RE: SMP problems with Thinkpad X60s

> 3.4.6 manages to start my roottask, but it prints only gibberish on the
> console before it dies. Just for reference I also attached the output of
the
> same kernel running on a KVM-emulated 2-CPU box (which works fine).
> 

OK, let's look at the 3.4.6 version first, since there's more to see:

> Pistachio:A:sPselrattfioonrm k:d:Pb_lactufrroermnt()_:cp u[ 3==3 m>CO
NsFetIGu_p_SMbPa_sMicAsX_C[P0Um
> S fvaoiilde dP iistna cfhilioe
::/Phlomate/fjourlmi:a:_n/ssetrucp/_bebalesigc/ls(4k):a-
pi[s3t3maLcheit'os/k echreneckl /kifd b/weg > luaree/v
4c-oxm8pl6/etpre.ep..ost[.0cm
> c, line 410 (fn=f012002a)
> void PAsisstearcthiiono: :kPdlb_actfurorremn:t:__sceptu u=p=_
bCasOiNcFsI(G)_S: MP[_3MA3Xm_DOCNPUE.S f[0ami
> lemd emi n 0f: i[le0 x/0h0o00me0/0j00u,l i0anx0/0s0rc0/00b0e0l)e
gt/lyp4eka=-0pxi4 s(taSchahireo/d)k erNonenl-V/ikdrbt/uagllu
> em/ve4m -1x8:6 [/p0xre0p0o0s0t00.c0c0,,  l0xin0e00 49f100 00()fn
t=fy0pe1=200x012 a()
> Co--- KD# unknown reason 2 ip f0115e6b ---
> > showtcb

That looks to me like a failed assertion printed together with startup code
of your rootserver. The gibberish probably comes from several cpus printing
stuff on the screen simultaneously. Since L4 synchronizes console output
printed via L4_KDB_Printchar and friends, there's two options: i) you're
using your own print routines and those get messed up with L4's debug
output, or ii) the L4 ASSERTION fails somewhere in the KDB, and the printf
synchronization fails as a fall-out.

Anyway, some questions:

What's at IP f0115e6b? What does the stack of idle say? Can you try to
enabling in the tracebuffer and use it to find out what happened before the
assertion failed? (The tracebuffer submenu is under 'y' in the KDB, you can
e.g., print a history of in-kernel events). Feel free to send me a binary
(including GDB debug symbols), maybe I can have a look at it. 

-Jan

--
Jan Stoess
System Architecture Group
University of Karlsruhe
Phone: +49 (721) 608-4056
Fax: +49 (721) 608-7664
eMail: [email protected]