PFS3 or PFS3 SCSI Direct - Page 3

Toni Wilen · 21 October 2012, 19:00

Really weird.. SCSI Read Capacity returns correct and full drive size.

Are you 100% sure you have standard ROM? Something strange is going on..

mfilos · 21 October 2012, 21:22

I have an 27C800 with dual Kickstart in it.
2x256KB of Kickstart 1.3 and 1x512KB of Kickstart 3.1.

In these picture I booted the stock Kickstart 3.1 without Startup-Sequence alas NO usage of ACATune to kick anything else as I normally do

I'm pretty sure that with the original PFS3 distribution that I had in the past (PFS3+Patch = 18.5) I wasn't seeing the second partition without the usage of newer scsi.device
For some reason your version of PFS3 recognizes the 2nd partition even with the stock scsi.device of 3.1!!!

Toni Wilen · 22 October 2012, 08:46

Quote:

Originally Posted by mfilos

I have an 27C800 with dual Kickstart in it.
2x256KB of Kickstart 1.3 and 1x512KB of Kickstart 3.1.

In these picture I booted the stock Kickstart 3.1 without Startup-Sequence alas NO usage of ACATune to kick anything else as I normally do

I'm pretty sure that with the original PFS3 distribution that I had in the past (PFS3+Patch = 18.5) I wasn't seeing the second partition without the usage of newer scsi.device
For some reason your version of PFS3 recognizes the 2nd partition even with the stock scsi.device of 3.1!!!

Official PFS3DS should also work with second partition, it uses exact same SCSI direct code. Normal PFS3 won't and can't work, it requires TD64.

What does version full scsi.device say?

fgh · 24 October 2012, 12:36

Good work guys!

Finally back home, I tried the debug version on a stock a600 with KS37.300.
A 16GB SD card was partitoned as (approx.) 2GB/8GB/5GB in WinUAE using a600 quickstart and booting from WB3.1 install disk.
Thereafter WB2.1 was installed from scratch onto DH0.

Results with scsi.device 37.34: Test 1=1, and tests 4=0, 3=0, 2=1 on both DH1 and DH2.
Results with scsi.device 44.2: Tests 4=1 on both DH1 and DH2.

Mfilos is right - I can also access DH2 using 37.34. (My DH2 even resides completely outside the first 8GB.)
Also tried copying and reassigning Dopus to DH2, and it runs as it should.
This also works with scsi.device 44.2.

Have a look at the attached screenshots..
(DH3 device was supposed to be DH2, I initially had one more partition but changed my mind as there would be sh*itloads of debug button presses on each reboot.)

Do you want me to try with some other version of PFS3 or scsi.device?

Edit: I did try scsi.device 43.47 by Cosmos, but my A600 just kept rebooting.
I wasn't even able to enter the early startup menu. I used the a600-1200 version, but I guess it needs ks3.x.

Toni Wilen · 24 October 2012, 12:53

Quote:

Originally Posted by fgh

Mfilos is right - I can also access DH2 using 37.34. (My DH2 even resides completely outside the first 8GB.)
Also tried copying and reassigning Dopus to DH2, and it runs as it should.
This also works with scsi.device 44.2.

Thanks, this 100% confirms that at least A600 scsi.device _does_ support >8G with Direct SCSI, at least when some unknown condition is true..

I have only done A1200 scsi.device tests, perhaps A600 is different enough. (If I remember correctly, A600 scsi.device uses LBA addressing and A1200 uses CHS for some strange reason which may explain the difference)

Quote:

Do you want me to try with some other version of PFS3 or scsi.device?

Try confirming if official PFS3DS (aminet, non-ks13 version) also mounts DH2 correctly. It should, if not, this will get even more strange..

mfilos · 24 October 2012, 12:58

I'm almost certain that my PFS3ds version (030) wasn't mounting my second (and outside 8GB) partition without using an updated scsi.device
With your version Toni, it even shows in stock 3.1 scsi.device.

I can tell for sure once I get home since @ work I only have WinUAE

fgh · 24 October 2012, 15:06

I'll let mfilos check the non-ks13 version then, since his a600 has both 030 and ks3.1.

Toni: Will you submit your changes to the pfs3 sourceforge project?

Toni Wilen · 24 October 2012, 17:27

Quote:

Originally Posted by mfilos

I can tell for sure once I get home since @ work I only have WinUAE

It won't work in WinUAE here. A600 or A1200 IDE, I don't even understand how it can work because old-style size query always returns 16383/63/16 "geometry" which is about 8G...

Do you see following messages in winuaelog.txt when it works:

Code:

IDE0 recalibrate
IDE0 initialize drive parameters, CYL=16383,SPT=63,HEAD=16
IDE0 drive multiple mode = 16

Quote:

Toni: Will you submit your changes to the pfs3

I am not sure. I have commited some generic updates already but I am not sure if m68k-only special features (especially 1.x support) should included in standard PFS3 tree.

fgh · 24 October 2012, 21:45

I dug out my ACA630 and used ACAtune maprom to boot into kick31. (scsi.device 40.5)
Cleaned the 16GB SD, repartitioned with PFS3-020DS v5.3 off aminet, and placed a 4GB partition at the very end.

The partition shows up and I can copy and run stuff off it!

Edit:

Quote:

Originally Posted by Toni Wilen

I am not sure. I have commited some generic updates already but I am not sure if m68k-only special features (especially 1.x support) should included in standard PFS3 tree.

Fair enough, but I'm sure they won't mind a unified TD64/NSD/SCSI version

Toni Wilen · 24 October 2012, 21:50

Quote:

Originally Posted by fgh

I dug out my ACA630 and used ACAtune maprom to boot into kick31. (scsi.device 40.5)
Cleaned the 16GB SD completely, repartitioned with PFS3-020DS v5.3 off aminet, and placed a 4GB partition at the very end.

The partition shows up and I can copy and run stuff off it!

Just to confirm that it is correctly working (and not some weird wraparound happening), does it mount correctly on WinUAE when using UAE controller (not IDE)?

And finally, does it work with identical WinUAE A600 config using IDE0? I hope it _does not_ work because it would explain a lot

fgh · 24 October 2012, 22:03

I seems to work well in WinUAE with UAE controller, no errors and can launch dopus from the partition.

With quickstart a600, changing to 68030 but staying at ks37.300 rom the partition is shown as NDOS

Toni Wilen · 24 October 2012, 22:19

Perfect

Only logical explanation is that some cards or adapters ignore ATA spec or there is some special CF/SD/whatever spec that allows it.

ATA spec says that drive bigger than 16515072 blocks must return static 16383/15/63 CHS (cylinders/heads/sectors, ancient way to calculate disk size and address its blocks) value. This is the value that IDE scsi.device uses to detect the drive size. (16383*15*63*512 = ~7.7G)

If drive would return more "correct" CHS values, for example 16383/15/127 for 16G drive, it should be fully accessible to IDE scsi.device in Direct SCSI mode.

I'll write short IDE identify device debugging program to confirm this..

fgh · 25 October 2012, 01:06

I have software that reads the ATA identify device replies, but I cannot find my 3.5-2.5" IDE adapter anymore, so I cannot test the 16GB SD I used above. (It doesn't work over USB)
But I do have a 16GB CF card, and already saved the replies for that card before. CHS given is like you said 3FFF/10/3F (16383/16/63).

So I gave this card a go with PFS3-020DS v5.3, and voila - this one shows up with DH2 as NDOS. DH1, a small PFS partition within first 4GB works well.
With scsi.device 44.2 DH2 also works well.

Guess the chinese IDE-SD adapter could be giving the full capacity in the CHS reply then.. I'll keep looking for the 3.5-2.5 adapter and let you know if I find it.

See attachments for the identify device replies for my 16GB CF card and another SD card in the IDE-SD adapter (Only 8GB though, so doesn't prove anything regarding CHS)

fgh · 25 October 2012, 10:52

Found the 3.5-2.5 adapter! (still connected inside the PC

)

Indeed, the IDE-SD adapter gives CHS values 7A22/10/3F (31266/16/63), so I guess that explains it. Attached the full replies..

Toni Wilen · 25 October 2012, 11:03

Excellent! Problem solved.

Now we only need to find max supported values, too large values will most likely overflow in some calculations.

Also I need to find some method that allows WinUAE to query correct CHS values from mounted harddrive/cf/sd/etc.

fgh · 25 October 2012, 11:26

I just confirmed what mfilos was probably experiencing as well.
On the 16GB CF card with 'ATA correct' ~7.8GB CHS values, PFS3DS v5.3 and scsi.device 40.5 still lets you see and work with partitions that start inside the first 8gb of the drive, but not partitions that reside completely outside it.
(Tried first with a 9GB partition starting around 3GB position, and then an 6GB partition starting around 6gb position.)

mfilos · 25 October 2012, 12:10

Exactly! Nice job there fgh!

Toni Wilen · 25 October 2012, 17:45

Do you see "correct" or "incorrect" geometry in winuae's log (Click "Add Harddrive" to generate disk log info)

You should see something like this:

device id string: 'Generic- Compact Flash 1.00 20060413092100000'
BPS=512 Cyls=996 TPC=255 SPT=63 MediaType=11
device size 8195604480 (0x1e87f0000) bytes

(I am quite sure this data isn't same as identify device cylinders value but better confirm it first..)

I don't currently have >8G CF cards or SD-IDE adapters.

Unfortunately Windows IOCTL_ATA_PASS_THROUGH (which can be used to send any ATA command) does not work with CF USB adapters (at least it won't work with my adapter). It returns 50 = ERROR_NOT_SUPPORTED.

It seems to be very difficult to get exact same size on real IDE based Amiga and WinUAE. Stupid "buggy" adapters that return "illegal" ATA cylinder size information

fgh · 25 October 2012, 18:19

Quote:

Originally Posted by WinUAE.log

47-734 [0 000x000]: WARNING: ':Generic- SD/MMC 1.00' FSCTL_ALLOW_EXTENDED_DASD_IO returned 1
47-734 [0 000x000]: hd accepted (rdb detected at block 0)
47-734 [0 000x000]: HDF ':Generic- SD/MMC 1.00' 04580B20 opened, size=15758336K mode=0 empty=1
47-734 [0 000x000]: GAYLE_IDE0 ':Generic- SD/MMC 1.00', CHS=16383,16,63. 15389M. LBA48=0

That's using the USB adapter. Should I try over IDE?

Toni Wilen · 25 October 2012, 18:22

Quote:

Originally Posted by fgh

That's using the USB adapter. Should I try over IDE?

Thats wrong part of the log. This shows IDE emulation generated CHS, not CHS queried from Windows.

Run with -log parameter, click add harddrive and the data should appear in log window.

24 October 2012, 12:36	#44
fgh Registered User Join Date: Dec 2010 Location: Norway Posts: 819	Good work guys! Finally back home, I tried the debug version on a stock a600 with KS37.300. A 16GB SD card was partitoned as (approx.) 2GB/8GB/5GB in WinUAE using a600 quickstart and booting from WB3.1 install disk. Thereafter WB2.1 was installed from scratch onto DH0. Results with scsi.device 37.34: Test 1=1, and tests 4=0, 3=0, 2=1 on both DH1 and DH2. Results with scsi.device 44.2: Tests 4=1 on both DH1 and DH2. Mfilos is right - I can also access DH2 using 37.34. (My DH2 even resides completely outside the first 8GB.) Also tried copying and reassigning Dopus to DH2, and it runs as it should. This also works with scsi.device 44.2. Have a look at the attached screenshots.. (DH3 device was supposed to be DH2, I initially had one more partition but changed my mind as there would be shitloads of debug button presses on each reboot.) Do you want me to try with some other version of PFS3 or scsi.device? Edit:* I did try scsi.device 43.47 by Cosmos, but my A600 just kept rebooting. I wasn't even able to enter the early startup menu. I used the a600-1200 version, but I guess it needs ks3.x. Attached Thumbnails Last edited by fgh; 24 October 2012 at 12:55.

25 October 2012, 11:26	#56
fgh Registered User Join Date: Dec 2010 Location: Norway Posts: 819	I just confirmed what mfilos was probably experiencing as well. On the 16GB CF card with 'ATA correct' ~7.8GB CHS values, PFS3DS v5.3 and scsi.device 40.5 still lets you see and work with partitions that start inside the first 8gb of the drive, but not partitions that reside completely outside it. (Tried first with a 9GB partition starting around 3GB position, and then an 6GB partition starting around 6gb position.) Last edited by fgh; 25 October 2012 at 11:53. Reason: Clarity

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Cant Install PFS3	videofx	support.Hardware	1	07 March 2013 18:14
PFS3 why cant i...	zharn	support.Apps	9	27 January 2013 06:27
PFS3 error: INVALID PFS3 COPY !!! WTF?	keropi	support.Apps	10	18 March 2008 22:30
Pfs3	Hewitson	request.Apps	3	22 December 2007 14:32
Installing PFS3 on 8.5GB SCSI HD	lopos2000	support.Apps	26	27 March 2007 19:31

21 October 2012, 19:00	#41
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,516	Really weird.. SCSI Read Capacity returns correct and full drive size. Are you 100% sure you have standard ROM? Something strange is going on..

21 October 2012, 21:22	#42
mfilos Paranoid Amigoid Join Date: Mar 2008 Location: Athens/Greece Age: 45 Posts: 1,978	I have an 27C800 with dual Kickstart in it. 2x256KB of Kickstart 1.3 and 1x512KB of Kickstart 3.1. In these picture I booted the stock Kickstart 3.1 without Startup-Sequence alas NO usage of ACATune to kick anything else as I normally do I'm pretty sure that with the original PFS3 distribution that I had in the past (PFS3+Patch = 18.5) I wasn't seeing the second partition without the usage of newer scsi.device For some reason your version of PFS3 recognizes the 2nd partition even with the stock scsi.device of 3.1!!!

24 October 2012, 12:58	#46
mfilos Paranoid Amigoid Join Date: Mar 2008 Location: Athens/Greece Age: 45 Posts: 1,978	I'm almost certain that my PFS3ds version (030) wasn't mounting my second (and outside 8GB) partition without using an updated scsi.device With your version Toni, it even shows in stock 3.1 scsi.device. I can tell for sure once I get home since @ work I only have WinUAE

24 October 2012, 15:06	#47
fgh Registered User Join Date: Dec 2010 Location: Norway Posts: 819	I'll let mfilos check the non-ks13 version then, since his a600 has both 030 and ks3.1. Toni: Will you submit your changes to the pfs3 sourceforge project?

24 October 2012, 22:03	#51
fgh Registered User Join Date: Dec 2010 Location: Norway Posts: 819	I seems to work well in WinUAE with UAE controller, no errors and can launch dopus from the partition. With quickstart a600, changing to 68030 but staying at ks37.300 rom the partition is shown as NDOS

24 October 2012, 22:19	#52
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,516	Perfect Only logical explanation is that some cards or adapters ignore ATA spec or there is some special CF/SD/whatever spec that allows it. ATA spec says that drive bigger than 16515072 blocks must return static 16383/15/63 CHS (cylinders/heads/sectors, ancient way to calculate disk size and address its blocks) value. This is the value that IDE scsi.device uses to detect the drive size. (163831563*512 = ~7.7G) If drive would return more "correct" CHS values, for example 16383/15/127 for 16G drive, it should be fully accessible to IDE scsi.device in Direct SCSI mode. I'll write short IDE identify device debugging program to confirm this..

25 October 2012, 11:03	#55
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,516	Excellent! Problem solved. Now we only need to find max supported values, too large values will most likely overflow in some calculations. Also I need to find some method that allows WinUAE to query correct CHS values from mounted harddrive/cf/sd/etc.

25 October 2012, 12:10	#57
mfilos Paranoid Amigoid Join Date: Mar 2008 Location: Athens/Greece Age: 45 Posts: 1,978	Exactly! Nice job there fgh!

25 October 2012, 17:45	#58
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,516	Do you see "correct" or "incorrect" geometry in winuae's log (Click "Add Harddrive" to generate disk log info) You should see something like this: device id string: 'Generic- Compact Flash 1.00 20060413092100000' BPS=512 Cyls=996 TPC=255 SPT=63 MediaType=11 device size 8195604480 (0x1e87f0000) bytes (I am quite sure this data isn't same as identify device cylinders value but better confirm it first..) I don't currently have >8G CF cards or SD-IDE adapters. Unfortunately Windows IOCTL_ATA_PASS_THROUGH (which can be used to send any ATA command) does not work with CF USB adapters (at least it won't work with my adapter). It returns 50 = ERROR_NOT_SUPPORTED. It seems to be very difficult to get exact same size on real IDE based Amiga and WinUAE. Stupid "buggy" adapters that return "illegal" ATA cylinder size information

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)