Screen corruption on nvidia
55 views
Skip to first unread message
johny...@sigaint.org
Aug 15, 2016, 8:10:42 AM8/15/16
to qubes...@googlegroups.com
I realize that nVidia's aren't the preferred video card, but (being
divorce-poor) one sometimes has to make do with what one has. :)
With my on-board nVidia (GeForce7100) and the nouveau driver (on both
Tails and Qubes), things work okay, then suddenly at some random point the
screen gets filled with diagonal line garbage, and things are
unresponsive. (The system seems to still be running, but the screen is
pooched and keyboard seems unresponsive, and no Alt-F1'ing or whatever
will get it back. Processes still seem to be running though, and the
ethernet light is blinking normally as traffic goes in an out, etc.)
So I put in a PCI card, nVidia Geforce 7300, which seems solid, works
perfectly under Tails.
However, under Qubes, I experience random screen corruption.
See: https://i.imgur.com/ovEFgYO.png
It's usually fewer horizontal lines than in that snap, but I thought I'd
include an extreme example to show the problem.
It usually happens when the system is a bit busier CPU-wise/memory-wise,
but otherwise functioning fine. Moving the window, or switching to
another window and back, clears up the corruption.
Screen corruption like that could be a sign of some wild pointer or other
memory management bug, so I thought I'd bring it up.
The paranoid half of me (okay, okay, I'll admit, it's more than half)
worries it might be some attempted DMA attack.
(I've been, and continue to be, the subject of some rather high-end
professional hacking over the years, which adds to the concern.)
I've read about attacks where screen memory is used to stuff code and run
it to escalate privileges. And the nature of the garbage doesn't look
like mis-placed screen stuff, but code or other binary data. Thankfully,
if this is the case, the corruption seems contained to the AppVM. The
garbage never spills outside an AppVm's window. It happens in both
Redhat-23 and Debian-8 based AppVMs. I've never seen evidence of it in
dom0/Qubes Manager and such.
It realize it's *probably* not an attack, but I'd like to track down the
nature of the problem whether it's a bug or an attack.
I've read in another thread about video corruption caused by accessing
memory who page has moved, and "echo 0
>/proc/sys/vm/compact_unevictable_allowed" supposedly helps. At first, I
thought that made a difference, but ultimately, the problem is still
there.
Any ideas on what I can try, short of giving up and hunting down an ATI
card or the like? (I might try the proprietary nVidia drivers when I get
a chance to compile them. Not comfortable with grabbing binary versions
for obvious reasons.)
Thanks. You guys rock!
divorce-poor) one sometimes has to make do with what one has. :)
With my on-board nVidia (GeForce7100) and the nouveau driver (on both
Tails and Qubes), things work okay, then suddenly at some random point the
screen gets filled with diagonal line garbage, and things are
unresponsive. (The system seems to still be running, but the screen is
pooched and keyboard seems unresponsive, and no Alt-F1'ing or whatever
will get it back. Processes still seem to be running though, and the
ethernet light is blinking normally as traffic goes in an out, etc.)
So I put in a PCI card, nVidia Geforce 7300, which seems solid, works
perfectly under Tails.
However, under Qubes, I experience random screen corruption.
See: https://i.imgur.com/ovEFgYO.png
It's usually fewer horizontal lines than in that snap, but I thought I'd
include an extreme example to show the problem.
It usually happens when the system is a bit busier CPU-wise/memory-wise,
but otherwise functioning fine. Moving the window, or switching to
another window and back, clears up the corruption.
Screen corruption like that could be a sign of some wild pointer or other
memory management bug, so I thought I'd bring it up.
The paranoid half of me (okay, okay, I'll admit, it's more than half)
worries it might be some attempted DMA attack.
(I've been, and continue to be, the subject of some rather high-end
professional hacking over the years, which adds to the concern.)
I've read about attacks where screen memory is used to stuff code and run
it to escalate privileges. And the nature of the garbage doesn't look
like mis-placed screen stuff, but code or other binary data. Thankfully,
if this is the case, the corruption seems contained to the AppVM. The
garbage never spills outside an AppVm's window. It happens in both
Redhat-23 and Debian-8 based AppVMs. I've never seen evidence of it in
dom0/Qubes Manager and such.
It realize it's *probably* not an attack, but I'd like to track down the
nature of the problem whether it's a bug or an attack.
I've read in another thread about video corruption caused by accessing
memory who page has moved, and "echo 0
>/proc/sys/vm/compact_unevictable_allowed" supposedly helps. At first, I
thought that made a difference, but ultimately, the problem is still
there.
Any ideas on what I can try, short of giving up and hunting down an ATI
card or the like? (I might try the proprietary nVidia drivers when I get
a chance to compile them. Not comfortable with grabbing binary versions
for obvious reasons.)
Thanks. You guys rock!
Andrew David Wong
Aug 15, 2016, 9:20:00 PM8/15/16
to johny...@sigaint.org, qubes...@googlegroups.com
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Looks like it could be this issue:
https://github.com/QubesOS/qubes-issues/issues/1028
As you can see from the qubes-builder-github comments, some patches for this
are already in the testing repos. You may want to give those a try.
- --
Andrew David Wong (Axon)
Community Manager, Qubes OS
https://www.qubes-os.org
-----BEGIN PGP SIGNATURE-----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=gpn+
-----END PGP SIGNATURE-----
Hash: SHA512
https://github.com/QubesOS/qubes-issues/issues/1028
As you can see from the qubes-builder-github comments, some patches for this
are already in the testing repos. You may want to give those a try.
- --
Andrew David Wong (Axon)
Community Manager, Qubes OS
https://www.qubes-os.org
-----BEGIN PGP SIGNATURE-----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=gpn+
-----END PGP SIGNATURE-----
johny...@sigaint.org
Aug 16, 2016, 9:28:03 AM8/16/16
to Andrew David Wong, johny...@sigaint.org, qubes...@googlegroups.com
>> However, under Qubes, I experience random screen corruption.
>>
>> See: https://i.imgur.com/ovEFgYO.png
>>
>> See: https://i.imgur.com/ovEFgYO.png
> Looks like it could be this issue:
>
> https://github.com/QubesOS/qubes-issues/issues/1028
>
> As you can see from the qubes-builder-github comments, some patches for
> this
> are already in the testing repos. You may want to give those a try.
Awesome! Will give that a shot. Thanks for the reply! :)
>
> https://github.com/QubesOS/qubes-issues/issues/1028
>
> As you can see from the qubes-builder-github comments, some patches for
> this
> are already in the testing repos. You may want to give those a try.
johny...@sigaint.org
Aug 18, 2016, 3:57:02 PM8/18/16
to qubes...@googlegroups.com
This problem persists in 3.2rc2.
JJ
> --
> You received this message because you are subscribed to the Google
Groups "qubes-users" group.
> To unsubscribe from this group and stop receiving emails from it, send
an email to qubes-users...@googlegroups.com.
> To post to this group, send email to qubes...@googlegroups.com. To
view this discussion on the web visit
> https://groups.google.com/d/msgid/qubes-users/2380dfd0340bdf4025341dff82056852.webmail%40localhost.
For more options, visit https://groups.google.com/d/optout.
>
JJ
> You received this message because you are subscribed to the Google
Groups "qubes-users" group.
> To unsubscribe from this group and stop receiving emails from it, send
an email to qubes-users...@googlegroups.com.
> To post to this group, send email to qubes...@googlegroups.com. To
view this discussion on the web visit
> https://groups.google.com/d/msgid/qubes-users/2380dfd0340bdf4025341dff82056852.webmail%40localhost.
For more options, visit https://groups.google.com/d/optout.
>
johny...@sigaint.org
Aug 19, 2016, 7:21:22 AM8/19/16
to johny...@sigaint.org, qubes...@googlegroups.com
>>>> However, under Qubes, I experience random screen corruption.
>>>>
>>>> See: https://i.imgur.com/ovEFgYO.png
> This problem persists in 3.2rc2.
>
> JJ
Actually, just FYI, the behavior seems to be a lot better under 3.2rc2.
>>>>
>>>> See: https://i.imgur.com/ovEFgYO.png
> This problem persists in 3.2rc2.
>
> JJ
I've only seen it a couple of times, versus seeing it consistently under
3.1.
Andrew David Wong
Aug 19, 2016, 2:17:11 PM8/19/16
to johny...@sigaint.org, qubes...@googlegroups.com
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Several packages were recently pushed to testing repos (see
Hash: SHA512
qubes-buider-github comments on the issue). Have you had a chance to try those?
P.S. - Please keep the list CCed and try not to top-post.
- --
Andrew David Wong (Axon)
Community Manager, Qubes OS
https://www.qubes-os.org
-----BEGIN PGP SIGNATURE-----
iQIcBAEBCgAGBQJXt00bAAoJENtN07w5UDAwsmMP/1sY/IiuOY8smMwFNWJlekd7
Andrew David Wong (Axon)
Community Manager, Qubes OS
https://www.qubes-os.org
-----BEGIN PGP SIGNATURE-----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=mf2l
-----END PGP SIGNATURE-----
johny...@sigaint.org
Aug 19, 2016, 10:35:19 PM8/19/16
to johny...@sigaint.org, a...@qubes-os.org, qubes...@googlegroups.com
>> Several packages were recently pushed to testing repos (see
>> qubes-buider-github comments on the issue). Have you had a chance to try
>> those?
>
> Cool, I will grab the latest qubes-gui-vm from current-testing and see if
>> qubes-buider-github comments on the issue). Have you had a chance to try
>> those?
>
> that helps.
Sorry, that was phrased wrong, and I hate to add any confusion for anyone.
(I really should read more carefully before posting.)
I see the updated packages are for qubes-gui-agent's in the fedora/debian
templates. Will grab those, fire up several AppVM's, and see if things
improve.
Cheers.
JJ
johny...@sigaint.org
Aug 20, 2016, 8:23:19 AM8/20/16
to qubes...@googlegroups.com, a...@qubes-os.org
> I see the updated packages are for qubes-gui-agent's in the fedora/debian
> templates. Will grab those, fire up several AppVM's, and see if things
> improve.
Added testing repos to (clones of) debian-23 and debian-8 templates (as
> templates. Will grab those, fire up several AppVM's, and see if things
> improve.
well as whonix-gw/whonix-ws), did upgrades/dist-updates, restarted, loaded
up a bunch of AppVM's, and have been pounding on things awhile.
No sign of screen garbage yet! :)
Looks promising.
Cheers.
JJ
johny...@sigaint.org
Aug 22, 2016, 11:19:17 AM8/22/16
to johny...@sigaint.org, qubes...@googlegroups.com, a...@qubes-os.org
> Added testing repos to (clones of) debian-23 and debian-8 templates (as
> well as whonix-gw/whonix-ws), did upgrades/dist-updates, restarted, loaded
> up a bunch of AppVM's, and have been pounding on things awhile.
>
> No sign of screen garbage yet! :)
>
> Looks promising.
Day 3 of banging on r3.2-testing, and zero sign of any screen corruption.
> well as whonix-gw/whonix-ws), did upgrades/dist-updates, restarted, loaded
> up a bunch of AppVM's, and have been pounding on things awhile.
>
> No sign of screen garbage yet! :)
>
> Looks promising.
I think it's safe to mark this one as fixed.
JJ
Andrew David Wong
Aug 22, 2016, 11:53:11 AM8/22/16
to johny...@sigaint.org, qubes...@googlegroups.com
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Hash: SHA512
- --
Andrew David Wong (Axon)
Community Manager, Qubes OS
https://www.qubes-os.org
-----BEGIN PGP SIGNATURE-----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=xHjg
-----END PGP SIGNATURE-----
Reply all
Reply to author
Forward
0 new messages