PFS3AIO v3.1 Final

[Toni Wilen]

Originally Posted by loonsta

Hi Toni

I see in the git repository there are some commits since the last release on Aminet including fixes from upstream.

As I'm in the process of setting up a new 3.1.4 install doing research I read that a pfs3aio bug was found in 3.1.4 testing, so I was wondering if there is a newer pfs3aio binary with those changes available?

If you mean delete returning not empty directory in some rare situations: no. (and I didn't even know about it until months later...) Workaround apparently is to disable DELDIR.

But as usual: I need a test case.

mfilos on amiga forums in 2013: "All-In-One means that both versions (PFS3 and PFS3ds) are merged into one version. Also yesterday Toni release a Romable version of his PFS3 handler."
What benefit, if any, is there of putting the handler in rom? (I am making my own custome Kickstart Roms)

There is probably only one reason: if HD controller boot ROM refuses to load any non-FFS filesystems (they do exist!) or has broken filesystem loader. And it is still sane enough to use FileSystem.resource correctly.

Also are there any gotcha's I shuold be aware of when using PFS with a FastATA IV on A1200?

I don't know. Most likely driver is fine.

[loonsta]

Originally Posted by Toni Wilen

If you mean delete returning not empty directory in some rare situations: no. (and I didn't even know about it until months later...) Workaround apparently is to disable DELDIR.

But as usual: I need a test case.

What would need to be done for a test case? (in case I can help)

unrelated question: Is there any way to add an if statement in startup-sequence that would get skipped if running under uae?

[Toni Wilen]

Originally Posted by loonsta

What would need to be done for a test case? (in case I can help)

For example lha archive that when unpacked to pfs3 partition and then deleting all files would trigger the problem.

unrelated question: Is there any way to add an if statement in startup-sequence that would get skipped if running under uae?

There are few options, you can check for existence of autoconfig devices or device drivers or other UAE only modules. There are one or two threads somewhere with examples.

[Foul]

hi !

I just launched PFSDoctor after a crash of my system (CF Card Reader disconnected from USB) and an error occurred :



i started anyway to repair.. and everything was fine... but what's could be the prob with this error ???

edit: i just saw the last line of the 1st post and tested PFSDoctor Experimental but :

pfsdoctor CHECK HD0:
DEVICE HD0: not found

??!?

[indigolemon]

Originally Posted by Foul

pfsdoctor CHECK HD0:

Try:

pfsdoctor HD0 check

No ':'

[Foul]

Perfect ! thx

[apex]

Would it be possible to add real support for 4096 byte blocksize? A flash medium is using this native size and till yet only the new FFS support it.

[Toni Wilen]

Originally Posted by apex

Would it be possible to add real support for 4096 byte blocksize? A flash medium is using this native size and till yet only the new FFS support it.

Experimental pfs3aio should already support it (1024, 2048 or 4096 block sizes) but there is still same problem: It won't become official until there is filesystem torture test.

[apex]

Where can I get it? Such a test could kill the card. Maybe I will do it on a small one.

[Toni Wilen]

I meant it won't become official until someone writes Amiga filesystem torture test program.

"Experimental" pfs3aio needs to survive proper torture test first without crashing or corrupting data. Until it happens, I refuse to release any official versions with experimental features, buggy filesystem is the worst thing that can happen in any platform.

There is no need to use any CF cards, UAE + hardfile (stored on fast SSD) is much easier and faster.

[apex]

Sadly I cannot write such a tool and I am using only original Amigas. Ok, thank you. Seems there is no demand for such a version, because very less people understand the problem behind.

[daxb]

Maybe because there is nearly no writes to a "normal" Amiga system compared to modern ones. So, the problem doesn't effect in real world situations. Most people doesn't really use their systems, mostly just using some programs or playing games. No massive logging or permanent database writes.

[apex]

If I use my Amiga, very often small data is written. Icon collection, emails (SimpleMail), downloading and extracting files... I hope I am not one of the last users

[Storm]

I too use my A4000 quite a lot and move a bit of data around. I back up my other Amigas to the 4000 and then back the whole lot up to an external drive using USB. If I ever get problems with PFS, it’s only on the 4000 and usually transferring over USB. Something to do with Poseidon maybe?

[apex]

There is no problem with the filesystem, but it was made for real harddisks with a native blocksize of 512 byte. If you use now a flash device (CF card) with a native blocksize of 4096, every block/cell will be updated 4 times on a writing process. Could be a reason for dying CF cards. It is a very complex topic, I hope my few lines are right.

[Toni Wilen]

Lets not discuss that topic in this thread. There can be lots of reasons for problems (getting as cheap as possible CF card = don't do it. Bad/unstable power supply. Something simply corrupting memory. It is also strangely common to assume it is "normal" if HD has problems but not if CF card has problems..)

I'll probably write the tester some day.

[daxb]

I usually write and delete around 1GB data (HD-Recording) on my A1200 HD each week. Icons, emails, downloading/extracting, backups are peanuts. If you want to limit write access to your CF you should remember your Ramdisk for extracting and temporary data but I guess that doesn't make much difference.

The problem with CF, USB-Sticks and SD is that you don't know what you get. Quality can differ a lot. Then there is the old hardware with often unknown PSU quality that can affect your flash memory. So, you can only guess what kills your card.

[sean_sk]

Originally Posted by apex

If you use now a flash device (CF card) with a native blocksize of 4096, every block/cell will be updated 4 times on a writing process. Could be a reason for dying CF cards.

I had a look into this because I too was concerned about this in my system. I have a brand new Sandisk 32GB CF cards on my A600 IDE as well as a new Sandisk 32GB Micro SD card in the Vampires SD card slot. Both have been set up with 512 block sizes in WB 3.1.4 using PFS3AIO. In regards to both, Sandisk says the following in their technical documents:

The 512-byte sector size of a card in the SanDisk microSD Card Product Family is the same as that in an IDE magnetic disk drive. To write or read a sector (or multiple sectors), the host software simply issues a read or write command to the card. The command contains the address and number of sectors to write or read. The host software then waits for the command to complete.

The host software does not get involved in the details of how the flash memory is erased, programmed or read. This is extremely important because flash devices are expected to get increasingly complex in the future. Because cards in the SanDisk microSD Card Product Family use an intelligent on-board controller, host system software will not need to be updated as new flash memory evolves. In other words, systems that support the microSD Card Product Family today will be able to access future SanDisk cards built with new flash technology without having to update or change host software.

This document was dated 2012. The technical document for Sandisk Compact Flash cards basically says the same thing. Not sure about other brands though.
So therefore I'm wondering whether it is fine to just simply set it up as 512 byte block size.

EDIT: OK, I stand corrected. I believe sector size and block size may NOT be the same thing apparently. It seems a block size will include a number of sectors.

[apex]

Interesting what they write. Only long term usage will show us the true.

[npagonis73]

Hi there

How can I update my pfs3 system? I can see in my L folder 2 relevant files a) PFS3 (version 18.5) and b) pfs3_aio-handler-custom (same version). In the 3.1 pfs3aio file there are 2 files after extracting it: one pfs3aio and pfs3aio-custom both version 19.2.

I would appreciate any help in order to update mine.