BUG/MEDIUM: task: Don't free a task that is about to be run.
While running a task, we may try to delete and free a task that is about to
be run, because it's part of the local tasks list, or because rq_next points
to it.
So flag any task that is in the local tasks list to be deleted, instead of
run, by setting t->process to NULL, and re-make rq_next a global,
thread-local variable, that is modified if we attempt to delete that task.
Many thanks to PiBa-NL for reporting this and analysing the problem.
This should be backported to 1.8.
diff --git a/include/proto/task.h b/include/proto/task.h
index cbc1a90..c1c4c07 100644
--- a/include/proto/task.h
+++ b/include/proto/task.h
@@ -90,6 +90,8 @@
extern unsigned int niced_tasks; /* number of niced tasks in the run queue */
extern struct pool_head *pool_head_task;
extern struct pool_head *pool_head_notification;
+extern THREAD_LOCAL struct task *curr_task; /* task currently running or NULL */
+extern THREAD_LOCAL struct eb32sc_node *rq_next; /* Next task to be potentially run */
__decl_hathreads(extern HA_SPINLOCK_T rq_lock); /* spin lock related to run queue */
__decl_hathreads(extern HA_SPINLOCK_T wq_lock); /* spin lock related to wait queue */
@@ -177,8 +179,11 @@
static inline struct task *task_unlink_rq(struct task *t)
{
HA_SPIN_LOCK(TASK_RQ_LOCK, &rq_lock);
- if (likely(task_in_rq(t)))
+ if (likely(task_in_rq(t))) {
+ if (&t->rq == rq_next)
+ rq_next = eb32sc_next(rq_next, tid_bit);
__task_unlink_rq(t);
+ }
HA_SPIN_UNLOCK(TASK_RQ_LOCK, &rq_lock);
return t;
}
@@ -230,7 +235,7 @@
* Free a task. Its context must have been freed since it will be lost.
* The task count is decremented.
*/
-static inline void task_free(struct task *t)
+static inline void __task_free(struct task *t)
{
pool_free(pool_head_task, t);
if (unlikely(stopping))
@@ -238,6 +243,18 @@
HA_ATOMIC_SUB(&nb_tasks, 1);
}
+static inline void task_free(struct task *t)
+{
+ /* There's no need to protect t->state with a lock, as the task
+ * has to run on the current thread.
+ */
+ if (t == curr_task || !(t->state & TASK_RUNNING))
+ __task_free(t);
+ else
+ t->process = NULL;
+}
+
+
/* Place <task> into the wait queue, where it may already be. If the expiration
* timer is infinite, do nothing and rely on wake_expired_task to clean up.
*/
diff --git a/src/task.c b/src/task.c
index fd9acf6..3d021bb 100644
--- a/src/task.c
+++ b/src/task.c
@@ -39,6 +39,7 @@
unsigned int niced_tasks = 0; /* number of niced tasks in the run queue */
THREAD_LOCAL struct task *curr_task = NULL; /* task currently running or NULL */
+THREAD_LOCAL struct eb32sc_node *rq_next = NULL; /* Next task to be potentially run */
__decl_hathreads(HA_SPINLOCK_T __attribute__((aligned(64))) rq_lock); /* spin lock related to run queue */
__decl_hathreads(HA_SPINLOCK_T __attribute__((aligned(64))) wq_lock); /* spin lock related to wait queue */
@@ -186,7 +187,6 @@
struct task *t;
int i;
int max_processed;
- struct eb32sc_node *rq_next;
struct task *local_tasks[16];
int local_tasks_count;
int final_tasks_count;
@@ -227,8 +227,14 @@
*/
if (likely(t->process == process_stream))
t = process_stream(t);
- else
- t = t->process(t);
+ else {
+ if (t->process != NULL)
+ t = t->process(t);
+ else {
+ __task_free(t);
+ t = NULL;
+ }
+ }
curr_task = NULL;
if (likely(t != NULL)) {
@@ -309,8 +315,14 @@
curr_task = t;
if (likely(t->process == process_stream))
t = process_stream(t);
- else
- t = t->process(t);
+ else {
+ if (t->process != NULL)
+ t = t->process(t);
+ else {
+ __task_free(t);
+ t = NULL;
+ }
+ }
curr_task = NULL;
if (t)
local_tasks[final_tasks_count++] = t;