I am trying Mark's changes on hardware. Here is my hardware plan
1) Run on js21 without the nohv boot argument.
2) Run on a js21 with the nohv boot argument.
3) run on a g5.
I am not using the js20 because I don't want to deal with the reversion
to hil problem.
On the js21 that I use there is the console peculiarity. So in addition
to Mark's patches, I run with a hard-coded hardware address for the
serial port. (hardcoded as in, boot_of has this assigment
@@ -796,6 +796,7 @@ static int __init boot_of_serial(void *o
of_panic("%s: no location for serial port\n", __func__);
}
ns16550.io_base = val[1];
+ ns16550.io_base = 0x2f8;
ns16550.baud = BAUD_AUTO;
ns16550.data_bits = 8;
)
(It does not get more hardcoded than this. I am being very specific
here in case someone needs to replicate this.) I should add that on a
js21 with the above change and a small change to linux configuration I
can up to a shell prompt.
Here are pieces of the console output:
OF: Xen/PPC version 3.0-unstable (butrico@localdomain) (gcc version
3.4.4) Wed Jun 28 13:16:47 EDT 2006
boot_of_init args: 0x0 0x0 0x111027c 0x0 0x0
boot msr: 0x1000000000003000
boot_of_init: _start 0000000003000100 _end 0000000003775a38 0x0
boot_of_probemem: memory 0x0000000000000000[0x80000000]
boot_of_probemem: memory 0x0000000100000000[0x80000000]
bootargs =
Notice no bootargs
(XEN) *** Serial input -> DOM0 (type 'CTRL-a' three times to switch
input to Xen).
OF stdout device is: /xen/console
command line:
Notice no args to linux
802.1Q VLAN Support v1.8 Ben Greear <greearb@xxxxxxxxxxxxxxx>
All bugs added by David S. Miller <davem@xxxxxxxxxx>
Root-NFS: No NFS server available, giving up.
VFS: Unable to mount root fs via NFS, trying floppy.
VFS: Cannot open root device "<NULL>" or unknown-block(2,0)
Please append a correct "root=" boot option
Kernel panic - not syncing: VFS: Unable to mount root fs on
unknown-block(2,0)
<0>Rebooting in 180 seconds..
My next step was to supply linux with arguments, by adding them to the
linux .config file. Again pieces of the console output:
OF: Xen/PPC version 3.0-unstable (butrico@localdomain) (gcc version
3.4.4) Wed Jun 28 13:16:47 EDT 2006
boot_of_init args: 0x0 0x0 0x111027c 0x0 0x0
boot msr: 0x1000000000003000
boot_of_init: _start 0000000003000100 _end 0000000003775a38 0x0
boot_of_probemem: memory 0x0000000000000000[0x80000000]
boot_of_probemem: memory 0x0000000100000000[0x80000000]
bootargs =
again no arguments to xen.
OF stdout device is: /xen/console
command line: console=ttyS1,19200 ro root=/dev/nfs
nfsroot=9.2.208.86:/home/kitchawa/linux.img/ppc64nfs-2005-06-18
ip=9.2.129.20::9.2.129.2:255.255.255.0:kpblade7:eth1:off
init=/sbin/quickinit noshell
arguments to linux for its root fs and network. (I am not too sure
about the console argument. I would like to build linux so it could
boot alone, i. e., without xen, but so far this has not worked for me.
I get no console output past prom_init).
802.1Q VLAN Support v1.8 Ben Greear <greearb@xxxxxxxxxxxxxxx>
All bugs added by David S. Miller <davem@xxxxxxxxxx>
IP-Config: Complete:
device=eth1, addr=9.2.129.20, mask=255.255.255.0, gw=9.2.129.2,
host=kpblade7, domain=, nis-domain=(none),
bootserver=255.255.255.255, rootserver=9.2.208.86, rootpath=
Looking up port of RPC 100003/2 on 9.2.208.86
tg3: eth1: Link is up at 1000 Mbps, full duplex.
tg3: eth1: Flow control is off for TX and off for RX.
Looking up port of RPC 100005/1 on 9.2.208.86
VFS: Mounted root (nfs filesystem) readonly.
Freeing unused kernel memory: 232k freed
Sun Jan 1 01:01:00 EST 2006
28 Jun 14:30:12 ntpdate[835]: step time server 9.2.1.64 offset
15424151.601078 sec
Starting system log daemon: syslogd.
Starting the hotplug events dispatcher: udevd.
Synthesizing the initial hotplug events...done.
Waiting for /dev to be fully populated...(XEN) h_enter: PTEG FULL
(XEN) h_enter: PTEG FULL
(XEN) h_enter: PTEG FULL
.....
I get about 5 pages of this. Then
oom-killer: gfp_mask=0x201d2, order=0
Call Trace:
[C000000001E87690] [C000000000010728] .show_stack+0x50/0x1cc (unreliable)
[C000000001E87740] [C00000000008ABD4] .out_of_memory+0x220/0x254
[C000000001E87800] [C00000000008C400] .__alloc_pages+0x324/0x348
[C000000001E878E0] [C00000000008F980] .__do_page_cache_readahead+0x130/0x378
[C000000001E87A50] [C0000000000871AC] .filemap_nopage+0x138/0x4ec
[C000000001E87B20] [C00000000009AA30] .__handle_mm_fault+0x658/0xff8
[C000000001E87C20] [C00000000003109C] .do_page_fault+0x5bc/0x84c
[C000000001E87E30] [C0000000000048DC] .handle_page_fault+0x20/0x54
Then more bad things happens and it all ends with many messages from
linux about being out of memory.
So 1 is not working, if we actually bring up linux. Then I tried 2.
One might wonder why I might go after 2 given than 1 is broken. The
code takes different paths if the nohv flag is on and it is possible
that the problems with see without the nohv flag be not there when we
use the nohv code. Specifying nohv the system stops (no output) at
(XEN) *** LOADING DOMAIN 0 ***
(XEN) xen_start_info: 0000000007ffe000
(X
Mark F Mergen wrote:
OK, Linux is running in problem state on Xen to the same point that
versions from xenbits.xensource.com run in my environment, using
Mambo, namely "Unable to mount root fs via NFS". It's time for a RAM
disk root fs. I'll be looking for a little help with this to make
fast progress. The DEC virtualization is partially done, but we need
to really engage on what HV-related hardware actually works on the Mac
G5 and what doesn't, to do this properly and to do whatever else is
needed to run on Mac G5. I have a few loose ends to tie up in HPT
eviction, hcall destroy VM area, and hcall flush HPTE, which I am
working.
The patches below represent the status described above.
Mark Mergen
------------------------------------------------------------------------
_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ppc-devel
_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ppc-devel
|
Home