tracing: remove on the fly allocator from function profiler

Impact: safer code

The on the fly allocator for the function profiler was to save
memory. But at the expense of stability. Although it survived several
tests, allocating from the function tracer is just too risky, just
to save space.

This patch removes the allocator and simply allocates enough entries
at start up.

Each function gets a profiling structure of 40 bytes. With an average
of 20K functions, and this is for each CPU, we have 800K per online
CPU. This is not too bad, at least for non-embedded.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>

diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index a141d8499ab0395d5620643b34d2fc66de61e725..4d90c916b2bbd75d8ba024a6da7f332a2e110a68 100644 (file)
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -401,6 +401,8 @@ static void ftrace_profile_reset(struct ftrace_profile_stat *stat)
 int ftrace_profile_pages_init(struct ftrace_profile_stat *stat)
 {
        struct ftrace_profile_page *pg;
+       int functions;
+       int pages;
        int i;
 
        /* If we already allocated, do nothing */
@@ -411,22 +413,46 @@ int ftrace_profile_pages_init(struct ftrace_profile_stat *stat)
        if (!stat->pages)
                return -ENOMEM;
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+       functions = ftrace_update_tot_cnt;
+#else
+       /*
+        * We do not know the number of functions that exist because
+        * dynamic tracing is what counts them. With past experience
+        * we have around 20K functions. That should be more than enough.
+        * It is highly unlikely we will execute every function in
+        * the kernel.
+        */
+       functions = 20000;
+#endif
+
        pg = stat->start = stat->pages;
 
-       /* allocate 10 more pages to start */
-       for (i = 0; i < 10; i++) {
+       pages = DIV_ROUND_UP(functions, PROFILES_PER_PAGE);
+
+       for (i = 0; i < pages; i++) {
                pg->next = (void *)get_zeroed_page(GFP_KERNEL);
-               /*
-                * We only care about allocating profile_pages, if
-                * we failed to allocate here, hopefully we will allocate
-                * later.
-                */
                if (!pg->next)
-                       break;
+                       goto out_free;
                pg = pg->next;
        }
 
        return 0;
+
+ out_free:
+       pg = stat->start;
+       while (pg) {
+               unsigned long tmp = (unsigned long)pg;
+
+               pg = pg->next;
+               free_page(tmp);
+       }
+
+       free_page((unsigned long)stat->pages);
+       stat->pages = NULL;
+       stat->start = NULL;
+
+       return -ENOMEM;
 }
 
 static int ftrace_profile_init_cpu(int cpu)
@@ -460,7 +486,7 @@ static int ftrace_profile_init_cpu(int cpu)
                        ftrace_profile_bits++;
        }
 
-       /* Preallocate a few pages */
+       /* Preallocate the function profiling pages */
        if (ftrace_profile_pages_init(stat) < 0) {
                kfree(stat->hash);
                stat->hash = NULL;
@@ -516,24 +542,21 @@ static void ftrace_add_profile(struct ftrace_profile_stat *stat,
        hlist_add_head_rcu(&rec->node, &stat->hash[key]);
 }
 
-/* Interrupts must be disabled calling this */
+/*
+ * The memory is already allocated, this simply finds a new record to use.
+ */
 static struct ftrace_profile *
-ftrace_profile_alloc(struct ftrace_profile_stat *stat,
-                    unsigned long ip, bool alloc_safe)
+ftrace_profile_alloc(struct ftrace_profile_stat *stat, unsigned long ip)
 {
        struct ftrace_profile *rec = NULL;
 
-       /* prevent recursion */
+       /* prevent recursion (from NMIs) */
        if (atomic_inc_return(&stat->disabled) != 1)
                goto out;
 
-       /* Try to always keep another page available */
-       if (!stat->pages->next && alloc_safe)
-               stat->pages->next = (void *)get_zeroed_page(GFP_ATOMIC);
-
        /*
-        * Try to find the function again since another
-        * task on another CPU could have added it
+        * Try to find the function again since an NMI
+        * could have added it
         */
        rec = ftrace_find_profiled_func(stat, ip);
        if (rec)
@@ -555,29 +578,16 @@ ftrace_profile_alloc(struct ftrace_profile_stat *stat,
        return rec;
 }
 
-/*
- * If we are not in an interrupt, or softirq and
- * and interrupts are disabled and preemption is not enabled
- * (not in a spinlock) then it should be safe to allocate memory.
- */
-static bool ftrace_safe_to_allocate(void)
-{
-       return !in_interrupt() && irqs_disabled() && !preempt_count();
-}
-
 static void
 function_profile_call(unsigned long ip, unsigned long parent_ip)
 {
        struct ftrace_profile_stat *stat;
        struct ftrace_profile *rec;
        unsigned long flags;
-       bool alloc_safe;
 
        if (!ftrace_profile_enabled)
                return;
 
-       alloc_safe = ftrace_safe_to_allocate();
-
        local_irq_save(flags);
 
        stat = &__get_cpu_var(ftrace_profile_stats);
@@ -586,7 +596,7 @@ function_profile_call(unsigned long ip, unsigned long parent_ip)
 
        rec = ftrace_find_profiled_func(stat, ip);
        if (!rec) {
-               rec = ftrace_profile_alloc(stat, ip, alloc_safe);
+               rec = ftrace_profile_alloc(stat, ip);
                if (!rec)
                        goto out;
        }