XPostFacto 3 boot problems |
September, 06, 2003 11:29 PM |
powderhaus |
I have been having some real problems with X. After my first attempt at installing X on my boot drive my computer crapped its pants and would not start on anything- except the PRAM restore disk so i had to reformat some drives and reinstall again from scratch. I was then amazed that Ryan included firewire CDROM booting in the first release of XPF 3 (Nice job) ten the install went fine first reboot went good, but then i installed 10.2.6 (along with a list of other software update programs) and sometimes my drive will not work (firewire, i have had this problem in the past and thought i found the solution...) other times i get a kernel panic (it seems when i do watch the verbose it does not panic) to some real strange errors (i will go get another free .mac to post them on) to the IOFireiwiresomething write fetch failed (something like that, and it fails about 50 times during the boot) to just not getting past the blue screen. Also (i don't think this is XPF) but my network has not been working, i messed with the settings (put them the way they should be) but for some random reason after about the 5th restart (from a successful boot) it started working... kinda odd... Other than that and the fact that it does not enable L2 Cache, it works... but have to restart my computer 1-4 times every reboot is not very cool, but i knew what i was doing when i got the alpha, i figure the more people that get it the faster it will be fixed. Thanks Ryan |
. |
RE: XPostFacto 3 boot problems |
September, 17, 2003 3:19 PM |
powderhaus |
. |
now i am getting a kernal panic just after it says the stuff about the color screen. at the begining of the kernal panic it says that it cannot find a driver for my computer. that is from os 9. i am going to try osx now. |
. |
RE: XPostFacto 3 boot problems |
September, 17, 2003 5:54 AM |
nick.ashton |
. |
I got the "file is in use or was left open" message a few times and found the workround was to delete the files from the Helper disk. It seems to be a problem updating or overwriting the helper files, not a problem reading the FW disk. |
. |
RE: XPostFacto 3 boot problems |
September, 16, 2003 4:39 PM |
powderhaus |
. |
after running diskwarrior on the FW disc (because it was crashing at the blue screen) i am no longer able to start to boot off the drive. When XPF is copying the BootX file it says that the file is still in use or was left open. i have restarted tried in 9 and X but that is all i get. I have been getting really odd errors when i did boot, so i need to know where the log is. |
. |
RE: XPostFacto 3 boot problems |
September, 16, 2003 8:36 AM |
OSXGuru |
. |
powderhaus: One thing you might try is reinstalling BootX, just to make sure that the latest version is installed. XPF should take care of this for you, but it's worth a try. Also, I should have another alpha version of XPF out soon, probably next weekend, with a significant number of bugs fixed--it may turn out that one of the bug fixes catches your situation. |
. |
RE: XPostFacto 3 boot problems |
September, 15, 2003 10:32 PM |
powderhaus |
. |
yeah i went to get CCC and found that i needed the subsystem, oops. anyway, after my install of OSX on my firewire drive (from my firewire CDROM) i rebooted into X from my other drive and rean the whole 9 yards in drive setup. I then tried to use XPF in X but apon reboot (and holding cmd v, verbose mode otherwise does not engage early enough and i get the no system folder icon, anyway it gave me an error that looks like this: Opening partition [scsi-int/@0:7]... HFSinitpartition: ff8d3500 loading hfs+ file [.xpostfactopci1@d@4@30e00e000008c@c000@0;3mach_kernal] from ff8d3500 failboot: 2 enter called open firmware, 1.0.5 to continue booting MacOS type: BYE to continue booting from defualt boot device type: Boot OK Then later i was getting alot of "waiting for directory services" (when using XPF from 9) Now i am booting 10.2 ( i have always been able to boot to that, just after the 10.2.6 install i can't boot anymore) and i do not get any FWFetchWrite errors at all. i do get a bunch of errors just before the end of the boot but i can still use the computer in OSX but for some reason nothing will mount except my FW drive. All other disks and disk images will not mount. It must be the errors late in the boot. i will post the later. Thanks for the help. |
. |
RE: BSD Subsystem |
September, 15, 2003 4:37 PM |
nick.ashton |
. |
Although the base system will run without the BSD subsystem installed you will find that many applications rely on its presence to run properly. Actually I thought CCC was one of the programs that required the BSD subsystem but I might be wrong. Anyway, I think it's probably a good idea to do a clean install on your Firewire drive as I've had problems with cloned systems retaining references to the hardware location of the original drive. For instance clicking on an application icon in the dock on the cloned system and having the application on the original drive start up instead. Dismounting the original disk results in an extra delay in application launching |
. |
RE: XPostFacto 3 boot problems |
September, 15, 2003 4:10 PM |
OSXGuru |
. |
I have fixed one problem with ownership of directories. The Mac OS 9 version was getting the ownership of directories like /System and /System/Library wrong. But I've got it fixed internally now, so it should work better in the next version I release (sometime this week). I haven't been able to reproduce problems with the permissions of the kernel extensions themselves, though. |
. |
RE: XPostFacto 3 boot problems |
September, 15, 2003 3:26 PM |
powderhaus |
. |
I will try again but i do want to know what he BSD subsystem is and is it all that impotant? i would lke to CCC my SCSI drive, but i did not install the BSD subsystem because i need to save space on my drive, so is it worth reinstalling the system again on the FW drive or is it almost useless? |
. |
RE: XPostFacto 3 boot problems |
September, 15, 2003 9:10 AM |
nick.ashton |
. |
Well I've been through everything again and I now have a system installed which seems to behave correctly even after I boot into OS9 and then use XPF to get back to OSX. I did find that running Repair Permissions in the OSX Disk Utility seemed to help but I haven't actually identified a single cause for all the strange behaviour I reported earlier. I did discover that the problem with VNC startup was actually timing related - the access to the Firewire disk is so much faster that VNC was starting quicker and trying to access the main display before the window server had configured it. Also, looking back at console logs of previous boots from a pure SCSI system shows that the "(ipc/send) invalid destination port" messages seem to be a feature of OSX so aren't related to Firewire booting. However, I did notice one thing which might finally track down the cause. The xpfbootsupportd process crashes with a bad access exception every time after b> it has successfully updated the Extensions folder on the helper disk. I've opened this problem in a new thread and also opened a bug report on Bugzilla. |
. |
RE: XPostFacto 3 boot problems |
September, 11, 2003 5:35 PM |
powderhaus |
. |
I am still getting all of the FW write errors, and i can not boot into 10.2.6, but i did figure out why, i noticed that your sincronizing stuff never turned on (once at the beginning before 10.2.6 was actually installed, it turned on for iphoto or something - via software update). I put this on bugzilla (well i thought i did....) but the way i think you should be doing the syncing is when the user trys to shut down, log out, or restart it should check for syncing, that way you don't have to worry about restarting early or anything. also i got the tango 2.0 card and it does not work any better than the belkin card, i still get 50+ fw fetchwrite errors and it fails at the same point my belkin did. I am going to install 10.2 on my normal SCSI drive then continue with nicks directions with the belkin card, so i can return the Sonnet one as i can see no advantage to it, besides USB 2, but FW is my game. |
. |
RE: XPostFacto 3 boot problems |
September, 10, 2003 6:44 PM |
OSXGuru |
. |
I should add that I had been assuming that the Firewire booting problem was at the Mac OS X end--your observations are now getting me to focus on what XPF does in Mac OS 9, and it will be interesting to see what comes of that. |
. |
RE: XPostFacto 3 boot problems |
September, 10, 2003 6:39 PM |
OSXGuru |
. |
Thanks, Nick. I'll have to take a close look at what you report and see if I can reproduce it. I appreciate the detail. A few comments: -- I am planning to change XPF so that you can make the appropriate settings and then quit without rebooting immediately. In that respect, it will behave more like Startup Disk does now, which is a good thing (in that it ought to match user expectations better). Here's a reference to the bug: Should offer to write NVRAM when closing -- Your idea about being able to change XPF's behaviour's for debugging purposes is a good one. It could be useful to make a "debugging" version available so that people don't have to recompile to get different behaviours (which is what I do, but it's more work for others :-) Make "debug" options available to facilitate testing -- Both the OS 9 and OS X version of XPostFacto should be copying only files that have been modified, which is why it goes by so quickly sometimes. -- You're quite right that the wrong ownership on the files could cause the problems that we see. But I thought I had that working :-) I'll have to test it and see if I can reproduce the problem. One question is whether you have "ignore ownership" selected on that volume. If so, the ownership may be reported to you incorrectly, but that would not affect the boot process, as BootX is the only thing that reads the helper disk, and it doesn't know about "ignore ownership." (It does enforce the root/wheel requirement, though, I believe). The other question would be why it works for some Firewire drives but not others. But if I can narrow this down to a reproducible problem, I should be able to fix it. Here's the URL for the bug report: Kernel extensions copied to "helper drive" have wrong owner And a related bug report Test copying of ownership when "ignore ownership" checked -- The "version.plist.bak" files should not be there. Are you sure they aren't a product of some disk utility that you are running? I did a search through the source code, and I can't find "bak" anywhere. But I'll open a bug and try to reproduce it: Helper disk contains version.plist.bak files -- XPF should be copying the Extensions.mkext file to the helper when it notices that it has changed. So when you first do the install, it will copy the Extensions.mkext from the CD to the helper. Once the install is finished, the computer will boot back into Mac OS 9 (because the Mac OS X Installer sets the NVRAM to boot directly from the Firewire drive, which won't work). At that point, when you run XPF again, it should copy the Extensions.mkext from the Firewire drive to the helper. (This would be the Extensions.mkext which the Mac OS X Installer created). Is what you're seeing different than what I describe?. One thing that may not be obvious is the structure of the .XPostFacto directory on the helper. Basically, it is organized according to the Open Firmware name of the target volume. So each target volume gets its own "space" in the .XPostFacto directory of the helper. Also, the install procedure actually uses the "helper" idea to do its work. That is, installing from a CD always uses the target volume as the "helper" (or, the target volume's helper, if it has one). |
. |
RE: XPostFacto 3 boot problems |
September, 10, 2003 3:58 PM |
nick.ashton |
. |
Ryan, thanks for your observations. Here are the reasons why I suspect that XPF3 under OS9 is not correctly copying the extensions to the helper disk - When I run XPF3 under OSX and select a FW disk and a SCSI helper it's clear from the on-screen feedback and the amount of time it takes that the entire contents of / System/Library/Extensions are being copied to the helper disk even if a full set of extensions is already present from a previous boot. By contrast, the same operation under OS9 often completes so quickly that it is physically impossible for all of the files (~45MB) to have been written. Either the OS9 version only updates files which it thinks are out of date or it doesn't complete the job. Having achieved a stable, working installation of OSX on a Firewire disk it will reboot happily time after time but will only fail if I reboot into OS9 and then use XPF3 to reboot from Firewire. The only thing that changes in these circumstances are the contents of the helper disk (and the NVRAM). Doing a comparison of the files on the helper disk and the files on the boot disk, reveals that there are several files called "version.plist.bak" on the helper disk in various extension directories which don't exist on the system disk itself. Presumably these are created by the mechanism that XPF uses to update the helper disk. Ownership of the extension files on the helper disk are usually wrong. OSX is fairly picky about kexts belonging to user "root" and group "wheel". Typically the kexts on the helper disk end up belonging to group "unknown". To do some further investigation it would be very useful if a version of XPF3 was made available which allowed the user to override the updating of the /System/ Library directories on the helper disk. That way I can make manual copies of the directories under OSX, ensure that they are consistent and belong to the correct user and group without having them zapped by XPF every time I reboot. Also the ability to have XPF make all the required changes but NOT reboot immediately would be helpful to speed up debugging and reduce the chance of corrupting disks by having to interrupt reboots with the power button. Out of interest does XPF3 employ some sort of comparison between what is on the helper disk and what is on the boot disk or does it attempt to blindly copy everything each time. The OS9 and OSX behaviours seem to suggest it varies with the parent OS. The final item which seems to be a factor is the extensions.mkext file. Systems seem to work best when the extensions.mkext on the helper disk is updated in line with what is on the boot disk. However, during an initial attempt to install there is a time when the extensions.mkext on the helper disk is empty or non-existent and is only updated when a successful boot into OSX is achieved, the OSX system rebuilds the extensions.mkext file, and the xpf background process then copies this back to the helper disk. I've tried forcing a rebuild of this file using the "kextcache -e" command. |
. |
RE: XPostFacto 3 boot problems |
September, 10, 2003 12:35 PM |
lei1 |
. |
I ran disk warrior on a fresh Jaguar/Xpostfacto3 installation that wasn't booting and it reported a repair tas follows: Folder: “.XPostFacto�! Repaired BSD Type! Location: “Storage/� Don't know what this means but it may be relevant. There were a number of other problems including overlapped files but after repair the system was sweet. I was installing OSX on another SCSI drive to use as a bootable utility disk (in preparation for OSX to OSX XPF booting). I also had a problem in XPF3 when reloading XPF extensions & BootX. I got the "grey screen of death" when trying to boot my regular OSX disk. The text read that it was "unavailable". I returned to XPF2 and have had no more problems since. I am steering a bit clear of the XPF3 alpha for now but I love the new interface and am eagerly awaiting OSX to OSX booting. Thanks again Ryan for your excellent work. |
. |
RE: XPostFacto 3 boot problems |
September, 10, 2003 8:19 AM |
OSXGuru |
. |
Hmm. The XPostFacto code that does the copying is basically the same in the Mac OS 9 version and the Mac OS X version. So I'll have to think about why it would work OK in one instance and not in the other. How do you determine that there are errors in the copy? That is, do you run some kind of disk utility that reports those errors? Or is it a theory? Basically, I'm trying to figure out how I can reproduce this and test it. If the problem is in the way that XPF copies the files, that would be great, as that can be fixed :-) |
. |
Found the problem! |
September, 10, 2003 7:21 AM |
nick.ashton |
. |
Well I think I've found the problem at least as it affects my machine. It appears that XPF3 is making a bad copy of the /System/Library/Extensions on the Helper disk when run under OS9. The copied extensions seem to be corrupt in a number of ways which cause all sorts of errors when booting from them. My workround requires that you already have a bootable OSX system on a SCSI or ATA disk. Currently you have to run XPF3 under OS9 at least once to make sure the BootX file is installed (there's a known restriction which prevents this being done under OSX). So my suggested method is to run XPF under OS9, select your preferred FW boot drive and SCSI/ATA helper drive and then click "Restart". Watch XPF copying BootX and a copy of your extensions to the helper disk and then, when the system begins to restart, interrupt it by pressing the power button on the CPU. Boot back into OS9 by holding down the option key and powering on the system. Now use XPF 2.2.5 to select your bootable OSX system on a SCSI or ATA drive and let the system reboot. Now run XPF3 under OSX and reselect your FW and Helper disks. XPF should now make a good copy of the extensions and the system will then boot cleanly from the FW drive. Just to be sure that you get a new copy of the extensions on the helper disk you could run a terminal session and explicitly remove the /System/Library/ Extensions folder and /System/Library/Extensions.mkext file from the helper disk. You'll find these buried inside a directory tree that starts .XpostFacto/pci/... on the helper disk. It's invisible in the OSX finder because it start with a dot character, hence the need to use the terminal. |
. |
RE: XPostFacto 3 boot problems |
September, 09, 2003 5:28 PM |
OSXGuru |
. |
Thanks, Nick, for the report. It's interesting that the throttle makes a difference. It really shouldn't, because it only is supposed to affect the very earliest part of the boot process (before the Firewire drive is accessed). I suppose it is possible that the throttle has a residual effect that I'm not thinking of. (What it does is set the ictc register in the NVRAM script to slow down the instruction cache. But both BootX and the ApplePowerSurgePE set it back to normal, so it shouldn't have any residual effect). But I'll give the throttle a try in some of my troubleshooting, just in case. The other thing that is interesting is that the same files caused a problem each time. If it was purely a timing issue, you wouldn't expect the same files to cause a problem--you'd expect more random variation. Though I suppose the boot process is relatively deterministic, so the same things might happen at more or less the same time. (Come to think of it, I suppose that the throttle would change the boot timing globally--that is, by lengthening the process. Though that isn't the kind of timing issue one would expect to be at fault here. Unless, for some odd reason, the Firewire cards work better if you access them later, which seems unlikely). Anyway, if the same files were causing a problem each time, then it seems likely that the files were corrupted in some way (which may have been fixed when you ran the Disk Utility). So the problem may lie in errors writing the files, rather than reading them as such. So, much food for thought. |
. |
RE: XPostFacto 3 boot problems |
September, 09, 2003 2:10 PM |
powderhaus |
. |
Have tried almost all of those suggestions but i cannot get 10.2.6 to boot, it will boot off the CD (FW, 10.2 installer) and into 10.2 (with plenty of errors...) but any update after that will crash. For a while i was being put into single users mode every time and i did not check that in XPF, now i just get a panic right after the 1699 stuff in verbose mode (that is right after the VIDEO stuff) i got it from OWC, and if i say it doesn't work at all i should be able to get my money back, if worst comes to worst then i will put the belkin card on ebay, the belkin one seems to be the culprit for all my FW problems (previous to this also) but i have never been able to say 100%. thanks |
. |
RE: XPostFacto 3 boot problems |
September, 09, 2003 7:50 AM |
nick.ashton |
. |
Sorry guys, forgot about the stupid way this forum deals with line spacing. Here are the two lists from my previous post properly formatted (I hope) What I did 1) Set throttle to 8 in XPF. 2) Set verbose booting as the default. 3) Disabled the L2 cache. 4) Ran Disk Utility on another OSX system and repaired the FW boot drive, I did see some errors saying "Invalid BSD file type" which were repaired. 5) Use DiskWarrior to rebuild the dirctories on the boot disk. 6) Manually deleted the /System/Library/extensions.mkext and extensions.kextcache files to force them to be rebuilt. 7) Changed the /etc/rc boot script to mount all file systems early in the boot sequence so I could then put some of the problematic startup items back on a SCSI disk. Some of these items then worked properly. 8) Switched to a different port and connecting cable on the FW card. 9) Removed all USB devices in case this was afftecting the FW/USB card. So I'd suggest that anyone who has problems tries the following order of actions Be prepared to do item 4) repeatedly if you encounter any failed boots or have to do a forced restart. 4) under OS9 Disk First Aid or under OSX Disk Utility on another m/c 5) (if you have DiskWarrior) 8) 1) 2) 3) 6) 9) |
. |
RE: XPostFacto 3 boot problems |
September, 09, 2003 7:45 AM |
nick.ashton |
. |
Well I have good and bad news to report. I am now successfully booting off the firewire drive error free every time, without any throttle (0), without having to resort to a verbose boot and with all startup processes running normally. The bad news is that I have no idea why it is now working. Here's the list of things I went through in trying to track down the problem but everything is now back to the way it was. 1) Set throttle to 8 in XPF. 2) Set verbose booting as the default. 3) Disabled the L2 cache. 4) Ran Disk Utility on another OSX system and repaired the FW boot drive, I did see some errors saying "Invalid BSD file type" which were repaired. 5) Use DiskWarrior to rebuild the dirctories on the boot disk. 6) Manually deleted the /System/Library/extensions.mkext and extensions.kextcache files to force them to be rebuilt. 7) Changed the /etc/rc boot script to mount all file systems early in the boot sequence so I could then put some of the problematic startup items back on a SCSI disk. Some of these items then worked properly. 8) Switched to a different port and connecting cable on the FW card. 9) Removed all USB devices in case this was afftecting the FW/USB card. I was as methodical as I could be, only making one change at a time and checking the boot reliability. Certainly setting the throttle to 8 got rid of the messages saying fetch agent write failed! retrying and setting verbose boot reduced the number of times that the boot sequence ended at a blank grey or blue screen due to crashed startup processes. However, having got it to a state after item 8) where it seemed to be booting reliably every time I started backtracking the changes hoping to find the point at which it broke again. Unfortunately that never happened. It's now back as close to its original condition as I can get it and boots cleanly every time in "qiuet" mode with no throttle, back on its ortiginal FW port and cable with all USB hubs & peripherals connected. So I'd suggest that anyone who has problems tries the following order of actions Be prepared to do item 4) repeatedly if you encounter any failed boots or have to do a forced restart. 4) under OS9 Disk First Aid on original m/c or under OSX Disk Utility on another m/c 5) (if you have DiskWarrior) 8) 1) 2) 3) 6) 9) Hopefully someone else may spot the vital step. Just to save people referring back to my other threads the m/c config is PM8600, G3/400/1MB, 352MB RAM, Sonnet Tango FW/USB1 Combo card LaCie 20GB FW/USB PocketDrive (using FW i/f), Internal SCSI disk as helper drive Built-in video, built-in Ethernet + 2xDEC DE500 10/100 Ethernet cards. I'm using PowerLogix CPU Director v1.1f1 to enable the L2 cache. |
. |
RE: XPostFacto 3 boot problems |
September, 09, 2003 12:39 AM |
OSXGuru |
. |
The main trick with New World support would be that the method for writing to NVRAM in Mac OS 9 is entirely different, so I would need to implement that. Also, the values you need to write may be different--I'm not sure about that. So it would take some investigation. |
. |
RE: XPostFacto 3 boot problems |
September, 08, 2003 7:21 PM |
powderhaus |
. |
for some reason my computer is booting into single users mode, i did not select it in XPF... all my drives are there but it just stops after an master password error, its reading a file from my FW drive so i am assuming that it is just the firewire PCI problem thing. How hard could it be to get Newworld support? do you actualy tweak the PRAM? or have you just never tested it? Coming from me it would seem that because it already boots normaly it only needs the helper drive. |
. |
RE: XPostFacto 3 boot problems |
September, 08, 2003 6:44 PM |
OSXGuru |
. |
No, XPF doesn't support the New World macs yet. Though it is something in the back of my mind that could be possible in the future. |
. |
RE: XPostFacto 3 boot problems |
September, 08, 2003 4:05 PM |
powderhaus |
. |
well does it boot at all?because i did order one from OWC, i will not open it unless you can conferm that it does work and that it does boot off OSX almost every time. When i put 2 drives on it sometimes it is unreliable and will not always show my drive, i think i tricked it and it usually works when i daisy chain the drives, so i am fairly sure it is the PCI card, but, it works fine in OS9, so i think it is the same problem as the OSX booting thing. And can you use XPF to boot off a firewire drive on a Blue and White G3? I ask because i know it has new world ROM... |