XFS In Linux 4.16 Continues With "Great Scads Of New Stuff"

[phoronix]

Phoronix: XFS In Linux 4.16 Continues With "Great Scads Of New Stuff"

Back during the Linux 4.15 kernel merge window XFS file-system maintainer Darrick Wong commented there was great scads of new stuff and now with Linux 4.16 he's repeating that line. XFS for Linux 4.16 brings several significant changes to this mature Linux file-system...

http://www.phoronix.com/scan.php?page=news_item&px=XFS-Features-For-Linux-4.16

[FireBurn]

Last time I used XFS I had issues with Thunderbird losing its settings if the machine didn't shut down properly

[boxie]

Great Scott! That is scads of new stuff!

it looks like a lot of work went into making the "everything has gone to shit" use case more pleasant. Good stuff!

[ngkaho1234]

Last time I used XFS I had issues with Thunderbird losing its settings if the machine didn't shut down properly

It is also common that developers tend to forget about using fsync() at the right moment and IN THE RIGHT PLACE to make sure that things would not go horribly wrong if power outage happens.

[nils_]

It is also common that developers tend to forget about using fsync() at the right moment and IN THE RIGHT PLACE to make sure that things would not go horribly wrong if power outage happens.

And a lot of SSD also simply ignore it, presumably so they can sell versions with "Power Loss Protection".

[starshipeleven]

And a lot of SSD also simply ignore it, presumably so they can sell versions with "Power Loss Protection".

More like to cheat on write speed.

Power loss protection is to ensure that all data in the internal buffers was flushed to the NAND before power off.

[nils_]

More like to cheat on write speed.

Power loss protection is to ensure that all data in the internal buffers was flushed to the NAND before power off.

which is exactly my point.

[starshipeleven]

which is exactly my point.

no your point is bullshit.

Internal caches and these fsync tricks are there to boost write performance, not just because they want to sell otherwise useless features.

The only SSDs that don't really need "power loss protection" (i.e. these tricks) are the latest 3D nand from Intel (Optane?) or similar as they don't need a cache at all.

[nils_]

no your point is bullshit.

Hold your horses there champ, maybe have a sip of coffee before posting.

Internal caches and these fsync tricks are there to boost write performance, not just because they want to sell otherwise useless features.

This still doesn't excuse not honoring the protocol specification. HDD vendors used to do this too, confirming a write when data was only written to cache.

The only SSDs that don't really need "power loss protection" (i.e. these tricks) are the latest 3D nand from Intel (Optane?) or similar as they don't need a cache at all.

Which is what I criticize, any SSD that needs power loss protection should have. Instead this is a feature very often only offered in enterprise product.

[starshipeleven]

Hold your horses there champ, maybe have a sip of coffee before posting.

Maybe you should have taken coffee before making bullshit statements such as

"And a lot of SSD also simply ignore it, presumably so they can sell versions with "Power Loss Protection"."

This still doesn't excuse not honoring the protocol specification.

That's a different thing from adding a non-feature so they could sell you a "feature", which was what you said above.

I reacted to that.

Instead this is a feature very often only offered in enterprise product.

And Crucial SSDs