From mboxrd@z Thu Jan 1 00:00:00 1970 From: Malcolm Beattie Date: Fri, 30 Oct 1998 16:52:36 +0000 Subject: Mixed news: new UP 1.0.9 on Ultra5 Message-Id: List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ultralinux@vger.kernel.org I had a go at installing the latest UltraPenguin 1.0.9 on my Ultra5. The previous version of 1.0.9 I tried had difficulties with booting due to various bits of the ultra and, thanks to a flurry of new kernel images from Eddie Dost, it got booting but didn't appear in installable form for a while. The spec of this thing is: Ultra5 270MHz UltraSPARC IIi CPU, 128MB RAM, 4GB IDE disk, SunSwift card (SCSI HBA + extra hme NIC). The boot PROM is "Openboot 3.11". I set up the new tftpboot.img and did a "boot net". The good news is that in this configuration, the new UP1.0.9 boots OK into the installation menus (539 BogoMIPS :-). The extra hme on the sunswift card gets seen as eth0 and the built-in hme as eth1. After switching cables, it turns out that Linux refuses to use the sunswift hme properly. Using tcpdump shows the ultra sending out lots of arp queries (and the replies heading for the ultra) but nothing else. It won't respond to ping requests, nor will NFS work. I confirmed this by setting up the box in a completely different location on another network and the same happens: the kernel log shows detection of both hme's OK (eth0: HAPPY..., eth1: HAPPY...) and the successful bringing up of the link (100/FD in the first test, 10/HD in the second) but no replies. I did configure IP properly. So then I removed the sunswift card and just used the builtin hme and eth0. Good news: it boots fine and gets a long way through installation. All the necessary RPMs were installed and I manually answered the X questions. Then it failed with Error during "Configure timezone" with the log on VC3 saying running /usr/sbin/timeconfig tool failed So I did the next steps manually from the installation menu. The "Install bootloader" step failed with cannot run /bin/rmmod: No such file or directory so I made a symlink to install2 just as /usr/bin/rmmod has. However, even after sorting out rmmos/insmod of loop.o the boot loader still fails to install. I ended up writing an attempted silo.conf from scratch: I can't remember what my old sparc at home has so I guessed boot=/dev/hda map=/boot/map install=/boot/ultra.b image=1/kernel/unix label=solaris solaris image=8/boot/vmlinux-2.1.125.gz label=linux root=/dev/hda8 read=only since I installed Linux on partition 8 (counting 1-8). Running silo -r /mnt gave me the error "failed to open superblock on /dev/hda8". So I did a "sync;sync;sync;halt". That leaves me in the state where Linux has successfully installed (as far as I'm aware) but I need to boot it somehow. I tried a net boot and a floppy boot: boot net linux root=/dev/hda8 boot net root=/dev/hda8 with the same UP1.0.9 installation tftpboot.img. Both put me straight into the installation dialog rather than mounting my installed partition as root. Then I wrote the UP1.0.9 boot.img to floppy and did boot floppy boot floppy linux root=/dev/hda8 boot floppy root=/dev/hda8 but "boot floppy" with anything just says Boot device: /pci@1f,0/pci@1,1/ebus@1/fdthree File and args: ... Bad magic number in disk label Can't open disk label package Can't open boot device Finally the question: how can I boot this thing now? If anyone wants more details of the problems along the way (sunswift, multiple ethernet cards, timeconfig, silo etc.) then let me know. --Malcolm -- Malcolm Beattie Unix Systems Programmer Oxford University Computing Services