If it passes universe then +1. I wonder if this was a workaround for bfd or just an old mistake.

Add comment noting there is still a race in lagg_ioctl()

Tested this with a builtin and agree it's redundant and dangerous.

Wow. Be sure to get buyin from sjg@ or it may get dropped by accident.

In D25284#557571, @markj wrote:

I think this is a step in the right direction, but it does not completely fix the problem. sc_destroying is set by lagg_clone_destroy(), which later drops the lock and frees the lagg softc. If lagg_port_create() is dereferencing the softc pointer, then it may be dereferencing freed memory. You could probably trigger a panic by enabling memguard on "lagg" and running your repro.

I suspect that we need to use the net epoch to ensure that if_softc points to valid memory, but I don't have any detailed suggestions at the moment.

I think deadlock at https://lists.freebsd.org/pipermail/freebsd-current/2020-May/076100.html may be related to this but I don't understand a lot of it.
Edit: Nevermind, reverting did not help. Sorry for the noise.

Expand tabs to 4 spaces

Handle invalid globs

Invalid exclude for common/ip/tst.*sctp*.ksh*blah

Meant to remove the default TESTBASE before submitting.

Wasn't intended to be pushed yet.

Set NO_CTFCONVERT for no-obj

While everyone's head is in this code have any ideas on https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225934 ?

Use proper version

bump required version

Upload proper diff

Bump CONFIGVERS for change

In D24537#539813, @bdrewery wrote:

Hmm I had this at Isilon but not here.

--- cddl/contrib/opensolaris/tools/ctf/cvt/merge.c
+++ cddl/contrib/opensolaris/tools/ctf/cvt/merge.c
@@ -347,7 +347,7 @@ equiv_node(tdesc_t *ctdp, tdesc_t *mtdp, equiv_data_t *ed)
        int (*equiv)(tdesc_t *, tdesc_t *, equiv_data_t *);
        int mapping;

-       if (ctdp->t_emark > ed->ed_clear_mark ||
+       if (ctdp->t_emark > ed->ed_clear_mark &&
            mtdp->t_emark > ed->ed_clear_mark)
                return (ctdp->t_emark == mtdp->t_emark);

Hmm I had this at Isilon but not here.

--- cddl/contrib/opensolaris/tools/ctf/cvt/merge.c
+++ cddl/contrib/opensolaris/tools/ctf/cvt/merge.c
@@ -347,7 +347,7 @@ equiv_node(tdesc_t *ctdp, tdesc_t *mtdp, equiv_data_t *ed)
        int (*equiv)(tdesc_t *, tdesc_t *, equiv_data_t *);
        int mapping;

This is old but was the last on the mirror I checked. Git has year newer files though. Hmm

I have no opinions on how this should work. I was just trying to follow what was there. The presence of -h is definitely inconsistent now. If you want to pick this up and pick the right route that's fine with me.

Use .sinclude for consistency

The test layout in Poudriere is chaos. I started converting to using OVERLAYS per test that needed special ports. bsd.overlay.mk helps with testing pre-FLAVORS support.

builtins.h does define these prototypes

Fix uninitialized kernel_app_name

Review updates

Fix handling of no LOG_PID on older releases. It is force enabled on head...

Can you still make -C share/mk with this?