Looks good to me, as well. Thank you.

• Drop the use of FLAVORS -- instead, gtk2/gtk3 are now just config options. FLAVORS are apparently not supported for this type of use in a meta port, and triggered an 'incorrectly depends on itself' error when building with poudriere. See also: https://github.com/freebsd/poudriere/issues/582
• Testing with poudriere also revealed a number of missing dependencies; these have now been added.

All requested changes were incorporated, if you wouldn't mind taking another look.

Add upstream fix for double-free in blacklist handling.

Incorporated latest feedback :-)

• Simplify COMMENT variable definition.

Incorporate review feedback:

Adopt FLAVORS, replacing the use of slave ports.

In D18342#394419, @markj wrote:

Sorry to nag, but did you get a chance to test this?

This looks good to me; I'll try to test on real hardware over the weekend, just to be sure.

This was closed by rS328912, but Phabricator doesn't seem to want to import the commit.

• Merge latest bhnd(4)/bwn(4) bug fixes from HEAD.
• Do not attempt to access DMA engine registers prior to performing a core reset; this fixes a machine check exception on PowerPC systems.

Sync up with latest changes to HEAD.

• Add mtx_assert(MA_OWNED) to uninorth_enable_config() for good measure.

Whoops, I got beat to it :-)

• Fix leak of the core register resource in the bwn_attach() error path.
• Fix over-release of the PMU provider.

Do you mind if I close this as superseded by rS326079?

This was closed by rS326454 (phabricator doesn't seem to want to import the commit).

• Removed PCI device ID's from bwn_pci(4)'s device table that are not actually supported by bwn(4).
• Added PCI device ID for the BCM43224.
• Delete bwn_pci(4)'s children on detach; otherwise, we're left with stale children on module unload.
• Use DRIVER_MODULE_ORDERED() to ensure that the bwn_pci-specific bhndb driver is registered before the bwn_pci driver.

• Add SPDX-License-Identifier, update copyright statements, normalize licenses on sources to which I hold copyright.
• Add missing __FBSDID()
• Fixed incorrect PCI ID in if_bwn_pci device table.

• Fix incorrect NVRAM variable type.
• Handle typecasting of pa0b* NVRAM variables for compatibility with the existing bwn(4) APIs.
• Fix 32-bit DMA mask assertion.

• Implement missing support for specifying I/O control flags during core reset.
• Add missing check for siba(4) target reject flags in bhnd_is_hw_suspended()
• Fix siba(4) core suspend/reset behavior:
  • We don't need to wait on SIBA_TMH_BUSY when modifying any target state register; only when waiting for initiated transactions to clear.
  • We do need to wait on SIBA_IM_BY when asserting SIBA_IM_RJ.
  • We need to overwrite any previously set SIBA_TML_REJ flag when bringing the core out of reset. This fixes a lockup that occured when we brought up a core (after reboot) that had previously been placed into RESET by siba_bwn(4).

• Implement missing spuravoid modes
• PWRCTL support is no longer untested; drop the warning message printed to the console
• Drop a couple of unnecessary whitespace deltas relative to HEAD.

Regenerated diff with git's incorrect rename tracking heuristics disabled.

Add missing bhnd(4) generic PMU method table entries.

Applied requested copy and line edits.

Thanks for reviewing; I've inlined a couple questions, and I'll follow up in a bit with the remaining requested edits.

If mips_pic_map_fixed_intr() fails when called from cpu_establish_intr(), include the returned error in the panic() message.