On Wed, 30 Aug 2017 12:19:30 +0800 "Ziqian SUN (Zamir)" <z...@redhat.com> wrote:
> From: "Ziqian SUN (Zamir)" <z...@redhat.com> > > The mmiotrace tracer cannot be enabled with ftrace=mmiotrace in > kernel commandline. With this patch, a trace_noboot_tracer_list > is implemented and will be checked during system boot. If the > tracer declares itself as not for boot up, system will print out > a message and continue booting. > > Link: https://bugzilla.kernel.org/show_bug.cgi?id=196557 > Signed-off-by: Ziqian SUN (Zamir) <z...@redhat.com> > --- > kernel/trace/trace.c | 34 +++++++++++++++++++++++++++++++++- > kernel/trace/trace.h | 11 +++++++++++ > kernel/trace/trace_mmiotrace.c | 4 ++++ > 3 files changed, 48 insertions(+), 1 deletion(-) > > diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c > index 44004d8..0d25b1e 100644 > --- a/kernel/trace/trace.c > +++ b/kernel/trace/trace.c > @@ -319,6 +319,33 @@ void trace_free_pid_list(struct trace_pid_list *pid_list) > kfree(pid_list); > } > > +struct mutex trace_noboot_tracer_lock; > + > +DEFINE_MUTEX(trace_noboot_tracer_lock); > +LIST_HEAD(trace_noboot_tracer_list); > + > +void trace_noboot_tracer_put(struct tracer_noboot *noboot_tracer) > +{ > + mutex_lock(&trace_noboot_tracer_lock); > + list_add(&noboot_tracer->list, &trace_noboot_tracer_list); > + mutex_unlock(&trace_noboot_tracer_lock); > +} > + > +bool trace_noboot_tracer_find(const char *str) > +{ > + mutex_lock(&trace_noboot_tracer_lock); > + struct tracer_noboot *ptr = NULL; > + > + list_for_each_entry(ptr, &trace_noboot_tracer_list, list) { > + if (strcmp(ptr->name, str) == 0) { > + mutex_unlock(&trace_noboot_tracer_lock); > + return true; > + } > + } > + mutex_unlock(&trace_noboot_tracer_lock); > + return false; > +} > + > /** > * trace_find_filtered_pid - check if a pid exists in a filtered_pid list > * @filtered_pids: The list of pids to check > @@ -1641,8 +1668,13 @@ int __init register_tracer(struct tracer *type) > if (strncmp(default_bootup_tracer, type->name, MAX_TRACER_SIZE)) > goto out_unlock; > > + /* If the tracer should not be enabled by kernel parameter */ > + if (trace_noboot_tracer_find(type->name)) { > + pr_warn("Tracer '%s' cannot be enabled during boot\n", > type->name); > + goto out_unlock; > + } > + > printk(KERN_INFO "Starting tracer '%s'\n", type->name); > - /* Do we want this tracer to start on bootup? */ > tracing_set_tracer(&global_trace, type->name); > default_bootup_tracer = NULL; > > diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h > index 490ba22..23bc668 100644 > --- a/kernel/trace/trace.h > +++ b/kernel/trace/trace.h > @@ -502,6 +502,17 @@ enum { > TRACE_IRQ_BIT, > }; > > +/* > + * For tracer that cannot be enabled during boot time. > + */ > + > +struct tracer_noboot { > + const char *name; > + struct list_head list; > +}; > + > +void trace_noboot_tracer_put(struct tracer_noboot *noboot_tracer); > + > #define trace_recursion_set(bit) do { (current)->trace_recursion |= > (1<<(bit)); } while (0) > #define trace_recursion_clear(bit) do { (current)->trace_recursion &= > ~(1<<(bit)); } while (0) > #define trace_recursion_test(bit) ((current)->trace_recursion & > (1<<(bit))) > diff --git a/kernel/trace/trace_mmiotrace.c b/kernel/trace/trace_mmiotrace.c > index cd7480d..06aa0a5 100644 > --- a/kernel/trace/trace_mmiotrace.c > +++ b/kernel/trace/trace_mmiotrace.c > @@ -286,6 +286,10 @@ static enum print_line_t mmio_print_line(struct > trace_iterator *iter) > > __init static int init_mmio_trace(void) > { > + struct tracer_noboot mmio_notrace; > + > + mmio_notrace.name = "mmiotrace"; > + trace_noboot_tracer_put(&mmio_notrace); > return register_tracer(&mmio_tracer); > } > device_initcall(init_mmio_trace); This is way overkill. Just add bool noboot; to struct tracer in trace.h and then to the mmiotracer: struct tracer mmio_tracer = { [..] .noboot = true, }; And then in tracing_set_tracer() add after: if (t == tr->current_trace) goto out; this: if (system_state < SYSTEM_RUNNING && t->noboot) { WARN(1, "Tracer %s not allowed on command line", t->name); goto out; } That's it. No need for locking or anything else. -- Steve
