volcreate should handle no acls more gracefully

[macrotex]

When you run volcreate one of the steps is to set ACLs via the volume_setacls() function. That function has this line:

system($FS, 'setacl', '-path', $mtpt, '-acl', @acls) == 0

However, if @acls is empty that call fails (and the whole script dies) with the error message

fs: The field '-acl' isn't completed properly
Failed to set acls (status 8)

So, I suggest skipping the fs setacl system call if @acls is empty. (Note: there are times when I need to create a volume and there are no ACLs to set at creation time.)

[jaltman]

I agree that there should be no attempt to set an acl if there is no acl to set.

If the local vos process is AuriStorFS, volcreate should be attempting to set acls for the volume root directory using "vos setrootacl" instead of using "fs setacl" as the former doesn't require the volume to be mounted. In which case, the "fs setacl" should be skipped if/when a mountpoint is created.

@whm can you look at this?

[macrotex]

Using "vos setrootacl" sounds like a good idea. Ideally we would like as many of our AFS administrative functions as possible to run in a container with no mounted AFS file system.

[jaltman]

To run without a kernel cache manager, the creation of the mountpoint will need to be performed using "afsio makemount". Removal can be performed using "afsio removemount".

[whm]

I'll look at it this week if I can.

[whm]

I have updated volcreate to use "vos setrootacl" and I am testing it now.

I see the afsio binary. I don't see a man page. From the help I am guessing that a command line like:

afsio makemount -path /afs/ir/public/whm -volume public.whm  -rw

is what I should implement. Are there other options I should consider? Specifically I am curious about the implications of the -mountcell and -dynroot switches.

Note, in looking at this code there are quite a few things that I would like to change that will not change the functionality, but will improve the code quality. Well, at least improve according to my style opinions. Specifically I dislike the code overriding the perl system command and the liberal use of 'local $_'. But, I save those changes for a separate update and give time for feedback.

[jaltman]

Specifically I am curious about the implications of the -mountcell and -dynroot switches.

"afsio makemount -mountcell " is the equivalent of the "fs mkmount -cell " option. Its used for creating a mountpoint to a volume in a cell other than the local workstation (aka "thiscell") cell.

The "afsio -dynroot <yes|no>" parameter permits the dynroot functionality to be disabled. This is the equivalent of the "afsd -dynroot <yes|no>" option. Dynamic root provides the ability to dynamically discover cells which are not present in the local cell configuration database aka CellServDB or the /etc/yfs/yfs-client.conf [cells] section.