FreeBSD 10-STABLE/sparc64 panic

[Chris Ross]

tl;dr : I’ve finished my testing and have a result, but see other things I
don’t understand.  Could use more help.

On Jun 29, 2014, at 22:46, Chris Ross <cross+freebsd at distal.com> wrote:
> On Jun 20, 2014, at 23:44 , Chris Ross <cross+freebsd at distal.com> wrote:
>> On Jun 16, 2014, at 14:46 , Chris Ross <cross+freebsd at distal.com> wrote:
>>> On Jun 9, 2014, at 16:18, Chris Ross <cross+freebsd at distal.com> wrote:
>>>> On Jun 9, 2014, at 14:13, Chris Ross <cross+freebsd at distal.com> wrote:
>>>> 
>>>> Oh well.  That was learned quickly.  r263478 booted once, but a second
>>>> attempt caused the multiple-boots-before-getting-to-multiuser.  So, moving
>>>> right along, I’ll try r263401 in the more “binary search” algorithm…  
>>> 
>>> [...] so I’m going to try r263407.  Mostly zfs changes from Illumos, but.
>> 
>> r263407 showed no problems.  I just rebooted a few times more and see
>> no issues.  I worry I'm chasing ghosts, but am going to try r263470 next.
> 
> I ran r263470 for a week or so, with quite a few reboots (7, it appears).
> Never tried more than once to boot successfully.  Next, I'm going to confirm
> I still get crashes with a fresh build of r263478.  But, there's only about 1000
> lines of diff between those two revisions, much of it changes to routing and
> related networking code.  So, could certainly be it.

  Easily confirmed.  The first attempt to boot an r263478 stable/10 kernel
panic’d, and there were 10 more boot-and-panic attempts before it
successfully booted.

http://svnweb.freebsd.org/base?view=revision&sortby=date&revision=263478

  I can try manually reverting each of the specified revisions (262763, 262767,
262771, and 262806) from the code, but I’d feel better having someone with
deeper knowledge of the networking code and/or sparc64 MD architecture
take a closer look at it from here.  Anyone available?

  Kurt, I know you said you saw this on head back in March:

http://lists.freebsd.org/pipermail/freebsd-sparc64/2014-March/009261.html

  ..which looks like it was before (r262669) the things listed above as
MFC’d.  And, I see another email from you in 2013 that looks like the same 
issue in head r257208.

http://lists.freebsd.org/pipermail/freebsd-sparc64/2013-October/009085.html

  So, I’m a bit confused.  It looks like this goes back much further than my
testing had revealed.  But, maybe the things MFC’d in stable/10 r263478
either (a) included bits that were also in other revisions, or (b) provoked
the same problem provoked earlier in head, just in a different way/place.

  I’m at a bit of a loss for what to do now, but would really like to see this
get resolved.  I don’t like the fact that numerous of us on the list have
seem this problem, and are having difficulty running machines (v240
only?) because of it.

  If there’s anything else I can do, I have a machine that appears to be
behaving semi-consistently with these two revisions of stable/10.  I’m
happy to help further.

                               - Chris