PFS3 All-in-One released

[Geijer]

Originally Posted by Toni Wilen

Why? Does some situation really exists where you don't have automount/boot harddrive (filesystem in RDB) and still need PFS3?

Us poor bastards with T-Rex II.

the tekscsi2.device is not NSD compliant, it autoboots FFS partitions only - it looks for filesystems in the Kickstart ROM, but it does not care about filesystems loaded into the RDB area

[Toni Wilen]

Originally Posted by Geijer

Us poor bastards with T-Rex II.

Is it confirmed that it really does not check any RDB filesystems or if it is only "locked" to DOSx dostypes? In other words, if updated FFS is installed to RDB, does it still use the (older) ROM FFS?

[mfilos]

@Geijer
Damn I didn't know about that!

[emufan]

here is a thread with same issue. that confirms the rdb issue with GVP 4060 (T-Rexx II).

[Toni Wilen]

Originally Posted by emufan

here is a thread with same issue. that confirms the rdb issue with GVP 4060 (T-Rexx II).

No, it does not confirm anything. I want 100% proof. Question that still needs an answer: If RDB has newer FFS compared to ROM version, does it use RDB or ROM FFS?

[videofx]

Hi toni,

I have a new Trex II 4060T coming to me and can test PFS3AIO andf let you know.

[Damion]

The 2060 and 4060 SCSI driver is broken by default, and only looks for filesystems in the KS ROM. You are stuck booting from FFS partitions, other filesystems don't even show in the bootmenu. You can boot from SFS partitions if SFS is stuck in a custom ROM (at least I did on my 2060).

Fortunately, Doobrey was kind enough to tackle the problem. You'll have to burn and install the fixed ROM(s) after running the patch. Been running the fix for a few years on my 2060, booting from PFS3 partitions.

I haven't had feedback yet on the 4060 fix, but presumably it works as both versions were riddled with the same bug.

Further details and patch available on Doobrey's site: http://www.doobreynet.co.uk/amiga.html

[Toni Wilen]

Finally good enough proof, thanks

It should be as simple as adding resident structure inside pfs3 aio that adds pfs3 dostype references to FileSystem.resource. (It won't have any effect with normal RDB usage)

But first I need to check if current pfs3 is fully rommable.

[Toni Wilen]

ROMmable test version attached.

Changes since previous public version:

- Includes resident structure that adds PFS\3 and PDS\3 dostypes to FileSystem.resource, also creates FileSystem.resource if it does not exist (Required by KS 1.3)
- 100% pure/rommable.

[emufan]

thanks toni.

[Damion]

Wow, thank you!

[Geijer]

Thank you very much!

[NovaCoder]

Toni,

Cool great work

Any plans to improve the performance for us classic Amiga users (we need all we can get).

[Toni Wilen]

Originally Posted by NovaCoder

Any plans to improve the performance for us classic Amiga users (we need all we can get).

No. I don't think it would make any noticeable difference. I don't believe in micro-optimizations that are only visible in some specific benchmarks. You only lose source code readability which is much more important.

[Toni Wilen]

Can anyone confirm that it does work with original&buggy 2060/4060 roms?

[breech]

I just wanted to say thanks for AIO, it's brilliant on my A500 -> A590 -> PCD50b -> 2g CF, WB2.1

[midway]

Originally Posted by Geijer

Us poor bastards with T-Rex II.

Or tekmagic on the a2000!

[Toni Wilen]

Attached is new test version.

- Reverted SVN rev#9 commit, it is not compatible with AmigaDOS and can break some programs. ACTION_FINDOUTPUT must use exclusive lock.
- 1.3 compatibility fix, AddSemaphore() is unreliable under 1.x. Caused crash under 1.3 if more than one partition was mounted.
- Added special version for braindead boot ROMs (pfs3 handler with -custom extension). This version is not rommable. pbscsi.device (perhaps others) have really bad RDB filesystem loader, it only works with filesystems that have single hunk and no reloc entries! (=1.x FFS only) Special version has been "wrapped" inside custom relocator that "fixes" pbscsi.device's requirement of no reloc entries.

This will be uploaded to aminet if no problems found.

Thanks to McTrinsic for testing special version with Phoenix Board pbscsi.device.

Horrible details of pbscsi.device:

It loads first LSEG block, looks for 0x0003E9 (code hunk) long word, next long word is used to allocate memory for filesystem, load remaining LSEG blocks. Done. Other hunks are ignored, reloc entries are ignored..

[Retrofan]

Wonderful. Thanks Toni. I will have to try if now it can work with SkinnableClock, maybe is the first fix you list.

[McTrinsic]

Thanks a LOT Toni!!!

Investigating further an issue that I would like to understand irst before posting about.

Cheers,
McT