[Devel] [PATCH 2/3] restart debug: add final process tree status

Serge E. Hallyn serue at us.ibm.com
Tue Sep 29 09:54:02 PDT 2009


Have tasks in sys_restart keep some status in a list off
of checkpoint_ctx, and print this info when the checkpoint_ctx
is freed.

This is mostly an RFC - in particular the error tracking is
pretty half-hearted so far.  But the info it does spit out
helped me to figured out the coordinator syncing problem
fixed by the previous patch.

Sample dmesg output:
[4568:4568:c/r:free_per_task_status:200] 4 tasks registered, nr_tasks was 0 nr_total 0
[4568:4568:c/r:free_per_task_status:202] active pid was 1, ctx->errno 0
[4568:4568:c/r:free_per_task_status:204] kflags 6 uflags 0 oflags 1
[4568:4568:c/r:free_per_task_status:206] task 0 to run was 4568
[4568:4568:c/r:free_per_task_status:209] pid 4566
[4568:4568:c/r:free_per_task_status:211] it was coordinator
[4568:4568:c/r:free_per_task_status:219] it was running
[4568:4568:c/r:free_per_task_status:209] pid 4570
[4568:4568:c/r:free_per_task_status:213] it was a ghost
[4568:4568:c/r:free_per_task_status:209] pid 4569
[4568:4568:c/r:free_per_task_status:213] it was a ghost
[4568:4568:c/r:free_per_task_status:209] pid 4568
[4568:4568:c/r:free_per_task_status:215] it was the root task
[4568:4568:c/r:free_per_task_status:221] it was a normal task

So, when one task died before hitting sys_restart, the first line would
show '3 tasks registered'.

Signed-off-by: Serge E. Hallyn <serue at us.ibm.com>
---
 checkpoint/restart.c             |  105 ++++++++++++++++++++++++++++++++++++++
 checkpoint/sys.c                 |   49 ++++++++++++++++++
 include/linux/checkpoint_types.h |   20 +++++++
 3 files changed, 174 insertions(+), 0 deletions(-)

diff --git a/checkpoint/restart.c b/checkpoint/restart.c
index 849bda5..1085ed5 100644
--- a/checkpoint/restart.c
+++ b/checkpoint/restart.c
@@ -26,6 +26,98 @@
 #include <linux/checkpoint.h>
 #include <linux/checkpoint_hdr.h>
 
+#ifdef CONFIG_CHECKPOINT_DEBUG
+static struct ckpt_task_status *ckpt_debug_checkin(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s;
+	s = kmalloc(sizeof(*s), GFP_KERNEL);
+	if (!s)
+		return NULL;
+	s->pid = current->pid;
+	s->error = 0;
+	s->flags = RESTART_DBG_WAITING;
+	if (current == ctx->root_task)
+		s->flags |= RESTART_DBG_ROOT;
+	list_add_tail(&s->list, &ctx->per_task_status);
+	return s;
+}
+
+static struct ckpt_task_status *getme(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s = NULL;
+	list_for_each_entry(s, &ctx->per_task_status, list) {
+		if (s->pid == current->pid)
+			break;
+	}
+	if (!s || s->pid != current->pid)
+		return NULL;
+	return s;
+}
+
+static void ckpt_debug_coord(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s;
+
+	s = ckpt_debug_checkin(ctx);
+	if (s)
+		s->flags |= RESTART_DBG_COORD;
+}
+
+static void ckpt_debug_ghost(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s;
+
+	s = ckpt_debug_checkin(ctx);
+	if (s)
+		s->flags |= RESTART_DBG_GHOST;
+}
+
+static void ckpt_debug_normal(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s;
+
+	s = ckpt_debug_checkin(ctx);
+	if (s)
+		s->flags |= RESTART_DBG_NORMAL;
+}
+
+static void ckpt_debug_log_error(struct ckpt_ctx *ctx, int err)
+{
+	struct ckpt_task_status *s = getme(ctx);
+	if (!s)
+		return;
+	s->error = err;
+	s->flags &= ~RESTART_DBG_WAITING;
+	s->flags &= ~RESTART_DBG_RUNNING;
+	if (err)
+		s->flags |= RESTART_DBG_FAILED;
+	else
+		s->flags |= RESTART_DBG_SUCCESS;
+}
+
+static void ckpt_debug_log_running(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s = getme(ctx);
+	if (!s)
+		return;
+	s->flags &= ~RESTART_DBG_WAITING;
+	s->flags |= RESTART_DBG_RUNNING;
+}
+#else
+static inline void ckpt_debug_checkin(struct ckpt_ctx *ctx)
+{}
+static inline void ckpt_debug_coord(struct ckpt_ctx *ctx)
+{}
+static inline void ckpt_debug_ghost(struct ckpt_ctx *ctx)
+{}
+static inline void ckpt_debug_normal(struct ckpt_ctx *ctx)
+{}
+static inline void ckpt_debug_log_error(struct ckpt_ctx *ctx, int err)
+{}
+static inline void ckpt_debug_log_running(struct ckpt_ctx *ctx)
+{}
+#endif
+
 static int _ckpt_read_err(struct ckpt_ctx *ctx, struct ckpt_hdr *h)
 {
 	char *ptr;
@@ -661,6 +753,8 @@ static int do_ghost_task(void)
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
+	ckpt_debug_ghost(ctx);
+
 	/*
 	 * Wait until coordinator task has completed prepare_descendants().
 	 * Note that prepare_descendants() wakes each task one a time
@@ -681,6 +775,8 @@ static int do_ghost_task(void)
 	if (atomic_dec_and_test(&ctx->nr_running))
 		complete(&ctx->all_ready);
 
+	ckpt_debug_log_running(ctx);
+
 	current->flags |= PF_RESTARTING;
 
 	/* Finally wait for all ghosts to be woken up - to die */
@@ -688,6 +784,7 @@ static int do_ghost_task(void)
 				 all_tasks_activated(ctx) ||
 				 ckpt_test_ctx_error(ctx));
 
+	ckpt_debug_log_error(ctx, 0);
 	current->exit_signal = -1;
 	ckpt_ctx_put(ctx);
 	do_exit(0);
@@ -704,6 +801,8 @@ static int do_restore_task(void)
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
+	ckpt_debug_normal(ctx);
+
 	/*
 	 * Wait until coordinator task has completed prepare_descendants().
 	 * Note that prepare_descendants() wakes each task one a time
@@ -731,6 +830,8 @@ static int do_restore_task(void)
 	if (ret < 0)
 		goto out;
 
+	ckpt_debug_log_running(ctx);
+
 	zombie = restore_task(ctx);
 	if (zombie < 0) {
 		ret = zombie;
@@ -759,6 +860,7 @@ static int do_restore_task(void)
 	if (old_ctx)
 		ckpt_ctx_put(old_ctx);
 
+	ckpt_debug_log_error(ctx, ret);
 	/* if we're first to fail - notify others */
 	if (ret < 0 && !ckpt_test_ctx_error(ctx)) {
 		restore_notify_error(ctx, ret);
@@ -945,6 +1047,9 @@ static int do_restore_coord(struct ckpt_ctx *ctx, pid_t pid)
 	struct ckpt_ctx *old_ctx;
 	int ret;
 
+	ckpt_debug_coord(ctx);
+	ckpt_debug_log_running(ctx);
+
 	ret = restore_read_header(ctx);
 	if (ret < 0)
 		return ret;
diff --git a/checkpoint/sys.c b/checkpoint/sys.c
index d48e261..fb3332a 100644
--- a/checkpoint/sys.c
+++ b/checkpoint/sys.c
@@ -188,10 +188,56 @@ static void task_arr_free(struct ckpt_ctx *ctx)
 	kfree(ctx->tasks_arr);
 }
 
+#ifdef CONFIG_CHECKPOINT_DEBUG
+static void free_per_task_status(struct ckpt_ctx *ctx)
+{
+	struct ckpt_task_status *s, *p;
+	int i, count = 0;
+
+	list_for_each_entry(s, &ctx->per_task_status, list)
+		count++;
+	ckpt_debug("%d tasks registered, nr_tasks was %d nr_total %d\n",
+		count, ctx->nr_tasks, atomic_read(&ctx->nr_total));
+	ckpt_debug("active pid was %d, ctx->errno %d\n", ctx->active_pid,
+		ctx->errno);
+	ckpt_debug("kflags %lu uflags %lu oflags %lu", ctx->kflags,
+		ctx->uflags, ctx->oflags);
+	for (i = 0; i < ctx->active_pid; i++)
+		ckpt_debug("task %d to run was %d\n", i, ctx->pids_arr[i].vpid);
+
+	list_for_each_entry_safe(s, p, &ctx->per_task_status, list) {
+		ckpt_debug("pid %d\n", s->pid);
+		if (s->flags & RESTART_DBG_COORD)
+			ckpt_debug("it was coordinator\n");
+		if (s->flags & RESTART_DBG_GHOST)
+			ckpt_debug("it was a ghost\n");
+		if (s->flags & RESTART_DBG_ROOT)
+			ckpt_debug("it was the root task\n");
+		if (s->flags & RESTART_DBG_WAITING)
+			ckpt_debug("it was still waiting to run restart\n");
+		if (s->flags & RESTART_DBG_RUNNING)
+			ckpt_debug("it was running\n");
+		if (s->flags & RESTART_DBG_NORMAL)
+			ckpt_debug("it was a normal task\n");
+		if (s->flags & RESTART_DBG_FAILED)
+			ckpt_debug("it finished with error %d\n", s->error);
+		if (s->flags & RESTART_DBG_FAILED)
+			ckpt_debug("it finished successfully");
+		list_del(&s->list);
+		kfree(s);
+	}
+}
+#else
+static inline void free_per_task_status(struct ckpt_ctx *ctx)
+{ }
+#endif
+
 static void ckpt_ctx_free(struct ckpt_ctx *ctx)
 {
 	BUG_ON(atomic_read(&ctx->refcount));
 
+	free_per_task_status(ctx);
+
 	if (ctx->deferqueue)
 		deferqueue_destroy(ctx->deferqueue);
 
@@ -237,6 +283,9 @@ static struct ckpt_ctx *ckpt_ctx_alloc(int fd, unsigned long uflags,
 
 	atomic_set(&ctx->refcount, 0);
 	INIT_LIST_HEAD(&ctx->pgarr_list);
+#ifdef CONFIG_CHECKPOINT_DEBUG
+	INIT_LIST_HEAD(&ctx->per_task_status);
+#endif
 	INIT_LIST_HEAD(&ctx->pgarr_pool);
 	init_waitqueue_head(&ctx->waitq);
 	init_waitqueue_head(&ctx->ghostq);
diff --git a/include/linux/checkpoint_types.h b/include/linux/checkpoint_types.h
index 2a854e0..1cda085 100644
--- a/include/linux/checkpoint_types.h
+++ b/include/linux/checkpoint_types.h
@@ -76,10 +76,30 @@ struct ckpt_ctx {
 	wait_queue_head_t waitq;	/* waitqueue for restarting tasks */
 	wait_queue_head_t ghostq;	/* waitqueue for ghost tasks */
 	struct cred *realcred, *ecred;	/* tmp storage for cred at restart */
+#ifdef CONFIG_CHECKPOINT_DEBUG
+	struct list_head per_task_status;
+#endif
 
 	struct ckpt_stats stats;	/* statistics */
 };
 
+#ifdef CONFIG_CHECKPOINT_DEBUG
+struct ckpt_task_status {
+	pid_t pid;
+#define RESTART_DBG_ROOT	(1 << 0)
+#define RESTART_DBG_GHOST	(1 << 1)
+#define RESTART_DBG_COORD	(1 << 2)
+#define RESTART_DBG_NORMAL	(1 << 3)
+#define RESTART_DBG_WAITING	(1 << 4)
+#define RESTART_DBG_RUNNING	(1 << 5)
+#define RESTART_DBG_FAILED	(1 << 6)
+#define RESTART_DBG_SUCCESS	(1 << 7)
+	int flags;
+	int error;
+	struct list_head list;
+};
+#endif
+
 #endif /* __KERNEL__ */
 
 #endif /* _LINUX_CHECKPOINT_TYPES_H_ */
-- 
1.6.1

_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers




More information about the Devel mailing list