[GIT PULL] kgdb regression fixes for 2.6.33

[GIT PULL] kgdb regression fixes for 2.6.33

[Jason Wessel]

Linus, please pull the kgdb git tree fixes for 2.6.33-rc5

  git://git.kernel.org/pub/scm/linux/kernel/git/jwessel/linux-2.6-kgdb.git for_linus

Summary:
  
  * Fix the kgdb regression in functionality caused by the new hw
    breakpoint API changes in 2.6.33.

This leaves us with some cleanup in 2.6.34, but all the previous
functionality works again.  In fact you can use hw breakpoints in user
space and the kernel at the same time now that the API is shared,
which is something that never previously worked.

Thanks,
Jason.

--
The following changes since commit be8cde8b24c9dca1e54598690115eee5b1476519:
  Linus Torvalds (1):
        Merge git://git.kernel.org/.../jejb/scsi-rc-fixes-2.6

--
Jason Wessel (2):
      x86,hw_breakpoint,kgdb: kgdb to use hw_breakpoint API
      perf,hw_breakpoint,kgdb: No mutex taken for kernel debugger

 arch/x86/kernel/kgdb.c        |  222 ++++++++++++++++++++++++++++++-----------
 include/linux/hw_breakpoint.h |    2 +
 kernel/hw_breakpoint.c        |   52 ++++++++--
 kernel/kgdb.c                 |    3 +
 4 files changed, 212 insertions(+), 67 deletions(-)

[PATCH 1/2] x86,hw_breakpoint,kgdb: kgdb to use hw_breakpoint API

[Jason Wessel]

In the 2.6.33 kernel, the hw_breakpoint API is now used for the
performance event counters.  The hw_breakpoint_handler() now consumes
the hw breakpoints that were previously set by kgdb arch specific
code.  In order for kgdb to work in conjunction with this core API
change, kgdb must use some of the low level functions of the
hw_breakpoint API to install, uninstall, and deal with hw breakpoint
reservations.

The kgdb core required a change to call kgdb_disable_hw_debug anytime
a slave cpu enters kgdb_wait() in order to keep all the hw breakpoints
in sync as well as to prevent hitting a hw breakpoint while kgdb is
active.

During the architecture specific initialization of kgdb, it will
pre-allocate 4 disabled (struct perf event **) structures.  Kgdb will
use these to manage the capabilities for the 4 hw breakpoint
registers, per cpu.  Right now the hw_breakpoint API does not have a
way to ask how many breakpoints are available, on each CPU so it is
possible that the install of a breakpoint might fail when kgdb
restores the system to the run state.  The intent of this patch is to
first get the basic functionality of hw breakpoints working and leave
it to the person debugging the kernel to understand what hw
breakpoints are in use and what restrictions have been imposed as a
result.  Breakpoint constraints will be dealt with in a future patch.

While atomic, the x86 specific kgdb code will call
arch_uninstall_hw_breakpoint() and arch_install_hw_breakpoint() to
manage the cpu specific hw breakpoints.

The net result of these changes allow kgdb to use the same pool of
hw_breakpoints that are used by the perf event API, but neither knows
about future reservations for the available hw breakpoint slots.

CC: Ingo Molnar <mingo@elte.hu>
CC: K.Prasad <prasad@linux.vnet.ibm.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
---
 arch/x86/kernel/kgdb.c |  171 ++++++++++++++++++++++++++++++++----------------
 kernel/kgdb.c          |    3 +
 2 files changed, 117 insertions(+), 57 deletions(-)

diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
index dd74fe7..62bea73 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -42,6 +42,7 @@
 #include <linux/init.h>
 #include <linux/smp.h>
 #include <linux/nmi.h>
+#include <linux/hw_breakpoint.h>
 
 #include <asm/debugreg.h>
 #include <asm/apicdef.h>
@@ -204,40 +205,38 @@ void gdb_regs_to_pt_regs(unsigned long *gdb_regs, struct pt_regs *regs)
 
 static struct hw_breakpoint {
 	unsigned		enabled;
-	unsigned		type;
-	unsigned		len;
 	unsigned long		addr;
+	int			len;
+	int			type;
+	struct perf_event	**pev;
 } breakinfo[4];
 
 static void kgdb_correct_hw_break(void)
 {
-	unsigned long dr7;
-	int correctit = 0;
-	int breakbit;
 	int breakno;
 
-	get_debugreg(dr7, 7);
 	for (breakno = 0; breakno < 4; breakno++) {
-		breakbit = 2 << (breakno << 1);
-		if (!(dr7 & breakbit) && breakinfo[breakno].enabled) {
-			correctit = 1;
-			dr7 |= breakbit;
-			dr7 &= ~(0xf0000 << (breakno << 2));
-			dr7 |= ((breakinfo[breakno].len << 2) |
-				 breakinfo[breakno].type) <<
-			       ((breakno << 2) + 16);
-			set_debugreg(breakinfo[breakno].addr, breakno);
-
-		} else {
-			if ((dr7 & breakbit) && !breakinfo[breakno].enabled) {
-				correctit = 1;
-				dr7 &= ~breakbit;
-				dr7 &= ~(0xf0000 << (breakno << 2));
-			}
-		}
+		struct perf_event *bp;
+		struct arch_hw_breakpoint *info;
+		int val;
+		int cpu = raw_smp_processor_id();
+		if (!breakinfo[breakno].enabled)
+			continue;
+		bp = *per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		info = counter_arch_bp(bp);
+		if (bp->attr.disabled != 1)
+			continue;
+		bp->attr.bp_addr = breakinfo[breakno].addr;
+		bp->attr.bp_len = breakinfo[breakno].len;
+		bp->attr.bp_type = breakinfo[breakno].type;
+		info->address = breakinfo[breakno].addr;
+		info->len = breakinfo[breakno].len;
+		info->type = breakinfo[breakno].type;
+		val = arch_install_hw_breakpoint(bp);
+		if (!val)
+			bp->attr.disabled = 0;
 	}
-	if (correctit)
-		set_debugreg(dr7, 7);
+	hw_breakpoint_restore();
 }
 
 static int
@@ -259,15 +258,23 @@ kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 static void kgdb_remove_all_hw_break(void)
 {
 	int i;
+	int cpu = raw_smp_processor_id();
+	struct perf_event *bp;
 
-	for (i = 0; i < 4; i++)
-		memset(&breakinfo[i], 0, sizeof(struct hw_breakpoint));
+	for (i = 0; i < 4; i++) {
+		if (!breakinfo[i].enabled)
+			continue;
+		bp = *per_cpu_ptr(breakinfo[i].pev, cpu);
+		if (bp->attr.disabled == 1)
+			continue;
+		arch_uninstall_hw_breakpoint(bp);
+		bp->attr.disabled = 1;
+	}
 }
 
 static int
 kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 {
-	unsigned type;
 	int i;
 
 	for (i = 0; i < 4; i++)
@@ -278,27 +285,38 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 
 	switch (bptype) {
 	case BP_HARDWARE_BREAKPOINT:
-		type = 0;
-		len  = 1;
+		len = 1;
+		breakinfo[i].type = X86_BREAKPOINT_EXECUTE;
 		break;
 	case BP_WRITE_WATCHPOINT:
-		type = 1;
+		breakinfo[i].type = X86_BREAKPOINT_WRITE;
 		break;
 	case BP_ACCESS_WATCHPOINT:
-		type = 3;
+		breakinfo[i].type = X86_BREAKPOINT_RW;
 		break;
 	default:
 		return -1;
 	}
-
-	if (len == 1 || len == 2 || len == 4)
-		breakinfo[i].len  = len - 1;
-	else
+	switch (len) {
+	case 1:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_1;
+		break;
+	case 2:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_2;
+		break;
+	case 4:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_4;
+		break;
+#ifdef CONFIG_X86_64
+	case 8:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_8;
+		break;
+#endif
+	default:
 		return -1;
-
-	breakinfo[i].enabled = 1;
+	}
 	breakinfo[i].addr = addr;
-	breakinfo[i].type = type;
+	breakinfo[i].enabled = 1;
 
 	return 0;
 }
@@ -313,8 +331,21 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
  */
 void kgdb_disable_hw_debug(struct pt_regs *regs)
 {
+	int i;
+	int cpu = raw_smp_processor_id();
+	struct perf_event *bp;
+
 	/* Disable hardware debugging while we are in kgdb: */
 	set_debugreg(0UL, 7);
+	for (i = 0; i < 4; i++) {
+		if (!breakinfo[i].enabled)
+			continue;
+		bp = *per_cpu_ptr(breakinfo[i].pev, cpu);
+		if (bp->attr.disabled == 1)
+			continue;
+		arch_uninstall_hw_breakpoint(bp);
+		bp->attr.disabled = 1;
+	}
 }
 
 /**
@@ -378,7 +409,6 @@ int kgdb_arch_handle_exception(int e_vector, int signo, int err_code,
 			       struct pt_regs *linux_regs)
 {
 	unsigned long addr;
-	unsigned long dr6;
 	char *ptr;
 	int newPC;
 
@@ -404,20 +434,6 @@ int kgdb_arch_handle_exception(int e_vector, int signo, int err_code,
 				   raw_smp_processor_id());
 		}
 
-		get_debugreg(dr6, 6);
-		if (!(dr6 & 0x4000)) {
-			int breakno;
-
-			for (breakno = 0; breakno < 4; breakno++) {
-				if (dr6 & (1 << breakno) &&
-				    breakinfo[breakno].type == 0) {
-					/* Set restore flag: */
-					linux_regs->flags |= X86_EFLAGS_RF;
-					break;
-				}
-			}
-		}
-		set_debugreg(0UL, 6);
 		kgdb_correct_hw_break();
 
 		return 0;
@@ -485,8 +501,7 @@ static int __kgdb_notify(struct die_args *args, unsigned long cmd)
 		break;
 
 	case DIE_DEBUG:
-		if (atomic_read(&kgdb_cpu_doing_single_step) ==
-		    raw_smp_processor_id()) {
+		if (atomic_read(&kgdb_cpu_doing_single_step) != -1) {
 			if (user_mode(regs))
 				return single_step_cont(regs, args);
 			break;
@@ -539,7 +554,42 @@ static struct notifier_block kgdb_notifier = {
  */
 int kgdb_arch_init(void)
 {
-	return register_die_notifier(&kgdb_notifier);
+	int i, cpu;
+	int ret;
+	struct perf_event_attr attr;
+	struct perf_event **pevent;
+
+	ret = register_die_notifier(&kgdb_notifier);
+	if (ret != 0)
+		return ret;
+	/*
+	 * Pre-allocate the hw breakpoint structions in the non-atomic
+	 * portion of kgdb because this operation requires mutexs to
+	 * complete.
+	 */
+	attr.bp_addr = (unsigned long)kgdb_arch_init;
+	attr.type = PERF_TYPE_BREAKPOINT;
+	attr.bp_len = HW_BREAKPOINT_LEN_1;
+	attr.bp_type = HW_BREAKPOINT_W;
+	attr.disabled = 1;
+	for (i = 0; i < 4; i++) {
+		breakinfo[i].pev = register_wide_hw_breakpoint(&attr, NULL);
+		if (IS_ERR(breakinfo[i].pev)) {
+			printk(KERN_ERR "kgdb: Could not allocate hw breakpoints\n");
+			breakinfo[i].pev = NULL;
+			kgdb_arch_exit();
+			return -1;
+		}
+		for_each_online_cpu(cpu) {
+			pevent = per_cpu_ptr(breakinfo[i].pev, cpu);
+			pevent[0]->hw.sample_period = 1;
+			if (pevent[0]->destroy != NULL) {
+				pevent[0]->destroy = NULL;
+				release_bp_slot(*pevent);
+			}
+		}
+	}
+	return ret;
 }
 
 /**
@@ -550,6 +600,13 @@ int kgdb_arch_init(void)
  */
 void kgdb_arch_exit(void)
 {
+	int i;
+	for (i = 0; i < 4; i++) {
+		if (breakinfo[i].pev) {
+			unregister_wide_hw_breakpoint(breakinfo[i].pev);
+			breakinfo[i].pev = NULL;
+		}
+	}
 	unregister_die_notifier(&kgdb_notifier);
 }
 
diff --git a/kernel/kgdb.c b/kernel/kgdb.c
index 2eb517e..c7ade62 100644
--- a/kernel/kgdb.c
+++ b/kernel/kgdb.c
@@ -583,6 +583,9 @@ static void kgdb_wait(struct pt_regs *regs)
 	smp_wmb();
 	atomic_set(&cpu_in_kgdb[cpu], 1);
 
+	/* Disable any cpu specific hw breakpoints */
+	kgdb_disable_hw_debug(regs);
+
 	/* Wait till primary CPU is done with debugging */
 	while (atomic_read(&passive_cpu_wait[cpu]))
 		cpu_relax();
-- 
1.6.4.rc1

[PATCH 2/2] perf,hw_breakpoint,kgdb: No mutex taken for kernel debugger

[Jason Wessel]

This patch fixes the regression in functionality where the kernel
debugger and the perf API do not nicely share hw breakpoint
reservations.

The kernel debugger cannot use any mutex_lock() calls because it can
start the kernel running from an invalid context.

A mutex free version of the reservation API needed to get created for
the kernel debugger to safely update hw breakpoint reservations.

The possibility for a breakpoint reservation to be concurrently
processed at the time that kgdb interrupts the system is improbable.
Should this corner case occur the end user is warned, and the kernel
debugger will prohibit updating the hardware breakpoint reservations.

Any time the kernel debugger reserves a hardware breakpoint it will be
a system wide reservation.

CC: Ingo Molnar <mingo@elte.hu>
CC: K.Prasad <prasad@linux.vnet.ibm.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
---
 arch/x86/kernel/kgdb.c        |   51 ++++++++++++++++++++++++++++++++++++++++
 include/linux/hw_breakpoint.h |    2 +
 kernel/hw_breakpoint.c        |   52 +++++++++++++++++++++++++++++++++--------
 3 files changed, 95 insertions(+), 10 deletions(-)

diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
index 62bea73..bfba601 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -239,6 +239,49 @@ static void kgdb_correct_hw_break(void)
 	hw_breakpoint_restore();
 }
 
+static int hw_break_reserve_slot(int breakno)
+{
+	int cpu;
+	int cnt = 0;
+	struct perf_event **pevent;
+
+	for_each_online_cpu(cpu) {
+		cnt++;
+		pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		if (dbg_reserve_bp_slot(*pevent))
+			goto fail;
+	}
+
+	return 0;
+
+fail:
+	for_each_online_cpu(cpu) {
+		cnt--;
+		if (!cnt)
+			break;
+		pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		dbg_release_bp_slot(*pevent);
+	}
+	return -1;
+}
+
+static int hw_break_release_slot(int breakno)
+{
+	struct perf_event **pevent;
+	int cpu;
+
+	for_each_online_cpu(cpu) {
+		pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		if (dbg_release_bp_slot(*pevent))
+			/*
+			 * The debugger is responisble for handing the retry on
+			 * remove failure.
+			 */
+			return -1;
+	}
+	return 0;
+}
+
 static int
 kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 {
@@ -250,6 +293,10 @@ kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 	if (i == 4)
 		return -1;
 
+	if (hw_break_release_slot(i)) {
+		printk(KERN_ERR "Cannot remove hw breakpoint at %lx\n", addr);
+		return -1;
+	}
 	breakinfo[i].enabled = 0;
 
 	return 0;
@@ -316,6 +363,10 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 		return -1;
 	}
 	breakinfo[i].addr = addr;
+	if (hw_break_reserve_slot(i)) {
+		breakinfo[i].addr = 0;
+		return -1;
+	}
 	breakinfo[i].enabled = 1;
 
 	return 0;
diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h
index 41235c9..070ba06 100644
--- a/include/linux/hw_breakpoint.h
+++ b/include/linux/hw_breakpoint.h
@@ -75,6 +75,8 @@ extern int __register_perf_hw_breakpoint(struct perf_event *bp);
 extern void unregister_hw_breakpoint(struct perf_event *bp);
 extern void unregister_wide_hw_breakpoint(struct perf_event **cpu_events);
 
+extern int dbg_reserve_bp_slot(struct perf_event *bp);
+extern int dbg_release_bp_slot(struct perf_event *bp);
 extern int reserve_bp_slot(struct perf_event *bp);
 extern void release_bp_slot(struct perf_event *bp);
 
diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c
index 50dbd59..e5aa380 100644
--- a/kernel/hw_breakpoint.c
+++ b/kernel/hw_breakpoint.c
@@ -243,38 +243,70 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable)
  *       ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *))
  *            + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM
  */
-int reserve_bp_slot(struct perf_event *bp)
+static int __reserve_bp_slot(struct perf_event *bp)
 {
 	struct bp_busy_slots slots = {0};
-	int ret = 0;
-
-	mutex_lock(&nr_bp_mutex);
 
 	fetch_bp_busy_slots(&slots, bp);
 
 	/* Flexible counters need to keep at least one slot */
-	if (slots.pinned + (!!slots.flexible) == HBP_NUM) {
-		ret = -ENOSPC;
-		goto end;
-	}
+	if (slots.pinned + (!!slots.flexible) == HBP_NUM)
+		return -ENOSPC;
 
 	toggle_bp_slot(bp, true);
 
-end:
+	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)
+{
+	toggle_bp_slot(bp, false);
+}
+
 void release_bp_slot(struct perf_event *bp)
 {
 	mutex_lock(&nr_bp_mutex);
 
-	toggle_bp_slot(bp, false);
+	__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;
+}
 
 int register_perf_hw_breakpoint(struct perf_event *bp)
 {
-- 
1.6.4.rc1

[tip:perf/urgent] perf, hw_breakpoint, kgdb: Do not take mutex for kernel debugger

[tip-bot for Jason Wessel]

Commit-ID:  5352ae638e2d7d5c9b2e4d528676bbf2af6fd6f3
Gitweb:     http://git.kernel.org/tip/5352ae638e2d7d5c9b2e4d528676bbf2af6fd6f3
Author:     Jason Wessel <jason.wessel@windriver.com>
AuthorDate: Thu, 28 Jan 2010 17:04:43 -0600
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Sat, 30 Jan 2010 08:42:21 +0100

perf, hw_breakpoint, kgdb: Do not take mutex for kernel debugger

This patch fixes the regression in functionality where the
kernel debugger and the perf API do not nicely share hw
breakpoint reservations.

The kernel debugger cannot use any mutex_lock() calls because it
can start the kernel running from an invalid context.

A mutex free version of the reservation API needed to get
created for the kernel debugger to safely update hw breakpoint
reservations.

The possibility for a breakpoint reservation to be concurrently
processed at the time that kgdb interrupts the system is
improbable. Should this corner case occur the end user is
warned, and the kernel debugger will prohibit updating the
hardware breakpoint reservations.

Any time the kernel debugger reserves a hardware breakpoint it
will be a system wide reservation.

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: kgdb-bugreport@lists.sourceforge.net
Cc: K.Prasad <prasad@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: torvalds@linux-foundation.org
LKML-Reference: <1264719883-7285-3-git-send-email-jason.wessel@windriver.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 arch/x86/kernel/kgdb.c        |   51 ++++++++++++++++++++++++++++++++++++++++
 include/linux/hw_breakpoint.h |    2 +
 kernel/hw_breakpoint.c        |   52 +++++++++++++++++++++++++++++++++--------
 3 files changed, 95 insertions(+), 10 deletions(-)

diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
index 62bea73..bfba601 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -239,6 +239,49 @@ static void kgdb_correct_hw_break(void)
 	hw_breakpoint_restore();
 }
 
+static int hw_break_reserve_slot(int breakno)
+{
+	int cpu;
+	int cnt = 0;
+	struct perf_event **pevent;
+
+	for_each_online_cpu(cpu) {
+		cnt++;
+		pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		if (dbg_reserve_bp_slot(*pevent))
+			goto fail;
+	}
+
+	return 0;
+
+fail:
+	for_each_online_cpu(cpu) {
+		cnt--;
+		if (!cnt)
+			break;
+		pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		dbg_release_bp_slot(*pevent);
+	}
+	return -1;
+}
+
+static int hw_break_release_slot(int breakno)
+{
+	struct perf_event **pevent;
+	int cpu;
+
+	for_each_online_cpu(cpu) {
+		pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		if (dbg_release_bp_slot(*pevent))
+			/*
+			 * The debugger is responisble for handing the retry on
+			 * remove failure.
+			 */
+			return -1;
+	}
+	return 0;
+}
+
 static int
 kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 {
@@ -250,6 +293,10 @@ kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 	if (i == 4)
 		return -1;
 
+	if (hw_break_release_slot(i)) {
+		printk(KERN_ERR "Cannot remove hw breakpoint at %lx\n", addr);
+		return -1;
+	}
 	breakinfo[i].enabled = 0;
 
 	return 0;
@@ -316,6 +363,10 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 		return -1;
 	}
 	breakinfo[i].addr = addr;
+	if (hw_break_reserve_slot(i)) {
+		breakinfo[i].addr = 0;
+		return -1;
+	}
 	breakinfo[i].enabled = 1;
 
 	return 0;
diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h
index 41235c9..070ba06 100644
--- a/include/linux/hw_breakpoint.h
+++ b/include/linux/hw_breakpoint.h
@@ -75,6 +75,8 @@ extern int __register_perf_hw_breakpoint(struct perf_event *bp);
 extern void unregister_hw_breakpoint(struct perf_event *bp);
 extern void unregister_wide_hw_breakpoint(struct perf_event **cpu_events);
 
+extern int dbg_reserve_bp_slot(struct perf_event *bp);
+extern int dbg_release_bp_slot(struct perf_event *bp);
 extern int reserve_bp_slot(struct perf_event *bp);
 extern void release_bp_slot(struct perf_event *bp);
 
diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c
index c030ae6..8a5c7d5 100644
--- a/kernel/hw_breakpoint.c
+++ b/kernel/hw_breakpoint.c
@@ -243,38 +243,70 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable)
  *       ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *))
  *            + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM
  */
-int reserve_bp_slot(struct perf_event *bp)
+static int __reserve_bp_slot(struct perf_event *bp)
 {
 	struct bp_busy_slots slots = {0};
-	int ret = 0;
-
-	mutex_lock(&nr_bp_mutex);
 
 	fetch_bp_busy_slots(&slots, bp);
 
 	/* Flexible counters need to keep at least one slot */
-	if (slots.pinned + (!!slots.flexible) == HBP_NUM) {
-		ret = -ENOSPC;
-		goto end;
-	}
+	if (slots.pinned + (!!slots.flexible) == HBP_NUM)
+		return -ENOSPC;
 
 	toggle_bp_slot(bp, true);
 
-end:
+	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)
+{
+	toggle_bp_slot(bp, false);
+}
+
 void release_bp_slot(struct perf_event *bp)
 {
 	mutex_lock(&nr_bp_mutex);
 
-	toggle_bp_slot(bp, false);
+	__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;
+}
 
 int register_perf_hw_breakpoint(struct perf_event *bp)
 {

[tip:perf/urgent] x86, hw_breakpoints, kgdb: Fix kgdb to use hw_breakpoint API

[tip-bot for Jason Wessel]

Commit-ID:  cc0967490c1c3824bc5b75718b6ca8a51d9f2617
Gitweb:     http://git.kernel.org/tip/cc0967490c1c3824bc5b75718b6ca8a51d9f2617
Author:     Jason Wessel <jason.wessel@windriver.com>
AuthorDate: Thu, 28 Jan 2010 17:04:42 -0600
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Sat, 30 Jan 2010 08:42:20 +0100

x86, hw_breakpoints, kgdb: Fix kgdb to use hw_breakpoint API

In the 2.6.33 kernel, the hw_breakpoint API is now used for the
performance event counters.  The hw_breakpoint_handler() now
consumes the hw breakpoints that were previously set by kgdb
arch specific code.  In order for kgdb to work in conjunction
with this core API change, kgdb must use some of the low level
functions of the hw_breakpoint API to install, uninstall, and
deal with hw breakpoint reservations.

The kgdb core required a change to call kgdb_disable_hw_debug
anytime a slave cpu enters kgdb_wait() in order to keep all the
hw breakpoints in sync as well as to prevent hitting a hw
breakpoint while kgdb is active.

During the architecture specific initialization of kgdb, it will
pre-allocate 4 disabled (struct perf event **) structures.  Kgdb
will use these to manage the capabilities for the 4 hw
breakpoint registers, per cpu.  Right now the hw_breakpoint API
does not have a way to ask how many breakpoints are available,
on each CPU so it is possible that the install of a breakpoint
might fail when kgdb restores the system to the run state.  The
intent of this patch is to first get the basic functionality of
hw breakpoints working and leave it to the person debugging the
kernel to understand what hw breakpoints are in use and what
restrictions have been imposed as a result.  Breakpoint
constraints will be dealt with in a future patch.

While atomic, the x86 specific kgdb code will call
arch_uninstall_hw_breakpoint() and arch_install_hw_breakpoint()
to manage the cpu specific hw breakpoints.

The net result of these changes allow kgdb to use the same pool
of hw_breakpoints that are used by the perf event API, but
neither knows about future reservations for the available hw
breakpoint slots.

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: kgdb-bugreport@lists.sourceforge.net
Cc: K.Prasad <prasad@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: torvalds@linux-foundation.org
LKML-Reference: <1264719883-7285-2-git-send-email-jason.wessel@windriver.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 arch/x86/kernel/kgdb.c |  171 ++++++++++++++++++++++++++++++++----------------
 kernel/kgdb.c          |    3 +
 2 files changed, 117 insertions(+), 57 deletions(-)

diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c
index dd74fe7..62bea73 100644
--- a/arch/x86/kernel/kgdb.c
+++ b/arch/x86/kernel/kgdb.c
@@ -42,6 +42,7 @@
 #include <linux/init.h>
 #include <linux/smp.h>
 #include <linux/nmi.h>
+#include <linux/hw_breakpoint.h>
 
 #include <asm/debugreg.h>
 #include <asm/apicdef.h>
@@ -204,40 +205,38 @@ void gdb_regs_to_pt_regs(unsigned long *gdb_regs, struct pt_regs *regs)
 
 static struct hw_breakpoint {
 	unsigned		enabled;
-	unsigned		type;
-	unsigned		len;
 	unsigned long		addr;
+	int			len;
+	int			type;
+	struct perf_event	**pev;
 } breakinfo[4];
 
 static void kgdb_correct_hw_break(void)
 {
-	unsigned long dr7;
-	int correctit = 0;
-	int breakbit;
 	int breakno;
 
-	get_debugreg(dr7, 7);
 	for (breakno = 0; breakno < 4; breakno++) {
-		breakbit = 2 << (breakno << 1);
-		if (!(dr7 & breakbit) && breakinfo[breakno].enabled) {
-			correctit = 1;
-			dr7 |= breakbit;
-			dr7 &= ~(0xf0000 << (breakno << 2));
-			dr7 |= ((breakinfo[breakno].len << 2) |
-				 breakinfo[breakno].type) <<
-			       ((breakno << 2) + 16);
-			set_debugreg(breakinfo[breakno].addr, breakno);
-
-		} else {
-			if ((dr7 & breakbit) && !breakinfo[breakno].enabled) {
-				correctit = 1;
-				dr7 &= ~breakbit;
-				dr7 &= ~(0xf0000 << (breakno << 2));
-			}
-		}
+		struct perf_event *bp;
+		struct arch_hw_breakpoint *info;
+		int val;
+		int cpu = raw_smp_processor_id();
+		if (!breakinfo[breakno].enabled)
+			continue;
+		bp = *per_cpu_ptr(breakinfo[breakno].pev, cpu);
+		info = counter_arch_bp(bp);
+		if (bp->attr.disabled != 1)
+			continue;
+		bp->attr.bp_addr = breakinfo[breakno].addr;
+		bp->attr.bp_len = breakinfo[breakno].len;
+		bp->attr.bp_type = breakinfo[breakno].type;
+		info->address = breakinfo[breakno].addr;
+		info->len = breakinfo[breakno].len;
+		info->type = breakinfo[breakno].type;
+		val = arch_install_hw_breakpoint(bp);
+		if (!val)
+			bp->attr.disabled = 0;
 	}
-	if (correctit)
-		set_debugreg(dr7, 7);
+	hw_breakpoint_restore();
 }
 
 static int
@@ -259,15 +258,23 @@ kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 static void kgdb_remove_all_hw_break(void)
 {
 	int i;
+	int cpu = raw_smp_processor_id();
+	struct perf_event *bp;
 
-	for (i = 0; i < 4; i++)
-		memset(&breakinfo[i], 0, sizeof(struct hw_breakpoint));
+	for (i = 0; i < 4; i++) {
+		if (!breakinfo[i].enabled)
+			continue;
+		bp = *per_cpu_ptr(breakinfo[i].pev, cpu);
+		if (bp->attr.disabled == 1)
+			continue;
+		arch_uninstall_hw_breakpoint(bp);
+		bp->attr.disabled = 1;
+	}
 }
 
 static int
 kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 {
-	unsigned type;
 	int i;
 
 	for (i = 0; i < 4; i++)
@@ -278,27 +285,38 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
 
 	switch (bptype) {
 	case BP_HARDWARE_BREAKPOINT:
-		type = 0;
-		len  = 1;
+		len = 1;
+		breakinfo[i].type = X86_BREAKPOINT_EXECUTE;
 		break;
 	case BP_WRITE_WATCHPOINT:
-		type = 1;
+		breakinfo[i].type = X86_BREAKPOINT_WRITE;
 		break;
 	case BP_ACCESS_WATCHPOINT:
-		type = 3;
+		breakinfo[i].type = X86_BREAKPOINT_RW;
 		break;
 	default:
 		return -1;
 	}
-
-	if (len == 1 || len == 2 || len == 4)
-		breakinfo[i].len  = len - 1;
-	else
+	switch (len) {
+	case 1:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_1;
+		break;
+	case 2:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_2;
+		break;
+	case 4:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_4;
+		break;
+#ifdef CONFIG_X86_64
+	case 8:
+		breakinfo[i].len = X86_BREAKPOINT_LEN_8;
+		break;
+#endif
+	default:
 		return -1;
-
-	breakinfo[i].enabled = 1;
+	}
 	breakinfo[i].addr = addr;
-	breakinfo[i].type = type;
+	breakinfo[i].enabled = 1;
 
 	return 0;
 }
@@ -313,8 +331,21 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype)
  */
 void kgdb_disable_hw_debug(struct pt_regs *regs)
 {
+	int i;
+	int cpu = raw_smp_processor_id();
+	struct perf_event *bp;
+
 	/* Disable hardware debugging while we are in kgdb: */
 	set_debugreg(0UL, 7);
+	for (i = 0; i < 4; i++) {
+		if (!breakinfo[i].enabled)
+			continue;
+		bp = *per_cpu_ptr(breakinfo[i].pev, cpu);
+		if (bp->attr.disabled == 1)
+			continue;
+		arch_uninstall_hw_breakpoint(bp);
+		bp->attr.disabled = 1;
+	}
 }
 
 /**
@@ -378,7 +409,6 @@ int kgdb_arch_handle_exception(int e_vector, int signo, int err_code,
 			       struct pt_regs *linux_regs)
 {
 	unsigned long addr;
-	unsigned long dr6;
 	char *ptr;
 	int newPC;
 
@@ -404,20 +434,6 @@ int kgdb_arch_handle_exception(int e_vector, int signo, int err_code,
 				   raw_smp_processor_id());
 		}
 
-		get_debugreg(dr6, 6);
-		if (!(dr6 & 0x4000)) {
-			int breakno;
-
-			for (breakno = 0; breakno < 4; breakno++) {
-				if (dr6 & (1 << breakno) &&
-				    breakinfo[breakno].type == 0) {
-					/* Set restore flag: */
-					linux_regs->flags |= X86_EFLAGS_RF;
-					break;
-				}
-			}
-		}
-		set_debugreg(0UL, 6);
 		kgdb_correct_hw_break();
 
 		return 0;
@@ -485,8 +501,7 @@ static int __kgdb_notify(struct die_args *args, unsigned long cmd)
 		break;
 
 	case DIE_DEBUG:
-		if (atomic_read(&kgdb_cpu_doing_single_step) ==
-		    raw_smp_processor_id()) {
+		if (atomic_read(&kgdb_cpu_doing_single_step) != -1) {
 			if (user_mode(regs))
 				return single_step_cont(regs, args);
 			break;
@@ -539,7 +554,42 @@ static struct notifier_block kgdb_notifier = {
  */
 int kgdb_arch_init(void)
 {
-	return register_die_notifier(&kgdb_notifier);
+	int i, cpu;
+	int ret;
+	struct perf_event_attr attr;
+	struct perf_event **pevent;
+
+	ret = register_die_notifier(&kgdb_notifier);
+	if (ret != 0)
+		return ret;
+	/*
+	 * Pre-allocate the hw breakpoint structions in the non-atomic
+	 * portion of kgdb because this operation requires mutexs to
+	 * complete.
+	 */
+	attr.bp_addr = (unsigned long)kgdb_arch_init;
+	attr.type = PERF_TYPE_BREAKPOINT;
+	attr.bp_len = HW_BREAKPOINT_LEN_1;
+	attr.bp_type = HW_BREAKPOINT_W;
+	attr.disabled = 1;
+	for (i = 0; i < 4; i++) {
+		breakinfo[i].pev = register_wide_hw_breakpoint(&attr, NULL);
+		if (IS_ERR(breakinfo[i].pev)) {
+			printk(KERN_ERR "kgdb: Could not allocate hw breakpoints\n");
+			breakinfo[i].pev = NULL;
+			kgdb_arch_exit();
+			return -1;
+		}
+		for_each_online_cpu(cpu) {
+			pevent = per_cpu_ptr(breakinfo[i].pev, cpu);
+			pevent[0]->hw.sample_period = 1;
+			if (pevent[0]->destroy != NULL) {
+				pevent[0]->destroy = NULL;
+				release_bp_slot(*pevent);
+			}
+		}
+	}
+	return ret;
 }
 
 /**
@@ -550,6 +600,13 @@ int kgdb_arch_init(void)
  */
 void kgdb_arch_exit(void)
 {
+	int i;
+	for (i = 0; i < 4; i++) {
+		if (breakinfo[i].pev) {
+			unregister_wide_hw_breakpoint(breakinfo[i].pev);
+			breakinfo[i].pev = NULL;
+		}
+	}
 	unregister_die_notifier(&kgdb_notifier);
 }
 
diff --git a/kernel/kgdb.c b/kernel/kgdb.c
index 2eb517e..c7ade62 100644
--- a/kernel/kgdb.c
+++ b/kernel/kgdb.c
@@ -583,6 +583,9 @@ static void kgdb_wait(struct pt_regs *regs)
 	smp_wmb();
 	atomic_set(&cpu_in_kgdb[cpu], 1);
 
+	/* Disable any cpu specific hw breakpoints */
+	kgdb_disable_hw_debug(regs);
+
 	/* Wait till primary CPU is done with debugging */
 	while (atomic_read(&passive_cpu_wait[cpu]))
 		cpu_relax();