Hi together,
As some of you may have seen, I've just put the kernel RPMs for the Intel
i386 distributions 6.3, 6.4, 7.0 and 7.1 in place. You can find them at
ftp://ftp.suse.com/pub/suse/i386/update/6.3/kernel/RPMs/
ftp://ftp.suse.com/pub/suse/i386/update/6.4/kernel/RPMs/
ftp://ftp.suse.com/pub/suse/i386/update/7.0/kernel/RPMs/
ftp://ftp.suse.com/pub/suse/i386/update/7.1/kernel/RPMs/
for download.
The announcement is waiting to be published, but you get the preliminary
version. A kernel update is not rocket science, but it is more sensitive
than exchanging a package that you have never heard of for a stupid
tempfile race. You should carefully follow each step before you reboot
your machine for the freshly installed kernel to become active.
We decided to update the kernel packages on all supported distributions to
the most recent version 2.2.19 instead of fixing the known security bugs
with patches. Tons of other problems have been fixed with the release of
the 2.2.19 kernel, and we believe that this is the most stable kernel in
the 2.2 release row.
To find out which package to download use the command
rpm -qf `awk '/^image =/ { print $3 }' < /etc/lilo.conf`
Additional stuff can be found in the respective misc/ directory under
ftp://ftp.suse.com/pub/suse/i386/update/<distvers>/kernel/misc, such as
the .config used for kernel configuration, the output of the compile run
and the specfile for the RPM packages.
To install the kernel, proceed as follows:
1. Download the kernel rpm for your installation.
2. Install the kernel using the command
rpm -Uhv k_deflt.rpm
or whatever the command necessary for your installation (see above).
It may be necessary to either deinstall the old kernel package and
modules package or to use the "--force --nodeps" commandline options
for rpm in addition to -Uhv.
3. If you use an Adaptec aic7xxx scsi host adapter, please change
"aic7xxx" in INITRD_MODULES on top of /etc/rc.config to read
"aic7xxx_old".
4. Execute
mk_initrd
An error indicates that the initrd is not written to disk. It is
possible that your machine will not boot if you have a scsi system.
5. Execute
lilo
If you did not change the line starting with "initrd" in
/etc/lilo.conf, then the freshly created init-Ramdisk will be loaded
together with the kernel by lilo upon system boot. This initrd contains
the drivers that are necessary for the kernel to mount the root
filesystem of the system. On scsi systems, you can't boot without the
modules contained in the initrd.
6. If you use lvm, update the lvm package as well. It is contained in the
same directory as the kernel rpms.
7. To avoid complaints from the kernel about an old mount, update the
util-linux (util for 7.0 and older), too, using rpm -Uhv <packname>.
8. The 7.0 distribution contained a bug in the glibc that is triggered by
2.2.19. The bug may result in mounted nfs filesystems to report -EIO
when listing directories in some rare cases. Get the update package
for shlibs (the other sub-packages are not needed) from
i386/update/7.0/a1/shlibs.rpm, and install it, again, using "rpm -Uhv
<packname>". Keep your machine in a quiet state during this upgrade; if
at all possible, bring the machine into single user mode, using the
command "init 1"
Execute "ldconfig" after completion of the rpm command.
9. Execute
init 6
or
shutdown -r now # which is the same without warning
...and if it fails, drop me a mail, please.
Thanks,
Roman.
--
- -
| Roman Drahtmüller
Which of these rpm's is the source for those of us that build our own kernels? Viel Spaß -- Paul Elliott 1(512)837-9345 1(512)837-1096 pelliott@io.com PMB 181, 11900 Metric Blvd Suite J http://www.io.com/~pelliott/pme/ Austin TX 78758-3117
Hi, On Sat, May 12, Roman Drahtmueller wrote:
As some of you may have seen, I've just put the kernel RPMs for the Intel i386 distributions 6.3, 6.4, 7.0 and 7.1 in place. You can find them at
Ahh, finally ;) I have some little comments... [...]
3. If you use an Adaptec aic7xxx scsi host adapter, please change "aic7xxx" in INITRD_MODULES on top of /etc/rc.config to read "aic7xxx_old".
Please note that this is not strictly needed. The wording suggests that the new driver does not work. In fact, it works quite well. We just found some hardware configurations that show some quirks in some combinations: When trying to mount a CD-ROM immediately after inserting the CD-ROM, the mount will fail one some drives. When waiting until the drive has completed its initializiation, everything works fine even on those affected machines. Something similar was experienced here on one single machine with a HP DAT drive: Accesses to the tape failed when being accessed immediately after inserting the tape. In no single case did we have problems accessing harddisks with the new driver. So you really only need to "downgrade" to the old driver if you experience this sort of problems. [...]
7. To avoid complaints from the kernel about an old mount, update the util-linux (util for 7.0 and older), too, using rpm -Uhv <packname>.
This is a cosmetic problem and only affects NFS mounts. When mounting an NFS volume with the old mount, the kernel will issue a warning: kernel: nfs warning: mount version older than kernel As said above, this is purely cosmetic; the mount will succeed and everything works ok. We decided to provide the update nevertheless as the kernel warning might scare some people.
Roman.
Thanks very much Roman. See you at Sur Mandir soon ;) -o) Hubert Mantel Goodbye, dots... /\\ _\_v
* Hubert Mantel
Hi,
util-linux (util for 7.0 and older), too, using rpm -Uhv <packname>.
Ok I have downloaded the util for 7.1 in addition to k_deflt-2.2.19 here are the steps I did rpm -e k_deflt rpm -Uhv k_deflt-2.2.19 rpm -Uhv --nodeps --force util-linux (fails due to conflicts with -Uhv) mk_initrd lilo reboot /dev/hdb3 was not cleanly unmounted ( ????? unable to mount /dev/hdb3 manual fsck require e2fsck -y /dev/hdb3 CONTROL D Ok so far everything looks fine so it was not cosmetic to me unless I did it wrong
As said above, this is purely cosmetic; the mount will succeed and everything works ok. We decided to provide the update nevertheless as the kernel warning might scare some people.
-- Togan Muftuoglu
Hello
here are the steps I did
rpm -e k_deflt rpm -Uhv k_deflt-2.2.19 rpm -Uhv --nodeps --force util-linux (fails due to conflicts with -Uhv) mk_initrd
Do I need to download all of these for an ordinary Intel based workstation ? k_deflt-2.2.19-11.i386.rpm k_eide-2.2.19-0.i386.rpm k_i386-2.2.19-12.i386.rpm Thank you -- Richard
Hi, On Sat, May 12, Richard Ibbotson wrote:
rpm -e k_deflt rpm -Uhv k_deflt-2.2.19 rpm -Uhv --nodeps --force util-linux (fails due to conflicts with -Uhv) mk_initrd
Do I need to download all of these for an ordinary Intel based workstation ?
k_deflt-2.2.19-11.i386.rpm k_eide-2.2.19-0.i386.rpm k_i386-2.2.19-12.i386.rpm
No, you only need one of them depending on the kernel you're using. 99% of all people will use k_deflt-2.2.19-11.i386.rpm. k_eide only is needed for some special EIDE controllers k_i386 is only needed for really old machines (I assume you have at least a Pentium class computer, so just use the default kernel).
Richard -o) Hubert Mantel Goodbye, dots... /\\ _\_v
On Sat, 12 May 2001, Hubert Mantel wrote:
No, you only need one of them depending on the kernel you're using. 99% of all people will use k_deflt-2.2.19-11.i386.rpm.
k_eide only is needed for some special EIDE controllers k_i386 is only needed for really old machines (I assume you have at least a Pentium class computer, so just use the default kernel).
What type of machine is k_pentiu-2.2.19-0.i386.rpm meant for? (I did not see this in the 6.3 tree before today. rpm -qip gives no clues.) dproc
Hi, On Sat, May 12, Togan Muftuoglu wrote:
* Hubert Mantel
[010512 15:07]: Hi,
util-linux (util for 7.0 and older), too, using rpm -Uhv <packname>.
Ok I have downloaded the util for 7.1 in addition to k_deflt-2.2.19
here are the steps I did
rpm -e k_deflt rpm -Uhv k_deflt-2.2.19 rpm -Uhv --nodeps --force util-linux (fails due to conflicts with -Uhv) mk_initrd lilo reboot /dev/hdb3 was not cleanly unmounted ( ????? unable to mount /dev/hdb3 manual fsck require e2fsck -y /dev/hdb3 CONTROL D
Ok so far everything looks fine so it was not cosmetic to me unless I did it wrong
This is something unrelated. You did update the util-package anyway, so you will not see the NFS mount complaints. The real question is: Why was /dev/hdb3 not cleanly unmounted on shutdown?
As said above, this is purely cosmetic; the mount will succeed and everything works ok. We decided to provide the update nevertheless as the kernel warning might scare some people.
-- Togan Muftuoglu -o) Hubert Mantel Goodbye, dots... /\\ _\_v
Ok I have downloaded the util for 7.1 in addition to k_deflt-2.2.19
here are the steps I did
rpm -e k_deflt rpm -Uhv k_deflt-2.2.19 rpm -Uhv --nodeps --force util-linux (fails due to conflicts with -Uhv) mk_initrd lilo reboot /dev/hdb3 was not cleanly unmounted ( ????? unable to mount /dev/hdb3 manual fsck require e2fsck -y /dev/hdb3 CONTROL D
Ok so far everything looks fine so it was not cosmetic to me unless I did it wrong
This is something unrelated. You did update the util-package anyway, so you will not see the NFS mount complaints.
The real question is: Why was /dev/hdb3 not cleanly unmounted on shutdown?
Good morning! :-)
The thing with the unclean shutdown can happen if init doesn't know the
runlevel it is in. This is the case when init crashed (you don't see
that), maybe after an upgrade of some library. It should not repeat,
though.
Roman.
--
- -
| Roman Drahtmüller
* Roman Drahtmueller
The thing with the unclean shutdown can happen if init doesn't know the runlevel it is in. This is the case when init crashed (you don't see that), maybe after an upgrade of some library. It should not repeat, though.
So if I bring the machine to init 1 for example do the upgrading and then issue another init 1 and then reboot will that solve the problem of e2fsck -- Togan Muftuoglu
So if I bring the machine to init 1 for example do the upgrading and then issue another init 1 and then reboot will that solve the problem of e2fsck
Runlevel 1 is not necessary for a kernel upgrade, since the kernel files
(and modules) are not actively used from the filesystem any more. It's
more about the libc upgrade (shlibs package) if you have a 7.0.
Two times init 1 in succession will render the second time useless (it's a
no-op). Try to reboot your machine with shutdown -r now, or with init 6.
If reboot can't determine the runlevel, it will issue the reboot(2) system
call, instead of telling init to go to runlevel 6 (at the end of which
there will be a reboot again). I experienced this behaviour a long while
ago, and I guess this has changed in the meanwhile, but it is a possible
explanation for the problem you see.
That you had to repair manually indicates another problem. But I can't
tell what's wrong with it.
Just see if it repeats.
Roman.
--
- -
| Roman Drahtmüller
* Roman Drahtmueller
no-op). Try to reboot your machine with shutdown -r now, or with init 6.
OK :-) I did upgrade the firewall/router and this time I issued the above and phew everything was unmounted with no problem and everything started properly. The only difference was I did not install the util-linux package ( I can live with a cosmetic error on the firewall). Also thanks for pointing out the nfs install of the rpm as I was so focused to do it by ssh I have simply forgoten other alternatives. Thanks -- Togan Muftuoglu A Happy SuSE Customer
* Roman Drahtmueller wrote on Sat, May 12, 2001 at 15:37 +0200:
The thing with the unclean shutdown can happen if init doesn't know the runlevel it is in. This is the case when init crashed (you don't see that), maybe after an upgrade of some library. It should not repeat, though.
init is linked statically, isn't it? IIRC init case that init cannot determine the current runlevel, it displays such a message. oki, Steffen -- Dieses Schreiben wurde maschinell erstellt, es trägt daher weder Unterschrift noch Siegel.
* Hubert Mantel
Hi,
This is something unrelated. You did update the util-package anyway, so you will not see the NFS mount complaints.
Good :-)
The real question is: Why was /dev/hdb3 not cleanly unmounted on shutdown?
If I had only known the answer. As I have to upgrade the kernel for my firewall/router PC which is low on space (well actually the whole thing is 180 MB Hard disk ) so I have to initiate rpm -e for both packages before I can do rpm -Uhv for the new ones so I can have space I have to scp these rpms first unless someone can show me a way to install the rpms over ssh without actually copying the rpm to the firewall -- Togan Muftuoglu
The real question is: Why was /dev/hdb3 not cleanly unmounted on shutdown?
If I had only known the answer. As I have to upgrade the kernel for my firewall/router PC which is low on space (well actually the whole thing is 180 MB Hard disk ) so I have to initiate rpm -e for both packages before I can do rpm -Uhv for the new ones so I can have space I have to scp these rpms first unless someone can show me a way to install the rpms over ssh without actually copying the rpm to the firewall
Make a ramdisk that is big enough and copy it there.
Mount the directory where the rpm is via NFS and install it from there.
Convert the rpm to a cpio archive and pipe this archive directly into
cpio, maybe using netcat or alike.
I shouldn't give you this advice, actually, because a consistent system is
worth something the next time. It's just that in this case it doesn't
break that many things...
Thanks,
Roman.
--
- -
| Roman Drahtmüller
participants (7)
-
dproc
-
Hubert Mantel
-
Paul Elliott
-
Richard Ibbotson
-
Roman Drahtmueller
-
Steffen Dettmer
-
Togan Muftuoglu