Solaris 10u5 hanging early in Jumpstart x86 install
Andrew Deason
whatever I do, it keeps hanging rather earliy in the process. Here's
what I see with -v (attempting to preserve long lines):
krtld: Unused kernel arguments: `install dhcp'.
module /platform/i86pc/kernel/unix: text at [0xfe800000, 0xfe882caf] data at 0xfec00000
module misc/krtld: text at [0xfe882cb0, 0xfe8915e7] data at 0xfec413b0
module /kernel/genunix: text at [0xfe8915e8, 0xfe9b330f] data at 0xfec46a20
SunOS Release 5.10 Version Generic_127128-11 32-bit
Copyright 1983-2008 Sun Microsystems, Inc. All rights reserved.
Use is subject to license terms.
features: 1003fdf<cpuid,sse,sep,pat,cx8,pae,mca,mmx,cmov,pge,mtrr,msr,tsc,lgpg>
Using default device instance data
And here it just hangs. I find this odd because the same box installed
fine with u4. In fact, if I replace the u5 x86.miniroot with the u4
x86.miniroot (leaving everything else the same), the install process
successfully starts; but obviously fails later on.
Although I didn't think this could be a problem with the tftp server
(since the miniroot is already transferred by this point, aiui), I tried
this with a few different tftp servers, both Linux and Solaris, with the
same result.
Any ideas on where to go from here?
--
Andrew Deason
adea...@uiuc.edu
Andrew Gabriel
Andrew Deason <adea...@uiuc.edu> writes:
> I'm trying to install Solaris 10u5 on an xSeries 330 via Jumpstart, but
Ah, that looks like a PIII system.
> whatever I do, it keeps hanging rather earliy in the process. Here's
> what I see with -v (attempting to preserve long lines):
>
> krtld: Unused kernel arguments: `install dhcp'.
> module /platform/i86pc/kernel/unix: text at [0xfe800000, 0xfe882caf] data at 0xfec00000
> module misc/krtld: text at [0xfe882cb0, 0xfe8915e7] data at 0xfec413b0
> module /kernel/genunix: text at [0xfe8915e8, 0xfe9b330f] data at 0xfec46a20
> SunOS Release 5.10 Version Generic_127128-11 32-bit
> Copyright 1983-2008 Sun Microsystems, Inc. All rights reserved.
> Use is subject to license terms.
> features: 1003fdf<cpuid,sse,sep,pat,cx8,pae,mca,mmx,cmov,pge,mtrr,msr,tsc,lgpg>
> Using default device instance data
>
> And here it just hangs. I find this odd because the same box installed
> fine with u4. In fact, if I replace the u5 x86.miniroot with the u4
> x86.miniroot (leaving everything else the same), the install process
> successfully starts; but obviously fails later on.
Try this...
Boot kadb:
Type 'e' on the relevant grub menu choice
Select the kernel line and type 'e'
append -k -d [return]
Type 'b' to boot kadb.
Patch cmi_no_init to 1:
cmi_no_init/W1
:c
If that works, make the change permanent once you've installed
by putting the following in /etc/system:
set cpu\.generic:gcpu_legacy_cpu_support=1
If that doesn't work, then put this in /etc/system instead:
set cmi_no_init=1
--
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]
Andrew Deason
and...@cucumber.demon.co.uk (Andrew Gabriel) wrote:
> In article <20080614195038.e...@uiuc.edu>,
> Andrew Deason <adea...@uiuc.edu> writes:
> > I'm trying to install Solaris 10u5 on an xSeries 330 via Jumpstart,
> > but
>
> Ah, that looks like a PIII system.
Your eyes do not deceive you. Dual PIII 1GHz.
> Try this...
>
[set cmi_no_init to 1]
>
> If that works,
It does.
> make the change permanent once you've installed by putting the
> following in /etc/system:
>
> set cpu\.generic:gcpu_legacy_cpu_support=1
Putting this in the x86.miniroot also works. If I enable this for all
boxen (just for the installation boot), even those that don't need it,
is there any downside? Enabling this all the time would make things
slightly easier to automate.
And many thanks! I see you've answered this before; apologies, but this
isn't an easy problem to search for until I knew the solution.
--
Andrew Deason
adea...@uiuc.edu
Dick Hoogendijk
> On 15 Jun 2008 08:18:25 GMT
> and...@cucumber.demon.co.uk (Andrew Gabriel) wrote:
>> Try this...
>>
> [set cmi_no_init to 1]
>>
>> If that works,
>
> It does.
>
>> make the change permanent once you've installed by putting the
>> following in /etc/system:
>>
>> set cpu\.generic:gcpu_legacy_cpu_support=1
> isn't an easy problem to search for until I knew the solution.
What I would like to know is if SUN has given up on PIII processors
officially. Patch 127128 (kernel) had issues, but he kernel I downloaded
today (a far newer one) still has the same issues.
Without the cpu.generic in /etc/system it would not boot.
--
Dick Hoogendijk -- PGP/GnuPG key: 01D2433D
++ http://nagual.nl/ | SunOS 10u5 05/08 ++
Andrew Gabriel
Dick Hoogendijk <di...@nagual.nl> writes:
>
> What I would like to know is if SUN has given up on PIII processors
> officially.
No.
I believe the official support line is any Pentium except those
with the FDIV bug (the workaround for that was pulled out in S10,
as I believe it had meant retaining the 387 emulation code).
I have some old Pentium systems which I used to test S10 on, but
they can't take more than 128Mb memory, so grub boot was the end
of them as far as Solaris is concerned. I suspect the problem is
finding old systems to test on.
John D Groenveld
Andrew Gabriel <and...@cucumber.demon.co.uk> wrote:
>of them as far as Solaris is concerned. I suspect the problem is
>finding old systems to test on.
The bug isn't triggered on Sun's Pentium3-based LX50 or Xeon-based
V60 and V65 systems?
John
groe...@acm.org
Dick Hoogendijk
I have the feeling you already know the answer to that question ;-)
Solaris used to start on my very old PIII. It still does, now that I
have changed the /etc/system file with the known workaround.
But I do feel this should not no longer be needed.
There is after all a new kernel patch available.