PFS3aio v3.2 test

[pcotter]

Screenshots of Mem usage of the same boot Partition with block size of 1024 and 4096. Booted without SS
WinUAE used for all tests

[TuKo]

DH0: is 280MB, formatted with pfsformat, 4KB block size

[Toni Wilen]

Originally Posted by TuKo

DH0: is 280MB, formatted with pfsformat, 4KB block size

Hmm.. <1G reserved block allocation has not changed.

Could you include pfsdoctor output using same partition size but 512 block size?

[janz]

@pcotter Can you boot from new PFS 0.20 partition on PiStorm32-lite? No matter what the block size is. I tried and it doesn't here...

[TuKo]

Originally Posted by Toni Wilen

Hmm.. <1G reserved block allocation has not changed.

Could you include pfsdoctor output using same partition size but 512 block size?

[pcotter]

Originally Posted by janz

@pcotter Can you boot from new PFS 0.20 partition on PiStorm32-lite? No matter what the block size is. I tried and it doesn't here...

Hi Jan, i confirm too.
Using the PiStorm32 it boots from the internal IDE A1200 (scsi.device) but not from the SD.
I have been using only WinUAE for the tests.

[janz]

Originally Posted by pcotter

Hi Jan, i confirm too.
Using the PiStorm32 it boots from the internal IDE A1200 (scsi.device) but not from the SD.
I have been using only WinUAE for the tests.

Thanks for your confirmation, Pedro. It seems that Emu68 struggles with new PFS the same way as it struggles with SFS: https://github.com/michalsc/Emu68/issues/219

[Rochabian]

@Toni

Could you share a version with 4GB support to test if it works or do you plan to add it later when other tests are done ?

[Toni Wilen]

Originally Posted by Rochabian

@Toni

Could you share a version with 4GB support to test if it works or do you plan to add it later when other tests are done ?

4G limit? There has never been any 4G limits in pfs3.

[Rochabian]

I mean Filesize limit, sry

Perhaps i have misunderstood the first post. The Test version allows files below 4GB but it's planned for files equal or superior to 4GB if needed ?

I'm right ? I just want to test with files from 2GB to 4GB

[Toni Wilen]

Yes, 4G minus 1 byte is the max supported file size in 3.2 betas.

Many programs probably get confused with >2G file sizes (negative size, seeking above 2G may not work etc). You are on your own if you create >2G files

[Toni Wilen]

There won't be anything official until someone writes filesystem torture test. I want to be very sure that it won't destroy data suddenly after being used for few months or even earlier.

There is also another issue, pfsdoctor reports off by one block size error when partition size is exactly "right" with >=1024 block size. Need to solve that first too.

[McTrinsic]

Any chance of a version for the braindead Phoenix-SCSI?

I tried using a small FFS partition and mount PFS3v20t5 partitions via mountfile.
Under KS1.3 and KS3.1, the Amiga freezes when adding a partition.
Here is the mountfile... "mount DATA" doesnt work, freezes the system (file is named data).

FileSystem = Lfs3t (in the middle is L: Pfs3t... : and P being converted to emoji)
Device = pbscsi.device
Unit = 0
BlockSize = 4096
Surfaces = 16
SectorsPerBlock = 8
BlocksPerTrack = 255
Reserved = 2
PreAlloc = 0
Interleave = 65536
MaxTransfer = 0x00FFFFFF
Mask = 0x7FFFFFFE
LowCyl = 655
HighCyl = 1654
Buffers = 30
BufMemType = 1
StackSize = 16384
Priority = 0
GlobVec = -1
DosType = 0x50445303
Activate = 1
Mount = 1
/* Bootable = FALSE */

Edit: I tried to define a mountfile for KS/OS1.3 and had the same effect.
Mount complains about some unsupported keywords, e.g. surfaces. Anyone have a good resource where I can learn about 1.3 mountfiles?

[pawelek]

Amiga 1200, Blizzard 1260/50

1.AmigaOS3.2:> ibm3:
1.IBM3:> DriveSpeed 1230scsi.device 2
Drive information:
Manufacturer Name: IBM
Drive Name: DNES-309170W
Drive Revision: SA30
Raw read: 9004137 bytes/sec

1.IBM3:> pfsdoctor_v3.2 ibm3: info
Autodetected disk access mode: Standard
Initializing
rootblock 100...
Checking rootblock
Checking disk IBM3
Option mask: 0000077f
Reserved blocksize: 1024
Reserved block cluster: 16
First reserved block: 2
Last reserved block: 117761
Reserved blocks: 58880
Free reserved blocks: 58084
Free blocks: 5761387
Always free blocks: 313911
Total blocks: 6396000
..
Checking rext
Max file name length: 32
Disk formatted with PFS-III 0.19

1.IBM3:> Diskspeed ALL
Diskspeed 1.2 (12 July 1998) - Programmed by John Hendrikx

Volume IBM3: 5964327 blocks (2912 MB), 512 bytes per block.
The volume has 300 buffers and the DosType is 0x444f5303.

Creates 1000 files of 128 bytes : 101 files/second
Scan all files in a directory : 9895 files/second
Lock and unlock a random file : 415 files/second
Open and close a random file : 392 files/second
Load a random file : 285 files/second
Deletes 1000 files : 314 files/second
Random seek/reads of 2048 bytes in
2097152 byte file : 176 times/second
Read data using 65536 byte buffer : 8230 kB/second
Write data using 65536 byte buffer : 5376 kB/second

----------------------------------------------------------------------------

1.IBM3:> pfsdoctor_v3.2 ibm3: info
Autodetected disk access mode: Standard
Initializing
rootblock 100...
Checking rootblock
Checking disk IBM3
Option mask: 000017ff
Reserved blocksize: 4096
Reserved block cluster: 1
First reserved block: 2
Last reserved block: 7937
Reserved blocks: 7936
Free reserved blocks: 7901
Free blocks: 726944
Always free blocks: 39578
Total blocks: 6396000
..
Checking rext
Max file name length: 32
Disk formatted with PFS-III 0.20
1.IBM3:> Diskspeed ALL
Diskspeed 1.2 (12 July 1998) - Programmed by John Hendrikx

Volume IBM3: 751984 blocks (2937 MB), 4096 bytes per block.
The volume has 300 buffers and the DosType is 0x444f5303.

Creates 1000 files of 128 bytes : 99 files/second
Scan all files in a directory : 9565 files/second
Lock and unlock a random file : 435 files/second
Open and close a random file : 410 files/second
Load a random file : 120 files/second
Deletes 1000 files : 343 files/second
Random seek/reads of 2048 bytes in
2097152 byte file : 165 times/second
Read data using 65536 byte buffer : 8032 kB/second
Write data using 65536 byte buffer : 5565 kB/second

[pawelek]

Pfsaio mode selection order:
If last block of partition is outside of first 4G of drive:
- Test if NSD is supported and can access last block
- Test if TD64 is supported and can access last block.
- Test if HD_SCSICMD Read(10) can access last block.

The 1230scsi.device driver supports TD64 and DirectSCSI,
but for me, for an IBM 9GB SCSI disk and Hitachi 70GB SCSI disk, it sets the mode:

Autodetected disk access mode: Direct SCSI

and it works very well, but...
Question: Which mode for BlizzardSCSI controller is better?
Does mode autodetection for scsi controllers work correctly?

[Toni Wilen]

Originally Posted by McTrinsic

Any chance of a version for the braindead Phoenix-SCSI?

Do you mean similar to "custom" version included with previous pfs3aio official release? It will happen when current version gets closer to release.

Originally Posted by pawelek

Amiga 1200, Blizzard 1260/50

1.AmigaOS3.2:> ibm3:
1.IBM3:> DriveSpeed 1230scsi.device 2
Drive information:
Manufacturer Name: IBM
Drive Name: DNES-309170W
Drive Revision: SA30

Don't really care what mechnical drives do but apparently 4k blocks are not that good idea with those...

Originally Posted by pawelek

Does mode autodetection for scsi controllers work correctly?

Pfsdoctor TD64/NSD detection was broken (possibly it worked accidentally when compiled with older compilers and depending on how driver responded to zero size TD64/NSD read attempt). test5 zip updated with fixed pfsdoctor.

[pipper]

I have a question: PFS3.1 managed to "duplicate" my SYS:Prefs folder.
I.e. I have 3 'SYS:Prefs' folders, all pointing to the same directory.
I can't delete the duplicates as it would delete the whole Prefs directory. pfsdoctor doesn't detect a problem...
I is there any way to fix this apart from creating a copy of Prefs, the delete all Prefs entries and eventually copying it back?
I do not know how what I did to create this situation, but I think it might have been due to a flakey A3000 SCSI host controller leading to corruption. I have since moved to an A4091.

[Toni Wilen]

Originally Posted by pipper

I have a question: PFS3.1 managed to "duplicate" my SYS:Prefs folder.
I.e. I have 3 'SYS:Prefs' folders, all pointing to the same directory.
I can't delete the duplicates as it would delete the whole Prefs directory. pfsdoctor doesn't detect a problem...
I is there any way to fix this apart from creating a copy of Prefs, the delete all Prefs entries and eventually copying it back?
I do not know how what I did to create this situation, but I think it might have been due to a flakey A3000 SCSI host controller leading to corruption. I have since moved to an A4091.

Sorry, no idea. I don't know that much about pfs3 internals. Maybe copy everything inside it to some other directory first, then delete all of them? But it could cause weird side-effects because above is probably "impossible" situation. (Unless they are hard/soft links but I suspect not)

[pawelek]

Originally Posted by Toni Wilen

Pfsdoctor TD64/NSD detection was broken ... test5 zip updated with fixed pfsdoctor.

Now auto detect mode works fine. Thank you for the update.

[McTrinsic]

Originally Posted by Toni Wilen

Do you mean similar to "custom" version included with previous pfs3aio official release? It will happen when current version gets closer to release.

Yes, that’s what I mean.
I guess I just have to be patient. ?