

	ret = arch_validate_hwbkpt_settings(bp);
	if (ret)
		return ret;

	if (arch_check_bp_in_kernelspace(bp)) {
		if (bp->attr.exclude_kernel)
			return -EINVAL;
		/*
		 * Don't let unprivileged users set a breakpoint in the trap
		 * path to avoid trap recursion attacks.
		 */
		if (!capable(CAP_SYS_ADMIN))
			return -EPERM;
	}

	return 0;
}

int register_perf_hw_breakpoint(struct perf_event *bp)
{
	int ret;

	ret = reserve_bp_slot(bp);
	if (ret)
		return ret;

	ret = validate_hw_breakpoint(bp);

	/* if arch_validate_hwbkpt_settings() fails then release bp slot */
	if (ret)
		release_bp_slot(bp);

	return ret;
}

/**
 * register_user_hw_breakpoint - register a hardware breakpoint for user space
 * @attr: breakpoint attributes
 * @triggered: callback to trigger when we hit the breakpoint
 * @tsk: pointer to 'task_struct' of the process to which the address belongs
 */
struct perf_event *
register_user_hw_breakpoint(struct perf_event_attr *attr,
			    perf_overflow_handler_t triggered,
			    struct task_struct *tsk)
{
	return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered);
}
EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);

/**
 * modify_user_hw_breakpoint - modify a user-space hardware breakpoint
 * @bp: the breakpoint structure to modify
 * @attr: new breakpoint attributes
 * @triggered: callback to trigger when we hit the breakpoint
 * @tsk: pointer to 'task_struct' of the process to which the address belongs
 */
int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
{
	u64 old_addr = bp->attr.bp_addr;
	u64 old_len = bp->attr.bp_len;
	int old_type = bp->attr.bp_type;
	int err = 0;

	perf_event_disable(bp);

	bp->attr.bp_addr = attr->bp_addr;
	bp->attr.bp_type = attr->bp_type;
	bp->attr.bp_len = attr->bp_len;

	if (attr->disabled)
		goto end;

	err = validate_hw_breakpoint(bp);
	if (!err)
		perf_event_enable(bp);

	if (err) {
		bp->attr.bp_addr = old_addr;
		bp->attr.bp_type = old_type;
		bp->attr.bp_len = old_len;
		if (!bp->attr.disabled)
			perf_event_enable(bp);

		return err;
	}

end:
	bp->attr.disabled = attr->disabled;

	return 0;
}
EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);

/**
 * unregister_hw_breakpoint - unregister a user-space hardware breakpoint
 * @bp: the breakpoint structure to unregister
 */
void unregister_hw_breakpoint(struct perf_event *bp)
{
	if (!bp)
		return;
	perf_event_release_kernel(bp);
}
EXPORT_SYMBOL_GPL(unregister_hw_breakpoint);

/**
 * register_wide_hw_breakpoint - register a wide breakpoint in the kernel
 * @attr: breakpoint attributes
 * @triggered: callback to trigger when we hit the breakpoint
 *
 * @return a set of per_cpu pointers to perf events
 */
struct perf_event * __percpu *
register_wide_hw_breakpoint(struct perf_event_attr *attr,
			    perf_overflow_handler_t triggered)
{
	struct perf_event * __percpu *cpu_events, **pevent, *bp;
	long err;
	int cpu;

	cpu_events = alloc_percpu(typeof(*cpu_events));
	if (!cpu_events)
		return (void __percpu __force *)ERR_PTR(-ENOMEM);

	get_online_cpus();
	for_each_online_cpu(cpu) {
		pevent = per_cpu_ptr(cpu_events, cpu);
		bp = perf_event_create_kernel_counter(attr, cpu, -1, triggered);

		*pevent = bp;

		if (IS_ERR(bp)) {
			err = PTR_ERR(bp);
			goto fail;
		}
	}
	put_online_cpus();

	return cpu_events;

fail:
	for_each_online_cpu(cpu) {
		pevent = per_cpu_ptr(cpu_events, cpu);
		if (IS_ERR(*pevent))
			break;
		unregister_hw_breakpoint(*pevent);
	}
	put_online_cpus();

	free_percpu(cpu_events);
	return (void __percpu __force *)ERR_PTR(err);
}
EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint);

/**
 * unregister_wide_hw_breakpoint - unregister a wide breakpoint in the kernel
 * @cpu_events: the per cpu set of events to unregister
 */
void unregister_wide_hw_breakpoint(struct perf_event * __percpu *cpu_events)
{
	int cpu;
	struct perf_event **pevent;

	for_each_possible_cpu(cpu) {
		pevent = per_cpu_ptr(cpu_events, cpu);
		unregister_hw_breakpoint(*pevent);
	}
	free_percpu(cpu_events);
}
EXPORT_SYMBOL_GPL(unregister_wide_hw_breakpoint);

static struct notifier_block hw_breakpoint_exceptions_nb = {
	.notifier_call = hw_breakpoint_exceptions_notify,
	/* we need to be notified first */
	.priority = 0x7fffffff
};

static int __init init_hw_breakpoint(void)
{
	return register_die_notifier(&hw_breakpoint_exceptions_nb);
}
core_initcall(init_hw_breakpoint);


struct pmu perf_ops_bp = {
	.enable		= arch_install_hw_breakpoint,
	.disable	= arch_uninstall_hw_breakpoint,
	.read		= hw_breakpoint_pmu_read,
};
Commit	Line	Data
62a038d3 P	1	/*
	2	* This program is free software; you can redistribute it and/or modify
	3	* it under the terms of the GNU General Public License as published by
	4	* the Free Software Foundation; either version 2 of the License, or
	5	* (at your option) any later version.
	6	*
	7	* This program is distributed in the hope that it will be useful,
	8	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	9	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	10	* GNU General Public License for more details.
	11	*
	12	* You should have received a copy of the GNU General Public License
	13	* along with this program; if not, write to the Free Software
	14	* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
	15	*
	16	* Copyright (C) 2007 Alan Stern
	17	* Copyright (C) IBM Corporation, 2009
24f1e32c	18	* Copyright (C) 2009, Frederic Weisbecker <fweisbec@gmail.com>
ba1c813a FW	19	*
ba1c813a FW	20	* Thanks to Ingo Molnar for his many suggestions.
ba6909b7 P	21	*
	22	* Authors: Alan Stern <stern@rowland.harvard.edu>
	23	* K.Prasad <prasad@linux.vnet.ibm.com>
	24	* Frederic Weisbecker <fweisbec@gmail.com>
62a038d3 P	25	*/
	26
	27	/*
	28	* HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
	29	* using the CPU's debug registers.
	30	* This file contains the arch-independent routines.
	31	*/
	32
	33	#include <linux/irqflags.h>
	34	#include <linux/kallsyms.h>
	35	#include <linux/notifier.h>
	36	#include <linux/kprobes.h>
	37	#include <linux/kdebug.h>
	38	#include <linux/kernel.h>
	39	#include <linux/module.h>
	40	#include <linux/percpu.h>
	41	#include <linux/sched.h>
	42	#include <linux/init.h>
88f7a890	43	#include <linux/cpu.h>
62a038d3 P	44	#include <linux/smp.h>
62a038d3 P	45
24f1e32c FW	46	#include <linux/hw_breakpoint.h>
24f1e32c FW	47
0102752e FW	48	enum bp_type_idx {
	49	TYPE_INST = 0,
	50	#ifdef CONFIG_HAVE_MIXED_BREAKPOINTS_REGS
	51	TYPE_DATA = 0,
	52	#else
	53	TYPE_DATA = 1,
	54	#endif
	55	TYPE_MAX
	56	};
	57
ba1c813a FW	58	/*
	59	* Constraints data
	60	*/
62a038d3	61
ba1c813a	62	/* Number of pinned cpu breakpoints in a cpu */
0102752e	63	static DEFINE_PER_CPU(unsigned int, nr_cpu_bp_pinned[TYPE_MAX]);
ba1c813a FW	64
ba1c813a FW	65	/* Number of pinned task breakpoints in a cpu */
0102752e	66	static DEFINE_PER_CPU(unsigned int, nr_task_bp_pinned[TYPE_MAX][HBP_NUM]);
ba1c813a FW	67
ba1c813a FW	68	/* Number of non-pinned cpu/task breakpoints in a cpu */
0102752e	69	static DEFINE_PER_CPU(unsigned int, nr_bp_flexible[TYPE_MAX]);
ba1c813a FW	70
	71	/* Gather the number of total pinned and un-pinned bp in a cpuset */
	72	struct bp_busy_slots {
	73	unsigned int pinned;
	74	unsigned int flexible;
	75	};
	76
	77	/* Serialize accesses to the above constraints */
	78	static DEFINE_MUTEX(nr_bp_mutex);
	79
0102752e FW	80	static inline enum bp_type_idx find_slot_idx(struct perf_event *bp)
	81	{
	82	if (bp->attr.bp_type & HW_BREAKPOINT_RW)
	83	return TYPE_DATA;
	84
	85	return TYPE_INST;
	86	}
	87
ba1c813a FW	88	/*
	89	* Report the maximum number of pinned breakpoints a task
	90	* have in this cpu
	91	*/
0102752e	92	static unsigned int max_task_bp_pinned(int cpu, enum bp_type_idx type)
62a038d3	93	{
ba1c813a	94	int i;
0102752e	95	unsigned int *tsk_pinned = per_cpu(nr_task_bp_pinned[type], cpu);
62a038d3	96
ba1c813a FW	97	for (i = HBP_NUM -1; i >= 0; i--) {
	98	if (tsk_pinned[i] > 0)
	99	return i + 1;
62a038d3 P	100	}
62a038d3 P	101
24f1e32c	102	return 0;
62a038d3 P	103	}
62a038d3 P	104
0102752e	105	static int task_bp_pinned(struct task_struct *tsk, enum bp_type_idx type)
56053170 FW	106	{
	107	struct perf_event_context *ctx = tsk->perf_event_ctxp;
	108	struct list_head *list;
	109	struct perf_event *bp;
	110	unsigned long flags;
	111	int count = 0;
	112
	113	if (WARN_ONCE(!ctx, "No perf context for this task"))
	114	return 0;
	115
	116	list = &ctx->event_list;
	117
e625cce1	118	raw_spin_lock_irqsave(&ctx->lock, flags);
56053170 FW	119
	120	/*
	121	* The current breakpoint counter is not included in the list
	122	* at the open() callback time
	123	*/
	124	list_for_each_entry(bp, list, event_entry) {
	125	if (bp->attr.type == PERF_TYPE_BREAKPOINT)
0102752e FW	126	if (find_slot_idx(bp) == type)
0102752e FW	127	count++;
56053170 FW	128	}
56053170 FW	129
e625cce1	130	raw_spin_unlock_irqrestore(&ctx->lock, flags);
56053170 FW	131
	132	return count;
	133	}
	134
ba1c813a FW	135	/*
	136	* Report the number of pinned/un-pinned breakpoints we have in
	137	* a given cpu (cpu > -1) or in all of them (cpu = -1).
	138	*/
56053170	139	static void
0102752e FW	140	fetch_bp_busy_slots(struct bp_busy_slots slots, struct perf_event bp,
0102752e FW	141	enum bp_type_idx type)
ba1c813a	142	{
56053170 FW	143	int cpu = bp->cpu;
	144	struct task_struct *tsk = bp->ctx->task;
	145
ba1c813a	146	if (cpu >= 0) {
0102752e	147	slots->pinned = per_cpu(nr_cpu_bp_pinned[type], cpu);
56053170	148	if (!tsk)
0102752e	149	slots->pinned += max_task_bp_pinned(cpu, type);
56053170	150	else
0102752e FW	151	slots->pinned += task_bp_pinned(tsk, type);
0102752e FW	152	slots->flexible = per_cpu(nr_bp_flexible[type], cpu);
ba1c813a FW	153
	154	return;
	155	}
	156
	157	for_each_online_cpu(cpu) {
	158	unsigned int nr;
	159
0102752e	160	nr = per_cpu(nr_cpu_bp_pinned[type], cpu);
56053170	161	if (!tsk)
0102752e	162	nr += max_task_bp_pinned(cpu, type);
56053170	163	else
0102752e	164	nr += task_bp_pinned(tsk, type);
ba1c813a FW	165
	166	if (nr > slots->pinned)
	167	slots->pinned = nr;
	168
0102752e	169	nr = per_cpu(nr_bp_flexible[type], cpu);
ba1c813a FW	170
	171	if (nr > slots->flexible)
	172	slots->flexible = nr;
	173	}
	174	}
	175
	176	/*
	177	* Add a pinned breakpoint for the given task in our constraint table
	178	*/
0102752e FW	179	static void toggle_bp_task_slot(struct task_struct *tsk, int cpu, bool enable,
0102752e FW	180	enum bp_type_idx type)
ba1c813a	181	{
11e66357	182	unsigned int *tsk_pinned;
56053170	183	int count = 0;
ba1c813a	184
0102752e	185	count = task_bp_pinned(tsk, type);
ba1c813a	186
0102752e	187	tsk_pinned = per_cpu(nr_task_bp_pinned[type], cpu);
ba1c813a	188	if (enable) {
11e66357	189	tsk_pinned[count]++;
ba1c813a	190	if (count > 0)
11e66357	191	tsk_pinned[count-1]--;
ba1c813a	192	} else {
11e66357	193	tsk_pinned[count]--;
ba1c813a	194	if (count > 0)
11e66357	195	tsk_pinned[count-1]++;
ba1c813a FW	196	}
	197	}
	198
	199	/*
	200	* Add/remove the given breakpoint in our constraint table
	201	*/
0102752e FW	202	static void
0102752e FW	203	toggle_bp_slot(struct perf_event *bp, bool enable, enum bp_type_idx type)
ba1c813a FW	204	{
	205	int cpu = bp->cpu;
	206	struct task_struct *tsk = bp->ctx->task;
	207
	208	/* Pinned counter task profiling */
	209	if (tsk) {
	210	if (cpu >= 0) {
0102752e	211	toggle_bp_task_slot(tsk, cpu, enable, type);
ba1c813a FW	212	return;
	213	}
	214
	215	for_each_online_cpu(cpu)
0102752e	216	toggle_bp_task_slot(tsk, cpu, enable, type);
ba1c813a FW	217	return;
	218	}
	219
	220	/* Pinned counter cpu profiling */
	221	if (enable)
0102752e	222	per_cpu(nr_cpu_bp_pinned[type], bp->cpu)++;
ba1c813a	223	else
0102752e	224	per_cpu(nr_cpu_bp_pinned[type], bp->cpu)--;
ba1c813a FW	225	}
	226
	227	/*
	228	* Contraints to check before allowing this new breakpoint counter:
	229	*
	230	* == Non-pinned counter == (Considered as pinned for now)
	231	*
	232	* - If attached to a single cpu, check:
	233	*
	234	* (per_cpu(nr_bp_flexible, cpu) \|\| (per_cpu(nr_cpu_bp_pinned, cpu)
6ab88863	235	* + max(per_cpu(nr_task_bp_pinned, cpu)))) < HBP_NUM
ba1c813a FW	236	*
	237	* -> If there are already non-pinned counters in this cpu, it means
	238	* there is already a free slot for them.
	239	* Otherwise, we check that the maximum number of per task
	240	* breakpoints (for this cpu) plus the number of per cpu breakpoint
	241	* (for this cpu) doesn't cover every registers.
	242	*
	243	* - If attached to every cpus, check:
	244	*
	245	* (per_cpu(nr_bp_flexible, ) \|\| (max(per_cpu(nr_cpu_bp_pinned, ))
6ab88863	246	* + max(per_cpu(nr_task_bp_pinned, *)))) < HBP_NUM
ba1c813a FW	247	*
	248	* -> This is roughly the same, except we check the number of per cpu
	249	* bp for every cpu and we keep the max one. Same for the per tasks
	250	* breakpoints.
	251	*
	252	*
	253	* == Pinned counter ==
	254	*
	255	* - If attached to a single cpu, check:
	256	*
	257	* ((per_cpu(nr_bp_flexible, cpu) > 1) + per_cpu(nr_cpu_bp_pinned, cpu)
6ab88863	258	* + max(per_cpu(nr_task_bp_pinned, cpu))) < HBP_NUM
ba1c813a FW	259	*
	260	* -> Same checks as before. But now the nr_bp_flexible, if any, must keep
	261	* one register at least (or they will never be fed).
	262	*
	263	* - If attached to every cpus, check:
	264	*
	265	* ((per_cpu(nr_bp_flexible, ) > 1) + max(per_cpu(nr_cpu_bp_pinned, ))
6ab88863	266	* + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM
ba1c813a	267	*/
5352ae63	268	static int __reserve_bp_slot(struct perf_event *bp)
ba1c813a FW	269	{
ba1c813a FW	270	struct bp_busy_slots slots = {0};
0102752e	271	enum bp_type_idx type;
ba1c813a	272
0102752e FW	273	/* Basic checks */
	274	if (bp->attr.bp_type == HW_BREAKPOINT_EMPTY \|\|
	275	bp->attr.bp_type == HW_BREAKPOINT_INVALID)
	276	return -EINVAL;
	277
	278	type = find_slot_idx(bp);
	279	fetch_bp_busy_slots(&slots, bp, type);
ba1c813a FW	280
ba1c813a FW	281	/* Flexible counters need to keep at least one slot */
5352ae63 JW	282	if (slots.pinned + (!!slots.flexible) == HBP_NUM)
5352ae63 JW	283	return -ENOSPC;
ba1c813a	284
0102752e	285	toggle_bp_slot(bp, true, type);
ba1c813a	286
5352ae63 JW	287	return 0;
	288	}
	289
	290	int reserve_bp_slot(struct perf_event *bp)
	291	{
	292	int ret;
	293
	294	mutex_lock(&nr_bp_mutex);
	295
	296	ret = __reserve_bp_slot(bp);
	297
ba1c813a FW	298	mutex_unlock(&nr_bp_mutex);
	299
	300	return ret;
	301	}
	302
5352ae63 JW	303	static void __release_bp_slot(struct perf_event *bp)
5352ae63 JW	304	{
0102752e FW	305	enum bp_type_idx type;
	306
	307	type = find_slot_idx(bp);
	308	toggle_bp_slot(bp, false, type);
5352ae63 JW	309	}
5352ae63 JW	310
24f1e32c	311	void release_bp_slot(struct perf_event *bp)
62a038d3	312	{
ba1c813a FW	313	mutex_lock(&nr_bp_mutex);
ba1c813a FW	314
5352ae63	315	__release_bp_slot(bp);
ba1c813a FW	316
ba1c813a FW	317	mutex_unlock(&nr_bp_mutex);
62a038d3 P	318	}
62a038d3 P	319
5352ae63 JW	320	/*
	321	* Allow the kernel debugger to reserve breakpoint slots without
	322	* taking a lock using the dbg_* variant of for the reserve and
	323	* release breakpoint slots.
	324	*/
	325	int dbg_reserve_bp_slot(struct perf_event *bp)
	326	{
	327	if (mutex_is_locked(&nr_bp_mutex))
	328	return -1;
	329
	330	return __reserve_bp_slot(bp);
	331	}
	332
	333	int dbg_release_bp_slot(struct perf_event *bp)
	334	{
	335	if (mutex_is_locked(&nr_bp_mutex))
	336	return -1;
	337
	338	__release_bp_slot(bp);
	339
	340	return 0;
	341	}
ba1c813a	342
b2812d03 FW	343	static int validate_hw_breakpoint(struct perf_event *bp)
	344	{
	345	int ret;
	346
	347	ret = arch_validate_hwbkpt_settings(bp);
	348	if (ret)
	349	return ret;
	350
	351	if (arch_check_bp_in_kernelspace(bp)) {
	352	if (bp->attr.exclude_kernel)
	353	return -EINVAL;
	354	/*
	355	* Don't let unprivileged users set a breakpoint in the trap
	356	* path to avoid trap recursion attacks.
	357	*/
	358	if (!capable(CAP_SYS_ADMIN))
	359	return -EPERM;
	360	}
	361
	362	return 0;
	363	}
	364
b326e956	365	int register_perf_hw_breakpoint(struct perf_event *bp)
62a038d3	366	{
24f1e32c	367	int ret;
62a038d3	368
24f1e32c FW	369	ret = reserve_bp_slot(bp);
	370	if (ret)
	371	return ret;
62a038d3	372
b2812d03	373	ret = validate_hw_breakpoint(bp);
62a038d3	374
b23ff0e9 MS	375	/* if arch_validate_hwbkpt_settings() fails then release bp slot */
	376	if (ret)
	377	release_bp_slot(bp);
	378
24f1e32c FW	379	return ret;
24f1e32c FW	380	}
62a038d3	381
62a038d3 P	382	/**
62a038d3 P	383	* register_user_hw_breakpoint - register a hardware breakpoint for user space
5fa10b28	384	* @attr: breakpoint attributes
24f1e32c	385	* @triggered: callback to trigger when we hit the breakpoint
62a038d3	386	* @tsk: pointer to 'task_struct' of the process to which the address belongs
62a038d3	387	*/
24f1e32c	388	struct perf_event *
5fa10b28	389	register_user_hw_breakpoint(struct perf_event_attr *attr,
b326e956	390	perf_overflow_handler_t triggered,
5fa10b28	391	struct task_struct *tsk)
62a038d3	392	{
5fa10b28	393	return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered);
62a038d3 P	394	}
	395	EXPORT_SYMBOL_GPL(register_user_hw_breakpoint);
	396
	397	/**
	398	* modify_user_hw_breakpoint - modify a user-space hardware breakpoint
24f1e32c	399	* @bp: the breakpoint structure to modify
5fa10b28	400	* @attr: new breakpoint attributes
24f1e32c	401	* @triggered: callback to trigger when we hit the breakpoint
62a038d3	402	* @tsk: pointer to 'task_struct' of the process to which the address belongs
62a038d3	403	*/
44234adc	404	int modify_user_hw_breakpoint(struct perf_event bp, struct perf_event_attr attr)
62a038d3	405	{
44234adc	406	u64 old_addr = bp->attr.bp_addr;
cd757645	407	u64 old_len = bp->attr.bp_len;
44234adc	408	int old_type = bp->attr.bp_type;
44234adc FW	409	int err = 0;
	410
	411	perf_event_disable(bp);
	412
	413	bp->attr.bp_addr = attr->bp_addr;
	414	bp->attr.bp_type = attr->bp_type;
	415	bp->attr.bp_len = attr->bp_len;
	416
	417	if (attr->disabled)
	418	goto end;
62a038d3	419
b2812d03	420	err = validate_hw_breakpoint(bp);
44234adc FW	421	if (!err)
	422	perf_event_enable(bp);
	423
	424	if (err) {
	425	bp->attr.bp_addr = old_addr;
	426	bp->attr.bp_type = old_type;
	427	bp->attr.bp_len = old_len;
	428	if (!bp->attr.disabled)
	429	perf_event_enable(bp);
	430
	431	return err;
	432	}
	433
	434	end:
	435	bp->attr.disabled = attr->disabled;
	436
	437	return 0;
62a038d3 P	438	}
	439	EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);
	440
	441	/**
24f1e32c	442	* unregister_hw_breakpoint - unregister a user-space hardware breakpoint
62a038d3	443	* @bp: the breakpoint structure to unregister
62a038d3	444	*/
24f1e32c	445	void unregister_hw_breakpoint(struct perf_event *bp)
62a038d3	446	{
24f1e32c FW	447	if (!bp)
	448	return;
	449	perf_event_release_kernel(bp);
	450	}
	451	EXPORT_SYMBOL_GPL(unregister_hw_breakpoint);
	452
62a038d3	453	/**
24f1e32c	454	* register_wide_hw_breakpoint - register a wide breakpoint in the kernel
dd1853c3	455	* @attr: breakpoint attributes
24f1e32c	456	* @triggered: callback to trigger when we hit the breakpoint
62a038d3	457	*
24f1e32c	458	* @return a set of per_cpu pointers to perf events
62a038d3	459	*/
44ee6358	460	struct perf_event * __percpu *
dd1853c3	461	register_wide_hw_breakpoint(struct perf_event_attr *attr,
b326e956	462	perf_overflow_handler_t triggered)
62a038d3	463	{
44ee6358	464	struct perf_event * __percpu cpu_events, pevent, bp;
24f1e32c FW	465	long err;
	466	int cpu;
	467
	468	cpu_events = alloc_percpu(typeof(*cpu_events));
	469	if (!cpu_events)
44ee6358	470	return (void __percpu __force *)ERR_PTR(-ENOMEM);
62a038d3	471
88f7a890 LZ	472	get_online_cpus();
88f7a890 LZ	473	for_each_online_cpu(cpu) {
24f1e32c	474	pevent = per_cpu_ptr(cpu_events, cpu);
dd1853c3	475	bp = perf_event_create_kernel_counter(attr, cpu, -1, triggered);
62a038d3	476
24f1e32c	477	*pevent = bp;
62a038d3	478
605bfaee	479	if (IS_ERR(bp)) {
24f1e32c FW	480	err = PTR_ERR(bp);
	481	goto fail;
	482	}
62a038d3	483	}
88f7a890	484	put_online_cpus();
62a038d3	485
24f1e32c FW	486	return cpu_events;
	487
	488	fail:
88f7a890	489	for_each_online_cpu(cpu) {
24f1e32c	490	pevent = per_cpu_ptr(cpu_events, cpu);
605bfaee	491	if (IS_ERR(*pevent))
24f1e32c FW	492	break;
	493	unregister_hw_breakpoint(*pevent);
	494	}
88f7a890 LZ	495	put_online_cpus();
88f7a890 LZ	496
24f1e32c	497	free_percpu(cpu_events);
44ee6358	498	return (void __percpu __force *)ERR_PTR(err);
62a038d3	499	}
f60d24d2	500	EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint);
62a038d3 P	501
62a038d3 P	502	/**
24f1e32c FW	503	* unregister_wide_hw_breakpoint - unregister a wide breakpoint in the kernel
24f1e32c FW	504	* @cpu_events: the per cpu set of events to unregister
62a038d3	505	*/
44ee6358	506	void unregister_wide_hw_breakpoint(struct perf_event * __percpu *cpu_events)
62a038d3	507	{
24f1e32c FW	508	int cpu;
24f1e32c FW	509	struct perf_event **pevent;
62a038d3	510
24f1e32c FW	511	for_each_possible_cpu(cpu) {
	512	pevent = per_cpu_ptr(cpu_events, cpu);
	513	unregister_hw_breakpoint(*pevent);
62a038d3	514	}
24f1e32c	515	free_percpu(cpu_events);
62a038d3	516	}
f60d24d2	517	EXPORT_SYMBOL_GPL(unregister_wide_hw_breakpoint);
62a038d3 P	518
	519	static struct notifier_block hw_breakpoint_exceptions_nb = {
	520	.notifier_call = hw_breakpoint_exceptions_notify,
	521	/* we need to be notified first */
	522	.priority = 0x7fffffff
	523	};
	524
	525	static int __init init_hw_breakpoint(void)
	526	{
	527	return register_die_notifier(&hw_breakpoint_exceptions_nb);
	528	}
62a038d3	529	core_initcall(init_hw_breakpoint);
24f1e32c FW	530
	531
	532	struct pmu perf_ops_bp = {
	533	.enable = arch_install_hw_breakpoint,
	534	.disable = arch_uninstall_hw_breakpoint,
	535	.read = hw_breakpoint_pmu_read,
24f1e32c	536	};