Re: Linux 6.8

From: Frank Scheiner <frank.scheiner@web.de>
To: torvalds@linux-foundation.org
Cc: "Tomáš Glozar" <tglozar@gmail.com>,
	"Sergei Trofimovich" <slyich@gmail.com>,
	gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org,
	Linux-Arch <linux-arch@vger.kernel.org>,
	linux-ia64@vger.kernel.org,
	debian-ia64 <debian-ia64@lists.debian.org>,
	t2@t2sde.org
Subject: Re: Linux 6.8
Date: Wed, 13 Mar 2024 20:30:46 +0100	[thread overview]
Message-ID: <145da253-b3bc-43da-a262-a3ebdfbea5a2@web.de> (raw)
In-Reply-To: <CAHk-=wiehc0DfPtL6fC2=bFuyzkTnuiuYSQrr6JTQxQao6pq1Q@mail.gmail.com>

Dear all,

as usual, an update for Linux/ia64:

As far as I can tell, the v6.8 development cycle for us looked not much
different to the v6.7 one: The ia64 patch set ([1]) was extended where
need was identified. All ia64 machines we have available for testing
continue to work, no system support was lost during this cycle. No, we
even got another "system" back: the HP Sim platform - up to mainline,
that is. Together with Ski [2], thankfully kept together and updated by
Sergei Trofimovich, this allows to run ia64 kernels and (light) ia64
userland software on for example x86_64 hosts. Like it's done for the
recently established auto-builds for Linux stable RCs and releases on
GitHub. Have a look on [3] for example: all building, all working (in
Ski). For the manual testing of the Linux mainline RCs and releases some
changes were introduced. Mainly that compilation always happens with the
latest gcc-14 snapshot starting with v6.8-rc1 - so far no surprises -
and recently, the enabling of LRA for the compiler.

[1]: https://github.com/lenticularis39/linux-ia64

[2]: https://github.com/trofi/ski

[3]:
https://github.com/johnny-mnemonic/linux-stable-rc/actions/runs/8258902207

Unfortunately there's one difference to v6.7 with v6.8 (actually
beginning with v6.8-rc1 as we found out later during the cycle): there
is a userland regression present that leads to segfaults with v6.8 where
it does not with v6.7. We collected the following examples for this
regression (if they are all related):

* Debian: apt(-get) segfaults (though not immediately) and is affected
to a different degree depending on non-usrmerged or usrmerged root FS
* Gentoo: segfaults happen when emerging different packages
* T2: compiling a specific Perl source code file with gcc leads it to
segfault

...and hope to find the cause of it and a solution. Before you ask, no,
this is not due to enabling LRA for the compiler, it already happened
before that was done.

****

This time no new distributions for ia64, but unfortunately one less
soon: Debian will close shop on ia64 ([4]). As much as this make me sad,
because this was the distribution that got me going on ia64 nearly ten
years ago, better switch to another option sooner than later. I am
switching to T2 ([5]) for example for future testing. For network boot
this was really simple to set up, similar to how you can create an
OpenBSD root FS by unpacking a list of tarballs plus some manual
configuration afterwards.

[4]: https://lists.debian.org/debian-ia64/2024/02/msg00002.html

[5]: https://t2sde.org/

****

Thank you all for your hard work on Linux!

Cheers,
Frank at al