In D54037#1234622, @imp wrote:

Why can't we make virtio_p9fs.ko depend on p9fs.ko? That would also cover this case, no? And would confine their visibility to that one module, no? It's how we do pci, etc.

if you confirm Rupesh Pilania <rupeshpilania@gmail.com> as the git commit author I'll push this, thanks!

Looks like:

Add notes from wulf@

LGTM

FYI I didn't look closely at the ones after XHCI_DCST

It is better to provide an explicit list of symbols than just EXPORT_SYMS=YESunless we are sure that all symbols should be exported. But, this enough to preserve the status quo at least.

If it renders like https://www.freebsd.org/releases/13.0R/relnotes/ and not https://www.freebsd.org/releases/14.0R/relnotes/ looks good to me

@mav reported a regression; at least one issue is linux.ko depends on kern_kmq_open as of 97add684f5306ebf93be238a0340597ba1898d0e. Now fixed by eb90470f5f2a0e5c820c47be329423f5c60ca247 and a897492770735d4e5a55cbc06a02e267ca41f5b8. @cy reported a regression when agp is loaded as a module, fixed by a87c1e2dd8fc997d6ee603c252c543afe8a4d2a4.

The issue I have with "maintenance" is that then 14.x releases are production releases that get maintenance updates and 13.x releases are maintenance releases that are suitable for use in production.

All the variables in here would benefit from documentation but we can start with these

Would be good to indicate in the commit message how far back in gcc and clang __builtin_align_up is supported.

Thanks. Thinking more about it, it does seem sensible to execute all of the teardown steps for this module before doing the dependencies.

LGTM. I do wonder if there's a reason the steps in the kld_unload EVENTHANDLER aren't (shouldn't be) handled in the same place as sysctls / sysuninits?

Writing seems fairly straightforward but indeed would be nice to have centralized infrastructure for it.

@wulf can we add a sentence or two about any user-facing impact, config changes, etc.? For most people it will be a no-op.

Will add to the commit message:

This isn't used by modern cards, but is needed for i915kms to load
on a system that has agp as a module not compiled into the kernel.

I've applied this to my local tree for testing

Are there kernconfs to compile the module? We also usually put that in synopsis.

I see this formulation in share/man/man4/mgb.4 and share/man/man4/muge.4 and I probably looked at one of those as a template. I'll take a look at updating those first. mgb(4) is a driver for PCIe GigE hw, muge(4) USB GigE.

Update HARDWARE introduction, thanks @ziaee

drop quotes, add spdx

I've been testing this with Binardat 10G Ethernet PCIe Network Adapter, AQUANTIA AQC107 chip LAN Controller, 10G/5G/2.5G/1G/100Mbps RJ45 NIC Card,Support PXE for Windows/Linux/Mac

Also please add it to the list in crypto/openssh/FREEBSD-upgrade

Delete even from minor to minor?

PR238324

Three patches obtained from net/aquantia-atlantic-kmod get the driver to build:

commit 74807fef8838c93f783d151f7200bfcc4e9e030a
Author:     Rozhuk Ivan <rozhuk.im@gmail.com>
AuthorDate: Wed Oct 26 15:37:57 2022 +0300
Commit:     Ed Maste <emaste@FreeBSD.org>
CommitDate: Tue Nov 18 18:27:39 2025 -0500

Cleanups and build infrastructure will follow; mainly wanted to pop this into review for a couple of points

• I didn't do this via a vendor branch. https://github.com/Aquantia/aqtion-freebsd looks like it is is abandoned and will not receive further updates. NetBSD and OpenBSD have derivatives of this driver with additional hardware support. Thus I think there is no value in bringing this into a vendor branch -- we'll have the initial import as a snapshot, but from there we will just iterate on it in the tree as our own code, including borrowed changes from NetBSD and/or OpenBSD.
• It has been restructured in NetBSD and OpenBSD, but I still think it makes sense to import this working driver as-is first rather than starting with a port of NetBSD's/OpenBSD's driver.
• I should probably strip out the non-src files (.cirrus.yml, .gitignore etc.) and just keep *.[ch]. The files have the license included so the standalone LICENSE is redundant.