CF / SD and large drives FAQ

[mark_k]

Ah that's the same value as given by HDToolBox 40.4; that probably uses the same READ CAPACITY command as SCSIQuery. Maybe it'd be best to use that lower figure for safety. I don't know whether scsi.device in the A600 Kickstart ROM would prevent access to blocks above 15,523,200.

Out of interest, if you connect the CF card to your PC via a card reader, what is the number of blocks shown as? I'm not familiar with how to show that figure. You could experiment with diskpart or the System control panel.

[ShK]

Max Logical block addressing: 15,523,839

[mark_k]

So Windows agrees with HDToolBox 44.23. Max LBA 15,523,839 means there are 15,523,840 blocks (numbered from 0 to 15,523,839 inclusive).

Hopefully someone who knows more about the details than me can post to clarify this. Does the A600 ROM scsi.device fail or prevent SCSI-direct access to blocks beyond the C/H/S fake geometry it returns for READ CAPACITY? That is, beyond the last whole multiple of 16×63 sectors.

If it does, then when partitioning the card use the HDToolBox 40.4 (or equivalent) fake geometry which limits the number of sectors to a multiple of 16×63.

[ShK]

I made with HDInstTools 12128×5×256 = 15,523,840 fake geometry and the DH2 partition at the end of the drive. Then I created Dummy File what is about same size as the partition. Then I copy that file to that last partition with 'FilecopyVerifyRetry'-program. No errors. This is all under the WinUAE and PFS3 AIO.

http://s24.postimg.org/c5bud2g79/HDInst_Tools.png
http://s24.postimg.org/srtem5951/DH2.png
http://s24.postimg.org/r2kbe2tfp/successfully.png

[mark_k]

You'd need to test with your A600, not WinUAE I think. Assuming you want to use the card with your A600 at some point of course.

[ShK]

Yes, I put the memory card to the A600 and I did the same test (from DH1 to DH2). All went well. SCSIQuery still reports 15,523,200 capacity for 15,523,840 blocks geometry.

http://s29.postimg.org/ph7h9lzyf/A600.png

[ShK]

HDToolBox 40.4 gives warning on the real A600 when I try give (12128×5×256) 15,523,840 blocks geometry:

[ShK]

I did a little comparison with the 2GB, 4GB and 8GB -cards. I tested two different adapters on Amiga and the same adapters on the PC, but there was no difference. I also tested the two different Amigas, A600 and A1200 (In both KS3.1). All the cards were emptied with diskpart before the test. Test programs were on the Amiga SCSIQuery and on the Win7x64 SeaTools for Windows.

http://s10.postimg.org/caw506qbd/2014_03_11_1426.jpg

	8GB		4GB		2GB
WinUAE	15,523,200	7,626,528	3,841,488
A600	15,523,200	7,626,528	3,841,488
A1200	15,523,200	7,626,528	3,841,488
Win7x64	15,523,839	7,626,751	3,842,047

[mark_k]

SeaTools reports the highest logical block address (i.e. number of blocks - 1), so you need to add 1 to each of the Windows numbers:
8GB: 15,523,840
4GB: 7,626,752
2GB: 3,842,048

For the WinUAE config you were using, was that with an emulated A600/1200 IDE controller?

scsi.device in the Kickstart ROM only uses C/H/S addressing. Windows and probably the later scsi.device which comes with AmigaOS 3.9 — which isn't available at boot time — use LBA addressing.

That's why the reported capacity of the cards is limited to a multiple of 16×63 sectors. For legacy/backward compatibility reasons, IDE devices effectively report two capacities, a C/H/S capacity and an LBA capacity. What I'm not clear on, is whether accessing past the C/H/S reported capacity works with the ROM scsi.device.

Try this with the 2GB card:
- Manually set the geometry to cover the entire 3,842,048 blocks. For example Cylinders = 7504, Heads = 4, Blocks per track = 128. Click Continue at the HDToolBox warning and create an FFS partition at the end of the card (HighCyl=7503).
- After rebooting, format the partition and copy files to it so the partition is over 50% full. (The way FFS fills up partitions, it starts in the middle and works towards the end. So you only need fill up more than half of a partition in order to cause the very end blocks to be written to.)
- Compare the source and copied files. I recommend Cloanto's DirDiff program.

[ShK]

Originally Posted by mark_k

SeaTools reports the highest logical block address (i.e. number of blocks - 1), so you need to add 1 to each of the Windows numbers:
8GB: 15,523,840
4GB: 7,626,752
2GB: 3,842,048

Yes, they are Zero-based numbering. I wrote it just as a SeaTools for Windows indicated.

Originally Posted by mark_k

For the WinUAE config you were using, was that with an emulated A600/1200 IDE controller?

Yes, It is the Amiga 1200 emulated IDE controller:

http://s17.postimg.org/cwvv2gfgf/Adv_Chipset.png
http://s17.postimg.org/4p348y2bj/ROM.png
http://s17.postimg.org/5jgh3hven/Floppy_Drives.png
http://s17.postimg.org/f2q5wyiwv/IDE0.png

Please note that at the last picture WinUAE gives the same LBA capacity as the SeaTools for Windows (3,842,048) !

Originally Posted by mark_k

scsi.device in the Kickstart ROM only uses C/H/S addressing. Windows and probably the later scsi.device which comes with AmigaOS 3.9 — which isn't available at boot time — use LBA addressing.

That's why the reported capacity of the cards is limited to a multiple of 16×63 sectors. For legacy/backward compatibility reasons, IDE devices effectively report two capacities, a C/H/S capacity and an LBA capacity. What I'm not clear on, is whether accessing past the C/H/S reported capacity works with the ROM scsi.device.

Version command gives scsi.device 40.12 for A600 KS3.1 (v40.68).

Originally Posted by mark_k

Try this with the 2GB card:
- Manually set the geometry to cover the entire 3,842,048 blocks. For example Cylinders = 7504, Heads = 4, Blocks per track = 128. Click Continue at the HDToolBox warning and create an FFS partition at the end of the card (HighCyl=7503).
- After rebooting, format the partition and copy files to it so the partition is over 50% full. (The way FFS fills up partitions, it starts in the middle and works towards the end. So you only need fill up more than half of a partition in order to cause the very end blocks to be written to.)
- Compare the source and copied files. I recommend Cloanto's DirDiff program.

The card is cleaned with the DiskPart.:
http://s17.postimg.org/lqmrjk2f3/diskpart.png

Partitioning for the full LBA capacity has been done with a real Amiga 600:
http://s17.postimg.org/vhk39j0vz/P_Mar15_183934.png
http://s17.postimg.org/9h3qswi7z/P_Mar15_183959.png
http://s17.postimg.org/xnee414cf/P_Mar15_184115.png
http://s17.postimg.org/jenydg4lb/P_Mar15_184146.png
http://s17.postimg.org/6qz09gawv/P_Mar15_184324.png

Then let's try to copy two Aminet CD-ROM directories from the PCMCIA memory card to the 2GB card hard disk with the DirDiff.
http://s17.postimg.org/k95wlqn27/P_Mar15_184834.png

Copying stops with an error when the disk is 50% full:
http://s17.postimg.org/ltftxvku7/P_Mar15_224250.png
http://s17.postimg.org/wcb89ay4f/P_Mar15_224615.png

The new partition to the C/H/S capacity with the default settings of the HDToolBox:
http://s17.postimg.org/mb6dzzl1b/P_Mar15_225242.png
http://s17.postimg.org/7o5fibmf3/P_Mar15_225344.png
http://s17.postimg.org/cvqims30f/P_Mar15_225448.png

And let's try again to copy two Aminet CD-ROM directories from the PCMCIA memory card to the 2GB card hard disk:
http://s17.postimg.org/poemmpem7/P_Mar15_230131.png
http://s17.postimg.org/g54xt8r3z/P_Mar16_054623.png

No error messages. Let's do an another test pass with the DirDiff:

http://s17.postimg.org/6w7cwv7lr/P_Mar16_110544.png
http://s17.postimg.org/5xmegu4vz/P_Mar16_065911.png
http://s17.postimg.org/famzky8n3/P_Mar16_110829.png

Directory 1 = "PCMCIA:2CD"
Directory 2 = "IDE:2CD"

Mode: BYTEDIFF EXACT DATES


PHASE I (searching directory 2 for items with no match in directory 1)


PHASE II (processing items in directory 1)


Errors/unresolved differences: 0
Medium errors: 0

Directories Files

Items examined: 910 35018
Items modified: 0 0
Items created : 0 0
Items deleted : 0 0

DirDiff complete (the directories are identical)

The disk is 66% full and all good. So, the full LBA capacity didn't work!

[mark_k]

Originally Posted by ShK

So, the full LBA capacity didn't work!

Thanks for doing those tests. That's very handy to know for anyone preparing a CF card on a PC for use with a real Amiga.

In your WinUAE/A1200 config, were you using any patched/updated version of scsi.device, or just the ROM built-in version? If the latter, that suggests that there's something about WinUAE's IDE drive emulation which doesn't match real hardware.

[ShK]

Originally Posted by mark_k

Thanks for doing those tests. That's very handy to know for anyone preparing a CF card on a PC for use with a real Amiga.

And you brought up some interesting information, thank you!

Originally Posted by mark_k

In your WinUAE/A1200 config, were you using any patched/updated version of scsi.device, or just the ROM built-in version? If the latter, that suggests that there's something about WinUAE's IDE drive emulation which doesn't match real hardware.

On WinUAE I used the original Kickstart ROM-image and Workbench 3.1 disks which are from the Petro Tyschtschenko NOS (New Old Stock) Amiga 1200 Magic Pack.

http://eab.abime.net/album.php?album...pictureid=2604

Workbench 3.1 floppy disks are read by Kryoflux.

http://img838.imageshack.us/img838/987/d6i9.png

Used scsi.device is version 40.12 from original KS ROM v3.1 (A1200) rev 40.68 (512k) [391773-01/391774-01].

WinUAE and real Amiga gives same C/H/S capacity results with the SCSIQuery:

        8GB         4GB        2GB
WinUAE  15,523,200  7,626,528  3,841,488
A600    15,523,200  7,626,528  3,841,488
A1200   15,523,200  7,626,528  3,841,488
Win7x64 15,523,840  7,626,752  3,842,048

WinUAE itself and SeaTools for Windows gives also same results. Different memory cards in images, but it is an LBA capacity:

http://s17.postimg.org/f2q5wyiwv/IDE0.png
http://s7.postimg.org/n30bwuvm3/LBA.png (Zero-based numbering)

[mark_k]

Just one thing is confusing me now. You posted earlier about HDToolBox 44.23 showing the LBA capacity. Were you using a slightly different WinUAE config then? Maybe the card was connected as a WinUAE native drive (uaehf.device) instead of emulated A600/1200 IDE (scsi.device)? That would explain why HDToolBox 44.23 was able to detect the LBA capacity.

[fgh]

IIRC Winuae uses size given by windows for both IDE0 and UAEHF.

Edit : That's why HDToolbox 44.23 will know the LBA size even when scsi.device only supports CHS

[fgh]

Interesting observation though.
I guess a drive or card partitioned on a real amiga with an updated scsi.device and the new hdtoolbox will also have same issue if used on a system with unpatched scsi.device..

[Solo761]

I had "fun" this weekend with Kingston 16GB 266x CF card. It is original, bought in store locally, not an ebay fake.

I prepped CWB lite 3.1 (this card is for my A600) in WinUAE, set doobrey's 44.20 scsi.device, connected CF via USB reader as usual. I used HDToolBox that comes as standard with CWB in tools folder. I used it as well with Transcend 16GB 133x CF card I set up for my A1200 and it worked without problems.

This time however... Everything looked fine, partitioned it, set first partition as 500MB FFS, rest space as one PFS3 partition, so far so good, set MaxTransferRate to 0x1fe00. Saved config, rebooted, also looked fine, formatted both partitions and when I tried to copy WB to it half way done got some error, I think it was something about blocks, and then partitions disappeared from WB. Rebooted and partitions are still missing. But in HDToolBox everything looks fine. Back to PC, formatted it again, put some files onto it to be sure to overwrite what amiga left (I had cases where I formatted it on PC, cleaned in diskpart, and when I connected it to amiga and created partitions as they were old ones would reappear although I didn't even format new ones, so I guess PC format it doesn't wire to the area amiga uses by default), then diskpart clean and back to amiga.

This time I only made 500MB DH0 partition, again everything looked fine, but this time I got an error while formatting, and again partition dissapeared... Changed transfer rate to lower value and partition reappeared, tried to format and again same story. To cut long story short, I tried more transfer rate values but it didn't help.

Then I tried with HDInstTools 6.9 from aminet (hdinst.lha) and it read drive geometry differently, as someone said, looked "fake". But as last try I tried to go with it and this time there were no problems. Partition was created, formatted, I copied WB files to it and everything worked perfectly. Made one large PFS3 partition from rest of the space and also no problems.

This is what HDToolBox said

Cylinders: 30167
Heads: 16
Blocks per Track: 63
Blocks per Cylinder: 1008

similar values as with transcend one, cylinders number was different, apparently kingstone CF has cca 500MB less.

While HDInstTools reported this

Cylinders: 104496
Heads: 291
Blocks per Track: 1
Blocks per Cylinder: 291 (this isn't displayed in HDInstTools, I just put it here for consistency, it's Heads * Blocks per Track)

For some reason with this values it works without problems, although it can be it's because HDInstTools and it works because of that. Too bad I didn't try manually entering values from HDToolBox (max disk size in both cases), but this worked and after a day of banging my head and reformating I just wanted to be done with it .

I had some "features" with HDInstTools. I couldn't load PFS3 aio with HDInstTools 6.9, for some reason it would read it as FastFileSystem type and then when I tried to format it it opened FFS type format dialog, not PFS3 one. In desperation I did a no-no and used HDToolBox to install PFS3 into RDB (or where ever it goes) and then created partition and set partition type with HDInstTools. So, partition creation with HDInstTools and PFS3aio install with HDToolBox and I finally had it working.

I formatted them regularly, from WorkBench. This is another "feature" HDInstTools had in my case . I did all of this from CWB image in WinUAE, it was DH0, first partition I created with HDInstTools I also named DH0, I tried formatting it from HDInstTools but instead of "my" DH0 it formatted the one with CWB . At first I thought I messed up something so I set CWB again, tried again, and same thing happened so that was when I kinda figured what's the issue here .

[ShK]

Originally Posted by mark_k

Just one thing is confusing me now. You posted earlier about HDToolBox 44.23 showing the LBA capacity. Were you using a slightly different WinUAE config then? Maybe the card was connected as a WinUAE native drive (uaehf.device) instead of emulated A600/1200 IDE (scsi.device)? That would explain why HDToolBox 44.23 was able to detect the LBA capacity.

No, it depends on the scsi.device -version. This table has been made under WinUAE with 2GB memory card:

AmigaOS      scsi.device  HDToolBox   Capacity
40.42(3.1)   40.12        40.4        3,841,488  OS31.png
44.2 (3.5)   43.30        44.18       3,842,048  OS35.png
45.1 (3.9)   43.35        44.23       3,842,048  OS39.png

44.2 (3.5)   40.12        44.18       3,841,488  OS35+scsi.device_disabled.png
45.1 (3.9)   40.12        44.23       3,841,488  OS39+scsi.device_disabled.png

WinUAE itself                         3,842,048  WinUAE.png

[fgh]

So you didn't try hdtoolbox 44.18 or 44.23 with scsi.device 40.12?

[ShK]

Originally Posted by fgh

So you didn't try hdtoolbox 44.18 or 44.23 with scsi.device 40.12?

No I didn't, because SCSIQuery shows the capacity results for all scsi.device versions. You'll notice that from the screenshots.

e. Ok, I updated the table. I run OS3.9 with SetPatch DISABLEROMMODULES "" SKIPROMUPDATES scsi.device -command.

[ShK]

I tested the HDToolbox v40.4 (AmigaOS3.1) in AmigaOS3.9 (scsi.device v43.35):

scsi.device v43.35 sees LBA 3,842.048 capacity, but old HDToolbox v40.4 only 3,841,488 Cylinder-head-sector (C/H/S or CHS) capacity.

OS39+HDToolbox_v40.4.png

Cylinders * Heads * Sectors = total number of blocks a.k.a. highest logical block address (LBA)
3811 * 16 * 63 = 3,841,488