Skip to contentSkip to conversation

Introducing generic hardware breakpoint handler interfaces

PATCH In review K.Prasad 3 participants 17 messages Mar 24, 2009 – Mar 27, 2009 1 message not in the archive

Subsystems (5)
lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

K.Prasad

author
#1 lore raw

PATCH 1/11 Introducing generic hardware breakpoint handler interfaces

This patch introduces two new files hw_breakpoint.[ch] which defines the generic interfaces to use hardware breakpoint infrastructure of the system.

---
 arch/Kconfig                        |    4 
 include/asm-generic/hw_breakpoint.h |  138 +++++++++++++
 kernel/Makefile                     |    1 
 kernel/hw_breakpoint.c              |  367 ++++++++++++++++++++++++++++++++++++
 4 files changed, 510 insertions(+)
kernel/hw_breakpoint.c new file +367 -0
Index: linux-2.6-tip/kernel/hw_breakpoint.c
===================================================================
@@ -0,0 +1,367 @@
+/*
+ * 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
+ */
+
+/*
+ * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
+ * using the CPU's debug registers.
+ *
+ * This file contains the arch-independent routines.  It is not meant
+ * to be compiled as a standalone source file; rather it should be
+ * #include'd by the arch-specific implementation.
+ */
+
+#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/mutex.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/smp.h>
+
+#include <asm/hw_breakpoint.h>
+#include <asm/processor.h>
+#include <asm/debugreg.h>
+
+/*
+ * Mutex that protects all (un)register operations over kernel/user-space
+ * breakpoint requests
+ */
+DEFINE_MUTEX(hw_breakpoint_mutex);
+
+/* Array of kernel-space breakpoint structures */
+struct hw_breakpoint *hbp_kernel[HB_NUM];
+
+/*
+ * Kernel breakpoints grow downwards, starting from HB_NUM
+ * 'hbp_kernel_pos' denotes lowest numbered breakpoint register occupied for
+ * kernel-space request
+ */
+unsigned int hbp_kernel_pos;
+
+/*
+ * An array containing refcount of threads using a given bkpt register
+ * Accesses are synchronised by acquiring hw_breakpoint_mutex
+ */
+unsigned int hbp_user_refcount[HB_NUM];
+
+struct task_struct *last_debugged_task;
+
+/*
+ * Install the debug register values for a new thread.
+ */
+void switch_to_thread_hw_breakpoint(struct task_struct *tsk)
+{
+	/* Set the debug register */
+	arch_install_thread_hw_breakpoint(tsk);
+	last_debugged_task = current;
+}
+
+/*
+ * Install the debug register values for just the kernel, no thread.
+ */
+void switch_to_none_hw_breakpoint(void)
+{
+	arch_install_none();
+}
+
+/*
+ * Load the debug registers during startup of a CPU.
+ */
+void load_debug_registers(void)
+{
+	unsigned long flags;
+	struct task_struct *tsk = current;
+
+	mutex_lock(&hw_breakpoint_mutex);
+
+	/* Prevent IPIs for new kernel breakpoint updates */
+	local_irq_save(flags);
+	arch_load_debug_registers();
+	local_irq_restore(flags);
+
+	if (test_tsk_thread_flag(tsk, TIF_DEBUG))
+		switch_to_thread_hw_breakpoint(tsk);
+
+	mutex_unlock(&hw_breakpoint_mutex);
+}
+
+/*
+ * Erase all the hardware breakpoint info associated with a thread.
+ *
+ * If tsk != current then tsk must not be usable (for example, a
+ * child being cleaned up from a failed fork).
+ */
+void flush_thread_hw_breakpoint(struct task_struct *tsk)
+{
+	int i;
+	struct thread_struct *thread = &(tsk->thread);
+
+	mutex_lock(&hw_breakpoint_mutex);
+
+	/* The thread no longer has any breakpoints associated with it */
+	clear_tsk_thread_flag(tsk, TIF_DEBUG);
+	for (i = 0; i < HB_NUM; i++) {
+		if (thread->hbp[i]) {
+			hbp_user_refcount[i]--;
+			kfree(thread->hbp[i]);
+			thread->hbp[i] = NULL;
+		}
+	}
+	thread->hbp_num_installed = 0;
+
+	/* Actually uninstall the breakpoints if necessary */
+	if (tsk == current)
+		switch_to_none_hw_breakpoint();
+	mutex_unlock(&hw_breakpoint_mutex);
+}
+
+/*
+ * Copy the hardware breakpoint info from a thread to its cloned child.
+ */
+int copy_thread_hw_breakpoint(struct task_struct *tsk,
+		struct task_struct *child, unsigned long clone_flags)
+{
+	/*
+	 * We will assume that breakpoint settings are not inherited
+	 * and the child starts out with no debug registers set.
+	 * But what about CLONE_PTRACE?
+	 */
+	clear_tsk_thread_flag(child, TIF_DEBUG);
+	return 0;
+}
+
+/*
+ * Validate the settings in a hw_breakpoint structure.
+ */
+static int validate_settings(struct hw_breakpoint *bp, struct task_struct *tsk)
+{
+	int ret;
+	unsigned int align;
+
+	if (!bp)
+		return -EINVAL;
+
+	ret = arch_validate_hwbkpt_settings(bp, &align, tsk);
+	if (ret < 0)
+		goto err;
+
+	/*
+	 * Check that the low-order bits of the address are appropriate
+	 * for the alignment implied by len.
+	 */
+	if (bp->info.address & align)
+		return -EINVAL;
+
+	/* Check that the virtual address is in the proper range */
+	if (tsk) {
+		if (!arch_check_va_in_userspace(bp->info.address, bp->info.len))
+			return -EFAULT;
+	} else {
+		if (!arch_check_va_in_kernelspace(bp->info.address,
+								bp->info.len))
+			return -EFAULT;
+	}
+ err:
+	return ret;
+}
+
+int __register_user_hw_breakpoint(int pos, struct task_struct *tsk,
+					struct hw_breakpoint *bp)
+{
+	struct thread_struct *thread = &(tsk->thread);
+	int rc;
+
+	/* Do not overcommit. Fail if kernel has used the hbp registers */
+	if (pos >= hbp_kernel_pos)
+		return -ENOSPC;
+
+	rc = validate_settings(bp, tsk);
+	if (rc)
+		return rc;
+
+	thread->hbp[pos] = bp;
+	thread->hbp_num_installed++;
+	hbp_user_refcount[pos]++;
+
+	arch_register_user_hw_breakpoint(pos, bp, tsk);
+
+	/*
+	 * Does it need to be installed right now?
+	 * Otherwise it will get installed the next time tsk runs
+	 */
+	if (tsk == current)
+		switch_to_thread_hw_breakpoint(tsk);
+	return rc;
+}
+
+/*
+ * Modify the address of a hbp register already in use by the task
+ * Do not invoke this in-lieu of a __unregister_user_hw_breakpoint()
+ */
+int __modify_user_hw_breakpoint(int pos, struct task_struct *tsk,
+					struct hw_breakpoint *bp)
+{
+	int rc;
+	struct thread_struct *thread = &(tsk->thread);
+
+	if ((pos >= hbp_kernel_pos) || (validate_settings(bp, tsk)))
+		return -EINVAL;
+
+	thread->hbp[pos] = bp;
+
+	/*
+	 * 'pos' must be that of a hbp register already used by 'tsk'
+	 * Otherwise arch_modify_user_hw_breakpoint() will fail
+	 */
+	rc = arch_modify_user_hw_breakpoint(pos, bp, tsk);
+	if (rc)
+		return rc;
+
+	if (tsk == current)
+		switch_to_thread_hw_breakpoint(tsk);
+	return 0;
+}
+
+/*
+ * Actual implementation of unregister_user_hw_breakpoint.
+ */
+void __unregister_user_hw_breakpoint(int pos, struct task_struct *tsk,
+						struct hw_breakpoint *bp)
+{
+	struct thread_struct *thread = &(tsk->thread);
+
+	if (!bp)
+		return;
+
+	hbp_user_refcount[pos]--;
+	thread->hbp_num_installed--;
+
+	arch_unregister_user_hw_breakpoint(pos, bp, tsk);
+
+	if (tsk == current)
+		switch_to_thread_hw_breakpoint(tsk);
+	kfree(tsk->thread.hbp[pos]);
+	tsk->thread.hbp[pos] = NULL;
+}
+
+/**
+ * register_kernel_hw_breakpoint - register a hardware breakpoint for kernel space
+ * @bp: the breakpoint structure to register
+ *
+ * @bp.info->name or @bp.info->address, @bp.info->len, @bp.info->type and
+ * @bp->triggered must be set properly before invocation
+ *
+ */
+int register_kernel_hw_breakpoint(struct hw_breakpoint *bp)
+{
+	int rc;
+
+	rc = validate_settings(bp, NULL);
+	if (rc)
+		return rc;
+
+	mutex_lock(&hw_breakpoint_mutex);
+
+	rc = -EINVAL;
+	/* Check if we are over-committing */
+	if ((hbp_kernel_pos > 0) && (!hbp_user_refcount[hbp_kernel_pos-1])) {
+		hbp_kernel_pos--;
+		hbp_kernel[hbp_kernel_pos] = bp;
+		arch_register_kernel_hw_breakpoint(hbp_kernel_pos);
+		rc = 0;
+	}
+
+	mutex_unlock(&hw_breakpoint_mutex);
+	return rc;
+}
+EXPORT_SYMBOL_GPL(register_kernel_hw_breakpoint);
+
+/**
+ * unregister_kernel_hw_breakpoint - unregister a HW breakpoint for kernel space
+ * @bp: the breakpoint structure to unregister
+ *
+ * Uninstalls and unregisters @bp.
+ */
+void unregister_kernel_hw_breakpoint(struct hw_breakpoint *bp)
+{
+	int i, j;
+
+	mutex_lock(&hw_breakpoint_mutex);
+
+	/* Find the 'bp' in our list of breakpoints for kernel */
+	for (i = hbp_kernel_pos; i < HB_NUM; i++)
+		if (bp == hbp_kernel[i])
+			break;
+	/*
+	 * We'll shift the breakpoints one-level above to compact if
+	 * unregistration creates a hole
+	 */
+	if (i > hbp_kernel_pos)
+		for (j = i; j == hbp_kernel_pos; j--)
+			hbp_kernel[j] = hbp_kernel[j-1];
+
+	/*
+	 * Delete the last kernel breakpoint entry after compaction and update
+	 * the pointer to the lowest numbered kernel entry after updating its
+	 * corresponding value in kdr7
+	 */
+	hbp_kernel[hbp_kernel_pos] = 0;
+	arch_unregister_kernel_hw_breakpoint();
+	hbp_kernel_pos++;
+
+	mutex_unlock(&hw_breakpoint_mutex);
+}
+EXPORT_SYMBOL_GPL(unregister_kernel_hw_breakpoint);
+
+/*
+ * Handle debug exception notifications.
+ */
+static int __kprobes hw_breakpoint_exceptions_notify(
+		struct notifier_block *unused, unsigned long val, void *data)
+{
+	if (val != DIE_DEBUG)
+		return NOTIFY_DONE;
+
+	return hw_breakpoint_handler(data);
+}
+
+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)
+{
+	int i;
+
+	hbp_kernel_pos = HB_NUM;
+	for (i = 0; i < HB_NUM; i++)
+		hbp_user_refcount[i] = 0;
+	load_debug_registers();
+
+	return register_die_notifier(&hw_breakpoint_exceptions_nb);
+}
+
+core_initcall(init_hw_breakpoint);
kernel/Makefile +1 -0
Index: linux-2.6-tip/kernel/Makefile
===================================================================
@@ -95,6 +95,7 @@ obj-$(CONFIG_FUNCTION_TRACER) += trace/
 obj-$(CONFIG_TRACING) += trace/
 obj-$(CONFIG_SMP) += sched_cpupri.o
 obj-$(CONFIG_PERF_COUNTERS) += perf_counter.o
+obj-$(CONFIG_HAVE_HW_BREAKPOINT) += hw_breakpoint.o
 
 ifneq ($(CONFIG_SCHED_OMIT_FRAME_POINTER),y)
 # According to Alan Modra <[email protected]>, the -fno-omit-frame-pointer is
include/asm-generic/hw_breakpoint.h new file +138 -0

138 lines not on this page · raw message

arch/Kconfig +4 -0
Index: linux-2.6-tip/arch/Kconfig
===================================================================
@@ -109,3 +109,7 @@ config HAVE_CLK
 
 config HAVE_DMA_API_DEBUG
 	bool
+
+config HAVE_HW_BREAKPOINT
+	bool
+
author
#2 lore raw

PATCH 2/11 x86 architecture implementation of Hardware Breakpoint interfaces

In reply to a message not in the archive

This patch introduces two new files named hw_breakpoint.[ch] inside x86 specific directories. They contain functions which help validate and serve requests for using Hardware Breakpoint registers on x86 processors.

---
 arch/x86/Kconfig                     |    1 
 arch/x86/include/asm/hw_breakpoint.h |   73 +++++
 arch/x86/kernel/Makefile             |    2 
 arch/x86/kernel/hw_breakpoint.c      |  441 +++++++++++++++++++++++++++++++++++
 4 files changed, 516 insertions(+), 1 deletion(-)
arch/x86/kernel/hw_breakpoint.c new file +441 -0
Index: linux-2.6-tip/arch/x86/kernel/hw_breakpoint.c
===================================================================
@@ -0,0 +1,441 @@
+/*
+ * 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) 2009 IBM Corporation
+ */
+
+/*
+ * HW_breakpoint: a unified kernel/user-space hardware breakpoint facility,
+ * using the CPU's debug registers.
+ */
+
+#include <linux/irqflags.h>
+#include <linux/notifier.h>
+#include <linux/kallsyms.h>
+#include <linux/kprobes.h>
+#include <linux/percpu.h>
+#include <linux/kdebug.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/smp.h>
+
+#include <asm/hw_breakpoint.h>
+#include <asm/processor.h>
+#include <asm/debugreg.h>
+
+/* Unmasked kernel DR7 value */
+static unsigned long kdr7;
+
+/*
+ * Masks for the bits corresponding to registers DR0 - DR3 in DR7 register.
+ * Used to clear and verify the status of bits corresponding to DR0 - DR3
+ */
+static const unsigned long	dr7_masks[HB_NUM] = {
+	0x000f0003,	/* LEN0, R/W0, G0, L0 */
+	0x00f0000c,	/* LEN1, R/W1, G1, L1 */
+	0x0f000030,	/* LEN2, R/W2, G2, L2 */
+	0xf00000c0	/* LEN3, R/W3, G3, L3 */
+};
+
+
+/*
+ * Encode the length, type, Exact, and Enable bits for a particular breakpoint
+ * as stored in debug register 7.
+ */
+static unsigned long encode_dr7(int drnum, unsigned len, unsigned type)
+{
+	unsigned long temp;
+
+	temp = (len | type) & 0xf;
+	temp <<= (DR_CONTROL_SHIFT + drnum * DR_CONTROL_SIZE);
+	temp |= (DR_GLOBAL_ENABLE << (drnum * DR_ENABLE_SIZE)) |
+				DR_GLOBAL_SLOWDOWN;
+	return temp;
+}
+
+/*
+ * Install the kernel breakpoints in their debug registers.
+ * If 0 <= pos < HB_NUM, then set the debug register corresponding to that number
+ * If 'pos' is negative, then all debug registers are updated
+ */
+void arch_install_kernel_hw_breakpoint(void *idx)
+{
+	int pos = *(int *)idx;
+	unsigned long dr7;
+	int i;
+
+	get_debugreg(dr7, 7);
+
+	/* Don't allow debug exceptions while we update the registers */
+	set_debugreg(0UL, 7);
+
+	for (i = hbp_kernel_pos; i < HB_NUM; i++) {
+		if ((pos >= 0) && (i != pos))
+			continue;
+		dr7 &= ~(dr7_masks[i]);
+		if (hbp_kernel[i])
+			set_debugreg(hbp_kernel[i]->info.address, i);
+	}
+
+	dr7 |= kdr7;
+	/* No need to set DR6 */
+	set_debugreg(dr7, 7);
+}
+
+void arch_load_debug_registers()
+{
+	int pos = -1;
+	/*
+	 * We want all debug registers to be initialised for this
+	 * CPU so pos = -1
+	 */
+	arch_install_kernel_hw_breakpoint((void *)&pos);
+}
+
+/*
+ * Install the thread breakpoints in their debug registers.
+ */
+void arch_install_thread_hw_breakpoint(struct task_struct *tsk)
+{
+	int i;
+	struct thread_struct *thread = &(tsk->thread);
+
+	for (i = 0;  (i < hbp_kernel_pos) && hbp_user_refcount[i]; i++)
+		if (thread->hbp[i])
+			set_debugreg(thread->hbp[i]->info.address, i);
+
+	/* No need to set DR6 */
+
+	set_debugreg((kdr7 | thread->dr7), 7);
+}
+
+/*
+ * Install the debug register values for just the kernel, no thread.
+ */
+void arch_install_none()
+{
+	/* Clear the user-space portion of dr7 by setting only kdr7 */
+	set_debugreg(kdr7, 7);
+}
+
+static int get_hbp_len(u8 hbp_len)
+{
+	unsigned int len_in_bytes = 0;
+
+	switch (hbp_len) {
+	case HW_BREAKPOINT_LEN_1:
+		len_in_bytes = 1;
+		break;
+	case HW_BREAKPOINT_LEN_2:
+		len_in_bytes = 2;
+		break;
+	case HW_BREAKPOINT_LEN_4:
+		len_in_bytes = 4;
+		break;
+#ifdef CONFIG_X86_64
+	case HW_BREAKPOINT_LEN_8:
+		len_in_bytes = 8;
+		break;
+#endif
+	}
+	return len_in_bytes;
+}
+
+/*
+ * Check for virtual address in user space.
+ */
+int arch_check_va_in_userspace(unsigned long va, u8 hbp_len)
+{
+	unsigned int len;
+
+	len = get_hbp_len(hbp_len);
+
+	return (va <= TASK_SIZE - len);
+}
+
+/*
+ * Check for virtual address in kernel space.
+ */
+int arch_check_va_in_kernelspace(unsigned long va, u8 hbp_len)
+{
+	unsigned int len;
+
+	len = get_hbp_len(hbp_len);
+
+	return ((va >= TASK_SIZE) && ((va + len) >= TASK_SIZE));
+}
+
+/*
+ * Store a breakpoint's encoded address, length, and type.
+ */
+void arch_store_info(struct hw_breakpoint *bp)
+{
+	/*
+	 * User-space requests will always have the address field populated
+	 * For kernel-addresses, either the address or symbol name can be
+	 * specified.
+	 */
+	if (bp->info.address)
+		return;
+	if (bp->info.name)
+		bp->info.address = (unsigned long)
+					kallsyms_lookup_name(bp->info.name);
+}
+
+/*
+ * Validate the arch-specific HW Breakpoint register settings
+ */
+int arch_validate_hwbkpt_settings(struct hw_breakpoint *bp,
+				unsigned int *align, struct task_struct *tsk)
+{
+	int ret = -EINVAL;
+
+	switch (bp->info.type) {
+
+	/*
+	 * Ptrace-refactoring code
+	 * For now, we'll allow instruction breakpoint only for user-space
+	 * addresses
+	 */
+	case HW_BREAKPOINT_EXECUTE:
+		if ((!arch_check_va_in_userspace(bp->info.address,
+							bp->info.len)) &&
+			bp->info.len != HW_BREAKPOINT_LEN_EXECUTE)
+			return ret;
+		break;
+	case HW_BREAKPOINT_WRITE:
+				break;
+	case HW_BREAKPOINT_RW:
+				break;
+	default:
+		return ret;
+	}
+
+	switch (bp->info.len) {
+	case HW_BREAKPOINT_LEN_1:
+		*align = 0;
+		break;
+	case HW_BREAKPOINT_LEN_2:
+		*align = 1;
+		break;
+	case HW_BREAKPOINT_LEN_4:
+		*align = 3;
+		break;
+#ifdef CONFIG_X86_64
+	case HW_BREAKPOINT_LEN_8:
+		*align = 7;
+		break;
+#endif
+	default:
+		return ret;
+	}
+
+	if (bp->triggered) {
+		ret = 0;
+		arch_store_info(bp);
+	}
+	return ret;
+}
+
+/*
+ * Register a new user breakpoint structure.
+ */
+void arch_register_user_hw_breakpoint(int pos, struct hw_breakpoint *bp,
+		struct task_struct *tsk)
+{
+	struct thread_struct *thread = &(tsk->thread);
+
+	thread->dr7 &= ~dr7_masks[pos];
+	thread->dr7 |= encode_dr7(pos, bp->info.len, bp->info.type);
+}
+
+/*
+ * Modify an existing user breakpoint structure.
+ */
+int arch_modify_user_hw_breakpoint(int pos, struct hw_breakpoint *bp,
+		struct task_struct *tsk)
+{
+	struct thread_struct *thread = &(tsk->thread);
+
+	/* Check if the register to be modified was enabled by the thread */
+	if (!(thread->dr7 & (1 << (pos * DR_ENABLE_SIZE))))
+		return -EINVAL;
+
+	thread->dr7 &= ~dr7_masks[pos];
+	thread->dr7 |= encode_dr7(pos, bp->info.len, bp->info.type);
+
+	return 0;
+}
+
+/*
+ * Unregister a user breakpoint structure.
+ */
+void arch_unregister_user_hw_breakpoint(int pos, struct hw_breakpoint *bp,
+					struct task_struct *tsk)
+{
+	struct thread_struct *thread = &(tsk->thread);
+
+	if (!thread->hbp[pos])
+		return;
+
+	thread->hbp[pos]->info.address = 0;
+	thread->dr7 &= ~dr7_masks[pos];
+}
+
+/*
+ * Register a kernel breakpoint structure.
+ */
+void arch_register_kernel_hw_breakpoint(int pos)
+{
+	struct hw_breakpoint *bp;
+
+	bp = hbp_kernel[pos];
+
+	kdr7 &= ~(dr7_masks[pos]);
+	if (bp)
+		kdr7 |= encode_dr7(pos, bp->info.len, bp->info.type);
+
+	on_each_cpu(arch_install_kernel_hw_breakpoint, (void *)&pos, 0);
+}
+
+/*
+ * Unregister a kernel breakpoint structure.
+ */
+void arch_unregister_kernel_hw_breakpoint()
+{
+	int i, pos = -1;
+
+	/*
+	 * Modify kdr7 to reflect the new layout of kernel-space breakpoints
+	 * and invoke the routine to write breakpoint addresses onto the
+	 * physical registers
+	 */
+	for (i = hbp_kernel_pos; i < HB_NUM; i++) {
+		kdr7 &= ~(dr7_masks[i]);
+		if (hbp_kernel[i]) {
+			struct hw_breakpoint *bp = hbp_kernel[i];
+			kdr7 |= encode_dr7(i, bp->info.len, bp->info.type);
+		}
+	}
+	on_each_cpu(arch_install_kernel_hw_breakpoint, (void *)&pos, 0);
+}
+
+
+/*
+ * Copy out the debug register information for a core dump.
+ *
+ * tsk must be equal to current.
+ */
+void dump_thread_hw_breakpoint(struct task_struct *tsk, int u_debugreg[8])
+{
+	struct thread_struct *thread = &(tsk->thread);
+	int i;
+
+	memset(u_debugreg, 0, sizeof u_debugreg);
+	for (i = 0; i < thread->hbp_num_installed && thread->hbp[i]; ++i)
+		u_debugreg[i] = thread->hbp[i]->info.address;
+	u_debugreg[7] = thread->dr7;
+	u_debugreg[6] = thread->dr6;
+}
+
+/*
+ * Handle debug exception notifications.
+ */
+int __kprobes hw_breakpoint_handler(struct die_args *args)
+{
+	int i, rc = NOTIFY_DONE;
+	struct hw_breakpoint *bp;
+	/* The DR6 value is stored in args->err */
+	unsigned long dr7, dr6 = args->err;
+
+	if (dr6 & DR_STEP)
+		return NOTIFY_DONE;
+
+	get_debugreg(dr7, 7);
+
+	/* Disable breakpoints during exception handling */
+	set_debugreg(0UL, 7);
+
+	/*
+	 * Assert that local interrupts are disabled
+	 * Reset the DRn bits in the virtualized register value.
+	 * The ptrace trigger routine will add in whatever is needed.
+	 */
+	current->thread.dr6 &= ~(DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3);
+
+	/* Lazy debug register switching */
+	if (last_debugged_task != current)
+		switch_to_none_hw_breakpoint();
+
+	/* Handle all the breakpoints that were triggered */
+	for (i = 0; i < HB_NUM; ++i) {
+		if (likely(!(dr6 & (DR_TRAP0 << i))))
+			continue;
+		/*
+		 * Find the corresponding hw_breakpoint structure and
+		 * invoke its triggered callback.
+		 */
+		if (hbp_user_refcount[i])
+			bp = current->thread.hbp[i];
+		else if (i >= hbp_kernel_pos)
+			bp = hbp_kernel[i];
+		else	/* False alarm due to lazy DR switching */
+			continue;
+
+		if (!bp)
+			break;
+
+		switch (bp->info.type) {
+		case HW_BREAKPOINT_WRITE:
+		case HW_BREAKPOINT_RW:
+			if (bp->triggered)
+				(bp->triggered)(bp, args->regs);
+
+			if (arch_check_va_in_userspace(bp->info.address,
+							bp->info.len))
+				rc = NOTIFY_DONE;
+			else
+				rc = NOTIFY_STOP;;
+			goto exit;
+		case HW_BREAKPOINT_EXECUTE:
+			/*
+			 * Presently we allow instruction breakpoints only in
+			 * user-space when requested through ptrace.
+			 */
+			if (arch_check_va_in_userspace(bp->info.address,
+							bp->info.len)) {
+				(bp->triggered)(bp, args->regs);
+				/*
+				 * do_debug will notify user through a SIGTRAP
+				 * signal So we are not requesting a
+				 * NOTIFY_STOP here
+				 */
+				rc = NOTIFY_DONE;
+				goto exit;
+			}
+		}
+	}
+
+	/* Stop processing further if the exception is a stray one */
+	if (!(dr6 & ~(DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)))
+		rc = NOTIFY_STOP;
+exit:
+	set_debugreg(dr7, 7);
+	return rc;
+}
arch/x86/include/asm/hw_breakpoint.h new file +73 -0

73 lines not on this page · raw message

arch/x86/kernel/Makefile +1 -1
Index: linux-2.6-tip/arch/x86/kernel/Makefile
===================================================================
@@ -36,7 +36,7 @@ obj-$(CONFIG_X86_64)	+= sys_x86_64.o x86
 obj-$(CONFIG_X86_64)	+= syscall_64.o vsyscall_64.o
 obj-y			+= bootflag.o e820.o
 obj-y			+= pci-dma.o quirks.o i8237.o topology.o kdebugfs.o
-obj-y			+= alternative.o i8253.o pci-nommu.o
+obj-y			+= alternative.o i8253.o pci-nommu.o hw_breakpoint.o
 obj-y			+= tsc.o io_delay.o rtc.o
 
 obj-$(CONFIG_X86_TRAMPOLINE)	+= trampoline.o
arch/x86/Kconfig +1 -0
Index: linux-2.6-tip/arch/x86/Kconfig
===================================================================
@@ -47,6 +47,7 @@ config X86
 	select HAVE_KERNEL_LZMA
 	select HAVE_ARCH_KMEMCHECK
 	select HAVE_DMA_API_DEBUG
+	select HAVE_HW_BREAKPOINT
 
 config ARCH_DEFCONFIG
 	string
author
#3 lore raw

PATCH 3/11 Modifying generic debug exception to use thread-specific debug registers

In reply to a message not in the archive

From: Alan Stern <[email protected]>

This patch modifies the breakpoint exception handler code to use the abstract register names.

[K.Prasad: Split-out from the bigger patch and minor changes following
re-basing]

---
 arch/x86/kernel/traps.c |   73 ++++++++++++++++--------------------------------
 1 file changed, 25 insertions(+), 48 deletions(-)
arch/x86/kernel/traps.c +25 -48
Index: linux-2.6-tip/arch/x86/kernel/traps.c
===================================================================
@@ -530,13 +530,14 @@ asmlinkage __kprobes struct pt_regs *syn
 dotraplinkage void __kprobes do_debug(struct pt_regs *regs, long error_code)
 {
 	struct task_struct *tsk = current;
-	unsigned long condition;
+	unsigned long dr6;
 	int si_code;
 
-	get_debugreg(condition, 6);
+	get_debugreg(dr6, 6);
+	set_debugreg(0, 6);	/* DR6 may or may not be cleared by the CPU */
 
 	/* Catch kmemcheck conditions first of all! */
-	if (condition & DR_STEP && kmemcheck_trap(regs))
+	if (dr6 & DR_STEP && kmemcheck_trap(regs))
 		return;
 
 	/*
@@ -545,61 +546,37 @@ dotraplinkage void __kprobes do_debug(st
 	clear_tsk_thread_flag(tsk, TIF_DEBUGCTLMSR);
 	tsk->thread.debugctlmsr = 0;
 
-	if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code,
+	/* Store the virtualized DR6 value */
+	tsk->thread.dr6 = dr6;
+
+	if (notify_die(DIE_DEBUG, "debug", regs, dr6, error_code,
 						SIGTRAP) == NOTIFY_STOP)
 		return;
 
 	/* It's safe to allow irq's after DR6 has been saved */
 	preempt_conditional_sti(regs);
 
-	/* Mask out spurious debug traps due to lazy DR7 setting */
-	if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) {
-		if (!tsk->thread.debugreg7)
-			goto clear_dr7;
-	}
-
-#ifdef CONFIG_X86_32
-	if (regs->flags & X86_VM_MASK)
-		goto debug_vm86;
-#endif
-
-	/* Save debug status register where ptrace can see it */
-	tsk->thread.debugreg6 = condition;
-
-	/*
-	 * Single-stepping through TF: make sure we ignore any events in
-	 * kernel space (but re-enable TF when returning to user mode).
-	 */
-	if (condition & DR_STEP) {
-		if (!user_mode(regs))
-			goto clear_TF_reenable;
+	if (regs->flags & X86_VM_MASK) {
+		handle_vm86_trap((struct kernel_vm86_regs *) regs,
+				error_code, 1);
+		return;
 	}
 
-	si_code = get_si_code(condition);
-	/* Ok, finally something we can handle */
-	send_sigtrap(tsk, regs, error_code, si_code);
-
 	/*
-	 * Disable additional traps. They'll be re-enabled when
-	 * the signal is delivered.
+	 * Single-stepping through system calls: ignore any exceptions in
+	 * kernel space, but re-enable TF when returning to user mode.
+	 *
+	 * We already checked v86 mode above, so we can check for kernel mode
+	 * by just checking the CPL of CS.
 	 */
-clear_dr7:
-	set_debugreg(0, 7);
-	preempt_conditional_cli(regs);
-	return;
-
-#ifdef CONFIG_X86_32
-debug_vm86:
-	/* reenable preemption: handle_vm86_trap() might sleep */
-	dec_preempt_count();
-	handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, 1);
-	conditional_cli(regs);
-	return;
-#endif
-
-clear_TF_reenable:
-	set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
-	regs->flags &= ~X86_EFLAGS_TF;
+	if ((dr6 & DR_STEP) && !user_mode(regs)) {
+		tsk->thread.dr6 &= ~DR_STEP;
+		set_tsk_thread_flag(tsk, TIF_SINGLESTEP);
+		regs->flags &= ~X86_EFLAGS_TF;
+	}
+	si_code = get_si_code(dr6);
+	if (tsk->thread.dr6 & (DR_STEP|DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3))
+		send_sigtrap(tsk, regs, error_code, si_code);
 	preempt_conditional_cli(regs);
 	return;
 }
author
#4 lore raw

PATCH 4/11 Introduce user-space debug registers

In reply to a message not in the archive

This patch introduces debug registers to used by the per-thread structure and wrapper routines to manage debug registers by process-related functions.

---
 arch/x86/include/asm/debugreg.h  |   23 +++++++++++++++++++++++
 arch/x86/include/asm/processor.h |   16 +++++++++-------
 2 files changed, 32 insertions(+), 7 deletions(-)
arch/x86/include/asm/debugreg.h +23 -0
Index: linux-2.6-tip/arch/x86/include/asm/debugreg.h
===================================================================
@@ -49,6 +49,8 @@
 
 #define DR_LOCAL_ENABLE_SHIFT 0    /* Extra shift to the local enable bit */
 #define DR_GLOBAL_ENABLE_SHIFT 1   /* Extra shift to the global enable bit */
+#define DR_LOCAL_ENABLE (0x1)      /* Local enable for reg 0 */
+#define DR_GLOBAL_ENABLE (0x2)     /* Global enable for reg 0 */
 #define DR_ENABLE_SIZE 2           /* 2 enable bits per register */
 
 #define DR_LOCAL_ENABLE_MASK (0x55)  /* Set  local bits for all 4 regs */
@@ -67,4 +69,25 @@
 #define DR_LOCAL_SLOWDOWN (0x100)   /* Local slow the pipeline */
 #define DR_GLOBAL_SLOWDOWN (0x200)  /* Global slow the pipeline */
 
+/*
+ * HW breakpoint additions
+ */
+#ifdef __KERNEL__
+
+/* For process management */
+void flush_thread_hw_breakpoint(struct task_struct *tsk);
+int copy_thread_hw_breakpoint(struct task_struct *tsk,
+		struct task_struct *child, unsigned long clone_flags);
+void dump_thread_hw_breakpoint(struct task_struct *tsk, int u_debugreg[8]);
+void switch_to_thread_hw_breakpoint(struct task_struct *tsk);
+
+/* For CPU management */
+void load_debug_registers(void);
+static inline void hw_breakpoint_disable(void)
+{
+	set_debugreg(0UL, 7);
+}
+
+#endif	/* __KERNEL__ */
+
 #endif /* _ASM_X86_DEBUGREG_H */
arch/x86/include/asm/processor.h +9 -7
Index: linux-2.6-tip/arch/x86/include/asm/processor.h
===================================================================
@@ -29,6 +29,7 @@ struct mm_struct;
 #include <linux/threads.h>
 #include <linux/init.h>
 
+#define HB_NUM 4
 /*
  * Default implementation of macro that returns current
  * instruction pointer ("program counter").
@@ -424,13 +425,14 @@ struct thread_struct {
 	unsigned long		ip;
 	unsigned long		fs;
 	unsigned long		gs;
-	/* Hardware debugging registers: */
-	unsigned long		debugreg0;
-	unsigned long		debugreg1;
-	unsigned long		debugreg2;
-	unsigned long		debugreg3;
-	unsigned long		debugreg6;
-	unsigned long		debugreg7;
+	/* Hardware breakpoint info */
+	struct hw_breakpoint	*hbp[HB_NUM];
+	unsigned int		hbp_num_installed;
+	/* Thread's view of debug reg 6 */
+	unsigned long		dr6;
+	/* Thread's view of debug reg 7 */
+	unsigned long		dr7;
+
 	/* Fault info: */
 	unsigned long		cr2;
 	unsigned long		trap_no;
author
#5 lore raw

PATCH 5/11 Use wrapper routines around debug registers in processor related functions

In reply to a message not in the archive

From: Alan Stern <[email protected]>

This patch enables the use of wrapper routines to access the debug/breakpoint registers.

[K.Prasad: Split-out from the bigger patch and minor changes following
re-basing]

---
 arch/x86/kernel/smpboot.c |    3 +++
 arch/x86/power/cpu_32.c   |   16 +++-------------
 arch/x86/power/cpu_64.c   |   15 +++------------
 3 files changed, 9 insertions(+), 25 deletions(-)
arch/x86/power/cpu_32.c +3 -13
Index: linux-2.6-tip/arch/x86/power/cpu_32.c
===================================================================
@@ -12,6 +12,7 @@
 #include <asm/mtrr.h>
 #include <asm/mce.h>
 #include <asm/xcr.h>
+#include <asm/debugreg.h>
 
 static struct saved_context saved_context;
 
@@ -47,6 +48,7 @@ static void __save_processor_state(struc
 	ctxt->cr2 = read_cr2();
 	ctxt->cr3 = read_cr3();
 	ctxt->cr4 = read_cr4_safe();
+	hw_breakpoint_disable();
 }
 
 /* Needed by apm.c */
@@ -79,19 +81,7 @@ static void fix_processor_context(void)
 	load_TR_desc();				/* This does ltr */
 	load_LDT(&current->active_mm->context);	/* This does lldt */
 
-	/*
-	 * Now maybe reload the debug registers
-	 */
-	if (current->thread.debugreg7) {
-		set_debugreg(current->thread.debugreg0, 0);
-		set_debugreg(current->thread.debugreg1, 1);
-		set_debugreg(current->thread.debugreg2, 2);
-		set_debugreg(current->thread.debugreg3, 3);
-		/* no 4 and 5 */
-		set_debugreg(current->thread.debugreg6, 6);
-		set_debugreg(current->thread.debugreg7, 7);
-	}
-
+	load_debug_registers();
 }
 
 static void __restore_processor_state(struct saved_context *ctxt)
arch/x86/power/cpu_64.c +3 -12
Index: linux-2.6-tip/arch/x86/power/cpu_64.c
===================================================================
@@ -15,6 +15,7 @@
 #include <asm/pgtable.h>
 #include <asm/mtrr.h>
 #include <asm/xcr.h>
+#include <asm/debugreg.h>
 
 static void fix_processor_context(void);
 
@@ -70,6 +71,7 @@ static void __save_processor_state(struc
 	ctxt->cr3 = read_cr3();
 	ctxt->cr4 = read_cr4();
 	ctxt->cr8 = read_cr8();
+	hw_breakpoint_disable();
 }
 
 void save_processor_state(void)
@@ -158,16 +160,5 @@ static void fix_processor_context(void)
 	load_TR_desc();				/* This does ltr */
 	load_LDT(&current->active_mm->context);	/* This does lldt */
 
-	/*
-	 * Now maybe reload the debug registers
-	 */
-	if (current->thread.debugreg7){
-                loaddebug(&current->thread, 0);
-                loaddebug(&current->thread, 1);
-                loaddebug(&current->thread, 2);
-                loaddebug(&current->thread, 3);
-                /* no 4 and 5 */
-                loaddebug(&current->thread, 6);
-                loaddebug(&current->thread, 7);
-	}
+	load_debug_registers();
 }
arch/x86/kernel/smpboot.c +3 -0
Index: linux-2.6-tip/arch/x86/kernel/smpboot.c
===================================================================
@@ -63,6 +63,7 @@
 #include <asm/apic.h>
 #include <asm/setup.h>
 #include <asm/uv/uv.h>
+#include <asm/debugreg.h>
 #include <linux/mc146818rtc.h>
 
 #include <asm/smpboot_hooks.h>
@@ -326,6 +327,7 @@ notrace static void __cpuinit start_seco
 	setup_secondary_clock();
 
 	wmb();
+	load_debug_registers();
 	cpu_idle();
 }
 
@@ -1250,6 +1252,7 @@ void cpu_disable_common(void)
 	remove_cpu_from_maps(cpu);
 	unlock_vector_lock();
 	fixup_irqs();
+	hw_breakpoint_disable();
 }
 
 int native_cpu_disable(void)
author
#6 lore raw

PATCH 6/11 Use the new wrapper routines to access debug registers in process/thread code

In reply to a message not in the archive

From: Alan Stern <[email protected]>

This patch enables the use of abstract debug registers in process-handling routines.

[K.Prasad: Split-out from the bigger patch and minor changes following
re-basing]

---
 arch/x86/kernel/process.c    |   23 ++++++-----------------
 arch/x86/kernel/process_32.c |   31 +++++++++++++++++++++++++++++++
 arch/x86/kernel/process_64.c |   33 +++++++++++++++++++++++++++++++++
 3 files changed, 70 insertions(+), 17 deletions(-)
arch/x86/kernel/process.c +6 -17
Index: linux-2.6-tip/arch/x86/kernel/process.c
===================================================================
@@ -14,6 +14,8 @@
 #include <asm/idle.h>
 #include <asm/uaccess.h>
 #include <asm/i387.h>
+#include <asm/debugreg.h>
+#include <asm/hw_breakpoint.h>
 
 unsigned long idle_halt;
 EXPORT_SYMBOL(idle_halt);
@@ -83,6 +85,8 @@ void exit_thread(void)
 		put_cpu();
 		kfree(bp);
 	}
+	if (unlikely(t->dr7))
+		flush_thread_hw_breakpoint(me);
 
 	ds_exit_thread(current);
 }
@@ -103,14 +107,9 @@ void flush_thread(void)
 	}
 #endif
 
-	clear_tsk_thread_flag(tsk, TIF_DEBUG);
+	if (unlikely(tsk->thread.dr7))
+		flush_thread_hw_breakpoint(tsk);
 
-	tsk->thread.debugreg0 = 0;
-	tsk->thread.debugreg1 = 0;
-	tsk->thread.debugreg2 = 0;
-	tsk->thread.debugreg3 = 0;
-	tsk->thread.debugreg6 = 0;
-	tsk->thread.debugreg7 = 0;
 	memset(tsk->thread.tls_array, 0, sizeof(tsk->thread.tls_array));
 	/*
 	 * Forget coprocessor state..
@@ -192,16 +191,6 @@ void __switch_to_xtra(struct task_struct
 	else if (next->debugctlmsr != prev->debugctlmsr)
 		update_debugctlmsr(next->debugctlmsr);
 
-	if (test_tsk_thread_flag(next_p, TIF_DEBUG)) {
-		set_debugreg(next->debugreg0, 0);
-		set_debugreg(next->debugreg1, 1);
-		set_debugreg(next->debugreg2, 2);
-		set_debugreg(next->debugreg3, 3);
-		/* no 4 and 5 */
-		set_debugreg(next->debugreg6, 6);
-		set_debugreg(next->debugreg7, 7);
-	}
-
 	if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^
 	    test_tsk_thread_flag(next_p, TIF_NOTSC)) {
 		/* prev and next are different */
arch/x86/kernel/process_32.c +31 -0
Index: linux-2.6-tip/arch/x86/kernel/process_32.c
===================================================================
@@ -61,6 +61,8 @@
 #include <asm/idle.h>
 #include <asm/syscalls.h>
 #include <asm/ds.h>
+#include <asm/debugreg.h>
+#include <asm/hw_breakpoint.h>
 
 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
 
@@ -265,7 +267,14 @@ int copy_thread(int nr, unsigned long cl
 
 	task_user_gs(p) = get_user_gs(regs);
 
+	p->thread.io_bitmap_ptr = NULL;
+
 	tsk = current;
+	err = -ENOMEM;
+	if (unlikely(tsk->thread.dr7)) {
+		if (copy_thread_hw_breakpoint(tsk, p, clone_flags))
+			goto out;
+	}
 	if (unlikely(test_tsk_thread_flag(tsk, TIF_IO_BITMAP))) {
 		p->thread.io_bitmap_ptr = kmemdup(tsk->thread.io_bitmap_ptr,
 						IO_BITMAP_BYTES, GFP_KERNEL);
@@ -285,10 +294,13 @@ int copy_thread(int nr, unsigned long cl
 		err = do_set_thread_area(p, -1,
 			(struct user_desc __user *)childregs->si, 0);
 
+out:
 	if (err && p->thread.io_bitmap_ptr) {
 		kfree(p->thread.io_bitmap_ptr);
 		p->thread.io_bitmap_max = 0;
 	}
+	if (err)
+		flush_thread_hw_breakpoint(p);
 
 	ds_copy_thread(p, current);
 
@@ -426,6 +438,25 @@ __switch_to(struct task_struct *prev_p, 
 		lazy_load_gs(next->gs);
 
 	percpu_write(current_task, next_p);
+	/*
+	 * There's a problem with moving the switch_to_thread_hw_breakpoint()
+	 * call before current is updated.  Suppose a kernel breakpoint is
+	 * triggered in between the two.  The hw-breakpoint handler will see
+	 * that current is different from the task pointer stored in the chbi
+	 * area, so it will think the task pointer is leftover from an old task
+	 * (lazy switching) and will erase it.  Then until the next context
+	 * switch, no user-breakpoints will be installed.
+	 *
+	 * The real problem is that it's impossible to update both current and
+	 * chbi->bp_task at the same instant, so there will always be a window
+	 * in which they disagree and a breakpoint might get triggered.  Since
+	 * we use lazy switching, we are forced to assume that a disagreement
+	 * means that current is correct and chbi->bp_task is old.  But if you
+	 * move the code above then you'll create a window in which current is
+	 * old and chbi->bp_task is correct.
+	 */
+	if (unlikely(test_tsk_thread_flag(next_p, TIF_DEBUG)))
+		switch_to_thread_hw_breakpoint(next_p);
 
 	return prev_p;
 }
arch/x86/kernel/process_64.c +33 -0
Index: linux-2.6-tip/arch/x86/kernel/process_64.c
===================================================================
@@ -55,6 +55,8 @@
 #include <asm/idle.h>
 #include <asm/syscalls.h>
 #include <asm/ds.h>
+#include <asm/debugreg.h>
+#include <asm/hw_breakpoint.h>
 
 asmlinkage extern void ret_from_fork(void);
 
@@ -248,6 +250,8 @@ void release_thread(struct task_struct *
 			BUG();
 		}
 	}
+	if (unlikely(dead_task->thread.dr7))
+		flush_thread_hw_breakpoint(dead_task);
 }
 
 static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr)
@@ -303,12 +307,18 @@ int copy_thread(int nr, unsigned long cl
 
 	p->thread.fs = me->thread.fs;
 	p->thread.gs = me->thread.gs;
+	p->thread.io_bitmap_ptr = NULL;
 
 	savesegment(gs, p->thread.gsindex);
 	savesegment(fs, p->thread.fsindex);
 	savesegment(es, p->thread.es);
 	savesegment(ds, p->thread.ds);
 
+	err = -ENOMEM;
+	if (unlikely(me->thread.dr7)) {
+		if (copy_thread_hw_breakpoint(me, p, clone_flags))
+			goto out;
+	}
 	if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) {
 		p->thread.io_bitmap_ptr = kmalloc(IO_BITMAP_BYTES, GFP_KERNEL);
 		if (!p->thread.io_bitmap_ptr) {
@@ -346,6 +356,9 @@ out:
 		kfree(p->thread.io_bitmap_ptr);
 		p->thread.io_bitmap_max = 0;
 	}
+	if (err)
+		flush_thread_hw_breakpoint(p);
+
 	return err;
 }
 
@@ -491,6 +504,26 @@ __switch_to(struct task_struct *prev_p, 
 	 */
 	if (tsk_used_math(next_p) && next_p->fpu_counter > 5)
 		math_state_restore();
+	/*
+	 * There's a problem with moving the switch_to_thread_hw_breakpoint()
+	 * call before current is updated.  Suppose a kernel breakpoint is
+	 * triggered in between the two.  The hw-breakpoint handler will see
+	 * that current is different from the task pointer stored in the chbi
+	 * area, so it will think the task pointer is leftover from an old task
+	 * (lazy switching) and will erase it.  Then until the next context
+	 * switch, no user-breakpoints will be installed.
+	 *
+	 * The real problem is that it's impossible to update both current and
+	 * chbi->bp_task at the same instant, so there will always be a window
+	 * in which they disagree and a breakpoint might get triggered.  Since
+	 * we use lazy switching, we are forced to assume that a disagreement
+	 * means that current is correct and chbi->bp_task is old.  But if you
+	 * move the code above then you'll create a window in which current is
+	 * old and chbi->bp_task is correct.
+	 */
+	if (unlikely(test_tsk_thread_flag(next_p, TIF_DEBUG)))
+		switch_to_thread_hw_breakpoint(next_p);
+
 	return prev_p;
 }
 
author
#7 lore raw

PATCH 7/11 Modify signal handling code to refrain from re-enabling HW Breakpoints

In reply to a message not in the archive

From: Alan Stern <[email protected]>

This patch disables re-enabling of Hardware Breakpoint registers through the signal handling code. This is now done during hw_breakpoint_handler().

---
 arch/x86/kernel/signal.c |    9 ---------
 1 file changed, 9 deletions(-)
arch/x86/kernel/signal.c +0 -9
Index: linux-2.6-tip/arch/x86/kernel/signal.c
===================================================================
@@ -800,15 +800,6 @@ static void do_signal(struct pt_regs *re
 
 	signr = get_signal_to_deliver(&info, &ka, regs, NULL);
 	if (signr > 0) {
-		/*
-		 * Re-enable any watchpoints before delivering the
-		 * signal to user space. The processor register will
-		 * have been cleared if the watchpoint triggered
-		 * inside the kernel.
-		 */
-		if (current->thread.debugreg7)
-			set_debugreg(current->thread.debugreg7, 7);
-
 		/* Whee! Actually deliver the signal.  */
 		if (handle_signal(signr, &info, &ka, oldset, regs) == 0) {
 			/*
author
#8 lore raw

PATCH 8/11 Modify Ptrace routines to access breakpoint registers

In reply to a message not in the archive

From: Alan Stern <[email protected]>

This patch modifies the ptrace code to use the new wrapper routines around the debug/breakpoint registers.

[K.Prasad: Adapted the ptrace routines and to changes post x86/x86_64 merger,
split the minor patch from bigger patch. Re-wrote ptrace_write_dr7() and ptrace_set_debugreg() functions to use new data-structures]

[K.Prasad: Changed code to suit the simplified HW breakpoint implementation]

---
 arch/x86/kernel/ptrace.c |  229 ++++++++++++++++++++++++++++-------------------
 1 file changed, 138 insertions(+), 91 deletions(-)
arch/x86/kernel/ptrace.c +138 -91
Index: linux-2.6-tip/arch/x86/kernel/ptrace.c
===================================================================
@@ -34,6 +34,7 @@
 #include <asm/prctl.h>
 #include <asm/proto.h>
 #include <asm/ds.h>
+#include <asm/hw_breakpoint.h>
 
 #include "tls.h"
 
@@ -134,11 +135,6 @@ static int set_segment_reg(struct task_s
 	return 0;
 }
 
-static unsigned long debugreg_addr_limit(struct task_struct *task)
-{
-	return TASK_SIZE - 3;
-}
-
 #else  /* CONFIG_X86_64 */
 
 #define FLAG_MASK		(FLAG_MASK_32 | X86_EFLAGS_NT)
@@ -263,15 +259,6 @@ static int set_segment_reg(struct task_s
 	return 0;
 }
 
-static unsigned long debugreg_addr_limit(struct task_struct *task)
-{
-#ifdef CONFIG_IA32_EMULATION
-	if (test_tsk_thread_flag(task, TIF_IA32))
-		return IA32_PAGE_OFFSET - 3;
-#endif
-	return TASK_SIZE_MAX - 7;
-}
-
 #endif	/* CONFIG_X86_32 */
 
 static unsigned long get_flags(struct task_struct *task)
@@ -462,95 +449,155 @@ static int genregs_set(struct task_struc
 }
 
 /*
- * This function is trivial and will be inlined by the compiler.
- * Having it separates the implementation details of debug
- * registers from the interface details of ptrace.
+ * Decode the length and type bits for a particular breakpoint as
+ * stored in debug register 7.  Return the "enabled" status.
  */
-static unsigned long ptrace_get_debugreg(struct task_struct *child, int n)
+static int decode_dr7(unsigned long dr7, int bpnum, unsigned *len,
+		unsigned *type)
 {
-	switch (n) {
-	case 0:		return child->thread.debugreg0;
-	case 1:		return child->thread.debugreg1;
-	case 2:		return child->thread.debugreg2;
-	case 3:		return child->thread.debugreg3;
-	case 6:		return child->thread.debugreg6;
-	case 7:		return child->thread.debugreg7;
-	}
-	return 0;
+	int temp = dr7 >> (DR_CONTROL_SHIFT + bpnum * DR_CONTROL_SIZE);
+
+	*len = (temp & 0xc) | 0x40;
+	*type = (temp & 0x3) | 0x80;
+	return (dr7 >> (bpnum * DR_ENABLE_SIZE)) & 0x3;
 }
 
-static int ptrace_set_debugreg(struct task_struct *child,
-			       int n, unsigned long data)
+static void ptrace_triggered(struct hw_breakpoint *bp, struct pt_regs *regs)
 {
+	struct thread_struct *thread = &(current->thread);
 	int i;
 
-	if (unlikely(n == 4 || n == 5))
-		return -EIO;
+	/* Store in the virtual DR6 register the fact that the breakpoint
+	 * was hit so the thread's debugger will see it.
+	 */
+	for (i = 0; hbp_user_refcount[i]; i++)
+		if (bp->info.address == thread->hbp[i]->info.address)
+			break;
 
-	if (n < 4 && unlikely(data >= debugreg_addr_limit(child)))
-		return -EIO;
+	thread->dr6 |= (DR_TRAP0 << i);
+}
 
-	switch (n) {
-	case 0:		child->thread.debugreg0 = data; break;
-	case 1:		child->thread.debugreg1 = data; break;
-	case 2:		child->thread.debugreg2 = data; break;
-	case 3:		child->thread.debugreg3 = data; break;
+/*
+ * Handle ptrace writes to debug register 7.
+ */
+static int ptrace_write_dr7(struct task_struct *tsk, unsigned long data)
+{
+	struct hw_breakpoint *bp;
+	struct thread_struct *thread = &(tsk->thread);
+	int i;
+	int rc = 0;
+	unsigned long old_dr7 = thread->dr7;
 
-	case 6:
-		if ((data & ~0xffffffffUL) != 0)
-			return -EIO;
-		child->thread.debugreg6 = data;
-		break;
+	data &= ~DR_CONTROL_RESERVED;
+	/* Loop through all the hardware breakpoints, making the
+	 * appropriate changes to each.
+	 */
+	thread->dr7 = data;
+	for (i = 0; i < HB_NUM; i++) {
+		int enabled;
+		unsigned len, type;
+
+		bp = thread->hbp[i];
+		if (!bp)
+			continue;
+
+		enabled = decode_dr7(data, i, &len, &type);
+		if (!enabled) {
+			if (bp->triggered)
+				__unregister_user_hw_breakpoint(i, tsk, bp);
+			continue;
+		}
 
-	case 7:
-		/*
-		 * Sanity-check data. Take one half-byte at once with
-		 * check = (val >> (16 + 4*i)) & 0xf. It contains the
-		 * R/Wi and LENi bits; bits 0 and 1 are R/Wi, and bits
-		 * 2 and 3 are LENi. Given a list of invalid values,
-		 * we do mask |= 1 << invalid_value, so that
-		 * (mask >> check) & 1 is a correct test for invalid
-		 * values.
-		 *
-		 * R/Wi contains the type of the breakpoint /
-		 * watchpoint, LENi contains the length of the watched
-		 * data in the watchpoint case.
-		 *
-		 * The invalid values are:
-		 * - LENi == 0x10 (undefined), so mask |= 0x0f00.	[32-bit]
-		 * - R/Wi == 0x10 (break on I/O reads or writes), so
-		 *   mask |= 0x4444.
-		 * - R/Wi == 0x00 && LENi != 0x00, so we have mask |=
-		 *   0x1110.
-		 *
-		 * Finally, mask = 0x0f00 | 0x4444 | 0x1110 == 0x5f54.
-		 *
-		 * See the Intel Manual "System Programming Guide",
-		 * 15.2.4
-		 *
-		 * Note that LENi == 0x10 is defined on x86_64 in long
-		 * mode (i.e. even for 32-bit userspace software, but
-		 * 64-bit kernel), so the x86_64 mask value is 0x5454.
-		 * See the AMD manual no. 24593 (AMD64 System Programming)
-		 */
-#ifdef CONFIG_X86_32
-#define	DR7_MASK	0x5f54
-#else
-#define	DR7_MASK	0x5554
-#endif
-		data &= ~DR_CONTROL_RESERVED;
-		for (i = 0; i < 4; i++)
-			if ((DR7_MASK >> ((data >> (16 + 4*i)) & 0xf)) & 1)
-				return -EIO;
-		child->thread.debugreg7 = data;
-		if (data)
-			set_tsk_thread_flag(child, TIF_DEBUG);
+		if (bp->triggered)
+			rc = __modify_user_hw_breakpoint(i, tsk, bp);
+		else {
+			bp->triggered = ptrace_triggered;
+			bp->info.len = len;
+			bp->info.type = type;
+			rc = __register_user_hw_breakpoint(i, tsk, bp);
+		}
+		if (rc < 0)
+			break;
 		else
-			clear_tsk_thread_flag(child, TIF_DEBUG);
-		break;
+			set_tsk_thread_flag(tsk, TIF_DEBUG);
 	}
+	/* If anything above failed, restore the original settings.
+	 * The original settings will take effect the next time the thread
+	 * is scheduled
+	 */
+	if (rc < 0)
+		data = old_dr7;
+	return rc;
+}
 
-	return 0;
+/*
+ * Handle PTRACE_PEEKUSR calls for the debug register area.
+ */
+unsigned long ptrace_get_debugreg(struct task_struct *tsk, int n)
+{
+	struct thread_struct *thread = &(tsk->thread);
+	unsigned long val = 0;
+
+	mutex_lock(&hw_breakpoint_mutex);
+	if (n < HB_NUM) {
+		if (thread->hbp[n])
+			val = thread->hbp[n]->info.address;
+	} else if (n == 6) {
+		val = thread->dr6;
+	} else if (n == 7) {
+		val = thread->dr7;
+	}
+	mutex_unlock(&hw_breakpoint_mutex);
+	return val;
+}
+
+/*
+ * Handle PTRACE_POKEUSR calls for the debug register area.
+ */
+int ptrace_set_debugreg(struct task_struct *tsk, int n, unsigned long val)
+{
+	struct thread_struct *thread = &(tsk->thread);
+	int rc = -EIO;
+
+	mutex_lock(&hw_breakpoint_mutex);
+
+	/* There are no DR4 or DR5 registers */
+	if (n == 4 || n == 5)
+		goto ret_path;
+
+	/* Writes to DR6 modify the virtualized value */
+	if (n == 6) {
+		tsk->thread.dr6 = val;
+		rc = 0;
+		goto ret_path;
+	}
+
+	/* Writes to DR0 - DR3 change a breakpoint address */
+	rc = 0;
+	if (n < HB_NUM) {
+		if (!val)
+			goto ret_path;
+		if (thread->hbp[n]) {
+			thread->hbp[n]->info.address = val;
+			rc = __modify_user_hw_breakpoint(n, tsk,
+							  thread->hbp[n]);
+			goto ret_path;
+		}
+		thread->hbp[n] = kzalloc(sizeof(struct hw_breakpoint),
+								GFP_KERNEL);
+		if (!thread->hbp[n]) {
+			rc = -ENOMEM;
+			goto ret_path;
+		} else
+			thread->hbp[n]->info.address = val;
+	}
+	/* All that's left is DR7 */
+	if (n == 7)
+		rc = ptrace_write_dr7(tsk, val);
+
+ret_path:
+	mutex_unlock(&hw_breakpoint_mutex);
+	return rc;
 }
 
 /*
@@ -871,7 +918,7 @@ long arch_ptrace(struct task_struct *chi
 		else if (addr >= offsetof(struct user, u_debugreg[0]) &&
 			 addr <= offsetof(struct user, u_debugreg[7])) {
 			addr -= offsetof(struct user, u_debugreg[0]);
-			tmp = ptrace_get_debugreg(child, addr / sizeof(data));
+			tmp = ptrace_get_debugreg(child, addr/sizeof(data));
 		}
 		ret = put_user(tmp, datap);
 		break;
@@ -889,7 +936,7 @@ long arch_ptrace(struct task_struct *chi
 			 addr <= offsetof(struct user, u_debugreg[7])) {
 			addr -= offsetof(struct user, u_debugreg[0]);
 			ret = ptrace_set_debugreg(child,
-						  addr / sizeof(data), data);
+						addr/sizeof(data), data);
 		}
 		break;
 
author
#9 lore raw

PATCH 9/11 Cleanup HW Breakpoint registers before kexec

In reply to a message not in the archive

From: Alan Stern <[email protected]>

This patch disables Hardware breakpoints before doing a 'kexec' on the machine.

[K.Prasad: Split-out from the bigger patch and minor changes following
re-basing]

---
 arch/x86/kernel/machine_kexec_32.c |    2 ++
 arch/x86/kernel/machine_kexec_64.c |    2 ++
 2 files changed, 4 insertions(+)
arch/x86/kernel/machine_kexec_32.c +2 -0
Index: linux-2.6-tip/arch/x86/kernel/machine_kexec_32.c
===================================================================
@@ -25,6 +25,7 @@
 #include <asm/desc.h>
 #include <asm/system.h>
 #include <asm/cacheflush.h>
+#include <asm/debugreg.h>
 
 static void set_idt(void *newidt, __u16 limit)
 {
@@ -202,6 +203,7 @@ void machine_kexec(struct kimage *image)
 
 	/* Interrupts aren't acceptable while we reboot */
 	local_irq_disable();
+	hw_breakpoint_disable();
 
 	if (image->preserve_context) {
 #ifdef CONFIG_X86_IO_APIC
arch/x86/kernel/machine_kexec_64.c +2 -0
Index: linux-2.6-tip/arch/x86/kernel/machine_kexec_64.c
===================================================================
@@ -18,6 +18,7 @@
 #include <asm/pgtable.h>
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
+#include <asm/debugreg.h>
 
 static int init_one_level2_page(struct kimage *image, pgd_t *pgd,
 				unsigned long addr)
@@ -282,6 +283,7 @@ void machine_kexec(struct kimage *image)
 
 	/* Interrupts aren't acceptable while we reboot */
 	local_irq_disable();
+	hw_breakpoint_disable();
 
 	if (image->preserve_context) {
 #ifdef CONFIG_X86_IO_APIC
author
#10 lore raw

PATCH 10/11 Sample HW breakpoint over kernel data address

In reply to a message not in the archive

This patch introduces a sample kernel module to demonstrate the use of Hardware Breakpoint feature. It places a breakpoint over the kernel variable 'pid_max' to monitor all write operations and emits a function-backtrace when done.

---
 samples/Kconfig                         |    6 ++
 samples/Makefile                        |    4 +
 samples/hw_breakpoint/Makefile          |    1 
 samples/hw_breakpoint/data_breakpoint.c |   79 ++++++++++++++++++++++++++++++++
 4 files changed, 89 insertions(+), 1 deletion(-)
samples/Kconfig +6 -0
Index: linux-2.6-tip/samples/Kconfig
===================================================================
@@ -39,5 +39,11 @@ config SAMPLE_KRETPROBES
 	default m
 	depends on SAMPLE_KPROBES && KRETPROBES
 
+config SAMPLE_HW_BREAKPOINT
+	tristate "Build kernel hardware breakpoint examples -- loadable modules only"
+	depends on HAVE_HW_BREAKPOINT && m
+	help
+	  This builds kernel hardware breakpoint example modules.
+
 endif # SAMPLES
 
samples/Makefile +3 -1
Index: linux-2.6-tip/samples/Makefile
===================================================================
@@ -1,3 +1,5 @@
 # Makefile for Linux samples code
 
-obj-$(CONFIG_SAMPLES)	+= markers/ kobject/ kprobes/ tracepoints/
+obj-$(CONFIG_SAMPLES)	+= markers/ kobject/ kprobes/ tracepoints/ \
+			   hw_breakpoint/
+
samples/hw_breakpoint/Makefile new file +1 -0
Index: linux-2.6-tip/samples/hw_breakpoint/Makefile
===================================================================
@@ -0,0 +1 @@
+obj-$(CONFIG_SAMPLE_HW_BREAKPOINT) += data_breakpoint.o
samples/hw_breakpoint/data_breakpoint.c new file +79 -0

79 lines not on this page · raw message

author
#11 lore raw

PATCH 11/11 ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

In reply to a message not in the archive

This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses.

---
 kernel/trace/Kconfig          |   21 +
 kernel/trace/Makefile         |    1 
 kernel/trace/trace.h          |   25 +
 kernel/trace/trace_ksym.c     |  555 ++++++++++++++++++++++++++++++++++++++++++
 kernel/trace/trace_selftest.c |   36 ++
 5 files changed, 638 insertions(+)
kernel/trace/Kconfig +21 -0
Index: linux-2.6-tip/kernel/trace/Kconfig
===================================================================
@@ -268,6 +268,27 @@ config POWER_TRACER
 	  power management decisions, specifically the C-state and P-state
 	  behavior.
 
+config KSYM_TRACER
+	bool "Trace read and write access on kernel memory locations"
+	depends on HAVE_HW_BREAKPOINT
+	select TRACING
+	help
+	  This tracer helps find read and write operations on any given kernel
+	  symbol i.e. /proc/kallsyms.
+
+config PROFILE_KSYM_TRACER
+	bool "Profile all kernel memory accesses on 'watched' variables"
+	depends on KSYM_TRACER
+	help
+	  This tracer profiles kernel accesses on variables watched through the
+	  ksym tracer ftrace plugin. Depending upon the hardware, all read
+	  and write operations on kernel variables can be monitored for
+	  accesses.
+
+	  The results will be displayed in:
+	  /debugfs/tracing/profile_ksym
+
+	  Say N if unsure.
 
 config STACK_TRACER
 	bool "Trace max stack"
kernel/trace/Makefile +1 -0
Index: linux-2.6-tip/kernel/trace/Makefile
===================================================================
@@ -46,5 +46,6 @@ obj-$(CONFIG_EVENT_TRACER) += trace_expo
 obj-$(CONFIG_FTRACE_SYSCALLS) += trace_syscalls.o
 obj-$(CONFIG_EVENT_PROFILE) += trace_event_profile.o
 obj-$(CONFIG_EVENT_TRACER) += trace_events_filter.o
+obj-$(CONFIG_KSYM_TRACER) += trace_ksym.o
 
 libftrace-y := ftrace.o
kernel/trace/trace.h +25 -0
Index: linux-2.6-tip/kernel/trace/trace.h
===================================================================
@@ -12,6 +12,10 @@
 #include <trace/kmemtrace.h>
 #include <trace/power.h>
 
+#ifdef CONFIG_KSYM_TRACER
+#include <asm/hw_breakpoint.h>
+#endif
+
 enum trace_type {
 	__TRACE_FIRST_TYPE = 0,
 
@@ -37,6 +41,7 @@ enum trace_type {
 	TRACE_KMEM_FREE,
 	TRACE_POWER,
 	TRACE_BLK,
+	TRACE_KSYM,
 
 	__TRACE_LAST_TYPE,
 };
@@ -212,6 +217,23 @@ struct syscall_trace_exit {
 	unsigned long		ret;
 };
 
+#ifdef CONFIG_KSYM_TRACER
+struct trace_ksym {
+	struct trace_entry	ent;
+	struct hw_breakpoint	*ksym_hbp;
+	unsigned long		ksym_addr;
+	unsigned long		ip;
+#ifdef CONFIG_PROFILE_KSYM_TRACER
+	unsigned long 		counter;
+#endif
+	struct hlist_node	ksym_hlist;
+	char			ksym_name[KSYM_NAME_LEN];
+	char			p_name[TASK_COMM_LEN];
+};
+#else
+struct trace_ksym {
+};
+#endif /* CONFIG_KSYM_TRACER */
 
 /*
  * trace_flag_type is an enumeration that holds different
@@ -330,6 +352,7 @@ extern void __ftrace_bad_type(void);
 			  TRACE_SYSCALL_ENTER);				\
 		IF_ASSIGN(var, ent, struct syscall_trace_exit,		\
 			  TRACE_SYSCALL_EXIT);				\
+		IF_ASSIGN(var, ent, struct trace_ksym, TRACE_KSYM);	\
 		__ftrace_bad_type();					\
 	} while (0)
 
@@ -593,6 +616,8 @@ extern int trace_selftest_startup_syspro
 					       struct trace_array *tr);
 extern int trace_selftest_startup_branch(struct tracer *trace,
 					 struct trace_array *tr);
+extern int trace_selftest_startup_ksym(struct tracer *trace,
+					 struct trace_array *tr);
 #endif /* CONFIG_FTRACE_STARTUP_TEST */
 
 extern void *head_page(struct trace_array_cpu *data);
kernel/trace/trace_ksym.c new file +555 -0

555 lines not on this page · raw message

kernel/trace/trace_selftest.c +36 -0
Index: linux-2.6-tip/kernel/trace/trace_selftest.c
===================================================================
@@ -16,6 +16,7 @@ static inline int trace_valid_entry(stru
 	case TRACE_BRANCH:
 	case TRACE_GRAPH_ENT:
 	case TRACE_GRAPH_RET:
+	case TRACE_KSYM:
 		return 1;
 	}
 	return 0;
@@ -749,3 +750,38 @@ trace_selftest_startup_branch(struct tra
 	return ret;
 }
 #endif /* CONFIG_BRANCH_TRACER */
+
+#ifdef CONFIG_KSYM_TRACER
+int
+trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr)
+{
+	unsigned long count;
+	int ret;
+
+	/* start the tracing */
+	ret = tracer_init(trace, tr);
+	if (ret) {
+		warn_failed_init_tracer(trace, ret);
+		return ret;
+	}
+
+	/* Sleep for a 1/10 of a second */
+	msleep(100);
+	/* stop the tracing. */
+	tracing_stop();
+	/* check the trace buffer */
+	ret = trace_test_buffer(tr, &count);
+	trace->reset(tr);
+	tracing_start();
+
+	/* read & write operations - one each is performed on the dummy variable
+	 * triggering two entries in the trace buffer
+	 */
+	if (!ret && count != 2) {
+		printk(KERN_CONT "Ksym tracer startup test failed");
+		ret = -1;
+	}
+
+	return ret;
+}
+#endif /* CONFIG_KSYM_TRACER */
maintainer
#12 lore raw

Re: [Patch 11/11] ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

Hi Prasad,

Quoted 11 lines · K.Prasad
kernel/trace/trace_ksym.c new file +11 -0
+
+static void ksym_trace_print_header(struct seq_file *m)
+{
+
+	seq_puts(m,
+		 "#       TASK-PID      CPU#      Symbol         Type    "
+		 "Function         \n");
+	seq_puts(m,
+		 "#          |           |          |              |         "
+		 "|            \n");
+}

kernel/trace/trace_ksym.c line 420

Quoted 10 lines
kernel/trace/trace_ksym.c new file +10 -0
+
+static enum print_line_t ksym_trace_output(struct trace_iterator *iter)
+{
+	struct trace_entry *entry = iter->ent;
+	struct trace_seq *s = &iter->seq;
+	struct trace_ksym *field;
+	char str[KSYM_SYMBOL_LEN];
+	int ret;
+
+	trace_assign_type(field, entry);

Looking at this, you do not test the iter->ent->type. If events or trace_printk is running, this will fail your tracer.

You want to test for your types first, because all trace types will be called for "print_line" function below.

Thus you want something like

	if (entry->type != TRACE_KSYM)
		return TRACE_TYPE_UNHANDLED;

Then the tracer infrastructure could print out the trace event or trace_printk.

If you only want to print your events and ignore all others. Return TRACE_TYPE_HANDLED.

Note, the if needs to be before the trace_assign_type.

-- Steve

Quoted 40 lines
kernel/trace/trace_ksym.c new file +40 -0
+
+	ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->p_name,
+				entry->pid, iter->cpu, field->ksym_name);
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	switch (field->ksym_hbp->info.type) {
+	case HW_BREAKPOINT_WRITE:
+		ret = trace_seq_printf(s, " W  ");
+		break;
+	case HW_BREAKPOINT_RW:
+		ret = trace_seq_printf(s, " RW ");
+		break;
+	default:
+		return TRACE_TYPE_PARTIAL_LINE;
+	}
+
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	sprint_symbol(str, field->ip);
+	ret = trace_seq_printf(s, "%-20s\n", str);
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	return TRACE_TYPE_HANDLED;
+}
+
+struct tracer ksym_tracer __read_mostly =
+{
+	.name		= "ksym_tracer",
+	.init		= ksym_trace_init,
+	.reset		= ksym_trace_reset,
+#ifdef CONFIG_FTRACE_SELFTEST
+	.selftest	= trace_selftest_startup_ksym,
+#endif
+	.print_header   = ksym_trace_print_header,
+	.print_line	= ksym_trace_output
+};
+
author
#13 lore raw

Re: [Patch 11/11] ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

Quoted 41 lines · Steven Rostedt

Hi Prasad,

Quoted 11 lines · K.Prasad
kernel/trace/trace_ksym.c new file +11 -0
+
+static void ksym_trace_print_header(struct seq_file *m)
+{
+
+	seq_puts(m,
+		 "#       TASK-PID      CPU#      Symbol         Type    "
+		 "Function         \n");
+	seq_puts(m,
+		 "#          |           |          |              |         "
+		 "|            \n");
+}

kernel/trace/trace_ksym.c line 420

Quoted 10 lines
kernel/trace/trace_ksym.c new file +10 -0
+
+static enum print_line_t ksym_trace_output(struct trace_iterator *iter)
+{
+	struct trace_entry *entry = iter->ent;
+	struct trace_seq *s = &iter->seq;
+	struct trace_ksym *field;
+	char str[KSYM_SYMBOL_LEN];
+	int ret;
+
+	trace_assign_type(field, entry);

Looking at this, you do not test the iter->ent->type. If events or trace_printk is running, this will fail your tracer.

You want to test for your types first, because all trace types will be called for "print_line" function below.

Thus you want something like

	if (entry->type != TRACE_KSYM)
		return TRACE_TYPE_UNHANDLED;

Then the tracer infrastructure could print out the trace event or trace_printk.

Quoted 7 lines

If you only want to print your events and ignore all others. Return TRACE_TYPE_HANDLED.

Note, the if needs to be before the trace_assign_type.

-- Steve

Thanks for pointing that out. I will make those changes and send out a new patchset.

However, looking at some of the existing plugins I find that many of them do this check after trace_assign_type() or not do at all (e.g. trace_branch_print()). They may need similar changes too.

Thanks,
K.Prasad

maintainer
#14 lore raw

Re: [Patch 11/11] ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

Quoted 7 lines · K.Prasad

Thanks for pointing that out. I will make those changes and send out a new patchset.

However, looking at some of the existing plugins I find that many of them do this check after trace_assign_type() or not do at all (e.g. trace_branch_print()). They may need similar changes too.

Ah, that was the other option I was going to mention. The trace_branch_print is not called by tracer->print_line, like you have. What you did is not wrong, but there are other options.

The other option is to register an event callback. That is, the ftrace infrastructure will call the event registered to an id. I register the trace_branch_event with the TRACE_BRANCH id. Then the ftrace infrastructure will call the trace_branch_print if it encounters the TRACE_BRANCH entry.

This means you can do the UNHANDLED return as I suggested. Or you remove the print_line from you tracer and register an event for TRACE_KSYM instead.

The choice is up to you ;-)

-- Steve

author
#15 lore raw

PATCH 11/11 ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v3

In reply to K.Prasad
Quoted 4 lines · K.Prasad

This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses.

Quoted 9 lines
---
 kernel/trace/Kconfig          |   21 +
 kernel/trace/Makefile         |    1 
 kernel/trace/trace.h          |   25 +
 kernel/trace/trace_ksym.c     |  555 ++++++++++++++++++++++++++++++++++++++++++
 kernel/trace/trace_selftest.c |   36 ++
 5 files changed, 638 insertions(+)

Submitting the revised patch 11/11 and not reposting the entire patchset.

Changelog
---------
- Checking for the appropriate tracer type in the .print_line handler function

Please find the new patch below.

Thanks,
K.Prasad

ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses.

---
 kernel/trace/Kconfig          |   21 +
 kernel/trace/Makefile         |    1 
 kernel/trace/trace.h          |   25 +
 kernel/trace/trace_ksym.c     |  558 ++++++++++++++++++++++++++++++++++++++++++
 kernel/trace/trace_selftest.c |   36 ++
 5 files changed, 641 insertions(+)
kernel/trace/Kconfig +21 -0
Index: linux-2.6-tip/kernel/trace/Kconfig
===================================================================
@@ -268,6 +268,27 @@ config POWER_TRACER
 	  power management decisions, specifically the C-state and P-state
 	  behavior.
 
+config KSYM_TRACER
+	bool "Trace read and write access on kernel memory locations"
+	depends on HAVE_HW_BREAKPOINT
+	select TRACING
+	help
+	  This tracer helps find read and write operations on any given kernel
+	  symbol i.e. /proc/kallsyms.
+
+config PROFILE_KSYM_TRACER
+	bool "Profile all kernel memory accesses on 'watched' variables"
+	depends on KSYM_TRACER
+	help
+	  This tracer profiles kernel accesses on variables watched through the
+	  ksym tracer ftrace plugin. Depending upon the hardware, all read
+	  and write operations on kernel variables can be monitored for
+	  accesses.
+
+	  The results will be displayed in:
+	  /debugfs/tracing/profile_ksym
+
+	  Say N if unsure.
 
 config STACK_TRACER
 	bool "Trace max stack"
kernel/trace/Makefile +1 -0
Index: linux-2.6-tip/kernel/trace/Makefile
===================================================================
@@ -46,5 +46,6 @@ obj-$(CONFIG_EVENT_TRACER) += trace_expo
 obj-$(CONFIG_FTRACE_SYSCALLS) += trace_syscalls.o
 obj-$(CONFIG_EVENT_PROFILE) += trace_event_profile.o
 obj-$(CONFIG_EVENT_TRACER) += trace_events_filter.o
+obj-$(CONFIG_KSYM_TRACER) += trace_ksym.o
 
 libftrace-y := ftrace.o
kernel/trace/trace.h +25 -0
Index: linux-2.6-tip/kernel/trace/trace.h
===================================================================
@@ -12,6 +12,10 @@
 #include <trace/kmemtrace.h>
 #include <trace/power.h>
 
+#ifdef CONFIG_KSYM_TRACER
+#include <asm/hw_breakpoint.h>
+#endif
+
 enum trace_type {
 	__TRACE_FIRST_TYPE = 0,
 
@@ -37,6 +41,7 @@ enum trace_type {
 	TRACE_KMEM_FREE,
 	TRACE_POWER,
 	TRACE_BLK,
+	TRACE_KSYM,
 
 	__TRACE_LAST_TYPE,
 };
@@ -212,6 +217,23 @@ struct syscall_trace_exit {
 	unsigned long		ret;
 };
 
+#ifdef CONFIG_KSYM_TRACER
+struct trace_ksym {
+	struct trace_entry	ent;
+	struct hw_breakpoint	*ksym_hbp;
+	unsigned long		ksym_addr;
+	unsigned long		ip;
+#ifdef CONFIG_PROFILE_KSYM_TRACER
+	unsigned long 		counter;
+#endif
+	struct hlist_node	ksym_hlist;
+	char			ksym_name[KSYM_NAME_LEN];
+	char			p_name[TASK_COMM_LEN];
+};
+#else
+struct trace_ksym {
+};
+#endif /* CONFIG_KSYM_TRACER */
 
 /*
  * trace_flag_type is an enumeration that holds different
@@ -330,6 +352,7 @@ extern void __ftrace_bad_type(void);
 			  TRACE_SYSCALL_ENTER);				\
 		IF_ASSIGN(var, ent, struct syscall_trace_exit,		\
 			  TRACE_SYSCALL_EXIT);				\
+		IF_ASSIGN(var, ent, struct trace_ksym, TRACE_KSYM);	\
 		__ftrace_bad_type();					\
 	} while (0)
 
@@ -593,6 +616,8 @@ extern int trace_selftest_startup_syspro
 					       struct trace_array *tr);
 extern int trace_selftest_startup_branch(struct tracer *trace,
 					 struct trace_array *tr);
+extern int trace_selftest_startup_ksym(struct tracer *trace,
+					 struct trace_array *tr);
 #endif /* CONFIG_FTRACE_STARTUP_TEST */
 
 extern void *head_page(struct trace_array_cpu *data);
kernel/trace/trace_ksym.c new file +558 -0

558 lines not on this page · raw message

kernel/trace/trace_selftest.c +36 -0
Index: linux-2.6-tip/kernel/trace/trace_selftest.c
===================================================================
@@ -16,6 +16,7 @@ static inline int trace_valid_entry(stru
 	case TRACE_BRANCH:
 	case TRACE_GRAPH_ENT:
 	case TRACE_GRAPH_RET:
+	case TRACE_KSYM:
 		return 1;
 	}
 	return 0;
@@ -749,3 +750,38 @@ trace_selftest_startup_branch(struct tra
 	return ret;
 }
 #endif /* CONFIG_BRANCH_TRACER */
+
+#ifdef CONFIG_KSYM_TRACER
+int
+trace_selftest_startup_ksym(struct tracer *trace, struct trace_array *tr)
+{
+	unsigned long count;
+	int ret;
+
+	/* start the tracing */
+	ret = tracer_init(trace, tr);
+	if (ret) {
+		warn_failed_init_tracer(trace, ret);
+		return ret;
+	}
+
+	/* Sleep for a 1/10 of a second */
+	msleep(100);
+	/* stop the tracing. */
+	tracing_stop();
+	/* check the trace buffer */
+	ret = trace_test_buffer(tr, &count);
+	trace->reset(tr);
+	tracing_start();
+
+	/* read & write operations - one each is performed on the dummy variable
+	 * triggering two entries in the trace buffer
+	 */
+	if (!ret && count != 2) {
+		printk(KERN_CONT "Ksym tracer startup test failed");
+		ret = -1;
+	}
+
+	return ret;
+}
+#endif /* CONFIG_KSYM_TRACER */
maintainer Acked-by on v1 11/11
#16 lore raw

Re: [Patch 11/11] ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v3

Quoted 43 lines · K.Prasad
Quoted 4 lines · K.Prasad

This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses.

Quoted 9 lines
---
 kernel/trace/Kconfig          |   21 +
 kernel/trace/Makefile         |    1 
 kernel/trace/trace.h          |   25 +
 kernel/trace/trace_ksym.c     |  555 ++++++++++++++++++++++++++++++++++++++++++
 kernel/trace/trace_selftest.c |   36 ++
 5 files changed, 638 insertions(+)

Submitting the revised patch 11/11 and not reposting the entire patchset.

Changelog
---------
- Checking for the appropriate tracer type in the .print_line handler function

Please find the new patch below.

Thanks,
K.Prasad

ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses.

---
 kernel/trace/Kconfig          |   21 +
 kernel/trace/Makefile         |    1 
 kernel/trace/trace.h          |   25 +
 kernel/trace/trace_ksym.c     |  558 ++++++++++++++++++++++++++++++++++++++++++
 kernel/trace/trace_selftest.c |   36 ++
 5 files changed, 641 insertions(+)

kernel/trace/trace_ksym.c line 421

Quoted 10 lines
kernel/trace/trace_ksym.c new file +10 -0
+static enum print_line_t ksym_trace_output(struct trace_iterator *iter)
+{
+	struct trace_entry *entry = iter->ent;
+	struct trace_seq *s = &iter->seq;
+	struct trace_ksym *field;
+	char str[KSYM_SYMBOL_LEN];
+	int ret;
+
+	if (entry->type != TRACE_KSYM)
+		return TRACE_TYPE_UNHANDLED;

Looks good.

Thanks,

-- Steve

Quoted 30 lines
kernel/trace/trace_ksym.c new file +30 -0
+
+	trace_assign_type(field, entry);
+
+	ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->p_name,
+				entry->pid, iter->cpu, field->ksym_name);
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	switch (field->ksym_hbp->info.type) {
+	case HW_BREAKPOINT_WRITE:
+		ret = trace_seq_printf(s, " W  ");
+		break;
+	case HW_BREAKPOINT_RW:
+		ret = trace_seq_printf(s, " RW ");
+		break;
+	default:
+		return TRACE_TYPE_PARTIAL_LINE;
+	}
+
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	sprint_symbol(str, field->ip);
+	ret = trace_seq_printf(s, "%-20s\n", str);
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	return TRACE_TYPE_HANDLED;
+}
+
#17 lore raw

Re: [Patch 11/11] ftrace plugin for kernel symbol tracing using HW Breakpoint interfaces - v2

In reply to K.Prasad

Hi!

Quoted 15 lines

This patch adds an ftrace plugin to detect and profile memory access over kernel variables. It uses HW Breakpoint interfaces to 'watch memory addresses.

kernel/trace/trace_ksym.c new file +10 -0
+static void ksym_trace_print_header(struct seq_file *m)
+{
+
+	seq_puts(m,
+		 "#       TASK-PID      CPU#      Symbol         Type    "
+		 "Function         \n");
+	seq_puts(m,
+		 "#          |           |          |              |         "
+		 "|            \n");
+}

...

Quoted 24 lines
kernel/trace/trace_ksym.c new file +24 -0
+	trace_assign_type(field, entry);
+
+	ret = trace_seq_printf(s, "%-15s %-5d %-3d %-20s ", field->p_name,
+				entry->pid, iter->cpu, field->ksym_name);
+	if (!ret)
+		return TRACE_TYPE_PARTIAL_LINE;
+
+	switch (field->ksym_hbp->info.type) {
+	case HW_BREAKPOINT_WRITE:
+		ret = trace_seq_printf(s, " W  ");
+		break;
+	case HW_BREAKPOINT_RW:
+		ret = trace_seq_printf(s, " RW ");
+		break;
+	default:
+		return TRACE_TYPE_PARTIAL_LINE;
+	}
+#ifdef CONFIG_PROFILE_KSYM_TRACER
+static int ksym_tracer_stat_headers(struct seq_file *m)
+{
+	seq_printf(m, "   Access type    ");
+	seq_printf(m, "            Symbol                     Counter     \n");
+	return 0;
+}

...

Quoted 9 lines
kernel/trace/trace_ksym.c new file +9 -0
+	switch (access_type) {
+	case HW_BREAKPOINT_WRITE:
+		seq_printf(m, "     W     ");
+		break;
+	case HW_BREAKPOINT_RW:
+		seq_printf(m, "     RW    ");
+		break;
+	default:
+		seq_printf(m, "     NA    ");

kernel/trace/trace_ksym.c line 518

Quoted 10 lines
kernel/trace/trace_ksym.c new file +10 -0
+	}
+
+	if (lookup_symbol_name(entry->ksym_addr, fn_name) >= 0)
+		seq_printf(m, "               %s                 ", fn_name);
+	else
+		seq_printf(m, "               <NA>                ");
+
+	seq_printf(m, "%15lu\n", entry->counter);
+	return 0;
+}

Quite as lot of ascii-art here. Is userland expected to parse it? If so, how? (docs would be nice) What if you'll want to make fields wider one day?

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html