[lttng-modules.git] / lttng-statedump-impl.c

/*
 * lttng-statedump.c
 *
 * Linux Trace Toolkit Next Generation Kernel State Dump
 *
 * Copyright 2005 Jean-Hugues Deschenes <jean-hugues.deschenes@polymtl.ca>
 * Copyright 2006-2012 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; only
 * version 2.1 of the License.
 *
 * This library 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
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * Changes:
 *	Eric Clement:                   Add listing of network IP interface
 *	2006, 2007 Mathieu Desnoyers	Fix kernel threads
 *	                                Various updates
 */

#include <linux/init.h>
#include <linux/module.h>
#include <linux/netlink.h>
#include <linux/inet.h>
#include <linux/ip.h>
#include <linux/kthread.h>
#include <linux/proc_fs.h>
#include <linux/file.h>
#include <linux/interrupt.h>
#include <linux/irqnr.h>
#include <linux/cpu.h>
#include <linux/netdevice.h>
#include <linux/inetdevice.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/wait.h>
#include <linux/mutex.h>
#include <linux/device.h>

#include <lttng-events.h>
#include <lttng-tracer.h>
#include <wrapper/irqdesc.h>
#include <wrapper/spinlock.h>
#include <wrapper/fdtable.h>
#include <wrapper/irq.h>
#include <wrapper/tracepoint.h>
#include <wrapper/genhd.h>
#include <wrapper/file.h>
#include <wrapper/fdtable.h>

#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
#include <linux/irq.h>
#endif

/* Define the tracepoints, but do not build the probes */
#define CREATE_TRACE_POINTS
#define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
#define TRACE_INCLUDE_FILE lttng-statedump
#define LTTNG_INSTRUMENTATION
#include <instrumentation/events/lttng-module/lttng-statedump.h>

DEFINE_TRACE(lttng_statedump_block_device);
DEFINE_TRACE(lttng_statedump_end);
DEFINE_TRACE(lttng_statedump_interrupt);
DEFINE_TRACE(lttng_statedump_file_descriptor);
DEFINE_TRACE(lttng_statedump_start);
DEFINE_TRACE(lttng_statedump_process_state);
DEFINE_TRACE(lttng_statedump_network_interface);

struct lttng_fd_ctx {
	char *page;
	struct lttng_session *session;
	struct task_struct *p;
	struct files_struct *files;
};

/*
 * Protected by the trace lock.
 */
static struct delayed_work cpu_work[NR_CPUS];
static DECLARE_WAIT_QUEUE_HEAD(statedump_wq);
static atomic_t kernel_threads_to_run;

enum lttng_thread_type {
	LTTNG_USER_THREAD = 0,
	LTTNG_KERNEL_THREAD = 1,
};

enum lttng_execution_mode {
	LTTNG_USER_MODE = 0,
	LTTNG_SYSCALL = 1,
	LTTNG_TRAP = 2,
	LTTNG_IRQ = 3,
	LTTNG_SOFTIRQ = 4,
	LTTNG_MODE_UNKNOWN = 5,
};

enum lttng_execution_submode {
	LTTNG_NONE = 0,
	LTTNG_UNKNOWN = 1,
};

enum lttng_process_status {
	LTTNG_UNNAMED = 0,
	LTTNG_WAIT_FORK = 1,
	LTTNG_WAIT_CPU = 2,
	LTTNG_EXIT = 3,
	LTTNG_ZOMBIE = 4,
	LTTNG_WAIT = 5,
	LTTNG_RUN = 6,
	LTTNG_DEAD = 7,
};

static
int lttng_enumerate_block_devices(struct lttng_session *session)
{
	struct class *ptr_block_class;
	struct device_type *ptr_disk_type;
	struct class_dev_iter iter;
	struct device *dev;

	ptr_block_class = wrapper_get_block_class();
	if (!ptr_block_class)
		return -ENOSYS;
	ptr_disk_type = wrapper_get_disk_type();
	if (!ptr_disk_type) {
		return -ENOSYS;
	}
	class_dev_iter_init(&iter, ptr_block_class, NULL, ptr_disk_type);
	while ((dev = class_dev_iter_next(&iter))) {
		struct disk_part_iter piter;
		struct gendisk *disk = dev_to_disk(dev);
		struct hd_struct *part;

		/*
		 * Don't show empty devices or things that have been
		 * suppressed
		 */
		if (get_capacity(disk) == 0 ||
		    (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
			continue;

		disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
		while ((part = disk_part_iter_next(&piter))) {
			char name_buf[BDEVNAME_SIZE];
			char *p;

			p = wrapper_disk_name(disk, part->partno, name_buf);
			if (!p) {
				disk_part_iter_exit(&piter);
				class_dev_iter_exit(&iter);
				return -ENOSYS;
			}
			trace_lttng_statedump_block_device(session,
					part_devt(part), name_buf);
		}
		disk_part_iter_exit(&piter);
	}
	class_dev_iter_exit(&iter);
	return 0;
}

#ifdef CONFIG_INET

static
void lttng_enumerate_device(struct lttng_session *session,
		struct net_device *dev)
{
	struct in_device *in_dev;
	struct in_ifaddr *ifa;

	if (dev->flags & IFF_UP) {
		in_dev = in_dev_get(dev);
		if (in_dev) {
			for (ifa = in_dev->ifa_list; ifa != NULL;
			     ifa = ifa->ifa_next) {
				trace_lttng_statedump_network_interface(
					session, dev, ifa);
			}
			in_dev_put(in_dev);
		}
	} else {
		trace_lttng_statedump_network_interface(
			session, dev, NULL);
	}
}

static
int lttng_enumerate_network_ip_interface(struct lttng_session *session)
{
	struct net_device *dev;

	read_lock(&dev_base_lock);
	for_each_netdev(&init_net, dev)
		lttng_enumerate_device(session, dev);
	read_unlock(&dev_base_lock);

	return 0;
}
#else /* CONFIG_INET */
static inline
int lttng_enumerate_network_ip_interface(struct lttng_session *session)
{
	return 0;
}
#endif /* CONFIG_INET */

static
int lttng_dump_one_fd(const void *p, struct file *file, unsigned int fd)
{
	const struct lttng_fd_ctx *ctx = p;
	const char *s = d_path(&file->f_path, ctx->page, PAGE_SIZE);
	unsigned int flags = file->f_flags;
	struct fdtable *fdt;

	/*
	 * We don't expose kernel internal flags, only userspace-visible
	 * flags.
	 */
	flags &= ~FMODE_NONOTIFY;
	fdt = files_fdtable(ctx->files);
	/*
	 * We need to check here again whether fd is within the fdt
	 * max_fds range, because we might be seeing a different
	 * files_fdtable() than iterate_fd(), assuming only RCU is
	 * protecting the read. In reality, iterate_fd() holds
	 * file_lock, which should ensure the fdt does not change while
	 * the lock is taken, but we are not aware whether this is
	 * guaranteed or not, so play safe.
	 */
	if (fd < fdt->max_fds && lttng_close_on_exec(fd, fdt))
		flags |= O_CLOEXEC;
	if (IS_ERR(s)) {
		struct dentry *dentry = file->f_path.dentry;

		/* Make sure we give at least some info */
		spin_lock(&dentry->d_lock);
		trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd,
			dentry->d_name.name, flags, file->f_mode);
		spin_unlock(&dentry->d_lock);
		goto end;
	}
	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd, s,
		flags, file->f_mode);
end:
	return 0;
}

static
void lttng_enumerate_task_fd(struct lttng_session *session,
		struct task_struct *p, char *tmp)
{
	struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .p = p };
	struct files_struct *files;

	task_lock(p);
	files = p->files;
	if (!files)
		goto end;
	ctx.files = files;
	lttng_iterate_fd(files, 0, lttng_dump_one_fd, &ctx);
end:
	task_unlock(p);
}

static
int lttng_enumerate_file_descriptors(struct lttng_session *session)
{
	struct task_struct *p;
	char *tmp;

	tmp = (char *) __get_free_page(GFP_KERNEL);
	if (!tmp)
		return -ENOMEM;

	/* Enumerate active file descriptors */
	rcu_read_lock();
	for_each_process(p)
		lttng_enumerate_task_fd(session, p, tmp);
	rcu_read_unlock();
	free_page((unsigned long) tmp);
	return 0;
}

#if 0
/*
 * FIXME: we cannot take a mmap_sem while in a RCU read-side critical section
 * (scheduling in atomic). Normally, the tasklist lock protects this kind of
 * iteration, but it is not exported to modules.
 */
static
void lttng_enumerate_task_vm_maps(struct lttng_session *session,
		struct task_struct *p)
{
	struct mm_struct *mm;
	struct vm_area_struct *map;
	unsigned long ino;

	/* get_task_mm does a task_lock... */
	mm = get_task_mm(p);
	if (!mm)
		return;

	map = mm->mmap;
	if (map) {
		down_read(&mm->mmap_sem);
		while (map) {
			if (map->vm_file)
				ino = map->vm_file->lttng_f_dentry->d_inode->i_ino;
			else
				ino = 0;
			trace_lttng_statedump_vm_map(session, p, map, ino);
			map = map->vm_next;
		}
		up_read(&mm->mmap_sem);
	}
	mmput(mm);
}

static
int lttng_enumerate_vm_maps(struct lttng_session *session)
{
	struct task_struct *p;

	rcu_read_lock();
	for_each_process(p)
		lttng_enumerate_task_vm_maps(session, p);
	rcu_read_unlock();
	return 0;
}
#endif

#ifdef CONFIG_LTTNG_HAS_LIST_IRQ

#if (LTTNG_LINUX_VERSION_CODE < LTTNG_KERNEL_VERSION(2,6,39))
#define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
#endif

static
int lttng_list_interrupts(struct lttng_session *session)
{
	unsigned int irq;
	unsigned long flags = 0;
	struct irq_desc *desc;

#define irq_to_desc	wrapper_irq_to_desc
	/* needs irq_desc */
	for_each_irq_desc(irq, desc) {
		struct irqaction *action;
		const char *irq_chip_name =
			irq_desc_get_chip(desc)->name ? : "unnamed_irq_chip";

		local_irq_save(flags);
		wrapper_desc_spin_lock(&desc->lock);
		for (action = desc->action; action; action = action->next) {
			trace_lttng_statedump_interrupt(session,
				irq, irq_chip_name, action);
		}
		wrapper_desc_spin_unlock(&desc->lock);
		local_irq_restore(flags);
	}
	return 0;
#undef irq_to_desc
}
#else
static inline
int lttng_list_interrupts(struct lttng_session *session)
{
	return 0;
}
#endif

/*
 * Called with task lock held.
 */
static
void lttng_statedump_process_ns(struct lttng_session *session,
		struct task_struct *p,
		enum lttng_thread_type type,
		enum lttng_execution_mode mode,
		enum lttng_execution_submode submode,
		enum lttng_process_status status)
{
	struct pid_namespace *pid_ns;

	pid_ns = task_active_pid_ns(p);
	do {
		trace_lttng_statedump_process_state(session,
			p, type, mode, submode, status, pid_ns);
		pid_ns = pid_ns ? pid_ns->parent : NULL;
	} while (pid_ns);
}

static
int lttng_enumerate_process_states(struct lttng_session *session)
{
	struct task_struct *g, *p;

	rcu_read_lock();
	for_each_process(g) {
		p = g;
		do {
			enum lttng_execution_mode mode =
				LTTNG_MODE_UNKNOWN;
			enum lttng_execution_submode submode =
				LTTNG_UNKNOWN;
			enum lttng_process_status status;
			enum lttng_thread_type type;

			task_lock(p);
			if (p->exit_state == EXIT_ZOMBIE)
				status = LTTNG_ZOMBIE;
			else if (p->exit_state == EXIT_DEAD)
				status = LTTNG_DEAD;
			else if (p->state == TASK_RUNNING) {
				/* Is this a forked child that has not run yet? */
				if (list_empty(&p->rt.run_list))
					status = LTTNG_WAIT_FORK;
				else
					/*
					 * All tasks are considered as wait_cpu;
					 * the viewer will sort out if the task
					 * was really running at this time.
					 */
					status = LTTNG_WAIT_CPU;
			} else if (p->state &
				(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)) {
				/* Task is waiting for something to complete */
				status = LTTNG_WAIT;
			} else
				status = LTTNG_UNNAMED;
			submode = LTTNG_NONE;

			/*
			 * Verification of t->mm is to filter out kernel
			 * threads; Viewer will further filter out if a
			 * user-space thread was in syscall mode or not.
			 */
			if (p->mm)
				type = LTTNG_USER_THREAD;
			else
				type = LTTNG_KERNEL_THREAD;
			lttng_statedump_process_ns(session,
				p, type, mode, submode, status);
			task_unlock(p);
		} while_each_thread(g, p);
	}
	rcu_read_unlock();

	return 0;
}

static
void lttng_statedump_work_func(struct work_struct *work)
{
	if (atomic_dec_and_test(&kernel_threads_to_run))
		/* If we are the last thread, wake up do_lttng_statedump */
		wake_up(&statedump_wq);
}

static
int do_lttng_statedump(struct lttng_session *session)
{
	int cpu, ret;

	trace_lttng_statedump_start(session);
	ret = lttng_enumerate_process_states(session);
	if (ret)
		return ret;
	ret = lttng_enumerate_file_descriptors(session);
	if (ret)
		return ret;
	/*
	 * FIXME
	 * ret = lttng_enumerate_vm_maps(session);
	 * if (ret)
	 * 	return ret;
	 */
	ret = lttng_list_interrupts(session);
	if (ret)
		return ret;
	ret = lttng_enumerate_network_ip_interface(session);
	if (ret)
		return ret;
	ret = lttng_enumerate_block_devices(session);
	switch (ret) {
	case 0:
		break;
	case -ENOSYS:
		printk(KERN_WARNING "LTTng: block device enumeration is not supported by kernel\n");
		break;
	default:
		return ret;
	}

	/* TODO lttng_dump_idt_table(session); */
	/* TODO lttng_dump_softirq_vec(session); */
	/* TODO lttng_list_modules(session); */
	/* TODO lttng_dump_swap_files(session); */

	/*
	 * Fire off a work queue on each CPU. Their sole purpose in life
	 * is to guarantee that each CPU has been in a state where is was in
	 * syscall mode (i.e. not in a trap, an IRQ or a soft IRQ).
	 */
	get_online_cpus();
	atomic_set(&kernel_threads_to_run, num_online_cpus());
	for_each_online_cpu(cpu) {
		INIT_DELAYED_WORK(&cpu_work[cpu], lttng_statedump_work_func);
		schedule_delayed_work_on(cpu, &cpu_work[cpu], 0);
	}
	/* Wait for all threads to run */
	__wait_event(statedump_wq, (atomic_read(&kernel_threads_to_run) == 0));
	put_online_cpus();
	/* Our work is done */
	trace_lttng_statedump_end(session);
	return 0;
}

/*
 * Called with session mutex held.
 */
int lttng_statedump_start(struct lttng_session *session)
{
	return do_lttng_statedump(session);
}
EXPORT_SYMBOL_GPL(lttng_statedump_start);

static
int __init lttng_statedump_init(void)
{
	/*
	 * Allow module to load even if the fixup cannot be done. This
	 * will allow seemless transition when the underlying issue fix
	 * is merged into the Linux kernel, and when tracepoint.c
	 * "tracepoint_module_notify" is turned into a static function.
	 */
	(void) wrapper_lttng_fixup_sig(THIS_MODULE);
	return 0;
}

module_init(lttng_statedump_init);

static
void __exit lttng_statedump_exit(void)
{
}

module_exit(lttng_statedump_exit);

MODULE_LICENSE("GPL and additional rights");
MODULE_AUTHOR("Jean-Hugues Deschenes");
MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Statedump");
MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
	__stringify(LTTNG_MODULES_MINOR_VERSION) "."
	__stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
	LTTNG_MODULES_EXTRAVERSION);
Commit	Line	Data
c337ddc2	1	/*
886d51a3 MD	2	* lttng-statedump.c
886d51a3 MD	3	*
c337ddc2 MD	4	* Linux Trace Toolkit Next Generation Kernel State Dump
	5	*
	6	* Copyright 2005 Jean-Hugues Deschenes <jean-hugues.deschenes@polymtl.ca>
	7	* Copyright 2006-2012 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
	8	*
886d51a3 MD	9	* This library is free software; you can redistribute it and/or
	10	* modify it under the terms of the GNU Lesser General Public
	11	* License as published by the Free Software Foundation; only
	12	* version 2.1 of the License.
	13	*
	14	* This library is distributed in the hope that it will be useful,
	15	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	16	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	17	* Lesser General Public License for more details.
	18	*
	19	* You should have received a copy of the GNU Lesser General Public
	20	* License along with this library; if not, write to the Free Software
	21	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
	22	*
c337ddc2 MD	23	* Changes:
	24	* Eric Clement: Add listing of network IP interface
	25	* 2006, 2007 Mathieu Desnoyers Fix kernel threads
	26	* Various updates
c337ddc2 MD	27	*/
	28
	29	#include <linux/init.h>
	30	#include <linux/module.h>
	31	#include <linux/netlink.h>
	32	#include <linux/inet.h>
	33	#include <linux/ip.h>
	34	#include <linux/kthread.h>
	35	#include <linux/proc_fs.h>
	36	#include <linux/file.h>
	37	#include <linux/interrupt.h>
	38	#include <linux/irqnr.h>
	39	#include <linux/cpu.h>
	40	#include <linux/netdevice.h>
	41	#include <linux/inetdevice.h>
	42	#include <linux/sched.h>
	43	#include <linux/mm.h>
c337ddc2 MD	44	#include <linux/swap.h>
	45	#include <linux/wait.h>
	46	#include <linux/mutex.h>
f0dbdefb	47	#include <linux/device.h>
c337ddc2	48
241ae9a8 MD	49	#include <lttng-events.h>
	50	#include <lttng-tracer.h>
	51	#include <wrapper/irqdesc.h>
	52	#include <wrapper/spinlock.h>
	53	#include <wrapper/fdtable.h>
241ae9a8 MD	54	#include <wrapper/irq.h>
	55	#include <wrapper/tracepoint.h>
	56	#include <wrapper/genhd.h>
	57	#include <wrapper/file.h>
dd970ee8	58	#include <wrapper/fdtable.h>
c337ddc2	59
29784493	60	#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
c337ddc2 MD	61	#include <linux/irq.h>
	62	#endif
	63
	64	/* Define the tracepoints, but do not build the probes */
	65	#define CREATE_TRACE_POINTS
241ae9a8	66	#define TRACE_INCLUDE_PATH instrumentation/events/lttng-module
c337ddc2	67	#define TRACE_INCLUDE_FILE lttng-statedump
3bc29f0a	68	#define LTTNG_INSTRUMENTATION
241ae9a8	69	#include <instrumentation/events/lttng-module/lttng-statedump.h>
c337ddc2	70
f0dbdefb	71	DEFINE_TRACE(lttng_statedump_block_device);
20591cf7 MD	72	DEFINE_TRACE(lttng_statedump_end);
	73	DEFINE_TRACE(lttng_statedump_interrupt);
	74	DEFINE_TRACE(lttng_statedump_file_descriptor);
	75	DEFINE_TRACE(lttng_statedump_start);
	76	DEFINE_TRACE(lttng_statedump_process_state);
	77	DEFINE_TRACE(lttng_statedump_network_interface);
	78
361c023a MD	79	struct lttng_fd_ctx {
	80	char *page;
	81	struct lttng_session *session;
	82	struct task_struct *p;
d561ecfb	83	struct files_struct *files;
361c023a MD	84	};
361c023a MD	85
c337ddc2 MD	86	/*
	87	* Protected by the trace lock.
	88	*/
	89	static struct delayed_work cpu_work[NR_CPUS];
	90	static DECLARE_WAIT_QUEUE_HEAD(statedump_wq);
	91	static atomic_t kernel_threads_to_run;
	92
	93	enum lttng_thread_type {
	94	LTTNG_USER_THREAD = 0,
	95	LTTNG_KERNEL_THREAD = 1,
	96	};
	97
	98	enum lttng_execution_mode {
	99	LTTNG_USER_MODE = 0,
	100	LTTNG_SYSCALL = 1,
	101	LTTNG_TRAP = 2,
	102	LTTNG_IRQ = 3,
	103	LTTNG_SOFTIRQ = 4,
	104	LTTNG_MODE_UNKNOWN = 5,
	105	};
	106
	107	enum lttng_execution_submode {
	108	LTTNG_NONE = 0,
	109	LTTNG_UNKNOWN = 1,
	110	};
	111
	112	enum lttng_process_status {
	113	LTTNG_UNNAMED = 0,
	114	LTTNG_WAIT_FORK = 1,
	115	LTTNG_WAIT_CPU = 2,
	116	LTTNG_EXIT = 3,
	117	LTTNG_ZOMBIE = 4,
	118	LTTNG_WAIT = 5,
	119	LTTNG_RUN = 6,
	120	LTTNG_DEAD = 7,
	121	};
	122
f0dbdefb HD	123	static
	124	int lttng_enumerate_block_devices(struct lttng_session *session)
	125	{
	126	struct class *ptr_block_class;
	127	struct device_type *ptr_disk_type;
	128	struct class_dev_iter iter;
	129	struct device *dev;
	130
	131	ptr_block_class = wrapper_get_block_class();
	132	if (!ptr_block_class)
	133	return -ENOSYS;
	134	ptr_disk_type = wrapper_get_disk_type();
	135	if (!ptr_disk_type) {
	136	return -ENOSYS;
	137	}
	138	class_dev_iter_init(&iter, ptr_block_class, NULL, ptr_disk_type);
	139	while ((dev = class_dev_iter_next(&iter))) {
	140	struct disk_part_iter piter;
	141	struct gendisk *disk = dev_to_disk(dev);
	142	struct hd_struct *part;
	143
5a91f3df MD	144	/*
	145	* Don't show empty devices or things that have been
	146	* suppressed
	147	*/
	148	if (get_capacity(disk) == 0 \|\|
	149	(disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
	150	continue;
	151
f0dbdefb HD	152	disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
	153	while ((part = disk_part_iter_next(&piter))) {
	154	char name_buf[BDEVNAME_SIZE];
	155	char *p;
	156
	157	p = wrapper_disk_name(disk, part->partno, name_buf);
	158	if (!p) {
	159	disk_part_iter_exit(&piter);
	160	class_dev_iter_exit(&iter);
	161	return -ENOSYS;
	162	}
	163	trace_lttng_statedump_block_device(session,
	164	part_devt(part), name_buf);
	165	}
	166	disk_part_iter_exit(&piter);
	167	}
	168	class_dev_iter_exit(&iter);
	169	return 0;
	170	}
	171
c337ddc2	172	#ifdef CONFIG_INET
f0dbdefb	173
c337ddc2 MD	174	static
	175	void lttng_enumerate_device(struct lttng_session *session,
	176	struct net_device *dev)
	177	{
	178	struct in_device *in_dev;
	179	struct in_ifaddr *ifa;
	180
	181	if (dev->flags & IFF_UP) {
	182	in_dev = in_dev_get(dev);
	183	if (in_dev) {
	184	for (ifa = in_dev->ifa_list; ifa != NULL;
	185	ifa = ifa->ifa_next) {
	186	trace_lttng_statedump_network_interface(
	187	session, dev, ifa);
	188	}
	189	in_dev_put(in_dev);
	190	}
	191	} else {
	192	trace_lttng_statedump_network_interface(
	193	session, dev, NULL);
	194	}
	195	}
	196
	197	static
	198	int lttng_enumerate_network_ip_interface(struct lttng_session *session)
	199	{
	200	struct net_device *dev;
	201
	202	read_lock(&dev_base_lock);
	203	for_each_netdev(&init_net, dev)
	204	lttng_enumerate_device(session, dev);
	205	read_unlock(&dev_base_lock);
	206
	207	return 0;
	208	}
	209	#else /* CONFIG_INET */
	210	static inline
	211	int lttng_enumerate_network_ip_interface(struct lttng_session *session)
	212	{
	213	return 0;
	214	}
	215	#endif /* CONFIG_INET */
	216
361c023a MD	217	static
	218	int lttng_dump_one_fd(const void p, struct file file, unsigned int fd)
	219	{
	220	const struct lttng_fd_ctx *ctx = p;
	221	const char *s = d_path(&file->f_path, ctx->page, PAGE_SIZE);
29021503	222	unsigned int flags = file->f_flags;
d561ecfb	223	struct fdtable *fdt;
361c023a	224
29021503 MD	225	/*
	226	* We don't expose kernel internal flags, only userspace-visible
	227	* flags.
	228	*/
	229	flags &= ~FMODE_NONOTIFY;
d561ecfb MD	230	fdt = files_fdtable(ctx->files);
	231	/*
	232	* We need to check here again whether fd is within the fdt
	233	* max_fds range, because we might be seeing a different
	234	* files_fdtable() than iterate_fd(), assuming only RCU is
	235	* protecting the read. In reality, iterate_fd() holds
	236	* file_lock, which should ensure the fdt does not change while
	237	* the lock is taken, but we are not aware whether this is
	238	* guaranteed or not, so play safe.
	239	*/
aa29f2d3	240	if (fd < fdt->max_fds && lttng_close_on_exec(fd, fdt))
29021503	241	flags \|= O_CLOEXEC;
361c023a MD	242	if (IS_ERR(s)) {
	243	struct dentry *dentry = file->f_path.dentry;
	244
	245	/* Make sure we give at least some info */
	246	spin_lock(&dentry->d_lock);
	247	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd,
29021503	248	dentry->d_name.name, flags, file->f_mode);
361c023a MD	249	spin_unlock(&dentry->d_lock);
	250	goto end;
	251	}
29021503 MD	252	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd, s,
29021503 MD	253	flags, file->f_mode);
361c023a MD	254	end:
	255	return 0;
	256	}
c337ddc2 MD	257
	258	static
	259	void lttng_enumerate_task_fd(struct lttng_session *session,
	260	struct task_struct p, char tmp)
	261	{
361c023a	262	struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .p = p };
d561ecfb	263	struct files_struct *files;
c337ddc2 MD	264
c337ddc2 MD	265	task_lock(p);
d561ecfb MD	266	files = p->files;
	267	if (!files)
	268	goto end;
	269	ctx.files = files;
	270	lttng_iterate_fd(files, 0, lttng_dump_one_fd, &ctx);
	271	end:
c337ddc2 MD	272	task_unlock(p);
	273	}
	274
	275	static
	276	int lttng_enumerate_file_descriptors(struct lttng_session *session)
	277	{
	278	struct task_struct *p;
cfcee1c7 MD	279	char *tmp;
	280
	281	tmp = (char *) __get_free_page(GFP_KERNEL);
	282	if (!tmp)
	283	return -ENOMEM;
c337ddc2 MD	284
	285	/* Enumerate active file descriptors */
	286	rcu_read_lock();
	287	for_each_process(p)
	288	lttng_enumerate_task_fd(session, p, tmp);
	289	rcu_read_unlock();
	290	free_page((unsigned long) tmp);
	291	return 0;
	292	}
	293
0658bdda MD	294	#if 0
	295	/*
	296	* FIXME: we cannot take a mmap_sem while in a RCU read-side critical section
	297	* (scheduling in atomic). Normally, the tasklist lock protects this kind of
	298	* iteration, but it is not exported to modules.
	299	*/
c337ddc2 MD	300	static
	301	void lttng_enumerate_task_vm_maps(struct lttng_session *session,
	302	struct task_struct *p)
	303	{
	304	struct mm_struct *mm;
	305	struct vm_area_struct *map;
	306	unsigned long ino;
	307
	308	/* get_task_mm does a task_lock... */
	309	mm = get_task_mm(p);
	310	if (!mm)
	311	return;
	312
	313	map = mm->mmap;
	314	if (map) {
	315	down_read(&mm->mmap_sem);
	316	while (map) {
	317	if (map->vm_file)
b06ed645	318	ino = map->vm_file->lttng_f_dentry->d_inode->i_ino;
c337ddc2 MD	319	else
	320	ino = 0;
	321	trace_lttng_statedump_vm_map(session, p, map, ino);
	322	map = map->vm_next;
	323	}
	324	up_read(&mm->mmap_sem);
	325	}
	326	mmput(mm);
	327	}
	328
	329	static
	330	int lttng_enumerate_vm_maps(struct lttng_session *session)
	331	{
	332	struct task_struct *p;
	333
	334	rcu_read_lock();
	335	for_each_process(p)
	336	lttng_enumerate_task_vm_maps(session, p);
	337	rcu_read_unlock();
	338	return 0;
	339	}
0658bdda	340	#endif
c337ddc2	341
29784493	342	#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
47faec4b	343
360d3efe	344	#if (LTTNG_LINUX_VERSION_CODE < LTTNG_KERNEL_VERSION(2,6,39))
47faec4b JN	345	#define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
	346	#endif
	347
c337ddc2	348	static
cfcee1c7	349	int lttng_list_interrupts(struct lttng_session *session)
c337ddc2 MD	350	{
	351	unsigned int irq;
	352	unsigned long flags = 0;
	353	struct irq_desc *desc;
	354
	355	#define irq_to_desc wrapper_irq_to_desc
	356	/* needs irq_desc */
	357	for_each_irq_desc(irq, desc) {
	358	struct irqaction *action;
	359	const char *irq_chip_name =
	360	irq_desc_get_chip(desc)->name ? : "unnamed_irq_chip";
	361
	362	local_irq_save(flags);
3a523f5b	363	wrapper_desc_spin_lock(&desc->lock);
c337ddc2 MD	364	for (action = desc->action; action; action = action->next) {
	365	trace_lttng_statedump_interrupt(session,
	366	irq, irq_chip_name, action);
	367	}
3a523f5b	368	wrapper_desc_spin_unlock(&desc->lock);
c337ddc2 MD	369	local_irq_restore(flags);
c337ddc2 MD	370	}
cfcee1c7	371	return 0;
c337ddc2 MD	372	#undef irq_to_desc
	373	}
	374	#else
	375	static inline
cfcee1c7	376	int lttng_list_interrupts(struct lttng_session *session)
c337ddc2	377	{
cfcee1c7	378	return 0;
c337ddc2 MD	379	}
	380	#endif
	381
4ba1f53c MD	382	/*
	383	* Called with task lock held.
	384	*/
73e8ba37 JD	385	static
	386	void lttng_statedump_process_ns(struct lttng_session *session,
	387	struct task_struct *p,
	388	enum lttng_thread_type type,
	389	enum lttng_execution_mode mode,
	390	enum lttng_execution_submode submode,
	391	enum lttng_process_status status)
	392	{
73e8ba37 JD	393	struct pid_namespace *pid_ns;
73e8ba37 JD	394
aba7d7fc MJ	395	pid_ns = task_active_pid_ns(p);
aba7d7fc MJ	396	do {
73e8ba37	397	trace_lttng_statedump_process_state(session,
aba7d7fc	398	p, type, mode, submode, status, pid_ns);
3e2bb820	399	pid_ns = pid_ns ? pid_ns->parent : NULL;
aba7d7fc	400	} while (pid_ns);
73e8ba37 JD	401	}
73e8ba37 JD	402
c337ddc2 MD	403	static
	404	int lttng_enumerate_process_states(struct lttng_session *session)
	405	{
	406	struct task_struct g, p;
	407
	408	rcu_read_lock();
	409	for_each_process(g) {
	410	p = g;
	411	do {
	412	enum lttng_execution_mode mode =
	413	LTTNG_MODE_UNKNOWN;
	414	enum lttng_execution_submode submode =
	415	LTTNG_UNKNOWN;
	416	enum lttng_process_status status;
	417	enum lttng_thread_type type;
	418
	419	task_lock(p);
	420	if (p->exit_state == EXIT_ZOMBIE)
	421	status = LTTNG_ZOMBIE;
	422	else if (p->exit_state == EXIT_DEAD)
	423	status = LTTNG_DEAD;
	424	else if (p->state == TASK_RUNNING) {
	425	/* Is this a forked child that has not run yet? */
	426	if (list_empty(&p->rt.run_list))
	427	status = LTTNG_WAIT_FORK;
	428	else
	429	/*
	430	* All tasks are considered as wait_cpu;
	431	* the viewer will sort out if the task
	432	* was really running at this time.
	433	*/
	434	status = LTTNG_WAIT_CPU;
	435	} else if (p->state &
	436	(TASK_INTERRUPTIBLE \| TASK_UNINTERRUPTIBLE)) {
	437	/* Task is waiting for something to complete */
	438	status = LTTNG_WAIT;
	439	} else
	440	status = LTTNG_UNNAMED;
	441	submode = LTTNG_NONE;
	442
	443	/*
	444	* Verification of t->mm is to filter out kernel
	445	* threads; Viewer will further filter out if a
	446	* user-space thread was in syscall mode or not.
	447	*/
	448	if (p->mm)
	449	type = LTTNG_USER_THREAD;
	450	else
	451	type = LTTNG_KERNEL_THREAD;
73e8ba37	452	lttng_statedump_process_ns(session,
c337ddc2 MD	453	p, type, mode, submode, status);
	454	task_unlock(p);
	455	} while_each_thread(g, p);
	456	}
	457	rcu_read_unlock();
	458
	459	return 0;
	460	}
	461
	462	static
	463	void lttng_statedump_work_func(struct work_struct *work)
	464	{
	465	if (atomic_dec_and_test(&kernel_threads_to_run))
	466	/* If we are the last thread, wake up do_lttng_statedump */
	467	wake_up(&statedump_wq);
	468	}
	469
	470	static
	471	int do_lttng_statedump(struct lttng_session *session)
	472	{
cfcee1c7	473	int cpu, ret;
c337ddc2	474
c337ddc2	475	trace_lttng_statedump_start(session);
cfcee1c7 MD	476	ret = lttng_enumerate_process_states(session);
	477	if (ret)
	478	return ret;
	479	ret = lttng_enumerate_file_descriptors(session);
	480	if (ret)
	481	return ret;
	482	/*
	483	* FIXME
	484	* ret = lttng_enumerate_vm_maps(session);
	485	* if (ret)
	486	* return ret;
	487	*/
	488	ret = lttng_list_interrupts(session);
	489	if (ret)
	490	return ret;
	491	ret = lttng_enumerate_network_ip_interface(session);
	492	if (ret)
	493	return ret;
	494	ret = lttng_enumerate_block_devices(session);
	495	switch (ret) {
84c7055e MD	496	case 0:
84c7055e MD	497	break;
cfcee1c7 MD	498	case -ENOSYS:
	499	printk(KERN_WARNING "LTTng: block device enumeration is not supported by kernel\n");
	500	break;
	501	default:
	502	return ret;
	503	}
c337ddc2 MD	504
	505	/* TODO lttng_dump_idt_table(session); */
	506	/* TODO lttng_dump_softirq_vec(session); */
	507	/* TODO lttng_list_modules(session); */
	508	/* TODO lttng_dump_swap_files(session); */
	509
	510	/*
	511	* Fire off a work queue on each CPU. Their sole purpose in life
	512	* is to guarantee that each CPU has been in a state where is was in
	513	* syscall mode (i.e. not in a trap, an IRQ or a soft IRQ).
	514	*/
	515	get_online_cpus();
	516	atomic_set(&kernel_threads_to_run, num_online_cpus());
	517	for_each_online_cpu(cpu) {
	518	INIT_DELAYED_WORK(&cpu_work[cpu], lttng_statedump_work_func);
	519	schedule_delayed_work_on(cpu, &cpu_work[cpu], 0);
	520	}
	521	/* Wait for all threads to run */
7a7128e0	522	__wait_event(statedump_wq, (atomic_read(&kernel_threads_to_run) == 0));
c337ddc2 MD	523	put_online_cpus();
c337ddc2 MD	524	/* Our work is done */
c337ddc2 MD	525	trace_lttng_statedump_end(session);
	526	return 0;
	527	}
	528
	529	/*
	530	* Called with session mutex held.
	531	*/
	532	int lttng_statedump_start(struct lttng_session *session)
	533	{
c337ddc2 MD	534	return do_lttng_statedump(session);
	535	}
	536	EXPORT_SYMBOL_GPL(lttng_statedump_start);
	537
dd8d5afb MD	538	static
	539	int __init lttng_statedump_init(void)
	540	{
d16aa9c9 MD	541	/*
	542	* Allow module to load even if the fixup cannot be done. This
	543	* will allow seemless transition when the underlying issue fix
	544	* is merged into the Linux kernel, and when tracepoint.c
	545	* "tracepoint_module_notify" is turned into a static function.
	546	*/
	547	(void) wrapper_lttng_fixup_sig(THIS_MODULE);
	548	return 0;
dd8d5afb MD	549	}
	550
	551	module_init(lttng_statedump_init);
	552
461277e7 MD	553	static
	554	void __exit lttng_statedump_exit(void)
	555	{
	556	}
	557
	558	module_exit(lttng_statedump_exit);
	559
c337ddc2 MD	560	MODULE_LICENSE("GPL and additional rights");
	561	MODULE_AUTHOR("Jean-Hugues Deschenes");
	562	MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Statedump");
13ab8b0a MD	563	MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
	564	__stringify(LTTNG_MODULES_MINOR_VERSION) "."
	565	__stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
	566	LTTNG_MODULES_EXTRAVERSION);