[PATCH] pi-futex: rt mutex debug
[pandora-kernel.git] / kernel / rtmutex-debug.c
diff --git a/kernel/rtmutex-debug.c b/kernel/rtmutex-debug.c
new file mode 100644 (file)
index 0000000..4aa8a2c
--- /dev/null
@@ -0,0 +1,513 @@
+/*
+ * RT-Mutexes: blocking mutual exclusion locks with PI support
+ *
+ * started by Ingo Molnar and Thomas Gleixner:
+ *
+ *  Copyright (C) 2004-2006 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
+ *  Copyright (C) 2006 Timesys Corp., Thomas Gleixner <tglx@timesys.com>
+ *
+ * This code is based on the rt.c implementation in the preempt-rt tree.
+ * Portions of said code are
+ *
+ *  Copyright (C) 2004  LynuxWorks, Inc., Igor Manyilov, Bill Huey
+ *  Copyright (C) 2006  Esben Nielsen
+ *  Copyright (C) 2006  Kihon Technologies Inc.,
+ *                     Steven Rostedt <rostedt@goodmis.org>
+ *
+ * See rt.c in preempt-rt for proper credits and further information
+ */
+#include <linux/config.h>
+#include <linux/sched.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <linux/kallsyms.h>
+#include <linux/syscalls.h>
+#include <linux/interrupt.h>
+#include <linux/plist.h>
+#include <linux/fs.h>
+
+#include "rtmutex_common.h"
+
+#ifdef CONFIG_DEBUG_RT_MUTEXES
+# include "rtmutex-debug.h"
+#else
+# include "rtmutex.h"
+#endif
+
+# define TRACE_WARN_ON(x)                      WARN_ON(x)
+# define TRACE_BUG_ON(x)                       BUG_ON(x)
+
+# define TRACE_OFF()                                           \
+do {                                                           \
+       if (rt_trace_on) {                                      \
+               rt_trace_on = 0;                                \
+               console_verbose();                              \
+               if (spin_is_locked(&current->pi_lock))          \
+                       spin_unlock(&current->pi_lock);         \
+               if (spin_is_locked(&current->held_list_lock))   \
+                       spin_unlock(&current->held_list_lock);  \
+       }                                                       \
+} while (0)
+
+# define TRACE_OFF_NOLOCK()                                    \
+do {                                                           \
+       if (rt_trace_on) {                                      \
+               rt_trace_on = 0;                                \
+               console_verbose();                              \
+       }                                                       \
+} while (0)
+
+# define TRACE_BUG_LOCKED()                    \
+do {                                           \
+       TRACE_OFF();                            \
+       BUG();                                  \
+} while (0)
+
+# define TRACE_WARN_ON_LOCKED(c)               \
+do {                                           \
+       if (unlikely(c)) {                      \
+               TRACE_OFF();                    \
+               WARN_ON(1);                     \
+       }                                       \
+} while (0)
+
+# define TRACE_BUG_ON_LOCKED(c)                        \
+do {                                           \
+       if (unlikely(c))                        \
+               TRACE_BUG_LOCKED();             \
+} while (0)
+
+#ifdef CONFIG_SMP
+# define SMP_TRACE_BUG_ON_LOCKED(c)    TRACE_BUG_ON_LOCKED(c)
+#else
+# define SMP_TRACE_BUG_ON_LOCKED(c)    do { } while (0)
+#endif
+
+/*
+ * deadlock detection flag. We turn it off when we detect
+ * the first problem because we dont want to recurse back
+ * into the tracing code when doing error printk or
+ * executing a BUG():
+ */
+int rt_trace_on = 1;
+
+void deadlock_trace_off(void)
+{
+       rt_trace_on = 0;
+}
+
+static void printk_task(task_t *p)
+{
+       if (p)
+               printk("%16s:%5d [%p, %3d]", p->comm, p->pid, p, p->prio);
+       else
+               printk("<none>");
+}
+
+static void printk_task_short(task_t *p)
+{
+       if (p)
+               printk("%s/%d [%p, %3d]", p->comm, p->pid, p, p->prio);
+       else
+               printk("<none>");
+}
+
+static void printk_lock(struct rt_mutex *lock, int print_owner)
+{
+       if (lock->name)
+               printk(" [%p] {%s}\n",
+                       lock, lock->name);
+       else
+               printk(" [%p] {%s:%d}\n",
+                       lock, lock->file, lock->line);
+
+       if (print_owner && rt_mutex_owner(lock)) {
+               printk(".. ->owner: %p\n", lock->owner);
+               printk(".. held by:  ");
+               printk_task(rt_mutex_owner(lock));
+               printk("\n");
+       }
+       if (rt_mutex_owner(lock)) {
+               printk("... acquired at:               ");
+               print_symbol("%s\n", lock->acquire_ip);
+       }
+}
+
+static void printk_waiter(struct rt_mutex_waiter *w)
+{
+       printk("-------------------------\n");
+       printk("| waiter struct %p:\n", w);
+       printk("| w->list_entry: [DP:%p/%p|SP:%p/%p|PRI:%d]\n",
+              w->list_entry.plist.prio_list.prev, w->list_entry.plist.prio_list.next,
+              w->list_entry.plist.node_list.prev, w->list_entry.plist.node_list.next,
+              w->list_entry.prio);
+       printk("| w->pi_list_entry: [DP:%p/%p|SP:%p/%p|PRI:%d]\n",
+              w->pi_list_entry.plist.prio_list.prev, w->pi_list_entry.plist.prio_list.next,
+              w->pi_list_entry.plist.node_list.prev, w->pi_list_entry.plist.node_list.next,
+              w->pi_list_entry.prio);
+       printk("\n| lock:\n");
+       printk_lock(w->lock, 1);
+       printk("| w->ti->task:\n");
+       printk_task(w->task);
+       printk("| blocked at:  ");
+       print_symbol("%s\n", w->ip);
+       printk("-------------------------\n");
+}
+
+static void show_task_locks(task_t *p)
+{
+       switch (p->state) {
+       case TASK_RUNNING:              printk("R"); break;
+       case TASK_INTERRUPTIBLE:        printk("S"); break;
+       case TASK_UNINTERRUPTIBLE:      printk("D"); break;
+       case TASK_STOPPED:              printk("T"); break;
+       case EXIT_ZOMBIE:               printk("Z"); break;
+       case EXIT_DEAD:                 printk("X"); break;
+       default:                        printk("?"); break;
+       }
+       printk_task(p);
+       if (p->pi_blocked_on) {
+               struct rt_mutex *lock = p->pi_blocked_on->lock;
+
+               printk(" blocked on:");
+               printk_lock(lock, 1);
+       } else
+               printk(" (not blocked)\n");
+}
+
+void rt_mutex_show_held_locks(task_t *task, int verbose)
+{
+       struct list_head *curr, *cursor = NULL;
+       struct rt_mutex *lock;
+       task_t *t;
+       unsigned long flags;
+       int count = 0;
+
+       if (!rt_trace_on)
+               return;
+
+       if (verbose) {
+               printk("------------------------------\n");
+               printk("| showing all locks held by: |  (");
+               printk_task_short(task);
+               printk("):\n");
+               printk("------------------------------\n");
+       }
+
+next:
+       spin_lock_irqsave(&task->held_list_lock, flags);
+       list_for_each(curr, &task->held_list_head) {
+               if (cursor && curr != cursor)
+                       continue;
+               lock = list_entry(curr, struct rt_mutex, held_list_entry);
+               t = rt_mutex_owner(lock);
+               WARN_ON(t != task);
+               count++;
+               cursor = curr->next;
+               spin_unlock_irqrestore(&task->held_list_lock, flags);
+
+               printk("\n#%03d:            ", count);
+               printk_lock(lock, 0);
+               goto next;
+       }
+       spin_unlock_irqrestore(&task->held_list_lock, flags);
+
+       printk("\n");
+}
+
+void rt_mutex_show_all_locks(void)
+{
+       task_t *g, *p;
+       int count = 10;
+       int unlock = 1;
+
+       printk("\n");
+       printk("----------------------\n");
+       printk("| showing all tasks: |\n");
+       printk("----------------------\n");
+
+       /*
+        * Here we try to get the tasklist_lock as hard as possible,
+        * if not successful after 2 seconds we ignore it (but keep
+        * trying). This is to enable a debug printout even if a
+        * tasklist_lock-holding task deadlocks or crashes.
+        */
+retry:
+       if (!read_trylock(&tasklist_lock)) {
+               if (count == 10)
+                       printk("hm, tasklist_lock locked, retrying... ");
+               if (count) {
+                       count--;
+                       printk(" #%d", 10-count);
+                       mdelay(200);
+                       goto retry;
+               }
+               printk(" ignoring it.\n");
+               unlock = 0;
+       }
+       if (count != 10)
+               printk(" locked it.\n");
+
+       do_each_thread(g, p) {
+               show_task_locks(p);
+               if (!unlock)
+                       if (read_trylock(&tasklist_lock))
+                               unlock = 1;
+       } while_each_thread(g, p);
+
+       printk("\n");
+
+       printk("-----------------------------------------\n");
+       printk("| showing all locks held in the system: |\n");
+       printk("-----------------------------------------\n");
+
+       do_each_thread(g, p) {
+               rt_mutex_show_held_locks(p, 0);
+               if (!unlock)
+                       if (read_trylock(&tasklist_lock))
+                               unlock = 1;
+       } while_each_thread(g, p);
+
+
+       printk("=============================================\n\n");
+
+       if (unlock)
+               read_unlock(&tasklist_lock);
+}
+
+void rt_mutex_debug_check_no_locks_held(task_t *task)
+{
+       struct rt_mutex_waiter *w;
+       struct list_head *curr;
+       struct rt_mutex *lock;
+
+       if (!rt_trace_on)
+               return;
+       if (!rt_prio(task->normal_prio) && rt_prio(task->prio)) {
+               printk("BUG: PI priority boost leaked!\n");
+               printk_task(task);
+               printk("\n");
+       }
+       if (list_empty(&task->held_list_head))
+               return;
+
+       spin_lock(&task->pi_lock);
+       plist_for_each_entry(w, &task->pi_waiters, pi_list_entry) {
+               TRACE_OFF();
+
+               printk("hm, PI interest held at exit time? Task:\n");
+               printk_task(task);
+               printk_waiter(w);
+               return;
+       }
+       spin_unlock(&task->pi_lock);
+
+       list_for_each(curr, &task->held_list_head) {
+               lock = list_entry(curr, struct rt_mutex, held_list_entry);
+
+               printk("BUG: %s/%d, lock held at task exit time!\n",
+                      task->comm, task->pid);
+               printk_lock(lock, 1);
+               if (rt_mutex_owner(lock) != task)
+                       printk("exiting task is not even the owner??\n");
+       }
+}
+
+int rt_mutex_debug_check_no_locks_freed(const void *from, unsigned long len)
+{
+       const void *to = from + len;
+       struct list_head *curr;
+       struct rt_mutex *lock;
+       unsigned long flags;
+       void *lock_addr;
+
+       if (!rt_trace_on)
+               return 0;
+
+       spin_lock_irqsave(&current->held_list_lock, flags);
+       list_for_each(curr, &current->held_list_head) {
+               lock = list_entry(curr, struct rt_mutex, held_list_entry);
+               lock_addr = lock;
+               if (lock_addr < from || lock_addr >= to)
+                       continue;
+               TRACE_OFF();
+
+               printk("BUG: %s/%d, active lock [%p(%p-%p)] freed!\n",
+                       current->comm, current->pid, lock, from, to);
+               dump_stack();
+               printk_lock(lock, 1);
+               if (rt_mutex_owner(lock) != current)
+                       printk("freeing task is not even the owner??\n");
+               return 1;
+       }
+       spin_unlock_irqrestore(&current->held_list_lock, flags);
+
+       return 0;
+}
+
+void rt_mutex_debug_task_free(struct task_struct *task)
+{
+       WARN_ON(!plist_head_empty(&task->pi_waiters));
+       WARN_ON(task->pi_blocked_on);
+}
+
+/*
+ * We fill out the fields in the waiter to store the information about
+ * the deadlock. We print when we return. act_waiter can be NULL in
+ * case of a remove waiter operation.
+ */
+void debug_rt_mutex_deadlock(int detect, struct rt_mutex_waiter *act_waiter,
+                            struct rt_mutex *lock)
+{
+       struct task_struct *task;
+
+       if (!rt_trace_on || detect || !act_waiter)
+               return;
+
+       task = rt_mutex_owner(act_waiter->lock);
+       if (task && task != current) {
+               act_waiter->deadlock_task_pid = task->pid;
+               act_waiter->deadlock_lock = lock;
+       }
+}
+
+void debug_rt_mutex_print_deadlock(struct rt_mutex_waiter *waiter)
+{
+       struct task_struct *task;
+
+       if (!waiter->deadlock_lock || !rt_trace_on)
+               return;
+
+       task = find_task_by_pid(waiter->deadlock_task_pid);
+       if (!task)
+               return;
+
+       TRACE_OFF_NOLOCK();
+
+       printk("\n============================================\n");
+       printk(  "[ BUG: circular locking deadlock detected! ]\n");
+       printk(  "--------------------------------------------\n");
+       printk("%s/%d is deadlocking current task %s/%d\n\n",
+              task->comm, task->pid, current->comm, current->pid);
+
+       printk("\n1) %s/%d is trying to acquire this lock:\n",
+              current->comm, current->pid);
+       printk_lock(waiter->lock, 1);
+
+       printk("... trying at:                 ");
+       print_symbol("%s\n", waiter->ip);
+
+       printk("\n2) %s/%d is blocked on this lock:\n", task->comm, task->pid);
+       printk_lock(waiter->deadlock_lock, 1);
+
+       rt_mutex_show_held_locks(current, 1);
+       rt_mutex_show_held_locks(task, 1);
+
+       printk("\n%s/%d's [blocked] stackdump:\n\n", task->comm, task->pid);
+       show_stack(task, NULL);
+       printk("\n%s/%d's [current] stackdump:\n\n",
+              current->comm, current->pid);
+       dump_stack();
+       rt_mutex_show_all_locks();
+       printk("[ turning off deadlock detection."
+              "Please report this trace. ]\n\n");
+       local_irq_disable();
+}
+
+void debug_rt_mutex_lock(struct rt_mutex *lock __IP_DECL__)
+{
+       unsigned long flags;
+
+       if (rt_trace_on) {
+               TRACE_WARN_ON_LOCKED(!list_empty(&lock->held_list_entry));
+
+               spin_lock_irqsave(&current->held_list_lock, flags);
+               list_add_tail(&lock->held_list_entry, &current->held_list_head);
+               spin_unlock_irqrestore(&current->held_list_lock, flags);
+
+               lock->acquire_ip = ip;
+       }
+}
+
+void debug_rt_mutex_unlock(struct rt_mutex *lock)
+{
+       unsigned long flags;
+
+       if (rt_trace_on) {
+               TRACE_WARN_ON_LOCKED(rt_mutex_owner(lock) != current);
+               TRACE_WARN_ON_LOCKED(list_empty(&lock->held_list_entry));
+
+               spin_lock_irqsave(&current->held_list_lock, flags);
+               list_del_init(&lock->held_list_entry);
+               spin_unlock_irqrestore(&current->held_list_lock, flags);
+       }
+}
+
+void debug_rt_mutex_proxy_lock(struct rt_mutex *lock,
+                              struct task_struct *powner __IP_DECL__)
+{
+       unsigned long flags;
+
+       if (rt_trace_on) {
+               TRACE_WARN_ON_LOCKED(!list_empty(&lock->held_list_entry));
+
+               spin_lock_irqsave(&powner->held_list_lock, flags);
+               list_add_tail(&lock->held_list_entry, &powner->held_list_head);
+               spin_unlock_irqrestore(&powner->held_list_lock, flags);
+
+               lock->acquire_ip = ip;
+       }
+}
+
+void debug_rt_mutex_proxy_unlock(struct rt_mutex *lock)
+{
+       unsigned long flags;
+
+       if (rt_trace_on) {
+               struct task_struct *owner = rt_mutex_owner(lock);
+
+               TRACE_WARN_ON_LOCKED(!owner);
+               TRACE_WARN_ON_LOCKED(list_empty(&lock->held_list_entry));
+
+               spin_lock_irqsave(&owner->held_list_lock, flags);
+               list_del_init(&lock->held_list_entry);
+               spin_unlock_irqrestore(&owner->held_list_lock, flags);
+       }
+}
+
+void debug_rt_mutex_init_waiter(struct rt_mutex_waiter *waiter)
+{
+       memset(waiter, 0x11, sizeof(*waiter));
+       plist_node_init(&waiter->list_entry, MAX_PRIO);
+       plist_node_init(&waiter->pi_list_entry, MAX_PRIO);
+}
+
+void debug_rt_mutex_free_waiter(struct rt_mutex_waiter *waiter)
+{
+       TRACE_WARN_ON(!plist_node_empty(&waiter->list_entry));
+       TRACE_WARN_ON(!plist_node_empty(&waiter->pi_list_entry));
+       TRACE_WARN_ON(waiter->task);
+       memset(waiter, 0x22, sizeof(*waiter));
+}
+
+void debug_rt_mutex_init(struct rt_mutex *lock, const char *name)
+{
+       void *addr = lock;
+
+       if (rt_trace_on) {
+               rt_mutex_debug_check_no_locks_freed(addr,
+                                                   sizeof(struct rt_mutex));
+               INIT_LIST_HEAD(&lock->held_list_entry);
+               lock->name = name;
+       }
+}
+
+void rt_mutex_deadlock_account_lock(struct rt_mutex *lock, task_t *task)
+{
+}
+
+void rt_mutex_deadlock_account_unlock(struct task_struct *task)
+{
+}
+