Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: Tenacious on January 06, 2011, 03:20:36 AM
-
I may have mentioned that my main computer (A3000 with two 4.3 Gig Quantum hard drives with PFS2) got sick in 2010. Basically, the partitions on the primary HD began to corrupt. Then, the partitions on the backup drive also began to corrupt within a month of the first. It is too strange that both drives would fail at the same time. Some partitions would report that some of their data was outside their boundaries. Strange! This has been a bit depressing for the last 9 months!
Today, I found another 4.3 Gig HD in the basement and it seems to be functional (What luck!). While partitioning it, I wondered what would happen if I formatted all of the available 4.3 Gigs. Would the system immediately report the drive unusable because it requires more than the available addressing range? Or, would my system happily chug along for years until I wrote a file that went beyond the theoretical address limit? Would that data overflow into another partition, corrupting both?
That's it for tonight's brain storm. I may be able to finally salvage my data to the new drive. Maybe a fresh repartitioning (inside the TRUE 4 Gig limit) of the other two supposedly bad drives will save them as well.
Yes, I know there are patches to allow HDs much larger than 4 Gigs. Being close to the limit I elected not to use them at the time. My mistake.
Any thoughts?
-
I may have mentioned that my main computer (A3000 with two 4.3 Gig Quantum hard drives with PFS2) got sick in 2010. Basically, the partitions on the primary HD began to corrupt. Then, the partitions on the backup drive also began to corrupt within a month of the first. It is too strange that both drives would fail at the same time. Some partitions would report that some of their data was outside their boundaries. Strange! This has been a bit depressing for the last 9 months!
Today, I found another 4.3 Gig HD in the basement and it seems to be functional (What luck!). While partitioning it, I wondered what would happen if I formatted all of the available 4.3 Gigs. Would the system immediately report the drive unusable because it requires more than the available addressing range? Or, would my system happily chug along for years until I wrote a file that went beyond the theoretical address limit? Would that data overflow into another partition, corrupting both?
That's it for tonight's brain storm. I may be able to finally salvage my data to the new drive. Maybe a fresh repartitioning (inside the TRUE 4 Gig limit) of the other two supposedly bad drives will save them as well.
Any thoughts?
Using something like the FastATA MKIII hardware or the 4xEIDE hardware they spilt your drives automatically into lots of 4GB partions, but upon actually checking them they are always just a few K under 4GB (never 4GB exactly).
Seems when you try to do this due to the actual way the filesystems work if they were to actually try and write data all the way up to the very last byte of an exact 4GB partition some of the data ends up being written to the previous partition if there is one or if not it overwrites another part of the current partition causing all sorts of data loss... :(
The applies to the max size of a FFS partition which is supposed to be 2GB but is always just a few K less than 2GB to prevent this overwriting and corruption of other partitions... :)
That's the simplest explanation I can give you cos it's really a bit more complicated that this... :)
-
Using something like the FastATA MKIII hardware or the 4xEIDE hardware they spilt your drives automatically into lots of 4GB partions, but upon actually checking them they are always just a few K under 4GB (never 4GB exactly).
What you say rings a bell with me. If I understand correctly, there was a natural 4 GB limit to total hard drive size (regardless of partitions) until OS3.5 or 3.9 released a patch (I didn't install these) to allow larger drives. 4 GB was a lot of HD when 3.5 was released, grin.
Seems when you try to do this due to the actual way the filesystems work if they were to actually try and write data all the way up to the very last byte of an exact 4GB partition some of the data ends up being written to the previous partition if there is one or if not it overwrites another part of the current partition causing all sorts of data loss... :(
It only occurred to me tonight that data written past the address limit might cycle through zero to an address in my first partition.
I feel like a boob. If this turns out to be true (I haven't yet confirmed my partition sizes), the solution will be easy enough.
It's late here. It must be very early morning in Scotland. You do sleep, right? ;)
-
What you say rings a bell with me. If I understand correctly, there was a natural 4 GB limit to total hard drive size (regardless of partitions) until OS3.5 or 3.9 released a patch (I didn't install these) to allow larger drives. 4 GB was a lot of HD when 3.5 was released, grin.
It only occurred to me tonight that data written past the address limit might cycle through zero to an address in my first partition.
I feel like a boob. If this turns out to be true (I haven't yet confirmed my partition sizes), the solution will be easy enough.
It's late here. It must be very early morning in Scotland. You do sleep, right? ;)
Not very much it's only 4:10 am here and I'm still wide awake... :lol:
As I said, most folk would reasonably expect 4GB to mean exactly 4GB but in AmigaLand it's not as simple as that, even if you managed to partition an HD to exactly 4GB the file system when your HD is just about full would end up writing data to a lower partition or as you say cycle though to zero... :)
-
I'm still looking around for a number. One site suggested the limit might be in the 3.7 to 3.9 GB range, depending on filesystem used.
Appearently, Thomas wrote a utility to check the limit. Maybe that's my answer. I need to read up more on PFS.
I'm going to turn in tonight.
Thanks for the help.
-
I'm still looking around for a number. One site suggested the limit might be in the 3.7 to 3.9 GB range, depending on filesystem used.
Appearently, Thomas wrote a utility to check the limit. Maybe that's my answer. I need to read up more on PFS.
I'm going to turn in tonight.
Thanks for the help.
The number doesn't have to be exact, just make sure it's slightly under the exact 4GB size... :)
(gawd... I'm still wide awake and it's now 7:01am... :))
-
Using my calculator - 4194303 ?
-
The exact limit is 2^32 bytes a.k.a. 4,294,967,296 bytes or 8,388,608 sectors.
To be sure to stay below the limit, you can change the geometry of the hdd, for example to
cylinders = 8322
heads = 16
blocks per track = 63
blocks per cylinder = 1008
This results in 8,388,576 sectors or 4,294,950,912 bytes.
There is fixhddsize (http://aminet.net/package/disk/misc/fixhddsize) program on Aminet which can do this for you.
Or you can change from PFS2 to PFS2ds which allows you to use 8 GB. (To be precise, 16383 cylinders, 16 heads, 63 sectors which results in 16,514,064 sectors or 8,455,200,768 bytes).
Regarding corrupted partitions, you should upgrade to PFS3. I never had a corrupted partition with PFS3 while it happened with PFS2 every now and then.
-
The exact limit is 2^32 bytes a.k.a. 4,294,967,296 bytes or 8,388,608 sectors.
To be sure to stay below the limit, you can change the geometry of the hdd, for example to
cylinders = 8322
heads = 16
blocks per track = 63
blocks per cylinder = 1008
This results in 8,388,576 sectors or 4,294,950,912 bytes.
There is fixhddsize (http://aminet.net/package/disk/misc/fixhddsize) program on Aminet which can do this for you.
Or you can change from PFS2 to PFS2ds which allows you to use 8 GB. (To be precise, 16383 cylinders, 16 heads, 63 sectors which results in 16,514,064 sectors or 8,455,200,768 bytes).
Thanks for the advise and the link, Thomas. I will use fixhddsize on the new drive. It should be able to check the old drives as well?
I thought it was so clever to have two HDs in the A3000, one as a mirror of the other. Now, I'm considering having one in an external SCSI box, so it can act as a backup to all of my Amigas.
Regarding corrupted partitions, you should upgrade to PFS3. I never had a corrupted partition with PFS3 while it happened with PFS2 every now and then.
Yea, well, I bet I'm not the only member of this board impatiently waiting for PFS3 to become available again!
-
I need to correct myself as I didn't notice you are using SCSI. The 8 GB limit mentioned above only applies to IDE. With SCSI + PFS2ds you can actually use hard drives up to 2 TB with partitions up to 104 GB :-)
-
I need to correct myself as I didn't notice you are using SCSI. The 8 GB limit mentioned above only applies to IDE. With SCSI + PFS2ds you can actually use hard drives up to 2 TB with partitions up to 104 GB :-)
Thanks again Thomas. I like your FixHHDsize program, added it to my system drawer.
@ all
I have been re-partitioning and formatting (with PFS2) the one good 4.3GB drive most of the weekend. I found at the end last night I had my desired partitions (all formatted and ready to go) plus a strange ghost partition. I say "ghost" because the intended partitions accounted for the whole capacity of the drive. Where did this extra partition come from? It had the same name as one of the others with a few extra characters at the end. It would even format. I've never noticed this before. If a user does not delete all old partitions, can the new and old overlap and corrupt each other? Considering my current circumstance, I re-did the whole drive. No ghosts so far. ;)
I've been rounding up all of my old SCSI HDs and trying to revive them. I have a suspicion that several are locked up as a result of corrupted parameters, not corrupted disks. Aren't some fundamental parameters about the drive stored on a EEPROM? Most manufacturers have jumpers marked "reserved" or "factory use". I wonder if some of these are for resetting the EEPROM to factory defaults. Anyone tried this?
-
@^%?}#!
The new drive just disappeared off the system! HDtoolBox reports "Sorry, this type of device cannot be handled by HDtoolBox". Same as one my other "dead" drives.
Can a person change partition too many times and corrupt the EEPROM?
I seriously need to find a master reset!
-
Partitions are not stored in an EEPROM. The parttion table is written to the first few blocks of the drive. The drive's firmware does not know about partitions. Partitions are only known to the operating system.
Could it be that it's too hot inside your Amiga?
-
Could it be that it's too hot inside your Amiga?
The drive has been sitting in the open air. Nothing felt hot.
When a drive is new, where is the initial information (number of cylinders, heads, block/track, etc) stored? This is the info that can't be read anymore.
-
Im a little confused here. I have partitions well beyond 4 gig without issues. The 4gig limit only applies to the boot partition, or is this what you mean ? (or perhaps it's an os3.1 issue that Ive forgotten about after having used 3.9 for so long?)
-
I'm probably more than a little confused. ;) I beleive there was a problem (before OS3.5 and also PFS2) addressing drives larger than 4GB. I'm beginning to beleive this was a limitation of FFS. I don't know if the limits were the same for SCSI and IDE.
It doesn't seem now that this was my original problem. I'm left at the moment with several possible causes: PFS2 has a bug, my poor understanding and setting of drive parameters (mask, maxtranfer, etc), too many small files (like those generated by iBrowse, Yam, Palm synchronization, etc) that are constantly being revised, drives do have a finite lifespan.
-
Up to 3.1 you're generally restricted to 4 GB due to the API - read: Do not partition beyond 4 GiB (=2^23 sectors)! Any write access beyond 4 GB will 'wrap around' and overwrite lower sectors (=destroy data). You can very well destroy RDB information where partitioning data is stored.
From 3.5 onwards (or with various other patches), you'll gain 64 bit addressing capability. There are some issues: http://www.youngmonkey.ca/nose/articles/NewTekniques_9810/AmigaInMotion.html (http://www.youngmonkey.ca/nose/articles/NewTekniques_9810/AmigaInMotion.html), esp. when your controller doesn't support 64 bit access nor is it automatically patched by 3.5+. Your boot partition must not exceed 4 GB as it is possibly written to during boot before patches are applied.
There are various further restrictions due to hardware architecture:
- A1200/4000/600 IDE is limited to 2 TB (32 bit block address)
- many SCSI controller (e.g. WD33C93 based ones) are limited to 2 TB (32 bit block address)
Some of these can be worked around.
-
There are various further restrictions due to hardware architecture:
- A1200/4000/600 IDE is limited to 2 GB (32 bit block address)
- many SCSI controller (e.g. WD33C93 based ones) are limited to 2 GB (32 bit block address)
Some of these can be worked around.
This should read 2 TB in both cases.
I'm probably more than a little confused. ;) I beleive there was a problem (before OS3.5 and also PFS2) addressing drives larger than 4GB. I'm beginning to beleive this was a limitation of FFS. I don't know if the limits were the same for SCSI and IDE.
It doesn't seem now that this was my original problem. I'm left at the moment with several possible causes: PFS2 has a bug, my poor understanding and setting of drive parameters (mask, maxtranfer, etc), too many small files (like those generated by iBrowse, Yam, Palm synchronization, etc) that are constantly being revised, drives do have a finite lifespan.
The problem is in the interface between file system and device driver. It is not specific to 3.1 and below and it is not magically fixed in 3.5 and above. You need to find a combination of file system and device driver which works together in order to bypass the 4GB limit.
The other problem is that the so-called standard was made without talking to the other parties which already had developed a fix. Therefore now there are two "standards" which can be used to bypass the limit, but they are not compatible.
The standard which came with 3.5 is called NSD.
The other command set invented by a group of third-party hardware manufacturers in the time before 3.5 came out is called TD64.
There is a third possibility called Direct-SCSI a.k.a. HD_SCSICMD which has some other limits.
Now fact is that PFS2/3 support TD64 but not NSD, so they do not work with OS 3.5's device driver for the internal IDE controllers on Amiga motherboards. For large hard drives, that is. Hard drives smaller than 4 GB can be accessed by every file system.
-
This should read 2 TB in both cases.
Err - yes. I've fixed my post, sorry about that.