On Tue, Oct 11, 2016 at 5:30 PM, Noah Watkins <noahwatkins@xxxxxxxxx> wrote:
> Some code using libcephfs running on Jewel isn't working on master.
> After creating a file I receive -EPERM when seeking on the file
> immediately after opening it. I would expect the seek to succeed, or
> the open to have failed.
>
> Here is the code. The check_zero call after ceph_lseek is getting
> permission denied.
>
> #include <cassert>
> #include <stdio.h>
> #include <string.h>
> #include <cstdlib>
> #include "include/cephfs/libcephfs.h"
>
> #define check_zero(ret) do { \
> if (ret) { \
> fprintf(stderr, "error: ret: %d/%s", ret, strerror(-ret)); \
> assert(0); \
> exit(1); \
> } } while (0)
>
> int main(int argc, char **argv)
> {
> struct ceph_mount_info *cmount;
>
> int ret = ceph_create(&cmount, "admin");
> check_zero(ret);
>
> ceph_conf_read_file(cmount, NULL);
> ceph_conf_parse_env(cmount, NULL);
>
> ret = ceph_mount(cmount, "/");
> check_zero(ret);
>
> int fd = ceph_open(cmount, "/hello", O_CREAT|O_RDWR, 0666);
> assert(fd >= 0);
>
> int64_t lret = ceph_lseek(cmount, fd, 0, SEEK_END);
> check_zero(lret);
>
> ceph_close(cmount, fd);
>
> ceph_unmount(cmount);
> ceph_release(cmount);
>
> return 0;
> }
Yeah, I just did this while tightening up our security model, and I
clearly did something wrong. Jeff has a revert
(https://github.com/ceph/ceph/pull/11421/commits/7238b68fdce50e70e3f5c48a1665f33591763b6d)
and says the kernel doesn't do any kind of perm checking in these
calls, but I'm still terribly confused about *why* it broke and how
not checking isn't a security hole...
Wait, I presume you mean libcephfs is master and the cluster is Jewel?
If not something else interesting is going on.
-Greg
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
