Needs rebase on top of D60004? (merged)

For the record: this is why code changes for utilities like this need tests written beforehand (@des has been leading by example in this area).

In D60271#1383040, @alc wrote:

@andrew, Is this what you were asking for?

The removal in openssh-portable is:

commit a32d28d792567253bb601362f36391f155f8f772
Author: djm@openbsd.org <djm@openbsd.org>
Date:   Tue May 6 05:40:56 2025 +0000

This commit likely caused a problem with one of the sigtramp tests that uses libunwind: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=289096 .

done

Add copyright text

removed extraneous space

As mentioned in D31489, the inability of LTO and DTrace to work together is unacceptable.

ice(4): Add HW reset check after DDP package load

Do you have a test program you can check in somewhere as a stress test? would be good to catch this if it creeps up again.

See https://reviews.freebsd.org/D60369 for an updated version of this patch.

In D31489#1384532, @0mp wrote:

In D31489#1384510, @vishwin wrote:

In D31489#1239048, @0mp wrote:

maybe helpful: https://antranigv.am/posts/2025/02/dtrace-python-freebsd-again/

DTrace and LTO being mutually exclusive according to this experience is a non-starter.

Hello Charlie,

What do you mean by a non-starter?

Note that DSA support was disabled in 0b707d5fe8b6b1920eb29c6375a6be92b7e44758 in early 2025 (prior to 15.x).
This MFC would remove the code in 14.x. We need to decide between:

• merging 0b707d5fe8b6b1920eb29c6375a6be92b7e44758 and related changes to disable DSA support in 14.x
• leaving the DSA support in place as a local patch set for the duration of 14.x

This may work, but it seems a little bit fragile. Would it not be more robust to omit the MSG_WAITALL for the call when recv is expecting a variable-length string?

Seems reasonable.

@jhb Is this ready for commit?

@jhb Is this ready to commit?

The diff is now just the functional change, the early return in iflib_txd_db_check() when ift_db_pending is 0. The inversion will be posted separately as a no-functional-change follow-up.

LGTM

Ping @jhb and @olce for mentor approval

@olce Would you review this as scheduler?

@olce Would you review this as scheduler?

Ping @jhb for mentor approval

@jhb Is this ready for commit?

Apply even more atomicity to loads.
Drop the mp != NULL check.

Looks good - please go ahead and commit.

simplify

In D31489#1384510, @vishwin wrote:

In D31489#1239048, @0mp wrote:

maybe helpful: https://antranigv.am/posts/2025/02/dtrace-python-freebsd-again/

DTrace and LTO being mutually exclusive according to this experience is a non-starter.