[ENBD] diskless enbd-client/enbd on initrd

[Peter T. Breuer]

"Also sprach Rudolph Bott:"
> Starting enbd-client...
> enbd-client   572: client says target 0 is 172.16.20.1:1111
> mount: No such device or address
> pivot_root: Device or resource busy

Here is your problem. A mount fails, and then a pivot root fails. You
want to strace this part of the procedure and pin the problem down
further.

> -- snip --
>    if ((pid =3D fork())) {
>      waitpid(pid,NULL,0);
>    } else {
>      fprintf(stderr, "Loading enbd module...\n");
>      system("/bin/insmod /modules/enbd.ko");
>      perror("exec");
>=20
>      fprintf(stderr, "Starting enbd-client...\n");
>      execl("/bin/enbd-client", "enbd-client", "172.16.20.1:1111", "-n",=
=20
> "1", "-i", "xenoo", "/dev/nda", NULL);
>      perror("exec");
>    }

You forked a process to connect to the enbd server, but did not wait
for it (you waited for the parent to die instead - that'll happen
rapidly since near the first thing the client does is go daemon).

>    if (mount("/dev/nda", "/mnt", "ext3", MS_RDONLY, "")) {
>      perror("mount");
>    }

And here the mount dies probably because there is no device yet, but
you want to strace it and see why it returns EBUSY. The man page says
that either /dev/nda is mounted already, or /mnt is occupied or
referenced. You want to strace and see some more.

  Specialfile is already mounted.  Or, it cannot be remounted read-only,
  because it still holds files open for writing.  Or, it cannot be
  mounted on dir because dir is still busy (it is the working direc=AD
  tory of some task, the mount point of another device, has open files,
  etc.).


If I were you I'd pipe from the client and see when it's ready. r  lok
at /proc/nbdinfo.

It maybe is a good idea to make the client daemon stay in foreground
for a while, until it knows the status of its child's connect. I
thought it almost did that (by accident) as it is, just to get error
messages to the console, but it's not deliberate. I could make it so.

Peter