On Wed, 25 Jul 2018 00:09:09 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
Hmm, your patch seems to leak a memory since event_trigger_init() will be called twice on same trigger_data (Note that event_trigger_init() does not init ref counter, but increment it.) So we should decrement it when we find it is succeeded. Moreover, if register_trigger()
Good catch, and easily fixed.
fails before calling data->ops->init() (see -EEXIST case), the ref counter will be 0 (-1 +1). But if it fails after data->ops->init(), the ref counter will be 1 (-1 +1 +1). It still be unstable. (Ah, that means we may have another trouble...)
I'm not sure there's a problem here. I now have:
out_reg: /* Up the trigger_data count to make sure reg doesn't free it on failuer */ event_trigger_init(trigger_ops, trigger_data); ret = cmd_ops->reg(glob, trigger_ops, trigger_data, file); /* * The above returns on success the # of functions enabled, * but if it didn't find any functions it returns zero. * Consider no functions a failure too. */ if (!ret) { ret = -ENOENT; } else if (ret > 0) ret = 0;
/* Down the counter of trigger_data or free it if not used anymore */ event_trigger_free(trigger_ops, trigger_data); out: return ret;
Thus we increment trigger_data before calling reg, and free it afterward. But if reg() did an init too, then the event_trigger_free() just decs the ref counter.
As for register_trigger()
P.S. This brings up another minor bug. The failure should return ENOMEM not ENOENT.
Hmm it seems we should review the register_trigger() implementation. It should return the return value of trace_event_trigger_enable_disable(), shouldn't it?
Yeah, that's not done well. I'll fix it up.
Thanks for pointing it out.
-- Steve