linux-kernel - x86_64, Haswell, unhandled vga irqs before i915 loaded

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-next>] [day] [month] [year] [list]

Message-ID: <5736DFFD.9030507@mail.ru>
Date:	Sat, 14 May 2016 11:21:17 +0300
From:	Nikolai Zhubr <zhubr@...l.ru>
To:	linux-kernel@...r.kernel.org
Subject: x86_64, Haswell, unhandled vga irqs before i915 loaded

Hello all,

Some of my motherboards exhibit the [in]famous bug of throwing unwanted 
(and unhandled) irqs from integrated intel video core on monitor cable 
plug/unplug events when there is no specific chipset driver loaded 
(yet). Searching some forums yields this apparently started happening 
(sometimes?) starting from Sandy Bridge, affecting both analog VGA 
output and digital (DVI, HDMI) outputs.

Currently, I observe this with 4.1.13 kernel (I could also test some 
later one but I suppose nothing relevant has changed since). 
Unfortunately the warning message from the kernel in this particular 
case is extremally unhelpfull and confusing, because the irq line in 
question usually happen to be shared with some usb and/or netcard etc., 
so kernel keeps talking about usb and/or netcard, so people end up being 
totally confused.

How to reproduce:
- get some relevant motherboard (Mine is gigabyte ga-b85-hd3 rev 2.1)
- ensure i915.ko (or any fb drivers) not automatically loaded on boot.
- watch dmesg and /proc/interupts
- start plugging/unplugging VGA or HDMI monitor cable.

What you will see then (approxmately):

/proc/interupts:
...
16:   100001       0      0     IR-IO-APIC-fasteoi   ehci_hcd:usb5
...
(100000 counter increase in a fraction of second)

dmesg:
irq 16: nobody cared (try booting with the "irqpoll" option)
...
Disabling IRQ #16

Now meanwhile freebsd people have apparently managed to fix the problem 
with a one-liner patch, but I can not test it yet because adapting this 
patch for linux is a bit beyond my capability.
This is the discussion:
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=156596
This is the patch:
https://svnweb.freebsd.org/base?view=revision&revision=284012
This is the commit log:
Disable VGA PCI interrupts until a chipset driver is loaded for VGA
   PCI devices. Else unhandled display adapter interrupts might freeze
   the CPU or consume a lot of CPU.

My understanding is that the problem is caused by buggy BIOS rather than 
the kernel, but e.g. in my case no usefull BIOS update is available 
(yet?), and still I'd like to "fix" it anyhow.

*** Please CC me, I'm not subscribed!

Thank you,

Regards,
Nikolai