Re: Panic on boot for armv7 on rpi2 [when was NETLINK added to kernels vs. such as failure starting]
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Mon, 26 Jan 2026 00:14:57 UTC
[Note: the message this replies to does not seem to have managed to appear on the list despite the cc: for freebsd-arm@ being in the Email that I got. May be the @net-art.cz address is not subscribed?] On Jan 25, 2026, at 09:12, Alex Samorukov <samm@net-art.cz> wrote: > I think race was always there, but NETLINK needs to be in kernel to trigger it. > > Which was added in FreeBSD 14, according to https://github.com/freebsd/freebsd-src/blob/releng/14.0/sys/arm64/conf/DEFAULTS. > Note the std.armv7 in: # grep -r NETLINK /usr/src/sys/*/conf/ /usr/src/sys/amd64/conf/DEFAULTS:options NETLINK # netlink(4) support /usr/src/sys/arm/conf/std.armv7:options NETLINK # netlink(4) support /usr/src/sys/arm64/conf/DEFAULTS:options NETLINK # netlink(4) support /usr/src/sys/i386/conf/DEFAULTS:options NETLINK # netlink(4) support /usr/src/sys/powerpc/conf/DEFAULTS:options NETLINK # netlink(4) support /usr/src/sys/riscv/conf/DEFAULTS:options NETLINK # netlink(4) support The below also applies to sys/arm64/conf/DEFAULTS as also being included in the MFC and the commit in main . NETLINK is not new on the scale of the change of behavior showing up. It was present for armv7 and the others in 13.3-RELEASE as well, not just 14.0-RELEASE . May be something in NETLINK changed later or something using NETLINK changed later? Supporting details for finding the 13.3-RELEASE status . . . The addition to /usr/src/sys/arm/conf/std.armv7 goes back to 2023-01-31 (UTC) in main : committer Alexander V. Chernikov <melifaro@FreeBSD.org> 2023-01-31 14:22:11 +0000 commit e80699a809a079c4e673940544d6ddd944e8d08e (patch) tree 0ec9651ad94814c4930b7a23a9c72f5438bb951c /sys/arm/conf/std.armv7 parent a3baca22d0b7789a040a60f389e3c132b6e1755b (diff) netlink: add NETLINK to GENERIC. This is a followup of 692e19cf5195 (add netlink to GENERIC@amd64). . . . # ~/fbsd-branches-containing.sh e80699a809a079c4e673940544d6ddd944e8d08e * main remotes/freebsd/HEAD -> freebsd/main remotes/freebsd/main remotes/freebsd/releng/14.0 remotes/freebsd/releng/14.1 remotes/freebsd/releng/14.2 remotes/freebsd/releng/14.3 remotes/freebsd/releng/15.0 remotes/freebsd/stable/14 remotes/freebsd/stable/15 But stable/13 has an MFC of that with its own hash: committer Alexander V. Chernikov <melifaro@FreeBSD.org> 2023-02-09 15:52:09 +0000 commit 6058f6cc48f5ec02476bf46693673bd71291a5c0 (patch) tree a0534e89df6ae7f8259565124a1a385ce2429964 /sys/arm/conf/std.armv7 parent 5e598a761c37dbd491579a0805968ae50ac72465 (diff) netlink: add NETLINK to GENERIC. This is a followup of 692e19cf5195 (add netlink to GENERIC@amd64). . . . # ~/fbsd-branches-containing.sh 6058f6cc48f5ec02476bf46693673bd71291a5c0 remotes/freebsd/releng/13.2 remotes/freebsd/releng/13.3 remotes/freebsd/releng/13.4 remotes/freebsd/releng/13.5 remotes/freebsd/stable/13 I'll note that: https://cgit.freebsd.org/src/blame/sys/arm/conf/std.armv7?h=releng%2F13.2 does not show a NETLINK line but: https://cgit.freebsd.org/src/blame/sys/arm/conf/std.armv7?h=releng%2F13.3 does show it. This is because of releng/13.2 having its own revert of the addition: committer Alexander V. Chernikov <melifaro@FreeBSD.org> 2023-02-27 16:33:29 +0000 commit a052a188766f022d4b0cb2484bda636a3345f644 (patch) tree 2ad4c1e3d572a72d238e6c21198ea46f1050e521 /sys/arm/conf parent 7dea81a4ed19e18a6f1104f14cd3e0ce57c8f5f5 (diff) Revert "netlink: add NETLINK to GENERIC." This is a Direct commit to releng/13.2. . . . > On 25.01.2026 0:37, Mark Millard wrote: >> On 1/9/26 16:27, bob prohaska wrote: >>> On Fri, Jan 09, 2026 at 01:14:56PM -0800, Mark Millard wrote: >>>> The call chain in the backtrace in your original report is very >>>> different than the one shown in buzilla 291294 (see above url). So >>>> separate bugzilla reports looks to be what is appropriate. >>> Reported as bug 292314 >>> >>> I have two rpi2's running current, both crash during shutdown -r >>> but they seem to issue different messages. Both are included >>> in the bugzilla report. One uses a relatively modern usb3-sata >>> bridge while the other uses an old usb2-pata bridge, but that's >>> the only obvious difference. >>> >>> Thanks for your help! >>> >>> bob prohaska >>> >>> >> There are other bugzilla reports for at least both RPi2B v1.1's (armv7) >> and RPi3B (aarch64) contexts that look similar (in at least part). >> >> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=292695 >> >> has a proposed patch. >> >> I do not remember seeing any reporting that indicates what changed to >> start to need the likes of that patch vs. previously not needing it. >> >> One of the reports goes back to: 2025-10-22 06:26 UTC : >> >> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=290417 >> >> > === Mark Millard marklmi at yahoo.com