PFS3AIO v3.1 Final

[Foul]

scsi.device 43.47b1 (Cosmos) attached if anyone need it

[apex]

Working fine using my A2000 with new AT-Bus Clone and Octapussy in ROM. :-)

[Toni Wilen]

PFS3AIO 3.0 Test #5 released. Restored lost v2 features.

[Romanujan]

Originally Posted by Toni Wilen

pfs3aio is again rommable and loadmodule etc.. compatible.

I wonder what can one gain from PFS in ROM? Is there a way to, for example, create a PFS formatted boot floppy this way?

[alexh]

Originally Posted by Romanujan

I wonder what can one gain from PFS in ROM?

You can boot from PFS partitions without PFS in the RDB?

[McTrinsic]

To be a bit more fair: you can boot from RDB- incompliant devices with PFS.

[Jope]

Yass, A2090 PFS DH0 :-)

[Turran]

Using Test1 in my CD32 and Test3 in my A1200. Both work great from what I've seen. Will update them when we get a release version.

Thanks Toni

[mark_k]

There might be a couple of issues in scsi_get_geometry().

struct SCSIBlockDescriptor *bd = (struct SCSIBlockDescriptor*)(ph + 1);
ULONG blocks = bd->scbd_NumberOfBlocks & 0x00ffffff;
if (blocks != 0x00ffffff)
	geom->dg_TotalSectors = blocks;
geom->dg_SectorSize = bd->scbd_BlockLength & 0x00ffffff;

The number of blocks field can contain zero: "A value of zero indicates that all of the remaining logical blocks of the logical unit shall have the medium characteristics specified."
So instead of if (blocks != 0x00ffffff) perhaps do if ((blocks != 0x00ffffff) && (blocks != 0))

Also it looks like there's an off-by-one bug with READ CAPACITY:

struct SCSICapacity *scc = (struct SCSICapacity*)buffer;
geom->dg_TotalSectors = scc->scc_Block;
geom->dg_SectorSize = scc->scc_BlockLength ? scc->scc_BlockLength : 512;

Instead of geom->dg_TotalSectors = scc->scc_Block; do something like

if (scc->scc_Block != 0xFFFFFFFF)
	geom->dg_TotalSectors = scc->scc_Block + 1;
else
	geom->dg_TotalSectors = 0xFFFFFFFF;

[Toni Wilen]

Originally Posted by mark_k

The number of blocks field can contain zero

This works as designed, target variable is already zero anyway.

Also it looks like there's an off-by-one bug with READ CAPACITY:

Oops, +1 and zero check (drive too large) got lost when I cleaned up the code It was "blind" assembly to C conversion originally, I didn't write it originally.

EDIT: This does not affect anyone. (Except maybe some OS 3.1.4 testers, I guess..)

[Toni Wilen]

Test #6, still some issues and missed AROS "port" features found..

[supaduper]

Originally Posted by Toni Wilen

Test #6, still some issues and missed AROS "port" features found..

Thanks again Toni,I`d be lost without pfs3aio !

[malko]

Originally Posted by Toni Wilen

Test #6, still some issues and missed AROS "port" features found..

Thank you Toni

[Toni Wilen]

Does anyone have some real world case which can be used to confirm working ACTION_DIE (unmount)?

[thomas]

The IDEfix package contains a KillDev program which can unmount CD drives mounted with CacheCDFS.

If you can get rid of a PFS partition with this program I would assume it works.

[apex]

Loading PFS via loadmodule is not working?

[huepper]

Originally Posted by apex

Loading PFS via loadmodule is not working?

Test #5 or #6 ?

[apex]

Tested #5.

[Toni Wilen]

Originally Posted by apex

Loading PFS via loadmodule is not working?

How exactly it is "not working?" It works as expected, pfs3 is added to FileSystem.resource.

[apex]

Ok, will try again using number 6. I used latest loadmodule from Aminet. FFS is working, for PFS I will check in the evening.