From: Richard Palethorpe Date: Wed, 25 Sep 2019 14:13:49 +0200 Subject: [LTP] [PATCH v3] fzsync: revoke thread_b if parent hits an accidental break In-Reply-To: References: <20190924105801.7616-1-liwang@redhat.com> <87wodxkgr4.fsf@rpws.prws.suse.cz> <87ef04ycxv.fsf@rpws.prws.suse.cz> Message-ID: <87d0foy3nm.fsf@rpws.prws.suse.cz> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ltp@lists.linux.it Hello, Li Wang writes: > Richard Palethorpe wrote: > >>> must wrap the user supplied function. You can create a function which >> >> accepts a pointer to some contiguous memory containing the user supplied >> >> function >> >> pointer and the user supplied arg pointer. >> >> >> > >> > Since you have fixed the function format of thread B as void >> *(*run_b)(void >> > *) in tst_fzsync_pair_reset(), which means we have no need to take care >> of >> > the function arg pointer anymore. >> >> I think the function pointer signature would be 'void *(*run_b)(void)' >> not 'void *(*run_b)(void *)'. >> > > Hmm, but at least we should respect the pthread_create()? It requires the > function prototype is ''void *(*func)(void *)'. > > int pthread_create(pthread_t *thread, const pthread_attr_t *attr, > void *(*start_routine) (void *), void *arg); > > We could unuse the arg in thread B, but declare the function prototype with > parameter "void *" is no harm. I'm not sure what you are saying. However you could do something like this (I haven't tested it): struct tst_fzsync_run_thread { void *(*run)(void *); void *arg; }; static void tst_fzsync_thread_wrapper(void *arg) { struct tst_fzsync_run_thread t = *(struct tst_fzsync_run_thread *)arg; pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS); pthread_setcancelstate(PTHREAD_CANCEL_ENABLE); t.run(t.arg); } static void tst_fzsync_pair_reset(..., struct tst_fzsync_run_thread *run_b) { ... if (run_b) SAFE_PTHREAD_fCREATE(..., tst_fzsync_thread_wrapper, (void *)run_b); ... } Note that in any case you can't reliably cast a function pointer to a void pointer without some magic. I am guessing wrapping it in a struct is the clearest way to do it. You can remove the arg altogether, but I kept it because we have a struct anyway to wrap the function pointer. > > >> I doubt any test would need the arg though, because we only use one >> thread and can store parameters in global variables. So you could remove >> it and update the tests. >> >> The user might need that arg if they are starting many threads, but for >> now we don't have explicit support for that in the library. >> > > Maybe we just need to note that in the lib document. > > >> > >> > So just like what I did in V2, the wrapper function could steal the real >> > run_b address from pthread_create(..., wrap_run_b, run_b) parameter. >> >> >> -- >> Thank you, >> Richard. >> -- Thank you, Richard.