Messages in this thread | | | Subject | Re: [PATCH v2] mm, oom: Tolerate processes sharing mm with different view of oom_score_adj. | From | Tetsuo Handa <> | Date | Sat, 2 Feb 2019 20:06:07 +0900 |
| |
On 2019/02/01 18:14, Michal Hocko wrote: > On Fri 01-02-19 05:59:55, Tetsuo Handa wrote: >> On 2019/01/31 16:11, Michal Hocko wrote: >>> This is really ridiculous. I have already nacked the previous version >>> and provided two ways around. The simplest one is to drop the printk. >>> The second one is to move oom_score_adj to the mm struct. Could you >>> explain why do you still push for this? >> >> Dropping printk() does not close the race. > > But it does remove the source of a long operation from the RCU context. > If you are not willing to post such a trivial patch I will do so. > >> You must propose an alternative patch if you dislike this patch. > > I will eventually get there. >
This is really ridiculous. "eventually" cannot be justified as a reason for rejecting this patch. I want a patch which can be easily backported _now_ .
If vfork() => __set_oom_adj() => execve() sequence is permitted, someone can try vfork() => clone() => __set_oom_adj() => execve() sequence. And below program demonstrates that task->vfork_done based exemption in __set_oom_adj() is broken. It is not always the task_struct who called vfork() that will call execve().
---------------------------------------- #define _GNU_SOURCE #include <stdio.h> #include <stdlib.h> #include <sys/types.h> #include <sys/stat.h> #include <fcntl.h> #include <unistd.h> #include <sched.h>
static int thread1(void *unused) { char *args[3] = { "/bin/true", "true", NULL }; int fd = open("/proc/self/oom_score_adj", O_WRONLY); write(fd, "1000", 4); close(fd); execve(args[0], args, NULL); return 0; } int main(int argc, char *argv[]) { printf("PID=%d\n", getpid()); if (vfork() == 0) { clone(thread1, malloc(8192) + 8192, CLONE_VM | CLONE_FS | CLONE_FILES, NULL); sleep(1); _exit(0); } return 0; } ----------------------------------------
PID=8802 [ 1138.425255] updating oom_score_adj for 8802 (a.out) from 0 to 1000 because it shares mm with 8804 (a.out). Report if this is unexpected.
Current loop to enforce same oom_score_adj is 99%+ ending in vain. And even your "eventually" will remove this loop.
| |