linux-ext4 - Re: simplify ext4_sb_read

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 02 Jun 2022 13:47:03 -0400
From:   Gabriel Krisman Bertazi <krisman@...labora.com>
To:     "Stephen E. Baker" <baker.stephen.e@...il.com>
Cc:     linux-ext4@...r.kernel.org
Subject: Re: simplify ext4_sb_read_encoding regression

"Stephen E. Baker" <baker.stephen.e@...il.com> writes:

> On Mon, May 30, 2022 at 10:21 AM Gabriel Krisman Bertazi
> <krisman@...labora.com> wrote:
>>
>> "Stephen E. Baker" <baker.stephen.e@...il.com> writes:
>>
>> > Hello,
>> >
>> > I have a Samsung Chromebook Plus (rk3399-gru-kevin) which boots linux
>> > off an external ssd plugged into USB. The root filesystem is ext4 with
>> > unicode support, case folding is enabled only on some directories in
>> > my home directory.
>> >
>> > Since 5.17 the system has been unbootable. I ran a git bisect and it
>> > pointed to aa8bf298a96acaaaa3af07d09cf7ffeb9798e48a ext4: simplify
>> > ext4_sb_read_encoding
>>
>> Hi Stephen,
>>
>> This series moved the UTF-8 data tables to a kernel module; before it,
>> the module had to be built-in.
>>
>> Since you have your rootfs as a case-insensitive filesystem, either the
>> utf8data module needs to be available in the initramfs or unicode
>> needs to be built-in.  Are you building your own kernel?
>>
>> Can you confirm that utf8data.ko exists in your initramfs, and
>> regenerate it if missing?  Alternatively, make sure that you have
>> CONFIG_UNICODE=y in your kernel configuration file.
>>
> Thanks Gabriel, I've verified that CONFIG_UNICODE=y, as well as
> CONFIG_UNICODE_UTF8_DATA which exists in this patch for that
> purpose, though it was removed earlier.

Thanks for checking.  Indeed, UNICODE_UTF8_DATA was just transitional,
and went away right before being part of a release.  CONFIG_UNICODE=y
should have adressed it, if it was a problem with module not being
available early enough.

>
>> If that doesn't work, can you provide the kernel log?  If you can't
>> collect the console output, a photo of the screen displaying the error
>> will suffice.
>>
> I don't have any output to provide unfortunately. It fails before the
> backlight turns on, and nothing is written to disk. I seem to remember
> that someone else at Collabora had figured out a way to get a serial
> console on this device - perhaps Tomeu. I'm not equipped for that
> personally, particularly if it involves soldering.

I'm following the discussion with Ted on the other subthread, but I
don't have anything to add at the moment other than what is already
said.

Nothing stands out on that commit specifically and I couldn't reproduce
it in a vm.  I've reached out to Tomeu to get my hands on that exact
chromebook, to try to reproduce it there.  I will report back with my
findings.

-- 
Gabriel Krisman Bertazi