GitLab

(Note: I didn't make Rob's ssh changes since jay is waiting on
this specifically.)

with the ifc and os setup scripts. Much more robust. Even works.

and so tit can be read by the tb scripts which are running as the
real user now.

except ifc_setup deals in node_ids and, where i make the call, i
am dealing with node IP addresses. also, we are  in the habit (i am
anyway) of playing fast and loose with scoping. specifically, i use
a global scope for my database queries, statement handles and ROWS.
turns out that while in the subroutines GetCurrentOS() and OSFeatureSupported(), i was overwriting a value i was setting before (before i had called
the subroutines) and using after the calls. I think the scripts are
getting sufficiently complicated that we should use better scoping --
we can think about that after the demo.

os_setup and ifc_setup read the features list from the database for
the currently assigned OS. os_setup will not enter a ping wait for the
node to come back alive if the node does not support ping! ifc_setup
will not try and and do ifconfig stuff if the OS does not support ifconfig.

Added calls to create paperbag shell account on paper as well as
add a group entry.

does a chown to the new user, and I was not thinking to pass in a
numeric uid, and this is easier to change than the script. So there.

not /root if we're setuid....

Also, set $< to 0 if $> is 0 - this seems to be necessary to stop sshtb
from dropping root permissions.

setuid root. Modified the Makefile so that it gets the setuid bit
when the post-install target is run.

via the create-os directive in the NS file.
tbsetup/ir/handle_os.tcl - Do a valid check for the image given with
	set-node-os in the NS file, and propogate that information through
        to the IR file. Nothing is added to the DB.
tbsetup/mkexpdir - Add a tftpboot to the list of experiment directories. The
	tftpd daemon now allows kernels from /proj.
tbsetup/os_setup - Very hacky changes to allow for multiboot kernels. Read
	local images table and cross check against that for nodeos spec.
        Hardwire in "mb" as a special partition tag that says to not try
	and do too much with it. This should be changed to a DB check of
	some kind. On reboot, do not wait for these nodes to come alive since
	there is no way to determine if an oskit kernel (or any foreign)
	kernel is running.

Two changes:

	- changed the uname ssh to accept ssh output in addition
	  to the uname output. Initially, i thought this was hackish
	  but, given ssh'es propensity to output messages for every
 	  conceivable occasion, now i think this is simply robust.

	- previously, i was using perl's grep() function to
	  detect duplicate IP addresses. testing with more complicated
       	  scenarios revealed that not only was i misusing the grep()
          function, but it wasn't sufficient even when used correctly.
          Many of the IP addresses aren't reachable from paper; the script
          would deal with this, but only after lengthy delays while
          waiting for input. i changed the script to only communicate
          with interfaces on the control net.

nodes and vlans.

has permission to mess with the nodes specified. Some general cleanup.
Redo the freebsd approach completely. Instead of copying over the current
rc.conf.local and munging it, I build them up locally and just overwrite
what is on the node. This cuts down on the number of ssh/scp commands that
get invoked (which appear to cause all kinds of problems). Its also more
rebust than trying to devine whats in that file. In fact, we should change
/etc/rc.conf to load, say, /etc/rc.conf.tbifc so avoid any clashes with
local stuff that might/will exist in that file.

for no more replies. Still not great, and this causes the loop to reboot
all the machines to get kinda long.
More important is that we have to wait until all the nodes reboot and come
back so that the next part tbrun does not fail. That adds a bunch of time
to this. Needs to parallelize the reboot and wait, but thats too hard too
deal with right now.

group or local root in that project.

approveuser.php3: Add in call to SUEXEC(mkacct-ctrl project user );

happens when tbprerun does its thing. Causes experiment creation to
stop, when using delay nodes.

fix at some point.

next_boot_path so that OSKit-NETBOOT.SILENT does the right thing.
This gets pulled from the disk_images table. Temporary though, since
we want to allow more flexibility in how the OS and related stuff
gets specified.

mkacct-ctrl: fixed the EUID/UID problem; removed all su1s from
'su1 ssh' calls.

tbrun.tcl: added call to mkacct.

rmacct: added a die if we fail to open uname pipe.

are checked against the disk_images table. There is not dnard support
and no create-os support.

(nodes that were allocated) so that subsequent runs don't run out
of resources. Not appropriate to force user to run tbend on failure.

checking, database queries and machine-pinging is done, i set
the UID = EUID for the remaining root-only commands. Cleaned up
the su1s from the commands as now they're unnecessary (and wouldn't
have worked for non-su1-able users, anyway).

I tested this as user tu1 - one of my test users with no
privileges (i set tu1's shell temporarily to tcsh) on paper. its
a good thing because this revealed some  typos in my queries for
the non-privileged-user path. Fixed those.

Still need to deal with "stop" button.

the same. Added chown/chmod lines in the post-install section.

rmuser call to 'pw userdel'. added call to killtip.

appropriate for the node type). This prevents the user from editing
the IR file directly and specifying a bogus image ID.