]> rtime.felk.cvut.cz Git - linux-imx.git/commitdiff
Merge tag 'trace-fixes-3.11-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
authorLinus Torvalds <torvalds@linux-foundation.org>
Wed, 7 Aug 2013 20:01:30 +0000 (13:01 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 7 Aug 2013 20:01:30 +0000 (13:01 -0700)
Pull tracing fixes from Steven Rostedt:
 "Oleg Nesterov has been working hard in closing all the holes that can
  lead to race conditions between deleting an event and accessing an
  event debugfs file.  This included a fix to the debugfs system (acked
  by Greg Kroah-Hartman).  We think that all the holes have been patched
  and hopefully we don't find more.  I haven't marked all of them for
  stable because I need to examine them more to figure out how far back
  some of the changes need to go.

  Along the way, some other fixes have been made.  Alexander Z Lam fixed
  some logic where the wrong buffer was being modifed.

  Andrew Vagin found a possible corruption for machines that actually
  allocate cpumask, as a reference to one was being zeroed out by
  mistake.

  Dhaval Giani found a bad prototype when tracing is not configured.

  And I not only had some changes to help Oleg, but also finally fixed a
  long standing bug that Dave Jones and others have been hitting, where
  a module unload and reload can cause the function tracing accounting
  to get screwed up"

* tag 'trace-fixes-3.11-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Fix reset of time stamps during trace_clock changes
  tracing: Make TRACE_ITER_STOP_ON_FREE stop the correct buffer
  tracing: Fix trace_dump_stack() proto when CONFIG_TRACING is not set
  tracing: Fix fields of struct trace_iterator that are zeroed by mistake
  tracing/uprobes: Fail to unregister if probe event files are in use
  tracing/kprobes: Fail to unregister if probe event files are in use
  tracing: Add comment to describe special break case in probe_remove_event_call()
  tracing: trace_remove_event_call() should fail if call/file is in use
  debugfs: debugfs_remove_recursive() must not rely on list_empty(d_subdirs)
  ftrace: Check module functions being traced on reload
  ftrace: Consolidate some duplicate code for updating ftrace ops
  tracing: Change remove_event_file_dir() to clear "d_subdirs"->i_private
  tracing: Introduce remove_event_file_dir()
  tracing: Change f_start() to take event_mutex and verify i_private != NULL
  tracing: Change event_filter_read/write to verify i_private != NULL
  tracing: Change event_enable/disable_read() to verify i_private != NULL
  tracing: Turn event/id->i_private into call->event.type

fs/debugfs/inode.c
include/linux/ftrace_event.h
include/linux/kernel.h
kernel/trace/ftrace.c
kernel/trace/trace.c
kernel/trace/trace_events.c
kernel/trace/trace_events_filter.c
kernel/trace/trace_kprobe.c
kernel/trace/trace_uprobe.c

index 4888cb3fdef76037b7af5f9db898bdcb82ce48c2..c7c83ff0f752ce6a86d6e42a420e34ed996ae8ca 100644 (file)
@@ -533,8 +533,7 @@ EXPORT_SYMBOL_GPL(debugfs_remove);
  */
 void debugfs_remove_recursive(struct dentry *dentry)
 {
-       struct dentry *child;
-       struct dentry *parent;
+       struct dentry *child, *next, *parent;
 
        if (IS_ERR_OR_NULL(dentry))
                return;
@@ -544,61 +543,37 @@ void debugfs_remove_recursive(struct dentry *dentry)
                return;
 
        parent = dentry;
+ down:
        mutex_lock(&parent->d_inode->i_mutex);
+       list_for_each_entry_safe(child, next, &parent->d_subdirs, d_u.d_child) {
+               if (!debugfs_positive(child))
+                       continue;
 
-       while (1) {
-               /*
-                * When all dentries under "parent" has been removed,
-                * walk up the tree until we reach our starting point.
-                */
-               if (list_empty(&parent->d_subdirs)) {
-                       mutex_unlock(&parent->d_inode->i_mutex);
-                       if (parent == dentry)
-                               break;
-                       parent = parent->d_parent;
-                       mutex_lock(&parent->d_inode->i_mutex);
-               }
-               child = list_entry(parent->d_subdirs.next, struct dentry,
-                               d_u.d_child);
- next_sibling:
-
-               /*
-                * If "child" isn't empty, walk down the tree and
-                * remove all its descendants first.
-                */
+               /* perhaps simple_empty(child) makes more sense */
                if (!list_empty(&child->d_subdirs)) {
                        mutex_unlock(&parent->d_inode->i_mutex);
                        parent = child;
-                       mutex_lock(&parent->d_inode->i_mutex);
-                       continue;
+                       goto down;
                }
-               __debugfs_remove(child, parent);
-               if (parent->d_subdirs.next == &child->d_u.d_child) {
-                       /*
-                        * Try the next sibling.
-                        */
-                       if (child->d_u.d_child.next != &parent->d_subdirs) {
-                               child = list_entry(child->d_u.d_child.next,
-                                                  struct dentry,
-                                                  d_u.d_child);
-                               goto next_sibling;
-                       }
-
-                       /*
-                        * Avoid infinite loop if we fail to remove
-                        * one dentry.
-                        */
-                       mutex_unlock(&parent->d_inode->i_mutex);
-                       break;
-               }
-               simple_release_fs(&debugfs_mount, &debugfs_mount_count);
+ up:
+               if (!__debugfs_remove(child, parent))
+                       simple_release_fs(&debugfs_mount, &debugfs_mount_count);
        }
 
-       parent = dentry->d_parent;
+       mutex_unlock(&parent->d_inode->i_mutex);
+       child = parent;
+       parent = parent->d_parent;
        mutex_lock(&parent->d_inode->i_mutex);
-       __debugfs_remove(dentry, parent);
+
+       if (child != dentry) {
+               next = list_entry(child->d_u.d_child.next, struct dentry,
+                                       d_u.d_child);
+               goto up;
+       }
+
+       if (!__debugfs_remove(child, parent))
+               simple_release_fs(&debugfs_mount, &debugfs_mount_count);
        mutex_unlock(&parent->d_inode->i_mutex);
-       simple_release_fs(&debugfs_mount, &debugfs_mount_count);
 }
 EXPORT_SYMBOL_GPL(debugfs_remove_recursive);
 
index 4372658c73ae5eddd3dd6570d4dfac8c5d8825c8..120d57a1c3a5dcb452a75d4fd21a6d338b128a0d 100644 (file)
@@ -78,6 +78,11 @@ struct trace_iterator {
        /* trace_seq for __print_flags() and __print_symbolic() etc. */
        struct trace_seq        tmp_seq;
 
+       cpumask_var_t           started;
+
+       /* it's true when current open file is snapshot */
+       bool                    snapshot;
+
        /* The below is zeroed out in pipe_read */
        struct trace_seq        seq;
        struct trace_entry      *ent;
@@ -90,10 +95,7 @@ struct trace_iterator {
        loff_t                  pos;
        long                    idx;
 
-       cpumask_var_t           started;
-
-       /* it's true when current open file is snapshot */
-       bool                    snapshot;
+       /* All new field here will be zeroed out in pipe_read */
 };
 
 enum trace_iter_flags {
@@ -332,7 +334,7 @@ extern int trace_define_field(struct ftrace_event_call *call, const char *type,
                              const char *name, int offset, int size,
                              int is_signed, int filter_type);
 extern int trace_add_event_call(struct ftrace_event_call *call);
-extern void trace_remove_event_call(struct ftrace_event_call *call);
+extern int trace_remove_event_call(struct ftrace_event_call *call);
 
 #define is_signed_type(type)   (((type)(-1)) < (type)1)
 
index 3bef14c6586bc9d7d875e6579352a62e53e13c6a..482ad2d84a32dae333c74cccb2d1394b247d5ae8 100644 (file)
@@ -629,7 +629,7 @@ extern void ftrace_dump(enum ftrace_dump_mode oops_dump_mode);
 static inline void tracing_start(void) { }
 static inline void tracing_stop(void) { }
 static inline void ftrace_off_permanent(void) { }
-static inline void trace_dump_stack(void) { }
+static inline void trace_dump_stack(int skip) { }
 
 static inline void tracing_on(void) { }
 static inline void tracing_off(void) { }
index 8ce9eefc5bb4f6aae3ffa520152a51eb033adb1e..a6d098c6df3f47f960004019fb48468ecea64410 100644 (file)
@@ -2169,12 +2169,57 @@ static cycle_t          ftrace_update_time;
 static unsigned long   ftrace_update_cnt;
 unsigned long          ftrace_update_tot_cnt;
 
-static int ops_traces_mod(struct ftrace_ops *ops)
+static inline int ops_traces_mod(struct ftrace_ops *ops)
 {
-       struct ftrace_hash *hash;
+       /*
+        * Filter_hash being empty will default to trace module.
+        * But notrace hash requires a test of individual module functions.
+        */
+       return ftrace_hash_empty(ops->filter_hash) &&
+               ftrace_hash_empty(ops->notrace_hash);
+}
+
+/*
+ * Check if the current ops references the record.
+ *
+ * If the ops traces all functions, then it was already accounted for.
+ * If the ops does not trace the current record function, skip it.
+ * If the ops ignores the function via notrace filter, skip it.
+ */
+static inline bool
+ops_references_rec(struct ftrace_ops *ops, struct dyn_ftrace *rec)
+{
+       /* If ops isn't enabled, ignore it */
+       if (!(ops->flags & FTRACE_OPS_FL_ENABLED))
+               return 0;
+
+       /* If ops traces all mods, we already accounted for it */
+       if (ops_traces_mod(ops))
+               return 0;
+
+       /* The function must be in the filter */
+       if (!ftrace_hash_empty(ops->filter_hash) &&
+           !ftrace_lookup_ip(ops->filter_hash, rec->ip))
+               return 0;
+
+       /* If in notrace hash, we ignore it too */
+       if (ftrace_lookup_ip(ops->notrace_hash, rec->ip))
+               return 0;
+
+       return 1;
+}
+
+static int referenced_filters(struct dyn_ftrace *rec)
+{
+       struct ftrace_ops *ops;
+       int cnt = 0;
 
-       hash = ops->filter_hash;
-       return ftrace_hash_empty(hash);
+       for (ops = ftrace_ops_list; ops != &ftrace_list_end; ops = ops->next) {
+               if (ops_references_rec(ops, rec))
+                   cnt++;
+       }
+
+       return cnt;
 }
 
 static int ftrace_update_code(struct module *mod)
@@ -2183,6 +2228,7 @@ static int ftrace_update_code(struct module *mod)
        struct dyn_ftrace *p;
        cycle_t start, stop;
        unsigned long ref = 0;
+       bool test = false;
        int i;
 
        /*
@@ -2196,9 +2242,12 @@ static int ftrace_update_code(struct module *mod)
 
                for (ops = ftrace_ops_list;
                     ops != &ftrace_list_end; ops = ops->next) {
-                       if (ops->flags & FTRACE_OPS_FL_ENABLED &&
-                           ops_traces_mod(ops))
-                               ref++;
+                       if (ops->flags & FTRACE_OPS_FL_ENABLED) {
+                               if (ops_traces_mod(ops))
+                                       ref++;
+                               else
+                                       test = true;
+                       }
                }
        }
 
@@ -2208,12 +2257,16 @@ static int ftrace_update_code(struct module *mod)
        for (pg = ftrace_new_pgs; pg; pg = pg->next) {
 
                for (i = 0; i < pg->index; i++) {
+                       int cnt = ref;
+
                        /* If something went wrong, bail without enabling anything */
                        if (unlikely(ftrace_disabled))
                                return -1;
 
                        p = &pg->records[i];
-                       p->flags = ref;
+                       if (test)
+                               cnt += referenced_filters(p);
+                       p->flags = cnt;
 
                        /*
                         * Do the initial record conversion from mcount jump
@@ -2233,7 +2286,7 @@ static int ftrace_update_code(struct module *mod)
                         * conversion puts the module to the correct state, thus
                         * passing the ftrace_make_call check.
                         */
-                       if (ftrace_start_up && ref) {
+                       if (ftrace_start_up && cnt) {
                                int failed = __ftrace_replace_code(p, 1);
                                if (failed)
                                        ftrace_bug(failed, p->ip);
@@ -3384,6 +3437,12 @@ ftrace_match_addr(struct ftrace_hash *hash, unsigned long ip, int remove)
        return add_hash_entry(hash, ip);
 }
 
+static void ftrace_ops_update_code(struct ftrace_ops *ops)
+{
+       if (ops->flags & FTRACE_OPS_FL_ENABLED && ftrace_enabled)
+               ftrace_run_update_code(FTRACE_UPDATE_CALLS);
+}
+
 static int
 ftrace_set_hash(struct ftrace_ops *ops, unsigned char *buf, int len,
                unsigned long ip, int remove, int reset, int enable)
@@ -3426,9 +3485,8 @@ ftrace_set_hash(struct ftrace_ops *ops, unsigned char *buf, int len,
 
        mutex_lock(&ftrace_lock);
        ret = ftrace_hash_move(ops, enable, orig_hash, hash);
-       if (!ret && ops->flags & FTRACE_OPS_FL_ENABLED
-           && ftrace_enabled)
-               ftrace_run_update_code(FTRACE_UPDATE_CALLS);
+       if (!ret)
+               ftrace_ops_update_code(ops);
 
        mutex_unlock(&ftrace_lock);
 
@@ -3655,9 +3713,8 @@ int ftrace_regex_release(struct inode *inode, struct file *file)
                mutex_lock(&ftrace_lock);
                ret = ftrace_hash_move(iter->ops, filter_hash,
                                       orig_hash, iter->hash);
-               if (!ret && (iter->ops->flags & FTRACE_OPS_FL_ENABLED)
-                   && ftrace_enabled)
-                       ftrace_run_update_code(FTRACE_UPDATE_CALLS);
+               if (!ret)
+                       ftrace_ops_update_code(iter->ops);
 
                mutex_unlock(&ftrace_lock);
        }
index 882ec1dd15155233f89c8c7bad0d2a6dbf58c45f..496f94d57698294966d2c0c9bd10de34854b34a5 100644 (file)
@@ -243,20 +243,25 @@ int filter_current_check_discard(struct ring_buffer *buffer,
 }
 EXPORT_SYMBOL_GPL(filter_current_check_discard);
 
-cycle_t ftrace_now(int cpu)
+cycle_t buffer_ftrace_now(struct trace_buffer *buf, int cpu)
 {
        u64 ts;
 
        /* Early boot up does not have a buffer yet */
-       if (!global_trace.trace_buffer.buffer)
+       if (!buf->buffer)
                return trace_clock_local();
 
-       ts = ring_buffer_time_stamp(global_trace.trace_buffer.buffer, cpu);
-       ring_buffer_normalize_time_stamp(global_trace.trace_buffer.buffer, cpu, &ts);
+       ts = ring_buffer_time_stamp(buf->buffer, cpu);
+       ring_buffer_normalize_time_stamp(buf->buffer, cpu, &ts);
 
        return ts;
 }
 
+cycle_t ftrace_now(int cpu)
+{
+       return buffer_ftrace_now(&global_trace.trace_buffer, cpu);
+}
+
 /**
  * tracing_is_enabled - Show if global_trace has been disabled
  *
@@ -1211,7 +1216,7 @@ void tracing_reset_online_cpus(struct trace_buffer *buf)
        /* Make sure all commits have finished */
        synchronize_sched();
 
-       buf->time_start = ftrace_now(buf->cpu);
+       buf->time_start = buffer_ftrace_now(buf, buf->cpu);
 
        for_each_online_cpu(cpu)
                ring_buffer_reset_cpu(buffer, cpu);
@@ -1219,11 +1224,6 @@ void tracing_reset_online_cpus(struct trace_buffer *buf)
        ring_buffer_record_enable(buffer);
 }
 
-void tracing_reset_current(int cpu)
-{
-       tracing_reset(&global_trace.trace_buffer, cpu);
-}
-
 /* Must have trace_types_lock held */
 void tracing_reset_all_online_cpus(void)
 {
@@ -4151,6 +4151,7 @@ waitagain:
        memset(&iter->seq, 0,
               sizeof(struct trace_iterator) -
               offsetof(struct trace_iterator, seq));
+       cpumask_clear(iter->started);
        iter->pos = -1;
 
        trace_event_read_lock();
@@ -4468,7 +4469,7 @@ tracing_free_buffer_release(struct inode *inode, struct file *filp)
 
        /* disable tracing ? */
        if (trace_flags & TRACE_ITER_STOP_ON_FREE)
-               tracing_off();
+               tracer_tracing_off(tr);
        /* resize the ring buffer to 0 */
        tracing_resize_ring_buffer(tr, 0, RING_BUFFER_ALL_CPUS);
 
@@ -4633,12 +4634,12 @@ static ssize_t tracing_clock_write(struct file *filp, const char __user *ubuf,
         * New clock may not be consistent with the previous clock.
         * Reset the buffer so that it doesn't have incomparable timestamps.
         */
-       tracing_reset_online_cpus(&global_trace.trace_buffer);
+       tracing_reset_online_cpus(&tr->trace_buffer);
 
 #ifdef CONFIG_TRACER_MAX_TRACE
        if (tr->flags & TRACE_ARRAY_FL_GLOBAL && tr->max_buffer.buffer)
                ring_buffer_set_clock(tr->max_buffer.buffer, trace_clocks[i].func);
-       tracing_reset_online_cpus(&global_trace.max_buffer);
+       tracing_reset_online_cpus(&tr->max_buffer);
 #endif
 
        mutex_unlock(&trace_types_lock);
index 898f868833f2d4df0d80838f9461385b5eedcd77..29a7ebcfb42621d05cb9a74a7bcf291515b38461 100644 (file)
@@ -409,33 +409,42 @@ static void put_system(struct ftrace_subsystem_dir *dir)
        mutex_unlock(&event_mutex);
 }
 
-/*
- * Open and update trace_array ref count.
- * Must have the current trace_array passed to it.
- */
-static int tracing_open_generic_file(struct inode *inode, struct file *filp)
+static void remove_subsystem(struct ftrace_subsystem_dir *dir)
 {
-       struct ftrace_event_file *file = inode->i_private;
-       struct trace_array *tr = file->tr;
-       int ret;
+       if (!dir)
+               return;
 
-       if (trace_array_get(tr) < 0)
-               return -ENODEV;
+       if (!--dir->nr_events) {
+               debugfs_remove_recursive(dir->entry);
+               list_del(&dir->list);
+               __put_system_dir(dir);
+       }
+}
 
-       ret = tracing_open_generic(inode, filp);
-       if (ret < 0)
-               trace_array_put(tr);
-       return ret;
+static void *event_file_data(struct file *filp)
+{
+       return ACCESS_ONCE(file_inode(filp)->i_private);
 }
 
-static int tracing_release_generic_file(struct inode *inode, struct file *filp)
+static void remove_event_file_dir(struct ftrace_event_file *file)
 {
-       struct ftrace_event_file *file = inode->i_private;
-       struct trace_array *tr = file->tr;
+       struct dentry *dir = file->dir;
+       struct dentry *child;
 
-       trace_array_put(tr);
+       if (dir) {
+               spin_lock(&dir->d_lock);        /* probably unneeded */
+               list_for_each_entry(child, &dir->d_subdirs, d_u.d_child) {
+                       if (child->d_inode)     /* probably unneeded */
+                               child->d_inode->i_private = NULL;
+               }
+               spin_unlock(&dir->d_lock);
 
-       return 0;
+               debugfs_remove_recursive(dir);
+       }
+
+       list_del(&file->list);
+       remove_subsystem(file->system);
+       kmem_cache_free(file_cachep, file);
 }
 
 /*
@@ -679,15 +688,25 @@ static ssize_t
 event_enable_read(struct file *filp, char __user *ubuf, size_t cnt,
                  loff_t *ppos)
 {
-       struct ftrace_event_file *file = filp->private_data;
+       struct ftrace_event_file *file;
+       unsigned long flags;
        char buf[4] = "0";
 
-       if (file->flags & FTRACE_EVENT_FL_ENABLED &&
-           !(file->flags & FTRACE_EVENT_FL_SOFT_DISABLED))
+       mutex_lock(&event_mutex);
+       file = event_file_data(filp);
+       if (likely(file))
+               flags = file->flags;
+       mutex_unlock(&event_mutex);
+
+       if (!file)
+               return -ENODEV;
+
+       if (flags & FTRACE_EVENT_FL_ENABLED &&
+           !(flags & FTRACE_EVENT_FL_SOFT_DISABLED))
                strcpy(buf, "1");
 
-       if (file->flags & FTRACE_EVENT_FL_SOFT_DISABLED ||
-           file->flags & FTRACE_EVENT_FL_SOFT_MODE)
+       if (flags & FTRACE_EVENT_FL_SOFT_DISABLED ||
+           flags & FTRACE_EVENT_FL_SOFT_MODE)
                strcat(buf, "*");
 
        strcat(buf, "\n");
@@ -699,13 +718,10 @@ static ssize_t
 event_enable_write(struct file *filp, const char __user *ubuf, size_t cnt,
                   loff_t *ppos)
 {
-       struct ftrace_event_file *file = filp->private_data;
+       struct ftrace_event_file *file;
        unsigned long val;
        int ret;
 
-       if (!file)
-               return -EINVAL;
-
        ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
        if (ret)
                return ret;
@@ -717,8 +733,11 @@ event_enable_write(struct file *filp, const char __user *ubuf, size_t cnt,
        switch (val) {
        case 0:
        case 1:
+               ret = -ENODEV;
                mutex_lock(&event_mutex);
-               ret = ftrace_event_enable_disable(file, val);
+               file = event_file_data(filp);
+               if (likely(file))
+                       ret = ftrace_event_enable_disable(file, val);
                mutex_unlock(&event_mutex);
                break;
 
@@ -825,7 +844,7 @@ enum {
 
 static void *f_next(struct seq_file *m, void *v, loff_t *pos)
 {
-       struct ftrace_event_call *call = m->private;
+       struct ftrace_event_call *call = event_file_data(m->private);
        struct list_head *common_head = &ftrace_common_fields;
        struct list_head *head = trace_get_fields(call);
        struct list_head *node = v;
@@ -857,7 +876,7 @@ static void *f_next(struct seq_file *m, void *v, loff_t *pos)
 
 static int f_show(struct seq_file *m, void *v)
 {
-       struct ftrace_event_call *call = m->private;
+       struct ftrace_event_call *call = event_file_data(m->private);
        struct ftrace_event_field *field;
        const char *array_descriptor;
 
@@ -910,6 +929,11 @@ static void *f_start(struct seq_file *m, loff_t *pos)
        void *p = (void *)FORMAT_HEADER;
        loff_t l = 0;
 
+       /* ->stop() is called even if ->start() fails */
+       mutex_lock(&event_mutex);
+       if (!event_file_data(m->private))
+               return ERR_PTR(-ENODEV);
+
        while (l < *pos && p)
                p = f_next(m, p, &l);
 
@@ -918,6 +942,7 @@ static void *f_start(struct seq_file *m, loff_t *pos)
 
 static void f_stop(struct seq_file *m, void *p)
 {
+       mutex_unlock(&event_mutex);
 }
 
 static const struct seq_operations trace_format_seq_ops = {
@@ -929,7 +954,6 @@ static const struct seq_operations trace_format_seq_ops = {
 
 static int trace_format_open(struct inode *inode, struct file *file)
 {
-       struct ftrace_event_call *call = inode->i_private;
        struct seq_file *m;
        int ret;
 
@@ -938,7 +962,7 @@ static int trace_format_open(struct inode *inode, struct file *file)
                return ret;
 
        m = file->private_data;
-       m->private = call;
+       m->private = file;
 
        return 0;
 }
@@ -946,14 +970,18 @@ static int trace_format_open(struct inode *inode, struct file *file)
 static ssize_t
 event_id_read(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos)
 {
-       struct ftrace_event_call *call = filp->private_data;
+       int id = (long)event_file_data(filp);
        char buf[32];
        int len;
 
        if (*ppos)
                return 0;
 
-       len = sprintf(buf, "%d\n", call->event.type);
+       if (unlikely(!id))
+               return -ENODEV;
+
+       len = sprintf(buf, "%d\n", id);
+
        return simple_read_from_buffer(ubuf, cnt, ppos, buf, len);
 }
 
@@ -961,21 +989,28 @@ static ssize_t
 event_filter_read(struct file *filp, char __user *ubuf, size_t cnt,
                  loff_t *ppos)
 {
-       struct ftrace_event_call *call = filp->private_data;
+       struct ftrace_event_call *call;
        struct trace_seq *s;
-       int r;
+       int r = -ENODEV;
 
        if (*ppos)
                return 0;
 
        s = kmalloc(sizeof(*s), GFP_KERNEL);
+
        if (!s)
                return -ENOMEM;
 
        trace_seq_init(s);
 
-       print_event_filter(call, s);
-       r = simple_read_from_buffer(ubuf, cnt, ppos, s->buffer, s->len);
+       mutex_lock(&event_mutex);
+       call = event_file_data(filp);
+       if (call)
+               print_event_filter(call, s);
+       mutex_unlock(&event_mutex);
+
+       if (call)
+               r = simple_read_from_buffer(ubuf, cnt, ppos, s->buffer, s->len);
 
        kfree(s);
 
@@ -986,9 +1021,9 @@ static ssize_t
 event_filter_write(struct file *filp, const char __user *ubuf, size_t cnt,
                   loff_t *ppos)
 {
-       struct ftrace_event_call *call = filp->private_data;
+       struct ftrace_event_call *call;
        char *buf;
-       int err;
+       int err = -ENODEV;
 
        if (cnt >= PAGE_SIZE)
                return -EINVAL;
@@ -1003,7 +1038,12 @@ event_filter_write(struct file *filp, const char __user *ubuf, size_t cnt,
        }
        buf[cnt] = '\0';
 
-       err = apply_event_filter(call, buf);
+       mutex_lock(&event_mutex);
+       call = event_file_data(filp);
+       if (call)
+               err = apply_event_filter(call, buf);
+       mutex_unlock(&event_mutex);
+
        free_page((unsigned long) buf);
        if (err < 0)
                return err;
@@ -1225,10 +1265,9 @@ static const struct file_operations ftrace_set_event_fops = {
 };
 
 static const struct file_operations ftrace_enable_fops = {
-       .open = tracing_open_generic_file,
+       .open = tracing_open_generic,
        .read = event_enable_read,
        .write = event_enable_write,
-       .release = tracing_release_generic_file,
        .llseek = default_llseek,
 };
 
@@ -1240,7 +1279,6 @@ static const struct file_operations ftrace_event_format_fops = {
 };
 
 static const struct file_operations ftrace_event_id_fops = {
-       .open = tracing_open_generic,
        .read = event_id_read,
        .llseek = default_llseek,
 };
@@ -1488,8 +1526,8 @@ event_create_dir(struct dentry *parent,
 
 #ifdef CONFIG_PERF_EVENTS
        if (call->event.type && call->class->reg)
-               trace_create_file("id", 0444, file->dir, call,
-                                 id);
+               trace_create_file("id", 0444, file->dir,
+                                 (void *)(long)call->event.type, id);
 #endif
 
        /*
@@ -1514,33 +1552,16 @@ event_create_dir(struct dentry *parent,
        return 0;
 }
 
-static void remove_subsystem(struct ftrace_subsystem_dir *dir)
-{
-       if (!dir)
-               return;
-
-       if (!--dir->nr_events) {
-               debugfs_remove_recursive(dir->entry);
-               list_del(&dir->list);
-               __put_system_dir(dir);
-       }
-}
-
 static void remove_event_from_tracers(struct ftrace_event_call *call)
 {
        struct ftrace_event_file *file;
        struct trace_array *tr;
 
        do_for_each_event_file_safe(tr, file) {
-
                if (file->event_call != call)
                        continue;
 
-               list_del(&file->list);
-               debugfs_remove_recursive(file->dir);
-               remove_subsystem(file->system);
-               kmem_cache_free(file_cachep, file);
-
+               remove_event_file_dir(file);
                /*
                 * The do_for_each_event_file_safe() is
                 * a double loop. After finding the call for this
@@ -1692,16 +1713,53 @@ static void __trace_remove_event_call(struct ftrace_event_call *call)
        destroy_preds(call);
 }
 
+static int probe_remove_event_call(struct ftrace_event_call *call)
+{
+       struct trace_array *tr;
+       struct ftrace_event_file *file;
+
+#ifdef CONFIG_PERF_EVENTS
+       if (call->perf_refcount)
+               return -EBUSY;
+#endif
+       do_for_each_event_file(tr, file) {
+               if (file->event_call != call)
+                       continue;
+               /*
+                * We can't rely on ftrace_event_enable_disable(enable => 0)
+                * we are going to do, FTRACE_EVENT_FL_SOFT_MODE can suppress
+                * TRACE_REG_UNREGISTER.
+                */
+               if (file->flags & FTRACE_EVENT_FL_ENABLED)
+                       return -EBUSY;
+               /*
+                * The do_for_each_event_file_safe() is
+                * a double loop. After finding the call for this
+                * trace_array, we use break to jump to the next
+                * trace_array.
+                */
+               break;
+       } while_for_each_event_file();
+
+       __trace_remove_event_call(call);
+
+       return 0;
+}
+
 /* Remove an event_call */
-void trace_remove_event_call(struct ftrace_event_call *call)
+int trace_remove_event_call(struct ftrace_event_call *call)
 {
+       int ret;
+
        mutex_lock(&trace_types_lock);
        mutex_lock(&event_mutex);
        down_write(&trace_event_sem);
-       __trace_remove_event_call(call);
+       ret = probe_remove_event_call(call);
        up_write(&trace_event_sem);
        mutex_unlock(&event_mutex);
        mutex_unlock(&trace_types_lock);
+
+       return ret;
 }
 
 #define for_each_event(event, start, end)                      \
@@ -2270,12 +2328,8 @@ __trace_remove_event_dirs(struct trace_array *tr)
 {
        struct ftrace_event_file *file, *next;
 
-       list_for_each_entry_safe(file, next, &tr->events, list) {
-               list_del(&file->list);
-               debugfs_remove_recursive(file->dir);
-               remove_subsystem(file->system);
-               kmem_cache_free(file_cachep, file);
-       }
+       list_for_each_entry_safe(file, next, &tr->events, list)
+               remove_event_file_dir(file);
 }
 
 static void
index 0c7b75a8acc8f9ccad53d75bc6a3fe0da425ae9f..97daa8cf958df8e73c9fa8090f89cbaf4e63893e 100644 (file)
@@ -637,17 +637,15 @@ static void append_filter_err(struct filter_parse_state *ps,
        free_page((unsigned long) buf);
 }
 
+/* caller must hold event_mutex */
 void print_event_filter(struct ftrace_event_call *call, struct trace_seq *s)
 {
-       struct event_filter *filter;
+       struct event_filter *filter = call->filter;
 
-       mutex_lock(&event_mutex);
-       filter = call->filter;
        if (filter && filter->filter_string)
                trace_seq_printf(s, "%s\n", filter->filter_string);
        else
                trace_seq_puts(s, "none\n");
-       mutex_unlock(&event_mutex);
 }
 
 void print_subsystem_event_filter(struct event_subsystem *system,
@@ -1841,23 +1839,22 @@ static int create_system_filter(struct event_subsystem *system,
        return err;
 }
 
+/* caller must hold event_mutex */
 int apply_event_filter(struct ftrace_event_call *call, char *filter_string)
 {
        struct event_filter *filter;
-       int err = 0;
-
-       mutex_lock(&event_mutex);
+       int err;
 
        if (!strcmp(strstrip(filter_string), "0")) {
                filter_disable(call);
                filter = call->filter;
                if (!filter)
-                       goto out_unlock;
+                       return 0;
                RCU_INIT_POINTER(call->filter, NULL);
                /* Make sure the filter is not being used */
                synchronize_sched();
                __free_filter(filter);
-               goto out_unlock;
+               return 0;
        }
 
        err = create_filter(call, filter_string, true, &filter);
@@ -1884,8 +1881,6 @@ int apply_event_filter(struct ftrace_event_call *call, char *filter_string)
                        __free_filter(tmp);
                }
        }
-out_unlock:
-       mutex_unlock(&event_mutex);
 
        return err;
 }
index 3811487e7a7a1d58f23b6e04b49fda0dbf06c9b0..243f6834d02675c9c1101e1eabe18e56eb7d3244 100644 (file)
@@ -95,7 +95,7 @@ static __kprobes bool trace_probe_is_on_module(struct trace_probe *tp)
 }
 
 static int register_probe_event(struct trace_probe *tp);
-static void unregister_probe_event(struct trace_probe *tp);
+static int unregister_probe_event(struct trace_probe *tp);
 
 static DEFINE_MUTEX(probe_lock);
 static LIST_HEAD(probe_list);
@@ -351,9 +351,12 @@ static int unregister_trace_probe(struct trace_probe *tp)
        if (trace_probe_is_enabled(tp))
                return -EBUSY;
 
+       /* Will fail if probe is being used by ftrace or perf */
+       if (unregister_probe_event(tp))
+               return -EBUSY;
+
        __unregister_trace_probe(tp);
        list_del(&tp->list);
-       unregister_probe_event(tp);
 
        return 0;
 }
@@ -632,7 +635,9 @@ static int release_all_trace_probes(void)
        /* TODO: Use batch unregistration */
        while (!list_empty(&probe_list)) {
                tp = list_entry(probe_list.next, struct trace_probe, list);
-               unregister_trace_probe(tp);
+               ret = unregister_trace_probe(tp);
+               if (ret)
+                       goto end;
                free_trace_probe(tp);
        }
 
@@ -1247,11 +1252,15 @@ static int register_probe_event(struct trace_probe *tp)
        return ret;
 }
 
-static void unregister_probe_event(struct trace_probe *tp)
+static int unregister_probe_event(struct trace_probe *tp)
 {
+       int ret;
+
        /* tp->event is unregistered in trace_remove_event_call() */
-       trace_remove_event_call(&tp->call);
-       kfree(tp->call.print_fmt);
+       ret = trace_remove_event_call(&tp->call);
+       if (!ret)
+               kfree(tp->call.print_fmt);
+       return ret;
 }
 
 /* Make a debugfs interface for controlling probe points */
index a23d2d71188e61f6ad3b706730b2a70d8c773103..272261b5f94ff84e3291be05aaa3b1a8c1ca00ca 100644 (file)
@@ -70,7 +70,7 @@ struct trace_uprobe {
        (sizeof(struct probe_arg) * (n)))
 
 static int register_uprobe_event(struct trace_uprobe *tu);
-static void unregister_uprobe_event(struct trace_uprobe *tu);
+static int unregister_uprobe_event(struct trace_uprobe *tu);
 
 static DEFINE_MUTEX(uprobe_lock);
 static LIST_HEAD(uprobe_list);
@@ -164,11 +164,17 @@ static struct trace_uprobe *find_probe_event(const char *event, const char *grou
 }
 
 /* Unregister a trace_uprobe and probe_event: call with locking uprobe_lock */
-static void unregister_trace_uprobe(struct trace_uprobe *tu)
+static int unregister_trace_uprobe(struct trace_uprobe *tu)
 {
+       int ret;
+
+       ret = unregister_uprobe_event(tu);
+       if (ret)
+               return ret;
+
        list_del(&tu->list);
-       unregister_uprobe_event(tu);
        free_trace_uprobe(tu);
+       return 0;
 }
 
 /* Register a trace_uprobe and probe_event */
@@ -181,9 +187,12 @@ static int register_trace_uprobe(struct trace_uprobe *tu)
 
        /* register as an event */
        old_tp = find_probe_event(tu->call.name, tu->call.class->system);
-       if (old_tp)
+       if (old_tp) {
                /* delete old event */
-               unregister_trace_uprobe(old_tp);
+               ret = unregister_trace_uprobe(old_tp);
+               if (ret)
+                       goto end;
+       }
 
        ret = register_uprobe_event(tu);
        if (ret) {
@@ -256,6 +265,8 @@ static int create_trace_uprobe(int argc, char **argv)
                group = UPROBE_EVENT_SYSTEM;
 
        if (is_delete) {
+               int ret;
+
                if (!event) {
                        pr_info("Delete command needs an event name.\n");
                        return -EINVAL;
@@ -269,9 +280,9 @@ static int create_trace_uprobe(int argc, char **argv)
                        return -ENOENT;
                }
                /* delete an event */
-               unregister_trace_uprobe(tu);
+               ret = unregister_trace_uprobe(tu);
                mutex_unlock(&uprobe_lock);
-               return 0;
+               return ret;
        }
 
        if (argc < 2) {
@@ -408,16 +419,20 @@ fail_address_parse:
        return ret;
 }
 
-static void cleanup_all_probes(void)
+static int cleanup_all_probes(void)
 {
        struct trace_uprobe *tu;
+       int ret = 0;
 
        mutex_lock(&uprobe_lock);
        while (!list_empty(&uprobe_list)) {
                tu = list_entry(uprobe_list.next, struct trace_uprobe, list);
-               unregister_trace_uprobe(tu);
+               ret = unregister_trace_uprobe(tu);
+               if (ret)
+                       break;
        }
        mutex_unlock(&uprobe_lock);
+       return ret;
 }
 
 /* Probes listing interfaces */
@@ -462,8 +477,13 @@ static const struct seq_operations probes_seq_op = {
 
 static int probes_open(struct inode *inode, struct file *file)
 {
-       if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC))
-               cleanup_all_probes();
+       int ret;
+
+       if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
+               ret = cleanup_all_probes();
+               if (ret)
+                       return ret;
+       }
 
        return seq_open(file, &probes_seq_op);
 }
@@ -968,12 +988,17 @@ static int register_uprobe_event(struct trace_uprobe *tu)
        return ret;
 }
 
-static void unregister_uprobe_event(struct trace_uprobe *tu)
+static int unregister_uprobe_event(struct trace_uprobe *tu)
 {
+       int ret;
+
        /* tu->event is unregistered in trace_remove_event_call() */
-       trace_remove_event_call(&tu->call);
+       ret = trace_remove_event_call(&tu->call);
+       if (ret)
+               return ret;
        kfree(tu->call.print_fmt);
        tu->call.print_fmt = NULL;
+       return 0;
 }
 
 /* Make a trace interface for controling probe points */