Grr. Framebuffer scrollback is gone

[miket]

There's a feature I don't use much--but when I do, I do depend on it.

I was setting up a new machine and had some problems at boot time (gentoo-sources-5.10.27). It seemed that an interesting message scrolled off the screen, so I hit Shift-Page Up to see it. Hmm. Didn't scroll up.

A search led to a Phoronix article quoting Linus saying that no one uses FBCON any more so the scroll functionality was coming out as of 5.9.rc5.

Ouch! How do I find that message? Dmesg doesn't record messages from either early boot or init. As far as that goes, /var/log/messages doesn't either (and the machine is borked enough that this isn't being written anyway).

[Zucca]

I do hope someone competent enough continues to maintain it.
_________________
..: Zucca :..

[Goverp]

Yup, scrollback's been gone for some time. The usual answer is to use "app-misc/screen" or IIUC tmux; I use screen, though I don't like it much (too many other features and a weird magic keys to get it to scroll back - "Ctl-[" "a" "PgUp" ?). Of course, you need to remember to start it before you type any other commands, or include it in a shell profile.

<edit> Sorry, of course you can't start screen that early.

I think there are some early printk and pstore kernel configuration options that can send compressed output to pstore, though I couldn't get it to work when I tried some time ago. I read somewhere (perhaps a note from Neddie Seagoon) that you need to spuriously enable serial port console stuff to get early printk to work.

IIUC you really don't want to run with this pstore stuff enabled all the time, as pstore (assuming you use efivars for storage - maybe there are better alternatives) is strictly limited, and overfilling it can brick you machine; hence specify xz compression or similar.
_________________
Greybeard

[NeddySeagoon]

miket,

Install over ssh. Every live media has that, don't they?
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[pa4wdh]

Removing framebuffer scrollback seems like a bad idea to me ...

[wjb]

Setting rc_logger in /etc/rc.conf gets you a little bit more than dmesg, but not the earliest stuff.

[pietinger]

[Hu]

Scrollback was removed for having known bugs, some of which had (somewhat obscure, iirc) security implications. Linus hoped that someone competent to maintain it would object and resurrect it. So far, despite having inconvenienced some users, nobody has stepped up to take over that subsystem, as far as I know.

Although a bit of a hack, in some cases, configuring a framebuffer driver might be helpful because it will give the kernel more total lines on screen. Depending on timing, OP might be able to use flow control to suspend output before the interesting message scrolls out. I've dealt with systems that produce output too fast to rely on a human to suspend output at the right time, though.

OP's problem is not that it is too early for dmesg, but that it is not a kernel-generated message, so it was never in dmesg and never will be. He needs to capture some user-program generated message(s).

[miket]

From what I found on lwn.net, I'm not alone. There are a couple of kernel devs who are working on it (or at least started). Linus' original beef was about odd bugs in a system that has accumulated a lot of cruft over the years; the security issue was reported later. He wants a clean refactoring of the system.

I like pietinger's rc_interactive="YES" suggestion. I also found it helpful to start the Busybox recovery shell in my initramfs: at least I can see the messages.

That brings up an issue that others have noted: the Busybox recovery shell also loses its scrollback.

Good luck trying to capture output into a file if init can't mount the HD r/w. At least I got that solved--but somehow nothing's being written to /var/log/messages.

I'll have to limp along with rc_interactive="YES".

[eccerr0r]

For people with bus connected serial ports or (ugh) management engine, serial console if you have another machine.

Yeah I noticed the scrollback disappearing. It's too bad, though the only reason why I use it is because it was in Linux since forever.
The NetBSD FB console on the pmax didn't have scrollback and only wished it because Linux supported it.

Looks like I will have to use gnu/screen more, even on console.

BTW gnu-screen scrollback is not actually scrollback, it's actually abusing the cut/paste feature... WAIT... you can cut and paste in gnu-screen? Without a mouse? (Yeah I'm a gnu-screen abuser, so many useful features, so many arcane key sequences to access them ...)
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[Leonardo.b]

[M95D]

I saw this conversation:
https://linux.gentoo.user.narkive.com/waFrRU83/gentoo-user-console-scrollback-is-back-again
There seems to be a patch available that restores console scrollback with a separate buffer for each console!
Unfortunately I can't see the patch file. Either the site doesn't store attachments or there's a problem with my browser.
Does anyone have that file? Please post it here. Thank you!

[pa4wdh]

Good to see someone is working on that.

Somehow attachments don't make it into the official archives But here's one that has it:
https://marc.info/?l=gentoo-user&m=161764274023755
Direct link to the patch itself: https://marc.info/?l=gentoo-user&m=161764274023755&q=p3
_________________
The gentoo way of bringing peace to the world:
USE="-war" emerge --newuse @world

My shared code repository: https://code.pa4wdh.nl.eu.org
Music, Free as in Freedom: https://www.jamendo.com

[CaptainBlood]

+1
Planning 2 test it this W.E.

Thks 4 ur attention, interest & support.
_________________
USE="-* ..." in /etc/portage/make.conf here, i.e. a countermeasure to portage implicit braces, belt & diaper paradigm
LT: "I've been doing a passable imitation of the Fontana di Trevi, except my medium is mucus. Sooo much mucus. "

[JustAnother]

[M95D]

Patch works!
Thank you and thanks to the original patch author!

[M95D]

The patch fails with kernel 5.13, but there is a new patch for 5.14 here:
https://marc.info/?l=gentoo-user&m=163250452914253&w=2

I didn't test it yet.

[figueroa]

[Zucca]

I use openrc-init instead of sysvinit.
Any guesses if it'll will be impossible to have bootlogd operating correctly with openrc-init?
_________________
..: Zucca :..

[figueroa]

[Zucca]

[M95D]

[flysideways]

I'm glad I found the recommendation for abusing screen.

[irets]

The patch does not apply anymore.
Does anyone have an updated patch?

[pietinger]

Alan Mackenzie postet a patch for 6.6.13 on our Gentoo User MailingList at 2024-01-24
_________________
https://wiki.gentoo.org/wiki/User:Pietinger