PFS3AIO v3.1 Final

[AMIGASYSTEM]

Sinphaltimus add StackAttack to your startup-sequence and you have no more problems with Stack

[kolla]

Let's talk about pfs3aio and buffers

Today, just for fun, I replicated the disk drive image I use on the Minimig only with pfs3aio instead of fastfilesystem. To my surprise, I noticed the RAM usage on the poor system (2MB of chipram, 1.5MB of slowram) increase with around 700kB for the 5 partitions. Of course I first checked with hdtoolbox, but all partitions had buffers set to 30, default, just like they were on the old image with FFS. But then I used addbuffers from the shell, and it said 150 on all the partitions. Oh well, so did "addbuffers dh0: -50", and it was down to 100, another -50 and it was 50, but -20 and it bounced up to 150 again. Hm, so I did -119 and it was 31, -1 and back up to 150. -121 and it was 29. Aha, so 30, the default, is magic. And so it seems - I can set the buffers to any number and it appears to be respected, except if I set it to 30 - it treats 30 as 150, which is mildly confusing. I know the pfs3aio changelog says quite a bit about adding buffers whenever needed, I just cannot recall anything about treating 30 buffers magically. When pfs3aio adds buffers on its own, is that visible using addbuffers?

Other than that, pfs3aio appears to work flawlessly on the Minimig. But it does eat considerably more RAM than FFS, even if I do silly stuff like setting buffers to 10 on all partitions, so I am not sure if it's a good fit for such a minimalistic system.

[daxb]

I guess a lower buffer value then 70 (at least for the old PFS3 version) shouldn't be possible but Toni can tell if he changed something and how it should work.

[oissery77]

Thanks Toni you're the best: )
Will update my old pfs3aio file system soon!

[Toni Wilen]

Originally Posted by kolla

Let's talk about pfs3aio and buffers

This is not pfs3aio feature, official pfs3 does the same. 30 becomes 150 because 30 is hdtoolbox default.

There is only 2 buffer releated changes: minimum buffers is smaller, buffers get automatically added (if memory is available) when/if pfs3 requires more. (pfs3 also uses buffers for caching modifications before they get written to the disk)

[kolla]

Thanks, I will read the original docs.

[oissery77]

updated! works great on my A500 KS 3.1, Classic WB, alfapower sidecar with CF HDD
thanks again Toni

[patrik]

Originally Posted by Toni Wilen

This is not pfs3aio feature, official pfs3 does the same. 30 becomes 150 because 30 is hdtoolbox default.

Is this a good feature and is it really needed? Inconsistent behavior is usually not a good thing.

[Romanujan]

Well - in the repository I can see a memory leak fix, some symlink fixes... https://github.com/tonioni/pfs3aio/commits/master

[Toni Wilen]

Originally Posted by patrik

Is this a good feature and is it really needed? Inconsistent behavior is usually not a good thing.

I don't know if it is good or bad idea but it is needed for backwards compatibility. "Fixing" it would mean everyone who has 30 buffers set, would suddenly get 30 buffers (instead of 150) and complaining that update made it slower!

Originally Posted by Romanujan

Well - in the repository I can see a memory leak fix, some symlink fixes... https://github.com/tonioni/pfs3aio/commits/master

They affect softlinks only and because they have always been broken (even in official pfs3) it isn't important fix.

[Amiga1992]

Originally Posted by kolla

Let's talk about pfs3aio and buffers

We discussed this in length last year, here's the thread so you see where we got to:
http://eab.abime.net/showthread.php?...light=mounting
You can see which results aI reached for a low memory system like an A600 or a Minimig.

Using 5 automount partitions is a bad start

[TuKo]

Ok, a bug report here.

It seems pfs3aio 3.0 (even with latest version built from GitHub) doesnt compute correctly free and full space reports to WB

(it should be 85% to be correct)

ShK tested same with FFS (thanks for his dedication for bug hunting) :

[Toni Wilen]

Sorry but that does not prove anything. There can be multiple reasons that can cause it.

EDIT: and as usual: not enough information, information content actually is non-existing.

[TuKo]

With debug infos

[Toni Wilen]

That won't help at all. There is no useful information whatsoever!

KS version? Any patches? WB version? Any patches? Drive size? RDB/partition information? FFS blocksize? (test is invalid unless block size is 1024). What does C:Info report? Does same happen with UAE controller? What scsi.device version?

Minimum information: empty/as small as possible vhd hardfile that shows the problem.

I still think this has nothing to do with PFS3 because filesystem only returns number of used and free blocks, it is WB that calculates the result.

[TuKo]

KS version = 3.1
Any patches = too many to list..
WB version = 3.9BB2
Drive size ~28GB
RDB/partition information = see attached files
FFS blocksize = 1024
C:Info report = valid
Does same happen with UAE controller = yes
scsi.device = 43.45p

With same setup, FFS does not trigger this computation error

[Toni Wilen]

Originally Posted by TuKo

Any patches = too many to list..

C:Info report = valid

Even more pointing to some bug or stupid assumption in some patch. I am going to ignore this unless it also happens with OS39BB2 + UAE controller. I don't trust most 3rd party patches, never have, never will.

[mpattonm]

Hello, I feel stupid, but cannt help it - is there some kind of how-to-install instructions?

[malko]

Originally Posted by mpattonm

Hello, I feel stupid, but cannt help it - is there some kind of how-to-install instructions?

Download the pfs3 package from aminet. It contains the documentation/faq/etc. But install the new pfs3aio from Toni of course

[antiriad76]

I am having issues with a 600GB+ PFS3 Partition on WinUAE using Fast speed. It seems that during the copy of large files UAE is freezing randomly. Cursor is still responding but there is no HDD activity.

PFSDoctor shows no errors.

Is this size way too big?

Buffers is set to 80
Max Transfer 0x1FE000 (Fast IDE Standard)
Mask is 7FFFFFFE

Thanks

Mike