Re: [for-next][PATCH 2/2] tracing: Use temp buffer when filtering events

[Steven Rostedt]

On Fri, 28 Aug 2020 18:49:55 -0400
Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:

> On Fri, 28 Aug 2020 15:53:06 +0800
> Wen Gong <wgong@xxxxxxxxxxxxxx> wrote:
> 
> > this patch commit id is : 0fc1b09ff1ff404ddf753f5ffa5cd0adc8fdcdc9 which 
> > has upstream.
> > 
> > how much size is the per cpu buffer?
> > seems it is initilized in trace_buffered_event_enable,
> > it is only 1 page size as below:
> > void trace_buffered_event_enable(void)
> > {
> > ...
> > 	for_each_tracing_cpu(cpu) {
> > 		page = alloc_pages_node(cpu_to_node(cpu),
> > 					GFP_KERNEL | __GFP_NORETRY, 0);
> > If the size of buffer to trace is more than 1 page, such as 46680, then 
> > it trigger kernel crash/panic in my case while run trace-cmd.
> > After debugging, the trace_file->flags in 
> > trace_event_buffer_lock_reserve is 0x40b while run trace-cmd, and it is 
> > 0x403 while collecting ftrace log.
> > 
> > Is it have any operation to disable this patch dynamically?  
> 
> It shouldn't be disabled, this is a bug that needs to be fixed.
> 
> Also, if an event is more than a page, it wont be saved in the ftrace
> ring buffer, as events are limited by page size minus the headers.
> 

Untested (not even compiled, as I'm now on PTO) but does this patch
work for you?

-- Steve

diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index f40d850ebabc..3a9b4422e7fc 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -2598,7 +2598,7 @@ trace_event_buffer_lock_reserve(struct trace_buffer **current_rb,
 	    (entry = this_cpu_read(trace_buffered_event))) {
 		/* Try to use the per cpu buffer first */
 		val = this_cpu_inc_return(trace_buffered_event_cnt);
-		if (val == 1) {
+		if (val == 1 || (len > (PAGE_SIZE - 8))) {
 			trace_event_setup(entry, type, flags, pc);
 			entry->array[0] = len;
 			return entry;