The combination of (IBM/Lenovo ThinkPad) T60 chassis and T61 mother board is known as FrankenPad, I learnt since I bought a none too well refurbished Lenovo T61 7659-CTO laptop for a moderate price. I had long wished I could get a T61, so that was some impulsive buying.
Maybe I should have spent the money (my guess is around $ 100 US) on a brand new Raspberry Pi I've also been wishing for, but I could not resist. The 9front fraternity may be pleased to hear that this new addition to my stable of obsolete equipment is currently capable of running both 32-bit and 64-bit versions of 9front - I just realised I'll need to compile executables for both architectures indefinitely, I wonder how many times that will bite me? To run under 32 bits I resorted to network booting (from my long suffering traditional Plan 9 network server), but that won't load the 64-bit kernel, complaining that it is too big. I tried compressing it, but netbooting no longer supports compression. I paired down the kernel a bit, but seemingly not enough. I presume 9front has ways to netboot a 64-bit kernel, but it isn't critical, yet - it would just fill what is a rather obvious hole that 9front and 9legacy (for want of a more suitable moniker - 9pf doesn't seem right) seem to suffer from differently. And, yes, this could be the start of a long, offline whinge about differences, I have long evolved a flameproof skin for this particular purpose. But first, let me tell my cautionary tale, it was quite an adventure and I am happy to act as proxy for those who may want to go in a similar direction. I had the idea to install both 9front and 9legacy on the T61 and thought I might run cwfs in the former case after discovering then that 9front has enhanced cwfs - which I have never used, but did for a long time use kenfs standalone - so I followed their lead for that. For 9legacy, I'm fine with fossil/venti, it has saved my bacon a few times and I respect its capabilities fully. So, where should I have started? Obviously, this being the non-deterministic world of New Computing (TM), I followed my head and installed 9front - no one argues that it is the one most likely to work on a T61. I can't quite recall how, but I managed to do something that in retrospect was not a great idea: I set up two Plan 9 primary partitions using Linux Mint off a USB stick - Windows was just not an option, in my experience, for editing partition tables. I left Windows 7 Professional installed, but shrank the partition to a safe, much smaller size - that left some scar tissue, incidentally, but irrelevant to this tale. The 9front installation completed without any memorable trouble and I left the boot loader unchanged (as instructed). Somehow, boot selection didn't work as I wished and I blamed the double partition for my woes. Time to start again, this time with the 9legacy bootstrap that I was in any case more comfortable with and only one, combined Plan 9 partition. It looks as if 9front used the same partitioning scheme as 9legacy. I got some idea of partition allocations to "other", "fscache" and "fsworm" from a recent disk/prep display, so I decided to configure the drive with fossil and venti partitions, leaving enough space for 9front, when I would eventually re-install it. The 9legacy installation almost, almost worked. I assigned half the plan9 partition to fossil, arenas, isect and bloom and left space for 9front. I assumed that nvram and 9fat would need to be shared. Except I got some spurious errors in the very last stage of writing the bootstrap loader and what looked like an otherwise happy installation simply could not be completed. I could not get past the final stage of 9legacy installation. The complaint was that 9fat could not be created, or perhaps something could not be written to that partition - from memory, it was the error one encounters after a server connection has failed. At that point only a reboot made sense to me. Of course, rebooting with the plan9 partition active didn't do anything useful. It's likely that this is when I also discovered that the Windows partitions were no longer recognised as bootable. That lost me the Windows recovery capability on the drive, but that was never an essential, no regrets. With a partially complete 9legacy installation, the time had come to see what 9front was good for. So I repeated that installation. When the time came to allocate disk space, however, 9front installation had no record of the previous content of the plan9 partition. As I had started to keep track of such things, I just proceeded with manual partitioning (not as wisely as I imagined, I am only now discovering). I set up all the partitions I could think of - and made a few judgemental mistakes, it turns out, but I didn't notice, so I could actually continue. The completed 9front installation this time included the 9front boot loader - which I will have to become more familar with, for obvious reasons. I have accepted that Windows will require special attention and will almost certainly not get it any time soon. With a working 9front installation, I was a lot more confident, ready to try 9legacy once more. If I made any additional preparations at this point, I do not remember them. Once again, 9legacy installation (fossil+venti) proceeded as expected, with manual disk preparation, using the space left by 9front. The installation had respected the previous settings, which needed some manual rearrangement. Once again, after a successful run, the penultimate step reported the same 9fat trouble and the last step simply failed altogether, just like before. Now, with a working 9front (32-bit) installation known to be working, thinking that I had sound foundations in place, I proceeded to do some post-installation configuration of the 9front system. It took a separate adventure to update the sources and regenerate the system including the amd64 version. The details also raise some issues, so expect a separate report for that. Small tweaks to plan9.ini - thank you, Stanley - allowed me to switch to the newer, more appropriate architecture, which is what I'm running now. I have had a netboot system in place just about forever and I thought I would use that to get past my lack of familiarity with 9front booting to be able to switch between 64-bit and 32-bit 9front and plausibly also 9legacy without a working boot loader and a suitable plan9.ini configuration. I think I mentioned that the Plan 9 bootloader refuses to work with the 64-bit kernel (that may be my error, in that I have no idea how the switch in architectures is likely to take place and where). It loads the 32-bit kernels adequately, so that it is possible for me to run normally under amd64 9front, with the 386 option available on demand. I can load the 9legacy kernel this way too, but it fails to access the SATA drive and fails quite miserably. If I boot off CD, though, using 9pcflop, I can access the 9legacy system and start both venti and fossil. I can also edit the 9fat partition, so that is food for thought. Venti works on 9front as well. For now, while the arenas are completely empty, the behaviour of venti is consistent across these flavours. In the long run, I'd like the two flavours (9front and 9legacy) to be both bootable in the available architectures (386 and amd64) and to be able to access each other's file systems at all times. That seems possible if fossil can be ported to 9front and cwfs64x to 9legacy - I haven't looked for such options yet, not while 9legacy is not an option. How other flavours (9atom, nix, etc.) can then be shoehorned into such a single ecosystem is a much more complex matter to resolve. There are many questions raised as a result of the efforts described above, I'll try to formulate them so that they can be resolved objectively. Private mail with suggestions, comments, insults and praise will be entertained as best I can. Lucio. ------------------------------------------ 9fans: 9fans Permalink: https://9fans.topicbox.com/groups/9fans/Tff7ec3b7f1114286-M6e8d1a56fa784994195e1a3b Delivery options: https://9fans.topicbox.com/groups/9fans/subscription
