mmmmn

版权声明：转载时请以超链接形式标明文章原始出处和作者信息及本声明
http://mmmmn.blogbus.com/logs/1889591.html

1、> I've been tweaking around Xorg but find no solution to this problem. The
> fact is the graphical performance is VERY poor, I mean, you can see how
> bitmaps are drawn in some programs, the cursor gets jerky sometimes (like
> it was refreshed at 4 fps), and some other annoying slowdowns.
>
> The relevant info about the system I have is:
> -SiS chipset based motherboard (supporting 400 MHz FSB CPUs and SDRAM)
> -ATI Radeon 9600 128 MB RAM AGP 8x
> -Athlon XP 3000+ Barton (400 MHz DDR FSB, 512 KB 2L cache)
> -1 GB 400 Mhz DDR-SDRAM
> -Xorg configured with both automatic configuration (without
> /etc/X11/xorg.conf) and with xorgconfig. Tested with both "ati" module and
> "radeon" one. Tested with a working Linux config (with the necessary
> changes) and touching the xorgconfig one.
>
> Not necessary to say it only happens in Solaris 10 (3/05 & 1/06 tested),
> but not in Windows nor Linux.
>
> ¿ Any clue ?
> Many thanks in advance,
> --
> Mik

This probably does not help your problem , but as a data point , Im using a
ATI Radeon 9200 SE 128M PCI Express
at 1920x1600x60, and it works fast on X86

Section "Device"
Identifier "Card2"
Driver "ati"
VendorName "ATI Technologies Inc"
BoardName "RV280 [Radeon 9200 SE]"
BusID "PCI:10:4:0"
EndSection

2、Well, my trusty old Sparc20/Solaris9 blew up a few weeks ago. Now I'm

working
on moving all of my services over to zones on Solaris10, with much success
and a lot of learning. However, I've run into a silly little problem with YP
and the automounter, and need some spare eyes to help me out here.

I have home directories coming off of one server (salvador), shared out as
/export/home. I have /home set up as an indirect map in auto_master, in NIS.
auto_home has the generic line:

* salvador:/export/home/&

This is all absolutely bog-standard stuff. The thing is, this works if I
have /etc/auto_home local, but NOT if I use the NIS auto.home table. The
syntax between the two is identical. I've played around with nsswitch.conf
and toyed with the +auto_home line in /etc/auto_home (as well as removing
/etc/auto_home entirely) but the end result is that /home/<username> is only
accessible if the above line appears in /etc/auto_home.

Stranger still--the entry MUST be in the local /etc/auto_home, even if
nsswitch.conf indicates that automount should only ever look in NIS.

Any ideas? I'm sure I'm missing something obvious, but I can't figure out
what it is for the life of me.

Sorry, now ideas, but it looks like I've been observing the same (or at least

very similar) weirdness here with Solaris Express snv_28:

[root@snaut] /etc 268 > grep auto nsswitch.conf
automount:  files nis

(files is required here 'cause on this host the local auto_home map should

take precedence over the NIS one)

[root@snaut] /etc 269 > ls -1 auto_*
auto_home
auto_shared

[root@snaut] /etc 270 > ypcat -k auto.master
/shared  auto_shared  -nobrowse
/home    auto_home    -nobrowse
/net     -hosts       -nosuid,nobrowse

(map auto_master happily comes from NIS and everything's OK)

[root@snaut] /etc 271 > ypcat -k auto.shared
* obender:/export/shared/&

Now, this does not work as it should: for whatever reason, NIS map

auto.shared is not used UNLESS I also have this as a local map:

[root@snaut] /etc 272 > cat auto_shared
+auto_shared

Of course, the same effect gives just making auto_shared a local map (file).

It certainly worked as expected in Solaris 8 and (maybe) 9: normally I don't

have any /etc/auto_* files at all, and nsswitch.conf always defines sequence

'files nis' just in case a map needs to be locally overridden. All maps used

to be found in NIS, but not anymore, now it looks like it suddenly became

'files [notfound=return] nis' instead of default 'files [notfound=continue]

nis'.

3、There was a recent thread about how to configure FQDN under DHCP in:
http://groups.google.com/group/comp.unix.solaris/browse_frm/thread/57920dc237

3e9d02/8e11d9a1753d388d?lnk=st&q=author%3Ak_dal2%40hotmail.com&rnum=5&hl=en#8

e11d9a1753d388d
(or search for "hostname with DHCP" in comp.unix.solaris if the link
above does not work)

4、I have USPIII single CPU system controllers with some running Solaris
>8, and other 9, that
>are locked up when I apply the real-time scheduling class to several
>threads in my processes. A process running at a higher priority in the
>RT scheduling class, say a vmstat, will not run either. I have a stack
>trace of the threads in the process that I believe has caused the
>system lockup.

A real time thread which spins will not allow any other forward
progress in the system. So locking up the system is certainly
a possibility. Note that RT priority is higher than the system so
starting new processes or sending I/O through a network daemon/X server
or xterminal will not be possible. (If everything in the path to
the higher priority RT process also high priority RT)

Casper

5、> Just installed a fresh copy of Solaris 10 01/06 build with all
> packages+OEM and I noticed it installed both Apache v1.3 and v2.0.
>
> Just curious why they didn't just standardize on one with the initial
> install and then offer the other as part of the Companion CD, etc...

Apache 1.3 is the default. The Apache 2 on Solaris 10
is by default disabled. Why would anybody want to have
to add packages from a CD when
# svcadm enable apache2
is so much easier?

6、take a look at pkginfo(4). There you can find a description of 3 new
Parameter.
SUNW_PACKAGE_ALLZONES is the on of interest for you.

HTH,

Michael

P wrote:
> Hi,
>
> I've installed a package by using pkgadd -G and now I want to
> uninstall something on a Solaris 10 system with a sub zone. I looked
> into Sun's documentation for pkgrm and it said I could use -G just as
> I could with pkgadd. However, when I did a man on pkgrm, -G is not
> listed. I came across a post that says -G isn't implemented yet.
>
> When I just use pkgrm, it'll display the following:
>
> ## Waiting for up to <300> seconds for package administration commands
> to become available (another user is administering packages on zone
> <ember>)
>
> (ember is the subzone)
>
> How do I get around this problem?
>
> Thanks

7、> However, we are not an all Solaris 10 shop; we have 2.6, 2.8, one or
> two SunOS 4.3 machines, a couple of AIX 4.3.3 machines, and a bunch of
> NeXT OS servers. I'm not worried about the NeXT OS machines, as
> migrating off these to SOlaris 10 is precicely the project I'm working
> on with the developer.
>
> I was trying to get the compiler bootstrapped to then use it to create
> compilers for these other platforms as Daniel suggested (is this what
> is termed a cross-compiler?). Not only that, but I will eventually be
> required to download other SW to compile for the various platforms, so
> I thought that this would be as good a time as any to figure it out.

For a cross compiler you will need much more. You will need a basic set of
libraries and header files of the target platform. For a cross compiler
you also need binutils.

For building a cross compiler Solaris/x86 -> Solaris/SPARC you need the
following.

Suppose you want to build a cross compiler for

target=sparc-sun-solaris2.8
prefix=/opt/cross
archdir=/opt/cross/sparc

(1) extract the needed headers/libraries/object files for the target

platform:
Use a Solaris 8 CD and extract the files from the following packages
to $archdir:
. SUNWarc (/usr/lib/values-*.o)
. SUNWcsu (/usr/lib/ld.so.1)
. SUNWcsl SUNWcslx (/usr/lib/*.so)
. SUNWhea (/usr/include)
. SUNWlibm (/usr/include/math.h)
. SUNWlibms, SUNWlmsx (Shared Math Library)
for X applications
. SUNWxwplt (/usr/openwin/lib)
. SUNWxwinc (/usr/openwin/include)
. SUNWmfrun (/usr/dt/{include,lib})

extract with:
cd $archdir
bzip2 -dc $PKGDIR/archive/none.bz2 | cpio -id

(2) build binutils
./configure --target=$target --prefix=$prefix
gmake; gmake install

(3) add $prefix/bin to your PATH

(4) build gcc
.../gcc-<version>/configure --prefix=$prefix --target=$target \
--with-sysroot=$archdir
gmake; gmake install

The above description may be incomplete or contain errors but these are the
the basic steps to build a cross compiler. Other architectures will differ
(building a cross compiler Solaris/SPARC -> SINIX/Z was a real nightmare)
-----
> Joe D. wrote:
>> Hello gurus;
>>
>> I am trying to install GCC (latest and greatest downloaded from GNU) on
>> a Solaris 10 server. I have searched the GNU.org and Google groups for
>> several days. I understand that I need to 'bootstrap' the compiler in
>> order to get it to actually BE a compiler. I do already have a compiler
>> on my test server (SFWgcc) to bootstrap the GNU download with.
>>
> A few things to consider.
>

And a few more...

> 1) gcc is probably already installed as /usr/sfw/bin/gcc on Solaris 10.
>

Yes. And have you noticed how libstdc++.la and libsupc++.la are suspicious-
ly empty (Bug ID: 6220191)? Enjoy libtool not working with them. This
issue might be gone with the HW 06/01 release. Haven't checked.

The next thing you might want to figure out using the SFW gcc is how to
properly encode a runpath if you decide to build a 64bit binary. Sun came
up with the idea to modify lib/gcc/sparc-sun-solaris2.10/3.4.3/specs
to always include /usr/sfw/lib in the RUNPATH of the binary. Always in
this case actually means - always:

$ /usr/sfw/bin/gcc hello.c -m64 -o hello
$ dump -Lv hello
...
[4] RUNPATH /usr/sfw/lib
[5] RPATH /usr/sfw/lib
...

Under the right circumstances the dynamic linker now will go ahead and
link a 32bit library to your 64bit program. This won't work too well of
course:

$ /usr/sfw/bin/gcc hello.c -lssl -m64 -o hello
$ dump -Lv hello
[1] NEEDED libssl.so.0.9.7
...
[5] RUNPATH /usr/sfw/lib
[6] RPATH /usr/sfw/lib
...
$ ./hello
ld.so.1: hello: fatal: /usr/sfw/lib/libssl.so.0.9.7: wrong ELF
class: ELFCLASS32
Killed

I wonder why '-z ld32='/'-z ld64=' (see ld(1)) were not used in the
specs file. OTOH this matches other instances of missing attention
to "minor" details in the SFW packages pretty well.

> 2) Sun Studio 11 is a better compiler, although it must be admitted some
> open-source software is only tested with gcc, so can in some
> circumstances be less hassle to build with gcc. Sun Studio 11 is a free
> download.
>

Why does funroll-loops.org invariably come to my mind if anyone talks
about how much better the code of the Studio compiler is. Yes, there are
situations in which a 10 to 20% increase in speed will matter and will
be worth real money. In most cases software isn't CPU bound though and
small code optimizations will yield no noticeable speedup whatsoever.
Under such circumstances "less hassle to build with gcc' rests the case
in favour of gcc. At least it does for me.

Helmut

--
Almost everything in life is easier to get into than out of.

8、The clock on my Ultra 2 workstation suddenly stopped keeping time. I'm
assuming that the lithium battery on the motherboard needs replacing,
but according to the Sun website this is not a "customer replaceable"
part. Has anyone here replaced one of these batteries? Are they readily
available and could they be replaced by someone with decent technical
skills?

Any advice would be much appreciated.

Sounds like your NVRAM battery died. Try searching for
"sun nvram faq" on google for information how to replace
the nvram with a new part or how to "refurbish" the one
already in the machine.

Regards,

Frank
Excellent information. Thanks!

Chris

9、> My plan was to partition mirror boot drive on "server b" to
> match our "server a" and ufsdump/ufsrestore over the network to the
> mounted mirror boot partitions on "server b". Once I had mirror boot
> disk bootable, I would boot from it and dd from one disk to the other.
>
> Once system restore was complete, I modified the vfstab on mounted
> mirror drive, /mnt/etc/vfstab, to point to correct disk and installed
> bootblk. I tried to boot from disk1 and the system failed to boot and
> panicked rebooting from primary boot disk. I have done some
> investigation and I think it has to do with the meta devices.
>
> boot message:
>
> Cannot open mirrored root device, error 19
> Cannot remount root on /pseudo/md@0:0,0,blk fstype ufs
>
> panic[cpu1]/thread=180e000: vfs_mountroot: cannot remount root
>
> Any help appreciated,
>
> Mike Dundas

If your description of what took place is complete I would say
that the failure is due to that you did NOT create a new set
of METADEVICE databases on server b.

There is at least three things that must be present for a successful
volumemanager boot:

/etc/system file must preload the MD devicedrivers and
the pointer to the metadevice rootfilesystem.

/etc/vfstab must reference the metadevices.

A valid and correct metadabase must exist that translates
the metadevice root-referens from /etc/system to a physical
disc/partition number.

The error seems to say that the /etc/system file that you copied
over from server A with its root metadevice pointer could not
find a metadb on server B to resolve it to a physical partition.
--
 >> My plan was to partition mirror boot drive on "server b" to
> >> match our "server a" and ufsdump/ufsrestore over the network to the
> >> mounted mirror boot partitions on "server b". Once I had mirror boot
> >> disk bootable, I would boot from it and dd from one disk to the other.
> >>
> >> Once system restore was complete, I modified the vfstab on mounted
> >> mirror drive, /mnt/etc/vfstab, to point to correct disk and installed
> >> bootblk. I tried to boot from disk1 and the system failed to boot and
> >> panicked rebooting from primary boot disk. I have done some
> >> investigation and I think it has to do with the meta devices.
> >>
> >> boot message:
> >>
> >> Cannot open mirrored root device, error 19
> >> Cannot remount root on /pseudo/md@0:0,0,blk fstype ufs
> >>
> >> panic[cpu1]/thread=180e000: vfs_mountroot: cannot remount root
> >>
> >> Any help appreciated,
> >>
> >> Mike Dundas
> >
> > If your description of what took place is complete I would say
> > that the failure is due to that you did NOT create a new set
> > of METADEVICE databases on server b.
> >
> >
> > There is at least three things that must be present for a successful
> > volumemanager boot:
> >
> > /etc/system file must preload the MD devicedrivers and
> > the pointer to the metadevice rootfilesystem.
> >
> > /etc/vfstab must reference the metadevices.
> >
> > A valid and correct metadabase must exist that translates
> > the metadevice root-referens from /etc/system to a physical
> > disc/partition number.
> >
> > The error seems to say that the /etc/system file that you copied
> > over from server A with its root metadevice pointer could not
> > find a metadb on server B to resolve it to a physical partition.
> >
> > //Lars
> >
>
> I did "not" create a new set of devices, that is correct. I did modify the
> newly
> restored vfstab to reference the /dev/dsk/cXXXXXXX etc, and commented out
> the metadevice references. The reason for this was I am unfamiliar with
> metadevices
> and most of my systems have a boot drive and a mirror drive that I manually
> create/update
> every month with a script using the dd command.
>
> Must I therefore modify the /etc/system file. Is this possible by hand?
>
> What is the best practice when trying to restore this box. As it is right
> now I am trying to
> boot from the mirror drive on a server "B". Should I install Sol 10 copying
> partitions to
> match server "A", then use volume manager to create the mirror.
>
> The vendor who supplied the system wasn't much help. They had said to
> backup server "A"
> and restore to server "B". Unfortunatley, our netbackup software isn't
> compatable with
> Solaris 10 yet and no standalone DLT's avail anymore. I was using
> ufsdump/restore across network
> but I am having to restore to a mounted mirror drive. This seems rather
> convoluted!
>
>
> thanks for the advice can you now please answer my new questions?
>
> Thanks
>
> Mike Dundas

Mike,

I Understnad then that you need some sort of crash course in volume
manager
to be able to get this system back online Here it is as good as I can
make it.
READ All of this before you try to do it :-)

To recapitulate you replaced the c0t0d0 disk drive and was force
to restore
the c0t1d0 drive because the root partition was corrupted.
( this was a very bad series of events that led to both mirrors being
destroyed at the same time )

so to get this system back at all you reinstalled solaris on c0t0d0
and started
a restore of c0t1d0 by copying filesystems from an identical server .

Question now is in what state are your metadevice databases after all
this
has taken place.
I think you should Login to you working server A and run

metadb -i

and

metastat

this will show you what the config is supposed to look like.

Try to verify if the metadb parttions shown by metadb still exists
on
server B, try to run metadb -i on server B to check if there is a
remnant
of the previous metadb configuration left on the system.
If metadb dosent work at this point on server B use
prtvtoc /dev/rdsk/cXtXdXs2 to list the partion info on each disk.
It is likely that you have lost the metadb partitions on you two
system
disks at this point as you have reinstalled them and that they still
exist on
you two data disk.
Probably the easiest thing to to at this point is to erase the old
metadb
information that may still exist on server B

metadb -d -f devicename ( for all of them )

After you have removed the remainng metadb partitions I suggest that
you do exactly as you tried to do from the beginnig.
Copy the System partitions from server A to the second disc
on server B. Also create the metadb partition on disk2
Install the bootstrap on the second disk.
Then Edit /etc/vfstab and /etc/system so that all info regarding
Volume manager is deleted from these files ( i.e. a Vanilla /etc/system

and /etc/vfstab )
Then verify that you can boot standalone on server B's second disk.
Next copy the VTOC of the second disk to the first systemdisk

Probably: ( you actual devicenames may vary )

prtvtoc /dev/rdsk/c0t1d0s2 > /disk2.vtoc
fmthard -s /disk2.vtoc /dev/rdsk/c0t0d0s2

Now the Partions layouts should be identical.

nOW YOU NEED TO RESTABLISH THE metadb PARTITIONS

Read the metadb manual page carefully.

then probably

metadb -a -f c0t0d0sX c0t1d0sX C0t2d0sX c0t3d0sX

metadb -i should now show four valid metadb devices

then you need to create one way mirrors for all the active partitions
on the
systemdisk an maybe the datadisks to if these are to be mirrored.

READ the example section of the metainit(1M) man page carefully
You need to to this with the devicenames from you SECOND disk.

metainit -f d12 1 1 c0t1d0s0 ( for root )
metainit -f d22 1 1 c0t1d0s1 ( for SWAP ????)
metainit -f d32 1 1 c0t1d0s3 ( for /var ????)
..
..
metainit -d d10 -m d12 ( root's oneway mirror )
metainit -d d20 -m d22 ( SWAP oneway mirror )
metainit -d d30 -m d32 ( /var's oneway mirror )

A.S.O. for all you system partitions

then set the root device this uppdates /etc/system

metaroot d10

Check that /etc/system has been updated.

Manually update /etc/vfstab with the above entered metadevice names.
so that it looks similar to server A.

run metastat and check that everything seems O.K:

Shutdown and restart the server with disk2 as the boot disk.

Login and start the mirroring
setup the first disks devices

metainit -f d11 1 1 c0t0d0s0 ( for root )
metainit -f d21 1 1 c0t0d0s1 ( for SWAP ????)
metainit -f d31 1 1 c0t0d0s3 ( for /var ????)
..
..
metattach d10 d11 ( root's twoway mirror )
metattach d20 d21 ( SWAP twoway mirror )
metattach d30 d31 ( /var's twoway mirror )

Then do the same thing for you datadisks if nessesary.
Decide which of the datadisks is going to be treated as the good copy
and work from there.

If this is to heavy stuff for you, you will need to get SUN or some
other consultant
to do it.

Regards //Lars

.
----
I've been building servers for the past month by mirroring a drive,
breaking the mirror and then booting off the cloned disk. Every time
there was a boot panic, I had had either removed the rootdev line from
/etc/system and had changed the vfstab... or vice versa, changed
vfstab, but forget to remove teh rootdev line. Once I got these two in
sync, it was hunky dory.

Also don't forget, if you break the mirror and run off the physcial
slice, remove the meta-objects. They will still exist in the the
/etc/lvm/md.tab.

SVS (SDS) is your friend... and the manual is a rather short read.

---
> metattach d10 d11 ( root's twoway mirror )
> metattach d20 d21 ( SWAP twoway mirror )
> metattach d30 d31 ( /var's twoway mirror )
>
>
> Then do the same thing for you datadisks if nessesary.
> Decide which of the datadisks is going to be treated as the good copy
> and work from there.
>
> If this is to heavy stuff for you, you will need to get SUN or some
> other consultant
> to do it.
>
> Regards //Lars

commenting out the line in the /etc/system allowed me to come closer to
booting, but I still got a ton of errors.

without giving a 100 line copy of the output I will give a shortened
version and if you need more details I can provide them.

here is output from boot:

Rebooting with command: boot disk1
Boot device: /pci@1c,600000/scsi@2/disk@1,0 File and args:
SunOS Release 5.10 Version Generic 64-bit
Copyright 1983-2005 Sun Microsystems, Inc. All rights reserved.
Use is subject to license terms.
Hardware watchdog enabled
/kernel/drv/sparcv9/ip: undefined symbol 'MD5Init'
/kernel/drv/sparcv9/ip: undefined symbol 'MD5Final'
/kernel/drv/sparcv9/ip: undefined symbol 'MD5Update'
WARNING: mod_load: cannot load module 'ip'
WARNING: ip: unable to resolve dependency, module 'misc/md5' not found
/kernel/fs/sparcv9/sockfs: undefined symbol 'sctp_disconnect'
/kernel/fs/sparcv9/sockfs: undefined symbol 'sctp_getsockname'
/kernel/fs/sparcv9/sockfs: undefined symbol 'sctp_create'
/kernel/fs/sparcv9/sockfs: undefined symbol 'sctp_close'
/kernel/fs/sparcv9/sockfs: undefined symbol 'sctp_listen'
..
..
..WARNING: ip: unable to resolve dependency, module 'misc/md5' not found
/kernel/drv/sparcv9/tcp: undefined symbol 'tcpinfo'
WARNING: mod_load: cannot load module 'tcp'
WARNING: tcp: unable to resolve dependency, module 'drv/ip' not found
/kernel/drv/sparcv9/ipsecah: undefined symbol 'ip_drop_packet'
/kernel/drv/sparcv9/ipsecah: undefined symbol 'ip_drop_register'
/kernel/drv/sparcv9/ipsecah: undefined symbol 'ip_hdr_length_nexthdr_v6
..
..
/kernel/drv/sparcv9/ipsecesp: undefined symbol 'ipsec_outbound_sa'
WARNING: mod_load: cannot load module 'ipsecesp'
WARNING: ipsecesp: unable to resolve dependency, module 'drv/ip' not
found
Failed to plumb IPv4 interface(s): bge0
Dec 30 05:38:17 svc.startd[7]: svc:/network/physical:default: Method
"/lib/svc/m
ethod/net-physical" failed with exit status 96.
[ network/physical:default misconfigured (see 'svcs -x' for details) ]
Hostname: scpIIIb
The / file system (/dev/rdsk/c1t1d0s0) is being checked.
Loading smf(5) service descriptions: 1/118
SC Alert: SC Request to XIR Host due to Watchdog
XIR failed - performing
SC Alert: Host System has Reset
SPOR
XIR/Watchdog Reset
Executing Power On Self Test
0>
0>@(#) Sun Fire[TM] V210/V240,Netra[TM] 240 POST 4.16.2 2004/10/04
22:43

/export/common-source/firmware_re/post/post-build-4.16.2/Fiesta/enxs/inte
grated (firmware_re)
0>Copyright © 2004 Sun Microsystems, Inc. All rights reserved
SUN PROPRIETARY/CONFIDENTIAL.
Use is subject to license terms.
0>OBP->POST Call with %o0=00000800.01012000.
0>Diag level set to MIN.

after this it boots to disk0 (solaris 9, primary boot device)
successfully.

I tar'ed/untar'ed /kernel -> mounted <mirror_disk>/kernel again and
tried another reboot, but still got same errors.

Hopefully this will give you guys some clues.

Thanks for your help so far!

Mike Dundas

-----
> I tar'ed/untar'ed /kernel -> mounted <mirror_disk>/kernel again and
> tried another reboot, but still got same errors.
>
> Hopefully this will give you guys some clues.
>
> Thanks for your help so far!
>
> Mike Dundas

No tthat wont work , there is dependencies between the kernel , the
device tree
and the /etc/path_to_inst file.

I think you need to do ufsdumps of your system partitions (/) ( /var
) (/usr) (/opt)
from the working Server A , transfer them to ServerB and do
ufsrestore of
the dumps to the second disk., Then clear /etc/system from
volumemanager
entries and remove md devises from /etc/vfstab , then install the
bootstrap
on the second disk .
then try to reboot.

//Lars

----
  > on the second disk .
> then try to reboot.
>
> //Lars

I will try that tonight. I did this originally, except I
ufsdump/restore from serverA boot disk to serverB mirror disk for /
(root partition) and tar for all other partitions. I did install
bootblk and changed references in vfstab to /dev/dsk instead of md.
Last night I changed the /etc/system file which allowed me "almost"
boot instead of the cpu panic and immediate reset. This is when I did
a comparison of /kernel (serverA) and /mnt/kernel (serverB) and saw the
discrepancies. This was the reason for my most recent tar attempt.

Should I be ufsrestoring from the mirror on ServerA to get the
identical config? Or is ufsrestore from serverA primary boot disk ->
serverB mirror OK?

Thanks for taking the time and offering you advice.

Mike Dundas

11、>I have a U10/Sol10, with a couple NFS shares. When I mount them with -t

nfs
>from linux, the mount command takes about 30 to 60 seconds to return for
>each FS, on a local subnet. Mounting nfs sparc/sol10 to sparc/sol10 is
>almost instantaneos.

Here I think we could need some snoop outputs. Try this on your U10

# snoop -tr -v -o /var/tmp/snoop-nfs u10 linux

and place the snoop-nfs file somewhere where we can fetch it.

>Why is this? Is this a fact of life, or is there something I can do to fix
>this.

I have never seen that behavour and I've mounted a lot of NFS ;-)
This seems fun to dig in to though

>I also posted this on a linux newsgroup, but figured people here probably
>could shed some insight too.

You have come to the right place for NFS knowledge... ;-) You can
hardly find as much
experts on NFS than here... (Me not really one of them, I just practise
my english skills)

Please also give us uname -a from both machines and the exact share and
mount commands you use.

I use to do like on my Ubuntu and it comes right back
# mount -t nfs -o rsize=8192,wsize=8182,tcp solaris:/usr/local
/usr/local
-----
The command I use to mount the share is this:
root@woody:/home/sheam # time mount /matilda

real 1m45.054s
user 0m0.002s
sys 0m0.047s

Looks like it took longer than I thought. The relevant line
from /etc/fstab is:
matilda:/export/home /matilda nfs rw 0 0

>From the solaris box:
# share
mhome /export/home rw "export home"

I started the snoop right before I executed the mount command.
After the mount finished, I copied a file to and from the share
(cp done from linux box).

> # snoop -tr -v -o /var/tmp/snoop-nfs u10 linux
> and place the snoop-nfs file somewhere where we can fetch it.
ftp://www.snowsquirrel.ca/pub/nfs_woody_mount_matilda.snoop

> Please also give us uname -a from both machines and the exact share and
> mount commands you use.
>From linux box:
Linux woody 2.6.12-10-k7-smp #1 SMP Fri Nov 18 13:05:01 UTC 2005 i686
GNU/Linux

>From Ultra 10:
SunOS matilda 5.10 Generic_118822-18 sun4u sparc SUNW,Ultra-5_10

>
> I use to do like on my Ubuntu and it comes right back
> # mount -t nfs -o rsize=8192,wsize=8182,tcp solaris:/usr/local
> /usr/local

I tried your options, but it didn't make a difference, timewise.

Thanks for the interest.

---------
>
>> So you'd want, for example:
>>
>> # mount -o nfsvers=4 sun:/foo /somedir
>
> This command tells me version 4 is not supported, and it looks like it

tries
> as v3, after. maybe I will build a kernel with v4 support tommorrow, and
> try that, if anyone thinks it may help.

Worth a try. :) You could also build the NFS stuff as a loadable module,
and then just 'modprobe nfs' as root instead of having to reboot for a
new kernel. But if you prefer to build stuff into a kernel directly,
then yeah, going to have to build+reboot.

The reason why I suggest it's worth a try for the v4 stuff is because in
case this is some kind of Linux NFS v3 client bug.

I'm lazy and didn't feel like getting to my Sun, so I looked at the
snoop output with Ethereal, which I already had on my Linux box.

(Ethereal can read snoop's packet dump file format, ditto for tcpdump's
and a few others.)

And, you're right, a long pause is *not* normal. Should be nearly
instantaneous, even Linux->Solaris.

192.168.0.2 is woody (Linux) and 192.168.0.8 is matilda (Solaris), I take it?

I read through the snoop capture with text mode Ethereal by doing:

$ tethereal -r nfs_woody_mount_matilda.snoop \!\(tcp.port == 22 or tcp.port

== 143\) > nfs.txt
$ more nfs.txt

(That excluded the ssh and IMAP traffic and made it much more readable.)

First packet from the Linux box for NFS was at 4.151757 second since
start of the dump, to matilda's rpcbind port (to do a query). Response
was immediate. Good.

The time from all traffic (15 packets total) from start to finish for
the RPC AND NFS mount request completed within 180 milliseconds. Very
good.

So where is it delaying? Easy to see:

[...]
24 4.169240 192.168.0.2 -> 192.168.0.8 Portmap V2 GETPORT Call NFS(100003)

V:3 UDP
25 4.169942 192.168.0.8 -> 192.168.0.2 Portmap V2 GETPORT Reply (Call In 24)

Port:2049
242 109.217501 192.168.0.2 -> 192.168.0.8 NFS V3 FSINFO Call, FH:0x4c88849e
243 109.218072 192.168.0.8 -> 192.168.0.2 NFS V3 FSINFO Reply (Call In 242)
344 159.323701 192.168.0.2 -> 192.168.0.8 NFS V3 GETATTR Call, FH:0x4c88849e
345 159.324172 192.168.0.8 -> 192.168.0.2 NFS V3 GETATTR Reply (Call In 344)
346 159.325585 192.168.0.2 -> 192.168.0.8 NFS V3 ACCESS Call, FH:0x4c88849e
347 159.325719 192.168.0.8 -> 192.168.0.2 NFS V3 ACCESS Reply (Call In 346)
348 159.326016 192.168.0.2 -> 192.168.0.8 NFS V3 READDIRPLUS Call,

FH:0x4c88849e
349 159.326292 192.168.0.8 -> 192.168.0.2 NFS V3 READDIRPLUS Reply (Call In

348) . .. lost+found sheam TT_DB pub postgres svn

No packets are listed between 25 and 242; this is ok and expected
because they are just the excluded ssh and IMAP traffic.

There is a very unusual 105 second pause where nothing happens at all.
VERY BAD!!

No retransmits, no NFS queries, no *nothing* between woody and matilda.
It's as if woody had decided to take a nap for some unknown reason.
matilda has been answering everything promptly.

woody then finally decides to ask matilda for filesystem information for
the mounted NFS filesystem.

Then another unusual 50 second pause after that, woody finally decides
to issue NFS GETATTR/ACCESS/READDIRPLUS requests. This is when you see
your files show up. The 50 sec delay is also VERY BAD, too!

EVERY single query or request from the client is promptly answered by
the server, and there are no retransmits for any of them, so we know
that the client saw and accepted the replies.

My tentative thought is that this problem is due to some sort of weird
bug with the Linux NFS client, somewhere. Or a misconfig of some kind on
the Linux side or missing (non-enabled) kernel bits.

Someone reported they had to start portmapper locally on the Linux box
before doing the mounts, as odd as that sounds... (Maybe a red herring,
but guess it doesn't hurt to try if you get desperate.)

Do triple check to make sure that NFS v3 is enabled in the kernel, too.

Linux NFS client is still rather buggy 'out of the box' and needs tons
of patches to make it usable. :) You may have to apply patches from:

http://client.linux-nfs.org/Linux-2.6.x/2.6.12/

I noticed you were running 2.6.12 -- you'll probably be surprised to see
how long the list of patches at that page is, and the kind of stuff they
fix. Reading over the list gives me the impression that it is quite very
possible that one or more of the patches there may fix your issue.

I'm not sure how you'd handle patching with your Linux distribution (I'm
guessing Debian or Ubuntu) or if they already have these patches
applied. Something you may need to check into further.

Just for reference, the home site of these NFS patches:

http://client.linux-nfs.org/

I have a suggestion for you. On woody (the Linux box), make sure tcpdump
is installed, then do this:

$ su -
# tcpdump -w /tmp/woody-matilda-nfs.tcp host woody and host matilda and not

port 143 and not port 22

Post the woody-matilda-nfs.tcp (or whatever filename) to your FTP server
and let us know. If you want to review it first before putting it up for
download, just do 'tcpdump -r /tmp/woody-matilda-nfs.tcp|more'.

The reason for this request is because, sometimes there is indeed a bona
fide networking issue but only visible on one host. We've seen matilda's
side of the story, but not woody's, just yet. We're fair jurists, after
all. :-)

-Dan
-----
> $ su -
> # tcpdump -w /tmp/woody-matilda-nfs.tcp host woody and host matilda and
> # not port 143 and not port 22
>
> Post the woody-matilda-nfs.tcp (or whatever filename) to your FTP server
> and let us know. If you want to review it first before putting it up for
> download, just do 'tcpdump -r /tmp/woody-matilda-nfs.tcp|more'.

ftp://snowsquirrel.ca/pub/woody-matilda-nfs.tcp

Nothing really jumped out at me. I guess I'll likely have to build a nfs4
module, and see if that helps. I'll also look into the portmapper.

---

11、> Has anyone tried adding the Sun MIBS to net-snmp ? ... or ...
> running Sun's SNMP as a subagent to net-snmp ?
>
> The net-snmp README.solaris has a section on having Sun's SNMP as a master
> and net-snmp as a subagent, but I want to do it the other way around.

Ever get an answer? I tried setting up net-snmp as a subagent, and mostly
it worked; but walking the tree didn't descend into the "host" stuff starting
from mib-2, but only if starting from host. Sure like to know if that
can be fixed, so that starting high enough could get it all in one walk.

12、I've done the same trip and I think most is covered at
http://docs.sun.com/app/docs/coll/47.16?q=solaris+10+administration&s=t
The things that are of most interest are zones/containers, dtrace,
services(SMF), RBAC (Role-Based Access control), and of course ZFS (in
SolarisExpress Nevada_b27)

Then you can google for a lot of things around Solaris 10. Then
blogs.sun.com are also
really useful I think.

http://mediacast.sun.com/share/bobn/SMF-migrate.pdf is a good start for
SMF

13、> Hello Everyone,
>
> I'm trying to install updates/patches on my solaris 9 box, where jass
> (solaris security toolkit 4.1) is already installed.
>
> I was told not to do a undo/rollback to re-run jass again, but this is
> the only way I know how to install patches with jass. I usually do a
> undo and copy the patches in /opt/SUNWjass/Patches directory and then
> re-run jass. this usually installs the patches.
>
> Any other ideas on how to install additional patches on an existing
> jass installation?

Jass installs the Recommended Patch SET Before doing the bulk of the
the security alterations.

If you now have additional patches not included in
a Recommended patch set, Jass wont know how to handle them.
unless you hack the
SUWjass/Finish/install_recommendedpatches.fin script.

You will have to check the individual README's of the patches
verify if a file in the Patch overwrites any active security config
file,
if so make a backup copy of the config file before
"patchadd" ing the patch and then restore the config file afterwards.

//Lars

.

--------
Thank Lars,

I guess there is no solution for this. I might need to install these
patches via patchadd and cross my fingers that JASS does not freak-out.

I'm surprise there is no utility for this with JASS. I tried looking
into their documentation and I was not able to find anything. The only
thing I found was add_patch function, which needs to be added to a
customized finish script. The problem with this solution is that I
will need to do a ./jass-execute -u (undo) and then ./jass-execute -d
(run jass again). This is not an option for me.

I will keep digging for the solution since I don't have much choice.
I'll keep this thread posted.
-------
> what configuration file?

Well that depends on the individual patch does'nt it.

If you install a SSH patch and it replaces /etc/ssh/sshd.conf
you better make a copy of your sshd.conf before applying the patch.

Right ?
-
thank you Lars. i can do that. it will be a pain in the butt, but i
guess i don't have a choice.

i'm looking into jass-execute options. i will try -u (undo) with -k
option (keep any manual changes i've made), on a test machine. i'll
keep you posted

13、Yes I am trying to login as root

If you want to allow that (and I'm no expert on the rights and wrongs of

that), you need to

a) Stop sshd

b) Change the line below from no to a yes.

sparrow /export/home/drkirkby % grep Root /etc/ssh/sshd_config
PermitRootLogin no

c) Restart sshd

There might be a way of getting sshd to re-read the config file without

actually stopping it. Perhaps:

# pkill -HUP sshd

will do that, but I have no idea.

You appear to be allowing telnet login as root. That is considered a

particularly bad practice. If someone has to use another account to get to

root, you can track what account they had access to.

--

14、> I just installed a zone on Solaris 10 on for the first time. I used my

U80
> for this. Naturally tried a few experiements, one of which was to executed
> "poweroff" whilst in the zone.

I take it you're using Solaris 10 3/5 or the 1/6 ?

Still, this is an odd situation as it shouldn't have the impact you

described,
it almost looks as if it didn't properly shut things down.

There should be no need to reboot the entire box, I'm wondering what happens

if
you simply try to shut down the zones process itself through svcadm.

Something
in the likes of 'svcadm disable system/zones' ?

> # zoneadm list -v
> ID NAME STATUS PATH
> 0 global running /
> 3 secondzone down /zones/myzone

The 'zones(5)' manpages describes this to a certain degree; the moment that
your zone is in the 'down' state its stuck between 'running' and 'installed',
so my guess is indeed that your poweroff had a big impact on the zone itself.

> # zlogin -S secondzone
> zlogin: login allowed only to running zones (secondzone is 'down').
> # zoneadm -z secondzone boot
> zoneadm: zone 'secondzone': zone is already booted

These are actually logical responses, even though it looks kind of weird. The
zone *is* booted, but its stuck in a shutdown sequence. It would almost seem

as
if some process is running amok.

> # zoneadm -z secondzone halt
> zoneadm: zone 'secondzone': unable to unmount '/zones/myzone/root/usr'
> zoneadm: zone 'secondzone': unable to unmount file systems in zone
> zoneadm: zone 'secondzone': unable to destroy zone

My best guess at this time would be to check if no global processes are

keeping
the zone's filesystems busy. If that isn't the case then perhaps the svcadm
command can help you out here, but it looks to me as if this is a matter of a
process running amok somewhere.

So I'd try to check global processes if something is keeping
'/zones/myzone/root/usr' busy, if you can't find any (which would be odd IMO)
then maybe you can shut down the entire zones process through svcadm.

> I assume (perhaps incorrectly) that rebooting the physical machine will

allow
> me access to the zone again, but that seems a sledehammer to crack a

wallnut.

Yes, rebooting is very likely to get the job done since it will eventually

also
kill several processes and reset your zones state to the default. But it is
indeed a little overkill.

--
I just installed a zone on Solaris 10 on for the first time. I used my U80
for this. Naturally tried a few experiements, one of which was to executed
"poweroff" whilst in the zone.

I take it you're using Solaris 10 3/5 or the 1/6 ?
No, Jan 2005.

sparrow /export/home/drkirkby % cat /etc/motd
Sun Microsystems Inc.   SunOS 5.10      Generic January 2005
sparrow /export/home/drkirkby % uname -a
SunOS sparrow 5.10 Generic_118822-23 sun4u sparc SUNW,Ultra-80

I downloaded Solaris 10 update 1 a week or two back, but for some reason the

CD would not boot. I did not spend much time on it. I was installing it on a

machine in the garage and it was too cold to mess about, so I installed the

older version.

Still, this is an odd situation as it shouldn't have the impact you

described,
it almost looks as if it didn't properly shut things down.

There should be no need to reboot the entire box, I'm wondering what happens

if
you simply try to shut down the zones process itself through svcadm.

Something
in the likes of 'svcadm disable system/zones' ?

It appeared to make no difference:

# svcadm disable system/zones
# svcadm enable system/zones
# zoneadm list -v
  ID NAME             STATUS         PATH
   0 global           running        /
   3 secondzone       down           /zones/myzone
# zoneadm -z secondzone boot
zoneadm: zone 'secondzone': zone is already booted
# zlogin -S secondzone
zlogin: login allowed only to running zones (secondzone is 'down').

So far no changes. But then I tried this, which I had not done before, which

was an oversight I guess.

# zlogin -C secondzone

[Connected to zone 'secondzone' console]

SunOS Release 5.10 Version Generic_118822-23 64-bit
Copyright 1983-2005 Sun Microsystems, Inc.  All rights reserved.
Use is subject to license terms.

Hostname: secondzone

secondzone console login: root
Password: Dec 27 23:53:20 secondzone sendmail[19235]: My unqualified host

name (secondzone) unknown; sleeping for retry
Dec 27 23:53:20 secondzone sendmail[19236]: My unqualified host name

(secondzone) unknown; sleeping for retry

Last login: Tue Dec 27 16:45:36 on console
Dec 27 23:53:24 secondzone login: ROOT LOGIN /dev/console
Sun Microsystems Inc.   SunOS 5.10      Generic January 2005

So the act of logging into the zone's console seems to have got it to boot.

Then, close that terminal and in another window:

#  zoneadm list -v
  ID NAME             STATUS         PATH
   0 global           running        /
   4 secondzone       running        /zones/myzone
#

So the zones are now running OK.

Since I'm only experimenting, repeated this, powering off again and that time

it behaved exactly as expected and would boot later.

# zoneadm -z secondzone boot
# zoneadm list -v
  ID NAME             STATUS         PATH
   0 global           running        /
   5 secondzone       running        /zones/myzone

So it seems okay now, even after another "poweroff" command. I know poweroff

is not a very clean shutdown, so perhaps that was the problem, although I am

surprised. It will not execute the shutdown scripts, that should not hang the

"OS".

Perhaps I should read more about the theory of the zones, then I'll perhaps

understand what is happening.
--
 So the zones are now running OK.

I'm glad to hear that.

> Perhaps I should read more about the theory of the zones, then I'll perhaps
> understand what is happening.

For whats it worth; this behaviour wouldn't have been prevented. I tried to
reproduce this behaviour on my blade box (setup a basic zone with only the
directory inheritances present) and ran "poweroff" on it after which is

simply
went to the "installed" state.

I have no idea what could have caused this.

--

14、Sean wrote:
Dan Foster wrote:

At the lom prompt, do this instead:

lom> poweroff

Wait 5-10 seconds.

lom> poweron

And let it sit for a few minutes. No need to physically push the power
button as you can remotely manage the system via LOM.

Do you see anything like this, anywhere?

LOM event: +0h3m52s overheat sensor 1 FATAL FAULT: overheat - shutdown req'd
LOM event: +0h3m52s host power on
LOM event: +0h3m52s Fault LED 3Hz
LOM event: +0h3m53s overheat sensor 1 no overheat

lom>
LOM event: +0h3m58s host power off

The reason why I ask is because the X1 is infamous for having a failed
temperature overheat sensor which is enclosed in the PSU. If you have a
failed temp overheat sensor, you're going to have to somehow find a new
Netra X1 PSU -- loads of fun there.

If you don't see these exact messages, then it's probably not the
infamous temp sensor overheat problem.

If you still aren't getting anywhere... then try:

<press enter key once>
<press #. keys>
lom> bootmode forth
lom> reset

...and see if you eventually get dropped into the OpenBoot PROM (OBP)'s
ok prompt.

If you get that far, you can boot off a CDROM by doing 'boot cdrom', or
off the network by doing 'boot net'.

-Dan
Dan,

Thank you for your response. I don't believe I've seen the error messages

about the PSU. I also tried the other set of instructions and I still have

got the OBP prompt. I can type the 'bootmode forth'
command but when I type 'reset' I get 'system in standby mode'

There is a SCC card inserted. When I remove it then I get the error message

that the card has been removed. Other than that, there is not much I can do

outside of the LOM. I hope this helps. Thanks for your help. This is first

Sun hardware I have ever own. Just wanted to see how it works.

Sean
A good first step would be to go to docs.sun.com and RTFM, particularly
the introductory stuff about LOM.  If you truly have the OBP prompt (not
to be confused with the LOM prompt), then you should be able to boot up
(unless the disk has been wiped, in which case you should still be able
to boot net after first setting up a boot server).

Nothing you've said gives me any reason to believe the hardware is
faulty, but I suspect you don't have an OS on the disk.

An X1 would not be my first choice of system to cut my teeth on, because
of its lack of a normal framebuffer/keyboard console.  You might do
better to sell it (to me, e.g.  :-)  ) and get a more regular desktop
type Sun like an Ultra 10, which might be similarly priced.

--
Mr. CTJ,

RTFM huh? I RTFM BEFORE I got the server. I've used Goggle. I took notes and

tried the commands, it hasn't worked. I know what the LOM prompt and the OBP

prompt look like. Which is why I'm wondering if this Netra is bad. I didn't

pay for it, I got it because someone I knew didn't know how to test it.

Seeing as how I never paid for this server it's not a total lost to me if it

doesn't work I have another Sun server to work on.

I'm only trying to install a OS on it that's why I am hoping to boot to a

CDROM. Or at least to troubleshoot the server and learn about Sun hardware.

I thought I could come to this newsgroup and ask for help. It's people like

you and you nasty comments that making this fustrating. Others have offered

advice and have emailed me directly and I thank them. You don't even know me

and you are posting smart-ass comments towards me. All I am asking for is a

little help. If it was the other way around I wouldn't be a smart ass like

you.

I don't want to start a flame war, I just want to learn about Sun hardware.

If you are want to help then I would apperciate it. If you want to be a

smartass then keep you comments to yourself. I know these newsgroups gets

flooded with newbie questions. I posted my stuff because I want help on what

to do when it doesn't work like the manual said.
-----
> RTFM huh? I RTFM BEFORE I got the server. I've used Goggle. I took
> notes and tried the commands, it hasn't worked. I know what the LOM
> prompt and the OBP prompt look like.

Obviously not, and google isn't the right place to look.
Why not use sunsolve.sun.com? Please read this:

http://www.sun.com/products-n-solutions/hardware/docs/pdf/806-5980-10.pdf

Especially chapter 6 and/or page 29. Page 52 may also be
interesting, if the previous owner seperated the console
from the lom.

15、> mnikhil@xxxxxxxxx <mnikhil@xxxxxxxxx> wrote:
>> lets say if one installs a patch on the solaris machine, and not sure
>> if that requires a reboot. how one can identify if there is such thing
>> which can tell us that there is a need of the reboot for the system ? (
>> I have seen such things in Windows atleast).
>
> Read the patch README (Section "Special Install Instructions")

On that point, I have a complaint. I've seen quite a few patches that
either claimed they needed a reboot, but probably didn't (although you
might not get the full benefit without it), or at least didn't need
it if you'd already installed a reasonably recent prior rev (say something
that affected the kernel interface for e.g. getrusage). Excessive reboots
are something that a number of folks would Really Prefer To Avoid, whether
to minimize impact on ongoing operation, or just because it too much
resembles Windows. Now some might require certain procedures, or awareness
of effects and consequences, to safely avoid a reboot, while obtaining
sensible results. Even that might require stopping and starting processes,
but it could well be quite a bit quicker, esp. on systems where the
PROM takes quite awhile to do its thing on reset, or where the total
time taken by system+installation rc scripts (or SMF methods) is already
rather long.

Other qc of README files isn't quite up to par, either. There are patches
that are mentioned as xxxxxx-yy (or later) that have in fact been obsoleted
by something else altogether (typically a kernel patch) that are listed in
the notes (i.e. to get feature A/for full fix of bug B, the following
patches are also required). For those who attempt to verify all of those
not-quite-dependencies have been met, it can be more tedious and
time-consuming than it ought to be.

16、I have learned
a bit that on UNIX platforms there are font servers used to render
fonts and font resources must be somehow prepared before first usage.
Maybe this is the point? How can I for example check whether fonts are
available in general?

You wrote in your first posting everything works fine expect xterm.
Does that includes other terminal emulators or is it just xterm
which has trouble ?

If you manage to run one of these, try:

"xset -q", will show you the fontpath, including font server if set.

"xlsfonts", will display all fonts available.

You can also try running "xfontsel" (graphic font selector).
.

-------
In article <1135542867.418790.326470@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
<grzegorz_manturzyk@xxxxxxxxxxxxxx> wrote:
>Hello Hans,
>
>Thank you for reply. As you guessed, I use Xvnc on Solaris side and
>vncviewer on Windows. I browsed FAQs and mailing list forth and back on
>VNC site and did not found any message regarding my problem. Therefore
>I suspect that it is not the issue of VNC but the operating system
>itself. I think that font resources are not configured properly, xterm
>is not able to find ANY font and simply does not start. I have learned
>a bit that on UNIX platforms there are font servers used to render
>fonts and font resources must be somehow prepared before first usage.

the font-server must not run. i did

# svcadm -v disable svc:/application/x11/xfs:default
svc:/application/x11/xfs:default disabled.

# svcs -vx svc:/application/x11/xfs:default
svc:/application/x11/xfs:default (X Window System font server)
State: disabled since Sun Dec 25 23:07:32 2005
Reason: Disabled by an administrator.
See: http://sun.com/msg/SMF-8000-05
See: man -M /usr/openwin/share/man -s 1 xfs
See: man -M /usr/openwin/share/man -s 1 fsadmin
Impact: This service is not running.

and xterm does start properly within an vnc-session.

how do you start Xvnc ?
did you install all packages from the companion-cd ?
what happens, if you start xterm on the local screen ?
is it possible to start xterm against a remote x-server without
using Xnvc ?

here my installed fonts:
# pkginfo | grep -i font
system SFWfnlib fnlib - X11 font rendering library
system SUNW1251f Russian 1251 fonts
ALE SUNW5xmft Traditional Chinese (BIG5) X Windows Platform minimum required

Fonts Package
system SUNWarrf X11 Arabic required fonts
ALE SUNWcxmft Simplified Chinese (EUC) X Windows Platform minimum Required

Fonts
system SUNWeurf Global fonts
system SUNWfdl Font Downloader
system SUNWfontconfig Fontconfig library
system SUNWfontconfig-docs Fontconfig library documentation
system SUNWfontconfig-root Fontconfig library configuration files
system SUNWfreetype2 FreeType2 Font library
GNOME2 SUNWgnome-fonts GNOME Unicode and Korean TrueType fonts
system SUNWgsfot GNU Ghostscript Fonts (Other)
system SUNWgsfst GNU Ghostscript Fonts (Standard)
system SUNWi13rf X11 ISO8859-13 required fonts
system SUNWi15rf X11 ISO8859-15 required fonts
system SUNWi1of ISO-8859-1 (Latin-1) Optional Fonts
system SUNWi2rf X11 ISO8859-2 required fonts
system SUNWi4rf X11 ISO8859-4 required fonts
system SUNWi5rf X11 ISO8859-5 required fonts
system SUNWi7rf X11 ISO8859-7 required fonts
system SUNWi8rf X11 ISO8859-8 required fonts
system SUNWi9rf X11 ISO8859-9 required fonts
system SUNWjxmft Japanese Required Bitmap Font
system SUNWkoi8f X11 KOI8-R fonts
ALE SUNWkxmft Korean (UTF-8) X Windows Platform minimum Required Fonts
system SUNWtxfnt Thai X Windows Platform required Fonts Package
system SUNWxwcft X Window System common (not required) fonts
system SUNWxwfa X Window System Font Administrator
system SUNWxwfnt X Window System platform required fonts
system SUNWxwfs X Window System Font server
system SUNWxwoft X Window System optional fonts
system SUNWxwpft X Window System Printer Fonts

i am sure, not all are needed, i.e. Thai X Windows here in austria :-)

best regards
hans

17、 Xterms are now showing a very strange behaviour. The CTRL+some key
> combinations I usually type (CTRL+e, CTRL+a mostly) are perfectly
> working, but for some reason CTRL+c is not. It sometimes print "^C"
> (when I try to break a running process), sometimes beeps (when I try to
> scrap the current line in the shell), but the processus behind never
> actually catches the signal! It's slightly annoying to say the least...

Your stty settings aren't using CTRL-c, e.g.,

stty intr \^C

For reference, "stty" shows what intr is assigned to, e.g.,

speed 38400 baud; -parity
rows = 40; columns = 80; ypixels = 0; xpixels = 0;
min = 1; time = 0;
eof = ^a; swtch = <undef>; start = <undef>; stop = <undef>; susp = <undef>;

dsusp = <undef>;
brkint -inpck -istrip -ixany imaxbel -opost
-icanon -echo -echoe -echok

----
 D. wrote:
>> The default one is intr = ^? BTW...
>
> This would be interesting, because AFAIK, "intr" was set to ^C
> since the beginning of SunOS (behaviour originating from BSD)
> and DEL was used for "erase".
>
> AFAIK, the _SVID_ comes with "intr" = DEL,
> this originates from Version 7.

The system-wide default on Solaris can be set in /kernel/drv/options.conf;
set them the way you want in one window, do an

stty -g

command, put the output in the o