Date: Sat, 6 May 2023 09:33:58 +0200 From: Miroslav Lachman <000.fbsd@quip.cz> To: Dimitry Andric <dim@FreeBSD.org> Cc: Dave Cottlehuber <dch@skunkwerks.at>, stable@freebsd.org Subject: Re: etcupdate created empty files = system unbootable Message-ID: <004e8a85-7bc1-6b33-8a42-70d7b43f9497@quip.cz> In-Reply-To: <7afb4128-3eef-f2d9-07c0-035a237cba45@quip.cz> References: <312ae63f-4fab-e194-c603-a85fd11f6edd@quip.cz> <bae5d97b-42af-48a6-abf8-ff88ac19c62e@app.fastmail.com> <0db3bbf1-9e98-d721-c734-fdfe2483cbf0@quip.cz> <95169DC0-5DB6-41D1-8B04-5A450094529F@FreeBSD.org> <19399711-4a71-7025-7ceb-8dd481ec0038@quip.cz> <7afb4128-3eef-f2d9-07c0-035a237cba45@quip.cz>
next in thread | previous in thread | raw e-mail | index | archive | help
On 04/05/2023 22:59, Miroslav Lachman wrote:
[..]
> I still don't know how to reproduce it / what is the root cause, but it
> has happend again. This is the resulting list of empty files after a
> "successful" run of etcupdate after upgrading from 12.3 to 13.2.
>
> # find /etc/ -type f -empty ! -mtime +15m -ls
> 65994 0 -rw-r--r-- 1 root wheel 0 May 4 22:19 /etc/rc.firewall
> 66060 0 -rw-r--r-- 1 root wheel 0 May 4 22:19 /etc/network.subr
> 65814 0 -rw-r--r-- 1 root wheel 0 May 4 22:20 /etc/ssh/ssh_config
OK, now I have all the details of why it failed, and why etcupdate
continued even when the file copy failed.
I found dozens of errors like this in /var/db/etcupdate/log after
running "etcupdate".
>>> cp -Rp /var/db/etcupdate/etcupdate-0PvaUI8/etc/rc.d/zfskeys
/etc/rc.d/zfskeys
cp: /var/db/etcupdate/etcupdate-0PvaUI8/etc/rc.d/zfskeys: Function not
implemented
>>> cp -Rp /var/db/etcupdate/etcupdate-0PvaUI8/etc/rc.d/zpool
/etc/rc.d/zpool
cp: /var/db/etcupdate/etcupdate-0PvaUI8/etc/rc.d/zpool: Function not
implemented
>>> cp -Rp /var/db/etcupdate/etcupdate-0PvaUI8/etc/rc.d/zpoolreguid
/etc/rc.d/zpoolreguid
cp: /var/db/etcupdate/etcupdate-0PvaUI8/etc/rc.d/zpoolreguid: Function
not implemented
As you can see, cp failed, but etcupdate continued without any sign of
error.
cp failed because I did installkernel & installworld without rebooting.
I have never run into a problem like this in the past, but I understand
the reason - running the new 13.2 binary "cp" on an older 12.3 kernel is
not supported.
However, there is a bug in etcupdate which causes it to continue
silently instead of stopping when the first copy fails.
This function needs to check the return status of cp and not blindly
return 0 even if the copy failed. In other words, the function
install_new() does not do what the comment says. This needs to be fixed.
# Install the "new" version of a file. Returns true if it succeeds
# and false otherwise.
#
# $1 - pathname of the file to install (relative to DESTDIR)
install_new()
{
if ! install_dirs $NEWTREE "$DESTDIR" $1; then
return 1
fi
log "cp -Rp ${NEWTREE}$1 ${DESTDIR}$1"
if [ -z "$dryrun" ]; then
cp -Rp ${NEWTREE}$1 ${DESTDIR}$1 >&3 2>&1
fi
post_install_file $1
return 0
}
I will open PR with all the details today.
Kind regards
Miroslav Lachman
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?004e8a85-7bc1-6b33-8a42-70d7b43f9497>