• Cosmetic fixes

• Remove spurious SNDSTAT_UNLOCK() in sndstat_create_devs_nvlist()

• Remove spurious pf->lock unlock and replace with SNDSTAT_UNLOCK() 2

• Remove spurious pf->lock unlock and replace with SNDSTAT_UNLOCK()

In D26884#601870, @hselasky wrote:

What about providing the default sample bit depth used?

• Add nameunit label to device items

• Address delphij's comments

• Cosmetic fixes on vtpci_attach

Does things sound better if we create another label called nameunit to hold something like pcm0, while prepending /dev/ to existing devnode string, so that everytime user only need to directly use the devnode string to open the node without the hassle of snprintf?

• Update the virtio-pci BAR0 detach path as well

Fix the differential

LGTM.
BTW, since I consider the API a draft, the API is subject to changes. :)

In D26884#599928, @hselasky wrote:

Do you have a patch for virtual_oss aswell?

In D26209#594999, @c.koehne_beckhoff.com wrote:

Updates:

• Reuse D24066

Tested. So far it runs well on my setup.

I am curious is it possible you split out mmio remap handling to another differential review? This may make the mmio remap handling part quicker to be reviewed and merged.

Seems not, although both are directories.

• Add .ccls-cache to .gitignore

In D26555#591358, @greg_unrelenting.technology wrote:

Ideally everyone would keep that kind of stuff in their global gitignore (core.excludesfile) :)

but if we're going with this, please also add .ccls-cache/ from the sort-of more popular these days server ccls

In D26514#590312, @cem wrote:

What drops this file? I don’t recognize it. No objection in principle.

In D26514#590175, @mhorne wrote:

Just curious, what tool did you use to generate the compile commands database for FreeBSD? I have a vim linter plugin that is kind of useless for FreeBSD code without one.

Abandoned as D24880 was landed

• security/pam_pkcs11: Fix .example file paths in PAM_PKCS11_EXAMPLEDOCS

• security/pam_pkcs11: Do path replacement of etc/*.example in post-patch target

• www/firefox: Some tailoring jobs

• www/firefox: Replace the occurrence of /usr/xxx with /usr/local/xxx for etc/*.example

In D26303#584645, @jbeich wrote:

Can you submit the change upstream? If you did rename the patch file after the bug.

• sysutils/beats7: Update to 7.9.1
• Update to the 20200906 snapshot of GCC 11.0.0.
• Update to 1.1.0
• Update to 1.0.7. This also fixes the -fno-common issue.
• devel/intel-graphics-compiler: update to 1.0.4879
• devel/gh: Update to 0.12.0
• net-im/tut: Update to 0.0.17
• New port: lang/hermes: JavaScript engine
• Update to 6.4.167
• - Update WWW
• Update lang/erlang-runtime23 to version 23.0.4.
• databases/pgrouting: Update to 3.1.0
• editors/jucipp: Update to 1.6.1
• Fix build on current.
• graphics/gmt: Update to 6.1.1
• - Reset MAINTAINER due to fatal bounces
• devel/tortoisehg: Update to 5.5.1 and allow Python 3 by flavorizing the port. Switch to automatic plist.
• devel/efl and x11-wm/enlightenment: Remove some files that I missed in the last update.
• lang/intel-compute-runtime: update to 20.36.17813
• devel/efl: Mark WAYLAND option BROKEN. Despite the code compiles, it is reported to fail on runtime.
• math/pdal: Update to 2.2.0
• Try to fix ports build.
• Update to the 20200910 snapshot of GCC 8.4.1.
• graphics/py-fiona: Update to 1.8.17
• www/nextcloud: Update to 19.0.3
• databases/virtualpg: Update to 2.0.0
• www/firefox: Fix tab crashing when not using pulse-rust backend

In D26303#584976, @khng300_gmail.com wrote:

In D26303#584645, @jbeich wrote:

Can you submit the change upstream? If you did rename the patch file after the bug.

I would try to seek feedback from upstream before moving in this direction.

Notes on FreeBSD guests (may be updated):
This change also works with:

• 13.0-CURRENT (20200903 snapshot) guest

• Update ObseleteFiles for removal of vm_map_create(9)

In D26209#586036, @c.koehne_beckhoff.com wrote:

Thank's for your feedback.

Yes, that's the expected behavior. As I mentioned in the Known Issues section, bhyve-edk2 is missing a working GOP driver. As a result, graphical output is visible after i915 driver is loaded and initialized.
If I boot into a shell instead of a desktop environment like GDM, I'll get access to a text console. However, the login prompt is the first graphical output that I get.

I back-ported this to my own khng300/freebsd@releng/12.1 and I also had a relatively successful GVT-d passthrough trial! However, was it expected when using this with Linux guest the console being blank? I only saw output right after I started GDM (in Wayland), without access to text console at all.

In D26303#584645, @jbeich wrote:

Can you submit the change upstream? If you did rename the patch file after the bug.

Unregister as long as the stream ended peacefully

• Unregister as long as the stream ended peacefully

At the mean time, the offending code was introduced in bug1647717.

In D25097#553630, @hselasky wrote:

Looks good, but won't that be a lot of patches?

And I see the OSS audio support could also be patched to use the new scaling function where it uses NSEC_PER_SEC .

Updating D23591: bhyve: use stream_read() to read 12 bytes of RFB client version

• bhyve: use stream_read() to read 12 bytes of RFB client version

Opps careless mistakes.

Update according to @jhb 's suggestions.

Add comments in regards to SM961/PM961.

Add comments in regards to SM961/PM961

It seems i get something wrong about the regarding nvme features. I have to look more carefully at it

• Updated according to @mckusick 's suggestion

Fix commit message

Now remove the two affected lines as they are implementation-specific. Instead
one should read vm_map_min.9 and vm_map_max.9.

In D16105#341436, @kib wrote:

This is wrong, it exposes an implementation detail which is really not relevant when _KBI wrappers are used.

These two lines should be removed, and vm_map_min(9)/vm_map_max(9) documented instead.