Samsung's Better exFAT Driver Gets Revised Ahead Of Mainline Linux Integration

[DrYak]

I don't think you understand what filesystems like FAT and exFAT are for.

I know! I know! I know! FAT and exFAT are filesystem that are for NOT-flash based media!

Like needing to constantly over-write in-place a file allocation table or two (hence their name) - which is a wonderful modern pinnacle of technology: dating back to 8bits and 12bits tables of FAT8/12 in the era of 5"1/2 floppy disks on 8bit home computers with less than 64k total RAM(*).

(exFAT is only a semi-modern rehash of a dinosaurean technology)

Let's be real. exFAT is based around a very ancient tech, and very ill adapted to flash media which is where nearly all industry is heading.
(and even the big storage industry sector which is traditionally staying with spinning rust tech: even they moving toward shingled magnetic hard drives which are just as averse to random in-place overwrites as flash is)

Log-structured file-systems like F2FS (and UDF) are better suited to these use-cases (and so are also CoW filesystems like Btrfs, ZFS, BCacheFS, etc.)

F2FS has been designed for flash media from the beginning (that where the "Flash Friendly" part of its name comes from). It is indeed a much better tech given where the future tech is heading.
Allocation-table based filesystem need to die.

---

(*) which leads to the only use-case where FAT still makes somewhat kind of sense. Given it's prehistorical personnal-computer roots, it's still pretty good with extremely low spec. FAT would make sense in (Arduino-class) 8bit microcontrollers with very limited RAM. But as soon as you move to ARM Cortex-M0 or anything befier than that, you are better of using F2FS.
That's why FAT-derivatives where used in embed application like point'n'shoot camera, back in the era when a M68k-derived micro controller seemed like a very powerful architecture.
Nowadays, most of these device either straight run Linux (lots of networked suverillance cameras, some Wifi servers embed inside point'n'shoot camera or even inside SD card) or some realtime OS on similar class CPU (e.g.: GoPro) - they would benefit from using F2FS instead.

[Ardje]

Which company are you talking of?

https://www.wired.com/2012/03/ms-patent/

Eh, we are of course talking about Microsoft. Microsoft has been known to threaten everybody with litigation over .*FAT.* . And in patent speak: microsoft stole the IP from linus torvalds, which means, they individually came to the same solutions, but you can't say that in patent court.
Microsoft has a lot of those dirty practices. That's why it's so weird to say that these days they are embracing the use of linux, publicly. But FUD parties in germany uphold the old mantra and forced german municipalities from linux back to windows, raising cost of ownership by licenses alone.
Governments are easy to sell into the windows fud, because you can make deals with dirty politicians to write the tender so it excludes certain open parties.

[duby229]

linux lacks any fsck utility for udf and other OSes may or may not have something but it's weird at best as it was never supposed to be used (as it's a filesystem for CD/DVD media that is most of the times never edited). I know it's more resistent to corruption because of its type (log structured filesystem), but that's still bad.

EDIT: hmm, it seems someone is working on that in more recent times https://github.com/pali/udftools/pull/7 and here is the most recent commit in a development repo of udftools https://github.com/jlmxyz/udftools/c...b2ee92b0ccb16a

I don't think you have a single clue how many filesystems tools like scandisk and chkdsk have corrupted. I absolutely guarantee it's far more than they ever corrected. It's the kind of tool that should -never- be automatically run and -no- end user should ever have access to. If a filesystem doesn't need error correction in userspace, then that is a much better concept by far.

[starshipeleven]

I don't think you have a single clue how many filesystems tools like scandisk and chkdsk have corrupted.
I absolutely guarantee it's far more than they ever corrected.

You don't have a single clue about what these tools can actually do. If the filesystem is trashed because of driver or hardware issues (not uncommon back then) they can't save it either.

But the same is true of any other filesystem tool.

It's the kind of tool that should -never- be automatically run and -no- end user should ever have access to. If a filesystem doesn't need error correction in userspace, then that is a much better concept by far.

Yeah, that's btrfs. But we are talking of UDF

[duby229]

You don't have a single clue about what these tools can actually do. If the filesystem is trashed because of driver or hardware issues (not uncommon back then) they can't save it either.

But the same is true of any other filesystem tool.

Yeah, that's btrfs. But we are talking of UDF

Ok, seems you do understand, my bad. But yeah, you're 100% right.

[rogerx]

I did a little more research on format-udf.sh" (git) and mkudffs.

I found mkudffs has supported creating fake partitions for MS Windows' compatability/readability since version mkudffs-2.0 released on Dec 29, 2017.

As such, no need for a heavy script for wiping prior partition table and writing a fake partition table.

# mkudffs --label={your_label} --bootarea=mbr /dev/sdx

Of course, "hexedit /dev/sdx" if you must to verify the partition table is correct.

Specifying "--blocksize=512" is also apparently not needed, at least for my case scenario here on Linux.

[rogerx]

linux lacks any fsck utility for udf and other OSes may or may not have something but it's weird at best as it was never supposed to be used (as it's a filesystem for CD/DVD media that is most of the times never edited). I know it's more resistent to corruption because of its type (log structured filesystem), but that's still bad.

EDIT: hmm, it seems someone is working on that in more recent times https://github.com/pali/udftools/pull/7 and here is the most recent commit in a development repo of udftools https://github.com/jlmxyz/udftools/c...b2ee92b0ccb16a

UDFTools udffsck main.c source has nothing in it, just a stub or as if there was something but now deleted.

https://github.com/pali/udftools/tree/master/udffsck

udftools/udffsck/main.c

int main()
{
return 0;
}

git log: Remove old non-working Ben's email address 13 months ago. (2020.01.21)

For a good review of open source and proprietary UDF fsck/check fs utils:
https://github.com/JElchison/format-...-tools-for-UDF

OK, found one of the initial commits for udffsck/main.c:

https://github.com/pali/udftools/commit/83cfbba4b758f3463fdfc921c55ad8434ede97af#diff-830507dbf7e75f6e7ab40670e50bbbf9

"udffsck: Do not install binary, it is only empty stub"

[starshipeleven]

UDFTools udffsck main.c source has nothing in it, just a stub or as if there was something but now deleted.

Yes that's what I said.
There is a beta fsck tool discussed in that issue on github and it seems to be found in the development repo I linked above (it is NOT a branch of the main repository)
see how that udftools/udffsck/main.c file is filled with stuff

https://github.com/jlmxyz/udftools/blob/fsck-udf-fixup/udffsck/main.c