Re: Two DOMU's hang on amd64 NetBSD 5.0_STABLE (XEN3_DOM0)

[Makoto Fujiwara <makoto%ki.nu@localhost>]

Thanks for the response, I really appreciate,
Below is still the same condition, 
   DOM0 = amd64 5.0_STABLE
   DOMU 
      (1) i386 5.0_STABLE
      (2) i386 5.99.27

(1) when dom0 hangs, can you still break into ddb?
I don't exactly know, but no response to keyboard, mouse
and the network. so answer may be NO.

(2) disk or file
I have converted the second DOMU to file,
--------------
#ernel  = "/export/xen/kernel/head/netbsd-XEN3PAE_DOMU.gz"
kernel  = "/export/xen/kernel/head/netbsd-INSTALL_XEN3PAE_DOMU.gz"
pae     = 1
memory  = 512
name    = "netbsd-HEAD"
#isk    = [     'phy:/dev/wd1g,0x1,w']
disk    = [     'file:/disk-1g/diskg-emul-2G,0x1,w' ]
vif     = [ 'mac=00:16:3e:02:00:37,bridge=bridge0' ]
root    = "xbd0"
--------------
but this does not seem to resolve the problem. 

(3) block access or raw manipulation:
It seems to me at-least raw access has a problem.
When sysinst starts newfs, it definitely hung.

(4) Is really disk access ?
The guess that disk access is the problem, may not be true.

The INSTALL (memory disk) kernel DOMU first started. 
And getting shell, then ready to check if dd access affects or
not, say type:
   dd if=/dev/rxbd0a of=/dev/null count=4
(not type 'Return' at the moment),

And the second DOMU, which is 5.99.27, is xm created.
I don't remember exactly, but not so far after 'xm create' typed,
It hungs. It is on amd64 Xorg Window, and everything freezes.
Screen stays visible, but mouse doesn't respond.
I did not memorize the screen, sorry.

(5) As you said it was not reproducable, and here always hungs,
and the point is not really fixed. I presume this is Hardware
related. My machine is ML115, and it has some curious behavior
not related to xen.  PS/2 port and interrupts are the most
suspicious point. I don't have non ML115 machine (of amd64)
handy for the moment, but I will look for another machine to
reproduce.

(6) One thing to add, I have checked the disk partition
(for the overlap etc). and I think it is OK. I have checked with
spreadsheel (OpenOffice scalc).

Here is the disklabel:
-------
16 partitions:
#        size    offset     fstype [fsize bsize cpg/sgs]
 a:  62516097        63     4.2BSD   1024  8192     0  # (Cyl.      0*-  62019)
 c: 312581745        63     unused      0     0        # (Cyl.      0*- 310100)
 d: 312581808         0     unused      0     0        # (Cyl.      0 - 310100)
 e:  62516160  62516160     4.2BSD      0     0     0  # (Cyl.  62020 - 124039)
 f:  62516160 125032320     4.2BSD      0     0     0  # (Cyl. 124040 - 186059)
 g:  31258080 187548480     4.2BSD      0     0     0  # (Cyl. 186060 - 217069)
 h:  31258080 218806560     4.2BSD      0     0     0  # (Cyl. 217070 - 248079)
 i:  31258080 250064640     4.2BSD      0     0     0  # (Cyl. 248080 - 279089)
 j:  31259088 281322720     4.2BSD      0     0     0  # (Cyl. 279090 - 310100)
-------
g: and h: are for DOMU for the moment.

Thanks a lot,

---
Makoto Fujiwara,