Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: Slash on September 08, 2003, 04:14:59 PM
-
Last night while compiling an application, my Amiga crashed as GCC was writing the executable to the disk.
Of course, after I reboot, the SCSI disk started thrashing away trying to validate itself.
However, after approx 3 minutes, instead of successfully fixing the disk, or giving an error such as "Error validating HD2..." etc, the machine just reset itself with a "Software Failure". I'm sure it's a "8000 0004" in the Guru.
Anyone come across this strange error?
Disk is partitioned and formatted with the standard 3.1 ROM version of FFS and is setup as DC-FFS.
Help please :-?
-
Sorry to say.... I had this exact same problem twice in one month (when i was using my A1200 as my main machine). After that I moved to SFS and never looked back.
In my case I was able to salvage data by booting with no startup sequence, and copying as many files as I could to another HDD in the time I had before the GURU/reset. I had about less than a minute each time. Got my important stuff off it, trashed the rest.
You will need to reformat AFAIK. Sorry :/
- Paul
-
Thanks for the info Paul, glad to see I'm not the only one experiencing this problem.
I'm a bit sceptical about installing SFS, I have it on my internet cache partition but I don't trust it with as my main file system yet. Although a lot of people seem to be giving it good reports.
I did manage to get the machine to stay up without guru'ing only by booting into the OS4 beta partition I have installed. Don't know what it does, but it stops it crashing.
I have since managed to salvage my data (all 18,000 files) and reformat the disk with FFS2 so I'll see how that goes. Maybe I will move to SFS if that fails.
I was just posting to see if anyone else had this problem, and what the actual guru code of 8000 0004 was, and how it was related to my validation process.
No salvage utility I have used successfully in the past (Disksalv, Quarterback, TurboVal) could repair the problem, as you say it needed a re-format. Nasty stuff.
-
@all -
Exactly the same problem here, too. I feel your pain. I can get it to boot on my backup drive only if I disable the unvalidated partition through early startup control. Then the only thing that sees it is HDToolBox, which doesn't leave me many options. Booting with no startup-sequence causes the reset within minutes, not enough time to even start DiskSalv. :-x
Mike
-
Anyone come across this strange error?
Yes, I have, and it's a pain to fix. This is your problem:
Disk is partitioned and formatted with the standard 3.1 ROM version of FFS and is setup as DC-FFS.
Do not use Directory Caching on hard drives! It provides no benefit and (I forget most of the techy details) if it goes wrong, the validation procedure takes longer and more memory. This results in the crash you have been experiencing.
I can't remember how I fixed mine. I think it involved booting with no startup-sequence and running DiskSalv so it locked the hard drive before the validation procedure crashed. Then you can use DiskSalv's validate option to fix the disk (maybe get DiskSalv on a bootable disk instead if the validate crashes too quickly?)
Next step is to use ReOrg to convert the filesystem to plain FFS International. Using ReOrg may also help if DiskSalv doesn't validate the disk.
Chris
-
My solution is just to boot with startup-seqence disabled. This lets the drive continue validating without system files loading and slowing things down and eventually crashing.
-
Ouch!
I've had similar problems in the past, both with FFS and SFS. I think something must get corrupted in the RDB as the first thing you know about it is the visit from the Guru when trying to reboot.
The only think that seems to sort it is to make sure you do a FULL format of your partition before you install anything on it. It may take all night, but prevention is much better than cure, especially when talking about your files.
-
The only think that seems to sort it is to make sure you do a FULL format of your partition before you install anything on it. It may take all night, but prevention is much better than cure, especially when talking about your files.
I haven't had any problems since using the OS3.5 FastFileSystem, with my partitions of type FFS Intl. My current drive is in fact a Quick Formatted one that was previously used in a Windows laptop. There must be loads of random bytes in free space, but I haven't had a single failed validation with it.
Chris
-
chris wrote:
Do not use Directory Caching on hard drives! It provides no benefit and (I forget most of the techy details) if it goes wrong, the validation procedure takes longer and more memory. This results in the crash you have been experiencing.
@Chris -
While I do not doubt what you are saying for one second, can you give us a clue where to get the technical details you mentioned? Just doing routine tasks, opening directory windows in Workbench, etc., my Amiga certainly "feels" faster when using DC-FFS.
...And I've got to try what you've said, I really don't want to install 3.9 again, not when I finally had it almost running right! Anybody going to be at the NCAUG picnic this weekend want to take a gander at validating a hard drive? I'll be the one rolling up in the big-a*s 31-year-old Buick, since my last running Oldsmobile is currently up on blocks. :-D
Thanks,
Mike
-
I once formatted and partitioned a new HD without looking so exactly what I did. I wondered why my A4000 did crash even during startup-sequence about 50% of the bootups. Yes - it was the FFS-DC which was installed by default. The crashes vanished after I changed the file system to FFS.
-
There might be some benefit but can you truly say it is worth it in this case?
DO NOT USE DIRCACHE ON LARGE PARTITIONS. THE VALIDATION WILL CRASH DUE TO LACK OF MEMORY.
Thank you for calling :-)
-
Oldsmobile_Mike wrote:
While I do not doubt what you are saying for one second, can you give us a clue where to get the technical details you mentioned?
Try Olaf Barthel :-) I think I originally found out from a mailing list, it is something to do with the directory structure having to be updated in two separate places.
Chris
-
DO NOT USE DIRCACHE ON LARGE PARTITIONS. THE VALIDATION WILL CRASH DUE TO LACK OF MEMORY.
Hummm... So any idea how much memory I would need to validate a 4gb partition? :-? (Ie. if I could find someone with enough memory... I really don't want to go through that reinstall again, and am using my 3.1 backup in the meantime)
Thanks much,
Mike
-
If your 4 GB is at the beginning of the HD, under the 4 GB "barrier", ReOrg can convert on the fly - I have done it.
-
olegil wrote:
There might be some benefit but can you truly say it is worth it in this case?
DO NOT USE DIRCACHE ON LARGE PARTITIONS. THE VALIDATION WILL CRASH DUE TO LACK OF MEMORY.
Thank you for calling :-)
Dont use dircache with hard disk partitions full stop. In the past it caused me nothing but grief and actually wound up slowing down a lot when the partitions started to get full.
@zipper
That's true, but if the drive is too full or there is not enough memory the conversion can fail. It happened to me once - reorg seems to have a memory ceiling of 32 megs, I never seen it allocate more than that.
-
ReOrg allocates 64 MB for me; I have about 110 MB free.
-
Slash wrote:
I did manage to get the machine to stay up without guru'ing only by booting into the OS4 beta partition I have installed. Don't know what it does, but it stops it crashing.
A late reply, but still a huge problem!
Well, finally this has happened to me too. Unfortionately I don't have OS4 beta to allow me to copy all the files I really need. I've got about 4 Gb data but this is above the 4.3 Barrier.
Can anyone tell me how to boot my machine and tell the system NOT to validate so I can copy all my files ?
I don't mind formatting after, but just need the data. Can I mount the drive on Linux and than access the files ??
(for example)
-
HellCoder wrote:
Slash wrote:
I did manage to get the machine to stay up without guru'ing only by booting into the OS4 beta partition I have installed. Don't know what it does, but it stops it crashing.
A late reply, but still a huge problem!
Well, finally this has happened to me too. Unfortionately I don't have OS4 beta to allow me to copy all the files I really need. I've got about 4 Gb data but this is above the 4.3 Barrier.
Can anyone tell me how to boot my machine and tell the system NOT to validate so I can copy all my files ?
I don't mind formatting after, but just need the data. Can I mount the drive on Linux and than access the files ??
(for example)
I've been told that you can install WinUAE on a PC, boot into Workbench that way (using a WinUAE partition or Workbench adf), and that WinUAE should then recognize the Amiga hard drive, and allow you to copy your files off. Of course, I haven't been able to try this yet, as I can't get WinUAE to work under Vista. :-?
Good luck to you!
-
That is an old story: if FFS crashes AND the problematic file is on 6th or more level of directory structure, the revalidating process is unable to repair the partition.
6th level means something like:
DH1:News/journal/classifieds/cars/gm/camaro/problematicFile.{bleep}
That's an UNDOCUMENTED:-o problem of FFS, please, don't ask for the fonts.
-
how about trying to boot from KS1.3, without disk-validator ?
-
Well,
I've managed to get all my data on another HD. What I did was this.
1: Remove the HD from my computer.
2: Put HD into PC machine.
3: Start Knoppix from CD (I used 3.3)
4: make a dir in /tmp
5: mount the partition using affs filesystem and copy all!
Whola,... now I can format the partition again and copy it all back. Atleast I've saved all my data.
-
I have one partition in os4 that has validation problems occassionally. So I made an os4 boot cd and I just boot from cd and when the system loads, I right click on the icon for that partition and select inf, then click on the size bar to validate the partition. That always solves the problem for a few months without loss of data. Now and then I backup and reformat that partition. I suspect the problem is caused by arteffect's swapfile.
painterly