On Mon, Sep 17, 2018 6:54 PM Amir Goldstein <amir73il@xxxxxxxxx> wrote: >On Mon, Sep 17, 2018 at 1:02 PM nixiaoming <nixiaoming@xxxxxxxxxx> wrote: >> >> Added FAN_EVENT_INFO_TID to select the thread id of the event trigger >> > >Maybe "to report the thread id of the task that caused the event". >Also in commit title, I think "reporting thread id" is more to the point >than "save thread id". > Very good advice, I will update the patch description based on your suggestions. > >> Signed-off-by: nixiaoming <nixiaoming@xxxxxxxxxx> >> --- >> --- a/include/uapi/linux/fanotify.h >> +++ b/include/uapi/linux/fanotify.h >> @@ -18,6 +18,7 @@ >> >> #define FAN_ONDIR 0x40000000 /* event occurred against dir */ >> >> +#define FAN_EVENT_INFO_TID 0x02000000 /* event save thread id replace tgid */ >> #define FAN_EVENT_ON_CHILD 0x08000000 /* interested in child events */ >> > >nixiaoming, > >You misunderstood me. > >I meant that you should add support for flag FAN_EVENT_INFO_TID >for the fanotify_init() syscall, not the fanotify_mark() syscall. > I mistakenly thought that adding a tag to group->fanotify_data in the fanotify_init system call would affect the system ABI. local tests found that Module.symvers did not change before and after the modification. According to your opinion, I have re-modified a version of fanotify_init I will send the patch later. >See this commit from Steve Grubb for example of adding new opt-in >behavior to fanotify: >de8cd83e91bc audit: Record fanotify access control decisions > >BTW, Steve, did you ever follow up with a man-pages patch? > >Thanks, >Amir. > thanks
