[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/fdtable.h>
#include <linux/swap.h>
#include <linux/wait.h>
#include <linux/mutex.h>

#include "lttng-events.h"
#include "lttng-tracer.h"
#include "wrapper/irqdesc.h"
#include "wrapper/spinlock.h"
#include "wrapper/fdtable.h"
#include "wrapper/nsproxy.h"
#include "wrapper/irq.h"
#include "wrapper/tracepoint.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
#include "instrumentation/events/lttng-module/lttng-statedump.h"

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

/*
 * 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,
};

#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);

	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);
		spin_unlock(&dentry->d_lock);
		goto end;
	}
	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd, s);
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 };

	task_lock(p);
	lttng_iterate_fd(p->files, 0, lttng_dump_one_fd, &ctx);
	task_unlock(p);
}

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

	/* 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->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 (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39))
#define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
#endif

static
void 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);
	}
#undef irq_to_desc
}
#else
static inline
void lttng_list_interrupts(struct lttng_session *session)
{
}
#endif

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 nsproxy *proxy;
	struct pid_namespace *pid_ns;

	rcu_read_lock();
	proxy = task_nsproxy(p);
	if (proxy) {
		pid_ns = lttng_get_proxy_pid_ns(proxy);
		do {
			trace_lttng_statedump_process_state(session,
				p, type, mode, submode, status, pid_ns);
			pid_ns = pid_ns->parent;
		} while (pid_ns);
	} else {
		trace_lttng_statedump_process_state(session,
			p, type, mode, submode, status, NULL);
	}
	rcu_read_unlock();
}

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;

	trace_lttng_statedump_start(session);
	lttng_enumerate_process_states(session);
	lttng_enumerate_file_descriptors(session);
	/* FIXME lttng_enumerate_vm_maps(session); */
	lttng_list_interrupts(session);
	lttng_enumerate_network_ip_interface(session);

	/* 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>
	44	#include <linux/fdtable.h>
	45	#include <linux/swap.h>
	46	#include <linux/wait.h>
	47	#include <linux/mutex.h>
	48
	49	#include "lttng-events.h"
13ab8b0a	50	#include "lttng-tracer.h"
c337ddc2	51	#include "wrapper/irqdesc.h"
3a523f5b	52	#include "wrapper/spinlock.h"
361c023a	53	#include "wrapper/fdtable.h"
3247f8bd	54	#include "wrapper/nsproxy.h"
29784493	55	#include "wrapper/irq.h"
dd8d5afb	56	#include "wrapper/tracepoint.h"
c337ddc2	57
29784493	58	#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
c337ddc2 MD	59	#include <linux/irq.h>
	60	#endif
	61
	62	/* Define the tracepoints, but do not build the probes */
	63	#define CREATE_TRACE_POINTS
	64	#define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
	65	#define TRACE_INCLUDE_FILE lttng-statedump
	66	#include "instrumentation/events/lttng-module/lttng-statedump.h"
	67
361c023a MD	68	struct lttng_fd_ctx {
	69	char *page;
	70	struct lttng_session *session;
	71	struct task_struct *p;
	72	};
	73
c337ddc2 MD	74	/*
	75	* Protected by the trace lock.
	76	*/
	77	static struct delayed_work cpu_work[NR_CPUS];
	78	static DECLARE_WAIT_QUEUE_HEAD(statedump_wq);
	79	static atomic_t kernel_threads_to_run;
	80
	81	enum lttng_thread_type {
	82	LTTNG_USER_THREAD = 0,
	83	LTTNG_KERNEL_THREAD = 1,
	84	};
	85
	86	enum lttng_execution_mode {
	87	LTTNG_USER_MODE = 0,
	88	LTTNG_SYSCALL = 1,
	89	LTTNG_TRAP = 2,
	90	LTTNG_IRQ = 3,
	91	LTTNG_SOFTIRQ = 4,
	92	LTTNG_MODE_UNKNOWN = 5,
	93	};
	94
	95	enum lttng_execution_submode {
	96	LTTNG_NONE = 0,
	97	LTTNG_UNKNOWN = 1,
	98	};
	99
	100	enum lttng_process_status {
	101	LTTNG_UNNAMED = 0,
	102	LTTNG_WAIT_FORK = 1,
	103	LTTNG_WAIT_CPU = 2,
	104	LTTNG_EXIT = 3,
	105	LTTNG_ZOMBIE = 4,
	106	LTTNG_WAIT = 5,
	107	LTTNG_RUN = 6,
	108	LTTNG_DEAD = 7,
	109	};
	110
	111	#ifdef CONFIG_INET
	112	static
	113	void lttng_enumerate_device(struct lttng_session *session,
	114	struct net_device *dev)
	115	{
	116	struct in_device *in_dev;
	117	struct in_ifaddr *ifa;
	118
	119	if (dev->flags & IFF_UP) {
	120	in_dev = in_dev_get(dev);
	121	if (in_dev) {
	122	for (ifa = in_dev->ifa_list; ifa != NULL;
	123	ifa = ifa->ifa_next) {
	124	trace_lttng_statedump_network_interface(
	125	session, dev, ifa);
	126	}
	127	in_dev_put(in_dev);
	128	}
	129	} else {
	130	trace_lttng_statedump_network_interface(
	131	session, dev, NULL);
	132	}
	133	}
	134
	135	static
	136	int lttng_enumerate_network_ip_interface(struct lttng_session *session)
	137	{
138	struct net_device *dev;
139
140	read_lock(&dev_base_lock);
141	for_each_netdev(&init_net, dev)
142	lttng_enumerate_device(session, dev);
143	read_unlock(&dev_base_lock);
144
145	return 0;
146	}
147	#else /* CONFIG_INET */
148	static inline
149	int lttng_enumerate_network_ip_interface(struct lttng_session *session)
150	{
151	return 0;
152	}
153	#endif /* CONFIG_INET */
154
361c023a MD	155	static
	156	int lttng_dump_one_fd(const void p, struct file file, unsigned int fd)
	157	{
	158	const struct lttng_fd_ctx *ctx = p;
	159	const char *s = d_path(&file->f_path, ctx->page, PAGE_SIZE);
	160
	161	if (IS_ERR(s)) {
	162	struct dentry *dentry = file->f_path.dentry;
	163
	164	/* Make sure we give at least some info */
	165	spin_lock(&dentry->d_lock);
	166	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd,
	167	dentry->d_name.name);
	168	spin_unlock(&dentry->d_lock);
	169	goto end;
	170	}
	171	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd, s);
	172	end:
	173	return 0;
	174	}
c337ddc2 MD	175
	176	static
	177	void lttng_enumerate_task_fd(struct lttng_session *session,
	178	struct task_struct p, char tmp)
	179	{
361c023a	180	struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .p = p };
c337ddc2 MD	181
c337ddc2 MD	182	task_lock(p);
361c023a	183	lttng_iterate_fd(p->files, 0, lttng_dump_one_fd, &ctx);
c337ddc2 MD	184	task_unlock(p);
	185	}
	186
	187	static
	188	int lttng_enumerate_file_descriptors(struct lttng_session *session)
	189	{
	190	struct task_struct *p;
	191	char tmp = (char ) __get_free_page(GFP_KERNEL);
	192
	193	/* Enumerate active file descriptors */
	194	rcu_read_lock();
	195	for_each_process(p)
	196	lttng_enumerate_task_fd(session, p, tmp);
	197	rcu_read_unlock();
	198	free_page((unsigned long) tmp);
	199	return 0;
	200	}
	201
0658bdda MD	202	#if 0
	203	/*
	204	* FIXME: we cannot take a mmap_sem while in a RCU read-side critical section
	205	* (scheduling in atomic). Normally, the tasklist lock protects this kind of
	206	* iteration, but it is not exported to modules.
	207	*/
c337ddc2 MD	208	static
	209	void lttng_enumerate_task_vm_maps(struct lttng_session *session,
	210	struct task_struct *p)
	211	{
	212	struct mm_struct *mm;
	213	struct vm_area_struct *map;
	214	unsigned long ino;
	215
	216	/* get_task_mm does a task_lock... */
	217	mm = get_task_mm(p);
	218	if (!mm)
	219	return;
	220
	221	map = mm->mmap;
	222	if (map) {
	223	down_read(&mm->mmap_sem);
	224	while (map) {
	225	if (map->vm_file)
	226	ino = map->vm_file->f_dentry->d_inode->i_ino;
	227	else
	228	ino = 0;
	229	trace_lttng_statedump_vm_map(session, p, map, ino);
	230	map = map->vm_next;
	231	}
	232	up_read(&mm->mmap_sem);
	233	}
	234	mmput(mm);
	235	}
	236
	237	static
	238	int lttng_enumerate_vm_maps(struct lttng_session *session)
	239	{
	240	struct task_struct *p;
	241
	242	rcu_read_lock();
	243	for_each_process(p)
	244	lttng_enumerate_task_vm_maps(session, p);
	245	rcu_read_unlock();
	246	return 0;
	247	}
0658bdda	248	#endif
c337ddc2	249
29784493	250	#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
47faec4b JN	251
	252	#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39))
	253	#define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
	254	#endif
	255
c337ddc2 MD	256	static
	257	void lttng_list_interrupts(struct lttng_session *session)
	258	{
	259	unsigned int irq;
	260	unsigned long flags = 0;
	261	struct irq_desc *desc;
	262
	263	#define irq_to_desc wrapper_irq_to_desc
	264	/* needs irq_desc */
	265	for_each_irq_desc(irq, desc) {
	266	struct irqaction *action;
	267	const char *irq_chip_name =
	268	irq_desc_get_chip(desc)->name ? : "unnamed_irq_chip";
	269
	270	local_irq_save(flags);
3a523f5b	271	wrapper_desc_spin_lock(&desc->lock);
c337ddc2 MD	272	for (action = desc->action; action; action = action->next) {
	273	trace_lttng_statedump_interrupt(session,
	274	irq, irq_chip_name, action);
	275	}
3a523f5b	276	wrapper_desc_spin_unlock(&desc->lock);
c337ddc2 MD	277	local_irq_restore(flags);
	278	}
	279	#undef irq_to_desc
	280	}
	281	#else
	282	static inline
2e7a0709	283	void lttng_list_interrupts(struct lttng_session *session)
c337ddc2 MD	284	{
	285	}
	286	#endif
	287
73e8ba37 JD	288	static
	289	void lttng_statedump_process_ns(struct lttng_session *session,
	290	struct task_struct *p,
	291	enum lttng_thread_type type,
	292	enum lttng_execution_mode mode,
	293	enum lttng_execution_submode submode,
	294	enum lttng_process_status status)
	295	{
	296	struct nsproxy *proxy;
	297	struct pid_namespace *pid_ns;
	298
	299	rcu_read_lock();
	300	proxy = task_nsproxy(p);
	301	if (proxy) {
3247f8bd	302	pid_ns = lttng_get_proxy_pid_ns(proxy);
73e8ba37 JD	303	do {
	304	trace_lttng_statedump_process_state(session,
	305	p, type, mode, submode, status, pid_ns);
af73f727	306	pid_ns = pid_ns->parent;
73e8ba37 JD	307	} while (pid_ns);
	308	} else {
	309	trace_lttng_statedump_process_state(session,
	310	p, type, mode, submode, status, NULL);
	311	}
	312	rcu_read_unlock();
	313	}
	314
c337ddc2 MD	315	static
	316	int lttng_enumerate_process_states(struct lttng_session *session)
	317	{
	318	struct task_struct g, p;
	319
	320	rcu_read_lock();
	321	for_each_process(g) {
	322	p = g;
	323	do {
	324	enum lttng_execution_mode mode =
	325	LTTNG_MODE_UNKNOWN;
	326	enum lttng_execution_submode submode =
	327	LTTNG_UNKNOWN;
	328	enum lttng_process_status status;
	329	enum lttng_thread_type type;
	330
	331	task_lock(p);
	332	if (p->exit_state == EXIT_ZOMBIE)
	333	status = LTTNG_ZOMBIE;
	334	else if (p->exit_state == EXIT_DEAD)
	335	status = LTTNG_DEAD;
	336	else if (p->state == TASK_RUNNING) {
	337	/* Is this a forked child that has not run yet? */
	338	if (list_empty(&p->rt.run_list))
	339	status = LTTNG_WAIT_FORK;
	340	else
	341	/*
	342	* All tasks are considered as wait_cpu;
	343	* the viewer will sort out if the task
	344	* was really running at this time.
	345	*/
	346	status = LTTNG_WAIT_CPU;
	347	} else if (p->state &
	348	(TASK_INTERRUPTIBLE \| TASK_UNINTERRUPTIBLE)) {
	349	/* Task is waiting for something to complete */
	350	status = LTTNG_WAIT;
	351	} else
	352	status = LTTNG_UNNAMED;
	353	submode = LTTNG_NONE;
	354
	355	/*
	356	* Verification of t->mm is to filter out kernel
	357	* threads; Viewer will further filter out if a
	358	* user-space thread was in syscall mode or not.
	359	*/
	360	if (p->mm)
	361	type = LTTNG_USER_THREAD;
	362	else
	363	type = LTTNG_KERNEL_THREAD;
73e8ba37	364	lttng_statedump_process_ns(session,
c337ddc2 MD	365	p, type, mode, submode, status);
	366	task_unlock(p);
	367	} while_each_thread(g, p);
	368	}
	369	rcu_read_unlock();
	370
	371	return 0;
	372	}
	373
	374	static
	375	void lttng_statedump_work_func(struct work_struct *work)
	376	{
	377	if (atomic_dec_and_test(&kernel_threads_to_run))
	378	/* If we are the last thread, wake up do_lttng_statedump */
	379	wake_up(&statedump_wq);
	380	}
	381
	382	static
	383	int do_lttng_statedump(struct lttng_session *session)
	384	{
	385	int cpu;
	386
c337ddc2 MD	387	trace_lttng_statedump_start(session);
	388	lttng_enumerate_process_states(session);
	389	lttng_enumerate_file_descriptors(session);
0658bdda	390	/* FIXME lttng_enumerate_vm_maps(session); */
c337ddc2 MD	391	lttng_list_interrupts(session);
	392	lttng_enumerate_network_ip_interface(session);
	393
	394	/* TODO lttng_dump_idt_table(session); */
	395	/* TODO lttng_dump_softirq_vec(session); */
	396	/* TODO lttng_list_modules(session); */
	397	/* TODO lttng_dump_swap_files(session); */
	398
	399	/*
	400	* Fire off a work queue on each CPU. Their sole purpose in life
	401	* is to guarantee that each CPU has been in a state where is was in
	402	* syscall mode (i.e. not in a trap, an IRQ or a soft IRQ).
	403	*/
	404	get_online_cpus();
	405	atomic_set(&kernel_threads_to_run, num_online_cpus());
	406	for_each_online_cpu(cpu) {
	407	INIT_DELAYED_WORK(&cpu_work[cpu], lttng_statedump_work_func);
	408	schedule_delayed_work_on(cpu, &cpu_work[cpu], 0);
	409	}
	410	/* Wait for all threads to run */
7a7128e0	411	__wait_event(statedump_wq, (atomic_read(&kernel_threads_to_run) == 0));
c337ddc2 MD	412	put_online_cpus();
c337ddc2 MD	413	/* Our work is done */
c337ddc2 MD	414	trace_lttng_statedump_end(session);
	415	return 0;
	416	}
	417
	418	/*
	419	* Called with session mutex held.
	420	*/
	421	int lttng_statedump_start(struct lttng_session *session)
	422	{
c337ddc2 MD	423	return do_lttng_statedump(session);
	424	}
	425	EXPORT_SYMBOL_GPL(lttng_statedump_start);
	426
dd8d5afb MD	427	static
	428	int __init lttng_statedump_init(void)
	429	{
d16aa9c9 MD	430	/*
	431	* Allow module to load even if the fixup cannot be done. This
	432	* will allow seemless transition when the underlying issue fix
	433	* is merged into the Linux kernel, and when tracepoint.c
	434	* "tracepoint_module_notify" is turned into a static function.
	435	*/
	436	(void) wrapper_lttng_fixup_sig(THIS_MODULE);
	437	return 0;
dd8d5afb MD	438	}
	439
	440	module_init(lttng_statedump_init);
	441
461277e7 MD	442	static
	443	void __exit lttng_statedump_exit(void)
	444	{
	445	}
	446
	447	module_exit(lttng_statedump_exit);
	448
c337ddc2 MD	449	MODULE_LICENSE("GPL and additional rights");
	450	MODULE_AUTHOR("Jean-Hugues Deschenes");
	451	MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Statedump");
13ab8b0a MD	452	MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
	453	__stringify(LTTNG_MODULES_MINOR_VERSION) "."
	454	__stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
	455	LTTNG_MODULES_EXTRAVERSION);