Bug 40902

Summary:	sata_nv driver leaves BIOS in strange state on reboot
Product:	IO/Storage	Reporter:	David Krider (david)
Component:	Serial ATA	Assignee:	Jeff Garzik (jgarzik)
Status:	RESOLVED OBSOLETE
Severity:	normal	CC:	alan
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	2.6.38	Subsystem:
Regression:	Yes	Bisected commit-id:

Description David Krider 2011-08-10 20:30:16 UTC

When rebooting from Linux, either to go into Windows or back into Linux, my computer will get to a point in the loading process, and then spontaneously reboot. In either OS, it's about the time to mount drives and get serious about loading the rest of the system. If I halt, and then power the computer back up, the BIOS will start to do it's thing, the keyboard lights will come on, the monitor power button briefly stops flashing, and then the BIOS itself reboots. When it finally gets back up, it tells me that the last boot failed.

This wasn't a problem for me in 2.6.36. I've downgraded and confirmed the problem goes away. It never does this when rebooting from Windows. I'm convinced it's not my hardware.

Twice, now, I've had to deal with a scrambled /home directory (fixed by a fsck), which has never happened to me before in the ~15 years I've run Linux. I'm nervous that I'm going to lose data here. I use a fakeraid to access a stripe on which I have both Windows and Linux loaded, and I suspect the sata_nv driver, but I have no way to catch the problem to try to diagnose if that's the problem at all, and, if it is, where it jumps the rails.

Can anyone suggest a way to capture the problem? Maybe it's not storage at all, but the dirty /home problems are the only thing I have to go on right now.

Comment 1 David Krider 2011-09-01 22:38:20 UTC

Well, it's not related to fakeraid. I bought an SSD, and disabled all the RAID stuff in the BIOS, and it's still doing it.

Comment 2 David Krider 2011-09-03 19:13:15 UTC

I'm on Ubuntu, but I've downloaded all the vanilla kernels from 2.6.36.4 through 2.6.37.6 to 2.6.38, built them by starting with Ubuntu's config (but taking out "paravirtualized guest support" because that always seems to fail for me), and then running `fakeroot make-kpkg --initrd kernel_image kernel_headers'. The problem starts on 2.6.38. I'm looking over the patch, and wondering if I can break out the various patches by driver subdirectories...

Comment 3 Alan 2013-12-23 14:16:13 UTC

Closing as obsolete, sorry nobody could figure this one out