XFS File-System With Linux 5.12 Has "A Lot Going On This Time"

[curfew]

filesystem interacts with other pieces of kernel and bugs in software often result from incorrect interaction between pieces. and rust doesn't make programs magically bug free, it can be seen as a silver bullet only by imbeciles like you or by people who never learned language more advanced than c

I don't think that guy has ever learned programming of any kind.

[Vistaus]

uid313 We already have BTRFS, Ext4, XFS and many more. What would be the value of yet another filesystem?

To be able to use the phrase "Yet Another xyz" as the name of a file system.

[Vistaus]

XFS' (lack of) shrink support is the main thing that has stopped me using it all these years.

That said, now we have btrfs I'm not sure if I would bother going back to XFS even if it did pick up shrink support. Maybe if it picked up subvolume support as well.

And yes, for typical laptop uses you *can* trust btrfs with your data, at least on *buntu and Manjaro.

What is this "shrink support": shrinking with e.g. GParted? If so, then I won't care 'cause I only use one distro at a time anyway, so there's no need for me to mess around with resizing partitions. And in that case, I might give XFS a try next time I do a reinstall.

[Vistaus]

Do you really think that a programming language has much to do with how reliable a filesystem is? You can write the most error prone algorithm in BASIC or wreck an entire database easily while coding in PHP.

If you've ever paid attention to his contributions on Phoronix, you'd know that uid313 is the main Rust advocate, so of course he thinks it's *the* universal solution to every problem of mankind

[flower]

i dont think shrinking is an important feature.
if you need something like that just make a thin pool below xfs and use trim.

[kreijack]

You can use dm-integrity together with any raid level >0 to defend against bitrot.
You will get correct data with any fs, but you need to scrub to write it back.

From what I read[*], dm-integirty is not "integrated" with raid. This means that:

1) dm integrity over raidX
if the data returned by raid is *corrupted*, dm-integrity can detect it but it can't repair it. Even a scrub at raid level, can't repair the data.

2) raidX over dm-integrity
a data corruption is detected (and corrected); if a raid copy is *misaligned* to the other one, dm-integrity (1) can't detect and/or repair it. A scrub at raid level, can repair the data.

3) I don't have any data, however dm-integrity uses journal, and also, in order to have a reliable raid5/6, MD has to use journal. Journal over journal... I don't think that the performances will be great. If someone has done a performance test, a link is appreciated.

(1) eg. due to a crash or a power failure[*]
- https://raid.wiki.kernel.org/index.php/Dm-integrity
- https://www.redhat.com/en/blog/what-...detect-it-rhel
- https://www.kernel.org/doc/html/late...integrity.html

[uid313]

Rust does not magically make a filesystem or a system driver more robust, stable or have less bugs. The Rust Cargo Cult needs culling.

Well, it kind of does. Sure you can still implement poor algorithms and make bad design decisions and use dynamic dispatch or unsafe calls, but you do eliminate entire categories of bugs, no such things as buffer overflows, memory safety problems, null references, etc.

[waxhead]

If you've ever paid attention to his contributions on Phoronix, you'd know that uid313 is the main Rust advocate, so of course he thinks it's *the* universal solution to every problem of mankind

You don't say.... and I who was under the impression that *I* was the universal solution to every problem of mankind. Well well.... thanks for the heads up then

[pal666]

You can use dm-integrity together with any raid level >0 to defend against bitrot.
You will get correct data with any fs, but you need to scrub to write it back.

first, dm-integrity doesn't know which blocks are really used by fs. second, you can't have metadata-only dm-integrity. third, it's slower than btrfs version. fourth, nobody uses it anyway, so this exercise is academic. in reality you either use btrfs and have protection, or you don't

[pal666]

Yes, it did, which is why Rust is seeing so much adoption.

what other browser switched to rust being scared by bugfree firefox? what are you smoking?

Is the rest of Btrfs stable?

status of rest is tracked here https://btrfs.wiki.kernel.org/index.php/Status

I've heard so much problems about Btrfs, I never heard so many complaints about any other file system.

i'm pretty sure you've heard more complaints about systemd than about sysvinit, what such scientific measure tells you?
and btw, how many complaints exactly you've heard about btrfs in 5.12(if someone will advise you to use kernel 3.x with btrfs, spit him in the face)?