[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 <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/nsproxy.h"
#include "wrapper/irq.h"
#include "wrapper/tracepoint.h"
#include "wrapper/genhd.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"

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 */

#ifdef FD_ISSET	/* For old kernels lacking close_on_exec() */
static inline bool lttng_close_on_exec(int fd, const struct fdtable *fdt)
{
	return FD_ISSET(fd, fdt->close_on_exec);
}
#else
static inline bool lttng_close_on_exec(int fd, const struct fdtable *fdt)
{
	return close_on_exec(fd, fdt);
}
#endif

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->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
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

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, 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 -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>
	44	#include <linux/fdtable.h>
	45	#include <linux/swap.h>
	46	#include <linux/wait.h>
	47	#include <linux/mutex.h>
f0dbdefb	48	#include <linux/device.h>
c337ddc2 MD	49
c337ddc2 MD	50	#include "lttng-events.h"
13ab8b0a	51	#include "lttng-tracer.h"
c337ddc2	52	#include "wrapper/irqdesc.h"
3a523f5b	53	#include "wrapper/spinlock.h"
361c023a	54	#include "wrapper/fdtable.h"
3247f8bd	55	#include "wrapper/nsproxy.h"
29784493	56	#include "wrapper/irq.h"
dd8d5afb	57	#include "wrapper/tracepoint.h"
f0dbdefb	58	#include "wrapper/genhd.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
	66	#define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
	67	#define TRACE_INCLUDE_FILE lttng-statedump
	68	#include "instrumentation/events/lttng-module/lttng-statedump.h"
	69
f0dbdefb	70	DEFINE_TRACE(lttng_statedump_block_device);
20591cf7 MD	71	DEFINE_TRACE(lttng_statedump_end);
	72	DEFINE_TRACE(lttng_statedump_interrupt);
	73	DEFINE_TRACE(lttng_statedump_file_descriptor);
	74	DEFINE_TRACE(lttng_statedump_start);
	75	DEFINE_TRACE(lttng_statedump_process_state);
	76	DEFINE_TRACE(lttng_statedump_network_interface);
	77
361c023a MD	78	struct lttng_fd_ctx {
	79	char *page;
	80	struct lttng_session *session;
	81	struct task_struct *p;
d561ecfb	82	struct files_struct *files;
361c023a MD	83	};
361c023a MD	84
c337ddc2 MD	85	/*
	86	* Protected by the trace lock.
	87	*/
	88	static struct delayed_work cpu_work[NR_CPUS];
	89	static DECLARE_WAIT_QUEUE_HEAD(statedump_wq);
	90	static atomic_t kernel_threads_to_run;
	91
	92	enum lttng_thread_type {
	93	LTTNG_USER_THREAD = 0,
	94	LTTNG_KERNEL_THREAD = 1,
	95	};
	96
	97	enum lttng_execution_mode {
	98	LTTNG_USER_MODE = 0,
	99	LTTNG_SYSCALL = 1,
	100	LTTNG_TRAP = 2,
	101	LTTNG_IRQ = 3,
	102	LTTNG_SOFTIRQ = 4,
	103	LTTNG_MODE_UNKNOWN = 5,
	104	};
	105
	106	enum lttng_execution_submode {
	107	LTTNG_NONE = 0,
	108	LTTNG_UNKNOWN = 1,
	109	};
	110
	111	enum lttng_process_status {
	112	LTTNG_UNNAMED = 0,
	113	LTTNG_WAIT_FORK = 1,
	114	LTTNG_WAIT_CPU = 2,
	115	LTTNG_EXIT = 3,
	116	LTTNG_ZOMBIE = 4,
	117	LTTNG_WAIT = 5,
	118	LTTNG_RUN = 6,
	119	LTTNG_DEAD = 7,
	120	};
	121
f0dbdefb HD	122	static
	123	int lttng_enumerate_block_devices(struct lttng_session *session)
	124	{
	125	struct class *ptr_block_class;
	126	struct device_type *ptr_disk_type;
	127	struct class_dev_iter iter;
	128	struct device *dev;
	129
	130	ptr_block_class = wrapper_get_block_class();
	131	if (!ptr_block_class)
	132	return -ENOSYS;
	133	ptr_disk_type = wrapper_get_disk_type();
	134	if (!ptr_disk_type) {
	135	return -ENOSYS;
	136	}
	137	class_dev_iter_init(&iter, ptr_block_class, NULL, ptr_disk_type);
	138	while ((dev = class_dev_iter_next(&iter))) {
	139	struct disk_part_iter piter;
	140	struct gendisk *disk = dev_to_disk(dev);
	141	struct hd_struct *part;
	142
5a91f3df MD	143	/*
	144	* Don't show empty devices or things that have been
	145	* suppressed
	146	*/
	147	if (get_capacity(disk) == 0 \|\|
	148	(disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
	149	continue;
	150
f0dbdefb HD	151	disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
	152	while ((part = disk_part_iter_next(&piter))) {
	153	char name_buf[BDEVNAME_SIZE];
	154	char *p;
	155
	156	p = wrapper_disk_name(disk, part->partno, name_buf);
	157	if (!p) {
	158	disk_part_iter_exit(&piter);
	159	class_dev_iter_exit(&iter);
	160	return -ENOSYS;
	161	}
	162	trace_lttng_statedump_block_device(session,
	163	part_devt(part), name_buf);
	164	}
	165	disk_part_iter_exit(&piter);
	166	}
	167	class_dev_iter_exit(&iter);
	168	return 0;
	169	}
	170
c337ddc2	171	#ifdef CONFIG_INET
f0dbdefb	172
c337ddc2 MD	173	static
	174	void lttng_enumerate_device(struct lttng_session *session,
	175	struct net_device *dev)
	176	{
	177	struct in_device *in_dev;
	178	struct in_ifaddr *ifa;
	179
	180	if (dev->flags & IFF_UP) {
	181	in_dev = in_dev_get(dev);
	182	if (in_dev) {
	183	for (ifa = in_dev->ifa_list; ifa != NULL;
	184	ifa = ifa->ifa_next) {
	185	trace_lttng_statedump_network_interface(
	186	session, dev, ifa);
	187	}
	188	in_dev_put(in_dev);
	189	}
	190	} else {
	191	trace_lttng_statedump_network_interface(
	192	session, dev, NULL);
	193	}
	194	}
	195
	196	static
	197	int lttng_enumerate_network_ip_interface(struct lttng_session *session)
	198	{
	199	struct net_device *dev;
	200
	201	read_lock(&dev_base_lock);
	202	for_each_netdev(&init_net, dev)
	203	lttng_enumerate_device(session, dev);
	204	read_unlock(&dev_base_lock);
	205
	206	return 0;
	207	}
	208	#else /* CONFIG_INET */
	209	static inline
	210	int lttng_enumerate_network_ip_interface(struct lttng_session *session)
	211	{
	212	return 0;
	213	}
	214	#endif /* CONFIG_INET */
	215
eaca9be6 MD	216	#ifdef FD_ISSET /* For old kernels lacking close_on_exec() */
	217	static inline bool lttng_close_on_exec(int fd, const struct fdtable *fdt)
	218	{
	219	return FD_ISSET(fd, fdt->close_on_exec);
	220	}
	221	#else
	222	static inline bool lttng_close_on_exec(int fd, const struct fdtable *fdt)
	223	{
	224	return close_on_exec(fd, fdt);
	225	}
	226	#endif
	227
361c023a MD	228	static
	229	int lttng_dump_one_fd(const void p, struct file file, unsigned int fd)
	230	{
	231	const struct lttng_fd_ctx *ctx = p;
	232	const char *s = d_path(&file->f_path, ctx->page, PAGE_SIZE);
29021503	233	unsigned int flags = file->f_flags;
d561ecfb	234	struct fdtable *fdt;
361c023a	235
29021503 MD	236	/*
	237	* We don't expose kernel internal flags, only userspace-visible
	238	* flags.
	239	*/
	240	flags &= ~FMODE_NONOTIFY;
d561ecfb MD	241	fdt = files_fdtable(ctx->files);
	242	/*
	243	* We need to check here again whether fd is within the fdt
	244	* max_fds range, because we might be seeing a different
	245	* files_fdtable() than iterate_fd(), assuming only RCU is
	246	* protecting the read. In reality, iterate_fd() holds
	247	* file_lock, which should ensure the fdt does not change while
	248	* the lock is taken, but we are not aware whether this is
	249	* guaranteed or not, so play safe.
	250	*/
eaca9be6	251	if (fd < fdt->max_fds && lttng_close_on_exec(fd, fdt))
29021503	252	flags \|= O_CLOEXEC;
361c023a MD	253	if (IS_ERR(s)) {
	254	struct dentry *dentry = file->f_path.dentry;
	255
	256	/* Make sure we give at least some info */
	257	spin_lock(&dentry->d_lock);
	258	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd,
29021503	259	dentry->d_name.name, flags, file->f_mode);
361c023a MD	260	spin_unlock(&dentry->d_lock);
	261	goto end;
	262	}
29021503 MD	263	trace_lttng_statedump_file_descriptor(ctx->session, ctx->p, fd, s,
29021503 MD	264	flags, file->f_mode);
361c023a MD	265	end:
	266	return 0;
	267	}
c337ddc2 MD	268
	269	static
	270	void lttng_enumerate_task_fd(struct lttng_session *session,
	271	struct task_struct p, char tmp)
	272	{
361c023a	273	struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .p = p };
d561ecfb	274	struct files_struct *files;
c337ddc2 MD	275
c337ddc2 MD	276	task_lock(p);
d561ecfb MD	277	files = p->files;
	278	if (!files)
	279	goto end;
	280	ctx.files = files;
	281	lttng_iterate_fd(files, 0, lttng_dump_one_fd, &ctx);
	282	end:
c337ddc2 MD	283	task_unlock(p);
	284	}
	285
	286	static
	287	int lttng_enumerate_file_descriptors(struct lttng_session *session)
	288	{
	289	struct task_struct *p;
cfcee1c7 MD	290	char *tmp;
	291
	292	tmp = (char *) __get_free_page(GFP_KERNEL);
	293	if (!tmp)
	294	return -ENOMEM;
c337ddc2 MD	295
	296	/* Enumerate active file descriptors */
	297	rcu_read_lock();
	298	for_each_process(p)
	299	lttng_enumerate_task_fd(session, p, tmp);
	300	rcu_read_unlock();
	301	free_page((unsigned long) tmp);
	302	return 0;
	303	}
	304
0658bdda MD	305	#if 0
	306	/*
	307	* FIXME: we cannot take a mmap_sem while in a RCU read-side critical section
	308	* (scheduling in atomic). Normally, the tasklist lock protects this kind of
	309	* iteration, but it is not exported to modules.
	310	*/
c337ddc2 MD	311	static
	312	void lttng_enumerate_task_vm_maps(struct lttng_session *session,
	313	struct task_struct *p)
	314	{
	315	struct mm_struct *mm;
	316	struct vm_area_struct *map;
	317	unsigned long ino;
	318
	319	/* get_task_mm does a task_lock... */
	320	mm = get_task_mm(p);
	321	if (!mm)
	322	return;
	323
	324	map = mm->mmap;
	325	if (map) {
	326	down_read(&mm->mmap_sem);
	327	while (map) {
	328	if (map->vm_file)
	329	ino = map->vm_file->f_dentry->d_inode->i_ino;
	330	else
	331	ino = 0;
	332	trace_lttng_statedump_vm_map(session, p, map, ino);
	333	map = map->vm_next;
	334	}
	335	up_read(&mm->mmap_sem);
	336	}
	337	mmput(mm);
	338	}
	339
	340	static
	341	int lttng_enumerate_vm_maps(struct lttng_session *session)
	342	{
	343	struct task_struct *p;
	344
	345	rcu_read_lock();
	346	for_each_process(p)
	347	lttng_enumerate_task_vm_maps(session, p);
	348	rcu_read_unlock();
	349	return 0;
	350	}
0658bdda	351	#endif
c337ddc2	352
29784493	353	#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
47faec4b JN	354
	355	#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39))
	356	#define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
	357	#endif
	358
c337ddc2	359	static
cfcee1c7	360	int lttng_list_interrupts(struct lttng_session *session)
c337ddc2 MD	361	{
	362	unsigned int irq;
	363	unsigned long flags = 0;
	364	struct irq_desc *desc;
	365
	366	#define irq_to_desc wrapper_irq_to_desc
	367	/* needs irq_desc */
	368	for_each_irq_desc(irq, desc) {
	369	struct irqaction *action;
	370	const char *irq_chip_name =
	371	irq_desc_get_chip(desc)->name ? : "unnamed_irq_chip";
	372
	373	local_irq_save(flags);
3a523f5b	374	wrapper_desc_spin_lock(&desc->lock);
c337ddc2 MD	375	for (action = desc->action; action; action = action->next) {
	376	trace_lttng_statedump_interrupt(session,
	377	irq, irq_chip_name, action);
	378	}
3a523f5b	379	wrapper_desc_spin_unlock(&desc->lock);
c337ddc2 MD	380	local_irq_restore(flags);
c337ddc2 MD	381	}
cfcee1c7	382	return 0;
c337ddc2 MD	383	#undef irq_to_desc
	384	}
	385	#else
	386	static inline
cfcee1c7	387	int lttng_list_interrupts(struct lttng_session *session)
c337ddc2	388	{
cfcee1c7	389	return 0;
c337ddc2 MD	390	}
	391	#endif
	392
73e8ba37 JD	393	static
	394	void lttng_statedump_process_ns(struct lttng_session *session,
	395	struct task_struct *p,
	396	enum lttng_thread_type type,
	397	enum lttng_execution_mode mode,
	398	enum lttng_execution_submode submode,
	399	enum lttng_process_status status)
	400	{
	401	struct nsproxy *proxy;
	402	struct pid_namespace *pid_ns;
	403
	404	rcu_read_lock();
	405	proxy = task_nsproxy(p);
	406	if (proxy) {
3247f8bd	407	pid_ns = lttng_get_proxy_pid_ns(proxy);
73e8ba37 JD	408	do {
	409	trace_lttng_statedump_process_state(session,
	410	p, type, mode, submode, status, pid_ns);
af73f727	411	pid_ns = pid_ns->parent;
73e8ba37 JD	412	} while (pid_ns);
	413	} else {
	414	trace_lttng_statedump_process_state(session,
	415	p, type, mode, submode, status, NULL);
	416	}
	417	rcu_read_unlock();
	418	}
	419
c337ddc2 MD	420	static
	421	int lttng_enumerate_process_states(struct lttng_session *session)
	422	{
	423	struct task_struct g, p;
	424
	425	rcu_read_lock();
	426	for_each_process(g) {
	427	p = g;
	428	do {
	429	enum lttng_execution_mode mode =
	430	LTTNG_MODE_UNKNOWN;
	431	enum lttng_execution_submode submode =
	432	LTTNG_UNKNOWN;
	433	enum lttng_process_status status;
	434	enum lttng_thread_type type;
	435
	436	task_lock(p);
	437	if (p->exit_state == EXIT_ZOMBIE)
	438	status = LTTNG_ZOMBIE;
	439	else if (p->exit_state == EXIT_DEAD)
	440	status = LTTNG_DEAD;
	441	else if (p->state == TASK_RUNNING) {
	442	/* Is this a forked child that has not run yet? */
	443	if (list_empty(&p->rt.run_list))
	444	status = LTTNG_WAIT_FORK;
	445	else
	446	/*
	447	* All tasks are considered as wait_cpu;
	448	* the viewer will sort out if the task
	449	* was really running at this time.
	450	*/
	451	status = LTTNG_WAIT_CPU;
	452	} else if (p->state &
	453	(TASK_INTERRUPTIBLE \| TASK_UNINTERRUPTIBLE)) {
	454	/* Task is waiting for something to complete */
	455	status = LTTNG_WAIT;
	456	} else
	457	status = LTTNG_UNNAMED;
	458	submode = LTTNG_NONE;
	459
	460	/*
	461	* Verification of t->mm is to filter out kernel
	462	* threads; Viewer will further filter out if a
	463	* user-space thread was in syscall mode or not.
	464	*/
	465	if (p->mm)
	466	type = LTTNG_USER_THREAD;
	467	else
	468	type = LTTNG_KERNEL_THREAD;
73e8ba37	469	lttng_statedump_process_ns(session,
c337ddc2 MD	470	p, type, mode, submode, status);
	471	task_unlock(p);
	472	} while_each_thread(g, p);
	473	}
	474	rcu_read_unlock();
	475
	476	return 0;
	477	}
	478
	479	static
	480	void lttng_statedump_work_func(struct work_struct *work)
	481	{
	482	if (atomic_dec_and_test(&kernel_threads_to_run))
	483	/* If we are the last thread, wake up do_lttng_statedump */
	484	wake_up(&statedump_wq);
	485	}
	486
	487	static
	488	int do_lttng_statedump(struct lttng_session *session)
	489	{
cfcee1c7	490	int cpu, ret;
c337ddc2	491
c337ddc2	492	trace_lttng_statedump_start(session);
cfcee1c7 MD	493	ret = lttng_enumerate_process_states(session);
	494	if (ret)
	495	return ret;
	496	ret = lttng_enumerate_file_descriptors(session);
	497	if (ret)
	498	return ret;
	499	/*
	500	* FIXME
	501	* ret = lttng_enumerate_vm_maps(session);
	502	* if (ret)
	503	* return ret;
	504	*/
	505	ret = lttng_list_interrupts(session);
	506	if (ret)
	507	return ret;
	508	ret = lttng_enumerate_network_ip_interface(session);
	509	if (ret)
	510	return ret;
	511	ret = lttng_enumerate_block_devices(session);
	512	switch (ret) {
	513	case -ENOSYS:
	514	printk(KERN_WARNING "LTTng: block device enumeration is not supported by kernel\n");
	515	break;
	516	default:
	517	return ret;
	518	}
c337ddc2 MD	519
	520	/* TODO lttng_dump_idt_table(session); */
	521	/* TODO lttng_dump_softirq_vec(session); */
	522	/* TODO lttng_list_modules(session); */
	523	/* TODO lttng_dump_swap_files(session); */
	524
	525	/*
	526	* Fire off a work queue on each CPU. Their sole purpose in life
	527	* is to guarantee that each CPU has been in a state where is was in
	528	* syscall mode (i.e. not in a trap, an IRQ or a soft IRQ).
	529	*/
	530	get_online_cpus();
	531	atomic_set(&kernel_threads_to_run, num_online_cpus());
	532	for_each_online_cpu(cpu) {
	533	INIT_DELAYED_WORK(&cpu_work[cpu], lttng_statedump_work_func);
	534	schedule_delayed_work_on(cpu, &cpu_work[cpu], 0);
	535	}
	536	/* Wait for all threads to run */
7a7128e0	537	__wait_event(statedump_wq, (atomic_read(&kernel_threads_to_run) == 0));
c337ddc2 MD	538	put_online_cpus();
c337ddc2 MD	539	/* Our work is done */
c337ddc2 MD	540	trace_lttng_statedump_end(session);
	541	return 0;
	542	}
	543
	544	/*
	545	* Called with session mutex held.
	546	*/
	547	int lttng_statedump_start(struct lttng_session *session)
	548	{
c337ddc2 MD	549	return do_lttng_statedump(session);
	550	}
	551	EXPORT_SYMBOL_GPL(lttng_statedump_start);
	552
dd8d5afb MD	553	static
	554	int __init lttng_statedump_init(void)
	555	{
d16aa9c9 MD	556	/*
	557	* Allow module to load even if the fixup cannot be done. This
	558	* will allow seemless transition when the underlying issue fix
	559	* is merged into the Linux kernel, and when tracepoint.c
	560	* "tracepoint_module_notify" is turned into a static function.
	561	*/
	562	(void) wrapper_lttng_fixup_sig(THIS_MODULE);
	563	return 0;
dd8d5afb MD	564	}
	565
	566	module_init(lttng_statedump_init);
	567
461277e7 MD	568	static
	569	void __exit lttng_statedump_exit(void)
	570	{
	571	}
	572
	573	module_exit(lttng_statedump_exit);
	574
c337ddc2 MD	575	MODULE_LICENSE("GPL and additional rights");
	576	MODULE_AUTHOR("Jean-Hugues Deschenes");
	577	MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Statedump");
13ab8b0a MD	578	MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
	579	__stringify(LTTNG_MODULES_MINOR_VERSION) "."
	580	__stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
	581	LTTNG_MODULES_EXTRAVERSION);