[Cluster-devel] Re: trace: fix multiple use of __print_flags and __print_symbolic

[Steven Whitehouse]

Hi,

On Wed, 2009-06-03 at 17:50 +0800, Li Zefan wrote:
> Steven Whitehouse wrote:
> > When using multiple calls to __print_symbolic and __print_flags in
> > the same tracer, I noticed that the output was incorrect. I think
> > the following patch is the correct fix (it works for me) but please
> > check it carefully since I'm not that familiar with this code,
> > and I may well have made a mistake somewhere.
> > 
> 
> I don't see there's bug in __print_symbolic() or __print_flags():
> 
> enum print_line_t
> ftrace_raw_output_##call(struct trace_iterator *iter, int flags)
> {
> 	...
> 	p = &get_cpu_var(ftrace_event_seq);
> 	/* here we call ftrace_print_flags_seq(p, ...) */
> 	ret = trace_seq_printf(s, #call ": " print);
> 	put_cpu();
> 	...
> }
> 
> ftrace_event_seq is percpu data, and here is used with preempt-disabled,
> so there shouldn't be problem of concurrent accessing.
> 
> 
Well without the "fix" every call to __print_symbolic and __print_flags
seems to result in printing the same string, so with:


        TP_printk("%u.%u glock %d:%lld state %s => %s tgt:%s dmt:%s flags:%s",
                  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->gltype,
                 (unsigned long long)__entry->glnum,
                  glock_trace_name(__entry->cur_state),
                  glock_trace_name(__entry->new_state),
                  glock_trace_name(__entry->tgt_state),
                  glock_trace_name(__entry->dmt_state),
                  show_glock_flags(__entry->flags))
and

#define dlm_state_name(nn) { DLM_LOCK_##nn, #nn }
#define glock_trace_name(x) __print_symbolic(x,         \
                            dlm_state_name(IV),         \
                            dlm_state_name(NL),         \
                            dlm_state_name(CR),         \
                            dlm_state_name(CW),         \
                            dlm_state_name(PR),         \
                            dlm_state_name(PW),         \
                            dlm_state_name(EX))

#define show_glock_flags(flags) __print_flags(flags, "",        \
        {(1UL << GLF_LOCK),                     "l" },          \
        {(1UL << GLF_DEMOTE),                   "D" },          \
        {(1UL << GLF_PENDING_DEMOTE),           "d" },          \
        {(1UL << GLF_DEMOTE_IN_PROGRESS),       "p" },          \
        {(1UL << GLF_DIRTY),                    "y" },          \
        {(1UL << GLF_LFLUSH),                   "f" },          \
        {(1UL << GLF_INVALIDATE_IN_PROGRESS),   "i" },          \
        {(1UL << GLF_REPLY_PENDING),            "r" },          \
        {(1UL << GLF_INITIAL),                  "I" },          \
        {(1UL << GLF_FROZEN),                   "F" })

every call to glock_trace_name and the call to glock_show_flags resulted in the
string "EX" when I was testing it. With the previous version of my patch where
I did the constant to string conversion with my own function, this didn't happen.
Applying the patch corrects the output, so there is a bug here I think, even if I've
not found the correct fix for it,

Steve.