Superseded by a version bump to 20210325.

I can confirm the console spam for non-UFS, ZFS, installations on bare metal (amd64) is gone.

For me, when booting (amd64, BIOS boot) off ZFS, this clears the 'BTX halted' showstopper during early bootup.
UFS is not involved, installer media is unplugged.

@hselasky do you expect me to write something in RELNOTES?

I can do more tests this Saturday.

In D30919#777117, @hselasky wrote:

Update: I tested with a bunch of USB sticks from Kingston and found issues. I will try to resolve those issues and get you an updated patch.

Sorry. This stuff is not easy!

In D30919#777108, @hselasky wrote:

I've now moved that "usbconfig -d xxx reset" into the kernel, so then all devices should work?

In D30919#777053, @hselasky wrote:

Can you show me the dmesg with only this patch applied and no quirks? What goes wrong? Does "usbconfig -d X.Y reset" fix the device again?

In D30919#777053, @hselasky wrote:

Can you show me the dmesg with only this patch applied and no quirks? What goes wrong? Does "usbconfig -d X.Y reset" fix the device again?

Ehm, the last two iterations only /appeared/ to work because I left in the already present quirk entries @ https://github.com/freebsd/freebsd-src/blob/main/sys/dev/usb/quirk/usb_quirk.c#L190

In D30919#776650, @hselasky wrote:

Could you test this version of the patch with your devices? Does it do what you want?

@hselasky I 'm looking at your last implementation at the moment.

@hselasky there seems to exist a trick for outside contributors (those without a commit bit) that I am unaware of, to attach Harbormaster. See https://reviews.freebsd.org/D34220

In D30919#769780, @hselasky wrote:

There is a couple of other places where zero-length SCSI commands are used, which should be DIR_NONE . Can you fix?

Rebased, removed leftover lines (extern ...).

I recall having tested (can dig this up in sent e-mail) 8 devices in the past, apart from the ones that I own. Those last two are:
https://github.com/freebsd/freebsd-src/blob/main/sys/dev/usb/quirk/usb_quirk.c#L227 https://github.com/freebsd/freebsd-src/blob/main/sys/dev/usb/quirk/usb_quirk.c#L406

@hselasky also, there should be a (c) line added to sys/dev/usb/quirk/usb_quirk.h, sys/dev/usb/quirk/usb_quirk.c and sys/dev/usb/usb_device.c I think.

Can you return ownership of this change? I have another revision of this change that doesn't add scsi_quirk_no_test_unit_ready[] and scsi_quirk_no_sync_cache[], and the option (in Phabricator) to attach an update to an existing Revision is no longer available.

In D30919#766846, @hselasky wrote:

Can you explain exactly why this patch is not enough?

In D30919#766773, @hselasky wrote:

Can you test this patch instead on your USB mass storage devices.

I think this is what you are trying to do.

In D30919#766516, @hselasky wrote:

Did you see my comment?

Ping hselasky.

This has my attention, so I tried booting a hybrid Debian ISO [1]. Booting this ISO file using the same firmware build as mentioned earlier successfully loaded the kernel (and beyond).

In D30919#701651, @hselasky wrote:

Try to remove existing mass storage quirks from usb_quirk.c by using:

usbconfig -d X.Y remove_quirk UQ_MSC_NO_SYNC_CACHE
usbconfig -d X.Y remove_quirk UQ_MSC_NO_TEST_UNIT_READY

Then re-plug the device.

--HPS

Did you mean DPRINTF() or DPRINTFN()?

In D30919#701631, @hselasky wrote:

Try:

hw.usb.debug=1

In D28227#653579, @tcberner wrote:

Moin moin

[00:00:22] [01] [00:00:14] Finished devel/tex-kpathsea | tex-kpathsea-6.3.2: Failed: patch
[00:00:22] [04] [00:00:14] Finished devel/tex-libtexluajit | tex-libtexluajit-2.0.3: Failed: patch
[00:00:23] [05] [00:00:15] Finished devel/tex-libtexlua | tex-libtexlua-5.2.4: Failed: extract

mfg Tobias

It was brought up that some files can't be patched because phabricator would handle deletions poorly.
Deleted are:
print/texlive-base/files/patch-texk-dvisvgm-dvisvgm-1.9.2-src-*
print/texlive-base/files/patch-texk-lcdf-typetools-lcdf-typetools-2.104-*

In D28227#631054, @lwhsu wrote:

General suggestion: please use make fetch and make makesum to generate distinfo file, there is also TIMESTAMP field missing.

I don't feel like I get to say what and what not goes into base/sys/.

(earlier I thought the breaking change was https://reviews.freebsd.org/D27579 )

In D27383#613546, @bcr wrote:

Can you describe your use case for this and why it should become part of the general installer?