Bug 194112 - [panic] Fatal trap 12: Page fault while in vm86 mode with kern.vty=vt console, hw.vga.textmode=1 and NVIDIA card
Summary: [panic] Fatal trap 12: Page fault while in vm86 mode with kern.vty=vt console...
Status: Closed Overcome By Events
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: 10.1-STABLE
Hardware: i386 Any
: --- Affects Only Me
Assignee: Mark Linimon
URL:
Keywords: vt
Depends on:
Blocks:
 
Reported: 2014-10-03 09:45 UTC by Juan Ramón Molina Menor
Modified: 2015-07-10 15:48 UTC (History)
1 user (show)

See Also:


Attachments
Screenshoot of panic with vt, textmode and NVIDIA (151.34 KB, image/jpeg)
2014-10-03 15:29 UTC, Juan Ramón Molina Menor
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Juan Ramón Molina Menor 2014-10-03 09:45:10 UTC
I’m getting a consistent panic with a NVIDIA GeForce FX 5600 XT (NV31) AGP card when enabling both kern.vty=vt and hw.vga.textmode=1 in /boot/loader.conf. FreeBSD 10.1-BETA3.

I wanted to test the VT console, enabled kern.vty=vt, but the graphics mode was unusable due to the buffer being drawn out of the screen (that’s another problem). I then added hw.vga.textmode=1 and kernel panicked.

Unfortunately, I’m not getting any core dumps despite having dumpdev="AUTO" in /etc/rc.conf and a 3 GB swap device enabled (actual memory is 1.5 GB). I’ve not found any erros related to savecore.

Here is some relevant text, copied by hand, from the panic:

vga0: <VGA GENERIC ISA VGA> ...

Fatal trap 12: Page fault while in vm86 mode
...
fault code: user read, page not present
...

I’m willing to test further if any developer is interested.
Comment 1 Ed Maste freebsd_committer 2014-10-03 15:03:35 UTC
Can you please add the backtrace if available (or just attach a picture of the full panic output)?
Comment 2 Juan Ramón Molina Menor 2014-10-03 15:29:51 UTC
Created attachment 147937 [details]
Screenshoot of panic with vt, textmode and NVIDIA

Sure, attached.

I don’t understand why it’s not dumping core… :(
Comment 3 Ed Maste freebsd_committer 2014-10-27 16:53:08 UTC
> I don’t understand why it’s not dumping core...

It happens too early for the coredump, unfortunately.  The kernel core infrastructure relies on the system getting to multiuser and executing /etc/rc.d/dumpon in order to set the dump device.
Comment 4 Juan Ramón Molina Menor 2014-10-27 19:08:58 UTC
> It happens too early for the coredump, unfortunately.  The kernel core
> infrastructure relies on the system getting to multiuser and executing
> /etc/rc.d/dumpon in order to set the dump device.

Thanks for the explanation, some new knowledge… :)

I hope you’ll resolve this issue soon. I’m ready to test.
Comment 5 Ed Maste freebsd_committer 2014-10-29 14:07:40 UTC
> It happens too early for the coredump, unfortunately.  The kernel core
> infrastructure relies on the system getting to multiuser and executing
> /etc/rc.d/dumpon in order to set the dump device.

Oh, see dumpon(8) - you can set the dumpdev variable from the loader to set the dumpdev in early boot.
Comment 6 Juan Ramón Molina Menor 2014-11-05 14:04:37 UTC
Updated to 10.1.-RC4-p1. I set kern.vty=vt at loader time and it boots without issues but panics at shut down with another Fatal trap 12. I can provide the core dump if useful.
Comment 7 Juan Ramón Molina Menor 2014-11-05 20:27:16 UTC
For completeness, please note that the panic is still there at boot if hw.vga.textmode=1. I just discovered that the panic seems to be "postponed" until shut down time if this option is not set.

Hope it helps!
Comment 8 Marcus von Appen freebsd_committer freebsd_triage 2015-02-18 11:54:21 UTC
Updated 10.1-BETA and 10.1-RC versioned bugs to 10.1-STABLE.
Comment 9 Glen Barber freebsd_committer 2015-07-08 13:24:28 UTC
Can you please test 10.2-PRERELEASE to see if this problem persists?
Comment 10 Juan Ramón Molina Menor 2015-07-09 10:12:08 UTC
(In reply to Glen Barber from comment #9)

Thanks for your interest, but I think you should close this PR: The legacy x11/nvidia-driver-173 port, required by this old AGP card, is forbidden since December 2014 because of vulnerabilities. Compiling manually is not working with current sources and I’m unable to debug those errors.

On the other hand, booting the 10.2-PRERELEASE memstick and setting kern.vty=vt and hw.vga.textmode=1 at loader time works with the built-in vga driver.

Thanks again for your efforts!
Juan
Comment 11 Mark Linimon freebsd_committer freebsd_triage 2015-07-10 15:48:16 UTC
Submitter notes that the only native driver supporting this card is marked as FORBIDDEN and suggests to close this.