The New Features & Exciting Changes Of The Linux 4.10 Kernel

[phoronix]

Phoronix: The New Features & Exciting Changes Of The Linux 4.10 Kernel

Linus Torvalds is expected to release the Linux 4.10-rc1 kernel this weekend ahead of Christmas and thereby marking the formal end of the 4.10 merge window, but with all of the major pull requests already submitted and Linus tending not to honor last-minute pull requests of big changes, here is our usual look at the exciting changes and new features you will be able to find with the Linux 4.10 kernel.

http://www.phoronix.com/vr.php?view=23913

[GrayShade]

Perhaps the writeback throttling changes are worth mentioning: https://lwn.net/Articles/682582/

[yurikoles]

Michael There is a typo

LG Nexus 6X

Must be 5X.

[cjcox]

As for me, I'm going to wait until Linux becomes feature complete.

[starshipeleven]

As for me, I'm going to wait until Linux becomes feature complete.

not going to happen. World does not stand still and projects evolve with it.

[geearf]

Perhaps the writeback throttling changes are worth mentioning: https://lwn.net/Articles/682582/

Is it going in for 4.10?
The article link makes it seem very promising, but I didn't find any link about it and 4.10.

Thank you!

[GrayShade]

Is it going in for 4.10?
The article link makes it seem very promising, but I didn't find any link about it and 4.10.

Thank you!

Yeah, sorry for not providing a better link. The only reference I have is this one https://lwn.net/Articles/709017/.

[geearf]

Yeah, sorry for not providing a better link. The only reference I have is this one https://lwn.net/Articles/709017/.

No worries, this link is great!

I read a bit the mailing list from your previous link, and I got one question that hopefully you can answer.
After having this in, would BFQ still be that useful? I always thought the goal of BFQ (as BFS) was interactivity, which this patchset seems to also tackle in its own way.
Since this patchset started before the last review of BFQ, and the kernel devs still wanted BFQ for blq-mq then, I'm sure I'm missing the difference...

[GrayShade]

No worries, this link is great!

I read a bit the mailing list from your previous link, and I got one question that hopefully you can answer.
After having this in, would BFQ still be that useful? I always thought the goal of BFQ (as BFS) was interactivity, which this patchset seems to also tackle in its own way.
Since this patchset started before the last review of BFQ, and the kernel devs still wanted BFQ for blq-mq then, I'm sure I'm missing the difference...

As far as I can see it, they are different issues. Disk schedulers reorder the submitted disk requests in order to improve throughput and/or latency. Writeback comes into play when modified data in RAM has to be written back to disk, like when a process wrote to a file but the data is still in memory waiting to be flushed. The throttling patches try to slow processes down to prevent them from generating a lot of write requests. This happens before the [disk] scheduler sees those.

There are, of course, other differences. Writeback has nothing to do with disk reads, while the disk scheduler also handles them.

[geearf]

As far as I can see it, they are different issues. Disk schedulers reorder the submitted disk requests in order to improve throughput and/or latency. Writeback comes into play when modified data in RAM has to be written back to disk, like when a process wrote to a file but the data is still in memory waiting to be flushed. The throttling patches try to slow processes down to prevent them from generating a lot of write requests. This happens before the [disk] scheduler sees those.

There are, of course, other differences. Writeback has nothing to do with disk reads, while the disk scheduler also handles them.

Right, but whenever I read tests about BFQ it's always doing something alongside of heavy writing, hence I wondered if what BFQ was tackling was just the symptoms of the writeback issue.

Thank you!