FreeNAS Code

enforce ro = false when wo = true and vice-versa

One-stop-shop this puppy.

Create the ISO after creating the full image.

Restore status quo for image naming scheme.

The problem after the build infrastructure was refactored by your's truly was that I didn't fully understand the importance of the path labeling for the full and GUI images.

So, let's restore the prefixing according to what's posted up on the FreeNAS SF site as the nuance of each image with the current naming scheme and previous naming scheme aren't incredibly intuitive.

Follow ${NANO_DISKIMGDIR} appropriate to create_*_diskimage, instead of ${NANO_OBJ} to more tightly couple the two functions.

Also, fix a legacy bug where nanobsd.sh -i would fail because it would always assume that the source file for the GUI upgrade image would always exist.

Clean up the ISO temporary directory so stale context doesn't accumulate in the 'temporary' directory (in reality this should be a legitimate temporary directory).

Don't regenerate the full install image in create_iso.sh. This is incredibly wasteful (both of CPU cycles and disk space): use a hardlink instead.

1. Fix logic error so $CREATE_IMAGE=true doesn't report build failures
incorrectly.
2. Shove image creation down into last_orders within nanobsd so that it will
better handle image creation for us.
3. Get rid of do_build.sh -i because this can be better handled/encapsulated
from within nanobsd.sh.
4. Turn off set -e before running the nanobsd build so that we can get a user
friendly fail message if it indeed fails.

r8499 was a good idea, but unfortunately -- as one discovers with a clean tree -- nanobsd.sh only overrides MODULES_OVERRIDE for installkernel and not buildkernel. So installkernel bombs today with a clean tree. Fun.

Move NANO_MODULES to build/nano_env to avoid polluting nanobsd.sh with too many local changes that can't or shouldn't be pushed back upstream to FreeBSD. Part of the reason why I created it :).

Also, remove MODULES_OVERRIDE from the KERNCONF. This gets us closer to GENERIC, and looks much cleaner.

Bump the number of mbuf clusters from the system default (25600) to 50000.

This should help some people that are using igb and ixgb based NICs.

Replication uses ssh to communicate data across the wire. We should ensure that both endpoints are present and talking when doing the replication, otherwise one side or the other could hang up, resulting in zombie connections, which could get our replication code into an odd funk where it wouldn't properly terminate, and thus could back up the system with multiple requests.

This should resolve some issues seen by a customer (Todd H.) as well as potentially other FreeNAS users.. but this is just a precautionary measure. The zfs send and recieve protocol may need to be cleaned up so that both ends check on each other periodically to ensure that they're in a semi-sane state. This may be a bug as well with ZFS v15 that has been resolved in later versions as well.

Lots of unknowns..

The knob is WITH_BDB_VERSION, not USE_BDB.

Move the specification of which kernel modules to build from the kernel config files to nanobsd configs

1. Use make.conf.install instead of make.conf.build when building the package name.
2. Add USE_BDB=46 to make.conf.install to break bsd.database.mk's 'cleverness' by defaulting to USE_BDB=42.