From: Zhihang Wei <wzh@ilbers.de>
To: Jan Kiszka <jan.kiszka@siemens.com>,
isar-users@googlegroups.com,
Quirin Gylstorff <quirin.gylstorff@siemens.com>
Subject: Re: [PATCH v2] qemuarm-trixie: Workaround with missing drivers in qemuarm-trixie initramfs
Date: Tue, 3 Mar 2026 14:12:24 +0100 [thread overview]
Message-ID: <8eb3ee5f-27f2-4c86-a546-e46074fc71c1@ilbers.de> (raw)
In-Reply-To: <79032cb0-c5e8-478b-8b99-84bd95195ab1@siemens.com>
On 2/20/26 17:43, Jan Kiszka wrote:
> On 20.02.26 16:55, Zhihang Wei wrote:
>>
>> On 2/20/26 16:36, Jan Kiszka wrote:
>>> On 20.02.26 15:42, Zhihang Wei wrote:
>>>> This is a workaround to fix the current qemuarm-trixie image unbootable
>>>> issue.
>>>>
>>>> Starting with Debian Trixie, update-initramfs invokes "dracut-install"
>>>> to collect and install required drivers into the generated initramfs.
>>>> "dracut-install" relies on fts_open() / fts_read() from glibc to
>>>> traverse directories and locate drivers.
>>>>
>>>> Due to a long-standing bug [1] between qemu and glibc, the fts_*
>>>> functions may fail to find files on certain 32-bit architectures. As a
>>>> result, required modules such as virtio_blk are not detected and not
>>>> added to the initramfs. The produced image then fails to boot under
>>>> qemu because the block device driver is missing.
>>>>
>>>> A similiar dracut bug report was filed in 2024 [2], pointing to this
>>>> upstream glibc issue reported in 2018 [1]. No upstream fix has been
>>>> applied, and the issue appears to affect only qemu builds for specific
>>>> 32-bit targets.
>>>>
>>>> As a temporary workaround, use a customized initramfs-hook to append
>>>> the neccessary drivers that are currently missed from the initramfs.
>>>>
>>>> For a complete fix, we either need to push for an upstream glibc/qemu
>>>> fix, or convince dracut to avoid using these non-POSIX fts_* functions
>>>> and use opendir() / readdir() instead.
>>>>
>>>> [1] https://sourceware.org/bugzilla/show_bug.cgi?id=23960
>>>> [2] https://bugs-devel.debian.org/cgi-bin/bugreport.cgi?bug=1079443
>>>>
>>>> Signed-off-by: Zhihang Wei <wzh@ilbers.de>
>>>> ---
>>>> Changes in v2:
>>>> - Use initramfs-hook to install the driver, instead of adding another
>>>> variable
>>>>
>>>> meta-isar/conf/multiconfig/qemuarm-trixie.conf | 4 +++-
>>>> .../initramfs-virtio-blk-hook_0.1.bb | 8 ++++++++
>>>> 2 files changed, 11 insertions(+), 1 deletion(-)
>>>> create mode 100644 meta/recipes-initramfs/initramfs-virtio-blk-
>>>> hook/initramfs-virtio-blk-hook_0.1.bb
>>>>
>>>> diff --git a/meta-isar/conf/multiconfig/qemuarm-trixie.conf b/meta-
>>>> isar/conf/multiconfig/qemuarm-trixie.conf
>>>> index 5600ab23..20eecd92 100644
>>>> --- a/meta-isar/conf/multiconfig/qemuarm-trixie.conf
>>>> +++ b/meta-isar/conf/multiconfig/qemuarm-trixie.conf
>>>> @@ -1,7 +1,9 @@
>>>> # This software is a part of Isar.
>>>> -# Copyright (C) 2024 ilbers GmbH
>>>> +# Copyright (C) 2024-2026 ilbers GmbH
>>>> #
>>>> # SPDX-License-Identifier: MIT
>>>> MACHINE ?= "qemuarm"
>>>> DISTRO ?= "debian-trixie"
>>>> +
>>>> +IMAGE_INSTALL += "initramfs-virtio-blk-hook"
>>> This covers the initramfs-tools based generation. What if someone
>>> enables dracut (kas/opt/dracut.yaml)?
>> A pure dracut generated initramfs for qemuarm-trixie cannot boot. It
>> surprisingly has virtio_blk, but still not bootable. Need to find the
>> reason.
>>
>> Actually, there should be more drivers installed into initramfs by
>> initramfs-tools, but they are now missing due to the same bug. Adding
>> virtio-blk just makes this initramfs-tools generated initramfs
>> "minimally" bootable. My guess is that the dracut generated initramfs
>> is missing some other necessary drivers.
> initramfs and "necessary" drivers is tricky for us. Normally, those are
> probed on the machine that will use the initramfs afterwards, but that
> does not work with Isar. With trixie, some unconditionally installed
> modules were dropped, and that already caused extra efforts downstream
> (unless you are on Debian kernels, it's often best to simply compile
> them in).
>
> But maybe there is more involved with plain dracut, and Quirin has some
> idea.
Any update on this issue?
I checked the pure dracut-generated initramfs for qemuarm-trixie. It is
missing virtio_mmio, so it cannot boot. Since it is the low-level
dracut call, "dracut-install", which introduced this bug, both
dracut-based and initramfs-tools-based initramfs generation are
affected.
Also, some targets (bananapi, nanopi-neo) cannot boot with trixie. It
seems their initramfs images are missing mmc drivers. I'll check if I
can extend this hook to allow them to boot as well.
Nevertheless, I would like to apply this patch as soon as possible. It
would be an example workaround until the upstream glibc/qemu issue is
resolved. I will also document the motivation in the hook recipe.
Zhihang
--
You received this message because you are subscribed to the Google Groups "isar-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isar-users+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/isar-users/8eb3ee5f-27f2-4c86-a546-e46074fc71c1%40ilbers.de.
prev parent reply other threads:[~2026-03-03 13:12 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-02-20 14:42 Zhihang Wei
2026-02-20 15:36 ` 'Jan Kiszka' via isar-users
2026-02-20 15:55 ` Zhihang Wei
2026-02-20 16:43 ` 'Jan Kiszka' via isar-users
2026-03-03 13:12 ` Zhihang Wei [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8eb3ee5f-27f2-4c86-a546-e46074fc71c1@ilbers.de \
--to=wzh@ilbers.de \
--cc=isar-users@googlegroups.com \
--cc=jan.kiszka@siemens.com \
--cc=quirin.gylstorff@siemens.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox