This is an absolute FIASCO !

[Michael Hohmuth]

Hi Sed!

roux at ecoledoc.lip6.fr (Cedric ROUX) writes:

> linux boots, everything seems alright, then it loads the deamons, then
> the keymap, then the login should be there, but it reboots.

> That was the first try.
> I tried it just after this boot. And i now have this :
> It boots, then at a time (don't remember exactly when), it shows that :
> thread.cc:304: failed assertion 't->kernel_sp==0 || reinterpret_cast<vm_offset_t>(t->kernel_op) > reinterpret_cast<vm_offset_t>(t)+sizeof(thread_t)+0x20'
> then it's over, asking me to reboot or run the debugger.
> 
> So my questions are :
> - what does it mean ?

This is a panic because a kernel invariant wasn't met.  In this case,
the kernel's thread switching code detected a stack overrun in the
target thread.  This may also have been the reason for the
crash/reboot you've seen.

Are you trying to run the X Window System at this point?  If so, this
is a know problem.  You should boot into a runlevel which doesn't run
X11 to try out Fiasco.

> - is it a problem in my compiling of l4-linux ? (i disabled serial support)

I don't think that your L4Linux server is at fault.  

> I have a second machine, so i could have tried the debugger, but some pieces
> are missing and i don't have a null-modem, but i'll try to solve that.

It would be great if you could generate a backtrace or even a state
analysis for the thread whose stack space runs out.

> Hope this is not an already known bug.

No, it's not (unless you're running X11).

> (if someone could send me his kernel configuration, you run make xconfig and
> you save it and you send it to me, of course if it works !)

My .config file is appended below.

> (don't know if it had to be done, but it checked my hd, and during
> this process i had a lot of 'wake_idle_if_needed : ipc failed')

This message should be removed: It's just informational.  (The reason
it doesn't bug me as much as you is that I have redirected it to a
second monitor (Hercules, using the "-hercules") option.)

> PS : i have the idea to do a distributed memory managment for my
> studies, do you think it's possible with fiasco ? 

Yes, I think it's possible, and I think this would be an interesting
application for Fiasco.

> and when you run under l4-linux, can you make process that directly
> talk to fiasco ?

Yes, that's both possible and easy to do.  However, we would need to
come up with some documentation, and it would help if long IPC
messages were available (I'm in the process of implementing them right
now) so that our local name server could be utilized.

Michael
-- 
hohmuth at innocent.com, hohmuth at inf.tu-dresden.de
http://home.pages.de/~hohmuth/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: .config
Type: application/octet-stream
Size: 5068 bytes
Desc: not available
URL: <http://os.inf.tu-dresden.de/pipermail/l4-hackers/attachments/19981201/a7f05530/attachment-0001.obj>