PFS3 file system and Tornado angst

[Angus]

Hi all.
I have finally completed a mission that was giving me a lot of trouble in Digital Integration's Tornado - I was pretty chuffed at having finally made it back to base intact, the enemy runway, a smoking ruin.


But my celebrations were premature, when I quit out of WHDLoad I got a whole string of error requesters referring to PFS3, which I relatively recently installed on all partitions of my PiStorm 32's SD card except the boot partition, SDH0:


Can anybody diagnose the problem? Here are the error messages in the sequence they appeared on Workbench:


(This also happened the other day, I tried writing a random file or two to SDH1: and it worked okay. I have also run DiskValid and PFSDoctor to analyse the problem but they found no errors.) :-/









PFSIII error requester
Device SDH1:
ALERT:
Wrong ablock id 00004102 block 21814







Device SDH1:
LRU update failed block 27668 error 214




Device SDH1:
LRU update failed block 27672 error 214




PFSIII error requester
Device SDH1:
ALERT:
Wrong ablock id 0000ffff block 27672





Device SDH1:
Deldir invalid




Device SDH1:
Disk update failed




DOS error #214
(Disk is write protected)
on deleting/writing file "DATA/LOGFILE.DAT"


(I hit retry here)





Device SDH1:
ALERT:
Wrong index block ID 0000ffff, expected 00004942, block 27684,0,253


(I hit cancel at this point)


Any thoughts?

[mja65]

Using a SD extender?

[Angus]

Originally Posted by mja65

Using a SD extender?

No, straight into the beast.

and this just in......

I seem to lose SDH1 completely after an error, and the PFS tools say "could not inhibit SDH1:" But if I reboot, it reappears - at least for the time being.

[Don_Adan]

Im not expert, but for me Tornado is saving any data after mission.
PFS3 needs more time for update, but no disk validation.
FFS need less time.
You can try to run this game from FFS or RAM.
Or someone must fix write delay.

Edit, or you can redirect save only to FFS partition via assign, if this game used save disk or save directory.

[malko]

Originally Posted by Angus

[...] Can anybody diagnose the problem? Here are the error messages in the sequence they appeared on Workbench: [...]

Difficult as is.
Maybe providing some informations about the CF card size, the used parameters to configure it (- mask, -maxtransfert, -buffers, pfs3aio version, used ID, MBR ? Mountlist ? etc. ?) may help.

[Toni Wilen]

If it only happens with that single game: my guess is that game randomly corrupts filesystem/dos buffers.

[Don_Adan]

Or you can change WriteDelay value for WHDload, maybe is too small for this game and PFS3?

https://whdload.de/docs/en/opt.html

[Angus]

Update.


I tried saving on another game, Steel Empire, and got similar results with a zillion error requesters.

The partition now seems to be much less stable.

I just tried to use the analyse feature in DiskValid and it did the checks and came up okay, then a requester appeared saying please insert volume SDH1: !

DiskValid also says the partition was formatted using PFS-II 19.2 I thought was PS-III?!!

It seems if I reboot SDH1: reappears, but if I check it with DiskValid or PFSDoctor it does an initial check which seems okay, but then the volume SDH1: is lost to the system. :-/

I think this is a PFS3 error rather than the games which have worked flawlesslly before this.

Although perhaps I did something wrong when I installed it?

I put in a Mask value of 0x7ffffffc

and a Maxtransfer of 0x1fe00

and set the buffers to 300

[Toni Wilen]

If it happens with other games, it starts to sounds like some kind of hardware or software problem and not directly pfs3 related.

EDIT: WriteDelay can't cause this. Only difference is that if "last write" is not done by PFS3 and system is rebooted, all modications in this "transaction" are rolled back by design. It won't corrupt anything.

[Angus]

Originally Posted by Toni Wilen

If it happens with other games, it starts to sounds like some kind of hardware or software problem and not directly pfs3 related.

EDIT: WriteDelay can't cause this. Only difference is that if "last write" is not done by PFS3 and system is rebooted, all modications in this "transaction" are rolled back by design. It won't corrupt anything.

Thanks Tony Possibly just a damaged SD card, then?

[DisasterIncarna]

Originally Posted by Angus

Update.
I tried saving on another game, Steel Empire, and got similar results with a zillion error requesters.

was this save attempt made on your already damaged pfs3 drive? or from something fresh where you then noticed that other game do the same thing?

If you state other games are doing it, that could just mean whatever broke things originally has not been fixed yet and is just snowballing damage from other game saves.

[malko]

Originally Posted by Angus

[...] Although perhaps I did something wrong when I installed it?
I put in a Mask value of 0x7ffffffc
and a Maxtransfer of 0x1fe00
and set the buffers to 300

You seems to have a huge amount of RAM in your config.
I cannot help precisely but here is a subset of my notes about PFS settings.

- the MaxTransfert value depend on the type of controller you have. Also the MaxTranfert value must be a multiple of 512 (something related with a bug in the WB "format" code). Usually for an IDE controller you can risk-free use "0x1FE00" (which is 255 × 512, about 128KB) and, for an SCSI controller, use "0xFFFE00" (which is 32767 × 512, about 16MB). Optimal values can be determined by setting values that are power of 2 minus the size of a block (512) until you get errors while copying large files. But speed gain seems to be very marginal (I never tried that and stayed with usual values).

- the DMA Mask value depend on the amount of memory installed and will depend also if your HD controller can access whole memory or not. Default value of 0xFFFFFF restrict DMA to Zorro2 fast-memory area (16MB max). For Zorro3 default value is 0xFFFFFFFF. Default may work fine but it is recommended to use the WB program "show config" to get the max value of memory that is displayed there. Then you define the mask value like :
- dma mask start with 1, 3, 7 or F
- all others digit are substituted by F
- last digit must be C
For example my config shows a max memory address of $9FFFFF (8 MB fast-ram). So mask -> 9 becomes F, others F stays, last becomes C -> mask is then 0x00FFFFFC.
If you change the amount of memory it is important to adapt the mask value before installing the memory.

- buffers 300 as you setted seem ok.

Hope it helps

[Don_Adan]

For me, to be sure that this is damaged SD card, you can try without PiStorm.
Or disable JIT, totally slowdown PiStorm.
PiStorm can be too fast or can be buggy too.
Of course you can use other SD card for test too.

[Angus]

Originally Posted by Don_Adan

For me, to be sure that this is damaged SD card, you can try without PiStorm.
Or disable JIT, totally slowdown PiStorm.
PiStorm can be too fast or can be buggy too.
Of course you can use other SD card for test too.

Disable JIT? Do I need WHDLoad wrapper for that?



I just copied Steel Empire on to another PFS3 partition (SDH2) and saved a game. It worked fine.


If I try the same with Tornado, I'm worried it might damage that partition.


Supposing I backup SDH1: (the problem partition) format it, and then try again with it?


Sorry for my ignorance, but if I format it from Workbench will it still be PFS3 afterwards?

[utri007]

I have found out that some games does not work with PFS partion. Bzsili's Doom RPG as a example.

[patrik]

Originally Posted by malko

- the MaxTransfert value depend on the type of controller you have. Also the MaxTranfert value must be a multiple of 512 (something related with a bug in the WB "format" code). Usually for an IDE controller you can risk-free use "0x1FE00"
…<SNIP>
- the DMA Mask value depend on the amount of memory installed and will depend also if your HD controller can access whole memory or not. Default value of 0xFFFFFF restrict DMA to Zorro2 fast-memory area (16MB max). For Zorro3 default value is 0xFFFFFFFF
…<SNIP>

Both MaxTransfer and Mask are values that the filesystem can use as a workaround for bugs/limitations in the hardware or the device driver.

MaxTransfer: The filesystem will split the transfer into chunks of this size

Mask: The filesystem will apply this mask to the address of the transfer and if it doesn’t match will transfer the data via chipmem in small blocks or similar “safe” region it deems safe.

A very known such bug is the C= scsi.device IDE driver corrupting data when more than 0x1fe00 bytes of data are transferred in one go.

If there is no such bugs/limitations, both values can be 0xFFFFFFFF.

[malko]

Found back some sources of my notes (PFS manual excepted)
https://eab.abime.net/showthread.php?t=45491
https://eab.abime.net/showthread.php?t=61666

@Patrik: thanks.

[Don_Adan]

Originally Posted by Angus

Disable JIT? Do I need WHDLoad wrapper for that?



I just copied Steel Empire on to another PFS3 partition (SDH2) and saved a game. It worked fine.


If I try the same with Tornado, I'm worried it might damage that partition.


Supposing I backup SDH1: (the problem partition) format it, and then try again with it?


Sorry for my ignorance, but if I format it from Workbench will it still be PFS3 afterwards?

I think only for tests.
PiStorm has control program, if I remember right.
If you want general solution, then you can create small (10-100MB) FFS partition.
And redirect to this partition all WHDload saves via WHDload config.
Even if validation data occured, it will be take a few secs for very fast CPU and small partition.
Can be invisible.

[Don_Adan]

And of course you can create small 10-100MB PFS partition for saves too, if you want.
If problem occured again, you can easy reformat this partition or change to other file system.
In general split/move game saves from big PFS partitions to special small partition.
This is less dangerous.

[thomas]

Originally Posted by Angus

please insert volume SDH1:

This message appears if you point DiskValid to the volume rather than the device. If you named both the volume and the device the same, then rename the volume and use the device name (which should now be unique).

Originally Posted by Angus

could not inhibit SDH1:

This means there are still open files or directories on the partition. Boot without startup-sequence and try again.

Originally Posted by Angus

using PFS-II 19.2 I thought was PS-III?!!

The PFS-text is hard-coded. Only the version number is relevant.