I am ok too. I can probably manage to get this committed within next week if there are no objections.

In D17955#421328, @yamori813_yahoo.co.jp wrote:

My understand is sys/gnu/dts have only openwrt(opneled) dts files. And sys/dts have freebsd original dts files. MZK-W04N-XX.dts is my missing add dts file then move to freebsd dts directory. Do you think how to sync openwrt(opneled) dts files ?

Please remove the dts file from this review. Let's import it with the rest of the dts/dtsi files in https://reviews.freebsd.org/D17955

Why was MZK-W04N-XX.dts moved to sys/dts instead of sys/gnu/dts ?

! In D18001#384940, @yamori813_yahoo.co.jp wrote:
OK I will delete Mediatek in this review.

! In D18001#384940, @yamori813_yahoo.co.jp wrote:
Mediatek soc is some time CPU print name and CHIPID is different. I have one that is CPU printed RT3050F but CHIPID is RT3350. I think Mediatek code need check SYSCTL_CHIPID0_3 and SYSCTL_CHIPID4_7. Sure thing first is CHIPID and second is FDT I believe.

Please split Mediatek and Atheros in separate reviews.

as far as if_rt changes and dts additions go, I am ok with this.

OK, v2 patch is here:
https://people.freebsd.org/~sgalabov/mtkswitch_v2.patch

In D9607#198883, @yamori813_yahoo.co.jp wrote:

This is always false in patch.

tmp = v->es_untagged_ports & v->es_member_ports;
/* fail if untagged members are not a subset of all members */
if (tmp != v->es_untagged_ports) {

Because of all v->es_untagged_ports bits contain in v->es_member_ports.

Can you please test with the patch found at:
https://people.freebsd.org/~sgalabov/mtkswitch.patch

OK, one issue exists with the existing code:

• if more ports are configured as "untagged" than are configured as "untagged members" for the given VLAN the check would fail.

Can you explain the problem in a bit more detail?
The code you're suggesting to change (from memory) is intended to do the following:

1. check which ports are configured to untag VLANs
2. get the subset of VLAN member ports that are also untagged.
3. compare the 2 values and if they match - go ahead; if they don't match - error out.

I would like to eventually get rid of all the ifdef's (where possible, of course) related to SoC type (e.g., if defined MT7620, etc).. unfortunately at the moment I have absolutely no bandwidth to start doing that, so I am ok with this for now.

In D8460#175967, @andrew wrote:

In D8460#175966, @andrew wrote:

Can you try with rS308408. It might pay to add an error if this file is included but _MACHINE_ASM_H_ is undefined.

Even better would be _SYS_CDEFS_H_.

Any other comments on this? If it's ok with everyone I'd like to commit it...

Hmm, this is weird, it still fails the same way for me, although I am using the default gcc:
gcc version 4.2.1 20070831 patched [FreeBSD]

In D8459#175948, @meloun-miracle-cz wrote:

Hmm, right.

Btw, which compiler you use? It seems that clang doesn't generate error for unused external variables...

Actual build failure message is:
/src/current.git/sys/sys/boot.h:40: warning: 'howto_names' defined but not used

All files, which report an issue of __ARM_ARCH not being defined (and include machine/armreg.h):
sys/arm/arm/cpufunc_asm_sheeva.S
sys/arm/arm/fiq_subr.S
sys/arm/arm/setstack.s

In D8460#175937, @meloun-miracle-cz wrote:

Are you sure? Also for D8459? I'm able to compile actual SHEEVAPLUG kernel without any issue.

In D8460#175936, @andrew wrote:

When is it the case that __ARM_ARCH is undefined? machine/acle-compat.h is included from sys/cdefs.h so this should have already been included.

Yes, I was planning to commit it on Wednesday if I get a chance, but wouldn't mind if someone else does it in the meantime...

In D7692#160754, @adrian wrote:

(Eg, it needs a bunch more testing on non-FDT, non-INTRNG platforms, but I bet that's going to be easy..)

The updated diff also fixes additional problems seen on Broadcom MIPS (INTRNG without FDT) as a result of the r304459 changes. The new diff was boot tested by mizhka.

Provide more diff context

In D7692#160205, @adrian wrote:

Hm, can you get someone who does FDT stuff on #bsdmips to look at this?

Previous diff was not compiling for BCM kernel config, as BCM only uses INTRNG but not FDT.
Rework the fix, so that cpu_establish_[hard|soft]intr uses a different intr mapping strategy, which doesn't require FDT-specific data to be exposed to non-FDT kernels.

update diff

update diffs

update diffs

Add MT7620/MT7621 first pass driver as well.

add copyright to *.h files.

Update diff

Update diffs to remove MT7621 GIC stuff as it will be unneeded after D6045.

Update diffs so MT7621 can be built