Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
4 * using the CPU's debug registers. Derived from
5 * "arch/x86/kernel/hw_breakpoint.c"
6 *
7 * Copyright 2010 IBM Corporation
8 * Author: K.Prasad <prasad@linux.vnet.ibm.com>
9 */
10
11#include <linux/hw_breakpoint.h>
12#include <linux/notifier.h>
13#include <linux/kprobes.h>
14#include <linux/percpu.h>
15#include <linux/kernel.h>
16#include <linux/sched.h>
17#include <linux/smp.h>
18#include <linux/spinlock.h>
19#include <linux/debugfs.h>
20#include <linux/init.h>
21
22#include <asm/hw_breakpoint.h>
23#include <asm/processor.h>
24#include <asm/sstep.h>
25#include <asm/debug.h>
26#include <asm/hvcall.h>
27#include <asm/inst.h>
28#include <linux/uaccess.h>
29
30/*
31 * Stores the breakpoints currently in use on each breakpoint address
32 * register for every cpu
33 */
34static DEFINE_PER_CPU(struct perf_event *, bp_per_reg[HBP_NUM_MAX]);
35
36/*
37 * Returns total number of data or instruction breakpoints available.
38 */
39int hw_breakpoint_slots(int type)
40{
41 if (type == TYPE_DATA)
42 return nr_wp_slots();
43 return 0; /* no instruction breakpoints available */
44}
45
46static bool single_step_pending(void)
47{
48 int i;
49
50 for (i = 0; i < nr_wp_slots(); i++) {
51 if (current->thread.last_hit_ubp[i])
52 return true;
53 }
54 return false;
55}
56
57/*
58 * Install a perf counter breakpoint.
59 *
60 * We seek a free debug address register and use it for this
61 * breakpoint.
62 *
63 * Atomic: we hold the counter->ctx->lock and we only handle variables
64 * and registers local to this cpu.
65 */
66int arch_install_hw_breakpoint(struct perf_event *bp)
67{
68 struct arch_hw_breakpoint *info = counter_arch_bp(bp);
69 struct perf_event **slot;
70 int i;
71
72 for (i = 0; i < nr_wp_slots(); i++) {
73 slot = this_cpu_ptr(&bp_per_reg[i]);
74 if (!*slot) {
75 *slot = bp;
76 break;
77 }
78 }
79
80 if (WARN_ONCE(i == nr_wp_slots(), "Can't find any breakpoint slot"))
81 return -EBUSY;
82
83 /*
84 * Do not install DABR values if the instruction must be single-stepped.
85 * If so, DABR will be populated in single_step_dabr_instruction().
86 */
87 if (!single_step_pending())
88 __set_breakpoint(i, info);
89
90 return 0;
91}
92
93/*
94 * Uninstall the breakpoint contained in the given counter.
95 *
96 * First we search the debug address register it uses and then we disable
97 * it.
98 *
99 * Atomic: we hold the counter->ctx->lock and we only handle variables
100 * and registers local to this cpu.
101 */
102void arch_uninstall_hw_breakpoint(struct perf_event *bp)
103{
104 struct arch_hw_breakpoint null_brk = {0};
105 struct perf_event **slot;
106 int i;
107
108 for (i = 0; i < nr_wp_slots(); i++) {
109 slot = this_cpu_ptr(&bp_per_reg[i]);
110 if (*slot == bp) {
111 *slot = NULL;
112 break;
113 }
114 }
115
116 if (WARN_ONCE(i == nr_wp_slots(), "Can't find any breakpoint slot"))
117 return;
118
119 __set_breakpoint(i, &null_brk);
120}
121
122static bool is_ptrace_bp(struct perf_event *bp)
123{
124 return bp->overflow_handler == ptrace_triggered;
125}
126
127struct breakpoint {
128 struct list_head list;
129 struct perf_event *bp;
130 bool ptrace_bp;
131};
132
133/*
134 * While kernel/events/hw_breakpoint.c does its own synchronization, we cannot
135 * rely on it safely synchronizing internals here; however, we can rely on it
136 * not requesting more breakpoints than available.
137 */
138static DEFINE_SPINLOCK(cpu_bps_lock);
139static DEFINE_PER_CPU(struct breakpoint *, cpu_bps[HBP_NUM_MAX]);
140static DEFINE_SPINLOCK(task_bps_lock);
141static LIST_HEAD(task_bps);
142
143static struct breakpoint *alloc_breakpoint(struct perf_event *bp)
144{
145 struct breakpoint *tmp;
146
147 tmp = kzalloc(sizeof(*tmp), GFP_KERNEL);
148 if (!tmp)
149 return ERR_PTR(-ENOMEM);
150 tmp->bp = bp;
151 tmp->ptrace_bp = is_ptrace_bp(bp);
152 return tmp;
153}
154
155static bool bp_addr_range_overlap(struct perf_event *bp1, struct perf_event *bp2)
156{
157 __u64 bp1_saddr, bp1_eaddr, bp2_saddr, bp2_eaddr;
158
159 bp1_saddr = ALIGN_DOWN(bp1->attr.bp_addr, HW_BREAKPOINT_SIZE);
160 bp1_eaddr = ALIGN(bp1->attr.bp_addr + bp1->attr.bp_len, HW_BREAKPOINT_SIZE);
161 bp2_saddr = ALIGN_DOWN(bp2->attr.bp_addr, HW_BREAKPOINT_SIZE);
162 bp2_eaddr = ALIGN(bp2->attr.bp_addr + bp2->attr.bp_len, HW_BREAKPOINT_SIZE);
163
164 return (bp1_saddr < bp2_eaddr && bp1_eaddr > bp2_saddr);
165}
166
167static bool alternate_infra_bp(struct breakpoint *b, struct perf_event *bp)
168{
169 return is_ptrace_bp(bp) ? !b->ptrace_bp : b->ptrace_bp;
170}
171
172static bool can_co_exist(struct breakpoint *b, struct perf_event *bp)
173{
174 return !(alternate_infra_bp(b, bp) && bp_addr_range_overlap(b->bp, bp));
175}
176
177static int task_bps_add(struct perf_event *bp)
178{
179 struct breakpoint *tmp;
180
181 tmp = alloc_breakpoint(bp);
182 if (IS_ERR(tmp))
183 return PTR_ERR(tmp);
184
185 spin_lock(&task_bps_lock);
186 list_add(&tmp->list, &task_bps);
187 spin_unlock(&task_bps_lock);
188 return 0;
189}
190
191static void task_bps_remove(struct perf_event *bp)
192{
193 struct list_head *pos, *q;
194
195 spin_lock(&task_bps_lock);
196 list_for_each_safe(pos, q, &task_bps) {
197 struct breakpoint *tmp = list_entry(pos, struct breakpoint, list);
198
199 if (tmp->bp == bp) {
200 list_del(&tmp->list);
201 kfree(tmp);
202 break;
203 }
204 }
205 spin_unlock(&task_bps_lock);
206}
207
208/*
209 * If any task has breakpoint from alternate infrastructure,
210 * return true. Otherwise return false.
211 */
212static bool all_task_bps_check(struct perf_event *bp)
213{
214 struct breakpoint *tmp;
215 bool ret = false;
216
217 spin_lock(&task_bps_lock);
218 list_for_each_entry(tmp, &task_bps, list) {
219 if (!can_co_exist(tmp, bp)) {
220 ret = true;
221 break;
222 }
223 }
224 spin_unlock(&task_bps_lock);
225 return ret;
226}
227
228/*
229 * If same task has breakpoint from alternate infrastructure,
230 * return true. Otherwise return false.
231 */
232static bool same_task_bps_check(struct perf_event *bp)
233{
234 struct breakpoint *tmp;
235 bool ret = false;
236
237 spin_lock(&task_bps_lock);
238 list_for_each_entry(tmp, &task_bps, list) {
239 if (tmp->bp->hw.target == bp->hw.target &&
240 !can_co_exist(tmp, bp)) {
241 ret = true;
242 break;
243 }
244 }
245 spin_unlock(&task_bps_lock);
246 return ret;
247}
248
249static int cpu_bps_add(struct perf_event *bp)
250{
251 struct breakpoint **cpu_bp;
252 struct breakpoint *tmp;
253 int i = 0;
254
255 tmp = alloc_breakpoint(bp);
256 if (IS_ERR(tmp))
257 return PTR_ERR(tmp);
258
259 spin_lock(&cpu_bps_lock);
260 cpu_bp = per_cpu_ptr(cpu_bps, bp->cpu);
261 for (i = 0; i < nr_wp_slots(); i++) {
262 if (!cpu_bp[i]) {
263 cpu_bp[i] = tmp;
264 break;
265 }
266 }
267 spin_unlock(&cpu_bps_lock);
268 return 0;
269}
270
271static void cpu_bps_remove(struct perf_event *bp)
272{
273 struct breakpoint **cpu_bp;
274 int i = 0;
275
276 spin_lock(&cpu_bps_lock);
277 cpu_bp = per_cpu_ptr(cpu_bps, bp->cpu);
278 for (i = 0; i < nr_wp_slots(); i++) {
279 if (!cpu_bp[i])
280 continue;
281
282 if (cpu_bp[i]->bp == bp) {
283 kfree(cpu_bp[i]);
284 cpu_bp[i] = NULL;
285 break;
286 }
287 }
288 spin_unlock(&cpu_bps_lock);
289}
290
291static bool cpu_bps_check(int cpu, struct perf_event *bp)
292{
293 struct breakpoint **cpu_bp;
294 bool ret = false;
295 int i;
296
297 spin_lock(&cpu_bps_lock);
298 cpu_bp = per_cpu_ptr(cpu_bps, cpu);
299 for (i = 0; i < nr_wp_slots(); i++) {
300 if (cpu_bp[i] && !can_co_exist(cpu_bp[i], bp)) {
301 ret = true;
302 break;
303 }
304 }
305 spin_unlock(&cpu_bps_lock);
306 return ret;
307}
308
309static bool all_cpu_bps_check(struct perf_event *bp)
310{
311 int cpu;
312
313 for_each_online_cpu(cpu) {
314 if (cpu_bps_check(cpu, bp))
315 return true;
316 }
317 return false;
318}
319
320int arch_reserve_bp_slot(struct perf_event *bp)
321{
322 int ret;
323
324 /* ptrace breakpoint */
325 if (is_ptrace_bp(bp)) {
326 if (all_cpu_bps_check(bp))
327 return -ENOSPC;
328
329 if (same_task_bps_check(bp))
330 return -ENOSPC;
331
332 return task_bps_add(bp);
333 }
334
335 /* perf breakpoint */
336 if (is_kernel_addr(bp->attr.bp_addr))
337 return 0;
338
339 if (bp->hw.target && bp->cpu == -1) {
340 if (same_task_bps_check(bp))
341 return -ENOSPC;
342
343 return task_bps_add(bp);
344 } else if (!bp->hw.target && bp->cpu != -1) {
345 if (all_task_bps_check(bp))
346 return -ENOSPC;
347
348 return cpu_bps_add(bp);
349 }
350
351 if (same_task_bps_check(bp))
352 return -ENOSPC;
353
354 ret = cpu_bps_add(bp);
355 if (ret)
356 return ret;
357 ret = task_bps_add(bp);
358 if (ret)
359 cpu_bps_remove(bp);
360
361 return ret;
362}
363
364void arch_release_bp_slot(struct perf_event *bp)
365{
366 if (!is_kernel_addr(bp->attr.bp_addr)) {
367 if (bp->hw.target)
368 task_bps_remove(bp);
369 if (bp->cpu != -1)
370 cpu_bps_remove(bp);
371 }
372}
373
374/*
375 * Perform cleanup of arch-specific counters during unregistration
376 * of the perf-event
377 */
378void arch_unregister_hw_breakpoint(struct perf_event *bp)
379{
380 /*
381 * If the breakpoint is unregistered between a hw_breakpoint_handler()
382 * and the single_step_dabr_instruction(), then cleanup the breakpoint
383 * restoration variables to prevent dangling pointers.
384 * FIXME, this should not be using bp->ctx at all! Sayeth peterz.
385 */
386 if (bp->ctx && bp->ctx->task && bp->ctx->task != ((void *)-1L)) {
387 int i;
388
389 for (i = 0; i < nr_wp_slots(); i++) {
390 if (bp->ctx->task->thread.last_hit_ubp[i] == bp)
391 bp->ctx->task->thread.last_hit_ubp[i] = NULL;
392 }
393 }
394}
395
396/*
397 * Check for virtual address in kernel space.
398 */
399int arch_check_bp_in_kernelspace(struct arch_hw_breakpoint *hw)
400{
401 return is_kernel_addr(hw->address);
402}
403
404int arch_bp_generic_fields(int type, int *gen_bp_type)
405{
406 *gen_bp_type = 0;
407 if (type & HW_BRK_TYPE_READ)
408 *gen_bp_type |= HW_BREAKPOINT_R;
409 if (type & HW_BRK_TYPE_WRITE)
410 *gen_bp_type |= HW_BREAKPOINT_W;
411 if (*gen_bp_type == 0)
412 return -EINVAL;
413 return 0;
414}
415
416/*
417 * Watchpoint match range is always doubleword(8 bytes) aligned on
418 * powerpc. If the given range is crossing doubleword boundary, we
419 * need to increase the length such that next doubleword also get
420 * covered. Ex,
421 *
422 * address len = 6 bytes
423 * |=========.
424 * |------------v--|------v--------|
425 * | | | | | | | | | | | | | | | | |
426 * |---------------|---------------|
427 * <---8 bytes--->
428 *
429 * In this case, we should configure hw as:
430 * start_addr = address & ~(HW_BREAKPOINT_SIZE - 1)
431 * len = 16 bytes
432 *
433 * @start_addr is inclusive but @end_addr is exclusive.
434 */
435static int hw_breakpoint_validate_len(struct arch_hw_breakpoint *hw)
436{
437 u16 max_len = DABR_MAX_LEN;
438 u16 hw_len;
439 unsigned long start_addr, end_addr;
440
441 start_addr = ALIGN_DOWN(hw->address, HW_BREAKPOINT_SIZE);
442 end_addr = ALIGN(hw->address + hw->len, HW_BREAKPOINT_SIZE);
443 hw_len = end_addr - start_addr;
444
445 if (dawr_enabled()) {
446 max_len = DAWR_MAX_LEN;
447 /* DAWR region can't cross 512 bytes boundary on p10 predecessors */
448 if (!cpu_has_feature(CPU_FTR_ARCH_31) &&
449 (ALIGN_DOWN(start_addr, SZ_512) != ALIGN_DOWN(end_addr - 1, SZ_512)))
450 return -EINVAL;
451 } else if (IS_ENABLED(CONFIG_PPC_8xx)) {
452 /* 8xx can setup a range without limitation */
453 max_len = U16_MAX;
454 }
455
456 if (hw_len > max_len)
457 return -EINVAL;
458
459 hw->hw_len = hw_len;
460 return 0;
461}
462
463/*
464 * Validate the arch-specific HW Breakpoint register settings
465 */
466int hw_breakpoint_arch_parse(struct perf_event *bp,
467 const struct perf_event_attr *attr,
468 struct arch_hw_breakpoint *hw)
469{
470 int ret = -EINVAL;
471
472 if (!bp || !attr->bp_len)
473 return ret;
474
475 hw->type = HW_BRK_TYPE_TRANSLATE;
476 if (attr->bp_type & HW_BREAKPOINT_R)
477 hw->type |= HW_BRK_TYPE_READ;
478 if (attr->bp_type & HW_BREAKPOINT_W)
479 hw->type |= HW_BRK_TYPE_WRITE;
480 if (hw->type == HW_BRK_TYPE_TRANSLATE)
481 /* must set alteast read or write */
482 return ret;
483 if (!attr->exclude_user)
484 hw->type |= HW_BRK_TYPE_USER;
485 if (!attr->exclude_kernel)
486 hw->type |= HW_BRK_TYPE_KERNEL;
487 if (!attr->exclude_hv)
488 hw->type |= HW_BRK_TYPE_HYP;
489 hw->address = attr->bp_addr;
490 hw->len = attr->bp_len;
491
492 if (!ppc_breakpoint_available())
493 return -ENODEV;
494
495 return hw_breakpoint_validate_len(hw);
496}
497
498/*
499 * Restores the breakpoint on the debug registers.
500 * Invoke this function if it is known that the execution context is
501 * about to change to cause loss of MSR_SE settings.
502 */
503void thread_change_pc(struct task_struct *tsk, struct pt_regs *regs)
504{
505 struct arch_hw_breakpoint *info;
506 int i;
507
508 for (i = 0; i < nr_wp_slots(); i++) {
509 if (unlikely(tsk->thread.last_hit_ubp[i]))
510 goto reset;
511 }
512 return;
513
514reset:
515 regs_set_return_msr(regs, regs->msr & ~MSR_SE);
516 for (i = 0; i < nr_wp_slots(); i++) {
517 info = counter_arch_bp(__this_cpu_read(bp_per_reg[i]));
518 __set_breakpoint(i, info);
519 tsk->thread.last_hit_ubp[i] = NULL;
520 }
521}
522
523static bool is_larx_stcx_instr(int type)
524{
525 return type == LARX || type == STCX;
526}
527
528static bool is_octword_vsx_instr(int type, int size)
529{
530 return ((type == LOAD_VSX || type == STORE_VSX) && size == 32);
531}
532
533/*
534 * We've failed in reliably handling the hw-breakpoint. Unregister
535 * it and throw a warning message to let the user know about it.
536 */
537static void handler_error(struct perf_event *bp, struct arch_hw_breakpoint *info)
538{
539 WARN(1, "Unable to handle hardware breakpoint. Breakpoint at 0x%lx will be disabled.",
540 info->address);
541 perf_event_disable_inatomic(bp);
542}
543
544static void larx_stcx_err(struct perf_event *bp, struct arch_hw_breakpoint *info)
545{
546 printk_ratelimited("Breakpoint hit on instruction that can't be emulated. Breakpoint at 0x%lx will be disabled.\n",
547 info->address);
548 perf_event_disable_inatomic(bp);
549}
550
551static bool stepping_handler(struct pt_regs *regs, struct perf_event **bp,
552 struct arch_hw_breakpoint **info, int *hit,
553 ppc_inst_t instr)
554{
555 int i;
556 int stepped;
557
558 /* Do not emulate user-space instructions, instead single-step them */
559 if (user_mode(regs)) {
560 for (i = 0; i < nr_wp_slots(); i++) {
561 if (!hit[i])
562 continue;
563 current->thread.last_hit_ubp[i] = bp[i];
564 info[i] = NULL;
565 }
566 regs_set_return_msr(regs, regs->msr | MSR_SE);
567 return false;
568 }
569
570 stepped = emulate_step(regs, instr);
571 if (!stepped) {
572 for (i = 0; i < nr_wp_slots(); i++) {
573 if (!hit[i])
574 continue;
575 handler_error(bp[i], info[i]);
576 info[i] = NULL;
577 }
578 return false;
579 }
580 return true;
581}
582
583static void handle_p10dd1_spurious_exception(struct arch_hw_breakpoint **info,
584 int *hit, unsigned long ea)
585{
586 int i;
587 unsigned long hw_end_addr;
588
589 /*
590 * Handle spurious exception only when any bp_per_reg is set.
591 * Otherwise this might be created by xmon and not actually a
592 * spurious exception.
593 */
594 for (i = 0; i < nr_wp_slots(); i++) {
595 if (!info[i])
596 continue;
597
598 hw_end_addr = ALIGN(info[i]->address + info[i]->len, HW_BREAKPOINT_SIZE);
599
600 /*
601 * Ending address of DAWR range is less than starting
602 * address of op.
603 */
604 if ((hw_end_addr - 1) >= ea)
605 continue;
606
607 /*
608 * Those addresses need to be in the same or in two
609 * consecutive 512B blocks;
610 */
611 if (((hw_end_addr - 1) >> 10) != (ea >> 10))
612 continue;
613
614 /*
615 * 'op address + 64B' generates an address that has a
616 * carry into bit 52 (crosses 2K boundary).
617 */
618 if ((ea & 0x800) == ((ea + 64) & 0x800))
619 continue;
620
621 break;
622 }
623
624 if (i == nr_wp_slots())
625 return;
626
627 for (i = 0; i < nr_wp_slots(); i++) {
628 if (info[i]) {
629 hit[i] = 1;
630 info[i]->type |= HW_BRK_TYPE_EXTRANEOUS_IRQ;
631 }
632 }
633}
634
635int hw_breakpoint_handler(struct die_args *args)
636{
637 bool err = false;
638 int rc = NOTIFY_STOP;
639 struct perf_event *bp[HBP_NUM_MAX] = { NULL };
640 struct pt_regs *regs = args->regs;
641 struct arch_hw_breakpoint *info[HBP_NUM_MAX] = { NULL };
642 int i;
643 int hit[HBP_NUM_MAX] = {0};
644 int nr_hit = 0;
645 bool ptrace_bp = false;
646 ppc_inst_t instr = ppc_inst(0);
647 int type = 0;
648 int size = 0;
649 unsigned long ea = 0;
650
651 /* Disable breakpoints during exception handling */
652 hw_breakpoint_disable();
653
654 /*
655 * The counter may be concurrently released but that can only
656 * occur from a call_rcu() path. We can then safely fetch
657 * the breakpoint, use its callback, touch its counter
658 * while we are in an rcu_read_lock() path.
659 */
660 rcu_read_lock();
661
662 if (!IS_ENABLED(CONFIG_PPC_8xx))
663 wp_get_instr_detail(regs, &instr, &type, &size, &ea);
664
665 for (i = 0; i < nr_wp_slots(); i++) {
666 bp[i] = __this_cpu_read(bp_per_reg[i]);
667 if (!bp[i])
668 continue;
669
670 info[i] = counter_arch_bp(bp[i]);
671 info[i]->type &= ~HW_BRK_TYPE_EXTRANEOUS_IRQ;
672
673 if (wp_check_constraints(regs, instr, ea, type, size, info[i])) {
674 if (!IS_ENABLED(CONFIG_PPC_8xx) &&
675 ppc_inst_equal(instr, ppc_inst(0))) {
676 handler_error(bp[i], info[i]);
677 info[i] = NULL;
678 err = 1;
679 continue;
680 }
681
682 if (is_ptrace_bp(bp[i]))
683 ptrace_bp = true;
684 hit[i] = 1;
685 nr_hit++;
686 }
687 }
688
689 if (err)
690 goto reset;
691
692 if (!nr_hit) {
693 /* Workaround for Power10 DD1 */
694 if (!IS_ENABLED(CONFIG_PPC_8xx) && mfspr(SPRN_PVR) == 0x800100 &&
695 is_octword_vsx_instr(type, size)) {
696 handle_p10dd1_spurious_exception(info, hit, ea);
697 } else {
698 rc = NOTIFY_DONE;
699 goto out;
700 }
701 }
702
703 /*
704 * Return early after invoking user-callback function without restoring
705 * DABR if the breakpoint is from ptrace which always operates in
706 * one-shot mode. The ptrace-ed process will receive the SIGTRAP signal
707 * generated in do_dabr().
708 */
709 if (ptrace_bp) {
710 for (i = 0; i < nr_wp_slots(); i++) {
711 if (!hit[i])
712 continue;
713 perf_bp_event(bp[i], regs);
714 info[i] = NULL;
715 }
716 rc = NOTIFY_DONE;
717 goto reset;
718 }
719
720 if (!IS_ENABLED(CONFIG_PPC_8xx)) {
721 if (is_larx_stcx_instr(type)) {
722 for (i = 0; i < nr_wp_slots(); i++) {
723 if (!hit[i])
724 continue;
725 larx_stcx_err(bp[i], info[i]);
726 info[i] = NULL;
727 }
728 goto reset;
729 }
730
731 if (!stepping_handler(regs, bp, info, hit, instr))
732 goto reset;
733 }
734
735 /*
736 * As a policy, the callback is invoked in a 'trigger-after-execute'
737 * fashion
738 */
739 for (i = 0; i < nr_wp_slots(); i++) {
740 if (!hit[i])
741 continue;
742 if (!(info[i]->type & HW_BRK_TYPE_EXTRANEOUS_IRQ))
743 perf_bp_event(bp[i], regs);
744 }
745
746reset:
747 for (i = 0; i < nr_wp_slots(); i++) {
748 if (!info[i])
749 continue;
750 __set_breakpoint(i, info[i]);
751 }
752
753out:
754 rcu_read_unlock();
755 return rc;
756}
757NOKPROBE_SYMBOL(hw_breakpoint_handler);
758
759/*
760 * Handle single-step exceptions following a DABR hit.
761 */
762static int single_step_dabr_instruction(struct die_args *args)
763{
764 struct pt_regs *regs = args->regs;
765 struct perf_event *bp = NULL;
766 struct arch_hw_breakpoint *info;
767 int i;
768 bool found = false;
769
770 /*
771 * Check if we are single-stepping as a result of a
772 * previous HW Breakpoint exception
773 */
774 for (i = 0; i < nr_wp_slots(); i++) {
775 bp = current->thread.last_hit_ubp[i];
776
777 if (!bp)
778 continue;
779
780 found = true;
781 info = counter_arch_bp(bp);
782
783 /*
784 * We shall invoke the user-defined callback function in the
785 * single stepping handler to confirm to 'trigger-after-execute'
786 * semantics
787 */
788 if (!(info->type & HW_BRK_TYPE_EXTRANEOUS_IRQ))
789 perf_bp_event(bp, regs);
790 current->thread.last_hit_ubp[i] = NULL;
791 }
792
793 if (!found)
794 return NOTIFY_DONE;
795
796 for (i = 0; i < nr_wp_slots(); i++) {
797 bp = __this_cpu_read(bp_per_reg[i]);
798 if (!bp)
799 continue;
800
801 info = counter_arch_bp(bp);
802 __set_breakpoint(i, info);
803 }
804
805 /*
806 * If the process was being single-stepped by ptrace, let the
807 * other single-step actions occur (e.g. generate SIGTRAP).
808 */
809 if (test_thread_flag(TIF_SINGLESTEP))
810 return NOTIFY_DONE;
811
812 return NOTIFY_STOP;
813}
814NOKPROBE_SYMBOL(single_step_dabr_instruction);
815
816/*
817 * Handle debug exception notifications.
818 */
819int hw_breakpoint_exceptions_notify(
820 struct notifier_block *unused, unsigned long val, void *data)
821{
822 int ret = NOTIFY_DONE;
823
824 switch (val) {
825 case DIE_DABR_MATCH:
826 ret = hw_breakpoint_handler(data);
827 break;
828 case DIE_SSTEP:
829 ret = single_step_dabr_instruction(data);
830 break;
831 }
832
833 return ret;
834}
835NOKPROBE_SYMBOL(hw_breakpoint_exceptions_notify);
836
837/*
838 * Release the user breakpoints used by ptrace
839 */
840void flush_ptrace_hw_breakpoint(struct task_struct *tsk)
841{
842 int i;
843 struct thread_struct *t = &tsk->thread;
844
845 for (i = 0; i < nr_wp_slots(); i++) {
846 unregister_hw_breakpoint(t->ptrace_bps[i]);
847 t->ptrace_bps[i] = NULL;
848 }
849}
850
851void hw_breakpoint_pmu_read(struct perf_event *bp)
852{
853 /* TODO */
854}
855
856void ptrace_triggered(struct perf_event *bp,
857 struct perf_sample_data *data, struct pt_regs *regs)
858{
859 struct perf_event_attr attr;
860
861 /*
862 * Disable the breakpoint request here since ptrace has defined a
863 * one-shot behaviour for breakpoint exceptions in PPC64.
864 * The SIGTRAP signal is generated automatically for us in do_dabr().
865 * We don't have to do anything about that here
866 */
867 attr = bp->attr;
868 attr.disabled = true;
869 modify_user_hw_breakpoint(bp, &attr);
870}
1/*
2 * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
3 * using the CPU's debug registers. Derived from
4 * "arch/x86/kernel/hw_breakpoint.c"
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
19 *
20 * Copyright 2010 IBM Corporation
21 * Author: K.Prasad <prasad@linux.vnet.ibm.com>
22 *
23 */
24
25#include <linux/hw_breakpoint.h>
26#include <linux/notifier.h>
27#include <linux/kprobes.h>
28#include <linux/percpu.h>
29#include <linux/kernel.h>
30#include <linux/sched.h>
31#include <linux/smp.h>
32
33#include <asm/hw_breakpoint.h>
34#include <asm/processor.h>
35#include <asm/sstep.h>
36#include <asm/uaccess.h>
37
38/*
39 * Stores the breakpoints currently in use on each breakpoint address
40 * register for every cpu
41 */
42static DEFINE_PER_CPU(struct perf_event *, bp_per_reg);
43
44/*
45 * Returns total number of data or instruction breakpoints available.
46 */
47int hw_breakpoint_slots(int type)
48{
49 if (type == TYPE_DATA)
50 return HBP_NUM;
51 return 0; /* no instruction breakpoints available */
52}
53
54/*
55 * Install a perf counter breakpoint.
56 *
57 * We seek a free debug address register and use it for this
58 * breakpoint.
59 *
60 * Atomic: we hold the counter->ctx->lock and we only handle variables
61 * and registers local to this cpu.
62 */
63int arch_install_hw_breakpoint(struct perf_event *bp)
64{
65 struct arch_hw_breakpoint *info = counter_arch_bp(bp);
66 struct perf_event **slot = &__get_cpu_var(bp_per_reg);
67
68 *slot = bp;
69
70 /*
71 * Do not install DABR values if the instruction must be single-stepped.
72 * If so, DABR will be populated in single_step_dabr_instruction().
73 */
74 if (current->thread.last_hit_ubp != bp)
75 set_breakpoint(info);
76
77 return 0;
78}
79
80/*
81 * Uninstall the breakpoint contained in the given counter.
82 *
83 * First we search the debug address register it uses and then we disable
84 * it.
85 *
86 * Atomic: we hold the counter->ctx->lock and we only handle variables
87 * and registers local to this cpu.
88 */
89void arch_uninstall_hw_breakpoint(struct perf_event *bp)
90{
91 struct perf_event **slot = &__get_cpu_var(bp_per_reg);
92
93 if (*slot != bp) {
94 WARN_ONCE(1, "Can't find the breakpoint");
95 return;
96 }
97
98 *slot = NULL;
99 hw_breakpoint_disable();
100}
101
102/*
103 * Perform cleanup of arch-specific counters during unregistration
104 * of the perf-event
105 */
106void arch_unregister_hw_breakpoint(struct perf_event *bp)
107{
108 /*
109 * If the breakpoint is unregistered between a hw_breakpoint_handler()
110 * and the single_step_dabr_instruction(), then cleanup the breakpoint
111 * restoration variables to prevent dangling pointers.
112 */
113 if (bp->ctx && bp->ctx->task)
114 bp->ctx->task->thread.last_hit_ubp = NULL;
115}
116
117/*
118 * Check for virtual address in kernel space.
119 */
120int arch_check_bp_in_kernelspace(struct perf_event *bp)
121{
122 struct arch_hw_breakpoint *info = counter_arch_bp(bp);
123
124 return is_kernel_addr(info->address);
125}
126
127int arch_bp_generic_fields(int type, int *gen_bp_type)
128{
129 *gen_bp_type = 0;
130 if (type & HW_BRK_TYPE_READ)
131 *gen_bp_type |= HW_BREAKPOINT_R;
132 if (type & HW_BRK_TYPE_WRITE)
133 *gen_bp_type |= HW_BREAKPOINT_W;
134 if (*gen_bp_type == 0)
135 return -EINVAL;
136 return 0;
137}
138
139/*
140 * Validate the arch-specific HW Breakpoint register settings
141 */
142int arch_validate_hwbkpt_settings(struct perf_event *bp)
143{
144 int ret = -EINVAL, length_max;
145 struct arch_hw_breakpoint *info = counter_arch_bp(bp);
146
147 if (!bp)
148 return ret;
149
150 info->type = HW_BRK_TYPE_TRANSLATE;
151 if (bp->attr.bp_type & HW_BREAKPOINT_R)
152 info->type |= HW_BRK_TYPE_READ;
153 if (bp->attr.bp_type & HW_BREAKPOINT_W)
154 info->type |= HW_BRK_TYPE_WRITE;
155 if (info->type == HW_BRK_TYPE_TRANSLATE)
156 /* must set alteast read or write */
157 return ret;
158 if (!(bp->attr.exclude_user))
159 info->type |= HW_BRK_TYPE_USER;
160 if (!(bp->attr.exclude_kernel))
161 info->type |= HW_BRK_TYPE_KERNEL;
162 if (!(bp->attr.exclude_hv))
163 info->type |= HW_BRK_TYPE_HYP;
164 info->address = bp->attr.bp_addr;
165 info->len = bp->attr.bp_len;
166
167 /*
168 * Since breakpoint length can be a maximum of HW_BREAKPOINT_LEN(8)
169 * and breakpoint addresses are aligned to nearest double-word
170 * HW_BREAKPOINT_ALIGN by rounding off to the lower address, the
171 * 'symbolsize' should satisfy the check below.
172 */
173 length_max = 8; /* DABR */
174 if (cpu_has_feature(CPU_FTR_DAWR)) {
175 length_max = 512 ; /* 64 doublewords */
176 /* DAWR region can't cross 512 boundary */
177 if ((bp->attr.bp_addr >> 10) !=
178 ((bp->attr.bp_addr + bp->attr.bp_len - 1) >> 10))
179 return -EINVAL;
180 }
181 if (info->len >
182 (length_max - (info->address & HW_BREAKPOINT_ALIGN)))
183 return -EINVAL;
184 return 0;
185}
186
187/*
188 * Restores the breakpoint on the debug registers.
189 * Invoke this function if it is known that the execution context is
190 * about to change to cause loss of MSR_SE settings.
191 */
192void thread_change_pc(struct task_struct *tsk, struct pt_regs *regs)
193{
194 struct arch_hw_breakpoint *info;
195
196 if (likely(!tsk->thread.last_hit_ubp))
197 return;
198
199 info = counter_arch_bp(tsk->thread.last_hit_ubp);
200 regs->msr &= ~MSR_SE;
201 set_breakpoint(info);
202 tsk->thread.last_hit_ubp = NULL;
203}
204
205/*
206 * Handle debug exception notifications.
207 */
208int __kprobes hw_breakpoint_handler(struct die_args *args)
209{
210 int rc = NOTIFY_STOP;
211 struct perf_event *bp;
212 struct pt_regs *regs = args->regs;
213 int stepped = 1;
214 struct arch_hw_breakpoint *info;
215 unsigned int instr;
216 unsigned long dar = regs->dar;
217
218 /* Disable breakpoints during exception handling */
219 hw_breakpoint_disable();
220
221 /*
222 * The counter may be concurrently released but that can only
223 * occur from a call_rcu() path. We can then safely fetch
224 * the breakpoint, use its callback, touch its counter
225 * while we are in an rcu_read_lock() path.
226 */
227 rcu_read_lock();
228
229 bp = __get_cpu_var(bp_per_reg);
230 if (!bp)
231 goto out;
232 info = counter_arch_bp(bp);
233
234 /*
235 * Return early after invoking user-callback function without restoring
236 * DABR if the breakpoint is from ptrace which always operates in
237 * one-shot mode. The ptrace-ed process will receive the SIGTRAP signal
238 * generated in do_dabr().
239 */
240 if (bp->overflow_handler == ptrace_triggered) {
241 perf_bp_event(bp, regs);
242 rc = NOTIFY_DONE;
243 goto out;
244 }
245
246 /*
247 * Verify if dar lies within the address range occupied by the symbol
248 * being watched to filter extraneous exceptions. If it doesn't,
249 * we still need to single-step the instruction, but we don't
250 * generate an event.
251 */
252 info->type &= ~HW_BRK_TYPE_EXTRANEOUS_IRQ;
253 if (!((bp->attr.bp_addr <= dar) &&
254 (dar - bp->attr.bp_addr < bp->attr.bp_len)))
255 info->type |= HW_BRK_TYPE_EXTRANEOUS_IRQ;
256
257 /* Do not emulate user-space instructions, instead single-step them */
258 if (user_mode(regs)) {
259 current->thread.last_hit_ubp = bp;
260 regs->msr |= MSR_SE;
261 goto out;
262 }
263
264 stepped = 0;
265 instr = 0;
266 if (!__get_user_inatomic(instr, (unsigned int *) regs->nip))
267 stepped = emulate_step(regs, instr);
268
269 /*
270 * emulate_step() could not execute it. We've failed in reliably
271 * handling the hw-breakpoint. Unregister it and throw a warning
272 * message to let the user know about it.
273 */
274 if (!stepped) {
275 WARN(1, "Unable to handle hardware breakpoint. Breakpoint at "
276 "0x%lx will be disabled.", info->address);
277 perf_event_disable(bp);
278 goto out;
279 }
280 /*
281 * As a policy, the callback is invoked in a 'trigger-after-execute'
282 * fashion
283 */
284 if (!(info->type & HW_BRK_TYPE_EXTRANEOUS_IRQ))
285 perf_bp_event(bp, regs);
286
287 set_breakpoint(info);
288out:
289 rcu_read_unlock();
290 return rc;
291}
292
293/*
294 * Handle single-step exceptions following a DABR hit.
295 */
296int __kprobes single_step_dabr_instruction(struct die_args *args)
297{
298 struct pt_regs *regs = args->regs;
299 struct perf_event *bp = NULL;
300 struct arch_hw_breakpoint *info;
301
302 bp = current->thread.last_hit_ubp;
303 /*
304 * Check if we are single-stepping as a result of a
305 * previous HW Breakpoint exception
306 */
307 if (!bp)
308 return NOTIFY_DONE;
309
310 info = counter_arch_bp(bp);
311
312 /*
313 * We shall invoke the user-defined callback function in the single
314 * stepping handler to confirm to 'trigger-after-execute' semantics
315 */
316 if (!(info->type & HW_BRK_TYPE_EXTRANEOUS_IRQ))
317 perf_bp_event(bp, regs);
318
319 set_breakpoint(info);
320 current->thread.last_hit_ubp = NULL;
321
322 /*
323 * If the process was being single-stepped by ptrace, let the
324 * other single-step actions occur (e.g. generate SIGTRAP).
325 */
326 if (test_thread_flag(TIF_SINGLESTEP))
327 return NOTIFY_DONE;
328
329 return NOTIFY_STOP;
330}
331
332/*
333 * Handle debug exception notifications.
334 */
335int __kprobes hw_breakpoint_exceptions_notify(
336 struct notifier_block *unused, unsigned long val, void *data)
337{
338 int ret = NOTIFY_DONE;
339
340 switch (val) {
341 case DIE_DABR_MATCH:
342 ret = hw_breakpoint_handler(data);
343 break;
344 case DIE_SSTEP:
345 ret = single_step_dabr_instruction(data);
346 break;
347 }
348
349 return ret;
350}
351
352/*
353 * Release the user breakpoints used by ptrace
354 */
355void flush_ptrace_hw_breakpoint(struct task_struct *tsk)
356{
357 struct thread_struct *t = &tsk->thread;
358
359 unregister_hw_breakpoint(t->ptrace_bps[0]);
360 t->ptrace_bps[0] = NULL;
361}
362
363void hw_breakpoint_pmu_read(struct perf_event *bp)
364{
365 /* TODO */
366}