Re: kern/53311: boot netbsd -d panics

[Valery Ushakov <uwe%stderr.spb.ru@localhost>]

The following reply was made to PR kern/53311; it has been noted by GNATS.

From: Valery Ushakov <uwe%stderr.spb.ru@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/53311: boot netbsd -d panics
Date: Sun, 27 May 2018 22:13:03 +0300

 On Sun, May 27, 2018 at 16:05:28 +0300, Valery Ushakov wrote:
 
 > On Sun, May 27, 2018 at 15:41:46 +0300, Andreas Gustafsson wrote:
 > 
 > > Valery Ushakov wrote:
 > > > >  By bisection, I found that the problem started with the following
 > > > >  commit:
 > > > >  
 > > > >    2018.03.16.04.48.19 ozaki-r src/sys/arch/x86/x86/db_memrw.c 1.6
 > > > >  
 > > > >  I suggest reverting it (and its pullup).
 > > > 
 > > > This is from March 2018.  The boot -d + bt triple fault that I
 > > > reported above is on i386 with 8.99.12 from late January 2018.  So
 > > > that's probably not the whole story.
 > > 
 > > To me it looks like "boot netbsd -d panics" (immediately) and
 > > "entering bt after boot -d causes triple fault" are two separate
 > > problems
 > 
 > The timeline I see is 
 > 
 > - entering "bt" after "boot -d" causes triple fault (virtualbox guru)
 >   with late January kernel on i386
 > 
 > - "bt" is now executed automatically when entering ddb, committed on
 >   February 17
 > 
 >   http://mail-index.netbsd.org/source-changes/2018/02/17/msg092275.html
 > 
 > - "boot -d" causes panic/reset on real hardware
 
 Ok, so what I was wrong about is the automatic "bt" in newer kernels.
 That only happens on panic, not every time ddb is entered.  But
 otherwise the problem is still the same: pmap is messed up.  The
 committ by ozaki-r@ is just an innocent victim here, it walks into the
 pmap mess when it checks the breakpoint address.  Reverting that
 change makes no sense.
 
 -uwe