[CRIU] ✗ travis-ci: failure for Per-thread seccomp support, v3

[Andrei Vagin]

On Fri, Apr 27, 2018 at 10:14:04AM +0300, Cyrill Gorcunov wrote:
> On Thu, Apr 26, 2018 at 04:14:16PM -0700, Andrei Vagin wrote:
> > ================ Run zdtm/static/seccomp_filter_threads in uns =================
> > Start test
> > Test is SUID
> > ./seccomp_filter_threads --pidfile=seccomp_filter_threads.pid --outfile=seccomp_filter_threads.out
> > Run criu dump
> > Run criu restore
> > Send the 15 signal to  240
> > Wait for zdtm/static/seccomp_filter_threads(240) to die for 0.100000
> > ######### Test zdtm/static/seccomp_filter_threads FAIL at result check #########
> > Test output: ================================
> > __NR_ptrace filtered inside a sole thread 7
> > Waiting thread 0
> > __NR_mincore filtered inside a sole thread 9
> > __NR_prctl filtered inside a sole thread 8
> > Triggering 0 __NR_ptrace thread 7
> > Triggering 2 __NR_mincore thread 9
> > Triggering 1 __NR_prctl thread 8
> > Waiting thread 1
> > Waiting thread 2
> > 21:30:42.531:     5: ERR: seccomp_filter_threads.c:50: fopen failed (errno = 13 (Permission denied))
> > 21:30:42.531:     5: FAIL: seccomp_filter_threads.c:201: seccomp mode mismatch -1 (errno = 13 (Permission denied))
> 
> int get_seccomp_mode(pid_t pid)
> {
> 	FILE *f;
> 	char buf[PATH_MAX];
> 
> 	sprintf(buf, "/proc/%d/status", pid);
> 	f = fopen(buf, "r+");
> 	if (!f) {
> -->		pr_perror("fopen failed");
> 		return -1;
> 	}
> 
> 	while (NULL != fgets(buf, sizeof(buf), f)) {
> 		int mode;
> 
> 		if (sscanf(buf, "Seccomp:\t%d", &mode) != 1)
> 			continue;
> 
> 		fclose(f);
> 		return mode;
> 	}
> 	fclose(f);
> 
> 	return -1;
> }
> 
> I suspect it's is due to "r+" mode in f = fopen(buf, "r+");. It didn't trigger on
> my machine, neither in travis tests I ran. Andrew, is there simple way to rerun
> this series with plain read-only (ie f = fopen(buf, "r");) without resending
> the whole series?

You can resend one patch in reply to its previous version.