remove submodule patch

ping

Especially if this will be per-CPU, it should probably be more compact. A \n per cache level seems like a lot of dmesg spam.

• rebase
• add SYSUNINIT for deallocating the strings
• remove useless null check (kfree supports null)

add missed case

As mentioned in the comment, around 3500 is what I measured for a really hard press, so I rounded up to the next power of two, that's why it's 4096.

Endian convert the initial 32 bit field, use given IDs, do not report width major/minor

switch to macro

remove junk

although PS3 also uses NetBSD

Patches work fine with the current version, just fix the distversion/portrevision conflict when rebasing

Perhaps it's reasonable to take a snapshot u2f.conf but also keep the port for updates?

It could also be used if we support Top Byte Ignore for the same purpose

Switch to sx lock

Use the linuxkpi allocator

add missing + 1 for strlen

format-patch w/o 9999 context: P538

Fix format (more %02d), move the new field to the end of the struct

Fix style

Please don't pull upstream patches into the tree, add the commit hash to PATCHFILES instead (and do a makesum)

rebase: gdm

Uploading the current rebase since @adridg pinged on bugzilla.

Yep, I've had basically the exact same opinion as @delphij about the copyright. Let's go with Intel.

So seems like it's easier to just do it all in core.lua, which is where lots of config accesses are anyway.

Yeah, we should probably make the iic-errnos public instead. In freebsd-embedded-hal the API I'm implementing also prefers more specific errors.

I wonder if https://github.com/freebsd/freebsd-src/blob/e5b431fc0c20771a2a18cb5169d75cf337d0f1dd/sys/kern/imgact_shell.c#L206-L214 can be removed now. (Not saying it *should*)

err, I have not addressed the "isUEFIBoot" thing and the "This file needs a copyright / license at the top" thing…

Updated P528 to rebase after the whitespace cleanup; did the diff without -U9999

Excellent, this very significantly improves the loader's performance on my desktop (AMD Radeon RX Vega GPU)

Another note, some ports (e.g. ruby, node) expect that dtrace can find CC etc via "cc" in path. These need to be patched to respect DTRACEFLAGS (TODO), so I just disabled DTRACE for now.

In D33053#747336, @manu wrote:

amdgpu_hw_i2c is not taken into account if the chip has DC support (so anything > Sea Island).
I hadn't realized that before so I might have a way to test lkpi_iicbb with my hardward.

In D33053#746767, @manu wrote:

But amdgpu found the screen or nothing works ?
May I ask for the pci id ?

You're missing a

(huh, you updated the revision without commandeering it? is that due to some admin rights? kinda confusing that it still belongs to me, please do click "commandeer revision" :D)

Full git patch with metadata for someone to commit: P528 (I'm still not exactly sure why sometimes arc can apply with metadata and sometimes it can't…)

Also add to Makefile.inc1 and SUBDIR_DEPEND_*. Should always work now.

Move the table into a .c file, use in bhyve too

In D32879#742594, @wulf wrote:

There is another one user of this crc16(): usr.sbin/bhyve/pci_nvme.c
May be we should move table to sys/libkern/crc16.c to use single copy across the users?

Well, this one does match the Apple SPI input controller too.

In D32881#742229, @jrtc27 wrote:

this needs resolving properly upstream by having the FreeBSD version propagated properly as cfg info, not by whack-a-mole patching of places all over that RISC-V is FreeBSD 12 and everything else is 11

Change error code to EIO