linux-kernel - Re: 4.13-rc7: WARNING at arch/x86/kvm/mmu.c:717 (and a crash thereafter)

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

Message-ID: <s5hinh7lton.wl-tiwai@suse.de>
Date:   Mon, 28 Aug 2017 15:15:36 +0200
From:   Takashi Iwai <tiwai@...e.de>
To:     Adam Borowski <kilobyte@...band.pl>
Cc:     Paolo Bonzini <pbonzini@...hat.com>, kvm@...r.kernel.org,
        x86@...nel.org, linux-kernel@...r.kernel.org
Subject: Re: 4.13-rc7: WARNING at arch/x86/kvm/mmu.c:717 (and a crash thereafter)

On Mon, 28 Aug 2017 15:06:00 +0200,
Adam Borowski wrote:
> 
> On Mon, Aug 28, 2017 at 02:26:06PM +0200, Takashi Iwai wrote:
> > I seem to get a kernel warning when running KVM on Dell desktop with
> > IvyBridge like below.  As you can see, a bad page BUG is triggered
> > after that, too.  The problem is not triggered always, but it happens
> > occasionally.
> 
> See the thread starting with 20170820231302.s732zclznrqxwr46@...band.pl
> 
> > I haven't seen this on 4.13-rc4 at all, and IIRC, it started happening
> > since rc5.  So this might be a regression at rc5.  But, as it doesn't
> > happen always, I can't be 100% sure about it, and it's quite difficult
> > to bisect (the test case isn't reliable), unfortunately.
> 
> Same here -- it sometimes takes a few hours of trying to reproduce, which
> makes proving the negative greatly unpleasant.
> 
> And all I've been able to tell so far is that the problem is between
> 4.13-rc4 and 4.13-rc5, just like you say.

Good to hear that we can chorus!
So if it's really a regression between rc4 and rc5, I see no obvious
changes in arch/x86, i.e. it's likely somewhere else.

(snip)
> The first WARN is always the above.  But the rest seems to be totally random
> -- a nasty case of fandango on core whose results range from harmless
> through crash to massive data loss (just guess what would happen if some
> idiot picked balancing the disk as a test load -- no one would be that
> stupid, right?  At least an incomplete idiot has checksums and backups).

Yeah, the crash after the WARNING seems quite random.


thanks,

Takashi