Hi Alex,

looking at the code changes, I am not sure if we could just replace
td_umount_and_delete_swap() with td_umount_all().
The reason is that there might be swap device added by previous
call to td_mount_and_add_swap() which we want to release once
we are done there.

Going through the code, I can see that as far as dealing with regular
filesystems is concerned, TD keeps track of 'mounting' operations
carried out in linked list 'umount_head' (see td_mountall.c).

When cleaning up is to be done, td_umount_all() walks through the
list and unmounts only filesystems recorded in that list.

The same mechanism exists for 'swap' - TD keeps track of added swap
devices in 'unswap_head' linked list.

However, it seems that when 'swap' devices are to be released,
td_unswap_all() ignores that list and removes all swap devices
available. I think that instead it should walk through the 'unswap_head'
list and release only those swap devices which were recorded there.

As far as testing procedures are concerned, these changes are in code
which takes care of discovery of existing Solaris instances installed
on UFS filesystem (e.g. Solaris 10).
Target Discovery goes through existing slices, mounts them and tries to
find Solaris instance on them. Please see following document for more 
details:

http://www.opensolaris.org/os/project/caiman/files/dwarf_td_design.pdf

Based on this, I might recommend following testing procedure:

[1] Install Solaris on UFS - e.g. you could install SXCE release 
obtained from
http://www.opensolaris.org/os/downloads/

[2] Create at least one non-global zone

[3] Boot LiveCD environment, create swap device on one of free slices
    and run tdmgtst test driver - verify that

* it finds and reports UFS Solaris instance
* things are cleaned up correctly after discovery, i.e.
  - no new mountpoints appeared
  - nothing disappeared (swap device)

Thank you,
Jan


Alexander Eremin wrote:
> Hi - here's a small webrev to remove swap deleting during os discovering
>
> webrev:
>
> http://cr.opensolaris.org/~alhazred/7915/
>


Reply via email to