[PATCH 1/7] PID namespace translation support

[Ákos Uzonyi]

On Fri, 17 Jul 2020 at 16:41, Dmitry V. Levin <ldv at altlinux.org> wrote:
> On Sun, Jul 12, 2020 at 09:46:30PM +0200, Ákos Uzonyi wrote:
> [...]
> > +/**
> > + * Returns a list of PID NS IDs for the specified PID.
> > + *
> > + * @param proc_pid PID (as present in /proc) to get information for.
> > + * @param ns_buf   Pointer to buffer that is able to contain at least
> > + *                 MAX_NS_DEPTH items.
> > + * @param last     ID of NS on which ascencion can be interrupted.
> > + *                 0 for no interruption.
> > + * @return         Amount of NS in list. 0 indicates error, MAX_NS_DEPTH + 1
> > + *                 indicates that ascension limit hasn't been reached (only
> > + *                 MAX_NS_DEPTH values have been written to the array, however).
> > + */
> > +static size_t
> > +get_ns_hierarchy(int proc_pid, uint64_t *ns_buf, size_t ns_buf_size)
>
> Looks like the description of get_ns_hierarchy is out of sync with the
> implementation.

Thanks, will be fixed.

> > +{
> > +     char path[PATH_MAX + 1];
> > +     struct stat st;
> > +     ssize_t ret;
> > +     size_t n = 0;
> > +     int fd;
> > +     int parent_fd;
> > +
> > +     ret = snprintf(path, sizeof(path), "/proc/%s/ns/pid",
> > +                    pid_to_str(proc_pid));
> > +
> > +     if ((ret < 0) || ((size_t) ret >= sizeof(path)))
> > +             return 0;
> > +
> > +     fd = open(path, O_RDONLY | O_NONBLOCK);
> > +     if (fd < 0)
> > +             return 0;
> > +
> > +     while (1) {
> > +             ret = fstat(fd, &st);
> > +             if (ret)
> > +                     break;
> > +
> > +             /* 32 is the hierarchy depth on modern Linux */
> > +             if ((n >= MAX_NS_DEPTH) || (n >= ns_buf_size)) {
>
> Is ns_buf_size allowed to exceed MAX_NS_DEPTH?

Yes, it's allowed to be larger than MAX_NS_DEPTH. Checking (n >=
MAX_NS_DEPTH) condition is useless, however. I'm going to refactor
this function a bit.