X-Git-Url: http://git.liburcu.org/?a=blobdiff_plain;f=lttng-statedump-impl.c;h=fef97812f2b87f2eb89e74783981ddc45441b1f2;hb=refs%2Fheads%2Fstable-2.11;hp=cfcf4febb97b3ceff2a2bcd7c28c53f2fe17cc3d;hpb=c337ddc219f608d4d35f461bdc9d2246324d6708;p=lttng-modules.git

diff --git a/lttng-statedump-impl.c b/lttng-statedump-impl.c
index cfcf4feb..fef97812 100644
--- a/lttng-statedump-impl.c
+++ b/lttng-statedump-impl.c
@@ -1,4 +1,7 @@
-/*
+/* 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>
@@ -8,8 +11,6 @@
  *	Eric Clement:                   Add listing of network IP interface
  *	2006, 2007 Mathieu Desnoyers	Fix kernel threads
  *	                                Various updates
- *
- * Dual LGPL v2.1/GPL v2 license.
  */
 
 #include <linux/init.h>
@@ -25,25 +26,78 @@
 #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 "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/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>
+#include <wrapper/sched.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>
+
+LTTNG_DEFINE_TRACE(lttng_statedump_block_device,
+	TP_PROTO(struct lttng_session *session,
+		dev_t dev, const char *diskname),
+	TP_ARGS(session, dev, diskname));
+
+LTTNG_DEFINE_TRACE(lttng_statedump_end,
+	TP_PROTO(struct lttng_session *session),
+	TP_ARGS(session));
+
+LTTNG_DEFINE_TRACE(lttng_statedump_interrupt,
+	TP_PROTO(struct lttng_session *session,
+		unsigned int irq, const char *chip_name,
+		struct irqaction *action),
+	TP_ARGS(session, irq, chip_name, action));
+
+LTTNG_DEFINE_TRACE(lttng_statedump_file_descriptor,
+	TP_PROTO(struct lttng_session *session,
+		struct task_struct *p, int fd, const char *filename,
+		unsigned int flags, fmode_t fmode),
+	TP_ARGS(session, p, fd, filename, flags, fmode));
+
+LTTNG_DEFINE_TRACE(lttng_statedump_start,
+	TP_PROTO(struct lttng_session *session),
+	TP_ARGS(session));
+
+LTTNG_DEFINE_TRACE(lttng_statedump_process_state,
+	TP_PROTO(struct lttng_session *session,
+		struct task_struct *p,
+		int type, int mode, int submode, int status,
+		struct pid_namespace *pid_ns),
+	TP_ARGS(session, p, type, mode, submode, status, pid_ns));
+
+LTTNG_DEFINE_TRACE(lttng_statedump_network_interface,
+	TP_PROTO(struct lttng_session *session,
+		struct net_device *dev, struct in_ifaddr *ifa),
+	TP_ARGS(session, dev, ifa));
+
+struct lttng_fd_ctx {
+	char *page;
+	struct lttng_session *session;
+	struct task_struct *p;
+	struct files_struct *files;
+};
 
 /*
  * Protected by the trace lock.
@@ -82,7 +136,156 @@ enum lttng_process_status {
 	LTTNG_DEAD = 7,
 };
 
+
+#if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(5,11,0))
+
+#define LTTNG_PART_STRUCT_TYPE struct block_device
+
+static
+int lttng_get_part_name(struct gendisk *disk, struct block_device *part, char *name_buf)
+{
+	const char *p;
+
+	p = bdevname(part, name_buf);
+	if (!p)
+		return -ENOSYS;
+
+	return 0;
+}
+
+static
+dev_t lttng_get_part_devt(struct block_device *part)
+{
+	return part->bd_dev;
+}
+
+#else
+
+#define LTTNG_PART_STRUCT_TYPE struct hd_struct
+
+static
+int lttng_get_part_name(struct gendisk *disk, struct hd_struct *part, char *name_buf)
+{
+	const char *p;
+	struct block_device bdev;
+
+	/*
+	 * Create a partial 'struct blockdevice' to use
+	 * 'bdevname()' which is a simple wrapper over
+	 * 'disk_name()' but has the honor to be EXPORT_SYMBOL.
+	 */
+	bdev.bd_disk = disk;
+	bdev.bd_part = part;
+
+	p = bdevname(&bdev, name_buf);
+	if (!p)
+		return -ENOSYS;
+
+	return 0;
+}
+
+static
+dev_t lttng_get_part_devt(struct hd_struct *part)
+{
+	return part_devt(part);
+}
+#endif
+
+#if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(5,12,0))
+static
+int lttng_statedump_each_block_device(struct lttng_session *session, struct gendisk *disk)
+{
+	struct block_device *part;
+	unsigned long idx;
+	int ret = 0;
+
+	/* Include partition 0 */
+	idx = 0;
+
+	rcu_read_lock();
+	xa_for_each(&disk->part_tbl, idx, part) {
+		char name_buf[BDEVNAME_SIZE];
+
+		/* Exclude non-partitions bdev and empty partitions. */
+		if (bdev_is_partition(part) && !bdev_nr_sectors(part))
+			continue;
+
+		if (lttng_get_part_name(disk, part, name_buf) == -ENOSYS) {
+			ret = -ENOSYS;
+			goto end;
+		}
+		trace_lttng_statedump_block_device(session, lttng_get_part_devt(part),
+				name_buf);
+	}
+end:
+	rcu_read_unlock();
+	return ret;
+}
+#else
+static
+int lttng_statedump_each_block_device(struct lttng_session *session, struct gendisk *disk)
+{
+	struct disk_part_iter piter;
+	LTTNG_PART_STRUCT_TYPE *part;
+
+	disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
+
+	while ((part = disk_part_iter_next(&piter))) {
+		char name_buf[BDEVNAME_SIZE];
+
+		if (lttng_get_part_name(disk, part, name_buf) == -ENOSYS) {
+			disk_part_iter_exit(&piter);
+			return -ENOSYS;
+		}
+		trace_lttng_statedump_block_device(session, lttng_get_part_devt(part),
+				name_buf);
+	}
+	disk_part_iter_exit(&piter);
+
+	return 0;
+}
+#endif
+
+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;
+	int ret = 0;
+
+	ptr_block_class = wrapper_get_block_class();
+	if (!ptr_block_class) {
+		ret = -ENOSYS;
+		goto end;
+	}
+	ptr_disk_type = wrapper_get_disk_type();
+	if (!ptr_disk_type) {
+		ret = -ENOSYS;
+		goto end;
+	}
+	class_dev_iter_init(&iter, ptr_block_class, NULL, ptr_disk_type);
+	while ((dev = class_dev_iter_next(&iter))) {
+		struct gendisk *disk = dev_to_disk(dev);
+
+		/*
+		 * Don't show empty devices or things that have been
+		 * suppressed
+		 */
+		if (get_capacity(disk) == 0 ||
+		    (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
+			continue;
+
+		ret = lttng_statedump_each_block_device(session, disk);
+	}
+	class_dev_iter_exit(&iter);
+end:
+	return ret;
+}
+
 #ifdef CONFIG_INET
+
 static
 void lttng_enumerate_device(struct lttng_session *session,
 		struct net_device *dev)
@@ -126,34 +329,61 @@ int lttng_enumerate_network_ip_interface(struct lttng_session *session)
 }
 #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 fdtable *fdt;
-	struct file *filp;
-	unsigned int i;
-	const unsigned char *path;
+	struct lttng_fd_ctx ctx = { .page = tmp, .session = session, .p = p };
+	struct files_struct *files;
 
 	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);
-		/* 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_unlock(&p->files->file_lock);
-unlock_task:
+	files = p->files;
+	if (!files)
+		goto end;
+	ctx.files = files;
+	lttng_iterate_fd(files, 0, lttng_dump_one_fd, &ctx);
+end:
 	task_unlock(p);
 }
 
@@ -161,7 +391,11 @@ static
 int lttng_enumerate_file_descriptors(struct lttng_session *session)
 {
 	struct task_struct *p;
-	char *tmp = (char *) __get_free_page(GFP_KERNEL);
+	char *tmp;
+
+	tmp = (char *) __get_free_page(GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
 
 	/* Enumerate active file descriptors */
 	rcu_read_lock();
@@ -172,6 +406,12 @@ int lttng_enumerate_file_descriptors(struct lttng_session *session)
 	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)
@@ -190,7 +430,7 @@ void lttng_enumerate_task_vm_maps(struct lttng_session *session,
 		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);
@@ -212,10 +452,16 @@ int lttng_enumerate_vm_maps(struct lttng_session *session)
 	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
 
-#ifdef CONFIG_GENERIC_HARDIRQS
 static
-void lttng_list_interrupts(struct lttng_session *session)
+int lttng_list_interrupts(struct lttng_session *session)
 {
 	unsigned int irq;
 	unsigned long flags = 0;
@@ -229,23 +475,46 @@ void lttng_list_interrupts(struct lttng_session *session)
 			irq_desc_get_chip(desc)->name ? : "unnamed_irq_chip";
 
 		local_irq_save(flags);
-		raw_spin_lock(&desc->lock);
+		wrapper_desc_spin_lock(&desc->lock);
 		for (action = desc->action; action; action = action->next) {
 			trace_lttng_statedump_interrupt(session,
 				irq, irq_chip_name, action);
 		}
-		raw_spin_unlock(&desc->lock);
+		wrapper_desc_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
 
+/*
+ * 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)
 {
@@ -267,7 +536,7 @@ int lttng_enumerate_process_states(struct lttng_session *session)
 				status = LTTNG_ZOMBIE;
 			else if (p->exit_state == EXIT_DEAD)
 				status = LTTNG_DEAD;
-			else if (p->state == TASK_RUNNING) {
+			else if (lttng_task_is_running(p)) {
 				/* Is this a forked child that has not run yet? */
 				if (list_empty(&p->rt.run_list))
 					status = LTTNG_WAIT_FORK;
@@ -278,7 +547,7 @@ int lttng_enumerate_process_states(struct lttng_session *session)
 					 * was really running at this time.
 					 */
 					status = LTTNG_WAIT_CPU;
-			} else if (p->state &
+			} else if (lttng_get_task_state(p) &
 				(TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE)) {
 				/* Task is waiting for something to complete */
 				status = LTTNG_WAIT;
@@ -295,7 +564,7 @@ int lttng_enumerate_process_states(struct lttng_session *session)
 				type = LTTNG_USER_THREAD;
 			else
 				type = LTTNG_KERNEL_THREAD;
-			trace_lttng_statedump_process_state(session,
+			lttng_statedump_process_ns(session,
 				p, type, mode, submode, status);
 			task_unlock(p);
 		} while_each_thread(g, p);
@@ -316,15 +585,37 @@ void lttng_statedump_work_func(struct work_struct *work)
 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);
-	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;
+	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); */
@@ -343,10 +634,9 @@ int do_lttng_statedump(struct lttng_session *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;
 }
@@ -356,11 +646,36 @@ int do_lttng_statedump(struct lttng_session *session)
  */
 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);