-/*
+/* SPDX-License-Identifier: (GPL-2.0 or LGPL-2.1)
+ *
* 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
#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 "wrapper/irqdesc.h"
-
-#ifdef CONFIG_GENERIC_HARDIRQS
+#include <linux/device.h>
+
+#include <lttng-events.h>
+#include <lttng-tracer.h>
+#include <wrapper/irqdesc.h>
+#include <wrapper/fdtable.h>
+#include <wrapper/namespace.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_PATH instrumentation/events/lttng-module
#define TRACE_INCLUDE_FILE lttng-statedump
-#include "instrumentation/events/lttng-module/lttng-statedump.h"
+#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_process_pid_ns);
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,6,0))
+DEFINE_TRACE(lttng_statedump_process_cgroup_ns);
+#endif
+DEFINE_TRACE(lttng_statedump_process_ipc_ns);
+#ifndef LTTNG_MNT_NS_MISSING_HEADER
+DEFINE_TRACE(lttng_statedump_process_mnt_ns);
+#endif
+DEFINE_TRACE(lttng_statedump_process_net_ns);
+DEFINE_TRACE(lttng_statedump_process_user_ns);
+DEFINE_TRACE(lttng_statedump_process_uts_ns);
+DEFINE_TRACE(lttng_statedump_network_interface);
+#ifdef LTTNG_HAVE_STATEDUMP_CPU_TOPOLOGY
+DEFINE_TRACE(lttng_statedump_cpu_topology);
+#endif
+
+struct lttng_fd_ctx {
+ char *page;
+ struct lttng_session *session;
+ struct files_struct *files;
+};
/*
* Protected by the trace lock.
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)
}
#endif /* CONFIG_INET */
-
static
-void lttng_enumerate_task_fd(struct lttng_session *session,
- struct task_struct *p, char *tmp)
+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;
- struct file *filp;
- unsigned int i;
- const unsigned char *path;
-
- task_lock(p);
- if (!p->files)
- goto unlock_task;
- spin_lock(&p->files->file_lock);
- fdt = files_fdtable(p->files);
- for (i = 0; i < fdt->max_fds; i++) {
- filp = fcheck_files(p->files, i);
- if (!filp)
- continue;
- path = d_path(&filp->f_path, tmp, PAGE_SIZE);
+
+ /*
+ * 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 */
- trace_lttng_statedump_file_descriptor(session, p, i,
- IS_ERR(path) ?
- filp->f_dentry->d_name.name :
- path);
+ spin_lock(&dentry->d_lock);
+ trace_lttng_statedump_file_descriptor(ctx->session,
+ ctx->files, fd, dentry->d_name.name, flags,
+ file->f_mode);
+ spin_unlock(&dentry->d_lock);
+ goto end;
}
- spin_unlock(&p->files->file_lock);
-unlock_task:
- task_unlock(p);
+ trace_lttng_statedump_file_descriptor(ctx->session,
+ ctx->files, fd, s, flags, file->f_mode);
+end:
+ return 0;
}
+/* Called with task lock held. */
static
-int lttng_enumerate_file_descriptors(struct lttng_session *session)
+void lttng_enumerate_files(struct lttng_session *session,
+ struct files_struct *files,
+ char *tmp)
{
- struct task_struct *p;
- char *tmp = (char *) __get_free_page(GFP_KERNEL);
+ struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .files = files, };
- /* 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);
+ lttng_iterate_fd(files, 0, lttng_dump_one_fd, &ctx);
+}
+
+#ifdef LTTNG_HAVE_STATEDUMP_CPU_TOPOLOGY
+static
+int lttng_enumerate_cpu_topology(struct lttng_session *session)
+{
+ int cpu;
+ const cpumask_t *cpumask = cpu_possible_mask;
+
+ for (cpu = cpumask_first(cpumask); cpu < nr_cpu_ids;
+ cpu = cpumask_next(cpu, cpumask)) {
+ trace_lttng_statedump_cpu_topology(session, &cpu_data(cpu));
+ }
+
+ return 0;
+}
+#else
+static
+int lttng_enumerate_cpu_topology(struct lttng_session *session)
+{
return 0;
}
+#endif
#if 0
/*
down_read(&mm->mmap_sem);
while (map) {
if (map->vm_file)
- ino = map->vm_file->f_dentry->d_inode->i_ino;
+ ino = map->vm_file->lttng_f_dentry->d_inode->i_ino;
else
ino = 0;
trace_lttng_statedump_vm_map(session, p, map, ino);
}
#endif
-#ifdef CONFIG_GENERIC_HARDIRQS
-
-#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39))
-#define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
-#endif
+#ifdef CONFIG_LTTNG_HAS_LIST_IRQ
static
-void lttng_list_interrupts(struct lttng_session *session)
+int lttng_list_interrupts(struct lttng_session *session)
{
unsigned int irq;
unsigned long flags = 0;
raw_spin_unlock(&desc->lock);
local_irq_restore(flags);
}
+ return 0;
#undef irq_to_desc
}
#else
static inline
-void list_interrupts(struct lttng_session *session)
+int lttng_list_interrupts(struct lttng_session *session)
{
+ return 0;
}
#endif
+/*
+ * Statedump the task's namespaces using the proc filesystem inode number as
+ * the unique identifier. The user and pid ns are nested and will be dumped
+ * recursively.
+ *
+ * Called with task lock held.
+ */
static
void lttng_statedump_process_ns(struct lttng_session *session,
struct task_struct *p,
{
struct nsproxy *proxy;
struct pid_namespace *pid_ns;
+ struct user_namespace *user_ns;
+
+ /*
+ * The pid and user namespaces are special, they are nested and
+ * accessed with specific functions instead of the nsproxy struct
+ * like the other namespaces.
+ */
+ pid_ns = task_active_pid_ns(p);
+ do {
+ trace_lttng_statedump_process_pid_ns(session, p, pid_ns);
+ pid_ns = pid_ns ? pid_ns->parent : NULL;
+ } while (pid_ns);
+
+
+ user_ns = task_cred_xxx(p, user_ns);
+ do {
+ trace_lttng_statedump_process_user_ns(session, p, user_ns);
+ /*
+ * trace_lttng_statedump_process_user_ns() internally
+ * checks whether user_ns is NULL. While this does not
+ * appear to be a possible return value for
+ * task_cred_xxx(), err on the safe side and check
+ * for NULL here as well to be consistent with the
+ * paranoid behavior of
+ * trace_lttng_statedump_process_user_ns().
+ */
+ user_ns = user_ns ? user_ns->lttng_user_ns_parent : NULL;
+ } while (user_ns);
+ /*
+ * Back and forth on locking strategy within Linux upstream for nsproxy.
+ * See Linux upstream commit 728dba3a39c66b3d8ac889ddbe38b5b1c264aec3
+ * "namespaces: Use task_lock and not rcu to protect nsproxy"
+ * for details.
+ */
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0) || \
+ LTTNG_UBUNTU_KERNEL_RANGE(3,13,11,36, 3,14,0,0) || \
+ LTTNG_UBUNTU_KERNEL_RANGE(3,16,1,11, 3,17,0,0) || \
+ LTTNG_RHEL_KERNEL_RANGE(3,10,0,229,13,0, 3,11,0,0,0,0))
+ proxy = p->nsproxy;
+#else
rcu_read_lock();
proxy = task_nsproxy(p);
+#endif
if (proxy) {
- pid_ns = proxy->pid_ns;
- 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);
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,6,0))
+ trace_lttng_statedump_process_cgroup_ns(session, p, proxy->cgroup_ns);
+#endif
+ trace_lttng_statedump_process_ipc_ns(session, p, proxy->ipc_ns);
+#ifndef LTTNG_MNT_NS_MISSING_HEADER
+ trace_lttng_statedump_process_mnt_ns(session, p, proxy->mnt_ns);
+#endif
+ trace_lttng_statedump_process_net_ns(session, p, proxy->net_ns);
+ trace_lttng_statedump_process_uts_ns(session, p, proxy->uts_ns);
}
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0) || \
+ LTTNG_UBUNTU_KERNEL_RANGE(3,13,11,36, 3,14,0,0) || \
+ LTTNG_UBUNTU_KERNEL_RANGE(3,16,1,11, 3,17,0,0) || \
+ LTTNG_RHEL_KERNEL_RANGE(3,10,0,229,13,0, 3,11,0,0,0,0))
+ /* (nothing) */
+#else
rcu_read_unlock();
+#endif
}
static
int lttng_enumerate_process_states(struct lttng_session *session)
{
struct task_struct *g, *p;
+ char *tmp;
+
+ tmp = (char *) __get_free_page(GFP_KERNEL);
+ if (!tmp)
+ return -ENOMEM;
rcu_read_lock();
for_each_process(g) {
+ struct files_struct *prev_files = NULL;
+
p = g;
do {
enum lttng_execution_mode mode =
LTTNG_UNKNOWN;
enum lttng_process_status status;
enum lttng_thread_type type;
+ struct files_struct *files;
task_lock(p);
if (p->exit_state == EXIT_ZOMBIE)
type = LTTNG_USER_THREAD;
else
type = LTTNG_KERNEL_THREAD;
+ files = p->files;
+
+ trace_lttng_statedump_process_state(session,
+ p, type, mode, submode, status, files);
lttng_statedump_process_ns(session,
p, type, mode, submode, status);
+ /*
+ * As an optimisation for the common case, do not
+ * repeat information for the same files_struct in
+ * two consecutive threads. This is the common case
+ * for threads sharing the same fd table. RCU guarantees
+ * that the same files_struct pointer is not re-used
+ * throughout processes/threads iteration.
+ */
+ if (files && files != prev_files) {
+ lttng_enumerate_files(session, files, tmp);
+ prev_files = files;
+ }
task_unlock(p);
} while_each_thread(g, p);
}
rcu_read_unlock();
+ free_page((unsigned long) tmp);
+
return 0;
}
static
int do_lttng_statedump(struct lttng_session *session)
{
- int cpu;
+ int cpu, ret;
- printk(KERN_DEBUG "LTT state dump thread start\n");
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);
+ ret = lttng_enumerate_process_states(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;
+ }
+ ret = lttng_enumerate_cpu_topology(session);
+ if (ret)
+ return ret;
/* TODO lttng_dump_idt_table(session); */
/* TODO lttng_dump_softirq_vec(session); */
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));
+ __wait_event(statedump_wq, (atomic_read(&kernel_threads_to_run) == 0));
put_online_cpus();
/* Our work is done */
- printk(KERN_DEBUG "LTT state dump end\n");
trace_lttng_statedump_end(session);
return 0;
}
*/
int lttng_statedump_start(struct lttng_session *session)
{
- printk(KERN_DEBUG "LTTng: state dump begin\n");
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_DESCRIPTION("LTTng statedump provider");
+MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
+ __stringify(LTTNG_MODULES_MINOR_VERSION) "."
+ __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
+ LTTNG_MODULES_EXTRAVERSION);