Messages in this thread | | | Date | Sun, 31 Jul 2022 12:47:30 -0400 | From | Steven Rostedt <> | Subject | Re: [PATCH V9 01/16] rv: Add Runtime Verification (RV) interface |
| |
On Sun, 31 Jul 2022 23:06:31 +0800 Tao Zhou <tao.zhou@linux.dev> wrote:
> > All things above are misled by the first interpretation but,,, > > Yeah, this is not that clear from my above words expression. I said the return > value of da_monitor_init_*() will be 0, but it is not right. Global and per-cpu > monitor will return 0, per-task monitor may return a positive value when the > slot is equal or greater than RV_PER_TASK_MONITOR_INIT(how possible this will > happen I do know yet). This is from reading the current code implementation. > I just want to say that there may be a bug here.
Well, rv_get_monitor_slot() can currently only return 0 or negative. This is because PER_TASK_MONITORS is just 1 and we can not return that or greater.
> If rv_enable_monitor() return a positive value and the error happened(as above
With the current code this can not happen, as we only allow for a single PER_TASK_MONITORS.
But in the future, if we increment this, then you are correct. We can not just check retval, but need to check retval < 0.
This does need to be fixed. But because it currently isn't an issue because we they can only return 0 or negative, I'm going to pull this series in.
But Daniel, these checks do need to be updated. Please send patches on top of this series to address it.
-- Steve
> said), user space will not know this is a error return value, but regard it as a > right writing. Even if the return value(the slot value not in [0..RV_PER_TASK_MONITOR_INIT)) > is equal to count of charaters that are writen to the file(the string length of monitor name), > it will still be not a right writing. >
| |